# INTERACTIONS BETWEEN DIETS, GUT MICROBIOTA AND HOST METABOLISM

EDITED BY : Jie Yin, Liwei Xie, Yuheng Luo and Helieh S. Oz PUBLISHED IN : Frontiers in Nutrition, Frontiers in Microbiology and Frontiers in Physiology

#### Frontiers eBook Copyright Statement

The copyright in the text of individual articles in this eBook is the property of their respective authors or their respective institutions or funders. The copyright in graphics and images within each article may be subject to copyright of other parties. In both cases this is subject to a license granted to Frontiers. The compilation of articles constituting this eBook is the property of Frontiers.

Each article within this eBook, and the eBook itself, are published under the most recent version of the Creative Commons CC-BY licence. The version current at the date of publication of this eBook is CC-BY 4.0. If the CC-BY licence is updated, the licence granted by Frontiers is automatically updated to the new version.

When exercising any right under the CC-BY licence, Frontiers must be attributed as the original publisher of the article or eBook, as applicable.

Authors have the responsibility of ensuring that any graphics or other materials which are the property of others may be included in the CC-BY licence, but this should be checked before relying on the CC-BY licence to reproduce those materials. Any copyright notices relating to those materials must be complied with.

Copyright and source acknowledgement notices may not be removed and must be displayed in any copy, derivative work or partial copy which includes the elements in question.

All copyright, and all rights therein, are protected by national and international copyright laws. The above represents a summary only. For further information please read Frontiers' Conditions for Website Use and Copyright Statement, and the applicable CC-BY licence.

ISSN 1664-8714 ISBN 978-2-88963-998-4 DOI 10.3389/978-2-88963-998-4

#### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

#### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

#### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

#### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# INTERACTIONS BETWEEN DIETS, GUT MICROBIOTA AND HOST METABOLISM

Topic Editors: Jie Yin, Institute of Subtropical Agriculture, China Liwei Xie, Guangdong Institute of Microbiology, China Yuheng Luo, Sichuan Agricultural University, China Helieh S. Oz, University of Kentucky, United States

Citation: Yin, J., Xie, L., Luo, Y., Oz, H. S., eds. (2020). Interactions Between Diets, Gut Microbiota and Host Metabolism. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-998-4

# Table of Contents

*08 Editorial: Diets, Gut Microbiota, and Host Metabolism* Jie Yin, Liwei Xie, Yuheng Luo and Helieh S. Oz *10 Jejunal Metabolic Responses to* Escherichia coli *Infection in Piglets* Hucong Wu, Jiaqi Liu, Siyuan Chen, Yuanyuan Zhao, Sijing Zeng, Peng Bin, Dong Zhang, Zhiyi Tang and Guoqiang Zhu *19 Effects of Probiotic* Bacillus *as an Alternative of Antibiotics on Digestive* 

*Enzymes Activity and Intestinal Integrity of Piglets* Shenglan Hu, Xuefang Cao, Yanping Wu, Xiaoqiang Mei, Han Xu, Yang Wang, Xiaoping Zhang, Li Gong and Weifen Li

*28 Rutin and Its Combination With Inulin Attenuate Gut Dysbiosis, the Inflammatory Status and Endoplasmic Reticulum Stress in Paneth Cells of Obese Mice Induced by High-Fat Diet*

Xiulan Guo, Renyong Tang, Shiyong Yang, Yurong Lu, Jing Luo and Zhenhua Liu

*39 Serine Alleviates Dextran Sulfate Sodium-Induced Colitis and Regulates the Gut Microbiota in Mice*

Haiwen Zhang, Rui Hua, Bingxi Zhang, Xiaomeng Zhang, Hui Yang and Xihong Zhou

*49 Anti-breast Cancer Enhancement of a Polysaccharide From Spore of*  Ganoderma lucidum *With Paclitaxel: Suppression on Tumor Metabolism With Gut Microbiota Reshaping*

Jiyan Su, Dan Li, Qianjun Chen, Muxia Li, Lu Su, Ting Luo, Danling Liang, Guoxiao Lai, Ou Shuai, Chunwei Jiao, Qingping Wu, Yizhen Xie and Xinxin Zhou

*63 Corrigendum: Anti-breast Cancer Enhancement of a Polysaccharide From Spore of* Ganoderma lucidum *With Paclitaxel: Suppression on Tumor Metabolism With Gut Microbiota Reshaping*

Jiyan Su, Dan Li, Qianjun Chen, Muxia Li, Lu Su, Ting Luo, Danling Liang, Guoxiao Lai, Ou Shuai, Chunwei Jiao, Qingping Wu, Yizhen Xie and Xinxin Zhou


Daniel N. Villageliú, David J. Borts and Mark Lyte

*87 Dietary Supplementation With Chinese Herbal Residues or Their Fermented Products Modifies the Colonic Microbiota, Bacterial Metabolites, and Expression of Genes Related to Colon Barrier Function in Weaned Piglets*

Jiayi Su, Qian Zhu, Yue Zhao, Li Han, Yulong Yin, Francois Blachier, Zhanbin Wang and Xiangfeng Kong


Zongxin Ling, Xia Liu, Shu Guo, Yiwen Cheng, Li Shao, Dexiu Guan, Xiaoshuang Cui, Mingming Yang and Xiwei Xu

*129 Intestinal Morphologic and Microbiota Responses to Dietary* Bacillus *spp. in a Broiler Chicken Model*

Cheng-liang Li, Jing Wang, Hai-jun Zhang, Shu-geng Wu, Qian-ru Hui, Cheng-bo Yang, Re-jun Fang and Guang-hai Qi


Yuanyuan Zhang, Aili Dong, Keliang Xie and Yonghao Yu


Feifei Zhang, Tong Ma, Peng Cui, Amin Tamadon, Shan He, Chuanbing Huo, Gulinazi Yierfulati, Xiaoqing Xu, Wei Hu, Xin Li, Linus R. Shao, Hongwei Guo, Yi Feng and Congjian Xu

*201 Matrine Mediates Inflammatory Response via Gut Microbiota in TNBS-Induced Murine Colitis*

Peiyuan Li, Jiajun Lei, Guangsheng Hu, Xuanmin Chen, Zhifeng Liu and Jing Yang

*208 Intestinal Bacteria Interplay With Bile and Cholesterol Metabolism: Implications on Host Physiology*

> Natalia Molinero, Lorena Ruiz, Borja Sánchez, Abelardo Margolles and Susana Delgado

*218 Dose-Dependent Effects of Aloin on the Intestinal Bacterial Community Structure, Short Chain Fatty Acids Metabolism and Intestinal Epithelial Cell Permeability*

Kuppan Gokulan, Pranav Kolluru, Carl E. Cerniglia and Sangeeta Khare


Yong Ma, Sujuan Ding, Gang Liu, Jun Fang, Wenxin Yan, Veeramuthu Duraipandiyan, Naif Abdullah Al-Dhabi, Galal Ali Esmail and Hongmei Jiang

*259* Astragalus *and Ginseng Polysaccharides Improve Developmental, Intestinal Morphological, and Immune Functional Characters of Weaned Piglets*

C. M. Yang, Q. J. Han, K. L. Wang, Y. L. Xu, J. H. Lan and G. T. Cao


Pengfei Yu, Shubo Yu, Juan Wang, Hui Guo, Ying Zhang, Xiyu Liao, Junhui Zhang, Shi Wu, Qihui Gu, Liang Xue, Haiyan Zeng, Rui Pang, Tao Lei, Jumei Zhang, Qingping Wu and Yu Ding

*300 Corrigendum:* Bacillus cereus *Isolated From Vegetables in China: Incidence, Genetic Diversity, Virulence Genes, and Antimicrobial Resistance*

Pengfei Yu, Shubo Yu, Juan Wang, Hui Guo, Ying Zhang, Xiyu Liao, Junhui Zhang, Shi Wu, Qihui Gu, Liang Xue, Haiyan Zeng, Rui Pang, Tao Lei, Jumei Zhang, Qingping Wu and Yu Ding

*302 Dietary Quercetin Increases Colonic Microbial Diversity and Attenuates Colitis Severity in* Citrobacter rodentium*-Infected Mice*

Rui Lin, Meiyu Piao and Yan Song

*310 IgA-Targeted* Lactobacillus jensenii *Modulated Gut Barrier and Microbiota in High-Fat Diet-Fed Mice* Jin Sun, Ce Qi, Hualing Zhu, Qin Zhou, Hang Xiao, Guowei Le,

Daozhen Chen and Renqiang Yu

*323 Microbiome-Metabolomics Analysis Investigating the Impacts of Dietary Starch Types on the Composition and Metabolism of Colonic Microbiota in Finishing Pigs*

Miao Yu, Zhenming Li, Weidong Chen, Ting Rong, Gang Wang and Xianyong Ma


Liang Gong, Haocheng He, Dongjie Li, Lina Cao, Tahir Ali Khan, Yanping Li, Lifei Pan, Liang Yan, Xuezhi Ding, Yunjun Sun, Youming Zhang, Ganfeng Yi, Shengbiao Hu and Liqiu Xia

*378 Probiotic Properties of Lactic Acid Bacteria Isolated From Neera: A Naturally Fermenting Coconut Palm Nectar*

Rakesh Somashekaraiah, B. Shruthi, B. V. Deepthi and M. Y. Sreenivasa

*389 Pinocembrin Protects Against Dextran Sulfate Sodium-Induced Rats Colitis by Ameliorating Inflammation, Improving Barrier Function and Modulating Gut Microbiota*

Lin Hu, Chao Wu, Zijian Zhang, Mingchang Liu, E. Maruthi Prasad, Yu Chen and Kai Wang

*399 Probiotic Properties of* Lactobacillus paracasei *subsp. paracasei L1 and Its Growth Performance-Promotion in Chicken by Improving the Intestinal Microflora*

Yunhe Xu, Yuan Tian, Yunfang Cao, Jianguo Li, Haonan Guo, Yuhong Su, Yumin Tian, Cheng Wang, Tianqi Wang and Lili Zhang


Wang Zhang, Dakai Gan, Jie Jian, Chenkai Huang, Fangyun Luo, Sizhe Wan, Meichun Jiang, Yipeng Wan, Anjiang Wang, Bimin Li and Xuan Zhu

*437 Gut Microbiota as an Objective Measurement for Auxiliary Diagnosis of Insomnia Disorder*

Bingdong Liu, Weifeng Lin, Shujie Chen, Ting Xiang, Yifan Yang, Yulong Yin, Guohuan Xu, Zhihong Liu, Li Liu, Jiyang Pan and Liwei Xie

*449 Corrigendum: Gut Microbiota as an Objective Measurement for Auxiliary Diagnosis of Insomnia Disorder*

Bingdong Liu, Weifeng Lin, Shujie Chen, Ting Xiang, Yifan Yang, Yulong Yin, Guohuan Xu, Zhihong Liu, Li Liu, Jiyang Pan and Liwei Xie

*450 Dietary Supplementation With Leucine or in Combination With Arginine Decreases Body Fat Weight and Alters Gut Microbiota Composition in Finishing Pigs*

Chengjun Hu, Fengna Li, Yehui Duan, Yulong Yin and Xiangfeng Kong

*462 Adhesive* Bifidobacterium *Induced Changes in Cecal Microbiome Alleviated Constipation in Mice*

Linlin Wang, Cailing Chen, Shumao Cui, Yuan-kun Lee, Gang Wang, Jianxin Zhao, Hao Zhang and Wei Chen


Sujuan Ding, Yong Ma, Gang Liu, Wenxin Yan, Hongmei Jiang and Jun Fang


Hong Shen, Zhihui Xu, Zanming Shen and Zhongyan Lu


Lujie Fan, Xiao Li, Jianhong Hu, Gongshe Yang and Xin'e Shi


Ryan C. McDonald, Joy E. M. Watts and Harold J. Schreier


# Editorial: Diets, Gut Microbiota, and Host Metabolism

#### Jie Yin<sup>1</sup> \*, Liwei Xie<sup>2</sup> , Yuheng Luo<sup>3</sup> and Helieh S. Oz <sup>4</sup>

*<sup>1</sup> College of Animal Science and Technology, Hunan Agriculture University, Changsha, China, <sup>2</sup> State Key Laboratory of Applied Microbiology Southern China, Guangdong Provincial Key Laboratory of Microbial Culture Collection and Application, Guangdong Open Laboratory of Applied Microbiology, Guangdong Institute of Microbiology, Guangdong Academy of Sciences, Guangdong, China, <sup>3</sup> Institute of Animal Nutrition, Sichuan Agricultural University, Chengdu, China, <sup>4</sup> Department of Medicine, University of Kentucky Medical Center, Lexington, KY, United States*

Keywords: diets, gut microbiota, host metabolism, dysbiosis, metagenomics, metabolomics

#### **Editorial on the Research Topic**

#### **Interactions Between Diets, Gut Microbiota and Host Metabolism**

Human gut hosts ∼100 trillion bacteria, which eventually fall into about 500–1,000 bacterial species. Microbial density and diversity are established very rapidly in the gut of the newborn and drive stabilization of the normal commensal microbiota. However, the gut microbiota diversity and compositions are affected by genetic background, but most importantly by diets and environment (1–3). Wang et al. explored the relationships among diet supplements, gut microbiota, host genetics and metabolic status, leading to the conclusion that diets more intensively disturbed the structure of gut microbiota in excess of genetic change, particularly under leptin deficient conditions. Protein is a major dietary nutrient and Li et al. systemically analyzed the colonic microbiota and metabolic responses to different dietary protein sources in a piglet model. Hu et al. further indicated a role of amino acids in gut microbiota compositions. Starch acts as a major energy source of the daily diet and is the largest fraction among human and monogastric animal diets. Yu et al. demonstrated that the different dietary starch types treatment altered the intestinal microbiota and metabolite profiles of the pigs, and dietary with higher amylose may offer potential benefits for gut health. In an amazonian catfish Panaque nigrolineatus model, McDonald et al. discussed the microbiome and predictive metagenomic profiles in response to dietary a wood alone or a less refractory mixed diet containing wood and plant nutrition and a marked change was noticed in the enteric bacterial community composition. Besides, rutin, polysaccharide, matrine, fiber, quercetin, ahlorogenic acid, aloin, pinocembrin, xylooligosaccharide, and other active components have been reported to shape gut microbiota compositions in this Research Topic.

#### Edited by:

*Giovanna Suzzi, University of Teramo, Italy*

Reviewed by: *Simone Guglielmetti, University of Milan, Italy*

> \*Correspondence: *Jie Yin yinjie2014@126.com*

#### Specialty section:

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Nutrition*

Received: *29 May 2020* Accepted: *12 June 2020* Published: *22 July 2020*

#### Citation:

*Yin J, Xie L, Luo Y and Oz HS (2020) Editorial: Diets, Gut Microbiota, and Host Metabolism. Front. Nutr. 7:108. doi: 10.3389/fnut.2020.00108*

Gut microbiota is also changed in various pathologic conditions and has been considered as an objective measurement for some diseases (4–6). For example, Liu et al. provided a comprehensive understanding of the link between the gut microbiota and insomnia disorder and constructed a prediction model utilizing artificial neural network, which might be used for auxiliary diagnosis of insomnia disorder. Meanwhile, the potential relationships between gut microbiota and childhood leukemia also have been discussed by Wen et al., which may help build a healthy gut microbiota by adjusting the diet construction to deal with childhood leukemia. Thus, manipulation of gut microbiota has been considered a potential target for treating diseases.

Probiotics are live microorganisms that confer health benefits on the host via an improvement of gut microbiota compositions, and metabolism, or directly interacting with host (7, 8). In this Research Topic, several studies focused on the dietary benefits of probiotics. A clinic study showed that co-administration of probiotics with azithromycin is a promising therapy for Mycoplasma pneumoniae pneumonia treatment which could prevent and treat antibiotic-associated diarrhea

**8**

effectively (Liang et al.). In the dextran sulfate sodium-induced colitis model, Lactobacillus brevis and Bacillus amyloliquefaciens are reported to alleviate colonic damage by reprograming gut microbiota ()Diang et al.; Cao et al.. Constipation is a common gastrointestinal disease, which is characterized by gut dysbiosis. However, Bifidobacterium colonization can alleviate constipation more efficiently by improving the water, propionate and butyrate content in feces, and overall gastrointestinal transit time (Wang et al.). In animal industry, probiotics are widely introduced to improve health and growth. Xu et al. reported that Lactobacillus paracasei subsp. paracasei L1 possesses probiotic properties (i.e., adhesion, aggregation, hydrophobicity, as well as survival upon exposure to various gastrointestinal conditions, and lack hemolytic and decarboxylation activities) and improves the growth performance in chickens. In weaning piglets, dietary Lactobacillus plantarum PFM 105 promotes intestinal development (intestinal villi) through modulation of gut microbiota and metabolism (Wang et al.). Similarly, Hu et al. probiotic Bacillus amyloliquefaciens has been indicated as an alternative of antibiotics on digestive enzymes activity and intestinal integrity of piglets.

We also received several articles about probiotic isolation. Somashekaraiah et al. isolated seven potential of lactic acid bacteria with the best probiotic attributes from the sap extract of the coconut palm inflorescence-Neera. Gong et al. also isolated a new Pediococcus pentosaceus strain (SL001) from soil samples and found it can be used as a dietary probiotic in freshwater fish aquaculture by enhancing immunity and promoter growth rate of grass carps.

#### REFERENCES


It is clear from these papers in this Research Topic that the field of diet-gut microbiota-host metabolism interplay has many avenues to explore. The impact of dietary nutrients (i.e., protein, starch, fiber, amino acids, probiotics, and other active substances) on the gut microbiome is currently an area of intense investigation and hopefully will lead a detailed understanding of how specific nutrient can influence the ecology of the gut microenvironment to produce metabolic benefits. Meanwhile, gut microbiota dysbiosis is associated with various kinds of diseases, thus manipulation of gut ecology has been widely incorporated into clinical methods. Bacterial metabolites, microRNA, bacteriocin, and microbiota sensing pathways are increasingly recognized to be involved in the diet-gut microbiota-host metabolism interplay, while the detailed mechanisms in different models are far from clear.

#### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

#### FUNDING

This study was supported by the Young Elite Scientists Sponsorship Program by CAST (2019-2021QNRC001) and Guangdong Basic and Applied Basic Research Foundation (2020B1515020046).


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Yin, Xie, Luo and Oz. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Jejunal Metabolic Responses to Escherichia coli Infection in Piglets

Hucong Wu<sup>1</sup>† , Jiaqi Liu<sup>1</sup>† , Siyuan Chen<sup>2</sup> , Yuanyuan Zhao<sup>2</sup> , Sijing Zeng<sup>2</sup> , Peng Bin<sup>2</sup> , Dong Zhang<sup>1</sup> , Zhiyi Tang<sup>2</sup> \* and Guoqiang Zhu<sup>1</sup> \*

<sup>1</sup> College of Veterinary Medicine, Jiangsu Co-Innovation Center for Important Animal Infectious Diseases and Zoonoses, Joint International Research Laboratory of Agriculture and Agri-Product Safety of Ministry of Education of China, Yangzhou University, Yangzhou, China, <sup>2</sup> Guangdong Provincial Key Laboratory of Animal Nutrition Control, Institute of Subtropical Animal Nutrition and Feed, College of Animal Science, South China Agricultural University, Guangzhou, China

#### Edited by:

Jie Yin, Institute of Subtropical Agriculture (CAS), China

#### Reviewed by:

Luchang Zhu, Houston Methodist Research Institute, United States Yongjie Liu, Nanjing Agricultural University, China Charles Z. Li, Agricultural Research Service (USDA), United States

#### \*Correspondence:

Zhiyi Tang 43591680@qq.com Guoqiang Zhu yzgqzhu@yzu.edu.cn †These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 12 July 2018 Accepted: 26 September 2018 Published: 16 October 2018

#### Citation:

Wu H, Liu J, Chen S, Zhao Y, Zeng S, Bin P, Zhang D, Tang Z and Zhu G (2018) Jejunal Metabolic Responses to Escherichia coli Infection in Piglets. Front. Microbiol. 9:2465. doi: 10.3389/fmicb.2018.02465 This study aimed to investigate the jejunal metabolic variations in enterotoxigenic Escherichia coli (ETEC)-infected piglets. Piglets were infected with 1 × 10<sup>10</sup> CFUs (colony-forming units) of ETEC W25K and assigned into diarrheal, recovered, control, and resistant groups. Jejunal samples were harvested at day 6 and metabolic profiles were analyzed via gas chromatography coupled to time-of-flight mass spectrometry (GC/TOFMS). The results showed that 33 metabolites in the jejunum were identified in ETEC-induced diarrhea, including amino acids, fatty acids, sugars, and organic acids. Compared with the control, resistant, and recovered piglets, diarrheal piglets showed higher concentrations of 4-aminobutyric acid (GABA) and glycine in the jejunum. Compared with the control and resistant piglets, six metabolites were markedly decreased in diarrheal piglets, including ornithine, asparagine, glutamine, citric acid, citrulline, and lysine. Collectively, this study provides insights into jejunal metabolic response to ETEC infection and ETEC induced diarrhea in piglets.

Keywords: jejunum, metabolism, ETEC, diarrhea, piglet

#### INTRODUCTION

Diarrheal illnesses are a severe public health problem and pathogenic enterotoxigenic Escherichia coli (ETEC) has been considered as a major cause of diarrhea in human and animals (Fleckenstein et al., 2010). After infection, ETEC rapidly colonizes in small intestine, including duodenum, jejunum, and ileum. ETEC colonization inhibits intestinal immune function and induces inflammatory response. In our previous report, we found that ETEC infection inhibits the mRNA expression of intestinal immune factors, such as polymeric immunoglobulin receptor (pIgR), cryptdin-related sequence 1C (CRS1C), and Reg3γ in mice (Liu et al., 2017). Meanwhile, ETEC infection upregulates intestinal IL-17 and causes dysbiosis of intestinal microbiota via increasing abundance of γ-aminobutyric acid (GABA)-producing Lactococcus lactis subsp. lactis (Ren et al., 2016b). The jejunal metabolite (e.g., amino acids and polyamine) participate in many important physiological process, such as the regulation of gene expression, synthesis and secretion of hormones, oxidative defense, and so on (Wu, 2009). The proteome analysis from our previous study identifies 92 differentially expressed proteins in the jejunum after exposure to ETEC and large body of these proteins were involved in metabolic process, such as protein turnover, nutrients (i.e., nucleotide, amino acids, carbohydrate, lipid, and inorganic ion) transport and metabolism, coenzyme metabolism, energy production and conversion, and secondary metabolite biosynthesis (Ren et al., 2016a). Metabolomics is an emerging analytical technique to seek global profiles of

metabolites in particular samples, including endogenous and exogenous metabolites (Khan et al., 2017). Therefore, we conduct this study to further investigate metabolic profiles in the jejunum after ETEC infection in piglets.

# MATERIALS AND METHODS

# Bacterial Strains

This study used the Escherichia coli F4-producing strain W25K (O149:K91, K88ac; LT, ST, EAST), which was originally isolated from a diarrheal piglet.

#### ETEC Infection

This study was conducted according to the guidelines of the Institute of Subtropical Agriculture, Chinese Academy of Sciences. Piglets (Landrace × Yorkshire; 18-day-old) were purchased from ZhengDa, Co., Chongqing, China and orally administrated with ETEC W25K at dose of 1 × 10<sup>10</sup> CFUs (colony-forming units) for five consecutive days (Ren et al., 2015, 2016a). The control piglets were treated with the same volume of LB medium. Fecal consistency was scored daily as: 0 = normal; 1 = soft; 2 = runny or watery. Piglets with the development of watery diarrhea were defined as diarrheal piglets, and piglets that were recovered from diarrhea were regarded as recovery piglets, while piglets that were challenged with ETEC but not suffered from diarrhea were defined as resistant piglets. Six control piglets, six diarrheal piglets, six recovered piglets, and six resistant piglets were randomly selected for collecting the samples.

# Sample Preparation

Twenty-four piglets were sacrificed at day 6 after ETEC infection and jejunal samples (100 mg) and extraction solvents (50 µL L-2-chlorophenylalanine and 350 µL methanol) were added and then homogenized using a Mini-BeadBeater-16 (Biospec, Co., Bartlesville, OK, United States) for 5 min. The mixture was placed on a shaker at 70◦C for 10 min and centrifuged at 12,000 × g and 4 ◦C for 10 min. The supernatant was separated, transferred into a GC vial, and then evaporated to dryness under a stream of N<sup>2</sup> gas.

Methoxyamine hydrochloride (20 µL, 20 mg/mL pyridine) was added to the dried fraction and incubated at 37◦C for 2 h. One hundred µL of bis-(trimethylsilyl) trifluoroacetamide (BSTFA) containing 1% TMCS was rapidly added and incubated

Permutation validation of PLS-DA model, blue dots and green dots represented Q<sup>2</sup> and R<sup>2</sup> , respectively; (D) diarrheal piglets vs. control piglets; (E) diarrheal piglets vs. resistant piglets; and (F) diarrheal piglets vs. recovered piglets.

at 70◦C for 1 h. Then the samples were kept at room temperature before analysis.

# GC-TOFMS Analysis

Metabolites in jejunal samples were derivatized prior to gas chromatography coupled to time-of-flight mass spectrometry (GC-TOFMS) analysis (Agilent 7890A, Agilent, United States; LECO Chroma TOF PEGASUS 4D, MI, LECO, United States). The system utilized a DB-5MS capillary column coated with 5% diphenyl cross-linked with 95% dimethylpolysiloxane (30 m × 250 µm inner diameter, 0.25 µm film thickness; J&W Scientific, Folsom, CA, United States). A 1 µL aliquot of the analyte was injected in splitless mode. Helium was used as the carrier gas, the front inlet purge flow was 3 mL min−<sup>1</sup> , and the gas flow rate through the column was 1 mL min−<sup>1</sup> . The initial temperature was kept at 90◦C for 0.25 min, then raised to 240◦C at a rate of 5◦C min−<sup>1</sup> , and finally to 285◦C at a rate of 20◦C min−<sup>1</sup> for 11.5 min. The injection, transfer line, and ion source temperatures were 280, 250 and 220◦C, respectively. The energy was −70 eV in electron impact mode. The mass spectrometry data were acquired in full-scan mode with the m/z range of 20–600

at a rate of 100 spectra per second after a solvent delay of 492 s.

#### Data Processing and Analysis

Each sample was represented by a GC-TOFMS chromatograph. The GC-TOFMS raw data were processed by Chroma TOF 4.3X software (LECO Corporation, St. Joseph, MI, United States) and LECO-Fiehn Rtx5 database for raw peaks extracting, data baselines filtering and calibration, peak alignment, deconvolution analysis, peak identification, and peak area integration. All the output data exported from Chroma TOF 4.3X software were imported into SIMCA-P software (version 11.0, Umetrics, Umeå, Sweden) for multivariate statistical analyses including a principal component analysis (PCA), partial least squares-discriminant analysis (PLS-DA), and pairwise orthogonal projections to latent structures discriminant analyses (OPLS-DA).

# RESULTS

#### PCA Model Analysis

Principal component analysis is an unsupervised mathematical procedure used to identify latent structures in the dataset and outliers (Caboni et al., 2014). PCA of jejunal samples from diarrheal piglets, recovered piglets, control piglets, and resistant piglets was shown in **Figure 1A**. The results showed that plots from diarrheal piglets, recovered piglets, control piglets, and resistant piglets were separated each other. As the points that were close to each other had similar metabolic profiles, our results indicated that there might be significant metabolic differences among the four groups. The modeling of the three datasets (diarrheal piglets vs. control piglets, diarrheal piglets vs. resistant piglets, and diarrheal piglets vs. recovered piglets) of separate pairs, revealed separation between subjects (**Figures 1B–D**).

# PLS-DA Model Analysis

To specify the metabolic variations produced by ETEC infection, PLS-DA models were constructed in jejunal samples (**Figure 2**). The results showed that the samples from each group were perfectly separated in three subjects: diarrheal piglets vs. control piglets (R2X = 0.90, Q<sup>2</sup> = 0.30), diarrheal piglets vs. resistant piglets (R2X = 0.73, Q<sup>2</sup> = 0.38) and diarrheal piglets vs. recovered piglets (R2X = 0.29, Q<sup>2</sup> = −0.16). This phenomenon indicated that the physiological metabolism was interrupted by ETEC infection. In addition, diarrheal piglets showed distinctive metabolic profiles compared with piglets that recovered from diarrhea and were resistant to ETEC infection.

# OPLS-DA Model Analysis

The variable importance in the projection (VIP) statistic of the first principal component of orthogonal partial least squares discriminant analysis (OPLS-DA) model (threshold > 1) coupled with the P-value of the Student's t-test (threshold < 0.05) were used for selecting significant variables responsible for group separation.

As shown in **Figures 3**, **4**, the OPLS-DA models showed a clear separation between the diarrheal piglets vs. control piglets (R2X = 0.41, R2Y = 0.96, Q<sup>2</sup> = 0.77), diarrheal piglets vs. resistant piglets (R2X = 0.45, R2Y = 0.97, Q<sup>2</sup> = 0.82) and diarrheal

piglets vs. recovered (R2X = 0.32, R2Y = 0.98, Q<sup>2</sup> = 0.62). We detected 26, 33, and 14 differential metabolites between diarrheal piglets vs. control piglets, diarrheal piglets vs. resistant piglets and diarrheal piglets vs. recovered piglets, respectively. However, only three differential metabolites were commonly altered (including increase and decrease) among them.

Compared with the control piglets, diarrheal piglets showed higher concentration of nine metabolites in the jejunum [4-aminobutyric acid (GABA), glycine, 8-aminocaprylic acid, taurine, 5-methoxytryptamine, lactamide, isocitric acid, L-threose, and malonic acid]. However, 17 metabolites showed a decreased trend in the diarrheal piglets (2-hydroxybutanoic acid, L-allothreonine, 2-amino-1-phenylethanol, methionine, ornithine, lauric acid, asparagine, glutamine, O-phosphorylethanolamine, citric acid, citrulline, lysine, tyrosine, myo-inositol, stearic acid, spermidine and arachidonic acid) (**Table 1**).

Compared with the resistant piglets, nine metabolites were significantly enhanced in the diarrheal piglets (GABA, glycine, pyruvic acid, lactic acid, ethanolamine, creatine, 8-aminocaprylic

TABLE 1 | The variation in content of metabolites in the jejunum between diarrheal and control piglets.


<sup>∗</sup>The extremely significant metabolites between diarrheal and control piglets, which P-value < 0.001.

acid, taurine, and noradrenaline), while 24 metabolites were significantly enhanced in the diarrheal piglets (beta-alanine, glutamine, L-malic acid, alanine, ornithine, glutamic acid, asparagine, lyxose, glucose-1-phosphate, citric acid, gluconic lactone, citrulline, fructose, sorbose, mannose, lysine, sorbitol, L-threose, spermidine, malonic acid, diglycerol, inosine, uridine monophosphate, and lactobionic acid) (**Table 2**).

Compared with the recovered piglets, 14 metabolites were significantly different in diarrheal piglets, and 8 metabolites were increased in the jejunum (GABA, glycine, glycolic acid, Dglyceric acid, xylitol, glucose, cis-gondoic acid, and malonic acid). Meanwhile, six metabolites were decreased in diarrheal piglets (methyl phosphate, fumaric acid, alanine, inosine, adenosine, and uridine monophosphate) (**Table 3**).

#### DISCUSSION

Infection with ETEC bacteria is the major cause of diarrhea in human and animals. After infection, ETEC rapidly colonizes the intestine and secretes exotoxins, which further disrupt intestinal barrier integrity and cause secretory diarrhea (Deng et al., 2015). In addition, ETEC colonization induces imbalance of intestinal microbiota and may dysregulate intestinal metabolism (Ren et al., 2016b). In this study, 33 metabolites have been identified in ETEC induced diarrhea, including amino acids, fatty acids, sugars, and organic acids.

Compared with the control, resistant and recovered piglets, diarrheal piglets have higher concentrations of GABA and glycine in the jejunum. GABA, a transmitter of enteric interneurons, has been noticed in the cytoplasm and the brush border of intestinal epithelial cells and regulates the function of the gastrointestinal tract (Wang et al., 2004; Li et al., 2012; Jung et al., 2017). The direct functions of intestinal GABAergic signaling system have been identified to be involved in fluid transport through luminal secretion of Cl<sup>−</sup> (Jin et al., 2006), which is a major driving force for fluid secretion and increased during diarrhea. Similar to our results in piglets, Li et al. (2012) reported that intestinal GABAergic signaling was upregulated in diarrheal mice caused by ovalbumin and blocking this GABA signaling decreased the occurrence of allergic diarrhea. In our previous study, we found that ETEC infection increased GABA-producing L. lactis subsp. lactis and GABA production, which further promotes IL-17 expression through mTORC1–S6K1–EGR-2–GFI-1 pathway and mediates intestinal inflammation (Ren et al., 2016b). Glycine serves as a precursor for glutathione along with cysteine and glutamic acid (Yin et al., 2016), while cysteine, glutamic acid and glutamine were markedly decreased in diarrheal piglets, suggesting that increased glycine failed to contribute to the glutathione synthesis and antioxidant effect in ETEC model. In intestinal injury, even after manifestation of a severe systemic

TABLE 2 | The variation in content of metabolites in the jejunum between diarrheal and resistant piglets.


<sup>∗</sup>The extremely significant metabolites between diarrheal and resistant piglets, which P-value < 0.001.

TABLE 3 | The variation in content of metabolites in the jejunum between diarrheal and recovered piglets.


impairment (Effenberger-Neidnicht et al., 2014) and causing pHdependent membrane damage to ETEC (Vanhauteghem et al., 2012), glycine has been demonstrated to protect intestine against subacute endotoxemia. Meanwhile, the glycine cleavage system contributes to the intracellular replication of virulent bacterium and pathogenesis (Brown et al., 2014). Thus, the increased GABA and glycine in the jejunum may mediate or promote diarrhea in ETEC infectious piglet model.

Compared with the control and resistant piglets, we found that six metabolites were markedly decreased in diarrheal piglets, including ornithine, asparagine, glutamine, citric acid, citrulline, and lysine. Ornithine, asparagine, and citrulline play roles in the urea cycle (Mikulski et al., 2015), thus decreased ornithine and asparagine may indicate urea cycle is altered in ETEC induced diarrhea. Aspartate, a precursor of asparagine, has been demonstrated to enhance intestinal integrity and energy status in weaning piglets after lipopolysaccharide challenge (Pi et al., 2014) and alleviate diquat-induced intestinal oxidative stress (Yin et al., 2015). Thus, the decreased jejunal asparagine in diarrheal piglets may exacerbate ETEC infection. Glutamine has been indicated to modulate intestinal permeability and tight junction expression in various diseases, and serves as a protective mechanism against radiation-induced diarrhea and diarrheapredominant irritable bowel syndrome (Kucuktulu et al., 2013; Bertrand et al., 2015). Lower jejunal glutamine in diarrheal piglets suggested that ETEC infection influenced glutamine synthesis, which may further disturbs the protective mechanism of glutamine against diarrhea. Citrate is an intermediate in the tricarboxylic acid cycle (TCA) and the reduced citrate in the diarrheal piglets suggested that the TCA was altered after ETEC infection.

Out of our expectation, we found that the lactic acid in diarrheal piglets was much higher than resistant piglets as a 1498647 fold change. Lactic acid was the metabolite fermented by various intestinal microorganisms (e.g., Klebsiella, Bacteroides, Lactobacillus), and presented in the intestine lumen (Zhao et al., 2011). As mentioned above, ETEC infection increased the relative abundance of lactic acid-producing bacteria (L. lactis subsp. lactis) significantly, thus the concentration of lactic acid in jejunum lumen was increased. But it failed to explain why lactic acid in jejunum tissue of diarrheal piglets was much higher than resistant piglets. In the healthy condition, the intestine

#### REFERENCES


of mammals was incapable of absorbing lactic acid, while the intestinal permeability might be changed in some pathological conditions, and the lactic acid was permitted to permeate the intestinal mucosa, thus the lactic acid may be regarded as an important indictor response to the intestinal barrier function (Qiu et al., 2013; Ikuta et al., 2017). We concluded that ETEC infection might impair the intestinal barrier of piglets and increase the permeability of jejunum (Yang et al., 2014), thus resulted in the high concentration of lactic acid in jejunum tissue.

#### ETHICS STATEMENT

All experimental protocols were approved by the Animal Care and Use Committee of Yangzhou University [approval ID: SYXK (Su) 2005-0005]. All animal care and use protocols in this study were performed in accordance with the approved current guidelines. At the end of the experimental period, all of the surviving or sick piglets were euthanized by potassium chloride.

#### AUTHOR CONTRIBUTIONS

GZ conceived and designed the study. HW, JL, SC, YZ, SZ, PB, and DZ conducted experiments. HW and JL carried out animal experiments, performed statistical analysis, and drafted the manuscript. GZ finalized the manuscript. ZT conceived and designed the study and finalized the manuscript. All authors read and approved the final version of the manuscript.

# FUNDING

This study was supported by National Basic Research Program of China (Grant No. 2013CB127301), National Natural Science Foundation of China (Grant Nos. 31272463 and 31472106), and Hunan Provincial Natural Science Foundation of China (Grant No. 12JJ2014). It was also funded by the Priority Academic Program of Development Jiangsu High Education Institution, and Innovation and Training Program of Jiangsu High Education Institution (201711117089X).



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Wu, Liu, Chen, Zhao, Zeng, Bin, Zhang, Tang and Zhu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Effects of Probiotic Bacillus as an Alternative of Antibiotics on Digestive Enzymes Activity and Intestinal Integrity of Piglets

Shenglan Hu1,2† , Xuefang Cao<sup>1</sup>† , Yanping Wu<sup>1</sup> , Xiaoqiang Mei<sup>1</sup> , Han Xu<sup>1</sup> , Yang Wang<sup>1</sup> , Xiaoping Zhang<sup>3</sup> , Li Gong<sup>1</sup> \* and Weifen Li<sup>1</sup> \*

<sup>1</sup> Key Laboratory of Molecular Animal Nutrition and Feed Sciences, College of Animal Science, Zhejiang University, Hangzhou, China, <sup>2</sup> State Key Laboratory of Livestock and Poultry Breeding, Key Laboratory of Animal Nutrition and Feed Science in South China, Ministry of Agriculture, Institute of Animal Science, Guangdong Academy of Agricultural Sciences, Guangzhou, China, <sup>3</sup> Key Laboratory of Resources and Utilization of Bamboo of State Forestry Administration, China National Bamboo Research Center, Hangzhou, China

#### Edited by:

Jie Yin, Institute of Subtropical Agriculture (CAS), China

#### Reviewed by:

Yiannis Kourkoutas, Democritus University of Thrace, Greece Yanchun Shao, Huazhong Agricultural University, China

#### \*Correspondence:

Li Gong 11617002@zju.edu.cn Weifen Li wfli@zju.edu.cn †These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 27 July 2018 Accepted: 21 September 2018 Published: 22 October 2018

#### Citation:

Hu S, Cao X, Wu Y, Mei X, Xu H, Wang Y, Zhang X, Gong L and Li W (2018) Effects of Probiotic Bacillus as an Alternative of Antibiotics on Digestive Enzymes Activity and Intestinal Integrity of Piglets. Front. Microbiol. 9:2427. doi: 10.3389/fmicb.2018.02427 The previous study in our team found that supplementation of probiotic Bacillus amyloliquefaciens (Ba) instead of antibiotics promote growth performance of piglets. Hence, the present study was carried out to further demonstrate the effect of Ba replacement of antibiotics on digestive and absorption enzyme activity and intestinal microbiota population of piglets. A total of 90 piglets were selected and divided into three groups: G1 group was fed with basal diet supplemented with 150 mg/Kg aureomycin, G2 group was fed with 1 × 10<sup>8</sup> cfu/Kg Ba and half dose of aureomycin, G3 group was used the diet with 2 × 10<sup>8</sup> cfu/Kg Ba replaced aureomycin. Each treatment had three replications of 10 pigs per pen. Results indicated that Ba replacement significantly increased the activities of amylase, disaccharides and Na+/K+-ATPase. And chymotrypsin activity in different section of intestine was dramatically enhanced in half replacement of aureomycin with Ba. Moreover, Ba replacement maintained the intestinal integrity with the significantly decreased activity of DAO compared with aureomycin group. Besides, supplementation with Ba increased the β-diversity of intestinal microbiota. Taken together, the current study indicated that diet supplementation with Ba instead of aureomycin increased the growth performance of piglets by improving the digestive and absorb enzyme activities, enhancing the intestinal integrity and regulating the population of intestinal micrbiota.

Keywords: piglets, antibiotics, Bacillus amyloliquefaciens, digestive enzyme activity, intestinal integrity

# INTRODUCTION

Antibiotics have long been used to promote the growth and health of piglets (van den Bogaard and Stobberingh, 2000). However, with the increasing phenomena of bacterial resistance and antibiotic residues in animal products, the use of antibiotics in feed industry has been prohibited in many nations, such as Europe, United States, Korea, and so on (Martin et al., 2015; Walsh and Wu, 2016).

**Abbreviations:** γ-GT, γ-glutamyltranspeptidase; AKPase, Alkaline phosphatase; BBM, Brush border membrane; DAO, Diamine oxidase; PEPT1, Peptide transporter 1; SGLT1, Sodium-dependent glucose transporter.

Therefore, many alternatives to antibiotics have been developed (Wang et al., 2012, 2017). It is well known that probiotics are an alternative strategy to antibiotics. Previous researches illustrated that probiotics enhance growth performance of poultry and swine (Wang et al., 2012), modulate immune system (González-Ortiz et al., 2013), and promote intestinal health (Tojo et al., 2014). Probiotics, with the definition of live micro-organisms, are considered to have potential benefits on the host health (Hickey et al., 2012; Djurasevic et al., 2017). Many researches demonstrated that probiotics improve growth performance (Giang et al., 2010), modulate host immunity (Deng et al., 2013), and decrease the diarrhea rate of weaned piglets (González-Ortiz et al., 2013). Bacillus amyloliquefaciens is one of probiotic strains, which produces a variety of commercially important enzymes to improve digestibility and absorption of nutrients (Farhadi et al., 2003; Lee et al., 2008). Recent studies of our research group have found that replacing aureomycin with B. amyloliquefaciens SC06 significantly improve the daily weight gain of piglets, increase antioxidant capacity (Wang et al., 2017) and decrease bacterial translocation (Ji et al., 2013). However, little information about effects of B. amyloliquefaciens SC06 replacement on digestibility and absorption of nutrients in piglets was found. Therefore, the aim of this study was to clarify effects of probiotic Bacillus amyloliquefaciens as an alternative of antibiotics on main digestive and absorb enzymes in piglet intestine.

# MATERIALS AND METHODS

The experimental procedures used in the present study were approved by the Animal Care and Use Committee of Zhejiang University, and strictly followed the guidelines of the Guide for the Care and Use of Agricultural Animals in Research and Teaching.

#### Animals and Experimental Treatments

A total of 90 male Duroc × Landrace × Yorkshire piglets at 42 days old were blocked by BW (average 14.57 ± 0.25 kg), and randomly divided into three groups with 10 piglets pre pan and 3 pans pre group. The three groups were (1) Group1 (G1) fed the basal diet supplemented with 150 mg/Kg aureomycin, (2) Group 2 (G2) fed the basal diet supplemented with 75 mg/Kg aureomycin and 1 × 10<sup>8</sup> cfu/Kg Ba, and (3) Group 3 (G3) fed the basal diet with 2 × 10<sup>8</sup> cfu/Kg Ba. The composition of the basal diet was shown in **Table 1**. The experimental period was 28 days. Piglets were housed in a temperature-controlled nursery and had ad libitum access to feed and water. Ingredient and chemical composition of the basal diet were listed in **Table 1**.

#### Bacterial Strain and Aureomycin

Bacillus amyloliquefaciens cells (China Center For Type Culture Collection, No: M2012280) (1 × 10<sup>8</sup> cfu/g) were prepared by the Laboratory of Microbiology, Institute of Feed Sciences, Zhejiang University, China. Starch was used to dilute Ba and the same amount of starch was also added to each group to compensate for the difference in nutrient composition of the diets. Aureomycin was obtained from Tongyi feed agriculture and animal husbandry Co., Ltd. (Qingdao, China).

# Sample Collection

Six pigs were randomly selected from each group for sample collecting at the end of the experiment. After the slaughter, the gastrointestinal tract was immediately removed. The segments of jejunum were removed and rinsed with sterilized saline, and then the Jejunal mucosa was scraped from a 10–15 cm segment of jejunum. The content samples of duodenum, jejunum, ileum, and cecal were also collected. All samples were frozen in liquid nitrogen immediately and then stored at −70◦C for further analysis.

### Enzyme Activity Analyses

Jejunal mucosa samples were homogenized in ice-cold 0.1 mol/L, pH = 6.8 maleic acid buffer (1:10, w/v) and centrifuged at 3000 × g for 10 min. Supernatants were collected to determinate the activities of surcrase, maltase, lactuase, AKPase, Na+, K+- ATPase, γ-glutamyl transferase (γ-GT) and DAO, following the protocol of assay kit purchased from Nanjing Jiancheng Bioengineering Institute (Nanjing, China). The contents samples of duodenum, jejunum and ileum were homogenized with icecold physiologic saline (1:4, w/v) and supernatants were obtained by centrifugation at 3500 × g for 15 min. The activities of chymotrypsin, amylase, trypsin, lipase in the supernatants of intestinal content were detected using the assay kit (Nanjing Jiancheng Bioengineering Institute, Nanjing, China).

#### Western Blot Analysis

0.1 g jejunal mucosa was lysed with 500 µL cell lysis buffer for Western and IP (Beyotime Co. Ltd., Nantong, China). The lysates were centrifuged at 12,000 × g for 5 min at 4◦C and the supernatants were transferred to 1.5 ml Eppendorf


<sup>a</sup>Premix supplied per kg: 8255 IU of vitamin A, 2000 IU of vitamin D3, 40 IU of vitamin E, 3 mg of vitamin K3, 2 mg of vitamin B1, 4 mg of vitamin B2, 10 mg of vitamin B6, 0.05 mg of B12, 3 mg of vitamin PP, 15 mg of pantothenic acid, 2 mg of folic acid, 0.2 mg of biptin, 800 mg of choline chloride, 100 mg of vitamin C, 100 mg of Fe (FeSO4), 20 mg of Cu (CuSO4), 55 mg of Mn (MnO), 50 mg of Zn (ZnO), 100 mg of I (KI), 2 mg of Se (Na2SeO3), and 2 mg of Co (CoSO4).

tubes. The concentration of total protein was quantified by BCA Protein Assay Kit (Beyotime Co. Ltd., Nantong, China). Equal amount of cell lysates were resolved by SDS-PAGE, and then transferred electrophoretically to nitrocellulose membranes. After blocking with TBS containing 5% nonfat dry milk (Wondersun Inc., Haerbing, China) and 0.1% Tween-20 for 1 h at room temperature, the membranes were incubated with a primary antibody at 4◦C overnight. After washing with TBST, membranes were incubated with secondary antibody linked to HRP. Detection by enzyme-linked chemiluminescence was performed follow the manufacturer's protocol (ECL, Beyotime Co. Ltd., Nantong, China). Mouse anti-β-actin monoclonal antibody and IgG-HRP secondary antibodies were purchased from Beyotime Biotechnology (Nantong, China). Rabbit anti-SGLT1 and anti-PEPT1 were obtained from Abcam (MA, United States). Quantification of protein bands were analyzed using the Image J software.

#### DNA Extraction and Illumina Miseq

Microbial genome DNA was extracted from fecal samples (using TIANamp Stool DNA kit; TIANGEN, DP328) following the manufacture's recommendation. The V3-V4 hyper variable regions of 16S rRNA were PCR amplified from microbial genome DNA which were harvested from fecal samples (n = 3) forward primers: 5<sup>0</sup> -CCTACGGGNGGCWGCAG-3<sup>0</sup> , reverse primers: 5<sup>0</sup> - GACTACHVGGGTATCTAATCC-3<sup>0</sup> ). A total volume of 20 µL was prepared, containing 1 × reaction buffer, 2 mM Mg2+, 0.2 mM dNTP, 0.1 µM primers, 1 U HotStarTaq polymerase (QIAGEN, cat#203203) and 2 µL DNA template. The PCR program initially started with 94◦C for 2 min; 94◦C for 20 s, 52◦C for 40 s and 72◦C for 1 min, 72◦C for 2 min, repeat for 30 cycles; 72◦C 2 min; stored in 4◦C. The PCR reaction system which was used to add specific tags sequence was 20 µL, containing 1 × reaction buffer (NEB Q5TM), 0.3 mM dNTP, 0.25 M of each primer, 1 U Q5TM DNA polymerase (NEB) and l µL of diluted template. The PCR condition were 98◦C for 30 s; 94◦C for 10 s, 65◦C for 30 s and 72◦C for 30 s, repeat for 30 cycles; 72◦C for 5 min. PCR product was excised from a 1.5% agarose gel, purified by QIAquick Gel Extraction Kit (QIAGEN, cat# 28706) and quantified by UV-Vis spectrophotometer (NanoDrop ND1000, United States). Library construction and Illumina MiSeq sequencing was carried out in G-Bio Biotech (Hangzhou) Co., Ltd. And the information of DNA sequences was analyzed by QIIME software (Caporaso et al., 2010).

#### Statistical Analysis

The data in present study were analyzed by one-way ANOVA using the IBM SPSS 16.0. The values of P < 0.05 or 0.01were considered a statistically significant difference.

# RESULTS

#### Digestive Enzyme Activity in Intestinal Contents of the Piglets

Compare with the G1 group, half replacing antibiotic with Ba significantly enhanced the activity of chymotrypsin (P < 0.05) in jejunum and ileum contents, while decreased the chymotrypsin activity in duodenum contents. The activities of chymotrypsin in different intestinal sections in G3 group were slightly increased in compare to the G1 group. It was much higher than that of the G2 group. However, in jejunum and ileum, they were dramatically lower when compared G2 group with G3 group (**Figure 1**).

The activity of amylase in intestinal contents was shown in **Figure 2**. Half replacing antibiotic with Ba significantly enhanced the amylase activity (P < 0.05) in the jejuna content when compared with the G1 group, and the piglets in G3 group had the highest amylase activity (P < 0.05) in the contents of duodenum and ileum (P < 0.05) among the three groups. In addition, it was observed that half replacing the antibiotic with Ba significantly increased the lipase activity in duodenal and jejunal content compared with the G1 group, and that in G3 group was dramatically enhanced in comparison with the G2 group (**Figure 3A**), while there was dramatical difference between G1 and G3 group (**Figure 3B**). However, there were no significant changes of trypsin activity in the content of duodenum and jejunum (**Figures 4A,B**).

#### Enzyme Activity Related to Absorption in Jejuna Mucosa of the Piglets

**Figure 5** showed the results of sucrase (A), lactase (B), and maltase (C) activities. Half replacing the antibiotic with Ba only significantly increased the activity of lactase (P < 0.05) when compared with G1 group (**Figure 5B**). The piglets in G3 group had much higher activity of sucrase (P > 0.05), maltase (P < 0.05) compared with G1 and G2 group (**Figures 5A,C**). Compared with G1 group, half replacing the antibiotic with Ba did not affect the activity of AKPase, however when the piglet fed with Ba instead of antibiotic, AKPase activity was significantly improved (**Figure 5D**). And the same result was found in the activity of Na+, K+-ATPase (**Figure 5E**).

Feed supplemented with antibiotic or Ba did not induce any change in γ-GT activity in each group (**Figure 5F**). No significant difference of DAO activity was found between G1 and G2 group, while DAO activity in the piglets fed with Ba instead of antibiotic was significantly decreased (P < 0.05) (**Figure 5G**).

# Effects of Ba on Transporter Expression

Compared with G1 group, half and total replacing the antibiotic with Ba both significantly reduced the SGLT1 expression in jejuna mucosa (P < 0.05), while no dramatical difference between G2 and G3 group (**Figure 6**). And half and total replacing the antibiotic with Ba did not affect the PEPT1 expression when compared with G1 group.

# Effects of Ba on Intestinal Microbiota

To further characterize the changes in microbial population imposed by the use of Ba and antibiotics, 16S rRNA were classified taxonomically to the genera level. Similar Shannon diversity indexes were found in G1–G3 group, as well as Chao1, PD whole tree and observed species (**Figures 7A–D**). To further measure the variability in species composition when the piglets were fed the diet with half and total replacing the antibiotic with

Frontiers in Microbiology | www.frontiersin.org

Ba, β-diversity indexes were analyzed. Based on the unweighted UniFrac distance analysis, the difference between the intestinal flora of the intestinal flora of G1, G2, and G3 was significant (p = 0.028, **Figure 8A**), but based on the weighted UniFrac distance, no differences were found (p = 0.199, **Figure 8B**). While no significant changes were found among the different treatments in the phylum and genus level of gut bacteria (**Supplementary Figures S1**, **S2** and **Supplementary Tables S1**, **S2**).

#### DISCUSSION

As more and more people are increasingly looking at food safety and environmental contamination, antibiotics replacement has become a trend. Probiotics have been widely used as antibiotic replacement to enhance animal growth and intestinal health (Caporaso et al., 2010; Deng et al., 2013). However, it remains unclear whether probiotics impact on the nutrients

FIGURE 2 | Effects of Ba on amylase activity in contents of duodenum (A), jejunum (B), and ileum (C). Results were expressed as mean ± SEM, n = 6. a and b means significantly different (p < 0.05).

fmicb-09-02427 October 17, 2018 Time: 13:56 # 4

digestion and absorption of piglets. Previous study found that diet supplemented with Bacillus amyloliquefaciens partly instead of antibiotics dramatically improved the growth performance of piglets (Wang et al., 2017). Hence, our study focused on the influences of probiotic Bacillus amyloliquefaciens as an alternative of antibiotics on activities of digestive enzymes in piglet intestine.

Carbohydrates are one of the most important components of the diet. In digestive tract, carbohydrates are mainly digested by salivary and pancreatic amylases, further broken down into monosaccharides by disaccridase, such as sucrase, maltase and lactase, which secreted by the BBM of enterocytes, and then are absorbed (Drozdowski and Thomson, 2006; Zhen et al., 2018). Wang and Gu (2010) found that amylase activity was remarkably higher in Arbor Acres broilers feed with Bacillus coagulans NJ0516 than that in control group. In current study, amylase activity in duodenum and ileum significantly higher in piglets administrated with Ba than the piglets fed the diet supplemented with antibiotics or Ba half replacing antibiotics, and Ba half replacing antibiotics dramatically increased amylase activity in jejunum compared with the antibiotic group. Compared with the piglets fed the diet containing antibiotic, an increase in sucrase, maltase and lactase in jejuna mucosa was observed when they were fed the diet supplemented with Ba. And the same results were reported that when rats administrated with probiotics Lactobacillus bulgaricus and Streptococcus thermophilus, sucrase and lactase activity was enhanced in intestinal mucosa (Southcott et al., 2008), and Goyal et al. (2013) found that probiotic Lactobacillus rhamnosus GG dramatically increased the activity of sucrase and lactase in BALB/c mice. In addition, Na+/K+- ATPase is a transmembrane protein and is responsible for driving the sodium-dependent glucose transporter (SGLT1) in BBM, which plays an important role in glucose transport. It has shown that the inhibition of SGLT1 was secondary to a reduction in Na+/K+-ATPase activity (Manoharan et al., 2018). However, we observed that instead of antibiotic, feeding Ba-supplemented diet significantly increased the activity of Na+/K+-ATPase in jejuna mucosa, though the expression of SGLT1 was remarkably decreased in piglets fed with Ba. The results indicate that as an alternative of antibiotic, Ba could influence the metabolism of carbohydrates metabolism, while the certain further research was needed to clarify the certain effect of Ba.

Dietary protein is digested by both mammalian and bacterial enzymes in the intestinal tract (Goyal et al., 2013; Switzar et al., 2013). The protease activity was significantly higher in the common carp fed with basal diets supplemented with Bacillus sp. compared with control group (Wang and Xu, 2006). It has been suggested that in our study, compared to antibiotic fed group, activities of chymotrpsin was significantly increased in jejunual and ileal contents of piglets fed with Ba half replacing antibiotics. There was a tendency for increased activity of chymotrpsin when the piglet fed diet supplemented with Ba instead of antibiotic, while no remarkable changes were observed. However, activities of trypsin in intestinal contents and γ-glutamyl transpeptidase (γ-GT) in jejuna mucosa of piglets were not affected by the diet supplemented with Ba. And Ba treatment did not change the expression of peptide transporter 1 (PEPT1), which is a kind of membrane transporter proteins and helps the cellular of oligopepetides (Daniel, 2004).

In addition, the increase of lipase activity leads to more effective fat absorption (Lowe, 1994; Karásková et al., 2015). However, Ogawa et al. (2015) indicated that Lactobacillus gasseri SBT2055 significantly decreased the lipase activity in order to increase size of fat emulsion droplet and suppress lipid absorption. In current study, we found that piglets in replacing half dosage of antibiotic with Ba had much higher activity of lipase in contents of duodenum and jejunum than piglets fed with antibiotic (P < 0.05), and piglets in Ba groups had the highest lipase activity in duodenum, although in jejunum, lipase activity was significantly decreased in piglets fed with Ba compared to Ba half replacing antibiotics group.

The intestinal environment plays a critical role in maintaining good health (Farhadi et al., 2003; Melo et al., 2015). DAO is one of the indicators of intestinal epithelial integrity (Tossou et al., 2016). It has been reported that antibiotic treatment significantly lowered DAO activity (Sato et al., 2016). However, in the present study, piglets administrated with Ba had much lower DAO activity compared with antibiotic group, which may indicate much better state of intestinal integrity. AKPase is a new factor which contributes to maintain gut homeostasis (Lalles, 2014; Melo et al., 2015). The deletion of intestinal AKPase gene caused a significant decrease of tight junction protein expression and function (Liu et al., 2016). In the present study, we found that Ba replacing antibiotic dramatically enhanced the activity of AKPase.

Intestinal microbiota is an important compartment of digestive tract of animals (Azad et al., 2018). Various researches have demonstrated that probiotic can positively regulate the composition of the intestinal macobiota (Cisek and Binek, 2014;

Kristensen et al., 2016; Hu et al., 2017). Dietary supplementation of Bacillus amyloliquefaciens dramatically decreased the population of Escherichia coli and increased Lactobacillus population in cecum (Lei et al., 2014). The administration of Lactobacillus B1 significantly decreased the number of E. coli and increased lactic acid bacteria in cecal digesta of chickens (Peng et al., 2016). While in the current study, no significant influences were found when the piglet fed with Ba instead of antibiotics.

#### CONCLUSION

fmicb-09-02427 October 17, 2018 Time: 13:56 # 8

In conclusion, results from the previous study indicate that supplementation with Ba enhanced growth performance of piglets (Wang et al., 2017). This enhancement was associated with the positive influence of tract digestibility and intestinal integrity. Therefore, according to our research, Bacillus amyloliquefaciens SC06 could be used as alternative to antibiotics in piglet diets.

#### AUTHOR CONTRIBUTIONS

WL and LG designed the experiment. SH and XC drafted the manuscript. XC carried out most of the experiments and analyzed the data. YWu, XM, and HX participated in the animal experiment. YWa helped to detect some parameters. XZ performed the analysis of the microbiota. WL and LG had primary responsibility for the final content.

#### REFERENCES


#### FUNDING

This research was jointly supported by the National Natural Science Foundation of China (Grant Nos. 31472128 and 31672460), and by the Key Project of Science and Technology of Zhejiang Province, China (Grant No. 2006C12086).

# ACKNOWLEDGMENTS

We are grateful to the staff in this laboratory for their valuable input to this study.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.02427/full#supplementary-material

FIGURE S1 | The changes of gut bacterial phyla in different treatments.

FIGURE S2 | The changes of gut bacterial genera in different treatments.

TABLE S1 | Detailed data for the gut bacterial phyla.

TABLE S2 | Detailed data for the gut bacterial genera.

healthy dogs. Arch. Anim. Nutr. 67, 406–415. doi: 10.1080/1745039X.2013. 830517


Junction Protein of Weaned Pig. Biomed. Res. Int. 2016, 2912418. doi: 10.1155/ 2016/2912418


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Hu, Cao, Wu, Mei, Xu, Wang, Zhang, Gong and Li. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fmicb-09-02427 October 17, 2018 Time: 13:56 # 9

# Rutin and Its Combination With Inulin Attenuate Gut Dysbiosis, the Inflammatory Status and Endoplasmic Reticulum Stress in Paneth Cells of Obese Mice Induced by High-Fat Diet

#### Edited by:

Yuheng Luo, Sichuan Agricultural University, China

#### Reviewed by:

Qingping Zhong, South China Agricultural University, China Zhaolai Dai, China Agricultural University, China Diana Soghomonyan, Yerevan State University, Armenia

#### \*Correspondence:

Xiulan Guo guoxiulan@cdu.edu.cn Zhenhua Liu zliu@nutrition.umass.edu

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 11 August 2018 Accepted: 17 October 2018 Published: 05 November 2018

#### Citation:

Guo X, Tang R, Yang S, Lu Y, Luo J and Liu Z (2018) Rutin and Its Combination With Inulin Attenuate Gut Dysbiosis, the Inflammatory Status and Endoplasmic Reticulum Stress in Paneth Cells of Obese Mice Induced by High-Fat Diet. Front. Microbiol. 9:2651. doi: 10.3389/fmicb.2018.02651 Xiulan Guo1,2 \*, Renyong Tang<sup>1</sup> , Shiyong Yang<sup>3</sup> , Yurong Lu<sup>1</sup> , Jing Luo<sup>1</sup> and Zhenhua Liu<sup>2</sup> \*

<sup>1</sup> School of Pharmacy and Biological Engineering, Chengdu University, Chengdu, China, <sup>2</sup> School of Public Health and Health Sciences, University of Massachusetts, Amherst, MA, United States, <sup>3</sup> College of Animal Science and Technology, Sichuan Agricultural University, Chengdu, China

Gut dysbiosis induced by high fat diet (HF) or obesity is a predisposing factor to develop diverse inflammatory diseases. Polyphenols and fibers, often eaten together, have been reported to have prebiotic actions, but their health promoting benefits still need to be further characterized and defined. This study attempted to understand how polyphenol rutin and polysaccharide inulin influence intestinal health in mouse model fed a HF (60 kcal%) diet. A total of 48 C57BL/6J mice were divided into four groups fed with a low fat (10% kcal%) control diet (LC), a high fat control diet (HC), a high-fat diet supplemented with rutin (HR), or a high-fat diet supplemented rutin and inulin (HRI) for 20 weeks. Rutin supplementation reduced the HF diet-induced increase of Firmicutes/Bacteroidetes (F/B) ratio, Deferribacteraceae population and plasma lipopolysaccharide (LPS) (p < 0.05); ameliorated inflammation as indicated by the decreased circulating inflammatory cytokines (p < 0.05) and the reduced expressions of intestinal inflammatory mediators (p < 0.05); and attenuated the endoplasmic reticulum (ER) stress in Paneth cells as indicated by the decreased expressions of the ER markers (p < 0.05). Compared to the rutin supplementation alone, the co-administration of rutin with inulin improved the utilization of rutin as indicated by its decreased excretion, suppressed a number of harmful bacteria including Deferribacteraceae and Desulfovibrionaceae (p < 0.05), and further reduced the expression of the key inflammatory cytokine TNF-α and increased the production of butyrate, despite the supplementation of inulin reversed the decrease of body weight induced by rutin supplementation due to an increased food intake. Taken together, our data demonstrated that rutin supplementation ameliorated the inflammatory status and ER stress in Paneth cells under a HF-induced obese state, and its co-administration

**28**

with inulin further mitigated the inflammatory status, indicating the potential to combine polyphenol rutin and the polysaccharide inulin as a dietary strategy to ameliorate gut dysbiosis, to improve inflammatory status and thereby to reduce medical disorders associated with HF-induced obesity.

Keywords: obesity, rutin, inulin, gut dysbiosis, inflammation, Paneth cells

#### INTRODUCTION

fmicb-09-02651 November 1, 2018 Time: 18:53 # 2

The prevalence of obesity has dramatically increased and has more than doubled since 1980 (Gregg and Shaw, 2017). Currently, one third of the world's population is overweight or considered obese, and a further increase is even predicted, ∼50% by 2030 (Kelly et al., 2008; Finkelstein et al., 2012). Obesity has now been considered as a major public health issue since it is associated with an array of medical complications, including increased risk of type 2 diabetes, hypertension, cardiovascular disease, inflammatory bowel disease, and cancers (Hotamisligil, 2006). One of the responsible mechanisms for this relationship is the systemic inflammatory status associated with high fat diet or obesity. Recently, growing evidence has implicated the intestinal immune system as an important contributor to those diseases (Garidou et al., 2015; Luck et al., 2015; Monteiro-Sepulveda et al., 2015; Winer et al., 2016).

A close relationship between the gut microbiota and obesity has been demonstrated in both human and animals, with enriched Firmicutes and reduced Bacteroidetes phyla in obesity (Ley et al., 2005, 2006). Our recent research revealed that high fat (HF) diet stimulated intestinal inflammation via altering gut microbiota, and it occurred prior to the potential influence by circulating inflammatory cytokines (Guo et al., 2017), indicating, in addition to adipose tissue, HF per se also drives intestinal inflammation via altering gut microbiota. Data from Lee et al. (2017) reiterated the role of HF on intestinal inflammation and health by showing that HF-fed mice were more susceptible to experimental colitis, and exhibited severe colonic inflammation, accompanied by the expansion of selected pathobionts such as Atopobium sp. and Proteobacteria.

A number of studies have demonstrated that a variety of polyphenols and dietary fibers exert the properties to mitigate the microbial dysbiosis produced by high-fat diets (Delzenne et al., 2011; Karlsson et al., 2013; Etxeberria et al., 2015). For instances, the beneficial Bacteroidetes community is favored by wine vinegar, polyphenol-rich fruits and green tea (Lee et al., 2006; Cerezo et al., 2008; Selma et al., 2009). Quercetin or grape polyphenols attenuated Firmicutes/Bacteroidetes ratio and suppressed the growth of bacterial species associated to diet-induced obesity, resulting in lower intestinal and systemic inflammation (Etxeberria et al., 2015; Roopchand et al., 2015). Non-digestible dietary fiber exerts a significant role in promoting beneficial bacteria and enhance microbial diversity in the gut (Cani et al., 2007), which might contribute to ameliorate obesity and obesity associated disorders (Dewulf et al., 2011).

Polyphenols and dietary fiber may interact to mediate gut microbiota. Jaganath et al. (2006) have shown that combining fermentable carbohydrates accelerate the breakdown of the polyphenol rutin in an in vitro model of colonic fermentation (Hou et al., 2015). Moreover, a range of fermentable fibers inhibited phenolic acid production from rutin (Mansoorian et al., 2015). Also, polyphenols could influence the fermentation of the dietary fiber as polyphenols have been shown to have both antibacterial (Taguri et al., 2004) and prebiotic actions (Tuohy et al., 2012). Thus, there would be a reciprocal interaction between dietary polyphenols and fibers in the mediation of gut mcirobiota. The present study was to assess the potential of rutin and its co-administration with inulin to ameliorate HF-diet-induced gut microbiota dysbiosis and inflammation.

#### MATERIALS AND METHODS

#### Animal Study

The animal protocol was approved by the Institutional Animal Care and Use Committee of Chengdu University. Forty-eight male C57BL/6J mice (4 weeks old) were randomly distributed into four groups: the low-fat control group (LC; n = 12), mice were fed a low-fat diet (LF) with 10% kcal from fat (D12450B; Research Diets Inc.); the high-fat control group (HC; n = 12), mice were fed a high-fat (HF) diet with 60% kcal from fat (D12492; Research Diets Inc.); a rutin group (HR; n = 12), mice fed a HF diet with rutin 6.4 mg/g diet; and a rutin + inulin group (HRI; n = 12), mice fed a HF diet with rutin 6.4 mg/g diet and inulin (30 mg/ml) via drinking water. The doses of rutin (about 0.01 mol/kg diet) and inulin were chosen based on previous studies (Jurgonski ´ et al., 2012; Hoek-van den Hil et al., 2013, 2015; Hamilton et al., 2017). Rutin (≥97% purity) was supplied by Chengdu Okay Pharmaceutical Co. Ltd (Sichuan, China); and inulin (≥95% purity) was supplied by Rui Hu Biological Co. Ltd (Qinghai, China).

The mice were fed ad libitum and housed under a 12 h light/12 h dark cycle. Body weights and food intake were recorded every other week. During the 19th week, food and feces samples were collected, and feces excretion was recorded. After 20 weeks of feeding, all mice were euthanized with CO2. The blood sample was collected via heart puncture after euthanasia, and plasma was separated and stored at −80◦C for later analysis of inflammatory cytokines and biochemical parameters. Following with the heart puncture, the abdomen was opened and visceral fat pads were harvested. Samples of the intestinal and colonic contents were collected, frozen immediately with liquid nitrogen, and stored at −80◦C for further microbial abundance and SCFA analysis. A segment (∼2 cm) of ileum was excised and formalin-fixed for immunohistochemistry, and the intestinal mucosa was collected as described in a previous publication (Guo et al., 2017), and then

immediately frozen with liquid nitrogen and stored at −80◦C for later real-time PCR analysis.

#### 16S rRNA Gene Sequencing and Analysis of Microbial Profile

16S rRNA gene sequencing was performed on the small intestinal contents from 32 C57BL/6J mice (n = 8 mice per diet group). Bacterial DNA was extracted from the intestinal contents using the QIAamp DNA Stool Mini Kit (Qiagen, Germany) according to the manufacturer's instructions. The concentration and purity of extracted DNA were determined using the NanoDrop-2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United States).

The V4 region of the 16S rRNA gene was amplified used universal bacterial primers 515F (5<sup>0</sup> -GTGCCAGCMGCCGCG GTAA-3<sup>0</sup> ) and 806R (5<sup>0</sup> -GGACTACHVGGGTWTCTAAT-3<sup>0</sup> ) with Phusion High-Fidelity PCR Master Mix (New England Biolabs, Ipswich, MA, United States) followed by purification with Qiagen Gel Extraction Kit (Qiagen, Germany). The sequencing libraries were prepared using TruSeq DNA PCR-Free Sample Preparation Kit (Illumina, San Diego, CA, United States), sequenced on an Illumina HiSeq 2500 platform (San Diego, CA, United States), and 250 bp paired-end reads were generated. Paired-end reads were merged using FLASH (Baltimore, MD, United States), and subsequently raw tags were filtered with a specific standard to obtain the high-quality tags using QIIME version 1.7.0 (Caporaso et al., 2010). Sequences with ≥97% similarity were assigned to the same operational taxonomic units (OTUs) using Uparse software version 7.0.1001 (Edgar, 2013), and were annotated with taxonomic information based on the RDP classifier version 2.2 (Wang et al., 2007) algorithm using the Greengene database. OTU abundance data was normalized using a standard sequence number corresponding to the sample with least sequences, and subsequent analysis of alpha diversity and beta diversity were performed with QIIME. To visualize the sample differences, principal coordinate analysis (PCoA) was performed with weighted Unifrac (Lozupone and Knight, 2005). The clustering of samples was explained with the principal coordinate (PC) values. Differences in OTU abundance between groups were identified using LDA (Linear Discriminant Analysis) Effect Size (LEfSe) (Segata et al., 2011) 1 .

# Gene Expression in Small Intestine Mucosa

The mRNA expressions of inflammatory factors, 4 antimicrobial peptides, α-defensin 5, lysozyme, angiogenin 4 (ANG 4) and regenerating islet derived 3-gamma (Reg IIIγ), and ER stress markers in the intestine were measured by real-time PCR. Briefly, total RNA was extracted from the small intestine mucosa with Trizol (Invitrogen, Carlsbad, CA, United States); the concentrations of total RNA were determined spectrophotometrically (NanoDrop-2000, Thermo Fisher Scientific, Waltham, MA, United States), and cDNA was synthesized with the ExScriptTM RT-PCR kit (TaKaRa, Dalian,

#### Immunohistochemistry of Lysozyme

Immunohistochemical analysis was performed on the formalinfixed sections of ileum. The paraffin-embedding slides were deparaffinized in xylene, followed by rehydration in ethanol. After blocking non-specific antibody binding with 5% bovine serum albumin, sections were incubated with the specific first antibody of rabbit monoclonal antibody lysozyme (ab108508; Abcam, Cambridge, United Kingdom) followed with the second antibody goat anti-rabbit IgG HRP. Immunoreactivity was detected with horseradish peroxidase-conjugated anti-goat EnVision kit (DAKO). All slides were counterstained with hematoxylin. Paneth cells were identified microscopically by their location just at the base of small intestinal crypts of Lieberkuhn (Elphick and Mahida, 2005).

#### Blood Plasma Analysis

For detecting inflammatory cytokines in the blood, a ProcartaPlex Mouse High Sensitivity Panel (5 plex) from Thermo Fisher Scientific for 5 cytokines, IFN-γ, IL-2, IL-4, IL-6, and TNF-α, was selected for mice, and the assays were performed on the MAGPIXTM platforms (Luminex, Austin, TX, United States) following the manufacturer's instruction. Plasma leptin level was determined using a leptin ELISA kit (Multisciences, Hangzhou, China) according to the manufacturer's instructions. Triglycerides and total cholesterol contents were quantified by Beckman Coulter DXC 600 Pro (Beckman Coulter, Inc., Brea, CA, United States) using standard spectrophotometric assays. Plasma LPS level was analyzed using a mouse LPS ELISA kit (Cusabio, Wuhan, China). All standards and samples were measured in duplicate.

# Short-Chain Fatty Acid (SCFA) Analysis

Short-chain fatty acids (acetate, propionate, and butyrate) in colonic contents were determined. Colonic content samples (150 mg) were vortexed with 350 µl deionized water and 2-ethylbutyric acid (internal standard, Sigma Chemical), after which 500 µl 2.5 M sulphuric acid was added to the sample. SCFA were then extracted with diethyl ether, silylated with n-(tert-butyldimethylsilyl)-n-methyltrifluoroacetamide (Sigma Chemical), and then centrifuged (5,000 × g, 10 min). The supernatant was filtered through a 0.45 µm filter, and 1 µl of clear filter solution was directly injected into gas chromatograph system (Agilent Technologies) for analysis, which was performed as described by Sheveleva and Ramenskaya (2010).

# Rutin Content and Rutin Degradation Analysis

Rutin content was measured according to Araujo et al. (2013). The food samples or feces samples were mixed with the 60%

China). Real-time PCR was performed on the ViiATM 7 System (Applied Biosystems, Foster City, CA, United States) utilizing the following thermal cycling conditions: 95◦C for 10 min, followed by 40 cycles of 95◦C for 15 s and 60◦C for 60 s. Primer sequences were chosen according to our previous studies (Liu et al., 2016; Guo et al., 2017) and listed in **Supplementary Data (Table S1)**.

<sup>1</sup>http://huttenhower.sph.harvard.edu/galaxy/

methanol solution, and the suspension was extracted under ultrasonication for 50 min and subsequently centrifuged for 10 min at 5,000 × g. The supernatant filtered through a 0.45 µm filter was used in high performance liquid chromatography (Shimadzu Co., Kanagawa, Japan) analysis. After detecting the rutin content in feed and feces, the degradation of rutin in the gut was determined as follows:

Degradation of rutin (%) = rutin content in feed (mg/g) × feed intake (g) − rutin content in feed (mg/g) rutin contentin feces(mg/g) × feces excretion (g) × feed intake (g) × 100%

#### Statistical Analysis

Data are expressed as means ± SEM. Data analysis was performed using SPSS 15.0 software (Chicago, IL, United States). Comparisons between groups were made using ANOVA followed by post hoc test and associations were assessed by the linear regression. Correlation and regression analysis of the microbiome and relative parameters were conducted using Spearman's rank correlation coefficient. For the gene expression data analysis, the expression of each gene was normalized to the housekeeping gene β-actin (CtTarget gene-Ctβ−actin). Statistical analyses were performed based on 1Ct. The relative abundance of relative gene expression was reported as 2−11Ct, where 11Ct = 1CtExperiment-1CtControl.

#### RESULTS

#### Food Consumption and Animal Growth

The body weights between the HC and HRI groups were similar (p > 0.05), but higher than that of LC group after 8 weeks and that of HR group after 14 weeks (p < 0.05, **Figure 1A**). As expected, the LC group had the highest food intake during the experiment and the HC group has the lowest (**Figure 1B**). The HR group had the similar food intake to the HC group (p > 0.05) whereas the further supplementation with inulin (the HRI group) increased food intake gradually, reaching to a significant level at 8 weeks when compared to the HC and HR group (p < 0.01), and close to LC after 16 weeks.

#### Plasma Metabolic Parameters

High fat diet (the HC group) increased plasma cholesterol, triglycerides, leptin and LPS contents when compared to the LC group (**Table 1**). The rutin supplementation numerically reduced the plasma triglyceride (HR vs. HC), and its co-administration with inulin magnified this effect into a statistically significant level (p < 0.05, HRI vs. HC). Surprisingly, the cholesterol contents were not reduced by rutin (HR vs. HC) and even increased by inulin (HRI vs. HC). The leptin content was reduced by rutin supplementation and the LPS contents were significantly reduced in HR and HRI groups when compared to the HC group (p < 0.05).

# The Inflammatory Factors in the Blood and Small Intestine

Five plasma inflammatory cytokines, IFN-γ, IL-4, IL-6, IL-2, and TNF-α, were measured using the ProcartaPlex Multiplex Immunoassay (**Figure 2A**). Compared to the LC group, the high fat diet (the HC group) increased plasma levels of those cytokines, with statistically significant differences for IL-6, TNF-α and IL-2 (p < 0.05), and the supplementation with rutin or rutin + inulin significantly reduced the plasma concentrations of those cytokines (HR and HRI vs. HC, p < 0.01). For the inflammatory mediators (**Figure 2B**), NFκB, MyD 88, TNF-α and LBP, the mRNA levels of them were higher in the HC group when compared to the LC group with a statistical difference for LBP, the gene for LPS binding protein (p < 0.05). The rutin supplement or its combination with inulin suppressed the production of these mediators (HR and HRI vs. HC, p < 0.05), and notably the combinatorial supplementation of rutin and inulin significantly reduced the intestinal expression of TNF-α (HRI vs. HR, p < 0.05).

#### Expression of Paneth Cell AMPs and ER Stress Markers

The relative mRNA expression levels of Paneth cell antimicrobial peptides (AMPs) were detected in the small intestinal tissue (**Figure 3A**). Except for Reg IIIγ, high fat diet (the HC group) significantly increased mRNA expressions of Paneth AMPs: α-cryptdin (1.8-fold), lysozyme (1.6-fold), and ANG 4 (1.9 fold) (p < 0.05). The mRNA expression levels of these Paneth cell AMPs were positively correlated with the plasma LPS level (**Figure 3B**, p < 0.01) and the gene expressions of inflammatory mediators NFκB (**Figure 3C**, p < 0.01).

However, when we measured the protein level of lysozyme by immunohistochemistry (**Figure 3D**) the results indicated HC had lower protein expression in Paneth cells when compared to the LC group, indicating discordance between protein expression and mRNA expression of lysozyme in the Paneth cells. Since endoplasmic reticulum directly participates in the translational expression of those AMPs and may be responsible for the inconsistency at the protein and mRNA level, we therefore detected mRNA expression of ER stress markers in the intestine tissue, such as chaperone protein binding protein (BiP), activating transcription factor 4 (ATF4) and c/EBP-homologous protein (CHOP). The results showed higher mRNA expressions of BiP and ATF4 in HC when compared to the LC group (**Figure 3E**).

With the addition of rutin or its combination with inulin ameliorated the elevation of the mRNA expression of Paneth AMPs and ER stress biomarkers, and reduction of the lysozyme protein expression induced high-fat diet (**Figures 3A,D,E**).

#### The Abundances of Gut Microflora

High fat diet (the HC group) significantly reduced the Bacteroidetes (p < 0.01)and its families (especially Bacteroidales\_ S24-7 group, p < 0.001, Porphyromonadaceae, p < 0.05), in favor of the presence of Lachnospiraceae family in Firmicutes

weight induced by high-fat diet, but its co-administration with inulin elevated gradually food intake toward no significant effect on body weight compared with HC. Black star indicated there was a significant difference compared with HC (p < 0.05).



LC, low-fat control group; HC, high-fat control group; HR, the group fed high-fat diet with rutin; HRI, the group fed high-fat diet with rutin and inulin. In each line, the values with different letters indicated significant difference among groups (p < 0.05).

(p < 0.01) and Deferribacteres (largely due to increase of Deferribacteraceae family, p < 0.05), and resulted in an increased Firmicutes/Bacteroidetes (F/B) ratio (p < 0.01) (**Figures 4A–C**, **5A,B**). However, high fat diet reduced Erysipelotrichaceae in Firmicutes phylum (p < 0.01) in the intestine when comparing to the LC group (**Figure 4C**). Correlation analysis indicated F/B ratio was positively associated with plasma LPS level (**Figure 4D**, p < 0.01). Rutin supplementation or its combinational supplement with inulin promoted the growth of the most families of Bacteroidetes (**Figures 4A,C**), such as Bacteroidales\_S24-7 group (HR and HRI vs. HC, p < 0.001), Bacteroidaceae (HR and HRI vs. HC, p < 0.01), Porphyromonadaceae (HR vs. HC, p < 0.01; HRI vs. HC, p = 0.09), and Rikenellaceae (HR vs. HC, p < 0.05; HRI vs. HC, p = 0.05 ), and decreased the F/B ratio (**Figure 4B, HR** and **HRI** vs. HC, p < 0.01) and Deferribacteraceae population (**Figures 4A,C**, HR vs. HC, p = 0.05; HRI vs. HC, p < 0.01). Rutin supplementation alone (the HR group) reduced the number of Firmicutes (p < 0.01), especially Lachnospiraceae family (p < 0.05), and promoted the growth of Proteobacteria phylum (p < 0.05) (largely due to Desulfovibrionaceae family) in comparison with HC group (**Figures 4A,C**, **5C,D**). The further supplementation with inulin (the HRI group) altered the composition of gut microflora (**Figures 4C**, **5E,F**): increased the proportions of Lachnospiraceae (p < 0.01) and Bacteroidaceae (p < 0.05), and suppressed Desulfovibrionaceae (p < 0.001), Ruminococcaceae (p < 0.01), and Deferribacteraceae (p < 0.001) at family level; at genus level, elevated (p < 0.05) the abundances of Lachnospiraceae\_NK4A136\_group, Lachnoclostridium, Roseburia, Blautia, Bacteroides and Lactobacillus, and reduced (p < 0.01) Desulfovibrio Ruminiclostridium\_9 and Mucispirillum (**Supplementary Figure S1**) when compared with HR group.

The characteristics of the gut microbiota in LC group and rutin-supplemented mice were similar according to PCoA plot (**Figure 4E**), besides rutin supplementation increased significantly Proteobacteria phylum (**Figure 4A**), and they were far away from HC and HRI groups.

#### The Content of SCFA and Rutin in Feces

High fat diet feeding (the HC group) reduced acetate, propionate, butyrate and their total contents in colonic contents when compared to the LC control group (**Table 2**). The supplementation of rutin or its combination with inulin significantly attenuated the reduction of SCFA. When compared to the HC group, the rutin supplementation (HR group) increased acetate and propionate contents (p < 0.05), and its co-administration with inulin (the HRI group) increased the production of all 3 SCFAs (p < 0.05). It is highly noteworthy that the co-administration with inulin (the HRI group) further significantly increased the production of butyrate (p < 0.05), which is considered as the key SCFA, and numerically increased propionate (p = 0.10) when compared to the HR group.

The HRI group indicated lower rutin left in feces (2.16 ± 0.47 vs. 2.90 ± 0.09 mg/g, p < 0.05), and increased the catabolism (the % of degradation) of rutin (95.37 ± 0.79 vs. 92.42 ± 0.64%, p < 0.05) in the feces compared with HR group.

FIGURE 2 | Effect of high fat diet, rutin and inulin supplementation on inflammatory cytokine profile. (A) Plasma inflammatory cytokine levels across the groups; (B) mRNA expression levels of inflammatory mediators in the small intestine tissue. The values with different letters indicated significant difference among groups (p < 0.05).

significant difference among groups (p<0.05).

(D) Ratio F/B was positively related with plasma LPS; (E) Principal coordinate analysis (PCoA) of 16S sequences from 32 intestine content samples of four treatments based on Weighted Unifrac. The microbiome of HR was similar to LC, HF feeding reduced significantly the abundance of Bacteroidetes and its families, favoring the presence of Lachnospiraceae family in Firmicutes and Deferribacteres, and resulted in higher Firmicutes/Bacteroidetes (F/B) ratio in intestine, while HR and HRI group increased the abundance of Bacteroidetes and its families, and attenuated the rise of F/B ratio and Deferribacteres, and HR group increased the growth of Proteobacteria and its family. <sup>∗</sup> indicated significant difference at p < 0.05, and ∗∗ indicated significant difference at p < 0.01 compared with LC.

TABLE 2 | Effect of high fat diet, rutin and inulin supplementation on colonic contents of SCFA in mice (µmol/g).


In each line, the values with different letters indicated significant difference among groups (p<0.05).

# DISCUSSION

The dysbiosis of gut microbiota is an important mediator in the high fat-induced inflammation, and thus manipulation of the gut microbiota, using prebiotics, may provide novel preventive strategies for obesity-associated inflammation and medical disorders. Both polyphenols and fibers, often eaten together, have been reported to have prebiotic actions. The present study demonstrated that the supplementation of polyphenol rutin suppressed the rise of F/B ratio, Deferribacteracea population and plasma LPS induced by HF diet, and thereby mitigated systematic and intestinal inflammation, and ER stress in Paneth cells. The co-administration with polysaccharide inulin improved the utilization of rutin and further increased the

production of butyrate and reduced the expression of the key inflammatory cytokine TNF-α, indicating the polyphenol rutin and polysaccharide inulin can be utilized combinatorially as a dietary strategy to ameliorate gut dysbiosis and inflammation associated with HF-induced obesity.

In the present study, after feeding a 60 kcal% HF diet for 20 weeks, the animals exhibited a significant increase of body weight with increased plasma cholesterol, triglycerides, leptin and LPS contents, and displayed a systematic and intestinal inflammation as indicated by increased circulating inflammatory cytokines and

elevated expressions of inflammatory mediators in the intestinal epithelial cells. Meanwhile, HF diet significantly increased mRNA expressions of Paneth AMPs, which were positively associated with plasma LPS and inflammatory mediators, suggesting that a critical role of Paneth cell AMPs in promoting obesity-associated inflammation. HF diet also induced a significantly increase of ER stress as indicated by the higher mRNA expressions of ER stress biomarkers, BiP and ATF4, which resulted in a discordance between protein expression and mRNA expression of lysozyme. Similarly, Hodin et al. (2011) reported ER stress was apparent in Paneth cells in obese subjects, indicated by diminished AMPs protein expression and increased mRNA levels of their corresponding genes.

Corresponding to those metabolic and inflammatory changes, a shift of the gut microbiota was observed in the animals fed with HF diet comparing to the LF diet group, with elevated F/B ratio, similar to the alteration observed in obese individuals induced by high-fat (Zhang et al., 2012; Guo et al., 2017) and high fat/high-sucrose diet (Parks et al., 2013). The increase of Firmicutes in HC was largely due to increase of Lachnospiraceae in this study, which was similar to results of Han et al. (2018) indicating Lachnospiraceae is the main intestinal bacterial family, and accounts over thirty percent of the total in the microbiota. Correlation analysis indicated F/B ratio and abundance of Lachnospiraceae (R = 0.41, p < 0.05) were correlated with plasma LPS in the present study. These findings were consistent with previous studies, which demonstrated that HF diet-induced gut microbiota dysbiosis (shifting to Firmicutes) led to an increase in gut permeability and plasma LPS concentration, and thereby promoted a low-grade inflammation (Cani et al., 2007, 2008; Creely et al., 2007).

Rutin supplementation (HR group) or its co-administration with inulin (HRI group) mitigated the increase of plasma triglycerides or leptin, attenuated inflammatory status, and improved the ER stress in Paneth cells induced by high-fat diet (the HC group) in this study. These data suggested that the inflammatory status and ER stress in Paneth cells in an obese state could be compromised by rutin supplementation or its co-administration with inulin in HF diet.

The characteristics of the gut microbiota in rutinsupplemented mice were similar to LC group according to PCoA plot, suggesting the addition of rutin could attenuate gut dysbiosis induced by HF diet. Rutin supplementation in HF diet suppressed the reduction of Bacteroidetes and its families, and ameliorated the elevation of Lachnospiraceae, Firmicutes/Bacteroidetes (F/B) ratio, Deferribacteraceae and plasma LPS induced by high-fat diet. Our finding was supported by a number of previous studies that demonstrated that the influences of polyphenols on intestinal bacteria: purple lettuces administration, which contains high flavonoid content, decreased Lachnospiraceae and Deferribacteraceae (Han et al., 2018); regular wine vinegar ingestion or polyphenol-rich fruits and green tea favored the growth of Bacteroidetes community (Lee et al., 2006; Cerezo et al., 2008; Selma et al., 2009); and supplementation of quercetin or grape polyphenols stimulated the proliferation of bifidobacteria and attenuated the rise of F/B ratio and thereby ameliorated obesity-associated inflammation and metabolic disorders (Parkar et al., 2013; Etxeberria et al., 2015; Roopchand et al., 2015).

Polyphenols and dietary fiber are often eaten together, and interact to mediate gut microbiota. In the present study, the co-administration of polysaccharide inulin with rutin increased the breakdown of the rutin, indicated by the lower rutin left in feces when comparing the HRI with the HR group. This result was consistent with a prior in vitro study, which demonstrated that fermentable fibers could speed up the breakdown of the rutin (Jaganath et al., 2006). As expected, the combinatorial supplementation of inulin with rutin magnified the effects of rutin supplementation alone: elevated the production of butyrate and reduced the intestinal expression of TNF-α (p < 0.05).

Unfavorably, in this study, rutin supplement ameliorated the increase of body weight triggered by high-fat diet, while the co-administration with inulin reversed the decrease of body weight induced by rutin supplementation. This might largely be due to food intake elevated by sweet taste of inulin. Nevertheless, the combination of rutin and inulin further elevated the growth of Bacteroidaceae and Lactobacillus, and recovered the numbers of SCFA -producing bacterium Lachnospiraceae reduced by rutin addition (**Figure 4C** and **Supplementary Figure S1**). The bloom of Lachnospiraceae was largely accounted by increases of Lachnospiraceae\_NK4A136\_group, Lachnoclostridium, Roseburia and Blautia, most of which are butyrate producers (Million et al., 2018). Previous studies also demonstrated that inulin supplementation favored Lactobacillus and SCFA-producing bacteria Lachnospiraceae or Roseburia, and Bacteroides (Aguirre et al., 2016; Zhang et al., 2018). The HRI group also reduced a number of harmful bacteria including Deferribacteraceae and Desulfovibrionaceae. The further decrease in Deferribacteraceae in HRI was largely accounted for reduction of Mucispirillum genus in this study, which is known as mucin degrader and associated with early disruption of the colonic surface mucus layer, prior to the onset of symptomatic colitis (Belzer et al., 2014). Most members of Desulfovibrionaceae are reported as lipopolysaccharide (LPS) producers and lead to a low grade and chronic inflammation in obese objects (Xiao et al., 2014; Zhang-Sun et al., 2015). These results were in agreement with previous studies that reported Mucispirillum (belong to Deferribacteracea) and Desulfovibrionaceae abundances were significantly suppressed by inulin treatment (Yu et al., 2018; Zhang et al., 2018).

#### CONCLUSION

In summary, the present study demonstrated that rutin supplementation suppressed the rise of F/B ratio, Deferribacteracea and plasma LPS induced by HF diet, and thereby mitigated systematic and intestinal inflammation, and ER stress in Paneth cells. The co-administration of rutin with inulin improved the utilization of rutin as indicated by its decreased excretion, further suppressed a number of harmful bacteria, reduced the expression of the key inflammatory cytokine TNF-α and increased the production of butyrate. Taken together, our data demonstrated the potential to combine polyphenol rutin and the polysaccharide inulin as a dietary strategy to ameliorate gut dysbiosis, to improve inflammatory status and thereby to reduce medical disorders associated with HF-induced obesity.

#### ETHICS STATEMENT

fmicb-09-02651 November 1, 2018 Time: 18:53 # 10

This study was carried out in accordance with the laboratory animal-guildline for ethical review for animal welfare in China. The animal protocol was approved by the Institutional Animal Care and Use Committee of Chengdu University (Permission No. 2018-23-01).

# AUTHOR CONTRIBUTIONS

XG and ZL designed the research. YL and JL performed the research. RT and SY analyzed the data. XG wrote the paper. ZL supervised the manuscript.

### REFERENCES


#### FUNDING

This project was supported in part by National Natural Science Foundation of China (31401488, XG) and USDA grant (2014- 67017-21762, ZL).

#### ACKNOWLEDGMENTS

We acknowledge the Novogene Genome Sequencing Company (Beijing, China) for 16S rRNA gene sequencing and analysis of microbial profile.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.02651/full#supplementary-material



gut bacterium akkermansia muciniphila and attenuate high-fat diet-induced metabolic syndrome. Diabetes Metab. Res. Rev. 64, 2847–2858. doi: 10.2337/ db14-1916


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling Editor declared a shared affiliation, though no other collaboration, with one of the authors SY at the time of review.

Copyright © 2018 Guo, Tang, Yang, Lu, Luo and Liu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Serine Alleviates Dextran Sulfate Sodium-Induced Colitis and Regulates the Gut Microbiota in Mice

Haiwen Zhang1,2, Rui Hua1,2, Bingxi Zhang1,2, Xiaomeng Zhang1,2, Hui Yang1,2 and Xihong Zhou<sup>3</sup> \*

<sup>1</sup> Key Laboratory of Tropical Animal Breeding and Epidemic Disease Research of Hainan Province, Hainan University, Haikou, China, <sup>2</sup> Key Laboratory of Tropical Biological Resources of Ministry of Education, Haikou, China, <sup>3</sup> Key Laboratory of Agro-ecological Processes in Subtropical Region, Institute of Subtropical Agriculture, Chinese Academy of Sciences, Changsha, China

Serine alleviates inflammatory responses and is beneficial for gut health; however, whether it exerts any effects on ulcerative colitis or regulates intestinal microbiota remains unknown. We investigated the effects of serine supplementation on colonic morphology, inflammation, and microbiota composition in dextran sulfate sodium (DSS) induced colitis model in mice. Acute colitis was induced through the oral intake of 3.5% DSS in water for 7 days. Mice with acute colitis were divided into two groups; The DSS and Ser-treated groups were rectally administrated with PBS or 1% (w/v) serine (40 mg/kg body weight) for 7 days. The results showed that serine decreased the disease activity index, as well as myeloperoxidase, eosinophil peroxidase, and proinflammatory cytokine concentrations in colonic tissue, while serine improved colonic morphology and inhibited cell apoptosis in colitis mice. In addition, 16S rRNA phylogenetic sequencing revealed a shift in bacterial community composition, and changes in microbiota functional profiles following serine supplementation, although no significant difference in α-diversity analysis was observed. The effects of serine supplementation helped on the recovery of major perturbations to macrobiotic functions, such as amino acids metabolism; tissue replication and repair; and cell growth and death. Serine might have great potential for the renewal of colonic tissue in DSS-induced colitis.

#### Keywords: colitis, inflammation, microbiome, morphology, serine

#### INTRODUCTION

Ulcerative colitis (UC) is one of the major forms of inflammatory bowel disease (IBD) characterized by inflammation of the rectal and colonic mucosa. Although the etiology of UC is unclear, the consensus is that the combined effects of several factors, such as genetic susceptibility factors, impaired intestinal integrity, dysfunctional immune responses and their interactions with intestinal microbiota, as well as environmental factors contribute to the occurrence and development of this disease (Munyaka et al., 2016). In particular, alterations in gut microbiota composition and functions have been demonstrated to play critical roles in human IBD at different stages of the disease (Tamboli et al., 2004; Morgan et al., 2012).

Edited by:

Helieh S. Oz, University of Kentucky, United States

#### Reviewed by:

Kai Wang, Institute of Apiculture Research (CAAS), China Zhenlong Wu, China Agricultural University, China

> \*Correspondence: Xihong Zhou xhzhou@isa.ac.cn

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 01 September 2018 Accepted: 27 November 2018 Published: 10 December 2018

#### Citation:

Zhang H, Hua R, Zhang B, Zhang X, Yang H and Zhou X (2018) Serine Alleviates Dextran Sulfate Sodium-Induced Colitis and Regulates the Gut Microbiota in Mice. Front. Microbiol. 9:3062. doi: 10.3389/fmicb.2018.03062

Various therapies involving alleviation of inflammation and restoration of the intestinal microbiota have been suggested for the prevention and treatment of UC, including probiotics, prebiotics and dietary supplements such as phytonutrients and propolis (Yeom et al., 2016; Rodriguez-Nogales et al., 2017; Wang et al., 2017). Serine, long considered as a nutritionally non-essential amino acid, has recently been suggested to act as a conditionally functional amino acid. Previous studies have demonstrated that dietary serine helps alleviate triglyceride accumulation and oxidative damage in the liver of several rodent models (Sim et al., 2015; Zhou et al., 2017a, 2018; Cao et al., 2018). Moreover, our recent study showed that dietary serine prevented inflammation in the small intestine and maintained barrier integrity (Zhou et al., 2017b). However, whether serine has any effects on colonic inflammation and gut microbiota is yet to be explored.

Most studies examining the effect of dietary supplements on colitis alleviation have used the most common rodent model involving UC-related erosion and inflammation induced by chemicals such as dextran sulfate sodium (DSS) (Hakansson et al., 2015; Munyaka et al., 2016). Consequently, the present study was conducted to determine the effects of serine on intestinal integrity, inflammation, and microbial dysbiosis in DSS-induced colitis. Furthermore, the related functional pathways were also explored. Our results will provide evidences for the therapeutic potential of serine for regulating dysbiosis in UC.

#### MATERIALS AND METHODS

#### Chemicals and Kits

Dextran sulfate sodium salt (DSS) and L-serine were purchased from Sigma-Aldrich, (Shanghai, China). The animal diet was purchase from Research Diets, Inc. (New Brunswick, NJ, United States). Cell death detection kit was purchased from Roche (Shanghai, China). ELISA quantitative kits for the detection of IgA, IgG, IgM, IL-1β, IL-6, TNF-α, myeloperoxidase (MPO), and eosinophil peroxidase (EPO) concentrations were purchased from Cusabio Biotech (Wuhan, Hubei, China<sup>1</sup> ). The QIAamp DNA stool MiniKit and Gel Extraction Kit were purchased from Qiagen (Hilden, Germany). The Protein Extract Kit was purchased from Keygen (Nanjing, China). Caspase 3, proliferating cell nuclear antigen (PCNA), mammalian target of rapamycin complex I (mTORC1), and general control nonderepressible 2 (GCN2) antibodies were purchased from Cell Signaling (Beverly, MA, United States). EZ-ECL was purchased from Biological Industries (Cromwell, CT, United States).

#### Animal Experiments

Twenty-one 9-week-old male C57BL/6 mice were obtained from SLAC Laboratory Animal Central (Changsha, China) and maintained in plastic cages under standard conditions. Standard pelleted diets (Research Diets, Inc.) were supplied ad libitum. Acute colitis was induced through the oral intake of 3.5% DSS (w/v, molecular mass of 6,500–10,000 Da; Sigma-Aldrich) in fresh running water ad libitum for 7 days. Mice with acute colitis were divided into two groups; The DSS and Ser-treated groups (n = 7) were rectally administrated with PBS or 1% (w/v) serine (40 mg/kg body weight) dissolved in PBS once daily for 7 days. Since previous report indicated that almost all of the amino acids were absorbed or catabolized by the small intestine (Wu, 2009) and our preliminary experiments suggested that oral administration of serine had no significant effects on DSS-induced colitis, we finally supplemented serine via rectal administration which is also widely used for the treatment of DSS-induce colitis in mouse model (Tai et al., 2007; Han et al., 2015). In addition, a Control group (n = 7) without acute colitis was also rectally administrated with PBS. Mice were kept inversely for 1 min after administration to prevent leakage from the anus as previously did (Tai et al., 2007; Han et al., 2015). Upon completion of the experiment on day 14, blood was obtained from the retro-orbital sinus. Thereafter, the mice were euthanized by cervical dislocation and the colon and colonic digesta were collected. The experimental protocol was approved by the Protocol Management and Review Committee of Institute of Subtropical Agriculture, Chinese Academy of Sciences and mice were cared for and sacrificed according to the animal care guidelines of Institute of Subtropical Agriculture (Changsha, China).

#### Assessment of Disease Activity

The disease activity index (DAI) based on body weight change, stool consistency, rectal bleeding, and overall condition of the animal during the experiment was calculated according to a standard scoring system (Zhang et al., 2015). Scores were recorded using the following criteria: Weight loss: 0, no weight loss; 1, weight loss of 0.1–5% (compared to baseline); 2, 5–10%; and 3, >10%. Stool consistency: 0, well-formed pellets; 2, pasty and semi-formed stools that did not adhere to the anus; and 3, liquid stools that adhered to the anus. Rectal bleeding: 0, no blood; 1, small amount of blood in some stool; 2, blood regularly observed in stool; and 3, blood in all stool.

#### Histological Analyses

The colon (3.5–4 cm proximal to the anus) was collected and flushed with chilled PBS. Subsequently, the samples were fixed with 10% buffered formalin and embedded in paraffin. Next, 8-µm sections were obtained and stained with hematoxylin and eosin (HE) (Hu et al., 2017). Finally, colonic morphology was observed under light microscopy and representative pictures were taken. Colonic samples were also fixed in 2.5% glutaraldehyde, then further fixed with osmium tetroxide and dehydrated with graded alcohol. Subsequently, an epon-araldite resin was used to embed the dehydrated samples. Ultrathin sections were obtained and stained with uranyl acetate and lead citrate. Finally, alteration of the mitochondrial structure was observed and representative images were obtained using a Zeiss 902 transmission electron microscope (Zeiss, Thornwood, NY, United States). The histological scoring of colitis was performed according to previous study did (Kitajima et al., 2000).

<sup>1</sup>https://www.cusabio.com/

### Assessment of Apoptosis

fmicb-09-03062 December 7, 2018 Time: 17:19 # 3

Colonic tissues were fixed in 10% formaldehyde and embedded in paraffin and then 5-µm sections were obtained. Apoptosis was detected by TUNEL staining using an in situ cell death detection kit (Roche). DAPI mounting solution (Vector, Burlingame, CA, United States) was used for nuclei staining. The results were observed under a light microscope and representative pictures were photographed.

#### Measurement of Immunoglobulin, Inflammatory Cytokine, Myeloperoxidase, and Eosinophil Peroxidase Concentrations

Immunoglobulin (Ig)A, IgG, and IgM concentrations in sera and IL-1β, IL-6, TNF-α, MPO, and EPO concentrations in the distal colon (2–3.5 cm proximal to the anus) were determined using ELISA quantitative kits (Cusabio Biotech) according to the manufacturer's instructions.

#### Gut Microbiota Profiling

All content within the colon were pooled and homogenized, and then colonic content DNA was extracted using the QIAamp DNA stool MiniKit (Qiagen). DNA samples were further purified from DSS using lithium chloride as previous study did (Viennois et al., 2013). Briefly, DNA were incubated with 0.1 volume of 8 M LiCl diluted in diethylpyrocarbonate (DEPC)-treated water on ice for 2 h and then centrifuged at 14,000 × g for 30 min at 4◦C. The supernatants were discarded and the remaining DNA were dissolved in DEPC-treated water. Then the abovementioned procedure was repeated once more. Bacterial 16S rRNA gene sequences (V3–V4 region) were amplified using specific primers containing a barcode. PCR reactions were performed in a final volume of 50 µL consisting of 12.5 µL of Phusion High-Fidelity PCR Master Mix (New England BioLabs Inc., Beverly, MA, United States), 50 ng of template DNA, 1 µL of each primer, and PCR-grade water. Further experiments were carried out with the 400–450 bp PCR products purified using the Qiagen Gel Extraction Kit (Qiagen). Subsequently, MiSeq Illumina sequencing was performed using the IlluminaHiSeq2500 platform (Illumina Inc., San Diego, CA, United States) and 250 bp paired-end reads were obtained. Next, the paired-end reads were merged using FLASH and then assigned to each sample based on their unique barcodes. Highquality clean tags were clustered into (OTUs) using USEARCH according to the QIIME quality-controlled process based on 97% sequence similarity and representative OTUs were used for further analysis using the Greengenes database with the RDP algorithm. Alpha and beta diversity and principal coordinate analysis (PCoA) were performed with QIIME. OTUs were also used for genome prediction of microbial communities by PICRUSt (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States). Each OTU was used to search the pre-calculated genome content for metagenome prediction. KEGG abundances for each metagenome sample were obtained and KEGG levels 1, 2, and 3 were identified to determine the metagenome response to DSS and serine supplementation.

#### Western Blot Analysis

Distal colonic samples (2–3.5 cm proximal to the anus) were ground and lysed using a total protein extract kit (Keygen). Each sample (30 µg protein) was separated by SDS-PAGE and

transferred to a nitrocellulose membrane (Wang et al., 2016). The membrane was first incubated with primary antibodies against Caspase 3, PCNA, mTORC1, and GCN2 (Cell Signaling) overnight at 4◦C and then incubated with the secondary antibody for 2 h at 25◦C. The results were visualized via chemiluminescence with EZ-ECL (Biological Industries).

#### Statistical Analysis

fmicb-09-03062 December 7, 2018 Time: 17:19 # 4

All statistical analyses were performed by one-way ANOVA using the general linear model procedures and a mixed procedure (PROCMIXED) of SAS software version 9.2 (SAS Institute Inc., Cary, NC, United States). Data are presented as least squares means ± SEM. Mean values were considered significantly different when P < 0.05, and 0.05 < P < 0.1 was considered as a tendency.

#### RESULTS

#### Serine Alleviates Colitis Symptoms

The effects of serine supplementation on colitis symptoms are shown in **Figure 1**. By day 14, serine supplementation reversed DSS-induced weight loss (**Figure 1B**). No significant differences in colon length, colon weight, and colon length/colon weight were observed between control mice and serinesupplemented mice (**Figures 1C–E**). During the experiment, serine supplementation lowered the DAI from day 9 to day 14 (**Figure 1F**).

### Serine Alleviates DSS-Induced Histopathological Changes and Apoptosis

The effects of serine supplementation on DSS-induced histopathological changes and apoptosis are shown in **Figure 2**. The HE staining results demonstrated pathological changes, such as obvious edema following DSS treatment, in the structure of the mucous layer in the distal colon tissue, while no such changes were observed in serine-supplemented mice. In addition, the TEM results revealed severe mitochondrial edema in the colon tissue of DSS-treated mice. However, serine supplementation alleviated these changes in the mitochondria. Moreover, TUNEL staining showed that the level of apoptosis was higher in DSS-induced mice than in serine-supplemented mice.

#### Serine Increases Immunoglobulin Content and Decreases Content of Inflammatory Cytokine and Colonic Infiltration Marker

Dextran sulfate sodium treatment significantly decreased IgA, IgG, and IgM concentrations (**Figures 3A–C**) in serum,

FIGURE 2 | Serine alleviated DSS-induced histopathological changes and apoptosis. (Upper) HE staining of ileum morphology (40×), histological scoring was obtained from the HE results. Arrow, obvious edema. (Middle) ultrastructural observation of colon (transmission electron microscopy, 10,000×). Arrow, mitochondria. (Lower) TUNEL staining (yellow) for assessment of apoptosis (400×), nuclei were stained with DAPI (blue) and cell apoptosis rate was calculated. Control, mice was rectally administrated with PBS without pretreatment with dextran sulfate sodium; DSS, mice was rectally administrated with PBS after pretreatment with dextran sulfate sodium; Ser, mice was rectally administrated with serine after pretreatment with dextran sulfate sodium.

while increasing IL-1β, IL-6, and TNF-α concentrations in colonic tissue (**Figures 3D–F**). However, serine supplementation alleviated these DSS-induced changes. In addition, there was a significant increase in MPO and EPO concentrations in the colonic tissue of DSS-treated mice (**Figures 3G,H**). Supplementing DSS-treated mice with serine significantly decreased MPO and EPO concentrations.

# Serine Alters Colonic Microbiota Composition in Colitis Mice

A 16S rRNA phylogenetic approach was used to compare the colonic microbial population of mice from different treatment groups. OTUs were generated from sequences with at least 97% similarity. The alpha diversity of the microbial communities, as indicated by the Shannon index, tended to decrease in DSS-treated mice, while no significant difference was observed between control mice and serine-supplemented mice (**Figure 4A**). Moreover, we observed a significant difference in beta-diversity among the three treatments based on the weighted PCoA results (**Figure 4B**). In addition, the results showed that at the class and order levels, Clostridia and Bacteroidia were the main orders in control mice and serine-supplemented mice, while a decrease of Clostridiales was observed in DSS-treated mice (**Figures 4C,D**). At the phylum level, Bacteroidetes and Firmicutes were the main phyla (**Figure 4E**). Firmicutes were significantly decreased in DSS-treated mice, while no significant difference was observed between control mice and serine-supplemented mice. The Venn diagram results show that among a total of 680 detected OTUs, 295 OTUs were universal to all samples and there were 75 unique OTUs in serine-supplemented mice (**Figure 4F**).

# Biofunction Prediction of Microbial Communities

PICRUSt was performed to predict the functional profiles of the microbial communities. PCoA was generated for different KEGG levels (level 2, **Figure 5A**; level 3, **Figure 5B**) to evaluate differences in KEGG abundance among the functional profiles of different treatments. The results showed that the metagenome was greatly modulated in response to DSS treatment and serine supplementation. The most significant KEGG pathway types were cellular processes (predominantly replication and repair, cell growth, and death in level 2) and metabolism (predominantly amino acid and nucleotide metabolism in level 2) gene pathways (**Figures 5C,D**). These pathways were further analyzed in KEGG level 3. The results showed that serine supplementation alleviated the increased activities of glycolysis/gluconeogenesis and glycine,

serine, and threonine metabolism, while improving the decreased activities of pyruvate and purine metabolism, DNA repair and recombination proteins, and DNA replication proteins, which were all induced by DSS treatment (**Figure 5E**).

### Effects of Serine Supplementation on the Expression of Apoptosis and Proliferation Markers and Proteins Involved in Amino Acid Metabolism

To further explore whether changes in the metabolic pathways of microbial communities affected mucosal renewal and amino acid metabolism in the small intestine, the protein expression of Caspase 3, PCNA, GCN2, and mTORC1 was examined. The results showed that serine supplementation alleviated the increased expression of Caspase 3 and GCN2 (**Figures 6A,B,D**), while improving the decreased expression of PCNA and mTORC1 expression in DSS-induced mice (**Figures 6A,C,E**).

# DISCUSSION

Dextran sulfate sodium induction is the most widely used experimental model for UC. In these animal models, DAI value

and colonic weight and length are critical markers for evaluating colitis severity (Osada et al., 2008). In the present study, DSS induced a significantly higher DAI (weight loss, gross rectal bleeding, and severe diarrhea) in mice. Moreover, in addition to decreased colonic weight and length, the colonic length and weight ratio was also decreased. These results indicate that DSS treatment results in overt features of colitis. Recently, L-serine has been generally considered as safe by the United States Food and Drug Administration since patients did not show any side effects in phase II clinical trials (Metcalf et al., 2018). In addition, our previous studies showed that mice supplemented with 1% serine in the drinking water did not showed any toxic effects on liver and small intestine (Zhou et al., 2017a,b). These results suggested that administration of serine is safe when it is used in mice. Although no previous reports have indicated that L-serine has any effects on UC, our previous study demonstrated an anti-inflammatory effect on LPS-treated

mice. Thus, as expected, we found that the DSS-induced changes were reversed following L-serine supplementation for 7 days.

Dextran sulfate sodium directly targets the intestinal mucosa, damaging the intestinal epithelial cells of the basal crypts and impairing mucosal barrier integrity (Wirtz and Neurath, 2007). Moreover, recent studies have highlighted that remarkable changes in gut microbiota composition trigger intestinal inflammation in DSS-induced colitis (Hakansson et al., 2015; Munyaka et al., 2016). Therefore, we investigated the effects of serine on histopathological changes in the colon and gut microbiota dysbiosis following DSS treatment; our results demonstrated that serine improved intestinal integrity and gut microbiota composition in DSS-induced colitis. In addition, the biofunction prediction of microbial communities further demonstrated DSS-induced functional alteration in colonic microbiota and that the bacterial functional activities or

metabolic pathways in serine-treated mice are more similar to those in control mice than in DSS-induced mice.

It has been widely reported that bacterial species diversity is reduced in both fecal and intestinal mucosa-associated microbiota samples from human patients or animal models with UC (Manichanh et al., 2006; Andoh et al., 2007; Samanta et al., 2012). However, a study using a pool of colonic and cecal content showed no significant difference in α-diversity analysis (Berry et al., 2012). Our results are consistent with this study as the results showed that DSS only tended to decrease intestinal microbial community evenness (Shannon H). These inconsistences suggest that the effects of DSS on microbiota composition may vary according to the anatomical site from which the sample is collected or whether the samples are from mucosa or feces. Nevertheless, a shift in bacterial community composition was confirmed in all these previous studies, as well as in the present study. In addition to reduced bacterial diversity, an increase in the Bacteroidetes to Firmicutes ratio also constitutes a significant characteristic of UC (Wang et al., 2017). Our results for samples from DSS-treated mice were consistent with this alteration (a reduction in Clostridia at the class level and in Firmicutes at the phylum level), while serine supplementation alleviated these changes. These results suggest that serine helps reverse the DSS-induced decrease in microbial evenness and community composition changes.

Prediction of metagenome functional composition showed that there was a difference in major perturbation in all three treatments. Alterations in amino acid metabolism were previously reported in UC patients (Morgan et al., 2012). In the present study, serine supplementation reversed the increased abundance of amino acid metabolism genes and decreased the abundance of nucleotide metabolism genes in DSS-treated mice. Serine not only links biosynthetic flux from glycolysis to purine synthesis (Parker and Metallo, 2016), but is also the major substrate for either glycine or pyruvate synthesis (El-Hattab, 2016). These critical steps of intermediary metabolism may be the reason that serine helps restore glycolysis; glycine, serine, and threonine metabolism; and pyruvate and purine metabolism to close to normal levels. In addition, alteration of microbiome amino acid metabolism may further affect amino acid requirements in colonic tissue. The increased expression of GCN2 and the decreased expression of mTORC1 suggest a lack of amino acids in the colonic tissue of DSS-treated mice. Interestingly, serine supplementation also alleviated these changes in colonic tissue in addition to its beneficial effects on amino acid metabolism in microbiota. A decrease in purine biosynthesis modules was reported in UC, consistent with our observations (Morgan et al., 2012), while serine supplementation alleviated this decrease. Serine is the major source of 1-C units for de novo synthesis of purine nucleotides and deoxythymidine monophosphate (de Koning et al., 2003). These substrates, as well as serine metabolism products, are essential for cell proliferation. The colonic microbiota of serinesupplemented mice was associated with increased DNA repair and recombination proteins and DNA replication proteins, leading us to further explore the effects of serine on cellular DNA

replication and repair and proliferation (PCNA) and apoptosis (Caspase 3) markers in colonic tissue. Unlike amino acid metabolism, the replication and repair and cell growth and death functions in the microbiome were consistent with decreased apoptosis and increased proliferation in colonic tissue. These results indicated that the increased DNA repair and replication proteins by the colonic microbiota of serine-supplemented mice, according to the prediction of metagenome functional composition, had beneficial effects on the renewal of colonic tissue.

Relapsing episodes of UC are related to over-production of proinflammatory cytokines, which further impair intestinal permeability and cause severe colonic infiltration and dysfunctional gut barrier. Impaired gut epithelial barrier function may lead to the augment of inflammatory responses and permeability (Hering and Schulzke, 2009). Following serine supplementation, the increased concentrations of proinflammatory cytokines, MPO and EPO (indicators of colonic infiltration with polymorphonuclear leukocytes and eosinophils, respectively) (Han et al., 2015; Zhang et al., 2015) in the colon tissue of DSS-treated mice were reversed. These results indirectly indicate that serine supplementation help the recovery of intestinal integrity and barrier function, although we did not detect the expression of tight junction proteins which can directly reflect gut barrier function. In addition, the recovery of colonic morphology and the decreased apoptotic rate also indicated that the colonic dysfunction may alleviate by serine supplementation in DSS-treated mice.

#### CONCLUSION

Our results show that serine supplementation reversed DSSinduced colitis in mice. To the best of our knowledge this is the first study to demonstrate that serine supplementation alters the composition of the colonic microbiota, as well as their functional profiles in colitis mice. Importantly, the effects of serine on the

#### REFERENCES


recovery of major perturbations of microbiome functions, such as glycine, serine, and threonine metabolism; replication and repair; and cell growth and death, may contribute to the renewal of colonic tissue in DSS-induced colitis. These results support the therapeutic potential of serine for regulating dysbiosis in UC. However, we unfortunately did not have a serine-only treated group since our results showed the unexpected and remarkable effects of serine on the microbiota composition in DSS-induced colitis mice in the present study. Because these results indicated that serine may be important for the composition of microbiota and their metabolism in the intestine, future works are suggested to explore the effects of dietary serine supplementation or serine deficiency on microbiota composition. In addition, future works are also needed to fully elucidate the mechanisms related to the effects of serine on colitis and to explore the beneficial effects of serine on other mouse model of colitis.

#### AUTHOR CONTRIBUTIONS

XhZ and HZ conceived and designed the research and critically revised the manuscript. HZ, RH, and BZ drafted the protocol. HZ, XhZ, HY, and XmZ contributed to the literature search, interpretation, writing, and proofreading of the manuscript.

#### FUNDING

This research was funded by Young Talents' Science and Technology Innovation Project of Hainan Association for Science and Technology (QCXM 201804), National Key Research and Development Program of China (2018YFD0500405), National Natural Science Foundation of China (31702125), Natural Science Foundation of Hunan Province (2017JJ3373), and the earmarked fund for China Agriculture Research System (CARS-35).

sulfate sodium (DSS) administration in mice. Clin. Exp. Med. 15, 107–120. doi: 10.1007/s10238-013-0270-5



modulates the gut microbiota in rats fed a Western Diet. Nutrients 9:E875. doi: 10.3390/nu9080875


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Zhang, Hua, Zhang, Zhang, Yang and Zhou. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Anti-breast Cancer Enhancement of a Polysaccharide From Spore of *Ganoderma lucidum* With Paclitaxel: Suppression on Tumor Metabolism With Gut Microbiota Reshaping

#### *Edited by:*

*Helieh S. Oz, University of Kentucky, United States*

#### *Reviewed by:*

*Yunping Qiu, Albert Einstein College of Medicine, United States Mingming Su, University of Hawaii, United States Michelle Martinez-Montemayor, Central University of the Caribbean, Puerto Rico*

#### *\*Correspondence:*

*Yizhen Xie xyzgdim@sina.com Xinxin Zhou xinxin\_zhou@163.com*

*†These authors have contributed equally to this work*

#### *Specialty section:*

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 27 July 2018 Accepted: 30 November 2018 Published: 17 December 2018*

#### *Citation:*

*Su J, Li D, Chen Q, Li M, Su L, Luo T, Liang D, Lai G, Shuai O, Jiao C, Wu Q, Xie Y and Zhou X (2018) Anti-breast Cancer Enhancement of a Polysaccharide From Spore of Ganoderma lucidum With Paclitaxel: Suppression on Tumor Metabolism With Gut Microbiota Reshaping. Front. Microbiol. 9:3099. doi: 10.3389/fmicb.2018.03099*

Jiyan Su1†, Dan Li 2,3†, Qianjun Chen<sup>4</sup> , Muxia Li 2,3, Lu Su<sup>5</sup> , Ting Luo<sup>6</sup> , Danling Liang2,3 , Guoxiao Lai 3,7, Ou Shuai <sup>3</sup> , Chunwei Jiao<sup>3</sup> , Qingping Wu<sup>1</sup> , Yizhen Xie1,3 \* and Xinxin Zhou<sup>2</sup> \*

*<sup>1</sup> State Key Laboratory of Applied Microbiology Southern China, Guangdong Provincial Key Laboratory of Microbial Culture Collection and Application, Guangdong Institute of Microbiology, Guangzhou, China, <sup>2</sup> School of Pharmaceutical Science, Guangzhou University of Chinese Medicine, Guangzhou, China, <sup>3</sup> Guangdong Yuewei Edible Fungi Technology Co. Ltd., Guangzhou, China, <sup>4</sup> Department of Breast Disease, Guangdong Provincial Hospital of Chinese Medicine, Guangzhou University of Chinese Medicine, Guangzhou, China, <sup>5</sup> School of Pharmacy and Chemistry, Dali University, Dali, China, <sup>6</sup> Guangdong Laboratory Animals Monitoring Institute, Guangzhou, China, <sup>7</sup> School of Pharmacy, Guangxi University of Chinese Medicine, Xining, China*

Increasing evidence highlights the cardinal role of gut microbiota in tumorigenesis and chemotherapy outcomes. Paclitaxel (PTX), although as a first-line chemotherapy reagent for breast cancer, still requires for improvement on its efficacy and safety due to drug resistance and adverse effects. The present work explored the enhancement of a polysaccharide derived from spore of *Ganoderma lucidum* (SGP) with PTX in a murine 4T1-breast cancer model. Results showed that the combination of PTX and SGP displayed an improved tumor control, in which mRNA expression of several Warburg effect-related proteins, i.e., glucose transporter 3 (*Glut3*), lactate dehydrogenase A (*Ldha*), and pyruvate dehydrogenase kinase (*Pdk*), and the metabolite profile of tumor was evidently altered. Flowcytometry analysis revealed that the combination treatment recovered the exhausted tumor infiltration lymphocytes (TILs) via inhibiting the expressions of immune checkpoints (PD-1 and Tim-3), while PTX alone evidently increased that of CTLA-4. 16S rRNA sequencing revealed a restoration by the combination treatment on gut microbiota dysbiosis induced by PTX, especially that *Bacteroides*, *Ruminococcus,* and other 5 genera were significantly enriched while the cancer-risk genera, *Desulfovibrio* and *Odoribacter*, were decreased. Moreover, spearman correlation analysis showed that abundance of *Ruminococcus* was significantly negative-associated with the amount of frucotose-6-phosphate within the tumor. Collectively, the present study suggests the clinical implication of SGP as an adjuvant candidate for PTX against breast cancer, which possibly relies on the regulation of tumor metabolism and gut microbiota.

Keywords: spore of *Ganoderma lucidum*, paclitaxel, tumor metabolism, immune checkpoints, gut microbiota

# INTRODUCTION

Breast cancer, as one of the most threatening carcinoma, has taken the first place of cancer-related deaths in women population worldwide (Li et al., 2016; DeSantis et al., 2017). With the advances of experimental and clinical researches, strategies against breast cancer mainly include surgical resection, adjuvant chemotherapy, radiotherapy, and hormone therapy (PDQ Adult Treatment Editorial Board). Nevertheless, their efficacies remain unsatisfactory, not only due to the increasing drug resistance and adverse effects, but also because there are limitations in the application of certain strategies, especially that triple-negative breast cancer does not respond hormonal or trastuzumab-based therapies (Reddy, 2011). On the other hand, it is quite inspiring that several immunnotherapy reagents are undergoing clinical trials against breast cancer due to their success in other types of cancer (Katz and Alsharedi, 2017). However, there are still challenges ahead to widen its optimal selection of ideal candidates, in which the response efficacy is closely related to the gut microbiome (Gopalakrishnan et al., 2018; Routy et al., 2018b). Therefore, chemotherapies, mainly employing paclitaxel (PTX), are still the most common and costeffective treatments for breast cancer control. But adverse effects of PTX, such as hypersensitivity reactions, myelosuppression, and peripheral neuropathy, still disturb most of patients (Gupta et al., 2014; Starobova and Vetter, 2017). Hence, continued hard work is required for update of the current regimens, and even a revolution, to improve the efficacy meanwhile eliminating the adverse effects.

Nowadays, gut microbiota has been recognized as a regulator of both host metabolism and immune response (Gill et al., 2006). In particular, increasing evidence highlights its cardinal role in tumorigenesis (Sivan et al., 2015; Wong et al., 2017) and that in the outcomes of chemotherapy and immunotherapy (Gopalakrishnan et al., 2018; Routy et al., 2018b). This is mainly attributed to their intrinsic capcities of drug metabolism and the influence on host metabolizing homeostasis (Haiser et al., 2013; Wilson and Nicholson, 2017). Moreover, gut microbiota is found to be able to interfere tumor, especially that it is necessary to reverse the Warburg effect by accumulating butyrate within colorectal adenocarcinomas (Donohoe et al., 2014). Warburg effect is indeed a shift from oxidative phosphorylation to aerobic glycolysis (Warburg et al., 1924), which represented the metabolic nature of tumor microenvironment (TME) to support tumor growth and to evade immune destruction. It is characterized by increased glucose uptake and accumulation of lactate, even under normoxic conditions, as well as the corresponding upregulation of transporters, glycolytic enzymes and the relative signaling pathway proteins (Warburg, 1956; Ward and Thompson, 2012). Therefore, pathways and activities involved in tumor metabolism have been considered as novel targets in cancer therapy. For instance, metformin, a classic antidiabetic agent for treating type 2 diabetes, has been undergoing phase 2/3 clinical trials as adjuvant reagent in several cancer types, attributing to a reduction on the expression of monocarboxylate transporter 4 (MCT4) in cancer-associated fibroblasts (Romero et al., 2015). Meanwhile, convincing evidence has revealed that metformin would significantly change the gut microbiota community as well as the gut metabolome (Forslund et al., 2015; Wu et al., 2017). However, the concrete correlations between microbiota and metabolism underlying these observations remain largely unknown.

Ganoderma lucidum (Leyss. et Fr.) Karst. is one of the most extensively studied basidiomycotina mushroom as a functional food and chemopreventive agent, especially in traditional Chinese medicine and other Asian folk medicine (Oliveira et al., 2014). Numerous documents have revealed that G. lucidum exerts anti-cancer effects not only via cancer cell-targeting approaches, such as cell cycle arrest (Wu et al., 2012), apoptosis induction (Dai et al., 2017), and migration inhibition (Tsao and Hsu, 2016), but also, more importantly, through ways of immune enhancement (Li et al., 2015; Sun et al., 2015). Recently, active components from the spore of G. lucidum (SG) have been unveiled versatile biological activities owing to the advance in sporoderm-breaking technology, especially the activities contributing to its anticancer potential (Wang et al., 2012; Na et al., 2017). In our previous study, it was found that a polysaccharide from SG (SGP) was able to potentiate the cytotoxicity T cell (Tc)-based tumor immune surveillance with a benefit reshaping on gut microbiota (Su et al., 2018). In the present study, the improvement potential of SGP on the antitumor activity of PTX was investigated from the perspective of tumor metabolism and gut microbiota.

#### MATERIALS AND METHODS

#### Animals

Female Balb/c mice (6 to 8 week old, weighting 18–22 g) were provided by Guangdong Medical Laboratory Animal Center (Guangzhou, Guangdong, China). The mice were raised in specific pathogen-free condition (23 ± 2 ◦C, 50 ± 5% humidity) in a 12 h light/dark cycle with food and water ad libitum. After 7-day acclimatization, the experiment was performed with the approval by Guang-dong Institute of Microbiology Laboratory Animal Ethics Committee according to the guidelines (permission number: GT-IACUC201708231).

#### Preparation for Polysaccharide of the Sporoderm-Breaking Spore of *G. lucidum* (SGP)

SGP were prepared as described previously (Su et al., 2018). The sporoderm-breaking SG was provided by Guangdong Yuewei Edible Fungi Technology Co. Ltd. In brief, the spore was extracted with boiling distilled water. The extract was then concentrated, following by 2–3 cycles of precipitation with anhydrous ethanol (final percentage of ethanol was 85%), and dialysis. Finally, the 3.5–100 kDa dialysate was pooled, concentrated, and lyophilized, to obtain SGP with a yield of 0.4%. Polysaccharide content of SGP is about 50%, which is mainly made up of glucose with an average molecular weight (Mw) of 3.6 kDa as reported previously (Su et al., 2018).

# Cell Culture

Murine metastatic breast cancer 4T1 cell line was bought from Cell bank of Chinese Academy of Sciences, Shanghai, China. 4T1 cells were cultured in high glucose DMEM medium (4.5 mg/mL, Gibco, NY, USA) containing 10% fetal bovine serum (FBS, Gibco, NY, USA) and 1% penicillin/streptomycin (Gibco, NY, USA), and maintained in incubators at 37◦C under an atmosphere of 5% CO2.

# 4T1-Breast Cancer Model Induction and Treatment

Murine 4T1-breast cancer model was established as described by Zhang et al. (2017) with mild modification. Briefly, 4T1 tumor cells were injected subcutaneously (s.c.) into the right forleg armpit of the subjected Balb/c mice (0.1 mL/mouse, 1 × 10<sup>5</sup> cells/mouse). The tumor-bearing mice were randomly divided into Model group, PTX group (Hainan Quanxing Pharmaceutical Co. Ltd., Hainna, China) and the two combination treatment groups (SLP and SHP groups), 9 for each. Another 9 mice (as Normal group) were injected subcutaneously (s.c.) with 0.1 mL complete DMEM medium at the similar site. Over the following 21 days, PTX group was intraperitoneally injected (i. p.) with PTX at a dose of 12.5 mg/kg twice a week. The SLP group was administrated orally with SGP (200 mg/kg, p.o.) once a day, in addition to the twice-a-week intraperitoneal injection of PTX (12.5 mg/kg). The SHP group was administrated orally with SGP (400 mg/kg, p.o.) once a day, together with the PTX treatment (12.5 mg/kg). Normal group and Model group received equal volume of saline. In the following 21 days, tumor volume was measured with an electronic vernier caliper every 3 days since 6th day. The volume was calculated as V = a × b 2 /2, where a indicated the longer diameter, and b indicated the shorter diameter. On the 22th day, all animals were blooded from orbital plexus, and then sacrificed by cervical dislocation to harvest tumors. Tumors were weighted, photographed, segmented, and then stored according to different purposes immediately.

# Tumor Infiltrating Lymphocyte (TIL) Isolation and Flow Cytometry Analysis

Tumor segments kept in pre-cold PBS were used for TIL isolation and analysis. The segments were minced and digested in 3 mL digestive medium, which was mainly composed of basic RPMI160 medium supplemented with 0.1% Type IV collegenase (Invitrogen, Thermo Fisher Scientific, Grand Isle, NY, USA), 350 U/mL DNAse I (Roche, Basel, Switzerland), and 1% penicillinstreptomycin. Then they were ground in pre-cold PBS by passing through a 70µm strainer, washed, and resuspended in basic RPMI160 medium. TILs from the obtained cell suspension were separated with Mouse 1× Lymphocyte Separation Medium (Dakewe Biotechnology Co. Ltd., Shenzhen, China) according to the manufacture' instruction. TILs were stained with FITC antimouse CD3 (2.5 µg/test), PE- Cyanine5 anti-mouse CD4 (0.0625 µg/test), APC-Cyanine7 anti-mouse CD8 (0.25 µg/test), PE antimouse CD 152 (cytotoxic T-lymphocyte-associated protein-4, CTLA-4, 0.25 µg/test), APC anti-mouse CD 273 (programmed cell death protein-1, PD-1, 1 µg/test), PE- Cyanine7 anti-mouse CD366 (T-cell immunoglobulin and mucin-domain containing-3, Tim-3, 0.25 µg/test), at 4◦C in dark for 30 min. All the above

antibodies were purchased from eBioscience, Thermo Fisher Scientific (Grand Isle, NY, USA). After two washes with pre-cold PBS, T cell subsets and the immune checkpoint expressions in TIL were enumerated with a FACS Canto II cytometer, and the data was analyzed by Diva software (version 6.1.3).

# Immunohistochemistry (IHC) for ki67

Immunohistochemistry for ki67 was performed with the formalin-fixed, paraffin-embedded tumor segment. Firstly, the slides were deparaffinized. Antigen retrieval was carried out by incubation in saline-sodium citrate buffer (pH 6.0) via autoclaving. After being washed with PBS, the sections were subjected to endogenous peroxidase blocking with 3% H2O<sup>2</sup> in dark for 25 min. Then they were blocked with 3% bovine serum albumin (BSA), incubated with primary antibody against mouse ki67 (1:300, Servicebio, Wuhan, China) at 4◦C overnight, following by the incubation with horse reddish peroxidase (HRP)-conjugated secondary antibodies at room temperature for 50 min. Lastly, the sections were stained with 3,3Ndiaminobenzidine tertrahydrochloride substrate (DAB) and counterstained with hematoxylin. The mean density of positive area was calculated as ratio of integrated optical density (IOD) to the total pixel of each picture (IOD/10<sup>6</sup> pixel), which was analyzed by Image Pro Plus 6.0 software (Media Cybernetics, Silver Spring, USA).

#### Total RNA Extraction and Quantitative Real-Time PCR

One of the tumor segments was kept in sample protector for RNA/DNA (Takara Bio, Inc., Shiga, Japan) for quantitative realtime PCR (q-PCR). Total RNAs from tumors were extracted with TRIzol according to the manufacturer's instructions (Invitrogen, Thermo Fisher Scientific, Grand Isle, NY, USA). Two micrograms of total RNA was reverse transcribed using the ReverAid First Strand cDNA Synthesis Kit (Thermo Scientific, Inc., MA USA) following the supplier's protocol. The reactions were incubated at 25◦C for 5 min, then at 42◦C for 60 min, then at 70◦C for 5 min, and the products were stored at −20◦C before used. The PCR primer sequences are listed in **Table 1**. q-PCR reactions were conducted with SYBR <sup>R</sup> Premix Ex TaqTM II (Takara Bio, Inc., Shiga, Japan), in an StepOnePlus Real-Time PCR system (Thermo Fisher Scientific, Grand Isle, NY, USA). The program was as follows: a precycling stage at 95◦C for 30 s, then 40 cycles of annealing at 95◦C for 5 s, 60◦C for 34 s. Fluorescence was measured at the end of each annealing step, and the melting curves were monitored to confirm the specificity of the PCR products. mRNA expression levels of target genes relative to control gene Gapdh was determined with the 2−<sup>11</sup> Ct method.

# Metabolomics of Tumor Tissue

For metabolomics analysis, tumor segment was frozen in liquid nitrogen immediately after harvested and stored at −80◦C until detection. The untargeted metabolomics profiling was performed TABLE 1 | Primers for quantitative real-time PCR.


by gas chromatograph–time-of-flight mass spectrum (GC-TOF/MS) on XploreMET platform (Metabo-Profile, Shanghai, China). The sample preparation procedure and GC-TOF/MS analysis were conducted as described previously with mild modification (Wang et al., 2013). More details are provided in the **Table S1**. The data result sets containing all the m/z value, retention time and ion peak area of each sample were exported to the multivariate statistical software SIMCA (version 14.0, Umetrics, Umea, Sweden) for the subsequent partial least squares discrimination analysis (PLS-DA). The differential metabolites between groups were obtained using a multi-criteria assessment in the orthogonal partial least square discrimination analysis (OPLS-DA) model, which combines the strength of both contribution (variable importance in projection, VIP) and variable reliability (correlation coefficients, Corr.). And then they were confirmed following by the univariate statistical analysis (student T-test) using the SPSS 22.0 software (IBM, USA). Finally, the differential metabolites that were responsible for the separation between two groups were selected and identified.

#### 16S rRNA Gene Sequence Analysis of Gut Microbiota in Cecum

Caecum content was frozen in liquid nitrogen immediately after harvested and stored at −80◦C until sequencing. Total DNA from the caecum content was extracted with Fast DNA SPIN extraction kits (MP Biomedicals, Santa Ana, CA, USA) following the manufacturer's recommendations. The bacterial 16S rRNA gene V3-4 region was amplified by PCR using the sense primer (5′ -ACTCCTACGGGAGGCAGCA-3′ ) and the anti-sense primer (5′ -GGACTACHVGGGTWTCTAAT-3′ ). PCR amplicons were purified with Agencourt AMPure Beads (Beckman Coulter, Indianapolis, IN) and quantified individually using the PicoGreen dsDNA Assay Kit (Invitrogen, Carlsbad, CA, USA). Next, amplicons were pooled in equal amounts, and pairend 2 × 300 bp sequencing was performed using the Illlumina MiSeq platform with MiSeq Reagent Kit v3 (Zhao et al., 2013). High quality sequencing data and predicted function data were obtained by QIIME (Version 1.8.0), MOTHUR (version 1.31.2), and PICRUst (http://picrust.github.io/picrust/). The following statistics was performed by R software. α-diversity was evaluated by Ace, chao, simpson, and shannon index. β-diversity was assessed with the uniFrac distance-based principal coordinates analysis (PCoA) and analysis of similarities (ANOSIM). Taxonbased analysis and linear discriminant analysis with effect size (LEfSe) were applied to identify specific taxa microbes among groups using the default parameters (Segata et al., 2011). The predicted genes and their functions were aligned to Kyoto Encyclopedia of Genes and Genomes (KEGG) database, and differences among groups were compared through software STAMP4. Relationship between the "metabolite-bacterial genus" were performed by Spearman correlation analysis based on the abundance datasets.

#### Statistics

Statistical analysis was performed with SPSS 22 (IBM Corp., NY, USA). Datasets from each experiment was subjected to normal distribution test firstly. If they followed normal distribution, oneway analysis of variance (ANOVA) was conducted following by different parametric test depending on test for homogeneity of variance, otherwise the data was compared by Kruskal–Wallis H-Test. In ANOVA, post–hoc LSD test was applied for difference analysis under homogeneity of variance, if not, a Dunnett's test would be applied. For tumor volume change comparison, repeat measurement ANOVA was performed with general linear model. For bioinformatics analysis, p-values were corrected (FDR < 0.05) to control multiple hypothesis testing errors.

# RESULTS

# SGP Promoted the Anti-cancer Activity of PTX

In the preliminary experiment in vitro, SGP showed no improvement on the cytotoxicity of PTX, in which SGP was up to 200µg/mL and PTX was 243.2 ng/mL (IC50), and the treatment time was 48 and 72 h (**Table S2**). Therefore, the enhancement possibility of SGP on PTX was studied in the murine 4T1 breast cancer model. Compared with PTX group, SHP, and SLP groups displayed improved tumor control in the murine breast cancer model (**Figures 1A,B**). Tumors of Model group kept growing throughout the 21-day observation, resulting in a weight of 0.907 ± 0.229 g. PTX alone inhibited the tumor growth since 15th day, resulting in a decreased tumor weight (0.663 ± 0.146 g, p < 0.05). By contrast, PTX plus SGP showed an earlier suppression on tumor growth, and curves of tumor volume change of SLP (200 mg/kg) and SHP (400 mg/kg) groups were significantly different from that of PTX group (both p < 0.01). Finally, tumors from SLP and SHP groups were 0.521 ± 0.127 g (p < 0.05) and 0.451 ± 0.200 g (p < 0.05), evidently smaller than that of PTX group. With the data of our previous data, the combination index (CI) of SGP (400 and 200 mg/kg) and PTX (12.5 mg/kg) were 1.29 and 1.12 (>0.85), indicating that there would be an additive effect or synergy in the combination of SGP and PTX against breast cancer (**Table S2**). Moreover, expression of ki67, the symbol protein for proliferation, was even more down-regulated in SHP group than that in PTX group (**Figures 1C,D**, p < 0.05).

#### SGP Supplement Boosted Tumor Immune Surveillance by Suppressing Immune Checkpoints

To explore the underlying mechanism of the anti-cancer promotion of SGP, we firstly took a look of its effect on TILs. The flowcytometry analysis scheme was presented in **Figure 2A**. Total T cells within TME were not apparently affected by PTX or the combination treatment, although percentage of Tc cells showed a decline in SHP group when compared with PTX group (**Figure 2B**). However, immune checkpoints in the TILs were obviously influenced. For the total T cells (**Figure 2B**, first row), it was in SHP group and SLP group that the proportions of Tim-3-positve ones were reduced, and that of PD-1-positive was tend to be decreased; while in PTX group, proportions of T cell that expressed CTLA-4 (p < 0.05) and Tim-3 (p = 0.069) were increased. For Tc cells (**Figure 2B**, second row), SHP and SLP groups displayed significant reductions in cells that were PD-1- and Tim-3-possitive when compared with those from PTX group. However, PD-1 and Tim-3 expressions in the helper T cells (Th) seemed to be not affected by the combination treatment groups, while the CTLA-4 expression was up-regulated (**Figure 2B**, third row).

# PTX Plus SGP Inhibited Tumor Metabolism Within Tumor Tissue

Firstly, it was found that the combination treatment exhibited significant inhibition on mRNA of several Warburg effectrelated proteins (**Figure 3A**). In contrast to the mild suppressing tendency in PTX group, evident down-regulation was found in SHP and SLP groups, including those on mRNA levels of glucose transporter 3 (Glut3), lactate dehydrogenase A (Ldha), and pyruvate dehydrogenase kinase (Pdk). However, neither treatment of SHP or SLP promoted the down-regulations of hypoxia-inducible factor 1-α (Hif1a) caused by PTX.

Tumor metabolism profile was analyzed by GC-TOF/MSbased metabolomics to explore the improvement potential of PTX combined with SGP. By PLS-DA, overall metabolite profile of SHP group displayed evident difference from that of Model group and PTX group (**Figure 3B**). Then OPLS-DA was applied to made a further confirmation, in which "Q<sup>2</sup> (cum) > 0" indicated a significant difference (**Table 2**). Results showed that metabolite profile of SHP group could be apparently discriminated from that of Model group [SHP vs. Model, Q<sup>2</sup> (cum) = 0.557] and that of PTX group [SHP vs. PTX, Q2 (cum) = 0.305], while the difference between model group and PTX group was not obvious [Q<sup>2</sup> (cum) = −0.272, **Table 1**]. Based on the relative peak area, differential metabolites between SHP and PTX group were obtained using a multi-criteria assessment in the OPLS-DA model (**Table S3** "OPLSA" sheets, VIP ≥ 1), and then were confirmed by the univariate statistical analysis (**Table S3** "T-test" sheets, p < 0.05). Among the most affected metabolite, typical intermediates within TME (glucose 6-phosphate and fructose 6-phosphate), malignant proliferation indicators [allantoin (Ahn et al., 2014) and dehydroascorbic acid (Spielholz et al., 1997)] and other 5 metabolites (6 phosphonoglucono-D-lactone, L-alpha-aminobutyric acid, L-sorbose, gluconolactone, 3-hydroxypyridine) were significantly reduced in tumors from SHP group when compared with those from PTX group; while hydroxylamine and decanoylcarnitine were enriched in SHP group (**Figure 3C**), suggesting that alternation of metabolic characteristic would be contribute to the tumor-control improvement by combination of PTX and SGP.

# SGP Supplement Restored the Gut Dysbiosis Induced by PTX

#### Overall Structural Modulation After Treatment

Increasing evidence suggests that gut microbiota stands the very core of therapeutic responses for tumors occurring outside of the intestinal tract (Gopalakrishnan et al., 2018; Routy et al., 2018b). Hence, we investigated its possible involvement in the efficacy of SGP in this part. Gut microbiota in caecum content from the Normal group, Model group, PTX group, and SHP group were compared by 16sRNA sequencing since the above data indicated that SHP group exhibited better tumor control, more evident suppression on TIL immunocheckpoints and tumor metabolism. Common OTU analysis presented by Venn diagram indicated that there existed 481 unique OTUs in Normal group, 428 in Model group, 347 in PTX group, and 3,398 in SPH group, respectively, while 2,286 common OTUs were shared by all samples (**Figure 4A**).

α-diversity analysis was applied to evaluated richness and diversity of microbiota community. Briefly, community richness could be indicated by Chao1 and ACE indices, while community diversity and uniformity was represented by Shannon, and Simpson indices. Compared with that of Normal group, obvious reduction in richness was demonstrated by lowered Chao1 and

ACE indices in the gut microbiome of Model group, PTX group, and SPH group (p < 0.05, **Figure 4B**). In particular, microbiome from SPH group displayed an ever lower richness than that of Model group.

Overall community structure was compared by β-analysis using PCoA. Both unweighted and weighted PCoA assessment showed that, despite the decrease of richness, microbiome structure of Model group shared similar pattern with that of Normal group, while neither samples of PTX group or SHP group displayed alike community structure when compared to Normal group (**Figure 4C**). Additionally, ANOSIM made a further confirmation on the structure change induced by PTX plus SGP (**Table 3**). In brief, unweighted and weighted R statistic of Normal-Model comparison were 0.2229 (p = 0.009) and −0.0233 (p = 0.553), suggesting an obvious structure similarity between them. By contrast, R statistics for the comparison of SHP group with any other groups were much higher with extremely low pvalue, indicating a significant inter-group difference among them.

#### Community Membership Shifts After Treatment

Taxonomy analysis revealed marked differences at both phylum and genus levels among Normal, Model, PTX and SHP groups. Overall, a total of 8 phyla were shared by samples from all groups (**Figure 5A**). Of them, Firmicutes, Bacteroidetes, and Proteobacteria compromised over 90% of the total classified sequences. Relative abundances (>0.1%) of Bacteroidetes, Proteobacteria, and Tenericutes displayed significant differences in the four groups. It is noteworthy that Bacteroidetes was



evidently increased in SHP group, while that of Firmicutes was not apparently affected, resulting in an increased ratio of Bacteroidetes to Firmicutes (**Figures 5B,C**). At genus level, a total of 70 genera were identified from all samples (**Table S4**). LEfSe analysis indicated that there existed several specific genera in each group (**Figures 6A,B**). In brief, relative abundances of Dorea and Ruminococcus were highest in Normal group. Coprococcus, Parabacteroides, and Prevotella were enriched in Model group.

metabolites comparison. Values were represented the means ± SD (*n* = 9). \**p* < 0.05 and \*\**p* < 0.01.

Desulfovibrio, Ochrobactrum, Odoribacter, and Turicibacter were specific in PTX group. Bacteroides, Ruminococcus, and other 5 genus were significantly enriched in SHP group. Moreover, based on the datasets of metabolites (**Table S3**) and bacterial genus (**Table S4**), spearman correlation analysis showed that content of fructose-6-phosphate within tumor was negatively related to the enrichment of Ruminococcus (**Figure 6C**, p < 0.05).


TABLE 3 | Result of the Analysis of similarities (ANOSIM).

# Microbiome Function Regulation After Treatment

Via comparing the sequencing data with those collected in KEGG pathway database by PICRUSt, gene profile responsible for function pathways demonstrated significant differences among the four groups, including 1 for cellular processes, 2 for environmental information processing, 3 for genetic information processing, and 7 for metabolism pathways (**Figure 7**, **Table S5**). Particularly, microbiota from the tumor-bearing mice of PTX group showed significant up-regulation on pathways of cell motility and signal transduction; while that from SHP group displayed evident suppression on these two pathways, and genes responsible for the metabolism of terpenoids and polyketides were enriched.

# DISCUSSION

PTX is a broad-spectrum natural anti-cancer reagent. Nevertheless, drug resistance and adverse effects call for elevation on its efficacy and safety (Gupta et al., 2014; Starobova and Vetter, 2017). It was previously demonstrated that SGP alone is able to potentiate the Tc-based tumor immune surveillance with a benefit reshaping on gut microbiota (Su et al., 2018). As has been found that enhanced responses to cancer therapy require an intact and optimal commensal microbiota (Viaud et al., 2013; Sivan et al., 2015), we took a further insight into the promotion on the anti-cancer activity of PTX by SGP.

In the murine breast cancer model, combination of PTX and SGP displayed an improved tumor control, with an earlier suppression of tumor growth and evident inhibition on ki67 expression than PTX alone. Under tumor development, TILs face a hostile environment that induces exhaustion to impair their antitumor activity (Boldajipour et al., 2016; Beckermann et al., 2017), which is characterized by the up-regulation of immune checkpoints, such as CTLA-4, PD-1, and Tim-3 (Ahmadzadeh et al., 2009). Results showed that combination of PTX and SGP restored the exhausted antitumorigenic

immune cells (especially Tc) via inhibiting the expressions of immune checkpoints (PD-1 and Tim3), while PTX alone even evidently increased that of CTLA-4, suggesting a recovered tumor immune surveillance. As previous evidence has proved that SGP is incapable of direct lymphocyte stimulation or cytotoxicity (Su et al., 2018), the present data suggested that PTX supplementing with SGP would bring about a beneficial alternation within TME, thus improving the cell sensitivity to PTX, meanwhile enhancing the antitumorigenic immune.

Metabolism features within TME not only fuels the rapid proliferation of tumor cells, but also pose suppression on the antitumor T cells. For tumor cells, they employ spontaneous aerobic glycolysis (the Warburg effect) to intake large amount of glucose and other nutrients (lipids, amino acids), not only for energy production (adenosine triphosphate, ATP), but also more importantly, for the molecular building blocks required for cell proliferation (Ward and Thompson, 2012). Critical metabolic transporters and enzymes, like GLUTs, LDHA, and PDK, are responsible for the Warburg effect (Warburg, 1956; Ward and Thompson, 2012). Not surprisingly, oncogenic transcription factor, such as p53, HIF-1α, c-Myc, involved in malignant transformation is directly linked to the altered tumor metabolism. For example, HIF-1α induced by hypoxia is able to enhance glucose transport by increasing expression of glucose transporters 1–3 along with the transcription of pyruvate dehydrogenase kinase (Guillaumond et al., 2013). Increased c-Myc elicits numerous metabolic effects through reprogrammed gene expression, including GLUTs, LDHA (Boroughs and DeBerardinis, 2015). Nevertheless, p53 blocks

FIGURE 7 | Metabolic pathway enrichment analysis. The predicted genes and their functions were aligned to KEGG database, and the relative expressions for each pathway were compared. (A) The most affected pathway for cellular processes (cell motility), environmental information processing (membrane transport, signal transduction), and genetic information processing (folding sorting and degradation, replication and repair, and translation). (B) The most affected pathway for metabolism. Values were represented the means ± SD. \**p* < 0.05 and \*\**p* < 0.01.

excessive entry of glucose through glycolytic flux by inhibiting expression of GLUT1, GLUT3, GLUT4, phosphoglycerate mutase 1 (PGM 1) (Kruiswijk et al., 2015). As showed in our data, combination of PTX and SGP made evident down-regulations on mRNAs of Glut3, Ldha, and Pdk, while PTX alone did not affect them, indicating a possible suppression on glucose uptake. Meanwhile, obvious changes on the tumor metabolic profiling was found under the combination treatment, of which the accumulation of typical intermediates within TME was inhibited, such as glucose 6-phosphate, fructose 6-phosphate, and 6-phosphonoglucono-D-lactone, and several malignant proliferation indicators, such as allantoin (Hammad et al., 2011; Ahn et al., 2014) and dehydroascorbic acid (Spielholz et al., 1997), were also decreased. On the other hand, lactic acid in tumors of none of the treated groups (PTX, SLP, or SHP group) showed difference from that of Model group. We speculate that the effect of combination on tumor metabolism was so slow that downstream of it (such as lactic acid) had not yet been exhausted during the experiment, so the relative concentration of it was not different among the four groups. The underlying details will be explored in the further study of SGP. All the above data suggested that PTX plus SGP would inhibit the tumor metabolism within TME and induce dramatic metabolic profile alternation, so as to improve the cell sensitivity to PTX in addition to the microtubule polymerization by PTX (Peng et al., 2014).

Besides the alternation of tumor metabolism profile, combination of PTX and SGP demonstrated a restoration on the gut dysbiosis induced by PTX, which not only remodeld the microbiota community, but also regulated the microbiome function. α-diversity data showed that Chao1 and ACE indices of SHP group was lower than those of Model group and PTX group, indicating that the combination treatment decreased community richness. β-diverstiy analysis displayed significant inter-group difference between SHP group and any other group, suggesting that the combination posed apparent structure shift on the overall microbiome community. As α-diversity and β-diverstiy analysis indicated that combination of PTX and SGP endowed the tumor-bearing mice with unique gut microbiota characteristics, taxonomy analysis was employed to figure out the responsible species. Certain bacterium has been proved to be benefit to the outcomes of chemotherapies or immunotherapies (Iida et al., 2013; Viaud et al., 2013; Gopalakrishnan et al., 2018; Routy et al., 2018b). For instance, the anticancer efficacy of CTLA-4 mono-antibody relies on the existence of Bacteroides fragilis, which is associated with the maturation and function-gaining of dentritic cells, as well as the induction of an MHC class II-restricted Th1 cell memory response (Vétizou et al., 2015; Routy et al., 2018a). The abundance of Ruminococcus is negative related to the incidence of colorectal cancer (Borges-Canha et al., 2015; Wang et al., 2017). Helicobacter and Rikenella were positively correlated to enhanced immune response (Huang et al., 2017). As showed by the data, it was the combination treatment that, at phylum level, significantly raised the ratio of Bacteroidetes to Firmicutes by enriching the Gram-negative Bacteroidetes but not affecting the Gram-positive Firmicutes. At genus level analyzed by LEfSe, it was found that Bacteroides, Ruminococcus, and other 5 genus were specifically enriched in the tumor-bearing mice under the combination treatment, while the two cancer-risk genera, Desulfovibrio and Odoribacter, were suppressed. And it is quite interesting that abundance of Ruminococcus was significantly negative-associated with the amount of frucotose-6-phosphate within the tumor, an important intermediate of tumor metabolism. Additionally, several microbiome function pathways displayed distinct regulations upon the combination treatment. It is noteworthy that complex polysaccharide is one of the major driving forces in shaping the function profile of the gut microbiota (El Kaoutari et al., 2013; Tamura et al., 2017). As has been proved in our previous work, SGP is mainly made up of 50% polysaccharide, and it displayed a regulation on the microbiome function (Su et al., 2018). So it is probable that by oral supplement, SGP was able to directly reshape the gut microbiome homeostasis through regulating its metabolism mode, and serves as selective promoters to certain species. Hence, the combination treatment reduced the community richness (with reduced Chao1 and ACE indexes) but not affected diversity, with an apparent community membership shift. All these data suggested that PTX supplemented with SGP was capable of restoring the gut dysbiosis induced by PTX monotherapy, which also contributed to the suppression of tumor metabolism, resulting in an improved tumor control.

Colletively, data from the present study revealed an adjuvant candidate of SGP with PTX. The combination of PTX and SGP was not only capable of suppressing the tumor metabolism within TME, but also made a benefit reshaping on the gut microbiota. Moreover, the combination also recovered the impaired TILs via down-regulating the co-inhibitory signaling (PD-1 and/or CTLA-4). Hence in all probability, the combination of PTX and SGP against breast cancer relies on the regulation of tumor metabolism and gut microbiota, highlighting important clinical implications for the application of SGP.

# AUTHOR CONTRIBUTIONS

JS, YX, and XZ conceived and designed the experiments. JS, DLi, ML, LS, TL, DLiang, GL, OS, and CJ performed the experiments. JS, DLi, QC, and QW analyzed the data. JS, DLi, YX, and XZ drafted and revised the manuscript.

# FUNDING

This work was financially supported by Natural Science Foundation of Guangdong province (2018A0303130102), National Natural Science Foundation of China (81701086), the Guangdong Science and Technology Plan Projects (201707020022; 201604020009; 2016A050502032; 2015A020211021; 201504281708257), High-level Leading Talent Introduction Program of GDAS (2016GDASRC-0102), GDAS' Special Project of Science and Technology Development (2017GDASCX-0102; 2018GDASCX-0102, 2017GDASCX-0820), Science and Technology Planning Project of Guangzhou (201707020022), and the Nanyue Microbial Talents Cultivation Fund of Guangdong Institute of Microbiology (GDIMYET20140203; GDIMYET20150203).

# ACKNOWLEDGMENTS

We thank Meta-profile Co., Ltd. (Shanghai, China) and Personal Biotechnology Co., Ltd. (Shanghai, China) for assistance with metabolomics analysis and sequencing, respectively.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.03099/full#supplementary-material

# REFERENCES


therapy by modulating the tumor microenvironment. Science 342, 967–970. doi: 10.1126/science.1240527


cereal polysaccharides. Cell Rep. 21, 417–430. doi: 10.1016/j.celrep.2017. 09.049


**Conflict of Interest Statement:** DLi, ML, DLiang, GL, OS, CJ, and YX were employed by Guangdong Yuewei Edible Fungi Technology Co. Ltd.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Su, Li, Chen, Li, Su, Luo, Liang, Lai, Shuai, Jiao, Wu, Xie and Zhou. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Corrigendum: Anti-breast Cancer Enhancement of a Polysaccharide From Spore of Ganoderma lucidum With Paclitaxel: Suppression on Tumor Metabolism With Gut Microbiota Reshaping

#### Edited by:

*Helieh S. Oz, University of Kentucky, United States*

#### Reviewed by:

*Gang Liu, Institute of Subtropical Agriculture (CAS), China*

#### \*Correspondence:

*Yizhen Xie xyzgdim@sina.com Xinxin Zhou xinxin\_zhou@163.com*

*†These authors have contributed equally to this work*

#### Specialty section:

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

> Received: *02 April 2019* Accepted: *16 May 2019* Published: *31 May 2019*

#### Citation:

*Su J, Li D, Chen Q, Li M, Su L, Luo T, Liang D, Lai G, Shuai O, Jiao C, Wu Q, Xie Y and Zhou X (2019) Corrigendum: Anti-breast Cancer Enhancement of a Polysaccharide From Spore of Ganoderma lucidum With Paclitaxel: Suppression on Tumor Metabolism With Gut Microbiota Reshaping. Front. Microbiol. 10:1224. doi: 10.3389/fmicb.2019.01224* Jiyan Su1†, Dan Li 2,3†, Qianjun Chen<sup>4</sup> , Muxia Li 2,3, Lu Su<sup>5</sup> , Ting Luo<sup>6</sup> , Danling Liang2,3 , Guoxiao Lai 3,7, Ou Shuai <sup>3</sup> , Chunwei Jiao<sup>3</sup> , Qingping Wu<sup>1</sup> , Yizhen Xie1,3 \* and Xinxin Zhou<sup>2</sup> \*

*<sup>1</sup> State Key Laboratory of Applied Microbiology Southern China, Guangdong Provincial Key Laboratory of Microbial Culture Collection and Application, Guangdong Institute of Microbiology, Guangzhou, China, <sup>2</sup> School of Pharmaceutical Science, Guangzhou University of Chinese Medicine, Guangzhou, China, <sup>3</sup> Guangdong Yuewei Edible Fungi Technology Co. Ltd., Guangzhou, China, <sup>4</sup> Department of Breast Disease, Guangdong Provincial Hospital of Chinese Medicine, Guangzhou University of Chinese Medicine, Guangzhou, China, <sup>5</sup> School of Pharmacy and Chemistry, Dali University, Dali, China, <sup>6</sup> Guangdong Laboratory Animals Monitoring Institute, Guangzhou, China, <sup>7</sup> School of Pharmacy, Guangxi University of Chinese Medicine, Xining, China*

Keywords: spore of Ganoderma lucidum, paclitaxel, tumor metabolism, immune checkpoints, gut microbiota

#### **A Corrigendum on**

**Anti-breast Cancer Enhancement of a Polysaccharide From Spore of Ganoderma lucidum With Paclitaxel: Suppression on Tumor Metabolism With Gut Microbiota Reshaping**

by Su, J., Li, D., Chen, Q., Li, M., Su, L., Luo, T., et al. (2018). Front. Microbiol. 9:3099. doi: 10.3389/fmicb.2018.03099

In the original article, there was a mistake in **Figure 2** as published . In Figure 2B, the second and third histograms in the last row were the same. The third histogram should be the result of the "CTLA-4<sup>+</sup> Th cell," rather than that of the "PD-1<sup>+</sup> Th cell." The corrected **Figure 2** appears below.

The authors apologize for this error and state that this does not change the scientific conclusions of the article in any way. The original article has been updated.

Copyright © 2019 Su, Li, Chen, Li, Su, Luo, Liang, Lai, Shuai, Jiao, Wu, Xie and Zhou. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Combination of Clostridium butyricum and Corn Bran Optimized Intestinal Microbial Fermentation Using a Weaned Pig Model

Jie Zhang1,2, Jian Sun1,2, Xiyue Chen<sup>1</sup> , Cunxi Nie1,3, Jinbiao Zhao<sup>1</sup> , Wenyi Guan<sup>2</sup> , Lihui Lei<sup>2</sup> , Ting He<sup>1</sup> , Yiqiang Chen<sup>1</sup> , Lee J. Johnston<sup>4</sup> , Jinshan Zhao<sup>5</sup> and Xi Ma1,5,6,7 \*

<sup>1</sup> State Key Laboratory of Animal Nutrition, College of Animal Science and Technology, China Agricultural University, Beijing, China, <sup>2</sup> Department of Animal Husbandry and Veterinary, Beijing Vocational College of Agriculture, Beijing, China, <sup>3</sup> College of Animal Science and Technology, Shihezi University, Xinjiang, China, <sup>4</sup> West Central Research and Outreach Center, University of Minnesota, Morris, MN, United States, <sup>5</sup> College of Animal Science and Technology, Qingdao Agricultural University, Shandong, China, <sup>6</sup> Department of Internal Medicine, University of Texas Southwestern Medical Center, Dallas, TX, United States, <sup>7</sup> Department of Biochemistry, University of Texas Southwestern Medical Center, Dallas, TX, United States

#### Edited by:

Liwei Xie, Guangdong Institute of Microbiology (CAS), China

#### Reviewed by:

Wei Wang, Jilin University, China Xuran Zhuang, ShanghaiTech University, China Chuan Wang, Auburn University, United States

#### \*Correspondence:

Xi Ma maxi@cau.edu.cn orcid.org/0000-0003-4562-9331

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 08 September 2018 Accepted: 29 November 2018 Published: 18 December 2018

#### Citation:

Zhang J, Sun J, Chen X, Nie C, Zhao J, Guan W, Lei L, He P, Chen Y, Johnston LJ, Zhao J and Ma X (2018) Combination of Clostridium butyricum and Corn Bran Optimized Intestinal Microbial Fermentation Using a Weaned Pig Model. Front. Microbiol. 9:3091. doi: 10.3389/fmicb.2018.03091 Experimental manipulation of the intestinal microbiota influences health of the host and is a common application for synbiotics. Here Clostridium butyricum (C. butyricum, C.B) combined with corn bran (C.B + Bran) was taken as the synbiotics application in a waned pig model to investigate its regulation of intestinal health over 28 days postweaning. Growth performance, fecal short chain fatty acids (SCFAs) and bacterial community were evaluated at day 14 and day 28 of the trial. Although the C.B + Bran treatment has no significant effects on growth performance (P > 0.05), it optimized the composition of intestinal bacteria, mainly represented by increased acetate-producing bacteria and decreased pathogens. Microbial fermentation in the intestine showed a shift from low acetate and isovalerate production on day 14 to enhanced acetate production on day 28 in the C.B + Bran treatment. Thus, C.B and corn bran promoted intestinal microbial fermentation and optimized the microbial community for pigs at an early age. These findings provide perspectives on the advantages of synbiotics as a new approach for effective utilization of corn barn.

Keywords: synbiotics, Clostridium butyricum, corn bran, intestinal bacteria, short chain fatty acids, weaned pig model

# INTRODUCTION

Great attention has been paid to the important influences of dynamic microbial communities on human health. Several animal studies have proven that experimental manipulations of the intestinal microbiota can modify many aspects of the host's health. Commonly, probiotics have been applied to manipulate the intestinal microbiota. Clostridium butyricum (C. butyricum, C.B) is an anaerobic, gram-positive bacillus found in the intestine of healthy animals and is commonly considered as a kind of probiotics. C.B plays an important role in optimizing the intestinal microbial community, especially in the colon, and maintains the harmonious intestinal microecology by inhibiting proliferation of harmful bacteria (Howarth and Wang, 2013). C.B prefers dietary fiber, which is not digested directly by enzymes of monogastric animals, as its fermentable substrate

(Sawicki et al., 2017). Orally administered C.B spores germinate and grow in intestinal tracts and produce a mass of short chain fatty acids (SCFAs) including acetate, propionate and butyrate (Sun et al., 2016; Jia et al., 2017) by fermenting non-digestible polysaccharides (Chen et al., 2013). SCFAs are major anions in the colon that are absorbed rapidly and stimulate absorption of water and sodium. SCFAs can be oxidized to serve as fuels for colonic cells. Among them, butyrate is the main end products of C.B (Cassir et al., 2016) and the major nutrient for energy metabolism (Jackie et al., 2013; van der Beek et al., 2017) of colonic epithelial cells (Ma et al., 2012; Gonçalves and Martel, 2013; Simeoli et al., 2017). SCFAs can also decrease colonic pH, stimulate intestinal peristalsis, improve the intestinal microenvironment and regulate the micro-ecological balance of the colon (Ma et al., 2018). In addition, SCFAs have an important role to play in proliferation and differentiation of colonocytes and regulation of gene expression in colonic epithelial cells (Nicholson et al., 2012). Theoretically, diets containing C.B could beneficially impact growth performance, SCFAs formation, and stability of microbial community in the gut of weaned pigs.

The combination of probiotics and fibrous prebiotics is called synbiotics. Dietary fiber includes a soluble part (SDF) and an insoluble part (IDF), and the later mainly consists of cellulose, hemicellulose, and lignin (Brownlee, 2011). In the past, weaned animals were considered unable to ferment carbohydrates. Recent research results suggest a proper addition of dietary fiber can enhance intestinal health, modulate the microbial community and support innate immunity of intestinal mucosa in weanling piglets (Han et al., 2017). Previous studies have proven that dietary fiber exerts its function by forming SCFAs from fermentation of saccharolytic microbiota, especially cellulose-degrading ones (Chen T. et al., 2017). An addition of dietary fiber can serves as the fermentation substrate of hindgut microorganisms and improves intestinal health by modulating gut microbial composition and function (Jeffery and O'Toole, 2013; Desai et al., 2016; Martens, 2016; Brahma et al., 2017). The intestinal microbiota contains highly diverse communities and has multiple roles in metabolism and health of the host (Chen et al., 2015; Wang et al., 2016).

As the main by-product of corn processing, corn bran is used widely as an ingredient for animal feed. Corn barn has the highest content of dietary fiber among all cereal brans (Liu et al., 2017). However, the high content of plant polysaccharides in corn bran limits its nutritive value for pigs. Several processing technologies such as solid-state fermentation, have been applied to corn barn as an effort to improve nutritive value (Liu et al., 2017). Use of saccharolytic bacteria might be another approach to enhance the nutritive value of corn barn. The combination of C.B and corn bran might be used as an effective synbiotics. Synbiotics are a mixture of probiotics and prebiotics that can exert the biogenic activity of probiotics, but also selectively increase the number of bacteria, making the probiotics more effective and lasting (Duncan and Flint, 2013). Thus, this experiment was conducted to compare the influences of synbiotics with C.B and corn bran or a single addition of C.B on intestinal health using a weaned piglet model. A long-term objective of this research is to evaluate the utility of synbiotics in improving the nutritional value of low quality, fibrous feed stuff.

# MATERIALS AND METHODS

# Ethics Approval and Consent to Participate

All procedures of this experiment were approved by the animal protection and utilization organization committee of China Agricultural University (CAU20171015-3).

# Pigs, Diets, and Experimental Protocol

Newly weaned pigs (n = 48; Landrace × Large White) were picked from 24 litter piglets at 28 day age. Pigs (8.09 ± 0.25 kg) were allotted randomly to a basal diet with 1% C.B or the basal diet with 1% C.B and 5% corn bran (C.B + Bran). One pen as a replicate, four replicates per treatment and six pigs per replicate. The standard corn-soybean basal diet was formulated based on the standard ileal digestible amino acids to satisfy 11–20 kg pigs' requirement (NRC, 2012. See **Table 1**).

The C.B supplement (China Microorganism Preservation Center, Strain No. 1.336) was included at 1% and consisted of 1 × 10<sup>8</sup> CFU/g in spore state.

Each animal was weighed on days 14 and 28 of the trial and feed intake was recorded weekly for every pen. ADFI, ADG, and F/G were calculated. Fresh fecal samples from 8 pigs per treatment were collected and immediately frozen in liquid nitrogen on day 14 and day 28. Fecal samples were stored at −80◦C for bacterial DNA and bacterial metabolite analysis.

# Extraction of Fecal DNA

E.Z.N.A Stool DNA Kit (Omega Bio-Tek Inc., United States) was used following the manufacturer's protocols to detect total bacterial DNA in fecal samples. A nanodrop 2000 spectrophotometer (Thermo Fisher Scientific, United States) was used for DNA micro-quantification and 1% agarose gel electrophoresis was used for detection of DNA size fragments. Finally, quantified DNA was kept at −20◦C for DNA sequencing analysis.

#### Polymerase Chain Reaction (PCR) Amplification

Amplification of V3–V4 regions of the bacterial 16S rRNA gene was accomplished via TransStart Fastpfu <sup>R</sup> DNA Polymerase (Takara, Japan) and a PCR procedure. The upstream primer was 5<sup>0</sup> -barcode-ACTCCTACGGGAGGCAGCA-3<sup>0</sup> and the downstream primer was 50GGACTACHVGGGTWTCTAAT-3<sup>0</sup> . The reaction system of PCR (20 µL) include 5 × FastPfu buffer, 4 µL; 2.5 mM dNTPs, 2 µL; each primer (5 µM), both 1.6 µL; FastPfu polymerase, 0.4 µL and template DNA, 10 ng. The PCR procedure included 95◦C denaturation for 3 min; then 26 cycles with 95◦C for 30 s, 55◦C for 30 s, and 72◦C for 45 s and finally 72◦C for 10 min.



<sup>1</sup>ADF, acid detergent fiber; C.B, basal diet + C. butyricum; C.B + Bran, basal diet + C. butyricum + corn bran; CON, basal diet; CP, crude protein; DE, digestible energy; EFFSB, extruded full fat soybeans; NDF, neutral detergent fiber; SID, standard ileal digestibility; Lys, lysine; Met, methionine; Thr, threonine; Trp, tryptophan; Val, valine; Ca, calcium; P, phosphorus. <sup>2</sup>Supplied per kilogram of complete diet: vitamin A, 12,000 IU; vitamin D3, 2,500 IU; vitamin E, 30 IU; vitamin K3, 3 mg; vitamin B12, 0.012 mg; riboflavin (vitamin B2), 4 mg; niacin (vitamin B3), 40 mg; pantothenic-acid (vitamin B5), 15 mg; choline chloride, 400 mg; folacin, 0.7 mg; thiamine (vitamin B1), 1.5 mg; (vitamin B6), 3 mg; biotin, 0.1 mg; Zn, 100 mg as ZnO; Mn, 40 mg; Fe, 90 mg; Cu, 200 mg; I, 0.35 mg; Se, 0.3 mg.

#### Illumina MiSeq Sequencing

After purification with the AxyPrep DNA Purification kit (Axygen Biosciences, United States), PCR products were detected by Agarose gel (2%) electrophoresis and were quantified using PicoGreen dsDNA Quantitation Reagent (Invitrogen, United States) on QuantiFluor-ST Fluorometer (Promega, United States). After that, collected amplicons for paired-end sequencing (2 × 300 bp) according to standard protocols. This process was completed on the Illumina MiSeq platform (Allwegene, China). The raw data in this manuscript have been uploaded to the NCBI SRA Database under an accession no. SRP159591.

#### Bioinformatics Analysis of Sequencing Data

For raw fastq files analysis, the first step was to demultiplex and quality-filter data via QIIME (version 1.17). basic principles used in this process were: (i) Sequencing reads were trimmed at the sites with an average quality score <20 over a 50 bp sliding window and deleted trimmed reads less than 50 bp; (ii) The reads that contained mismatching barcode were deleted; and (iii) Removing the paired reads with less than 10 bp overlapping.

UPARSE (version 7.1<sup>1</sup> ) was used to gather OTUs with a 97% similarity. UCHIME was used to identify and delete chimeric sequences. RDP Classifier<sup>2</sup> based on Silva (SSU115) 16S rRNA database was used to complete the taxonomic analysis for each 16S rRNA gene sequence with a confidence threshold of 70%. Venn diagrams software of R tools generated Venn figures (**Figures 1A,B**), which represented visually of the similarity and overlap of the OTU samples. The alpha diversity indexes, including Chao index and Shannon index, were all calculated using qiime software (version v.1.8<sup>3</sup> ) of Mothur v.1.21.1 and produced **Figures 1C,D**. Vegan and ggplot2 package of R tools conducted the Non-metric multidimensional scaling (NMDS) analysis and produce **Figures 1E,F**. Based on the results of taxonomic analysis, using R tool to produce the diagram of species composition in different samples (**Figures 2A,B**). To clustering data for abundance similarity between species or samples, using vegdist and hclust of vegan package of R tools to do distance calculation and clustering analysis, which distance algorithm did by Bray-Curtis and clustering method did by complete-linkage. Diagram of results shown as **Figures 2C–F**, **3**.

#### Detection of SCFAs

About 0.5 g feces were put into a 10 mL polypropylene tube and diluted with 8 mL deionized water. Tubes containing samples rested in an ultrasonic bath for 30 min, and were centrifuged at 8,000 rpm for 10 min. The supernatant was drained into an empty tube and diluted 50 times and then filtered with a 0.22 µm filter. High performance ion chromatography of ICS-3000 (Dionex, United States) was used to analyze the components of 25 µL of extracted sample solution. Separate organic acids used AS11 analytical column (250 mm × 4 mm); separate the other gradient conditions used an AG11 guard column. Varying concentrations of potassium hydroxide was used for gradient contrast. Those concentrations were: 0.8–1.5 mM for 0–5 min; 1.5–2.5 mM for 5–10 min; 2.5 mM for 10–15 min. The flow rate is 1.0 mL/min.

#### Statistical Analysis

The data analysis and graphic analysis of growth performance and organic acid data were performed by unpaired t-test of SPSS 19.0

<sup>1</sup>http://drive5.com/uparse/

<sup>2</sup>http://rdp.cme.msu.edu/

<sup>3</sup>http://qiime.org/scripts/alpha\_rarefaction.html

with C. butyricum (C.B) group and basal diet with the combination of C. butyricum and corn bran (C.B + Bran) group at the 14th day (A) and 28th day (B) after weaning. Bacterial richness was estimated by the Chao1 value (C). Bacterial diversity was estimated by Shannon index (D). The diff-NMDS plot comparative analysis of sample in bacterial community between two groups were showed on day 14 (E) and 28 (F) after weaning.

and GraphPad prism 6.0. Results are shown as means ± SEMs. P-value <0.05 was considered a significant difference.

# RESULTS

# Effect of C.B and Corn Bran on Growth Performance

From the day 0 to 14 and day 14 to 28, growth performance indicated by average daily feed intake (ADFI), average daily gain (ADG) and the ratio of ADFI to ADG (F/G) between two treatments showed no significant difference (**Table 2**).

# Effects of C.B and Corn Bran on Intestinal Bacterial Richness, Diversity, and Similarity

To understand changes in intestinal bacteria, we performed 16S rRNA gene sequencing of fecal samples on day 14 and 28 after weaning. After quality control, size filtering, and chimera removal, 449,014 and 463,345 clean reads were obtained from feces collected on day 14 and day 28, respectively. The total

operational taxonomic units (OTU) numbers were classified at 97% similarity, with 626 OTUs and 669 OTUs detected in fecal samples on days 14 and 28, respectively. Fecal bacterial communities of the two groups shared about 86.42% on day 14 and 85.35% on day 28 (**Figures 1A,B**). Interestingly, the number of unique OTUs in the C.B + Bran treatment was well above that in C.B group on day 28.

The Chao1 index and Shannon index were detected to study the effect of C.B and corn bran inclusion on bacterial abundance and diversity. Between C.B and C.B + Bran groups, no significant differences were observed on both day 14 and 28 (**Figures 1C,D**). The β diversity of OTU community comparisons done by hierarchical clustering showed no differences between the two groups on day 14 (**Figure 1E**). But on day 28, intestinal microbiota of two treatments were clustered separately (**Figure 1F**), indicating a significant effect of corn bran in the later period of the experiment.

# Effects of C.B and Corn Bran on Community Structure of Fecal Bacteria

The most prevalent phyla were Firmicutes and Bacteroidetes in the present fecal samples, accounting for more than 95% of the total microbiota (**Figure 2A**). On day 14 after weaning, no significant differences were found in the dominant phyla among the two treatments. On day 28, the proportion of Firmicutes dramatically increased from 69.67% in the C.B group to 88.14% in the C.B + Bran group, while the proportion of Bacteroidetes sharply decreased from 25.72 to 7.83%.

At the order level, Firmicutes were mainly composed of Clostridiales, Lactobacillales, and Selenomonadales, while Bacteroidales was the dominant order of Bacteroidetes (**Figure 2B**). Erysipelotrichales of Firmicutes decreased significantly in the C.B + Bran group on day 14. On day 28, Clostridiales and Lactobacillales increased dramatically from 28.37 to 48.91% and 14.94 to 30.46%, respectively in the C.B + Bran group. However, Selenomonadales of Firmicutes dropped its proportion significantly from 24.53% in the C.B group to 6.49% in the C.B + Bran group on day 28. Bacteroidales as the predominant order of Bacteroidetes were markedly lower in the C.B+ Bran group on day 28.

At the family level, the only change on day 14 occurred in the proportion of Erysipelotrichaceae that declined from 2.27% in the C.B group to 0.61% in the C.B + Bran group (P < 0.05) (**Figure 2C**). On day 28, changes between the two groups were multiple and various (**Figure 2D**). In the order of Clostridiales, Ruminococcaceae and Lachnospiraceae increased by 9% (P < 0.05), while Veillonellaceae and Acidaminococcaceae decreased from 18.18 and 6.35% to 5.86 and 0.64%, respectively (P < 0.05) in the C.B + Bran group compared with the C.B group. Additionally, Prevotellaceae showed a similar significant decrease with its order Bacteroidales.

Genera in fecal samples on day 14 displayed slight changes with increased Lactobacillus and decreased Megasphaera in the C.B + Bran group without difference (**Figure 2E**). On day 28, there was no significant difference in the dominant genera including Lactobacillus and Streptococcus. Prevotellaceae\_NK3B31\_group, Prevotella\_9 and Prevotella\_1 of Prevotellaceae, as well as Ruminococcaceae\_UCG\_005 of Ruminococcaceae changed resembled to their change in family level (P < 0.05) (**Figure 2F**).

#### Effects of C.B and Corn Bran on Concentration of Fecal SCFAs

To evaluate the effect of combining C.B with corn bran on intestinal fermentation, the concentration of fecal SCFAs, including acetate, propionate, butyrate, isobutyrate, and isovalerate were measured (**Table 3**). On day 14, concentration of acetate and isovalerate were lower (P < 0.05) in the C.B + Bran group than the C.B group. On day 28, the concentration of acetate increased with the combined addition of C.B and corn bran compared with the single addition of C.B (P < 0.05).

#### Correlation Analysis Between the Varied Index (Growth Performance and Fecal SCFAs) and Corresponding Intestinal Flora

To further discover whether the effects of C.B with corn bran on the intestinal microbiota were associated with the fluctuating growth performance and fecal SCFAs, the correlation analysis between the differentially abundant intestinal bacteria at the order, family and genus level and ADG and acetate on day 28 was completed. The community abundance of the orders Clostridiales, Lactobacillales and Bacteroidales were correlated negatively with ADG on day 28 (**Figure 3**). Down to the family and genus level, Ruminococcaceae with its genus

TABLE 2 | Effect of dietary C.B and C.B+ corn bran inclusion on weaned pigs growth performance<sup>1</sup> .


<sup>1</sup>Values are means ± SEMs, n = 24/treatment. ADFI, average daily feed intake; ADG, average daily gain; C.B, basal diet + C. butyricum; C.B + Bran, basal diet + C. butyricum+ corn bran; F/G, the ratio of ADFI to ADG.

TABLE 3 | Effect of dietary C.B and C.B + corn bran inclusion on concentration of fecal SCFAs (mg/g feces)<sup>1</sup> .


<sup>1</sup>Values are means ± SEMs, n = 8/treatment. Different superscripts in same row mean a significant difference, P < 0.05. SCFAs, short chain fatty acids; C.B, basal diet + C. butyricum; C.B + Bran, basal diet + C. butyricum + corn bran. a,b Different superscript within a row means significantly different (P < 0.05).

(Ruminococcaceae\_UCG-005, Ruminococcaceae\_UCG-014) and Lachnospiraceae in the order Clostridiales as well as Lactobacillus of Lactobacillaceae in the order Lactobacillales were correlated negatively with ADG on day 28. However, Bacteroidales including Prevotellaceae\_NK3B31\_group and Phascolarctobacterium, as well as Burkholderiales and Fibrobacterales were correlated positively with ADG on day 28. For the increased fecal acetate, the genus Subdoligranulum of Ruminococcaceae in Clostridiales was correlated positively with it on day 28.

#### DISCUSSION

Previous researches on the effects of corn bran on body health varied according to many factors with some of them indicating that fiber-rich diets would enhance growth performance (Gerritsen et al., 2012) while others showed reduced or unchanged digestibility of nutrients and energy (Jaworski et al., 2017; Morowitz et al., 2017). Dietary factors including source, solubility, processing and dose (Williams et al., 2017) can affect intestinal fermentation. Considering the low utilization of corn bran, it is necessary to link corn barn with new treatments such as combining with probiotics that will improve nutritional value. Thus, this manuscript aimed at investigating the effect of adding C.B and corn bran for intestinal health via using a weaned pig model, which is an ideal alternative model for humans (Heinritz et al., 2013).

Previous studies shown that C.B addition alone or corn barn addition alone both have no significant effects on ADG and ADFI (Liu et al., 2018; Zhang et al., 2018). Here, the combination of C.B and corn bran keep the consistent effects on them, indicating the combination has no negative effects on pig growth. However, we noticed that separate addition of these two substances both reduced the specific microbial flora in pigs, especially for C.B addition on day 28 (Liu et al., 2018; Zhang et al., 2018). However, the combination application of them showed different effects. Herein, increase of microbial diversity motivated our interest to study their combination how to affect intestinal microbiota structure in present study.

Microbial changes caused by C.B and corn bran should be discussed separately by period. On day 14 after weaning, both within- and between-habitat diversity of fecal samples remained stable. As for specific alterations in the microbial community, reduced Erysipelotrichaceae in the C.B and Bran group, suggests a positive effect of corn barn and a reduced potential for erysipelas infection (Ding et al., 2015). The intestinal microbial structure of newly weaned pigs is immature, and not firmly established. So, weaning stress can easily disturb the dynamic balance of intestinal microbiota (Chen L. et al., 2017, 2018). In present study, lack of difference in the intestinal microbiota between treatments on day 14 may due to successful establishment of C.B in the early period after weaning (Zhang et al., 2018). Also, intestinal function is not mature enough to successful digest dietary fiber (Xu et al., 2014). Thus, if we want to investigate the additive effects of corn bran on intestinal microbiota, select of the appropriate period is essential.

Since diversity is considered as an indicator of healthy microbiota (Salonen et al., 2012), the increased diversity from day 14 to 28 suggests at least 28 days are required for the gut to adapt to weaning stresses (i.e., change in diet, social structure, and environment). Significant changes of microbial composition on day 28 indicated modulation of

corn bran mainly occurred in the later period after weaning when a relatively stable microbial community has been established (Zhao et al., 2018). Addition of corn bran with high concentration of IDF provided fermentation substrates for intestinal microbiota such as C.B and increased the amount of unique OTUs and microbial variance. The present increase of Firmicutes has been proven to ferment polysaccharides to SCFAs (D'hoe et al., 2018) and orders of Clostridiales and Lactobacillales also showed a great boost in cellulose degradation. Clostridiales in the intestinal mucosa is a pivotal mediator for fiber fermentation, butyrate production, and mucosal immunity (David et al., 2015; Bensoussan et al., 2016). In addition, some bacteria of Clostridium such as Ruminococcus flavefaciens, Ruminococcus bromii, and Faecalibacterium prausnitzii can use the cellulosome system to degrade cellulose (Zhang et al., 2015; Hu et al., 2016). Among the Clostridiales, Ruminococcaceae increases in prevalence in diets enriched in resistant starch, while Lachnospiraceae is improved in a diet rich in wheat bran (Louis et al., 2014). In the C.B + Bran group, the higher proportion of Ruminococcaceae and Lachnospiraceae suggested an elevated demand for fiber degradation. Additionally, Ruminococcaceae and Lachnospiraceae are associated with lean phenotypes (Menni et al., 2017), which is consistent without performance effects in C.B + Bran group. When considering bacterial function, several changes merit attention. Prevotellaceae is reported to be associated with several human diseases, such as asthmatic airway inflammation and arthritis (Scher et al., 2013; Clarke et al., 2014). So, for humans, Prevotellaceae is thought to be an opportunistic pathogen (Lukens et al., 2014; Pianta et al., 2017). Decreased Prevotellaceae in the C.B + Bran group showed the benefits of combining C.B with corn bran. In sum, the addition of corn bran optimized the intestinal microbiota with increasing fiber-degrading bacteria and decreasing pathogens.

The wave of microbial fermentation in the intestine caused by the combination of C.B and corn bran deserves attention. Intestinal production of SCFAs depends on composition of intestinal microbes, substrate source and chyme transit time (Wichmann et al., 2013). In this study, fecal samples were used for SCFAs analysis. Unlike chyme samples, feces mainly reflect the nutritional difference between production and consumption. In the present study, we found that fecal acetate content declined with the C.B + Bran treatment on day 14 but increased on day 28. Acetate is the most abundant SCFA, and its concentration in the lumen is influenced by dynamic balance of production, use, and mucosal uptake (Elamin et al., 2013; Louis et al., 2014). Food with low viscosity such as bran could alter the intestinal microenvironment with reduced activity of amylase in small intestine (Desai et al., 2016; Martens, 2016), which could explain the decreased concentration of SCFAs on day 14. However, on day 28 the increased anaerobic bacteria in the C.B + Bran treatment, such as Ruminococcaceae and Lachnospiraceae are known to produce acetate and suppress the growth of Bacteroidales which is the preferential producer for propionate (Flint et al., 2008). Butyrate serves as a major energy source for intestinal enterocytes and exerts health-promoting effects on the colon (Huang et al., 2015). Bacteria synthetizes butyrate through two primary pathways. One pathway is a conversion of acetate to butyrate via butyryl-CoA (Duncan et al., 2002; Besten et al., 2013; Louis et al., 2014). The second pathway is a direct synthesis via butyrate kinase. Lactobacillus, Megasphaera, Blautia, and Prevotella are considered to participate in the butyrate producing (Berni Canani et al., 2016; Zhang et al., 2018). Among that, Lactobacillus was thought contact with butyrate production via expands butyrate-producing bacterial strains, like Blautia, Roseburia, and Coprococcus (Berni Canani et al., 2016). But here, the fluctuation of proportion in Megasphaera, and Prevotella made it is difficult to contact them with butyrate production. In present study, we have not observed any significant change in butyric acid content on day 14 and day 28, despite content of fecal butyrate increased slightly on day 28 along with increasing of fecal acetate. These results were consistent with previous study (Liu et al., 2018; Zhang et al., 2018). It should explains two things. First, the relationship of SCFAs concentrations in digesta and in feces should not be positively associated (Fan et al., 2017). Moreover, these results reminded us that producing butyrate maybe not the main ways of C.B or corn barn on improving intestinal environment. The specifically mechanism need further been illuminated.

Given our results, the effects of combination of C.B and corn bran should lie in providing substrates for intestinal fermentation, increasing acetate to reduce colonic pH, and optimizing intestinal microbiota which suppressed harmful bacteria.

# CONCLUSION

Addition of corn bran to C.B changed the intestinal microbial community greatly with increasing fiber-degrading bacteria including Ruminococcaceae and Lachnospiraceae and decreasing pathogens such as Erysipelotrichaceae and Prevotellaceae. IDF in the corn bran provided fermentable substrates for colonic microbiota and enhanced intestinal fermentation with elevated acetate content in feces on day 28. Thus, the combination of C.B and corn bran enhanced the benefits of the single addition of C.B with optimized intestinal microbiota and fermentation in the later period after weaning. Additionally, it suggested a new application for the use of corn bran as with synbiotics.

# AUTHOR CONTRIBUTIONS

XM conceived and designed the research. JZ, JS, and JbZ conducted the research. JZ wrote the manuscript and analyzed the data. JS and XC wrote a part of manuscript and assisted in analysis of data. CN, WG, and LL contributed to sample analysis. TH, YC, JsZ, and XM critically reviewed the manuscript. LJ contributed to language review. All authors read and approved the final manuscript.

# FUNDING

fmicb-09-03091 December 14, 2018 Time: 14:38 # 9

This work was supported by the National Key R&D Program of China (2018YFD0500601 and 2017YFD0500501), the National Natural Science Foundation of China (31829004, 31722054, and 31472101), the College of Animal Science and Technology "Young Talents Program" at China Agricultural University (2017DKA001), the Beijing Nova Programme Interdisciplinary Cooperation Project (xxjc201804), the 111 Project (B16044), the Developmental Fund for Animal Science by Shenzhen Jinxinnong Feed Co., Ltd., and the Apply Basic Research Project in Xinjiang Production and Construction Crops of

#### REFERENCES


China (2016AG009) and the Beijing Municipal Education Commission Project (KM201612448003) and the Beijing Agricultural Vocational College Project (XY-YF-18-13).

#### ACKNOWLEDGMENTS

We thank Professor Wenqing Lu to give us some helpful suggestions, thank Dr. Xiangli Sun and Wenjun Yang in Ministry of Agriculture Feed Industry Center, China Agricultural University, for their excellent assistance and support in chemical analysis.

the colonic mucus barrier and enhances pathogen susceptibility. Cell 167, 1339–1353. doi: 10.1016/j.cell.2016.10.043



application of evidence mapping methodology. Nutrients 9:E125. doi: 10.3390/ nu9020125


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Zhang, Sun, Chen, Nie, Zhao, Guan, Lei, He, Chen, Johnston, Zhao and Ma. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Production of the Neurotoxin Salsolinol by a Gut-Associated Bacterium and Its Modulation by Alcohol

#### Daniel N. Villageliú1,2, David J. Borts<sup>3</sup> and Mark Lyte<sup>1</sup> \*

<sup>1</sup> Department of Veterinary Microbiology and Preventive Medicine, College of Veterinary Medicine, Iowa State University, Ames, IA, United States, <sup>2</sup> Interdepartmental Microbiology Graduate Program, College of Veterinary Medicine, Iowa State University, Ames, IA, United States, <sup>3</sup> Department of Veterinary Diagnostic and Production Animal Medicine, College of Veterinary Medicine, Iowa State University, Ames, IA, United States

Utilizing a simulated gastrointestinal medium which approximates physiological conditions within the mammalian GI tract, experiments aimed at isolating and identifying unique microbial metabolites were conducted. These efforts led to the finding that Escherichia coli, a common member of the gut microbiota, is capable of producing significant quantities of salsolinol. Salsolinol is a neuroactive compound which has been investigated as a potential contributor to the development of neurodegenerative diseases such as Parkinson's disease (PD). However the origin of salsolinol within the body has remained highly contested. We herein report the first demonstration that salsolinol can be made in vitro in response to microbial activity. We detail the isolation and identification of salsolinol produced by E. coli, which is capable of producing salsolinol in the presence of dopamine with production enhanced in the presence of alcohol. That this discovery was found in a medium that approximates gut conditions suggests that microbial salsolinol production could exist in the gut. This discovery lays the ground work for follow up in vivo investigations to explore whether salsolinol production is a mechanism by which the microbiota may influence the host. As salsolinol has been implicated in the pathogenesis of PD, this work may be relevant, for example, to investigators who have suggested that the development of PD may have a gut origin. This report suggests, but does not establish, an alternative microbiota-based mechanism to explain how the gut may play a critical role in the development of PD as well other conditions involving altered neuronal function due to salsolinol-induced neurotoxicity.

Keywords: salsolinol, Parkinson's disease, microbial metabolic activity, gut-brain-axis communication, gut origin for Parkinson's disease

# INTRODUCTION

Emerging evidence has suggested that there is an association between certain diseases and an altered gut-microbiome. Metabolites produced in the gut have the capacity to modulate disease (Lyte, 2016). It has been noted that bacteria produce and utilize many of the same neurochemicals that are used by the host's neurophysiological system with clinical implications ranging from infection

#### Edited by:

Jie Yin, Institute of Subtropical Agriculture (CAS), China

#### Reviewed by:

Xiang Duan, Northwest A&F University, China Kai Wang, Institute of Apiculture Research (CAAS), China Hui Han, Chinese Academy of Sciences, China

> \*Correspondence: Mark Lyte mlyte@iastate.edu

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 23 September 2018 Accepted: 29 November 2018 Published: 18 December 2018

#### Citation:

Villageliú DN, Borts DJ and Lyte M (2018) Production of the Neurotoxin Salsolinol by a Gut-Associated Bacterium and Its Modulation by Alcohol. Front. Microbiol. 9:3092. doi: 10.3389/fmicb.2018.03092

to behavioral modification through the microbiota-gut-brain axis (Lyte, 2014; Lyte and Brown, 2018). Our lab has sought to identify metabolites, which are produced by the microbiota and have the capacity to influence the host. As part of our approach, we utilize a food-based simulated small intestinal medium (sSIM) which simulates the gastrointestinal environment following food consumption. This enabled us to report the ability of common bacterial genera present within the gut to produce neurochemicals that otherwise would not be detected in more common laboratory media [Luria-Bertani broth (LB) as a prototypical example] which do not accurately reflect the gut in vivo milieu (Villageliu et al., 2018). During these experiments we noted the appearance of a prominent microbial product of unknown identity (**Figure 1**) when cultures were inoculated with Escherichia coli. As detailed herein, we determined that this microbial product was salsolinol.

One of the most investigated chemicals implicated in the development of PD is salsolinol, a chemical which was first associated with PD in the 1970s when it was detected in the urine of patients receiving L-dopa therapy (Kurnik-Łucka et al., 2018). Pharmacologically, salsolinol is a tetrahydroisoquinoline derivative with neuroactive properties most often viewed in the context of its potential as a neurotoxin (Kang, 2013; Mozd˙ ze˙ n et al., 2015 ´ ; Chen et al., 2018). Possessing the group defining catechol moiety of catecholamines, salsolinol can interact with various adrenergic receptors. Additionally, it has been shown that (R)-SAL and (S)-SAL activate the µ-opioid receptor by the classical G protein-adenylate cyclase pathway in a morphine-like interaction (Berríos-Cárcamo et al., 2016). There is some evidence that salsolinol may cross the blood brain barrier. In vivo, intraperitoneal administration resulted in accumulation of salsolinol within the neostriatum in dialysate (Quintanilla et al., 2014). It is known that other tetrahydroisoquinolone derivatives can cross the blood–brain barrier and it has also been reported that salsolinol administered systemically, can alter laboratory animal behavior which indirectly suggests that salsolinol could cross the blood-brain barrier (Kurnik-Łucka et al., 2018).

Salsolinol may lead to oxidative stresses that culminate in the protein aggregation of alpha-synuclein, an activity highly relevant in PD. Salsolinol induces an oxidative modification in cytochrome c. When cytochrome c is incubated with salsolinol, cytochrome c is oxidized and aggregates in a concentration dependent manner (Kang, 2013). Cytochrome c may co-localize with alpha-synuclein and initiate the oligomerization of alpha-synuclein (Kumar et al., 2016) and it is conceivable that cytochrome c altered by salsolinol may trigger or enhance this effect. More generally, salsolinol may be a neurotoxic and oxidative insult that could initiate a cascade leading or contributing to disease.

To date, the routes through which salsolinol is produced and accumulates within the body, particularly within the brain, have been contested. The presence of salsolinol within the environment has long been recognized and the synthesis of simple tetrahydroisoquinolines under mild physiological conditions such as those that occur in plants was described as far back as 1936 (Georg and Stiehl, 1936). Many dietary sources of salsolinol have been reported, with some sources like bananas having in excess of 5 µg salsolinol/gram wet weight (Kurnik-Łucka et al., 2018). An unequal distribution in the R vs. S enantiomers within the brain has led some to suggest the possibility of in situ synthesis by a "salsolinol synthase" which selectively forms the R enantiomer. Within the last year, a salsolinol synthase enzyme has been identified and described (Chen et al., 2018). Conversely, it has also been shown that salsolinol can be formed by the Pictet-Spengler mechanism in which aldehyde and dopamine react to form a racemic mixture of both the R and S forms of salsolinol (Kurnik-Łucka et al., 2018).

The finding that enterobacteria can produce salsolinol is significant. In addition to offering a possible explanation as to where salsolinol may be generated with the body, the production of salsolinol suggests a novel mechanism by which the microbiota might impact host health. There are many examples of the microbiota modulating inflammatory conditions, and proteobacteria (of which enterobacteriacae is a family) are commonly associated with human disease (Rizzatti et al., 2017). The finding that enterobacteriacae can produce salsolinol corroborates studies which show that metabolites derived from the microbiota may impact neurological conditions as well. For example, in a study which utilized microbiome transplants from healthy and Parkinsonian human donors into mice which overexpressed the alpha-synuclein protein, it was demonstrated that mice with a microbiota from Parkinson's patients increased impairment (Sampson et al., 2016). In particular, it has recently been recognized that enterobacteriaceae are positively associated with the severity of postural instability and gait difficulty in Parkinson's disease (Scheperjans et al., 2015; Tremlett et al., 2017).

While we have chosen to detail the significance of salsolinol to Parkinson's disease as a prototypic example, it should also be noted that the production of salsolinol carries implications beyond those of Parkinson's disease. For example, the ability of salsolinol to modulate the action of ethanol on neuronal activity in the region of the brain associated with motivation has been shown to occur in dopaminergic neurons (Melis et al., 2015). It has been suggested that an opioid action of salsolinol or its enantiomers is involved in the rewarding effects of ethanol (Berríos-Cárcamo et al., 2016).

# MATERIALS AND METHODS

#### Microorganisms

Twelve isolates of enterobacteria were obtained. Of these twelve, eight of these were isolates of E. coli: four were environmental isolates from livestock including chickens (ML1160-ML1162) and swine (ML1084); strains designated BW25113 and JW1228 were obtained from the Coli Genetic

Stock Center (New Haven, CT, United States) and represent the parent strain of the Keio knock collection and and alcohol dehydrogenase mutant, respectively (Baba et al., 2006). Isolates of Enterobacter cloacae and Citrobacter freundii were obtained from the Iowa State Veterinary Medicine Diagnostic Laboratory.

# Chemicals and Reagents

Salsolinol hydrobromide was purchased from Sigma-Aldrich (St. Louis, MO, United States). LC-MS grade acetonitrile, methanol, water, formic acid, and ammonium acetate were purchased from Fisher Scientific (Pittsburgh, PA, United States).

# Processing of Samples for HPLC With Electrochemical Detection

Optimized conditions for the processing, separation and quantification of catecholamines from bacterial cultures were previously developed (Villageliu et al., 2018). Post growth, cultures were acidified with the addition of 10 µL of 10N hydrochloric acid (HCl) for every 1 mL of medium. Culture medium was centrifuged (3000 × g, 4◦C for 15 min) to remove insoluble fiber, denatured proteins and other precipitates. The sample supernatant was further purified by passage through a 2 kDa molecular weight cut off filter (MWCO). Samples were stored at −80◦C.

Quantification of neurochemicals was performed by UHPLC-ECD on a Dionex UHPLC system which consisted of the following components: a Dionex Ultimate 3000 autosampler, a Dionex Ultimate 3000 pump and a Dionex Ultimate 3000 RS electrochemical detector (Thermo Scientific, Sunnyvale, CA, United States). Separation was achieved using buffered 10% acetonitrile mobile phase (MD-TM mobile phase, Thermo Scientific), a 150 mm, 3 µm Hypersil BDS C18 column (Thermo Scientific) and flow rate of 0.6 mL min−<sup>1</sup> . Prior to injection, samples were held at 4 ◦C by the autosampler. Electrochemical detection (ECD) was achieved with a 6041RS glassy carbon electrode set to 400 mV.

# Isolation of Salsolinol

Escherichia coli was grown anaerobically in sSIM (Villageliu et al., 2018) for 24 h in the presence of 1 mM dopamine. Following growth, samples were acidified with the addition of 10 µL of 10 M HCl for every 1 mL of medium. To ensure acidification did not contribute to salsolinol formation, a second subset of samples was processed without acid treatment. Samples were centrifuged at 3000 × g for 15 min at 4◦C. Supernatant was passed through a 2 kDa molecular weight cut off filter and analyzed by UHPLC-ECD. Chromatography demonstrated the presence of a distinct chemical response with a retention time of 4.1 min in both acidified and centrifuge only subsets. 5 mL of filtered supernatant was further treated by mixing with 400 mg of affinity beads specific for catechols (Bio-Rad, Hercules, CA, United States). UHPLC was done on eluted samples to ensure the presence of the peak at 4.1 min. Fractional collection based on retention time was used to isolate the peak from remaining impurities. Fractions were pooled and concentrated by centrifugal rotoevaporation using a (CentriVap <sup>R</sup> Labconco, Kansas City, MO, United States). Concentrated samples were reanalyzed by HPLC-ECD to ensure that no deviation in retention time had occurred before delivery of the sample for analysis by LC-MS. Upon identification of the peak as salsolinol, a salsolinol standard (Sigma-Aldrich) was ordered and tested on UHPLC-ECD. The retention time of this standard was an identical match.

#### LC-MS Analysis

fmicb-09-03092 December 15, 2018 Time: 15:13 # 4

LC-MS analyses were performed using an Ultimate 3000 UHPLC system coupled to a Q Exactive Focus hybrid quadrupoleorbitrap mass spectrometer (Thermo Fisher Scientific, San Jose, CA, United States).

For reversed phase chromatography, the LC column was an Accucore aQ, 2.6 µm, 2.1 × 100 mm from Thermo Fisher Scientific (Waltham, MA, United States). The mobile phase consisted of (A) 0.1% formic acid in water and (B) 0.1% formic acid in methanol. The LC gradient was: start to 1.0 min: 0% B, linear ramp to 11.0 min: 98% B, held for 2 min, and then returned to initial conditions and held for 2 min prior to the next injection. The flow rate was 300 µL/min and the column temperature was maintained at 35◦C. An injection volume of 10 µL was used.

For HILIC (hydrophilic interaction liquid chromatography), the LC column was a SeQuant ZIC-HILIC, 5 µm, 200 Å, 2.1 × 150 mm from Millipore Sigma (Darmstadt, Germany). The mobile phase consisted of (A) 95/5 (v/v) water/acetonitrile containing 10 mM ammonium acetate and (B) 5/95 (v/v) water/acetonitrile containing 10 mM ammonium acetate. The LC gradient was: start to 0.5 min: 99% B, linear ramp to 15.5 min: 50% B, held for 3 min, returned to initial conditions over 1 min, and held for 8 min prior to the next injection. The flow rate was 500 µL/min and the column temperature was maintained at 40◦C. An injection volume of 10 µL was used in full mass range mode and a 1 µL injection volume was used in MS/MS mode.

For LC-MS analysis with reversed phase chromatography, the mass spectrometer was operated in positive and negative electrospray ionization (ESI) mode. In positive ESI mode the following conditions were used: spray voltage: 3.5 kV, ion transfer capillary temperature: 256◦C, S-lens RF level: 55, sheath gas flow rate: 48, auxiliary gas flow rate: 11, and sweep gas flow rate: 2. In negative ESI mode, the following conditions were used: spray voltage: 2.5 kV, ion transfer capillary temperature: 256◦C, S-lens RF level: 55, sheath gas flow rate: 48, auxiliary gas flow rate: 11, and sweep gas flow rate: 2. (S-lens RF level and gas flow rates are in arbitrary units.) For full mass range analysis, the orbitrap was operated at a resolution setting of 70,000 over a m/z range of 100–1000 with an automatic gain control setting of 1E10<sup>6</sup> and a maximum ion time setting of auto. For MS/MS analysis, the mass spectrometer was operated in parallel reaction monitoring (PRM) mode at a resolution of 35,000. Collision energy settings of 20, 30, or 40 were used in both positive and negative ESI mode.

For LC-MS analysis with HILIC chromatography, the mass spectrometer was operated in positive electrospray ionization (ESI) mode using the following conditions: spray voltage: 3.0 kV, ion transfer capillary temperature: 350◦C, S-lens RF level: 45, sheath gas flow rate: 60, auxiliary gas flow rate: 15, and sweep gas flow rate: 2. (S-lens RF level and gas flow rates are in arbitrary units.) For full mass range analysis, the orbitrap was operated at a resolution setting of 70,000 over a mass range of 65–975 with an automatic gain control setting of 1E10<sup>6</sup> and a maximum ion time setting of auto. For MS/MS analysis, the mass spectrometer was operated in parallel reaction monitoring (PRM) mode at a resolution of 35,000 and collision energy settings of 20, 30, or 40.

Data were acquired using Xcalibur 4.0 software and product ion spectra were searched against the mzCloud online mass spectral database<sup>1</sup> using FreeStyle 1.1 software (all from Thermo Fisher Scientific, San Jose, CA, United States).

The isolated fraction was initially analyzed using reversed phase LC-MS in positive ion ESI mode with full mass range analysis. The total ion chromatogram (TIC) revealed significant peaks at retention times of 0.66 and 0.95 min and a smaller peak at a retention time of 1.53 min. Inspection of the mass spectrum of the peaks at 0.66 and 0.95 min indicated that these peaks were consistent with salt and solvent clusters. Inspection of the mass spectrum of the peak at 1.53 showed an ion with a measured accurate m/z of 180.1015. An extracted ion chromatogram (XIC) for the m/z 180.1015 ion (±2.5 ppm) is shown in **Figure 2A**. Product ion spectra were acquired on the m/z 180.1015 ion at collision energy (CE) settings of 20, 30, and 40. The product ion spectrum for the m/z 180.1015 ion at CE = 20 is shown in **Figure 2C**. FreeStyle 1.1 software was used to process this spectrum and submit it for database searching against the mzCloud<sup>1</sup> online orbitrap-based product ion spectral database. (The mzCloud database contains data on over 8000 small molecule compounds and has nearly 3 million spectra). The top database search result returned was for the compound salsolinol. The modified NIST match score for salsolinol was 92.9. The compound with the next closest modified NIST match had a score of 78.2. Salsolinol was also the top database search result, with similar match scores, for the product ion spectra acquired at CE = 30 and 40. In addition, the exact m/z measured for the compound, 180.1015, was consistent with the calculated exact m/z for salsolinol, 180.1019, with 2.2 ppm error. All of these results are consistent with the unknown compound in the isolated fraction corresponding to salsolinol.

The isolated fraction was next analyzed using reversed phase LC-MS in negative ion ESI mode with full mass range analysis. The TIC revealed a significant peak at a retention time of 0.66 min and a smaller peak at a retention time of 1.53 min. Inspection of the mass spectrum of the peak at 0.66 min indicated that this peak was consistent with salt and solvent clusters. Inspection of the mass spectrum of the peak at 1.53 showed an ion with a measured accurate m/z of 178.0862. Product ion spectra were acquired on the m/z 178.0862 ion at collision energy (CE) settings of 20, 30, and 40 (data not shown). FreeStyle 1.1 software was used to process the product ion spectrum collected with a CE setting of 20 and to submit it for database searching against the mzCloud<sup>1</sup> online orbitrap-based product ion spectral database. The top database search result returned was for the compound salsolinol. The modified NIST match score for salsolinol was 93.8. The compound with the next closest modified NIST match had a score of 36.6. Salsolinol was also the top database search result, with similar match scores, for the product ion spectra acquired at CE = 30 and 40. In addition, the exact m/z measured for the compound, 178.0862, was consistent with the calculated exact

<sup>1</sup>mzcloud.org

m/z for salsolinol, 178.0873, with 6.2 ppm error. All of these results are also consistent with the unknown compound in the isolated fraction corresponding to salsolinol.

To confirm the putative identification of the compound contained in the fraction, an analytical standard of salsolinol was acquired and analyzed using reversed phase LC-MS positive ion ESI mass spectrometry procedures identical to the unknown

fraction. The salsolinol analytical standard material showed a retention time of 1.50 min with the reversed phase LC-MS system. This slight difference in retention time compared with the compound in the isolated fraction (1.53 min) is likely due to the high salt content in the isolated fraction. The measured exact m/z for the salsolinol analytical standard was 180.1019, which matches the calculated exact mass of 180.1019 with 0.0 ppm error. An XIC for the m/z 180.1019 ion is shown in **Figure 2B**. Product ion spectra for the salsolinol analytical standard were acquired at CE settings for 20, 30, and 40. The spectra at each of these CE settings were virtually identical to the compound in the isolated fraction. The product ion spectrum for the salsolinol analytical standard at CE = 20 is shown in **Figure 2D**.

To further confirm the identity of the compound in the isolated fraction as salsolinol, both the isolated fraction and the analytical standard salsolinol material were analyzed using an orthogonal chromatography mode known at hydrophilic interaction liquid chromatography or HILIC. In HILIC, polar compounds are generally more highly retained and elute later while more non-polar compounds are generally less highly retained and elute earlier. This is the reverse of the elution order with reversed phase liquid chromatography (McCalley, 2017).

The isolated fraction and salsolinol analytical standard were analyzed using HILIC LC-MS in positive ion ESI mode with full mass range analysis and MS/MS analysis. The XICs for m/z 180.1019 (± 2.5 ppm) for the isolated fraction and salsolinol analytical standard are shown in **Figures 3A,B**, respectively. Again, we believe the slight difference in retention times, 5.74 min for the isolated fraction versus 5.83 min for the analytical standard, can be attributed to the high salt content in the isolated fraction. The measured accurate mass for the m/z 180 component in the isolated fraction was 180.1020 while the measured accurate mass for the salsolinol analytical standard was 180.1019. These measured masses have 0.6 and 0.0 ppm error compared to the calculated exact m/z for salsolinol and 0.6 ppm error compared to one another.

Product ion spectra for the m/z 180 component in the isolated fraction and the salsolinol analytical standard were acquired in HILIC LC-MS positive ion ESI mode at CE settings of 20, 30, and 40. For all three CE settings, the product ion spectra for the m/z 180 component of the isolated fraction and the salsolinol analytical standard were virtually identical. The product ion spectra at CE settings of 20 for the m/z 180 component in the isolated fraction and the salsolinol analytical standard are shown in **Figures 3C,D**, respectively. Upon identification of molecule as salsolinol, a salsolinol standard was ordered and tested on UHPLC-ECD. The retention time of this standard was an identical match to the original unknown peak.

# Experimental Culture Conditions

Isolates used for inoculation were grown on TSA agar with 5% ovine blood overnight. Harvested colonies were suspended in peptone water and standardized to an OD<sup>600</sup> of 0.2. Inoculation was achieved by mixing 100 µL of organism suspension with 5 mL of medium. The production of salsolinol by various species of Enterobacteriaceae was tested in five types of media: Luria broth (LB), De Man, Rogosa, Sharpe (MRS) broth, brain heart infusion broth (BHI), tryptic soy broth (TSB), and sSIM. All media were tested and found to have little or no dopamine in their native state. Dopamine was supplemented by adding 100 µL of a 0.2 µm filtered 0.05 M solution of dopamine hydrochloride to a total volume of 5 mL of medium.

Culture broths were grown anaerobically, at 37◦C for 24 h. In the case of sSIM, which is a complex suspension, agitation was provided with magnetic stir bars to keep particles evenly distributed. After 24 h, samples were processed for analysis by UHPLC.

# RESULTS

Escherichia coli was grown anaerobically in sSIM for 24 h in the presence of 1 mM dopamine. Following growth, samples were acidified with the addition of 10 µL of 10M HCl for every 1 mL of medium. To ensure acidification did not contribute to salsolinol formation, a second subset of media samples was processed without acid treatment. Samples were centrifuged at 3000 × g for 15 min at 4◦C. Supernatant was passed through a 2 kDa molecular weight cut off filter and analyzed by UHPLC (ultra-high performance liquid chromatography)- ECD (electrochemical detection). UHPLC-ECD demonstrated the presence of a distinct chemical response with a retention time of 4.1 min in both acidified and centrifuge only subsets (**Figure 1**).

5 mL of filtered supernatant was further treated by mixing with 400 mg of affinity beads specific for catechols (Bio-Rad, Hercules, CA, United States). UHPLC was performed on eluted samples to ensure the presence of the peak at 4.1 min remained. Fractional collection based on retention time was used to isolate and further purify the peak. Fractions were pooled and concentrated by a centrifugal concentrator (Labconco, Kansas City, MO, United States) and then subjected to analysis by LC-MS (liquid chromatography-mass spectrometry) for identification of the collected peak.

The compound in the isolated fraction was identified and confirmed as salsolinol by analysis in full mass range and tandem mass spectrometry (MS/MS) modes at multiple collision energies using a UHPLC LC-MS system (Thermo Fisher Scientific, San Jose, CA, United States). Results from these analyses were consistent with those obtained using an authentic analytical standard of salsolinol. Chromatographically, both reversedphase and orthogonal HILIC (hydrophilic interaction liquid chromatography) separation modes were employed. In all cases, the chromatographic retention times matched between the compound in the isolated fraction and the salsolinol standard. In addition, the measured accurate masses (m/z ratios) for all pseudomolecular [(M+H)<sup>+</sup> and (M-H)−] ions and product ions were consistent between the compound in the isolated fraction and the salsolinol standard. The measured accurate masses were also consistent with calculated exact masses within expected experimental error. Product ion relative intensities were consistent between the compound in the isolated fraction and the salsolinol standard at all collision energies. Extracted ion chromatograms and product ion spectra for the compound in the isolated fraction and the salsolinol analytical standard are

shown in **Figures 2**, **3**. Following the assignment of the isolated fraction as salsolinol by mass spectrometry, we were able to match the retention time of a commercial salsolinol standard (Sigma-Aldrich, St. Louis, MO, United States) to that of experimental samples.

Medium choice appears to play a significant impact on the final concentration of salsolinol observed. Twelve isolates of enterobacteria were obtained. Of these twelve, eight of these were isolates of E. coli: four were environmental isolates from livestock including chickens (ML1160-ML1162) and swine (ML1084); strains designated BW25113 and JW1228 were obtained from the Coli Genetic Stock Center and represent the parent strain of the Keio knock out collection and an alcohol dehydrogenase mutant, respectively (Baba et al., 2006). Isolates of Enterobacter cloacae and Citrobacter freundii were obtained from the Iowa State Veterinary Medicine Diagnostic Laboratory.

In 8 of the 10 trials done with enterobacteria, the greatest amount of salsolinol was produced in MRS, averaging 220 µM among E. coli strains. In un-inoculated controls supplemented with 1 mM dopamine, only trace salsolinol was detected. Production in MRS showed comparatively little variation, with the standard deviation of the mean averaging 6.6 µM across six E. coli strains. Production of salsolinol in sSIM was also robust, averaging 174 µM across the E. coli strains. Production in BHI and TSB was relatively limited, averaging 90 µM in BHI and only 76 µM in TSB. For E. coli, there was no discernable production of salsolinol in LB and salsolinol levels averaging 7 µM were comparable to the LB controls. The addition of ethanol substantially impacted the salsolinol production of five tested enterobacteria. The effect was particularly noticeable in E. coli from the Keio knock out collection in which the wild type generated an average of 206 µM (±8 µM) in the absence of alcohol, but a remarkable 920 µM (±8 µM) in the presence of 4% ethanol. This production was essentially unaffected in the single alcohol dehydrogenase knock out JW1228 which generated an average of 232 µM (±16 µM) in the absence of alcohol and 890 µM (±28 µM) in the presence of ethanol.

#### DISCUSSION

This study provides the first evidence that common and abundant members of the gut microbiota, namely E. coli and several related enterobacteria, have the capacity to produce salsolinol and that production is enhanced in the presence of alcohol (**Figure 4**). When E. coli is inoculated in a medium containing dopamine, distinct chromatographic evidence of the production of a new compound is evident (**Figure 1**). Isolation of this chemical through retention time based fractional collection yielded a sample suitable for mass spectrometry. As shown, the results of mass spectrometry, taken collectively, confirm the identification of the positive ion m/z 180 component in the isolated fraction as salsolinol (**Figure 2**). The data presented here are consistent with and fulfill the criteria for an endogenous metabolite to be considered an identified compound as set forth by groups including the Chemical Analysis Working Group of the Metabolomics Standards Initiative (Sumner et al., 2007) and others (Creek et al., 2014; Schymanski et al., 2014).

The identification of novel microbial metabolic activity by E. coli (the most highly studied biological organism in the world) that potentially impacts the understanding of the mechanisms by which neurodegeneration in the gut may influence the development of PD raises several questions. First, since Enterobacteriaceae, and E. coli in particular, are among the most highly studied biological organisms, why has the production of salsolinol not been described before either from actual in vitro experiments or from bioinformatic mining of the extensive databases that are now available? Most likely, the reason(s) is due to the environmental growth conditions in which the present work was conducted in that bacteria were evaluated in complex medium reflective of the host-based milieu (Villageliu et al., 2018). This was achieved through the simulated digestion of foodstuff in a manner consistent with host physiology (Mackie and Rigby, 2015). For many decades, LB broth has been the media most commonly used to grow E. coli (Paul De Vos, 2009). As shown in **Figure 4**, when grown in LB, salsolinol is not produced by E. coli. Presumably, the medium either lacks a fundamental co-factor or does not encourage the activation of requisite genes. That we found a large difference in the production of salsolinol across various media (**Figure 4**) demonstrates the value of growing organisms in conditions that closely mimic the digested food-containing in vivo milieu of the gut. As such, it should not therefore be surprising that no bioinformatics-based databases contain any indication that E. coli can produce the neurotoxin salsolinol.

A second question concerns the genetic mechanism(s) by which salsolinol is made. Previously, several pathways leading to the production of salsolinol have been suggested (Chen et al., 2011). We have expanded on this proposal in order to incorporate the finding that alcohol enhances the production of salsolinol in microbes (**Figure 5**). It is well established that alcohol is converted to acetaldehyde via alcohol dehydrogenase in E. coli. An increase in acetaldehyde concentration would, by Le Chatelier's principle (Treptow, 1980), increase the production of salsolinol for either an enzymatic or non-enzymatic mechanism. While an enzyme driven synthesis seems plausible, the nonenzymatic Pictet-Spengler mechanism (Stockigt et al., 2011) cannot be ruled out. Via the Pictet-Spengler reaction, a microorganism could facilitate the formation of salsolinol by producing aldehyde in an environment that contains dopamine. A common pathway for the production of aldehyde is via alcohol dehydrogenase which generates aldehyde from alcohol. We found that the presence of additional alcohol (**Figure 4**) does indeed favor the production of salsolinol. This could be consistent with either an enzyme that uses aldehyde as a substrate or the Pictet-Spengler mechanism. It is important to note however, that a single alcohol dehydrogenase mutant JW1228 from the Keio knock out collection (Baba et al., 2006) showed no statistical difference in salsolinol production in comparison to the wild type alcohol dehydrogenase mutant. This finding argues against mechanisms that rely on aldehyde. It is conceivable that an enzyme catalyzed reaction could form salsolinol in the absence of aldehyde, the only known salsolinol synthase enzyme

forms salsolinol by a reaction with pyruvic acid, not aldehyde (Chen et al., 2018). Perhaps alcohol influences the microbial production of salsolinol by some other less direct pathway. It seems unlikely that aldehyde would be made by microorganisms in media like MRS, TSB, and BHI but not in LB. However, Pictet-Spengler mechanism still cannot be completely ruled out because the parent strain for this species is also known to have two copies of the alcohol dehydrogenase gene and it remains conceivable that one knock out is not sufficient to see a difference. Ultimately, mechanistic determinations will rely on follow up work.

As a putative salsolinol synthase has recently been described (Chen et al., 2018), we conducted a tblastn against the E. coli genome. Using the relatively loose parameters of BLOSUM45

and extension gap cost of 1, we were unable to find any homologous sequences. If there is a microbial enzyme, it is likely structurally distinct from the mammalian enzyme. We are currently conducting further research including the use of transposon mutagenesis to identify the genetic mechanism(s) by which E. coli produces salsolinol.

A third question concerns the conditions tested herein in vitro and whether they are representative of the conditions likely to be found in vivo. We tested what may seem like a relatively high exposure of 1 mM of dopamine and allowed the organisms twenty-four hours to interact with dopamine. Consider however, that within the localized pockets where microbes reside, this concentration could still be feasible. Dopamine is readily available and secreted throughout certain portions of the GI tract (Mezey et al., 1996, 1998) and a biofilm living in close proximity to dopamine secreting cells will experience a much higher effective concentration than cells which experience only the distant diffused concentration of the molecule. Research into the effects of dopamine on neural cell lines, including astroglial cells has covered the range of 1 mM (Hirrlinger et al., 2002) and some research has even tested concentrations as high as 100 mM (Chase et al., 2004).

The demonstration that one of the most abundant members of the gut microbiota, E. coli, can facilitate the synthesis of salsolinol does not by itself prove a causative link between the microbiota and PD. However, it is worth noting that a gut-based origin for PD has already been proposed in which dysregulation of the neuro-immune brain-gut axis could lead to the occurrence of enteric neuro-inflammatory conditions (Lionnet et al., 2018; Pellegrini et al., 2018). Interestingly, PD is frequently associated with functional gastrointestinal disorders including infrequent bowel movements, abdominal distension and constipation which can occur throughout all stages of the neurodegenerative process. Emerging evidence has suggested that there is an association between PD and an altered gutmicrobiome, metabolites produced in the gut may modulate the disease. As one of the most investigated chemicals implicated in the development of PD is salsolinol (Kurnik-Łucka et al., 2018), it follows that the production of salsolinol by a gut microbe could prove highly relevant.

To date the role of salsolinol has been brain-centric (Kurnik-Łucka et al., 2018) with little or no role for its ability to effect neurodegeneration described outside of the central nervous system. Further, the demonstration that alcohol dramatically increases the production of salsolinol (**Figure 4**) suggests a possible mechanism by which to explore the purported association between excessive alcohol consumption and PD (Liu et al., 2013; Bettiol et al., 2015). It should be noted, however, that by itself this study does not offer sufficient evidence to establish that salsolinol production contributes to the pathogenesis of neurodegenerative diseases such as PD. However, in light of previous reports showing a correlation between salsolinol and various conditions (Kurnik-Łucka et al., 2018), the potential capacity for the microbiota to produce salsolinol in vivo should be considered. Further, previous reports which have measured salsolinol production in alcoholics have done so using urine (Matsubara et al., 1985; Dostert et al., 1991). Subsequent studies have produced conflicting results (Feest et al., 1992; Haber et al., 1996). The results of the present study suggest that fecal matter in addition to urine should be utilized in such studies.

In summary, this is the first known report of a microbe producing salsolinol. While dietary and host derived sources of salsolinol may still be relevant, the capacity for certain members of the microbiota to produce significant quantities of salsolinol suggests that they may be a source for much of the salsolinol that has been reported in host tissues. Follow up studies may use this as a basis upon which to examine the role of microbiota in the gut and potentially on the pathogenesis of disease. Further, this report may be viewed as complementary to that of the Braak hypothesis (Braak et al., 2003a,b, 2006) since it is conceivable that salsolinol produced within the gut could effect changes locally in the enteric nervous system. Though the total salsolinol burden experienced by a host might be manageable in a healthy host, it is possible that in some individuals a dysregulated microbiome or other pre-existing condition may facilitate the initiation of a disease cascade.

# AUTHOR CONTRIBUTIONS

All authors contributed to the intellectual development of this paper. DV under the principal investigator ML made the initial

# REFERENCES


observation and isolation of a unique metabolite derived from microbial activity. Characterization of this molecule was done in collaboration with the analytic expertise of DB. Follow up experiments were planned and conducted by both ML and DV.

# FUNDING

This study was supported by the United States Department of Defense, Office of Naval Research award #N00014-15-1-2706 to ML and internal Iowa State University funds provided by the W. Eugene Lloyd Chair in Toxicology to ML.

#### ACKNOWLEDGMENTS

The assistance of Professor Gregory Phillips (Iowa State University) in providing guidance in the selection and procurement of alcohol dehydrogenase mutants is gratefully acknowledged. The technical assistance of Meicen Liu and Karrie Wright is also acknowledged.


interkingdom communication within the microbiota-gut-brain axis. PLoS One 13:e0191037. doi: 10.1371/journal.pone.0191037


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Villageliú, Borts and Lyte. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Dietary Supplementation With Chinese Herbal Residues or Their Fermented Products Modifies the Colonic Microbiota, Bacterial Metabolites, and Expression of Genes Related to Colon Barrier Function in Weaned Piglets

Jiayi Su1,2, Qian Zhu<sup>1</sup> , Yue Zhao1,2, Li Han1,2, Yulong Yin<sup>1</sup> , Francois Blachier<sup>3</sup> , Zhanbin Wang<sup>2</sup> and Xiangfeng Kong<sup>1</sup> \*

<sup>1</sup> Hunan Provincial Key Laboratory of Animal Nutritional Physiology and Metabolic Process, Key Laboratory of Agro-ecological Processes in Subtropical Region, National Engineering Laboratory for Pollution Control and Waste Utilization in Livestock and Poultry Production, Institute of Subtropical Agriculture, Chinese Academy of Sciences, Changsha, China, <sup>2</sup> College of Animal Science and Technology, Henan University of Science and Technology, Luoyang, China, <sup>3</sup> Nutrition Physiology and Ingestive Behavior, UMR 914 INRA/AgroParisTech/Universite Paris-Saclay, Paris, France

#### Edited by:

Yuheng Luo, Sichuan Agricultural University, China

#### Reviewed by:

Jun He, Sichuan Agricultural University, China Yong Su, Nanjing Agricultural University, China

> \*Correspondence: Xiangfeng Kong nnkxf@isa.ac.cn

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 20 August 2018 Accepted: 07 December 2018 Published: 21 December 2018

#### Citation:

Su J, Zhu Q, Zhao Y, Han L, Yin Y, Blachier F, Wang Z and Kong X (2018) Dietary Supplementation With Chinese Herbal Residues or Their Fermented Products Modifies the Colonic Microbiota, Bacterial Metabolites, and Expression of Genes Related to Colon Barrier Function in Weaned Piglets. Front. Microbiol. 9:3181. doi: 10.3389/fmicb.2018.03181 To explore the feasibility of dietary Chinese herbal residue (CHR) supplementation in swine production with the objective of valorization, we examined the effects of dietary supplementation with CHR or fermented CHR products on the colonic ecosystem (i.e., microbiota composition, luminal bacterial metabolites, and expression of genes related to the intestinal barrier function in weaned piglets). We randomly assigned 120 piglets to one of four dietary treatment groups: a blank control group, CHR group (dose of supplement 4 kg/t), fermented CHR group (dose of supplement 4 kg/t), and a positive control group (supplemented with 0.04 kg/t virginiamycin, 0.2 kg/t colistin, and 3000 mg/kg zinc 0.04 kg/t virginiamycin, 0.2 kg/t colistin, and 3000 mg/kg zinc oxide). Our results indicate that dietary supplementation with CHR increased (P < 0.05) the mRNA level corresponding to E-cadherin compared with that observed in the other three groups, increased (P < 0.05) the mRNA level corresponding to zonula occludens-1, and decreased (P < 0.05) the quantity of Bifidobacterium spp. When compared with the blank control group. Dietary supplementation with fermented CHR decreased (P < 0.05) the concentration of indole when compared to the positive control group; increased (P < 0.05) the concentrations of short-chain fatty acids compared with the values measured in the CHR group, as well as the mRNA levels corresponding to interleukin 1 alpha, interleukin 2, and tumor necrosis factor alpha. However, supplementation with fermented CHR decreased (P < 0.05) interleukin 12 levels when compared with the blank control group. Collectively, these findings suggest that dietary supplementation with CHR or fermented CHR modifies the gut environment of weaned piglets.

Keywords: bacterial metabolites, Chinese herbal residues, colon barrier function, fermentation, microbiota, weaned piglets

# INTRODUCTION

fmicb-09-03181 December 21, 2018 Time: 19:3 # 2

Weaning is a critical stage for piglets, as this process is associated with alterations in the architecture and function of the gut and changes in the enteric microbiota (Boudry et al., 2004). Weaned piglets often exhibit underdeveloped immune systems, digestive disorders, and post-weaning diarrhea (Liu et al., 2016), all these events decreasing growth performance and causing economic loss for the swine industry (Amezcua et al., 2002). Antibiotics and the pharmacological addition of ZnO in the diets have long been used to solve post-weaning problems (Barton, 2000). However, their continuous use and misuse have led to the emergence of drug resistant bacteria, the risk of residual antibiotics in animal products, and the release of Zn-residues in the environment (Schwarz et al., 2001). Therefore, many alternatives to antibiotics and ZnO have been explored in response to this situation (Hu et al., 2013; Kong et al., 2014).

Among potential alternatives, Chinese herbs (CH) are one of the most promising candidates (Kong et al., 2011). Recent findings indicated that the use of CH as a dietary additive is able to enhance gastrointestinal health indicators in weaned piglets (Kong et al., 2007b; Ding et al., 2011). Moreover, dietary supplementation with Chinese herbal ultra-fine powder enhances both cellular and humoral immunity in early weaned piglets (Kong et al., 2007a). Chinese herbal residues (CHR) are produced during CH production and processing. The use of partial CH extraction technology leads to CHR containing high levels of nutrients and bioactive ingredients, and the resulting CHR exhibits the same efficiency as CH with regards to enhanced gastrointestinal health, increased nutrient absorption, and immune system stimulation. However, the presence of cell walls in CHR results in the presence of a large number of active components that may act on the gut ecosystem. Therefore, limited utilization of CHR by animals and environmental pollution from byproducts represent two major drawbacks of using CHR as a feed supplement.

The process adopted for CHR fermentation involves inoculation of only one or a few types of probiotics from beneficial microbiota to produce fermented CHR. These products contain compounds originating from the microbiota, including bioactive ingredients, and bacterial metabolites (Sun et al., 2011). In addition, the levels of anti-nutritional factors and potentially toxic components are reduced, and numerous functional secondary metabolites are produced, thus resulting in better characteristics of fermented CHR compared to non-fermented CHR (Wen et al., 2013).

The production of CHR and fermented CHR remains limited despite the increasing popularity of herbal fermentation. Our previous study showed that fermented CHR can improve growth performance and digestion in weaned piglets up to a certain extent, without, however, any significant impact on the rate of diarrhea and intestinal morphology (Su et al., 2016). Recent studies have highlighted the pivotal role of the gut microbiota in animal health and diseases (Tremaroli and Bäckhed, 2012). We hypothesized that dietary supplementation with CHR or fermented CHR may exert beneficial effects on the colonic ecosystem, and subsequently tested the effects of these compounds on the colonic microbiota, bacterial metabolite concentration, and gene expression related to the intestinal barrier function in weaned piglets. A reduction in pathogen concentration may result in a reduction in toxic bacterial metabolite production, and of competition with host for nutrient utilization, thus improving host weight gain. Therefore, the aim of this study was to further test this hypothesis and to evaluate the interest of CHR and their fermented products utilization as feedstuff or feed additives for swine production.

# MATERIALS AND METHODS

#### Ethics Statement

The experimental design and procedures in this study were reviewed and approved by the Animal Care and Use Committee of the Institute of Subtropical Agriculture, Chinese Academy of Sciences. The processing of animal experiments and sample collection strictly followed the relevant guidelines.

# Preparation and Composition of Fermented CHR

The fermentation substrates used in the present study, including residues of Radix Rehmanniae Preparata, Fructus Crataegi, Pericarpium Citri Reticulatae, Fructus Hordei Germinatus, and Radix Glycyrrhizae, were supplied by Hunan Sunaccord Biological Technical Co., Ltd., China. The ratio of Radix Rehmanniae Preparata residues, Fructus Crataegi residues, Pericarpium Citri Reticulatae residues, Fructus Hordei Germinatus residues, and Radix Glycyrrhizae residues was 4:2:2:1:1.

After sterilization, mixed CHR containing 40–60% water was inoculated with Bacillus subtilis and yeast, incubated at 28–32◦C for 22–26 h in anaerobic conditions, inoculated with Lactobacillus and Bifidobacterium and incubated at 31–34◦C for 24 h in anaerobic conditions. The ratio of B. subtilis, yeast, Lactobacillus, and Bifidobacterium was 5:2:2:1, and the concentration of live bacteria was > 2 × 10<sup>10</sup> colony forming units per gram (CFU/g). After fermentation, the fermented products were dried in a vacuum, pulverized, and packed for use. The analyzed contents (%) of nutrients based on dry matter (with the exception of dry matter) in CHR and fermented CHR are listed in **Table 1**.

# Animals, Housing and Treatment

This study used a total of 120 piglets (Duroc × Landrace × Large White), weaned at 21 days of age with an average body weight of 6.12 ± 0.05 kg. Piglets were randomly assigned to one of four treatment groups, with five replicates per group and six piglets per replicate. The four treatment groups consisted of a blank control group (basal diet), CHR group (basal diet supplemented with 4 kg/t CHR), fermented CHR group (basal diet supplemented with 4 kg/t fermented CHR), and a positive control group (the basal diet supplemented with 0.04 kg/t virginiamycin, 0.2 kg/t colistin, and 3000 mg/kg zinc oxide). All pigs were housed in 2.0 m × 2.5 m pens with hard plastic slatted flooring, and


pigs had ad libitum access to drinking water and experimental diets. Each pen was equipped with a stainless-steel feeder and a nipple drinker. The room temperature was maintained at 25–27◦C. The composition and nutrient levels of the basal diet met the nutritional needs that National Research Council (2012) recommended for nursery piglets, which are shown in **Table 2**. The experiments lasted for 28 days.

#### Sample Collection and Preparation

At the end of the 28-day experimental period and 12 h after the last feeding, a medium-sized piglet from each replicate was sacrificed using general anesthesia (Kong et al., 2007b). After colon recovery, the intestinal contents from each colon (10 cm from the posterior to the ileocecal valve) were collected and stored at −20◦C for analysis of short-chain fatty acids (SCFAs), indoles, skatoles, ammonia (considered as the sum of NH<sup>3</sup> and NH<sup>4</sup> <sup>+</sup>), and the composition of microbiota. Colon tissue samples (approximately 2 cm) were collected, washed with cold physiological saline, immediately frozen in liquid nitrogen and


<sup>1</sup>The premix provided the following per kg of the diet: VA 6 200 IU, VD<sup>3</sup> 700 IU, VE 88 IU, VK 4.4 mg, VB<sup>2</sup> 8.8 mg, Pantothenate 24.2 mg, nicotinic acid 33 mg, Chloride choline 330 mg, Cu 10 mg, Zn 100 mg, Fe 145 mg, Mn 40 mg, Se 0.1 mg, I 0.3 mg. <sup>2</sup>Nutrient contents are calculated values.

stored at −80◦C until further analysis by RNA extraction was conducted.

#### DNA Extraction and Analysis of the Quantity of Colonic Microbiota

Total microbial DNA was extracted and purified using a QIAamp DNA Stool Kit (Qiagen, Hilden, Germany) and stored at −80◦C. The 16S rRNA gene sequences of Bacteroidetes, Bifidobacterium spp., Clostridium cluster IV, Clostridium cluster XIVa, Escherichia coli, Firmicutes, Lactobacillus, and total bacteria (Raveh-Sadka et al., 2015) were cloned into the pMD19-T vector. Gene sequences were amplified from total colonic DNA using the primers listed in **Table 3**. A total of eight clones with 16S rRNA gene sequences belonging to different taxa were used as templates to test primer specificity. Standard curves were constructed with DNA from representative species for a concentration range from 10<sup>2</sup> to 10<sup>10</sup> DNA copies/mL using a Lightcycler 480II instrument (Applied Biosystems). General microbial DNA extracted from colonic contents and specific DNA from recombinant microbiota were quantified using RT-PCR. The reaction conditions were as follows: 2 min at 50◦C; an initial denaturation step at 95◦C for 5 min; 40 cycles of denaturation at 94◦C for 20 s, primer annealing at a species-specific temperature for 30 s, and primer extension at 60◦C for 1 min (Decroos et al., 2006).

#### Analysis of the Colonic Concentrations of Bacterial Metabolites

The contents of each colon were recovered by expulsion, homogenized, and centrifuged at 1000 × g for 10 min (Zhou et al., 2012). A mixture of supernatant fluid and 25% metaphosphoric acid solution (1:0.25 mL) was prepared to determine SCFAs by gas chromatography (Zhou et al., 2014). The concentration of ammonia in the supernatant fluid was measured at a wavelength of 550 nm using a UV-2450 spectrophotometer (Shimadzu,

TABLE 3 | 16S rRNA gene-targeted group-specific primers used in this study.


Japan) (Kong et al., 2014). Indoles and skatoles were analyzed as described previously (Kong et al., 2016).

#### Analysis of the Levels of mRNAs Corresponding to Epithelial Cell Proteins, Cytokines, and TLR4 Signaling Pathway Proteins in Colonic Tissues

Total RNA was isolated from colonic tissues using TRIzol reagent (Invitrogen, Carlsbad, CA, United States), and samples were treated with DNAse. The RNA quality was checked using 1% agarose gel electrophoresis followed by staining with 10 µg/mL ethidium bromide. The OD260/OD280 ratio of extracted RNA was between 1.8 and 2.0. Reverse transcription was performed using a Prime Script RT Reagent Kit with gDNA Eraser (Takara, Dalian, China). The mRNA levels of the selected genes in the colon were determined by real-time quantitative (RT-qPCR) as described previously (Li et al., 2015; Wan et al., 2017, Wan et al., 2018b). Selected genes were corresponding to the following proteins: epithelial cell proteins, including E-cadherin, occludin, and zonula occludens-1 (ZO-1); anti-inflammatory cytokines, including interleukin-4 (IL-4), IL-10 and transforming growth factor-beta-1 (TGF-β1); proinflammatory cytokines, including granulocyte-macrophage colony-stimulating factor (GM-CSF), IL-1α, IL-1β, IL-2, IL-12, and tumor necrosis factor alpha (TNF-α); toll-like receptor 4 (TLR4) signaling pathway proteins, including cluster of differentiation 14 (CD14) and TLR4. The primers used to amplify these genes are shown in **Table 4**. Relative expression was reported as a ratio of the expression of the target gene to that of beta-actin (β-actin), and data were expressed relative to the data recorded in basal diet-treated piglets. The relative expression ratio (R) of mRNA was calculated as R = 2−11Ct (sample – control), where −11Ct (sample – control) = (Cttarget gene – Ctβ−actin) sample – (Cttarget gene – Ctβ−actin) control. RT-qPCR was performed using a SYBR Green detection kit (Thermo Fisher Scientific, Waltham, MA, United States), and the conditions were as follows: 30 s denaturation at 94◦C, 30 s annealing at 60◦C, and 30 s extension at 72◦C for 40 cycles, followed by a melting curve program (60–99◦C with a heating rate of 0.1◦C/s and fluorescence measurement). A melting temperature (Tm) peak of 85 ± 0.8◦C was used to determine the specificity of amplification. T<sup>m</sup> values are reported as the mean of three replicates (Liu et al., 2016).

#### Statistical Analysis

Data were statistically analyzed by one-way ANOVA using SPSS 17.0 software (SPSS, Inc., Chicago, IL, United States). Data are presented as means ± SEM, and P-values < 0.05 indicate statistical significance.

#### RESULTS

#### Colonic Microbiota in Piglets

The quantities of Bacteroidetes, Clostridium cluster IV, E. coli, Firmicutes, Lactobacillus, and total bacteria in the colon contents did not change significantly (P > 0.05) after supplementation TABLE 4 | Primers used for quantitative reverse transcription PCR.


GM-CSF, granulocyte-macrophage colony-stimulating factor; IL, interleukin; TGFβ1, transforming growth factor-beta-1; TLR, toll-like receptor; TNFα, tumor necrosis factor alpha; ZO-1, zonula occludens-1.

with CHR or fermented CHR. The quantity of Clostridium cluster XIVa decreased (P < 0.05) in the positive control group compared with the other three groups (**Table 5**). The quantities of Bifidobacterium spp. in the positive control and CHR groups were significantly lower (P < 0.05) than in the blank control group.

#### Concentrations of Bacterial Metabolites in the Colons of Piglets

As shown in **Table 6**, there were no significant differences (P > 0.05) between colonic luminal concentrations of ammonia and skatoles in the four treatment groups. Dietary supplementation with fermented CHR significantly decreased (P < 0.05) the concentration of indoles in the colonic contents compared with that in the positive control group. When compared with the CHR group, the results indicated that fermented CHR increased (P < 0.05) the colonic luminal concentrations of straight-chain fatty acids (including acetate, propionate, butyrate, and valerate) and branched-chain fatty acids (including isobutyrate and isovalerate).


TABLE 5 | Effects of dietary supplementation with Chinese herbal residues (CHR) or fermented CHR on colonic microbiota quantities in weaned piglets (n = 5; lg copies/g).

Data in the same row with different superscripts differ significantly (P < 0.05).

TABLE 6 | Effects of dietary supplementation with Chinese herbal residues (CHR) or fermented CHR on bacterial metabolite concentrations in the colonic content in weaned piglets (n = 5; mg/g).


Data in the same row with different superscripts differ significantly (P < 0.05). Straight-chain fatty acids, including acetate, propionate, butyrate, and valerate; Branchedchain fatty acids, including isobutyrate and isovalerate.

#### Levels of mRNA Corresponding to Epithelial Cell Proteins, Cytokines, and Proteins of the TLR4 Signaling Pathway in the Colons of Piglets

As shown in **Table 7**, dietary supplementation with CHR increased (P < 0.05) the mRNA level of E-cadherin compared with that in the other three groups and increased ZO-1 relative to that in the blank control group. Dietary supplementation with CHR or fermented CHR increased (P < 0.05) the mRNA levels of IL-1α, IL-2, and TNF-α, while decreasing (P < 0.05) the level of IL-12 compared with that in the blank control group. Piglets fed a diet supplemented with CHR exhibited higher (P < 0.05) levels of IL-4 and TGF-β1, but a lower (P < 0.05) level of GM-CSF was observed compared with those in the fermented CHR group. The level of CD14 mRNA was significantly higher (P < 0.05) in the positive control group than in the other three groups.

#### DISCUSSION

Gut microbiota are the resident microorganisms in the digestive tracts of animals and humans, which affect nutrient digestion and the bioconversion of food compounds in the host organisms (Seo et al., 2015). Recently, the intestinal microbiota composition and metabolic activities have emerged as important parameters affecting either positively or negatively "intestinal health" (Blachier et al., 2017). Components such as microbiota composition and diversity and bacterial metabolite concentrations can affect the epithelial integrity, barrier function, immunity, and enteroendocrine peptides. It is known that the intestinal microbiota has genomic characteristics that allow it to use "providential" undigested nutrients in feedstuff. In return, it benefits the host metabolism by providing energy through the production of metabolites that can be utilized and absorbed by colonic cells (Cani and Delzenne, 2007). The intestinal microbiota and associated metabolites affect positively or negatively intestinal barrier permeability, depending on their chemical structure and concentrations, as well as activate immune cells and produce pro- or anti-inflammatory molecules in the gut by cell-associated pattern recognition receptors that recognize molecules unique to the microbiota components or its metabolites (Macpherson and Harris, 2004). Indeed, changes in the luminal environment of the intestinal epithelial cells, following notable modifications in dietary intake, can negatively or positively affect the homeostatic process of colonic epithelia renewal and barrier function (Blachier et al., 2017).

Firmicutes and Bacteroidetes are regarded as the main microbiota phyla in the pig gut, representing approximately 90% of all phylogenetic types (Guo et al., 2008). The present study


TABLE 7 | Effects of dietary supplementation with Chinese herbal residues (CHR) or fermented CHR on colon mRNA levels corresponding to epithelial cell proteins, cytokines, and TLR4 signaling pathway in weaned piglets (n = 5).

Data in the same row with different superscripts differ significantly (P < 0.05). GM-CSF, granulocyte-macrophage colony-stimulating factor; IL, interleukin; TGF-β1, transforming growth factor-beta-1; TLR, toll-like receptor; TNFα, tumor necrosis factor alpha; ZO-1, zonula occludens-1.

validated this report by showing quantities of Firmicutes and Bacteroidetes that were similar to the total bacterial quantity. Blais et al. (2015) reported that the most abundant members of Firmicutes, including Clostridium clusters IV and XIVa, along with Lactobacillus and Bifidobacterium, appear beneficial for host health. Although microbiota belonging to Clostridium clusters IV and XIVa are more abundant than Lactobacillus and Bifidobacterium, they have received far less attention. Moreover, current knowledge of the mucosa-associated microbial communities in the colon is limited because studies have focused on the characterization of fecal diversity (Decroos et al., 2006). Therefore, the present study determined the quantities of several important colonic microbiota groups based on real-time PCR. Although these data do not reflect the total bacterial community, our data show that the quantities of all tested microorganisms, especially Clostridium cluster XIVa and Bifidobacterium, were lower in the positive control group than in the blank control group. This is likely due to the addition of antibiotics and ZnO in the positive control group because these compounds affect the balance of intestinal microbiota. In addition, although solid-state anaerobic fermentation was inoculated with Lactobacillus and Bifidobacterium, there was no difference in the quantities of these two probiotics with regard to the colon luminal contents between the CHR and fermented CHR groups. It is thus tempting to speculate that fermentation methods explain this result, but further studies outside the scope of the present study are necessary to test such a hypothesis.

The gut microbiota benefits the host by breaking down undigestible or not fully digested nutrients, as well as regulating the immune, endocrine and mesenteric nervous systems (Guarner and Malagelada, 2003). It is well known that the SCFAs are major compounds produced by the microbial fermentation of fibers and resistant starch (Laparra and Sanz, 2009), and to a lower extent, by some amino acids. The large intestine is the main site of SCFA production and absorption in pigs. SCFAs, especially butyrate, are the main energy source for colonocytes (Perrin et al., 2001), and they presumably play an important role in colon epithelium renewal (Nicholson et al., 2012). Butyrate produced by Clostridium clusters IV and XIVa is not only used by colonocytes for energy production, but it also effects gene expression, in relationship with the control of colonocyte proliferation and differentiation (Carneiro et al., 2008). Importantly, it is worth to keep in mind that SCFA concentrations represent the net result of microbiota production and colon mucosa absorption.

The results of the present study suggest that dietary supplementation with CHR decreased the butyrate concentration, but it did not significantly modify the quantities of Clostridium clusters IV and XIVa. It is likely that in our study a large amount of butyrate was absorbed by colon mucosa, resulting in increased mRNA levels of E-cadherin and ZO-1 in colonic tissue. In addition, dietary fibers may be fermented by the gut microbiota to produce health-promoting SCFAs that are believed to modify intestinal barrier function and microbiota composition (Chen et al., 2017). Although numerous types of fibers are present in CHR (crude fiber, 2.64%; acid detergent fiber, 49.51%; and neutral detergent fiber, 40.50% [based on dry matter]), the SCFAs concentration in the colonic contents originating from the CHR group was the lowest. A possible mechanism is that non-digestible fibers reach the large intestine and provide a source of energy for colonocytes through the metabolic activity of the microbiota, which produces SCFAs. Colonocytes then absorb and partially metabolize SCFAs (Konstantinov et al., 2003). This interpretation needs further validation.

The gut microbiota can also catabolize nitrogenous compounds to putrefactive catabolites, such as ammonia, sulfide, bioamines, indoles, and phenols (Blachier et al., 2007). The ammonia concentration in the lumen of the large intestine is primarily attributed to microbiota amino acid deamination and urea hydrolysis (Warren and Newton, 1959), microbiota utilization of ammonia, and ammonia absorption through

epithelial cells (Eklou-Lawson et al., 2009; Andriamihaja et al., 2010). Ammonia in excess has been considered as a metabolic troublemaker because of its ability to inhibit mitochondrial oxygen consumption in a dose-dependent manner. In addition, a high concentration of ammonia can inhibit SCFA oxidation in colonic epithelial cells (Cremin et al., 2003). In the present study, there were no significant differences in the concentration of ammonia in the colonic contents recovered from the four treatment groups. This suggests that dietary supplementation with CHR or fermented CHR does not lead to increased protein fermentation in the colon, and thus is not associated with a marked increase of protein-derived amino acid deamination and/or urea hydrolysis in the colonic content. Compared with other amino acids, aromatic amino acids are metabolized and fermented slowly by anaerobes, including Bacteroides, Lactobacillus, Bifidobacterium, and Clostridium (Jensen et al., 1995). Tryptophan yields indoleacetate and indole, with the former subsequently yielding skatole (Smith and Macfarlane, 1997). Among these bacterial metabolites, indole has been shown to be beneficial for colonic epithelial barrier function by increasing epithelial cell tight junction (TJ) resistance (Le et al., 2005). From an environmental perspective, the odor resulting from indole and skatole emissions during intensive periods of swine production has a negative impact on meat quality and a serious nuisance to people living in close proximity (Le et al., 2005). In the present study, the fermented CHR significantly decreased the concentration of indole in the colonic content, and this could be the result of a decreased production of indole and/or enhanced absorption through the colonic epithelium. Nonetheless, this decrease is presumably detrimental to the colonic epithelium since this particular bacterial metabolite, as discussed above, was shown to contribute to the maintenance of colonic barrier function (Shimada et al., 2013). In contrast, the co-metabolite indoxyl sulfate is considered a uremic toxin and can activate the aryl hydrocarbon receptor (AhR)-mediated transcription of several enzymes belonging to the cytochrome P450 family, with associated negative effects. Additionally, the reduction in the amount of fecal indole released into the environment is potentially beneficial. Further studies, including the measurement of related polluting substances in fecal material, are necessary to fully validate the latter proposition (Zhou et al., 2014), although they are outside the scope of the present study.

The surface of the gut is protected by a layer of epithelial cells covered by mucous layers, which is in constant contact with an abundant population of microbes and their metabolites (Hartsock and Nelson, 2008). The intestinal barrier formed by the epithelial cells and junctional complex, consisting of TJs (including occludin, claudin, and ZO-1), adherens junctions (including E-Cadherin and catenins), gap junctions, and desmosomes, excludes the vast majority of these microbes and some of their metabolites from access to the subepithelial cells (Ukena et al., 2007). A previous study reported that microbiota can influence the integrity of the intestinal epithelium, mucosal immunity function, and host health (Cerf-Bensussan and Gaboriau-Routhiau, 2010), and this in turn can affect intestinal barrier function and, consequently, intestinal permeability. Moreover, several pathogens have been shown to modulate the epithelial TJs, and this is achieved via the production of proteins that engage signaling mechanisms in epithelial cells, modulate the actin cytoskeleton, or degrade TJ proteins (Su et al., 2011). Hence, modulation of epithelial cell protein function, particularly by increasing the mRNA levels of E-Cadherin, occludin, and ZO-1, is a target for novel therapeutics or prophylactic treatments against a range of diseases involving alterations of the epithelial barrier function (Yang et al., 2015). The mechanism for the increased expression of E-cadherin and ZO-1 after dietary CHR supplementation is unknown but may be related to increases in the mRNA levels of anti-inflammatory cytokines, including IL-4, IL-10, and TGF-β1, which in turn affects the mRNA levels of E-cadherin and ZO-1. In the present study, the ZO-1 mRNA level in the positive control group increased relative to the blank control group, and this is consistent with a previous finding that zinc supplementation reduced intestinal permeability by enhancing occluding and ZO-1 expression in weaning piglets (Zhang and Guo, 2009).

Cytokines are small peptides that play important roles in the regulation of immune and inflammatory responses (Pié et al., 2004). On one hand, proinflammatory cytokines (e.g., IL-1α, IL-2, and TNF-α) are necessary to initiate the inflammatory response during infection, but the overexpression of such cytokines may lead to pathological responses (Clark, 2007; Zelnickova et al., 2008). Therefore, regulating the expression levels of proinflammatory cytokines could potentially reduce intestinal mucosal inflammation (Liu et al., 2008; Wan et al., 2018a). On the other hand, anti-inflammatory cytokines (e.g., TGFβ1 and IL-4) are capable of inhibiting the overexpression of proinflammatory cytokines and other mediators that could lead to hyperactivation of the immune response in weaned piglets (Clark, 2007; Zelnickova et al., 2008). In addition, the intestinal immune system has been reported to be closely linked to the intestinal microbiota, to the TLR4 signaling pathway, and to the intestinal TJ barrier (Wu and Wu, 2012). For example, Clostridium clusters IV and XIVa are capable of promoting the induction of colonic regulatory T cells (Atarashi et al., 2011); and TLR4/CD14 signaling pathways are known to mediate an inflammatory response to microbial stimuli (Takeda and Akira, 2005). Furthermore, most proinflammatory cytokines (TNF-a and IL-1β) induce a pathological opening in the intestinal TJ barrier and increase intestinal epithelial permeability (Ai-Sadi et al., 2009). In the present study, the colons of piglets fed a diet supplemented with CHR exhibited higher levels of IL-1α, IL-2, TNF-α, and IL-4, but a lower level of IL-12. Finally, we proposed that the CHR-induced changes in microbiota composition and metabolic activity act directly on the epithelium or indirectly by activating immune cells that, in turn, affect epithelial function. Additional studies are required to define the complex relationships between these different components.

#### CONCLUSION

In the present study, we show that dietary supplementation with CHR or fermented CHR has no major adverse effects on the colon ecosystem in weaned piglets. Supplementation with CHR

modified the expression of genes related to TJ epithelial cell proteins, as well as several cytokines in the colon. The fermented CHR supplementation modified the expression of genes encoding several cytokines and decreased indole concentrations in the colonic content, this latter bacterial metabolites being involved in colon epithelial barrier function. However, it should be kept in mind that this latter compound is also associated with deleterious effects, notably as precursor of the uremic toxin indoxyl sulfate. Although our study demonstrates that diets supplemented with CHR or fermented CHR modify the colon ecosystem in terms of microbiota, bacterial metabolite composition, and gene expression, further study is needed to elucidate the overall effects of such dietary supplementation in terms of beneficial and deleterious effects on the colon and overall piglet health. Finally, future studies investigating the mechanisms of action by which CHR and fermented CHR impact the gut ecosystem should determine the possible causal links between the different parameters measured in our study.

#### AUTHOR CONTRIBUTIONS

JS, QZ, YZ, LH, and XK performed the experiments. JS, YY, FB, ZW, and XK wrote the manuscript. JS

#### REFERENCES


and XK performed the statistical analysis. JS, QZ, YZ, and LH fed the animals. All authors reviewed the manuscript.

# FUNDING

This study was jointly supported by the Changsha Key Project of Science and Technology Plan (kq1703007), Talent Projects of Guangxi Science and Technology Department (AD17195043), Hunan Provincial Key R&D Program (2016NK2208), and STS regional key project of the Chinese Academy of Sciences.

#### ACKNOWLEDGMENTS

We thank staffs and postgraduate students of Hunan Provincial Engineering Research Center of Healthy Livestock for collecting samples and technicians from Key Laboratory of Agro-ecological Processes in Subtropical Region for providing technical assistance. We also would like to thank Editage (http://online.editage.cn/) for English language editing.



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Su, Zhu, Zhao, Han, Yin, Blachier, Wang and Kong. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# *Bacillus amyloliquefaciens* Ameliorates Dextran Sulfate Sodium-Induced Colitis by Improving Gut Microbial Dysbiosis in Mice Model

Guangtian Cao<sup>1</sup> , Kangli Wang<sup>2</sup> , Zhanming Li <sup>3</sup> , Fei Tao<sup>1</sup> , Yinglei Xu<sup>2</sup> , Junhong Lan<sup>2</sup> , Guangyong Chen<sup>2</sup> and Caimei Yang2,4 \*

*<sup>1</sup> College of Standardization, China Jiliang University, Hangzhou, China, <sup>2</sup> Key Laboratory of Applied Technology on Green-Eco-Healthy Animal Husbandry of Zhejiang Province, College of Animal Science and Technology, Zhejiang A and F University, Hangzhou, China, <sup>3</sup> Department of Food Science, China Jiliang University, Hangzhou, China, <sup>4</sup> Zhejiang Provincial Engineering Laboratory for Animal Health Inspection and Internet Technology, College of Animal Science and Technology, Zhejiang A and F University, Hangzhou, China*

#### *Edited by:*

*Jie Yin, Institute of Subtropical Agriculture (CAS), China*

#### *Reviewed by:*

*Xinyan Han, Zhejiang University, China Yulan Liu, Wuhan Polytechnic University, China*

> *\*Correspondence: Caimei Yang yangcaimei2012@163.com*

#### *Specialty section:*

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 13 October 2018 Accepted: 14 December 2018 Published: 08 January 2019*

#### *Citation:*

*Cao G, Wang K, Li Z, Tao F, Xu Y, Lan J, Chen G and Yang C (2019) Bacillus amyloliquefaciens Ameliorates Dextran Sulfate Sodium-Induced Colitis by Improving Gut Microbial Dysbiosis in Mice Model. Front. Microbiol. 9:3260. doi: 10.3389/fmicb.2018.03260* Several *Bacillus* strains exert beneficial effects on the maintenance of intestinal homeostasis and host health. However, whether *Bacillus amyloliquefaciens* (BA) can improve gut microbial dysbiosis and ameliorate colitis is unknown. Therefore, we conducted the present study to investigate the effects of BA administration on intestinal morphology, inflammatory response, and colonic microbial composition in a mouse model of dextran sulfate sodium (DSS)-induced colitis. Results showed that BA administration significantly ameliorated body weight loss, decreased disease activity index, and improved colonic tissue morphology in DSS-treated mice. In addition, levels of immunoglobulins, as well as pro-inflammatory cytokines, were decreased after BA administration. Importantly, colonic microbiota profiling indicated a significant (*p* < 0.05) difference in beta-diversity between BA-administrated and DSS-treated mice, according to weighted principal coordinate analysis (PCoA) results. The relative abundance of the *Firmicutes* genus was increased, whereas that of *Bacteroidetes* was decreased by BA administration. Furthermore, phylogenetic investigation of communities by reconstruction of unobserved states (PICRUSt) analysis showed that the most significantly changed pathways between the four groups of mice were carbohydrate, lipid, and amino acid metabolism. In conclusion, our results showed that BA administration has beneficial effects on DSS-induced colitis, suggesting that this strategy might be useful for the treatment of dysbiosis during ulcerative colitis. Further, the changes in metabolism, especially amino acid metabolism, might contribute to the beneficial effects of BA on the amelioration of DSS-induced colitis.

Keywords: *Bacillus amyloliquefaciens*, colitis, inflammation, intestinal morphology, microbiota profiling

# INTRODUCTION

Inflammatory bowel disease (IBD) is a chronic inflammatory disease that mostly occurs in the rectal and colonic mucosa and even deeper layers of the intestinal wall. Ulcerative colitis (UC) and Crohn's disease (CD) are the two main forms of IBD, and are characterized by clinical symptoms such as weight loss, diarrhea, and rectal bleeding. Although evidence suggested that mucosal edema and altered innate and adaptive immune responses contribute to the development of these diseases (Munyaka et al., 2016), the precise mechanism of the IBD pathogenesis is still unknown.

Beyond inherent basic nutrition, oral probiotics also exert many beneficial effects on health. Several species of Bacillus can function as antibiotic alternatives and growth promoters as they improve the digestibility and absorption of nutrients in the intestines of pigs and birds (Hong et al., 2005; Cao et al., 2018). Especially, emerging evidence has suggested that Bacillus amyloliquefaciens (BA) is beneficial for the amelioration of diarrhea and inflammation (Li et al., 2018). One study also reported the beneficial effects of BA on IBD as it ameliorated the body weight loss of dextran sulfate sodium salt (DSS)-induced colitis animals, in addition to reducing the protein and mRNA levels of pro-inflammatory cytokines in colonic tissues (Hairul Islam et al., 2011). However, they did not report any related mechanisms including whether BA supplementation affected the composition of the gut microbiota.

IBD is characterized as functional dysbiosis in the intestine including dysbiosis of microbiota. As the most widely used experimental model of UC, DSS-induced colitis is associated with alterations to the gut microbiota. Therefore, we conducted the present study to investigate the effects of B. amyloliquefaciens (CGMCC 9384) on DSS-induced colitis mice and to determine whether it has any effects on gut microbiota dysbiosis.

# MATERIALS AND METHODS

#### Animal Experiments

Twenty-four C57/BL6 male mice aged 9 weeks and weighing 21–23 g were purchased from SLAC Laboratory Animal Central (Changsha, China). All mice were individually maintained under standard conditions (lighting cycle, 12 h/d; temperature, 22 ± 2 ◦C; relative humidity, 50 ± 5%). Animals were randomly assigned into four groups as follows: mice were untreated for 7 days and then orally gavaged PBS for 7 days (Control group); mice were supplemented with 3.5% DSS (w/v; molecular mass = 6,500–10,000 Da; Sigma-Aldrich, Shanghai, China) dissolved in fresh running water ad libitum for 7 days and then orally gavaged PBS for 7 days (DSS group); mice were supplemented with 3.5% DSS for 7 days and then orally gavaged B. amyloliquefaciens (1.0 × 10<sup>8</sup> CFU/kg in 200 µL of PBS/mouse/day) for 7 days (BaL (low-dose) group); mice were supplemented with 3.5% DSS for 7 days and then orally gavaged with B. amyloliquefaciens (1.0 × 10<sup>9</sup> CFU/kg in 200 µL of PBS/mouse/day) for 7 days (BaH (highdose) group). The probiotic B. amyloliquefaciens (CGMCC9384) was provided by Zhejiang Huijia Biological Technology Ltd., Anji, China. According to our previous study, after fermentation (37◦C, 48 h) and drying, the strain was granulated and used at a level of approximately 10<sup>10</sup> CFU/g (Cao et al., 2018). Diet was supplied by Research Diets Inc (New Brunswick, NJ, USA). At the end of the experiment, blood was obtained from the retroorbital sinus, centrifuged at 4◦C at 800 g for 10 min for serum collection. Then mice were euthanized by cervical dislocation for colon and colonic digesta sample collection. Body weight, colonic length, rectal bleeding, and stool consistency were recorded and scored according to standard protocols (Zhang et al., 2015; Wang

TABLE 1 | Scoring system for the disease activity index.


et al., 2017). The scoring system for disease activity index (DAI) is shown in **Table 1**. All procedures in the present study were approved by the Animal Welfare Committee of China Jiliang University and all procedures were carried out according to the rules established by the committee.

#### Histological Analyses

Colonic samples were fixed with 4% formaldehyde overnight, and then embedded in paraffin and cut into 8µm thick sections as previous study described (Zhou X. et al., 2017; Zhou et al., 2018b). Following this, the sections were stained with hematoxylin and eosin (HE). Additionally, colonic samples were fixed with 2.5% glutaraldehyde at 4◦C and embedded in Epon-Araldite resin. The ultrathin sections were stained with uranyl acetate and lead citrate and observed under a Zeiss 902 transmission electron microscope.

#### Assessment of Apoptosis

The TUNEL method was used to stain apoptotic cells using an in situ cell death detection kit (Roche, Shanghai, China). Nuclei was stained with DAPI mounting solution (Vector, Burlingame, CA, USA). Apoptotic cells were observed under a light microscope (Leica, Solms, Germany) and representative pictures were taken.

#### Measurement of Immunoglobulins (Ig), Inflammatory Cytokines, Myeloperoxidase (MPO), and Eosinophil Peroxidase (EPO) Concentrations

Commercially available kits from Cusabio Biotech Co., Ltd. (Wuhan, Hubei, China) were used to the determination of serum concentrations of IgG, IgA, and IgM, as well as the concentrations of IL-1β, TNF-α, IL-6, EPO, and MPO in the colon, which were performed in accordance with manufacturer's instructions.

#### 16S rRNA Gene Sequencing

DNA from the colonic contents was isolated using the MoBio Power soil DNA Isolation Kit (Mo Bio Laboratories, Carlsbad, CA, United States) according to the manufacturer's manual. The concentration and purity of DNA were tested by 1% agarose gel electrophoresis, and DNA was diluted with sterile water to 1 ng/µL, which was stored at −80◦C until sequencing (Yin et al., 2017, 2018). The V4 region of the colonic bacterial 16S rRNA genes was amplified using the specific primers 515F (5′ -GTGCCAGCMGCCGCGGTAA-3′ ) and 806R (5′ -GGACTACVSGGGTATCTAAT-3′ ). Purified PCR amplicons with a bright main band of 300 bp were sequenced with the Illumina MiSeq platform at Novogene (Beijing, China). After the filtering of low-quality genes, the chimeras in raw reads were removed using Cutadapt software (V1.9.1). The sequence database was built by the Ion Plus Fragment Library Kit 48 rxns (Thermo Fisher Scientific, US), and Ion S5TMXLplatform was used for further sequencing. The operational taxonomic units (OTUs) with a similarity threshold of 97% were selected with the UPARSE software package (v7.0.1001).

#### Western Blot Analysis

Western blot analysis was conducted as previously described (Zhou X. H. et al., 2017; Zhou et al., 2018a). Extracts from colonic tissue samples containing equal quantities of proteins (30 µg) were electrophoresed on a polyacrylamide gel. Next, the separated proteins were transferred onto a PVDF membrane. Firstly, the membrane was incubated overnight at 4◦C with antibodies against AMP-activated protein kinase (AMPK, rabbit monoclonal, diluted at 1:1,000), general amino acid control nonderepressible 2 (GCN2, rabbit polyclonal, diluted at 1:1,000), and mammalian target of rapamycin complex 1 (mTORC1, rabbit monoclonal, diluted at 1:1,000) (Cell Signaling, Beverly, MA, United States). Then, second antibody (diluted at 1:5,000; Cell Signaling) was incubated with the blots used for the incubation at 4◦C for 2 h. Protein bands were developed with EZ-ECL (Biological Industries, Cromwell, CT, United States) by the ChemiDoc MP system (BIO-RAD, Hercules, CA, USA). All protein measurements were normalized to GAPDH (1:1,000; Proteintech, Rosemont, IL, USA) and data are expressed relative to the values in control mice.

# Sequencing Data Analysis

Based on the results of all sample species annotations using Mothur software and the SILVA database (SSUrRNA), the top 10 phyla and the top 10 genera were analyzed (relative abundance over 1%). The alpha diversity (Shannon index) and the unweighted pair-group method with Arithmetic Means (UPGMA) clustering were investigated using QIIME software (version 1.7.0). The beta diversity was calculated from unweighted and weighted UniFrac distances, in which principal coordinate analysis (PCoA) and non-metric multidimensional scaling (NMDS) analysis were performed to analyze the colonic microbial community data using R software (Version 2.15.3). In addition, analysis of similarity was performed to test significant differences between sample groups based on unweighted UniFrac distance matrices. Finally, the taxonomic files were achieved from QIIME software, and Linear discriminant analysis with effect size (LEfSe) analysis and phylogenetic investigation of communities by reconstruction of unobserved states (PICRUSt) were performed online (http://huttenhower.sph.harvard.edu/ galaxy/).

# Statistical Analysis

All statistical analyses were performed by one-way ANOVA, using the general linear model and the MIXED procedure (PROC MIXED) from SAS software version 9.2 (SAS Institute Inc., Cary, NC, United States). Data are presented as least square means plus pooled SEM. P < 0.05 was considered statistically significant and 0.05 < P < 0.1 was considered as a tendency.

# RESULTS

#### Effects of *B. amyloliquefaciens* on Colitis Signs

The effects of BA on colitis signs in mice are shown in **Figure 1**. On day 14, both low and high levels of BA supplementation reversed DSS-induced feed intake and weight loss (**Figures 1A,B**). BA-supplemented mice exhibited increased colon lengths, colon weights, and colon length/colon weight ratios compared to DSS-treated mice (**Figures 1C–E**). Moreover, BA supplementation lowered the DAI from day 9 to day 14 (**Figure 1F**).

#### Effects of *B. amyloliquefaciens* on DSS-induced Histopathological Damage and Apoptosis

The effects of BA supplementation on DSS-induced histopathological damage and apoptosis are shown in **Figure 2**. H&E staining results indicated pathological damage in the colon tissue of DSS-treated mice, such as obvious edema in the structure of the mucous layer, whereas BA-supplemented mice showed no such changes. In addition, severe mitochondrial edema in the colon tissue of DSS-treated mice was observed according to TEM results. However, both low and high levels of supplementation alleviated these detrimental changes. Moreover, TUNEL staining indicated that DSS-induced mice had higher levels of apoptosis compared to both BA-supplemented mice and control mice.

#### Effects of *B. amyloliquefaciens* on Immunoglobulins, Inflammatory Cytokines, and Colonic Infiltration Markers

IgA, IgG, and IgM concentrations (**Figures 3A–C**) in serum were decreased, whereas IL-1β, IL-6, and TNF-α concentrations in colonic tissue were increased in DSS-treated mice (**Figures 3D–F**) when compared with control mice. However, BA supplementation significantly alleviated these DSSinduced changes. In addition, MPO and EPO concentrations were increased in the colonic tissue of DSS-treated mice (**Figures 3G,H**) and those were significantly alleviated by supplementing DSS-treated mice with BA.

#### Colonic Microbiota in Mice Is Altered After *B. amyloliquefaciens* Supplementation

The hypervariable V3 + V4 regions of 16S rRNA genes were sequenced from colonic content and an average of 81,351 ± 15,327 reads were obtained. An average of 329 ± 61 operational taxonomic units (OTUs) were obtained after sequencing with at least 97% similarity. First, we determined the alpha diversity of colonic microflora and then investigated the differences in QIIME. Surprisingly, there was no significant alterations in alpha diversity as reflected by the observed species among all the groups, except that the Shannon index

FIGURE 1 | Effects of *B. amyloliquefaciens* on Colitis Signs. (A) Average daily feed intake. (B) Body weight gain. (C) Colon length. (D) Colon weight. (E) Colon length/colon weight ratio. (F) Disease activity index. Control, mice were untreated for 7 days and then orally gavaged PBS for 7 days; DSS, mice were supplemented with 3.5% DSS for 7 days and then orally gavaged PBS for 7 days; BaL, mice were supplemented with 3.5% DSS for 7 days and then orally gavaged *B. amyloliquefaciens* (Ba, 9.0 × 10<sup>8</sup> CFU/kg in 200 µL of PBS/mouse/day) for 7 days; BaH, mice were supplemented with 3.5% DSS for 7 days and then orally gavaged *B. amyloliquefaciens* (Ba, 10.0 × 10<sup>8</sup> CFU/kg in 200 µL of PBS/mouse/day).Values are expressed as mean ± SEM, *n* = 6. a,bMeans of the bars with different letters were significantly different among groups (*P* < 0.05).

tend to change in DSS-treated mice (**Figure 4A**). However, a significant difference in beta-diversity among the four treatment groups were observed according to the weighted PCoA results (**Figure 4B**). To analyze the composition of bacterial communities, the top 10 phyla in terms of relative abundance among the colonic bacteria that presented in Control, DSS, BaL, and BaH groups were determined (**Figure 4C**). The dominant bacterial phyla in all samples collected from different mice were

Bacteroidetes, Firmicutes, Actinobacteria, and Verrucomicrobia, which accounted for 98.2–99.7% of OTUs in mice. Remarkably, Firmicutes was decreased in DSS-treated mice when compared to that in mice of the other three groups, whereas no significant differences were observed among Control, BaL, and BaH groups. A Venn diagram, displaying the overlapping OTU data for all treatments, showed that 254 OTUs were shared among the colonic samples from all groups, based on this, the numbers of unique OTUs in Control, Dss, BaL, and BaH were 36, 77, 25, and 26, respectively (**Figure 4D**).

According to non-metric multidimensional scaling (NMDS) analysis, bacterial communities had similar structures in BaL and BaH groups, whereas DSS samples were extremely separated from the other three groups (**Figure 4E**). To fully understand the effects of BA treatments on DSS-induced mice, a metagenomic biomarker discovery approach (LEfSe) was used (**Figure 4F**). In the present study, taxa with an average relative abundance over >0.0001 in LEfSe to retain meaningful taxa. Only Clostridiales was observed to be significantly less abundant in the DSS group when compared to that in the Control group. Moreover, Clostridiales was significantly overrepresented in BaH and BaL groups when compared to its relative abundance in the Control group.

#### Biofunction Predictions in Microbial Communities

To characterize the distinct functions of colonic microbiota, we performed PICRUSt analysis combined with the Kyoto Encyclopedia of Genes and Genomes (KEGG) database of microbial genomic information (**Figure 5**). This analysis permitted a comparison of the differences in the functional profiles among all groups and revealed pathways that were significantly different among Control, DSS, BaL, and BaH groups. The most significantly different KEGG pathway types were metabolism (predominantly amino acid, carbohydrate and lipid metabolism in level 2) pathways (**Figures 5A**,**B**). These pathways were further analyzed in KEGG level 3. The results showed that BA supplementation alleviated the increased activities of pyruvate, arginine, proline, glycine, serine, and threonine metabolism, as well as glycolysis/gluconeogenesis, and the citrate cycle, which were all induced by DSS treatment (**Figure 5C**). To further explore if metabolism in the small intestinal epithelium was changed in association with changes in the metabolic pathways of microbial communities, the protein expression of AMPK, GCN2, and mTORC1 was examined. We observed a significant increase in AMPK expression and a decrease in GCN2 and mTORC1 expression in BAsupplemented mice when compared to that in DSS-treated mice (**Figures 5D–G**).

# DISCUSSION

Bacillus amyloliquefaciens have been widely used in animal feed as an alternative to antibiotics and no side effects were have previously been reported (Cao et al., 2018). In the present study, a strain of B. amyloliquefaciens (CGMCC 9384) was used to treat mice with DSS-induced colitis. The results showed that BA administration significantly ameliorated loss of body weight and colon weight, and decreased DAI in DSS-treated mice. Impaired intestinal structure and intestinal dysfunction such as mucosal and mitochondrial edema were associated with a disturbance in the balance of enterocyte apoptosis and renewal (Li et al., 2018). TUNEL staining showed higher rates of apoptotic cells and histological observation showed impaired intestinal villi in the colonic tissue of DSS-treated mice. However, we did not observe a difference between control mice and BA mice, which indicated the beneficial effects of BA treatment on intestinal morphology.

Impaired intestinal structure were also associated with immune activation and increased inflammatory response (Deniz et al., 2004; Xu et al., 2009). A previous study reported that the supplementation with BA, which was isolated and characterized from northeast Himalayan soil, decreased levels of pro-inflammatory cytokines in DSS-treated mice (Hairul Islam et al., 2011). Similarly, our data also showed that serum IgA, IgM, and IgG concentrations were also normalized by BA administration. In addition, infiltration of neutrophils and macrophages into the mucosa was also previously found to result in intestinal dysfunction related to colitis (Deniz et al., 2004). MPO and EPO activity are major factors that reflect granulocyte infiltration into colonic tissues (Han et al., 2015; Zhang et al., 2015). In the present study, the decreased levels of MPO and EPO in BA-administrated mice, when compared to those in DSS-treated mice, indicated the beneficial effects of alleviating inflammatory responses caused by the granulocytes.

For the new therapies to treat functional bowel disorders, dietary modifications focused on the gut microenvironment, mainly aimed at modulating the gut microbiota, have been studied most extensively (Simrén et al., 2013; Simrén and Tack, 2018). Trials also suggested that changes to the gut

microbiota might play a role in the development of DSS-induced colitis (Nagalingam et al., 2011; Gkouskou et al., 2014; Jin et al., 2017). The administration of a mixture of probiotics including Lactobacillus and Bifidobacterium showed beneficial effects on colitis mouse models (Isaacs and Herfarth, 2008). Hence, dietary probiotics might alleviate colitis by modulating the intestinal microflora. In our present study, the Venn diagram showed the distinct OTUs appeared in the colonic microbiota in different groups. PCoA and NMDS analyses resulted in a clear separation between clusters of BA-treated and DSStreated mice, indicating that BA administration transformed the colonic bacterial composition. Additionally, PCoA plots based on weighted UniFrac distance and the LEFSe analysis of pair-wise comparisons among BaL and BaH groups came to the agreement that there were no obvious differences between the different doses of BA used in the present study. From these perspectives, BA administration modulated the imbalanced colonic flora of DSStreated mice. The Bacteroidetes to Firmicutes ratio is a reliable index to assess the composition of the gut microflora (Yelehe-Okouma et al., 2018). A shift to an increased Firmicutes to Bacteroidetes ratio was previously considered to be responsible for protection against inflammatory bowel disease (Cho and Blaser, 2012). Our present study found that Firmicutes was increased, whereas Bacteroidetes was decreased following BA administration, suggesting an increase in this ratio. These results indicated that BA might exert beneficial effects on the modulation of colonic microflora in DSS-treated mice.

PICRUSt analysis showed that metabolic pathways were most significantly differentially represented among all groups. Utilizing the KEGG database, marked increases in pathways of carbohydrate, lipid, and amino acid metabolism, as well as the metabolism of their secondary metabolites, were found in DSS-treated mice. Enhanced amino acid metabolism and the consumption of nutrients by microbial communities reduces the nutritional supply for intestinal cells. Meanwhile, the regeneration and proliferation of intestinal cells require a large amount of energy in the colon of DSS-treated mice. Hence, enterocyte renewal might be seriously affected in the colons of DSS-treated mice. However, this might not be a problem for mice administrated BA, as metabolism was not significantly different when compared to that in the control mice. GCN2 senses the absence of one or more amino acids, while the mTOR pathway could be activated by the presence of certain amino acids, such as leucine. In addition, AMPK, as the cellular energy sensor, plays a key role in regulating metabolism (Zhou et al., 2015). Consequently, the reduced expression of GCN2 and the enhanced expression of mTORC1 in the colonic tissue of mice administrated BA, as compared to levels in DSS-treated mice, suggested an increased supply of amino acids (Gallinetti et al., 2013). In addition, the expression of AMPK, the sensor of cellular energy, was increased, which indicated an increase in energy requirements for the enterocyte renewal.

In conclusion, our results showed that BA administration exerted beneficial effects on ameliorating weight loss, intestinal dysfunction, and inflammatory response during DSS-induced colitis in mice, suggesting that it could be useful for the treatment of dysbiosis during UC. Importantly, colonic microbiota composition in mice with colitis was significantly affected by BA administration. In addition, the biofunctional prediction of microbial communities indicated that changes of metabolism, and especially amino acid metabolism, might contribute to the

#### REFERENCES


beneficial effects of BA on the amelioration of DSS-induced colitis.

#### AUTHOR CONTRIBUTIONS

GCa and CY designed the trials, performed the experiments, and edited the manuscript. ZL and FT performed the samples detection. YX, JL, and GCh analyzed the data. GCa and CY wrote the manuscript, which was edited by ZL, FT, and KW. All authors read and approved the final manuscript.

#### FUNDING

The present study was supported by Key Research Project of Zhejiang Province (No. 2017C02005), talent project of zhejiang A & F university (No. 2034020001) and the National Natural Science Foundation of China (No. 31471636).

dysbiosis in mice. J. Basic Microbiol. 56, 986–998. doi: 10.1002/jobm.2015 00726


and epigenetically modulating the expression of glutathione synthesisrelated genes. Biochim. Biophys. Acta Mol. Basis Dis. 1864, 488–498. doi: 10.1016/j.bbadis.2017.11.009


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Cao, Wang, Li, Tao, Xu, Lan, Chen and Yang. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Fermentation Products of Paenibacillus bovis sp. nov. BD3526 Alleviates the Symptoms of Type 2 Diabetes Mellitus in GK Rats

Zhenyi Qiao<sup>1</sup>† , Jin Han1,2† , Huafeng Feng<sup>1</sup> , Huajun Zheng3,4, Jiang Wu<sup>1</sup> , Caixia Gao<sup>1</sup> , Meng Yang1,5, Chunping You<sup>1</sup> , Zhenmin Liu<sup>1</sup> and Zhengjun Wu<sup>1</sup> \*

<sup>1</sup> State Key Laboratory of Dairy Biotechnology, Shanghai Engineering Research Center of Dairy Biotechnology, Dairy Research Institute, Bright Dairy & Food Co., Ltd., Shanghai, China, <sup>2</sup> College of Food Science and Technology, Shanghai Ocean University, Shanghai, China, <sup>3</sup> Shanghai-MOST Key Laboratory of Health and Disease Genomics, Chinese National Human Genome Center at Shanghai, Shanghai, China, <sup>4</sup> Key Laboratory of Reproduction Regulation of NPFPC, Shanghai Institute of Planned Parenthood Research, IRD, Fudan University, Shanghai, China, <sup>5</sup> School of Life Sciences, Shanghai University, Shanghai, China

#### Edited by:

Yuheng Luo, Sichuan Agricultural University, China

#### Reviewed by:

Natalia Shulzhenko, Oregon State University, United States Vishal Singh, University of Toledo, United States

#### \*Correspondence:

Zhengjun Wu wuzhengjun@brightdairy.com †These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 14 August 2018 Accepted: 18 December 2018 Published: 09 January 2019

#### Citation:

Qiao Z, Han J, Feng H, Zheng H, Wu J, Gao C, Yang M, You C, Liu Z and Wu Z (2019) Fermentation Products of Paenibacillus bovis sp. nov. BD3526 Alleviates the Symptoms of Type 2 Diabetes Mellitus in GK Rats. Front. Microbiol. 9:3292. doi: 10.3389/fmicb.2018.03292 Gut microbiota is closely related to type 2 diabetes mellitus (T2DM). The gut microbiota of patients with T2DM is significantly different from that of healthy subjects in terms of bacterial composition and diversity. Here, we used the fermentation products of Paenibacillus bovis sp. nov. BD3526 to study the disease progression of T2DM in Goto-kakisaki (GK) rats. We found that the symptoms in GK rats fed the fermentation products of BD3526 were significantly improved. The 16S rRNA sequencing showed that the fermentation products of BD3526 had strong effects on the gut microbiota by increasing the content of Akkermansia. In addition, the interaction of the genus in the gut of the BD3526 group also significantly changed. Additional cytokine detection revealed that the fermentation products of BD3526 can reduce the inflammatory factors in the intestinal mucus of GK rats and thereby inhibit the inflammatory response and ameliorate the symptoms of T2DM.

Keywords: type 2 diabetes, gut microbiota, Paenibacillus bovis sp. nov. BD3526, Akkermansia muciniphila, cytokine

#### INTRODUCTION

According to data published by the International Diabetes Federation, there were 425 million people with diabetes worldwide in 2017. Type 2 diabetes mellitus (T2DM) is characterized by a sustained decrease in the insulin secretion of pancreatic β-cells, which leads to insufficient insulin to fulfill the requirement of the body. Long-term T2DM in the human body can cause serious complications, for example in the kinds of kidney disease, cancer and cardiovascular disease (Stern, 1995; Coughlin et al., 2004).

In the last two decades, evidence has accumulated that the pathogenesis of T2DM and its complications may be related to inflammation factors (i.e., IL-1β, IL-6, MCP-1, TNF-α and IFNγ) (Cuman et al., 2001; Rotter et al., 2002; Masters et al., 2010; Westwellroper et al., 2011; WestwellRoper et al., 2014; Greer et al., 2016). For example, IL-6 is the most endocrine cytokine and is not only produced by immunocompetent cells but also by fat cells involved in the body's inflammatory response and energy metabolism (Rotter et al., 2002). In patients with inflammation, the IL-6 levels are elevated. Excess IL-6 promotes pancreatic islet β lymphocyte differentiation and overexpresses the IgG gene, which promotes excessive activation of T lymphocytes and thereby

causes destruction and death of islet β cells (André-Schmutz et al., 2010; Donath, 2013). Therefore, the therapeutic targets for T2DM have shifted from simple hypoglycemic drugs (e.g., insulin and acarbose) to drugs suppressing inflammatory factors (such as IL-1 receptors blockade Anakinra and IKKβ-NF-κB inhibition Salsalate) (Donath, 2013).

The gut microbiota plays an essential role in the development of T2DM by skewing the host immune response to the inflammatory reaction (Karlsson et al., 2013). Wu et al. (2017) found that clinical patients with T2DM had significant changes in the composition and diversity of the gut microbiota after taking metformin being beneficial to improve the symptoms of T2DM. Some sources claim that Akkermansia muciniphila in the feces is significantly enriched in patients and mice with T2DM after taking metformin (Karlsson et al., 2013; Lee and Ko, 2014; Shin et al., 2014; Zhang et al., 2015; de la Cuesta-Zuluaga et al., 2017; Forslund et al., 2017). Low levels of A. muciniphila in the intestine may correlate with the thinning of the mucosal layer, which leads to a weakened intestinal barrier function (Derrien et al., 2004). In addition, A. muciniphila was reduced in patients with obesity and T2DM (Derrien et al., 2004; Everard et al., 2013). Further out, a recent study also claims that A. muciniphila can improve the response rate of tumor patients to PD-1/PD-L1 immunotherapy (Routy et al., 2018).

We have since isolated a novel bacterium designated as Paenibacillus bovis sp. nov. BD3526 from Tibetan yak milk in a previous work (Hang et al., 2016). This strain can synthesize exopolysaccharides with immunomodulatory activity in vitro (Xu et al., 2016). Instead, attempts to identify the relationship between exopolysaccharides and T2DM are not yet resolved. In this work, the lyophilized supernatant of the fermented milk by the BD3526 strain was fed to GK rats and its effect on the diabetic phenotype of the rats was observed. It was found that the symptoms of diabetes in the rats fed the supernatant were significantly improved. The content of A. muciniphila in the intestines was significantly increased, which coincided with a decrease in the inflammatory response in the intestine.

# MATERIALS AND METHODS

#### Bacterial Strain and Cultivation

Paenibacillus bovis sp. nov BD3526 (CGMCC 8333 = DSM28815 = ATCC BAA-2746) was provided by the State Key Laboratory of Dairy Biotechnology, Shanghai 200436, China. The bacterial strain was routinely cultivated aerobically on milk agar at 30◦C for 24 h. The medium was prepared by adding 10 mL sterile 10% (w/w) reconstituted skim milk to 100 mL melted 1.5% (v/w) agar solution. The strain was stored in sterile 10% (w/w) reconstituted skim milk supplement with 10% (v/v) glycerol at −80◦C.

# Preparation of the BD3526 Strain Fermentation Products

A loop of freshly cultivated BD3526 on milk agar was inoculated into a 100-mL flask containing 20 mL sterile 10% (w/w) reconstituted skim milk and cultivated at 30◦C at 180 rpm for 24 h. The culture was then transferred to a 250-mL flask containing 50 mL sterile 10% (w/w) reconstituted milk at a ratio of 4% (v/v) and cultivated at the same conditions as mentioned above for 72 h. Samples at different intervals were boiled for 5 min and then centrifuged at 12000 × g at 4◦C for 20 min. The supernatant was neutralized with 1 M NaOH to pH 6.8 and assayed for the inhibitory activity to alpha glucosidase (EC 3.2.1.20) utilizing p-nitrophenyl a-D-glucopyranoside (PNPG) as the substrate. The 72 h culture with an inhibitory activity to alpha glucosidase of approximately 60% (data unpublished) was boiled and centrifuged at the same condition mentioned above. The supernatant was lyophilized under a vacuum. The lyophilized powder was then cold stored. Before gavage of the experimental animals, the powder was redissolved in distilled water at a concentration of 50 mg/mL.

#### Animal Experiments

A total of 10 twelve-week-old Goto-kakisaki (GK) rats were adopted for 1 week and then randomly divided into two groups. The rats in the BD3526 group were gavaged daily with 2 mL 50 mg/mL lyophilized fermentation products of BD3526, whereas the rats in the control group were gavaged with 2 ml 50 mg/mL skim milk powder. The animals were individually caged with free access to a normal chewing bar and drinking water. On the day of the assay of postprandial blood glucose, the animals were first fed a normal chewing bar for 1 h and then gavage was conducted. After the gavage, the chewing bars were removed from the cage and a blood sample from the tail was collected 2 h after the gavage and analyzed with commercial blood glucose test strips (Sannuo, Shenzhen, China).

The postprandial blood glucose and body weight were measured weekly and fecal samples were collected. Glycated hemoglobin was measured at week 5. From the 6th week and thereafter, the gavage was terminated and the animals were restored to normal management. Postprandial blood glucose and body weight was continuously measured (weekly) and fecal samples were collected until the 9th week. After the 9th week, the intestinal mucus of the rats was collected for detection of the immunological factor. Feeding of the rates was conducted at Shanghai Slac Experimental Animal Co., Ltd.

# DNA Extraction and PCR Amplification

Microbial DNA was extracted from the fecal samples using the E.Z.N.A. <sup>R</sup> stool DNA Kit (Omega Biotek, Norcross, GA, United States), according to the manufacturer's protocols. The final DNA concentration and purification were determined by a NanoDrop 2000 UV-vis spectrophotometer (Thermo Scientific, Wilmington, DE, United States), and the DNA quality was checked by 1% agarose gel electrophoresis. The V3-V4 hypervariable regions of the bacteria 16S rRNA gene were amplified by a thermocycler PCR system (GeneAmp 9700, ABI, United States). The PCRs were performed in triplicate with the 20-µL mixture containing 4 µL 5 × FastPfu Buffer, 2 µL 2.5 mM dNTPs, 0.8 µL of each primer (5 µM), 0.4 µL FastPfu Polymerase and 10 ng template DNA. The resulting PCR products were extracted from a 2% agarose gel, further purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, United States) and quantified using QuantiFluor TM-ST (Promega, United States), according to the manufacturer's protocol.

#### High-Throughput Sequencing

fmicb-09-03292 January 4, 2019 Time: 15:28 # 3

The purified amplicons were pooled in equimolar and pairedend sequenced (2 × 300 bp) on an Illumina MiSeq platform (Illumina, San Diego, CA, United States), according to the standard protocols by Sinotech Genome Technology Co., Ltd. (Shanghai, China).

#### Bioinformatic Analysis and Statistical Analysis

In the bioinformatics analysis of 16S rRNA sequencing samples, we used an online cloud platform from Sinotech Genome Technology Co., Ltd.

Specifically, we used Usearch (version 7.0)<sup>1</sup> to perform cluster analysis on the OTUs. The diversity of the BD3526 group and the control group was analyzed using mother software (version v.1.30.1)<sup>2</sup> . Furthermore, we also used the PLS-DA analysis in the R language mixOmics package and created a distance map for each sample.

In the comparison of different strains, we used metagenome seq derived from the R language package to perform the Zero-inflated Gaussian distribution to process the impact of sequencing depth, and we finally found a difference based on the linear model.

In the random forest analysis, we used the Random Forest package and the plotROC package to quickly and efficiently select the species category that is most important for sample classification and to perform relevant ROC verification.

We used Networkx software<sup>3</sup> in the collinear network analysis and correlation network analysis.

The 16S function prediction is used to standardize the OTU abundance table by PICRUSt (the PICRUSt software stores the KO information corresponding to the greengene id), that is, to remove the influence of the number of copies of the 16S marker gene in the species genome the. We then pass the greengene id corresponding to each out, obtain the KEGG Ortholog (KO) information corresponding to the out, and calculate the KO abundance. According to the KEGG database information, the KO, Pathway, and EC information can be obtained, and the abundance of each functional category can be calculated based on OTU abundance. In addition, for the Pathway, PICRUSt can be used to obtain three levels of information on the metabolic pathways and to obtain abundance tables for each level.

In addition, we also used Graphpad Prism6 software to create statistical images of the statistical data.

#### In vitro Gut Simulator

The basic protocols were referred to the published article (Wu et al., 2017). Briefly, 2% fecal samples were gathered and cultured in 5 mL BHI broth for 3 h at 37◦C and the 2% preculture was

<sup>1</sup>http://drive5.com/uparse/

<sup>3</sup>http://networkx.github.io

seeded into the feed medium. The medium was composed by: (in g/liter) arabinogalactan (1.0), pectin (2.0), xylose (1.5), starch (3.0), glucose (0.4), yeast extract (3.0), peptone (1.0), mucin type II (4.0), and cysteine (0.5). The medium was acidified to around pH 2 with 6-M HCl to simulate digestion processes, and neutralized with simulated pancreatic juice to a pH of around 6.9. The simulated pancreatic juice contained: (in g/liter) NaHCO<sup>3</sup> (12.5), Oxgall bile salts (6.0), and pancreatin (0.9). Unless the experiment performed for 3 days, all the cultures were under the condition of anaerobic.

#### Cytokine Detection

The protocol of cytokine detection can be referred to from the instructions of a rat Cytokine Antibody Array Kit (Abcam, ab133992). Briefly, the membrane was incubated with blocking buffer. Then, 1 mL large intestine mucus was added to bind the antibodies that are fixed at the surface of the membrane. Following this step, the membrane was washed with washing buffer five times and then incubated with 1X biotin-conjugated anti-cytokines and 1X HRP-conjugated streptavidin. Finally, the membrane was exposed by X-ray film.

#### Accession Numbers

The Paenibacillus bovis sp. nov. BD3526 mentioned in this article can be found in the ATCC database. The ATCC number of Paenibacillus bovis sp. nov. BD3526 is BAA-2746. Highthroughput sequencing data from this study was deposited in the Sequence Read Archive (SRA) databases under the following accession number: SRP151163 and PRJNA508215.

#### Ethics Statement

The project use and care of the animals in this research was reviewed and approved by the Shanghai Laboratory Animal Management Office (SYXK [Shanghai] 2017-0008).

The animals used in the research were utilized based on appropriate experimental procedures. All of the animals were lawfully acquired, and their retention and use were in compliance with federal, state and local laws and regulations in every case and in accordance with the Institutional Animal Care and Use Committee of SLAC (IACUC) Guide for Care and Use of Laboratory Animals.

Animals used in this research received every consideration for their comfort and were properly housed, fed, and their surroundings kept in a sanitary condition.

The use of animals was in accordance with the IACUC Guide for Care and Use of Laboratory Animals. A minimal number of mice were used during the experiments. Appropriate anesthetics were used to eliminate sensibility to pain during all of the surgical procedures.

#### RESULTS

#### Diabetic Symptoms Were Alleviated in the BD3526 Group

In our previous work, we found that the BD3526 strain could synthesize a large amount of exopolysaccharides (36.25 g/L)

<sup>2</sup>http://www.mothur.org/wiki/Schloss\_SOP#Alpha\_diversity

with in vitro immunomodulatory activity (Xu et al., 2016), which might also play roles in retarding the development of diabetes in vivo. Besides EPS, monosaccharides, for example in the shapes of fructose, could also be identified in the fermentation products of BD3526. Therefore, to observe the effect of the BD3526 strain fermentation products in skim milk on blood glucose, we selected ten GK rats as subjects. GK rats are a commonly used model of spontaneous nonobese type 2 diabetes with mildly elevated fasting blood glucose, elevated blood glucose after eating, and stable glucose-stimulated insulin secretion disorders and glucose intolerance (**Figure 1A**). In addition, it has similar changes to human microvascular complications of type 2 diabetes. Its phenotypes are approaching stability when closing to adulthood. The results demonstrated that postprandial blood glucose in the BD3526 group showed a significant decrease in the fourth and fifth week (∗P-value < 0.05, mean ± SEM) (**Figure 1B**). The lowest point of postprandial blood glucose in the BD3526 group appeared in the fifth week, and the

FIGURE 1 | Observation of type 2 diabetes mellitus symptoms. (A) A total of 10, 13-week-old GK rats were used in this work and randomly divided into two groups. The rats in the BD3526 group were gavaged daily with 2 mL lyophilized powder of the BD3526 strain fermentation products (50 mg/mL), whereas the GK rats in the control group were gavaged daily with 2 mL physiological saline. All of the animals were permitted free access to normal chewing bars and water. At intervals of 0, 2, 3, and 6 weeks, the changes of diabetes indexes in these rats were observed. At the sixth week, the gavage of either BD3526 strain fermentation products or physiological saline was interrupted and the rats were restored to normal diets. After the restoration, the animals were anesthetized and killed, and the mucus in the intestine and colon was scratched. It is desirable to observe changes in the phenotype after stopping the ingestion of the BD3526 strain fermentation products. (B) Postprandial blood glucose measurements were performed in the BD3526 group and the control group. The x-axis represents the time of the week. The y-axis represents the postprandial blood glucose concentration (mM/L). The asterisk represents a significant difference in blood glucose concentration between the BD3526 group and the control group (∗P-value < 0.05, mean ± SEM). (C) The glycated hemoglobin (HbA1c) test confirmed that the blood glucose concentration in the BD3526 group was significantly lower than that in the control group (∗∗P-value < 0.01, mean ± SEM). (D) Body weight indicators were used to assess weekly body weight changes in the BD3526 group and control group models (mean ± SEM).

blood glucose concentration was 13.12 mM/L. Correspondingly, the postprandial blood glucose concentration of the control group of GK rats was 17.32 mM/L in the fifth week. In addition, the glycosylated hemoglobin index of the BD3526 group was also significantly lower than that of the control group ( ∗∗P-value < 0.01, mean ± SEM) (**Figure 1C**). The average concentration of glycated hemoglobin in the BD3526 group was 199.58 nM/L, whereas the value in the control group was 231.25 nM/L. This finding suggests that the diabetic symptoms of the BD3526 group were significantly alleviated. Furthermore, in the body mass index test, we subtracted the body weight of the control group rats from that of the BD3526 group and observed that the weight gain rate of GK rats in the BD3526 group was significantly lower than that in the control group (**Figure 1D**). These results indicate that the BD3526 strain fermentation products had the ability to reduce diabetes-related indicators in GK rats of T2DM.

#### The Diversity of Gut Microbes in GK Rats Fed With the Fermentation Products of the BD3526 Strain Increased

In addition to weekly testing of physiological and biochemical indicators of GK rats in the BD3526 group and the control group, fecal samples were also collected at weeks 0, 2, 3, and 6 for 16S rRNA sequencing. It is hoped that the effect on the gut microbes of GK rats can be observed. Among them, week zero and the sixth week were the standards of the initial and washout value of the GK rat fecal microbiota in the BD3526 group, respectively, and the second week and the third week represented the microbiota affected by the intake of the BD3526 fermentation products.

The microbiota in the fecal samples of either the BD3526 group or the control were assayed by 16S rRNA sequencing performed on the Illumina Miseq platform. After sequencing, the biodiversity of the gut microbes of the two groups of GK rats were analyzed. At weeks 2 and 3, Pan/Core analysis showed that in the BD3526 group, the total number of OTUs was significantly higher than that in the control group (**Figure 2A**). Correspondingly, the number of OTUs shared in both the control group and the BD3526 group showed a significant decrease (**Figure 2B**). The Shannon curve also showed that the number of OTUs in the BD3526 group and that in the control group were saturated with the increase in sequencing reads, and the number of OTUs in the BD3526 group was greater than that in the control group (**Figure 2D**). These results indicate that the gut microbiota diversity of GK rats in the BD3526 group was significantly higher than that in control GK rats. This result was also confirmed by alpha diversity verification (∗P < 0.05, mean ± SEM) (**Figure 2C**).

To perform a difference analysis within and between groups, a PLS-DA (Partial Least Squares Discriminant Analysis) algorithm was chosen to compare the two groups of GK for week 0, week 2, week 3, and week 6. As shown in **Figure 2E**, no significant difference between the BD3526 group and the control group was observed. Instead, with the administration of the BD3526 fermentation products, the gut microbiota was diversified, and at the third week, the difference between the BD3526 group and the control group was the most significant. Nevertheless, at the 6th week and after termination of administration of the BD3526 fermentation products for 1 week, the difference in the gut microbiota between the two groups was partially restored.

# Akkermansia Is Greatly Enriched in the Gut of GK Rats Fed BD3526 Fermentation Products

The 16S rRNA sequencing is mainly concerned with differences in gut microbiota composition between groups. Therefore, we used the metagenomeseq algorithm to analyze the differences in gut microbiota composition between BD3526 GK rats and control GK rats at weeks 2 and 3. Based on the statistics of the counts of each group of sequencing samples, we found that a total of 23 genera changed in the BD3526 group (P < 0.05). Among them, Akkermansia, Ruminococcaceae\_NK4A214, Ruminclostridum\_1 and No rank\_Peptococcaceae were elevated in the BD3526 group (**Figure 3A** and **Supplementary Table S1**). A statistical analysis of the FDR-corrected P-values of these 23 genera revealed that Akkermansia demonstrated the most significant difference among all of the changed genera [−10 Lg(P-value) = 2.848796] (**Figure 3B**). In the BD3526 group, Akkermansia, Ruminococcaceae\_NK4A214, Ruminiclostridium\_1 and Lachnospiraceae\_ND3007 were enriched, while the four genera were not detected in the other group. Instead, Alkaliphilus, Sulfurimonas, Amphritea, Photobacterium, Arenibacter, Anaerolineaceae, Flavobacteriaceae, Sulfurovum, Colwellia, Clostridium sensu, Magnetovibrio, Thiotrichaceae, and Rhodobacteraceae disappeared in the BD3526 group. To be precise, after correcting Akkermansia to species, we found that it corresponds to A. muciniphila.

However, the metagenomeseq algorithm itself has certain limitations. To eliminate the ostensible impact of gut microbiota diversity caused by the genus due to its low abundance or low differential fold and efficiently identify the most important OTUs for the BD3526 group, we also constructed a model of random forest distribution (**Figure 3C**). In this model, the gut microbiota in the BD3526 group and the control group were enriched in two different sites. We found that Akkermansia had the most important position in this model based on the genus abundance of random forest distribution (**Figure 3D** and **Supplementary Table S2**). This suggests that Akkermansia may be the biggest influencing factor for the difference between the BD3526 group and the control group. In addition to Akkermansia, No rank\_Lachnospiraceae, Prevotella\_9, Ruminiclostridium, Lachnospiraceae\_UCG\_001 and Candidatus\_Saccharimonas were also among the top five genera. Among them, Akkermansia and Prevotella\_9 were also the two genera screened by the metagenomeseq algorithm that showed a significant difference between the two rat groups.

In the random forest distribution analysis, we also used the top 6 genera of the mean reduction accuracy (Akkermansia, No rank\_Lachnospiraceae, Prevotella\_9, Ruminiclostridium, Lachnospiraceae\_UCG\_001 and Candidatus\_Saccharimonas) as models to perform the receiver operating characteristic curve

analysis (ROC curve) (**Supplementary Figure S1**). The results show that in the model containing only these six genera, the Receiver Operating Characteristic Curve (ROC) reached 0.81. Correspondingly, in the model containing all genera, the ROC is 0.48. When the six genera were removed from the model of all genera, the ROC was 0.51. This illustrated that our model was accurate to assess the difference between the two groups.

Interestingly, among the six genera, Akkermansia displayed the most significant difference in 16S rRNA sequencing, which suggests that Akkermansia may play an important role in ameliorating the symptoms of the GK rats fed BD3526 fermentation products. Although short-chain fatty acids (SCFAs) had been reported as key effectors in T2DM by other researchers (Zhao et al., 2018), no enrichment of SCFAs-producing microbiota was observed in the BD3526 group.

For further confirm the direct relationship between fermentation products and Akkermansia, we re-cultured the fecal samples in vitro which treated with 5% fermentation products or 5% skim milk. In this in vitro experiment, we focused Akkermansia was significantly enriched after treated with fermentation products for 1 day (**Supplementary Figure S2**). This phenomenon was consistent to the conclusions we claimed before that Akkermansia is greatly enriched in the gut of GK rats fed BD3526 fermentation products.

# The BD3526 Strain Fermentation Products Shifted the Interactions of the Gut Microbes

In the gut, microbe interactions between different genera and between different species often occurred. Therefore, to further evaluated the effect of the BD3526 strain fermentation products on the gut microbiota of GK rats, we used Networkx software for interaction network analysis. The interaction between microorganisms of different genera in the same sample and species correlation between different samples were evaluated. In the species interaction network of the BD3526 group and the control group, we found that the dominant species belong to Firmicutes. At the phylum level, there was no significant difference between the two groups (**Figures 4A,B**). However, when we constructed the node centrality diagram at the genus level, we found that the

random forest distribution. (D) Rapid selection of species categories that are most important for sample classification through random forest distribution. The highest mean decrease accuracy of the genus represented the greatest impact on the BD3526 and control components.

interaction network of the BD3526 group and the control group were significantly different. In the node centrality diagram, the number on the x-axis represents a different node. Each node corresponds to the interaction of the individual genus to the other genera. When a peak simultaneously occurred on the three curves of the degree centrality, closeness centrality and between centrality, it was considered that the corresponding genus of the node might be of great significance to the entire network. According to this principle,

9 genera were identified in the BD3526 group including No rank\_Ruminococcaceae, Butyrivibrio, Prevotellaceae\_NK3B31, Parasutterella and Lactobacillus, Ruminococcaceae\_UCG-014, Ruminococcaceae\_UCG-013, Treponema\_2 and Ruminiclostridium. In the control group, only five genera were identified including Unclassified\_Veillonellaceae, Sulfurovum, Oscillibacter, No rank\_Gastranaerophilales, and [Ruminococcus]\_gauvreauii\_group (**Figures 4C,D**). There were no associations between the 9 genera of the BD3526 group

curves reach the peak at the same time may be the genus that plays an important role in the interaction network. Detailed genus name and parameter information are listed in the following table. The black triangles represent the important genus in this network.

and the 5 genera of the control group. This result suggests that the fermentation products of the strain BD3526 may exert an important impact on the gut microbiota interaction network of GK rats. This may be another important factor to improve the symptoms of T2DM in the BD3526 group.

At the same time, the species correlation revealed that the rats in the BD3526 strain group shared a portion of the species with those of the control group. However, the points corresponding to the BD3526 group alone were significantly higher than those of the control group alone (**Supplementary Figure S3**), which indicated that the species in the gut microbiota of GK rats in the BD3526 group demonstrated more specificity.

# The BD3526 Strain Fermentation Products Decreased the Genera in the Gut Microbiota That Were Closely Related to Diseases in GK Rats

To assess the effect of different microbes on the stability of the gut microbiota in the BD3526 group, we obtained 16S rRNA sequencing data through the corresponding Greengene ID of each OTU. The functional composition profiles were predicted with PICRUSt. The results show that in the BD3526 group, the most significant changes in KEGG (level 2) categories were focused on five pathways including immune

system diseases [−10 Lg (P-value) = 17.798790], cancer [−10 Lg (P-value) = 16.241620], cell growth and death [−10 Lg (P-value) = 15.919890], translation [−10 Lg (Pvalue) = 16.241620] and infectious diseases [−10 Lg (Pvalue) = 15.785920] (**Figure 5**). These five KEGG pathways were all associated with diseases including immune disease, cancer and infectious diseases. This suggests that the intake of the BD3526 strain fermentation products may have a profound effect on the development of diseases and maintain the balance of the gut microbiota of GK rats.

#### The BD3526 Strain Fermentation Products Lowered the Intestinal Mucosal Inflammation Response in GK Rats

The causes of T2DM are often complicated. Studies have reported that the onset of T2DM was correlated with a chronically lowgrade inflammatory response (Donath et al., 2003; Ehses et al., 2009; Masters et al., 2010; Westwellroper et al., 2011; Lerner et al., 2012; Oslowski et al., 2012; Jourdan et al., 2013; WestwellRoper et al., 2014). The inflammatory response plays an important role in the occurrence and development of diabetes. Another study reported that A. muciniphila can increase intestinal mucus thickness and reduce the inflammation response (Wu et al., 2017).

As aforementioned, we found that the BD3526 strain fermentation products significantly increased A. muciniphila in GK rats and reduced the genera in the gut microbiota closely related to diseases by the KEGG analysis. Therefore,

we postulated that the BD3526 strain fermentation products could reduce the inflammatory factors of intestinal mucosal by increasing the A. muciniphila content and ultimately play a role in lowering blood glucose. Assuming this postulation, we extended our work to the inflammatory factors in the intestinal mucosa of the BD3526 group and the control group by using a rat Cytokine Antibody Array Kit (Abcam, ab133992). After a period of 5 weeks of washout, the cytokines expressed in the large intestinal mucus of the rats in the BD3526 group and those in the control group were assayed. The kit can simultaneously detect 34 cytokines in one array. In the large intestinal mucus of the rats in the control group, 32 cytokines, including inflammatory factors (IL-1β, IL-6, MCP-1 and TNF-α) and other oncogenic factors, were expressed at a high level in the GK rats, whereas the Neg and Fas ligand could hardly be detected (**Figures 6A,B**). Among the 32 cytokines increasingly expressed, an oncogenic factor Agrin (A6 and B6 spots in the array map) was remarkably enriched in GK rats. In the other 31 cytokines, IL-1β (C7 and D7 spots in the array map) is regarded as a cytokine that activates multiple immune cells and promotes insulin resistance (Dinarello, 2009). IFN-γ (C5 and D5 spots in the array map) is reported to be a mediator in the regulation of glucose metabolism by A. muciniphila (Greer et al., 2016). IL-6 (C11 and D11 spots in the array map) is confirmed to be a promoter of the death of islet β cells, which leads to T2DM (Donath, 2013). Correspondingly, the expression of cytokines in the BD3526 group was generally decreased (**Figures 6A,B**). To further semi-quantify the expression of inflammatory factors, we used ImageJ software to perform gray-scale analysis on the two images. The results showed that the intensity of the positive controls in the upper left and lower right are similar in the two arrays. However, the only expressed gene Agrin in the BD3526 group decreased by approximately 50% compared with the control group (**Figure 6C**). This result implied that the BD3526 strain fermentation products could reduce the expression level of most cytokines in the large intestinal mucus, such as IL-1β, IFN-γ and IL-6, and thus lower the blood glucose in GK rats.

# DISCUSSION

The gut microbiota is widely involved in the development of various diseases, and alteration of the gut microbiota through the intake of some specific microbial supplements has been adopted for the treatment of various diseases. These diseases include autism, Parkinson's disease and cancer (Hsiao et al., 2013; Buffington et al., 2016; Thaiss et al., 2016; Yu et al., 2017; Chung et al., 2018; Dejea et al., 2018; Routy et al., 2018; Tilg et al., 2018). In the study of gut microbiota and diabetes, the diversity and composition of the gut microbiota of patients changed significantly compared with healthy people (Cani et al., 2008; Larsen et al., 2010; Musso et al., 2010; Qin et al., 2012). In several intervention experiments, metformin has been found to increase the content of A. muciniphila and other short-chain fatty acid-producing microorganisms in the intestine of diabetic patients (Shin et al., 2014; de la Cuesta-Zuluaga et al., 2017; Wu et al., 2017; Wang et al., 2018). Metformin also has a profound effect on the interaction of microorganisms within the

gut microbiota. In this work, when GK rats were administered with the BD3526 strain fermentation products, the content of A. muciniphila in the intestine increased significantly compared with those in the control group. Similarly, A. muciniphila showed an increase after the metformin treatment in T2DM subjects (Shin et al., 2014; de la Cuesta-Zuluaga et al., 2017). This outcome indicates that metformin and the BD3526 strain fermentation products might behave similarly in alleviating the symptoms of diabetes. However, administration of metformin not only increase the content of A. muciniphila but also the content of other SCFAs-producing microorganisms in the intestine, e.g., Butyrivibrio, Bifidobacterium bifidum, and Megasphaera (de la Cuesta-Zuluaga et al., 2017), which provide a rational explanation for the higher levels of SCFAs being detected in the stool of clinical patients taking metformin (Wu et al., 2017). A large amount of dietary fiber has also been demonstrated to promote the growth of some probiotics in the intestine, stimulate the production of SCFAs, and alleviate the symptoms of diabetes (Zhao et al., 2018). Nevertheless, GK rats fed the BD3526 strain fermentation products did not exhibit a similar phenomenon, i.e., SCFAs-producing microorganisms in the BD3526 group were not wholly increased. Therefore, it could be postulated that both A. muciniphila and SCFAs-producing microorganisms are important in the regulation of blood glucose and behaved differently.

Akkermansia muciniphila is an anaerobic bacterium attached to the intestinal mucosa (Derrien et al., 2004, 2008; Collado et al., 2007; van Passel et al., 2011), and a lower level of A. muciniphila has been reported in obesity and T2DM subjects compared with normal ones. In addition, A. muciniphila can also increase the patient's response to drugs by increasing the recruitment of CCR9<sup>+</sup> CXCR3<sup>+</sup> CD4<sup>+</sup> T lymphocytes during the immunotherapy of tumor patients with PD-1/PD-L1 and increase the progression-free survival (PFS) of patients after treatment (Routy et al., 2018). Although the molecular mechanism by which A. muciniphila ameliorates diabetes/obesity is not fully understood, it is widely believed that this bacterium can play a positive role in health promotion by increasing the integrity of the intestinal mucosa (Derrien et al., 2004, 2008; Collado et al., 2007).

In T2DM, although the direct cause of the syndrome is insufficient for insulin secreted by pancreatic β-cells, the intrinsic cause is the inflammatory response (Cuman et al., 2001; Naguib et al., 2004). The chronically low grade of inflammatory responses not only causes pancreatic β-cells to be attacked but also reduces the liver and muscle sensitivity to insulin, which results in insulin resistance (Cerasi, 1995; Kahn, 1998; Bonner-Weir, 2000; Kahn et al., 2006). Even less optimistic is that the inflammatory response not only causes diabetes but also shows correlation between T2DM, insulin resistance, cardiovascular disease and

obesity (Donath, 2013). Therefore, reducing inflammation has been an important strategy for diabetes treatment. In our work, we found that the BD3526 strain fermentation products could lower the expression of intestinal mucosal inflammatory factors in GK rats and the level of HaB1c and blood glucose.

In this work, we used the BD3526 strain isolated from Tibetan yak milk to ferment skim milk (Hang et al., 2016). Although our understanding of this strain is not comprehensive, it has already demonstrated some potential in health promotion, e.g., secreting antimicrobial agents and levan, which is a recognized dietary fiber (Xu et al., 2016). Since these fermentation products could alleviate the symptoms of T2DM and selectively stimulate the propagation of A. muciniphila in the intestine, we believe that the biological function of P. bovis sp. nov. BD3526 might not be restricted to improving diabetes. As only the metabolites of P. bovis sp. nov. BD3526 in skim milk are employed in this work, the effect of the bacterium itself on the intestinal tract is still unknown and worthy of further investigation.

# AUTHOR CONTRIBUTIONS

ZQ, JH, ZW, and ZL conceived and designed the experiments. JH, HF, MY, CG, JW, and CY performed the experiments.

#### REFERENCES


ZQ, HZ, CG, and CY analyzed the data. ZQ and JH wrote the manuscript. All the authors read and approved the final manuscript.

#### FUNDING

This project was sponsored by Shanghai Engineering Center of Dairy Biotechnology (16DZ2280600) and the Shanghai Rising-Star Program (18QB1400100).

#### ACKNOWLEDGMENTS

We would like to thank Huifang An from Sinotech Genome Technology Co., Ltd., for the 16S rRNA sequencing data analysis for this work.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.03292/full#supplementary-material



**Conflict of Interest Statement:** ZQ, JH, HF, CG, JW, CY, ZL, and ZW were employed by Bright Dairy & Food Co., Ltd.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Qiao, Han, Feng, Zheng, Wu, Gao, Yang, You, Liu and Wu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Role of Probiotics in Mycoplasma pneumoniae Pneumonia in Children: A Short-Term Pilot Project

Zongxin Ling<sup>1</sup> \* † , Xia Liu<sup>1</sup>† , Shu Guo<sup>2</sup> , Yiwen Cheng<sup>1</sup> , Li Shao<sup>1</sup> , Dexiu Guan<sup>2</sup> , Xiaoshuang Cui<sup>2</sup> , Mingming Yang<sup>3</sup> and Xiwei Xu<sup>2</sup> \*

<sup>1</sup> Collaborative Innovation Center for Diagnosis and Treatment of Infectious Diseases, State Key Laboratory for Diagnosis and Treatment of Infectious Diseases, The First Affiliated Hospital, School of Medicine, Zhejiang University, Hangzhou, China, <sup>2</sup> Department of Gastroenterology, Affiliated Beijing Children's Hospital, Capital Medical University, Beijing, China, <sup>3</sup> School of Pharmacy, Shenyang Pharmaceutical University, Shenyang, China

#### Edited by:

Liwei Xie, Guangdong Institute of Microbiology (CAS), China

#### Reviewed by:

Wei Lee, Guangdong Institute of Microbiology (CAS), China Chuan Wang, Auburn University, United States

#### \*Correspondence:

Zongxin Ling lingzongxin@zju.edu.cn Xiwei Xu xxiwei@aliyun.com †These authors have contributed

#### equally to this work Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 14 November 2018 Accepted: 14 December 2018 Published: 09 January 2019

#### Citation:

Ling Z, Liu X, Guo S, Cheng Y, Shao L, Guan D, Cui X, Yang M and Xu X (2019) Role of Probiotics in Mycoplasma pneumoniae Pneumonia in Children: A Short-Term Pilot Project. Front. Microbiol. 9:3261. doi: 10.3389/fmicb.2018.03261 Mycoplasma pneumoniae is one of the most common pathogens causing communityacquired pneumonia in children. Mycoplasma pneumoniae pneumonia (MPP) can be successfully treated with azithromycin; however, antibiotic-associated diarrhea (AAD) is a common adverse effect. Increasing evidence suggests that some probiotics may prevent the development of AAD. The present study determined the effects of probiotics (live Clostridium butyricum plus Bifidobacterium infantis) on the prevention and treatment of AAD in children with MPP when co-administered with intravenous azithromycin. Fifty-five children with MPP were enrolled and received azithromycin (10 mg/kg/day; once daily for 7 days) combined with probiotics (starting on the third day of azithromycin treatment; 1,500 mg three times daily); 50 healthy children served as controls. At the end of the trial, the incidence of AAD, fecal microbiota, intestinal mucosal barriers, and systemic inflammation were analyzed using recommended systems biology techniques. No cases of AAD or other adverse events occurred in children with MPP after co-administration of probiotics with azithromycin. A live C. butyricum plus B. infantis preparation partly reconstructed the gut microbiota, especially restoration of bacterial diversity. The indicators of intestinal mucosal barrier function, such as Dlactate, endotoxin, and diamine oxidase, were significantly improved and the systemic inflammation (interleukin 10) was attenuated after probiotic therapy. The present study indicated that co-administration of probiotics with azithromycin is a promising therapy for MPP treatment which could prevent and treat AAD effectively.

Keywords: azithromycin, Bifidobacterium, Clostridium, Mycoplasma pneumoniae pneumonia, probiotics

#### INTRODUCTION

Community-acquired pneumonia (CAP) is the most common cause of death in children worldwide. Globally, CAP accounts for 15% of deaths in children <5 years of age and 922,000 deaths in children of all ages (Haq et al., 2017). Among all causative organisms, Mycoplasma pneumoniae (MP) is one of the most common pathogens causing CAP in children (Defilippi et al., 2008). M. pneumoniae pneumonia (MPP) not only causes pulmonary complications, such

Ling et al. Beneficial Effects of Probiotics on MPP

as bronchiolitis obliterans and necrotizing pneumonia, but also leads to a number of extra-pulmonary complications, such as encephalitis, arthritis, pericarditis, and hemolytic anemia, which can develop into severe life-threatening pneumonia. Due to the absence of a cell wall, MP is usually treated with antibiotics, such as quinolones, tetracyclines, and macrolides; however, only macrolides (erythromycin, azithromycin, clarithromycin, and roxithromycin) are used for children because of the potential side effects of alternative drugs, such as tetracyclines and fluoroquinolones (Youn and Lee, 2012). Azithromycin is generally the first-line treatment choice for MPP for children in our hospital because azithromycin is well-tolerated in the presence of a wide variety of concurrent illnesses and medications. It cannot be ignored, however, that the gastrointestinal adverse effects, such as diarrhea, are 72% higher with long-term azithromycin therapy (Florescu et al., 2009). Antibiotic-associated diarrhea (AAD) may be associated with dysbiosis of the gut microbiota that is disturbed by azithromycin (Xinli et al., 2013). Wei et al. (2018) reported that short-term azithromycin administration caused a 23% reduction in observed richness and 13% reduction in Shannon diversity. Recent studies have shown that the gut microbiota plays vital roles in numerous aspects of normal host physiology, from nutritional status to behavioral and stress responses (Subramanian et al., 2014; Jiang et al., 2015). A previous study showed that the decrease in short-chain fatty acid (SCFA)-producing bacteria cannot prevent the overgrowth of some potentially pathogenic intestinal microbes such as Shigella and Escherichia, which are associated with the development of diarrhea (Clausen et al., 1991; De Filippo et al., 2010). Johnston et al. (2016) demonstrated that co-administration of antibiotics with probiotics is associated with lower rates of AAD compared with controls, without an increase in clinically important adverse events. Our previous study showed that SCFA-producing probiotic strains, such as Clostridium butyricum, Bifidobacterium infantis, and mixtures of C. butyricum and B. infantis can restore gut microbiota and attenuate systemic inflammation in mice with AAD (Ling et al., 2015). The effects of probiotics on the prevention and treatment of AAD in children with MPP co-administered antibiotics, however, have not been thoroughly investigated. The purpose of the present study was to evaluate the effects of probiotics (live C. butyricum plus B. infantis) on children with MPP who are simultaneously treated with intravenous azithromycin, which will provide new adjuvant therapy for MPP in clinical practice.

# MATERIALS AND METHODS

#### Recruitment of Subjects

From January 2015 to June 2017, a total of 55 children with a final diagnosis of MPP were admitted to the Department of Gastroenterology at the Affiliated Beijing Children's Hospital in China; 50 age- and gender-matched healthy children served as controls. All patients had symptoms and signs indicative of pneumonia at the time of admission, including a fever (>37.5◦C), cough, abnormal breath sounds on auscultation, and an abnormal chest X-ray. MP infection was confirmed in nasopharyngeal secretions and serum samples using PCR and ELISA (Wang et al., 2014). The following exclusion criteria were established: age <1 month or >14 years; refractory MPP based on the presence of persistent fever and clinically, as well as radiologic deterioration after azithromycin treatment for ≥7 days; use of antibiotics, probiotics, prebiotics, or synbiotics in the previous month; and other respiratory tract infections, such as bacterial, fungal, chlamydial, or viral infections (such as respiratory syncytial virus, adenovirus, metapneumovirus, influenza virus A and B, and parainfluenza virus 1, 2, and 3); other diseases, such as asthma, chronic cardiac and pulmonary disease, rheumatic diseases, and immunodeficiency. The protocols for the present study were approved by the Ethics Committee of Affiliated Beijing Children's Hospital at Capital Medical University (Beijing, China) and the methods were carried out in accordance with the approved guidelines (number: 2014-5). Written informed consent was obtained from the parents or guardians of all participants prior to enrollment.

#### Treatment

All patients were treated with azithromycin intravenously at a dose of 10 mg/kg/day once daily for 7 days. On the third day of azithromycin treatment, the children were instructed to administer probiotics orally at least 2 h after antibiotic treatment until discharge from the hospital. A C. butyricum (CGMCC0313- 1) combined with B. infantis (CGMCC0313-2) probiotic mixture (Changlekang <sup>R</sup> ; Shandong Kexing Bioproducts Co., Ltd., Jinan, China) was used to treat the potential AAD. The freeze-dried probiotic mixture had >1.0 × 10<sup>7</sup> CFU/g of viable C. butyricum and >1 × 10<sup>6</sup> CFU/g of viable B. infantis per capsule. Children who were treated with probiotics received a dose of 1,500 mg three times daily.

#### Sample Collection

On the third day of azithromycin treatment and the day of hospital discharge, fresh fecal samples (approximately 2 g) and blood samples were collected from each child for gut microbiota and intestinal mucosal barrier function analyses. These samples were transferred immediately to the laboratory and stored at −80◦C after preparation within 15 min until use.

# Fecal Bacterial Genomic DNA Extraction

Bacterial genomic DNA was extracted using a QIAamp <sup>R</sup> DNA Stool Mini Kit (Qiagen, Hilden, Germany) according to our previous study (Ling et al., 2015). The amount of bacterial genomic DNA was analyzed using a NanoDrop ND-1000 spectrophotometer (Thermo Scientific, Wilmington, DE, United States). The integrity and size of the bacterial genomic DNA were checked by electrophoresis. All bacterial genomic DNA were stored at −20◦C for further use.

# Amplicon Library Construction and High-Throughput Sequencing

Amplicon libraries were constructed with Illumina sequencingcompatible and barcode-indexed bacterial PCR primers

319F/806R, which target the V3−V4 regions of the 16S rRNA gene (Fadrosh et al., 2014). All PCR reactions were performed with KAPA HiFi HotStart ReadyMix using the manufacturer's protocol (Kapa Biosystems, Boston, MA, United States) and approximately 50 ng of extracted DNA per reaction. Thermocycling conditions were set at 95◦C for 1 min, 55◦C for 1 min, then 72◦C for 1 min for 30 cycles, followed by a final extension at 72◦C for 5 min. All PCR reactions were performed in 50 ml triplicates and combined after PCR. The amplicon library was prepared using a TruSeqTM DNA sample preparation kit (Illumina, Inc., San Diego, CA, United States). Prior to sequencing, the DNA concentration of each PCR product was extracted with the MiniElute <sup>R</sup> Gel Extraction Kit (Qiagen) and quantified on a NanoDrop ND-1000 spectrophotometer and Qubit 2.0 Fluorometer (Invitrogen, Carlsbad, CA, United States). The purified amplicons were then pooled in equimolar concentrations and the final concentration of the library was determined by Qubit. Negative DNA extraction controls (lysis buffer and kit reagents only) were amplified and sequenced as contamination controls. Sequencing was performed on a MiSeq instrument (Illumina, Inc.) using a 300 × 2 V3 kit together with PhiX Control V3 (Illumina, Inc.).

#### Bioinformatic Analysis

The 16S rRNA gene sequence data sets generated from the MiSeq run were first merged and demultiplexed into per samples using QIIME (version 1.9.0) with default parameters (Caporaso et al., 2010). Chimera sequences were detected and removed using the USEARCH software based on the UCHIME algorithm (Edgar et al., 2011). An open-reference operational taxonomic unit (OTU) pick was then performed with USEARCH (version 7) referenced against the Greengenes database (version 13.8) at 97% sequence similarity (Edgar, 2010; Mcdonald et al., 2012), which was used for subsequent microbiota composition analysis. Alpha diversity was calculated using QIIME software with Python scripts based on the sequence similarity at the 97% level, including an index of observed species, abundance-based coverage estimator (ACE), Chao1 estimator, Shannon, Simpson, Evenness and PD whole tree. Sequence coverage was assessed in mothur by rarefaction curves and Good's coverage (Good, 1953; Schloss et al., 2009). Beta diversity was measured by unweighted and weighted UniFrac distance calculated with 10 times of sub-sampling by QIIME. These distances were visualized by principal coordinate analysis (PCoA) with a different algorithm (Lozupone and Knight, 2005). Hierarchical clustering was performed and heatmap was generated using a Spearman's rank correlation coefficient as a distance measure and a customized script developed in the R statistical package. Characterization of microorganismal features differentiating the gut microbiota was performed using the linear discriminant analysis (LDA) effect size (LEfSe) method<sup>1</sup> for biomarker discovery, which emphasizes both statistical significance and biological relevance (Segata et al., 2011).

# Intestinal Mucosal Barrier Function Analysis

The parameters of intestinal mucosal barrier function, such as D-lactate, endotoxin (LPS), and diamine oxidase (DAO), were detected using the dry chemical method of the Intestinal Mucosal Barrier Biochemical Index Analysis System (JY-DLT; Beijing Zhongsheng Jinyu Diagnostic Technology Co., Ltd., Beijing, China). The anti-inflammatory cytokine, interleukin 10 (IL-10), in serum was detected using a human IL-10 immunoassay kit (eBioscience, San Diego, CA, United States). The data are presented as the mean ± standard deviation (SD) and the differences between the two groups were evaluated by Student's t-test using SPSS (version 20.0; SPSS, Inc., Chicago, IL, United States).

#### Statistical Analysis

Statistical analysis was performed using SPSS (version 20.0). GraphPad Prism (version 6.0; San Diego, CA, United States) was used for preparation of graphs. For bacterial diversity indices, bacterial composition at different taxonomic levels and parameters of intestinal mucosal barrier function, Student's t-test and Mann–Whitney U test were applied. All tests of significance were two-sided, and a p < 0.05 was considered statistically significant for all analyses.

#### Accession Number

The sequence data from this study are deposited in the GenBank Sequence Read Archive with the accession number SRP171107.

# RESULTS

#### General Information of Participants and Sequencing Data

A total of 55 children who were confirmed to have MPP were enrolled in the present study. Fifty age- and gender-matched healthy children were recruited as controls. The ages of both patients and healthy control subjects ranged from 3 to 6 years, with a female-to-male ratio of 0.58 in the MPP group (32/23) and 0.62 (31/19) in the control group. After successful treatment with azithromycin and probiotics, no AAD and other adverse events were observed in the MPP children. The average length of stay for these patients was 12.3 days (range, 11–14 days), while the average time for probiotic administration was 8.5 days (range, 8–11 days). On the third day of azithromycin treatment and the day of hospital discharge, feces and blood samples were collected from 55 patients; identical samples were also collected from the healthy children.

Based on the DNA amount and quality appropriate for 16S rRNA gene amplification and sequence analysis, 41 samples on the third day of azithromycin treatment (pre-treatment), 40 samples on the day of discharge (post-treatment) and 47 samples from healthy controls (control) were used for microbiota and intestinal mucosal barrier function analyses. Using the Illumina MiSeq sequencing platform, we generated 6,461,307 16S rRNA gene sequences from 128 samples, with an average

<sup>1</sup>http://huttenhower.sph.harvard.edu/lefse/

sequence length of 438 nt, following paired-end merging and trimming. The average sequence depth per sample was 42,045 (minimum = 26,139; maximum = 62,103). All libraries had a Good's coverage score ≥ 99.9% at the rarefaction point of 26,139 sequences, indicating that deep sequence coverage of the fecal microbiome was achieved for each sample. Thus, a total of 1,837,836 sequences were obtained from healthy controls for downstream analysis, while 1,819,648 sequences (pre-treatment) and 1,724,335 sequences (post-treatment) were obtained from children with MPP.

#### Alterations in the Overall Structure of Fecal Microbiota After Probiotic Treatment

In the present study, the total number of unique sequences from the three groups was 798, and represented all phylotypes. **Figure 1** shows the alterations in the overall structure of fecal microbiota among control and pre- and post-treatment groups. The alpha diversity indices, such as Shannon and Simpson, showed that the fecal microbiota in the pre-treatment group were significantly lower than the control and posttreatment groups (**Figures 1A,B**; p < 0.05); however, there was no apparent difference between the control and posttreatment groups. The richness estimators, such as observed OTUs, Chao1, and ACE, showed similar changing patterns (**Figures 1C–E**). Our data indicated that the richness estimators were significantly decreased in the pre-treatment group when compared with control and post-treatment groups (p < 0.05). No significant differences between the control and post-treatment groups were observed for the diversity and richness indices (p > 0.05). The diversity indices and richness estimators of the fecal microbiota suggested a tendency toward microbiota restoration after probiotic treatment. Beta diversity analysis, such as PCoA based on the unweighted UniFrac, weighted UniFrac, and Bray–Curtis distances, indicated that there was a significant distinct clustering in the fecal microbiota between control and MPP patients, and a similar clustering between the pre-treatment and post-treatment groups (**Figures 2A–C**). Furthermore, another beta diversity analysis, the non-metric multi-dimensional scaling (NMDS) analysis, also revealed a similar clustering of the microbiota composition with the PCoA analysis (**Figure 2D**). A Venn diagram indicated that children shared a core set of bacteria in fecal microbiota regardless of the health status (**Figure 2E**). Despite significant interindividual variability, three clusters were found in fecal microbiota of children using unweighted UniFrac, which indicated that fecal microbiota was divided into cluster I in healthy controls, while fetal microbiota was divided into clusters II and III in pre- and post-treatment groups (**Supplementary Figure S1**).

The sequences from the fecal microbiota could be classified as 17 phyla; however, the majority (≥95%) of sequences of the fecal microbiota were classified into four phyla, including Firmicutes, Bacteroidetes, Proteobacteria, and Actinobacteria (**Figure 3A**). Among the top four most abundant phyla in the fecal microbiota, Firmicutes and Actinobacteria were significantly increased and Bacteroidetes was decreased in the pre- and post-treatment groups when compared with the control group, while Proteobacteria was only increased in the pre-treatment group. **Supplementary Figure S2** demonstrates the relative abundance of phyla in each sample among the three groups. At the family level, the relative abundance of Lachnospiraceae, Streptococcaceae, and Actinomycetaceae was significantly higher, while the relative abundance of Bacteroidaceae, Porphyromonadaceae, Alcaligenaceae, Rikenellaceae, and Prevotellaceae was clearly lower in both preand post-treatment groups when compared with the control group. A relatively higher abundance of Enterobacteriaceae, Enterococcaceae, Erysipelotrichaceae, and Lactobacillaceae was observed in the pre-treatment group, while a relatively higher abundance of Bifidobacteriaceae, Coriobacteriaceae, Carnobacteriaceae, and Clostridium XI existed in the posttreatment group (**Figure 3B**; p < 0.05). At the genus level, Bacteroides, Faecalibacterium, and Parabacteroides were significantly decreased, while Enterococcus, the Ruminococcus gnavus group, Streptococcus, and Lachnoclostridium were significantly increased in the pre- and post-treatment groups when compared with the control group. After probiotic treatment, we also showed that Erysipelatoclostridium was lower and Blautia was higher than in the pre-treatment group (**Figure 3C**; p < 0.05). **Figure 4** shows a heatmap of bacterial genera in the pre- and post-treatment and control groups, which represented the relative percentages of the most abundant genera identified in each sample. Our data indicate that there were significant differences in the upper heatmap between the control and other two groups. **Supplementary Figure S3** shows the relative abundance of genera in each sample among the three groups. LEfSe was used to compare the estimated phylotypes of the fecal microbiota among the three groups (**Supplementary Figures S4**–**S7**). In agreement with the previous analysis, opportunistic pathogens, such as Enterococcus and Enterobacteriaceae, were increased after antibiotic treatment, while butyrate-producing bacteria, such as Clostridium and Ruminococcus, were also increased after probiotic treatment. In summary, the diversity of the fecal microbiota in children with MPP showed a trend toward microbiota restoration after short-term probiotic treatment, yet there was no significant improvement in the composition of the fecal microbiota.

#### Improvement in Intestinal Mucosal Barrier Function After Probiotic Treatment

The intestinal mucosal barrier function was damaged in children with MPP after antibiotic treatment. The levels of D-lactate, DAO, and LPS were significantly increased, and the concentration of the anti-inflammatory cytokine, IL-10, was dramatically decreased (**Figure 5**). The increased parameters of intestinal mucosal barrier function, such as D-lactate, DAO, and LPS, represented the increased intestinal permeability, which has been reported to be related to bacterial translocation from the lumen to extra-intestinal sites. At the end of the trial, the parameters of intestinal mucosal barrier function and systemic inflammation were significantly improved after probiotic treatment. The

present study indicated that a C. butyricum combined with B. infantis probiotic mixture repairs damaged intestinal mucosal permeability and attenuates systemic inflammation in children with MPP.

#### DISCUSSION

among the fecal microbiota.

As a common pathogen of lower respiratory tract infections in children, M. pneumoniae accounts for up to one third of all cases of atypical CAP (Principi et al., 2001). With no cell wall, M. pneumoniae is resistant to beta-lactams and to all antimicrobials targeting the cell wall, but sensitive to macrolides and related antibiotics, including tetracyclines and fluoroquinolones. Due to the low toxicity, low MIC against the pathogen, and absence of a contraindication in young children, macrolides and related antibiotics, especially azithromycin, are usually the first-line treatment for MPP in children (Pereyre et al., 2016). The present study also showed that azithromycin successfully treated M. pneumoniae infection in children; however, the antimicrobial agents that we used against M. pneumoniae also disrupted co-evolved microbial communities that are integral to human health. Azithromycin led to microbiota dysbiosis and disturbed the colonization resistance of gastrointestinal microbiota, which induces clinical symptoms, most commonly diarrhea. A previous study has shown that the use of azithromycin is correlated with the increased incidence of AAD (Erdeve et al., 2004). Engelbrektson et al. (2009) demonstrated that a probiotic mixture consisting of B. lactis Bl-04, B. lactis Bi-07, L. acidophilus NCFM, L. paracasei Lpc-37, and B. bifidum Bb-02 minimizes the disruption of fecal microbiota in healthy subjects undergoing antibiotic therapy. Our previous data also indicated that a C. butyricum combined with B. infantis probiotic mixture restores fecal microbiota and attenuates systemic inflammation in mice with AAD (Ling et al., 2015). The present study determined the role of the

abovementioned probiotic mixture in children with MPP treated with azithromycin.

In the present study, our data showed that there was no evident AAD for MPP children after azithromycin treatment with additional probiotic therapy. A previous study reported that azithromycin is most often associated with the development of AAD (Gorenek et al., 1999). Improvement of the therapeutic effects and reduction of side effects might be attributed to probiotic therapy, which was aimed at gut microbiota restoration. With high-throughput sequencing techniques, the diversity indices such as Shannon and Simpson indicated that the bacterial diversity of fecal microbiota showed clear trends in microbiota restoration, while the richness indices, such as observed OTUs and Chao1, demonstrated that the estimated phylotypes increased significantly after probiotic treatment. The overall structure of the fecal microbiota was reconstructed after azithromycin treatment when combined with probiotic therapy, which might be associated with the decreased incidence of AAD for children with MPP. Unfortunately, the composition of the fecal microbiota was not effectively restored with similar bacterial diversity after probiotic treatment. PCoA showed that the fecal microbiota could not be separated between antibiotic treatment and combined therapy, which might be related to short-term probiotic administration. Our previous study showed that the dysbiosis of the gut microbiota disturbed by antibiotics in mice did not recover quickly and naturally (Ling et al., 2015), although the native microbiota of the host displayed considerable resilience to the normal state after the antibiotic perturbation (Dethlefsen and Relman, 2011). Zaura et al. (2015) also reported that a single course of antibiotics is sufficient to disrupt the normal make-up of microorganisms in the gut for up to 1 year, potentially leading to antibiotic resistance (Jernberg et al., 2010). Based on our previous data, long-term probiotic therapy, but not short-term course, exerted beneficial effects on restoration of the gut microbiota (Ling et al., 2015). The present study was a short-term pilot clinical trial that evaluated the efficacy of probiotic therapy, which might affect the outcomes of microbiota restoration; however, short-term probiotic therapy was effective in improving clinical symptoms, ameliorating intestinal mucosal barrier function, and attenuating systemic inflammation. Despite successful clinical recovery, Song et al. (2013) also noted that the fecal microbiota of patients with recurrent C. difficile infection did not fully recover after fecal microbiota transplantation over 16 weeks. Rinne et al. (2006) also reported that probiotic intervention has short-term effects on gastrointestinal symptoms and long-term effects on gut microbiota. In agreement with the previous study, our data suggest that fecal microbiota restoration after probiotic intervention was slower than clinical recovery.

The probiotic mixture (live C. butyricum plus B. infantis preparation) has been used to modulate gut microbiota for children in China for many years, and can help achieve new eubiosis through supplementation of beneficial bacteria, stimulation of commensal bacteria growth, and consequent reduction of pathogenic species. Our previous studies have shown that the probiotic mixture, but not C. butyricum or B. infantis alone, exert beneficial effects on restoration of the gut

microbiota, recovery of intestinal mucosal barrier function, and attenuation of systemic inflammation in animals and humans (Ling et al., 2015; Xia et al., 2018). Our animal study showed that the probiotic mixture helped restore the gut microbiota better than C. butyricum or B. infantis alone (Ling et al., 2015). C. butyricum (belonging to Clostridium cluster I) is a typical butyrate-producing, endospores-forming, Gram-positive obligate anaerobe that is isolated from soil and guts of healthy animals and humans. The longevity of endospores and the resistance to both chemical and physical stresses have determined the survival of C. butyricum at lower pH and relatively higher bile concentrations (Kong et al., 2011). These properties are a

benefit of the use of C. butyricum as a probiotic in clinical practice. A previous study has shown that C. butyricum MIYAIRI 588 is effective for the treatment and the prophylaxis of AAD in children in Japan, as C. butyricum MIYAIRI 588 normalizes the gut microbiota disturbed by antibiotics (Seki et al., 2003). Previous studies have shown that C. butyricum promotes the growth of beneficial strains of Lactobacillus and Bifidobacterium and inhibits the harmful strain of C. perfringens (Seki et al., 2003; Kong et al., 2011; Zhang et al., 2011), which is consistent with our previous study on mice. Our previous study also showed that C. butyricum attenuates cerebral ischemia/reperfusion injuries in diabetic mice via modulation of gut microbiota, which regulates the bidirectional communication of the gut-brain axis (Sun et al., 2016). The potential underlying mechanism might be associated with elevated levels of SCFAs, especially butyrate after the administration of C. butyricum, which plays an important role in recovering intestinal tight junctions and maintaining gut integrity. Bifidobacterium, the dominant commensal bacteria of colonic microbiota, accounts for up to 25% of the cultivable fecal bacteria in adults and 80% in infants. Owing to the rare association with infection, Bifidobacteria have been studied for efficacy in the prevention and treatment of a broad spectrum of animal and/or human gastrointestinal disorders as probiotic strains. A previous study has shown that B. infantis, alone or in combination with other bacteria, can specifically relieve various symptoms of irritable bowel syndrome (Whorwell et al., 2006), and reduce the incidence and severity of necrotizing enterocolitis in very low birth weight infants (Lin et al., 2005). Charbonneau et al. (2013) demonstrated that supplementation with B. infantis potentially altered the composition of the fecal microbiota in patients with irritable bowel syndrome. The present study also showed that Bifidobacterium and Clostridium

increased significantly after probiotic treatment. A previous study has found that secreted bioactive factors from B. infantis, such as SCFAs and peptides, retain their biological activity in vivo, and are effective in normalizing gut permeability and improving disease in an animal model of colitis (Ewaschuk et al., 2008). In agreement with our previous study on mice, our data indicated that live C. butyricum plus B. infantis preparation repaired the intestinal mucosal barrier function and significantly alleviated the host inflammatory response, even though probiotic therapy was administered relatively shortterm. Chen et al. (2010) also reported that the short-term use of probiotics preparation, including live C. butyricum, not only improved the composition of the gut microbiota, but also attenuated the severity of acute diarrhea in hospitalized children and was associated with a reduced length of hospital stay; however, the efficacy of probiotics in treating infections and AAD has been recently questioned (Suez et al., 2018; Zmora et al., 2018) and some clinical studies have even reported probiotic-associated morbidity and mortality (Besselink et al., 2008; Vogel, 2008). Probiotics have been proposed to constitute an effective preventive treatment for antibioticinduced dysbiosis; however, the adverse effects associated with probiotic consumption may be under-reported in clinical trials (Bafeta et al., 2018). Recent studies conducted by one group from Israel have demonstrated that post-antibiotic probiotic benefits may be offset by compromised gut mucosal recovery, and empiric probiotics supplementation may delay gut microbiome and transcriptome reconstitution post-antibiotic treatment (Suez et al., 2018; Zmora et al., 2018). They consider that probiotics treatment should be person-, strain-, and disease-specific, but not an empiric "one size fits all" probiotic regimen design. Even so, the available evidence does not completely refute the positive roles of probiotics. The health benefits conferred by probiotic bacteria are strain-specific, depending on the strain and disease tested (Whelan and Quigley, 2013). The present study has demonstrated that the two bacterial strains have shown very promising results, preventing the occurrence of AAD in children.

There were several limitations to the present study. First, the present study only observed the therapeutic effects of short-term probiotic intervention on children with MPP, while additional doses of the probiotics administered for longer durations are needed to further explore the role in gut microbiota restoration and relief of clinical symptoms for children discharged from the hospital with MPP. Second, the number of patients in the present study was small and the follow-up time was relatively short; children with MPP who are treated with probiotics with a longer duration of follow-up would strengthen our findings and conclusions. Third, the potential protective effects of a live C. butyricum plus B. infantis preparation to prevent AAD in children did not withstand intention-to-treat analysis. Future randomized controlled trials will be designed with probiotics and other types of gastrointestinal protective agents which can help determine the actual role and mechanism underlying these beneficial bacteria. Fourth, the possibility of natural recovery of gut microbiota in these children with MPP was not considered in the present clinical experimental design, although the placebo control helped to determine the role of the probiotics more accurately. Thus, the present report can be regarded as a pilot study that should be verified in further clinical trials.

# CONCLUSION

fmicb-09-03261 January 7, 2019 Time: 17:47 # 10

In conclusion, the present study has shown that short-term use of a live C. butyricum plus B. infantis preparation was effective in preventing the development of AAD in children hospitalized with MPP who were treated with azithromycin. The probiotic mixture partly reconstructed the gut microbiota, especially restoration of bacterial diversity, which might be due to short-term probiotic intervention. Interestingly, the intestinal mucosal barrier function was significantly improved and systemic inflammation was attenuated after probiotic therapy. Therefore, the administration of a live C. butyricum plus B. infantis preparation may become a promising therapy for prevention and treatment of AAD.

#### AUTHOR CONTRIBUTIONS

ZL and XX conceived and designed the study. ZL, XL, YC, and LS generated the sequencing data. XL, DG, SG, XC, MY, and XX collected the samples. ZL, XL, LS, and XX analyzed the data, carried out the computational analysis, interpreted the data, and drafted the manuscript.

# FUNDING

This present work was funded by the grants of the National Natural Science Foundation of China (81771724, 31700800, 81790633, and 31870839) and National S&T Major Project of China (2018YFC2000501).

#### REFERENCES


#### ACKNOWLEDGMENTS

We gratefully acknowledge the volunteers who participated in our study.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.03261/full#supplementary-material

FIGURE S1 | Differentiation in the fecal microbiota from each sample of control and pre- and post-treatment groups (interpersonal variations). Community differentiation was measured using the unweighted UniFrac algorithm; the scale bar indicates the distance between clusters in UniFrac units. All of the branch nodes shown were significant (p < 0.001).

FIGURE S2 | Comparison of the relative abundances of bacterial phyla of each sample among control and pre- and post-treatment groups.

FIGURE S3 | Comparison of the relative abundances of bacterial genera of each sample among control and pre- and post-treatment groups.

FIGURE S4 | Taxonomic differences of the fecal microbiota among control and pre- and post-treatment groups using LEfSe.

FIGURE S5 | Taxonomic differences of the fecal microbiota between control and pre-treatment groups using LEfSe.

FIGURE S6 | Taxonomic differences of the fecal microbiota between control and post-treatment groups using LEfSe.

FIGURE S7 | Taxonomic differences of the fecal microbiota between pre- and post-treatment groups using LEfSe.


transplantation for recurrent Clostridium difficile infection. PLoS One 8:e81330. doi: 10.1371/journal.pone.0081330


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Ling, Liu, Guo, Cheng, Shao, Guan, Cui, Yang and Xu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fmicb-09-03261 January 7, 2019 Time: 17:47 # 11

# Intestinal Morphologic and Microbiota Responses to Dietary Bacillus spp. in a Broiler Chicken Model

Cheng-liang Li1,2, Jing Wang<sup>2</sup> , Hai-jun Zhang<sup>2</sup> , Shu-geng Wu<sup>2</sup> , Qian-ru Hui<sup>3</sup> , Cheng-bo Yang<sup>3</sup> , Re-jun Fang<sup>1</sup> \* and Guang-hai Qi1,2 \*

<sup>1</sup> College of Animal Science and Technology, Hunan Agricultural University, Changsha, China, <sup>2</sup> Key Laboratory of Feed Biotechnology of Ministry of Agriculture and Rural Affairs, Feed Research Institute, Chinese Academy of Agricultural Sciences, Beijing, China, <sup>3</sup> Department of Animal Science, Faculty of Agricultural and Food Sciences, University of Manitoba, Winnipeg, MB, Canada

#### Edited by:

Yuheng Luo, Sichuan Agricultural University, China

#### Reviewed by:

Tarique Hussain, Pakistan Institute of Engineering and Applied Sciences, Pakistan Tao Yang, University of Florida, United States Deguang Song, Yale School of Medicine, Yale University, United States

\*Correspondence:

Re-jun Fang fangrj63@126.com Guang-hai Qi qiguanghai@caas.cn

#### Specialty section:

This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology

Received: 30 September 2018 Accepted: 31 December 2018 Published: 17 January 2019

#### Citation:

Li C-l, Wang J, Zhang H-j, Wu S-g, Hui Q-r, Yang C-b, Fang R-j and Qi G-h (2019) Intestinal Morphologic and Microbiota Responses to Dietary Bacillus spp. in a Broiler Chicken Model. Front. Physiol. 9:1968. doi: 10.3389/fphys.2018.01968 Dietary inclusion of probiotic Bacillus spp. beneficially affect the broiler chickens by balancing the properties of the indigenous microbiota causing better growth performance. The effects of three Bacillus spp. on the growth performance, intestinal morphology and the compositions of jejunal microflora were investigated in broiler chickens. A total of 480 1-day-old male Arbor Acres broilers were randomly divided into four groups. All groups had six replicates and 20 birds were included in each replicate. The control birds were fed with a corn-soybean basal diet, while three treatment diets were supplemented with Bacillus coagulans TBC169, B. subtilis PB6, and B. subtilis DSM32315 with a dosage of 1 × 10<sup>9</sup> cfu/kg, respectively. The experiment lasted for 42 days. The compositions and diversity of jejunal microflora were analyzed by MiSeq high-throughput sequencing. The B. coagulans TBC169 group showed marked improvements of growth performance, nutrient digestibility and intestinal morphology compared with the other B. subtilis treatments. B. coagulans TBC169 supplementation improved the average body weight (BW), average daily weight gain (ADG), total tract apparent digestibility of crude protein and gross energy (GE), and reduced feed conversion rate (FCR) compared with the control group (P < 0.05). The villus height to crypt depth ratio (VH/CD) of jejunum and duodenum was increased in the birds fed with B. coagulans TBC169 compared with the control group (P < 0.05). However, two B. subtilis treatments presented more positive variation of the jejunum microflora of chickens than that in the B. coagulans TBC169 group. B. subtilis PB6 and B. subtilis DSM32315 treatments improved the diversity of jejunal microbiota on day 21 compared with the control (P < 0.05), while which were decreased on day 42 (P < 0.05). The supplementation with B. coagulans TBC169 significantly improved the proportion of Firmicutes, otherwise two B. subtilis significantly improved the proportion of Proteobacteria, Bacteroidetes, Actinobacteria, and Acidobacteria at the phylum level during starter phase and decreased the proportion of Bacteroidetes during growing phase compared with the control. The supplementation with B.subtilis DSM32315 significantly improved the proportion of Clostridiales during starter phase, whereas two B. subtilis significantly improved the proportion of Pseudomonas, Burkholderia, Prevotella, DA101 during growing phase at the genus level compared with the control. In conclusion, the dietary supplementation with probiotic Bacillus spp. strains improved body weight and intestinal morphology in broiler chickens, which might be associated with the gut microbiota.

Keywords: probiotics, growth performance, intestinal morphology, jejunum microbiota, broiler

#### INTRODUCTION

fphys-09-01968 April 1, 2019 Time: 19:1 # 2

Broiler chickens have been reared and consumed widely around the world since they can provide high-quality meat and eggs for human beings. During the last 6 or 7 decades, with the development of production system, chickens can convert feed into muscle mass efficiently (Clavijo and Flórez, 2017). The worldwide demand for chicken meat continues to grow considerably (Tallentire et al., 2018). Meanwhile, because of the detrimental side effects of antibiotics on both poultry products and human well-being, an increasing number of countries have implemented the withdrawal of antibiotic growth promoters (AGPs), which was previously recognized as growth promoters to boost animal growth performance and inhibit the spread of certain diseases (Mashayekhi et al., 2018).

Probiotics are non-pathogenic bacterial cultures that can adjust intestinal microflora and in turn improve the gastrointestinal environment of the host. In addition, probiotics have positive impacts on colonized beneficial bacterial and growth performance in broilers and pigs (Jeong and Kim, 2014; Valeriano et al., 2017). Noticeably, probiotics, as the alternatives for antibiotics used to prevent poultry diseases and improve production performance, have been demonstrated to be beneficial to chickens' growth performance and health, such as the increases of body weight (BW), feed conversion efficiency, immune response, resistance to bacterial infection, and regulation of intestinal microflora (Xu et al., 2012; Cao et al., 2013; Song et al., 2014; Clavijo and Flórez, 2017; Haque et al., 2017; Hussain et al., 2017a; Azad et al., 2018; Mashayekhi et al., 2018).

Presently, spore-forming bacteria, such as Bacillus spp. including Bacillus subtilis, B. coagulans, and B. licheniformis etc. have been widely used as commercialized probiotic products for humans and animals (Barbosa et al., 2005; Zhang et al., 2013; Park and Kim, 2014; Haque et al., 2017; Xu et al., 2017). Bacillus spp. have been also considered to be promising probiotics, due to the high stability of spores, which is resistant to high temperature and harsh gastrointestinal conditions during feed processing and that can confer health benefits to the host (Mazanko et al., 2017). Previous study showed that B. subtilis and B. coagulans had positive effects on tilapia growth and immune response (Zhou et al., 2010). Moreover, it has been reported that dietary supplementation with B. subtilis exerted a beneficial role in the digestibility and intestinal microbes of weaning piglets, and finally improving their growth performance (Tsukahara et al., 2013). Lee S.H. et al. (2014) suggested that dietary supplementation with B. subtilis in pigs exhibited significant effects on gut morphology, microbiota compositions and immune function. Feeding broilers with B. coagulans diets can improve the feed conversion ratio (FCR) and beneficially modulate the composition of the microflora, which markedly enhanced the relative abundance of lactobacilli and tended to lower coliform bacteria composition (Hung et al., 2012). As a result, oral administration with B. subtilis and B. coagulans may have potential to improve the growth state, intestinal function and microflora compositions of broilers. But many studies reported that probiotics had no significant effect on growth in broilers (Arslan et al., 2004; Chitra et al., 2004; Das et al., 2005).

Researches on the classification and identification of intestinal microbes in poultries were conducted progressively because of the conventional molecular ecology techniques such as denaturing gradient gel electrophoresis (DGGE) fingerprints (Li et al., 2017; Yang et al., 2018). However, these techniques can just detect minority dominant population and it is difficult to study the composition, structure and diversity of microflora. In recent years, with the technical development, the high-throughput sequencing has been promoted widely, which realized the parallel comparison among multiple samples on the level of metagenome, and can detect the microbial diversity including rare species more sensitively (Zhou et al., 2011; Micucci et al., 2017; Yin et al., 2017).

However, as three typical strains of B. subtilis and B. coagulans, B. coagulans TBC169, B. subtilis PB6 and B. subtilis DSM32315, have been rarely studied as probiotics to improve the wellbeing of broiler chickens. Furthermore, little is known about the effects of B. subtilis and B. coagulans on gastrointestinal tract (GIT) microflora compositions and intestinal morphology. Therefore, the objective of this study was to evaluate the effects of B. coagulans TBC169, B. subtilis PB6 and B. subtilis DSM32315 supplementation on the growth performance, nutrient utilization and morphological development of the small intestine in broilers. The microflora compositions in the jejunum of broilers were further studied by MiSeq high-throughput sequencing to reveal the relationship among the growth performance, intestinal morphology and microflora in order to promote new evidences for the mechanism of action of these probiotics.

#### MATERIALS AND METHODS

#### Probiotics Strains

Three kinds of commercial probiotics strains were B. coagulans TBC169, B. subtilis PB6 and B. subtilis DSM32315. The probiotic product contains at least 2.0 × 10<sup>9</sup> cfu/g of Bacillus spp. and was stored in a sterilized container. The concentration of each Bacillus spp. product was 1 × 10<sup>9</sup> cfu/kg.

#### Experimental Design and Dietary Treatments

A total of 480 healthy 1-day-old male Arbor Acres (Isolauri et al., 2001) broilers (Beijing Huadu Broiler Company, Beijing, China) with average body weight of 48 g were randomly allotted into four treatments. There were six replicates (20 birds per replicate) for each treatment. The diets, without any antibiotics and growth promoters, were based on corn– soybean meal and formulated to meet starter (days 1–21) and grower–finisher (days 22–42) growth requirements (**Table 1**) (Chinese Feeding Standard of Chicken, Ministry of Agriculture of China, 2004; National Research Council, 1994). Dietary treatments consisted of basal diet with B. coagulans TBC169; basal diet with B. subtilis PB6; (3) basal diet with B. subtilis DSM32315 and the basal diet with no probiotic supplementation was set as the control. Treatments were supplemented with 200 mg/kg Bacillus spp. All experimental protocols were approved by Animal Care and Use Committee of the Feed Research Institute of the Chinese Academy of Agricultural Sciences. All management of birds in this study was according to the guideline of raising AA broilers (Delezie et al., 2012).



<sup>1</sup>The vitamin premix supplied the following per kg of complete feed: vitamin A, 12,500 IU; vitamin D3, 2,500 IU; vitamin K3, 2.65 mg; vitamin B1, 2 mg; vitamin B2, 6 mg; vitamin B12, 0.025 mg; vitamin E, 30 IU; biotin, 0.0325 mg; folic acid, 1.25 mg; pantothenic acid, 12 mg; niacin, 50 mg. <sup>2</sup>The mineral premix supplied the following per kg of complete feed: Cu, 8 mg; Zn, 75 mg; Fe, 80 mg; Mn, 100 mg; Se, 0.15 mg; I, 0.35.

#### Growth Performance

Body weight and feed intake were recorded (days 1–21 and day 22–42). Average daily feed intake (ADFI), average daily weight gain (ADG), and feed conversion ratio (FCR, feed/weight gain, g/g) were calculated.

#### Apparent Total Tract Nutrients Digestibility

All droppings (five pens for each treatment) were sampled daily for 5 consecutive days from day 37 in this study. Dry matter (DM), crude protein (CP), crude ash, calcium (Ca) and phosphorus (P) in the diet and excreta samples were analyzed according to the method of the Association of Official Analytical Chemists International, 2007. Gross energy (GE) of these samples was tested by a bomb calorimetry (Gallenkamp Autobomb, London, United Kingdom).

### Histology and Morphometric Analysis of the Intestine

On days 21 and 42, five chicks close to average weight from each treatment were killed and intestinal sections were fixed to measure the intestinal villus height (VH) and crypt depth (CD) (Sun et al., 2005).

#### Sampling, DNA Extraction and PCR (Polymerase Chain Reaction) Amplification

Four jejunum samples each treatment were selected from the five chickens slaughtered above as the next study about microbiota on days 21 and 42. The jejunum was ligated by light twine, removed and finally collected in cryogenic vials. All samples were quickly put into liquid nitrogen and stored at −80◦C until DNA extraction. The jejunum content of each group was collected and homogenized for further experiments.

Total jejunal bacterial genomic DNA was extracted from content samples by using the Fast DNA SPIN extraction kits (MP Biomedicals, Santa Ana, CA, United States) following the manufacturer's instructions. Subsequently, The quantity and quality of extracted DNAs were measured using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United States) and agarose gel electrophoresis, respectively. The DNA was used as templates to amplify the V4 hyper variable region of 16S rRNA gene by PCR using barcoded fusion primers [forward primer: 520 (5-AYTGGGYDTAAAGNG-3), reverse primer: 802 (5-TACNVGGGTATCTAATCC-3)]. Sample-specific 7-bp barcodes were incorporated into the primers for multiplex sequencing. The PCR components contained 5 µl of Q5 reaction buffer (5×), 5 µl of Q5 High-Fidelity GC buffer (5×), 0.25 µl of Q5 High-Fidelity DNA Polymerase (5 U/µl), 2 µl (2.5 mM) of dNTPs, 1 µl (10 µM) of each Forward and Reverse primer, 2 µl of DNA Template, and 8.75 µl of ddH2O. Thermal cycling consisted of initial denaturation at 98◦C for 2 min, followed by 25 cycles consisting of denaturation at 98◦C for 15 s, annealing at 55◦C for 30 s, and extension at 72◦C for 30 s, with a final extension of 5 min at 72◦C. PCR amplicons were purified with Agencourt AMPure Beads (Beckman Coulter, Indianapolis, IN, United States) and

quantified using the PicoGreen dsDNA Assay Kit (Invitrogen, Carlsbad, CA, United States). The final sequencing library was prepared by mixing the equal amount of purified PCR products, followed by an end reparation with the addition of a poly (A) tail, and the amplicons were connected with each other with the sequencing adapters.

#### MiSeq High-Throughput Sequencing and Analysis

Purified PCR products from the 31 samples were mixed with equal concentrations, which were performed using the Illumina MiSeq platform with MiSeq Reagent Kit v3 at Shanghai Personal Biotechnology Co., Ltd. (Shanghai, China). Sequencing libraries were generated and analyzed according to previous studies (Yin et al., 2017, 2018a,b).

The Quantitative Insights Into Microbial Ecology (QIIME, v1.8.0) pipeline was employed to process the sequencing data, as previously described (Caporaso et al., 2010). Briefly, raw sequencing reads with exact matches to the barcodes were assigned to respective samples and identified as valid sequences. The low-quality sequences were filtered through following criteria (Gill et al., 2006; Chen and Jiang, 2014): sequences that had a length of <150 bp, sequences that had average Phred scores of <20, sequences that contained ambiguous bases, and sequences that contained mononucleotide repeats of >8 bp. Paired-end reads were assembled using FLASH (Magoc and ˇ Salzberg, 2011). After chimera detection, the remaining highquality sequences were clustered into operational taxonomic units (OTUs) at 97% sequence identity by UCLUST (Edgar, 2010). A representative sequence was selected from each OTU using default parameters. OTU taxonomic classification was conducted by BLAST searching the representative sequences set against the Greengenes Database (DeSantis et al., 2006) using the best hit (Altschul et al., 1997). An OTU table was further generated to record the abundance of each OTU in each sample and the taxonomy of these OTUs. OTUs containing less than 0.001% of total sequences across all samples were discarded. To minimize the difference of sequencing depth across samples, an averaged, rounded rarefied OTU table was generated by averaging 100 evenly resampled OTU subsets under the 90% of the minimum sequencing depth for further analysis.

To investigate the diversity of the jejunum microbiota, alpha diversity analysis was made by using the OUT table. Diversity indexes (Shannon, Simpson) (Chao and ShenMing, 1992) were calculated. Sequence data analyses were mainly performed using QIIME and R packages (v3.2.0). OTU-level ranked abundance curves were generated to compare the richness and evenness of OTUs among samples. Beta diversity analysis was performed to investigate the structural variation of microbial communities across samples using UniFrac distance metrics (Lozupone and Knight, 2005; Lozupone et al., 2007) and visualized via principal coordinate analysis (PCoA) and non-metric multidimensional scaling (NMDS) (Ramette, 2007). The significance of differentiation of microbiota structure among groups was assessed by PERMANOVA (McArdle and Anderson, 2001) and ANOSIM (Clarke, 1993; Warton et al., 2012) using R package "vegan". Taxa abundances at the phylum, class, order, family and genus levels were statistically compared among groups by LEfSe was performed to detect differentially abundant taxa across groups using the default parameters (Segata et al., 2011). PLS-DA was also introduced as a supervised model to reveal the microbiota variation among groups, using the "plsda" function in R package "mixOmics" (Chen et al., 2011). Spearman correlation coefficients were calculated for correlation between growth performance (i.e., BW and ADG) and change of microbiota, which aimed to establish suitable microbial composition for better growth performance.

TABLE 2 | Effect of dietary probiotic supplementation on growth performance of broiler chickens1,2,<sup>3</sup> . Items<sup>4</sup> Control Bacillus coagulans TBC169 Bacillus subtilis PB6 Bacillus subtilis DSM32315 p-value Initial BW (g) 48.16 ± 0.12 47.98 ± 0.21 48.09 ± 0.15 48.08 ± 0.12 0.866 BW on day 21 (g) 918.44 ± 17.11<sup>a</sup> 986.72 ± 19.30<sup>c</sup> 963.93 ± 15.14bc 947.54 ± 7.53abc 0.035 BW on day 42 (g) 2715.49 ± 34.71<sup>a</sup> 2849.49 ± 86.35<sup>c</sup> 2786.96 ± 52.38ab 2741.20 ± 42.35ab 0.024 Starter phase (days 1–21) ADG (g) 41.48 ± 0.84<sup>a</sup> 44.70 ± 0.92<sup>c</sup> 43.61 ± 0.72ac 42.83 ± 0.36ac 0.039 ADFI (g) 56.08 ± 0.83 56.59 ± 0.73 56.97 ± 0.83 55.30 ± 1.02 0.561 FCR (F/G, g/g) 1.36 ± 0.04 1.27 ± 0.02 1.31 ± 0.02 1.29 ± 0.02 0.142 Grower phase (days 22–42) ADG (g) 85.58 ± 1.45 94.32 ± 4.17 86.81 ± 2.43 85.41 ± 1.70 0.086 ADFI (g) 149.42 ± 3.89 150.93 ± 5.96 154.22 ± 3.74 144.34 ± 3.56 0.254 FCR (F/G, g/g) 1.75 ± 0.03 1.67 ± 0.05 1.78 ± 0.02 1.69 ± 0.06 0.223 Whole phase (days 1–42) ADG (g) 63.52 ± 0.83<sup>a</sup> 69.51 ± 2.05<sup>c</sup> 65.21 ± 1.25ab 64.12 ± 1.01ab 0.024 ADFI (g) 108.25 ± 2.89 108.99 ± 4.51 107.46 ± 2.65 105.16 ± 3.67 0.590 FCR (F/G, g/g) 1.70 ± 0.03 1.61 ± 0.03 1.65 ± 0.02 1.64 ± 0.07 0.536

a,b,cValues within a row with no common superscript differ significantly (P < 0.05). <sup>1</sup>Groups of chickens fed the corresponding diet. <sup>2</sup>n = 6 replicates per treatment. <sup>3</sup>Data are the mean of six replicates with 20 birds each. <sup>4</sup>BW, body weight; ADG, average daily weight gain; ADFI, average daily feed intake; FCR, feed conversion ratio (feed:gain, g:g).

#### Statistical Analysis

fphys-09-01968 April 1, 2019 Time: 19:1 # 5

Data were analyzed by one-way ANOVA and subsequent Duncan's multiple range test (SPSS 19.0 for Windows; SPSS Inc., Chicago, IL, United States). Results are expressed as means ± SEM. Probability values of less than or equal to 0.05 (P ≤ 0.05) were considered significant, whereas a trend for a treatment effect was noted for P ≤ 0.10.

#### RESULTS

#### Growth Performance

The effects of dietary probiotics supplementation on the growth performance of broilers were shown in **Table 2**. The supplementation with B. coagulans TBC169 in feeds increased BW on days 21 and 42 (P < 0.01) and ADG (P < 0.01) during starter phase (days 1–21) and the whole phase (days 1–42) compared with the control. The supplementation with B. subtilis PB6 in diets increased BW on day 21 (P < 0.05). Results showed that compared with subtilis PB6 group and B. subtilis DSM32315 group, the addition of B. coagulans TBC169 in broiler's diets improved (P < 0.05) BW on day 42 and ADG during the overall period (days 1–42). Dietary probiotic supplementations tended to increase the ADG (P = 0.086) during grower phase (days 22–42). However, ADFI and FCR were no difference (P > 0.05) among the groups of B. coagulans TBC169, B. subtilis PB6 and B. subtilis DSM32315.

#### Apparent Total Tract Nutrient Digestibility

Dietary supplementation with different Bacillus spp. influenced (P = 0.045, P = 0.011, respectively) the apparent total tract digestibility of GE and CP during the feeding phase (**Table 3**). Supplementation with B. coagulans TBC169 and B. subtilis DSM32315 improved (P < 0.05) the apparent total tract digestibility of GE and CP compared with the control group. Supplementation with B. coagulans

.

.

TABLE 3 | Effect of dietary probiotic supplementation on the apparent total tract nutrients digestibility in broiler chickens (%)<sup>1</sup>


a, b,cMeans values in the same row with different letters differ significantly (P < 0.05). <sup>1</sup>n = 6 replicates per treatment.

TABLE 4 | Effect of dietary Bacillus spp. supplementation on the intestinal morphology of broiler chickens on the age of days 21 and 421,<sup>2</sup>


a,b,cMeans with a row with no common superscripts differ significantly (P < 0.05). <sup>1</sup>Groups of chickens fed the corresponding diet. <sup>2</sup>n = 5 replicates per treatment.

**133**

TBC169 increased (P < 0.05) the apparent total tract digestibility of CP compared with the B. subtilis PB6 group. There were no statistical differences (P > 0.05) in the apparent total tract digestibility of DM, Ca and P among groups.

#### Intestinal Morphology

The intestinal morphology of small intestine of broilers in different treatments on days 21 and 42 was shown in **Table 4** and **Figure 1**. On day 21, Dietary supplementation with Bacillus spp. tended to influence (P = 0.068) the CD of jejunum. The jejunal CD of B. coagulans TBC169 group was lower (P < 0.05) than those of the control group. Dietary supplementation with Bacillus spp. increased (P = 0.05) the VH/CD of jejunum. The jejunal VH/CD ratio of B. coagulans TBC169 group was higher (P < 0.05) than those of the control group. There were no differences (P > 0.05) among other groups. On day 42, Dietary supplementation with Bacillus spp. influenced (P < 0.05) the VH and VH/CD ratio of jejunum. The jejunal VH of B. coagulans TBC169 group and B. subtilis PB6 group were higher (P < 0.05 and P < 0.01) than that of the control group. The jejunal VH/CD ratio of B. coagulans TBC169 group and B. subtilis PB6 group were greater (P < 0.05) than that of control group. The jejunal VH of B. subtilis PB6 group were higher (P < 0.05) than those of B. coagulans TBC169 group and B. subtilis DSM32315 group. Meanwhile, there were no differences (P > 0.05) among other groups. The dietary probiotics supplementation did not affect (P > 0.05) the intestinal parameters in the ileum of broilers.

#### Impact of Bacillus spp. Supplementation on Abundance of Microbial Taxa

Based on the V4 region of the 16S rDNA sequence, 114506 amplicons were used for this study with the average of


TABLE 5 | The effect of different Bacillus spp. on OTUs of gut microbiota of broiler chickens on days 21 and 42<sup>1</sup> .

a,b,c,dMeans within a row with no common superscripts differ significantly (P < 0.05). <sup>1</sup>n = 4 replicates per treatment. <sup>2</sup>n = 3 replicates per treatment on day 21.

36936 amplicons for each sample (ranging from 30256 to 47874).

As shown in **Table 5**, probiotics increased (P < 0.004) the number of OTUs in the jejunum microbiota at five different taxonomic levels (phylum, class, order, family, genus) on days 21 and 42. On day 21, the microbial abundance in broiler chickens in two B. subtilis (B. subtilis PB6 and B. subtilis DSM32315) groups was higher (P < 0.01) than B. coagulans TBC169 group and the control group, whereas there was no difference (P > 0.05) between the B. coagulans TBC169 group or the control group at five different taxonomic levels (phylum, class, order, family, genus). Two B. subtilis strains (B. subtilis PB6 and B. subtilis DSM32315) improved (P < 0.001) the microbial abundance of jejunum in broiler chickens compared with the B. coagulans TBC169 treatment. On day 42, three Bacillus spp. (B. coagulans TBC169, B. subtilis PB6, and B. subtilis DSM32315) influenced (P < 0.004) the microbial abundance of jejunum in broiler chickens than the control group at different taxonomic levels. The microbial abundance of broiler chickens in B. subtilis DSM32315 group was lower (P < 0.01) than B. coagulans TBC169 group and the control group, whereas there was no difference (P > 0.05) between two B. subtilis strains. The curves of OTU rank and rarefaction were calculated. The rarefaction curves showed that the total richness of the microbial community of all samples achieved a high sampling coverage (**Figure 2**).

The PCoA and NMDS plot of the jejunum microbiota were based on the weighted UniFrac metric measured as Adonis (P = 0.002 on day 21 and P = 0.012 on day 42) and Anosim (P = 0.003 on day 21 and P = 0.019 on day 42). These methods showed differences of chickens' microbiota with different probiotics supplementation compared to that of the control (**Figure 3**). The

compositions of the jejunal microbiota of the probiotics supplemented chicken were significantly different in comparison to that of the jejunal microbiota of the control.

#### Impact of Bacillus spp. Supplementation on Microbial Diversity

Alpha diversity (sample OTU richness) including Simpson index and Shannon index (**Table 6**) was measured to detect the diversity and structure of jejunal microbial communities with different Bacillus spp. supplementations. According to Shannon and Simpson index, the number of B. subtilis PB6 and B. subtilis DSM32315 was higher (P < 0.032) than the control on day 21, which illustrated that the jejunum microbial diversity of broiler chickens was greater than the control. But, according to Shannon index, the number of B. subtilis PB6 and B. subtilis DSM32315 was higher (P < 0.05) than B. coagulans TBC169 group. By contrast, on day 42, the figure of the control was the most compared with other treatments, and that was higher (P < 0.05) than the supplementation with B. subtilis DSM32315.

#### Impact of Bacillus spp. on the Number of Indicated Taxonomic Rank

At different taxonomic levels (phylum, class, order, family, and genus), B. subtilis PB6 and B. subtilis DSM32315 increased (P < 0.01) the number of indicated taxonomic rank on day 21, whereas that decreased (P < 0.01) the number of indicated taxonomic rank on day 42. The number of indicated taxonomic rank of B. coagulans TBC169 group showed no significant differences (P > 0.05) compared with the control group on both days 21 and 42 (**Table 7**).

#### Impact of Bacillus spp. Supplementation on the Gut Microbial Compositions

As shown in **Figure 4**, the phylum level analysis demonstrated that three kinds of Bacillusspp. significantly (P < 0.05) influenced the percentage of Firmicutes, Proteobacteria, Bacteroidetes, Actinobacteria, and Acidobacteria on 21 days. The control and all treatments held the largest share in Firmicutes at around 50% (**Figure 3**). Meanwhile, the percentage of Firmicutes was higher (P < 0.01) in the B. coagulans TBC169 group (70.10%)

.


TABLE 6 | The effect of different sources of Bacillus spp. on diversity index of jejunum microbiota of broiler chickens on days 21 and 42<sup>1</sup>

a,b,cMeans within a row with no common superscripts differ significantly (P < 0.05). <sup>1</sup>n = 4 replicates per treatment. <sup>2</sup>n = 3 replicates per treatment on day 21.


TABLE 7 | The effect of different sources of Bacillus spp. on the number of indicated taxonomic rank of broiler chickens on days 21 and 42<sup>1</sup> .

a,b,c,dMeans within a row with no common superscripts differ significantly (P < 0.05). <sup>1</sup>n = 4 replicates per treatment. <sup>2</sup>n = 3 replicates per treatment on day 21.

than that in the control group (47.08%). As for the rest phyla (Proteobacteria, Bacteroidetes, and Actinobacteria), dietary supplementation with B. subtilis PB6 and B. subtilis DSM32315 group were higher (P < 0.05) than both the control and B. coagulans TBC169 group, and there were no differences (P > 0.05) between the control and B. coagulans TBC169 group. Lactobacillus, which belongs to the phylum of Firmicutes, was the most abundant genus in chicken jejunal microbiota with three kinds of Bacillusspp. supplementation, followed by Bacillus, Lactococcus, and Bacillaceae on day 21. Interestingly, Three kinds of Bacillus spp. tended to decrease (P = 0.077) the percentage of Bacillus (phylum of Firmicutes), which was the most abundant genus in the microbiota of chicken jejunum in the control group at the same situation. Dietary supplementation with B. subtilis DSM32315 improved (P < 0.01) the composition of Clostridiales and Pseudomonas compared control group and B. coagulans TBC169 group at genus level.

The composition of jejunal microflora obviously changed with time. On day 42, the percentage of the phylum of Firmicutes dropped compared with on day 21, but it still occupied the dominant position in B. subtilis PB6 and B. subtilis DSM32315 groups, followed by Proteobacteria and Acidobacteria (**Figure 5**). But, the percentage of phylum of Proteobacteria occupied the dominant position in control group and B. coagulans TBC169 group, followed by Firmicutes and Acidobacteria. Similarly, Lactobacillus was also the most abundant genus in jejunal microbiota of chicken with three kinds of Bacillus spp. supplementation on day 42. On the ability of decreasing the percentage of Pseudomonas and Burkholderia, B. subtilis PB6 and B. subtilis DSM32315 were stronger than B.subtilis TBC169, whereas B.subtilis DSM32315 was stronger than B. subtilis TBC169 and B. subtilis PB6 in decreasing the percentage of DA101. The supplementation with probiotics B. subtilis DSM32315 decreased (P < 0.05) the percentage of Pseudomonas, Burkholderia, Prevotella, DA101 than control group, whereas the supplementation with probiotics B. subtilis PB6 decreased (P < 0.05) the percentage of Burkholderia, Prevotella, DA101 than control group.

A cladogram representative of the structure of the jejunum microbiota and the predominant bacteria were shown in **Figure 6**. LEfSe detected a marked increase (LDA score > 2) in the relative abundance among the four treatment groups. PLS-DA analysis showed some significant differences in the bacterial composition at genus levels among four treatment groups (**Figure 7**). It shows that bacterial composition on day 21 (**Figure 7A**) and 42 (**Figure 7B**) exhibiting the tendency of separation in the profiles among the four treatments and indicating the degree of reliability of PCoA analysis.

Data (**Figure 8**) from spearman correlation coefficients showed that change of Firmicutes have significantly (P < 0.05) positive correlations with growth performance (i.e., BW and ADG), whereas Armatimonadets, Chlorobi, and Cyanobacteria had significantly (P < 0.05) negative correlations with BW and ADG at the phylum level on day 21. The changes of

Macrococcus, Clostridium and Brevundimonas have positive (P < 0.05) correlations with BW and ADG, whereas Rhodococcus, Phenylobacterium, Neisseria, Azoarcus, Idiomarina, Erwinia, B-42 have significantly negative (P < 0.05) correlations with BW and ADG at the genus level on day 21. Agromyces, Cryocola, JG37-AG-70, Pirellula, Beijerinckia, Amaricoccus, Plesiocystis, Sorangium, Pseudoxanthomonas, Leptonema, and Thermus have positive (P < 0.05) correlations with BW and ADG at the genus level on day 42.

# DISCUSSION

Due to the high biological safety and beneficial functions in the regulation of intestinal microflora and micro-ecological balance, B. subtilis and B. coagulans have been widely used in the field of animal feed (Selvam et al., 2009; Hung et al., 2012; Pandey and Vakil, 2016). Previous studies indicated that the addition of B. subtilis and B. coagulans as probiotics to broilers' diets can obviously improve the growth performance and FCR (Hung et al., 2012; Jeong and Kim, 2014; Lee K.-W. et al., 2014; Hossain et al., 2015; Li et al., 2016). In this study, B. subtilis PB6, B. subtilis DSM32315 and B. coagulans TBC169 were used for probiotic supplementations in chickens' diets. Although there was no significant difference in ADFI between birds fed diet with probiotics and the control group at the age of 3rd and 6th week, the ADG and BW of birds fed with B. coagulans TBC169 diets were higher than that of chickens fed with the basal diets (**Table 2**) (P < 0.05). Generally, the reason why the addition of Bacillus spp. can improve the growth performance of broilers is that these probiotics may regulate the composition of intestinal microflora, and exert antibiotics and organic acids to inhibit the survival and colonization of harmful bacteria and further facilitate the balance of intestinal micro-ecosystem (Pedroso et al., 2006). However, the supplementation with B. subtilis did not affect the growth performance of chickens including the indexes of ADFI, ADG, and FCR (**Table 2**) (P > 0.05). In addition, there were also reports that dietary B. subtilis failed to improve ADG and decrease FCR in broiler chickens (Knap et al., 2011; Lee K.-W. et al., 2014). The reasons such as the bacterial strains, animal conditions, methods of using probiotics and environmental factors may also contribute to these results (Landy and Kavyani, 2013; Lee S.H. et al., 2014).

In addition, another characteristic of Bacillus spp. is that they can produce various digestive enzymes such as proteases, celluloses and amylases etc. and promoting the transformation of pepsinogen into protease after colonizing in intestinal tract of hosts. They can further stimulate the peristalsis of the intestine to improve nutritional digestion (Giang et al., 2011), which is related to the improvement of broilers' growth performance. A xylanase gene from the chicken caecum has been isolated and over expressed which focused on the potential development of novel and optimized feed additives for industrial use (Al-Darkazali et al., 2017). B. coagulans can produce bacteriocins, which have been widely reported to exhibit antibacterial function in various models (Riazi et al., 2009; Nobutani et al., 2017). Meanwhile,

B. coagulans is associated with lactic acid and other organic acids production, while lactic acid can inhibit gut colonizations of harmful bacteria (Cui et al., 2005) and activate the peristalsis in the intestine to mediate nutrient digestion (Giang et al., 2010; Ma et al., 2014). More recently, B. coagulans-produced dysprosium has been identified to exhibit broad antibacterial spectrum (Honda et al., 2011). Also, B. coagulans can maintain gut microbiota balance via breaking down polysaccharides into oligosaccharides (Zheng et al., 2011). The apparent digestibility of CP and GE were higher for birds in the B. coagulans TBC169 treatment group compared with those in the control (P < 0.05) in the present study. The ability of improving the apparent total tract digestibility of CP, B.subtilis TBC169 was even stronger than B. subtilis PB6. Dietary B. subtilis DSM32315 supplementation influenced the apparent digestibility of GE (**Table 3**) (P = 0.05). This was in accordance with the findings of Pelícia et al. (2004) who suggested that the improvement in broiler growth performance was likely linked to a better ileal digestibility of nutrients. Similar research reported the increasing levels of dietary supplementation with B. subtilis LS 1–2 product was linked to the improvement in the retention of DM, GE, and CP in broilers (Sen et al., 2012). The dietary supplementation with Bacillus spp. can improve GE and CP digestibility, which is highly associated with gut health and subsequent digestive capacity. The increased VH and VH/CD ratio were directly correlated with an enhanced epithelial turnover (Sen et al., 2012) and high VH and VH/CD ratio suggested an improved intestinal nutrient digestibility and absorption capacity (Montagne et al., 2003). The present study showed that B. coagulans TBC169 increased (P < 0.07) VH or VH/CD ratio mainly in jejunum compared with the control diet at different phases (**Table 4**). The B. coagulans TBC169 supplementation might improve the gut structure and further resulted in a greater absorption surface and high-speed turnover of epithelial cells. Similarly, an increase of VH/CD ratio was observed in broilers fed with B. coagulans (Hung et al., 2012). This situation was also consistent with the FCR and BW results in the B. coagulans TBC169 group, which indicated the positive growth performance might be associated with good intestinal morphology. However, at the same condition, the dietary B. subtilis PB6 supplementation improved VH and VH/CD ratio only in jejunum compared with the control diet on day 42. The dietary B. subtilis DSM32315 supplementation did not affect the structure of intestine compared with the control on days 21 and 42 (**Table 4**). Therefore, different Bacillus spp. has distinguished effects on stimulating the differentiation and proliferation of intestinal epithelial cells and improving nutrient utilization.

This study found that B. subtilis PB6 and B. subtilis DSM32315 more strongly affected the number of indicated taxonomic rank in jejunum microflora than the control or B. coagulans TBC169 group (**Table 7**). Furthermore, it is generally agreed that the improvement in growth performance and feed conversion efficiency rely on a healthy intestinal morphology, which

may be related to a balance of entire intestinal microflora, resulting in a better intestinal environment (Pedroso et al., 2006; Mountzouris et al., 2010). High-throughput sequencing of the V4 region of the 16S rRNA gene was used for detecting the jejunum microbiota of individual broiler chickens fed diets without or with Bacillus spp. supplementation in the current study. B. subtilis PB6 and B. subtilis DSM32315 treatments exhibited more OTUs than the control on day 21, but the OUTs were higher (P < 0.005) in the control compared with other treatments on day 42 (**Table 5**). In addition, the number of indicated taxonomic rank varied in accordance with OUT variation (**Table 7**). As the researches exploring the relationship between the growth performance and intestinal microbial composition have become more popular, recent

reports showed that alpha index was strongly correlated with diversity (Martiny et al., 2011). Many studies suggested that gut microbiota with high diversity could be more stable and healthier than those with low diversity (Konstantinov et al., 2004). Simpson and Shannon indexes can reflect the community diversity of the microbiota. In this study, Simpson and Shannon indexes showed the similar trend to the number of OTUs (**Table 6**). B. subtilis PB6 and B. subtilis DSM32315 treatments enhanced the diversity of jejunal microbiota on day 21 compared with the control (P ≤ 0.032). Diet probiotic supplementation influenced diversity of the microbiota according to Simpson and Shannon indexes (P = 0.071 and P = 0.018) on day 42. Simpson and Shannon index are respond to Bacillus spp. were different on days 21 and 42. The number of Shannon in the treatments with B. subtilis DSM32315 was lower (P < 0.05) than that in the control. Nevertheless, B. coagulans TBC169 strains did not influence the diversity of jejunal microbiota on days 21 and 42. Changes in taxonomic diversity are the most used indicator to infer changes in microbiological activity, but it is becoming apparent that many of the functions of a normal microbiome can be carried out by a number of microbial groups (Kurokawa et al., 2007; Human Microbiome Project Consortium, 2012). Many studies have shown that the microbial diversity of the chicken microbiota is relatively lower compared to the intestinal microbiota of other animals, which is attributed to the rapid transit of food through the digestive system, with short retention times (Wei et al., 2013). Microbial diversities increased during chicken development, reaching at the peak approximately on day 14 for the foregut and then remaining stable or decreasing slightly thereafter (Huang et al., 2018). Data from a previous study suggest a microbiome more affected by age rather than treatment (Ballou et al., 2016).

The complex microbiota colonized in a chicken' GIT with more than over 900 different bacterial species (Torok et al., 2011). Basically, the most abundant phylum of the chicken intestinal microbiota is Firmicutes (44–55%), followed by Proteobacteria and Bacteroidetes, which is consistent with the present study (**Figures 3**, **4**) (Qu et al., 2008). However, different sections GIT of chickens are highly interconnected and thus also influence each other's microbiota compositions (Sklan et al., 1978). The composition and function of these communities has been shown to vary depending on the age of the birds, location in the GI tract and on the dietary components (Oakley et al., 2014; Pan and Yu, 2014; Oakley and Kogut, 2016). In addition, the variability in results may be due to sample types (feces vs. cecum), and/or conventional microbiological and molecular methods that have limited coverage and accuracy. The jejunum of a chicken was dominated by Lactobacillus species, mainly L. salivarius and L. aviarius (Gong et al., 2007; Zotta et al., 2017). According to a previous report, at the genus level, Lactobacillus is the predominant genus in the foregut, Lactobacillus provides nutrients to the host and defends against opportunistic pathogens (Huang et al., 2018). Nutrient absorption mainly occurs in the intestine where occupies by a large number of Lactobacillus spp. (Witzig et al., 2015). Various trials have demonstrated that probiotics can positively modulate the composition of the gut microflora of chickens via the stimulation of potentially beneficial populations and the reduction of potentially pathogenic bacteria (Higgins et al., 2008; Hussain et al., 2017b). Dietary inclusion of B. subtilis PB6 and B. subtilis DSM32315 mainly decreased the number of Bacillus genus bacterial and increased lactobacilli population both on days 21 and 42 compared with the control (**Figures 3**, **4**). This phenomenon likely illustrated that the three probiotic treatments can regulate the intestinal bacterial flora by increasing the quantity of beneficial bacterial such as lactobacilli

performance (BW and ADG) on day 21 (A) and day 42 (B).

and decreasing the number of harmful bacterial like coli bacillus etc. Study have also shown that Bacillus spp. regulated the composition of intestinal bacterial flora, maintained the balance of GIT microflora and improved the immune function of the intestinal mucosa (Isolauri et al., 2001; Barbosa et al., 2005).

Animals with high FCR exhibited a higher abundance of Acinetobacter, Bacteroides, Streptococcus, Clostridium and Lactobacillus whereas Escherichia, Shigella, and Salmonella were more abundant in low FCR animals (Singh et al., 2014). Although the supplementation with probiotics B. subtilis DSM32315 increased (P < 0.05) the percentage of Clostridiales than control group on day 21, which did not affect the FCR in the present study. Individual FCR in broiler growers appears to vary, which may in part be due to variation in their gut microbiota. In this paper we analyzed the jejunal microbiota whereas most of the other studies analyzed fecal and cecal microorganisms. Clostridium spp., Enterococcus spp., Streptococcus spp., and Bacteroides spp. were shown to have polysaccharide degrading activity against non-starch polysaccharide (NSPs) found in grain (Beckmann et al., 2006). In mice and humans, Firmicutes have been shown to have a positive relationship with the ability to harvest energy from the diet (Turnbaugh et al., 2006; Jumpertz et al., 2011), and the Firmicutes/Bacteroides ratio may also be important for optimum physiology and nutrition (Bervoets et al., 2013). An increase in fecal Firmicutes was associated with an increase in nutrient absorption, whereas an increase in fecal Bacteroidetes was associated with a decrease in nutrient absorption (Jumpertz et al., 2011). Similar experiments on pig and poultry showed that the increase of Firmicutes abundance and the decrease of Bacteroidetes abundance were positively correlated with the increase of host weight and fat deposition, indicating that intestinal microorganisms were related to the absorption and utilization of energy by the host (Angelakis and Raoult, 2010). In the lowest portion of the small intestine, Lactobacillus spp. have been implicated as a causal factor in low performance (DeLange and Wijtten, 2010), suggesting the location of colonization by probiotic strains may affect performance.

There were many consistencies among the indexes determined in this study. Dietary supplementation with three probiotics

#### REFERENCES


mostly showed the beneficial effects on chicken's growth and intestinal health. Among that, the B. coagulans TBC169 group showed the better growth performance, nutrients digestibility and intestinal morphology compared with the two B. subtilis treatments, while two B. subtilis treatments presented more positive variation of the jejunum microflora of chickens than that in the B. coagulans TBC169 group (**Figure 7**). The possible explanations might be the different characteristics of different strains, the environmental and individual differences etc. In addition, the investigation of jejunum microflora of chickens might not stand for the whole gut microbiota conditions. Therefore, the internal relationships and underlying mechanism are worth further exploring.

#### CONCLUSION

Dietary addition of three probiotic Bacillus spp. strains affect body weight and intestinal morphology through altering intestinal microbiota composition in broiler chickens. The findings highlight the importance of intestinal microbiota in mediating the various physiological functions of probiotics in the host. However, the effect of different strains of Bacillus on intestinal microbial composition was different.

#### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

#### FUNDING

This research was financed by the National Key R&D Program of China (2018YFD0500600 and 2018YFD0501401), China Agriculture Research System-Beijing Team for Poultry Industry (CARS-PSTP, Beijing, China), and the Agricultural Science and Technology Innovation Program (ASTIP) of the Chinese Academy of Agricultural Sciences.



454 pyrosequencing. PLoS One 10:e0143442. doi: 10.1371/journal.pone.014 3442


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Li, Wang, Zhang, Wu, Hui, Yang, Fang and Qi. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Corrigendum: Intestinal Morphologic and Microbiota Responses to Dietary Bacillus spp. in a Broiler Chicken Model

Cheng-liang Li 1,2, Jing Wang<sup>2</sup> , Hai-jun Zhang<sup>2</sup> , Shu-geng Wu<sup>2</sup> , Qian-ru Hui <sup>3</sup> , Cheng-bo Yang<sup>3</sup> , Re-jun Fang<sup>1</sup> \* and Guang-hai Qi 1,2 \*

#### Approved by:

*Frontiers in Physiology, Frontiers Media SA, Switzerland*

#### \*Correspondence:

*Re-jun Fang fangrj63@126.com Guang-hai Qi qiguanghai@caas.cn*

#### Specialty section:

*This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology*

Received: *08 March 2019* Accepted: *11 March 2019* Published: *02 April 2019*

#### Citation:

*Li C-l, Wang J, Zhang H-j, Wu S-g, Hui Q-r, Yang C-b, Fang R-j and Qi G-h (2019) Corrigendum: Intestinal Morphologic and Microbiota Responses to Dietary Bacillus spp. in a Broiler Chicken Model. Front. Physiol. 10:332. doi: 10.3389/fphys.2019.00332* *<sup>1</sup> College of Animal Science and Technology, Hunan Agricultural University, Changsha, China, <sup>2</sup> Key Laboratory of Feed Biotechnology of Ministry of Agriculture and Rural Affairs, Feed Research Institute, Chinese Academy of Agricultural Sciences, Beijing, China, <sup>3</sup> Department of Animal Science, Faculty of Agricultural and Food Sciences, University of Manitoba, Winnipeg, MB, Canada*

Keywords: probiotics, growth performance, intestinal morphology, jejunum microbiota, broiler

#### **A Corrigendum on**

#### **Intestinal Morphologic and Microbiota Responses to Dietary Bacillus spp. in a Broiler Chicken Model**

by Li, C-l., Wang, J., Zhang, H-j., Wu, S-g., Hui, Q-r., Yang, C-b., et al. (2019). Front. Physiol. 9:1968. doi: 10.3389/fphys.2018.01968

In the original article, there was a mistake in **Figure 1** as published. Due to poor image quality a new image was prepared during the production stage. Unfortunately, the incorrect image was uploaded and used in the published article. The corrected **Figure 1** appears below.

Additionally, there was a mistake in **Figure 6** as published. We have uploaded **Figure 6A** was uploaded during production of the article instead of **Figure 6**. Thus, **Figure 6** should include both **Figures 6A,B**. The corrected **Figure 6** appears below.

The authors apologize for this error and state that they do not change the scientific conclusions of the article in any way. The original article has been updated.

Copyright © 2019 Li, Wang, Zhang, Wu, Hui, Yang, Fang and Qi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Dietary Supplementation With High Fiber Alleviates Oxidative Stress and Inflammatory Responses Caused by Severe Sepsis in Mice Without Altering Microbiome Diversity

Yuanyuan Zhang\*, Aili Dong, Keliang Xie and Yonghao Yu\*

Department of Anesthesiology, Tianjin Medical University General Hospital, Tianjin, China

#### Edited by:

Jie Yin, Institute of Subtropical Agriculture (CAS), China

#### Reviewed by:

Lei Sun, The George Washington University, United States Hengjia Ni, Institute of Subtropical Agriculture (CAS), China Xingyu Liu, General Research Institute for Nonferrous Metals (China), China

#### \*Correspondence:

Yuanyuan Zhang yzhang10@tmu.edu.cn Yonghao Yu zyymzkzyy@126.com

#### Specialty section:

This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology

Received: 31 October 2018 Accepted: 21 December 2018 Published: 18 January 2019

#### Citation:

Zhang Y, Dong A, Xie K and Yu Y (2019) Dietary Supplementation With High Fiber Alleviates Oxidative Stress and Inflammatory Responses Caused by Severe Sepsis in Mice Without Altering Microbiome Diversity. Front. Physiol. 9:1929. doi: 10.3389/fphys.2018.01929 In this study, we demonstrated the effects of a high-fiber diet on intestinal lesions, oxidative stress and systemic inflammation in a murine model of endotoxemia. C57BL/6 mice were randomly assigned to four groups: the control group (CONTROL), which received a commercial normal-fiber rodent diet comprising normal fiber; a CLP group, which received a commercial normal-fiber rodent diet and underwent caecal ligation puncture (CLP); a high-fiber group (HFG), which received a commercial high-fiber rodent diet; and a high fiber + CLP group (HFCLP) which received a commercial high-fiber rodent diet and underwent CLP (30%). The sepsis model was created via CLP after 2 weeks of dietary intervention. Notably, dietary highfiber supplementation in HFCLP group improved survival rates and reduced bacterial loads, compared with CLP alone. In the HFCLP group, dietary fiber supplementation decreased the serum concentrations of pro-inflammatory cytokines such as tumor necrosis factor-α (TNF-α), interleukin 6 (IL-6) and high-mobility group protein 1 (HMG-1) but raised the concentration of interleukin 10 (IL-10), compared with the levels in CLP mice. Meanwhile, high-fiber supplementation increased the relative proportions of Akkermansia and Lachnospiraceae. These data show that dietary high-fiber supplementation may be therapeutic for sepsis-induced lesions.

#### Keywords: fiber, sepsis, mitochondrion, HO-1, Nrf2

# INTRODUCTION

In 2011, more than \$20 billion in hospital costs in the United States were attributed to sepsis (Singer et al., 2016), a life-threatening organ pathology caused by a dysregulated host response to infection. Sepsis pathogenesis is typically classified as an initial pro-inflammatory phase, followed by an anti-inflammatory or immunosuppressive phase (Haak et al., 2018; Shankar Hari and Summers, 2018). During the last 30 years, researchers have investigated a number of unsuccessful immunotherapeutic strategies aimed at circumventing the unregulated pro-inflammatory host response during the initial phases of sepsis. However, most of these strategies focused on the cascade of pro-inflammatory cytokines, including tumor necrosis factor α (TNF-α), interleukin (IL)-1 and high-mobility group box 1 (HMGB1), which have all been shown to be of little practical therapeutic value.

Dietary modifications can affect systemic inflammation via changes in the gut microbiota (Kau et al., 2011; Tilg and Moschen, 2015). Generally, fiber is classified as either 'fermentable' or 'non-fermentable' (i.e., resistant), and studies have investigated the anti-inflammatory properties and mechanisms of the former type (Wedlake et al., 2014; Simpson and Campbell, 2015). By contrast, the protective anti-inflammatory properties of cellulose, a non-fermentable fiber, have yet to be elucidated. Dietary fiber has well-documented anti-inflammatory characteristics, which can be partly attributed to fiber-induced actions on the gut microbiota (Kuo, 2013; Simpson and Campbell, 2015). Morowitz et al. (2017) demonstrated that the benefits associated with dietary cellulose intake correlate with enrichment of the gut microbiome taxon Akkermansia, a genus typically associated with improved metabolic health. This finding led us to hypothesize that supplementation with cellulose would enhance survival in murine sepsis models by reducing intestinal lesions, modulating oxidative stress and reducing systemic inflammation.

In this investigation, we demonstrated the effects of a highfiber diet on intestinal lesions, oxidative stress and systemic inflammation in a murine model of endotoxemia.

# MATERIALS AND METHODS

#### Use and Care of Animals

All animal investigations were approved by the Tianjin Medical University General Hospital, Tianjin, China. Animals were cared for in accordance with the Chinese guidelines for animal use and treatment. C57BL/6 mice were randomly assigned to four groups (n = 20 each): control (CONTROL), which was fed a commercial normal-fiber rodent diet (5% cellulose); CLP, which was also fed a commercial normal-fiber rodent diet and underwent caecal ligation and puncture (CLP); high-fiber (HFG), which was fed a commercial high-fiber rodent diet (30% cellulose); and high-fiber + CLP (HFCLP), which was fed a commercial high-fiber rodent diet and subjected to CLP. The mouse weights were monitored daily. After a 2-week dietary intervention, a sepsis model was created by CLP according to our previous report (Yu et al., 2017). Mice in all groups were subjected to hypodermic peritoneal injection with 1 mL of a 0.9% saline solution immediately after the operation. The resulting lavage fluid was serially diluted with sterile saline, and 100 µL aliquots of the dilutions were placed on agar plates and incubated at 37◦C for 16 h. Subsequently, colony-forming units (CFUs) in the samples of peritoneal lavage fluid were calculated in accordance with previous studies (Morowitz et al., 2017). Serum, tissue and fecal samples were stored at −80◦C for further analysis.

#### Morphology Analysis of Intestinal Tissue

The small intestines of all mice were fixed in 10% paraformaldehyde, embedded in paraffin and stained with haematoxylin and eosin (HE). The disease scores were then rated by two pathologists who were blind to the experimental design and grouping to assess the extent of intestinal lesions (Shrum et al., 2014; Yu et al., 2017).

# Measurement of Oxidative Products, Antioxidant Enzymes and Inflammatory Cytokines

Twenty-four hours postoperatively, 10-mL blood samples were collected from 8 mice per group and centrifuged at 3500 rpm for 8 min. Subsequently, serum samples were collected and stored at −80◦C, after which the levels of oxidative products (e.g., malondialdehyde [MDA] and 8-iso-15(S)-prostaglandin F2α [8-iso-PGF2α]) were detected using a commercial kit (Nanjing Jiancheng Bio Co., Ltd., Nanjing, China). Catalase (CAT) and superoxide dismutase (SOD) activities in the sera were also detected using kits (Nanjing Jiancheng Co., Ltd., Nanjing, China) according to the manufacturer's instructions. Enzyme-linked immunosorbent assay (ELISA) kits were used to determine the serum concentrations of TNF-α, IL-6, IL-10 (R&D Systems) and HMGB1 (Nanjing Jiancheng Co., Ltd., Nanjing, China) in accordance with the manufacturers' instructions (Yu et al., 2017).

# Real-Time Quantitative PCR

The levels of nuclear factor (erythroid-derived 2)-like 2 (Nrf2) and heme oxygenase-1 (HO-1) mRNA were detected using real-time quantitative PCR. Expression of Gapdh mRNA was used as a reference. The following gene-specific primer sequences were used: Nrf2-F 5<sup>0</sup> -CGACAGAAACCTCCATCTACTGAA-3<sup>0</sup> , Nrf2-R 5<sup>0</sup> -CCTCATCACGTAACATGCTGAAG-3<sup>0</sup> ; HO-1-F 5<sup>0</sup> - ACAGATGGCGTCACTTCG-3<sup>0</sup> , HO-1-R 5<sup>0</sup> -TGAGGACCCAC TGGAGGA-3<sup>0</sup> ; GAPDH-F 5<sup>0</sup> -CATCACTGCCACCCAGAAG AC-3<sup>0</sup> , GAPDH-R 5<sup>0</sup> -CCAGTGAGCTTCCCGTTCAG-3<sup>0</sup> (Yu et al., 2017).

#### Sequencing and Analysis of Bacterial 16S rRNA Genes

Total DNA was extracted from fecal samples and purified, and the V4 regions of 16S rRNA genes were amplified using specific primers (515F-806R). All PCR reactions were performed using Phusion <sup>R</sup> High-Fidelity PCR Master Mix (New England Biolabs, Ipswich, MA, United States). Sequencing was performed on an Illumina MiSeq device (Illumina, Inc., San Diego, CA, United States), and QIIME (v1.9) was used to demultiplex the raw sequence reads. UPARSE (v8.0) was then used to filter the reads for quality. Sequences with a similarity >97% were classified in the same operational taxonomic unit (OTU). UCLUST and Greengenes reference database (v13.8) were then used to assign taxonomies to the predicted OTUs. Alpha diversity and QIIME (Version 1.7.0) were used to analyze the complexity of each sample.

#### Statistical Analysis

Mouse survival rates are expressed as percentages (%). Other data are presented as means ± standard deviations (SDs). The log-rank (Mantel–Cox) test was used to evaluate differences in survival rates between the groups; the unpaired t-test or Mann– Whitney test was also used if the results were approximately normally distributed (e.g., Gaussian distribution) or not normally distributed, respectively. A P-value <0.05 was considered to indicate a statistically significant difference. The statistical

analyses were performed using SPSS, version 21.0 (IBM Corp., Armonk, NY, United States).

# RESULTS

### Survival Rate and Bacterial Load

**Figure 1** suggests that a minimal number of mice died in each of these groups, the survival rates in the CLP and HFCLP mouse groups decreased significantly (P < 0.05). However, the mouse groups subject to dietary high-fiber supplementation exhibited an enhanced sepsis survival rate (P < 0.05). We further analyzed the CFUs in peritoneal lavage fluid and found significantly higher numbers in the CLP and HFCLP groups (P < 0.05). Notably, the HFCLP group exhibited a marked reduction in CFUs, compared to the CLP group (P < 0.05) (**Figure 2**).

#### Small Intestinal Morphology and Disease Scores

To determine the severity of intestinal lesions, the small intestines were subjected to HE staining, and appropriate histopathological scores were assigned to rate the severities of the observed intestinal injuries (**Figure 3**). In the CONTROL and HFG groups, the intestinal mucosa did not exhibit any abnormal morphological changes. However, shortening and atrophy of the intestinal mucosal villi were observed in the CLP and HFCLP groups. The Intestinal lesions were less severe in the HFCLP

group than in the CLP group (P < 0.05). In addition, the intestinal disease scores of mice in the CLP and HFCLP groups were much higher than those in the CONTROL and HFG groups (P < 0.05).

# Oxidative Products and Antioxidative Enzymes

**Figure 4** indicates that the levels of MDA and 8-iso-PGF2α were higher in the CLP and HFCLP groups than in the CONTROL and HFG groups (P < 0.05). Meanwhile, the levels of both oxidative products were lower in the HFCLP group than in the CLP group (P < 0.05). However, the activities of the anti-oxidative enzymes CAT and SOD were lower in the CLP and HFCLP groups than in the CONTROL and HFG groups (P < 0.05). Moreover, the activities of both enzymes were higher in the HFCLP group than in the CLP group (P < 0.05).

#### Serum Inflammatory Cytokines

Next, the serum levels of inflammatory cytokines were investigated. **Figure 5** demonstrates that the serum concentrations of TNF-α, IL-6 and HMGB1 were significantly increased in the CLP group, compared to the CONTROL and HFG groups (P < 0.05). However, the dietary fiber supplementation provided to the HFCLP group reduced TNF-α, IL-6 and HMGB1 levels markedly, compared to those in the CLP group (P < 0.05). The CLP and HFCLP groups exhibited significantly higher serum IL-10 concentrations relative to those in the CONTROL and HFG groups (P < 0.05). Furthermore, the serum IL-10 level was significantly lower in the CLP group than in the HFCLP group (P < 0.05).

# HO-1 and HMGB1 Expression

The levels of HO-1 mRNA were significantly higher in the CLP and HFCLP groups than in the CONTROL and HFG groups (P < 0.05). Furthermore, the level of HO-1 mRNA was higher in the HFCLP group, compared to the CLP group (P < 0.05) (**Figure 6**). Similarly, the levels of Nrf2 mRNA were significantly higher in the CLP and HFCLP groups than in the CONTROL and HFG groups (P < 0.05). However, no significant difference in this transcript was observed between the CLP and HFCLP groups.

# Microbial Diversity in the Fecal Samples

Bacterial 16S rRNA gene sequencing was used to profile the gut microbiota in mice from each group. Notably, no statistical differences in alpha diversity were observed among the four groups (**Table 1**). However, significant between-group differences were observed in the community compositions of the fecal samples. In particular, samples collected from the HFG and HFCLP groups contained highly abundant bacteria from the family Lachnospiraceae, which are typically associated with a healthy colon (P < 0.05). Meanwhile, the relative abundance of Akkermansia, a bacterial genus with known health-enhancing characteristics, was significantly higher in the HFG and HFCLP groups (P < 0.05) (**Figure 7**).

# DISCUSSION

Septic shock is a frequent cause of mortality in critical patients (Mayr et al., 2014; Rickard et al., 2014). Previous reports have shown that sepsis morbidity might result from an extreme pro-inflammatory response and/or extreme anti-inflammatory response which generates a state of immunosuppression (Mayr et al., 2014; Rickard et al., 2014). In this investigation, we successfully generated a sepsis model and demonstrated that dietary high-fiber supplementation led to an improved survival rate with lower bacterial loading, compared to CLP treatment alone. In addition, supplementation with a high-fiber dietalleviated intestinal lesions and oxidative injuries, thereby enhancing survival and reducing the serum levels of

FIGURE 6 | Effects of dietary fibre supplementation on the levels of (A) HO-1 and (B) Nrf2 mRNA. <sup>∗</sup>P < 0.05 vs. the CONTROL group, #P < 0.05 vs. the HFG group and &P < 0.05 vs. the CLP group.

pro-inflammatory cytokines in a CLP-induced murine sepsis model.

The generation of pro- and anti-inflammatory mechanisms has been suggested to represent a vital stage in sepsis survival


(Munford and Pugin, 2001; Rickard et al., 2014). In our study, mice with severe CLP-induced sepsis in the CLP and HFCLP groups exhibited more severe intestinal injuries, compared to untreated mice. Excessive cytokines secretion and elevated oxidative species levels underpin the pathogenesis of sepsis (Xie et al., 2014). To observe the effects of dietary fiber supplementation on intestinal lesions induced by severe sepsis, the concentrations of inflammatory factors (e.g., pro- and antiinflammatory cytokines) were monitored. Similarly, dietary fiber supplementation was shown to reduce the concentrations of proinflammatory cytokines, including TNF-α, IL-6 and HMGB1, and increase the concentration of IL-10 in sera from HFCLP mice relative to sera from the CLP group. Previous investigations have demonstrated that HMGB1 is a useful marker of severe sepsis,

and several reports have demonstrated that once activated and secreted into the extracellular milieu, this cytokine can mediate sepsis-related inflammatory responses (Wang et al., 2014; Stevens et al., 2017). Consistent with earlier studies (Bae, 2012; Nogueira-Machado and de Oliveira Volpe, 2012; Cho and Choi, 2014), our investigation demonstrated a correlation of the HMGB1 level with intestinal lesion severity. These data demonstrate that dietary fiber supplementation improves the clinical outcomes of mice subjected to sepsis.

The transcription factor Nrf2 is a key regulator of suitable antioxidant and anti-inflammatory responses (Vriend and Reiter, 2015; Ren et al., 2018). This investigation demonstrated greater mRNA levels of HO-1 and Nrf2 in mice subjected to CLP injection relative to normal mice. However, even higher levels were observed in septic mice subjected to the dietary highfiber intervention. Severe sepsis can cause disintegration of the intestinal tight junctions, resulting in systemic inflammation and oxidative stress (Ren et al., 2018). At this time, Nrf2 may be activated to translocate from the cytoplasm to the nucleus, where it binds to the ARE gene and thereby regulates the expression of SOD and CAT (Liu et al., 2014). HO-1, which is generated downstream of Nrf2, exerts beneficial actions against and thus downregulates pro-inflammatory responses (Yu et al., 2009; Vijayan et al., 2011; Bortscher et al., 2012).

The configuration of the gut microbiota has been shown to influence therapeutic responses in a variety of clinical conditions, including cancer and diabetes (Taur et al., 2014; Forslund et al., 2015; Vetizou et al., 2015). To date, however, clinical investigations of sepsis have not considered the status of the gut microbiota (i.e., they did not assess individual gut microbiota species present within the gut over the disease period). Diet is known to represent a robust connection between the gut microbiota and immune function (Kau et al., 2011; Tilg and Moschen, 2015). It would seem that this association is relevant to sepsis survival. In this investigation, the high-fiber intervention partly protected against systemic inflammation and mortality in a murine sepsis model. Earlier work by Peck et al. indicated that calorie restriction also enhanced survival in mice challenged with S. typhimurium (Peck et al., 1992). In this study, mice also exhibited clear alterations in gut microbiota, particularly an enrichment of the genus Akkermansia. These anaerobic microorganisms are typically found in both human and rodent gut microbiomes, and their abundance in humans has been shown to inversely correlate with body weight and inflammatory activity in patients with inflammatory bowel disease (Png et al., 2010; Santacruz et al., 2010).

#### CONCLUSION

This study has demonstrated that dietary supplementation with high fiber alleviates intestinal injuries. The mechanism of action is thought to be partially attributable to modifications of both the microbiota and host physiology by fiber supplementation, which thereby permit an appropriate and survival-promoting inflammatory response to the injury. Possibly, an improved comprehension of the correlations between the diet, microbiota and systemic pathology could lead to novel diet-based therapeutic approaches for sepsis. However, additional investigations are needed to assess the potential benefits of an intervention comprising dietary high-fiber supplementation for the treatment of severe sepsis.

#### DATA AVAILABILITY

All the data are available at YZ (yzhang10@tmu.edu) upon request.

#### AUTHOR CONTRIBUTIONS

YZ and YY designed the research and proofread the manuscript. AD and KX carried out the study. AD, KX, and YY analyzed the data. YZ wrote the manuscript.

#### FUNDING

This work was supported by National Natural Science Foundation of China (81471842).

#### REFERENCES

fphys-09-01929 January 16, 2019 Time: 18:44 # 7


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer HN and handling Editor declared their shared affiliation.

Copyright © 2019 Zhang, Dong, Xie and Yu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# A Data Integration Multi-Omics Approach to Study Calorie Restriction-Induced Changes in Insulin Sensitivity

Maria Carlota Dao<sup>1</sup> \* † , Nataliya Sokolovska<sup>1</sup> , Rémi Brazeilles<sup>2</sup> , Séverine Affeldt<sup>1</sup> , Véronique Pelloux<sup>1</sup> , Edi Prifti3,4, Julien Chilloux<sup>5</sup> , Eric O. Verger<sup>1</sup> , Brandon D. Kayser<sup>1</sup> , Judith Aron-Wisnewsky1,6, Farid Ichou<sup>7</sup> , Estelle Pujos-Guillot<sup>8</sup> , Lesley Hoyles5,9 , Catherine Juste10, Joël Doré10, Marc-Emmanuel Dumas<sup>5</sup> , Salwa W. Rizkalla<sup>1</sup> , Bridget A. Holmes<sup>2</sup> , Jean-Daniel Zucker3,4, Karine Clément1,6 \* and the MICRO-Obes Consortium‡

<sup>1</sup> Sorbonne University, French National Institute for Health and Medical Research, NutriOmics Unit, Institute of Cardiometabolism and Nutrition, Paris, France, <sup>2</sup> Danone Nutricia Research, Palaiseau, France, <sup>3</sup> Institute of Cardiometabolism and Nutrition, Integromics, ICAN, Paris, France, <sup>4</sup> Sorbonne University, IRD, UMMISCO, Bondy, France, <sup>5</sup> Section of Biomolecular Medicine, Division of Integrative Systems Medicine and Digestive Disease, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, London, United Kingdom, <sup>6</sup> Assistance Publique Hôpitaux de Paris, Nutrition Department, CRNH Ile-de-France, Pitié-Salpêtrière Hospital, Paris, France, <sup>7</sup> Institute of Cardiometabolism and Nutrition, ICANalytics, Paris, France, <sup>8</sup> Institut National de la Recherche Agronomique, Unité de Nutrition Humaine, Plateforme d'Exploration du Métabolisme, MetaboHUB, Université Clermont Auvergne, Clermont-Ferrand, France, <sup>9</sup> Department of Bioscience, School of Science and Technology, Nottingham Trent University, Clifton Campus, Nottingham, United Kingdom, <sup>10</sup> National Institute of Agricultural Research, Micalis Institute, AgroParisTech, Université Paris-Saclay, Jouy-en-Josas, France

Background: The mechanisms responsible for calorie restriction (CR)-induced improvement in insulin sensitivity (IS) have not been fully elucidated. Greater insight can be achieved through deep biological phenotyping of subjects undergoing CR, and integration of big data.

Materials and Methods: An integrative approach was applied to investigate associations between change in IS and factors from host, microbiota, and lifestyle after a 6-week CR period in 27 overweight or obese adults (ClinicalTrials.gov: NCT01314690). Partial least squares regression was used to determine associations of change (week 6 – baseline) between IS markers and lifestyle factors (diet and physical activity), subcutaneous adipose tissue (sAT) gene expression, metabolomics of serum, urine and feces, and gut microbiota composition. ScaleNet, a network learning approach based on spectral consensus strategy (SCS, developed by us) was used for reconstruction of biological networks.

Results: A spectrum of variables from lifestyle factors (10 nutrients), gut microbiota (10 metagenomics species), and host multi-omics (metabolic features: 84 from serum, 73 from urine, and 131 from feces; and 257 sAT gene probes) most associated with IS were identified. Biological network reconstruction using SCS, highlighted links between changes in IS, serum branched chain amino acids, sAT genes involved in endoplasmic reticulum stress and ubiquitination, and gut metagenomic species (MGS). Linear regression analysis to model how changes of select variables over the CR period

#### Edited by:

Jie Yin, Institute of Subtropical Agriculture (CAS), China

#### Reviewed by:

Zongxin Ling, Zhejiang University, China Yiliang Wang, Jinan University, China

#### \*Correspondence:

Maria Carlota Dao Carlota.Dao@tufts.edu Karine Clément karine.clement@inserm.fr

#### †Present address:

Maria Carlota Dao, Energy Metabolism Laboratory, Jean Mayer USDA Human Nutrition Research Center on Aging at Tufts University, Boston, MA, United States

‡The MICRO-Obes Consortium authors are listed at the end of the article

#### Specialty section:

This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology

Received: 01 November 2018 Accepted: 27 December 2018 Published: 05 February 2019

#### Citation:

Dao MC, Sokolovska N, Brazeilles R, Affeldt S, Pelloux V, Prifti E, Chilloux J, Verger EO, Kayser B, Aron-Wisnewsky J, Ichou F, Pujos-Guillot E, Hoyles L, Juste C, Doré J, Dumas M-E, Rizkalla SW, Holmes BA, Zucker J-D, Clément K and the MICRO-Obes Consortium (2019) A Data Integration Multi-Omics Approach to Study Calorie Restriction-Induced Changes in Insulin Sensitivity. Front. Physiol. 9:1958. doi: 10.3389/fphys.2018.01958

contribute to changes in IS, showed greatest contributions from gut MGS and fiber intake.

Conclusion: This work has enhanced previous knowledge on links between host glucose homeostasis, lifestyle factors and the gut microbiota, and has identified potential biomarkers that may be used in future studies to predict and improve individual response to weight-loss interventions. Furthermore, this is the first study showing integration of the wide range of data presented herein, identifying 115 variables of interest with respect to IS from the initial input, consisting of 9,986 variables.

Clinical Trial Registration: clinicaltrials.gov (NCT01314690).

Keywords: data integration, insulin sensitivity, lifestyle factors, microbiota, omics

#### INTRODUCTION

Obesity-associated impairment of glucose homeostasis is influenced by multiple elements including diet, physical activity, pharmacology and other lifestyle factors, predisposition from the host due to genetics, epigenetics, physiology and, as discovered more recently, gut microbiota alterations (Mutch and Clément, 2006; Khan et al., 2014; Mozaffarian, 2016). Although these elements have been implicated in different facets of the etiology of type 2 diabetes (T2D), specific mechanisms showing how they interact remain partially understood. Through high-throughput generation of biological data and analytical approaches involving data integration of elements from multiple sources, we endeavored to shed light on these complex interactions and highlight relevant mechanistic pathways.

Gut microbiota composition and function may play an important role in the prevention but also progression of insulin resistance (Qin et al., 2012; Cotillard et al., 2013; Karlsson et al., 2013; Le Chatelier et al., 2013; Dao et al., 2015; Forslund et al., 2015; Pedersen et al., 2016). Some bacterial species, such as Akkermansia muciniphila and Faecalibacterium prausnitzii, as well as high gut microbial gene richness are known to be associated with better metabolic and intestinal health (Cotillard et al., 2013; Le Chatelier et al., 2013; Dao et al., 2015; Plovier et al., 2017). Interestingly, studies considering the microbiome's functional potential have identified a series of pathways that are linked with different stages of insulin resistance. Among these pathways, branched chain amino acid (BCAA) metabolism by microbiota stands out. A recent study of the metagenome found increased abundance of bacterial genes coding for BCAA production and decreased abundance of genes coding for BCAA transporters on the bacterial wall in relation to insulin resistance in non-diabetic Danish adults (Pedersen et al., 2016).

High circulating concentrations of BCAA may disrupt glucose homeostasis in situations of stress on the body such as in high caloric intake (especially high-fat diets) or diets rich in foods with high glycemic index and excess body weight (Newgard et al., 2009; Roberts et al., 2014). BCAA have been repeatedly found to be elevated in T2D and obesity, and have been considered as early predictors of T2D (Roberts et al., 2014). Through activation of the nutrient-sensitive mTOR pathway and shift toward anabolism, BCAA may be redirected from protein synthesis toward gluconeogenesis with disruption of glucose homeostasis partly through chronically increased insulin secretion. Excessive fat mass storage in obesity may lead to not only in situ metabolic disruption, with some evidence of decreased BCAA catabolism (Roberts et al., 2014), but also ectopic fat deposition and inflammation. These modifications play an important role in the development of insulin resistance.

Lifestyle interventions resulting in weight loss, which involve calorie restriction (CR) together with improved dietary quality, and increased physical activity, lead to amelioration of IS (Després et al., 2001; Speakman and Mitchell, 2011; Kong et al., 2013). However, response to these interventions greatly varies from one person to another and a greater understanding at the individual level is needed to predict and enhance response (Kong et al., 2013; Dao et al., 2016). CR has been shown to significantly modify the transcriptional gene profile in subcutaneous adipose tissue (sAT) (Clément et al., 2004; Viguerie et al., 2005; Capel et al., 2009; Rizkalla et al., 2012), as well as other AT depots, and to modulate microbiota composition and function (reviewed in Dao et al., 2016).

So far, knowledge about the relationship between lifestyle factors, host biology, gut microbiota and gut-derived metabolites is somewhat fragmented. A better understanding of their interconnection may be achieved through data integration approaches across different disciplines of biological sciences and medicine (Kussmann et al., 2013; Kidd et al., 2014; Wu et al., 2015; Dao et al., 2016), and great efforts to develop data integration methods are currently being undertaken. Data integration can provide greater biological insight, a better

**Abbreviations:** AAAs, aromatic amino acids; ALDH6A1, aldehyde dehydrogenase 6 family, member A1; AT, adipose tissue; BCAT2, branched chain amino acid; transaminase 2; BCAA, branched chain amino acid; BCKDH complex, branched chain ketoacid dehydrogenase; BCKDHA, branched chain ketoacid dehydrogenase E1; BH, Benjamini-Hochberg; BMI, body mass index; CR, calorie restriction; DBT, dihydrolipoamide branched chain transacylase E2; DDRGK1, DDRGK domain containing 1; FIRI, fasting insulin resistance index; HOMA-B, homeostasis model assessment of beta cell function; HOMA-IR, homeostasis model assessment of insulin resistance; HOMA-S, homeostasis model assessment of insulin sensitivity; IL-6, interleukin-6; IS, insulin sensitivity; MGS, metagenomic species; MSE, mean squared error; NEFAs, non-esterified fatty acids; NF-κB, nuclear factor-κB; non-HDL, non-high density lipoprotein = total cholesterol – HDL; PLSR, partial least squares regression; pSVA, permuted-surrogate variable analysis; QUICKI, quantitative insulin sensitivity check index; revised QUICKI, revised quantitative insulin sensitivity check index; sAT, subcutaneous adipose tissue; SCS, spectral consensus strategy; T2D, type 2 diabetes; UFM1, ubiquitin-fold modifier 1.

understanding of relationships between variables, and may lead to innovative data-driven hypotheses paving the way to the discovery of mechanistic links. In fact, new data integration approaches are increasingly required to analyze the vast amounts of data being collected from deep phenotyping of patients (Piening et al., 2018).

The goal of this study was to improve our understanding of the inter-connectivity of IS and lifestyle, gut microbiome and host factors through the use a data integration approach. We aimed to conduct an in-depth exploration of associations between CRinduced improvement in IS and changes in host biology, gut microbiota and lifestyle factors. Novel connections between these elements were identified.

# MATERIALS AND METHODS

### Study Population and Dietary Intervention

A subgroup (N = 27) of the MICRO-Obes study (Cotillard et al., 2013; Kong et al., 2013, 2014) has been included in this report, based on AT sample availability. All analyses have been conducted using the delta (week 6 – baseline) of the different variables. The participants were overweight and obese adults [3 men and 24 women, median [interquartile range (IQR)] age and BMI: 41 (26) years and 33.9 (5.8) kg/m<sup>2</sup> ] with no chronic conditions and no medication use. The dietary intervention was completed at the Pitié-Salpêtrière Hospital in Paris, France and consisted of a 6-week individually prescribed hypocaloric diet (1200 kcal/d for women and 1500 kcal/d for men) followed by a 6-week weight-maintenance period. Only information collected at baseline and week 6 of the intervention has been included in the present analysis. This diet consisted of habitually consumed foods plus meal replacement with four dietary products daily (60–75 kcal; designed by CEPRODI-KOT Laboratory). These supplements consisted of lyophilized powder enriched in protein and soluble fiber (mainly inulin) and composed of low glycemic index carbohydrates. This intervention has been previously described in detail (Rizkalla et al., 2012; Kong et al., 2013). This study has been registered in ClinicalTrials.gov (NCT01314690) and approved by the Ethical Committee of Hôtel-Dieu Hospital in Paris, France in 2008 (under the number 0811792). All participants provided written informed consent. Data collection occurred in 2009 and 2010.

#### Assessment of Clinical Parameters

Body composition, sAT adipocyte diameter, and fasting blood lipids, insulin, glucose, inflammatory markers, and IS/secretion markers were assessed as described previously (Dao et al., 2015). Body composition measures included anthropometric measurements (weight, height, waist, and hip circumference), and dual energy X-ray (DXA) absorptiometry. Blood samples were collected after a 12-h fast.

#### Assessment of Food and Drink Intake

Diet was assessed with a 7-day unweighted food diary administered at baseline and week 6, as previously described in Kong et al. (2014). The software PROFILE DOSSIER V3 was used for analysis of the diaries and consumed foods were grouped into 26 food groups according to French dietary surveys. Average intake of energy, nutrients, and food groups were calculated from intake over the 7 days. The PROFILE DOSSIER X029 computer software (Audit Conseil en Informatique Médicale, Bourges, France) was used to analyze all food diaries and generate nutrient intakes. This software has a food composition database consisting of 400 food items which are representative of the French diet (Bouché et al., 2002).

#### Change in Diet and Clinical Variables

Changes in clinical and diet data were estimated using SAS 9.4. Median and IQR are reported for all variables. Statistical significance in change in clinical outcomes was determined with Wilcoxon signed rank sum test. Correction for multiple testing per class of variable was done using BH with FDR at the 5% level, and adjusted P-values are reported. Post hoc statistical power calculations for the change in IS markers from baseline to week 6 of CR is presented in **Supplementary Table 1**.

#### Adipose Tissue Biopsies and Gene Expression With Microarray Analysis

Subcutaneous adipose samples were collected at baseline and week 6 by needle biopsy from the periumbilical region under local anesthesia. Samples were frozen rapidly in liquid nitrogen then stored at −80◦C until they were further processed for microarray analysis as described in Rizkalla et al. (2012). Briefly, total RNA was extracted by using the RNeasy total RNA Mini kit (QIAGEN) with one-column DNase digestion. RNA quality and concentration were assessed by using an Agilent 2100 Bioanalyzer (Agilent Technologies). An Illumina RNA amplification kit (Ambion) was used according to the manufacturer's instructions to obtain biotin-labeled complementary RNA from 250 ng total RNA. Hybridization processes were performed with Illumina Human HT-12 version 3.0 Expression BeadChips (Illumina, Inc.). Hybridized probes were detected with cyanin-3 streptavidin (1 mg/mL; Amersham Biosciences, GE Health Care) and scanned by using an Illumina BeadArray Reader. Raw data were extracted with GenomeStudio 2011.1 Software by using the default settings and quantile normalization. Relevant genes associated with changes in IS were annotated using the FunNet tool (Prifti et al., 2008), using the whole human genome as the reference gene set.

# Batch Correction of Microarray Data

Transcriptomics data were adjusted for batch effects that are technical artifacts inherent to DNA/RNA sample processing (**Supplementary Figure 1A**) (Leek et al., 2010). We chose the pSVA approach (Chen et al., 2011; Parker et al., 2014), to remove these effects based on an in-house micro-array quality control procedure. The pSVA adjustment method, which is based on singular value decompositions, has been shown to be very efficient in preserving biological heterogeneity (Parker et al., 2014).

We have evaluated the adjustment quality of our data following the approach described in Gagnon-Bartsch and Speed (Gagnon-Bartsch and Speed, 2012). Specifically, two sets of control genes have been determined from the literature and narrowed down by statistical analysis: positive controls (18 genes) that were shown to be differentially expressed between male and female AT, and negative controls (47 genes) whose expression is independent of the gender biological signal (**Supplementary Figures 1B,C**). P-value rank of controls were assessed before and after batch effect correction to identify the optimal adjustment approach (**Supplementary Figures 1D,E**). A complementary qPCR analysis confirmed the batch effect correction improvement (**Supplementary Figures 1F,G**). The final data set included only genes that were 100% present in all 27 subjects at both time points.

# Analysis of sAT Gene Expression by qPCR

RNAs from AT were prepared using RNeasy Lipid Tissue kit (Qiagen) and cDNAs were synthesized using SuperScript II and random hexamers (Promega). Quantitative PCR with SybrGreen was performed and relative quantification of each transcript in comparison to 18S was determined using the 2−11Ct method. All primer sequences are available upon request.

# Metagenomic Sequencing for Assessment of Gut Microbiota

Fecal microbiota was characterized with shotgun quantitative metagenomics (N = 27). Total fecal DNA was analyzed with high throughput SOLiD sequencing, as described in Cotillard et al. (2013). Reads were mapped and counted onto the 3.9 million gene catalog, as described in Dao et al. (2015). Only MGS with more than 700 genes were considered.

# Serum and Urine Metabolic Phenotyping by <sup>1</sup>H NMR Spectroscopy

Urine (N = 21) and serum (N = 24) samples were randomized, prepared and measured on a NMR spectrometer (Bruker) operating at 600.22 MHz <sup>1</sup>H frequency using previously published experimental parameters (Dona et al., 2014). The <sup>1</sup>H NMR spectra were then pre-processed and analyzed as previously reported (Dumas et al., 2006) using Statistical Recoupling of Variables-algorithm (Blaise et al., 2009). Structural assignment was performed as reviewed in Dona et al. (2016), using in-house and publicly available databases.

# Fecal Metabolomics

Analysis of frozen fecal samples (N = 26) from patients was performed using an Agilent 7890A gas chromatography system coupled to an Agilent 5975C inert XL EI/CI MSD system (Agilent, Santa Clara, CA, United States). A HP-5MS fused-silica capillary column (30 m × 250 µm × 0.25 µm; Agilent J & W Scientific, Folsom, CA, United States) was used. Sample preparation and GC/MS methods were performed as detailed in the paper of Gao et al. (2009). Peaks obtained from fecal water samples were aligned, grouped and corrected according to the mass to charge ratio (m/z) and retention time using XCMS R tools (Smith et al., 2006). Resulting MS data was a data matrix in which 835 features were found and characterized by a retention time (RT), mass to charge ratio (m/z) and their corresponding intensities per patient. Spectra were deconvoluted using Automated Mass Spectral Deconvolution and Identification System (AMDIS) before being compared with reference libraries (in-house and NIST05). Annotation and identification of features were based on standards proposed by Sumner et al. (2007). Eighty-five features were found statistically relevant and identified based on the retention time, m/z and reference spectra of standards from a local database. Features were considered as putatively annotated when only the m/z and RT and main fragments were matched with the reference standards. The remaining features were putatively characterized (m/z and RT were verified with metabolites in our local database) or noted as unknown features.

### Inputs From Host, Lifestyle, and Microbiota for Partial Least Squares Regression Analysis

A large set of variables measured from elements in host, lifestyle patterns or microbiota have been included in the data integration analysis presented herein. Specifically, as shown in **Figure 1A**, the following numbers of variables have been included: 26 food groups and 34 nutrients derived from 7 day food diaries, three physical activity variables obtained from the validated Baecke questionnaire, 741 MGS derived from metagenomic sequencing, 835 features from fecal metabolic features obtained using GC–MS, 562 urine and 180 serum metabolic features measured with NMR, 45 clinical parameters (including anthropometric measures, markers of IS, circulating concentrations of lipids, inflammatory markers, adipokines, creatine, and cystatin C, and adipocyte diameter measured from sAT biopsies).

#### Partial Least Squares Regression With Canonical Mode

Multivariate methods such as PLSR are well-suited and often applied to big biological data, including situations where the number of variables (clinical parameters, metabolic features, genes, etc.) is much larger than the number of observations (patients). In particular, PLSR is used to model relationships between two groups of variables (Esposito Vinzi et al., 2010). It maximizes the covariance between latent variables which are linear combinations of the observed pairs of variable groups. The results can be efficiently visualized as graphs that enable better understanding of underlying biological relations and processes. In our analysis, we have used the R "mixOmics" package (Omics Data Integration Project), and the function pls(), with the mode = "canonical," which assumes no directional relationships between data sets. PLSR usually refers to an asymmetric deflation of the two groups of variables, if the deflation is symmetric, then the mode is called canonical mode. A threshold between | 0.7| and | 0.75| for association coefficients was set in order to only consider the strongest

connections between each variable block and IS. Networks from PLSR were generated using Cytoscape (Shannon et al., 2003).

#### Spectral Consensus Strategy for Accurate Reconstruction of Large Biological Networks

Developing statistical methods to accurately reconstruct biological interaction networks is a central goal of integrative systems biology (Herrgård et al., 2008). In the networks that are of interest in our analysis, nodes represent variables and edges indicate statistical dependency. While many network reconstruction methods exhibit competitive results on various types of data, most state-of-the art methods are challenged to scale to very high dimension datasets where the number of variables might be several orders of magnitude larger than the number of observations. In our analysis, this is precisely the setting we encounter as the number of variables is 9,986 and the number of observations is 27.

To accurately reconstruct such a network of interactions we have developed an original method (Affeldt et al., 2016) that first reduces the reconstruction problem into a large number of much simpler reconstruction problems, then let the lower-dimension problem be solved by state-of-the art reconstruction methods and finally adopt a consensual voting strategy between these methods to identify the most accurate sub-graphs. The different sub-graphs are then connected when they share common nodes like pieces of a larger puzzle. The main originality of the method lies in its powerful problem reduction that, thanks to a socalled spectral decomposition, both identifies non-exclusive sets of most likely dependent variables and enables the approach to scale-up to large biological networks. ScaleNet networks were generated using Cytoscape (Shannon et al., 2003). For the network presented herein, we used 14% of the eigenvector, and the size of the subgraphs used for the reconstruction was 0.03.

#### Prediction and Regression

fphys-09-01958 February 4, 2019 Time: 16:1 # 6

The target variables being predicted (i.e., revised QUICKI index and HOMA-B as markers of IS) are continuous, therefore, we performed a linear regression to estimate the impact of each set of independent of variables on the variable of interest. The MSE provides the average of the squares of the errors of deviations, and quantifies the difference between the real and estimated data. To measure the impact of each data source on IS, we ran linear regressions and computed the MSE for each type of data (10-fold cross validation) (Hastie et al., 2009). We inferred that the inverse values of the MSE represent the magnitude of the impact. The intuition behind such a visualization is that the lower the error, the more significant the data are.

#### Analysis of Subcutaneous Adipose Tissue DDRGK1 in a Separate Obese Population

To explore some of the novel links identified with PLSR and ScaleNet, we used existing data from an unpublished CR study and determined association between sAT DDRGK1 and IS. The methodologies for these variables were the same that in the present study. The validation study consisted of a 6-week very low calorie diet, followed by 6 weeks of weight stabilization. In this group of overweight/obese adults, only baseline and week 12 values (i.e., after the weight regain and stabilization period) are presented.

#### Availability of Data

Data used in this manuscript can be accessed as follows: (1) AT Illumina microarray data with Gene Expression Omnibus (GSE112307), (2) serum metabolomics NMR data with Metabolights (pending accession code), (3) urine metabolomics NMR data with Metabolights (pending accession code), (4) fecal metabolomics GC–MS data with Metabolights (MTBLS653), and (5) metagenomics raw solid read data with the European Bioinformatics Institute European Nucleotide Archive (ERP003699).

# RESULTS

#### Dietary Intervention-Induced Weight Loss and Improved Insulin Sensitivity: Multiple Inputs Used for Data Analysis and Integration

**Figure 1** shows an overview of the different steps taken to analyze the data presented in this study. Details are provided in Section "Materials and Methods." The study population is a subgroup from the MICRO-Obes study (Cotillard et al., 2013; Kong et al., 2013, 2014), consisting of 27 overweight or obese adults who underwent a 6-week CR intervention (1,200 kcal/day for women and 1,500 kcal/day for men). The diet was low in fat (25% of energy intake), high in protein (35% of energy intake), and rich in total fiber and in carbohydrates with a low glycemic index. The diet composition has been described elsewhere (Rizkalla et al., 2012). The variable inputs as described previously (Cotillard et al., 2013; Kong et al., 2013, 2014), were divided into the following blocks (**Figure 1A**): 45 clinical variables that included anthropometry, adipocyte diameter, measures from blood chemistry profiles (including 10 IS markers, ), and inflammatory markers; ❶ lifestyle factors (26 food groups, 34 macro- and micronutrients, 3 indexes representing degree of physical activity at work, sports and leisure time); ❷ 741 MGS; ❸ 835 fecal metabolic features; ❹ 562 urine metabolic features; ❺ 180 serum metabolic features; and ❻ 7,560 genes expressed in sAT. The change (week 6 – baseline) in each variable was exclusively used in all analytical steps (**Figures 1B– E**).

TABLE 1 | Change in clinical outcomes with calorie restriction.


Wilcoxon signed rank sum test, adjusted P-values (Padj, BH correction) are shown. BMI, body mass index; NEFA, non-esterified fatty acids; HDL, high density lipoprotein; LDL, low density lipoprotein; CRP, C-reactive protein; IL-6, interleukin-6; FIRI, new fasting insulin resistance index; revised QUICKI, revised quantitative insulin sensitivity check index; IQR, interquartile range. Bold values are significant adjusted p-values.

As expected, the dietary intervention led to a significant reduction in BMI, with an average weight loss of 5.4 ± 2.5%, as well as waist-to-hip ratio, % fat mass and android fat (**Table 1**). There was a significant decrease in blood lipids, and a marked improvement in surrogate markers of insulin secretion (HOMA-B, calculated as described in Levy et al., 1998); or IS (HOMA of insulin resistance, HOMA-IR, HOMA-S, Disse index, the quantitative IS check index, QUICKI, the revised QUICKI, insulin to glucose ratio, and the new FIRI, **Table 1**). It is important to note that there was a wide range of change in IS, with both responders and non-responders to the intervention, as is commonly the case with lifestyle and CR interventions (Kong et al., 2013). Indeed, while all participants experienced a decrease in BMI during this period, there was individual variability in change of the revised QUICKI, with 19 individuals having an increase (improving IS as measured by this index) and 8 individuals having a decrease in this index. As such there was no significant improvement in mean revised QUICKI (**Table 1**). **Supplementary Figure 2** shows individual trajectories for IS indexes used in the analysis.

#### Change in Insulin Sensitivity After Diet-Induced Calorie Restriction: Association With a Myriad of Lifestyle, Gut Microbiota, Metabolic Features, and Adipose Tissue Genes

Using PLSR with canonical mode, we examined association networks between the changes in IS from this CR intervention and changes in lifestyle factors (i.e., diet and physical activity), blood, urine, and fecal metabolomics, gut metagenomics, and sAT transcriptomics. Although there is redundancy in some of the IS markers, they were included to ensure analytical relevance. In other words, we would expect similar indexes such as FIRI and HOMA-IR to be associated with consistent sets of variables (which was the case, as seen in **Supplementary Figure 3**).

The networks depicting PLSR results (**Figures 2A,B**) show strong associations of change between several groups of variables and IS markers. We considered association coefficients greater than | 0.7| or, in some cases, | 0.75| . These thresholds represent a trade-off between analytical stringency and retention of useful information (**Supplementary Table 2**). Above the selected thresholds, there were associations between improvements in IS and change in nutrient intake that included fiber, carbohydrate, alcohol, β-carotene, vitamins A, B6, and B9, magnesium, phosphorus, and iron (**Figure 2C** and **Supplementary Figure 4A**, **Supplementary Table 3**). No food groups were associated with IS above the selected threshold.

Regarding gut microbiota changes, 10 MGS, each detected in at least two subjects, were associated (in some cases negatively and in others positively) with improvement in IS after CR. Three of these MGS have not been annotated at any taxonomic level (**Figure 2C**). Of the annotated ones, when considering the lowest taxonomic annotation available, one was related to the species Collinsella intestinalis (having negative association of change with revised QUICKI). At the genus level there were two Alistipes (having positive association of change with revised QUICKI, Disse index or glucose), one Bacteroides and one Prevotella (both having positive association of change with glucose). At the family level one Lachnospiraceae (having negative association of change with glucose); and at the order level one Clostridiales (having negative association of change with revised QUICKI) (**Supplementary Figure 4C**). Of the unannotated MGS, there was one having a negative association of change with the Disse index (Best Hit: uncultured Faecalibacterium sp.), and two having a negative association of change with fasting glucose (Best Hit: Opitutaceae bacterium TAV5 and uncultured Faecalibacterium sp.).

There were 84 serum metabolic features whose changes in concentration were most strongly associated (81 negatively and 3 positively) with improvement in IS. Among these were BCAA and AAAs, as well as other amino acids, namely lysine, glutamate, and glutamine (**Figure 2C** and **Supplementary Figure 4B**). IS improvement was associated with a decrease in BCAA, consistent with the literature (Roberts et al., 2014). Several explanations could account for the implication of BCAA in glucose homeostasis. One hypothesis is the potential role of AT, which has been shown to have decreased expression of BCAA catabolic enzymes in obesity and insulin resistance (Roberts et al., 2014). In agreement with this hypothesis, the PLSR analysis between variation in serum metabolic features and sAT transcriptomics show strong significant associations between sAT genes and BCAA as well as with other metabolites including 3 hydroxyisobutyrate and 3 methyl-2-oxovalerate (**Supplementary Figure 5**). When considering change in AT genes coding for enzymes involved in BCAA catabolism (**Supplementary Figure 5F**) (Herman et al., 2010), there was a positive association of change between valine and isoleucine with the BCAT2 gene. This enzyme converts BCAA into α-ketoacids in the first step of BCAA catabolism (Herman et al., 2010). There was a negative association between all BCAA and gene expression of DBT, which is part of the BCKDH complex. This complex catalyzes the second step of BCAA catabolism, namely the conversion of α-ketoacids into acyl-CoA esters. Only isoleucine was positively associated with BCKDHA, which is also part of the BCKDH complex. ALDH6A1, involved in a downstream step of the pathway, was also negatively associated with variations in all BCAA. These results suggest that while substrates to the first step of BCAA catabolism may decrease with weight loss, subsequent steps of the pathway increase, reflecting an upregulation of BCAA catabolism.

In urine, change in hippurate, a metabolite previously found to be inversely associated with metabolic syndrome risk, gut microbiota richness (Pallister et al., 2017), and obesity (Elliott et al., 2015), was positively associated with improvement in IS, specifically, with a decrease in plasma insulin, HOMA-IR and FIRI (**Supplementary Figure 4B**). Fecal metabolites associated with change in IS included isoleucine, alanine, β-alanine, 2 hydroxyhexanoic acid, arabinose methoxyamine, valeric acid, phenylacetic acid, 3-hydroxypyridine, 4-cresol, malonic acid, and lactate (**Supplementary Figure 4B**).

PLSR analysis also revealed that the clinical factors most strongly associated with AT genes were Disse, QUICKI, FIRI and revised QUICKI indexes, NEFAs, non-HDL, adipocyte diameter, interleukin-6 (IL-6), BMI, hip circumference and fat mass (in kg) (**Supplementary Figure 4D**). There were 257 genes associated

sensitivity (1INS. SEN.), where nodes are arranged by (A) betweenness centrality and (B) variable type. The green edges correspond to positive correlations of change and the red edges correspond to negative correlations of change. (C) Summary of variables from host, microbiota, and lifestyle factors associated with 1INS. SEN. Association coefficient threshold = [0.7] for lifestyle factors, and | 0.75| for metabolomics and sAT gene expression. NA, not annotated; BCAAs, branched chain amino acids; AAAs, aromatic amino acids; AAs, amino acids; MGS, metagenomic species. No association with change in physical activity or food groups was found above the selected threshold.

with these clinical factors (summarized in **Figure 2C** and listed in **Supplementary Table 4**).

In summary, this PLSR analysis, where we have applied stringent association coefficient thresholds, has resulted in the identification of a large set of factors (N = 565) from multiple sources that are connected to changes in IS after CR.

### Data Reduction and Integration: Novel Connections Between Insulin Sensitivity, BCAA and DDRGK1, an Adipocyte Expressed Gene Involved in Protein Ubiquitination

Our next objective was to gain a deeper insight into these large sets of interactions between variables and highlight novel connections based on the strongest associations between in CR-induced changes in IS, metabolic features, sAT genes and MGS. To reduce the number of features, we examined the top 20% associations which was a trade-off between stringency and retention of strong associations with change in IS (**Supplementary Table 5**).

We used ScaleNet, a network learning approach based on SCS, developed by our group (Affeldt et al., 2016), to examine and visualize the strongest statistical dependencies between CRinduced IS changes and the selected variables. The data reduction approach led to focus on the following features: three nutrients (carbohydrate in % of energy intake, fiber in grams, and alcohol in grams), MGS (three species with highest prevalence in subjects, namely GU:41 from the Clostridiales order, GU:84 from the Alistipes genus, and GU:228 which is not annotated but whose strongest hit from the gene catalog is Faecalibacterium sp.), 22 serum metabolic features, 19 urine metabolic features, 24 fecal metabolic features, the markers of IS, adipocyte diameter, NEFA, and 32 sAT genes.

ScaleNet thus established statistical dependencies between these variables based on the consensus taken from different network reconstruction approaches. The dependencies derived through ScaleNet are therefore more robust than if any one single network reconstruction method were used. These statistical dependencies represent the pairwise mutual information between variables, and so they capture both linear and non-linear relationships.

As shown in **Figure 3A**, the reconstructed network is composed of a large cluster (top) and several smaller clusters not connected to the large cluster (bottom). This network showed several within-class modules, such as for fecal metabolic features for which a distinct cluster of nodes (variables) can be seen, sAT genes, or serum metabolic features, where several features corresponding to serum glucose clustered together. Importantly, we observed statistical dependencies depicting already established connections between classes of variables (**Figure 3A**). Indeed, the IS markers that clustered together were strongly connected with a serum BCAA cluster (leucine, valine, and valine + isoleucine).

orange ellipse on the network in A). When significant, adjusted P-values (BH correction) from Wilcoxon Signed Rank test are shown. (C) Linear regression model, where each section of the pie chart shows the relative contribution of change in selected variables (grouped by class) to change in revised QUICKI and HOMA-B. Variables included in the model were those found in the main cluster of the network in part A (except serum metabolic features annotated as glucose). MGS, metagenomic species; metab., metabolic features; DDRGK1, DDRGK domain containing 1 (also known as Dashurin).

This network showed a novel direct connection between the sAT gene DDRGK1 (also known as UFM1-Binding Protein 1, or UFBP1 or Dashurin) and three markers of IS (glucose, HOMA-B, and revised QUICKI), as well as with serum acetylcarnitine and GU:228 (unannotated, with closest relevance to Faecalibacterium sp.), and an indirect connection with other markers of IS and serum BCAA. The DDRGK1 gene has not been previously explored in human AT. Revisiting existing "in house" microarray data from obese women (unpublished) revealed that DDRGK1 is 1.6-fold more highly expressed in isolated adipocytes than stromal vascular fraction in sAT (**Supplementary Figure 6**). To further examine the relevance of this relationship between DDRGK1 and IS variation, we explored another group of 27 overweight and obese patients with comparable clinical characteristics (unpublished data, **Supplementary Table 6**). Here, the subjects had undergone a very low-calorie diet for 6 weeks followed by a 6-week weight maintenance period, and DDRGK1 expression was examined after 12 weeks (**Supplementary Figure 7**). After this weight maintenance phase, there was a significant positive correlation of change between the revised QUICKI and DDRGK1 expression in sAT (**Supplementary Figure 7C**).

The changes in selected variables having dependencies with the IS cluster is displayed in **Figure 3B**, showing a significant decrease in serum BCAA and acetylcarnitine, a tendency to decrease in DDRGK1, and no significant change in GU:228, or unannotated urine and fecal metabolic features directly connected with the IS cluster. The network also showed dependencies between serum lysine and different elements including blood glucose, unannotated fecal metabolic features, an MGS from the Clostridiales order, and acetylcarnitine.

# Insulin Sensitivity Improvement After CR: Important Contribution of Metagenomic Species and Fiber Consumption in the Diet

The ScaleNet network showing the statistical dependencies of highly connected variables (**Figure 3A**) suggests that not only lifestyle factors, fecal, urine, and serum metabolic features, but also AT genes and gut microbiota composition contribute to improvements in IS after CR. We then examined the relative contribution of these elements to improved IS after CR. We used linear regression and computed the inverse of the MSE by variable class.

The pie charts on **Figure 3C** show that the greatest contribution (around 50%) to improvement in revised QUICKI and HOMA-B index came from change in abundance of the three MGS included in the network, increase in dietary fibers and, in the case of HOMA-B, change in concentration of fecal metabolic features. Interestingly, there were similar contributions from sAT gene expression (including DDRGK1) and selected serum and urine metabolic features (between 5 and 17%). These results also suggest important contributions by gut microbiota MGS and dietary intake of fiber to improvements in IS.

# DISCUSSION

In this study, we have identified novel links between host glucose homeostasis, lifestyle factors and gut microbiota in the context of a 6-week CR intervention. This is to our knowledge the first study integrating information from lifestyle factors, gut microbiota composition, clinical factors, sAT gene expression, and metabolomics from serum, urine and feces, to investigate their relationship with IS changes after CR. The variables that were found to be more strongly associated with changes in IS should be explored in future studies as potential biomarkers or predictors of individual responses to this kind of intervention.

The analytical approach presented herein has led to: (1) integration and visualization of multiple associations between changes in gut microbiota, human omics data, lifestyle factors, and IS; (2) reduction of data dimensionality to highlight key variables of interest (from 9,986 variables to 115 variables included in the ScaleNet reconstruction); (3) identification of a robust inverse association between changes in circulating BCAA with IS during CR; and (4) identification of new links between IS and different elements such as those seen with metabolic features in urine and feces, sAT genes and gut microbiota composition. (5) Finally, we show an important contribution of MGS in improved IS after CR in link with diet modulation.

# Decreased Serum BCAA Associated With Increased Insulin Sensitivity Are Connected With Features From Host, Lifestyle, and Gut Microbiota

One of the strongest associations with IS improvement was a significant decrease in BCAA with CR (**Figure 3B**). This differs from previous findings comparing metabolomics profiles between a gastric bypass intervention and a weight-matched dietary intervention (Laferrère et al., 2011). The authors observed a decrease in BCAA and their metabolites in the surgical group, concomitant with amelioration in IS. However, this was not the case in the dietary intervention group. The baseline BMI for this group was considerably higher than in our study group (42.8 ± 3.8 vs. 34.0 ± 4.1 kg/m<sup>2</sup> , respectively), and their insulin resistance markers remained higher than what we have observed in this study after the dietary intervention. Therefore, greater weight loss may have been necessary to yield significant decrease in BCAA related to an improvement in IS, as shown in the present study. Associations were identified not only between BCAA and IS, but also with sAT gene expression, corroborating the hypothesis of a potential role of sAT in insulin resistance through inhibition of BCAA catabolism in AT which may accompany the known alteration of other metabolic pathways, including mTOR signaling in skeletal muscle, increase in gluconeogenesis, decrease in glycolysis and β-oxidation, and increased mitochondrial stress (Newgard et al., 2009; Herman et al., 2010; Roberts et al., 2014).

Our data integration and reduction approach led to the identification of robust dependencies between the IS markers/BCAA cluster in the ScaleNet network with DDRGK1, a gene involved in protein ubiquitination. The relationship between DDRGK1 and IS was confirmed in an independent group of overweight/obese subjects following a different dietary intervention setting. Indeed, this second group was explored during a weight maintenance phase after a period of weight loss. Thus the fact that AT gene expression profiles greatly vary depending on the stage of weight loss and regain (Capel et al., 2009; Piening et al., 2018) could account for the difference in the direction of correlation between this study population and the independent clinical group.

DDRGK1 has been shown to become activated and bind to UFM1 in pancreatic beta cell lines for protection against ER stress-induced apoptosis (Lemaire et al., 2011). There is evidence in cell lines that DDRGK1 regulates NF-κB through interaction with IκBα (Xi et al., 2013). Limited evidence is available about this gene's function in AT. We show here that it is more highly expressed in isolated adipocytes than in the stromal vascular fraction calling for deeper exploration of this gene in adipocyte metabolic pathways. The ScaleNet network also shows a link between DDRGK1 and an unannotated MGS whose strongest alignment is with Faecalibacterium sp. Interestingly, this microbial genus has been shown to be anti-inflammatory and associated with IS (Sokol et al., 2008; Le Chatelier et al., 2013; Munukka et al., 2014).

The network also showed elements from sAT, microbiota composition and fecal and urine metabolomics connected to the IS and BCAA cluster. Acylcarnitines are intermediates of fatty acid oxidation that, when elevated, may be involved in lipidinduced pathways of insulin resistance, although this hypothesis has been questioned (Schooneman et al., 2013). Studies have conversely shown that acetyl-L-carnitine supplementation leads to improved IS in insulin-resistant overweight and obese individuals (Ruggenenti et al., 2009). Acylcarnitines may also derive from other metabolites, including ketone bodies, BCAA, lysine and, in the case of acetylcarnitine, glucose. Acetylcarnitine,

the two-carbon acylcarnitine, may be an indicator of glucose and fatty acid catabolism status and has been previously associated with insulin resistance (Schooneman et al., 2013). A recent weight loss study in obese subjects showed that acylcarnitines, including acetylcarnitine, were positively correlated with NEFAs at baseline (Schooneman et al., 2016). In this study, contrary to our findings and other published evidence (Ruggenenti et al., 2009; Muoio et al., 2012), acetylcarnitine increased after a weight loss intervention and was not correlated with IS. The design of this study differed from ours in that it was a 12-week intervention including a group on diet with physical activity and another on pharmacological treatment. The differences in study design (including diet) and effects induced on host metabolism from these interventions could explain the discrepancies between the studies. Here, concentration of acetylcarnitine decreased significantly, and on the ScaleNet network (**Figure 3A**) it was directly connected with DDRGK1, and indirectly connected with BCAA, IS markers and a MGS from the Clostridiales order (GU: 41). This suggests a possible link between ER stress in sAT, and insulin resistance-associated dysregulation of AT metabolism. This hypothesis calls for confirmation in future studies, both involving large human cohorts, as well as mechanistic exploration in in vivo and in vitro models.

#### Cross-Talk Between Gut Microbiota Composition and Insulin Sensitivity After CR

The cross-talk between microbiota and metabolism in various organs and tissues, including AT, has been described mostly in mice but less so in humans (Tremaroli and Bäckhed, 2012; Geurts et al., 2014; Fetissov, 2017). Even though we focused our study on peripheral organs, there is current interest in the cross-talk between gut microbiota and brain signaling (Wang et al., 2018). Gut microbiota likely has a role in BCAAassociated development of insulin resistance, as shown by Pedersen et al. (2016), where a connection between circulating BCAA, increased BCAA synthesis and decreased BCAA import in microbiota was identified in insulin resistant adults. In the same study, a causal role for gut microbiota in increase of circulating BCAA and insulin resistance was suggested in a mouse model. Consistently, our results demonstrate that a reduction of caloric intake, body weight and improvement in dietary quality results in a significant decrease in BCAA, which is connected with improvement in IS and potentially also sAT metabolism and protein ubiquitination, and with compositional changes in the gut microbiota. The relative contribution to circulating BCAA, as well as other metabolites, from the gut microbiome and host and the degree of gut permeability before and after weight loss interventions in human obesity are to be considered in future studies (Genser et al., 2018).

#### Challenges in Integration of Data With High Dimensionality

A limitation of this study is the small sample size, an issue generally encountered in studies using big data and deep phenotyping. To circumvent this issue, we have used stringent approaches in the different steps of the analysis, including high thresholds for selection of relevant variables and methods adapted to high dimensionality such as the local reconstruction aspect of ScaleNet. The advantage of this method is that it goes far beyond pairwise associations of variables, taking into account all long-range associations within subsets of variables, considering both linear and non-linear relationships. This method is adapted to high-dimensional settings since each reconstruction method uses many local reconstructions based on a spectral setting. However, it is important to note that the links we have identified need to be corroborated by future research, including comparable study designs and similar population characteristics. In fact, this analytical approach will be applied to the Metacardis study population<sup>1</sup> .

This study included only three men, which prevented the assessment of sex differences in our outcomes. Future studies should be balanced for gender to identify potential sex-specific effects relevant in the treatment of obesity. Another limitation is that some metabolic features included in this analysis could not be annotated due to lack of knowledge in public databases and/or limitations in the identification of low intensity metabolites, so interpretation of these results could evolve depending on the depth of our abilities to annotate further features.

The PLSR analysis takes into account pairwise associations between two sets of variables rather than all possible connections between and within both groups of variables. Our approach started with a pairwise assessment of connections with IS in order to prioritize variables of greatest relevance, followed by the ScaleNet analysis, which included all possible connections between these variables. However, the fact that we have initially considered associations between two groups of variables at a time (i.e., IS markers versus each block of variables), instead of taking a more global approach, may have masked other important connections between variables. Conversely, the approach we have used makes the results more interpretable.

# CONCLUSION

Our data integration approach has provided new insights and revealed new associations between IS and AT genes, metabolites, lifestyle factors, and microbiota during CR. Although here we present some initial steps toward validation of our observations, this approach should be repeated in other populations with similar characteristics, and ideally with larger sample sizes, to determine the robustness and reproducibility of the more novel links we have identified. A recent study has also taken a data integration approach to study multiomics associations during weight gain and weight loss, and have uncovered connections between biomarkers and the gut microbiota in individuals undergoing weight gain followed by weight loss (Piening et al., 2018). Similarly to our study, Piening et al. (2018) highlight connections between

<sup>1</sup>http://www.metacardis.net/

IS status and circulating BCAA and acylcarnitines. It is through reproducing such strategies in different studies that a reliable and more complete picture of the connection between obesity, metabolism, gut microbiota, and lifestyle factors will be obtained.

This type of integrative approach should lead to innovative applications for personalized care. Individual response profiles from weight loss interventions could be optimized by targeting markers of interest identified through this analysis. Data integration has the potential for the development of personalized approaches that would allow us to tackle obesity and other associated cardiometabolic diseases.

#### MICRO-OBES CONSORTIUM MEMBERS

Aurélie Cotillard, Sean P. Kennedy, Nicolas Pons, Emmanuelle Le Chatelier, Mathieu Almeida, Benoit Quinquis, Nathalie Galleron, Jean-Michel Batto, Pierre Renault, Stanislav Dusko Ehrlich, Hervé Blottière, Marion Leclerc, Tomas de Wouters, Patricia Lepage.

# AUTHOR CONTRIBUTIONS

MD, SR, J-DZ, and KC have designed the overall study. KC and SR have designed and conducted the clinical research and clinical data management. MD managed analysis and interpretation of results. MD, NS, RB, SA, EP, J-DZ, and KC developed the analysis strategy. JA-W and BK contributed to clinical data analysis. VP carried out adipose tissue gene expression analysis and validation. BH and EV contributed to dietary data analysis. M-ED, JC, and LH carried out the serum and urine metabolomics analysis. FI, EP-G, and CJ carried out the fecal metabolomics analysis. EP, JD, and J-DZ conducted the microbial data analysis. All authors

#### REFERENCES


contributed to the preparation, writing, and approval of the manuscript.

#### FUNDING

This work was supported by Agence Nationale de la Recherche (ANR MICRO-Obes and ANR-10-IAHU-05) and the European Union's Seventh Framework Programme for research, technological development and demonstration under grant agreement HEALTH-F4-2012-305312 (METACARDIS) and Fondation Leducq (to KC's team). The clinical work received support from KOT-Ceprodi, Danone Nutricia Research and the Foundation Coeur et Artère. LH was in receipt of an MRC Intermediate Research Fellowship in Data Science (Grant No. MR/L01632X/1) and Heart and Stroke Foundation of Canada.

#### ACKNOWLEDGMENTS

We thank the subjects for their participation in this study. We also thank Christine Baudoin who contributed to the clinical investigation study, Soraya Fellahi and Jean-Philippe Bastard (Department of Biochemistry and Hormonology, Tenon hospital) for analyses of biological markers, Dominique Bonnefont-Rousselot and Randa Bittar (Department of Metabolic Biochemistry, Pitié-Salpêtrière Hospital) for help with the analysis of plasma lipid profile.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphys. 2018.01958/full#supplementary-material

adipose tissue of obese subjects. FASEB J. 18, 1657–1669. doi: 10.1096/fj.04- 2204com


intervention in obese diabetic subjects despite identical weight loss. Sci. Transl. Med. 3:80re2. doi: 10.1126/scitranslmed.3002043


fphys-09-01958 February 4, 2019 Time: 16:1 # 13

increased cardiovascular risk: effects of acetyl-L-carnitine therapy. Hypertens 54, 567–574. doi: 10.1161/HYPERTENSIONAHA.109.132522


**Conflict of Interest Statement:** Co-authors from Danone Nutricia Research are part of the METACARDIS Consortium and they have contributed to the work presented herein in this capacity.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Dao, Sokolovska, Brazeilles, Affeldt, Pelloux, Prifti, Chilloux, Verger, Kayser, Aron-Wisnewsky, Ichou, Pujos-Guillot, Hoyles, Juste, Doré, Dumas, Rizkalla, Holmes, Zucker, Clément and the MICRO-Obes Consortium. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Lactobacillus plantarum PFM 105 Promotes Intestinal Development Through Modulation of Gut Microbiota in Weaning Piglets

Tianwei Wang1,2† , Kunling Teng<sup>1</sup>† , Yayong Liu1,2, Weixiong Shi1,2, Jie Zhang<sup>1</sup> , Enqiu Dong<sup>3</sup> , Xin Zhang<sup>3</sup> , Yong Tao1,2 and Jin Zhong1,2 \*

<sup>1</sup> State Key Laboratory of Microbial Resources, Institute of Microbiology, Chinese Academy of Sciences, Beijing, China, <sup>2</sup> School of Life Science, University of Chinese Academy of Sciences, Beijing, China, <sup>3</sup> LongDa Foodstuff Group Co., Ltd, Laiyang, China

#### Edited by:

Jie Yin, Institute of Subtropical Agriculture (CAS), China

#### Reviewed by:

Hongmei Jiang, Hunan Agricultural University, China Hao Xiao, Guangdong Academy of Agricultural Sciences, China

> \*Correspondence: Jin Zhong zhongj@im.ac.cn

†These authors have contributed equally to this work as co-first authors

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 16 November 2018 Accepted: 16 January 2019 Published: 05 February 2019

#### Citation:

Wang T, Teng K, Liu Y, Shi W, Zhang J, Dong E, Zhang X, Tao Y and Zhong J (2019) Lactobacillus plantarum PFM 105 Promotes Intestinal Development Through Modulation of Gut Microbiota in Weaning Piglets. Front. Microbiol. 10:90. doi: 10.3389/fmicb.2019.00090 Lactobacillus plantarum is a widespread bacterial species and is commonly used as a probiotic. L. plantarum PFM105 was isolated from the rectum of a healthy sow. Here we found that L. plantarum PFM105 showed probiotic effect on weaning piglets in which intestinal inflammation and unbalanced gut microbiota happened frequently. L. plantarum PFM105 was identified to improve the growth of weaning piglet and promote the development of small intestinal villi. Antibiotics are often used in weaning piglet to prevent intestinal infection and promote the growth of animal. We found that weaning piglets feeding with L. plantarum PFM105 showed similar growth promotion but decreased diarrhea incidence compared with those feeding with antibiotics. High-throughput sequencing was used to analyze the gut microbiota in weaning piglets treated with L. plantarum PFM105 or antibiotics. The relative abundance of beneficial microbes Prevotellaceae and Bifidobacteriaceae were increased in colon of weaning piglet feeding L. plantarum PFM105, while antibiotics increased the relative abundance of bacteria associated with pathogenicity, such as Spirochaeta and Campylobacteraceae. L. plantarum PFM 105 increased indicators of intestinal health including serum levels of IgM, IL-10, and TGF-β, and colonic levels of SCFAs. We found strong correlations between the alterations in gut microbiota composition caused by feeding antibiotics and probiotics and the measured growth and health parameters in weaning piglets. The addition of L. plantarum PFM105 could significantly increase the relative abundance of metabolic genes which may important to intestinal microbiota maturation. Altogether, we demonstrated here that L. plantarum PFM 105 could promote intestinal development through modulation of gut microbiota in weaning piglets.

Keywords: Lactobacillus plantarum, antibiotics, weaning piglets, intestinal development, microbiota

# INTRODUCTION

Weaning is a critical and stressful event in the life cycle of mammals, including pigs, and is frequently associated with severe enteric infections and subsequent overuse of antibiotics (Gresse et al., 2017; Lariviere-Gauthier et al., 2018). Such periods of multiple stressors may induce transient anorexia, intestinal inflammation, and unbalanced gut microbiota (Su et al., 2008). The weaning

transition generally causes gastrointestinal (GI) infections, mainly by opportunistic pathogens, such as Proteobacteria, which are associated with the death of around 17% of piglets born in Europe in each year (Lalles et al., 2007; Gresse et al., 2017). In China, about 24 million weaning piglets die from bacterially induced diarrhea annually, resulting in economic losses of 12 billion yuan each year (Yin, 2010).

Antibiotics have been recognized as one of the most successful therapies in both human and veterinary medicine, but they can lead to the development of resistant bacterial strains within human and animal gut microbiota (Toutain et al., 2016). Infeed antibiotics can also reduce α diversity and cause shifts in gut microbiota due to their wide spectrum activity and their potential ability to kill or prevent the growth of both pathogenic and beneficial microbes (Looft et al., 2012; Neuman et al., 2018). Prolonged use of subtherapeutic doses of antibiotics can increase opportunities for pathogenic microorganisms to colonize and trigger diseases (Schokker et al., 2014). Overzealous feeding of antibiotics to food animals can lead to the emergence of antibiotic-resistant microbes (Modi et al., 2014; Gao et al., 2017). Antibiotic apramycin sulfate is widely used to prevent piglet diarrhea due to its broad-spectrum antibacterial activity (Herrero-Fresno et al., 2016). However, apramycin administration is likely driving the increasing occurrence of apramycin/gentamicin cross-resistance of Escherichia coli (Jensen et al., 2006) and Salmonella enterica serotype Typhimurium (Lim et al., 2013) in swine. The use of apramycin may also lead to enhanced spread of gentamicin-resistant E. coli (Herrero-Fresno et al., 2016). These processes turn food animal systems into reservoirs of antibiotic resistance genes, which can transfer to the human population through consumption and lead to serious public health problems (Modi et al., 2014; Toutain et al., 2016). Abuse of human and animal antibiotics has also led to the development of antibiotic-associated diarrhea. As a result, many countries are banning or have banned the inclusion of antibiotics in swine diets as growth promoters (Samanidou and Evaggelopoulou, 2008; Thacker, 2013).

During past two decades, numerous studies have focused on the development of alternatives to antibiotics to maintain swine health and performance (Thacker, 2013; Wang J. et al., 2018). The most widely researched non-antibiotic alternatives include probiotics, prebiotics, acidifiers, and essential oils (Valeriano et al., 2016; Gresse et al., 2017; Wang W. et al., 2018). Among these alternatives, probiotics have higher potential to act as feed additives against pathogens (Gresse et al., 2017; Azad et al., 2018). Probiotics are defined as "a live microorganism that, when administered in adequate amounts, confers a health benefit on the host," and are generally recognized as safe (GRAS) (FAO/WHO, 2002). Lactobacillus plantarum is a bacterium used as a probiotic, and is found in diverse ecological niches, such as mammal gastrointestinal tracts, dairy products, and vegetables. It has high adaptability and diversity of metabolic pathways (Seddik et al., 2017). L. plantarum has many probiotic characteristics including the ability to ferment a broad spectrum of plant carbohydrates, growth to high densities, tolerance of bile salts and low pH, and antagonistic potential against intestinal pathogens (Suo et al., 2012; van den Nieuwboer et al., 2016). L. plantarum ZJ316 can improve pig growth and pork quality, likely through inhibiting the growth of opportunistic pathogens and promoting increased villus height, rather than by altering the gut bacterial community (Suo et al., 2012). The metabolite combinations of mixed L. plantarum can improve growth performance and increase the population of gut lactic acid bacteria (LAB) and concentration of fecal short-chained fatty acids (SCFAs) of postweaning piglets (Thu et al., 2011). L. plantarum JC1 can increase villus height and the number of goblet cells, and improve the immune and inflammatory response by reducing intraepithelial lymphocytes and plasma TNF-α (Guerra-Ordaz et al., 2014). Although L. plantarum strains appear to have a high potential for replacement of antibiotics, few published studies have examined the effects of L. plantarum and antibiotics on weaning piglets.

In this study, we isolated the strain PFM105 from the rectum of a healthy sow and identified it as L. plantarum using 16S rDNA. We evaluated the effects of L. plantarum PFM105 and antibiotics on growth performance, clinical status, and intestinal morphology in weaning piglets. The colonic microbiota composition, metabolic capacity and the potential link between alterations in gut microbiota composition and health parameters in piglets feeding PFM 105 or antibiotics were also assessed. Weaning piglets feeding with L. plantarum PFM 105 showed elevated intestinal health and improved gut microbiota rather than those feeding with or without antibiotics.

# RESULTS

### Growth Performance and Clinical Status

The weaning piglets were grouped as NC group (the negative control group) in which the piglets were fed the base diet without any antibiotics or probiotics, PC group (positive control group) in which the piglets were fed the base diet plus antibiotics, and LP group in which the piglets were fed the base diet plus probiotic L. plantarum PFM105. All the tested weaning piglets were fed for 21 days. During the first 2 weeks, there were no significant differences among the different groups in the metrics of average daily gain (ADG) and feed conversion ratio (FCR) (**Table 1**). However, during the third week, although there was no significant difference in the average daily feed intake (ADFI) in three groups, the ADG (P < 0.05) was significantly increased and the FCR was trend toward reduced in piglets in LP group compared to those in NC group (**Table 1**). These results demonstrated that L. plantarum PFM 105 might improve piglet growth performance while antibiotics did not.

Piglets of the PC group did not show obvious difference in fecal score and diarrhea incidence with those of NC group. While piglets in LP group showed significantly decreased fecal score (P = 0.01) and diarrhea incidence (P = 0.01) compared to piglets in PC group (**Table 1**). Antibiotics (apramycin sulfate) and L. plantarum PFM105 reduced mortality of weaning piglets by 2.08 and 4.17%, respectively, compared to that of the NC group (**Table 1**). Thus, both L. plantarum PFM105 and antibiotics could reduce piglet mortality, but compared to antibiotics, L. plantarum


ADG, average daily gain; ADFI, average daily feed intake; FCR, feed conversion ratio. Different superscript letters within a row represent statistically significant differences among groups (P < 0.05). The fecal score of 0 corresponded to firm and dry, 1 to pasty, 2 to thick and fluid, and 3 to watery. The fecal score was calculated as the sum of the diarrhea score over the period divided by the number of piglet days in the period. The incidence of diarrhea (%) was calculated as the sum of the total number of diarrheal piglets over the period divided by the number of piglet days in the period multiplied by 100. The mortality (%) was calculated as the sum of the total number of dead piglets over the period divided by the number of piglets multiplied by 100.

PFM105 improved the clinical performance of piglets by reducing incidence of diarrhea and mortality.

#### Effects of Probiotics and Antibiotics on Intestinal Morphology

Weaning stress is known to induce remarkable morphological alterations in the small intestine, such as villus atrophy and crypt hyperplasia (Montagne et al., 2007; Gresse et al., 2017). L. plantarum PFM105 treatment significantly increased the villus length over that of the NC (P = 0.0445) and PC groups (P = 0.0209), but had little effect on the crypt depth and the ratio of villus to crypt in the jejunum (**Table 2**). L. plantarum PFM105 treatment showed a trend toward increased villus length over the antibiotic treatment (P = 0.0518) in the ileum (**Table 2**), while it had little effect on crypt depth and the ratio of villus to crypt in the ileum. Antibiotics had no effect on villus length, crypt depth, or the ratio of villus to crypt in piglet small intestine. These results demonstrated that L. plantarum PFM 105 may promote the development of small intestinal villi, while antibiotics do not.

#### Summary of Bacterial Community Richness and Biodiversity and β-Diversity

The microbiota of colonic contents in the three groups of piglets was analyzed by sequencing the bacterial 16S rDNA V3+V4 region. High-throughput pyrosequencing of the samples (n = 6) produced a total of 1,365,826 raw reads. After removing the low-quality sequences, 468,232 clean tags were identified as a total of 484 operational taxonomic units (OTUs) present in at least six samples. This sequencing depth almost reflected the total microbial species richness, and the majority of OTUs were present at low abundance, as demonstrated by the rarefaction, Shannon index, and rank abundance curves (**Supplementary Figure S1**). There were 385, 465, and 384 OTUs obtained from the NC, PC, and LP groups, respectively, of which 337 were common across the three experimental groups (**Supplementary Figure S2**). Moreover, a total of 71 unique OTUs were found within the NC, PC, and LP groups (5, 60, and 6, respectively). The Simpson alpha diversity index was


VH/CD, ratio of villus height to crypt depth. Different superscript letters within a row represent statistically significant differences among groups (P < 0.05).

higher for the LP group than for the NC group (P = 0.03). The Shannon, Chao 1, observed species (OS), and ACE values were not affected by treatment with either L. plantarum PFM105 or antibiotics (**Table 3**). To analyze the β-diversities of the colonic samples, we compared the Unweight Unifrac distances among colonic content samples collected from piglets. The microbial community structures of the NC and LP groups were mixed in the hierarchical clustering tree, while they were clearly distinguished from the PC group (**Supplementary Figure S3A**). Principal component analysis (PCA) based on phylum level (**Supplementary Figure S3B**) and OUT level (**Supplementary Figure S3C**) revealed that the gut microbiota in the NC and LP groups segregated from that of the PC group. The structure of the gut microbiota of piglets in the PC group was altered by use of antibiotics, while the probiotic L. plantarum PFM 105 did not detectably affect the gut microbiota structure.

#### Characterization of the Colonic Microbiota of Piglets

We measured relative abundance of colonic microbiota that occurred as more than 1% of the microbiota at the phylum (**Figure 1A**), family (**Figure 1B**), and genus (**Figure 1C**) levels. The colonic microbiota was dominated by the phyla Bacteroidetes and Firmicutes (regardless of treatment), which constituted 66.7 and 27.4% of the total abundance, respectively (**Figure 1A** and **Supplementary Table S1**). Proteobacteria was also common, making up 3.6% of the total abundance (**Supplementary Table S1**). The dominant families within the phylum Bacteroidetes consisted of Prevotellaceae, Bacteroidaceae, Bacteroidales S24-7 group, Porphyromonadaceae, and Rikenellaceae. The main families within the phylum Firmicutes were Lachnospiraceae, Ruminococcaceae, Acidaminococcaceae, Veillonellaceae, and Lactobacillaceae. The dominant families belonging to phylum Proteobacteria were Campylobacteraceae, Enterobacteriaceae, and Neisseriaceae (**Figure 1B**). Other phyla (Tenericutes, Actinobacteria, Spirochaetae, Cyanobacteria, Fibrobacteres, Fusobacteria, and Deferribacteres) were present at very low relative abundances (**Supplementary Table S1**).

The differences in the microbial communities at the phylum, family, and genus levels are shown in **Figures 2A–C**. Antibiotics led to a decrease in Bacteroidetes compared to their relative abundance in the NC group (P = 0.0497). Bacteroidetes occurred at higher levels in the LP group (75.98%) than in the PC group (57.42%, P = 0.0046, **Figure 2A** and **Supplementary Table S1**), though there was no significant difference between the LP and NC groups. The relative abundance of Proteobacteria in the PC group (7.54%) was higher than that in the other two groups (NC group: 0.865% and LP group: 0.954%), though there were no significant differences (**Figure 2A** and **Supplementary Table S1**). In addition, piglets fed with antibiotics (PC group) had a higher abundance of Spirochaetae (0.315%) compared to the NC (0%) and LP groups (0%) (P < 0.01, **Figure 2A** and **Supplementary Table S1**). We further compared the microbial community at the family level. Antibiotics and probiotics did not differ significantly in their effect on the relative abundance of Prevotellaceae; however, the relative abundance was higher in the LP group (67.66%) than in the PC group (50.87%) (P = 0.004, **Figure 2B** and **Supplementary Table S2**). The relative abundance of Campylobacteraceae was higher in the PC group (0.624%) compared to the NC (0.013%) and LP (0%) groups (P = 0.034, **Figure 2B** and **Supplementary Table S2**). The relative abundance of Bifidobacteriaceae was higher in the LP group (0.015%) than in the NC group (0%) (P = 0.0047), while there was no significant difference between the PC and NC groups (**Figure 2B** and **Supplementary Table S2**). The microbial communities were also compared at the genus level. Antibiotics and probiotics did not have any significant effect on the relative abundance of the Prevotellaceae NK3B31 group, the most abundant genus in colonic microbiota, but it trended toward higher abundance in the LP group than in the PC group (P = 0.056). The genera Phascolarctobacterium, Treponema\_2, Sutterella, and Parasutterella exhibited increased relative abundances in the PC group compared to the NC group (**Figure 2C** and **Supplementary Table S3**). The relative abundance of genus Bifidobacterium was increased and the genus Eubacterium\_hallii was decreased in piglets of the LP group compared to the NC group (**Figure 2C** and **Supplementary Table S3**). In addition, there was a remarkable increase in the relative abundance of the genus Campylobacter in the PC group compared to the LP group. Linear discriminant analysis (LDA) effect size (LefSe) analysis was also performed to confirm the different effects of antibiotics and probiotic on intestinal microbiota in piglets (**Figure 3**). Interestingly, symbiotic (Prevotellaceae) and beneficial (Bifidobacteriaceae) bacteria were elevated in the LP group, while harmful bacteria (Spirochaetae and Campylobacteraceae) were increased in the PC group.

#### TABLE 3 | Alpha diversity indices of the colonic microbiota of weaning piglets.


Coverage percentage, richness estimators (Chao1, OS, and ACE), and diversity indices (Shannon and Simpson) were calculated using the mothur program. OS, observed species. Data in the same column that do not share a common superscript differ significantly (P < 0.05).

#### Comparison of Metabolic Pathway Abundances

We predicted the microbial metagenome with 16S rRNA gene sequencing using phylogenetic investigation of communities by reconstruction of unobserved states (PICRUSt) (Langille et al., 2013), and found that even with widespread differences in bacterial composition, most functional genes were largely conserved across different groups (**Supplementary Table S4**). However, the relative abundance of genes related to metabolism was higher in the LP group (49.99%) than in the PC group (48.03%) (P < 0.01, **Supplementary Table S4**). To further study which metabolic genes changed after treatment with probiotics, 40 KEGG Orthology (KO) groups were selected (**Supplementary Table S5**). We found that, in terms of metabolic pathways, genes that regulated metabolism of cofactors and vitamins, glycan biosynthesis and metabolism, metabolism of other amino acids, metabolism of terpenoids and polyketides, and biosynthesis of other secondary metabolites were more abundant in the LP group than in the PC group (**Figure 4**). We additionally found that genes related to lipid metabolism were less frequent in the LP group than in the NC group (**Figure 4**).

#### Effects of Probiotics and Antibiotics on SCFA in Colonic Content

To analyze the effects of probiotic L. plantarum PFM105 and antibiotics on intestinal microbiota metabolism, we focused on the colonic content of Short-chained fatty acids of weaning piglets. Levels of acetic acid, butyric acid, and total SCFAs were higher in the LP group than in the NC and PC groups (P < 0.05, **Table 4**). However, the acetic and butyric acids and total SCFAs

did not differ between the PC and NC groups. There was no significant difference in the colonic content concentrations of propanoic, isobutyric, valeric, and isovaleric acids among the different groups (**Table 4**). These results demonstrated that L. plantarum PFM 105 may promote microbial metabolism and result in increased production of SCFAs (acetic acid and butyric acid), while antibiotics did not have this effect.

#### Effects of Probiotics and Antibiotics on Immunoglobulins, Cytokines, and Intestinal Permeability-Related Biomarkers

The effects of antibiotics and probiotics on the humoral immunity levels were evaluated by detecting the content of Immunoglobulin G (IgG), Immunoglobulin A (IgA), and Immunoglobulin M (IgM) in serum. The impacts on intestinal immunity were evaluated by detecting the content of secretory IgA (sIgA) in colonic samples (**Figure 5A**). L. plantarum PFM105 treatment significantly increased the total serum IgM (P = 0.0447) antibody levels when compared to the NC group (**Figure 5A**). L. plantarum PFM105 treatment also trended toward increasing the intestinal sIgA (P = 0.0690) antibody levels as compared to the NC group (**Figure 5A**). However, the content of serum IgM and intestinal sIgA did not differ between the PC and NC groups. No differences were observed in serum IgG and IgA concentration among the different groups (**Figure 5A**). We next detected six biomarkers related to gut health of weaning piglets. These biomarkers included a set of serum cytokines [interleukin 2 (IL-2), interleukin 6

TABLE 4 | Concentrations of short-chain fatty acids (SCFAs) in the colonic contents of weaning piglets under the different treatments (mg/g dry weight).


Different superscript letters within a row represent statistically significant differences among groups (P < 0.05).

(IL-6), interleukin 10 (IL-10), and transforming growth factor β (TGF-β)] as markers for the immune system activation and systemic inflammatory response, serum diamine oxidase (DAO) as a marker for intestinal mucosal integrity, and colonic content of lipocalin-2 as a marker for intestinal inflammation. L. plantarum PFM105treatment increased the levels of antibody production mediated IL-2 (P = 0.0322), anti-inflammatory mediator IL-10 (P = 0.0437), and immune tolerance mediator TGF-β (P = 0.0204) over those of the NC group (**Figure 5B**). However, the levels of IL-2, IL-10, and TGF-β did not differ between the PC and NC groups. Differences in systemic pro-inflammatory cytokines IL-6 (**Figure 5B**), intestinal permeability marker DAO, and intestinal inflammation marker lipocalin-2 (**Figure 5C**) were not observed among these groups. These results demonstrated that L. plantarum PFM 105, but not antibiotics, may enhance humoral immunity, and prevent intestinal inflammation and excessive systemic immune response, thus promote intestinal health of piglets.

#### Alterations in Intestinal Microbiota Composition Were Correlated With Health Parameters

A Spearman's rank correlation analysis was performed to evaluate the potential link between alterations in gut microbiota composition and growth and health parameters of weanling piglets (**Figure 6**). The genus Bifidobacterium was positively correlated with increased levels of TGF-β (P < 0.01). The genus Prevotellaceae NK3B31 group was positively correlated with increased biosynthesis of other secondary metabolites (P < 0.05) and acetic acid (P < 0.05). The genus Campylobacter was positively correlated with fecal score (P < 0.05) but negatively correlated with genes of metabolism of cofactors and vitamins

groups (n = 6). (B) Serum cytokines IL-2, IL-6, IL-10, and TGF-β levels among the three groups (n = 6). (C) Levels of intestinal permeability-related biomarkers DAO and colonic content of Lipocalin-2 among the three groups (n = 6). <sup>∗</sup>P < 0.05. IgG, immunoglobulin G; IgA, immunoglobulin A; IgM, immunoglobulin M; sIgA, secretory IgA; TGF-β, transforming growth factor β; IL-2, interleukin 2; IL-6, interleukin 6; IL-10, interleukin 10; DAO, diamine oxidase.

and genes of metabolism of other amino acids (P < 0.05). The phylum Spirochaetae and the genus Treponema\_2 both were positively correlated with fecal score, and negatively correlated with isovaleric acid (P < 0.01).

#### DISCUSSION

Lactobacillus plantarum PFM105 significantly improved the ADG of piglets as compared to that of the NC group during the third week. Previous studies have shown L. plantarum can improve the growth performance of piglets, which aligns with our results (Lee et al., 2012; Suo et al., 2012). Probiotics may improve growth via promoting nutrient absorption by increasing villus height (Suo et al., 2012; Liu et al., 2014). We found that L. plantarum PFM105 significantly increased the villus height in the jejunum and trended toward increasing villus height in the ileum, which may be the reason for the increased body weight of these piglets. In this study, antibiotics did not show effects on the development of small intestinal villi or promote growth. Compared to antibiotics, L. plantarum PFM105 is likely a better food additive to

promote the intestinal development and growth of weaning piglets.

Weaning is usually associated with intestinal disorders as the piglet intestinal microbiota undergoes substantial dynamic changes (Chen et al., 2017). During the experiments, there were two piglets died in the NC group at day 18 and 20, respectively, and a piglet died in the PC group at day 11. The sample size was insufficient to detect differences between groups, but both L. plantarum PFM105 and antibiotics could be able to reduce the deaths caused by weaning stress. Previous study showed that piglets fed with L. plantarum ZJ316 experienced reduced mortality than those fed with antibiotics, which aligns with our findings (Suo et al., 2012). Additionally, L. plantarum PFM105 reduced the diarrhea rate compared to that of the PC group. A previous study also showed L. plantarum ZJ316 decreased the diarrhea rate more than antibiotics did (Suo et al., 2012). Other probiotics, such as Bacillus licheniformis-B. subtilis mixture ameliorated enteritis caused by an enterotoxigenic E. coli strain (F4<sup>+</sup> ETEC) (Zhang et al., 2017). Probiotics, including L. plantarum PFM105, may be a more effective means of reducing mortality and diarrhea in weaning piglets.

Piglets treated with L. plantarum PFM105 showed an increased Simpson's diversity index in the gut microbiota compared to that of piglets in the NC group. This could represent a benefit for the weaned animals because of the possible link between the diversity (i.e., degree of simplification) of ecosystems and their ability to respond to perturbations (McCann, 2000; Pieper et al., 2009). Similar results were also reported in previous studies demonstrating that probiotics can increase the Simpson's diversity index of the microbial ecosystem in piglets (Pieper et al., 2009; Liu et al., 2014). In previous studies, Firmicutes, and Bacteroidetes were the dominant groups in bacterial communities at the phylum level and were not significantly altered by use of probiotics (Wang J. et al., 2018) or of antibiotics (Looft et al., 2012; Holman and Chenier, 2014) in weaning piglets. However, here we found that antibiotics led to a decrease in the relative abundance of Bacteroidetes and an increase in the relative abundance of Proteobacteria and Spirochaetae. Representatives of Bacteroidetes provide the host with SCFAs that can supply up to 10% of daily calories through the fermentation of indigestible polysaccharides (McNeil, 1984; Johnson et al., 2017). The anaerobic Proteobacteria are usually associated with an impaired microbiota, or dysbiosis (Litvak et al., 2017). Feeding antibiotics

could cause impaired intestinal eubiosis; E. coli (phylum Proteobacteria) increased after pigs were treated with antibiotics (Zhang et al., 2018). We also analyzed lower taxonomic levels and found that the relative abundance of Prevotellaceae in the LP group was higher than in the PC group. Prevotellaceae was the dominant bacteria at the family level (accounting for nearly 60%) in our study, and recent metagenomic studies confirmed the prevalence of Prevotellaceae in the cecum, colon, and feces of pigs (Looft et al., 2014). Representatives of Prevotellaceae are associated with hemicellulose degradation and are the predominant bacteria in piglets at the nursery stage (Konstantinov et al., 2004). A high Prevotella spp. abundance may be essential for post-weaning piglets to be able to digest plant-based diets (Wang J. et al., 2018). High abundance of this bacterium in the LP group suggested that piglets supplemented with L. plantarum PFM105 may have strong digestion and absorption capacity.

Opportunistic pathogens, such as Phascolarctobacterium (significantly correlated with systemic inflammatory cytokines, such as TNF-α (Ling et al., 2016)), Campylobacter (ubiquitous in nature and in domestic animals, but also important in infections in animals) and Treponema\_2 which belongs to phylum Spirochaetae and has been isolated from pig lesions (Svartstrom et al., 2013) were increased after antibiotic treatment compared to the NC group. Certain Campylobacter (C. coli and C. hyoilei) can cause gastrointestinal infection and thus lead to gastroenteritis (Humphrey et al., 2007). Campylobacter, Treponema\_2, and Spirochaeta were positively correlated with fecal score; thus, we hypothesized that these pathogens disturbed the gut microbiome in piglets treated with antibiotics. These opportunistic pathogens were not changed in the LP group when compared to the NC group. By contrast, the well-studied probiotic Bifidobacterium increased after L. plantarum PFM105 treatment when compared to the NC group, and the relative abundance of Bifidobacterium was positively correlated with TGF-β. These beneficial bacteria were not changed in the PC group when compared to the NC group. These results were supported by previous reports indicating that Bifidobacterium shows higher anti-inflammatory capacity by inducing intestinal production of IL-10 and TGF-β (Finamore et al., 2012; Herfel et al., 2013). Thus, we hypothesized that L. plantarum PFM105 could increase beneficial bacteria, which may assist with energy harvesting and boost anti-inflammatory capacity, while antibiotics could increase pathogenic bacteria, potentially leading to intestinal dysbiosis.

We utilized inferred metagenomics by PICRUSt (Langille et al., 2013) which can reflect the metabolic activities of the microbiota (Peng et al., 2018) to investigate functional differences in the microbiota of piglets in order to determine the metabolic alterations caused by antibiotics or probiotics. We found that even with widespread differences in bacterial community composition, overall function was largely conserved across individuals, and these results were consistent with previous studies in humans (Human Microbiome and Project, 2012). There were no significant differences in metabolic genes between the PC and NC groups. However, metabolic genes were more abundant in the LP group compared to PC group, implying that microbial metabolism tended to be more vigorous after treatment with L. plantarum PFM105. We found that genes of cofactors and vitamins metabolism and glycan biosynthesis and metabolism were more abundant in the LP group than in the PC group. Previous studies have shown that vitamins and cofactors, especially of the vitamin B family, are critical for the bioconversion of nutrients to energy and for maintaining homeostasis (McDonald, 2009; Hu et al., 2016). The glycan biosynthesis and metabolism genes are important for carbohydrate metabolism (Hu et al., 2016). These results indicated that L. plantarum PFM105 may promote energy metabolism. Previous studies have also shown that these genes were more abundant in aged pigs (Hu et al., 2016). Gao et al. (2017) found that the maturation of intestinal microbiota was greatly accelerated by probiotic (L. plantarum LP-8) feeding, yet significantly delayed by antibiotic feeding. Probiotic L. plantarum PFM105 might accelerate intestinal microbiota maturation by increasing important metabolic genes.

Short-chained fatty acids are important for gut integrity, glucose homeostasis, and immune function (Morrison and Preston, 2016). They are produced by the colonic anaerobic microbial community through fermenting indigestible fiber matter and some luminal amino acids (Blachier et al., 2007; Kong et al., 2016). In our study, acetic and propanoic acids were the major SCFAs produced in the colon, which was consistent with previous findings in the colon of pregnant Huanjiang minipigs (Kong et al., 2016) and in the feces of primiparous sows (Paßlack et al., 2015). In the current study, the levels of these SCFAs were higher in the LP group than in the NC and PC groups. Acetic acid is reported to inhibit pathogenic bacteria, and butyric acid acts as a major energy source for colonic epithelial cells (Morrison and Preston, 2016). The increasing quantity of butyric acid in colonic content after the administration of L. plantarum PFM105 is consistent with previous findings in pigs administrated with Lactobacillus reuteri I5007 (Liu et al., 2014, 2017). The increased content of acetic acid and butyric acid in the LP group might due to the increased proportion of family Prevotellaceae which is known to be important for polysaccharide degradation and SCFAs formation. All identified enzymes involved in polysaccharide (starch) degradation are associated with Prevotellaceae (Ivarsson et al., 2014; Heinritz et al., 2016). Our results showed that the Prevotellaceae NK3B31 group was positively correlated with acetic acid, which was also demonstrated in a recent study in piglets administrated with L. plantarum ZLP001 (Wang J. et al., 2018). Though the unique interaction between L. plantarum and Prevotellaceae is still unclear, we hypothesize that L. plantarum PFM105 might promote the production of SCFAs by increasing Prevotellaceae.

Immunoglobulin G, Immunoglobulin A, and Immunoglobulin M represent the main antibody isotype found in blood and extracellular fluid, the predominant immunoglobulin isotype expressed in mucosal tissues, and the major component of natural antibodies, respectively. They are the main immunoglobulins involved in humoral immunity (Luo et al., 2013). In our study, the expression level of IgM was increased in LP group piglets compared to NC group piglets, indicating that the humoral immunity

level of piglets was improved by L. plantarum PFM105. Zhu et al. (2017) found that feeding pigs a diet containing probiotics could increase serum IgM levels, which supports our findings. Considering recent evidence suggesting that microbiota can modulate intestinal barrier integrity, and improve immunology tolerance of newborn individuals (Huang et al., 2015; Cheng et al., 2018), we further assessed the gut permeability and intestinal or systemic inflammatory response of piglets. Cytokines play a crucial role in immune and inflammatory responses, and their balance is important for protection against infection. Cytokine IL-2 is critical for regulating lymphoid homeostasis (Ma et al., 2006). IL-10, an anti-inflammatory cytokine, can prevent over-activation of the immune response and suppress the production of pro-inflammatory cytokines and thus plays an integral role in maintenance of immune homeostasis (Opal and Depalo, 2004). TGF-β exerts systemic immune suppression and inhibits host immunosurveillance (Yang et al., 2010). In the present study, we found that the expression levels of IL-2, IL-10, and TGF-β were increased in the LP group over those of the NC group. We hypothesized that the increased pro-inflammatory and anti-inflammatory cytokines caused by L. plantarum PFM105 treatment may have decreased susceptibility to pathogenic infection. Previous studies showed that probiotic Bifidobacterium can increase serum TGF-β levels (Ouwehand et al., 2008) and probiotic L. plantarum enhance IL10 production (Bosch et al., 2012; Noguchi et al., 2012), which also supports our results.

# CONCLUSION

Weaning represents a major challenge to a developing pig acclimating to gastrointestinal microbial colonization, and is often associated with gastrointestinal disorders. In pig husbandry, antibiotics are commonly used to alleviate weaning stress, but our study found that the use of probiotic L. plantarum PFM 105 may be more effective. L. plantarum PFM 105 could strongly improve the development of small intestinal villi and the growth performance of weaning piglets. Compared to antibiotics, L. plantarum PFM 105 enhanced piglet clinical performance, reduced mortality, and lowered incidence of diarrhea. L. plantarum PFM 105 might also enhance humoral immunity, prevent intestinal inflammation, and avert excessive systemic immune response. L. plantarum PFM 105 modulated the piglet gut microbiota by increasing the abundance of symbiotic and beneficial bacteria (e.g., Prevotellaceae and Bifidobacteriaceae), while antibiotics increased the occurrence of harmful bacteria (e.g., Spirochaetae and Campylobacteraceae). L. plantarum PFM 105 increased the expression levels of genes related to metabolism of cofactors and vitamins, and to glycan biosynthesis and metabolism, which may enhance the metabolic capacity of the microbiota in piglets. Our results demonstrate the possibility of using L. plantarum PFM 105 instead of antibiotics to promote intestinal development and modulate gut microbiota in weaning piglets.

# MATERIALS AND METHODS

# Statement of Ethics for the Care and Use of Animals

The experimental procedures used in this study were approved by the Laboratory Animal Ethical Commission of the Chinese Academy of Sciences and performed according to its guidelines. Humane animal care was practiced throughout the trial and every effort was made to minimize suffering for piglets.

#### Animals, Diets and Sampling

A total of 144 normal weaning piglets (72 males and 72 females) from 28 litters (Landrace × Large White, 28 days of age, 8.22 ± 0.38 kg) were obtained from the LongDa Foodstuff Group Co., Ltd (Shandong Province, China) and were allocated randomly to three groups for the 21 days trial, balancing for litter and gender. The negative control (NC) group was fed the base diet without any antibiotics or probiotics. The positive control (PC) group was fed the base diet plus the antibiotic apramycin sulfate at 25 mg/kg BW in feed. The probiotic (LP) group was fed the base diet plus the probiotic strain L. plantarum PFM105 (CGMCC 16113, isolated from healthy sow intestine) in feed. Prior to the start of the trial, no clinical signs of diarrhea or other diseases were observed in any of the piglets. All pigs in this study were selected from one delivery room and had similar genetic backgrounds and husbandry practices. Each group included 48 piglets in six replicates (8 piglets in each replicate and housed in a pen). These 18 pens were in the same nursing house. Room temperature was maintained at 26◦C, and the humidity was maintained constant at 65–75%. All pigs were fed four times a day with customized corn-soybean feed (free of probiotics and antibiotics) containing 19% crude protein, and details are provided in the supplementary material for ingredients and nutrient composition (**Supplementary Table S6**) (Yin et al., 2017). Water was available ad libitum from nipple drinkers. Lyophilized LP was provided by the Center for Technology Transfer and Transformation of Institute of Microbiology, Chinese Academy of Sciences (Beijing, China) and was added to the feed for piglets at a final concentration of 2 × 10<sup>7</sup> CFU/g. To ensure dose accuracy, the concentration of live bacteria in the powder was verified based on culture-based counting. Moreover, to verify the purity of the probiotic preparation, 10 clones were randomly picked from a de Man, Rogosa and Sharpe (MRS) plate derived from the bacterial freeze-dried powder. Genomic DNA was extracted from each of the clone and 16S genes were amplified via a universal bacterial 16S PCR primer (27F/1492R, listed in **Supplementary Table S7**) and sequenced by Beijing Ruiboxingke Biotechnology Co. Ltd (Beijing, China). All 10 clones were confirmed as L. plantarum, which validated the purity of our probiotic preparation.

All piglets in each pen were weighed individually at days 0, 14, and 21 during the trial. The feed consumed by each pen was monitored daily. ADG, ADFI, and feed conversion ratio (FCR; feed consumed/weight gain) were calculated for the periods of 1–14, 14–21, and 1–21 days. The health status of piglets during the experiments was assessed by fecal consistency scoring using

a four-grade system, where 0 corresponded to firm and dry, 1 to pasty, 2 to thick and fluid, and 3 to watery (De Cupere et al., 1992). The fecal score was calculated as the sum of the diarrhea score over the period divided by the number of piglet days in the period. The occurrence of diarrhea was defined as maintaining a score of 3 for 1 day (Liu et al., 2017). The incidence of diarrhea (%) was calculated as the sum of the total number of diarrheal piglets over the period divided by the number of piglet days in the period multiplied by 100. The mortality (%) was calculated as the sum of the total number of dead piglets over the period divided by the number of piglets multiplied by 100.

On day 21, one median-weight piglet from each replicate from different pens was sacrificed and 10 ml of blood was collected. The luminal content samples were collected at the same site for each gut location. Briefly, after opening the visceral cavity, esophagus and rectum were clamped to avoid spilling of gastrointestinal digesta and thus contamination of other intestinal parts. Immediately after removing the gastrointestinal tract (GIT) from the visceral cavity, the mid-jejunum, mid-ileum, mid-colon, and rectum were separated by clamping to avoid mixing of digesta from adjacent segments of the GIT. The luminal contents were separately gathered from the middle section of the colon. Subsequently, intestinal segments were disclosed at the mesentery with sterile instruments and digesta was removed. The experimental platform was disinfected before each sample was collected to avoid cross-contamination between samples. All samples were harvested within 30 min of slaughtering and transferred immediately to liquid nitrogen for temporary storage. Samples were then sent to the laboratory where they were stored at −80◦C until analysis.

#### Detection of Immunoglobulin, Cytokines, and DAO in Serum and of Intestinal SIgA and Lipocalin-2 in Colonic Contents

Serum was collected and centrifuged at 4000 rpm for 10 min at 4◦C before being stored at -80◦C until IgG, IgA, and IgM were quantified. Serum total IgG and IgA were detected by porcine enzyme-linked immunosorbent assay (ELISA) kits (Nanjing Jiancheng Bioengineering Institute, Nanjing, China). Colonic content (1 g) was collected and mixed with an equal volume of PBS and centrifuged at 1000 rpm for 15 min. The supernatant was then stored at -80 ◦C until sIgA was quantified by porcine ELISA kits (Jiangsu Meimian Industry Co., Ltd, Nanjing, China). IL-2, IL-6, IL-10, TGF-β, DAO, and lipocalin-2 concentrations were determined using porcine ELISA kits according to the manufacturer's instructions (Jiangsu Bo Deep Biological Technology Co., Ltd., Nanjing, China). Their concentrations were then calculated from the standard curves. All procedures were performed with 3 repetitions.

# Quantification of SCFAs in Colonic Content Samples

Short-chained fatty acids including acetic acid, propanoic acid, butyric acid, isobutyric acid, valeric acid, and isovaleric acid were analyzed as described previously with minor modifications (Wang K. et al., 2018; Yin et al., 2018). Colonic contents were collected from the middle segment of the colon from piglets after 21 days of treatment. Each sample was lyophilized and then pestled using a mortar. 100 mg of the homogenic powders was extracted with 1 mL of methanol (gradient grade for liquid chromatography LiChrosolv <sup>R</sup> Reag. Ph Eur EA). After 10 min sonication, the samples were centrifuged (6000rpm for 10 min), and the supernatants were used for GC-MS analysis. GC-MS was performed on a GC-MS-QP2010 Ultra with an autosampler (SHIMADZU) and the Rtx-wax capillary column (30 m, 0.25 mm i.d., 0.25 µm film thickness; SHIMADZU). Oven temperature was programmed from 60 to 100◦C at 5◦C/min, with a 1 min hold; to 150◦C at 5 ◦C/min, with a 5 min hold; to 225◦C at 30◦C/min, with a 20 min hold. Injection of a 2 µL sample was performed at 230◦C. Helium, at a flow of 1.2 mL/min, was the carrier gas. Electronic impact was recorded at 70 eV.

# Genomic DNA Extraction and 16S rRNA Gene Sequencing

Colonic content DNA was extracted from 0.2 g of sample using the protocol of the QIAamp PowerFecal DNAKit (Qiagen, Dusseldorf, Germany). DNA was eluted in ddH2O and stored at -80◦C until use. The V3-V4 hypervariable region of the 16S rDNA gene was targeted and the primers used are listed in **Supplementary Table S7**. The bacteria 16S ribosomal RNA gene was amplified. PCR reactions were performed in triplicate in a 20 µL mixture containing 4 µL of 5 × FastPfu Buffer, 2 µL of 2.5 mM dNTPs, 0.8 µL of each primer (5 µM), 0.4 µL of FastPfu Polymerase, and 10 ng of template DNA. Amplicons were extracted from 2% agarose gels and purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, United States) according to the manufacturer's instructions and quantified using QuantiFluorTM -ST (Promega, Madison, WI, United States). Purified amplicons were pooled in equimolar concentrations and paired-end sequenced (2 × 250) on an Illumina MiSeq platform according to the standard protocols.

#### Bioinformatics Analysis

Paired-end reads were merged using FLASH (Magoc and Salzberg, 2011), which was designed to merge paired-end reads when at least some of the reads overlap the read generated from the opposite end of the same DNA fragment, and the splicing sequences were called raw tags. Quality filtering on the raw tags were performed under specific filtering conditions to obtain the high-quality clean tags according to the QIIME quality-control process (Bokulich et al., 2013). The tags were compared with the reference database (Gold database) using the UCHIME algorithm to detect chimera sequences which were later removed (Edgar et al., 2011; Haas et al., 2011). OTUs were clustered with a 97% similarity cutoff using UPARSE (Edgar, 2013) and chimeric sequences were identified and removed using UCHIME. The taxonomy of each 16S rRNA gene sequence was analyzed by RDP Classifier against the SILVA (SSU115) 16S rRNA database using a confidence threshold of 70% (DeSantis et al., 2006; Wang et al., 2007). Sequences with higher than 97% similarity were assigned to the same OTU. A representative sequence for each OTU was screened for further annotation. OTU abundance information was normalized using a standard of sequence number corresponding to the sample with the fewest sequences. Subsequent analysis of alpha diversity was performed based on this output normalized data. Alpha diversity was analyzed through six indices, including observed-species (OS), Chao1, Shannon, Simpson, ACE, and Good-coverage. β-diversity was analyzed by PCA was conducted based on phylum and OTU, and the hierarchical clustering tree was constructed based on Unweight Unifrac distances. All of these indices in our samples were calculated with QIIME (Version 1.7.0) and displayed with R software (Version 2.15.3). The dominant bacterial community difference between groups was detected using LDA effect size (LefSe). The biomarkers used in the present study had an effectsize threshold of three.

#### Morphological Analyses

fmicb-10-00090 February 5, 2019 Time: 13:20 # 14

Piglet jejunums and ileum were prepared using our previous described methods (Wang T. et al., 2018). Briefly, they were fixed with 10% paraformaldehyde-PBS overnight and then dehydrated and embedded in paraffin blocks. A 5 µm section was cut, deparaffinized, hydrated, and then stained with hematoxylin and eosin (H&E). Villus length and crypt depth of at least three villi or crypts per slide were measured using Image-Pro Plus software 6. Six piglets were studied from each group. The data collectors were unaware of the treatment status of the examined slides.

#### Statistical Analysis

Data shown are means ± standard deviation (SD) or standard error of the mean (SEM). Data were analyzed by one-way ANOVA followed by Dunnett multiple comparisons (Prism 7.0) if the data were in Gaussian distribution and had equal variance or analyzed by the Kruskal-Wallis test followed by Dunn's multiple comparisons (Prism 7.0) if the data were not normally distributed. The Gaussian distribution of data was analyzed by the Kolmogorov-Smirnov test (Prism 7.0). The variance of data

#### REFERENCES


was analyzed by homogeneity of variance test (SPSS 19.0) or the Brown-Forsythe test (Prism 7.0). Differences with P < 0.05 were considered significant. Statistical evaluation of the incidence of diarrhea was performed using Pearson's chi-square test. The nonparametric Friedman's test using procedure FREQ was carried out to compare non-normally distributed and repeated-measure diarrhea scores. Correlations were analyzed by using Spearman's correlation in R 3.4.4 (The R Foundation) with the RStudio psych package and pheatmap for the heat map. Correlation results were corrected by FDR analysis according to the Benjamini-Hochberg procedure, with an α of <0.05.

# AUTHOR CONTRIBUTIONS

JinZ, KT, and TW designed the study. TW and KT wrote the manuscript. TW, YL, JieZ, ED, WS, and MZ performed the experiments. YT and JinZ edited the manuscript. All authors have discussed the results and reviewed the manuscript.

# FUNDING

This work was supported by grants from the Special Fund for Agro-scientific Research in the Public Interest (No. 201503134) and the National Natural Science Foundation of China (Nos. 31570114 and 31772642) and Guangxi Major Science and Technology project (AA18118041).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.00090/full#supplementary-material

weaned piglets. Zentralbl. Veterinarmed. B 39, 277–284. doi: 10.1111/j.1439- 0450.1992.tb01169.x


a potential complementary synbiotic to control postweaning colibacillosis in piglets. Appl. Environ. Microbiol. 80, 4879–4886. doi: 10.1128/AEM.00770-14


cytokine response of weaning pigs challenged with enterotoxigenic Escherichia coli. J. Anim. Sci. 90, 3709–3717. doi: 10.2527/jas.2011-4434



metabolic dysfunction in vivo. J. Med. Chem. 61, 3609–3625. doi: 10.1021/acs. jmedchem.8b00107


**Conflict of Interest Statement:** XZ and ED were employed by company LongDa Foodstuff Group Co., Ltd.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Wang, Teng, Liu, Shi, Zhang, Dong, Zhang, Tao and Zhong. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Diversity of the Gut Microbiota in Dihydrotestosterone-Induced PCOS Rats and the Pharmacologic Effects of Diane-35, Probiotics, and Berberine

Feifei Zhang1,2,3, Tong Ma<sup>4</sup> , Peng Cui<sup>4</sup> , Amin Tamadon<sup>4</sup> , Shan He<sup>4</sup> , Chuanbing Huo<sup>4</sup> , Gulinazi Yierfulati<sup>4</sup> , Xiaoqing Xu<sup>4</sup> , Wei Hu<sup>4</sup> , Xin Li2,3, Linus R. Shao<sup>5</sup> , Hongwei Guo<sup>1</sup> , Yi Feng<sup>4</sup> \* and Congjian Xu2,3 \*

<sup>1</sup> School of Public Health, Fudan University, Shanghai, China, <sup>2</sup> Obstetrics and Gynecology, Shanghai Medical College, Fudan University, Shanghai, China, <sup>3</sup> Shanghai Key Laboratory of Female Reproductive Endocrine Related Diseases, Shanghai, China, <sup>4</sup> Department of Integrative Medicine and Neurobiology, School of Basic Medical Sciences, Institutes of Brain Science, Brain Science Collaborative Innovation Center, State Key Laboratory of Medical Neurobiology, Fudan Institutes of Integrative Medicine, Fudan University, Shanghai, China, <sup>5</sup> Department of Physiology/Endocrinology, Institute of Neuroscience and Physiology, The Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden

#### Edited by:

Yuheng Luo, Sichuan Agricultural University, China

#### Reviewed by:

Adam C. N. Wong, University of Florida, United States Dingkun Gui, Shanghai Sixth People's Hospital, China

#### \*Correspondence:

Yi Feng fengyi17@fudan.edu.cn Congjian Xu xucongjian@fudan.edu.cn

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 10 August 2018 Accepted: 22 January 2019 Published: 08 February 2019

#### Citation:

Zhang F, Ma T, Cui P, Tamadon A, He S, Huo C, Yierfulati G, Xu X, Hu W, Li X, Shao LR, Guo H, Feng Y and Xu C (2019) Diversity of the Gut Microbiota in Dihydrotestosterone-Induced PCOS Rats and the Pharmacologic Effects of Diane-35, Probiotics, and Berberine. Front. Microbiol. 10:175. doi: 10.3389/fmicb.2019.00175 Polycystic ovary syndrome (PCOS) is a frequent endocrine and metabolic syndrome in reproductive-age women. Recently, emerging evidence has shown that gut microbiota is closely related to metabolic diseases such as type 2 diabetes, obesity and PCOS. In the present study, we established dihydrotestosterone (DHT)-induced PCOS rats and used Illumina MiSeq sequencing (PE300) to examine the composition, diversity, and abundance of the gut microbiota in PCOS. We compared the effects of three PCOS treatments: Diane-35 (estrogen and progesterone), probiotics and berberine. The DHTinduced rats showed constant estrous cycles, the loss of mature ovarian follicles, insulin resistance and obesity. The reproductive and metabolic functions in the PCOS rats were improved by treatment with Diane-35 and probiotics. Diane-35 and probiotics could restore the diversity of the gut microbiota, and the recovery of gut microbiota disorders improved the reproductive function in PCOS-like rats. However, berberine drastically reduced the species diversity and amount of gut microbiota and showed no improvement in PCOS. The findings of this study will help us to better understand the influence of the gut microbiota in the metabolic and reproductive alterations in PCOS

Keywords: gut microbiota, PCOS, DHT, Diane-35, probiotics, berberine

# INTRODUCTION

Polycystic ovary syndrome (PCOS) is a common endocrine and metabolic syndrome among women of reproductive-age, with a worldwide incidence of 5–15% (Michelmore et al., 1999; March et al., 2010; Li R. et al., 2013). The principal characters are clinical or biochemical androgen excess, ovulation disorders and polycystic ovarian change (Conway et al., 2014; Skubleny et al., 2016), and it is associated with abdominal obesity, insulin resistance, impaired glucose

as well as suggest opportunities for future personal dietary guidance for PCOS.

metabolism, and dyslipidemia (Tan et al., 2010). The pathologic state of PCOS is a life-long condition, leading to an increased risk of hyperlipidemia, cardiovascular disease, hypertension, metabolic syndrome and endometrial cancer (Ali, 2015).

The pathogenesis of PCOS is still poorly understood, but the majority of researchers agree that it is a multifactorial disorder mainly induced by genetic and environmental factors. Even in lean PCOS patients, hepatic insulin resistance is also present (Conway et al., 2014). Compared to healthy lean and obese women, both lean and obese women with PCOS show a clustering of metabolic disorders, including high of total cholesterol, lowdensity lipoprotein cholesterol, triglycerides and low levels of high-density lipoprotein cholesterol levels (Diamanti-Kandarakis et al., 2007). Lifestyle modification, anti-androgenic agents, insulin-sensitizing agents, anti-hypertensives, and statins are all optional therapies in the clinic (Moghetti et al., 2000; Lord et al., 2003; Ibanez and de Zegher, 2004; Teede et al., 2010; Bargiota and Diamanti-Kandarakis, 2012), but caloric restriction and regular exercise are the primary recommendations (Morgante et al., 2018).

Evidences have shown that the increased risk of adiposity in PCOS compared with that in the general population (Hoeger and Oberfield, 2012). Changes in diet that increase the protein/carbohydrate ratio have been shown to have small metabolic and reproductive improvements in PCOS patients (Moran et al., 2003), while obese women with PCOS treated with a 1,000 kcal, low fat diet for 6–7 months showed body weight (BW) loss on the order of 5–10% and considerable reductions in the clinical manifestations of PCOS by restoring ovulation, increasing the pregnancy rate, and reducing levels of insulin and androgens (Kiddy et al., 1992). Another study found that modification of the gut microbiota is one means to treat obesity by regulating the metabolic system (DiBaise et al., 2008). Therefore, a new hypothesis was raised that the intestinal flora might be strongly associated with the occurrence and maintenance of PCOS symptoms. Compared with healthy women, the diversity of gut microbiota in patients with PCOS were reduced, and hyperandrogenism, total testosterone, hirsutism were inversely associated with alpha diversity (Torres et al., 2018). Exposure to hyperandrogenemia during fetal development might thus result in long-term changes in gut microbiota and subsequent changes in cardiac and metabolic function in the daughters of PCOS patients.

Prenatal androgen exposure can cause infant gut dysbiosis and altered abundance of bacteria that produce short-chain fatty acid metabolites, suggesting that androgen excess in immature fetuses could result in long-term alterations in gut microbiota and lead to increased risk of developing PCOS (Sherman et al., 2018).

Around 1013–10<sup>14</sup> microorganisms populate the adult intestines (Ley et al., 2005, 2006), and the enteric nervous system has been referred to as "the second brain" (Mayer, 2011). The gut microbiota play a key role in regulating energy balance and are involved in the development and progression of obesity and metabolic diseases (Parekh et al., 2014). The diversity of the gut microbiota is related to multiple features of metabolic syndrome, and the gut microbiota contribute to endotoxemia and are involved in inflammation, glucose tolerance, and insulin secretion (Cani et al., 2007a,b). Some of these associations suggest that the gut microbiome might even be more important for some of these changes than the patient's age, sex, and genetic background (Rial et al., 2016). Le Chatelier et al. (2013) have found that reduced diversity in the gut microbiota is related with adiposity, insulin resistance and dyslipidemia. Decreased intestinal flora diversity can also be seen in metabolic syndrome alone, which is associated with genetic variation of the apolipoprotein A5 gene. In addition, Lactobacillus has been found to be correlated with central adiposity, fasting blood sugar and negatively correlated with HDL-C levels (Lim et al., 2017).

In order to explore the role of the gut microbiota in PCOS, we used Illumina Miseq sequencing to study the composition and diversity of the gut microbiota in dihydrotestosterone (DHT) induced PCOS rats in comparison with high-fat diet (HFD) induced obese rats. At the same time, we also analyzed the correlation of the gut microbiota with circulating steroid levels and various metabolic parameters and evaluated the effects of three clinically relevant PCOS treatments – Diane-35, probiotics, and berberine.

#### MATERIALS AND METHODS

#### Animals and Treatments

Female Wistar rats of 21 days were randomly divided into the following six groups: Control, HFD, DHT, DHT + Diane-35, DHT + Probiotics, and DHT + Berberine (n = 6 for all groups). All rats were placed in 12 h light/12 h dark, 22 ± 2 ◦C constant temperature and 45–55% humidity, free to eat and drink. The Control, DHT, DHT + Diane-35, DHT + Probiotics, and DHT + Berberine groups were fed with standard chow, energy%: 10.3% from fat, 65.5% from carbohydrate and 24.2% from protein, 3.52kcal/g (Shanghai SLAC Laboratory Animals), while the HFD group was fed a high-fat chow, energy%: 60% from fat, 20% from carbohydrates and 20% from protein, 5.24 kcal/g (Research Diets, D12492).

This study was carried out in accordance with the local ethics committee of Shanghai Medical College, Fudan University, approved the experimental procedure and protocols (No. 20150119-019).

The control rats received empty cervical silicone tubes subcutaneously (length = 1 cm, diameter = 2 mm) at 21 days of age (**Figure 1A**). At the same time, silicone tubes with DHT (15 mg, slow releasing for 75 days) were implanted subcutaneously into the neck in the DHT, DHT + Diane-35, DHT + Probiotics, and DHT + Berberine groups. At 7 weeks after implantation, the DHT-induced PCOS rats received Diane-35 [one tablet containing 2.0 mg cyproterone acetate and 35 µg ethinylestradiol dissolved in 50 ml 1% carboxymethyl cellulose (CMC) solution and administered at 0.005 ml/kg BW], probiotics Bifid Triple Viable (trade name Pei Feikang, a combination of Bifidobacterium, Lactobacillus acidophilus, and Enterococcus faecali, Shanghai Xinyi Inc., China) was dissolved in 1% CMC solution to a concentration of 42 mg/ml and administered at 210 mg/kg BW, or berberine (berberine hydrochloride tablets were dissolved in 1% CMC solution to a final concentration of

FIGURE 1 | (A) Schematic of animals and treatments. Wistar rats were randomly divided into six group as shown. The control rats received blank cervical silicone tubes subcutaneously at 21 days of age. The HFD group was fed a diet in which 60% of the calories came from fat daily throughout the whole experiment. At the same time, silicone tubes with DHT (15 mg, slow releasing for 75 days) were implanted subcutaneously into the neck in the DHT, DHT + Diane-35, DHT + Probiotics, and DHT + Berberine groups. The DHT-induced PCOS rats (at 7 weeks after implantation) received Diane-35 (0.005 ml/kg BW), probiotics (Bifidobacterium triple viable, 210 mg/kg BW), or berberine (150 mg/kg BW) by intragastric administration. (B) Growth curves of rats from 21 days of age to 96 days of age. <sup>∗</sup>p < 0.05, ∗∗p < 0.01 versus the control group and #p < 0.05, ##p < 0.01 versus the DHT group using two-way ANOVA and Tukey's post hoc test. (C) Oral glucose tolerance test (OGTT). The basal blood glucose level was measured before being given the oral D-glucose (1.5 g/kg BW), and then measurements were made at 30, 60, 90, and 120 min. <sup>∗</sup>p < 0.05, ∗∗p < 0.01 versus controls.

30 mg/ml and administered at 150 mg/kg BW), and all three were given by daily intragastric administration (**Figure 1A**, bottom panel). BW, vaginal opening, and estrus cycle were monitored throughout the experiment. An oral glucose tolerance test was performed at 8 a.m the day before the rats were sacrificed. At the termination of the experiment, all rats were deeply anesthetized to sacrificed, serum, bilateral ovaries, adipose tissue, muscles, liver, pancreas, and rectum fecal matter were collected. The

control and HFD group were sacrificed at the diestrus stage of the estrous cycle.

#### Hormone Profile and Biochemical Indexes

Trunk circulation blood samples were obtained and were allowed to incubate for 4 h at room temperature. All samples were then centrifuged at 2,500 rpm for 15 min, collected into 1.5 ml Eppendorf tubes, and kept at −80◦C for subsequent experiments. Progesterone, estradiol, total testosterone, follicle stimulating hormone (FSH), luteinizing hormone (LH), sex hormone-binding globulin (SHBG), and c-reactive protein (CRP) levels were measured using an enzyme-linked immunosorbent assay (ELISA) kit (Sino-UK Institute of Biological Technology, Beijing, China) with a STAT FAX 2100 Microplate Reader (Awareness Technology Inc., United States). TG, TC, HDL-C, LDL-C, insulin, aspartate transaminase (AST), and alanine aminotransferase (ALT) were measured using colorimetric kits (BioSino Bio-Technology & Science Inc., China) with a BS-420 Chemistry Analyzer (MINDRAY, China).

#### Vaginal Smears

The stage of the estrus cycle was determined by the predominant cell type in vaginal smears that were obtained from 7 weeks of age until the end of the experiment every day (Marcondes et al., 2002).

#### Oral Glucose Tolerance Test

After an overnight fasted (10–12 h), glucose levels in tail vein blood were measured with a glucometer (ACCU-CHECK Performa, Roche). One researcher held the animal securely and cleaned the tail, and a second researcher prepared the glucometer and took the measurements. Basal blood glucose levels were measured before administration of 50% oral D-glucose. After the measurement, the tail was covered with gauze, and the room temperature was kept constant throughout the process (2 g/kg BW), and measurements were then taken at 30, 60, 90, and 120 min.

#### Microbial Diversity Analysis

#### Sample Collection

Fresh fecal samples were taken from the colons of all rats, collected into 1.5 ml sterile EP tubes, rapidly snap-frozen in liquid nitrogen, and stored at −80◦C until further analysis.

#### DNA Extraction and PCR Amplification

Microbial DNA was extracted from the fecal samples with the E.Z.N.A. <sup>R</sup> soil DNA Kit (Omega Bio-Tek, Norcross, GA, United States) according to the manufacturer's protocol. The final DNA concentration and purity were determined with a NanoDrop 2000 UV-vis spectrophotometer (Thermo Scientific, Wilmington, NC, United States), and DNA quality was checked by 1% agarose gel electrophoresis. The V3-V4 hypervariable regions of the bacterial 16S rRNA gene were amplified with primers 338F (5<sup>0</sup> -barcode-ACTCCTACGGGAGGCAGCAG-3<sup>0</sup> ) and 806R (5<sup>0</sup> -GGACTACHVGGGTWTCTAAT-3<sup>0</sup> ) (Huang et al., 2016; Zheng et al., 2018) using a Thermocycler PCR system (GeneAmp 9700, ABI, Waltham, MA, United States). Each primer contained 8–13 bp paired-end error-correcting barcodes (Fadrosh et al., 2014). The barcodes were synthesized by Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China). The PCR reactions were performed as follows: 3 min of denaturation at 95◦C, 28 cycles of 30 s at 95◦C, 30 s at 55◦C, and 45 s at 72◦C, and a final extension at 72◦C for 10 min. The PCR reactions were performed in triplicate as 20 µL mixtures containing 4 µL 5× FastPfu Buffer, 2 µL 2.5 mM dNTPs, 0.8 µL each primer (5 µM), 0.4 µL of FastPfu Polymerase, 0.2 µL BSA, and 10 ng of template DNA. The resulting PCR products were separated on a 2% agarose gel and purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, United States) and quantified using QuantiFluorTM-ST (Promega, United States) according to the manufacturer's protocol (Zheng et al., 2018). Sterile water was used as the negative control. The results of agarose gel electrophoresis of PCR products showed that the sterile water had no electrophoretic bands, indicating no contamination.

#### Library Preparation and Illumina MiSeq Sequencing

Library preparation involved four steps. (1) 'Y' adapters were linked. (2) Adapter dimers were removed by using beads. (3) The products were PCR amplified for library concentrations. (4) Single-stranded DNA fragments were generated using sodium hydroxide.

Sample libraries were pooled in equimolar amounts and paired-end (PE300, 2<sup>∗</sup> 300 bp) sequenced on an Illumina MiSeq platform (Illumina, San Diego, CA, United States) according to the standard protocols<sup>1</sup> in Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China) (McElhoe et al., 2014). The raw reads were deposited into the NCBI Sequence Read Archive database (the accession number: SRP170100).

#### Processing of Sequencing Data

Raw fastq files were demultiplexed, quality-filtered by Trimmomatic, and merged by FLASH, and the reads were truncated where the average quality score was < 20 over a 50 bp sliding window. Sequences of each sample were separated according to the barcodes (exactly matching) and primers (allowing 2 nucleotide mismatching). We removed reads containing ambiguous bases, and we merged sequences with overlaps longer than 10 bp based on their overlap sequence.

We used UPARSE (version 7.1)<sup>2</sup> with a 97% similarity cutoff to cluster operational taxonomic units (OTUs), and we used UCHIME to remove chimeric sequences. Finally, we used the RDP Classifier algorithm<sup>3</sup> with a confidence threshold of 70% to analyze the taxonomy of each 16S rRNA gene sequence by comparing it against the Silva (SSU128) 16S rRNA database.

#### Statistical Analysis

Statistics analysis was applied with SPSS 20.0 for Windows (SPSS Inc., Chicago, IL, United States).

<sup>1</sup>https://support.illumina.com/downloads/miseq\_system\_user\_guide\_15027617. html

<sup>2</sup>http://drive5.com/uparse/

<sup>3</sup>http://rdp.cme.msu.edu/

All data were presented as mean ± SEM. One-way ANOVA followed by LSD test was used to determine the significance. P < 0.05 was set as statistically significant.

#### RESULTS

#### Reproductive and Metabolic Disorders

Body weight was measured once a week. At 3 weeks of age, the BWs of all groups were not significantly different. After 2 weeks of DHT implantation, the BW of the DHT group was increased significantly compared to the controls (p < 0.05 or p < 0.01). The HFD group also had increased BW, but to a lesser degree than the DHT-induced rats. At 10 weeks of age, Diane-35, probiotics, and berberine were administered to the DHT-induced rats. Diane-35 and probiotics decreased the BW compared to the DHT group, but the effect of Diane-35 was the most obvious (**Figure 1B**).

The oral glucose tolerance test showed that circulating glucose levels were increased in the HFD group compared to the control group at 30, 60, 90, and 120 min after sugar taken and were higher in the DHT + Berberine group at 60, 90, and 120 min compared to the control group. Compared with the control group, glucose levels in the DHT + Diane-35 group were only higher at 90 min and were only higher at 60 min in the DHT + Probiotics group (**Figure 1C**).

Ovarian function was evaluated based on estrous cyclicity, ovarian histology, and follicle number and morphology. As shown in Figure, the majority of the rats in the DHT and DHT + Berberine group were in the diestrus stage compared with the control and HFD group. Proestrus and estrus stages were observed to a greater degree in the DHT + Diane-35 and DHT + Probiotics group compared to the DHT group, but to a lesser degree compared to the control group (**Figure 2A**).

Morphological observation of the ovarian follicles showed that the HFD group had no differences in follicle number or composition compared with controls, but there were fewer mature follicles and corpus luteum in the DHT group compared to controls. The DHT + Diane-35 and DHT + Probiotics groups showed some corpus luteum and follicles at the antral and preovulatory stage, and follicular granulosa cells were arranged in an orderly fashion (**Figure 2B**). However, estrous cycle stage, ovarian morphology, and mature follicle number in

the DHT + Berberine group showed no difference compared with the DHT group. The quantitation analysis of total follicles by Clarity 3D imaging showed that the number and the percentage of preovulatory follicles increased significantly in the DHT + Diane-35 and DHT + Probiotics groups compared with DHT, but the number of corpus luteum only increased in the DHT + Diane-35 group (**Figures 3A–C**).

The total testosterone concentrations were similar in DHT rats and controls, and progesterone decreased significantly in the DHT, DHT + Diane-35, DHT + Probiotics, and DHT + Berberine groups compared with the controls. The subcutaneous, visceral, and gonadal fat depot/BW ratios were higher in the DHT group compared to controls, and the subcutaneous fat depot/BW ratio was higher in the HFD group compared with the control group. The ovarian tissue/BW ratio was lower in the DHT, DHT + Berberine, DHT + Diane-35, and DHT + Probiotics groups compared with the control group. The uterine tissue/BW ratio was lower in the DHT group compared to controls (p < 0.05) (**Supplementary Table 1**).

#### Diversity of the Gut Microbiota

V3 + V4 16S rRNA sequences reads with high quality and classification were obtained from the 36 samples with an average of 36,133 sequences per sample (the minimum of one sample was 30,242 reads and the maximum was 44,561 reads) (**Supplementary Table 2**).

In addition, we assessed the alpha diversity (ACE, Chao1, and Shannon and Simpson indices) in the samples. The Shannon– Wiener curve (smooth tendency) showed sufficient sample numbers. When the curve tended to be flat, it indicated that the amount of sequencing data was enough to reflect the vast majority of microbial diversity of the sample (**Supplementary Figure 1**). At a 97% similarity level, an average of 515 OTUs per sample were identified. The ACE, together with OTU, Chao1 coverage, and Shannon and Simpson analysis, showed the community diversity of all groups. Compared with the control, the DHT + Berberine group had significantly reduced alpha diversity of gut microbiota that clustered differently from the other two groups. The second lowest diversity was seen in the HFD group. Other groups showed no significant differences (**Figure 4A** and **Supplementary Table 3**). On the family level, the top 20 differentiated taxa with high relative abundance after the different treatments are shown in the heat-map. The community of the gut microbiota in the DHT + Berberine group was different from the other groups (**Figure 4B**), and unweighted-UniFrac-based principal coordinates analysis (PCoA) showed a distinct clustering of the microbiota composition in each group (R <sup>2</sup> = 0.78, P = 0.001, Adonis) (**Figure 4C**).

# The Composition of Gut Microbiota in Different Groups

Firmicutes and Bacteroidetes were the two dominant genera in the fecal samples from all groups, and the relative abundance of the top eight phyla is shown in **Figure 5A**. The ratio of Firmicutes to Bacteroidetes was 2.06 in controls, 2.32 in the HFD group, 1.60 in the DHT group, 1.27 in the DHT + Berberine group, 1.44 in the DHT + Diane-35 group, and 1.90 in the DHT + Probiotics group. **Figure 5B** shows the composition of the gut microbiota in different groups at the family level. The abundance of Prevotellaceae was 8.73% in controls, 3.66% in the HFD group, 10.96% in the DHT group, 8.27% in the DHT + Berberine group, 7.22% in the DHT + Diane-35 group, and 10.05% in the DHT + Probiotics group. The relative abundance of Spirochaetais shown in **Figure 5C**. The abundance of Spirochaetaceaein controls (2.6%) and the DHT + Diane-35 group (1.8%) was significantly higher than that of the DHT (0.7%), DHT + Probiotics (0.8%), and DHT + Berberine (0.0007%) groups (**Figure 5D**).

We used Student's t-test to test for differences between groups and adjusted the p-values. Compared to control, the Ruminococcus\_1, Bacteroides, Prevotella\_9, Treponema\_2, and Ruminococcaceae\_UCG-005 were significantly increased in the HFD group (**Figure 6A** and **Supplementary Figure 2A**), while the genus Prevotella\_9, Bacteroides was significantly increased and the genus Prevotellaceae\_Ga6A1\_groupwas decreased in the DHT group (**Figure 6B** and **Supplementary Figure 2B**). The different genera were similar between the controls and the DHT + Diane-35 and DHT + Probiotics groups (**Figures 6C,D** and **Supplementary Figures 2C,D**). In addition, Lachnospiraceae\_NK4A136\_group, Unclassified\_f\_Ruminococcaceae, Ruminococcus\_1, Ruminiclostridium\_9, Lachnospiraceae\_UCG-005, Ruminococcaceae\_UCG-014, and \_UCG-005were significantly decreased in the DHT + Berberine group (**Figure 6E** and **Supplementary Figure 2E**).

#### Correlation Between Gut Microbiota and Clinical Variables

Correlation analyses were performed to determine the potential associations between microbial genera and metabolic disorders changes (**Figure 7**). Microbes that maintain the intestinal microecological balance such as Bacteroidales-S24-7 showed a negative relationship with CRP. Lachnospiraceae NK4A136 showed a negative relationship with AST and TG. Ruminococcus\_1 showed a negative relationship with ALT, but a positive relationship with progesterone. Candidate pathogens have opposite relationship with some clinical variables. For example, Prevotella\_9 was negative correlation with progesterone, and all of the Desulfovibrio displayed a positive correlation with TC and LH.

#### DISCUSSION

In addition to reproductive disorders, the DHT-induced PCOS model rats in this study showed many metabolic syndromes similar to women with PCOS, including increased BW, abnormal glycolipid metabolism, and intestinal flora disturbance. We found that the abundance and diversity of intestinal flora decreased significantly in the HFD group due to the high fat diet and in the DHT + Berberine group likely because of the strong antibacterial action of berberine. Smaller differences were seen

FIGURE 3 | (A) 3D rendering after staining by TH and DAPI, with the bottom row showing the spot identification at the different follicle stages. Different stages of follicles are represented as different colors of spheres, of which the red and yellow represent preovulatory follicles and corpora lutea, respectively. (B) The numbers of total follicles and follicles at different stages in the six groups. (C) The percentages of different follicle stages in the six groups. TH, tyrosine hydroxylase; CL, corpus luteum. <sup>∗</sup>p < 0.05, ∗∗p < 0.01 vs. control, ##p < 0.05 vs. DHT.

microbiota, which clustered differently from the other groups.

among other groups. However, at the phylum level of gut flora, there was a consistent composition in the HFD and DHT groups, such as lower diversification of Proteobacteria, Spirochaetae, and Verrucomicrobia, and more diversified Tenericutes. The Bacteroidetes were increased in the DHT group. At family level, the abundances of the bacterial taxa in the HFD group were obviously different compared to the control group. In contrast, the DHT group had more similar abundances of different taxa compared to the controls except for Bacteroidates\_S24-7\_group, Prevotellaceae, Enterobacteriaceae, and Desulfovibrionaceae. Of

the three treatments, Diane-35 and probiotics could restore the diversity and abundance of the gut flora, showing significant effects on Firmicutes, Tenericutes, Proteobacteria, and Spirochaetae. However, berberine significantly reduced the species diversity and number of intestinal flora and had no obvious effect on improving the symptoms of PCOS.

Long-term HFDs have been shown to result in the development of obesity-related metabolic disorders and concurrent diseases due to changes in the gut microbiota. HFDs decrease the diversity of the gut microbiota and alter bacteria composition, showing a decrease in Bacteroidetes and an increase in Firmicutes and Protebacteria or changes in the Firmicutes to Bacteroidetes ratio (Hildebrandt et al., 2009; Le Chatelier et al., 2013). A previous study showed that if the gut microbiota of an obesity-prone individual was transferred into bacteria-free mice, the latter gained BW and became obese, and in that work Phascolarctobacterium, Proteus mirabilis, and Veillonellaceae were positively correlated with all metabolic parameters, but only Lactobacillus intestinalis was negatively correlated with BW and fat mass (Lecomte et al., 2015). Regarding PCOS-related obesity, lower alpha diversity has been observed in women with PCOS and to be correlated negatively with hyperandrogenism, total testosterone, and the other primary reproductive and metabolic characteristics of PCOS (Torres et al., 2018). It was also reported

that letrozole-induced PCOS rats show lower Lactobacillus, Ruminococcaceae, and Clostridium and higher Prevotellaceae (Guo et al., 2016), which is consistent with the results of our DHT-induced PCOS rats. Therefore, although the causes of obesity are different, HFD-induced and DHT-induced changes in intestinal flora are very similar.

Studies have shown that many genes related to the etiology of PCOS are involved in steroid synthesis and carbohydrate metabolism, suggesting that metabolic factors are involved in the development of PCOS (Ben-Shlomo, 2003; Fernandez-Real and Pickup, 2012; Shi et al., 2012). Research has also shown an increase of serum levels of zonulin, which is a biomarker for gut permeability in PCOS patients. Insulin resistance and menstrual disorders also demonstrate relevance with serum levels of zonulin, suggesting that changes in gut permeability might be involved in the pathophysiology of PCOS (Zhang et al., 2015). Sun analyzed the metabolic profile differences between PCOS patients and healthy women and showed that plasma levels of dimethylamine were significantly increased in PCOS patients, demonstrating that gut microflora was more active in the patients with PCOS.(Sun et al., 2012). Tremellen and Pearce (2012) suggested that dysbiosis of gut microbiota (the DOGMA theory) which induced by a high-fat, high-sugar diet, therefore contributes to higher intestinal permeability in women with PCOS. Lipopolysaccharides produced by gram-negative bacteria could then transfer through gut wall into the blood and caused a long term, low level inflammation. Subsequent immune system disrupted insulin receptors and derived up insulin levels, finally boosts testosterone produced by the ovaries causing PCOS. Thus the DOGMA theory describes how the gut microbiota might be directly related to the pathogenesis of PCOS (Tremellen and Pearce, 2012). Changes in gut microbiota have also been seen in letrozole-induced mouse models of PCOS (Conway et al., 2014; Guo et al., 2016), and fecal microbiota transplantation (FMT) from the PCOS-like mice to normal mice showed that the gut microbiota is also closely related with the host's level of sex hormone, estrous cycles, and ovarian morphology (Guo et al., 2016). Steroids have also been shown to regulate the composition of the gut microbiome and metabolism and that "dysbiosis" in the gut microbiome might also occur in PCOS patients (Kelley et al., 2016).

We observed significant differences in Spirochaetaceae between the control group and the other groups except for the DHT + Diane-35 group, and a plausible mechanism might be that this species plays a role in insulin resistance. However, there is no research to prove that the Spirochaetaceae have a regulatory role in the intestinal flora, and this theory requires further experimental work. We also found that Prevotella-9 was significantly increased in the DHT group, and Prevotella species have been found to be prevalent in the respiratory system, oral cavity, and genital tract and to contribute to infections such as chronic sinusitis, periodontitis, and bacterial vaginosis (Brook, 2005; Oakley et al., 2008; Teles et al., 2012). Males appear to be more likely to carry Prevotella intermedia in the oral cavity compared to females, suggesting that there is an association between sex hormones and Prevotella (Umeda et al., 1998). Among women with PCOS, the level of Prevotella intermedia serum antibodies were higher than in healthy women (Akcali et al., 2014), and PCOS rats treated with FMT and Lactobacillus transplantation had decreased prevalence of Prevotella and decreased levels of testosterone and androstenedione (Kelley et al., 2016).

Diane-35 is a common medication to treat PCOS, which composed of 2 mg cyproterone acetate and 0.035 mg ethinyl estradiol each tablet. Cyproterone acetate is not only the progesterone agonist, but also androgen antagonist. In addition, ethinyl estradiol might suppress ovarian estradiol synthesis through regulation of the hypothalamic-pituitary-ovarian axis. In the present study, Diane-35 not only restored the reproductive function significantly, but also increased the diversity and abundance of the gut microflora compared with the DHT group, showing a significant effect on the genus Prevotella. The abundance of Eubacterium\_ventriosum\_group, Prevotellaceae\_Ga6A1\_group, Adlercreutzia, and Turicibacter were significantly greater in the DHT + Diane-35 group compared to the DHT group, but these were all similar to the control group. Recent research has shown that prenatal androgen exposure in rats causes gut microbiota dysbiosis (Sherman et al., 2018), mainly in terms of changes in bacterial species that are involved in the short-chain fatty acids. The gut bacteria have been suggested to be another source of male steroids (Ridlon et al., 2013), and many bacterial species are known to convert glucocorticoids into androgens. Taken together, all of the evidence above indicates that there is a close interaction between androgens and the gut microbiota, and that this combined with hereditary and environmental factors is a primary risk factor for developing PCOS.

Probiotics have a long history and are a widely used food supplement to improve intestinal health, but only the strains classified as lactic acid bacteria are significant nutritionally, and the genera Lactococcus and Bifidobacterium are the most important (Holzapfel et al., 2001). Bifid Triple Viable (commercial name: Pei Feikang), which is a combination of Bifidobacterium, Lactobacillus acidophilus, and Enterococcus faecali, has been used in patients with diarrhea and ulcerative colitis (Li et al., 2008, 2012; Zhang et al., 2012), and these probiotics increased the diversity and added Lactobacillus in the DHT + Probiotics group compared with the DHT group, and this leads to dramatic decrease in the serum levels of cholesterol and increase in the serum levels of progesterone. In letrozole-induced PCOS rats, Lactobacillustreated groups with increased numbers of Lactobacillus and Clostridium and decreased numbers of Prevotella had decreased androgen biosynthesis and normalized ovarian morphologies (Guo et al., 2016). These results indicated that probiotics are beneficial to the treatment of PCOS, and the possible mechanism is consistent with the DOGMA theory, including maintenance of the gut microbiota, improvement in intestinal permeability, and prevention of bacterial translocation from the gut.

Berberine is a quaternary ammonium salt derived from the protoberberine group of benzylisoquinoline alkaloids, extracted from the roots, rhizomes, stems, and bark of Berberis, Coptis chinensis, Phellodendri chinensis, etc. It has strong anti-bacterial, anti-lipid peroxidation, and anti-tumor activities, and it is under investigation to determine whether it might have applications for treating arrhythmia, diabetes, hyperlipidemia, and PCOS (Li Y. et al., 2013). Li recruited 120 patients with PCOS who were treated for 12 weeks with berberine or placebo, and berberine significantly improved insulin resistance in the PCOS patients (Li Y. et al., 2013). Another clinical trial also confirmed that 3 months of berberine treatment significantly decreased waist circumference and waistto-hip ratio and reduced metabolic parameters compared with metformin and placebo (Wei et al., 2012). However, in our PCOS-like rats 4 weeks of berberine administration clearly decreased the composition and diversity of the gut microbiota and gave no improvement of any metabolic or reproductive phenotypes.

Dietary changes are considered the first line of treatment for those with PCOS (Moran et al., 2011). However, little is known about how often clinicians recommend this treatment, and there is currently no standard dietary approach recommended as optimal for such women (Moran et al., 2009). An ideal dietary plan recommended for PCOS patients should be composed of average amounts of polyunsaturated fatty acids, omega-3, and sufficient intake of fiber-rich foods such as whole grains, legumes, vegetables, and fruits with hypoglycemic index (Faghfoori et al., 2017).

PCOS and its progress to metabolic syndromes, type 2 diabetes mellitus, have traditionally been considered with intake of excess caloric a decrease of physical activity, and certain genetic factors. However, increasing evidence suggests that the intestinal microbiota are closely associated with these diseases (Turnbaugh et al., 2009; Theriot et al., 2014; Paramsothy et al., 2017).

FMT has been shown in animal models to be a potential treatment for improving obesity and its associated comorbidities, but there is little evidence regarding the use of FMT in treating obesity and metabolic syndrome in humans (Marotz and Zarrinpar, 2016). There are only two registered clinical trials using FMT as a treatment for obesity (clinicaltrials.gov identifiers: NCT02530385 and NCT02741518), and one for treating metabolic syndrome by FMT with donor fecal matter of lean people (Vrieze et al., 2012). The data from these trials and animal models suggest that FMT is a promising treatment for metabolic syndrome.

The associations between the gut microbiota and clinical variables presented here suggest that certain groups of microbes likely play important roles in the development of obesity-related metabolic disorders, and our results show that these bacteria might be of help for the prognosis, prevention, and treatment of PCOS. The findings of this study enable us to develop better understanding the influence of the gut microbiota in the metabolic and reproductive alterations in PCOS, and they also suggest opportunities for future personal dietary guidance, new biomarkers, and new treatments with specific groups of bacteria, and they support replacement with pathogen-free gut microbiota as a potential treatment for PCOS.

# AUTHOR CONTRIBUTIONS

FZ, HG, YF, and CX conceived the experiments, designed the project and protocols, and developed the collaborations. FZ, TM, PC, SH, CH, GY, WH, XX, AT, and YF performed the

experiments and analyzed the data. FZ, AT, CX, and YF wrote the manuscript. FZ, AT, LS, XL, HG, and CX provided scientific oversight and guidance and edited the manuscript. FZ, PC, TM, HG, AT, YF, and CX are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

#### FUNDING

This research was funded by the National Natural Science Foundation of China (NSFC 81673766) and Development Project

#### REFERENCES


of Shanghai Peak Disciplines-Integrative Medicine (20180101) to YF, the Shanghai Municipal Commission of Health and Family Planning (201640362) to FZ, the National Natural Science Foundation of China (NSFC 81572555) to XL, the Shanghai Municipal Commission of Health and Family Planning (2017ZZ01016) to CX.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.00175/full#supplementary-material



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Zhang, Ma, Cui, Tamadon, He, Huo, Yierfulati, Xu, Hu, Li, Shao, Guo, Feng and Xu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Matrine Mediates Inflammatory Response via Gut Microbiota in TNBS-Induced Murine Colitis

Peiyuan Li<sup>1</sup>† , Jiajun Lei<sup>2</sup>† , Guangsheng Hu<sup>1</sup> , Xuanmin Chen<sup>1</sup> , Zhifeng Liu<sup>3</sup> \* and Jing Yang<sup>1</sup> \*

<sup>1</sup> Department of Gastroenterology, The First Affiliated Hospital of University of South China, Hengyang, China, <sup>2</sup> Xiangya School of Medicine, Central South University, Changsha, China, <sup>3</sup> Department of Otorhinolaryngology, The First Affiliated Hospital of University of South China, Hengyang, China

#### Edited by:

Jie Yin, Institute of Subtropical Agriculture (CAS), China

#### Reviewed by:

Md. Abul Kalam Azad, Institute of Subtropical Agriculture (CAS), China Hui Han, Chinese Academy of Sciences, China Jin Zhong, Institute of Microbiology (CAS), China

#### \*Correspondence:

Jing Yang yangjing@usc.edu.cn Zhifeng Liu liuzf@usc.edu.cn †These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology

Received: 20 December 2018 Accepted: 11 January 2019 Published: 08 February 2019

#### Citation:

Li P, Lei J, Hu G, Chen X, Liu Z and Yang J (2019) Matrine Mediates Inflammatory Response via Gut Microbiota in TNBS-Induced Murine Colitis. Front. Physiol. 10:28. doi: 10.3389/fphys.2019.00028 This study mainly investigated the effect of matrine on TNBS-induced intestinal inflammation in mice. TNBS treatment caused colonic injury and gut inflammation. Matrine (1, 5, and 10 mg/kg) treatment alleviated colonic injury and gut inflammation via reducing bleeding and diarrhea and downregulating cytokines expression (IL-1β and TNF-α). Meanwhile, serum immunoglobulin G (IgG) was markedly reduced in TNBS treated mice, while 5 and 10 mg/kg matrine alleviated IgG reduction. Fecal microbiota was tested using 16S sequencing and the results showed that TNBS caused gut microbiota dysbiosis, while matrine treatment markedly improved gut microbiota communities (i.e., Bacilli and Mollicutes). Functional analysis showed that cell motility, nucleotide metabolism, and replication and repair were markedly altered in the TNBS group, while matrine treatment significantly affected cell growth and death, membrane transport, nucleotide metabolism, and replication and repair. In conclusion, matrine may serve as a protective mechanism in TNBS-induced colonic inflammation and the beneficial effect may be associated with gut microbiota.

Keywords: matrine, inflammation, gut microbiota, colitis, mouse

#### INTRODUCTION

Inflammatory bowel diseases (IBD), an intestinal chronic inflammatory response or ulceration, is characterized by various pathologic symptoms, including bloody diarrhea, intestinal motility dysfunction, and intestinal shortening (Lee et al., 2014; Hirai and Matsui, 2015). The prevalence and incidence of IBD in China has markedly increased in recent years (Zhu et al., 2013). In the United States, about 1.0–1.5 million patients were estimated to suffer from IBD occurring between 2003 and 2004 (Kappelman et al., 2008). Although, the pathological mechanism of IBD is still unclear, compelling evidence suggests that inflammation and gut microbiota dysbiosis may serve as the major contributor in IBD (Ferguson et al., 2016). Thus, improving inflammatory status and gut microbiota communities may serve as a potential therapy for IBD patients.

Matrine, a kind of alkaloid substance, isolates from the roots of Sophora species in China. Compelling pieces of evidence have indicated that matrine exhibits various pharmacological activities, such as anti-inflammation, anti-oxidative stress, anti-infection, and cardiovascular protective effects (Liu et al., 2014; Cordero-Herrera et al., 2015; Yan et al., 2016). However, the merit of matrine on 2,4,6-trinitrobenzene sulfonic acid (TNBS)-induced murine colitis has not been fully studied. In this study, effects of matrine of intestinal inflammatory and gut microbiota in TNBS-induced murine colitis were mainly investigated.

# MATERIALS AND METHODS

#### Animal Model and Groups

fphys-10-00028 February 7, 2019 Time: 15:19 # 2

This study was carried out in accordance with the recommendations of the Declaration of Helsinki. The protocol involving animal subjects was approved by the Animal Welfare Committee of the University of South China. Fifty female Balb/c mice (20.41 ± 1.68 g) were randomly divided into five groups with ten mice for each: normal control group (N group, n = 10), the TNBS group (TNBS group, n = 10), 1 mg/kg matrine plus TNBS (ML group), 5 mg/kg matrine plus TNBS (MM group), and 10 mg/kg matrine plus TNBS (MH group). Chronic colitis in mice was induced by weekly administration of increasing dosages of TNBS eight times (1.0–2.3 mg in 45% ethanol) according to previous report (Weiss et al., 2015; Levit et al., 2018). After 8 weeks, all mice were sacrificed for sample collection. Colonic length and weight were recorded.

# Clinical Evaluation of TNBS Colitis

Rectal bleeding and diarrhea of all mice in this study were recorded. Stool bloody level was determined by haemoccult kits (Beckman Coulter). Bloody stool was evaluated by the following scoring system: 0 means no blood in the stool; 2 means positive haemoccult in the stool; and 4 means gross bleeding in the stool. Diarrhea was evaluated by the following scoring system: 0 means well-formed pellets; 2 means pasty and semiformed stools; and 4 means liquid stools (Vlantis et al., 2015).

#### Serum Immunoglobulins (Igs)

Blood samples were harvested by eye blooding and serum was separated by centrifugation (3,000 × g, 10 min, 4◦C). Serum samples were stored at −80◦C before Igs (IgA, IgG, and IgM) analysis by spectrophotometric kits (Nanjing Jiangcheng Biotechnology Institute, China).

# Real-Time PCR

Gut pro-inflammatory cytokines were determined to evaluate inflammation by real-time PCR. One piece of jejunum, ileum, and colon were harvested and stored at −80◦C. Total RNA of these tissues was isolated using TRIZOL regent and reverse transcribed into the first strand (cDNA) with DNase I, oligo (dT)<sup>20</sup> and Superscript II reverse transcriptase (Invitrogen, United States). The reverse transcription reaction was carried at 37◦C for 15 min, 85◦C 5 s. Primers in this study were designed with Primer 5.0 (**Table 1**). β-actin was selected as the housekeeping gene to normalize the expression of target genes. The PCR cycling used followed these conditions: 40 cycles at 94◦C for 40 s, 60◦C for 30 s, and 72◦C for 35 s. The relative expression of target genes was normalized as a ratio to the expression of β-actin in the control group using the formula 2−(11Ct) , where 11Ct = (CtTarget−Ctβ–actin)Treatment−(CtTarget−Ctβ–actin)control.

#### Microbiota Sequencing

Total genome DNA from fecal samples was extracted for amplification using a specific primer (16S rRNA genes of distinct regions [Primer 16S V4, 515F:5<sup>0</sup> -GTGCC AGCMGCCGCGGTAA-3<sup>0</sup> and 806R:5<sup>0</sup> -GGACTACHVGGG TWTCTAAT-3 <sup>0</sup> )] (Burbach et al., 2017). Sequencing libraries were generated and analyzed according to our previous study. Observed-species, Chao1, Shannon, and Simpson are used to evaluate the complexity of species diversity. Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) was further used for genome prediction of microbial communities in this study (Douglas et al., 2018; Wilkinson et al., 2018).

#### Statistical Analysis

All data in this study were analyzed using IBM SPSS 21.0 software. Comparisons between groups were analyzed by Tukey's multiple comparison test after testing the homogeneity of variances via Levene's test. Values in the same row with different superscripts (a, b, c) are significant (P < 0.05) (Liu et al., 2018a,b).

# RESULTS

#### Effects of Matrine on TNBS-Induced Colonic Injury

In this study, final body weight, colonic weight and length, rectal bleeding score, and diarrhea score were studied to evaluate clinical status TNBS-induced murine colitis. As shown in **Table 2**, TNBS markedly reduced body weight (27.72 ± 2.12 g) compared with the N group (33.47 ± 2.38 g) (P < 0.05). 5 and 10 mg/kg matrine significantly alleviated TNBS-induced growth suppression (P < 0.05). TNBS caused a marked colonic injury evidenced by the reduced colonic length and elevated colonic weight (P < 0.05). Although matrine failed to influence colonic length (P > 0.05), colonic weight was





The values having different superscript letters were significantly different (P < 0.05; n = 10). FBW: final body weight (g); CL: colonic length (cm); CW: colonic weight (mg); RBS: rectal bleeding score; DS: diarrhea score.



Data are presented as mean ± SEM. The values having different superscript letters were significantly different (P < 0.05; n = 10).

significant lower in the MM and MH groups than that in TNBS group (P < 0.05).

Rectal bleeding and diarrhea score are two major clinical indexes and we found that TNBS treatment markedly increased rectal bleeding score and diarrhea score (P < 0.05), while 10 mg/kg matrine (MH) alleviated colonic bleeding and diarrhea (P < 0.05).

# Effects of Matrine on Serum Igs

As shown in **Table 3**, TNBS markedly reduced serum IgG level compared with the N group (P < 0.05), while 5 and 10 mg/kg matrine increased serum IgG level compared with the TNBS group (P < 0.05). In addition, dietary supplementation tended to enhance IgM production.

#### Effects of Matrine on Intestinal and Colonic Expression of Proinflammatory Cytokines

mRNA abundances of interleukin-1β (IL-1β), interleukin-10 (IL-10), interleukin -17 (IL-17), and tumor necrosis factorα (TNF-α) were determined in the jejunum, ileum, and colon to evaluate gut inflammatory response (**Table 4**). In the jejunum, TNBS treatment upregulated IL-1β expression (P < 0.05), 5 and 10 mg/kg matrine alleviated TNBS-induced IL-1β over-expression (P < 0.05). Meanwhile, compared with the TNBS group, 10 mg/kg matrine markedly inhibited TNF-α expression (P < 0.05).

In the ileum, TNBS treatment markedly increased IL-1β and TNF-α mRNA abundances (P < 0.05), although matrine failed to mediate TNF-α expression in TNBS-induced murine colitis. Matrine (1, 5, and 10 mg/kg) significantly alleviated the overexpression of IL-1β (P < 0.05). In the colon, IL-1β, IL-10, and TNF-α were significantly upregulated in TNBS group compared with the N group (P < 0.05), and matrine (5 and 10 mg/kg) reduced IL-1β and TNF-α mRNA abundances (P < 0.05).

# Effects of Matrine on Intestinal and Colonic Expression of TLR4/Myd88

TNBS treatment markedly upregulated ileal Myd88 expression compared with the N group (P < 0.05) and matrine (5 and 10 mg/kg) inhibited ileal Myd88 expression (P < 0.05) (**Table 5**). Meanwhile, the mRNA abundances of TLR4 and Myd88 were significantly higher in the TNBS group than that in the N group in the colon (P < 0.05), while 10 mg/kg matrine alleviated colonic TLR4 activation (P < 0.05).

#### Effects of Matrine on Gut Microbiota in TNBS-Induced Murine Colitis

16S rRNA sequencing yielded an average of 53,364 filtered partial sequences per sample with an average length of ∼300 bp. Alpha-diversity was tested by analyzing observed species, Chao1, Shannon, and Simpson (**Table 6**). Observed species, Chao1, and Simpson indexes were not altered in the TNBS and matrine groups (P > 0.05). Shannon value in the TNBS-treated mice was markedly lower than that in the normal group (P < 0.05), while matrine tended to increase the Shannon index (P > 0.05).

The overall microbial compositions in the TNBS and matrine groups were markedly changed at the class and family levels (**Table 7**). At class level, TNBS treatment markedly reduced the relative abundances of Bacilli and Mollicutes (P < 0.05), while matrine significantly restored the reduction of Bacilli and Mollicutes levels (P < 0.05). Meanwhile, matrine enhanced Betaproteobacteria and Bacteroidia levels compared with the N and TNBS groups (P < 0.05). At family level, Peptostreptococcaceae, Erysipelotrichaceae, Methylobacteriaceae, Sphingomonadaceae, and Lachnospiraceae were markedly reduced in response to TNBS-induced murine colitis, matrine treatment improved the relative abundances of Peptostreptococcaceae, Methylobacteriaceae, Sphingomonadaceae, and Lachnospiraceae (P < 0.05). Also, Bifidobacteriaceae was increased and Mycoplasmataceae was reduced in matrine-fed mice compared with the TNBS group (P < 0.05).

TABLE 4 | Effects of matrine on intestinal and colonic expression of proinflammatory cytokines.


Data are presented as mean ± SEM. The values having different superscript letters were significantly different (P < 0.05; n = 10).

TABLE 5 | Effects of matrine on intestinal and colonic expression of TLR4/Myd88.


Data are presented as mean ± SEM. The values having different superscript letters were significantly different (P < 0.05; n = 10).

FIGURE 1 | Genome prediction of microbial communities by PICRUSt analysis. Data are expressed as relative abundance of genes. The values having different superscript letters were significantly different (P < 0.05; n = 6).


Data are presented as mean ± SEM. The values having different superscript letters were significantly different (P < 0.05; n = 6).

TABLE 7 | List of significantly changed gut microbiota in response to TNBS and matrine treatments.


Data are presented as mean ± SEM. The values having different superscript letters were significantly different (P < 0.05; n = 6).

PICRUSt was further used for genome prediction of microbial communities and the results showed that cell motility, nucleotide metabolism, and replication and repair were markedly altered in the TNBS group, while matrine treatment significantly affected cell growth and death, membrane transport, nucleotide metabolism, and replication and repair (**Figure 1**).

#### DISCUSSION

Previous reports indicated that mice receiving TNBS administration showed significantly increased clinical scores of rectal bleeding score and diarrhea score and body weight loss (Weiss et al., 2015; Zhang et al., 2015). In this study, we found that TNBS influenced final body weight, colonic weight and length, rectal bleeding score, and diarrhea score, suggesting a colonic colitis model. In addition, matrine exhibited a positive role in TNBS-induced colonic injury.

Igs are glycoproteins and one of the vital components of the immune system and previous reports suggested a beneficial role of Igs in inflammatory response (Brimelow et al., 2015; Elluru et al., 2015; Li et al., 2018). In this study, we found that IgG involves in colonic colitis as TNBS significantly inhibited serum IgG production. Meanwhile, matrine enhanced the serum IgG level, suggesting a protective role in TNBS-induced colonic injury. Matrine was demonstrated to be an immune enhancer via inducing T cell anergy in human Jurkat cells (Li et al., 2010). In this study, we firstly reported that matrine regulates serum IgG in TNBS-induced colonic injury.

IBD, including Crohn's disease (CD) and ulcerative colitis (UC), are characterized by intestinal inflammatory response (Sands, 2015). In this study, TNBS caused intestinal inflammation and matrine exhibited an antiinflammatory effect via mediating intestinal cytokines expression. In asthmatic mice, matrine attenuates allergic airway inflammation and eosinophil infiltration by suppressing eotaxin and Th2 cytokine production (Huang et al., 2014). In addition, matrine inhibits ovalbumin-induced airway hyperresponsiveness, inflammatory cell infiltration, and goblet cell differentiation via regulating Il-4, IL-13, and TNF-α expression (Sun D. et al., 2016).

NF-κB plays critical roles in development, survival, oxidative stress, inflammation and activation of B lymphocytes (Herder et al., 2015; D'Addio and Fiorina, 2016; Sasaki and Iwai, 2016; Jin et al., 2018). NF-κB was identified as one of the key regulators in the immunological setting. Its activation is markedly induced in IBD patients and promotes the expression of various pro-inflammatory genes (Atreya et al., 2008; Ruhl and Landrier, 2016). Thus, inhibition or inactivation of NF-κB serves as a potential therapy for IBD patients. In this study, we found that dietary matrine inhibited TLR4/Myd88 expression, the upstream signal of NF-κB. TLR4 is widely expressed in the intestine. Once activated by its ligands, TLR4 can activate NF-κB signaling pathway linked to the transcription of many proinflammatory genes (Tang et al., 2015). Compelling evidence has demonstrated that matrine regulates TLR4 expression (Liu et al., 2015; Sun N. et al., 2016). Furthermore, matrine can target NF-κB signal to regulate gene expression. For example, Lu et al. (2015) reported that matrine inhibits IL-1β-induced expression of matrix metalloproteinases by suppressing the activation of NF-κB in human chondrocytes in vitro. Similarly, matrine has been demonstrated to inactivate NF-κB signal in various cancer cells (Kim et al., 2013; Li et al., 2014).

Various previous studies have confirmed the role of gut microbiota in the pathophysiology of IBD (Peloquin and Nguyen, 2013; Gkouskou et al., 2014; Arora et al., 2018; Palamidi and Mountzouris, 2018; Roelofs et al., 2018; Zhang et al., 2018). The potential mechanism may be associated with the gut microbiota and host metabolism interaction as gut bacteria often target host metabolism, which further drive immune activation and chronic inflammation (Weingarden and Vaughn, 2017). Similar to previous studies, the current results showed that TNBS treatment caused microbiota dysbiosis by reducing alpha-diversity and the relative abundances of Bacilli and Mollicutes. Functional analysis showed that cell motility, nucleotide metabolism, and replication and repair were markedly altered in the TNBS group, while matrine treatment significantly affected cell growth and death, membrane transport, nucleotide metabolism, and replication and repair. The microbiota plays a fundamental role on the induction, training, and function of the host immune system and inflammatory response (Belkaid and Hand, 2014). Meanwhile, NF-κB activity has been reported to be affected by gut microbiota (Topol and Kamyshny, 2013). Thus, the gut microbiota might serve as a potential mechanism of the protective role of matrine in IBD models.

In conclusion, TNBS treatment induced colonic injury and inflammatory response in mice. Dietary matrine exhibited a protective role via enhancing serum IgG abundance and alleviating intestinal cytokines expression. The mechanism might be associated with gut microbiota as matrine improved gut microbiota communities in TNBS-induced murine colitis.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

### FUNDING

This work was supported by the grants from the Hunan provincial Health and Family Planning Commission (B20180186) and the Hunan Provincial Natural Science Foundation of China (2018JJ2355).


fphys-10-00028 February 7, 2019 Time: 15:19 # 6

Peloquin, J. M., and Nguyen, D. D. (2013). The microbiota and inflammatory bowel disease: insights from animal models. Anaerobe 24, 102–106. doi: 10.1016/j. anaerobe.2013.04.006

fphys-10-00028 February 7, 2019 Time: 15:19 # 7


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Li, Lei, Hu, Chen, Liu and Yang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Intestinal Bacteria Interplay With Bile and Cholesterol Metabolism: Implications on Host Physiology

Natalia Molinero, Lorena Ruiz\*, Borja Sánchez, Abelardo Margolles and Susana Delgado

Department of Microbiology and Biochemistry of Dairy Products, Instituto de Productos Lácteos de Asturias – Consejo Superior de Investigaciones Científicas (IPLA-CSIC), Villaviciosa, Spain

Bile is a biological fluid synthesized in the liver, mainly constituted by bile acids and cholesterol, which functions as a biological detergent that emulsifies and solubilizes lipids, thereby playing an essential role in fat digestion. Besides, bile acids are important signaling molecules that regulate key functions at intestinal and systemic levels in the human body, affecting glucose and lipid metabolism, and immune homeostasis. Apart from this, due to their amphipathic nature, bile acids are toxic for bacterial cells and, thus, exert a strong selective pressure on the microbial populations inhabiting the human gut, decisively shaping the microbial profiles of our gut microbiota, which has been recognized as a metabolic organ playing a pivotal role in host health. Remarkably, bacteria in our gut also display a range of enzymatic activities capable of acting on bile acids and, to a lesser extent, cholesterol. These activities can have a direct impact on host physiology as they influence the composition of the intestinal and circulating bile acid pool in the host, affecting bile homeostasis. Given that bile acids are important signaling molecules in the human body, changes in the microbiotaresiding bile biotransformation ability can significantly impact host physiology and health status. Elucidating ways to fine-tune microbiota-bile acids-host interplay are promising strategies to act on bile and cholesterol-related disorders. This manuscript summarizes the current knowledge on bile and cholesterol metabolism by intestinal bacteria, as well as its influence on host physiology, identifying knowledge gaps and opportunities to guide further advances in the field.

Keywords: gut microbiota, bile acids, cholesterol, gut microbiota-host interplay, bile signaling

# INTRODUCTION

The human gastrointestinal tract (GIT) is colonized by a vast array of microbes which dynamically interact with dietary and host-derived molecules in the intestinal lumen, significantly contributing to host physiology. Indeed, several animal and human studies have demonstrated that specific gut microbiota configurations contribute to inflammatory and metabolic diseases (Wu et al., 2015), although the precise molecular mechanisms behind the microbiota-host interactions impacting host health remain largely unknown. Cholesterol and bile acids (BAs) are important signaling molecules that, apart from exerting digestive functions, regulate multiple physiological processes

Edited by:

Yuheng Luo, Sichuan Agricultural University, China

#### Reviewed by:

Simona Bertoni, University of Parma, Italy Manlio Vinciguerra, International Clinical Research Center (FNUSA-ICRC), Czechia

> \*Correspondence: Lorena Ruiz lorena.ruiz@ipla.csic.es

#### Specialty section:

This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology

Received: 28 November 2018 Accepted: 14 February 2019 Published: 14 March 2019

#### Citation:

Molinero N, Ruiz L, Sánchez B, Margolles A and Delgado S (2019) Intestinal Bacteria Interplay With Bile and Cholesterol Metabolism: Implications on Host Physiology. Front. Physiol. 10:185. doi: 10.3389/fphys.2019.00185

in the host (Hegyi et al., 2018). Besides, the interaction of cholesterol and BAs with gut bacteria has been known for decades, although the role of these interactions in host health, and the possibility to modulate them through targeting the gut microbiota composition to improve human health, have only started to be recently explored.

Bile acids are synthesized in hepatocytes from cholesterol and conjugated to glycine and taurine before being secreted into the small intestine with the bile flow, which plays a major role in fat emulsification and absorption. Bile composition depends on the diet and intrinsic characteristics of the individuals, but usually contains over 50% BAs, over 20% fatty acids and cholesterol, and lower amounts of other molecules such as bilirubin or phospholipids (Farina et al., 2009). During its gastrointestinal transit, most BAs and cholesterol are reabsorbed in the distal small intestine, though a significant proportion evades this process, being excreted with feces (Islam et al., 2011).

Bile acids and cholesterol reaching the large intestine dynamically interact with our gut microbes. Indeed, BAs strongly compromise bacterial survival in the GIT, thus gut microbes must have developed mechanisms to counteract bile toxicity (Ruiz et al., 2013). Besides, gut microbial communities are capable of chemically modifying cholesterol and BAs, transformations that impact the gut microbiota and the BAs pool and, consequently, the signaling mechanisms they mediate. Accordingly, changes in this gut microbiota-bile axis are now acknowledged to have decisive implications in human health (Long et al., 2017).

The present minireview examines the current knowledge on the enzymatic activities of intestinal bacteria over BAs and cholesterol, and their implications in human physiology, with a particular emphasis on their impact on gastrointestinal disorders and aging-associated decline. Opportunities and limitations to translate this body of knowledge into novel microbiome-based applications for some of these diseases are also discussed.

#### CHOLESTEROL METABOLISM BY INTESTINAL BACTERIA

Cholesterol is a terpenoid lipid with a carbon skeleton formed by four fused alicyclic rings. It is an essential component of the mammalian cell membranes and precursor of steroid hormones, vitamin D, and primary BAs (García et al., 2012). Following its GIT passage, most cholesterol is absorbed in the duodenum and proximal jejunum by a passive diffusion process. Reabsorbed cholesterol is incorporated with triglycerides and lipoproteins into transportable complexes called chylomicrons, which return to the liver through the enterohepatic circulation. The cholesterol escaping this re-absorption reaches the colon, where it can be metabolized by the intestinal microbiota and/or excreted with feces (Gérard, 2013).

The metabolism of cholesterol by gut microbes has been described since the 30s (Schoenheimer, 1931) and has been supported by studies on germ-free animal models (Gérard et al., 2007). The microbial activities on cholesterol are based on its enzymatic reduction to produce coprostanone and coprostanol (**Figure 1**), which is poorly absorbable in the intestine. Thus, coprostanol production leads to increased cholesterol excretion into feces, contributing to reduce blood cholesterol level (Lye et al., 2010). Two different pathways have been proposed for this microbial reduction of cholesterol. The first pathway involves the direct reduction of the double bond 5–6 to give coprostanol, by cholesterol reductases (Gérard et al., 2004). The second pathway involves the oxidation of the 3β-hydroxy group and the isomerization of the double bond to produce 4-cholesten-3-one by cholesterol oxidases (ChOx) or 3β-hydroxysteroid dehydrogenases/isomerases (HSD) (García et al., 2012), followed by two reductions to form coprostanone and finally coprostanol (Gérard, 2013). Very limited information is available on the occurrence and distribution of the latter enzymes, although sequences belonging to ChOx are frequently found in the genomes of intestinal bacteria and gut/fecal metagenomes, indicating that cholesterol oxidation is a common activity in the gut microbiota. Remarkably, ChOx-encoding genes are found in the phyla Bacteroidetes, Proteobacteria and Actinobacteria, displaying a lower degree of conservation in Actinobacteria, but are absent in Firmicutes, one of the dominant phyla in the human gut microbiota (**Figure 2** and **Supplementary Figure 1**).

Several factors throughout life, including changes in diet or antibiotics consumption (Korpela and Adlercreutz, 1985; Norin, 1997), have been suggested to affect the gut microbiota's ability to reduce cholesterol to coprostanol, which exhibits higher rates of conversion in elderly individuals (Benno et al., 2009). Indeed, these factors are known to affect the gut microbiota composition in humans, although the real impact of lifestyle and other clinical factors in the microbial reduction of cholesterol, and the particular gut bacteria/activities implicated warrant further investigation.

Several cholesterol-reducing strains have been isolated from the intestine and feces of mammals (Eyssen et al., 1973; Brinkley et al., 1980, 1982). The first described cholesterol-reducing isolate of human origin was the Bacteroides sp. strain D8 (Gérard et al., 2007). Otherwise, only a few cholesterol-reducing intestinal bacteria have been identified, most of them belonging to the genus Eubacterium, although the genes or enzymes involved in this metabolism have not been well characterized yet.

Some other gut bacterial inhabitants, including lactobacilli and bifidobacteria species usually used as probiotics, have been long studied for their possible cholesterol-lowering activities. Although different mechanisms of action (involving removal, coprecipitation or assimilation) have been proposed (Pereira and Gibson, 2002; Liong and Shah, 2005; Tomaro-Duchesneau et al., 2014; Zanotti et al., 2015), to date, the real contribution of these microbial groups toward cholesterol-lowering and the molecular activities involved remain mostly unknown.

#### BACTERIAL BILE METABOLISM: IMPLICATIONS ON HEALTH AND DISEASE

The metabolism of BAs by the gut microbiota has been known for decades, although its consequences on human health have only started to be considered (Farina et al., 2009;

(Continued)

#### Molinero et al. Bile/Cholesterol Metabolism by Gut Microbes

#### FIGURE 1 | Continued

fphys-10-00185 March 13, 2019 Time: 16:25 # 4

the enzymes bile acid CoA synthetase and bile acid-CoA: amino acid N-acyltransferase. These conjugated BAs are excreted into bile by a BA export pump (BSEP) and stored in the gallbladder. (B) Bile composition. Conjugated primary BAs (glycocholic, taurocholic, glycochenodeoxycholic and taurochenodeoxycholic acids) are the main components of bile. Cholesterol, fatty acids, bilirubin and phospholipids are present in lower amounts. (C) Metabolism of BAs and cholesterol by intestinal bacteria. (4) The first reaction in the metabolism of BAs is the deconjugation or hydrolysis of conjugated BAs, catalyzed by bile salt hydrolases (BSHs). (5) Then, a bile salt 7α-dehydroxylase carries out the conversion of primary BAs to secondary BAs, deoxycholic and lithocholic acids. A part of the cholesterol is absorbed in the duodenum and proximal jejunum, returning to the liver. Remaining cholesterol reaches the large intestine, where it can be further metabolized by the intestinal microbiota or excreted with the feces. (6) Regarding cholesterol metabolism, the main gut microbial activity reaction involves the direct reduction of cholesterol to produce coprostanol, a reaction carried out by cholesterol reductases. (7) The indirect pathway begins with the oxidation of the 3β-hydroxy group by cholesterol oxidases (ChOx) or 3β-hydroxysteroid dehydrogenases/isomerases (HSD) to form 4-cholesten-3-one, and then cholesterol dehydrogenases produce coprostanone. Finally, cholesterol reductases form coprostanol. (D) BAs and sterols in feces. The main BAs in feces are secondary BAs, deoxycholic acid and lithocholic acid, with a lower concentration of primary BAs. Feces do also contain products of cholesterol metabolism such as coprostanol and coprostanone, that represent more than 50% of the total fecal sterols.

Islam et al., 2011; Gérard, 2013; Jia et al., 2017; Long et al., 2017), opening a new area of research in the microbiome-host interactions field. Key findings on this microbiota-BA signaling and host health are presented below.

#### Metabolism of BAs by Intestinal Bacteria

The composition of the BAs pool in humans is determined by the enterohepatic cycle and the microbial metabolism of intestinal BAs. Briefly, the liver synthesizes two primary BAs from cholesterol, cholic acid and chenodeoxycholic acid, which are conjugated to either taurine or glycine before being poured into the bile flow. Conjugated BAs are the primary components of bile, which is stored in the gallbladder before being excreted into the small intestine during digestion. Over 95% of the BAs secreted in bile are reabsorbed in the terminal ileum, returning to the liver through the enterohepatic circulation, and only 5% reach the large intestine, being excreted in feces. In the large intestine, BAs can suffer several microbial-mediated transformations including deconjugation, carried out by bile salt hydrolases (BSHs) that hydrolyze the amide bond, and transformation of primary deconjugated BAs into secondary BAs mainly by a 7α-dehydroxylation (**Figure 1**). Whereas deconjugation reactions are carried out by a broad spectrum of colonic bacteria (**Figure 2** and **Supplementary Figure 1**), 7α-dehydroxylation appears to be restricted to a limited number of intestinal bacteria (Ridlon et al., 2006). Thus, the BAs profile excreted in feces, mainly composed of secondary BAs, largely depends on the gut microbiota metabolism (Perwaiz et al., 2002).

#### Deconjugation of BAs

Bile salt hydrolases encoding genes have been detected and characterized in diverse gut microbes including species belonging to the genera Bacteroides, Clostridium, Lactobacillus, and Bifidobacterium, among others, being more diverse in members of the phylum Firmicutes (**Figure 2** and **Supplementary Figure 1**) (Jones et al., 2008). BSH activity has been suggested as a BA detoxification mechanism for bacteria, although they may also obtain carbon, nitrogen and even sulfur from BA deconjugation. This latter element has relevance in the production of hydrogen sulfide that may have lasting health consequences as it increases colonocyte turnover and has been associated with inflammation and cancer (Carbonero et al., 2012). Through regulation of key genes involved in cholesterol metabolism and gastrointestinal homeostasis, BSH activity was proposed as a gut microbial activity with capacity to profoundly alter local (gastrointestinal) and systemic (hepatic) host functions as revealed by different studies in mice (Joyce et al., 2014).

#### 7-Dehydroxylation of BAs

The conversion of primary BAs to secondary BAs by 7α-dehydroxylases is probably one of the most physiologically relevant microbial transformations of BAs in humans (Duboc et al., 2013). Through 7α-dehydroxylation, the primary cholic acid is transformed into the secondary deoxycholic acid, and the primary chenodeoxycholic acid is transformed into the secondary lithocholic acid. To date, 7α-dehydroxylation activities have been characterized only in species belonging to the genera Eubacterium and Clostridium, including the species Clostridium scindens and Clostridium hylemonae (Ridlon et al., 2010). C. scindens is also capable of performing a 7β-dehydroxylation on ursodeoxycholic acid (the 7β epimer of chenodeoxycholic acid), yielding lithocholic acid (Ridlon et al., 2006, 2016).

#### Other Microbial Enzymatic Activities Acting on BAs

Other BA modifications such as amidation, oxidation-reduction, epimerization, esterification and desulfatation, can be carried out by intestinal microbes. Among them, oxidation-reduction and epimerization have received particular attention as some intestinal microbes synthesize HSD capable of performing reversible oxidation/reduction reactions and hydroxyl groups epimerization (Ridlon et al., 2016). Indeed, BA epimerization reactions have been largely overlooked due to the lack of appropriate analytical methods, although some iso-BAs have been suggested to represent the most abundant BAs in human feces (Hamilton et al., 2007). HSD activities are present in the four major phyla of the intestinal microbiota Actinobacteria, Proteobacteria, Firmicutes, and Bacteroidetes (Wahlströ et al., 2016), and the capability to carry out epimerization reactions has been characterized in several intestinal bacteria, including Clostridium, Collinsella, Ruminococcus or Eubacterium species (White et al., 1982; Lepercq et al., 2004; Liu et al., 2011; Lee et al., 2013). However, the physiological and functional significance of this metabolic activity remains largely unclear.

#### Host Health Implications of Microbial Bile Metabolism

The microbial-mediated transformations of BAs at the intestinal level have been shown to be essential for intestinal and

systemic health maintenance as the intestinal BAs and the gut microbiota mutually influence each other and, accordingly, BA-microbiota crosstalk disruption has been associated with several gastrointestinal, metabolic and inflammatory disorders, including those associated with aging-related decline (Jia et al., 2017), as summarized below.

#### BAs Metabolism and Inflammation

The gut microbiota-mediated biotransformation of the BA pool regulates BAs signaling by affecting the activation of host BA receptors such as the nuclear receptor farnesoid X receptor (FXR), which governs bile, glucose and lipid metabolism (Gadaleta et al., 2011). Indeed, a disrupted gut microbiota including reduced bile metabolizing bacteria significantly impairs BA metabolism and, consequently, the host metabolic pathways regulated by BA signaling, affecting glucose and cholesterol homeostasis, as well as immune states. Indeed, disorders associated with chronic low-grade inflammation have been linked to gut dysbiosis and altered BA profiles in humans (Chavez-Talavera et al., 2017), although few works have established a connection among specific activities of the microbiota on bile and cholesterol and the physiological alterations observed. As an example, analysis of existing gut metagenomic datasets evidenced that the abundance of the BSH gene bsh was significantly reduced in inflammatory bowel disease (IBD) and type-2 diabetes patients (Labbé et al., 2014). Accordingly, IBD patients evidenced increased fecal conjugated and sulphated BAs, and reduced fecal secondary BAs, suggesting the existence of characteristic alterations of bile metabolism associated with gut microbial shifts in IBD (Duboc et al., 2012, 2013). Indeed, some of these changes might be linked to dietary factors such as a diet high in saturated fat and increased sulfur-rich taurine conjugate BAs, which in turn promoted the expansion of the sulphite-reducing pathobiont Bilophila wadsworthia in mice. The resulting dysbiosis lead to an associated pro-inflammatory Th1 response and acute colitis in a mouse model, further demonstrating how microbial activity on a particular BA can impact inflammatory states and host health (Devkota et al., 2012).

#### BAs Metabolism and Colorectal Cancer

The relation between diet, microbial metabolism of BAs and human disorders, including colorectal cancer risk (CRC), is further supported by the fact that dietary fat increases biliary hepatic synthesis and, thus, the quantity of BAs that reach the colon, providing substrate for the synthesis of secondary BAs. These have been described as proinflammatory (Bernstein et al., 2011) and their increase may contribute to the pathogenesis of several gastrointestinal diseases, having been associated with colon polyps (de Kok et al., 1999) and CRC (Bernstein et al., 2005; O'Keefe et al., 2015). Indeed, fecal secondary BAs and microbial genes encoding for 7α-dehydroxylases were more common in African Americans who had a high risk of suffering CRC as compared with rural native Africans (Ou et al., 2013).

#### BAs Metabolism and Liver Diseases

Several chronic liver-related disorders, including non-alcoholic fatty liver disease (NAFLD), primary sclerosing cholangitis, steatosis and hepatic cancer – frequently associated with obesity – have been related to different intestinal microbial patterns (Adolph et al., 2018). In some of these diseases, an altered livermicrobiota-BAs crosstalk has also been defined. For instance, the ratio between primary and secondary BAs in feces and the levels of conjugated and unconjugated BAs in serum are higher in NAFLD patients (Kakiyama et al., 2013; Mouzaki et al., 2016; Jiao et al., 2018). Interestingly, an increase in taurine metabolizing activities has been evidenced in the gut microbiota of these patients, associated with increased representation of Bilophila species, and increased secondary BAs production (Jiao et al., 2018). Additionally, NAFLD is frequently associated with obese patients, for whom specific dysbiosis signatures have been defined (Gao et al., 2018). Consequently, in addition to affecting bile metabolism within the gut, the microbiota might also contribute to NAFLD pathogenesis through other mechanisms including increased energy intake, intestinal permeability and contribution to chronic pro-inflammatory states (Han et al., 2018), which go beyond the scope of this mini review.

#### Gut Microbiota Shifts in Aging Impact BAs Metabolism and Signaling

Gut microbiota changes throughout life, including loss of diversity, are associated with lifestyle and dietary changes in the elderly population, though they may also modulate elements of aging frailty such as innate immunity or cognitive function. Indeed, recent studies have evidenced that alterations in BAs metabolism accompany these aging-associated microbiota shifts and health decline. For instance, increased fecal excretion of deconjugated BAs has been observed in old mice in association with a shift toward pro-inflammatory states in the gut (Becker et al., 2019). In addition, a reduction in cholic acid and an increase in secondary BAs have been noticed in the serum of patients with Alzheimer disease (AD) (MahmoudianDehkordi et al., 2019), presumably reflecting augmented 7α-dehydroxylase activity in the gut microbiota. In fact, a mice model of AD has evidenced changes in the gut microbiota, including an increase in members of the Clostridium group, among which 7α-dehydroxylase activity is frequent (Brandscheid et al., 2017). Nevertheless, comprehensive studies of the gut microbiota and concomitant BAs metabolic changes in AD human cohorts are still lacking.

#### MICROBIOTA MODULATION OF BILE AND CHOLESTEROL METABOLISM: INFLUENCE ON HOST PHYSIOLOGY AND SIGNALING MECHANISMS INVOLVED

Several studies on germ-free animal models have evidenced the microbiota's involvement in cholesterol and bile metabolism. For instance, the lack of gut microbiota in mice deficient in ApoE (a protein involved in the metabolism of fats) increased the plasma and liver cholesterol levels and reduced hepatic BAs synthesis (Kasahara et al., 2017). Also, the reverse cholesterol

transport from peripheral tissues to the liver is augmented in germ-free mice (Mistry et al., 2017). These observations suggest that specific targeting of the intestinal microbiota could significantly impact cholesterol metabolism and cardiovascular diseases. Furthermore, germ-free animals lack secondary BAs production, and their microbial colonization modifies intestinal and serum BA fingerprinting, increasing total BAs concentrations (Joyce et al., 2014).

Since BAs are ligands of bile-responsive receptors involved in host metabolism, changes in BAs composition orchestrated by the intestinal microbiota activity, may affect their interaction with specific receptors, such as pregnane-activated receptor, vitamin D receptor, sphingosine-1-phosphate receptor, muscarinic receptor (Ridlon et al., 2016). Additionally, FXR, a nuclear transcription factor that regulates a wide range of genes (Teodoro et al., 2011), as well as the plasma membrane-bound G-protein coupled receptor TGR5 (Kawamata et al., 2003), have been remarkably characterized in relation to bile signaling. Both receptors are ubiquitously distributed in several tissues and have different affinity for individual BAs. TGR5 is mainly activated by the secondary BAs litocholic and deoxycholic acids, and recognizes both conjugated and deconjugated forms (Long et al., 2017). The most potent ligand for FXR is chenodeoxycholic acid, with cholic acid, deoxycholic acid and litocholic acid having a lower effect (Wahlströ et al., 2016). FXR activation can induce innate immune genes, promote the synthesis of antimicrobial agents acting on the gut microbiota (Inagaki et al., 2006), and regulate BA synthesis (Sinal et al., 2000). On the other hand, TGR5 plays a role in the regulation of BA and energy homeostasis (Wahlströ et al., 2016). Therefore, through these receptors, BAs act as signaling factors beyond the GIT. Further, considering that the gut microbiota deeply influences the BAs signature, different microbial communities can differentially impact bile signaling and determine the degree of activation of these receptors, with a concomitant impact on host metabolism. Indeed, BA receptors are currently considered therapeutic targets for several gastrointestinal and hepatic diseases (Firoucci et al., 2007); thus, microbiota-based approaches to modulate their activation may represent novel alternatives for certain disorders and warrant further investigation.

#### FUTURE PERSPECTIVES: POTENTIAL OF MICROBIOTA-BASED APPROACHES TO MODULATE BILE METABOLISM AND ASSOCIATED CONDITIONS

In light of the recently unearthed gut microbiota-BA-host signaling interactions, microbiota-based approaches, from probiotics to dietary interventions, may become novel strategies to manage specific diseases linked to BAs metabolism dysregulation, as suggested by some in vivo studies (Devkota and Chang, 2015; Fukui, 2017). Most studies to date have focused on the potential of probiotics administration to reduce serum cholesterol levels. In this context, administration of probiotic strains to healthy mice increased deconjugation of BAs and fecal excretion (Jeun et al., 2010; Degirolamo et al., 2014) in association with increased BSH activity in the gut and overall modification of the microbiota composition (Degirolamo et al., 2014; Joyce et al., 2014; Tsai et al., 2014; Lye et al., 2017), changes that may have implications for host lipid metabolism. Indeed, a cholesterol-lowering effect was also observed following supplementation of a BSH-positive Lactobacillus strain to mice fed high-fat diets (Michael et al., 2017). However, limited studies have been conducted in human subjects in this regard. Remarkably, consumption of a BSH-positive Lactobacillus strain significantly reduced cholesterol in hypercholesterolemic subjects (Jones et al., 2012), although the observed effect might be the result of a complex metabolic re-arrangement, rather than solely a consequence of an increase in bile excretion.

Some probiotic interventions have also demonstrated their efficacy to ameliorate liver and inflammatory markers in models of NAFLD and IBD, although results are not yet conclusive (Han et al., 2018; Kobyliak et al., 2018). Besides, the strains tested in most studies were not specifically selected for their activities over bile metabolism, and the impact of the intervention on the fecal or serum BAs profiles, on the fecal microbiota composition or on their metabolic capability over bile and cholesterol, was not always evaluated. This strongly hampers establishing causal relationships between the metabolic activities of the microbiota over these compounds and the physiological effects observed.

Diet is another factor known to affect the gut microbiota and the BAs host signature. For instance, in a dietary intervention study in humans, a diet rich in animal-based fats was associated with increased excretion of secondary BAs, in accordance with an increased overall expression of bsh encoding genes in the gut microbiota, and an increase in the representation of potential pathobiont species such as B. wadsworthia (David et al., 2014). Thus, dietary strategies aimed at modulating BA metabolism through balancing the microbiota may represent alternative approaches to manage diseases linked to BA dysmetabolism (Ghaffarzadegan et al., 2018). Though these have been scarcely studied in humans, studies in mice models have showed the potential of certain dietary ingredients to modulate gut microbiota and BAs profile. For instance, Akkermansia muciniphila enrichment through administration of epigallocatechin-3-gallate prevented diet-induced obesity and regulated bile signaling (Sheng et al., 2018), although the contribution of changes in specific microbial metabolic activities over bile and cholesterol in this model has not been determined.

#### CONCLUSION

In summary, it has become increasingly clear that BAs exert a much wider range of biological activities than initially recognized and that BAs, gut microbiota and health status are closely linked and hold a yet- underexplored valuable potential to design novel diagnostic and therapeutic approaches based on specific gut microbiota activities. Elucidating the molecular mechanisms underlying the gut microbiota-BA-host health interplay will establish the basis to fully understand the gut microbiota potential to

modulate bile metabolism and host health. Further studies using specifically designed in vivo models or human trials, and exploiting microorganisms or activities with demonstrated capacity to specifically act on selected BAs, are necessary for aiding the development of novel microbiome-based approaches for disorders associated with BAs dysregulation.

#### AUTHOR CONTRIBUTIONS

AM, SD, and BS conceived and organized the manuscript. NM designed the figures. NM, LR, BS, AM, and SD contributed to the writing, critically reviewed the manuscript, and approved the final version of the manuscript.

#### REFERENCES


#### FUNDING

This study was supported by MINECO under grant number AGL2013-44761-P. LR is a postdoctoral researcher supported by the Juan de la Cierva Postdoctoral Trainee Program (MINECO, JCI-2015-23196) and NM is the recipient of an FPI Predoctoral Grant (BES-2014-068736).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphys. 2019.00185/full#supplementary-material


Gao, R., Zhu, C., Li, H., Yin, M., Pan, C., Huang, L., et al. (2018). Dysbiosis signatures of gut microbiota along the sequence from healthy, young patients to those with overweight and obesity. Obesity 26, 351–361. doi: 10.1002/oby.22088



gnavus N53 to ursodeoxycholic acid formation in the human colon. J. Lipid Res. 54, 3062–3069. doi: 10.1194/jlr.M039834



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Molinero, Ruiz, Sánchez, Margolles and Delgado. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Dose-Dependent Effects of Aloin on the Intestinal Bacterial Community Structure, Short Chain Fatty Acids Metabolism and Intestinal Epithelial Cell Permeability

#### Kuppan Gokulan\*, Pranav Kolluru, Carl E. Cerniglia and Sangeeta Khare\*

Division of Microbiology, National Center for Toxicological Research, United States Food and Drug Administration, Jefferson, AR, United States

#### Edited by:

Yuheng Luo, Sichuan Agricultural University, China

#### Reviewed by:

Alex Galanis, Democritus University of Thrace, Greece Jing Zhang, Shanghai Jiao Tong University, China

#### \*Correspondence:

Kuppan Gokulan kuppan.gokulan@fda.hhs.gov Sangeeta Khare sangeeta.khare@fda.hhs.gov

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 31 July 2018 Accepted: 25 February 2019 Published: 26 March 2019

#### Citation:

Gokulan K, Kolluru P, Cerniglia CE and Khare S (2019) Dose-Dependent Effects of Aloin on the Intestinal Bacterial Community Structure, Short Chain Fatty Acids Metabolism and Intestinal Epithelial Cell Permeability. Front. Microbiol. 10:474. doi: 10.3389/fmicb.2019.00474 Aloe leaf or purified aloin products possess numerous therapeutic and pharmaceutical properties. It is widely used as ingredients in a variety of food, cosmetic and pharmaceutical products. Animal studies have shown that consumption of aloe or purified aloin cause intestinal goblet cell hyperplasia, and malignancy. Here, we tested antibacterial effects of aloin, against intestinal commensal microbiota. Minimum inhibitory concentration of aloin for several human commensal bacterial species (Gram-positive and Gram-negative) ranged from 1 to 4 mg/ml. Metabolism studies indicated that Enterococcus faecium was capable of degrading aloin into aloe-emodin at a slower-rate compared to Eubacterium spp. As a proof of concept, we incubated 3% rat fecal-slurry (an in vitro model to simulate human colon content) with 0.5, 1, and 2 mg/ml of aloin to test antimicrobial properties. Low aloin concentrations showed minor perturbations to intestinal bacteria, whereas high concentration increased Lactobacillus sp. counts. Aloin also decreased butyrate-production in fecal microbiota in a dosedependent manner after 24 h exposure. The 16S rRNA sequence-data revealed that aloin decreases the abundance of butyrate-producing bacterial species. Transepithelial resistant result revealed that aloin alters the intestinal barrier-function at higher concentrations (500 µM). In conclusion, aloin exhibits antibacterial property for certain commensal bacteria and decreases butyrate-production in a dose -dependent manner.

#### HIGHLIGHTS


**218**


Keywords: aloin, commensal bacterial community, antimicrobial activity, epithelial cell barrier function, short chain fatty acids

# INTRODUCTION

fmicb-10-00474 March 23, 2019 Time: 17:20 # 2

Aloe Vera is a tropical medicinal plant that belongs to the Liliaceae family (Ding et al., 2014). Aloe is a genus that consists of 548 recognized species. Aloe leaf has been used for medicinal purposes since ancient times (Morton, 1961). The two major components of aloe leaves are the mucilaginous inner gel and the outer dermal layer. The inner gel is composed of 99% water and several bioactive molecules, vitamins, sugars, amino acids, lipids, sterols, and phenolic compounds, most of which have medicinal values (Park and Kwon, 2006). Phenolic molecules, chromones, anthraquinones glycosides, and Aloin (A and B isomer) make up approximately 30% of the outer cutaneous layer (Gutterman and Chauser-Volfson, 2000; Hamman, 2008). It is important to note that the composition of bioactive molecules in Aloe vera varies depending on the age of the plant.

In Asian and African countries, the aloe leaves or extracted juice are often used to treat infectious and inflammatory diseases (Thiruppathi et al., 2010; Akinduti et al., 2013; Megeressa et al., 2015). Aloe vera derived products are also used to treat hair loss and constipation. Studies have suggested that in certain regions of the world, aloe leaf is used to treat asthma, constipation, hypertension, and excessive perspiration (Lans, 2006). More recently, the jelly products have been incorporated as an ingredient in cosmetics, skin care products, health supplements, health beverages, herbal remedies, and also used for their wound healing, anti-inflammatory, antimicrobial, and anti-oxidant activities (Briggs, 1995). The United States Food and Drug Administration (FDA) has approved the use of aloe leaf product ingredients for natural food coloring purpose, cosmetics, and dietary polyphenolic supplements (FDA, 2008); however, the exact molecular mechanisms of its beneficial effects is poorly understood.

The intestinal epithelial layer of the gastrointestinal system operates as a physical barrier between the internal system and outside environment (Odenwald and Turner, 2013). The intestinal system also provides a microbial habitat for 10<sup>10</sup> to 10<sup>14</sup> commensal microbiota (Ley et al., 2006; Qin et al., 2010) per gram feces. In healthy individuals, the composition of commensal bacteria is a stable microbial community, which shapes the development of the local immune system in the gut. In addition, the microbiome aids short chain fatty acid (SCFA) production by degrading undigested polysaccharides and other food materials. Bacterial derived butyrate serves as a major energy source for intestinal epithelial cell growth and maintenance of barrier function (Zheng et al., 2017). Progressions of intestinal disorders and diseases have been associated with shifts in the normal intestinal microbiome and dysbiosis in butyrate-producing intestinal bacteria (Clemente et al., 2012; Zheng et al., 2017). The metabolites of the ingested aloin can be absorbed by intestinal epithelial cells (Park et al., 2009) while traveling through the gastrointestinal tract (GIT) and these molecules may interact with commensal microbiota and alter intestinal homeostasis.

In the GIT, ingested aloin can be metabolized into deglycosylated aloe-emodin by intestinal bacteria (Che et al., 1991). The purified phenolic aloe products and its metabolites have free hydroxyl methyl groups which may interact with other hydroxyl groups or glucuronide enzymes to form glucuronide conjugates. Glucuronate derivatives are more stable than sulfate derivatives and require more time to clear from the system. In vitro permeability studies have revealed that intestinal epithelial cells absorb aloin and its metabolites in the following order: alosin > aloe-emodin > aloin (Park et al., 2009). These metabolites retain the phenolic structures that are considered bioactive molecules. Studies have also demonstrated that aloin exhibits antimicrobial and antimalarial properties against many pathogenic organisms (Asamenew et al., 2011; Abeje et al., 2014; Oumer et al., 2014).

Few studies provide evidence that Aloe vera products are safe to use and do not cause toxic effects (Sehgal et al., 2013). In contrast, many other studies have shown that the ingestion of aloe-products causes intestinal abnormalities (Boudreau et al., 2013, 2017; Sehgal et al., 2013). Since aloin and its metabolites exhibit antimicrobial properties, it is reasonable to conduct safety assessment of aloe-derived products for intestinal microbes, as well. Previous rodent studies have provided evidence that administration of aloe leaf or purified polyphenolic compounds causes lesions, cytotoxicity, and the progression of adenocarcinoma in the intestine (Xia et al., 2007; Boudreau et al., 2017). Interestingly, these intestinal abnormalities were found to be more predominant in males than females (Xia et al., 2007). Thus, systematic approaches are necessary to determine the concentration of aloin that can cause detrimental effects on intestinal microbiota. Specifically, the concentration of aloin responsible for the alteration of the intestinal microbial community and the integrity of the gastrointestinal system is not known. Additionally, there is a knowledge-gap regarding which experimental approaches should be used to assess whether the barrier function of the intestinal epithelial layer is compromised due to xenobiotic exposure and/or its metabolites. Furthermore, the antibacterial effect of aloin on intestinal commensal bacteria and production of SCFAs from undigested fiber food materials also remains to be answered.

In the present study, we hypothesized that aloin and its metabolites may cause detrimental effects to the gastrointestinal system. To test our hypothesis, we have employed in vitro experimental approaches to evaluate aloin toxicity toward

commensal bacteria and intestinal epithelial cells. In this study, our main objectives were fivefold: (i) to evaluate if there are any differences in the antibacterial properties of aloin between aerobic, facultative, and anaerobic human intestinal bacteria that are predominant in the intestine, (ii) to examine the effect of aloin on changes in the aerobic and anaerobic bacterial community structures using an in vitro rat fecal culture model and 16s rRNA sequencing, (iii) to examine the effect of aloin on SCFA production by rat intestinal bacterial community, (iv) to evaluate the effect aloin on intestinal permeability by in vitro methodology, and (v) to analyze the metabolism of aloin by intestinal commensal bacteria.

#### MATERIALS AND METHODS

#### Antibacterial Activity of Aloin on Pathogenic Bacteria and Human Intestinal Bacteria Growth

To test the antimicrobial properties of aloin, we used a reference Escherichia coli strain as a test organism. Overnight grown E. coli J53 strain was sub-cultured in tryptic soya agar (TSA) media to reach mid log phase, then diluted and plated (1 × 10<sup>5</sup> cells/well) into 96 well culture plate. Aloin stock (CAS 1415-73-2 procured from Pure Chemistry Scientific Inc., Houston, TX, United States) solution was prepared (10 mg/ml) by adding ethanol and culture media in 1:4 ratio; then a twofold dilution method was used to obtain 2mg/ml to 0.015 µg/ml concentration in culture plates. Control wells received only media and ethanol (equivalent concentration of ethanol present in 2 mg/ml aloin solution). The negative control was media alone without bacteria and served as a correction value for background reading. Control wells containing 2 mg/ml aloin and media were also included to detect background absorption of aloin. Plates were placed in BioTek plate reader (Cytation, BioTek Instruments, Inc., Winooski, VT, United States) and OD was recorded at 600 nm with 5 min intervals for a 24 h period to evaluate growth.

We selected six commensal bacterial strains from five major bacterial phyla based on its functionality in the intestine. These human commensal bacteria were obtained from ATCC (ATCC, Manassas, VA, United States) and included Bifidobacterium longum (ATCC <sup>R</sup> 15707TM), Lactobacillus acidophilus (strain ATCC 700396/NCK56/N2/NCFM), Enterococcus faecium (ATCC <sup>R</sup> 19434TM), Bacteroides thetaiotaomicron (ATCC <sup>R</sup> 29148TM), Akkermansia muciniphila (ATCC <sup>R</sup> BAA-835TM), and Eubacterium sp. (ATCC <sup>R</sup> BAA-148TM). These strains were cultured in specific bacterial growth media as recommended by ATCC. The anaerobic bacteria were cultured in the anaerobic chamber. The facultative bacteria were cultured in partially anaerobic condition at 37◦C.

#### Evaluation of Minimum Inhibitory Concentration (MIC) of Aloin Against Intestinal Commensal Bacteria

Bacterial strains were inoculated overnight in the respective culture media at 37◦C in a shaker as recommended by ATCC. Periodic OD<sup>600</sup> measurements were taken to evaluate bacterial growth and the incubation was halted when the organism growth reached mid-log phase. Bacterial cultures were diluted using respective culture media and adjusted to a density of approximately 1 × 10<sup>5</sup> cells/ml, after which, 100 µl (10,000 cells) were plated in each well of the 96 well plates. Aloin was prepared with 1:4 ratios of ethanol and respective culture media and added twofold dilutions starting from 4 mg/ml (9.56 mM) to 3.98 µg/ml (0.019 mM). Positive (bacterial culture without aloin) and negative controls (media alone and aloin alone without bacterial culture) were included in each experiment. Plates were placed in BioTek plate reader and OD was recorded at 600 nm with 5 min intervals for 24 h period to evaluate MIC. The MIC value (i.e., when aloin concentration totally inhibited bacterial growth) of aloin against intestinal commensal bacteria was determined.

#### Visualization of L. acidophilus Viability Using Acridine Orange and Ethidium Bromide Fluorescence

After the assessment of bacterial growth, the wells containing L. acidophilus were stained for detection of cell death according to the protocol developed by Ribble et al. (2005). Acridine orange (2 µg/ml) and ethidium bromide (2 µg/ml) dyes were added to each well and incubated at 37◦C for 15 min. The stained cells were visualized and photographed using an inverted fluorescence microscope (EVOS, Life Technologies, San Diego, CA, United States) to differentiate between live (green) and dead (red) bacteria.

#### In vitro Microgravity Rotary Culture Conditions of Fecal Microbiota

Fecal samples were collected post-mortem from the colon of three male Sprague Dawley rats (aged 4–6 months) who undergo regular surveillance at NCTR. As these fecal samples were collected post-mortem from the animals, an approval from the Institutional Animal Ethics Committee was not required.

To mimic the intestinal continuous movements, we chose a rotary cell culture system (RCCS) (Synthecon Inc., Houston, TX, United States) for the interaction of aloin with intestinal bacteria. The RCCS was developed for zero gravity (space) research by Wolf et al. (1988). This system was specifically designed to grow long term culture of functional primary human liver cells under low microgravity environment with low shear force, high mass transfer and 3-D cell culture of dissimilar cell types. In the present study, we used the same technology, however, customized the vessel to mimic an enclosed system with an anaerobic environment. The RCCS was placed inside of anaerobic chamber with low speed to mimic peristaltic movement of the intestine. We employed this system to study the effect of aloin on intestinal bacteria to mimic anaerobic environment of colon. We prepared aloin at 10 mg/ml concentration as described in "Materials and Methods" Section "Antibacterial Activity of Aloin on Pathogenic Bacteria and Human Intestinal Bacteria Growth." Fecal slurry (3%) was prepared using low carbohydrate medium (LCM) under anaerobic conditions as described earlier (Kim et al., 2011;

Jung et al., 2018). The fecal slurry was aliquoted into four equal volumes. The final concentrations of all four fecal slurries were adjusted to 3% using LCM. The fecal slurry was transferred into 10 ml custom made sterile anaerobic disposable culture vessels (four vessels) using a syringe. Care was taken to remove air bubbles. These vessels were fixed into a RCCS and the rotation speed was adjusted to 7 rpm/min. The experiment preparations and setup were conducted in an anaerobic chamber. The control sample received 400 µl of ethanol, which is equivalent to the highest concentration of ethanol that is present in 2 mg/ml aloin solution. In the remaining three vessels, aloin was added (0.5, 1, and 2 mg/ml, respectively) to analyze changes in the intestinal bacterial community structure, the concentrations were based on the MIC obtained from E. coli bacterial growth curve. A portion of the samples from these vessels were collected at 3, 6, and 24 h time points to analyze changes in the bacterial structural community for both aerobic and anaerobic bacteria, SCFAs production, and aloin metabolism by intestinal bacteria.

# Bacterial Cultures From Rat Fecal Slurry Treated With Aloin

Fecal samples were collected from all rotary vessels as indicated in Section "In vitro Microgravity Rotary Culture Conditions of Fecal Microbiota." To analyze the bacterial composition, 100 µl from each sample was serially (10-fold) diluted in dilution blank solution. The total aerobic and anaerobic bacterial populations were quantified using selective culture media plates (TSA and Brucella Blood Agar, respectively). These samples were also plated in a selective bacterial culture medium for anaerobic [BBE (for Bacteroides) and Bifido (for Bifidobacterium genus)], and facultative bacteria (LRMS). Culture plates were incubated at 37◦C. The aerobic and facultative bacterial plates were removed after 24 h. The anaerobic bacterial plates (BBE and Bifido) were removed after 72 h from the anaerobic chamber. Bacteria were then enumerated using a colony counter. The bacterial count obtained from control group was compared to that of the three experimental groups at each time point.

# Sample Preparation to Measure Short Chain Fatty Acids

The HPLC equipment, 1260 DAD LC Agilent (Agilent Technologies, Santa Clara, CA, United States), was used to measure SCFAs. Standards of succinic acid, lactic acid, acetic acid, propionic acid, isobutyric acid butyric acid, valeric acid, isovaleric acid, and hexanoic acid were obtained from Sigma-Aldrich (St. Louis, MO, United States). Standards were prepared freshly at 1 M concentration and 20 mM spiked in control fecal supernatant. Before analyzing the test samples, retention times for standards were calibrated using Aminex HPX-87H column (300 × 7.88 mm) (Bio-Rad, Richmond, CA, United States). The column was maintained at 65◦C and the column flow rate (0.6 ml/min) and molecules were monitored continuously at 210 nm. The mobile phase was composed of an isocratic H2SO<sup>4</sup> solution (2.5 mM) for 50 min. An aliquot (1 ml) of fecal slurry was removed at 3, 6, and 24 h for SCFA analysis by HPLC method (Gordon et al., 1982). Samples were immediately centrifuged at 15,000 rpm for 10 min at 4◦C and supernatants were collected and filtered through a 0.45 µm membrane-filter. A 20 µl of sample solution was injected into HPLC column to analyze SCFAs metabolites.

# Sample Preparation for Assaying Aloin Metabolism by Commensal Bacteria

After evaluating the MIC value from pure bacterial culture, samples were centrifuged to remove the bacterial pellet. The supernatant was transferred in a new tube then kept in speedvac to remove culture media. Aloin and its metabolites were extracted using methanol. Aloin and aloe-emodin standards were prepared freshly and evaluated for the retention time using Zorbox-SB C18 HPLC column. These standards were also spiked in control bacterial culture media, retention time for extracted aloin and aloe-emodin was analyzed as stated earlier. Experimental samples were analyzed by HPLC column to evaluate aloin and its metabolites. Column temperature was kept at 40◦C. Buffer A was 95% water and 5% Acetonitrile containing 0.1% TFA. Buffer B was 95% Acetonitrile and 5% water containing 0.1% TFA. Flow rate was 0.2 ml/min and absorbance was monitored at 260 nm.

# Extraction of DNA From Fecal Samples

Fecal DNA was extracted following the protocol described earlier with slight modifications (Khare et al., 2004; Williams et al., 2015). One ml of fecal sample was collected from each experimental group at 3, 6, and 24 h. DNA was extracted from fecal samples as described earlier. The pellet was suspended in the DNase and RNase free water and DNA was quantified using a Nanodrop ND-1000 spectrophotometer (Nanodrop, Wilmington, DE, United States). In addition, dsDNA was quantified by Qubit (Thermo Fisher Scientific, Waltham, MA, United States). This highly purified DNA was used as template for 16s sequencing for microbial population.

# 16S rRNA Gene Sequencing of Bacterial Population

Highly purified fecal DNA was used to analyze change in bacterial community due to aloin exposure. The V4-variable region of 16S rRNA gene was amplified using PCR primers 515/806. Pooled and purified PCR products were used to prepare Illumina DNA library. The output of the DNA sequence reads were joined and subsequently, the barcodes were depleted and then sequences <150 bp were removed. Sequences with ambiguous base calls were also removed for data analysis. Sequences were denoized, operational taxonomic units (OTUs) generated and chimeras were removed. OTUs were defined by clustering at 3% divergence (97% similarity). Final OTUs were taxonomically classified using BLASTn against a curated database derived from RDPII and NCBI.

# Intestinal Epithelial Cell Culture

T84 cells (ATCC <sup>R</sup> CCL-248TM), a human colorectal carcinoma cell line, were obtained from ATCC. Cells were cultured in complete growth media reported previously (Gokulan et al., 2016, 2017), which was composed of Dulbecco's Modified

Eagle Medium (DMEM)/F-12 medium supplemented with Lglutamine and HEPES (ATCC), with added 5% fetal bovine serum, penicillin/streptomycin, and Fungizone. Initially, a 75 cm<sup>2</sup> cell culture flask was used to grow the cells until they reached 70–80% confluency. Cells were detached with 0.25% trypsin-EDTA solution and washed twice with DMEM/F-12 medium. The media was decanted, and cell pellet was suspended with culture media. Cells were counted, then seeded (2.0 × 10<sup>5</sup> cells/well) into transwells and maintained in a 37◦C incubator with 5% CO<sup>2</sup> and 95% humidity until transepithelial electrical resistance (TER) values were stabilized.

#### Transepithelial Electrical Resistance (TER)

The TER measurement was performed as reported previously (Adams et al., 1993; Youakim and Ahdieh, 1999; Bruewer et al., 2003; Donato et al., 2011; Khare et al., 2012; Williams et al., 2016). Briefly, T84 cells were seeded in the apical compartment of 6.5 mm, PFTE, collagen-coated transwells inserts (Corning, Corning, NY, United States) at a concentration of 2 × 10<sup>5</sup> cells/well. Complete cell culture media was added to apical (0.2 ml) and basal reservoirs (0.8 ml) and cells were allowed to grow for 5–7 days to polarize (Williams et al., 2016). Then antibiotic-free medium was added to apical and basal reservoirs and TER was measured periodically using a STX electrode probe and EVOM2 Epithelial Voltohmmeter (World Precision Instruments, Sarasota, FL, United States). Once the wells had reached approximately 800–1000 /cm<sup>2</sup> the medium was changed and cells were allowed to equilibrate for 3 h. Aloin was dissolved in ethanol and media in a 1:4 ratio (10 mg Aloin dissolved in 0.2 ml ethanol and 0.8 ml of culture media at pH 7.2 or citrate buffer pH 4.9) and further diluted in cell culture media for the TER studies. The control wells contained 0.8% of ethanol (500 µM Aloin contains 0.8% ethanol). Baseline TER reading was taken before adding test compound as an initial value, then with the following aloin concentrations (0.05, 0.5, 1.5, 5, 50, and 500 µM) in apical compartments. TER measurements of individual transwells were taken before and after exposure to aloin in all wells (including controls) at 1, 2, 3, 4, 24, and 48 h post-exposure.

#### Statistical Analysis

The bacterial growth curve data was analyzed using one-way ANOVA to determine statistical significance in culture treated with different concentration of aloin. One-way ANOVA was also used to find out statistical significance on SCFAs production.

During 16S rRNA analysis, data were normalized by median and auto-scaling (mean centered/standard deviation of each variable). The effect of aloin-exposure on the bacteria representative of specific-phylum, genus and species level was calculated as the percent abundance. Next, a one-way ANOVA was performed to obtain significant differences between the experimental groups and control for Phyla, genes, and species level. For Genus and Species level analysis, Orthogonal Partial Least Squares Discriminant Analysis (PLS-DA) was used to observe the separation between different experimental groups and to assess the similarities within an experimental group, as described earlier (Al-Momani et al., 2016; Rabbi et al., 2016; Gokulan et al., 2018) The beta diversity was provided as the intensity of heatmap, where the clustering result shown as heatmap (distance measure using euclidean, and clustering algorithm using ward's method). The heatmap here is expressed by the abundance of genus or species while comparing the controls and treatment groups (individual columns depict the average of one experimental group).

#### RESULTS

#### Antimicrobial Activity of Aloin Against E. coli J53 Strains

To assess the antimicrobial activity of aloin against the reference strain E. coli J53, a sub-culture was incubated with aloin from 2 mg/ml to 0.015 µg/ml by twofold serial dilution. The bacterial growth was monitored for 24 h with data collection at 5 min intervals. The growth curve kinetic data of control E. coli at low aloin concentrations indicated that there was an initial 3 h lag phase, an exponential phase for 5–6 h, and stationary phase at 14 h. At the two higher concentrations of aloin (1 and 2 mg/ml), treated wells showed the complete inhibition of E. coli growth, which was statistically significant (p < 0.001) as compared to bacteria exposed to lower concentrations (0.015 µg/ml to 0.5 mg/ml) of aloin (**Figure 1A**). Wells containing only 2 mg/ml of aloin were measured and had absorbance values between 0.1 and 0.15, which was subtracted from experimental wells treated with 2 mg/ml of aloin. The aloin MIC value for E. coli strain was 1 mg/ml.

#### Minimum Inhibitory Concentration of Aloin Against Intestinal Commensal Bacterial Strains

The same experimental approach was carried out to determine the minimal inhibitory concentration of aloin for human commensal bacteria (i.e., L. acidophilus, B. longum; E. faecium, B. thetaiotaomicron, A. muciniphila, and Eubacterium sp.). For some bacteria, aloin concentration was increased above 2 mg/ml to 4 or 8 mg/ml since several bacterial species were not susceptible to 1 or 2 mg/ml of aloin. These experiments were conducted aerobically, anaerobically, or microaerophilic based on the growth condition requirements of the bacteria.

Bifidobacterium longum growth curve revealed that 1 and 2 mg/ml of aloin exposure slightly reduced bacterial growth, whereas lower concentrations of aloin had no effect on bacterial growth (**Figure 1B**). B. longum exposed to 4 mg/ml aloin also showed growth pattern similar to 2 mg/ml aloin exposure. Overall, the difference in the growth of B. longum was not statistically significant. This result showed that the aloin showed poor antibacterial activity toward B. longum; hence, no MIC value was recorded in the bacterial growth curve. The growth kinetic results suggest that under aerobic conditions, 1 and 2 mg/ml concentrations of aloin efficiently inhibited L. acidophilus growth (**Figure 1C**). The antimicrobial effect of these two concentrations

FIGURE 1 | Antimicrobial property of aloin on in vitro cultured bacterial species. All bacterial cultures were incubated with two-fold serial dilution of aloin starting a concentration of 2 mg/ml to 0.03 µg/ml (A–D) or 4g/ml to 0.015 µg/ml (E–I). The data presented here were average of five independent experiments. Values of p < 0.05 or below indicated as a statistical significance compared to control group or among the experimental groups. Asterisks indicates statistical significance ( <sup>∗</sup>p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001). (A) E. coli J53 was grown under aerobic conditions. (B,C) L. acidophilus was grown under aerobic conditions or anaerobic conditions, respectively. Aloin was unable to kill L. acidophilus under anaerobic conditions. (D) B. longum was grown under anaerobic conditions. The figure shows that only higher concentration of aloin was able to decrease Bifidobacterium growth under anaerobic conditions. (E,F) E. faecium was grown under aerobic or anaerobic conditions, respectively. The figure shows that aloin was able to decrease E. faecium growth under aerobic conditions even at 4 mg/ml concentration but unable to prevent bacterial growth, however, aloin was able to kill E. faecium growth under anaerobic conditions. (G) B. thetaiotaomicron was grown under anaerobic conditions. The figure shows that aloin was able to kill B. thetaiotaomicron growth under anaerobic condition and the MIC value was 1 mg/ml. (H) A. muciniphila was grown under anaerobic conditions. The figure shows that aloin decreases A. muciniphila growth under anaerobic conditions. (I) Eubacteria sp. was grown under anaerobic conditions. The figure shows that aloin decreases Eubacterium sp. growth under anaerobic conditions.

were statistically significant (p < 0.05) compared to lower concentrations (0.015 µg/ml to 0.5 mg/ml). The growth curve revealed an initial 11 h lag phase for samples exposed to all concentrations of aloin. L. acidophilus grew faster after the lag phase and its exponential growth phase was 3–4 h, followed by a stationary phase. Under anaerobic conditions, the lag and log phase times were similar for aerobic incubations; however, under anaerobic conditions, aloin did not have a similar antimicrobial effect on L. acidophilus (**Figure 1D**). Wells treated with higher concentrations of aloin (above 2 mg/ml) also had a similar growth curve to that observed in wells treated with lower concentrations. These data clearly suggest differential antibacterial properties of aloin under aerobic versus anaerobic culture conditions. L. acidophilus grown aerobically had an MIC value of 1 mg/ml.

In contrast, the result of aerobic growth data of E. faecium in the presence of aloin was significantly different than L. acidophilus. Specifically, the log phase of E. faecium varied between the lower and higher concentrations of aloin (**Figure 1E**). At lower concentrations (125–250 µg/ml) the lag phase was approximately 2 h and the stationary phase was within the 6 h time point. During exposure at higher concentrations (500 µg/ml to 4 mg/ml), the exponential growth starting time was delayed by 3–4 h and bacterial growth remained in the exponential phase until 20 h with a slower growth rate. E. faecium exposed to 1, 2, and 4 mg/ml had a statistical significant difference (p < 0.001) in a dose-dependent manner compared to groups treated with low concentrations of aloin. However, in 1 mg/ml treated samples, statistically significance was observed until 7 h, due to increased bacterial growth, but no significance difference was observed after the 7 h time point. Cultures treated with 2 and 4 mg/ml of aloin had lag phase for 8 h, after which, bacterial growth started slowly; hence, no MIC value was recorded for E. faecium under aerobic conditions. Under anaerobic conditions, the growth curve data revealed that lag phase was dose dependent. At low concentration (0.125 mg/ml), lag phase was 2 h, but at a higher concentration (1 mg/ml) the lag

phase was 7 h (**Figure 1F**). The higher concentration (2 mg/ml) of aloin completely inhibited the bacterial growth. In 1 mg/ml exposed wells, the log phase was delayed more than 5–9 h and reached the stationary phase after 14 h. The MIC value of aloin for E. faecium fell between 2 and 4 mg/ml concentrations.

We also examined the effect of aloin on Bacteriodes growth. The growth curve data indicates that MIC concentration of aloin for Bacteriodes was 2 mg/ml; this concentration completely inhibited the bacterial growth and was statistically significant (p < 0.001) when compared to the effects of lower aloin concentrations (**Figure 1G**). In Bacteroides cultures that were exposed to 1 mg/ml of aloin, the lag phase was approximately 11 h, then the log phase began and continued for 24 h. For the 500 µg/ml treated cultures, the lag phase was 5 h, then log phase continued for 17 h and was statistically significantly different from the effects of lower aloin concentrations. After 18 h, it reached stationary phase and growth was similar to that of the other lower aloin concentrations.

Akkermansia muciniphila is a Gram-negative anaerobic bacterium that belongs to phylum Verrucomicrobia, which colonizes in the intestine and accounts for approximately 1–5% of commensal bacteria in the colon. In healthy individuals, its main function is degrading the mucin and producing several metabolic products (Derrien et al., 2008). We tested the antimicrobial effect of aloin on A. muciniphila. The bacterial growth curve showed a dose-dependent growth inhibition. The growth curve data also revealed that the lag phase was dose-dependent and the highest two concentrations were statistically different (p < 0.05) compared to lower concentrations (**Figure 1H**). The MIC value of aloin for A. muciniphila was 2 mg/ml and the bacterial growth was completely inhibited.

The growth curve data of Eubacterium sp. revealed that during the initial 5 or 6 h time period, bacterial growth rate was very slow in wells incubated with 2 and 4 mg/ml aloin concentrations, which was statistically significant (p < 0.05) compared to those in wells treated with lower aloin concentrations. Eubacterium spp. ability to metabolize aloin slowly (after 5 h) could result in accumulation of glucose, which bacteria use as a nutrient for growth and multiplication. The bacterial growth of 4 and 2 mg/ml aloin treated wells was higher than in lower aloin concentrations after 13–14 h exposure and continued the log phase up to 20 h (**Figure 1I**). The MIC values were unable to be determined likely because aloin was metabolized as noted previously (Che et al., 1991; Pogribna et al., 2008). We confirmed this observation by analyzing metabolism of aloin in Eubacterium cultures by HPLC (Lobbens et al., 2016) (**Figure 2A**). The HPLC analysis revealed disappearance of aloin-A and aloin-B (retention time was 10.30 and 8.50 min, respectively) and appearance of aloeemodin (retention time was 22.1 min), which provides evidence that Eubacterium sp. metabolized aloin (**Figure 2B**).

#### Effect of Aloin on Total Aerobic Bacterial Population in Rat Fecal Samples

Bacterial counts revealed that at the 3 h time point there was not a significant difference in total aerobic bacterial counts between the control group and samples treated with aloin

(**Figure 3A**). Live total aerobic bacterial count obtained from rat fecal slurry incubations was similar to human commensal pure culture aerobic bacterial growth curve data, in which, most of the aerobic bacterial growth showed initial 3 h lag phase except E. faecium. In E. faecium, the growth curve showed that lag phase was 1.5 h. This can be correlated to differential effect of aloin on commensal bacteria, as we observed in the growth curve. Samples collected at the 6 h time point showed a slight increase in bacterial populations in control samples and samples incubated with 0.5 mg aloin. In contrast, aerobic bacterial count decreased in the 1 and 2 mg/ml aloin treated samples, but was statistically insignificant. This result further supports the in vitro bacterial growth curve results. For example, the B. longum growth curve indicated that aloin (2 mg/ml) decreased the bacterial growth by 10% and in E. faecium growth was inhibited at a higher concentration. The samples collected at 24 h had similar bacterial counts as that of 6 h control groups. In contrast, the bacterial count of aloin treated samples decreased and the effect was significant (p < 0.05) in the 2 mg/ml treated group as compared to the 0.5 and 1 mg/ml treated samples (**Figure 3A**). Overall, the bacterial count revealed that 2 mg/ml

aloin treated samples experienced a time and dose dependent growth inhibition of aerobic bacteria. This observation was consistent with in vitro bacterial growth culture results, which suggested higher concentrations had antimicrobial property.

# Effect of Aloin on Total Anaerobic Bacterial Populations in Rat Fecal Samples

Anaerobic bacterial count data suggests that at the 3 h time point, 0.5 and 1 mg/ml aloin treated samples had low bacterial counts compared to control groups (**Figure 3B**). In contrast, 2 mg/ml treated samples had higher bacterial counts than other groups (p < 0.05).

# Effect of Aloin on Lactobacillus in Rat Fecal Samples

At the 3 h time point, samples treated with 2 mg/ml aloin had slightly higher bacterial counts compared to other groups, which was statically significant (p < 0.01) (**Figure 3C**). In contrast, at the 24 h time point live bacterial counts indicated that all experimental groups had increased bacterial population; however, only 0.5 and 2 mg/ml treated samples were statistically significant.

# Effect of Aloin on Bifidobacterium in Rat Fecal Samples

Bifidobacterium bacterial culture dosed with aloin revealed that there was not a significant change in bacterial counts in samples that were collected at 3 h. Samples that were collected at the 6 h time point had slight increase in bacterial count in the 0.5 and 1 mg/ml aloin treatments. Samples treated with 2 mg/ml aloin had less bacterial counts, but were statistically insignificant (**Figure 3D**). Samples that were collected at the 24 h time point had higher bacterial counts compared to control groups, but were found to be statistically insignificant due to individual variation.

# The Effect of Aloin on the Production of Short Chain Fatty Acids

Intestinal bacteria metabolize the undigested food materials and produce SCFAs (Wong et al., 2006). Hence, we analyzed acetic, butyric, succinic, lactic, propionic, isobutyric, valeric, hexnoic, and isovaleric acid SCFAs production from the fecal slurry treated with aloin. Importantly, butyrate has been known to play a central role in the homeostasis of intestinal epithelial cells (Kelly et al., 2015). Our results suggest that butyrate production decreased due to aloin exposure (**Figure 4A**); however, we could not observe any statistical significance for the other SCFAs [acetic acid and succinic acid (**Figures 4B,C**)]. The result revealed that at the 3 and 6 h time points, there was no change in the butyrate production either in the control or in experimental groups. In contrast, the 24 h samples revealed a decrease in butyrate production in a dose dependent manner. Specifically, 1 and 2 mg/ml aloin treated samples showed statistical significant as compared to 0.5 mg aloin treated samples (**Figure 4A**).

# The Effect of Aloin on Bacterial Phyla as Assessed by 16s rRNA Sequencing

The human intestinal bacterial growth curve data revealed that antibacterial property of aloin varies between aerobic and

measured by HPLC method. The presented data are averages of four independent experiments. (A) Butyrate production, (B) acetic acid production, and (C) succinate production. Values of P < 0.05 or below considered as statistically significant. Asterisks indicates statistical significance (∗p < 0.05).

anaerobic conditions for the same bacterial species (for example Lactobacillus and Enterococcus). For global analysis of the bacterial populations, 16s rRNA sequencing was conducted to evaluate antimicrobial effect of aloin in rat intestinal commensal bacteria. The phyla level analysis revealed that Firmicutes, Bacteroidetes, and Proteobacteria were the major phyla that contributed more than 95% of the bacterial population in all samples (**Figure 5A** upper panel). In addition, the analysis also revealed the contribution of Verrucomicrobia and Actinobacteria to a small percentage. Evaluation of abundance changes at the phyla level for controls and aloin treated samples suggested variations between treatment groups and exposure times. Some major differences in Firmicutes, Bacteroidetes, and Actinobacteria phyla were observed in aloin treated samples (**Figure 5** lower panel). Specifically, the abundance of these three bacterial phyla was significant changes in samples collected at the 24 h time point for all three aloin concentrations (0.5, 1, and 2 mg/ml). The abundance of phylum Firmicutes decreased as the aloin concentration increased with time (6 and 24 h). An interesting observation was that the abundance of Actinobacteria was more prominent at 24 h time point than at 3 and 6 h time points. This phylum was found more abundant in aloin treated samples than the control (see **Figure 5A** upper panel).

#### The Effect of Aloin on Bacterial Genus Level as Assessed by 16s rRNA Sequencing

Next, we assessed similarity/differences among aloin treated experimental groups using Orthogonal Partial Least Squares Discriminant Analyses (PLS-DA) (Rabbi et al., 2016). The principal component analysis revealed that samples collected at 3 and 6 h (0.5, 1, and 2 mg/ml) were more tightly grouped together (**Figure 6A**). In contrast, samples collected at the 24 h time point (0.5, 1, and 2 mg/ml) separated from the other time points and distributed in wider range. Next, we analyzed abundance of bacterial genera by one-way ANOVA, which revealed marked differences among nine genera. The analysis provided the trend of bacterial abundance on Bacteriodetes, Afipia, Marivita, Turicibacter, Microcystis, Haemophilus, Sneathia, Lactobacillus, and Alkalibacter in samples collected at the 24 h time point for all three concentrations (0.5, 1, and 2 mg/ml) were compared to samples collected at the 3 and 6 h time points (**Figure 6B**). Among 24 h samples, the control had a higher abundance of these bacterial genera than experimental groups for all three concentrations. More specifically, bacterial abundance in samples exposed to higher concentration of aloin (2 mg/ml) was affected the most. In contrast, the genus Alkalibacter was abundant in samples collected at 3 and 6 h time points for all three aloin concentrations. These results were consistent with bacterial growth curve data (for example L. acidophilus) generated in the present study. Subsequently, heat map was generated to define beta diversity (**Figures 6C,D**) that reveals overall genus abundance in the experimental groups, and abundance of top 50 genera, respectively.

#### The Effect of Aloin on Bacterial Species Level as Assessed by 16s rRNA Sequencing

Next, analysis was focused on the bacterial species level to see whether there was any separation between experimental groups regarding specific bacterial species. The PLS-DA analysis displayed the differences in species level in samples that were exposed to aloin. The species separation was similar to genus level. Specifically, samples that were collected at the 24 h time point (all three concentrations) grouped together. In contrast, samples that were collected at 3 and 6 h grouped together and separated from 24 h time point samples (**Figure 7A**). We

also observed a clear difference at the species level for the experimental groups.

To provide a clear picture on the effect of aloin at species level, the heat map was generated to provide beta diversity in the experimental groups (**Figure 7B**). The heat map provides a snapshot of bacterial abundance at the species level and the difference within an experimental group and among the experimental groups. Here, we provided heat map data for the top 50 bacterial species that were affected by aloin treatment. The most significant representative bacterial species, in which aloin treatments had no adverse effect on Lactobacillus rhamnosus, Lactobacillus similis, Lactobacillus Intestinalis, and L. reuteri species, which showed abundance in the 24 h samples (**Figure 7C**) for all three aloin concentrations. Interestingly, aloin treatment increased the abundance of these bacterial species as compared to control. The 16s rRNA sequencing result is consistent with L. acidophilus growth curve data and live bacterial counts, where aloin either lacks microbicidal property or increases the bacterial growth. Similarly, Bifidobacterium choerinum and B. thermophilum also increased due to the aloin treatment at 24 h samples. In contrast, Clostridium indolis, Clostridium sulfatireducens, Bacteroides xylanolytics, and Alkalibacter saccharofermentans species decreased in abundance in samples collected at the 24 h time point (**Figure 7C**) for all three aloin concentrations. These bacterial species were abundant in samples collected at the 3 and 6 h time points for all three concentrations. We also observed that in live bacterial counts and bacterial growth curve, aloin had antimicrobial property at higher concentration, as well as during prolonged exposure (**Figures 1C,F**, **3A**). The sequencing data very well support the in vitro experimental data of the present study.

# Effect of Aloin on Transepithelial Resistance

The cell cytotoxicity study revealed that aloin at higher concentrations induced cell death. Next, we tested whether low concentrations of aloin can have impact on the permeability of intestinal epithelial cells or not? In this experiment, we dissolved aloin at two different pH solutions (pH 4.9 and 7.2) and tested the intestinal barrier integrity to mimic the intestinal pH. When cells were exposed with aloin (0.05–500 µM) that was dissolved in pH 7.2 buffer, the transepithelial resistance increased similarly to the control indicating the barrier integrity remains intact (**Figure 8A**). In contrast, when cells were exposed to aloin at (0.05–500 µM) dissolved in pH 4.9 (then diluted into cell culture media so final pH was equal to culture media) the TER value decreased in a dose-dependent manner

indicating compromised barrier integrity (**Figure 8B**). As aloin is metabolized by intestinal bacteria into aloe-emodin, we also tested the effect of aloe-emodin on intestinal cell integrity. Aloeemodin was dissolved in ethanol and diluted into culture media (pH 7.2). The results revealed that aloe-emodin did not decrease the TER value indicating that intestinal barrier integrity remained intact (**Figure 8C**).

#### DISCUSSION

Pharmacological and phytochemical molecules extracted from Aloe Vera plants exhibit several beneficial properties (Briggs, 1995; Moore and Cowman, 2008). However, several studies also report adverse effects of aloin, for example, one human case study showed that oral consumption of 500 mg capsule of aloe extract for 4 weeks by a 56 years old woman resulted in acute hepatotoxicity; however, upon discontinuation of aloe capsules a rapid improvement from liver damage was noticed (Rabe et al., 2005). In another case study, a 24 years young adult consumed 500 mg of Aloe vera capsules for 3 weeks; liver biopsy from this individual revealed liver toxicity that was similar to drug induced liver toxicity as well as other clinical abnormalities (Kanat et al., 2006). Several other studies also provide evidence that human oral consumption of aloe vera resulted in the hepatotoxicity (Bottenberg et al., 2007; Yang et al., 2010; Lee et al., 2014).

Orally consumed aloe products reach the intestine where they interact with intestinal epithelial cells and inhabitant intestinal microbiota. Earlier studies have shown that aloin exhibits antimicrobial properties against pathogenic bacteria (Asamenew et al., 2011; Abeje et al., 2014); however, limited information is available on aloin antimicrobial properties on human intestinal commensal microbiota. The gut-liver axis with a focus on commensal microbiota, as well as, microbiotaderived factors has emerged as key player in the hepatic as well as intestinal diseases (Adolph et al., 2018). A balanced intestinal microbiota is essential for intestinal homeostasis

treated experimental groups. (D) The right panel (heat map) shows the top 50 bacterial genera that altered (either increased or decreased) due to aloin treatment.

and overall health. Thus, there is a need to address the knowledge gap between interactions of aloe derived products with effects on the intestine commensal microbiota. In this study, we addressed whether aloin has antimicrobial activity against intestinal microbiome. The experimental approaches used in this study were fivefold: (1) evaluate aloin antimicrobial property against selected intestinal commensal bacteria that are known to positively contribute to human health; (2) assess the effect of aloin on rat fecal samples that mimic the bacterial composition of human intestine under anaerobic condition to delineate the effect on intestinal commensal bacterial community; (3) to determine the effect of aloin on SCFAs production by intestinal commensal bacteria; (4) to assess if aloin changes the intestinal epithelial cell permeability and (5) calculate metabolism of aloin by intestinal commensal bacteria.

The human intestine is a residence for 35,000 bacterial species (Frank et al., 2007) that belong to 50 bacterial phyla (Schloss et al., 2005). Among them Bacteroidetes and Firmicutes are two major phyla that account for 90–95%, of the bacterial population (Gordon and Dubos, 1970). In addition, Proteobacteria, Verrucomicrobia, Actinobacteria, and other phyla also contribute to various functions in the intestine (Eckburg et al., 2005). In the present study, the 16S rRNA sequencing data indicates that Bacteroidetes, Firmicutes, and Proteobacteria contribute 90%, while Verrucomicrobia and Actinobacteria contribute for the remaining phyla found in rat fecal samples (**Figure 3**), which aligns with earlier reports (Fujio-Vejar et al., 2017). Commensal bacterial species are distributed differentially and colonized throughout the intestine. Rat bacterial phyla are more closely related to human than other murine species (Li et al., 2017). In this study, we tested aloin antibacterial effects against seven bacterial species (one pathogenic and six commensal bacteria) that belongs to five dominant bacterial phyla of human microbiota (**Table 1**). For the in vitro bacterial growth curve study, we used representative Gram-positive and Gram-negative commensal bacteria. The growth conditions tested of these bacterial species ranged from strict anaerobic to facultative anaerobic (**Table 1**).

Earlier, it was reported that aloin exhibits antibacterial activity toward Gram-negative bacteria (Asamenew et al., 2011; Mariappan and Shanthi, 2012), but is incapable of killing Grampositive strain, such as Bacillus pumilus and Bacillus subtilis (Megeressa et al., 2015). In the present study, the in vitro bacterial growth curve shows that aloin also exhibits antimicrobial properties toward Gram-positive intestinal commensal bacteria, specifically E. faecium and L. acidophilus. However, aloin had limited antibacterial activity toward B. longum, another member of Gram-positive bacteria. Several intestinal bacteria are important for vitamin synthesis and degradation of undigested food materials (Wong et al., 2006). Earlier studies have shown that Eubacterium sp. is capable of degrading aloin into aloe-emodin and glucose, which is then utilized as an energy source for bacterial multiplication (Pogribna et al., 2008). Our results are consistent with an earlier report that Eubacterium spp. multiplication was directionally proportional to the concentration of aloin in the culture media (Che et al., 1991). Using non-animal model we also confirmed previous reports that aloin is metabolized into aloe-emodin or aloesin by intestinal bacteria (Che et al., 1991). These metabolites are structurally similar to anthraquinones and chromones, which exhibit antibacterial activity against Grampositive and Gram-negative pathogenic bacteria (Budzisz et al., 2001; Cock, 2007). Aloin and its metabolites may function similar to anthraquinones or chromones toward intestinal

commensal bacteria. The growth of L. acidophilus, E. faecium, and B. thetaiotaomicron were completely inhibited at 1 mg/ml concentration (**Figures 1C,E,G**), which indicates that aloin has similar levels of antibacterial activities to both Gram-positive and Gram-negative bacteria.

The molecular mechanism of antibacterial activity of aloin or its metabolites and anthraquinones is not well understood. However, several mechanisms have been put forth for the antimicrobial properties of aloin or its metabolites including the inhibition of membrane and respiration transport (Ubbink-Kok et al., 1986; Hamman, 2008). Aloin and its metabolites contain phenolic structures that are considered bioactive molecules, which may exhibit antimicrobial property against commensal bacteria. A recent study provided underlying molecular mechanisms of emodin's (metabolites of aloin) antimicrobial property. Emodin, an anthraquinone, interacts with bacterial cell wall proteins, thereby increasing the permeability due to alteration in cell wall structural integrity. This causes an outflow of intracellular contents that may result in bacterial death (Li et al., 2016). Emodin causes bacteriostatic effects by inhibiting bacterial growth or multiplications at low concentrations (16 and 32 µg/ml) and causes the bactericidal effect at higher concentration (64 µg/ml) (Li et al., 2016). The growth kinetic results of the present study also provide evidence that aloin at low concentrations causes the bacteriostatic effect on L. acidophilus

was more at species level than genera level. (B) Heat map shows the diversity among aloin treated experimental groups. The heat map shows the top 50 bacterial species that altered (either increased or decreased) due to aloin treatment. Here we have shown with red arrow mark of few bacterial species, which were decreased due to aloin treatment. These species are major contributors for butyrate production in the intestine. (C) These figures show that aloin exposure alters the bacterial community increased or decreased abundance at species level. Here we show representative examples for high as well as low abundance on bacterial species due to aloin treatment.


TABLE 1 | Aloin MIC value for intestinal commensal both Gram-positive and Gram-negative bacteria.

(aerobic condition), E. faecium, and B. thetaiotaomicron and bactericidal effect at higher concentrations. Nevertheless, the antibacterial property of aloin is species specific and is dependent on environmental factors such as aerobic and anaerobic growth conditions.

In the present study, the L. acidophilus grown in the presence of 1 mg/ml concentration of aloin experienced growth inhibition only under aerobic condition, whereas 2 mg/ml and above caused a bactericidal effect (**Supplementary Figure S1**). Interestingly, at low aloin concentrations, L. acidophilus appeared as a long chain of bacteria. This "chain-like" appearance could be correlated to bacterial aggregation caused by aloin or newly dividing bacteria unable to detach from each other in the presence of low concentrations of aloin (0.5 and 1 mg/ml). In contrast, bacteria exposed to a higher concentration of aloin lost their rod-shaped morphology and appeared as a "bristle-like" structure. We also examined the bactericidal mechanism of aloin by staining L. acidophilus with acridine orange and ethidium bromide after the completion of the growth curve experiment. Wells exposed to 0.5 mg/ml aloin revealed that all of L. acidophilus was stained green, whereas orange-stained bacteria were not observable. Acridine orange permeates intact cell membrane of rod shape bacteria, intercalates with DNA and fluoresces green in color, which is indicative of live L. acidophilus. Additionally, all bacteria retained rod shape morphology. This result suggests that low aloin concentration (0.5 mg/ml) is unable to induce killing mechanism or cytotoxicity. Wells exposed to 1 mg/ml showed 50% of bacteria with green and the remaining red in color. Bacteria red in color indicate that the ethidium bromide permeates bacteria that lost membrane integrity or in the process. In contrast, wells exposed with 2 mg/ml contained bacteria that were all red in color. Likely, they lost their cell wall structural integrity and formed bristle-like structures (**Supplementary Figure S1**) indicating that higher aloin concentration caused bactericidal effects. The possible mechanism of bactericidal effects that aloin may cause on bacterial membranes is very similar to emodin (Li et al., 2016).

Rat fecal samples that were exposed to various concentrations of aloin experienced changes in intestinal bacterial population. Fecal samples exposed to low aloin concentrations (0.5 and 1 mg/ml) underwent minor perturbation at the 3 and 6 h time points, whereas at 24 h bacterial colony forming units (CFU) were either increased or decreased depending upon the bacterial species. For example, under anaerobic growth conditions Lactobacillus CFU increased at 24 h time point, but not at the 3 or 6 h time point. These data are consistent with the in vitro L. acidophilus anaerobic growth curve result, in which we observed an initial 10 h lag phase following by an exponential growth (**Figure 3C**). The 16S rRNA sequencing data revealed the abundance of Lactobacillus species (**Figure 6C**) and supports the in vitro bacterial growth curve data and fecal live bacterial growth curve data. In the present study, we demonstrated that aloin exhibits antibacterial activity for some of the intestinal commensal bacteria, and the MIC value ranged from 1 mg to 4 mg/ml. Concentrated aloe gel or powdered form of leaf extracts are available in the market as a capsule containing 100–500 mg dose. Whereas, liquid consumption results in exposure to approximately 14.4 g whole leaf extract. In these consumer use products, aloin, levels range from 0.1 to 6.6% of leaf dry weight (Groom and Reynolds, 1987); thus indicating presence of aloin in a concentration range between 0.1 and 35 mg. However, to our knowledge, data is lacking on what concentration levels reach the GIT.

Next, we analyzed the effect of aloin on SCFAs production by intestinal commensal bacteria, which is an essential energy source for intestinal epithelial cells and gut homeostasis. The HPLC data provides evidence that microbial derived butyrate production had decreased at the 24 h time point in fecal samples treated with aloin. The sequencing data further confirms a decrease abundance of butyrate producing bacterial species (Clostridium spp., Roseburia spp., Coprococcus spp., and Eubacterium spp.) at 24 h time point, (**Figure 6C**). The depletion of bacterial species occurs more readily in aloin treated samples than control samples. A recent study suggests that streptomycin exposure in mice reduced butyrate producing Clostridia in the intestine within a day, which resulted in decreasing butyrate concentration by fourfold in cecum and facilitated the expansion of aerobic pathogenic S. Typhimurium (Rivera-Chavez et al., 2016). In the intestine Clostridia spp. contributes a substantial amount of butyrate (Louis and Flint, 2009; Vital et al., 2014), which is consumed by colonocytes to create a hypoxic environment (Kelly et al., 2015) in the intestine by converting butyrate into CO<sup>2</sup> (Donohoe et al., 2012). Change in the intestinal bacterial community structure has been strongly correlated to colonic disease and irritable bowel syndrome (IBS) primarily due to a decrease in the

FIGURE 8 | The effect of aloin on the permeability of human intestinal epithelial cells. T84 cells were grown in transwells till a TER value of 1,000 or above was reached. The TER value was recorded before and after incubation with five concentrations of aloin (0.05, 0.5, 5, 50, and 500 µM) at 0, 1, 2, 3, 24, and 48 h. The TER values recorded before adding aloin served as a baseline value. Data are presented here as % increase or decrease from the baseline readings. These data represent an average of seven independent experiments. The asterisk <sup>∗</sup> indicates statistical significance. (A) This graph shows the TER value for aloin dissolved in pH 7.2. (B) It shows the TER value for aloin dissolved in pH 4.9 citrate buffer and further diluted in cell culture media (antibiotic free media). (C) TER value for aloe-emodin. Asterisks indicates statistical significance (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001).

production of butyrate (Sokol et al., 2009; Eeckhaut et al., 2013). In ulcerative colitis patients with a reduced butyrate production, the depletion of Clostridium coccoides and Clostridium leptum was a probable contributing factor for etiology (Kumari et al., 2013). In addition, intestinal microbial derived butyrate increases the mitochondrial dependent oxygen consummation and inhibits the proinflammatory mediator expression through histone deaceytlase (Davie, 2003; Furusawa et al., 2013; Zheng et al., 2017). These studies suggest the importance of intestinal microbial derived butyrate on intestinal homeostasis. In the present study, decreased butyrate production could be correlated to the decreased abundance of butyrate-producing bacterial species. We show that the high concentration and long-term exposure of aloin decreased the butyrate production in rat fecal samples. The decreased butyrate production by intestinal commensal bacteria can be correlated to development of intestinal abnormalities reported in animal studies (Boudreau et al., 2013, 2017). The sequencing data complements SCFAs production and provides evidence for the antibacterial property of aloin on SCFA producing commensal bacteria. Furthermore, butyrate provides an energy source for intestinal epithelial cells and promotes epithelial barrier formation by decreasing the expression of pore forming claudin-2 genes (Zheng et al., 2017) and suppressing the colonic inflammation and carcinogenesis through the activation of GPA109a receptor (Singh et al., 2014).

Intestinal transepithelial resistance experiments revealed that aloin decreases the permeability barrier function only at high concentrations. Aloin dissolved in pH 4.9 decreased TER value indicating that barrier function is compromised within 2 h of exposure. It has been shown that aloin is more stable in acidic pH (pH 2.0) for a longer period compared to in basic pH (pH 8.0) (Ding et al., 2014). Aloin dissolved at pH 4.9 may be more stable than at neutral pH 7.2. The compromised barrier function may be associated with the stability of aloin at an acidic pH 4.9. Aloe-emodin treated cells maintained the intestinal barrier integrity. The focus was to evaluate the effect of aloin on intestinal epithelial cells; hence, we have not tested its properties by dissolving aloe-emodin in acidic pH 4.9. Aloin cytotoxicity was also assessed using polarized intestinal epithelial cells. Polarized intestinal epithelial cells were exposed to the same concentrations that were used in intestinal commensal bacteria. To differentiate live and dead cells, acridine orange and ethidium bromide were added as described earlier (Williams et al., 2015). The microscopic examination reveals that 2 mg/ml (4.78 mM concentration) aloin exposed cells were stained 50% green in color and remaining cells red color. In the TER experiment, 500 µM aloin compromised the barrier function. Thus, a greater concentration of aloin (i.e., 2 mg/ml or 4.87 mM) will have a greater detrimental effect on epithelial cells and may result in cytotoxicity. Earlier, it has been shown that Jarkat T-lymphocytes exposed to aloin caused several abnormalities including altered cellular morphology, cell cycle arrest, loss of membrane integrity, and induced cytotoxicity by apoptotic mechanism (Buenz, 2008). The present study also demonstrated that aloin induced cytotoxicity in intestinal epithelial cells required at least two times higher

concentration than the effect observed in Jarkat T-lymphocytes. The possible mechanism may involve intestinal epithelial cells secreting mucin, which could serve as a protective layer and prevent immediate interaction with epithelial cells. In contrast, T-lymphocytes lack a mucus layer, hence, aloin may interact with T-cells more quickly and require a smaller aloin concentration to induce cytotoxicity.

Bifidobacterium contributes several beneficial functions to the intestine. To date, 48 species have been recognized in this genus. Genome sequence analysis reveals presence of genes that can encode cell surface macromolecule proteins, which possibly play a role in bacterial attachment and colonization in the intestine (Dethlefsen et al., 2008). Bifidobacterium constantly encounters oxidative stress, and exposure to free radicals, various intestinal enzymes, and bile acids that can have a detrimental effect on bacterial survival and intestinal mucosa attachment. In addition, consumption of antibiotics and xenobiotic compounds has a negative impact on colonization and survival of Bifidobacterium. The present study provides evidence for the antimicrobial property of aloin to commensal bacteria that were cultured under anaerobic conditions. Specifically, pure in vitro cultured Bifidobacterium showed decreased bacterial growth or CFU in a dose- dependent manner. Aloin may bind to bacterial cell wall and cause the bactericidal effect by altering the bacterial membrane structural integrity; that could result in dysbiosis state. Due to this dysbiosis intestinal mucosa could become more prone to aloin mediated injury. This could probably be also a reason for the development of intestinal lesions and goblet cell hyperplasia in ascending colon in rats exposed to aloe whole leaf extracts or purified aloin compound in earlier animal studies (Boudreau et al., 2013, 2017).

In animal studies, the concentration of aloin available to interact with intestinal microbiota is very difficult to determine, as the oral exposure is usually through drinking water, where the aloin consumption and availability in the intestine to interact with intestinal microbiome could vary in each animal. Moreover, information is required for the SCFA metabolism to correlate active microbiome and the healthy gut. The non-animal models used in the current study (specific bacteria-aloin interaction and fecal slurry-aloin interaction model) address these knowledge gaps. Furthermore, using in vitro cultured human intestinal epithelial cells, we also showed a direct effect of aloin on the gut permeability. This study provides clear evidence that due to aloin (1 mg/ml and above) treatment several bacterial species show a low abundance that are involved in the butyrate production. Furthermore, we show that the butyrate production was decreased due to aloin treatment. Thus, the microbial 16s sequencing data complements the biochemical data (butyrate production) during the aloin treatment. Overall our study shows that aloin possibly may cause toxic effect; however, it depends upon aloin concentration.

#### CONCLUSION

The present study provides evidence that aloin exhibits antibacterial properties toward intestinal commensal bacteria, depending upon the growth conditions and concentration of aloin. In addition, aloin exposure decreased butyrate production by decreasing the abundance of butyrate producing bacterial species. The quantification of butyrate by HPLC analytical method confirmed decreased level of aloin exposure for 24 h. The sequencing data further supports microbiological and biochemical data observed in the present study. Transepithelial resistance results provide evidence that aloin dissolved at pH 4.9 compromised the intestinal barrier function in a dose dependent manner.

# AUTHOR CONTRIBUTIONS

KG, CC, and SK conceived and designed the experiments. KG, PK, and SK performed the experiments. KG and SK analyzed the data and contributed reagents, materials, and analysis tools. KG, CC, and SK wrote, reviewed, and approved the manuscript.

# FUNDING

This study was funded by the National Toxicology Program under an Interagency Agreement between FDA and NIEHS (FDA IAG # 224-17-0502 and NIH IAG #AES12013).

### ACKNOWLEDGMENTS

The authors would like to acknowledge Microbiology Surveillance Staff at NCTR for providing fecal samples from post-mortem animals. The authors would like to thank Dr. Jyotshnabala Kanungo, Dr. Rajesh Nayak, Ms. Sarah Orr, and Ms. Abhilasha Gokulan for reviewing the manuscript and providing valuable comments and suggestions. PK was a participant of Oak Ridge Institute for Science and Education (ORISE) Summer Student Research program at NCTR.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.00474/full#supplementary-material

FIGURE S1 | Effect of aloin on the morphology and survival of Lactobacillus acidophilus after incubation aloin treatment. The figures shown are the picture taken from the wells treated without aloin (A) or with aloin [B–D (0.5, 1, and 2 mg/ml)]. Control wells had individual rod shape bacteria, in contrast, wells treated with 0.5 mg and 1 mg/ml had long chain like appearance due to bacterial attachment or unseparated multiplying bacteria. At higher concentration of aloin treatment the bacteria lost its morphology and cellular content and appeared as a bristle like structure. The bottom panels (E–G) was stained with acridine orange and ethidium bromide for 15 min before images were captured. Bacteria stained green in color were live, whereas dead bacteria stained orange. Wells with 0.5 mg/ml of aloin showed all bacteria stained green, indicating the live bacteria; whereas, 2 mg/ml treated wells had bristle like appearance and stained orange indicating dead bacteria with change in morphology.

### REFERENCES

fmicb-10-00474 March 23, 2019 Time: 17:20 # 18


in bovine milk and feces by a combination of immunomagnetic bead separation-conventional PCR and real-time PCR. J. Clin. Microbiol. 42, 1075–1081. doi: 10.1128/JCM.42.3.1075-1081.2004


Am. J. Physiol. 276(5 Pt 1), G1279–G1288. doi: 10.1152/ajpgi.1999.276.5. G1279

Zheng, L., Kelly, C. J., Battista, K. D., Schaefer, R., Lanis, J. M., Alexeev, E. E., et al. (2017). Microbial-derived butyrate promotes epithelial barrier function through IL-10 receptor-dependent repression of Claudin-2. J. Immunol. 199, 2976–2984. doi: 10.4049/jimmunol.1700105

**Disclaimer:** The findings and conclusions presented in this manuscript are those of the authors and do not necessarily represent the views of the United States Food and Drug Administration.

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Gokulan, Kolluru, Cerniglia and Khare. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Chlorogenic Acid Ameliorates Colitis and Alters Colonic Microbiota in a Mouse Model of Dextran Sulfate Sodium-Induced Colitis

Peng Zhang, Huanli Jiao, Chunli Wang, Yuanbang Lin and Shengyi You\*

Department of General Surgery, Tianjin Medical University General Hospital, Tianjin, China

This study evaluated the mitigating effects of dietary chlorogenic acid (CGA) on colon damage and the bacterial profile in a mouse model of dextran sulfate sodium (DSS) induced colitis. C57BL/6J mice were randomly assigned to receive one of the following treatments: (i) basal diet; (ii) basal diet with 2% CGA; (iii) basal diet with 2.5% DSS or (iv) basal diet with 2% CGA and 2.5% DSS. Following a 2-week pre-treatment period, mice in the DSS and CGA-DSS groups received 2.5% DSS in drinking water for 5 days, while the other two groups received sterile water. Compared to DSS alone, CGA was found to reduce the disease activity index, myeloperoxidase activity and tumor necrosis factor-α levels in colon tissues (P < 0.05). CGA also ameliorated DSS-induced inflammatory responses, reduced colon shortening and decreased the histological scores (P < 0.05). In an evaluation of the relative abundances of bacteria in the fecal microbiota, we found that CGA reversed the decrease in diversity caused by DSS and improved the relative abundance of organisms in the genus Lactobacillus (P < 0.05). These results indicate that CGA maintains intestinal health and reduces DSS-induced colon injury by decreasing the production of pro-inflammatory cytokines and restoring intestinal microbial diversity.

Keywords: chlorogenic acid, colitis, MPO, TNF-α, microbiota

# INTRODUCTION

Inflammatory bowel disease (IBD) involves a complex and chronic inflammatory process. Although the evidence increasingly suggests that IBD results from the exposure of a genetically susceptible host to a combination of environmental factors, the exact aetiology and pathogenesis remain unclear (Tibble et al., 2000; Zhang et al., 2014). The existing evidence indicates that IBD is caused by an abnormal mucosal immune response to intestinal microorganisms and the inappropriate secretion of cytokines in the mucosa (Cario and Podolsky, 2000; Weichselbaum and Klein, 2018). Therefore, IBD treatment primarily aims to reduce recurrent inflammation and achieve a prolonged remission (Levesque et al., 2015; Sgambato et al., 2017). Traditionally, both treatment targets have been determined by clinical symptoms rather than objective evidence of inflammatory activity. However, IBD symptoms are often a direct consequence of the inflammatory process and may differ depending on the location of inflammation (Tegtmeyer et al., 2017).

#### Edited by:

Jie Yin, Institute of Subtropical Agriculture (CAS), China

#### Reviewed by:

Wutai Guan, South China Agricultural University, China Guiping Guan, Hunan Agricultural University, China

> \*Correspondence: Shengyi You shengyiyou019@163.com

#### Specialty section:

This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology

Received: 24 January 2019 Accepted: 11 March 2019 Published: 27 March 2019

#### Citation:

Zhang P, Jiao H, Wang C, Lin Y and You S (2019) Chlorogenic Acid Ameliorates Colitis and Alters Colonic Microbiota in a Mouse Model of Dextran Sulfate Sodium-Induced Colitis. Front. Physiol. 10:325. doi: 10.3389/fphys.2019.00325

**238**

Bioactive polyphenols, which have anti-oxidant and antiinflammatory effects and can regulate cellular signaling, could potentially be used as adjuvants to the treatment of metabolic syndrome (Santino et al., 2017). Chlorogenic acid (CGA) is an important bioactive dietary phenolic substance found widely in coffee, fruits and vegetables (Naveed et al., 2018). Studies have shown that CGA has anti-oxidant, anti-bacterial and antiinflammatory effects (Zhao et al., 2008; Kamiyama et al., 2015; Tsang et al., 2016; Zheng et al., 2016). For example, CGA inhibited the production of interleukin (IL)-8 in Caco-2 human intestinal cells in response to tumor necrosis factor (TNF)-α and H2O2, while studies in C57BL/6 mice found that CGA could alleviate dextran sulfate sodium (DSS)-induced weight loss, diarrhea, fecal blood and colon shortening. These data suggest that CGA can prevent intestinal inflammation (Shin et al., 2015). Other studies of rodent models have shown that CGA can inhibit lipopolysaccharide-induced myeloperoxidase (MPO) activity in the lung (Zatorski et al., 2015).

Colon microbes play an essential role in human health and have been associated with various diseases, including irritable bowel syndrome (Kassinen et al., 2007), autism (Li and Zhou, 2016) and obesity (Zhang et al., 2009). Most ingested CGA is not absorbed in the small intestine and reaches the colon, where it is converted into several metabolites by the local microbial community. Studies have shown that Bifidobacterium animalis can hydrolyse and alter the fate of CGA and thus affect the composition of the microbiota (Tomas-Barberan et al., 2014). Another study found that a 10-h course of CGA treatment increased the numbers of Bifidobacterium spp. and members of the Clostridium coccoides–Eubacterium rectale group in vitro and promoted the expansion of select bacteria, compared with the control group (Mills et al., 2015). Although colitis has been very well studied, evidence regarding the role of CGA in the alleviation of this disease is scarce. In this study, the effects of CGA on colonic microbial composition and pro-inflammatory factors were explored in a rodent model of DSS-induced colitis. In vivo experiments were also conducted to investigate whether CGA could prevent intestinal inflammation.

#### MATERIALS AND METHODS

#### Animals and Experimental Treatments

For this study, all procedures involving animals were approved by the Animal Ethics Committee of General Hospital of Tianjin Medical University and conformed in all respects with the Guidelines for the Care and Use of Laboratory Animals of Tianjin Medical University. The female C57BL/6J mice utilized in this study were obtained from the SLAC Laboratory Animal Centre (Shanghai, China) at 6–7 weeks of age. Mice were housed in a pathogen-free colony in accordance with standard laboratory conditions, including a temperature of 22–24◦C, humidity of 40– 60% and 12-h daily light/dark cycle. After a 7-day adaptation period, 40 mice were randomly assigned to one of four discrete treatment groups: (1) basal diet (CON); (2) basal diet with 2% CGA (CGA); (3) basal diet with 2.5% DSS (DSS, MW 5000, KAYON Biotechnology Co., Ltd.) and (4) basal diet with 2% CGA and 2.5% DSS (CGA-DSS). CGA was dissolved in sterile drinking water for administration. All of the mice had access to food and water ad libitum, and the latter was changed twice weekly. During a 20-day experiment, the drinking water was supplemented or not with 2.5% DSS for the last 5 days. The body weight of each mouse was recorded at the end of the experiment.

# Sample Collection

All mice were sacrificed according to standard procedures. The colon was removed from each animal and rinsed in physiological saline to remove fecal residue. The weight and length of each organ were recorded to determine the inflammation index. Next, samples from various segments of each colon were fixed in 4% buffered formaldehyde and embedded in paraffin. Fourmicrometer-thick slices of the paraffin-embedded tissues were stained with haematoxylin and eosin (H&E) in accordance with standard procedures for the histological evaluation of colonic damage. The remaining colonic tissue samples and digests were stored in frozen liquid nitrogen for further measurements of biological parameters.

#### Colitis Disease Activity Index (DAI)

Tissue samples from the distal colon (5 µm) were stained with H&E and subjected to a microscopic analysis to determine the colon histological score, as proposed by Cooper et al. (1993). The following scoring system was used (maximum score = 10): 0 (rare) = severe inflammatory cell infiltration; 1 = marginally dispersed cell infiltrate; 2 = moderately increased cell infiltrate with the formation of occasional cell foci and 3 = large areas of cell infiltration causing a severe loss of tissue architecture. Additionally, the following scoring system was used to determine the extent of injury: 0 = none; 1 = mucosal; 2 = mucosal and submucosal, and 3 = transmural. Crypt damage was scored as follows: 0 = intact crypts; 1 = damage to the basal one-third; 2 = damage to the basal two-thirds damaged, 3 = only surface epithelium remains intact, and 4 = loss of entire crypt and epithelium (Backer et al., 2017).

#### Assessment of Leukocyte Involvement

Myeloperoxidase activity, a marker of neutrophilic infiltration, was assessed using the method described by Grisham et al. (1990) with slight modifications. A distal colon sample from each mouse was excised, immediately rinsed with ice-cold saline, blotted dry and frozen at −70◦C. Subsequently, the tissue samples were thawed, weighed and homogenized in 10 volumes of 50 mM phosphate-buffered saline (PBS; pH 7.4). The homogenates were then centrifuged at 20,000 g and 4 ◦C and re-homogenized in 10 volumes of 50 mM PBS (pH 6.0) containing 0.5% hexadecyltrimethylammonium bromide (HETAB) and 10 mM ethylenediamine tetra-acetic acid (EDTA). Subsequently, the homogenates were subjected to a freeze/thaw cycle and a brief period of sonication, diluted in 50 volumes of 50 mM PBS (pH 6) and added to 50 ml of a solution containing o-Dianisidine dihydrochloride (0.067%), HETAB (0.5%) and hydrogen peroxide (0.003%). Complete sample reaction mixtures were placed in separate wellsand incubated in darkness for 5 min. A microplate reader (Labsystem Multiskan EX, Helsinki, Finland) was used to measure the changes in absorbance at 450 nm in accordance with the user's manual. The results are expressed in units of U/mg protein.

#### Determination of TNF-α Level

fphys-10-00325 March 27, 2019 Time: 12:11 # 3

Colon samples were homogenized in ice-cold PBS as previously described and centrifuged at 3,000 g for 10 min. The TNF-α levels in the supernatants were determined using an enzymelinked immunosorbent assay kit (Dou et al., 2013). Moreover, the degree of tissue inflammation was monitored using the tissue TNF-α activity level, which exhibits a linear relationship with neutrophilic infiltration in inflamed tissues. The TNF-α activity was determined in colonic samples adjacent to the installation point with a kits in accordance with the manufacturer's instructions (CytoStore, Calgary, AB, Canada). The results from each sample are expressed in units of pg/mg of protein.

#### 16S rRNA Sequencing

Thawed samples of colonic contents were mixed with PBS and homogenized using a high-speed homogenizer. The QIAamp DNA Stool Mini Kit was used to extract the total bacterial genome from each sample according to the manufacturer's protocol (Qiagen, Hilden, Germany). A polymerase chain reaction (PCR) assay containing the Phusion <sup>R</sup> High-Fidelity PCR Master Mix and GC Buffer kit (New England Biolabs, Ipswich, MA, United States) and the primers 341F (5<sup>0</sup> -CCTAYGGGRBGCASCAG-3<sup>0</sup> ) and 806R (5<sup>0</sup> - GGACTACNNGGGTATCTAAT-3<sup>0</sup> ) was used to amplify the 16S rRNA (V3–V4 region) genes in each sample. The following PCR conditions were applied: initial denaturation at 98◦C for 30 s; 30 cycles of 98◦C for 10 s, 55◦C for 30 s and 72◦C for 30 s and a final extension at 72◦C for 5 min. The resulting PCR products were stored at 4◦C until required. Later, the target fragments were detected by subjecting the PCR products to 2% agarose gel electrophoresis and were collected using the QIAquick Gel Extraction Kit (Qiagen, Germantown, MD, United States). A gene library was constructed using the Illumina TruSeq <sup>R</sup> DNA PCR-Free Sample Preparation Kit (Illumina Inc., San Diego, CA, United States), quantified using Qubit and real-time PCR and sequenced on a HiSeq2500 PE250 device (Illumina Inc.).

# Statistical Analysis

SPSS, version 20 (IBM Corp., Armonk, NY, United States) was used to conduct the statistical analysis. A one-way analysis of variance with Duncan's multiple range test was used to determine significant differences between the groups. A P-value of <0.05 was considered to indicate a significant difference.

# RESULTS

# CGA Ameliorates DSS-Induced Colitis

According to the initial random group allocation, mice in the CGA and CGA-DSS groups were treated with CGA for 14 days. The other groups received the basal diet for the same length of time. Mice in the DSS and CGA-DSS groups then received 2.5% DSS for the last 5 days of the study period, after which the final body weight and DAI were recorded (**Figure 1**). The results demonstrate that in the CGA-DSS group dietary CGA significantly reduced the DAI, compared to the DSS group (P < 0.05), and alleviated DSS-induced colitis. Despite CGA failing to reverse the DSS-induced weight loss, compared with mice in the DSS group, the mice in the CGA group gained weight (P < 0.05).

#### The Effects of CGA on MPO Activity and TNF-α Levels in Mice With DSS-Induced Colitis

Myeloperoxidase is a potential marker of tissue inflammation, tissue injury and neutrophil infiltration (Krawisz et al., 1984). The experimental results indicate increased MPO activity in the DSS group relative to the other groups (P < 0.05), as shown in **Figure 2A**. TNF-α also appears to play a significant role in the inflammatory process associated with ulcerative colitis (Moldoveanu et al., 2015). In this study, mice exposed to DSS exhibited a significant increase in TNF-α levels in the colon, whereas CGA treatment significantly mitigated this response (P < 0.05) as shown in **Figure 2B**. These data suggest that CGA exerts an anti-inflammatory effect by reducing neutrophil infiltration and pro-inflammatory cytokine production in the colon.

indicates a P-value <0.05.

a P-value <0.05.

# The Effects of CGA on Macroscopic and Histologic Observations in DSS-Induced Colitis

A histological comparison of colonic tissues from the control and DSS groups revealed multiple erosive lesions and an extensive infiltration of inflammatory cells that predominantly comprised macrophages, lymphocytes and neutrophils, along with irregular eosinophils, in the latter (**Figures 3A–D**). The levels of inflammatory cell infiltration and observable degree of tissue damage in the colonic tissues were assessed using the previously described histological scoring system. After a 5-day period of exposure to DSS in drinking water, mice in the DSS group exhibited an advanced inflammatory response. However, CGA had a mitigating effect on DSS-induced inflammatory markers (**Figure 3G**). Colon shortening is an indicator of colitis and inflammation (Siegmund, 2002). Notably, the DSS group exhibited significant colon shortening, compared to the other groups (**Figure 3E**, P < 0.05). However, no significant intergroup differences in colon weight were observed (**Figure 3F**).

### The Effect of CGA on Diversity in the Colonic Microbiota

Next, the V3–V4 regions of 16S rRNA isolated from the colonic digest samples were sequenced. The Shannon (**Figure 4A**), Simpson (**Figure 4B**), Chao (**Figure 4C**) and ACE indices (**Figure 4D**) were used to measure colonic microbial diversity. Treatment with DSS led to decreases in both the Shannon (P = 0.042) and Simpson indices (P = 0.035), whereas CGA treatment reversed these changes. By contrast, no significant differences were observed in the Chao and ACE indices.

# The Effect of CGA on Microbial Abundance at the Phylum Level

The major bacterial phyla in the colonic digest samples included Bacteroidetes, Firmicutes, Proteobacteria, and Verrucomicrobia, which accounted for approximately 90% of the microbiota. Bacteroidetes was most prevalent, accounting for 60.06, 49.72, 29.85, and 37.29% of the colonies in the CON, CGA, DSS, and CGA-DSS groups, respectively. The corresponding proportions of Firmicutes were 27.14, 29.45, 33.78, and 36.95%, respectively, while those of Proteobacteria were 8.62, 10.00, 8.37, and 8.20%, respectively (**Figure 5A**). DSS treatment decreased the relative abundance of Bacteroidetes (**Figure 5B**, P < 0.05), and this change was not reversed by CGA treatment. DSS treatment also decreased the relative abundance of Verrucomicrobia (**Figure 5C**).

Frontiers in Physiology | www.frontiersin.org

indicates a P-value <0.05.

∗

CGA-DSS groups. <sup>∗</sup> indicates a P-value <0.05.

# The Effect of CGA on Microbial Abundance at the Genus Level

The 10 most abundant microbial genera are shown in **Figure 6**. Bacteroides (14.86), Parasutterella (2.72%), and Helicobacter (3.44%) were predominant in the CON group, whereas Akkermansia, Bacteroides and Lactobacillus were the three major strains in the CGA, DSS, and CGA-DSS groups. In the latter groups, the proportions of Bacteroides were 12.56, 14.05, and 9.44%, respectively, while those of Lactobacillus and Akkermansia were 6.58, 9.39, and 12.47%, respectively and 6.27, 15.94, and 18.87%, respectively. The proportion of Akkermansia increased in the CGA-DSS group relative to the CON and CGA groups and in the DSS group relative to the CON group. CGA also increased the relative abundance of Lactobacillus.

# DISCUSSION

Continuous exposure to external stressors, such as foods, bacteria and environmental chemicals, induces a certain level of intestinal inflammation, although intestinal tissue damage and dysfunction may also play a role (Shimizu, 2017). CGA, an esterification product of caffeic acid and quinic acid, is one of the most abundant polyphenols in the human diet and has been shown to inhibit inflammatory responses in intestinal cells (Zhang et al., 2010). In this study, treatment with 2% CGA was shown to prevent weight loss and reduce the DAI. Furthermore, CGA reduced MPO activity and TNF-α levels in IBD-affected colons and ameliorated DSS-induced colon shortening and inflammatory responses. Moreover, CGA reduced the negative effects of DSS on intestinal microbial diversity and increased the relative abundance of Lactobacillus spp.

Intestinal dysfunction may affect the absorption of nutrients and, consequently, body weight. Furthermore, colon shortening is a sign and symptom of inflammation (Gearry et al., 2009). Studies of wild-type mice have shown that DSS-induced colitis significantly reduces body weight gain, increases epithelial permeability, rectal bleeding, colon length shortening, ulcer formation, inflammatory cell infiltration and goblet cell loss (Gadaleta et al., 2011). In a BALB/c mouse model of DSS-induced colitis, green tea polyphenols appeared to attenuate colitis by reducing the levels of TNF-α and serum amyloid A. Notably, the effects of green tea polyphenols on colitis were similar to those of sulfasalazine (Oz et al., 2013). Another study found that polyphenols from extra virgin olive oil could reduce the DAI and decrease the expression of monocyte chemoattractant protein

(MCP)-1, TNF-α, cyclooxygenase (COX)-2 and inducible nitric oxide synthase (iNOS) in colon tissues when compared with DSS treatment alone. These findings suggest that dietary polyphenols may be beneficial for the treatment of ulcerative colitis (Sanchez-Fidalgo et al., 2013). Our study revealed that the addition of CGA to drinking water reduced the DAI and the severity of colon damage and shortening in response to DSS.

control (CON), chlorogenic acid (CGA), dextran sulfate sodium (DSS), and CGA-DSS groups. <sup>∗</sup>

Cytokines are characteristic factors of IBD, wherein they act as key pathophysiological regulators of the occurrence, development and final resolution of inflammation (Strober and Fuss, 2011). In the past few decades, studies of cytokines in the context of IBD and other mucosal inflammatory conditions have been fruitful. These studies not only have provided us with important insights into the mechanisms of these diseases, but have also indicated directions for new treatment methods. Interactions between TNF family receptors and their ligands play important roles in the formation of key immune responses, including programmed cell death and lymphocyte co-stimulation (Meylan et al., 2011). CGA can inhibit the production of TNF, IL-6, INF-γ, MCP-l and macrophage inflammatory protein-l α by human peripheral blood monocytes in response to pathogenic bacteria (Krakauer, 2002). Ellagic acid was shown to reduce the expression of NF-κB, COX-2, iNOS, TNF-α, and IL-6 in 1, 2-dimethylhydrazine-induced colon cancer (Umesalma and Sudhandiran, 2010). MPO, a primarily a neutrophilic enzyme, is used as a quantitative indicator of inflammation because of the correlation between MPO activity and the histological detection of neutrophilic infiltration in the colon (Cetinkaya et al., 2006;

indicates a P-value <0.05.

Dost et al., 2009; Guan and Lan, 2018). In a previous rodent model study, 50 mg/kg CGA inhibited an increase in MPO activity in response to LPS and suppressed the migration of polymorphonuclear neutrophils to the lungs, as detected in bronchoalveolar lavage fluid. CGA also significantly reduced iNOS activity in lung tissues and thus inhibited the release of nitric oxide by LPS (Zhang et al., 2010). Another study found that treatment with 50 mg/kg CGA reduced the levels of MPO in gastric tissues, while a morphological analysis revealed the inhibition of neutrophilic infiltration into damaged tissues (Shimoyama et al., 2013). In this study, 2% CGA was shown to reduce MPO activity and DSS-induced damage in colon tissues.

Studies of animal models of IBD have suggested that intestinal inflammation relies heavily on a triggering event mediated by the intestinal microflora. Changes in the intestinal microflora can significantly affect both host immunity and mucosal inflammation (Hooper and Gordon, 2001; Ott et al., 2004; Huang et al., 2018; Li et al., 2018). Some changes in the microbial community are consistently observed in the context of intestinal inflammation, including reduced diversity (especially Firmicutes) and the presence of uncommon bacteria and increased concentrations of Escherichia coli (including pathogenic strains) (Sokol et al., 2006, 2008; Frank et al., 2007). Studies have shown that the oral administration of DSS can reduce the abundance of Akkermansia muciniphila and Bacteroides acidifaciens in feces (Kang et al., 2013; Yin et al., 2018). Recent studies of an IL10−/<sup>−</sup> mouse model of IBD have identified NLRP6 as an important inhibitor of spontaneous colitis, whereas a lack of NLRP6 leads to the enrichment of Akkermansia muciniphila (Seregin et al., 2017; Zhang et al., 2018). Further studies of mouse models of colitis found that Lactobacillus paracasei could reduce intestinal inflammation and the expression of pro-inflammatory factors in the mucosa (Oliveira et al., 2011; Wu and Tian, 2017; Azad et al., 2018). Treatment with a polyphenol-rich cranberry extract was shown to significantly increase the proportion of mucus-degrading Akkermansia spp. in a metagenomic sample (Anhe et al., 2015). Moreover, a phenolic extract of grape pomace/grape cider (1 mg/mL) significantly increased the Lactobacillus acidophilus biomass in liquid medium, while exposure to the highest concentration of phenolic compounds (5000 g/disk) had no

#### REFERENCES


inhibitory effect on the growth of Lactobacillus acidophilus in an agar diffusion assay (Hervert-Hernandez et al., 2009). This study demonstrated that 2% CGA could improve microbial diversity in the colon, particularly the relative abundances of Akkermansia and Lactobacillus, and suggests that CGA could potentially restore the microecological disorder induced by DSS.

#### CONCLUSION

In conclusion, our data provide insights into the role of CGA as a regulator of immunity and microbial diversity in the colon. Although these experiments were conducted in a mouse model of DSS-induced colitis, CGA could similarly inhibit the release of proinflammatory cytokines, such as MPO and TNF-α, in ulcerative colitis. Moreover, CGA was shown to have positive effects on the intestinal microbiota. Specifically, this polyphenol led to increases in microbial diversity and the relative abundances of Akkermansia and Lactobacillus. Our study thus demonstrates the possible use of CGA as a therapeutic adjunct to colitis treatment.

#### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and/or the supplementary files.

#### AUTHOR CONTRIBUTIONS

PZ, HJ, CW, and YL finished all the experiments. PZ and HJ performed the statistical analysis. PZ finished the first draft of the manuscript. SY critically revised the manuscript. All the authors read and approved the manuscript.

# FUNDING

This Project was supported by Science and Technology Development Fund Project of Tianjin no. 20130138.

bowel disease. Infect. Immun. 68, 7010–7017. doi: 10.1128/IAI.68.12.7010-7017. 2000



transfer mouse model of experimental colitis. Clin. Dev. Immunol. 2011:807483. doi: 10.1155/2011/807483



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Zhang, Jiao, Wang, Lin and You. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Egg Protein Transferrin-Derived Peptides IRW and IQW Regulate Citrobacter rodentium-Induced, Inflammation-Related Microbial and Metabolomic Profiles

Yong Ma<sup>1</sup> , Sujuan Ding<sup>1</sup> , Gang Liu1,2 \*, Jun Fang<sup>1</sup> \*, Wenxin Yan<sup>1</sup> , Veeramuthu Duraipandiyan<sup>3</sup> , Naif Abdullah Al-Dhabi<sup>3</sup> , Galal Ali Esmail<sup>3</sup> and Hongmei Jiang<sup>1</sup>

<sup>1</sup> College of Bioscience and Biotechnology, Hunan Agricultural University, Changsha, China, <sup>2</sup> Hunan Provincial Key Laboratory of Animal Nutritional Physiology and Metabolic Process, CAS Key Laboratory of Agro-ecological Processes in Subtropical Region, National Engineering Laboratory for Pollution Control and Waste Utilization in Livestock and Poultry Production, Institute of Subtropical Agriculture, Chinese Academy of Sciences, Changsha, China, <sup>3</sup> Department of Botany and Microbiology, College of Sciences, King Saud University, Riyadh, Saudi Arabia

Edited by: Helieh S. Oz, University of Kentucky, United States Reviewed by: Qiuxiang Zhang,

Jiangnan University, China Yi Cai, Sichuan Agricultural University, China

\*Correspondence:

Gang Liu gangle.liu@gmail.com Jun Fang fangjun1973@hunau.edu.cn

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 29 October 2018 Accepted: 14 March 2019 Published: 03 April 2019

#### Citation:

Ma Y, Ding S, Liu G, Fang J, Yan W, Duraipandiyan V, Al-Dhabi NA, Esmail GA and Jiang H (2019) Egg Protein Transferrin-Derived Peptides IRW and IQW Regulate Citrobacter rodentium-Induced, Inflammation-Related Microbial and Metabolomic Profiles. Front. Microbiol. 10:643. doi: 10.3389/fmicb.2019.00643 Bioactive peptides that target the gastrointestinal tract can strongly affect the health of animals and humans. This study aimed to evaluate the abilities of two peptides derived from egg albumin transferrin, IRW and IQW, to treat enteritis in a mouse model of Citrobacter rodentium-induced colitis by evaluating serum metabolomics and gut microbes. Forty-eight mice were randomly assigned to six groups: basal diet (CTRL), intragastric administration Citrobacter rodentium (CR), basal diet with 0.03%IRW (IRW), CR with 0.03% IRW (IRW+CR), basal diet with 0.03%IQW (IQW) and CR with 0.03% IQW (IQW+CR). CR administration began on day 10 and continued for 7 days. After 14 days of IRW and IQW treatment, serum was collected and subjected to a metabolomics analysis. The length and weight of each colon were measured, and the colon contents were collected for 16srRNA sequencing. The colons were significantly longer in the CR group, compared to the CTRL group. A serum metabolomics analysis revealed no significant difference in microbial diversity between the six groups. Compared with the CTRL group, the proportions of Firmicutes and Actinobacteria species decreased significantly and the proportions of Bacteroidetes and Proteobacteria species increased in the CR group. There were no significant differences between the CTRL and other groups. The serum metabolomics analysis revealed that Infected by CR increased the levels of oxalic acid, homogentisic acid and prostaglandin but decreased the levels of L-glutamine, L-acetyl carnitine, 1-methylhistidine and gentisic acid. Therefore, treatment with IRW and IQW was shown to regulate the intestinal microorganisms associated with colonic inflammation and serum metabolite levels, thus improving intestinal health.

Keywords: IRW, IQW, Citrobacter rodentium, inflammation, microbial, metabolomic

# INTRODUCTION

fmicb-10-00643 April 1, 2019 Time: 18:5 # 2

Citrobacter rodentium (CR) is a natural mouse pathogen that can be used to cause intestinal inflammation replaced enteropathogenic Escherichia coli (EHEC) and enteropathogenic E. coli (EPEC) (Hart et al., 2008). Although CR is a nonmotile pathogen, it can infect the host by passing through the intestinal mucosa to reach the intestinal epithelial cells. It adheres to the intestinal epithelial cells through surface structures called fimbriae when the CR started to infect the host (Mundy et al., 2003; Hart et al., 2008). These fimbriae can mediate infections in target cells that result in disorders of the host immune system. Therefore, the fimbriae are widely considered an important pathogenic factor in several diseases, particularly urinary, genital, and gastrointestinal infections (Petty et al., 2010). CR successfully colonizes intestinal epithelial cells, where it causes significant increases in the levels of cytokines and infiltrating immune cells in the intestinal environment (Collins et al., 2014). Consequently, infected intestinal epithelial cells promote the apical expression of antimicrobial peptides and inducible nitric oxide synthase, thereby modulating the inflammatory signal to produce NO, which inhibits CR (Vallance et al., 2002; Lopez et al., 2016). Although CR induces inflammatory infiltration, proinflammatory factor production, intestinal mucosal injury, and similar processes associated with inflammatory bowel disease in mice, this pathogen does not cause clinical diarrhea in humans (MacDonald et al., 2003). It may be an ideal model for mice but may not translate to human colonic inflammation, as it does not cause diarrhea in man (Zhang W. et al., 2015).

Many studies have provided substantial evidence indicating the regulatory roles of active peptides in intestinal inflammation (Kvidera et al., 2017; Eissa et al., 2018; Lv et al., 2018). The results of mouse experiments showed that porcine β-defensin 2 might improve mucosal lesions and extracellular permeability by targeting the NF-κB pathway and could thus alleviate inflammatory enteritis (Han et al., 2015). Cathelicidin-related antimicrobial peptide (CRAMP) restore body weight loss caused by ulcerative colitis and maintain colonic epithelial integrity (Li et al., 2018). Cathelicidin-BF can effectively inhibit the phosphorylation of NF-κB and increase the intestinal barrier function to reduce inflammation (Zhang H. et al., 2015). Egg protein transferrin-derived peptides, such as IRW and IQW, have been shown to provide effective relief of cardiovascular disease symptoms. The intact tripeptide structure of these molecules can alleviate endothelial inflammation and oxidative stress (Chen et al., 2017). IRW and IQW have also been shown to alleviate inflammatory responses induced by tumor necrosis factor (TNF) and may therefore be useful as nutraceuticals (Majumder et al., 2013). IRW can inhibit the TNF-α-induced expression of both ICAM-1 and VCAM-1, whereas IQW can only inhibit the expression of ICAM-1. Treatment with IRW and IQW also inhibits the expression of antioxidant enzymes and promotes the activity of superoxide dismutase (SOD), catalase (CAT), and glutathione peroxidase (GPx) (Liu et al., 2018; Ma et al., 2018). Therefore, we hypothesized that IRW and IQW could reduce intestinal damage and thus slow the inflammatory process. This study mainly analyzed the levels of serum metabolites in a CR-induced model of IBD in response to these two peptides, as well as the effects on microbial diversity and composition in the colonic contents.

# MATERIALS AND METHODS

# Animal and Experimental Design

The Animal Care and Use Committees of Hunan Agricultural University provided approval for the experiments. IRW and IQW (Ontores, Zhejiang, China) were dissolved in sterile water at 4 ◦C until the solution temperature reached equilibrium with the ambient temperature. Both peptides were synthetic and had purity levels of 99%. A CR bacterial solution was mixed with glycerol at a ratio of 3:7 and stored at −80◦C before inoculation into Luria Bertani broth to generate an activation culture before gavage. The mature CR culture was centrifuged at 4000 rpm for 10 min, and the supernatant was mixed with physiological saline to yield a 5 × 10<sup>9</sup> CFU/ml bacterial suspension that was set aside (Guan et al., 2016). Forty-eight mice (average body weight: 23 g; age: 8 weeks) were randomly assigned to six groups (n = 8 per group): basal diet (CTRL), basal diet with intragastric administration of 0.1 mL bacterial suspension (CR), basal diet with 0.03%IRW (IRW), CR with 0.03% IRW (IRW+CR), basal diet with 0.03%IQW (IQW) and CR with 0.03% IQW (IQW+CR) (Liu et al., 2018). All mice in one group were housed and fed in seperate cages. The cages were stored in a sterile environment under a 12-h light-dark cycle, relative humidity of 53%, and temperature of 24◦C. After 3 days of acclimation, IRW and IQW were fed to mice. CR administration began on day 10 and continued for 7 days. All mice were slaughtered on day 17. The experimental process adhered strictly to the animal experimental guidelines. Serum samples were collected for metabolomics analysis. The length and weight of the colon were measured, and a section was collected for a morphological analysis. The colon contents were also collected for 16s rDNA sequencing.

# Histopathology

Colon fragments were rinsed with saline and fixed in 12% formalin. The fixed colon fragments were dehydrated using an ethanol gradient and embedded in paraffin. Ten-micrometer sections were stained with hematoxylin and eosin according to standard protocols. The samples were observed under blind conditions using an optic Olympus BX41 microscope (Münster, Germany).

# Serum Metabolomics

The serum samples were thawed at room temperature and 100 µL aliquots were transferred into centrifuge tubes (1.5 mL). All serum samples were extracted using 300 µL of methanol and mixed with 10 µL of an internal standard (3.1 mg/mL, DL-ochlorophenylalanine). The samples were then vortexed for 30 s and centrifuged at 12000 rpm and 4◦C for 15 min. ACQUITYTM UPLC -QTOF analysis system and Waters ACQUITY UPLC HSS T3 chromatographic column (2.1 mm × 100 mm, 1.8 covering m) were used for LC-MS detection. The following chromatographic separation conditions were applied: column temperature, 40◦C; mobile phase A, water+0.1% formic acid; mobile phase B: acetonitrile+0.1% formic acid; flow rate: 0.35 mL/min; injection volume: 6 µL. The data were first transformed to CDF files using CDF bridge and imported into XCMS software for peak picking, peak alignment, peak filtering and peak filling. The data, including the retention time (RT), MZ, observations (samples) and peak intensity, were normalized using Excel 2007.

# 16S rDNA and Illumina MiSeq Sequencing

fmicb-10-00643 April 1, 2019 Time: 18:5 # 3

The colon contents were collected from all mice immediately after sacrifice. The V3–V4 region of the 16sRNA gene was sequenced in the 48 content samples. Microbial DNA was extracted from the colon contents using the QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany). The microbial V3-V4 region was amplified by PCR using the primers 5<sup>0</sup> -ACTCCTACGGGAGGCAGCA and 3<sup>0</sup> -GGACT ACHVGGGTWTCTAAT. Amplicons were extracted from a 2% agarose gel using the TIANgel MIdi Purification Kit (TIANGEN BIOTECH, Beijing, China) and were subsequently purified. For detailed experimental steps on Illumina MiSeq sequencing, please see our previous article (Zhu et al., 2018); The general data analysis was performed by a commercial company (Novogene, Beijing, China).

### Correlation Analysis Between Differential Metabolite Levels and Genus-Level Intestinal Microbes

SPSS 16.0 was used to conduct a correlative analysis of the proportions of microorganisms in each genus and details of the metabolic differences among all samples. GraphPad Prism was used to conduct a regression analysis of the results from the correlation analysis and draw graphs.

# Data Analysis

All data are presented as means ± standard errors of the means (SEM) and were analyzed using SPSS 16.0 software. Differences between mean values were evaluated using a one-way analysis of variance. If dissimilarities were detected, Tukey's multiple comparisons test was used. A P value <0.05 was considered to indicate a significant difference.

# RESULTS

# Effects of IRW and IQW on Changes in the Colon

Infected by CR induced ulcers and other inflammatory reactions that caused mucosal damage in the colon and reduced the colon length. The average colon length in the CTRL group was 9.36 ± 1.68 cm, compared to 7.38 ± 0.61 cm in the CR group. In other words, CR treatment significantly reduced the colon length (P < 0.05), whereas co-treatment with IRW and IQW maintained the colon length (P > 0.05) (**Figure 1A**). However, the colon weights did not differ between the six groups (P > 0.05) (**Figure 1B**). More severe histological damage was observed in colon tissues from the CR group, compared with the CTRL group, as demonstrated by broad disruption of the tissue architecture and the disappearance of intestinal crypts and goblet cells. Compared with the CTRL group, colon tissue from the CR group exhibited more extensive cellular infiltration (**Figure 1C**).

# Diversity of Bacterial Community of Colonic Contents

**Table 1** presents the original readings obtained by sequencing the V3–V4 region of 16s rRNA from each sample of colon contents. The sample size we sequenced is sufficient to reflect the abundance of microorganisms. The OTU (Operational Taxonomic Units) number of microbes in the colon is about 1200 (**Figure 2**). Raw reads and Clean Reads in the CR group were lower than other groups (**Table 1**). The six groups of goods coverage have reached more than 99%. There were no statistically significant differences in community richness (chao1 and ACE), community diversity (Shannon and Simpson) or species coverage (cargo coverage) (**Table 1**) (P > 0.05).

#### Composition of Bacterial Communities in the Colon Contents

We performed a taxonomic analysis of the sequencing results to determine the composition and community structure of different populations (or samples) at each classification level (e.g., domain, kingdom, gate, class, sequence, family, genus, species, OTU). At the phylum level, 10 phyla were detected at the highest abundance. Firmicutes, Bacteroidetes, Proteobacteria, Saccharibacteria and Actinobacteria were the most abundantly expressed phyla in all groups, accounting for 99% of all sequencing results. In the CTRL group, Firmicutes (59.7%), Bacteroidetes (27.5%) and Proteobacteria (8.2%) were the most abundant phyla (**Figure 3**), whereas Firmicutes (34.8%), Bacteroidetes (41.4%), Proteobacteria (17.2%) and Saccharibacteria (4.2%) were the most abundant phyla in the CR group (**Figure 3**). Firmicutes (56.1%), Bacteroidetes (34.6%), Proteobacteria (2.3%) and Actinobacteria (5.6%) were the most abundant phyla in the IRW group (**Figure 3**), while Firmicutes (47.8%), Bacteroidetes (42.6%), Proteobacteria (3.4%) and Actinobacteria (4.7%) were the most abundant phyla in the IRW+CR group (**Figure 3**). Firmicutes (77.1%), Bacteroidetes (15.9%) and Proteobacteria (5.0%) were the most abundant phyla in the IQW group (**Figure 3**), while Firmicutes (58.4%), Bacteroidetes (28.5%) and Proteobacteria (5.0%) were the most abundant in the IQW+CR group (**Figure 3**). Compared with the CTRL group, the CR group exhibited significantly lower percentages of Firmicutes (59.7% vs. 34.8%) and Actinobacteria species (2.3% vs. 1.7%) (**Figure 3**) but significantly higher percentages of Bacteroidetes (27.5% vs. 41.4%), Proteobacteria (8.2% vs. 17.2%) and Saccharibacteria species (1.5% vs. 4.2%) (**Figure 3**). However, treatment with IRW and IQW restored the percentages of Firmicutes, Actinobacteria, Bacteroidetes, Proteobacteria and Saccharibacteria in CR animals to levels that did not



differ significantly from the CTRL group (**Figure 3**). **Figure 4** lists the 10 most abundant genera in the colonic contents. Notably, the percentage of Lactobacillus decreased and the percentages of Helicobacter and Odoribacter increased in the CR group, compared with the CTRL group (**Figure 4**). Similarly, IRW and IQW treatment restored the normal percentages of Lactobacillus, Helicobacter and Odoribacter species in CR animals.

FIGURE 3 | (A) Taxonomic compositions of fecal bacterial communities at the phylum level. (B) Percentage of Firmicutes species in a sample from each of the four groups. (C) Percentage of Bacteroidetes species in a sample from each the four groups. <sup>∗</sup>P < 0.05.

CR-IQW+CR (ESI+). (C2) OPLS-DA score plot of CR-IQW+CR (ESI-).

#### Analysis of Serum Metabolomics

To identify differences between the six groups, a PCA (Principal Component Analysis) modeling method was used to analyze the serum samples. In this analysis, six principal components were obtained in the positive mode, and the performance characteristics of this multivariate model in the positive ion mode were as follows: R2X = 0.705, Q <sup>2</sup> = 0.591. Nine principal components were obtained in the negative mode. Similarly, the performance characteristics of the model in the negative ion mode were as follows: R2X = 0.531, Q <sup>2</sup> = 0.189. In addition, the total PCA score demonstrates that the scatters of the six groups were completely separated in both the ESI+ (**Figure 5A1**) and ESI- modes (**Figure 5A2**). There were 10 metabolic differences between the CR and CTRL groups, including four in the ESI+ mode and six in the ESI- mode (**Table 2**). The results of a plasma metabolite analysis revealed decreased levels of oxalic acid, allantoin, homogentisic acid, α-linolenic Acid, prostaglandin D3 and lysoPC and increased amounts of L-glutamine, gentisic acid, L-acetylcarnitine and 1-methylhistidine in the CR group. However, normal amounts of L-glutamine, gentisic acid, L-acetylcarnitine and 1-methylhistidine were observed in the IRW+CR and IQW+CR groups (**Table 2**).

When the CR group received IRW treatment, we further adopted a supervised multidimensional statistical method, namely the partial least square discriminant analysis (PLS-DA), to conduct statistical analysis of the two groups of samples. The model quality parameters were two principal components in the positive mode, R2X = 0.449, R2Y = 0.967, Q <sup>2</sup> = 0.904, and two principal components in the negative mode, R <sup>2</sup>X = 0.34, R2Y = 0.998, Q <sup>2</sup> = 0.942. The PLS-DA loading plot demonstrates the complete separation of the scatters of the two groups (**Figures 5B1,B2**) due to differences in the detected metabolites. When the CR group received IQW treatment, the OPLS-DA method was again used for the modeling analysis. One principal component and one orthogonal component, R <sup>2</sup>X = 0.481, R2Y = 0.979, Q <sup>2</sup> = 0.918, were obtained in the positive mode. One principal component and two orthogonal components, R2X = 0.472, R2Y = 1, Q <sup>2</sup> = 0.941, were obtained in the negative mode. The OPLS-DA scoring diagrams are shown in **Figures 5C1,C2**. In order to verify whether the model is "overfitting", the model is sorted (**Figure 5B3**), and the results show that the model is reliable.

#### Correlation Analysis Between Differential Metabolites and Genus-Level Intestinal Microbes

The correlations between different metabolites in the serum metabolome results and the profiles of genus-level intestinal microbes. Gentisic acid, L-glutamine, 1-methylhistidine, lysoPC and L-acetylcarnitine were found to associate significantly with Lactobacillus. Specifically, negative correlations were detected between Lactobacillus and gentisic acid (Pearson correlation: −0.347, P = 0.016) (**Figure 6A**), L-glutamine (Pearson correlation: −0.412, P = 0.004) (**Figure 6B**), 1 methylhistidine (Pearson Correlation: −0.438, P = 0.002) (**Figure 6C**), L-acetylcarnitine (Pearson correlation: −0.338, P = 0.019) (**Figure 6E**). However, a positive correlation was observed between Lactobacillus and LysoPC (Pearson correlation: 0.328, P = 0.023) (**Figure 6D**).

#### DISCUSSION

Citrobacter rodentium causes severe inflammatory infections in the mouse colon (Guan et al., 2016). These infections cause mucosal damage, reduce the colonic length, increased the colonic wall thickness, and cause hyperemia and even ulceration (Liu et al., 2015). Some studies also suggest that shortening of the colon is closely related to histological changes and even suggest that this morphological parameter indicates the degree of colonic inflammation (Yuge et al., 2014; Oz, 2017a). In our study, the colon lengths of mice infected intragastrically with CR were significantly shorter than those of normal mice, and this change was eliminated in the presence of IRW and IQW. The shortening of colon length may be an early warning of



inflammation. A large number of experimental results indicate that colonic inflammation is accompanied by a decrease in colon length and other different degrees of morphological damage (Gearry et al., 2009; Gadaleta et al., 2011). Accordingly, IRW and IQW treatment appears to play a role in colonic inflammation (Jiao et al., 2019).

Mouse fecal microbes may be contaminated, so we try to be aseptic as possible during the sampling process and the samples are stored in sterile, enzyme-free EP tubes. Our microbiological sequencing results are consistent with the composition of the gut microbiota in most inflammatory studies, so the possibility of contamination is low. The intestinal microbes change significantly at the time of intestinal inflammation. However, different degrees of inflammation lead to different results in dissimilar individuals; for example, IBD ecological disorders differ widely (Walker et al., 2011; Weichselbaum and

Klein, 2018). Notably, colonic inflammation commonly results in decreases in microbial diversity and the proportion of Firmicutes, as well as an increase in the proportion of Proteobacteria (Matsuoka and Kanai, 2015; Hong and Piao, 2018). In this study, the lack of significant changes in microbial diversity may be attributable to large differences in diversity within the group, a consequence of poor homogeneity among the mice. Fortunately, the proportion of Firmicutes was lower and the proportion of Proteobacteria was higher in the CR group, compared to the CTRL group. CR played a role in establishing a model of colonic inflammation. The proportions of Firmicutes and Proteobacteria returned to normal, however, in mice treated with IRW and IQW, which demonstrates the usefulness of these treatments. Lactobacillus and Bifidobacterium are the most commonly used probiotics (Pace et al., 2015; Azad et al., 2018b). Some studies have shown that Lactobacillus degrades pro-inflammatory factors secreted in milk proteins, thus eliminating inflammation (von Schillde et al., 2012; Yin et al., 2018). Furthermore, a study by Hormannsperger et al. (2013) found that Lactobacillus not only degraded pro-inflammatory chemokines and reduced immune cell infiltration in colonic inflammation (Di Cerbo and Palmieri, 2013; Lightfoot et al., 2015) but could also be used to identify protective microbial structures and even in new drug applications (Hormannsperger et al., 2013; Li and Gui, 2018; Zhang and Gui, 2018). This study also confirmed that the proportions of Lactobacillus and Bifidobacterium were significantly reduced in the CR group but returned to normal after treatment with IRW and IQW. Helicobacter spp. cause more severe intestinal inflammation and injury via M1 macrophages (Krakowiak et al., 2015; Azad et al., 2018a; Ding et al., 2019). Other studies have shown that Helicobacter can improve ROS levels by inducing regeneration synthesis and neutrophil infiltration of the epithelium both in vivo and in vitro (Davies et al., 1994; Bagchi et al., 1996; Mannick et al., 1996). The proportion of inflammatory Helicobacter species was significantly increased in the CR group, whereas this increase was inhibited by IRW and IQW. Accordingly, these peptides improved the intestinal condition.

When serum metabolomics were measured in normal and colitis mice, significant differences were observed (Schicho et al., 2010). Some studies have found that inflammatory infections can alter the structure of the gut microbiota and regulate the corresponding serum amino acids accordingly. As the proportion of Helicobacter increased, the concentration of tryptophan decreased, consistent with the changing trends in our study results. This finding suggests that Helicobacter can inhibit tryptophan metabolism (Gadaleta et al., 2011). Catabolic enzymes and indoleamine 2,3-dioxygenase (IDO) can metabolize tryptophan to kynurenine (Munn et al., 1998; Puccetti and Grohmann, 2007), and tryptophan metabolism was shown to affect immune modulation in recent studies (Romani et al., 2008; Fu et al., 2018). The tryptophan concentration may be inversely proportional to the disease activity and CRP level; in other words, a decrease in tryptophan is associated with inflammation (Hisamatsu et al., 2012). Hashimoto et al. found that caprylic acid could inhibit transcription of the gene encoding IL-8 in Caco-2 cells (Hoshimoto et al., 2002), thereby inhibiting the chemotaxis and proliferation of inflammatory cells. This phenomenon was also found to increase mitochondrial respiration in inflammatory cells (Hecker et al., 2014; Oz, 2017b). The expression of caprylic acid improved significantly after IRW and IQW were administered to CR mice, indicating that these protein transferrin-derived peptides play a role in CR-induced intestinal inflammation. Specifically, these peptides alter the composition of gut microbes that regulate metabolites. In turn, these metabolites increase the level of caprylic acid and thus improve the symptoms of inflammation. Lactobacillus can have good antibacterial activity against Clostridium (Monteiro et al., 2019), and can well reduce the oxidative stress caused by high-fat diet (Hsu et al., 2019). Lactobacillus has been shown to produce SCFA (short chain fatty acid) from L-glutamine (Botta et al., 2017). In humans, SCFA can regulate immune and metabolic functions in multiple tissues and organs (Ohira et al., 2017; Bin et al., 2018). Therefore, we believe that the amount of L-glutamine decreases (due to increased consumption) as the number of Lactobacillus increases, leading to a negative correlation between these parameters. The correlation between these species and metabolites indicates that the strain can promote the production of some inhibitory substances or can use these substances to produce metabolites to inhibit inflammation, thereby alleviating the inflammation of the colon.

# CONCLUSION

In conclusion, supplementation of IRW and IQW improves the morphological damage of the colon caused by CR, while regulating the microbial composition in the colon and altering the serum metabolites through changes in microbial composition. Finally, colonic inflammation is improved by changes in these metabolite components. Further studies are needed to clarify and confirm the benefits of IRW and IQW treatment in the host gastrointestinal tract and other organ systems.

# AUTHOR CONTRIBUTIONS

YM and SD carried out the study and performed the statistical analysis. VD, NA-D, and GE provided assistance for the experiments. GL, JF, and HJ designed the research. YM, SD, and WY prepared the first draft of the manuscript. GL and JF read and revised the manuscript.

# FUNDING

This research was supported by National Natural Science Foundation of China (Nos. 31672457 and 31772642), National Key Research and Development Program of China (2016YFD0500504 and 2016YFD0501201), Ministry of Agricultural of the People's Republic of China (2015-Z64 and 2016-X47), Local Science and Technology Development Project Guided by The Central Government (YDZX20184300002303 and 2018CT5002), and Hunan Provincial Science and Technology Department (2017NK2322, 2018TP1031, 2016NK2101, 2016WK2008, 2016TP2005, and 2018WK4025), China Postdoctoral Science Foundation (2018M632963), and Double first-class construction project of Hunan Agricultural University (SYL201802003, YB2018007).

# REFERENCES

fmicb-10-00643 April 1, 2019 Time: 18:5 # 10


# ACKNOWLEDGMENTS

The authors would like to extend their sincere appreciation to the Deanship of Scientific Research at King Saud University for its funding of this research through the Research Group Project No. RGP-213.

controls virulence gene expression in Citrobacter rodentium. Infect. Immun. 76, 5247–5256. doi: 10.1128/IAI.00770-08


mouse chronic granulomatous disease. Nature 451, 211–215. doi: 10.1038/ nature06471


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Ma, Ding, Liu, Fang, Yan, Duraipandiyan, Al-Dhabi, Esmail and Jiang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fmicb-10-00643 April 1, 2019 Time: 18:5 # 11

# Astragalus and Ginseng Polysaccharides Improve Developmental, Intestinal Morphological, and Immune Functional Characters of Weaned Piglets

C. M. Yang<sup>1</sup>† , Q. J. Han<sup>1</sup>† , K. L. Wang<sup>1</sup> , Y. L. Xu<sup>1</sup> , J. H. Lan<sup>1</sup> and G. T. Cao<sup>2</sup> \*

<sup>1</sup> Key Laboratory of Applied Technology on Green-Eco-Healthy Animal Husbandry of Zhejiang Province, The Zhejiang Provincial Engineering Laboratory for Animal Health and Internet Technology, College of Animal Science and Technology, Zhejiang A & F University, Hangzhou, China, <sup>2</sup> College of Standardization, China Jiliang University, Hangzhou, China

#### Edited by:

Jie Yin, Institute of Subtropical Agriculture (CAS), China

#### Reviewed by:

Caihong Hu, Zhejiang University, China Heng-wei Cheng, USDA, Purdue University, United States

\*Correspondence: G. T. Cao 15a1903025@cjlu.edu.cn †These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology

Received: 21 February 2019 Accepted: 27 March 2019 Published: 12 April 2019

#### Citation:

Yang CM, Han QJ, Wang KL, Xu YL, Lan JH and Cao GT (2019) Astragalus and Ginseng Polysaccharides Improve Developmental, Intestinal Morphological, and Immune Functional Characters of Weaned Piglets. Front. Physiol. 10:418. doi: 10.3389/fphys.2019.00418 Antibiotic resistance is a major issue in animal industries and antibiotic-free alternatives are needed to treat infectious diseases and improve performance of pigs. Plant extracts have been suggested as a potential solution. The present study was conducted to investigate the effects of Astragalus polysaccharides (Aps) and ginseng polysaccharide (Gps) on growth performance, intestinal morphology, immune function, volatile fatty acids (VFAs), and microfloral community in weaned piglets. A total of 180 weaned piglets were randomly divided into three treatment groups during a 28-days feeding experiment, including a basal diet (Con), basal diet supplemented with 800 mg/kg Aps (Aps), and basal diet supplemented with 800 mg/kg Gps (Gps). Results showed that both Aps and Gps increased body weight, average daily gain and feed conversion rate, and reduced the rate of diarrhea. Gps also decreased aspartate aminotransferase compared to the Con piglets after 14 days. No significant effects on alanine aminotransferase were observed. Both Aps and Gps piglets exhibited higher serum immunoglobulin M levels after 14 and 28 days, and also decreased jejunal crypt depth, increased jejunal villus length and villus height/crypt depth ratio, and increased expression of toll-like receptor 4, myeloid differentiation primary response 88, nuclear factor-kappa B proteins in the jejunum. Aps and Gps piglets also had higher concentrations of acetic acid, isobutyric acid, and butyrate in their colon. Data of high-throughput sequencing revealed that Aps and Gps affected bacterial quantity and diversity in the colon. Species richness and evenness were higher in both Aps and Gps piglets than the control piglets. Aps and Gps piglets also had a higher relative abundance of Lachnospiraceae and Anaerostipes, and the Aps piglets had a higher relative abundance of Lactobacillus gasseri and L. amylovorus. Therefore, dietary supplementation with Aps and Gps could be beneficial for optimizing the performance of industry pigs and reducing dependence on antibiotics. Furthermore, Plant polysaccharides play a great role in promoting the sustainable development of animal husbandry.

Keywords: Astragalus polysaccharide, ginseng polysaccharide, volatile fatty acids, microflora community, piglet

# INTRODUCTION

fphys-10-00418 April 10, 2019 Time: 20:3 # 2

Early weaning is a stressor, and sometimes exhibit intestinal development, digestive, or immune dysfunction in weaned piglets, which leads to diarrhea and growth inhibition (Hampson, 1986; Wu et al., 1996). Antibiotics are usually used to control the incidence of infectious diseases and improve growth in weaned piglets (Tang et al., 2005; Kong et al., 2007; Niekamp et al., 2007). However, antibiotic resistance has become a major issue in recent years. Hence, effective antibiotic substitutes are urgently required to reduce antibiotic dependence in the animal industry.

Plant polysaccharides have great potential as an antibiotic substitute owing to their natural characteristics (Leon et al., 2008). Reports have shown that Astragalus polysaccharides (Aps) extracted from Astragalus membranaceus, act as mild immunomodulators, restoring intestinal morphology and alleviating intestinal inflammation in acetic acid-induced colitis rats (Li et al., 2011; Zhao et al., 2014). Xu et al. (2019) reported that Aps decreased cytotoxicity and enhanced splenic lymphocyte proliferation in mice. Ginseng polysaccharides (Gps) has been shown to rescue intestinal epithelial cell viability from rotavirus infection dose-dependently (Baek et al., 2010). Kim et al. (2017) revealed that Gps exerted an anti-immunosenescent effect by suppressing thymic involution, reduced serum levels of IL-2, and modulated several types of immune cells in mice. Li et al. (2019) also showed that Gps altered the composition and diversity of gut microbiota in mice with antibiotic-associated diarrhea, restored gut microbiota, balanced metabolic processes, and promoted the recovery of mucosa.

Although the anti-inflammatory and immunomodulatory effects of plant polysaccharides have been recognized, more previous studies have focused on the immune function in vitro and in mice, while less on the effects of plant polysaccharides on gut volatile fatty acids (VFAs) and microbiota in weaned piglets. The present experiment was conducted to assess the effects of Aps and Gps on the growth performance, serum index, intestinal morphology, immune function, volatile fatty acid content, and colonic microbial community composition in weaned piglets. The aim was to provide valuable insights into antibiotic-free alternatives for improving weaned piglet performance.

#### MATERIALS AND METHODS

#### Animals and Treatments

A total of 180 weaned piglets (Duroc × Landrace × Yorkshire), which weaned at day 28, were randomly assigned to three treatment groups, each group contained 6 pens per treatment, and 10 animals per pen. The experiment lasted for 29 days. The control group (Con) received a basal diet; the Aps group received the basal diet with 800 mg/kg Aps; the Gps group received the basal diet with 800 mg/kg Gps. The basal diet was formulated to meet the nutrient requirements suggested by NRC (1998) and contained no antibiotics (**Table 1**). Water and feed were provided ad libitum. All pigs were housed in a room with the temperature controlled between 26◦C and 28◦C. This study was carried out in accordance with the recommendations of the care and use of laboratory animals, Institutional Animal Care and Use Committee of Zhejiang Agricultural and Forestry University. The protocol was approved by the Ethics Committee of Zhejiang Agricultural and Forestry University, Hangzhou, China (SYXKzhe 2016-087).

#### Preparation of Aps and Gps

The polysaccharide product had a purity of 80% and a molecular weight of 20,000–60,000. Aps was composed of hexanoic acid, glucose, fructose, rhamnose, arabinose, and galacturonan. Gps was composed of rhamnolipid, xylose, glucose, and galactose. Aps and Gps were provided by Vegamax Biotechnology Co., Ltd. (Anji, China).

#### Growth Performance

Piglets were weighed individually at the beginning and end of experiment. Feed intake was recorded each day. Average daily feed intake (ADFI), average daily gain (ADG), and feed to gain (F:G) for each pen were calculated. The amount of diarrhea was observed daily and recorded for calculating the rate of diarrhea.

# Sample Collection

On day 14 and 28, six piglets per treatment (one piglet per pen) were randomly selected for sample collection. Blood samples were taken from the front cavity vein and centrifuged (3,000 × g, 10 min) at 4◦C, and then the serum was separated and promptly stored at −20◦C for further analysis. At 28 days, one pig per pen were slaughtered and sampled at a local commercial slaughterhouse. About 1 cm jejunal segment was immediately collected after slaughtering, and colon contents were collected

TABLE 1 | Composition and nutrient levels of the basal diet.


<sup>1</sup>Supplied the following per kg of diet: vitamin A, 10,000 IU; vitamin D3, 400 IU; vitamin E, 10 mg; pantothenic acid, 15 mg; vitamin B6, 2 mg; biotin, 0.3 mg; folic acid, 3 mg; vitamin B12, 0.009 mg; ascorbic acid, 40 mg; Fe, 150 mg; Cu, 130 mg; Mn, 60 mg; Zn, 120 mg; I, 0.3 mg; and Se, 0.25 mg.

Yang et al. Plant Extracts Improve Pig Growth

and promptly stored at −80◦C for the detection of VFAs and high throughput sequencing.

#### Serum Parameter Analysis

fphys-10-00418 April 10, 2019 Time: 20:3 # 3

Immunoglobulin A (IgA), immunoglobulin M (IgM), and immunoglobulin G (IgG) were assayed using porcine-specific immunoturbidimetry kits. Aspartate aminotransferase (AST) and alanine aminotransferase (ALT) were assayed using kits (Nanjing Jiancheng Bioengineering Institute, China), following the manufacturer's instructions.

#### Morphological and Immunohistochemical Analysis of Jejunum

The jejunum was fixed with 10% formaldehyde solution, routinely sampled, dehydrated, paraffin-embedded, sliced (4 µm thick), stained with hematoxylin-eosin, observed under an optical microscope for description (200 × ) with the main descriptive parts being photographed (Microscope: Nikon Eclipse ci; imaging system: Nikon digital sight DS-FI2). The villus height and crypt depth were measured at 10 visual fields from each intestinal sample and villus height/crypt depth ratio (VCR) values were calculated for each treatment group. Pathological changes were also observed in the same field of vision.

Immunohistochemical TLR4, MyD88, and NF-κB were performed on each sample, which antibodies were anti-TLR4 (ab8376), anti-MyD88 (ab119048) and anti-NF-kB (ab86299). Three 400-fold visual fields were randomly selected for one photography. Image-Pro Plus 6.0 software was used to select the same brown-yellow color in the blue circle as a unified criterion for judging all positive photographs. The accumulated optical density of each positive circle was obtained by analyzing the three circles of each photograph, and the average value was calculated.

#### VFAs Analysis

VFA levels were tested by Headspace Sampler Gas Chromatography (Agilent Technologies, United States) using the method of Thanh et al. (2009). The commercial standards of acetic acid, butyrate, propionic acid, isovalerate, isobutyric acid, and valerate treated as external standards, were purchased from China Sinopharm Chemical Reagents Co., Ltd. (Beijing, China). Metaphosphoric acid were acquired from Shanghai Aladdin Biotech Inc. (Shanghai, China). A mixture of 1 g colon content with 6% phosphorous acid (m/v, 1:3) was injected into an Agilent Technologies GC7890 Network System (Agilent Technologies, United States) equipped with a 30 m × 0.25 mm × 0.25 µm column (HP-FFAP, Agilent Technologies) for flame ionization (Yang et al., 2018).

# Colonic Microflora Content by 16S rRNA Sequencing

Colonic content samples from each pig were used for the microbial community analysis. The V4 region of the 16S rRNA gene was detected using the Illumina-HiSeq platform (Novogene Bioinformatics Technology Co., Ltd., Beijing, China). The analysis was carried out following the method of Wang et al. (2007). The specific primer sequences 515F (5<sup>0</sup> -GTGCCAGCMGCCGCGGTAA-3<sup>0</sup> ) and 806R (50 -GGACTACHVGGGTWTCTAAT-3<sup>0</sup> ), were used for the analysis of V4 region of 16S rRNA gene. Briefly, through "quantitative insights into microbial ecology" (QIIME) and UPARSE software, we allocated 97% similarity between taxonomy and ribosomal database project (RDP) classifiers. Operational taxonomic units (OTUs) were clustered and species classified based on valid data. According to OTU clustering results, species annotations were made for each OTU sequence to obtain the corresponding species information and speciesbased abundance distribution. The alpha diversity calculations and Venn diagrams were analyzed to obtain species richness and evenness counts and information on the common and specific OTUs among different groups. Furthermore, OTUs were subjected multiple sequence alignment and a phylogenetic tree was constructed. A principal component analysis (PCA) and unweighted pair-group method with arithmetic mean (UPGMA) cluster tree displayed the differences in community structure among different groups, and results were explored in MetaStat.

#### Statistical Analysis

SPSS Statistics 21.0 (SPSS Inc., United States) and GraphPad Prism 7 (GraphPad Software Inc., United States) were used for the statistical analysis. The growth performance, serum parameters, jejunal morphology, concentration of VFAs in colon contents and relative abundance of fecal microbial communities were compared and analyzed by one-way ANOVA, and results were deemed significant if P < 0.05.

# RESULTS

#### Effects of Aps and Gps on the Growth Performance

Piglets fed Aps or Gps had higher final body weight (P < 0.05) and ADG compared to the Con group (**Table 2**). A reduced F:G (P < 0.01) and diarrhea rate (P < 0.01) were observed in Aps and Gps piglets.

#### Effects of Aps and Gps on Liver Function Indexes

The effect of Aps or Gps on liver function indices in serum are shown in **Figure 1**. Gps decreased the concentration of serum AST compared to the control piglets (P < 0.01) at 14 days. There were no significant effects observed on the levels of ALT.

#### Effects of Aps and Gps on Immunoglobulins in Serum

The effect of diet supplementation with Aps or Gps on serum immunoglobulins are shown in **Figure 2**. Compared to the control piglets, Aps piglets showed a higher concentration of IgA (P < 0.05) by 14 days, and IgM (P < 0.05) by 14 and 28 days. Furthermore, Gps piglets showed higher serum levels of IgM (P < 0.05) after 14 days.


TABLE 2 | Effects of the APS and GPS on growth performance of the weaned piglets<sup>1</sup> .

In the same row, values with different letter superscripts mean significant difference (P < 0.05). <sup>1</sup>Con, Aps and Gps represents the piglets supplemented with basal diet, piglets supplemented with the astragalus polysaccharides and piglets supplemented with ginseng polysaccharide, respectively. <sup>2</sup>Pooled SEM; n = 6 per treatment.

# Effects of Aps and Gps on the Jejunal Morphology

The effect of diet supplementation with Aps or Gps on jejunal morphology are shown in **Figure 3**. Compared to control group piglets, Aps or Gps decreased jejunal crypt depth (P < 0.01), and increased jejunal villus length (P < 0.01) and VCR (P < 0.01). In addition, there was no obvious difference regarding the jejunal villus structure between the control group and the piglets supplemented with plant polysaccharides (**Supplementary Figure S1**).

#### Effects of Aps and Gps on Protein Expression Regarding TLR4 Signaling Pathways in Jejunal Mucosa

Expression of proteins involved in the TLR4 signaling pathway in the jejunal mucosa is shown in **Figure 4**. Relative to the control piglets, both Aps and Gps increased (P < 0.05) the expression of jejunal TLR4, MyD88, and NF-κB proteins. This means that the plant polysaccharides promoted the expression of proteins involved in the TLR4 signaling pathway. Furthermore, the levels of proteins involved in the TLR4 signaling pathways in the Gps piglets was higher than those in the Aps piglets.

#### Effects of Aps and Gps on VFAs in Colon Contents

The levels of VFAs in colon contents are shown in **Figure 5**. The concentrations of acetic acid, isobutyric acid, and butyrate in Aps and Gps piglets were higher (P < 0.01) than those in the control piglets. Propionic acid was higher in Aps (P < 0.01) than it was in the control piglets. No significant differences in isovalerate and valerate concentrations were observed among all the different treatments.

# Effects of Aps and Gps on Colonic Microbial Community Composition

The abundance and diversity of colonic microorganisms were described by 16S-rRNA high-throughput sequencing. The colonic microbiota composition is shown in the Venn diagram in **Figure 6A**. A total of 1133 OTUs were shared between the three treatment groups. The Con piglets had 66 unique OTUs, Aps piglets had 46 unique OTUs, and the Gps piglets had 41 unique OTUs. The UPGMA cluster tree and PCA both indicated that the microflora of both plant polysaccharide groups were more similar (**Figures 6B,C**). The alpha diversity (Shannon) was higher in both the Aps and Gps piglets than in the control group (**Figure 6D**).

A taxonomic classification of the microbial composition of colonic contents in piglets revealed that Firmicutes, Bacteroidetes, Proteobacteria, Tenericutes and Euryarchaeota were the dominant bacterial phyla (**Figures 6E,F**). The relative abundance of Firmicutes and Negativicutes in the Aps group was significantly higher (P < 0.05) than that in the control group (**Figure 6G**). At the family level, we observed that Clostridiales, Ruminococcaceae, Lachnospiraceae, Prevotellaceae, and Muribaculaceae were the dominant strains (**Figures 6H,I**). The abundance of Melainabacteria in the control group tended to be lower than that in the Aps and Gps groups

FIGURE 2 | Effect of the Aps and Gps on the serum immunoglobulins in weaned piglets. Con represents the control piglets on d 14 and 28, respectively; Aps represents the piglets supplemented with the astragalus polysaccharide on d 14 and 28, respectively; Gps represents the piglets supplemented with ginseng polysaccharide on d 14 and 28, respectively. <sup>∗</sup>Means significant different (P < 0.05), ∗∗Means significant difference (P < 0.01). Values means n = 6 for the analysis of serum immunoglobulins.

(**Figure 6J**). At the genus level, the results showed that Clostridiales, Subdoligranulum, Faecalibacterium, Prevotellaceae, and Lactobacillus were the dominant genera in all samples (**Figures 6K,L**). The relative abundance of Lachnospiraceae and Anaerostipes in the Aps groups was significantly higher (P < 0.01) than that in control group (**Figure 6M**). At the species level, we found that Ruminococcus sp., Alloprevotella sp., Clostridium butyricum, Lactobacillus gasseri, and L. amylovorus were the dominant species in all samples. L. gasseri and L. amylovorus were higher in the Aps group (P < 0.01) than they were in the Con group (**Figures 6N-P**).

# DISCUSSION

In recent decades, the polysaccharides from medicinal plants have attracted much attention due to their significant bioactivities, such as increasing growth performance, antioxidant

activity, anti-viral activity, and immunomodulatory activities, which make them suitable as antibiotic replacements (Xie et al., 2015). Yuan et al. (2006) reported that dietary Aps increased the average daily gain of weaned piglets. The optimal Aps supplemental level was found to be between 381 mg/kg and 568 mg/kg. Han et al. (2014) reported adding 0.1% ginseng or Acanthopanax senticosus polysaccharides improved the growth performance and feed utilization rate in weaned piglets. Supplementation with Atractylodes macrocephala polysaccharides led to lower diarrheal incidence in early weaned piglets (Li et al., 2011). A study of (Li et al., 2018) showed that the supplementation of Gps modulated the intestinal microflora in rats. Current results showed that two plant polysaccharides increased growth performance, feed conversion ratio, and decreased the incidence of diarrhea in weaned piglets, especially those fed with Aps.

Plant polysaccharides have been confirmed to act as mild immunomodulators by previous studies (Yan et al., 2009; Jiao et al., 2016). Pu et al. (2015) reported that Aps decreased the activities of AST, ALT, and malondialdehyde and enhanced the activity of superoxide dismutase in carbon tetrachloride-treated mice. Liu et al. (2014) reported that 100 mg/kg of Aps not only restored the activity of serum transaminase, but also recovered intestinal pathology and ultrastructure of mice stimulated by docetaxel. Furthermore, Hu et al. (2017) reported that black ginseng significantly decreased the levels of serum ALT and AST in mice stimulated by acetaminophen. These effects were partly due to the antioxidant properties of Aps and Gps, which can scavenge free radicals, improve oxidative stress, and inhibit lipid peroxidation. Similarly, our results showed that ginseng polysaccharide reduced the levels of AST in serum on day 14.

Animal immune response to infection is closely related to immunoglobulin levels. Li et al. (2018) found that Aps alleviated a decreased serum IgG concentration in broilers treated with cyclophosphamide. Similarly, administration of Aps significantly increased the values of immune organ indices and serum IgM and IgG in mice and rats (Li et al., 2009; Meng et al., 2017). Tan et al. (2017) found that dietary supplementation with Aps significantly enhanced the presence of IgG and IgM in sow blood. Lan et al. (2016) indicated that pigs fed with A. membranaceus had higher IgG and IgA concentrations than those fed the control diet. The Aps injection increased the content of IgG, IgM, and IgA in weaned piglets (Xie et al., 2015). In addition, research found that the Gps increased the number of NK cells in the blood of immunosuppressed mice and promoted the expression of perforin and granzyme (Sun et al., 2016). We found that adding Aps could increase IgA and IgM levels in piglets after 14 and 28 days compared with control piglets. The results indicated that Aps and Gps could enhance immune function in piglets.

Wang et al. (2015) evaluated the immunomodulating activities of Aps in chicks. Aps was found to increase jejunal villus height at 8 mg/kg body weight. In rat, Aps in the range of 50–200 µg/ml

significantly increased the number of intestinal epithelial cells (IEC-6), remarkably increased cell migration rate at the wound site and accelerated the recovery of tissue calluses (Zhang et al., 2014). Lv et al. (2017) indicated that the addition of 200 µg/ml Aps significantly increased the density of intestinal cells and the regular distribution of intestinal microvilli by increasing the transcription and expression of cytokeratin (CK18, ZO-2, etc.). Cui et al. (2018) found that A. membranaceus decoction restored villus height and V/C of duodenum, jejunum, and ileum under lipopolysaccharides stimulation in mice. We found that both dietary Aps and Gps enhanced the intestinal healthy development of weaned piglets by decreasing crypt depth and increasing villus height and intestinal gland ratio in weaned piglets.

TLR4 to nuclear factor-kappa B (TLR4-NF-κB) pathway is involved in pathogen identification and infection defense in mammals (Zhang et al., 2017). Aps enhanced immune response including anti-tumor activities (Li et al., 2018). Wang et al. (2017)reported that Aps-mediated immunomodulatory activities in macrophages are involved in TLR4/NF-κB signaling pathways. Zhou et al. (2017) found that Aps stimulated the key nodes in the TLR4-MyD88-dependent signaling pathway, including TLR4, MyD88, TRAF-6, and NF-κB, both in vitro and in mice. Wang et al. (2018) confirmed that the NF-κB mRNA and protein levels were significantly increased in macrophages treated with Aps. These findings were similar to ours, in that both Aps and Gps increased expression of jejunal TLR4, MyD88, and NF-κB proteins, which means that the plant polysaccharides promoted the expression of proteins involved in the TLR4 signaling pathway. Wei et al. (2016) revealed that Aps enhanced the production of the tumor necrosis factor (TNF-α) that is blocked by the inhibitor of TLR4. The above results demonstrated that plant polysaccharides may regulate host immune functioning by activating the TLR4-mediated MyD88-dependent signaling pathway.

Generally, cecal and colonic VFAs were indicators of intestinal health and microbial activity (Bindelle et al., 2009). The cecal and colonic microorganisms ferment polysaccharides to produce VFAs such as acetic acid, propionic acid, and butyric acid. Che et al. (2018) indicated that supplementation with A. membranaceus significantly increased intestinal acetic acid, isobutyric acid, butyrate, and total VFAs in piglets. While, no trials have been conducted to investigate the effects of Gps on the colonic VFAs in weaned piglets. In the present study, piglets supplemented with Aps or Gps tended to have markedly higher concentrations of acetic acid, isobutyric acid, and butyrate.

The colon is the main habitat for microorganisms that affect host growth, development, and health (Pajarillo et al., 2014; Kim et al., 2015). Che et al. (2018) established that supplementation with 2.5–5% of A. membranaceus fiber decreased intestinal pH and increased intestinal levels of VFAs and gut microbial population and diversity, which enhances the health and wellbeing of piglets. In our case, Firmicutes and Bacteroidetes were the advantage bacterial phyla. Lamendella et al. (2011) pointed out that Firmicutes and Bacteroides were the dominant

FIGURE 6 | Summary of microbial community in colon contents of weaned piglets. (A) Represents the Venn diagram summarizing the numbers of common and

fphys-10-00418 April 10, 2019 Time: 20:3 # 8

bacteria in pig feces, which was supported by our data. Moreover, supplementation with Aps and Gps significantly increased the levels of Firmicutes and Negativicutes, which may improve unsaturated fatty acid synthesis in geese (Chen et al., 2018). Furthermore, Aps and Gps increased the relative abundance of Melainabacteria and Selenomonas in the colon. Also, Melainabacteria was found to be beneficial to the host because it fermented plant fibers in the gut and synthesized vitamin B and vitamin K (Di et al., 2013; Soo et al., 2014).

At the genus level, we observed that the levels of Lanchnospiraceae and Anaerostipes in the Aps piglets were significantly higher than those in the control group. Anaerostipes was shown to increase the level of acetic acid, propionic, and isobutyric acid in human feces (Bier et al., 2018). At the species level, our sequencing results indicated that L. gasseri and L. amylovorus were the dominant bacterial species in Aps piglets. Lactobacillus are recognized probiotics. Oh et al. (2016, 2018) reported that L. gasseri showed greater inhibition of nitric oxide production, considerably higher antioxidant activity and inhibition of α-glucosidase activity than other strains, which led to improved immune health of cows by fermenting lactic acid, propionic acid, and modulating pro-inflammatory cytokines. Similar findings were reported by Nobutani et al. (2017), who explained that L. gasseri significantly improved intestinal irritable bowel syndrome severity. L. amylovorus has also been found to modulate the negative regulators Tollip, interleukin-1 receptorassociated kinase, and extracellular heat shock proteins (Hsp72 and Hsp90) to achieve anti-inflammatory effects (Finamore et al., 2014). Furthermore more, Martinez et al. (2013) reported that L. amylovorus promoted short chain fatty acids and ammonia in feces of humans and piglets. In the present study, piglets supplemented with Aps or Gps enhanced the concentration of VFAs by increasing the microbial population and diversity in the colon of piglets. The present results suggest that growth performance was increased by modulating the microbial populations and increasing colonic microflora, which may be relative to the improvement of VFAs and intestinal morphology.

#### CONCLUSION

In conclusion, dietary supplementation with Aps or Gps could improve the growth performance, liver function, and intestinal villus morphology, and may regulate host immune functioning by activating the TLR4-mediated MyD88-dependent signaling pathway, and also enhance the concentration of VFAs in the colon and increase the colonic microbial population and diversity in weaned piglets. Plant polysaccharides may regulate the final fermentation products and the number and diversity of bacteria in the colon to promote the healthy growth of piglets.

#### REFERENCES

Baek, S.-H., Lee, J. G., Park, S. Y., Bae, O. N., Kim, D.-H., and Park, J. H. (2010). Pectic polysaccharides from panax ginseng as the antirotavirus principals in ginseng. Biomacromolecules 11, 2044–2052. doi: 10.1021/ bm100397p

# DATA AVAILABILITY

The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

# ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the care and use of laboratory animals, Institutional Animal Care and Use Committee of Zhejiang Agricultural and Forestry University. The protocol was approved by the Ethics Committee of Zhejiang Agricultural and Forestry University, Hangzhou, China (SYXKzhe 2016-087).

# AUTHOR CONTRIBUTIONS

CY, QH, and GC designed the trials, performed the experiments, and edited the manuscript. KW and YX performed the samples detection. YX, KW, and JL analyzed the data. QH wrote the manuscript, which was edited by CY and GC. All authors read and approved the final manuscript.

# FUNDING

The present research was supported by the Key Research Project of Zhejiang Province (No. 2017C02005) and the National Natural Science Foundation of China (No. 31501985).

# ACKNOWLEDGMENTS

We acknowledge Vegamax Biotechnology Co., Ltd. (Anji, Zhejiang, China) for providing the product of astragalus polysaccharides and ginseng polysaccharides.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphys. 2019.00418/full#supplementary-material

FIGURE S1 | Effect of Aps and Gps on the morphology of jejunum villus in weaned piglets. Con represents the control piglets; Aps represents the piglets supplemented with the astragalus polysaccharide; Gps represents the piglets supplemented with ginseng polysaccharide. Shooting multiples: 200×.


in vivo nitrogen excretion pathways in pigs as reflected by in vitro fermentation and nitrogen incorporation by fecal bacteria123. J. Anim. Sci. 87, 583–593. doi: 10.2527/jas.2007-0717


activities in rats with gastric cancer. Carbohydr. Polym. 78, 738–742. doi: 10. 1016/j.carbpol.2009.06.005



growth performance, immune system, faecal volatile fatty acids and microflora community in weaned piglets. J. Anim. Sci. 97, 133–143. doi: 10.1093/jas/ sky426


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Yang, Han, Wang, Xu, Lan and Cao. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Dietary Modulation of Intestinal Microbiota: Future Opportunities in Experimental Autoimmune Encephalomyelitis and Multiple Sclerosis

#### Yuying Fan and Junmei Zhang\*

Department of Pediatrics, Shengjing Hospital of China Medical University, Shenyang, China

#### Edited by:

Jie Yin, Institute of Subtropical Agriculture (CAS), China

#### Reviewed by:

Junjun Wang, China Agricultural University, China Hong-Hui Wang, Hunan University, China

> \*Correspondence: Junmei Zhang zhangjm@sj-hospital.org

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 20 January 2019 Accepted: 25 March 2019 Published: 16 April 2019

#### Citation:

Fan Y and Zhang J (2019) Dietary Modulation of Intestinal Microbiota: Future Opportunities in Experimental Autoimmune Encephalomyelitis and Multiple Sclerosis. Front. Microbiol. 10:740. doi: 10.3389/fmicb.2019.00740 Multiple sclerosis (MS) is an autoimmune disease that affects the functioning of the central nervous system (CNS). Recent studies on MS and its animal model, experimental autoimmune encephalomyelitis (EAE), have shown that the composition and abundance of microbes in the intestinal microbiota are an environmental risk factor for the development of MS and EAE. Changes in certain microbial populations in the gastrointestinal tract can cause MS in humans, but MS inflammation can be reduced or even prevented by introducing other commensal microbes that produce beneficial metabolites. Other risk factors for MS include the presence of an altered gut physiology and the interaction between the intestinal microbiota and the immune system. Metabolites including short-chain fatty acids (SCFAs), such as butyrate, are the primary signaling molecules produced by the intestinal microbiota that interact with the host immune system, suggesting an association between MS pathophysiology and gut microbiota. In addition, several host microRNAs present in the gut have been found to interact with the intestinal microbial community, these interactions may indirectly affect the neurological system. Increasing evidence has shown that regulation of the intestinal microbiota is an important approach for reducing MS inflammation. Thus, here we review the use of diet to alter the gut microbiota and its application in the treatment and prevention of MS.

Keywords: multiple sclerosis, experimental autoimmune encephalomyelitis, microbiota, dietary habits, shortchain fatty acid, microRNA

# INTRODUCTION

Multiple sclerosis (MS) is a chronic autoimmune disease that causes demyelination and degeneration in the central nervous system (CNS) and is characterized by inflammation and white matter lesions. These lesions consist of CNS cells, such as astrocytes and microglia, along with activated immune cells. The most popular method of studying MS is to use the animal model of experimental autoimmune encephalomyelitis (EAE) (Berer et al., 2011; Lee et al., 2011). More research is needed to understand how an etiological agent might trigger MS and/or EAE.

Certain genetic constitutions increase the susceptibility to MS, and these genetic factors may interact with environmental factors to cause MS (Ascherio, 2013; Koch et al., 2013). Environmental factors include dietary habits (Riccio and Rossano, 2015; Haghikia and Linker, 2018), which can alter the intestinal microbial community, and thus there is considerable variability among individuals. Dietary habits are the primary determinant of the microbial composition and function in the gut (Tilg and Moschen, 2015; Martinez et al., 2016) and thus greatly responsible for shaping the microbial structure. Studies have suggested that the presence of certain gut microbiomes could promote or prevent MS development (Jangi et al., 2016; Liu et al., 2016; Rothhammer and Quintana, 2016). For example, certain bacterial species, such as those of the genus Bifidobacterium and the phylum Firmicutes, are associated with good health, whereas those of the phylum Bacteroidetes have been linked to disease development (Johnson et al., 2017; Azad et al., 2018; Hiippala et al., 2018). Therefore, it would be beneficial to explore how dietary modifications can be used to mitigate the effects of MS. Individualized nutrition plans can be used to restore desirable gut flora populations to reverse microbial dysbiosis, thereby helping to prevent or treat MS. In this review, we explore how the intestinal microbiota affects the pathophysiology of MS and EAE and how this knowledge could be applied to develop therapies for MS treatment.

#### INTERACTIONS BETWEEN THE INTESTINAL MICROBIOTA AND A NORMAL CNS

In humans, the gut microbiota plays a crucial role in regulating physiological processes such as metabolism and immunity. They can also affect brain functions through interactions with the immune, nervous and endocrine systems (Crane et al., 2015; Yano et al., 2015). The metabolites of gut microbiota have the potential to prevent inflammation via their interactions with the CNS. One study confirmed the presence of a relationship between the intestinal mucosa and the brain by demonstrating that polysaccharide A (PSA) from Bacillus fragilis mediate microflora migration (Ochoa-Reparaz et al., 2010; Wang et al., 2014b). This finding suggests the presence of a gut–microbiota–brain axis in the human body. This axis is thought to control the biochemical communication between the CNS and the enteric nervous system (ENS) of the gastrointestinal (GI) tract that appears to be mediated by the intestinal microbiota (Gareau, 2014). The ENS communicates with the CNS via both afferent and efferent pathways of the autonomic nervous system. The afferent pathway communicates signals from the GI tract to the CNS, whereas the efferent pathway communicates signals from the CNS to the GI tract. The vagus nerves are involved in the afferent functions and can recognize the presence of microbial products and cell wall components, and efferent neural signals can affect GI motility, secretion and epithelial permeability. The effects of such efferent signals alter the composition of the intestinal microbiota by changing the physical environment (Bjelobaba et al., 2017; In't Veld et al., 2017).

Changes in the gut microbiota composition can also result from the release of glucocorticoids, mineralocorticoids, or catecholamines by the pituitary and adrenal glands, which are regulated by the hypothalamus. This release can also increase the gut epithelial permeability and immune responses (Bellavance and Rivest, 2014; Yin et al., 2018). Furthermore, the gut microbiota can alter cytokine secretion, thereby regulating neurotransmitter release in the central and peripheral nervous systems. The regulation of neurotransmitter release and other interactions between gut bacteria and the host can lead to the production of biogenic amines and neuroactive substances, such as γ-aminobutyric acid (GABA), dopamine and serotonin, with immunoregulatory effect from host cells (Wong et al., 2015). These neuroactive amines can affect important processes within the body including the digestive, immune, and nervous systems, thereby aiding the maintenance of homeostasis (Rodriguez et al., 2015).

The intestinal microbiota can influence not only the function but also the development of the host immune system (Belkaid and Harrison, 2017). Different subsets of cells in the immune system are affected by different microbiomes. In particular, the gut microbiota plays an important role in the fermentation process, converting indigestible carbohydrates into acetate, propionate and butyrate – the three primary SCFAs. SCFAs are known to decrease inflammation and inhibit the production of histone deacetylase, and recent evidence suggests that SCFAs are instrumental in neuroimmune homeostasis (Erny et al., 2015). Notably, SCFAs can activate the brain's immune response to inhibit histone deacetylase through epigenetic mechanisms. This inhibition then induces regulatory T cell (Treg cell) production in the intestine. Treg cells play a role in maintaining the blood–brain barrier (BBB) and regulating the activity of CNS microglia by simultaneously limiting the microglial size and effect in the brain while also activating them as needed (Erny et al., 2015; Sampson et al., 2016). This suggests that SCFAs can affect CNS homeostasis and maturation. In addition, SCFAs are considered antiinflammatory molecules as they induce cytokine production and interact with the G-protein coupled receptor 45. The intestinal microbiota can release immune antigens, such as peptidoglycan and PSA, that induce immune responses and thus regulate brain function (Desbonnet et al., 2015). SCFA production varies based on the gut microbiota composition, which, in turn, is affected by the type and amount of dietary fiber consumed. The gut microbiota also facilitates amino acid metabolism in the digestive system. Building upon these findings, more studies are investigating how microbial tryptophan metabolites affect immune and neurological processes by functioning as mediators in the human body (Desbonnet et al., 2015).

#### THE ROLE OF THE INTESTINAL MICROBIOTA IN MS PATHOPHYSIOLOGY

Multiple sclerosis is a heterogenous neurological disease that is mediated by the immune system and is attributable to both genetic and environmental factors. It is considered that

MS onset is triggered in genetically susceptible individuals by environmental factors (Belbasis et al., 2015), such as disturbed gut microbiota. A compromised immune system due to alterations in gut microbial composition (e.g., intestinal dysbiosis) can trigger MS development or aggravate its effects. Studies comparing age- and gender-matched groups of MS patients and healthy individuals demonstrated that the gut microbiomes in MS patients were significantly different from those in healthy individuals (Chen et al., 2016). The study reported an increased abundance of Pseudomonas, Mycoplana, Haemophilus, Blautia, and Dorea microbes and a decreased abundance of Parabacteroides, Adlercreutzia, and Prevotella microbes in MS patients (Chen et al., 2016).

#### Intestinal Dysbiosis and EAE

Studies using the animal model of MS, EAE, have reported strong evidence suggesting a link between the gut microbiota and MS development. One study showed that germ-free mice and antibiotic-treated mice had a significantly lower rate of EAE than specific pathogen-free mice (Kennedy et al., 2018). The intestinal microbiota has been shown to play a vital role in maintaining the balance between inflammatory and antiinflammatory responses of the immune system during EAE development. It also plays a critical role in regulating the BBB permeability, limiting astrocyte pathogenicity, activating microglia and expressing myelin genes (Hoban et al., 2016; Rothhammer et al., 2016).

One study on non-obese diabetic mice showed that the intestinal microbiota could affect EAE development and severity. In antibiotic-treated mice, EAE development was delayed and disease progression was stalled, thereby reducing the disease severity (Colpitts et al., 2017). Another group of mice developed a second form of EAE that was more severe than the first form. These mice, but not those with the first form of EAE, exhibited intestinal dysbiosis, and this difference could be observed from an early stage of disease progression (Colpitts et al., 2017). In addition, the commensal microbiota facilitated the responses of inflammatory T helper (Th) cells, such as Th1 and Th17, which prevent future infections. These Th cells could be recruited by B cells and activated by dendritic cells in EAE. These findings indicate that the commensal microbiota are involved in the prevention of EAE exacerbation in diseased mice (Wang et al., 2014a).

Building on these findings of the role of the gut microbiota in EAE onset and severity, several studies have examined how commensal microbes and their products affect the disease. B. fragilis, a gram-negative anaerobe and universal gut microbe, produces PSA on its surface. Pure PSA administration has been found to decrease EAE severity (Ochoa-Reparaz et al., 2010; Wang et al., 2014b), delay the disease onset and decrease its cumulative score, suggesting that it plays a crucial role in protection against EAE. Butyrate is a microbial metabolite of particular interest because it is produced in the gut and has shown to increase the effectiveness and production of circulating Treg cells in mice (Haghikia et al., 2015). The effect of butyrate is important as Treg cells play a crucial role in the development of peripheral tolerance, which prevents the onset of autoimmune diseases. Treg cells can differentiate into antiinflammatory and pro-inflammatory Treg cells. In one study, anti-inflammatory Treg cell levels could be enhanced in mice by SCFA administration, which suppressed the production of pro-inflammatory Treg cells (Haghikia et al., 2015; Chitrala et al., 2017). To further investigate this observation, Haghikia et al. (2015) studied the effects of dietary fatty acids of different lengths on the immune system and consequently on EAE pathophysiology in mice. They found that fatty acids with shorter chains were more effective at regulating and reducing EAE symptoms, thereby decreasing the disease severity. This effect could be promoted by feeding the mice with a diet richer in fiber than normal chow, which increased the propionate and fecal acetate contents responsible for this effect. A positive correlation has been reported between butyrate levels and FoxP3+ Treg cells present in the lymph nodes and spleen of mice. Notably, butyrate can also significantly decrease the production of interferon gamma (IFN-γ), a Th1 cytokine, and increase that of interleukin 17, another Th1 cytokine.

#### Intestinal Dysbiosis and MS

Recent studies comparing intestinal microbiota present in the feces of MS patients and healthy individuals have revealed intestinal dysbiosis in MS patients, in whom certain microbial populations such as Pseudomonas, Mycoplana, Haemophilus, Blautia, Dorea, Pedobacter, and Flavobacterium were enriched and others such as Prevotella, Parabacteroides, Adlercreutzia, Collinsella, Lactobacillus, Coprobacillus, and Haemophilus were depleted compared with those in healthy controls (**Figure 1**; Chen et al., 2016; Tremlett et al., 2016b). Similarly, Miyake et al. conducted a longitudinal study to compare the intestinal microbiota of Japanese patients with relapsing-remitting MS (RRMS) with healthy Japanese people (control). Compared with the control group, the RRMS group demonstrated a moderate level of dysbiosis and significant changes in the abundance of 21 microbial species. However, the bacterial diversity in MS patients remained similar to that in the control patients, which is an interesting finding because the bacterial diversity is known to decrease in other diseases such as inflammatory bowel disorders. To further evaluate the bacterial diversity, Schirmer et al. (2016) examined the effect of Dorea species in the body and suggested that certain Dorea species promote inflammation by facilitating IFN-γ production, metabolizing sialic acids and degrading mucin. Dorea populations have been suggested to be linked to MS development as MS patients exhibit increased Dorea abundance. This finding suggests that Dorea microbes plays pro-inflammatory roles. Notably, Dorea species, similar to T cells, can exhibit both pro-inflammatory and anti-inflammatory properties, the selection between which could be guided by environmental factors, such as the surrounding gut microbiota or nutritional composition.

Research has shown that MS patients exhibit decreased Faecalibacterium abundance compared with healthy individuals (Machiels et al., 2014). This decrease has been found to be attributable to glatiramer acetate administration used for treatment in MS patients. This treatment also decreases the abundance of Bacteroidaceae, Ruminococcus, Lactobacillaceae,

Clostridium, and other Clostridiales microbes. To further investigate this effect, Tremlett performed three experiments on children with MS and found that their gut microbiota were significantly different from that of healthy children (Tremlett et al., 2016a; McKay et al., 2017; Tremlett and Waubant, 2018). Although the abundances of Firmicutes, Archaea, Euryarchaeota, and Proteobacteria (Desulfovibrionaceae) microbes were increased in children with MS, those of Lachnospira (Lachnospiraceae), Verrucomicrobia (Ruminococcaceae), and Fusobacteria microbes were decreased. Furthermore, children with MS who did not exhibit Fusobacteria population in the gut were more likely to relapse, indicating a possible link between Fusobacteria abundance and MS relapse (Tremlett et al., 2016c). Although the study by Tremlett showed no variations in immune markers between children with and without MS, it showed a negative correlation between Bacteroidetes abundance and Th17 level in the MS group and a positive correlation between Fusobacteria abundance and the Treg cell population in the control group.

Branton et al. (2016) examined brain biopsy specimens of 23 MS patients and 21 non-MS patients with other diseases and found that compared with the MS patients, non-MS patients exhibited a greater diversity of bacterial RNA in the cerebral white matter but no significant variations in the gut bacterial diversity. However, the gut microbial structures were different between MS and non-MS patients. The MS patients were also likely to exhibit decreased abundance of Bacteroidetes and Firmicutes microbes in the gut compared with non-MS patients. Notably, healthy digestive systems harbor abundant populations of Firmicutes and Bacteroidetes microbes. Thus, the evidence presented by Branton et al. suggests the presence of differences in the abundance of microbiota at sub-phylum levels between MS and non-MS patients.

An increased abundance of Akkermansia spp. (phylum Verrucomicrobia) in MS patients compared with that in non-MS patients has also been reported. Members of this genus degrade mucin to produce SCFAs, which, as noted above, can aid the suppression of inflammation (Jangi et al., 2016). However, Akkermansia may also possess a pro-inflammatory property as it can upregulate genes associated with innate and adaptive immune responses. This property can affect antigen presentation, adaptive cell adhesion molecules and T cell production. As explained above, butyrate can also regulate T cell production – an increased butyrate level has been linked to a greater Treg cell population. Thus, it is hypothesized that changes in intestinal microbiota composition by certain mechanisms could increase the risk of developing MS. To investigate this hypothesis, Cantarel et al. (2015) compared intestinal microbiota between seven RRMS patients and eight healthy individuals with vitamin D deficiency and found that compared with the healthy individuals, the RRMS patients had significantly lower Faecalibacterium abundance and higher Ruminococcus abundance. Species of these two genera are known to produce butyrate in the human body.

In summary, many studies have demonstrated that MS patients exhibit intestinal dysbiosis. Compared with healthy individuals, MS patients often have lower abundance of Faecalibacterium, Bacteroidaceae and Prevotella populations, suggesting a link between gut microbiota and MS pathophysiology. To deepen our understanding of this

relationship, more large-scale longitudinal studies are needed to evaluate the effect of various endogenous and environmental factors, including age, gender, geographic location, dietary habits and genetic background, on gut microbiota and consequently MS outcome.

# Is the Gut Microbiota–MicroRNA Interaction Related to MS/EAE?

MicroRNAs (miRNAs) are increasingly used as biomarkers for several autoimmune diseases because these molecules are stable in the body and are small. In addition to their function as biomarkers, miRNAs have been suggested to affect MS pathophysiology (Jagot and Davoust, 2016). One study could identify the clinical progression of MS by examining the expression profiles of miRNAs in the body (Mancuso et al., 2015). In MS patients and EAE subjects, miRNAs have shown to mediate the upregulation of miR-29b, miR-141, miR-200a, miR-155, miR-223, miR-326, let-7e, and miR-448 expression along with a significant downregulation of miR-15a/16-1 and miR-15b expression in CD4+ T cells (Ifergan et al., 2016; Liu et al., 2017; Chen et al., 2018). In addition, miRNA has been shown to decrease the miR-20b expression and significantly increase the miR-21 and miR-590 expression in Th17 cells compared with those in Th1, Th2 and inducible Treg cells (Chen et al., 2018). This suggests that miRNAs and the gut microbiota interact to regulate disease progression in the body. Studies have suggested that miRNAs produced by host cells regulate the gut microbiota and that the gut microbiota, in turn, can trigger miRNA production in the host.

The relationships between the gut microbiota and miRNAs have been examined in liver diseases, cancers and intestinal epithelial disease. **Table 1** describes the findings of relevant articles. Notably, no study has yet analyzed the effect of the gut microbiota–miRNA interaction on MS and EAE, and there is little knowledge on how gut microbiota affects brain function. Nonetheless, knowledge from studies on the gut– brain axis can be applied here as there are notable overlaps between some parts of the axis and miRNA functions. These overlaps include the immune system, hypothalamic–pituitary– adrenal (HPA) axis and vagus nerve. In addition, MS severity in humans could be limited by microbial metabolites, such as butyrate, which triggers miR-375 expression. Microbial metabolites could also have beneficial effects via the regulation of tryptophan metabolism.

A recent study discovered a new mechanism underlying MS pathogenesis that involves exosome miRNAs from the plasma (Kimura et al., 2018). Exosomes miRNAs, such as let-7i, circulate through the bloodstream and prevent Th1 and Th17 cells from differentiating during MS onset (Kimura et al., 2018; Tse et al., 2018). This finding suggests that the transfer of extrinsic miRNAs by exosomes is critical for the onset of autoimmune diseases. The relationships between miRNAs and gut microbes likely impact the pathophysiology of MS and EAE. The suggested mechanism


is shown in **Figure 2**. This theoretical mechanism will need to be tested in the future.

The intestinal microbial composition varies based on dietary habits. As mentioned above, compared with healthy controls, MS patients exhibit intestinal dysbiosis with decreased abundance of Clostridium, Bacteroidetes and Adlercreutzia microbes in addition to other microbes known for their role in regulating the body's immune responses (Miyake et al., 2015). Clostridium species are the major producers of SCFAs, which aid Treg cells in suppressing inflammation in the body. Fecal miRNAs also influence the gut microbial composition, which, in turn, influences the miRNAs present in intestinal epithelial cells. The effects of gut microbiota–miRNA interaction overlap with those of the gut microbiota–hippocampus axis and are related to the development of the hippocampus, cognitive function and neuropsychological functions, such as anxiety. Studies have shown a link between MS/EAE and the following miRNA molecules: miR-29b, miR-141, miR-200a, miR-155, miR-223, miR-326, miR-448, miR-15a/16-1, miR-15b, miR-20b, miR-21, miR-590, miR let-7e, and miR let-7i (Chen et al., 2018; Ntranos et al., 2019). Taken together, these results suggest that MS/EAE development is significantly affected by the interactions between certain miRNAs and gut microbiota.

# MICROBIOTA-TARGETED THERAPIES FOR MS

Many modern holistic approaches for health care focus on promoting beneficial gut microbiota (Gibson et al., 2017). Microbiota-targeted treatments include dietary modifications, fecal microbiota transplants (FMTs) and administration of probiotics and prebiotics. However, as with many therapies, microbiota-targeted treatments are not effective in all individuals

and can also cause unintentional adverse effects, which must be minimized. Therefore, further studies are warranted to develop treatments with fewer adverse effects.

#### Dietary Modifications

fmicb-10-00740 April 12, 2019 Time: 17:35 # 7

Dietary habits could be the primary determinant of gut microbial composition and function (Tilg and Moschen, 2015; Martinez et al., 2016), consequently shaping the microbial structure (**Figure 3**). In general, hypercaloric, high-animal fat Western diets may accelerate anabolism, alter gut microbiota composition and result in intestinal dysbiosis. Conversely, a vegetarian diet rich in fiber promotes gut eubiosis (Riccio and Rossano, 2018). Mice with high-body fat percentages that were fed a typical Western diet consisting of high levels of salt, saturated fat, protein, sugar and calories showed increased EAE severity (Hucke et al., 2016; Hammer et al., 2017); the gut flora in the mice was altered, with a notable

increase in the levels of free pro-inflammatory fatty acids in the plasma. This finding suggests that Western diets reduce SCFA production by desirable gut microbiota. This reduction in SCFAs, which facilitate the production of protective Treg cells, could thus increase the incidence of autoimmune diseases (Haghikia et al., 2015).

In contrast, other studies have shown that low-calorie diets comprising high levels of fruits, vegetables and fish promote beneficial gut microbiota and reduce inflammation in the body (Riccio and Rossano, 2018). Diets that are effective at preventing EAE, reducing inflammation and increasing neuroprotection include the ketogenic diet, intermittent fasting and calorierestricted diets. In a recent study, calorie-restricted diet was found to be more effective in reducing the symptoms of EAE than ketogenic diet (Dupree and Feinstein, 2018; Kap et al., 2018). However, this result has not yet been confirmed, and further studies to investigate the different effects of specific diets are ongoing.

Studies on MS should not be limited to studying the abundance of gut microbiota alone. Vitamin D, a nutrient that promotes Treg cell differentiation and mediates gut microbiome balance, should also be examined. Reportedly, vitamin D levels in gut microbiota vary in different diseases, such as MS (Riccio and Rossano, 2018). Similarly, a close relationship between dietary tryptophan and EAE severity in mice has been suggested, which needs to be confirmed in a future study.

#### Probiotics

The use of probiotics, which produce essential vitamins and cofactors not naturally produced by the host, is a popular treatment with several health benefits, such as improvement of the immune system and inhibition of non-commensal microbiota growth (Li P. et al., 2017). The administration of probiotics is also thought to reduce the symptoms of CNS diseases. For instance, in one study, the oral administration of Lactobacillus paracasei and two strains of L. plantarum was effective in preventing EAE development in mice (Li P. et al., 2017). However, very few studies have evaluated the effects of probiotics on MS patients.

# Prebiotics and Polyphenols

Prebiotic substances also have a significant effect on microbiota. Foods with a prebiotic effect include colorful fruits, cocoa and tea leaves. Several fermentable carbohydrates also have prebiotic properties, but two oligosaccharides, fructans, and galactans, are known to be the most beneficial for Bifidobacterium growth. A high-fiber diet is commonly recommended due to its health benefits (Liu et al., 2015). Its protective effects in MS/EAE are associated with the ability of gut microbiota to produce SCFAs by fermenting dietary fiber. Some reports have suggested that circulating butyrate directly alters CNS function (Li W. et al., 2017; Zhang et al., 2018). Braniste et al. showed that the BBB permeability was significantly increased in germ-free mice, but after administering butyrate-producing bacteria, Clostridium tyrobutyricum, or an oral gavage of sodium butyrate, the BBB permeability was restored to the level present in pathogen-free mice (Braniste et al., 2014).

Similarly, another study suggested that a diet rich in polyphenols can protect against EAE (Miyake et al., 2006). In this study, mice with EAE were treated with a polyphenol extracted from Jatoba, a medicinal plant found in South America, and the polyphenol extract was found to facilitate the suppression of Th1 immunity. These results suggest that MS patients can benefit from more dietary polyphenols in addition to probiotics and prebiotics.

### FMT and Stool Substitute Transplant Therapy

Fecal microbiota transplant involves the replacement of the entire host gut microbiome to restore the desirable microbial balance. It is being increasingly used to combat a range of diseases. Benefits of FMT have been reported in three MS patients, and another study investigated how FMT affects the symptoms of secondary progressive MS (Makkawi et al., 2018). Stool substitute transplant therapy (SSTT) is similar to FMT. More research is required to determine the effects of FMT and SSTT on autoimmune diseases, such as MS.

#### Other Therapies

Broad-spectrum antibiotics have been used to modify gut microbiota and treat EAE by delaying EAE development. Antibiotics can reduce the effect of EAE by modifying the T cell population in the gut assisted lymphoid tissue and nearby lymphoid tissues. For each unit of increase in IL-10 production, a simultaneous increase in Foxp3+ Treg cell production has been demonstrated (Nakamura et al., 2016; Shi and Mu, 2017).

A new alternative therapy uses helminths, which are eukaryotic worms, to combat auto-immune diseases. Although there is evidence of success in treating EAE using helminths, there is limited evidence to support the use of this therapy in MS patients.

# REFERENCES


#### PERSPECTIVES

The theory that the intestinal microbiota plays a crucial role in regulating the gut–brain axis, influencing disease development and maintaining human health is gaining support. However, understanding the detailed effects of the gut microbiota in relation to MS pathophysiology requires more research. In particular, little knowledge is available regarding the mechanisms implicating the gut microbiota in MS onset and progression. These mechanisms could be elucidated by studying the gut microbiota-mediated miRNA–MS/EAE axis.

As an increasing number of studies confirm the association between gut microbiota and neuroimmune inflammatory diseases, treatment involving alteration of gut microbiota will become more appealing. However, the widespread application of such treatments may not happen soon as there is limited evidence of their potential beneficial effects in humans. Thus, more large-scale trials focusing on identifying the microbes that are integral to MS development and understanding how they function are warranted. The results of such studies will assist in developing effective treatments to prevent MS by altering the gut microbiome.

# AUTHOR CONTRIBUTIONS

YF finished the first draft of the manuscript. JZ critically revised the manuscript. Both authors approved the submission of the manuscript.

#### FUNDING

This work was supported by National Natural Science Foundation of China (81501299) and the National Key Research and Development Program of China (No. 2016YFC1306203).


clinical treatment of cardiovascular diseases. Oxid. Med. Cell Longev. 2017, 1–6. doi: 10.1155/2017/6504518


a striking depletion of species belonging to clostridia XIVa and IV clusters. PLoS One 10:e0137429. doi: 10.1371/journal.pone.0137429


in pediatric multiple sclerosis and controls. BMC Neurol. 16:182. doi: 10.1186/ s12883-016-0703-3


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Fan and Zhang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Biological Effects and Applications of Chitosan and Chito-Oligosaccharides

Guiping Guan1,2, Md. Abul Kalam Azad3,4, Yuanshan Lin<sup>1</sup> , Sung Woo Kim<sup>2</sup> , Yun Tian<sup>1</sup> , Gang Liu1,3 \* and Hongbing Wang<sup>5</sup> \*

<sup>1</sup> College of Bioscience and Biotechnology, Hunan Agricultural University, Changsha, China, <sup>2</sup> Department of Animal Science, North Carolina State University, Raleigh, NC, United States, <sup>3</sup> Hunan Province Key Laboratory of Animal Nutritional Physiology and Metabolic Process, Key Laboratory of Agro-ecological Processes in Subtropical Region, Institute of Subtropical Agriculture, Chinese Academy of Sciences, National Engineering Laboratory for Pollution Control and Waste Utilization in Livestock and Poultry Production, Changsha, China, <sup>4</sup> University of Chinese Academy of Sciences, Beijing, China, <sup>5</sup> Hunan Institute of Animal Husbandry and Veterinary Medicine, Changsha, China

The numerous functional properties and biological effects of chitosan and chito-oligosaccharides (COS) have led to a significant level of interest, particularly with regard to their potential use in the agricultural, environmental, nutritional, and pharmaceutical fields. This review covers recent studies on the biological functions of COS and the impacts of dietary chitosan and COS on metabolism. The majority of results suggest that the use of chitosan as a feed additive has favorable biological effects, such as antimicrobial, anti-oxidative, cholesterol reducing, and immunomodulatory effects. The biological impacts reviewed herein may provide a new appreciation for the future use of COS.

#### Edited by:

Yuheng Luo, Sichuan Agricultural University, China

#### Reviewed by:

Guillermo Tellez, University of Arkansas, United States Tatiana V. Kirichenko, Ministry of Health of the Russian Federation, Russia

#### \*Correspondence:

Gang Liu gangle.liu@gmail.com Hongbing Wang hongbingwanggg@gmail.com

#### Specialty section:

This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology

Received: 17 January 2019 Accepted: 11 April 2019 Published: 07 May 2019

#### Citation:

Guan G, Azad MAK, Lin Y, Kim SW, Tian Y, Liu G and Wang H (2019) Biological Effects and Applications of Chitosan and Chito-Oligosaccharides. Front. Physiol. 10:516. doi: 10.3389/fphys.2019.00516 Keywords: chitosan, chito-oligosaccharides, biological activity, application, microbiota

# INTRODUCTION

The health and performance of modern intensively reared farm animals, such as poultry and swine, are promoted with the aid of various feed additives. The non-toxic linear polysaccharide chitosan made up of β-1-4 linked D-glucosamine and N-acetyl-D-glucosamine units, and its derivatives [chito-oligosaccharides (COS)] are comparatively novel and less frequently used as feed additives in animal nutrition. Chitosan, in varying levels of deacetylation, forms the basis of the common natural substance chitin, found in the exoskeletons of insects, crabs, and shrimps (Koide, 1998; Younes and Rinaudo, 2015). The biocompatibility and biodegradable properties of chitin have led to numerous reports of biomedical applications (Park and Kim, 2010). The major applications of chitin are in the production of monosaccharides, which constitute the primary dietary supplement in the United States, and in the relief of osteoarthritic pain (Aam et al., 2010). The chemical distinction between chitin and chitosan is based solely on the extent of the acetylation of the D-glucosamine units, with chitin being over 70% acetylated and chitosan under 30% acetylated. However, the application of chitin in living systems is comparatively limited by its insolubility in water, whereas chitosan is soluble in acidic solutions (Shahidi et al., 1999). Although chitosan is less commonly found in nature, its presence along with that of chitin in the cell walls and septa of filamentous fungi and yeast (Muzzarelli et al., 2012). The production of commercial chitosan via deacetylation of chitin involves the high-temperature treatment of chitin with a strong solution of sodium hydroxide (Singla and Chawla, 2001; Lemma et al., 2016). Approximately 150,000 tons of commercially applicable chitosan is produced annually by the conversion of chitin acquired as a by-product of seafood production (Fernandez and Ingber, 2014; Younes et al., 2014).

The de-polymerization of chitosan by acid hydrolysis, physical hydrolysis, and enzymatic degradation results in the production of COS (Lodhi et al., 2014). The weak glycosidic bonds in chitosan facilitate cleavage in the presence of hydrolyzing agents to generate chitosan oligomers incorporating various monomer units (Kim and Rajapakse, 2005). COS (also termed chitosan oligomers or chito-oligomers) are chitosan with an average molecular weight (MW) under 3.9 kDa and containing less than 20 monomer units per polymer chain (Lodhi et al., 2014). The low MW and the solubility of COS generates significantly more interest than the precursor species, hence with increasing commercial production.

The numerous biological properties of COS suggest a variety of possible uses in a wide range of areas, such as agriculture, cosmetics, food, and medicine (Xia et al., 2011). As no single type of chitosan or COS displays all of the observed biological activities, an increasing number of studies have aimed to examine the specific activities of this group of compounds. Furthermore, the various structures and physicochemical activities of chitosan derivatives and enzymatic products may furnish new biological activities or help generate a new understanding of previously known bioactive substances.

This review examines recent research focused on two important biological properties of chitosan and COS as applied to swine nutrition, thus providing a new understanding of these biological functions and pointing the way to the development of chitosan and oligosaccharides as swine-feed additives.

#### ANTIMICROBIAL PROPERTIES AND REGULATION OF MICROBIOTA

The antimicrobial properties of COS and chitosan are well known. However, the exact mechanism of the antimicrobial activity of COS and chitosan is still unknown. Several studies have been suggested that the actual processes of antimicrobial activity could be occurred by changing the bacterial membrane permeability, cytoplasmic membrane barrier function or nutrient transport. In addition, the mechanism of antimicrobial activity mostly depends on MW, the degree of de-acetylation (DD), type of bacterium, pH, and the concentration of active compounds connected to chitosan and its derivatives (Jarmila and Eva, 2011; Guan et al., 2016).

The antimicrobial activity of chitosan or its derivatives usually depends on various factors including MW, DD, and other physicochemical characteristics along with microorganism type (Gram-negative or Gram-positive) (Liaqat and Eltem, 2018). COS has been broadly studied to improve the antimicrobial activities both in vivo and in vitro. The COS with positively charged can bind or absorb into the cell wall of microbes through negatively charged components of microorganisms present in the microbial cell. According to this concept, a study was conducted with 180 weaning pigs with average body weight were divided into five treatment groups and fed a control diet, treatment diet with COS (200, 400, or 600 mg/kg), and a diet with Colistin Sulfate (CSE) for 14 days. Results revealed that COS supplementation increases the population of Bifidobacteria and Lactobacilli, and decreases S. aureus in the cecum of the weaning pigs. The authors concluded that two possible mechanisms for the observed antimicrobial activity of COS. The first possible explanation was the positive charge on the NH<sup>3</sup> <sup>+</sup> group of the COS glucosamine monomer interactions with a negatively charged microbial cell membrane. Secondly, COS may exert an indirect influence via increasing the populations of Bifidobacteria and Lactobacilli and the exclusion of S. aureus (Yang et al., 2012). Similarly, Kong et al. (2014) have found that dietary COS (0.5 g/Kg) supplementation in weaned Huanjiang minipiglets increased the microbial population of Bifidobacterium spp., Bifidobacterium breve, Faecalibacterium prausnitzii, and Lactobacillus spp. in the ileum and colon. In addition, the number of Fusobacterium prausnitzii, Methanobrevibacter smithii, and Roseburia were increased in the colonic content of the dietary COS supplemented piglets. Furthermore, dietary COS supplementation decreased the microbial population of Firmicutes and Streptococcus in the ileum and colon, and Bacteroides fragilis, Clostridium coccoides, C. leptum subgroup., and Eubacterium rectale in the ileum, and Escherichia coli in the colonic content of the treated piglets. Of note, chemically modified COS has also been improved anti-microbial properties. For example, NO-releasing secondary amine-modified COS has been reported to readily penetrate the biofilm and associated with Pseudomonas aeruginosa and resulting in the effective killing of the P. aeruginosa biofilm through the effect of the released NO with the minimal inhibitory content (MIC) of 200 µg/mL (Lu et al., 2014).

In addition to antimicrobial activity, chitosan and COS have been shown to possess anti-fungal and anti-viral activities. However, with numerous studies attesting to the antifungal or anti-viral action of COS against a range of fungi and virus, the outcomes of studies on the anti-fungal or anti-viral properties of COS have been found somewhat inconsistent. These inconsistencies may be due to differences in the purity, quality, and properties of the COS used and/or the use of different microorganisms and methodologies. Although less potent than chitosan, COS has been shown to exhibit anti-fungal effects against several types of fungus including Saccharomyces cerevisiae, Aspergillus niger, Trichophyton rubrum, and Candida spp. with the MIC of 1.3 mg/mL (Seyfarth et al., 2008; Mei et al., 2015; Muanprasat and Chatsudthipong, 2017). Thus the potential clinical application of COS is desirable for its biocompatibility, biodegradability, and safety.

The gut microbiota is a very crucial factor that interacts with the host physiology and health (Niewold et al., 2010; Li et al., 2018); Wang et al., 2018). Alteration of gut microbiota plays an important role in host health, including vitamin synthesis, improve digestion, and promotion of angiogenesis and nerve function (Soler et al., 2014; Azad et al., 2018a; Wang et al., 2018). Chitosan and its derivatives have shown advantageous biological function in gut microbiota alteration. A study aimed to evaluate the effects of different levels of dietary COS (100, 200, and 400 mg/kg) supplementation during weaning period on growth performance, fecal shedding of E. coli and Lactobacillus, nutrient digestibility and small intestinal

morphology. COS supplementation revealed an increase in the amount of fecal Lactobacillus along with a decrease in the amount of E. coli (Liu et al., 2008). Similarly, pigs were given 400 mg/kg supplementary COS in a study by Yang et al. (2012) also displayed enhanced populations of Bifidobacteria and Lactobacilli in the caecum on the 7th day after weaning compared to those weaned on the basal diet. On the 14th day after weaning, the same study also revealed a higher quantity of Bifidobacteria in the caeca of pigs given 600 mg/kg COS relative to those given the basal diet (Yang et al., 2012).

Chitosan or COS have been shown a potential activity on anti-obesity by altering the gut microbiota populations. In an obese animal model, (Egan et al., 2015) aimed to evaluate the effect of prawn shell derived chitosan (1000 ppm) in a pig model. The study was carried out with 125 days of age pigs (70 ± 0.09 kg body weight) were a fed basal diet or treatment diet (1000 ppm chitosan with basal diet) for 63 days. Results revealed that dietary chitosan supplementation reduced the populations of phylum Firmicutes in the colon and of Lactobacillus spp. in both the colon and the caecum, whereas the amounts of the Bifidobacteria genera in the caecum increased. Furthermore, sows fed with dietary chitosan exhibited lower feed intake and final body weight (Egan et al., 2015). Yan and Kim (2011) reported an enhanced blood lymphocyte count along with a decreased fecal population of E. coli in weaned pigs given 3 g/kg dietary COS (Yan and Kim, 2011), whereas Wan et al. (2017) reported that 100 mg/kg COS both enhanced the ileal Bifidobacterium population and decreased the E. coli and total bacteria populations of the colon and caecum (Wan et al., 2017). In earlier, corresponding results were obtained by Wang et al. (2009) who reported that dietary supplementation with COS (0.50%) decreased the populations of fecal E. coli in growing pigs, whereas the count of fecal Lactobacillus was unaffected. The glucosamine monomer unit of COS may interact with negative charges on the microbial cell membranes, resulting in the leakage of the cells' internal constituents. In addition, COS has also been shown an indirect impact on the cell membrane by promoting Bifidobacteria and Lactobacilli populations which tend to exclude S. aureus (Yang et al., 2012).

A recent study examined the impact of dietary supplementation with COS of low MW (20,000 to 30,000 Da) on the gut microbiota of piglets and found significant differences in the composition of the gut microbiomes of the pigs, while pigs were given 50 mg/kg COS supplementation for 28 days of experimental periods along with the control group and in a group given antibiotics (Yu et al., 2017). Compared to the control group, the relative abundance of Prevotella increased, whereas the abundance of Lactobacillus decreased in both the COS supplemented group and antibiotic groups. In addition, the relative abundance of both Succinivibrio and Anaerovibrio were increased in the COS supplemented group and decreased in the antibiotics group (Yu et al., 2017). According to this study, microbial function prediction suggests that more pathways in cofactor and vitamin metabolism would be more enriched by the presence of COS compared than by the presence of antibiotics or the basal diet.

#### IMMUNOSTIMULATORY, IMMUNOREGULATORY AND ANTI-INFLAMMATORY PROPERTIES

The immune system is composed of innate immunity and adaptive immunity which plays an important role to prevent the foreign pathogenic substance from the body. In immune function enrichment, immunostimulating medicine, and nutraceuticals are of particular interest (Soler et al., 2014; Azad et al., 2018b; Liaqat and Eltem, 2018). Dietary COS has been demonstrated effective and promising immunostimulator activities in both in vivo and in vitro models. According to Zhang et al. (2014), the immunostimulatory properties of COS may occur via interaction with membrane receptors on the macrophage surface and depend on toll-like receptor 4 (TLR4). A significant dose- and MW dependent immunoregulatory responses have been observed in the presence of COS with MWs of 3 and 50 kDa. The presence of COS can boost the expression of the gene molecules essential to the NF-κB and AP-1 pathways and trigger protein phosphorylation in the RAW264.7 macrophage. In this study, COS with a MW of 3 kDa demonstrated more promise as a new treatment for immune suppressive conditions with potential application in vaccines (Zhang et al., 2014). This also suggests the potential use of COS as a component of functional food designed to combat diet-related and age-related conditions. The clinical testing of immunostimulation by orally administered COS has already conducted.

Immunomodulatory feed additives such as chitosan or its derivatives may act as alternatives to antimicrobial growth promoters in pig production. Therefore, (Yin et al., 2008) designed an experiment of the pig model to examine the immunoregulatory function of early weaned piglets. They fed 0.025% of dietary COS along with 0.2% of galacto-mannan-oligosaccharides (GMOs) or 0.11% of lincomycin. After the end of 2 weeks experimental period, the results revealed that the weaning challenge led to reduced levels of serum antibodies and cytokines, the administration of 250 mg/kg COS resulted in higher expression of the IL-1β gene in the lymph nodes and jejunal mucosa and higher concentrations of interleukins IL-2, IL-6, and IL-1β and immunoglobulins IgA, IgG, and IgM in the serum. Hence, the authors concluded that dietary COS promotes the cell-mediated immune reaction in early weaned piglets by regulating the generation of antibodies and cytokines (Yin et al., 2008). Similarly, Wan et al. (2017) indicated that the administration of 100 mg/kg dietary COS enhanced superoxide dismutase (SOD) and catalase (CAT) activities, total antioxidant capacity and the serum levels of IL-6, IgG, and TNF-α. Additionally, a 26.59% decrease in the concentration of serum malondialdehyde (MDA) was noted for the pigs given dietary COS. A recent study by Li et al. (2017) reported that the dose-dependent dietary chitosan (100, 500, 1000, and 2000 mg/kg feed) enriched the linear or quadratic levels

of prostaglandin E2, leukotriene B4 and arachidonic acid in piglets given dietary chitosan. Linear or quadratic enhancements in the activity of serum cytosolic-phospholipase A2 were also observed, as well as a quadratic enhancement in the activity of COX-2 and a linear enhancement in the activity of 5-lipoxygenase. These observations suggest that arachidonic acid metabolism is modulated by chitosan in a dose-dependent manner, which may partly explain why chitosan influences the immune function of weaned piglets through the AA pathway (Li et al., 2017).

Sun et al. (2009) evaluated the effects of chitosan (250 mg/kg, MW = 10<sup>3</sup> to 10<sup>4</sup> ), GMOs (2000 mg/kg) on the growth performance, serum immune parameters of 28-day weaned piglets challenged with pathogenic E. coli. Feed gain ratio and IgA, IgG, and IgM levels were increased in an E. coli challenged model by dietary COS supplementation. Similarly, Xiao et al. (2013, 2014) investigated the effects of dietary COS on growth performance, jejunal morphology, jejunal mucosal secretory IgA, occludin, claudin-1, and TLR4 expression in weaned piglets challenged by enterotoxic E. coli. A total of thirty piglets were fed a corn-soybean diet as a control diet, 50 mg/kg chlortetracycline, or 300 mg/kg COS. After 21 days experimental period, the findings showed that dietary COS and chlortetracycline reduced FCR, villus height, crypt depth, and the TLR4mRNA expression but increased the villus length, villus length/crypt depth, and goblet cells. In addition, the secretory IgA was observed higher in the dietary COS group compared with the other groups. Therefore, the authors concluded that chitosan showed similar effects with antibiotics in promoting the growth and reducing the intestinal inflammation in weaning piglets. Later, the same authors used a similar model to evaluate the effects of dietary COS on intestinal inflammation. The finding supports the previous work, and additionally, dietary supplementation with 300 mg/kg COS improved the mRNA expression of IL-1β and IL-6 in the jejunal mucosa (Xiao et al., 2013, 2014). Thus it proves that as a feed additive, dietary chitosan may influence different mechanism to alleviate inflammation in weaning piglets.

Several studies have examined the impacts of dietary chitosan supplements on antioxidative enzymes and stress hormones, and on humoral and cellular immune function in weaned piglets. Li et al. (2013) found a dose-dependent quadratic enhancement in the levels of serum IgG, and a dose-dependent linear or quadratic enhancement in the levels of serum specific ovalbumin IgG (Fan et al., 2013; Li et al., 2013). However, the levels of serum IgA and IgM were unaffected (Fan et al., 2013). The same authors reported a linear dose-dependent reduction in the levels of serum adrenocorticotropic hormone along with a dose-dependent linear or quadratic reduction in the levels of serum cortisol. Enhancements in the levels of CAT, SOD, and serum glutathione peroxidase with increasing chitosan were also noted, demonstrating that dietary chitosan enhances the activity of antioxidative enzymes and reduces weaning stress in piglets (Li et al., 2013). According to Huang et al. (2016) the impacts of dietary COS (300 µg/kg) on intestinal inflammation and the NF-κB signaling pathways in an LPS-challenged piglet model demonstrated the significant easing of LPS-induced intestinal injury. Furthermore, dietary COS reduced serum concentrations of IL-6, IL-8, and TNF-α, decreased intestinal levels of pro-inflammatory cytokine mRNA and increased levels of anti-inflammatory cytokine mRNA relative to the control group. The protein expression of IKKα/β, IκB, and phospho-NF-κB p65 also reported for the LPS-challenged piglets in the COS diet group (Huang et al., 2016). Therefore, dietary chitosan or its derivatives may play a crucial role in oxidative stress, intestinal inflammatory response, as well as by the inhibition of NF-κB signaling pathways under an inflammatory stimulus.

The anti-inflammatory properties of COS have been widely reported in the view of the potentially damaging effects of a disproportionate and protracted inflammatory response in a range of illnesses (Ngo et al., 2011). Efforts to explain the anti-inflammatory properties of chitosan and COS have focused on numerous potential mechanisms, for example, the acid hydrolysis of chitosan to glucosamine hydrochloride, sulfate, phosphate or other salts by salt conversion. Alternatively, the suppression of LPS-induced inflammatory gene expression by COS has been linked to the decreased nucleus translocation of the nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) (Li et al., 2011). Hence, COS effectively reduces both inflammations due to infection by enterotoxigenic E. coli and LPS-induced vascular endothelial inflammation (Liu et al., 2016; Xiao et al., 2016). Moreover, COS significantly reduces the LPS-induced phosphorylation of p38 mitogen-activated protein kinase (MAPK) and extracellular signal-related protein kinase <sup>1</sup>/<sup>2</sup> and may hinder the activation of NF-κB and activator protein-1 (AP-1). The results of a study by Xiao et al. (2013) in weaned piglets challenged with enterotoxigenic E. coli revealed that the dietary COS (0.03%) and antibiotic (chlortetracycline) had similar beneficial effects in reducing intestinal inflammation and promoting growth. Similar to an antibiotic, dietary COS supplementation increased, the concentration of intraepithelial lymphocytes, goblet cells, villus length, villus length to crypt depth ratio, occluding protein and secretory IgA protein expression, decreased TLR4 mRNA expression. Finally, the authors concluded that COS has the potential against inflammation Xiao et al. (2013). Furthermore, dietary supplementation of COS can activate the expression of inducible nitric oxide synthase and cyclooxygenase-2 (COX-2) induced by TNF-α in synoviocytes was inhibited via COS-reduced AMPK activation, resulting in the attenuation of synovial inflammation (Kunanusornchai et al., 2016).

Chronic inflammation of the gut is involved in various forms of inflammatory bowel disease (IBD), such as Crohn's disease and ulcerative colitis (Herfarth, 2013; Hill, 2014; Guan and Lan, 2018; Hong and Piao, 2018; Zhang Y. et al., 2018). The occurrence of IBD has steadily increased in certain parts of the world in the last 40 years, perhaps as a result of changing dietary practices, including the preference for low-fiber diets (Rose et al., 2007; Umakanthan et al., 2016; Weichselbaum and Klein, 2018; Zhang X. et al., 2018). The known anti-inflammatory properties of COS have therefore prompted scientists to examine its potential as an adjuvant

treatment for inflammatory illnesses. For example, tissue damage and reduction in colon length have been ameliorated and the inflammation of the colonic mucosa has been prevented in mice given COS orally, suggesting the potential application of COS as a functional food for individuals with IBD (Azuma et al., 2015). A recent study was carried out to investigate the effect of dietary COS supplementation on pig growth. The results revealed that the pigs consumed COS for 21 days increased average daily body weight gain compared to those in the control group. Besides the improvement of the activities of superoxide dismutase (SOD), catalase (CAT), and total antioxidant activity dietary COS also increased the IL-6, TNF-α, and IgG concentrations in the serum. In addition, dietary COS were found to increase the total bacterial populations of Bifidobacterium in the ileum and colon. Finally, the outcomes suggested that the growth of pigs during weaning can be accelerated by dietary COS supplementation because dietary COS can enhance the antioxidant and immune properties, as well as intestinal development (Wan et al., 2017). Therefore, the potential application of dietary COS should be further investigated as an anti-inflammatory compound in animal diets, food, and pharmaceutical industries.

#### EFFECTS ON PERFORMANCE, DIGESTION AND INTESTINAL STRUCTURE IN SWINE NUTRITION

The effectiveness and nutritional significance of COS as an animal-feed additive are summarized in **Table 1**. Importantly, a number of beneficial impacts have been noted during the weaning stage—a vital time for growing pigs, during which they are subject to environmental, immunological and nutritional pressures that frequently exert detrimental effects on a range of metabolic functions, resulting in digestive illnesses, diarrhea, limited growth and increased mortality (Swiatkiewicz et al., 2015; Oliveira et al., 2017; Zhao et al., 2017). For instance, (Liu et al., 2008) performed a study aimed to evaluate the effect of dietary COS supplementation (100, 200, and 400 mg/kg) on growth performance, fecal shedding of E. coli and Lactobacillus, nutrient digestibility, and small intestinal morphology in weaned pigs. Results revealed that dietary COS (100 and 200 mg/kg) supplementation improved average body weight gain (BWG), beneficial effects on feed intake (FI), and feed conversion ratio (FCR) compared to the control pigs. Additionally, dietary COS supplementation decreased diarrhea scores and increased Lactobacillus counts than those from control pigs (Liu et al., 2008). Yang et al. (2012) found the pigs were given dietary COS from the first to seventh day after weaning to have increased ADG and ADFI values compared to the control group. Although dietary supplementation with COS (400 mg/kg or 600 mg/kg) from the first to the 14th day after weaning resulted in enhanced ADG and gain/feed ratio, no effect on the crypt depth and villous height of the ileum, jejunum or duodenum was observed (Yang et al., 2012). The aim of an experiment by Zhou et al. (2012) was to assess the growth performance, nutrient digestibility, and incidence of diarrhea in weaned pigs. According to their purposes of the study, a total of 120 weaned pigs (21 ± 1 days of age) with average body weight (7.10 ± 0.48 kg) were divided into four dietary treatment groups; (a) CON, basal diet, (b) ANT: basal diet with antibiotic treatment, (c) COS1, basal diet with 1 g/kg COS, and (d) COS2, basal diet with 2 g/kg COS. At the end of study results showed that the higher addition of COS (2 g/kg) enhanced the total tract apparent digestibility of dry matter and nitrogen and growth performance and reduced the incidence of diarrhea. However, digestibility and growth performance were both reduced for the pigs' given dietary additions of antibiotics (Zhou et al., 2012).

Chitosan have the ability to enhance their bioavailability in the extraintestinal tissues by reducing oxygen consumption, as well as the dietary amino acid (AA) absorption into the portal vein in young pigs (Yin et al., 2010). Research has consistently demonstrated the enhanced digestibility of the ileal contents, enhanced adsorption capacity, and increased cell division, thus clearly indicating the potential applicability of COS as a dietary additive in raising the efficiency of the digestive process and stimulating nutrient adsorption (Suthongsa et al., 2017). A study conducted by Xu et al. (2013) examined the growth performance, small intestinal structure and serum growth hormone (GH) concentration of weaned pigs, the dietary administration of COS (100, 500, 1,000 and 2,000 mg/kg) enhanced BWG quadratically. Moreover, the dietary administration of COS led to quadratic increases in the serum GH concentration, the ileum and jejunum villus heights, and the villus height to crypt depth ratios of the ileum, jejunum, and duodenum (Xu et al., 2013). This study concluded that the enhanced growth performance of the following dietary administration of COS could be the direct result of enhanced serum GH levels and the improved morphology of the small intestine (Xu et al., 2013). These conclusions were substantiated by the same authors in another investigation in which they examined the positive impacts on the growth of weaned pigs given 1 g/kg or 2 g/kg dietary COS. The study suggested that the enhanced growth of the weaned pigs given dietary COS could also link to the enhanced digestibility of calcium, phosphorus, crude protein, and dry matter and enhanced levels of amylase in the jejuna (Xu et al., 2014). Nevertheless, the above outcomes have been disputed by other studies. For example, (O'Shea et al., 2011) observed no effect on nutrient digestibility or nitrogen utilization following chitosan consumption. Similarly, the administration of 0.30% dietary COS in a study by Yan and Kim (2011) had no impact on nutrient digestibility or growth performance.

Research evidence has found that dietary low-dosage of COS with high purity not only experienced on growth-enhancing effects but also displayed a tendency toward decreased villus height in the jejunum or duodenum (Xiong et al., 2015; Yang et al., 2016). A recent study was aimed to evaluate the effects of low-dosage COS (30 mg/kg, MW = 800–2000 Da, water solubility = 99%) on intestinal mucosal AA profiles and alkaline phosphatase (ALP) activities, and serum biochemical variables in weaned piglets. For these purposes of the experiment, a total of 24 piglets (25 days of age) assigned into two



(Continued)

#### TABLE 1 | Continued

fphys-10-00516 May 3, 2019 Time: 18:31 # 7


groups (control group and treatment group) for 14 days. The results demonstrated that the dietary COS supplementation increased serum IgG, calcium, and serum urea nitrogen contents. Moreover, dietary COS increased the contents of some AA in the mucosa of jejunum and ileum, ileal mucosal ALP activity, and luminal short-chain fatty acids (SCFA) in the cecum and of the weaned piglets (Yang et al., 2016). Earlier, the same authors used similar dietary COS to evaluate the intestinal morphology, immune response, antioxidant capacity, and intestinal barrier function of weaned piglets. Results showed that dietary COS increased stomach pH, IL-6 (duodenum, jejunum, and ileum), and secretory IgA (duodenum and ileum), and reduced villus height and villus height to crypt depth ratio in the ileum. Thus the outcomes suggest that supplemental COS at low dosage may lead to immunological and oxidative stress in the small intestine and damage the integrity of the intestinal barrier in weaned piglets (Xiong et al., 2015; Guan et al., 2016).

The quantity of studies assessing the effects of dietary COS on weaned piglets far outnumbers those dealing with pigs or sow's coming to the end of their growth phase. In a study in which dietary chitosan was administered to sows with approximate body masses of 70 kg, (Egan et al., 2015) found reductions in FI, final body weight, the ileal digestibility of dry matter, gross energy, and the coefficient of apparent total tract digestibility of the gross energy of nitrogen relative to the control group (Egan et al., 2015). Furthermore, (Xie et al., 2015, 2016b,c) investigated the effects on plasma glucose levels in suckling piglets following the dietary administration of COS (30 mg/kg) to maternal sows during gestation and lactation. In one of their studies, the daily gain and weaning weight of the piglets were enhanced, and AA concentration was increased in sow milk (Xie et al., 2015). Higher plasma glucose levels and lower hepatic glycogen levels also noted in the piglets of the COS-fed sows relative to those of the control group. In another study, the piglets of sows given dietary COS displayed increased villus length, an increased villus length to crypt depth ratio in the jejunum, and ileum and increased activity of plasma glutathione peroxidase (Xie et al., 2016a). The mRNA levels of the transcription-translation negative feedback element period 1 were enhanced and the mRNA levels of the positive feedback elements, the gene encoding the basic helix-loop-helix-PAS transcription factor (CLOCK) and brain and muscle Arnt-like protein-1 were reduced (Xie et al., 2016c).

It is worth noting that the mRNA expression of genes for certain antioxidants was enhanced in the placenta following the administration of dietary COS, whereas the levels of pro-inflammatory cytokines decreased. Further investigation indicated that the administration of dietary COS triggered the mTOR signaling pathway and enhanced the expression of AA transporters in the placenta (Xie et al., 2016c). These results were backed up by Wan et al. (2016) in their examination of the reproductive performance and gene expression of specific biochemical markers in the fetuses and placentas of sows following dietary COS (100 mg/kg) administration after 35 days of gestation. In addition, a 100-mg/kg dose of dietary COS supplementation after 35 days of gestation considerably increased the fetal survival rate and size. Furthermore, the number of viable piglets born per litter and the average weights of the live piglets at birth also increased considerably following dietary COS administration during gestation (Wan et al., 2016). Therefore, dietary COS is effective in increasing the growth performance, improving intestinal structures, and utilization of dietary protein by pigs.

#### CONCLUSION

fphys-10-00516 May 3, 2019 Time: 18:31 # 8

Chitosan and COS display a significantly broad range of biological properties that confer a possible potential for a variety of commercial uses. This review demonstrates that the use of chitosan and its derivatives as a pig-feed additive provides positive antimicrobial, anti-oxidative, immunoregulatory, and blood cholesterol limiting effects. Nevertheless, it is important to realize that various structures of chitosan and COS displayed different biological properties, with no single type of chitosan displaying the full range of properties. The majority of the studies have demonstrated the beneficial effects of chitosan, such as enhanced nutrient digestibility and enhanced the growth performance (in terms of FCR and/or BWG) in weaned piglets. Nevertheless, the molecular mechanisms of these bioactivities and the precise influences of the physicochemical properties of these substances on their various bioactivities remain to be understood. In the majority of published studies, the available experimental data suggest that the growthenhancing effects of chitosan are comparable to those of dietary antibiotics. Hence, chitosan is a promising and effective alternative to antibiotics.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

GG and GL initiated the idea and the outline of this review manuscript. GG and MA wrote the manuscript. MA, SK, YL, YT, and HW provided intellectual oversight, suggestions, and editing. MA revised the manuscript critically for intellectual content. All authors read and approved the final manuscript.

#### FUNDING

This review was funded by the China Scholarship Council (No. 201708430008), National Natural Science Foundation of China (Nos. 31402092, 31772642, and 31872991), the Scientific Research Fund of Hunan Provincial Education Department (Nos. 17K043 and 16A096), Hunan Provincial Natural Science Foundation of China (No. 2018JJ1012), the Hunan Provincial Science and Technology Department (Nos. 2017NK2322 and 2018RS3086), the National Key Research and Development Program of China (Nos. 2016YFD0500504 and 2016YFD0501201), and the International Partnership Program of the Chinese Academy of Sciences (No. 161343KYSB20160008), Double first-class construction project of Hunan Agricultural University (No. SYL201802002), and The Science and Technology Department of Changsha (No. kq1706025).



chitosan, chitosan oligosaccharide and N-acetyl-D-glucosamine against Candida albicans, Candida krusei and Candida glabrata. Int. J. Pharm. 353, 139–148. doi: 10.1016/j.ijpharm.2007.11.029


amino acids transport of sows. BMC Vet. Res. 12:243. doi: 10.1186/s12917-016- 0872-8


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Guan, Azad, Lin, Kim, Tian, Liu and Wang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Bacillus cereus Isolated From Vegetables in China: Incidence, Genetic Diversity, Virulence Genes, and Antimicrobial Resistance

#### Edited by:

Helieh S. Oz, University of Kentucky, United States

#### Reviewed by:

Qingli Dong, University of Shanghai for Science and Technology, China Adriana Vivoni, Oswaldo Cruz Foundation (Fiocruz), Brazil Yong Zhao, Shanghai Ocean University, China

#### \*Correspondence:

Qingping Wu wuqp203@163.com Yu Ding dingyu@jnu.edu.cn

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 05 January 2019 Accepted: 15 April 2019 Published: 15 May 2019

#### Citation:

Yu P, Yu S, Wang J, Guo H, Zhang Y, Liao X, Zhang J, Wu S, Gu Q, Xue L, Zeng H, Pang R, Lei T, Zhang J, Wu Q and Ding Y (2019) Bacillus cereus Isolated From Vegetables in China: Incidence, Genetic Diversity, Virulence Genes, and Antimicrobial Resistance. Front. Microbiol. 10:948. doi: 10.3389/fmicb.2019.00948 Pengfei Yu1,2, Shubo Yu<sup>2</sup> , Juan Wang<sup>3</sup> , Hui Guo1,2, Ying Zhang1,2, Xiyu Liao1,2 , Junhui Zhang1,2, Shi Wu<sup>2</sup> , Qihui Gu<sup>2</sup> , Liang Xue<sup>2</sup> , Haiyan Zeng<sup>2</sup> , Rui Pang<sup>2</sup> , Tao Lei<sup>2</sup> , Jumei Zhang<sup>2</sup> , Qingping Wu<sup>2</sup> \* and Yu Ding1,2 \*

<sup>1</sup> Department of Food Science and Technology, Institute of Food Safety and Nutrition, Jinan University, Guangzhou, China, <sup>2</sup> State Key Laboratory of Applied Microbiology Southern China, Guangdong Provincial Key Laboratory of Microbial Culture Collection and Application, Guangdong Open Laboratory of Applied Microbiology, Guangdong Institute of Microbiology, Guangzhou, China, <sup>3</sup> College of Food Science, South China Agricultural University, Guangzhou, China

Bacillus cereus is a food-borne opportunistic pathogen that can induce diarrheal and emetic symptoms. It is widely distributed in different environments and can be found in various foods, including fresh vegetables. As their popularity grows worldwide, the risk of bacterial contamination in fresh vegetables should be fully evaluated, particularly in vegetables that are consumed raw or processed minimally, which are not commonly sterilized by enough heat treatment. Thereby, it is necessary to perform potential risk evaluation of B. cereus in vegetables. In this study, 294 B. cereus strains were isolated from vegetables in different cities in China to analyze incidence, genetic polymorphism, presence of virulence genes, and antimicrobial resistance. B. cereus was detected in 50% of all the samples, and 21/211 (9.95%) of all the samples had contamination levels of more than 1,100 MPN/g. Virulence gene detection revealed that 95 and 82% of the isolates harbored nheABC and hblACD gene clusters, respectively. Additionally, 87% of the isolates harbored cytK gene, and 3% of the isolates possessed cesB. Most strains were resistant to rifampicin and β-lactam antimicrobials but were sensitive to imipenem, gentamicin, ciprofloxacin, kanamycin, telithromycin, ciprofloxacin, and chloramphenicol. In addition, more than 95.6% of the isolates displayed resistance to three kinds of antibiotics. Based on multilocus sequence typing, all strains were classified into 210 different sequence types (STs), of which 145 isolates were assigned to 137 new STs. The most prevalent ST was ST770, but it included only eight isolates. Taken together, our research provides the first reference for the incidence and characteristics of B. cereus in vegetables collected throughout China, indicating a potential hazard of B. cereus when consuming vegetables without proper handling.

Keywords: Bacillus cereus, food-borne pathogen, vegetables, incidence, MLST

# INTRODUCTION

fmicb-10-00948 April 28, 2020 Time: 15:30 # 2

Bacillus cereus is a Gram-positive, spore-forming opportunistic pathogen that is widespread in different environments and known to cause foodborne outbreaks in humans (Bottone, 2010; Osimani et al., 2018). B. cereus in food products at concentrations exceeding 10<sup>4</sup> spores or vegetative cells per gram can cause food poisoning (Ehling-Schulz et al., 2006; Fricker et al., 2007; Meldrum et al., 2009). Prevalence of potential emetic and diarrheal B. cereus in different foods has been reported in Finland (Shaheen et al., 2010), Belgium (Rajkovic et al., 2006), Thailand (Chitov et al., 2008), the United Kingdom (Altayar and Sutherland, 2006; Meldrum et al., 2009), the United States (Ankolekar et al., 2009), South Korea (Park et al., 2009), and Africa (Ouoba et al., 2008). B. cereus is also one of the most prevalent foodborne pathogens in France and China (Glasset et al., 2016; Paudyal et al., 2018). From 1994 to 2005, 1,082 food poisoning cases caused by foodborne pathogens had been reported in China. B. cereus caused 145 (13.4%) of these cases, leading to six deaths (Wang et al., 2007).

Vegetables are an indispensable part for human food and nutrition. The World Health Organization recommends taking 400 g of fresh vegetables and fruits daily to promote human health (World Health Organization, 2003). Vegetables are often consumed directly or only with minimal processing that does not eliminate pathogenic bacteria, such as B. cereus (Goodburn and Wallace, 2013; Mogren et al., 2018). Since consumption of fresh vegetable has increased dramatically over the last few decades (Olaimat and Holley, 2012; Hackl et al., 2013), more foodborne outbreaks resulting from contaminated vegetables have simultaneously emerged (Berger et al., 2010; Castro-Ibáñez et al., 2016). For example, food poisoning outbreaks associated with vegetables contaminated by foodborne pathogens in Korea increased from 119 in 1998 to 271 in 2010 (Park et al., 2018). Therefore, it is necessary to monitor the contamination level of B. cereus in vegetables.

Consuming food contaminated by B. cereus can lead to gastrointestinal diseases, including diarrhea and emesis. Diarrhea is caused by different enterotoxins, including non-hemolytic enterotoxin (Nhe; Ehling-Schulz et al., 2006), hemolysin BL (Hbl; Ehling-Schulz et al., 2006), and cytotoxin K (CytK; Fagerlund et al., 2004), and emesis is due to a thermo- and acidic-stable non-ribosomal peptide, cereulide, which is encoded by the ces gene cluster (Ehling-Schulz et al., 2005; Ehlingschulz et al., 2015). In addition, B. cereus can induce other non-gastrointestinal-tract infections (Bottone, 2010; Rishi et al., 2013) and may even lead to death (Lund et al., 2000; Posfay-Barbe et al., 2008).

Antimicrobial treatment is the main method to eliminate foodborne pathogens, including B. cereus, in patients with food poisoning. However, antibiotic resistance in B. cereus has already emerged due to the abuse of antibiotics. The therapeutic effect of some antibiotics against antimicrobial-resistant isolates decreases or even disappears, leading to the failure of clinical treatment (Brown et al., 2003; Friedman, 2015; Torkar and Bedenic, 2018 ´ ). As the consumption of vegetables contaminated with antimicrobial-resistant isolates may lead to more severe infection (Berthold-Pluta et al., 2017), it is important to test the antibiotic resistance of B. cereus in vegetables for food safety and human health.

B. cereus is widely distributed in nature and can contaminate foods primarily through soil and air (European Food Safety Authority [EFSA], 2005; Arnesen et al., 2008). Vegetables are generally planted in fields, where they are exposed to soil, and they are exposed to air during transportation and sale; they can therefore be easily contaminated by this pathogenic bacterium. B. cereus contamination in vegetables has been reported. The contamination rate in different vegetables ranged from 29.0 to 70.0% in South Korea (Chon et al., 2012, 2015; Kim H. J. et al., 2016; Kim Y. J. et al., 2016). In Mexico City, B. cereus was identified in 57% of the 100 analyzed samples (Flores-Urban et al., 2014). The contamination rate in different vegetables in southeast of Spain varied greatly (Valero et al., 2002).

As an essential daily nutrient, to date, no study has evaluated the occurrence rate of B. cereus in vegetables accounting for the whole of China. Therefore, in this study, we analyzed the contamination, genotypic diversity, pathogenic potential, and antimicrobial resistance of B. cereus isolated from vegetables in China to obtain an overview on the potential risk.

# MATERIALS AND METHODS

#### Vegetable Sample Collection

A total of 419 vegetable samples (89 Coriandrum sativum L samples, 85 var. ramosa Hort. samples, 134 Cucumis sativus L. samples, and 111 Lycopersicon esculentum Mill. samples) were collected from the local markets and supermarkets of 39 major cities in China (**Table 1** and **Supplementary Figure S1**) from 2011 to 2016, according to the general sample collection guidelines of the National Food Safety Standard (The Hygiene Ministry of China, 2010b). The samples were placed in sealed bags, transferred to the laboratory in a low-temperature (below 4 ◦C) sampling box, and immediately subjected to microbiological analysis after sending back to the laboratory.

#### Isolation and Identification of B. cereus

Twenty-five grams of vegetable samples was cut into pieces, transferred into a sterile homogenizer containing 225 ml phosphate buffered saline (PBS, 0.01 mol/L), and then homogenized at 8,000 rpm for 2 min using a rotary blade homogenizer. The homogenized solution and its 1/10 and 1/100 dilutions were used to detect B. cereus qualitatively and quantitatively according to the B. cereus test rules given by the National Food Safety Standard (The Hygiene Ministry of China, 2010a), as described previously (Gao et al., 2018). Mannitol yolk polymyxin (MYP) agar plate test, parasporal crystal observation (to distinguish between B. cereus and Bacillus thuringiensis), root growth observation, hemolysis test, catalase test, motility test, nitrate reduction test, casein decomposition test, lysozyme tolerance test, glucose utilization test, and acetyl methyl alcohol test were conducted for species detection. The most probable number (MPN) method was adopted for quantitative detection of species. Briefly, 1 ml of the homogenized solution and its 1/10 and 1/100 dilutions were inoculated into three tubes each

TABLE 1 | Prevalence and contamination level of B. cereus in different vegetables.


<sup>a</sup>Contamination rate = number of positive samples/total samples. <sup>b</sup>MPN value (MPN/g) = most probable number of B. cereus per gram sample.

containing 10 ml peptone soy polymyxin broth medium. The nine cultures were incubated at 30◦C for 48 h. Then, the cultures were streaked onto MYP plates and incubated at 30◦C for at least 24 h. Presumptive colonies were picked for species identification. MPN was determined based on the MPN table provided by the National Food Safety Standard (The Hygiene Ministry of China, 2010a) and the number of positive culture(s) to calculate the MPN of B. cereus per gram sample (MPN/g).

#### Detection of Emetic and Enterotoxin Toxin Genes

Genomic DNA was obtained using the HiPure Bacterial DNA Kit (Magene, United States) following the manufacturer's specifications. Polymerase chain reaction (PCR) amplification was conducted to detect the cereulide synthetase gene (cesB) and seven enterotoxin genes (nheA, nheB, nheC, hblA, hblC, hblD, and cytK) with a 20 µl reaction mixture consisting of 50 ng genomic DNA, 12.5 µl PCR Premix TaqTM (Takara, China), and 2 µM of each primer (Hansen and Hendriksen, 2001; Fagerlund et al., 2004; Ehling-Schulz et al., 2005; Oltuszak-Walczak and Walczak, 2013). The primers used in this study are listed in **Table 2**.

#### Antimicrobial Resistance Testing

The Kirby-Bauer disk diffusion method was employed to evaluate the antimicrobial resistant, intermediate, and sensitive profiles of the isolates to 20 selected antibiotics as previously described (The Clinical and Laboratory Standards Institute [CLSI], 2010; Gao et al., 2018). The zone diameter interpretive standards were referred to the standard for Staphylococcus aureus (The Clinical and Laboratory Standards Institute [CLSI], 2010).

#### Multilocus Sequence Typing (MLST) Gene Amplification, Sequencing, and Determination

Seven housekeeping genes, namely glp, gmk, ilvD, pta, pur, pycA, and tpi, were amplified with different primers and conditions (**Table 2**) according to MLST protocol for B. cereus in PubMLST<sup>1</sup> . The sequence of each PCR product was sequenced and submitted to the PubMLST database to get the corresponding allele number. The multilocus sequence type (ST) of each isolate was obtained by ranking and submitting seven housekeeping gene allele numbers. New STs were assigned by the MLST website administrator. A minimum spanning tree was constructed with PHYLOViZ 2.0 software (Instituto de Microbiologia, Portugal) according to the relationships between MLST alleles (Ribeiro-Gonçalves et al., 2016) and to visualize the relatedness and genetic diversity of different isolates.

#### RESULTS

#### Prevalence Analysis of B. cereus in Vegetables

B. cereus was detected in 211 of 419 (50%) vegetable samples (**Table 1**), and the contaminated samples were distributed in all 39 different cities in China from where the samples were collected. The contamination rate was higher than 60% in 15 cities presented, and only three cities had a contamination rate below 20% (**Supplementary Figure S1**).

The positive rates of B. cereus were 62.92% (56/89) for C. sativum L, 57.65% (49/85) for var. ramosa Hort., 47.01% (63/134) for C. sativus L., and 38.74% (43/111) for L. esculentum Mill., respectively. Contamination levels of 9.95% (21/211) of all the samples exceeded 1,100 MPN/g. Among all positive samples, the contamination levels for the C. sativum L (30.36%; 17/56) and var. ramosa Hort. (6.12%; 3/49) samples exceeded 1,100 MPN/g, which was higher than the contamination levels for the C. sativus L. and L. esculentum Mill. samples.

#### Distribution of Virulence Genes Among B. cereus Isolates

The presence of toxin genes is summarized in **Figure 1**. The cereulide synthetase gene cesB was detected in only 3% of isolates. In contrast, the rate of enterotoxin gene detection was very high. hbl genes encoding the Hbl toxin complex were detected in 81% of the samples. Additionally, 99, 100, and 96% of all isolates harbored nheA, nheB, and nheC, respectively. However, only 95% of the isolates harbored the integrated Nhe-encoding gene cluster nheABC. cytK was detected in 87% of the strains.

The virulence gene distribution could be divided into 20 different profiles. Only five isolates, namely, 2841-1B, 3713, 3715, 3715-2A, and 3740, possessed all eight virulence genes. Two isolates (3265 and 3463) harbored the least virulence gene list (nheA-nheB). The main gene profile (70.1% of all isolates) was hblA-hblC-hblD-nheA-nheB-nheC-cytK.

<sup>1</sup>http://pubmlst.org/bcereus/info/primers.shtml

#### TABLE 2 | Primers used in this study.

fmicb-10-00948 April 28, 2020 Time: 15:30 # 4


#### Antimicrobial Susceptibility Test of B. cereus Isolates

The antimicrobial susceptibilities of all isolates were tested with 20 antimicrobials. Most isolates were found to be resistant to amoxicillin-clavulanic (AMC; 97.6%), cephalothin (KF; 86.7%), penicillin (P; 99.7%), ampicillin (AMP; 99.7%), cefoxitin (FOX; 95.6%), which belong to β-lactams, as well as rifampin (RD, 83.0%), an ansamycin. On the other hand, most isolates were sensitive to some other antimicrobials, such as kanamycin (K; 83.3%), gentamicin (CN; 97.6%), telithromycin (TEL; 84.7%), imipenem (IPM; 99.7%), ciprofloxacin (CIP; 92.9%), chloramphenicol (C; 94.6%), and teicoplanin (TEC; 81.0%). Besides, most isolates exhibited intermediate resistance to quinupristin (QD; 61.9%) and clindamycin (DA; 74.8%; **Figure 2**).

There were 74 antimicrobial resistant profiles for all isolates. The strain 41-1 and 1515-1A turned out to be the most highly resistant isolates, which were resistant to 12 antibiotics (AMP-KF-FOX-P-AMC-CTT-SXT-DA-RD-QD-TEL-FD and AMP-KF-FOX-P-AMC-TE-CTT-DA-RD-QD-TEL-FD, respectively). In contrast, the most sensitive strain, 3763, showed resistance to only two antibiotics (FOX-TE). AMP-KF-FOX-P-AMC-RD was the most common antimicrobial resistant profile (91 of 294 strains). We also evaluated the multidrug resistance (MDR; Magiorakos et al., 2012) profiles and found that 95.6, 76.2, and 35.4% of isolates displayed simultaneous resistance to more than three, four, and five types of antimicrobials, respectively (**Figure 2**).

# Multilocus Sequence Typing and Cluster Analysis

Genetic diversity was analyzed by the MLST method. Among all 294 strains, 210 STs were assigned, and 145 strains were assigned to 137 new STs. Additionally, 175 of all the 210 (83%) STs included a single strain, 35 STs included two to eight isolates, and only ST-770 included eight isolates, followed by ST-1605, which included seven strains. Five isolates belonged to ST-26, which is associated with clinical isolates. All 210 STs were grouped into six clonal complexes (CCs) and 189 singletons. The ST-142 complex was most frequent, including 41 isolates, while the ST-18, ST-23,

FIGURE 1 | Detection rate of virulence genes in B. cereus from vegetables. The number at the top of the bars represents the positive rate of corresponding toxin genes. hblACD and nheABC mean that the strains are positive for hblA, hblC, and hblD or for nheA, nheB, and nheC at the same time, respectively. "All eight genes" presents the strains with all the detected toxin genes.

ST-97, ST-111, and ST-205 complexes contained 28, 10, 4, 10, and 12 isolates, respectively **(Figure 3**).

#### DISCUSSION

#### Prevalence of B. cereus Isolates in Vegetables

Few studies have evaluated pathogenic B. cereus in vegetables worldwide and no report has focused on the whole of China. In our study here, we found that 50% of all vegetable samples collected from 39 major cities in China contained B. cereus. The contamination level was more or less the same as those of previous surveys in other countries, i.e., 20–48% in Korea (Chon et al., 2015; Kim H. J. et al., 2016; Park et al., 2018), 57% in Mexico City (Flores-Urban et al., 2014), and 52% in the southeast of Spain (Valero et al., 2002). These reports, together with ours, indicate that B. cereus contamination in vegetables is common in several countries and suggest that consumption of vegetables contaminated with B. cereus is a potential health hazard (Valero et al., 2002; Kim Y. J. et al., 2016). The high level of contamination by B. cereus may be partly attributed to contact with soil or air during field planting (European Food Safety Authority [EFSA], 2005;

Arnesen et al., 2008) and exposure to air during transportation and sale. Upon contamination, B. cereus may form a biofilm on the surface of vegetables, resulting in its persistent difficulty to be eliminated (Majed et al., 2016). B. cereus-positive samples may also contaminate other vegetables by contact transmission. According to the Microbiological Guidelines for Food of Hong Kong, China (Food and Environmental Hygiene Department, 2014), even though the amounts of B. cereus between 10<sup>3</sup> and 10<sup>5</sup> CFU/g in ready-to-eat foods are considered to be "acceptable," they pose potential risks,

and the raw materials, processing period, and environment should be examined to investigate the reason why these foods are contaminated; if the amounts of B. cereus in ready-to-eat foods are more than 10<sup>5</sup> CFU/g, their quality is considered to be "unsatisfactory" and their sale should be stopped. The standards of microbiological limits for ready-to-eat foods in Australia and New Zealand (New South Wales Food Authority, 2009), however, stipulate that the "acceptable" level of B. cereus is 102–10<sup>3</sup> CFU/g, and an "unsatisfactory" level is 103–10<sup>4</sup> CFU/g. The United Kingdom microbiological testing standards (Health Protection Agency, 2009) for ready-to-eat foods stipulate that the "acceptable" level of B. cereus in ready-to-eat foods is 103–10<sup>5</sup> CFU/g, and a level of more than 10<sup>5</sup> CFU/g of B. cereus is considered to be "unsatisfactory." Of all the B. cereus-positive samples in this study, 9.95% (21/211), mainly of C. sativum L and var. ramosa Hort., had contamination levels of more than 1,100 MPN/g. The contamination levels of these samples were at least at the "acceptable" level according to Hong Kong and United Kingdom standards and at the "unsatisfactory" level according to Australia and New Zealand standards. This suggests that B. cereus-contaminated vegetables pose a potential risk of causing foodborne disease, and care needs to be taken when consuming them directly or with minimal processing.

# Multilocus Sequence Typing and Genetic Diversity

MLST is a crucial epidemiological typing method based on the sequences of seven different housekeeping gene loci; it is used in studies of evolution and population diversity of B. cereus isolates (Erlendur et al., 2004; Cardazzo et al., 2008; Liu et al., 2017; Yang et al., 2017). In this study, we employed MLST to analyze genetic polymorphism in isolates from vegetables. Most of the isolates were assigned to singleton (**Figure 3**). The six CCs were distributed in separate samples, except ST-23 complex and ST-97 complex. ST-18 complex, ST-23 complex, ST-142 complex, and ST-205 complex even crossed with some singletons, indicating high genetic diversity of the isolates. However, we could not find any unique STs that existed in only one particular vegetable variety. Five strains, two of which were isolated from var. ramosa Hort, one from C. sativum L, and the remaining two from C. sativus L., were assigned to ST26, the same molecular type of clinical isolates NC7401 and F4810/72 (Agata et al., 2002; Fricker et al., 2007). Three out of these five isolates were identified as potential emetic strains. As the preformed cereulide in foods is persistent and may also lead to food poisoning (Agata et al., 1995), there is a potential risk when consuming these vegetables directly or with minimal processing. Interestingly, we found that all isolates that belonged to the ST-18 complex, ST-97 complex, and ST-142 complex harbored the same virulence gene profile (hblAhblC-hblD-nheA-nheB-nheC-cytK), while other CCs showed no such phenomenon. Additionally, 30 of 32 isolates belonging to ST-18 and ST-97 complexes showed resistance to cefotetan (CTT, 30 µg), whereas only 10 of 73 isolates belonging to the other four CCs showed similar resistance, which may be explained by the properties of founder clones of different CCs.

# Virulence Gene Detection and Potential Toxicity

Diarrhea caused by B. cereus is attributed to different enterotoxins produced by these strains in the small intestine (Jeßberger et al., 2015), including Hbl, Nhe (Lund and Granum, 1997), and CytK (Lund et al., 2000). In this study, we evaluated seven enterotoxin genes in B. cereus, and the positive rates of nheABC, hblACD, and cytK were 95, 81, and 87%, respectively (**Figure 2** and **Supplementary Table S1**). The positive rates of nheABC, hblACD, and cytK were higher than those found in Korea (69.5, 41.7, and 74.2%; Park et al., 2018), but the hblA gene was detected in only 82% of isolates, which is lower than that reported in Mexico City (Flores-Urban et al., 2014). When considering different kinds of food, the nheABC frequency in our vegetables was lower than in Sunsik from Korea (Chon et al., 2012) or in pasteurized milk from China (Gao et al., 2018), but the cytK detection rate was much higher than those in rice and cereals from Korea (55%; Park et al., 2009), in Sunsik from Korea (77%; Chon et al., 2012), and in pasteurized milk from China (73%; Gao et al., 2018). Owing to the properties of foods consumed raw or processed minimally, the wider distribution of diarrheal B. cereus in these vegetables and their potential hazard cannot be neglected.

Emetic symptoms are caused by the emetic toxin cereulide. The positive rate of cesB was 3%, which is higher than those reported in Korea and Mexico City (Flores-Urban et al., 2014; Chon et al., 2015; Park et al., 2018), almost the same as that (2.9%) in Sunsik of Korea (Chon et al., 2012), but slightly lower than the 5% in pasteurized milk of China (Gao et al., 2018). Although the positive rate of cesB was quite low when compared with the rates of enterotoxins, cereulide is very persistent and heat-stable. Even emetic toxin remaining in sterilized food can cause emetic symptoms accordingly (Agata et al., 1995), the emetic isolates in these raw consuming vegetables are still potential risks.

# Antimicrobial Resistance of B. cereus Isolates

B. cereus infection may lead to diarrhea, vomiting, and even death (Hilliard et al., 2003; Evreux et al., 2007; Jean-Winoc et al., 2013; Ramarao et al., 2014). Antibiotic resistance test may provide a theoretical reference for the clinical treatment of B. cereus food poisoning and infection. Detailed antimicrobial resistance information of the isolates from vegetables is shown in **Supplementary Table S2**. More than 72.5% of all isolates were susceptible to six classes of antimicrobial agents, including aminoglycosides (CN, K), ketolide (TEL), glycopeptides (TEC), quinolones (CIP), phenylpropanol (C), tetracyclines (TE), and folate pathway inhibitors (SXT). A total of 60.2 and 99.7% of the strains were also susceptible to third-generation cephalosporin (CTT) and penems (IPM), respectively. More than 83.0% of the isolates showed resistance to other β-lactam antibiotics, including penicillins (AMP, P), β-Lactam/β-lactamase inhibitor combinations (AMC), cephems (KF, FOX), and ansamycins (RD). The high resistance of B. cereus to β-lactam antimicrobials has been widely reported and may be due to the synthesis of β-lactamase (Philippon et al., 1998; Chen et al., 2004; Park et al., 2018). Notably, 83.0% of the isolates showed resistance to

rifampin (RD), which is much higher than the resistance rates of vegetables in Korea (only 48.7%; Park et al., 2018) and even higher than those of different kinds of foods, including rice and cereal (62%; Park et al., 2009), ready-to-eat foods (0%; Agwa et al., 2012), and traditional dairy products (0%; Owusu-Kwarteng et al., 2017). This may be ascribed to the usage of antibiotics in different countries or the evolution of strains toward rifampicin resistance. These results emphasize the need for caution when using β-lactams and ansamycins (such as RD) for the clinical treatment of B. cereus. On the other hand, 95.6, 76.2, and 35.4% of isolates showed resistance to more than three, four, and five classes of antibiotics simultaneously, respectively, suggesting the need to monitor multiple drug resistance in B. cereus.

#### CONCLUSION

Vegetables such as L. esculentum Mill., C. sativus L., var. ramosa Hort., and so on are usually consumed directly or with minimal processing, so potential hazards associated with B. cereuscontaminated vegetables should not be ignored. The results in this study revealed a high incidence of B. cereus in vegetable samples collected from across China, for the first time as we know. Of all the samples, 21/211 (9.95%), mainly of C. sativum L and var. ramosa Hort., had contamination levels of more than 1,100 MPN/g. According to the Microbiological Guidelines for Food of Hong Kong, China (Food and Environmental Hygiene Department, 2014), the United Kingdom microbiological testing standards for ready-to-eat foods (Health Protection Agency, 2009), and the standards of microbiological limits for readyto-eat foods in Australia and New Zealand (New South Wales Food Authority, 2009), the contamination levels of these samples were at the "acceptable" level according to Hong Kong and United Kingdom standards and at the "unsatisfactory" level according to the Australia and New Zealand standards. These contamination levels indicate a potential risk caused by the consumption of B. cereus-contaminated vegetables either directly or with minimal processing and should be kept in mind while assessing the quality of vegetables. If necessary, the reason for the contamination should also be traced. The pathogenic ST ST26 was detected in vegetable isolates. Seven enterotoxin genes associated with diarrheal symptoms were widespread among isolates, and the emetic toxin gene was also detected. In addition, most isolates were resistant to β-lactam antimicrobials, such as amoxicillin-clavulanic acid, penicillin, ampicillin, cephalothin, cefoxitin, and rifampin. Our results indicate a potential risk of consuming vegetables without sufficient processing and the

#### REFERENCES


increasing difficulties in eliminating B. cereus with antibiotics. To ensure the health and safety for the public, it is therefore necessary to develop new methods to prevent contamination and consequent potential foodborne outbreak induced by B. cereus in raw consuming vegetables.

#### AUTHOR CONTRIBUTIONS

QW, YD, JW, JMZ, and PY conceived the project and designed the experiments. PY, SY, HG, YZ, XL, JHZ, SW, QG, LX, HZ, RP, and TL performed the experiments. QW and YD supervised the project. PY and YD analyzed the data and wrote the manuscript. QW, JW, and YD complemented the writing.

#### FUNDING

We would like to acknowledge the financial support of the National Key R&D Program of China (Grant No. 2018YFC1602500), National Natural Science Foundation of China (Grant No. 31730070), Science and Technology Program of Guangzhou, China (Grant No. 201604016068), State Key Laboratory of Applied Microbiology Southern China (Grant Nos. SKLAM004-2016 and SKLAM006-2016), GDAS' Special Project of Science and Technology Development (Grant No. 2017GDASCX-0201), and the Fundamental Research Funds for the Central Universities and the 111 Project.

#### ACKNOWLEDGMENTS

YD is awarded the 1000-Youth Elite Program (The Recruitment Program of Global Experts in China).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.00948/full#supplementary-material

FIGURE S1 | Sampling cities where the vegetables were collected.

TABLE S1 | Prevalence of virulence genes in B. cereus isolated from vegetables in China.

TABLE S2 | Results of antimicrobial resistance test for B. cereus isolates in the study.

markets in Portharcourt, Rivers State, Nigeria. Asian J. Microbiol. Biotechnol. Environ. Sci. 14, 13–18.



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Yu, Yu, Wang, Guo, Zhang, Liao, Zhang, Wu, Gu, Xue, Zeng, Pang, Lei, Zhang, Wu and Ding. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fmicb-10-00948 April 28, 2020 Time: 15:30 # 10

# Corrigendum: Bacillus cereus Isolated From Vegetables in China: Incidence, Genetic Diversity, Virulence Genes, and Antimicrobial Resistance

Pengfei Yu1,2, Shubo Yu<sup>2</sup> , Juan Wang<sup>3</sup> , Hui Guo1,2, Ying Zhang1,2, Xiyu Liao1,2 , Junhui Zhang1,2, Shi Wu<sup>2</sup> , Qihui Gu<sup>2</sup> , Liang Xue<sup>2</sup> , Haiyan Zeng<sup>2</sup> , Rui Pang<sup>2</sup> , Tao Lei <sup>2</sup> , Jumei Zhang<sup>2</sup> , Qingping Wu<sup>2</sup> \* and Yu Ding1,2 \*

*<sup>1</sup> Department of Food Science and Technology, Institute of Food Safety and Nutrition, Jinan University, Guangzhou, China, <sup>2</sup> State Key Laboratory of Applied Microbiology Southern China, Guangdong Provincial Key Laboratory of Microbial Culture Collection and Application, Guangdong Open Laboratory of Applied Microbiology, Guangdong Institute of Microbiology, Guangzhou, China, <sup>3</sup> College of Food Science, South China Agricultural University, Guangzhou, China*

Keywords: Bacillus cereus, food-borne pathogen, vegetables, incidence, MLST

#### **A Corrigendum on**

#### Edited and reviewed by:

*Giovanna Suzzi, University of Teramo, Italy*

#### \*Correspondence:

*Qingping Wu wuqp203@163.com Yu Ding dingyu@jnu.edu.cn*

#### Specialty section:

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

Received: *21 January 2020* Accepted: *08 April 2020* Published: *29 April 2020*

#### Citation:

*Yu P, Yu S, Wang J, Guo H, Zhang Y, Liao X, Zhang J, Wu S, Gu Q, Xue L, Zeng H, Pang R, Lei T, Zhang J, Wu Q and Ding Y (2020) Corrigendum: Bacillus cereus Isolated From Vegetables in China: Incidence, Genetic Diversity, Virulence Genes, and Antimicrobial Resistance. Front. Microbiol. 11:848. doi: 10.3389/fmicb.2020.00848*

#### **Bacillus cereus Isolated From Vegetables in China: Incidence, Genetic Diversity, Virulence Genes, and Antimicrobial Resistance**

by Yu, P., Yu, S., Wang, J., Guo, H., Zhang, Y., Liao, X., et al. (2019). Front. Microbiol. 10:948. doi: 10.3389/fmicb.2019.00948

In the original article, the reference for Andreja et al., 2010 was incorrectly written as "Andreja, R., Mieke, U., Tine, C., Mark, H., and Johan, D. (2010). Prevalence and characterisation of Bacillus cereus in vacuum packed potato puree. Int. J. Food Sci. Technol. 41, 878–884. doi: 10.1111/j.1365-2621.2005.01129.x" It should be "Rajkovic, A., Uyttendaele, M., Courtens, T., Heyndrickx, M., and Debevere, J. (2006). Prevalence and characterisation of Bacillus cereus in vacuum packed potato puree. Int. J. Food Sci. Technol. 41, 878– 884. doi: 10.1111/j.1365-2621.2005.01129.x" And the citation Andreja et al., 2010 in the INTRODUCTION Paragraph 1 should read: "(Rajkovic et al., 2006)."

The reference for Altayar and Sutherland, 2010 was incorrectly written as "Altayar, M., and Sutherland, A. D. (2010). Bacillus cereus is common in the environment but emetic toxin producing isolates are rare. J. Appl. Microbiol. 100, 7–14. doi: 10.1111/j.1365-2672.2005.02764.x" It should be "Altayar, M., and Sutherland, A. D. (2006). Bacillus cereus is common in the environment but emetic toxin producing isolates are rare. J. Appl. Microbiol. 100, 7–14. doi: 10.1111/j.1365-2672.2005.02764.x" And the citation Altayar and Sutherland, 2010 in the INTRODUCTION Paragraph 1 should read: "(Altayar and Sutherland, 2006)."

The reference for Wang and Zhang, 2013 was incorrectly written as "Wang, S., and Zhang, D.W. (2013). Analysis of bacterial foodborne disease outbreaks in China between 1994 and 2005. FEMS Immunol. Med. Microbiol. 51, 8–13. doi: 10.1111/j.1574-695x.2007.00305.x" It should be "Wang, S., Duan, H., Zhang, W., and Li, J. W. (2007). Analysis of bacterial foodborne disease outbreaks in China between 1994 and 2005. FEMS Immunol. Med. Microbiol. 51, 8– 13. doi: 10.1111/j.1574-695X.2007.00305.x" And the citation Wang and Zhang, 2013 in the INTRODUCTION Paragraph 1 should read: "(Wang et al., 2007)."

The reference for Barbara et al., 2008 was incorrectly written as "Barbara, C., Enrico, N., Lisa, C., Leonardo, A., Tomaso, P., and Valerio, G. (2008). Multiple-locus sequence typing and analysis of toxin genes in Bacillus cereus foodborne isolates. Appl. Environ. Microbiol. 74, 850–860. doi: 10.1128/AEM.01495-07" It should be "Cardazzo, B., Negrisolo, E., Carraro, L., Alberghini, L., Patarnello, T., and Giaccone, V. (2008). Multiple-locus sequence typing and analysis of toxin genes in Bacillus cereus food-borne isolates. Appl. Environ. Microbiol. 74, 850–860. doi: 10.1128/aem.01495-07" And the citation Barbara et al., 2008 in the DISCUSSION Multilocus Sequence Typing and Genetic Diversity Paragraph 1 should be "(Cardazzo et al., 2008)."

In the original article **"**(Osimani et al., 2018)" and "(Fricker et al., 2007)**"** were not cited in the Introduction part (Paragraph 1).

In the original article, there was an error in the Introduction part (Paragraph 1). The strains in the literatures we referenced to were not all isolated from food poisoning outbreaks, so our statement "Outbreaks with vomiting and diarrheal syndromes caused by B. cereus" in the manuscript may lead to misunderstanding. Besides, we think that the references on the foodborne outbreaks and pathogenic concentration of Bacillus cereus were not specific enough, so we prefer to add two more references (Fricker et al., 2007; Osimani et al., 2018) for the first two sentences.

The corrections have been made to INTRODUCTION Paragraph 1:

"Bacillus cereus is a Gram-positive, spore-forming opportunistic pathogen that is widespread in different environments and known to cause foodborne outbreaks in humans (Bottone, 2010; Osimani et al., 2018). B. cereus in food products at concentrations exceeding 10<sup>4</sup> spores or vegetative cells per gram can cause food poisoning (Ehling-Schulz et al., 2006; Fricker et al., 2007; Meldrum et al., 2009). Prevalence of potential emetic and diarrheal B. cereus in different foods has been reported in Finland (Shaheen et al., 2010), Belgium (Rajkovic et al., 2006), Thailand (Chitov et al., 2008), the United Kingdom (Altayar and Sutherland, 2006; Meldrum et al., 2009), the United States (Ankolekar et al., 2009), South Korea (Park et al., 2009), and Africa (Ouoba et al., 2008). B. cereus is also one of the most prevalent foodborne pathogens in France and China (Glasset et al., 2016; Paudyal et al., 2018). From 1994 to 2005, 1,082 food poisoning cases caused by foodborne pathogens had been reported in China. B. cereus caused 145 (13.4%) of these cases, leading to six deaths (Wang et al., 2007).

The authors apologize for these errors and state that this does not change the scientific conclusions of the article in any way. The original article has been updated.

#### REFERENCES


Copyright © 2020 Yu, Yu, Wang, Guo, Zhang, Liao, Zhang, Wu, Gu, Xue, Zeng, Pang, Lei, Zhang, Wu and Ding. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Dietary Quercetin Increases Colonic Microbial Diversity and Attenuates Colitis Severity in *Citrobacter rodentium*-Infected Mice

#### *Rui Lin\*, Meiyu Piao and Yan Song*

*Department of Gastroenterology and Hepatology, General Hospital, Tianjin Medical University, Tianjin, China*

#### *Edited by:*

*Jie Yin, Institute of Subtropical Agriculture (CAS), China*

#### *Reviewed by:*

*Longhuo Wu, Gannan Medical University, China Bi E. Tan, Institute of Subtropical Agriculture (CAS), China*

> *\*Correspondence: Rui Lin pubmed1128@126.com*

#### *Specialty section:*

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 25 March 2019 Accepted: 30 April 2019 Published: 16 May 2019*

#### *Citation:*

*Lin R, Piao M and Song Y (2019) Dietary Quercetin Increases Colonic Microbial Diversity and Attenuates Colitis Severity in Citrobacter rodentium-Infected Mice. Front. Microbiol. 10:1092. doi: 10.3389/fmicb.2019.01092*

Disturbed balance between microbiota, epithelial cells, and resident immune cells within the intestine contributes to inflammatory bowel disease (IBD) pathogenesis. The *Citrobacter rodentium*-induced colitis mouse model has been well documented. This model allows the analysis of host responses to enteric bacteria and facilitates improved understanding of the potential mechanisms of IBD pathogenesis. The current study evaluated the effects of dietary 30 mg/kg quercetin supplementation on *C. rodentium*-induced experimental colitis in C57BL/6 mice. Following dietary quercetin supplementation, the mice were infected with 5 × 108 CFU *C. rodentium*, and the pathological effects of *C. rodentium* were measured. The results showed that quercetin alleviated the effects of *C. rodentium*induced colitis, suppressed the production of pro-inflammatory cytokines, such as interleukin (IL)-17, tumor necrosis factor alpha, and IL-6 (*p* < 0.05), and promoted the production of IL-10 in the colon tissues (*p* < 0.05). Quercetin supplementation also enhanced the populations of *Bacteroides*, *Bifidobacterium*, *Lactobacillus,* and *Clostridia* and significantly reduced those of *Fusobacterium* and *Enterococcus* (*p* < 0.05). These findings indicate that dietary quercetin exerts therapeutic effects on *C. rodentium*-induced colitis, probably due to quercetin's ability to suppress pro-inflammatory cytokines and/or modify gut microbiota. Thus, these results suggest that quercetin supplementation is effective in controlling *C. rodentium*-induced inflammation.

Keywords: *Citrobacter rodentium*, colitis, gut microbiota, diet, quercetin, inflammatory bowel disease

# INTRODUCTION

In the last decade, inflammatory bowel disease (IBD) has been one of the most frequently investigated human health issues associated with the gut microbiota (Kostic et al., 2014). More than 3.6 million people worldwide (Loftus, 2004) are affected with IBD, including Crohn's disease (CD) and ulcerative colitis (UC), and the incidence of these diseases has been increasing in recent decades. The latter fact emphasizes the contribution of environmental factors to these diseases (Pillai, 2013). Thus, gut microbial communities have become a prominent research subject because of their effect on multiple aspects of health, such as IBD pathogenesis (Valdes et al., 2018).

The intestinal tract contains 10 trillion microorganisms (Dave et al., 2012) that are separated from the host's mucosal immune cells by single layer of polarized epithelial cells play a crucial role in the development of the mucosal immune system. These symbiotic inhabitants, collectively known as the gut microbiota, also supply vital nutrients and limit the colonization of pathogenic microbes in the gut (Honda and Takeda, 2009). Evidence from both IBD patients and mouse models has shown that profound changes in the gut, such as intestinal microbiota development, play a major role in IBD pathogenesis (Gkouskou et al., 2014; Matsuoka and Kanai, 2015). Similar to enteropathogenic *Escherichia coli* (EPEC) and enterohemorrhagic *E. coli* (EHEC), *Citrobacter rodentium* is a member of the noninvasive group of attaching and effacing (A/E) bacteria that attach themselves to the intestinal epithelium and colonizes the host's gut. At this point, the A/E pathogens induce alterations in the colonic tissue similar to those observed in cases of EPEC or EHEC infections in murine and human IBD (Law et al., 2013). A few models of infectious colitis exist, but in particular, the *C. rodentium*-induced colitis model (Law et al., 2013; Guan et al., 2016) has been well documented for studying the pathogenesis of host responses to enteric bacteria. This model can promote the understanding of the mechanism underlying IBD pathogenesis. Therefore, research related to *C. rodentium* is a key step for developing innovative prophylactic and therapeutic treatments.

Naturally found in fruits and vegetables, dietary antioxidant flavonoids are natural polyphenols. Recent studies have revealed that natural polyphenols exert potential preventative and therapeutic effects on various diseases (Ding et al., 2018; Hong and Piao, 2018). In certain organs, antioxidants provide inflammatory relief. Thus, natural polyphenols could be potential treatment options for IBD (Azuma et al., 2013). Quercetin is a flavonoid with antioxidant properties that is naturally present in most citrus fruits. It is considered to exert antidiabetic, antidepressant, and anti-inflammatory effects on cellular signaling pathways. In addition, quercetin inhibits tumor necrosis factor alpha (TNF-α) and interleukin (IL)-4 production in type I allergic reactions and decreases Th2-type cytokine production by basophils (Kim et al., 2014). It has also been proven to exert therapeutic effects on asthma, arthritis, and lung injury (Townsend and Emala, 2013); however, the precise mechanism by which it affects colitis is still unknown. Thus, this study was aimed at determining the potential effects of quercetin on *C. rodentium*-induced colitis in C57BL/6 mice.

# MATERIALS AND METHODS

#### *C. rodentium* Infection and Treatment/ Animals and Experimental Design

This study was conducted according to the Chinese animal welfare guidelines after receiving approval from the Animal Care and Use Committee of General Hospital of the Tianjin Medical University. The study sample comprised pathogenfree female C57BL/6 mice that were kept under controlled conditions at 24 ± 2°C with a relative humidity of 60 ± 5% and a 12-h light/dark cycle (06:00–18:00). The control group (CTRL, *n* = 10) and the *C. rodentium*-infection group (CR-infection, *n* = 10) received a basal rodent diet. The quercetin group (QUE, *n* = 10) received a basal rodent diet supplemented with 30-mg/kg quercetin (Q0125, Sigma).

*C. rodentium* for infection was grown for 14 h in Luria Bertani (LB) broth containing 0.05-g/L nalidixic acid/mL. The cultures were then centrifuged at 3,000 × *g* for 10 min, and the pellets were resuspended in sterile phosphate-buffered saline (PBS). This *C. rodentium* culture with a final concentration of 5 × 108 CFU/ml was used for infection. Briefly, mice were fed quercetin and/or basal rodent diet for 2 weeks as per the group allocation, and the mice in the CR-infection and QUE groups were infected with the 5 × 108 -CFU *C. rodentium* culture by gavage at 9:00 the next day. Subsequently, each mouse was housed in an individual cage to avoid reinfection from littermates. On day 7 post-infection, all mice were euthanized by CO2 asphyxiation.

The colonic mucosal tissues of all mice were removed using razors in ice and then stored frozen in liquid nitrogen. Colonic contents and feces were collected, weighed, and re-suspended in PBS, and their serial dilutions were then plated onto LB agar plates containing nalidixic acid. After 24 h, *C. rodentium* colonies were counted, and whether the colonies were of *C. rodentium* was confirmed using PCR with *C. rodentium*-specific primers (Bhinder et al., 2013).

#### Histopathological Analysis

Excised colonic mucosal tissue specimens were fixed in 10% formalin, embedded in paraffin, and cut into 3-μm sections. The sections were then stained with hematoxylin and eosin (HE) for visualization under a microscope (100× magnification). Blind histological scoring was performed by a pathologist using a six-grade system as described by Varshney et al. (2013).

#### Detection of Inflammatory Cytokines

The colonic mucosal tissues were pulverized using surgical scissors and homogenized in ice-cold PBS. The tissue homogenates were then centrifuged at 1,900× *g* at 4°C for 15 min to obtain the supernatant. The levels of pro-inflammatory cytokines IL-17, IL-6, TNF-α, and IL-10 were measured using commercial ELISA kits (eBioscience) following the manufacturer's instructions.

#### DNA Extraction and Sequence Analysis

Immediately after collection, the colonic contents were frozen, their genomic DNA was extracted, and the DNA was amplified using primers specific to the V3-V4 region of 16S rRNA gene barcodes. The samples were combined and subjected to sequencing on the Illumina MiSeq platform in accordance with the manufacturer's instructions (Fadrosh et al., 2014). In addition, quality filtering, chimera removal, and *de novo* operational taxonomic unit (OTU) clustering were conducted using the UPARSE pipeline (Edgar, 2013). Readings were replicated, organized, and grouped into candidate OUTs, and chimeric OUTs were removed.

The taxonomic assignment of the OUTs was annotated using RDP reference (version 16) with an identity threshold of 97% in the UPARSE pipeline. The OTU table with taxonomic assignments was converted into the "biom" format for compatibility with the QIIME software (Navas-Molina et al., 2013). Alpha diversity was calculated using QIIME, for which the existing significant difference between case/control was calculated with 999 Monte Carlo permutation and Bonferroni multiple correction.

#### Statistical Analysis

Data are presented as the standard error of the mean. One-way analysis of variance was used for comparison between multiple experimental groups, and the significance of the differences between the groups was determined using Duncan's multiple range test. The sample sizes were measured to ensure statistical validity, and *p* < 0.05 was considered statistically significant.

#### RESULTS

The body weights of the mice in the CTRL and QUE groups were not significantly different before and after the experiment, whereas those of the mice in the CR-infection group were lower after the experiment (**Figure 1A**). On day 7 postinfection, the *C. rodentium* count in the colonic contents or feces was not different between the CR-infection and QUE groups (**Figures 1B,C**). While comparing with the mice in the CTRL and QUE groups, the mice in the CR-infection group showed signs of colitis (**Figures 2A–D**).

Compared with the mice in the CTRL group, those in the CR-infection and QUE groups showed significantly elevated IL-10 (**Figure 3A**), IL-17 (**Figure 3B**), IL-6 (**Figure 3C**), and TNF-α (**Figure 3D**) levels (all *p* < 0.05). Quercetin supplementation also significantly increased the IL-10 level in the QUE mice (**Figure 3A**) compared with the level found in the CR-infected mice (*p* < 0.05). However, quercetin supplementation in the QUE group appeared to mitigate the *C. rodentium*-induced increases in the IL-17 (**Figure 3B**), IL-6 (**Figure 3C**), and TNF-α (**Figure 3D**) levels (*p* < 0.05).

Amplification of the V3-V4 region of the 16S rRNA gene obtained from the colonic contents of CTRL, CR-infected, and QUE groups provided raw readings (35,673, 37,894, and 33,511, respectively) to facilitate the assessment of the effects of *C. rodentium* infection and dietary quercetin on bacterial communities. Following trimming, assembly, and quality filtering, 2,895 OTUs were detected. **Figure 4** presents the Shannon and Simpson diversity indices and microbial richness indices (Chao1 and ACE) in the groups. Compared with the CTRL group, the CR-infection group showed significantly increased Simpson index (*p* < 0.05) (**Figure 4A**) and decreased Chao1 (**Figure 4B**), Shannon (**Figure 4C**) and ACE indices (**Figure 4D**) (all, *p* < 0.05). Compared with the CR-infection group, the QUE group showed increased Chao1 (**Figure 4B**), Shannon (**Figure 4C**), and ACE indices (**Figure 4D**) (all, *p* < 0.05) and decreased Simpson index (**Figure 4A**) (*p* < 0.05). However, the observed alpha diversity of the microbiota showed no significant difference between the CTRL and QUE groups.

Taxon-dependent analysis was used to determine the intestinal microbiota taxonomy, and Bacteroidetes, Firmicutes, Proteobacteria, and Verrucomicrobia were found to be the most abundant phyla (**Figures 5A**–**D**). Their relative abundances were 56.32, 33.18, 2.42 and 2.45%, respectively, in the CTRL group; 67.54, 24.47, 2.45, and 4.31%, respectively, in the CR-infection group; and 57.24, 29.87, 2.89, and 3.13%, respectively, in the QUE group.

Microbial populations at the genus level in the colonic contents were also investigated (**Figures 6A–F**). The populations of *Bacteroides*, *Bifidobacterium*, *Lactobacillus*, and *Clostridia* were decreased (*p* < 0.05), and those of *Fusobacterium* and *Enterococcus* were increased in the CR-infection group (*p* < 0.05) compared with those in3 the CTRL group. Notably, compared with the CR-infection group, the QUE group showed enhanced populations of *Bacteroides*, *Bifidobacterium*, *Lactobacillus*, and *Clostridia* but suppressed populations of *Fusobacterium* and *Enterococcus* (*p* < 0.05) because of quercetin supplementation.

#### DISCUSSION

Because of their inherent susceptibility to EPEC or EHEC, mice have been the most frequently used model for studying *C. rodentium*-induced intestinal infection or intestinal EPEC or EHEC infection. Although *C. rodentium* rarely causes intestinal diseases in humans, it can colonize all mice strains, causing either fatal or virtually asymptomatic illness depending on the

diversity between the CTRL, CR-infection and QUE groups in terms of the (A) Simpson, (B) chao1, (C) Shannon, and (D) ACE indices. \* indicates *p* < 0.05 compared with the CTRL group. # indicates *p* < 0.05 compared with the CR-infection group.

Verrucomicrobia members in the mouse colonic contents (*n* = 10 in each group). \* indicates *p* < 0.05 compared with the CTRL group. # indicates *p* < 0.05 compared with the CR-infection group.

mouse strain (Coleman et al., 2014; Scholz et al., 2016). The CD-1 and C57BL/6 mouse strains develop only subclinical symptoms and are considered resistant to *C. rodentium*-induced colitis, whereas FVB/N and C3H/HeJ mice strains exhibit *C. rodentium* infection and are considered susceptible (Borenshtein et al., 2008). In particular, mouse models of *C. rodentium*-induced infection are the most useful for studying infectious diseases and colitis in mice because they have been well documented for host responses to pathogenic bacteria. Our previous study showed that quercetin or quercetin monoglycoside supplementation can prevent dextran sulphate sodium-induced colitis (Hong and Piao, 2018). In that study, mice were fed quercetin to protect the gut and allow rapid recovery after *C. rodentium* infection. The results indicated that quercetin supplementation provided therapeutic benefits in the *C. rodentium*-induced infection model of gastrointestinal injury. Inflammatory responses and intestinal microflora composition were the most important determinants of host susceptibility.

Dietary preference has a major impact on the gut microbial composition throughout human life (Conlon and Bird, 2014; Yin et al., 2018). IBD has been shown to be associated with alterations in the human gut microbial composition (Willing Lin et al. Effects of Quercetin on Colitis

et al., 2010; Tong et al., 2013). Decreased microbiome diversity has been observed in CD patients (Ma et al., 2018) and in monozygotic discordant twins with CD (Dicksved et al., 2008). Decreased microbiome diversity has mainly been attributed to reduced diversity of Firmicutes members and has been linked to temporal instability in both CD and UC (Coleman et al., 2014). Decreased diversity has also been observed in inflamed and noninflamed tissues, and CD patients generally exhibit reduced bacterial loads at inflammation sites (Willing et al., 2010; Zhu et al., 2018).

Dysregulation of the mucosal immune system can cause IBD and a pathogenic immune response against gut flora (Xu et al., 2014; Haag and Siegmund, 2015). The present study suggests that quercetin affects the progress of microbiota-associated diseases. Notably, quercetin supplementation increased gut microbial diversity, which may improve gut protection. In IBD patients, gut microbiota dysbiosis is a common occurrence, typically manifesting as a superfluity of facultative anaerobic bacteria and a simultaneous deficiency of obligate anaerobic bacteria of the classes Bacteroidia and Clostridia (Minamoto et al., 2015; Stecher, 2015). According to Wu et al., long-term dietary patterns may affect the ratios of Bacteroides, Firmicutes, and Prevotella populations, whereas short-term dietary changes may show limited effects (Wu et al., 2011). In addition, Zimmer et al. stated that strict vegan or vegetarian diets significantly decrease *Bacteroides*, Enterobacteriaceae, and *Bifidobacterium* populations (Zimmer et al., 2012). Enterobacteriaceae populations have been consistently found to be elevated in IBD patients. Therefore, further studies are warranted to evaluate the effects of both long-term and short-term dietary changes on gut microbiota and consequently on IBD (Azad et al., 2018; Guan and Lan, 2018).

Inflammation can also considerably affect the gut microbiota. Severe inflammation has been reported to increase the relative abundance of *Salmonella* and similar pathogens (Varshney et al., 2013; Saltzman et al., 2018). In the present study, *C. rodentium*infected mice showed significantly increased levels of proinflammatory cytokines IL-6, TNF-α, and IL-17 in the colonic mucosal tissues compared with the CTRL mice (not infected). It has been demonstrated that quercetin protects against colonic damage linked to increases in the TNF-α level (Donder et al., 2018; Ju et al., 2018). In the present study, quercetin supplementation reduced localized production of inflammatory cytokines, which

#### REFERENCES


in turn promoted alterations in bacterial flora composition associated with rapid repair. This finding indicates that quercetin might restore the appropriate host-microbe relationship required to manage colitis by restoring the proinflammatory, anti-inflammatory, and bactericidal functions of enteric macrophages (Ju et al., 2018).

In summary, the findings of this study on the *C. rodentium*infected mouse model demonstrated that quercetin could reduce the pathological effects of *C. rodentium*. This finding suggests that dietary quercetin can directly stimulate the immune system to reduce inflammation and restore gut microbial balance. Future studies using human subjects are desirable to confirm these effects of quercetin on inflammatory markers and provide a more comprehensive understanding of the quercetin-induced variations in the human gut microbiota.

#### DATA AVAILABILITY

All the data are available upon reasonable request at Dr. Rui Lin, pubmed1128@126.com.

#### ETHICS STATEMENT

The current research was conducted according to Chinese animal welfare guidelines and following the granting of approval by the Animal Care and Use Committee of General Hospital of the Tianjin Medical University.

#### AUTHOR CONTRIBUTIONS

RL designed the experiment. RL, MP, and YS performed the experiment and statistical analysis. RL finished the draft of the manuscript. MP and YS revised the manuscript. All the authors read and approved the manuscript.

#### FUNDING

The project was supported by the National Natural Science Foundation of China (81600509).


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2019 Lin, Piao and Song. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# IgA-Targeted Lactobacillus jensenii Modulated Gut Barrier and Microbiota in High-Fat Diet-Fed Mice

Jin Sun1,2† , Ce Qi<sup>2</sup>† , Hualing Zhu<sup>2</sup> , Qin Zhou<sup>3</sup> , Hang Xiao<sup>4</sup> , Guowei Le<sup>2</sup> , Daozhen Chen<sup>3</sup> \* and Renqiang Yu<sup>3</sup> \*

<sup>1</sup> State Key Laboratory of Food Science and Technology, Jiangnan University, Wuxi, China, <sup>2</sup> School of Food Science and Technology, Jiangnan University, Wuxi, China, <sup>3</sup> The Affiliated Wuxi Maternity and Child Health Care Hospital of Nanjing Medical University, Wuxi, China, <sup>4</sup> Department of Food Science, University of Massachusetts, Amherst, MA, United States

#### Edited by:

Yuheng Luo, Sichuan Agricultural University, China

#### Reviewed by:

Manoj Kumar, ICMR-National Institute for Research in Environmental Health, India Giorgio Giraffa, Research Centre for Animal Production and Aquaculture (CREA), Italy

#### \*Correspondence:

Renqiang Yu yurenqiang@njmu.edu.cn Daozhen Chen chendaozhen@163.com

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Microbial Symbioses, a section of the journal Frontiers in Microbiology

Received: 06 March 2019 Accepted: 08 May 2019 Published: 24 May 2019

#### Citation:

Sun J, Qi C, Zhu H, Zhou Q, Xiao H, Le G, Chen D and Yu R (2019) IgA-Targeted Lactobacillus jensenii Modulated Gut Barrier and Microbiota in High-Fat Diet-Fed Mice. Front. Microbiol. 10:1179. doi: 10.3389/fmicb.2019.01179 IgA-coated Lactobacillus live in the mucous layer of the human or mammalian intestine in close proximity to epithelial cells. They act as potential probiotics for functional food development, but their physiological regulation has not yet been studied. We isolated IgA-targeted (Lactobacillus jensenii IgA21) and lumen lactic acid bacterial strains (Pediococcus acidilactici FS1) from the fecal microbiota of a healthy woman. C57BL/6 mice were fed a normal (CON) or high fat diet (HFD) for 6 weeks and then treated with IgA21 or FS1 for 4 weeks. HFD caused dyslipidemia, mucosal barrier damage, and intestinal microbiota abnormalities. Only IgA21 significantly inhibited dyslipidemia and gut barrier damage. This was related to significant up-regulation of mucin-2, PIgR mRNA expression, and colonic butyrate production (P < 0.05 vs. HFD). Unlike IgA21, FS1 caused a more pronounced gut dybiosis than did HFD, and, in particular, it induced a significant decrease in the Bacteroidales S24-7 group and an increase in Desulfovibrionaceae (P < 0.05 vs. CON). In conclusion, IgA-coated and non-coated lactic acid bacteria of gut have been demonstrated to differentially affect the intestinal barrier and serum lipids. This indicates that IgA-bound bacteria possess the potential to more easily interact with the host gut to regulate homeostasis.

Keywords: high fat diet, immunoglobulin A, Lactobacillus jensenii, mucosal barrier, microbiota, hyperlipidemia

# INTRODUCTION

The availability of inexpensive, processed fatty foods promotes the spread of obesity worldwide, which is one of the greatest risk factors for the development of metabolic diseases (Ford et al., 2017). Disruption of homeostasis in the gut microbiota is causally linked to the development of host metabolic diseases, including obesity (Rajani and Jia, 2018). Chronic, modest elevations in fasting serum endotoxins can be induced in lean mice that consume a high-fat diet (HFD) (Cani et al., 2007). This results in unfavorable alterations in the gut microbial composition, leading to increased intestinal permeability (Cani et al., 2009). The subsequent translocation of bacteria or their products can result in chronic tissue inflammation, ultimately triggering metabolic diseases

and insulin resistance. Therefore, due to its safety and ease of operation, the regulation of intestinal microbiota by food containing probiotics, prebiotics, or polyphenols has attracted wide attention. It has been found that oral administration of specific strains of Lactobacillus and Bifidobacterium prevented or alleviated metabolic syndrome in animal experiments (Wang et al., 2015). The effect of Lactobacillus on HFD-induced obesity, however, is strain-dependent (Qiao et al., 2015), indicating that the interaction between bacteria and their hosts is very complex, highlighting the necessity to understand its mechanism of action to properly screen strains. There is also a need to establish a targeted screening method for the protection of host mucosal barriers.

Secretory immunoglobulin A (sIgA) is an antibody secreted by the mucosal tissue, and it is concentrated in the outer layer of the gut mucus coating specific local microbiota (Rogier et al., 2014). IgA-targeted microbiota derived from healthy individuals may protect host barrier integrity (Kau et al., 2015; Macpherson et al., 2015), and this is related to specific nonpathogenic commensals, including lactobacilli, clostridial species, and Akkermansia muciniphila (Bunker et al., 2015; Planer et al., 2016). IgA-coated bacteria often reside in close proximity to the epithelial surface, in contrast to IgA-free bacteria that live within the intestinal lumen (Pabst et al., 2016). IgA-coated bacteria possess a greater chance of acting on intestinal epithelial cells at a closer distance than lumen bacteria due to a more potent uptake of secreted IgA from these bacteria. Conversely, mucous layer renewal speed is very rapid (Johansson, 2012), and the bacteria bound by sIgA will slough off and be excreted with the feces, allowing for their subsequent isolation. Many Lactobacillus species are known to possess the most probiotic potential. A number of Lactobacillus species can be used in food production, and it is, therefore, of value to isolate IgA-bound Lactobacillus. Although Lactobacillus is a rare species that is estimated to constitute approximately 0.3% of all bacteria within the human colon (Almonacid et al., 2017), it may be enriched in IgA-targeted microbiota, allowing for easier isolation. This study aimed to isolate IgA-targeted Lactobacillus from healthy humans and to provide insights into their potential to protect the mucosal barrier in HFD-fed mice.

# MATERIALS AND METHODS

#### Materials

Pig gastric mucin and fluorescein isothiocyanate-labeled 4.4 kDa dextran (FD4) were both purchased from Sigma-Aldrich (St. Louis, MO, United States). Kits for plasma total cholesterol (TC), triglyceride (TG), low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C), and a kit for catalase (CAT) activity were purchased from Nanjing Jiancheng Bioengineering Institute (Nanjing, Jiangsu, China). ELISA kits for tumor necrosis factor α (TNF-α), interleukin 6 (IL-6), lipopolysaccharide (LPS), and lipopolysaccharide-binding protein (LBP) were purchased from Huijia Biotechnology Co., Ltd. (Xiamen, China). The BCIP/NBT kit for intestinal alkaline phosphatase (IAP) was purchased from Beyotime Biotechnology (Jiangsu, China). The other solvents and reagents were all analytical grade (Sinopharm Chemical Reagent, Shanghai, China).

# Isolation and Identification of IgA-Coated Lactic Acid Bacteria

IgA-coated bacteria were enriched and cultured from fecal samples obtained from 12 healthy females provided by the affiliated Changzhou Maternity and Child Health Care Hospital of Nanjing Medical University according to a protocol (JN20130918) approved by the Internal Ethics Committee of the Institute of Chinese Medical Sciences, Jiangnan University. Written informed consent was obtained from all donors. Participants were women of 20–35 years of age in good health according to self-report who did not smoke or drink alcohol. Exclusion criteria included mastitis, any infectious disease (especially tuberculosis, viral hepatitis, and human immunodeficiency virus infection), cardiovascular disease, metabolic disease (such as diabetes), mental health disorders, cancer or other malignant or degenerative diseases, inability to answer questions, and current participation in any other study related to nutrition or drug intervention. The basic physical characteristics and blood lipid profiles of the patients were within normal range (**Supplementary Table S1**). The composition of cultured microbiota was analyzed by 16S rRNA gene amplicon sequencing.

IgA-coated bacteria were collected from feces using a magnetic bead-based enrichment. Briefly, feces suspended at 20% in pre-reduced phosphate-buffered saline (PBS) containing 0.5% Tween 20 (PBST) and protease inhibitors (1 mg/mL leupeptin, 1.6 mg/mL aprotinin, Sigma-Aldrich, St. Louis, MO, United States) were homogenized and centrifuged at 400 × g to remove large debris. The supernatant was centrifuged at 8,000 × g to pellet bacteria and washed with PBST for three times. The bacterial pellet was resuspended in prereduced PBS supplemented with 0.25% bovine serum albumin (BSA), 5% goat serum, and biotinylated goat anti-human IgA. After washing, biotin was linked to streptavidin-coated magnetic beads. IgA-coated bacteria were separated from the suspension with the aid of a magnet. Collected bacteria were washed three times with BPST and were cultured in De Man Rogosa Sharpe (MRS) broth and gut microbiota medium (GMM) (Goodman et al., 2011). The above procedure was performed in an anaerobic glove cabinet (DROID Instruments and Equipment Co., Ltd., Shanghai, China). Lactic acid bacteria were isolated using MRS agar from the IgA-coated and IgAfree portion of the samples exhibiting the highest abundance of Lactobacillus as revealed by 16S rRNA gene amplicon sequencing. Cultures obtained from the more dilute samples were spread onto Lactobacillus anaerobic MRS plates with lactobacillus vancomycin and bromocresol green agar (LAMVAB). Single colonies randomly selected from LAMVAB were purified by streaking out twice on MRS agar under anaerobic conditions at 37◦C. Purity of the isolates was confirmed by repeated streaking and sub-culturing in fresh MRS agar followed by microscopic examination. Bacterial universal primers 27F and 1492R (**Supplementary Table S2**) were used to amplify the 16S rRNA from genomic DNA. 16S rRNA sequences were blasted at the EzBioCloud web site for bacterial identification. All isolated strains were further typed by amplified fragment length polymorphism (AFLP).

# AFLP Typing

fmicb-10-01179 May 22, 2019 Time: 17:41 # 3

For discrimination of the strains, the AFLP analysis method was performed using chromosomal DNA as described previously (Watanabe et al., 2009). Briefly, total DNA was digested with EcoRI and MseI restriction enzymes, and the DNA fragments were ligated to double-stranded restriction site-specific adaptors, specifically EcoRI-adaptors and MseI-adaptors (**Supplementary Table S1**). For the pre-selective and selective PCR amplification, primers EcoR1-core/Mse1-core and EcoRI-A/MseI-CA were used, respectively. The 5<sup>0</sup> ends of EcoRI primers were labeled with 6-carboxy-fluorescine (FAM). PCR products were analyzed on an ABI PRISM 3130xl Genetic Analyzer (Applied Biosystems), and the AFLP patterns were analyzed and extracted with GeneMapper software v4.0 (Applied Biosystems). Peak height thresholds were set at 200. Bands of the same size in different individuals were assumed to be homologous and to represent the same allele. Bands of different sizes were treated as independent loci, and data were exported in a binary format with '1' representing the presence of a band/peak and '0' representing its absence. Data were analyzed using NTSYS-pc software (Exeter Software, Biostatistics, Inc., NY, United States) version 2.1. The similarity coefficient was determined using the similarity program for qualitative data (SIMQUAL) by incorporating the Dice similarity coefficient. Cluster analysis was performed to construct a tree plot using the unweighted pair-group method with arithmetic averages (UPGMA) in the SAHN program of the NTSYS-pc software.

#### Mucus Adhesion and Cell Surface Hydrophobicity of lgA-Coated and IgA-Free Lactic Acid Bacteria

Wells of microtiter plates were coated with pig gastric mucin or BSA (control). Bacteria were suspended in PBS containing 0.5% tween 20 (PBST) to an OD<sup>600</sup> of 0.5. Bacterial suspensions (100 µL) were added to each well and incubated overnight at 4 ◦C. The wells were washed with PBST. The buffer was poured off, and after the wells dried, bound bacteria were stained with safranin., The absorbance at OD<sup>492</sup> was then measured in an enzyme-linked immunosorbent assay (ELISA) plate reader. All measurements were performed in triplicate.

Bacterial surface hydrophobicity was measured using the bacterial adhesion to hydrocarbon (BATH) assay as previously described (Rosenberg, 1984). Bacterial cultures were collected at stationary phase and were pelleted by centrifugation. The pellet was washed twice and suspended in KH2PO<sup>4</sup> (0.01 mmol/L, pH 7.0) to an OD<sup>600</sup> of 0.5 ± 0.05 (A0). A volume of 0.15 mL of hexadecane was added to the 4-mL cell suspension, and the mixture was vortexed for 2 min. Samples were stored for 30 min to let phases separate. The absorbance (OD600) of the aqueous phase (A) was again determined. Results were expressed as percentage attachment to hexadecane = (1 - A/A0)/100.

# Animal Experiments

The animal experiments were performed according to the National Guidelines for Experimental Animal Welfare (MOST of PR China, 2006), and the Jiangnan University Animal Ethics Committee approved all experiments under protocol number 128/16. Four-week-old male C57BL/6 mice were fed ad libitum with chow diet or HFD (45% energy from fat, **Supplementary Table S3**) for 10 weeks. Control mice fed with chow diet (CON) and then gavaged daily with 0.1 mL of sterile PBS. Mice in the experimental groups were fed with HFD for 6 weeks and then gavaged daily with 0.1 mL of sterile PBS (HFD), 10<sup>9</sup> CFU/mL of Lactobacillus jensenii IgA21 (HFD + J), or Pediococcus acidilactici FS1 (HFD + P) in PBS, for 4 weeks. Mice were allowed free access to food and water and maintained under a 12-h light-dark cycle at 24◦C and constant humidity in soundproof cages. Body weight and water intake were measured every week during the 10 weeks of diet. Food intake was monitored from the first week of treatment. After 10 weeks of treatment, a Comprehensive Laboratory Animal Monitoring System (CLAMS) (Columbus Instruments, Columbus, OH, United States) was used to determine the respiratory exchange ratio (RER). Ambulatory locomotor activity was measured by consecutive beam breaks in adjacent beams. Heat production was calculated by multiplying the calorific value (CV) (3.815 + 1.232 × RER) by the observed VO<sup>2</sup> (Heat = CV × VO2). Mice were sacrificed by cervical dislocation, and a blood sample of ∼100–200 µL was obtained by cardiac puncture. The liver and white adipose tissue were then removed, cleared of blood, and transferred to pre-chilled Eppendorf tubes on ice for weighing. Tissue samples were weighed and fixed in 4% formalin solution or stored at −80◦C for further experiments.

A portion of the inguinal white adipose tissue (WAT) and liver was immediately fixed using 4% neutral buffered formalin for 3 days. Tissues were dehydrated and embedded into paraffin for preparation tissue slice (6 µm) and haematoxylin and eosin (H&E) staining. Other samples were stored at −80◦C for further analysis.

#### Determination of Intestinal Permeability

Mice were fasted for 4.5 h and then gavaged with 100 µL of 22 µg/µL FD4 at 13:00. Serum was obtained at 14:00 by decapitation. The serum FD4 concentration was calculated by comparing samples to serial dilutions of known standards using a Synergy HT fluorometer (BioTek, Winooski, VT, United States) with excitation at 485 nm and emission at 530 nm. A gain of 50 was used for all experiments.

#### Tissue and Blood Collection and Plasma Analysis

TC, TG, LDL-C, and HDL-C of serum were examined using the corresponding enzymatic colorimetric assay kits according to the manufacturer's instructions. Blood glucose was measured using a glucometer (Accu-Check; Roche Diagnostics, Madrid, Spain). TNF-α, IL-6, LPS, and LBP were assayed by ELISA kit according to the manufacturer's instructions. IAP activity was determined using the BCIP/NBT kit.

# Analysis of Community Structure by 16rRNA Gene Amplicon Sequencing

The community structure of colonic microbiota was analyzed by 16S rRNA gene amplicon sequencing. Total genomic DNA was extracted from two BAC samples using a soil DNA extraction kit (Mo Bio Laboratories, Carlsbad, CA, United States) following the manufacturer's protocol. PCR amplicon libraries were constructed for Illumina MiSeq sequencing using bacterial primers targeting the V3-4 hypervariable regions of the 16S rRNA genes (Goodman et al., 2011). Each 20 µL reaction mixture included 5 × FastPfu Buffer, 2.5 mM dNTPs, FastPfu Polymerase, 5 µM of each primer, and 10 ng of template DNA. The PCR profile was set as follows: 95◦C for 5 min and 27 cycles at 95◦C for 30 s, 55◦C for 30 s, and 72◦C for 45 s, with a final extension at 72◦C for 10 min. Reads from the original DNA fragments were merged using FLASH<sup>1</sup> (Magoc and ˇ Salzberg, 2011), and quality filtering was performed according to the literature (Caporaso et al., 2010). Sequencing data were processed using the Quantitative Insights Into Microbial Ecology (QIIME) pipeline. Operational Taxonomic Units (OTUs) were selected using a de novo OTU selection protocol with a 97% similarity threshold. The taxonomic identities of the bacterial sequences were determined using the RPD classifier (Wang et al., 2007). The microbial diversity was analyzed using QIIME software15 with Python scripts. Taxonomy assignment of OTUs was performed by comparing sequences to the Greengenes database (gg\_13\_5\_otus). The raw data were uploaded to the SRA database, and the BioProject ID is PRJNA523678.

#### Statistical Analyses

The statistical significance of the comparisons between multiple groups was performed by ANOVA followed by Tukey's or Duncan post hoc tests with normal distributions. Nonparametric data were analyzed with the Kruskal-Wallis H test. Data from microbiota sequencing were analyzed online by MicrobiomeAnalyst (Dhariwal et al., 2017) to calculate alpha diversity scores and Bray–Curtis distances. The UniFrac metric was used to determine the dissimilarity between any pair of bacterial communities. The similarity relationship, assessed using the UniFrac metric, was presented in PCoA (Principal Coordinate Analysis) plots. To identify differences in microbial communities between the three groups, analysis of similarities (ANOSIM) was performed between each pair. Differences in a specific genus were analyzed using the EdgeR package provided by MicrobiomeAnalyst that uses shrinkage estimators, fold change values, and controls false discovery rate by calculating adjusted P-Values.

# RESULTS

# Community Structure of Cultivable IgA-Coated Microbiota

IgA-coated microbiota was separated using a magnetic sorting method from one feces sample provided by a healthy female.

<sup>1</sup>http://ccb.jhu.edu/software/FLASH/

To recover IgA-coated bacteria with the best diversity, enrichment of bacteria was performed using MRS medium and gut microbiota medium (GMM). Typical subgroup bacteria were cultured with MRS medium and GMM according to Venn diagrams comparing the operational taxonomic unit (OTU) memberships (**Supplementary Figure S1a**). Seven rare OTUs were also enriched. The composition of IgA-free microbiota was similar to that of total feces microbiota. GMM-enriched IgA-coated and free microbiota both similar compositions (**Supplementary Figure S1b**). The MRS-enriched IgA-coated microbiota, however, was clearly different, and it was composed of Bifidobacterium, Escherichia, and Lactobacillus. A genus phylogenetic tree was constructed from sequencing data of MRSand GMM-cultured IgA-coated microbiota obtained from 12 healthy female donors. The IgA-coated microbiota primarily consisted of Firmicutes, Proteobacteria, and Actinobacteria (**Figure 1A**). The most frequently occurring genera were Enterococcus (6/12), Streptococcus (6/12), Bifidobacterium (7/12), and Lactobacillus (5/12) (**Figure 1B**).

#### Isolation and Identification of IgA-Coated and Lumen Lactic Acid Bacteria

An enriched sample of IgA-coated bacteria containing a high Lactobacillus level was chosen as a source from which to isolate lactic acid bacteria. A portion of the larger milky white opaque colonies on MRS agar that exhibited irregular edges were identified as Weissella confusa. Other colonies were transparent and smaller with regular edges. All strains isolated from IgA-positive microbiota were identified as L. jensenii (**Supplementary Table S4**). All strains isolated from IgA-free (lumen) microbiota were identified as Pediococcus acidilactici, which is similar to Pediococcus lolii NGRI 0510QT (Wieme et al., 2012). L. jensenii and P. acidilactici were included in 3 AFLP patterns (**Supplementary Figure S3** and **Supplementary Table S4**). We then measured the cell surface hydrophobicity of these bacteria, as this property is important for bacteria adhesion within the gut (Krasowska and Sigler, 2014). The typical strains of L. jensenii present in different AFLP patterns exhibited a higher cell surface hydrophobicity than that of Pediococcus acidilactici (**Figures 2A,B**). It has been suggested that the aggregation and adhesion characteristics of beneficial bacteria contribute to intestinal colonization, thereby enhancing the mucosal barrier to resist pathogenic bacteria infection (Tareb et al., 2013). L. jensenii isolated in our study displayed typical self-aggregation, while P. acidilactici did not. At the same time, L. jensenii exhibited straindependent mucus adhesion, with IgA21 exhibiting the highest adhesion (**Figure 2C**). All P. acidilactici strains, however, showed very weak mucus adhesion in vitro (**Figure 2D**). Additionally, IgA21 and P. acidilactici FS1 are typical slow-growing and fast-growing bacteria, respectively. (**Supplementary Figure S2**).

#### General Features of HFD-Fed Mice

From the 4th week of the animal experiments, the weight of the mice in the group fed with HFD was significantly

higher from that of the normal diet fed group (P < 0.01) (**Supplementary Figure S4a**). Notably, IgA21 and FS1 exhibited no significant effect on weight gain compared to that observed with HFD fed control mice (**Supplementary Figure S4b**). The liver of mice in the HFD + P group exhibited slight fat infiltration and white adipose tissue cell enlargement, but the

FIGURE 3 | Effects of Lactobacillus jensenii IgA-21 and Pediococcus acidilactici FS-1 on the respiratory exchange ratio (RER) (A,B), heat (C,D), and ambulatory activity (E,F) in high-fat diet (HFD) mice. Bar graphs represent median range values during the active cycle over 48 h. CON, control group; HFD, high-fat diet group; HFD + J, high-fat diet group treated with L. jensenii IgA-21; HFD + P, high-fat diet group treated with P. acidilactici FS1. RER, respiratory exchange ratio. Data of A, C and E are presented as mean ± SD (n = 8). <sup>∗</sup>P < 0.05 (HFD versus CON); ∗∗P < 0.01 (HFD versus CON); #P < 0.05 (bacterial treatment versus HFD). Comparisons were made with Kruskal-Wallis ANOVAs and the Bonferroni post hoc method to arrive at adjusted P-values.

liver structure and adipocytes of mice treated with J were normal (**Supplementary Figures S4c,d**).

During the 24-h monitoring period in CLAMS, the RER of mice in the HFD group ranged from 0.72 to 0.81 (**Figures 3A,B**), and heat production was significantly lower than that of mice in CON group (P < 0.01, **Figures 3C,D**). HFD induced a significant increase in ambulatory activity (P < 0.01), which was inhibited by both IgA21- and FS1-treatment (P < 0.05) (**Figures 3E,F**).

The serum T-CHOL (P < 0.05), TG (P < 0.01), and LDL-C (P < 0.01) of mice fed with HFD were all increased significantly. Only IgA21 treatment significantly inhibited the increase in TG (P < 0.01) (**Table 1**).

#### Effects of IgA21 on Endotoxemia and Systemic Inflammation in Mice Fed With HFD

For marker of endotoxemia, serum endotoxin and LBP were measured using TNF-α and IL-6 levels as indices of systemic inflammation (**Figure 4**). HFD feeding increased circulating concentrations of serum endotoxin (P < 0.01), LBP (P < 0.05), and TNF-α (P = 0.058) (**Figure 4**). Only IgA21 significantly inhibited HFD-induced endotoxemia and significantly reduced serum LBP and TNF-a (P < 0.04 vs. HFD) (**Figure 4D**).

TABLE 1 | Effect of Lactobacillus jensenii IgA-21 and Pediococcus acidilactici FS-1 on serum lipid profile and glucose in HFD feeding mice.


Data are mean ± SD (n = 8 for each group), <sup>∗</sup>P < 0.05, ∗∗P < 0.01 compared with CON; #P < 0.05, ##P < 0.01 compared with HFD; Comparisons were made with Kruskal-Wallis ANOVAs and the Bonferroni post hoc method to arrive at adjusted p-values. CON, control group; HFD, high-fat diet group; HFD + J, highfat diet group treated with L. jensenii IgA-21; HFD + P, high-fat diet group treated with P. acidilactici FS1. LDL-C, Low Density Lipoprotein Cholesterol; HDL-C, High Density Lipoprotein Cholesterol, TG, triglyceride; T-CHOL total cholesterol.

# Effects of IgA21 on Mucosal Barrier-Related Marker in HFD-Fed Mice

To confirm if reversed metabolic endotoxemia by IgA21 was related to improved gut integrity, we measured in vivo intestinal permeability to FD4, a selective marker of paracellular permeability (Laukoetter et al., 2006). Serum FD4 concentrations were significantly increased in HFD mice 4 h after gavage, and these concentrations were significantly decreased after IgA21 treatment (P < 0.05). To identify candidates targeted by IgA21 in modulation of gut barrier, we initially studied colonic mRNA expression of several tight junction proteins, mucin-2, polymeric immunoglobulin receptor (PIgR), and mRNA levels of RegIIIγ antimicrobial peptides that play an important role in the establishment and maintenance of the mucosal barriers (Okumura and Takeda, 2017). IgA21 significantly prevented HFD-induced down-regulation of Mucin-2 and PIgR mRNA expression (**Figure 5A**). FS1 could not inhibit the downregulation of these two genes, and it even led to a more significant decrease of mucin 2 expression compared to CON levels (P < 0.05) (**Figure 5A**). We further determined the activity of IAP enzyme, which is capable of detoxifying LPS (Bates et al., 2007). IAP activity in the colons of mice fed with HFD decreased significantly (P < 0.05) (**Figure 5B**), and this was accompanied by a significant increase in TNF-a production (P < 0.05, **Figure 5C**). Both bacterial treatments significantly inhibited the decrease in IAP activity (P < 0.01 vs. HFD), while IgA21 treatment also inhibited the excessive production of TNF-a (P < 0.05). Finally, we studied the effects of two bacteria on the production of short-chain fatty acids within the colon, where butyric acid is reported to play a role in maintaining mucosal barrier (Peng et al., 2009). The production of lactic acid, acetic acid, propionic acid, and butyric acid in colon lumens decreased significantly after HFD feeding (P < 0.05 vs. CON) (**Figure 5D**). IgA21 inhibited the decrease in butyric acid, while FS1 only alleviated the decrease in propionic acid production (P < 0.05 vs. HFD). Additionally, colonic CAT activity was also sensitive to HFD and was significantly decreased (P < 0.05), and this was reversed after IgA21 treatment (P < 0.05 vs. HFD, **Supplementary Figure S5**).

A significant negative correlation was observed between mucin-2 mRNA expression and serum LBP concentration (r = −0.42, P = 0.04) (**Figure 5E**). Colon butyric acid and serum endotoxin concentration exhibited a negative correlation trend (r = −0.47, P = 0.051). PIgR mRNA expression was negatively correlated with LPS (r = −0.463, P = 0.02) and LBP (r = −0.478, P = 0.02). This further indicates that mucin-2, the PIgR gene, and butyric acid may mediate the protective effect of IgA21 on the mucosal barrier.

#### Differential Modulation of Gut Microbiota Structure by IgA21 and FS1

We studied the gut microbiota of four groups of mice by DNA sequencing. PCR was used to amplify the V3-V4 hypervariable regions of the 16S rRNA gene, and products were then sequenced after multiplexing. yielding yield of 621,991 high quality sequences was obtained. Sequences were clustered into operational taxonomic units (OTUs) with 97% pairwise sequence identity, and they were then assigned with taxonomies. The weighted UniFrac analysis, a method sensitive to taxa abundances for beta-diversity analysis, expanded dramatically after HFD feeding, especially in the HFD+P group (R = 0.85, P < 0.01 by ANOSIM test) (**Figure 6A**). The unweighted UniFrac analysis, which is sensitive to rarer taxa, showed a similarity in the microbiota between HFD and HFD+J group, and a significant shift of HFD+P from other groups was observed (R = 0.68, P < 0.01 by ANOSIM test) (**Figure 6B**). Dendrogram of Hierarchical clustering analysis with distance measure based on Bray-Curtis index also confirmed differences between all HFDfed mice and the CON group (**Figure 6C**). Alpha diversity analysis of gut microbiota indicated that HFD induced a significant decrease in Shannon index (P < 0.05) and in simpson index (P < 0.05). No significant difference existed between the two indicators in mice treated with the two bacteria and in mice of the CON or the HFD group (**Figures 6D,E**). A significant increase was observed in the ratio of Bacteroidetes to Firmicutes in the HFD group compared to that of CON (P < 0.05, **Figure 6F**).

LEfSe analysis showed that Bacteroidales S24-7 group and Lachnospiraceae were the marker family of the CON group (**Figures 6G,H**). The typical marker genus of the HFD group is Akkermansia of the Verrucomicrobiaceae family, and this was significantly increased in all mice fed with HFD (p < 0.05) (**Figure 6F**). The HFD+J group was enriched with the genus Rikenellaceae RC9 gut group, consisting of an uncultured Bacteroidales bacterium and

FIGURE 5 | Effects of Lactobacillus jensenii IgA-21 and Pediococcus acidilactici FS-1 on mucosal barrier related markers. (A) mRNA levels of tight junction-related proteins, mucin-2, polymeric immunoglobulin receptor (PIgR), and Reg-γ of ileum measured using quantitative reverse transcription polymerase chain reaction (RT-qPCR). Gene expression was normalized to β-actin. (B) Alkaline phosphatase activity of ileum. (C) Tumor necrosis factor α (TNF-α) and interleukin 6 (IL-6) production in colon determined by ELISA. (D,E) Correlation between serum lipopolysaccharide-binding protein (LBP) and mucin-2 and PIgR mRNA expression; between serum lipopolysaccharide (LPS) and fecal butyric acid and mucin-2 mRNA expression. CON, control group; HFD, high-fat diet group; HFD + J, high-fat diet group treated with Lactobacillus jensenii IgA-21; HFD + P, high-fat diet group treated with Pediococcus acidilactici FS-1. All data are presented as mean ± SD (n = 8). <sup>∗</sup>P < 0.05, ∗∗P < 0.01 (HFD versus CON); #P < 0.05, ##P < 0.01 (bacterial treatment versus HFD). Comparisons were made with one-way ANOVA followed by Tukey's multicomparison test.

Coriobacteriaceae\_UCG-002 (**Figure 6G**). Alistipes was the highest marker genus for the HFD+P group (**Figure 6G**). There was a greater inhibition of the marker genus in the CON group, including lactobacillus, than was detected in mice fed a HFD (**Table 2**).

At the family level, FS1 treatment induced a significant decrease in the Bacteroidales 24-7 group and an increase in Desulfovibrionaceae (P < 0.05 vs. CON) (**Figure 6F**). At the genus level (**Table 2**), IgA21 significantly increased the Eubacterium coprostanoligenes group, the Clostridium sensu stricto 19 group,



Data were expressed as log2 transformed fold change (log2FC); The differential expression analysis was carried out by edgeR. Only species with FDR ≤ 0.05 and absolute log2FC ≥ 1.5 are shown; FDR, Adjusted P-value.

and the Rikenellaceae RC9 gut group while causing a decrease in Tyzzerella and Erysipelatoclostridium compared to levels detected in CON (FDR < 0.01). FS1 mainly promoted the proliferation of Lachnospiraceae UCG 001, Oscillibacter, Ruminococcaceae UCG 005, and Mucispirillum, but it inhibited Staphylococcus and Enterorhabdus (FDR < 0.01).

#### DISCUSSION

This study confirmed that Lactobacillus was the commonly found genus of IgA-coated bacteria in human feces, which is in agreement with other studies (Bjursell et al., 2008). L. jensenii is the main species that we isolated from a healthy woman. This species was previously found to be dominant in the vaginal microbiota of healthy women (Pavlova et al., 2002), and it is known to show a strong ability to form biofilms (Ventolini et al., 2015). It is, however, rarely encountered in human feces (Rossi et al., 2016) and almost disappears in individuals with IgA defects (Lonnermark et al., 2012). Compared to the specific lumen lactic acid bacteria FS1, IgA21 exhibits stronger surface hydrophobicity, mucus adhesion, and a higher growth rate, suggesting that it may be more adaptable to a mucus environment. This demonstrates that IgA -coated bacteria may live near epithelial cells in the mucous layer of the human or mammalian intestine (Pabst et al., 2016).

In the following short term HFD feeding experiment, mice showed significant weight gain without development of obesity (weight gain less than 20%). We found that HFD induced a significant increase in ambulatory activity. This is in agreement with a study by Kohsaka et al. (2007) who found an increase in the free-running period in mice within 6 weeks on a HFD. Although the two bacteria significantly reduced the ambulatory activity of mice fed with HFD, they did not significantly increase body weight, suggesting that the ambulatory activity of mice was independent of body weight.

Probiotics may prevent the occurrence of HFD-induced dyslipidemia by regulating intestinal flora. (Aron-Wisnewsky et al., 2013). In the present study, it was demonstrated that IgA21, but not FS1, significantly lowered hypertriglyceridemia and hypercholesterolemia in HFD-fed mice. The results also confirmed alteration of gut barrier in HFD-fed mice following endotoxemia and low-grade inflammation marked by increased serum TNF-α, and treatment with IgA21 restored the damaged barrier. This low-grade inflammation alters lipid metabolism (Cani et al., 2007). It was found that TNF-α can suppress

lipoprotein lipase synthesis, contributing to hypertriglyceridemia due to decrease of peripheral clearance rate of triglyceride (Kawakami et al., 1987). De novo fatty acid synthesis would be stimulated by low doses of LPS in the liver in conjunction with increased lipolysis, providing more fatty acids for hepatic TG production (Feingold et al., 1992). The role of IgA21 in alleviating lipid disorders may be related to its enhancement of the mucosal barrier, which can inhibit endotoxemia and chronic inflammation. Intestinal barrier is the result of interaction between host and microorganism, including physical barrier, immune barrier, and microbial colonization resistance. In this study, IgA21 may regulate the mucosal barrier in three ways. First, it upregulated the expression of mucin 2, which is important for the formation of the mucous layer that separates epithelial cells from harmful antigens and is an important chemical barrier to prevent pathogenic bacterial infection (Wagner et al., 2018). It has been found that a HFD decreases gut goblet cell expression of mucin-2 associated with reduced expression of Zonula occludens-1 and Occludin mRNA (Lee et al., 2017). Second, PIgR is a gene related to mucus secretion of sIgA (Johansen and Kaetzel, 2011), which is a major target of IgA21. IgA secreted within the gut plays an important role in maintaining gut immunological barrier function by inhibiting pathogens from adhering to the mucous, thus exerting effects in the lumen [29]. Third, IgA21 elicits anti-inflammatory activity within the intestine that is marked by decreased TNF-alpha production and increased activity of IAP. Other L. jensenii strains such as TL2937 were also found to mediate the induction of negative regulators of TLRs and prevent intestinal inflammatory damage (Shimazu et al., 2012). TNF-α is a key cytokine that can lead to mucosal barrier damage by inducing cytoskeleton depolymerization (Gibson, 2004). Last, IgA21 promotes the production of butyrate to enhance gut barrier function (Peng et al., 2009). Up-regulation of Claudin-2 can potentially affect the structure and function of tight junctions, resulting in barrier dysfunction (Wang et al., 2017). IgA21 administration also prevented HFD induced over-expression of Claudin-2, which may be related to stimulation of butyrate production. It has found that butyrate represses claudin-2 mRNA expression, a gene related to permeability-promoting tight-junction proteins (Zheng et al., 2017), through an IL-10 receptor A-dependent mechanism (Zheng et al., 2017).

This study also demonstrated that short-term HFD caused significant changes in the structure of intestinal microbiota, and FS1 treatment caused further disturbance, while IgA21 did not exert a significant impact according to beta diversity. The mucus digesting bacteria Akkermansia was the typical marker bacterium that was increased by HFD. The impact of HFD on Akkermansia is controversial, where some studies have demonstrated that a HFD reduced gut Akkermansia (Everard et al., 2013), while other studies demonstrated the opposite result (Carmody et al., 2015). Recently, it has also been found that mice receiving a highfat, ketogenic diet for 2 weeks exhibited a significant increase in gut Akkermansia (Olson et al., 2018), which suggested that mucus of mice fed a high-fat diet became the main carbon source to support their growth. The main difference between FS1 and IgA21 is that the former significantly inhibits the proliferation of the Bacteroidales S24-7 group and promotes the proliferation of Desulfovibrionaceae at the family level (**Figure 6H**). S24-7, an IgA-targeted gut commensal in humans and mouse, is involved in carbohydrate fermentation and utilization (Palm et al., 2014; Ormerod et al., 2016). Desulfovibrionaceae are sulfate-reducing and endotoxin-producing bacteria that were highly enriched in mice by long term high-fat feeding (6 months) and were associated with the development of metabolic syndromes (Zhang et al., 2010). They play a role in reducing sulfate to H2S to damage the gut barrier. These changes in the two types of bacteria indicated that FS1 promoted intestinal bacterial imbalance induced by HFD.

At the genus level, IgA21 is associated with enriched Eubacterium coprostanoligenes group, a typical cholesterolreducing bacterium (Li et al., 1998), which may explain to some extent its ability to prevent hyperlipidemia. Mucispirillum is a genus specially enriched by FS1treatment, and this bacteria is a pathobiont within the mucus layer of the rodent gut which is associated with inflammation (Loy et al., 2017). This may be partially related to the failure of FS1 to maintain the mucosal barrier effectively.

# CONCLUSION

In conclusion, IgA-coated and non-coated lactic acid bacteria within the gut have been shown to differentially affect the intestinal barrier and serum lipids. This indicates that IgA-bound bacteria possess the potential to more easily interact with the host gut to regulate homeostasis.

# DATA AVAILABILITY

The raw data were uploaded to SRA database and the BioProject ID is PRJNA523678.

# ETHICS STATEMENT

IgA-coated bacteria were enriched and cultured from samples of 12 healthy female provided by the affiliated Changzhou Maternity and Child Health Care Hospital of Nanjing Medical University according to a protocol approved by the Internal Ethics Committee of the Institute of Chinese Medical Sciences, Jiangnan University. Written informed consent was obtained from all donors. This animal experiments was performed according to the National Guidelines for Experimental Animal Welfare (MOST of PR China, 2006), and approved by the Jiangnan University Animal Ethics Committee approved all experiments under protocol number 128/16.

#### AUTHOR CONTRIBUTIONS

CQ and RY participated in data collection, performed the analyses, and wrote the manuscript. HZ designed the study,

collected the data, and participated in data analysis. QZ, DC, HX, RY, and GL participated in data analysis and writing of the manuscript. HX, GL, JS, and HZ performed data preprocessing. JS participated in study design and writing of the manuscript and was responsible for overall study coordination.

#### FUNDING

This research was supported by National Natural Science Foundation of China (No. 31201805), the China Postdoctoral Science Foundation (No. 172774), Wuxi Municipal Science and Education Strengthening Health Engineering Medical Key Discipline Construction Program (No. ZDXK003), Young Talent Project (No. QNRC039), and Wuxi Municipal

#### REFERENCES


Commission of Health and Family Planning Medical Research Project (No. Q201613).

#### ACKNOWLEDGMENTS

We thank Huiyan Wang of Changzhou Maternal and Child Health Hospital for her help in recruiting volunteers to collect fecal samples.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.01179/full#supplementary-material


Toll-like receptor signaling pathway. Infect. Immun. 80, 276–288. doi: 10.1128/ IAI.05729-11


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Sun, Qi, Zhu, Zhou, Xiao, Le, Chen and Yu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fmicb-10-01179 May 22, 2019 Time: 17:41 # 13

# Microbiome-Metabolomics Analysis Investigating the Impacts of Dietary Starch Types on the Composition and Metabolism of Colonic Microbiota in Finishing Pigs

Miao Yu, Zhenming Li, Weidong Chen, Ting Rong, Gang Wang and Xianyong Ma\*

Institute of Animal Science, Guangdong Academy of Agricultural Sciences, State Key Laboratory of Livestock and Poultry Breeding, Key Laboratory of Animal Nutrition and Feed Science in South China, Ministry of Agriculture, Guangdong Public Laboratory of Animal Breeding and Nutrition, Guangdong Engineering Technology Research Center of Animal Meat Quality and Safety Control and Evaluation, Guangzhou, China

#### Edited by:

Jie Yin, Institute of Subtropical Agriculture (CAS), China

#### Reviewed by:

Shihai Zhang, South China Agricultural University, China Huihua Zhang, Foshan University, China Zhiru Tang, Southwest University, China

> \*Correspondence: Xianyong Ma maxianyong@gdaas.cn

#### Specialty section:

This article was submitted to Systems Microbiology, a section of the journal Frontiers in Microbiology

Received: 20 March 2019 Accepted: 06 May 2019 Published: 29 May 2019

#### Citation:

Yu M, Li Z, Chen W, Rong T, Wang G and Ma X (2019) Microbiome-Metabolomics Analysis Investigating the Impacts of Dietary Starch Types on the Composition and Metabolism of Colonic Microbiota in Finishing Pigs. Front. Microbiol. 10:1143. doi: 10.3389/fmicb.2019.01143 The present study used a combination of 16S rRNA MiSeq sequencing strategy and gas chromatograph time of flight mass spectrometer (GC-TOF/MS) technique to investigate the effects of starch sources on the colonic microbiota and their metabolites in finishing pigs. A total of 72 crossbred barrows were allocated to three different experimental diets with eight replicates and three pigs per replicate. The diet types included tapioca starch (TS), corn starch (CS), and pea starch (PS) (amylose/amylopectin were 0.11, 0.25, and 0.44, respectively). Results showed that the PS diet markedly increased (adjusted P < 0.05) the abundance of short-chain fatty acids (SCFAs) and lactate producers, such as Lactobacillus, Prevotella, Faecalibacterium, and Megasphaera, while decreased (adjusted P < 0.05) the abundance of Escherichia coli when compared with the TS diet. The metabolomic and biochemistry analyses demonstrated that the PS diet increased (adjusted P < 0.05) the concentrations of organic acids (acetate, propionate, butyrate, valerate, and lactate) and some macronutrients (sugars and long-chain fatty acids), and decreased (adjusted P < 0.05) the amino acids and their derivatives (leucine, glycine, putrescine, cadaverine, skatole, indole, and phenol) when compared with the TS diet. Additionally, Spearman's correlation analysis revealed that the changes in the colonic metabolites were associated with changes in the microbial composition. Correlatively, these findings demonstrated that the different dietary starch types treatment significantly altered the intestinal microbiota and metabolite profiles of the pigs, and dietary with higher amylose may offer potential benefits for gut health.

Keywords: colon, metabolic profiles, microbiota, pigs, starch sources

# INTRODUCTION

Starch, acts as a major energy source of the daily diet and is the largest fraction among human and monogastric animal diets (Yin et al., 2010). Dietary starches from different sources can affect digestion and absorption at different rates and to different extents, depending on the physicochemical properties of the starch (Giuberti et al., 2015). Moreover, the rate, extent, and

site at which the starch is degraded can cause different physiological impacts on the physiological function of the gastrointestinal system and the gut health of the host (Giuberti et al., 2015; Metzler-Zebeli et al., 2018). Generally, starch contains two types of molecules, amylose and amylopectin (Tester et al., 2004), and the digestion rate of starch is largely dependent on the proportion of amylose to amylopectin that the starch molecule contains. Amylopectin is recognized to be rapidly digested because its branched structure provides multiple sites for enzymatic hydrolysis by amylase. In contrast, amylose is a more linear glucose polymer and is not degraded in the small intestine by either pancreatic α-amylase or brush border disaccharide hydrolases. It then reaches the large intestine where it can be fermented by the resident microbiota, providing a carbon source and energy for the bacteria (Englyst et al., 1992; Lafiandra et al., 2014). Starch with a higher proportion of amylose can decrease endogenous digestibility in the small intestine and subsequently increase the digesta mass reaching the large intestine for microbial fermentation (Topping et al., 1997). Alterations in substrate degradation by intestinal microbiota can induce changes in the microbiota as well as the metabolic end products of microbial degradation (Fouhse et al., 2015; Yin et al., 2018a). Accumulating evidence has indicated that diets containing starch with a higher content of amylose can increase distal digesta mass, short-chain fatty acid (SCFA) concentration, and commensal microbial populations in the gut, including Bifidobacterium spp. and Lactobacillus (Bird et al., 2007, 2009; Regmi et al., 2011). Increases in SCFAs, especially butyrate, have many important nutritional and physiological effects on maintaining intestinal health (Newman et al., 2018). However, until now, information on the effects of different starch sources on other microbial metabolites in the gut is limited, and the relationships among starch sources, microbial community, and microbial activity is not clearly understood.

Therefore, diets containing three purified starches with clear differences in the ratio between amylose and amylopectin, were fed to pigs to test the hypothesis that dietary starches with high amylose can exert different impacts on the gut bacterial community and microbial metabolites in pigs. The present study used a 16S rRNA MiSeq sequencing strategy and combined with gas chromatograph time of flight mass spectrometer (GC-TOF/MS) technique to investigate the effects of starch sources (tapioca starch, corn starch, and pea starch) on the colonic microbial composition and microbial metabolites in pigs.

# MATERIALS AND METHODS

#### Ethics Statement

The experimental proposals and procedures for the care and treatment of the pigs were approved by the Animal Care and Use Committee of Guangdong Academy of Agricultural Sciences (Authorization No. GAASIAS-2016-017).

#### Animals, Diets, and Sampling

Seventy-two crossbred (Duroc × Landrace × Large White) growing barrows were randomly allocated to three different experimental diets based on their body weight (BW, 77 ± 0.52 kg). Each dietary group consisted of eight pens (replicates), with three pigs per pen. Pigs in the three treatments were fed tapioca starch (TS group), corn starch (CS group), or pea starch (PS group), respectively, as their dietary starch sources. The ratio of amylose to amylopectin of the three diets were 0.11, 0.25, and 0.44, respectively. The experimental diets were formulated to meet or exceed the nutrient recommendations of the National Research Council (NRC) (**Table 1**) (NRC, 2012). The diets and water were provided with ad libitum throughout the 40-day feeding trial. The feed consumption per pen was recorded every day to calculate average daily feed intake (ADFI). The BWs of all pigs were recorded at the beginning and the end of the study period to determine average daily gain (ADG).

At the end of the experiment, eight pigs from each group (n = 8 barrows, based on the average body weight in each pen) were selected and then sampled. After fasting for approximately 12 h, the pigs were euthanized by electrical stunning and exsanguination. The digesta of the colon was collected and homogenized and the pH values of the digesta was immediately determined. About 5 g of mixed colonic digesta were frozen in liquid nitrogen and then stored at −80◦C for later bacterial DNA isolation and metabolites analysis. Another 10 g of mixed colonic digesta were stored at −20◦C for starch, amylose, and amylose/amylopectin ratio analyses.

#### DNA Extraction, Illumina MiSeq Sequencing, and Data Processing

Total genomic DNA from the individual samples of colonic digesta was extracted using a QIAamp PowerFecal DNA Kit (QIAGEN, Hilden, Germany) according to the manufacturer's instructions. DNA concentrations of every sample were quantified using a Nanodrop 2000 spectrophotometer (Thermo Fisher Scientific, Wilmington, DE, United States). The genes of all bacterial 16S rRNA in the region of V3–V4 were amplified by polymerase chain reaction (PCR) using a universal forward primer 338F (5<sup>0</sup> -ACTCCTRCGGGAGGCAGCAG-3<sup>0</sup> ) and a reverse primer 806R (5<sup>0</sup> -GGACTACCVGGGTATCTAAT-3<sup>0</sup> ) (Mao et al., 2015b). PCR amplicons were purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, United States), according to the manufacturer's instructions. The purified amplicons were pooled in equimolar from each sample and paired-end sequenced (2 × 250) on an Illumina MiSeq platform (Majorbio, Shanghai, China) according to the standard protocols (Caporaso et al., 2012).

The QIIME (version 1.17) software package was used to demultiplex and quality-filter raw sequence data generated from 16S rRNA MiSeq sequencing (Campbell et al., 2010). Gaps in each sequence were discarded from all the samples to decrease the noise generated the screening, filtering, and pre-clustering processes as described previously (Gao et al., 2018). Operational taxonomic units (OTUs) were clustered as a similarity cutoff of 97% using UPARSE (version 7.1<sup>1</sup> ) and unnormal gene sequences were identified and deleted using UCHIME (Edgar, 2010). With each OTU, the representative sequence was analyzed

<sup>1</sup>http://drive5.com/uparse/

TABLE 1 | Feed ingredient and nutrient composition of experimental diets (%, as-fed basis).


<sup>1</sup>TS, tapioca starch; CS, corn starch; PS, pea starch. <sup>2</sup>Provided per kilogram of complete diet: vitamin A, 15,000 IU; vitamin D3, 3,000 IU; vitamin E, 150 mg; vitamin K3, 3 mg; vitamin B1, 3 mg; vitamin B2, 6 mg; vitamin B6, 5 mg; vitamin B12, 0.03 mg; niacin, 45 mg; vitamin C, 250 mg; calcium pantothenate, 9 mg; folic acid, 1 mg; biotin, 0.3 mg; choline chloride, 500 mg; Fe (FeSO4.H2O), 170 mg; Cu (CuSO4.5H2O), 150 mg; I (KI), 0.90 mg; Se (Na2SeO3),0.2 mg; Zn (ZnSO4.H2O), 150 mg; Mg (MgO), 68 mg; Mn (MnSO4.H2O), 80 mg; Co (CoCl2), 0.3 mg. <sup>3</sup>Values were estimated based on database of NRC (2012). <sup>4</sup>ME, metabolized energy. <sup>5</sup>Analytical results obtained according to AOAC (2007).

using the Ribosomal Database Project (RDP) classifier (**RRID: SCR\_006633**) against the Silva (SSU119) 16S rRNA database employing a confidence level of 90%.

The bacterial diversity, such as rarefaction analysis, the number of observed OTUs, coverage abundance estimator, richness estimator (Chao 1 and ACE), and diversity indices (Shannon and Simpson) were calculated using MOTHUR software (version 1.35.1<sup>2</sup> ) according to previous instructions (Schloss et al., 2009). Principal coordinates analysis (PCoA) was performed based on the Bray–Curtis distance, and analysis of molecular variance (AMOVA) was performed to compare the dissimilarities among samples using the MOTHUR (Schloss et al., 2009).

The 16S sequencing data generated in this study were deposited into the National Center of Biotechnology Information (NCBI) Sequence Read Archive (SRA) database under Accession No. PRJNA517450.

#### Chemical Composition Analysis

The pH value of the colonic digesta was detected using a protable pH meter (HI 9024C; HANNA Instruments, Woonsocket, RI, United States). The starch and amylose concentrations of diets and colonic digesta were analyzed using the Total Starch Kit and Amylose/Amylopectin Kit (Megazyme International, Wicklow, Ireland), respectively. SCFA concentrations in the colon were determined by gas chromatography (GC) according to the method described in a previous study (Yu et al., 2017). Colonic lactate concentration was measured using a commercial kit

<sup>2</sup>http://www.mothur.org

according to the manufacturer's instructions (Nanjing Jiancheng Biological Engineering Institute, Nanjing, China).

Ammonia concentration in the colon was measured using a spectrophotometer (UV-2450; Shimadzu, Tokyo, Japan) according to a previous method (Chaney and Marbach, 1962). The biogenic amines concentrations in the colonic digesta were measured using high-performance liquid chromatography (HPLC) with precolumn dansylation according to a previous study (Yang et al., 2014). The concentrations of phenolic and indolic compounds in the colonic digesta were analyzed by HPLC, as described previously (Schüssler and Nitschke, 1999), with slight modifications. Briefly, 0.1 g of colonic digesta was mixed with 1.0 mL of acetonitrile. The mixture was vortexed, stored at −20◦C for 20 min, and then centrifuged at 3000 × g for 10 min at 4◦C. The supernatant was filtered through a 0.22-µm membrane and then analyzed on a Waters alliance HPLC System (e2695 separation module: Waters, Milford, MA, United States) with a Multi λ Fluorescence Detector (2475: Waters, Milford, MA, United States). Gradient elution of two mobile phases was used: mobile phase A consisted of HPLC grade water, and mobile phase B was acetonitrile. The gradient program was: 82% A initially, 55% A at 12 min, 10% A at 22 min, and 100% B at 23 min. The flow rate was 1.0 mL/min and the column temperature were 30◦C.

### Gas Chromatograph–Time-of-Flight Mass Spectrometry Analysis

GC-TOF/MS was used to measure the colonic metabolites. All of the samples were pretreated, extracted, and identified using the procedure a previously described (Mao et al., 2015a). The LECO Chroma TOF4.3X software, LECO-Fiehn Rtx5 database, and commercial databases, including KEGG<sup>3</sup> and HMDB<sup>4</sup> , were utilized to extract the raw peak and filter data baseline, as well as further identify and validate the different metabolites. The peaks area of each metabolite was standardized using internal standard normalization methods before further analysis. The resulting data containing the peak number, sample name, and normalized peak were imported into SIMCA-P 13.0 (Umetrics, Umeå, Sweden) for partial least squares discriminant analysis (PLS-DA) and orthogonal partial least-squared discriminant analysis (OPLS-DA). In the present study, the discriminated metabolites were selected based on variable importance in the projection (VIP) value from the OPLS-DA model. VIP > 1 and q < 0.05 [false discovery rate (FDR)] were used to select the significant metabolites among the three dietary treatment groups.

#### Data Analysis

Statistical calculations for all the experimental data were conducted using the SPSS software package (SPSS v. 20.0: SPSS, Chicago, IL, United States). Before assessing the differences between the groups, the Shapiro–Wilk test was used to confirm whether the variables exhibited a normal distribution. The variables that showed a non-normal distribution (some data of taxa richness and metabolomics parameters) were analyzed

<sup>3</sup>http://www.genome.jp/kegg

by Kruskal–Wallis one-way analysis of variance (ANOVA) with the Benjamini and Hochberg false discovery rate (FDR) multiple-testing correction (Benjamini and Hochberg, 1995). The variables that showed a normal distribution (pH, and metabolite concentrations) were analyzed by one-way ANOVA with a Tukey post hoctest. Significant differences were declared at P ≤ 0.05. The correlation between significantly changed bacteria by diet types (at the genus level, adjusted P < 0.05) and pH values, metabolites (VIP > 1.5, adjusted P < 0.05, and similarity > 600), main SCFA, and amines were analyzed by Spearman's rank correlation test using GraphPad Prim version 5.0 (GraphPad Software, San Diego, CA, United States). To elucidate potential systemic properties, we focused on the absolute Spearman's correlation coefficient > 0.5 with statistical significance at P < 0.05. These correlation networks were visualized using Cytoscape 3.5.1 software (Smoot et al., 2011).

# RESULTS

#### Growth Performance, the Content of Starch, Amylose, Amylose/Amylopectin Ratio of Colonic Digesta

In this study, the pigs in the PS group showed a greater BW and ADG than those in the TS group (BW: 115.88 ± 1.21 vs. 109.52 ± 0.91 kg/d; ADG: 0.97 ± 0.03 vs. 0.85 ± 0.01 kg/d) and a lower F:G than those in TS group (3.01 ± 0.04 vs. 3.34 ± 0.09 kg/d) during the whole experimental period (P < 0.05). However, there was no difference in ADFI (P = 0.741) among the TS group (2.83 ± 0.07), CS group (2.85 ± 0.05), and PS group (2.89 ± 0.05).

As shown in **Supplementary Table S1**, pigs in the PS group showed a higher starch and amylose/amylopectin ratios in colonic digesta than those in the TS group (P < 0.05). The starch content of the CS group in colonic digesta was lower than that of the PS group, but higher than that of the TS group. Additionally, there was no significant difference on the content of amylose in colonic digesta among the three groups.

# Colonic Bacterial Community Structure

To evaluate the impact of the different starch diets on the microbial composition of colonic digesta, a total of 998,521 V3–V4 16S rRNA effective sequences from the 24 samples, with an average of 41,605 sequences per sample were used for subsequent analysis. The flattened rarefaction curves showed that the sampling in each group provided sufficient OTU coverage (**Supplementary Figure S1**). The richness and diversity of the colonic digesta bacteria are shown in **Table 2**. Pigs in the CS group had a higher species richness and diversity indices compared to that in the PS group, as reflected by the OTU numbers, Chao1, and Shannon index with statistical differences. However, the ACE richness index, coverage, and Simpson index did not differ among the different groups. The PCoA with Bray–Curtis distance results showed that the PS group separated from the TS and CS groups (**Figure 1**). AMOVA analysis also showed significant dissimilarities among the three groups (Fs = 2.84, P < 0.001,

<sup>4</sup>http://www.hmdb.ca

TABLE 2 | Summary statistic of colonic digesta bacterial community at the 3% dissimilarity level.


Values are means ± SEM (n = 8). Results were analyzed by one-way analysis of variance (ANOVA) with Turkey's test, and the variant letter in the same row indicated significant difference when P < 0.05. OTU, operational taxonomic units; ACE, abundance-based coverage estimator. TS, tapioca starch; CS, corn starch; PS, pea starch.

among TS, CS, and PS groups; Fs = 3.31, P < 0.001, TS vs. PS; Fs = 3.45, P < 0.001, CS vs. PS; Fs = 1.80, P < 0.05, TS vs. CS).

At the phylum level, the Firmicutes and Bacteroidetes were the two predominant phyla, contributing 77.79 and 17.59% in the TS group, 74.17 and 20.97% in the CS group, and 78.05 and 18.96% in the PS group, respectively (**Figure 2A**). Proteobacteria and Actinobacteria were the next two most dominant phyla, accounting for 3.04 and 0.38% in the TS group, 1.59 and 0.50% in CS group, and 1.01 and 0.99% in the PS group, respectively. The abundance of the phyla Proteobacteria in the PS group was significant decreased (P < 0.05) compared with that in the TS group. There was a higher abundance of Tenericutes in the PS group than that in CS group (P < 0.05, **Figure 2B**). In addition, the abundance of Actinobacteria in the PS group was increased (P < 0.05) compared with that in the TS and CS groups. However, no significant changes were found in the abundance of Firmicutes, Bacteroidetes, and SHA-109 among the three groups.

At the genus level, the 30 most dominating genera of the colonic digesta are presented in a heat map (**Supplementary** **Figure S2**). The eight most dominating genera (those with a relative abundance ≥ 5% in at least one treatment) were the Clostridium\_sensu\_stricto\_1, unclassified Ruminococcaceae, unclassified Peptostreptococcaceae, unclassified S24-7, Lactobacillus, Streptococcus, unclassified Lachnospiraceae, and Prevotella. The pigs in the PS group showed a lower relative abundance of unclassified Ruminococcaceae, unclassified Lachnospiraceae, unclassified Christensenellaceae, Escherichia– Shigella, unclassified Family-XIII, and Anaerotruncus compared with those in the TS group (adjusted P < 0.05), while the relative abundance of Lactobacillus, Prevotella, Faecalibacterium, and Megasphaera were higher (adjusted P < 0.05) (**Figure 3**). Meanwhile, the pigs in the PS group had a lower relative abundance of unclassified Ruminococcaceae, Anaerotruncus, and Parabacteroides (adjusted P < 0.05) compared with the pigs in the CS group, while had a higher relative abundance of Faecalibacterium and Megasphaera (adjusted P < 0.05). In addition, the abundance of Lactobacillus and Parabacteroides were also increased in the CS group compared with the TS group (adjusted P < 0.05).

#### Metabolite Profiles in the Colonic Digesta

As shown in **Figure 4A**, pigs in the CS and PS groups presented with significantly decreased pH values compared with the TS group (P < 0.05). Pigs in the PS group had a higher lactate concentration than it in the TS group (P < 0.05) (**Figure 4B**). For SCFA (**Figure 4C**), the concentrations of total SCFA, acetate, propionate, butyrate, and valerate were higher in the PS group compared with the TS group (P < 0.05). Meanwhile, the concentrations of total SCFA and valerate were also increased in the CS group compared with the TS group (P < 0.05). However, the concentrations of branched-chain fatty acid (BCFA), isobutyrate, and isovalerate were not affected by the dietary treatments (P > 0.05).

For biogenic amines (**Figure 4D**), the pigs in the PS group had a lower total amines, putrescine, and cadaverine concentrations than in the TS group (P < 0.05), and a lower methylamine concentration than in the CS group (P < 0.05). The pigs in the CS group also had lower putrescine and cadaverine

concentrations than in the TS group (P < 0.05). However, there were no differences of tryptamine, spermine, spermidine, or tyramine concentrations among different dietary treatments (P > 0.05). For phenolic and indole compounds (**Figure 4E**), the concentrations of skatole, indole, and phenol were lower in the PS group than those in the TS group (P < 0.05). The concentration of indole was also lower in the CS group than in the TS group (P < 0.05). The concentration of p-cresol was not affected by the dietary treatments (P > 0.05). The dietary treatments also did not affect the ammonia concentration (P > 0.05; **Figure 4F**).

To further predict whether the feeding different starch diets affected the metabolite profiles of the colonic digesta, GC-TOF/MS was used to analyze the metabolite profiles. A total of 689 valid peaks were detected, and 135 reliable metabolite compounds were quantified in all the samples (Similarity > 600), and these mainly included amino acids, amines, fatty acids, carbohydrates, organic acids, purines, lipids, and others. The PLS-DA (**Figure 5A**) and OPLS-DA (**Figure 5B**) models showed that the three groups were well-separated.

To assess which compounds were responsible for the differences among the three groups, the parameters of VIP > 1.0 and adjusted P < 0.05 were used as key lineages for separating the colonic compounds among the three groups (**Figure 5C** and **Supplementary Table S2**). In total, twenty compounds with a VIP > 1.0 and adjusted P < 0.05 were identified. Among these, seven metabolites (leucine, glycine, putrescine, tyramine, indole-3-acetic acid, p-cresol, and hydroxylamine) were reduced and 10 metabolites (galactose, fucose, N-acetylgalactosamine, glycerate, stearic acid, capric acid, linoleic acid, lactate, uracil, and pantothenic acid) were enriched in the pigs fed the PS diet compared with the TS diet. Meanwhile, three metabolites (glycine, cholesterol, and hydroxylamine) were reduced in the pigs fed with the PS diet compared with the CS diet. Additionally, seven metabolites (fucose, glucose, ribose, N-acetylgalactosamine, glycerate, stearic

acid, and capric acid) were enriched in the pigs fed with the CS diet compared with the TS diet. Overall, these results indicated that the PS diet (containing a high ratio of amylose) markedly increased the concentrations of organic acids (acetate, propionate, butyrate, valerate, lactate), carbohydrates, and lipids related compounds, and decreased the concentrations of amino acid related compounds (leucine, glycine, amines, phenol, and indole compounds), suggesting a strong impact of the PS diet on carbohydrate, lipid, and amino acid metabolism characteristics in the colon.

#### Correlation Analysis Between the Colonic Metabolome and Microbiome

To explore the functional correlation between changes in the colonic microbiome and metabolite profiles, a Spearman's rank correlation analysis matrix was generated by calculating the Spearman's correlation coefficient among the microbial composition affected by the diet treatments (at the genus level, adjusted P < 0.05), pH values, and metabolites (**Figure 6**). A clear significant correlation (P < 0.05) and an absolute value of the Spearman's correlation coefficient of r > 0.5 was identified between the changes in the colonic microbiome and the metabolome. The correlation analysis revealed that Lactobacillus was positively correlated with glycerate, linoleic acid, total SCFA, butyrate, valerate, and lactate (P < 0.05), while negatively correlated with cadaverine, p-cresol, and phenol (P < 0.05). Prevotella was positively correlated with capric acid, linoleic acid, uracil, total SCFA, acetate, and butyrate (P < 0.05), while was negatively correlated with putrescine and p-cresol (P < 0.05). Unclassified Christensenellaceae was positively correlated with putrescine, cadaverine, p-cresol, skatole, and phenol (P < 0.05), while was negatively correlated with glucose, glycerate, capric acid, linoleic acid, uracil, total SCFA, acetate, propionate, butyrate, valerate, and lactate (P < 0.05). Turicibacter was positively correlated with putrescine, indole-3-acetic acid, and indole (P < 0.05), while was negatively correlated with glycerate, butyrate, and valerate (P < 0.05). Escherichia–Shigella was positively correlated with leucine, putrescine, cadaverine, and p-cresol (P < 0.05), while was negatively correlated with glycerate and butyrate (P < 0.05).

Megasphaera was positively correlated with lactate (P < 0.05), while was negatively correlated with skatole, indole, and phenol (P < 0.05). Faecalibacterium was negatively correlated with leucine, putrescine, cadaverine, and p-cresol (P < 0.05), while was positively correlated with linoleic acid, uracil, total SCFA, acetate, propionate, butyrate, and lactate (P < 0.05). Meanwhile, our results also revealed that colonic pH was positively correlated with leucine, putrescine, indole-3-acetic acid, p-cresol, skatole, and indole (P < 0.05), while negatively correlated with total SCFA, acetate, butyrate, and lactate (P < 0.05). Collectively, these results indicated that the changes in the colonic digesta microbiota were correlated with alterations of metabolites in pigs.

#### DISCUSSION

Starch is the main dietary energy source for humans and monogastric animals, and previous studies have indicated a close relationship between the structure of dietary starch types and their utilization efficiency (Pieper et al., 2008; Jha et al., 2011). However, there is a paucity of information on the microbial community and the metabolic profile after treatment with different starch sources. In the present study, we investigated the response of the microbes and metabolites of colonic digesta of pigs fed different starch sources using 16S rRNA MiSeq sequencing, GC-TOF/MS, and biochemical analyses. Our results showed that treatment with different dietary starch led to different responses regarding microbial composition and metabolism in the colon. The PS diet (containing high ratio of amylose) markedly increased the abundance of some probiotics (such as Lactobacillus), while decreased the abundance of Escherichia coli compared with the TS diet (containing a low ratio of amylose). Moreover, our results also demonstrated that the PS diet increased the concentrations of organic acids (SCFAs and lactate) and some macronutrients (galactose, fucose, glucose, ribose, stearic acid, and linoleic acid) compared with the TS diet, and decreased the amino acids and their derivatives (leucine, glycine, amines, phenolic and indole compounds). These findings indicated a marked influence of the different dietary starch sources on the intestinal microbial community and metabolic profiles in the colon of pigs.

#### Diets With Different Starch Sources Altered the Colonic Microbiota Structure in Pigs

Substrate availability and the preferential substrate utilization of microbes are the major factors that affects the composition of gastrointestinal tract microbiota (Castillo et al., 2007). In the present study, we found that the PS diet resulted in a lower pH and reduced the bacterial richness and diversity, as indicated by the Chao 1 values and Shannon index. Moreover, the results of Bray–Curtis PCoA and AMOVA analyses further revealed difference in the colonic bacterial communities among the three groups. Many amylolytic bacteria can produce bacteriocins

and antimicrobial molecules and then prevent the colonization of bacteria that cannot utilize starch (Harlow et al., 2016). Thus, one potential explanation for altered the colonic bacterial richness and diversity may be due to the PS diet increased the abundance of some amylolytic bacteria, such as Lactobacillus, Prevotella, Faecalibacterium, and Megasphaera as mentioned below, and then inhibit the colonization of many bacteria that cannot utilize starch. Bacteria-specific factors, such as substrate affinity, substrate preference, and pH tolerance, could influence competition among amylolytic bacteria. A lower pH value in the colon might decrease the richness of some bacteria (such as Escherichia–Shigella) due to their susceptibility to low pH and this can also increase the abundance of several low-pH-tolerant colonic digesta bacteria (Daniëlle et al., 2013; Mao et al., 2015a). Therefore, other potential explanation for the decreased colonic bacterial richness and diversity in the pigs fed the PS diet may be due to the low pH. Furthermore, the digestion of TS allows for fast digestion (Weurding et al., 2001), CS is partially protected by the endosperm protein matrix (Svihus et al., 2005), and PS has a high amylose proportion and cell structures enclosing the starch granules, making it resistance to α-amylase digestion in the small intestine to a certain extent (Sun et al., 2006). Therefore, the alteration in the structure of the microbial population may also be due to the PS diet resulting in some amount of fermentable substrate (starch) entering the colon, thus promoting the growth of amylolytic and other starch-digesting bacterial species.

At the genus level, univariate statistical analysis indicated that the PS diet marked increased the abundance of Lactobacillus, Prevotella, Faecalibacterium, and Megasphaera in the colonic digesta compared with the TS diet. Similarly, previous studies also demonstrated that diets rich in amylose increased the abundance of Lactobacillus (Bird et al., 2007; Luo et al., 2015; Newman et al., 2018), Prevotella (Sun et al., 2015; Maier et al., 2017), Faecalibacterium (Daniëlle et al., 2013; Wang et al., 2018), and Megasphaera (Newman et al., 2018) in both humans and pigs. Among these various taxa, several species of Lactobacillus have many beneficial effects on the gut health of both humans and animals (Azad et al., 2018; Yu et al., 2018), normalizing the ratio of anti-inflammatory to pro-inflammatory cytokines and inhibiting the infection or colonization of pathogens via the productions of antimicrobial factors, such as bacteriocins and lactate (O'Mahony et al., 2005; Yu et al., 2018). Prevotella is well-known as a gut colonizer, one of the predominant starch-degrading bacteria in the intestine, and confirming the producing capacity of SCFAs (Flint et al., 2012). Faecalibacterium and Megasphaera are starch-utilizing commensal bacteria that can ferment starch to produce butyrate (Harry et al., 2008; Kamke et al., 2016). Some species in the genus Faecalibacterium and Megasphaera together with butyrate, have many beneficial effects in regard to colonic homeostasis via enhancement of epithelial energy metabolism and stimulating immune system balance (Sokol et al., 2008; Petra et al., 2014). Thus, the higher relative abundances of some beneficial bacteria (Lactobacillus, Prevotella, Faecalibacterium, and Megasphaera) in the PS group indicated that feeding of PS diet (rich in amylose) might have beneficial effects on the colonic health of pigs.

Additionally, pigs feed the PS diet demonstrated a decrease in the abundance of unclassified Ruminococcaceae, unclassified Lachnospiraceae, unclassified Christensenellaceae, Escherichia– Shigella, and Anaerotruncus compared with the TS diet. Previous study also indicated that a high amylose diet decreased several of above bacterias in the colon (Bird et al., 2007; Newman et al., 2018). Ruminococcaceae and Lachnospiraceae are the main families in the gut of mammals and have been associated with the maintenance of gut health (Donaldson et al., 2016). Previous studies have found that the enrichment of these families is associated with colonic mucosal inflammation, which can trigger colitis upon disruption of the barrier function of colonic epithelial cell (Willing et al., 2011; Nakanishi et al., 2015). Escherichia–Shigella is involved in protein utilization and is sensitive to acidic environment (Chen et al., 2018). Thus, the decrease in Escherichia–Shigella may be explained by the shortage of protein substrates for fermentation and the lower pH after feeding with the PS diet feeding. Some species within the genus Escherichia–Shigella, such as E. coli, are known as opportunistic pathogens and are associated with numerous infections and diseases, such as bacillary dysentery or colitis disease (Barnich et al., 2007). The enrichment of some species of Anaerotruncus, which belongs to Clostridium cluster IV, are associated with inflammatory bowel disease in the feces and rectal mucosa of humans (Satokari et al., 2014). Therefore, these findings suggested that feeding of a PS diet (rich in amylose) inhibited the abundance of several potential pathogens, and this may also have beneficial effects on the health of pigs.

#### Diets With Different Starch Sources Significantly Altered the Colonic Metabolite Profiles of Pigs

In the intestine, differences in substrate fermentation by microbiota also lead to different microbial metabolic process and metabolite profiles (Fouhse et al., 2015). In our study, PLS-DA and OPLS-DA analyses showed a clear separation of colonic metabolites due to the different starch diets, indicating significant differences in the metabolic profiles. The univariate statistical analysis indicated that the carbohydrates, such as galactose, fucose, glucose, ribose, and N-acetylgalactosamine were increased in the CS and PS groups compared with the TS group, indicating that carbohydrate metabolism was influenced at the local level (**Figure 5C** and **Supplementary Table S2**). A previous study also demonstrated that a raw potato starch diet (rich in amylose) could also increase the concentrations of fructose, glucose, and maltose (Sun et al., 2016). Diets containing starch with a higher content of amylose can decrease the digestibility of starch in the small intestine and lead to most of the starch being extensively passed into the hindgut (Regmi et al., 2011), where it can be fermented by microbes to produce sugars. Indeed, our study also found that the colonic starch content in the pigs fed with the CS and PS diets were significantly higher than that in the pigs fed with the TS diet (**Supplementary Table S1**). Thus, these changes in sugar concentrations may be deemed a fundamental alteration caused by the CS and PS diets rich in amylose. Meanwhile, several fatty acids, such as stearic acid,

capric acid, and linoleic acid, were also increased in the colon of the PS group when compared with the TS group. Resistant starch (with a high ratio of amylose) can regulate lipid metabolism and decrease the absorption of fatty acids (Lee et al., 2012), which may partly explain this observation.

Our study also showed that the PS diet significantly increased the concentrations of organic acids, such as total SCFAs, acetate, propionate, butyrate, and lactate. Similarly, gut SCFAs and lactate concentrations increased when pigs are fed with high amounts of amylose (Topping et al., 1997; Bird et al., 2007; Fouhse et al., 2015) and this shift in metabolites may be attributed to increase in SCFA- and lactateproducing bacteria. The correlation analysis also showed a positive correlation between these metabolites and the abundance of Prevotella, Faecalibacterium, Megasphaera, and Lactobacillus. SCFAs and lactate have a beneficial role in the metabolic functions and health of the gut. Acetate and propionate are the energy substrates of peripheral tissues, butyrate is the major energy source for colonic epithelial cells and exerts an anti-inflammatory function, and lactate can inhibit the activity of pathogens that invade the gut, such as Escherichia– Shigella (Tremaroli and Bäckhed, 2012). Therefore, the increase in SCFAs and lactate concentrations in the present study suggest the presence of a host-friendly gut environment after feeding of the PS diet.

Additionally, results of colonic metabolomics and biochemical analyses showed that the PS diet feeding significantly decreased the amino acid relatives compared with the TS group, such as leucine and glycine, indicating fewer nitrogen sources left for the microbial fermentation. This alteration may be due to the PS diet which increased the amount of fermentable substrate (starch) entering the colon and the carbon: nitrogen ratio of the substrates for microbial fermentation. Furthermore, ammonia, several amines, as well as phenolic and indole compounds were also decreased in the pigs fed with the PS diet compared to those in the pigs fed with the TS diet. Biogenic amines are formed from decarboxylation of amino acids by gut bacteria, and phenolic and indole compounds are produced from decarboxylation of aromatic amino acids by gut bacteria, such as Escherichia– Shigella (Blachier et al., 2007). Correspondingly, our results also showed that the PS diet markedly decreased the abundance of Escherichia–Shigella, and this could explain the lower levels of amines as well as phenolic and indole compounds. On the other hand, amino acid fermentation is favored at a neutral pH, as the proteases secreted by the bacteria are more active at a neutral or slightly alkaline pH than an acidic pH (Pi et al., 2018). In the current study, the pH in the colonic digesta was maintained at a more acidic level in the PS group compared with that in the TS group (mean of 6.21 and 6.61 in the PS and TS group, respectively; **Figure 4A**). Thus, the lower pH in the PS group may influence the protease activity, which may further support the conclusion that the lower concentrations of amines, as well as phenolic and indole compounds in the PS group originated from changes in the microbial proteolytic activity. High level of amines (such as cadaverine and putrescine), phenol, and skatole might be toxic to gut health (Davila-Gay et al., 2013; Yin et al., 2018b). Thus, decreasing the concentrations of these compounds via a PS diet may exert a beneficial effect on gut health. In general, our findings clearly indicate an evident change in microbial metabolic activity, higher microbial carbohydrate fermentation and lower microbial catabolism of amino acids after feeding of a PS diet.

# CONCLUSION

The present study combining microbiome, metabolome, and biochemical analyses demonstrated that a PS diet (with greater amylose content) selectively altered the gut microbial composition and metabolic profiles in the colon of the pigs, likely toward a more host-friendly gut environment. Colonic bacteria, such as Lactobacillus and many SCFAs-producing bacteria increased, whereas the abundance of Escherichia–Shigella decreased after feeding a PS diet. The intestinal metabolites were changed by the different dietary starch sources, as evidenced by the increase in the concentrations of organic acids and carbohydrates and the decrease in the concentrations of metabolites involved in amino-acid metabolism. These findings may help us to understand the effects of dietary starches with higher amylose/amylopection ratios on the nutrition and health of animals and humans.

# DATA AVAILABILITY

All datasets for this study are included in the manuscript and/or the **Supplementary Material**.

# AUTHOR CONTRIBUTIONS

MY, WC, and XM conceived and designed the whole trial. ZL and TR conducted the pig trial. MY and ZL conducted laboratory analyses. MY, XM, and GW wrote the manuscript.

# FUNDING

This study was supported by the Presidential Foundation of the Guangdong Academy of Agricultural Sciences (201802B) and Guangdong Modern Agro-Industry Technology Research System (2016LM1080 and 2017LM1080).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.01143/full#supplementary-material

#### REFERENCES

fmicb-10-01143 May 29, 2019 Time: 14:55 # 12


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Yu, Li, Chen, Rong, Wang and Ma. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fmicb-10-01143 May 29, 2019 Time: 14:55 # 13

# Diet Control More Intensively Disturbs Gut Microbiota Than Genetic Background in Wild Type and ob/ob Mice

Jing-Hua Wang1,2, Na Rae Shin<sup>1</sup> , Soo-Kyoung Lim<sup>1</sup> , Ungjin Im<sup>3</sup> , Eun-Ji Song<sup>4</sup> , Young-Do Nam<sup>4</sup> and Hojun Kim<sup>1</sup> \*

<sup>1</sup> Department of Rehabilitation Medicine of Korean Medicine, Dongguk University, Goyang-si, South Korea, <sup>2</sup> Department of Research and Development, Cure Pharmtech, Goyang-si, South Korea, <sup>3</sup> Department of East-West Medical Science, Graduate School of East-West Medical Science, Kyung Hee University, Seoul, South Korea, <sup>4</sup> Research Group of Healthcare, Korea Food Research Institute (KFRI), Seongnam-si, South Korea

#### Edited by:

Yuheng Luo, Sichuan Agricultural University, China

#### Reviewed by:

Isabel Gordo, Gulbenkian Institute of Science, Portugal Maryam Dadar, Razi Vaccine and Serum Research Institute, Iran

> \*Correspondence: Hojun Kim kimklar@gmail.com

#### Specialty section:

This article was submitted to Systems Microbiology, a section of the journal Frontiers in Microbiology

Received: 25 February 2019 Accepted: 23 May 2019 Published: 07 June 2019

#### Citation:

Wang J-H, Shin NR, Lim S-K, Im U, Song E-J, Nam Y-D and Kim H (2019) Diet Control More Intensively Disturbs Gut Microbiota Than Genetic Background in Wild Type and ob/ob Mice. Front. Microbiol. 10:1292. doi: 10.3389/fmicb.2019.01292 Changes in environmental and genetic factors are vital to development of obesity and its complications. Induction of obesity and type 2 diabetes by both leptin deficiency (ob/ob) and high fat diet (HFD) has been verified in animal models. In the present experiment, three types of diets (normal diet; ND, HFD and high sucrose diet; HSD) and two types of genetic mice (Wild type: WT and ob/ob) were used to explore the relationship among diet supplements, gut microbiota, host genetics and metabolic status. HFD increased the body, fat and liver weight of both ob/ob and WT mice, but HSD did not. HFD also resulted in dyslipidemia, as well as increased serum transaminases and fasting glucose in ob/ob mice but not in WT mice, while HSD did not. Moreover, HFD led to brain BDNF elevation in WT mice and reduction in ob/ob mice, whereas HSD did not. Both HFD and HSD had a greater influence on gut microbiota than host genotypes. In detail, both of HFD and HSD alteration elucidated the majority (≥63%) of the whole structural variation in gut microbiota, however, host genetic mutation accounted for the minority (≤11%). Overall, diets more intensively disturbed the structure of gut microbiota in excess of genetic change, particularly under leptin deficient conditions. Different responses of host genotypes may contribute to the development of metabolic disorder phenotypes linked with gut microbiota alterations.

Keywords: gut microbiota, high fat diet, high sucrose diet, overweight, sequencing

# INTRODUCTION

Obesity and its related complications, such as type 2 diabetes and hyperlipidemia, have become an issue worldwide and the associated morbidity rate has been increasing rapidly in recent decades (Afshin et al., 2017; Zheng et al., 2018). Although the fundamental reasons for being overweight and obese are disruptions of energy balance, many etiological factors, such as genes, metabolism, the environment and dietary habits directly or indirectly lead to overweight and obesity (Brantley et al., 2005; Hill et al., 2012; Weinsier et al., 2012). Leptin is a representative energy expenditure hormone that is primarily secreted by adipocytes (Pandit et al., 2017). Because of its appetite suppressing effect, leptin deficient obese mice (ob/ob) have been deemed an excellent mutant model and utilized

extensively in studies of obesity and diabetes (Drel et al., 2006). Moreover, immoderate ingestion of high fat (HFD) or HFD together with high sucrose diet (HSD) results in dietary obesity and type 2 diabetes (Xi et al., 2004; Yang et al., 2012). Interestingly, ingestion of HSD alone without over ingestion of calories also causes glucose intolerance (Sakamoto et al., 2012); however, it is still unclear if obesity can be induced by HSD alone rather than excess calories.

Over 100 trillion microbes live in the gastrointestinal tract in a symbiotic relationship with the host, and these organisms, like Firmicutes, Bacteroidetes, and Actinobacteria, etc., have been shown to play vital roles in physiological and pathological process, such as immunomodulation and maintenance of homeostasis in both animal and human subjects (Ley et al., 2006; Thursby and Juge, 2017; Park, 2018). The development of systemic metabolic disorders such as obesity and type 2 diabetes has been shown to be closely linked with changes in gut microbiota as well, representatively Firmicutes and Bacteroidetes ratio involved in obesity (Cani et al., 2007b; Parks et al., 2013; Tai et al., 2015). For example, gut microbiota reduce leptin sensitivity and the fat-suppressing neuropeptides proglucagon and brain-derived neurotrophic factor (BDNF) (Schele et al., 2013). Therefore, gut microbiota have been deemed a non-negligible factor that can contribute to obesity and its complications.

Several previous preclinical studies have shown that gut microbiota is changed in obese individuals in response to deficiencies in leptin and over consumption of HFD or HFD together with HSD (Boulange et al., 2016; Collins et al., 2016; Guo et al., 2017). In detail, host leptin deficiency changes the gut microbiota correlated with variation in long-term glucose levels, glucose intolerance and mucosal regulatory immunity and HFD plus HSD treatment more rapidly alter gut microbiota together with muscle integrity, inflammation even 3 days. As we know, genetic and environmental factors have been shown to be essential to the development of chronic metabolic disorders in numerous studies (Maes et al., 1997; Speakman, 2004). Moreover, the gut microbiota have been shown to be shaped by both genetic and dietary factors (Goodrich et al., 2014; Graf et al., 2015). Development of obesity and a tight relationship with gut microbiota alteration by both leptin deficiency (ob/ob) and HFD/HFD+HSD has also been verified (Park et al., 2016; Lu et al., 2017). However, it is not clear if gut microbiota can be altered by HSD alone when coupled with ingestion of normal calories. Hypothetically, consumption of HSD within a normal calorie range should not induce fat accumulation; therefore, it is worth investigating whether such increased accumulation occurs as a result of alterations in gut microbiota.

A previous study revealed that HFD alone induced changes in gut microbiota relevant to metabolic syndrome phenotype development, and that these changes were more important than host gene mutations in ApoA-I knockout mice (Zhang et al., 2010). In addition, gut microbiota-associated bile acid deconjugation accelerates fat synthesis in normal diet fed ob/ob mice (Park et al., 2016). However, no studies have been conducted to evaluate the effects of different diets on host-gut microbiota interactions in ob/ob mice and their wildtype (WT) lean control mice as they relate to the development of obesity. Consequently, in this study, we investigated the importance of host gene and diet on gut microbiota as it relates to obesity and other metabolic factors in ob/ob and wild type mice. Moreover, HFD and HSD were employed as different diet controls that provided different levels of calories to evaluate the contributions of diet perturbed gut microbiota and host gene mutations relevant to the development of obesity.

# MATERIALS AND METHODS

# Animals and Experimental Schedule

Twenty-one male C57BL/6J mice (4-weeks old, wild type) and 21 C57BL/6J-ob/ob mice (4 weeks old, genetically obese with leptindeficiency) were obtained from the Korea Research Institute of Bioscience and Biotechnology (Ochang-eup, South Korea). The normal diet (AIN-93G, 16 cal% as fat; 20 cal% as protein, 64 cal % as carbohydrate, 4000 Kcal/kg), HFD (60 cal% as fat; 20 cal% as protein, 20 cal % as carbohydrate, 5333 Kcal/kg), and HSD (12 cal% as fat; 19 cal% as protein, 70 cal % as carbohydrate, 3702 Kcal/kg) were purchased from Todobio (**Supplementary Table S1**, Guri-si, South Korea). Seven animals with same genetic background in each group were housed in the same cage. To reduce individual variations in gut microbiota, the bedding of animals was thoroughly mixed twice per week for 6 weeks within the same genetic type until the start of experiment. After 6 weeks of acclimation at 22 ± 2 ◦C under a 12-h light/12 h dark cycle and 40–60% relative humidity with free access to water and normal diet, the wildtype and ob/ob mice were divided into a normal, HFD and HSD group that received a normal diet (AIN-93G), an HFD diet and an HSD diet, respectively. All of the diets were provided ad libitum for 10 weeks (**Figure 1A**).

The body weight and food intake were recorded once a week for 10 weeks. On the last experimental day, the animals were sacrificed under the anesthetics Zoletil (tiletamine-zolazepam, Virbac, Carros, France) and Rompun (xylazine-hydrochloride, Bayer, Leverkusen, Germany) (1:1, v/v). Blood was collected from the abdominal aorta and rapidly transferred into a BD Vacutainer (Franklin Lakes, NJ, United States), after which the brains and fats were removed and quickly stored in liquid nitrogen. The livers were then removed and weighed. After 2 h of clotting, sera were separated from whole blood by centrifugation at 3000 × g for 15 min. The Food Efficiency Ratio (FER) was computed by dividing the average body weight gain by the average food intake for each group.

The animal experimental protocol was approved by the Institutional Animal Ethical Committee of Dongguk University (Approval No. IACUC-2014-030). In addition, all experiments were conducted according to the Guide for the Care and Use of Laboratory Animals (Institute of Laboratory Animal Resources, Commission on Life Sciences, National Research Council, United States; National Academy Press: Washington, DC, 1996).

# Serum Biochemical Analysis

Blood was collected from ventral aorta under Zoletil and Rompun anesthesia. After 1 h of clotting at room temperature, blood was centrifuged at 3000 × g for 15 min for serum separation. The serum levels of triglyceride (TG), total cholesterol (TC), high density lipoprotein (HDL), low density lipoprotein (LDL), aspartate transaminase (AST), and alanine transaminase (ALT) were determined using commercial enzymatic assay kits according to the manufacturer's instructions (Asan Pharmaceutical Co., Seoul, South Korea).

# Oral Glucose Tolerance Test (OGTT)

Mice were orally dosed with glucose solution (2 g/kg, Sigma, United States) before 12 h of fasting in the start of the last week. The blood glucose levels in tail blood drops were then measured using an ACCU-CHEK Active blood glucose meter (Mannheim, Germany) at 0, 30, 60, 90, and 120 min after administration. The OGTT results were also expressed as areas under the curves (AUC) to estimate the extent of the glucose tolerance impairment.

# Western Blot

Mouse brain cortexes were homogenized in RIPA buffer containing protease and phosphatase inhibitor. The supernatants were subsequently isolated, after which total protein concentrations were measured using a BCA kit (Thermo Scientific, Rockford, LL, United States). Denatured proteins were then separated in 10% SDS-PAGE gel, after which they were transferred to a polyvinylidene fluoride (PVDF) membrane (GE Healthcare Life Science, Germany) using a Mini-PROTEAN Tetra Cell system (BioRad Laboratories Inc., Hercules, CA, United States). Next, membranes were blocked with 5% skim milk containing TBST, Tris–buffered saline and Tween 20 for 1 h, after which they were treated overnight with primary antibody (anti-BDNF) at 4◦C (1:200; Santa Cruz Biotechnology, Inc., Santa Cruz, CA, United States) and anti-β-actin (1:1,000; Santa Cruz Biotechnology, Inc.), then incubated with anti-rabbit IgG-peroxidase conjugated secondary antibody for 1 h (1:2,000; Santa Cruz Biotechnology, Inc.). Finally, the membranes were detected using SUPEX ECL solution and photographed with a LAS3000 Imager (FUJIFILM, FUJI, Japan).

# Metagenomic Analysis of Gut Microbiota and Functional Predictive Annotation

Fecal samples were collected and kept at −80◦C before DNA extraction. The genomic DNA was then isolated using a QIAamp stool DNA mini kit (QIAGEN, Hilden, Germany) according to the manufacturer's instructions. The V1–V2 region of the 16S rRNA genes was amplified using a Thermal Cycler PCR system (BioRad, Hercules, CA, United States), after which amplicons were purified using a LaboPass PCR purification kit (COSMO GENTECH, Seoul, South Korea). An equimolar concentration of each amplicon from different samples was pooled to equal proportions based on their molecular weight and purified using Agencourt AMPure XP PCR purification beads (Agencourt Bioscience, Beverly, MA, United States). The DNA concentration and quality were confirmed on a BioAnalyzer 2100 microfluidics device (Agilent, Santa Clara, CA, United States) using a DNA 100 lab chip (Agilent, Santa Clara, CA, United States). The mixed amplicons were amplified on sequencing beads by emulsion PCR (emPCR). Sequencing reactions were performed using an Ion Torrent Next-Generation Sequencing Platform (Ion PGM, Life Technologies, Carlsbad, CA, United States).

Operational taxonomic units selection, taxonomic assignment and phylogenetic reconstruction were conducted using the QIIME1 (Version 1.9.1, University of Colorado, Boulder, CO, United States) software package and visualized with the LEfSe (linear discriminant analysis effect size) program (Hutlab, Boston, MA, United States), STAMP v2.1.3 software (Dalhousie University, Halifax, Canada) and Graphpad Prism 5. All metagenomics data were predictably profiled with PICRUSt-1.0.0 (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States) (Langille et al., 2013) and HUMAnN2 (The HMP Unified Metabolic Analysis Network 2) (Abubucker et al., 2012). The functional and taxonomic differences in predicted results were statistically analyzed and represented graphically using the STAMP v2.1.3 software (Dalhousie University, Halifax, Canada<sup>1</sup> ).

All sequencing raw reads have been deposited in the European Nucleotide Archive (ENA) under the project accession number PRJEB28486 with a unique name (ena-STUDY-DGU-05-09- 2018-03:21:13:852-167).

# Statistical Analysis

All animal experimental data were analyzed by one-way ANOVA followed by Bonferroni's post hoc test using GraphPad Prism version 5.01 (La Jolla, CA, United States). The results were expressed as the means ± standard error of the mean (SEM), and a P < 0.05 and P < 0.01 was considered statistically significant. Relationship strength between parameters was evaluated by the two tailed Pearson's correlation test. An absolute value of Pearson's correlation coefficient R > 0.4 was considered to indicate a positive correlation with connected by red line while R < −0.4 was considered a negative correlation with connected by blue line. The boldness of the line indicates the strength of the correlation.

# RESULTS

#### HFD, but Not HSD, Changed Body and Liver Weight in Both ob/ob and WT Mice

All ob/ob mice and their WT lean control mice were fed either ND, HFD, or HSD (n = 7 for each group) for 10 weeks. Each WT or ob/ob mouse had a similar body weight at week 6; however, the body weight of mice in the HFD groups was significantly increased by HFD relative to the ND groups after 10 weeks, regardless of genetic type, but not by HSD (**Figures 1B–D**). Additionally, feeding ob/ob mice HFD resulted in markedly elevated liver weight relative to the corresponding normal group, but no significant augmentation of liver weight was observed in any WT groups (**Figure 1E**).

#### HFD, but Not HSD, Elevated Caloric Intake and FER in Both ob/ob and WT Mice, but Did Not Change Food Intake by WT Mice

Although noticeable differences in absolute volume of food intake were not found in WT groups, absolute volume of food intake was markedly increased by HFD relative to ND and HSD in ob/ob groups (**Figure 2A**). Additionally, treatment with HFD resulted in notably increased caloric intake and FER relative to other diets in both WT and ob/ob mice (**Figures 2B,C**).

# HFD, but Not HSD, Increased the Fat Weights in Both ob/ob and WT Mice

Providing WT and ob/ob mice with HFD resulted in increased total fat weight, pararenal fat weight and visceral fat weight relative to other diet groups (**Figures 3A–D**). Although HFD treatment led to marked increases in gonadal fat weight relative to other diet groups in WT mice, no significant enhancement of gonadal fat weight was found in ob/ob mice (**Figures 3A–D**).

# HFD, but Not HSD, Led to Dyslipidemia in ob/ob Mice, but Not in WT Mice

Feeding ob/ob mice HFD resulted in a marked increase in the level of serum TG, TC, LDL, AST, and ALT relative to normal diet, but no such changes were observed in response to HSD (**Figures 4A–F**). Moreover, feeding WT mice HFD led to slight elevation of serum TG, TC, LDL, AST, and ALT levels, but these changes were not statistically significant (**Figures 4A–F**). In contrast, feeding WT mice HFD notably reduced the serum HDL level relative to the corresponding normal diet, but these changes were not observed in HSD. However, feeding ob/ob mice different types of diet did not induce a significant difference in serum HDL levels.

#### HFD, but Not HSD, Resulted in Elevated Fasting Glucose Levels in ob/ob Mice and OGTT (AUC) in WT Mice

Feeding ob/ob mice a HFD resulted in significant elevation of the fasting blood glucose level when compared to the

<sup>1</sup>http://kiwi.cs.dal.ca/Software/STAMP

Frontiers in Microbiology | www.frontiersin.org

corresponding normal diet, but HSD did not (**Figure 5A**). However, in WT mice groups, HFD only weakly increased the level of fasting blood glucose relative to other diet groups, but this increase was not significant. In addition, impairment of the AUC of OGTT was calculated to compare the extent of glucose tolerance (**Figures 5B,C**). Feeding WT mice HFD resulted in a

FIGURE 4 | Serum biochemistry parameters. At the end of the experiment, blood was collected from the ventral aorta under Zoletil and Rompun anesthesia. After separation of serum, the serum level of triglyceride (TG) (A), total cholesterol (TC) (B), high density lipoprotein (HDL) (C), low density lipoprotein (LDL) (D), aspartate transaminase (AST) (E), and alanine transaminase (ALT) (F) were determined using commercial enzymatic assay kits. Data were expressed as the means ± SEM and statistically evaluated using one-way ANOVA followed by Bonferroni's post hoc test. <sup>∗</sup>P < 0.05; ∗∗P < 0.01 (n = 7).

significant increase in AUC when compared to other diet groups (**Figure 5C**). In ob/ob mice, all groups showed a high level of OGTT (AUC) relative to WT groups; however, no significant difference in OGTT (AUC) was detected among these ob/ob groups (**Figure 5C**).

# HFD, but Not HSD, Resulted in Brain BDNF Elevation in WT Mice and Reduction in ob/ob Mice

Feeding WT mice HFD resulted in a significant increase of BDNF content in the brain cortex relative to the corresponding normal diet (**Figures 6A,B**). Conversely, feeding ob/ob mice HFD resulted in a significant decrease of BDNF content in the brain cortex when compared to the corresponding normal diet (**Figures 6A,B**). However, the level of BDNF in the brain cortex was not altered in either WT or ob/ob mice treated with HSD (**Figures 6A,B**).

### Whole Structural Responses of Gut Microbiota to Intake of Various Diets and Host Leptin Deficiency

Principal coordinate analysis based on 16S rRNA gene sequence data revealed obvious differences in the composition of gut microbiota among animal types, food types and time points. Comparison of samples collected on the initial and final day of experiment revealed time-associated differences (**Figure 7** and **Supplementary Figure S2**). Before starting the experiment, nonobvious differences were found among all groups, regardless of

β-actin (B). Data were expressed as mean ± SEM and statistically evaluated using one-way ANOVA followed by Bonferroni's post hoc test. ∗∗P < 0.01 compared to ND and HSD (n = 7).

whether mice were ob/ob or WT. However, at the end of the experiment, HFD-related differences were mainly found along PC1, which accounted for 63.2% of the total variations, whereas the two genotypes showed an obviously smaller difference along PC3, which accounted for only 10.6% of the total variations (**Supplementary Table S2**). Moreover, HSD-related differences were mainly observed along PC1, which accounted for 66.4% of the total variations, whereas the two genotypes showed smaller differences along PC3, which accounted for only 5.4% of the variations (**Supplementary Table S2**).

At the phylum-level, feeding WT mice with HSD resulted in a significant decrease in Firmicutes and an increase in Bacteroidetes relative to the corresponding normal diet, but HFD did not. However, feeding ob/ob mice with HSD generated a similar pattern as WT mice, but this difference was not statistically significant (**Figure 8**). At the low taxonomic-level, ob/ob mice showed a significantly higher level of Alistipes and a lower level of Blautia in the two ND groups, while ob/ob mice showed a significantly higher level of Lactobacillus and lower level of Lachnospiraceae and Muribaculaceae in the two HFD groups (**Figure 8**). However, no significant differences in genera were found between WT and ob/ob mice in the HSD groups. Thus, it is supposed that HSD more sensitively regulated the phylum-level of gut microbiota in WT mice than ob/ob mice. Although the mechanism is still not clear, it is shown that leptin gene co-effect with HSD play a critical role in changing gut microbiota in phylum-level. However, genus-level of gut microbiota was not obviously influenced by HSD together with leptin gene expression or not. Moreover, various alpha diversity indexes revealed no notable differences in response to any diets, genotypes or time points (**Supplementary Figure S1**). Taken together, HSD co-effect with leptin only changed gut microbiota in high taxonomic level. Nevertheless diets, genotypes didn't evidently change the gut microbial community richness, evenness and diversity even in different time points in the scale of entire ecosystem.

In addition, according to the result of Pearson's correlation analysis between gut microbiota composition and host metabolic parameters, Alistipes showed a evident positive correlation (Pearson r > 0.4) with extensive weights and serum biochemical parameters rather than other gut bacteria, inversely Blautia and Enterorhabdus showed a distinct negative correlation (Pearson r < 0.4) (**Figure 8E**). However, relative high percentage of Bacteroides only positively correlated with fat weight, especially pararenal fat weight (**Figure 8E**).

# DISCUSSION

Epidemiological data for the past 30 years reveal that obesity, a popular chronic metabolic disease according to the American Medical Association (Stoner and Cornwall, 2014), is becoming an increasing problem in the worldwide in developed and developing countries (Ng et al., 2014). In the last decade, many investigators have focused on gut microbiota and chronic metabolic diseases, including obesity and type 2 diabetes, in an

attempt to identify complex mechanisms that can be targeted to treat the condition (Baothman et al., 2016).

Turnbaugh et al. (2006) showed that changes in the ratio of Firmicutes and Bacteroidetes contributed to obesity; specifically, notable elevation of Firmicutes and reduction of Bacteroidetes were observed in obese subjects relative to lean subjects, while HFD also promoted an increase in Firmicutes and reduction of Bacteroidetes and the changes in the dominant phyla promoted more effective caloric intake, leading to increased weight and obesity (Turnbaugh et al., 2006; Tilg and Kaser, 2011). However, Zhang et al. found no phylum-wide gut microbiota alteration in HFD induced obese animals, regardless of genotypes (Apoa-I−/−) (Zhang et al., 2010). Interestingly, we also did not observe significant changes in Firmicutes and Bacteroidetes in HFD-induced obese animals, regardless of genotypes (ob/ob−/−). Nevertheless, our study revealed that feeding both WT and ob/ob mice HSD decreased Firmicutes and increased Bacteroidetes relative to ND, but this genotype (ob/ob−/−) showed a different degree of phylum-wide gut microbiota change when compared to the WT. A previous study showed the gut microbiota structure could be significantly changed by variations in caloric intake (Jumpertz et al., 2011), while another study reported that calorie restriction can modulate the balance of gut microbiota in a way that exerts health benefits to the host (Zhang et al., 2013). Therefore, in the present study, the dissimilar caloric intake between the HFD and HSD group might have been a vital factor influencing gut microbiota alterations. The results of the present study also suggest that reducing the caloric intake will be beneficial to maintenance of a healthy structure of gut microbiota and conducive to amelioration of obesity. Functional predictions from PICRUSt based on the 16S rRNA bacteria gene sequences indicated that HFD more widely regulated gene function than HSD, regardless of genotypes (**Supplementary Figure S3**). Moreover, HFD contributed more to the difference in gene function between WT and ob/ob mice than ND and HSD. Additionally, KEGG pathway predictions from HUMAnN2 showed the same pattern as the PICRUSt results (**Supplementary Figure S4**). Taken together, these findings indirectly illustrated that HFD exerts more important actions on functional alterations of gut microbiota.

In fact, alpha diversity is a method to evaluate the species richness and evenness in a single group. Similar to the results of other studies (Kubeck et al., 2016), alpha-diversity analysis revealed that HFD did not cause any dissimilarity in the richness and evenness of species in each single group between 0 and 10 weeks, regardless of genetic background. HSD also showed the same pattern as HFD upon various alpha-diversity analysis. In the present study it has been demonstrated that the species richness and evenness of gut microbiota in each single group cannot be changed by feeding with HFD or HSD for at least 10 weeks, regardless of genetic difference.

Endotoxin is a complex lipopolysaccharide generated by the cell membranes of Gram-negative bacteria (Erridge et al., 2007). The role of endotoxin in the development and beginning of chronic metabolic diseases has been acknowledged (Cani et al., 2007a; Gomes et al., 2017), and it has been reported that HFD/HFD+HSD altered gut microbiota triggered inflammation and prompted low grade endotoxemia, which is a vital mechanism for accelerating development of obesity and related metabolic disorders (Cani et al., 2007a; Sanchez-Tapia et al., 2017). Our study also revealed that both HFD and HSD altered the composition of gut microbiota relative to ND. When compared to ND, HFD markedly augmented

FIGURE 8 | Phylum and family level composition of fecal microbiota and Pearson's correlation analysis. Gut microbiota taxonomic profiles. The relative abundance (%) of fecal bacterial phyla (A) and families (B) were compared according to the 16S rRNA gene sequencing data. The proportion of sequences (%) of firmicutes (C) and Bacteroidetes (D) were compared. Taxonomic differences in results were statistically analyzed and represented graphically using the STAMP v2.1.3 software (Dalhousie University, Halifax, Canada). (E) All of the gut microbiota composition and host parameters data were evaluated using Pearson's correlation analysis, with a R value greater than 0.4 indicating a positive correlation (red line) and a R value less than –0.4 indicating a negative correlation (blue line). The boldness of the line indicates the strength of the correlation. The size of each circle indicates the average relative abundance of each genus.

endotoxin producing bacteria belonging to Bacteroidaceae and Tannerellaceae while diminishing the non-endotoxin producing bacteria Erysipelotrichaceae, both in WT and ob/ob mice. However, HFD only significantly elevated the fasting glucose levels in ob/ob mice. Body weight and liver weight also showed a similar pattern as fasting glucose. Interestingly, differences in the caloric intake of HFD and HSD did not lead to glucose intolerance or elevation of fasting glucose, indicating that high caloric intake elevated the risk of obesity, hepatosteatosis and type 2 diabetes.

Although some studies have shown that environment dominates over host genetics in shaping gut microbiota in humans (Kubeck et al., 2016), other studies have demonstrated that host genetics influence the composition of the gut microbiome and impact host metabolism (Goodrich et al., 2014; Ussar et al., 2016). According to our PCoA results, changes in HFD explained 63.2% of the total structural variation in gut microbiota, whereas genetic mutation (ob/ob: leptin deficiency) accounted for only 10.6%. A previous study of apolipoprotein A-I gene mutation (Apoa-I) mice fed HFD indicated similar findings regarding the ability of various factors to induce gut microbiota changes. Similarly, HSD changes explained 66.4% of the total structural variation in gut microbiota, whereas genetic mutation accounted for only 5.4%. Therefore, our study demonstrated that diet exerts stronger effects than genetics, and that the impact of HSD is slightly greater than that of HFD in terms of alteration of gut microbiota structure. However, in the present study food combination was not applied for exploring the interference of gut microbiota. It will be examined whether synergistic disturbance of gut microbiota by food collocation in future. All the critical consequence can be applied in precise nutrition to improve health of human being.

Commensal microbiota have been shown to alter the level of neurotrophic factors (Diaz Heijtz et al., 2011), including BDNF. Previous studies demonstrated that BDNF exerts a potent role in cognitive abilities (Borrelli et al., 2016), emotional regulation (Bjorkholm and Monteggia, 2016) and gastrointestinal function (Wang et al., 2015). Moreover, HFD or hepatocytes and hippocampal neurons prompted elevation of BDNF mRNA expression in mice (Genzer et al., 2016). In the present study, HFD resulted in significant elevation of brain BDNF in WT mice and a notable reduction in ob/ob mice, but HSD did not. Therefore, it can be speculated that HFD more noticeably regulated BDNF than HSD through regulation of the gut microbiota; however, the ability was modulated by genotypes.

Although a previous study indicated that Apoa-I knockout mice with HFD showed less severe metabolic disorder phenotypes than WT mice with HFD (Zhang et al., 2010), our results did not show the above pattern in the level of fasting glucose between ob/ob-HFD mice and WT-HFD mice. Moreover, no difference in glucose tolerance was observed between WT-HFD mice and ob/ob-HFD mice. It is possible that different genotypes had only slight effects on gut microbiota relative to diets, and that different genotypes made different contributions to the development of metabolic disorder phenotypes.

# CONCLUSION

The results of our study demonstrate that diet more intensively disturbs gut microbiota than genetic change, particularly deficiency of leptin in mice. Long term feeding with HFD and HSD had different impacts on the structure of gut microbiota. Specifically, HSD feed without superfluous consumption obviously changed the structure of gut microbiota; however, obesity and its complication-related parameters were not visibly altered. Leptin deficiency changes gut microbiota in response to high-fat diet, while host genotypes show different responses to contribute to the development of metabolic disorder phenotypes, possibly linked with gut microbiota alteration. Overall, interactions between host genetics, gut microbiota and diet are closely linked to the development of obesity and its complications.

# DATA AVAILABILITY

The datasets generated for this study can be found in European Nucleotide Archive (ENA), ena-STUDY-DGU-05-09- 2018-03:21:13:852-167.

# ETHICS STATEMENT

The animal experimental protocol was approved by the Institutional Animal Ethical Committee of Dongguk University (Approval No. IACUC-2014-030). In addition, all experiments were conducted according to the Guide for the Care and Use of Laboratory Animals (Institute of Laboratory Animal Resources, Commission on Life Sciences, National Research Council, United States; National Academy Press: Washington, DC, 1996).

# AUTHOR CONTRIBUTIONS

HK perceived and designed the study. J-HW wrote the manuscript. S-KL performed the experiments. J-HW and NS analyzed the data. UI, Y-DN, and E-JS participated in study design and data analysis. All authors have read and approved the final manuscript.

# FUNDING

This work was supported by a National Research Foundation (NRF) of Korea Grant funded by the Korean Government (NRF-2016R1A2B4014225) and also supported by the Main Research Program (E0170602-02) of the Korea Food Research Institute (KFRI) funded by the Ministry of Science and ICT.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2019. 01292/full#supplementary-material

#### REFERENCES

fmicb-10-01292 June 7, 2019 Time: 9:36 # 11



development of metabolic syndromes in mice. ISME J. 4, 232–241. doi: 10.1038/ ismej.2009.112

Zheng, Y., Ley, S. H., and Hu, F. B. (2018). Global aetiology and epidemiology of type 2 diabetes mellitus and its complications. Nat. Rev. Endocrinol. 14, 88–98. doi: 10.1038/nrendo.2017.151

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Wang, Shin, Lim, Im, Song, Nam and Kim. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Interactions Between Gut Microbiota and Acute Childhood Leukemia

*Yuxi Wen, Runming Jin\* and Hongbo Chen\**

*Department of Pediatrics, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China*

Childhood leukemia, the commonest childhood cancer, mainly consists of acute lymphoblastic leukemia (ALL) and acute myeloid leukemia (AML). Though great progresses have been made in the survival rates of childhood leukemia, the long-term health problems of long-term childhood leukemia survivors remain remarkable. In addition, the deep links between risk factors and childhood leukemia need to be elucidated. What can be done to improve the prevention and the prognosis of childhood leukemia is an essential issue. Gut microbiota, referred to as one of the largest symbiotic microorganisms that is accommodated in the gastrointestinal tract of human or animals, is found to be involved in the progression of various diseases. It is reported that microbiota may keep people in good health by participating in metabolism processes and regulating the immune system. Studies have also explored the potential relationships between gut microbiota and childhood leukemia. This review is meant to illustrate the roles of gut microbiota in the onset of acute childhood leukemia, as well as in the progress and prognosis of leukemia and how the treatments for leukemia affect gut microbiota. Besides, this review is focused on the possibility of building or rebuilding a healthy gut microbiota by adjusting the diet construction so as to help clinicians deal with childhood leukemia.

#### *Edited by:*

*Liwei Xie, Guangdong Institute of Microbiology, China*

#### *Reviewed by:*

*Zongxin Ling, Zhejiang University, China Chuan Wang, Auburn University, United States*

#### *\*Correspondence:*

*Runming Jin jinrunm@qq.com Hongbo Chen hbchen@hust.edu.cn*

#### *Specialty section:*

 *This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 19 March 2019 Accepted: 24 May 2019 Published: 19 June 2019*

#### *Citation:*

*Wen Y, Jin R and Chen H (2019) Interactions Between Gut Microbiota and Acute Childhood Leukemia. Front. Microbiol. 10:1300. 10.3389/fmicb.2019.01300*

Keywords: gut microbiota, acute childhood leukemia, immune system, long-term health problem, diet construction

#### INTRODUCTION

Leukemia, the commonest childhood malignancy, mainly consists of acute lymphoblastic leukemia (ALL) and acute myeloid leukemia (AML) (Steliarova-Foucher et al., 2017). Over the past decades, tremendous progresses have been made in the cure of childhood leukemia. The mortality rates of childhood cancers in the United States have decreased by more than 50% from 1975 to 2010, and the 5-year survival rate for ALL children <15 years old has increased to 91%, while the survival rate for AML children has increased to 68% (Smith et al., 2014). Genetic background, birth weight, birth order (Crump et al., 2015a,b; Paltiel et al., 2019), caesarean delivery (Marcotte et al., 2016), breastfeeding (Amitay and Keinan-Boker, 2015), low dose of ionizing radiation (Little et al., 2018), and some other exposures are reported to influence the incidence of childhood leukemia. However, the deep links between these factors and acute childhood leukemia lack exploration, and the exact mechanisms for acute childhood leukemia are still not clear.

Gut microbiota is recently recognized as a factor that could be important in regulating the progress of diseases (including gut diseases, diabetes, and others). Microbiota, microorganisms that accommodate at various sites of the human or animal body, develops during the first few years of life and then lives in symbiosis with humans all their life (Arrieta et al., 2014; Hollister et al., 2015; Cheng et al., 2016). Gut microbiota is considered to be one of the largest and most complex ecosystem that is coevolved with the gastrointestinal tract. Since several factors such as genetic, environmental, and lifestyle can influence microbial constitutions (Little et al., 2018; Rothschild et al., 2018), these factors along with gut microbiota should be evaluated as integrated for cancer progress. The aim of this review was to figure out current understandings on interactions between gut microbiota and acute childhood leukemia and make out what can be done in future studies for the management of childhood leukemia.

#### GUT MICROBIOTA CHANGES RAPIDLY DURING CHILDHOOD AND SHOULD BE TAKEN INTO ACCOUNT IN THE FUTURE STUDY DESIGN

With the development of gene sequencing methods, studies are able to be carried out to identify the constitution and diversity of microbiota as well as the crucial roles of microbiota in maintaining body health and regulating the progress of diseases. Several body sites (including the gastrointestinal tract, oral, skin, etc.) have been identified to harbor microbiota, among which the gastrointestinal tract is the largest and most complex one, which harbors approximately 100 trillion microorganisms (mainly composed of bacteria) in the human body (Bull and Plummer, 2014; Valdes et al., 2018). The roles of gut microbiota in maintaining body health have been explored (Beaumont et al., 2016; Falony et al., 2016; De Palma et al., 2017; Little et al., 2018), and gut microbiota is found to be vital for humans, probably by participating in metabolism, regulating the movement and development of the intestinal tract, promoting the development of the brain, as well as regulating the immune system (Zhang et al., 2015b; Levy et al., 2017; Chen et al., 2018; Rothschild et al., 2018; Valdes et al., 2018). With so many diseases discovered to be associated with gut microbiota, researchers are passionate about figuring out the roles of gut microbiota in various fields so as to find out something new for the clinical diagnosis and management of diseases.

Ever since birth, gut microbiota has interacted with the host's conditions and shaped by numerous factors, such as genetic, diet construction, drugs, and others (**Figure 1**) (Little et al., 2018; Rothschild et al., 2018). Though the gut microbiome is divergent from people to people, the constitution and function of an individual's gut microbiota remain relatively stable, which is quite important for maintaining health (Lozupone et al., 2012; Moya and Ferrer, 2016). Studies believed that gut microbiota is established at the first few years of life and keeps developing during the childhood until adulthood (Arrieta et al., 2014; Hollister et al., 2015; Cheng et al., 2016). A cohort study carried out among healthy preadolescent children with ages from 7 to 12 years found that the diversity of gut microbiota was at a similar level in healthy children and adults, while the composition and function of the microbiome differed. It mainly consists of *Bifidobacterium* spp. and *Faecalibacterium* spp. for children, while it mainly consists of *Bacteroides* spp. for adults. As for functional differences, most of the children's microbiota are found to be able to promote development, while those of adults mostly participate in inflammation, obesity, etc. (Hollister et al., 2015). Another study that included children 1–4 years old drew similar results (Cheng et al., 2016). The gut microbiota community is found to change rapidly during the first few years and stay stable in the following years of adulthood until the decline of stability and function of the microbiota community for elders (Lan et al., 2013; Arrieta et al., 2014; Hollister et al., 2015; Cheng et al., 2016; An et al., 2018).

Differences in composition of the gut microbiota between healthy children and healthy adults, as well as the rapid change in childhood gut microbiota composition, stressed the importance of figuring out the specific vital leukemia-causing microorganisms based on a baseline childhood microbiota diversity and constitution. Arrieta et al. (2014) believed that the development of gut microbiota can be divided into six stages (at birth, 1 month, 6 month, 12 months, 3 years, and more than 3 years old) since the composition and diversity differed for every stage. For future studies on gut microbiota and childhood leukemia, the altering of gut microbiota along with aging should be taken into account. A recent study of gut microbiota, which involved 7,009 individuals from 14 districts in Guangdong province, showed that the locations of the hosts were associated with the variations of microbiota (He et al., 2018). Though there could be some confounding for this result, it reminds us that regional diversity is another factor which should be taken into account during the designing of future studies about microbiota.

#### STUDIES ABOUT THE POSSIBLE RELATIONSHIPS BETWEEN GUT MICROBIOTA AND LEUKEMIA

Since the incidence of ALL for children is much higher than that of AML (Steliarova-Foucher et al., 2017), studies for childhood AML are quite rare, and for better understanding, some studies about adult AML are included here. Nearly all children with leukemia are treated with systemic chemotherapy, and some may even receive allogeneic hematopoietic stem cell transplantation (allo-HSCT). Drugs used in chemotherapeutic and antibiotic treatments are known to disturb the host gut microbiota (Holler et al., 2014; Keeney et al., 2014) and, as a result, damage the mucosal protection and immunologic balance, and then contribute to the inflammation of the intestine (Holler et al., 2014).

### WHICH SPECIFIC TYPES OF INFECTION ARE THE PROTECTIVE OR DETRIMENTAL FACTORS FOR THE OCCURRENCE OF CHILDHOOD LEUKEMIA

According to a landscape study, childhood cancers are frequently driven by a single disease-specific mutation, which is quite different from the mechanisms for adulthood cancers (Bandopadhayay and Meyerson, 2018). The most common alterations for pediatric leukemia are CDKN2A, IKZF1, ETV6, and RUNX1, which mainly participate in the regulation of cell cycle and transcription (Ma et al., 2018). Mutations in PAX1 (transcription) and NOTCH1 (notch) were only found in ALL, while mutations in CBFB (transcription) were only found in AML (Ma et al., 2018). For some subtypes of pediatric leukemia, the "two genetic hits" hypothesis proposes that a secondary genetic change is indispensable for the arisen on the basis of a fusion gene or hyperdiploidy ever since in the utero (Knudson, 2001; Greaves, 2018). It can be best verified by the decreasing concordance rate with aging for monochorionic twins. Although monochorionic twins are considered to share the same initial genetic change (with equal preleukemic stem cells), a secondary genetic change is believed to be the cause for the condition that only one of the twins develops ALL in children (Cazzaniga et al., 2011; Bateman et al., 2015) or AML in adults (Jaiswal et al., 2014; Shlush et al., 2014).

More than 20 possible exposures, such as prenatal factors (Marcotte et al., 2014; Crump et al., 2015a,b), caesarean delivery (Marcotte et al., 2016), breastfeeding (Amitay and Keinan-Boker, 2015), low dose of ionizing radiation (Little et al., 2018), as well as infections, have been reported to be related with the occurrence of acute childhood leukemia. The "delayed infection hypothesis" believed that earlier exposures to microbiome are protective factors for childhood ALL, while the later infections without earlier exposures may contribute to the vital secondary variation that causes leukemia (Greaves, 2018). Studies of factors that are associated with exposures to infections (such as birth order, timing of birth, and caesarean delivery by which children were not exposed to the microbes in the maternal vaginal) can support this hypothesis to a certain degree (Marcotte et al., 2016). The reduction of exposures to early common infections as well as factors that relate with microbiota colonization (such as breastfeeding, vaginal delivery) were considered to increase the risk of childhood leukemia (Ajrouche et al., 2015), while some other studies claimed that medically diagnosed infections in infancy or before diagnose of ALL or AML indicate increased risks for childhood leukemia (Chang et al., 2012; Rudant et al., 2015). As for the development of *de novo* adult AML, only gastrointestinal infections were considered to be risk factors (Ostgard et al., 2018).

The different conclusions of the different studies may come from the sample selection. By figuring out the earlier stage of infection and analyzing the relationship between acute childhood leukemia and earlier or later stage infections separately, conclusions might be more reliable. Studies stressed the fact that the immune system and microbial infectious exposures influence each other both *in utero* and in infancy (Olszak et al., 2012; Lim et al., 2015; Laforest-Lapointe and Arrieta, 2017; Torow and Hornef, 2017; Haas, 2018). However, how to define early-stage infections as well as how to prove the existence of it are still unsolved. Signe found that newborns who develop B-ALL later are characterized by abnormal concentrations of several inflammatory markers (Soegaard et al., 2018). The abnormal concentration of inflammatory markers represents an abnormal immune function at birth and reminds us that the immune function might play important roles in the development of acute childhood leukemia. The early childhood exposures to infection were associated with the proliferation and expansion of B or T cell clones (Olszak et al., 2012), and early common infections before the maturity of CD4 T cells are likely to adjust the constitution of symbiotic gut microbiota and contribute to an immune tolerance state toward some antigens with the aid of regulatory T cells and sIgA from mothers (Torow et al., 2015). The excessive reaction toward later-stage infection is likely to be the trigger of acute childhood leukemia. However, recent retrospective studies which rely on maternal recall are limited to figure out the existence of earlier-stage infections. Further researches are in great need to identify whether the earlier exposure to specific microorganisms reduces the incidence of childhood ALL, as well as childhood AML, by regulating the gut microbiota and thus contributing to the building of a healthy immune system.

#### INTERACTIONS BETWEEN GUT MICROBIOTA AND TREATMENTS FOR CHILDHOOD LEUKEMIA

Myelosuppression and immunosuppression are common conditions for children with leukemia during anticancer therapeutics. Infections (mostly bloodstream infections) that are followed by myelosuppression and immunosuppression play important roles in the morbidity and mortality for childhood leukemia. The disturbance of the gut microbiota during chemotherapy procedures and allo-HSCT in children with leukemia has been explored. Changes of stool microbiota were examined in a large cohort study of children with ALL to reflect the changes of gut microbiota. The diversity of fecal microbiota reduced remarkably after induction and reinduction chemotherapy (Hakim et al., 2018). Hakim et al. (2018) believed that the presentation of *Proteobacteria* including *Enterobacteriaceae* and *Pseudomonas* species, and other bacteria in the gut microbiome before or during chemotherapy could be used for predicting subsequent outcomes such as diarrhea, bloodstream infections, or febrile neutropenia for childhood leukemia. Similar conclusions have also drawn in some adult allo-HSCT, AML, and non-Hodgkin lymphoma (Montassier et al., 2015; Taur et al., 2015; Galloway-Pena et al., 2016). In the large cohort study, though the diversity of gut microbiota could recover to the initial level, the composition was differed. The composition of the microbiota in children instead of the diversity in adults was identified to be independently predictive of infections caused by immunosuppression during chemotherapy (Hakim et al., 2018). Another study believed that the diversity and composition of gut microbiota before treatment can be applied to predict chemotherapy-related bloodstream infections (Montassier et al., 2016). However, a study proposes that stool microbiota is quite different from the microbiota that is detected from intestinal mucosa (Zmora et al., 2018). So, more studies are needed to confirm this opinion and to identify the representativeness of stool microbiota for gut microbiota.

Several studies illustrated that the disturbance of microbiome caused by antibiotics is not always temporarily, but in some cases continues (Hernandez et al., 2013; Perez-Cobas et al., 2013; Vangay et al., 2015). Antibiotic-induced shifts can increase the susceptibility toward *Clostridioides difficile* infection (Hernandez et al., 2013). Methotrexate (MTX) is widely used in the treatment for childhood leukemia. Studies showed that the gastrointestinal toxic induced by MTX is vital for patient management (Paci et al., 2014). A mice study showed that the disturbance of the gut microbiota for wild-type mice resulted in a tendency of suffering from MTX-induced mucosal injuries (Frank et al., 2015). A review claims that the microbiota interacts with anticancer drugs mainly in three aspects: improving the drug efficacy, reducing the anticancer effect, and increasing or reducing the toxicity (Panebianco et al., 2018).

The interactions between gut microbiota and therapeutic processes for childhood leukemia can be identified from divergent aspects: (1) whether the diversity and composition of gut microbiota can influence the efficacy or toxicity of drugs used during the therapeutic processes and how; (2) whether the treatments (chemotherapy or allo-HSCT) disturb the gut microbiota and how; (3) whether the gut microbiota can be used for predicting therapy-related complications (such as infections and diarrhea); and (4) whether it is possible for clinicians to deal with long-term health problems or therapy-related complications by regulating the gut microbiota.

#### WHAT CAN DIET REGULATION DO BOTH FOR GUT MICROBIOTA AND ACUTE CHILDHOOD LEUKEMIA

Since the balance of gut microbiota is rather important in childhood leukemia, efforts made to regulate or adjust the gut microbiota to a healthy state are in great need. Compared with administration of a multistrain probiotic preparation, the postantibiotic gut microbiota both in human and murine was rebuilt to the initial state more quickly by autologous fecal microbiome transplantation (Suez et al., 2018), similarly to three other researches (Yan et al., 2016; Bidu et al., 2018; Smillie et al., 2018; Suez et al., 2018). Besides, symbiosis between host and bacterium was believed to be dominantly driven by the bacterium's adaptation to the host's diet in a *Drosophila* model (Martino et al., 2018). Another study believed that melatonin (which is sufficient in several foods) supplementation can increase the diversity and regulate the composition of gut microbiota in mice (Ren et al., 2018). A meta-analysis that involved 18 studies believed that by breastfeeding for at least 6 months, the risks of childhood leukemia were reduced significantly compared with no or shorter-time breastfeeding (Amitay et al., 2016). There are sufficient prebiotic and antibodies for specific pathogens (which each infant's mother is exposed to) and much more natural killer cells in breast milk, which are essential for building a healthy microbiota in the gastrointestinal tract (Benno et al., 1984; Bode and Jantscher-Krenn, 2012; Brown, 2013). Thus, we can assume that the different outcomes of divergent diet constructions might come from the altered gut microbiota which evolved with the human immune system, especially for infants and children.

Long-term childhood leukemia survivors are faced with many long-term health problems such as obesity, cardiopulmonary toxicity, secondary malignancy, late neurotoxic effect, and others (Essig et al., 2014; Zhang et al., 2014; Cheung and Krull, 2015; Withycombe et al., 2015; Duncan et al., 2018). The unhealthy dietary behaviors (such as high intake of fat, sodium, sweets, and low intake of fruit, vegetables, and whole grains) (Tylavsky et al., 2010; Badr et al., 2011; Fuemmeler et al., 2013; Zhang et al., 2015a; Duncan et al., 2018) of childhood cancer survivors and the usage of antibiotics are known to increase weight by regulating the composition of the gut microbiota (Blaser, 2016). Obesity was identified to be a risk for childhood leukemia, while fasting was found to reduce the incidence and even reverse the progression of ALL in mouse models (Orgel et al., 2016; Lu et al., 2017). Hopefully, the incidence and progression of leukemia are likely to be stopped or reversed by simply fasting; prolonging the duration of breastfeeding; adjusting to a low intake of fat, sodium, and sweets; and a high intake of fruit, vegetables, and whole grains. Besides, whether the supplement of melatonin or probiotics, as well as fecal microbiome transplantation, can help keep a healthy gut microbiota in children with leukemia remains to be explored.

#### CONCLUSIONS

Recent studies have explored the possible relationships between gut microbiota and acute childhood leukemia. The "delayed infection hypothesis" highlights a favorable role of early infection in preventing the onset of childhood leukemia. This is somehow consistent with the idea that an oversanitized condition may lead to some noninfectious and immunological diseases like asthma, obesity, and diabetes. It is also found that gut microbiota develops with ageing. However, which specific microorganisms help push the onset and progression of childhood leukemia is still unclear. Besides, diet regulation such as fasting and breastfeeding may lower the incidence of childhood leukemia and reverse the progression by adjusting the gut microbiota. All of the above implies that acute childhood leukemia may

#### REFERENCES


be somewhat preventable by changing the lifestyle. In turn, antileukemia therapies can disturb the gut microbiota and bring short- and long-term health problems. This may be addressed in the future by gut microbiota transplantation and probiotic supplements, which demands some prospective studies in leukemia patients first.

There is also an urgent need to figure out the specific microorganisms for the onset and progression of childhood leukemia and dig out the latent mechanisms for the links between gut microbiota and childhood leukemia. By doing so, more targeted therapy (such as specific probiotic supplements, specific microbiome transplantation, as well as diet regulation that benefits specific microorganisms) instead of nontargeted probiotic supplements or others for reducing the incidence of leukemia and eliminating antileukemia treatment-related longterm health problems could be put forward. As for study design, both regions and ages should be taken into account to generate a confident baseline childhood microbiome diversity and constitution. In addition, the representativeness of stool microbiota for gut microbiota is in doubt and needs further confirmation.

#### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

#### FUNDING

This work was supported by a grant from National Natural Science Foundation of China (31701207 to HC).


pre-adolescent pediatric gut microbiome. *Microbiome* 3:36. doi: 10.1186/ s40168-015-0101-x


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2019 Wen, Jin and Chen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Effects of Blidingia sp. Extract on Intestinal Inflammation and Microbiota Composition in LPS-Challenged Mice

Wei Song1,2, Yan Li1,2, Xuelei Zhang1,2 and Zongling Wang1,2 \*

<sup>1</sup> Key Laboratory of Science and Engineering for Marine Ecology and Environment, The First Institute of Oceanography, Ministry of Natural Resources of the People's Republic of China, Qingdao, China, <sup>2</sup> Laboratory of Marine Ecology and Environmental Science, Qingdao National Laboratory for Marine Science and Technology, Qingdao, China

#### Edited by:

Jie Yin, Institute of Subtropical Agriculture (CAS), China

#### Reviewed by:

Liuqin He, Hunan Normal University, China Deguang Song, Yale University, United States Qingbiao Xu, Huazhong Agricultural University, China

> \*Correspondence: Zongling Wang kaso1986@126.com

#### Specialty section:

This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology

Received: 14 April 2019 Accepted: 31 May 2019 Published: 25 June 2019

#### Citation:

Song W, Li Y, Zhang X and Wang Z (2019) Effects of Blidingia sp. Extract on Intestinal Inflammation and Microbiota Composition in LPS-Challenged Mice. Front. Physiol. 10:763. doi: 10.3389/fphys.2019.00763 Blidingia sp. is a green alga that has spread rapidly in Subei Shoal, China. To explore the potential beneficial effects of Blidingia sp., we investigated the anti-inflammatory activity of its water–methanol extract of Blidingia sp. in a mouse model of lipopolysaccharide (LPS)-induced intestinal inflammation. The results revealed that the administration of Blidingia extract significantly alleviated the LPS-induced increase of the inflammatory cytokine content in the serum, as well as latter's gene expression in the ileum. Moreover, the extract inhibited the phosphorylation of NF-κB and IκBα in LPS-challenged mice. Apart from these changes, the extract also averted intestinal morphology damage(s) and cell apoptosis in mice. Interestingly, the extract also had beneficial effects on the diversity and composition of caecal microbiota in LPS-challenged mice. In conclusion, the results suggested that Blidingia extract had beneficial effects on the recovery of intestinal function by reducing the inflammatory response, improving the maintenance of intestinal morphology, and decreasing cell apoptosis in LPS-induced intestinal inflammation. In addition, the beneficial effects of the extract on caecal microbiota composition may play a role in its anti-inflammatory activity. These results suggested that Blidingia extract could be potentially used in preventing intestinal inflammation.

#### Keywords: apoptosis, Blidingia sp., inflammation, microbiota, morphology

# INTRODUCTION

Inflammation is an innate defense mechanism in response to tissue injury, stress, and infection. However, an excessive and acute inflammation could be potentially damaging. Moreover, a chronic and prolonged inflammation could promote the development of diabetes, obesity, cancer, and cardiovascular diseases (Wang Z. et al., 2017). The gastrointestinal tract is a primary site that faces exogenous materials and serves as a vital defense barrier against harmful substances and microbiota (Zhang H. et al., 2018). As a result, it is susceptible to inflammatory responses. Acute treatment of lipopolysaccharides (LPS), (produced by the Gram-negative bacterial cell wall) induces rapid accumulation of various cytokines that play an important role in the inflammatory process (Zhou et al., 2012, 2017b). Particularly, the use of an LPS-induced intestinal inflammation rodent model is common for evaluating the anti-inflammatory activity of natural products (Raetz and Whitfield, 2002).

Seaweeds, such as Ulva prolifera, have been used as a health food for ages due to their richness in polysaccharides, polyphenols, essential amino acids, and mineral elements (Song et al., 2018). Polysaccharides, flavonoids, polyphenols, and other components extracted from the seaweeds including U. prolifera, have exhibited anti-inflammatory effects since long (Okai and Higashi-Okai, 1997; Jung et al., 2013; Song et al., 2018). In addition, these extracts exhibit profound effects on the intestinal microbiota composition which is closely associated with the development of intestinal inflammatory response (Ren et al., 2018; Zhang Z. et al., 2018). Blidingia sp. is ubiquitous on several coastlines and often grows together with Ulva spp. (Woolcott et al., 2000). It is one of the dominant fouling green macroalgae in the Pyropia aquaculture facilities of Subei Shoal (Jiangsu, China) (Wang et al., 2015). Until now, little research has reported the effects of Blidingia extract on LPS-induced inflammation. To explore the beneficial effects of this extract, we scraped Blidingia sp. from the Pyropia aquaculture facilities for extraction. Furthermore, the components of the extract were analyzed through UHPLC-Q-Extractive-Orbitrap/MS and its effects on intestinal inflammation and microbiota composition were determined.

# MATERIALS AND METHODS

#### Sample Preparation

The green macroalgae Blidingia sp. was scraped from the Pyropia aquaculture facilities in the Subei Shoal. The collected Blidingia sp. was rinsed with sterile seawater and then dried in the sun. The air-dried powder of Blidingia sp. (1 kg) was extracted with 5 L volume of distilled water at 90◦C for 3 h in an ultrasonic bath (200 W, 45 kHz). The supernatant was collected by filtering through the siliceous earth and then submitted to the adsorption chromatography column (120 cm L × 150 mm ID, Huamei Experiment Instrument Plant, Shanghai, China) filled with AB-8 macroporous adsorption resin. After eluted with distilled water at a flow rate of 60 mL/min, the extract was further eluted with 70% methanol. The small molecule mixture (BSE) was obtained by freeze-drying the eluent.

# UHPLC-Q-Extractive-Orbitrap/MS Analysis

First, 50 mg of BSE was dissolved in 1 mL of extract solvent (acetonitrile-methanol-water, 2:2:1, containing internal standard 1 µg/mL) and then centrifugated at 8,000◦ g at 4◦C for 15 min. The resulting supernatants were transferred to LC-MS vials for UHPLC-QE-Orbitrap/MS analysis as previously described (Li et al., 2017). The analyses were performed using an 1290 UHPLC system (Agilent, Palo Alto, CA, United States) with a UPLC HSS T3 column (2.1 mm × 100 mm, 1.8 µm) coupled to Q Exactive (Orbitrap MS, Thermo, Somerset, NJ, United States). The mobile phase A was 0.1% formic acid in water for positive, and 5 mmol/L ammonium acetate in water for negative, and the mobile phase B was acetonitrile. The elution gradient was set as follows: 0 min, 1% B; 1 min, 1% B; 8 min, 99% B; 10 min, 99% B; 10.1 min, 1% B; and 12 min, 1% B. The flow rate was 0.5 mL/min and the injection volume was 2 µL.

# Animals and Treatment

Thirty nine-week-old male C57BL/6J mice were obtained from the SLAC Laboratory Animal Central (Changsha, China) and were acclimatized for 2 weeks under an environmental cycle of 12 h light/12 h dark. All animals were fed ad libitum and free to obtain water during the experiment. All mice were orally gavaged with either Blidingia sp. extract (BSE) (10 mg/kg body weight, n = 10) or the same volume of saline (n = 20) for 14 days. The dosage of Blidingia sp. extract used in the present study was based on our preliminary experiments. At 10:00 am on day 15, the mice gavaged with saline were challenged with intraperitoneally injection of either LPS (0.5 mg/kg, Escherichia coli serotype 055:B5; Sigma Chemical, Inc., St. Louis, MO, United States; n = 10) or saline (n = 10), while the mice gavaged with BSE were all challenged with LPS (0.5 mg/kg; n = 10). At 2 h after treatment with LPS or saline, samples of blood, ileum (1–2 cm proximal to the ileocecal valve) and caecal digesta were collected for further analysis. The experimental protocol was approved by the Protocol Management and Review Committee of the First Institute of Oceanography of China, and the mice were cared for and sacrificed according to the animal care guidelines of the First Institute of Oceanography of China.

# Determination of Inflammatory Cytokine Content in Serum

Tumor necrosis factor a (TNF-a), Interleukin 6 (IL-6), IL-8, and IL-10 content in serum were determined using ELISA quantitative kits (Cusabio Biotech, Wuhan, China) according to the manufacturer's instructions.

# RT-qPCR Analysis

The ileum samples were used for total RNA extraction using TRIzol reagent (Invitrogen, Shanghai, China) and then cDNA was obtained using the PrimeScript RT reagent kit (Takara, Dalian, China) (Zhou et al., 2017a, 2018). RT-qPCR was performed in 10 µL assay volumes containing 3 µL of DEPC-treated H2O, 0.2 µL of ROX, 1 µL of cDNA template, 0.4 µL forward primer and 0.4 µL reverse primer, and 5 µL of SYBR Green mix (Takara). All samples were run in triplicate and the results were obtained by calculating the average values. The primer sequences are presented in **Supplementary Table S1**.

# Protein Qualification by the Wes Simple Western System

Protein expression were qualified using the Wes Simple Western System (Proteinsimple, San Jose, CA, United States). Proteins extracted from the ileum samples were mixed with Master Mix, dithiothreitol, fluorescent standards and Simple Western Sample Buffer (Proteinsimple) and then were loaded into Wes 25-well plates. Primary antibodies (β-actin, phopho NFκB, NFκB, phopho IκBα and IκBα, Abcam, Cambridge, MA, United States), secondary antibodies, luminol-peroxide mixture, stacking gel matrix, and separation gel matrix were

added according to the manufacturer's instructions. Results were collected using the "gel view" function of the Protein Simple software (Proteinsimple).

# Haematoxylin-Eosin (HE) Staining and Transmission Electron Microscopy

The ileum samples were opened longitudinally, fixed with 4% formaldehyde, and then embedded in paraffin. Thereafter, samples were sliced as sections with 8-µm thickness for HE staining (Yin et al., 2018). The histological scoring was performed based on previous description (Zhang H. et al., 2018). Meanwhile, the ileum samples were also fixed in 2.5% glutaraldehyde and post-fixed in osmium tetroxide. After washed with PBS, the ileum samples were dehydrated with graded alcohol and then embedded in Epon-Araldite resin. Finally, samples were sliced as ultrathin sections with 50-nm thickness, stained with uranyl acetate and lead citrate, and observed with a Zeiss 902 transmission electron microscope.

# Assessment of Apoptosis

The ileum samples were opened longitudinally, fixed with 10% formaldehyde, and then embedded in paraffin. Thereafter, samples were sliced as sections with 5-µm thickness for TUNEL staining using an in situ cell death detection kit (Roche, Shanghai, China). Nuclei were stained using DAPI mounting solution (Vector, Burlingame, CA, United States). Representative results were collected using a light microscope.

#### Measurement of the Caecal Microbiota

Samples of all the content within the caecal digesta from treated mice were pooled and homogenized and then used for DNA extraction using the QIAamp DNA stool MiniKit (Qiagen, Shanghai, China). Bacterial 16S rRNA gene sequences (V3–V4 region) were amplified using specific primers with Premix Ex TaqTM Hot Start Version (Takara, Dalian, China). A total volume of 50 mL consisting of 12.5 mL of Phusion High-Fidelity PCR Master Mix (New England BioLabs Inc., Beverly, MA, United States), 50 ng of template DNA, 1 mL of each primer, and PCR-grade water were mixed for the performance of PCR reaction. Then, MiSeq Illumina sequencing was performed on the sequencing reaction (Illumina Inc., San Diego, CA, United States) for paired-end reads. Following, the paired-end reads were assembled, merged and assigned to each sample based on the unique barcodes. Based on a 97% sequence similarity, high-quality tags were clustered into operational taxonomic units (OTUs), which were used for further analysis using database of Greengenes by RDP algorithm. The results of alpha and beta diversity and principal coordinate analysis (PCoA) were obtained using QIIME software. The results of linear discriminant analysis (LDA) effect size (LEfSe) were collected using the LEfSe tool.

# Statistical Analyses

All data were analyzed by one-way ANOVA using the general linear model procedures and a mixed procedure (PROCMIXED) of SAS software version 9.2 (SAS Institute Inc., Cary, NC, United States). Data are presented as least squares means ± SEM. Mean values were considered significantly different when P < 0.05.

# RESULTS

### Characterization of Components of Blidingia sp. Extract

In total, over two hundred compounds were detected in the Blidingia sp. extract with the high-resolution UPLC-QE-Orbitrap/MS system (**Supplementary Table S2**). Twenty-eight species of amino acids, 15 species of nucleotides, 103 species of peptides, 49 species of organic acids, 10 species of alkaloids, 3 species of phenols, and other compounds were detected in positive ionization mode.

#### Blidingia sp. Extract Alleviates Inflammatory Response in LPS-Challenged Mice

LPS challenge induced significant increases of TNF-a, IL-6, IL-8, and IL-10 contents in serum, while Blidingia sp. extract significantly decreases their contents (**Figure 1**). Moreover, mRNA expression of TNF-a, IL-6, IL-8, and IL-10 in ileum were also increased. However, administration of Blidingia sp. extract alleviated these LPS-induced changes (**Figures 2A–D**). Additionally, expression of phosphorylated NFκB and IκBα in LPS-challenged mice were significantly higher when compared with control mice, while no significant difference was observed in their expression between control mice and mice administrated with Blidingia sp. extract (**Figures 2E,F**) (**Supplementary Figures S1–S4**).

#### Blidingia sp. Extract Alleviates LPS-Induced Histopathological Changes and Apoptosis

The HE staining results showed pathological changes, such as shedding and obvious edema, following LPS challenge in the ileum tissue, while no such changes were observed in control mice and mice administrated with Blidingia sp. extract (**Figures 3A–F**). Mice challenged with LPS had a significant higher histological index of ileum when compared with control mice or mice administrated with Blidingia sp. extract (**Figure 3G**). Moreover, the TEM results showed that irregularly arranged microvilli were only observed in ileum of LPS-challenged mice (**Figures 3H–J**). TUNEL staining revealed that the level of apoptosis was higher in LPS-challenged mice when compared with control mice, while no difference was observed between control mice and mice administrated with Blidingia sp. extract (**Figure 4A**). Moreover, LPS challenge induced significant increases of mRNA expression of Bax and Caspase 3, while significant decreases of mRNA expression of cFLIP and Bcl2 (**Figures 4B–E**). However, administration of Blidingia sp. extract alleviated these LPS-induced changes.

# Effects of Blidingia sp. Extract on Microbial Diversity

Based on the V3 + V4 region of the 16D rDNA sequence, an average of 80,012 (71,129-86,714) effective tags were used for the analysis of the study, OTUs were generated from sequences with at least 97% similarity. Alpha diversity including Observed species, Ace, Chao1, Shannon and Simpson index, as well as weight PCoA analysis were measured to detect the diversity and structure of caecal microbial communities in mice after treatment of LPS and BSE. The alpha diversity of microbial communities, as indicated by the index of Observed species, Ace, Chao1, Shannon and Simpson, was decreased significantly by the treatment of LPS, while BSE supplementation significantly increased the microbial diversity (**Figures 5A–E**). In addition, the PCoA plot based on the weighted UniFrac metric showed that caecal microbiota was significantly regulated by LPS and BSE treatment (**Figure 5F**).

# Effects of Blidingia sp. Extract on Microbial Compositions

The order level analysis demonstrated that the percentage of Clostridiales and Campylobacterales were significantly increased in the cecum of mice after the treatment of LPS, while BSE treatment had no effects on these increases (**Figures 6A,B**); LPS treatment also caused significant decreases of percentage of Lactobacillales and Erysipelotrichales, while BSE treatment alleviated these decreases (**Figures 6A,B**). In the class level, the percentage of Lachnospiraceae and Helicobacteraceae were significantly increased after the treatment of LPS, while BSE treatment had no effects on these increases (**Figures 6C,D**); LPS treatment also caused significant decreases of percentage of Lactobacillaceae and Erysipelotrichaceae, as well as an increase of Ruminococcaceae, while BSE treatment alleviated these changes in some extent (**Figures 6C,D**). The Venn diagram, revealing the overlapping OTUs data, displaying that 504 OTUs were universal to all samples, and there are 48 unique OTUs in control mice, 45 unique OTUs in LPS-treated mice and 33 unique OTUs in BSE-treated mice (**Figure 7A**).

To identify the bacterial taxa associated with the beneficial effects by BSE treatment, we used a LEfSe analysis to compare caecal microbiota (**Figures 7B,C**). A significant difference in the relative abundance was determined when LDA score >4. We observed higher relative abundance of Bacilli, as well as its lower taxa Lactobacillales and Lactobacillaceae in both Control and BSE group when compared with the LPS group. In addition, a higher relative abundance of Erysipelotrichaceae, as well as its lower taxa Erysipelotrichales and Erysipelotrichaceae were also observed in Control group. The LPS group showed higher abundance of Ruminococcaceae in the family level, Clostridia in the class level and Clostridiales in the order level when compared with the Control group.

# DISCUSSION

The green alga, Blidingia sp., is widely distributed in Subei Shoal, China. In the present study, Blidingia sp. was collected and the effects of its extract on the intestinal inflammatory response and caecal microbiota composition were evaluated.

expressed as mean±SEM, n = 3; <sup>∗</sup>p < 0.05.

It was observed that Blidingia sp. extract decreased the inflammatory cytokine content and inhibited the activation of the NF-κB signaling pathway in a mouse model of LPS-induced intestinal inflammation. The extract is majorly constituted of amino acids, polypeptides, organic acids, phenols, flavonoids, nucleotides, and alkaloids. These components are speculated to play critical effects on its anti-inflammatory activity. LPSinduced inflammation is associated with intestinal morphological damages and elevated apoptosis levels that act as key indicators in characterizing the dysfunctional condition of intestine. These changes were, however, not observed after the mice were administrated with the Blidingia extract. This observation suggested that the extract could have supported intestinal function by protecting the intestine from morphological damage and decreasing cell apoptosis besides alleviating the inflammatory response.

High diversity of intestinal microbiota is suggested as more stable and healthier (Konstantinov et al., 2004), whereas,

were stained with DAPI (blue). Relative mRNA expression of Caspase 3 (B), Bax (C), cFLIP (D), and Bcl2 (E). CONT, mice gavaged with sterile saline; LPS, mice injected with lipopolysaccharide; and BSE, mice gavaged with Blidingia sp. extract and injected with lipopolysaccharide. Values are expressed as mean±SEM, n = 8; <sup>∗</sup>p < 0.05.

human patients and animal models with intestinal inflammation displayed reduced bacterial species diversity (Wang K. et al., 2017; Zhang H. et al., 2018). In the present study, the results proposed that LPS treatment not only decreases the community richness (Chao1 and ACE indices) but also community diversity (Shannon and Simpson indices) of the caecal microbiota. In addition, PCoA analysis revealed a shift in the bacterial community composition. These results were corroborated by a previous study on LPSchallenged piglet model (Wang et al., 2019). Importantly, LPS treatment caused a significant increase in the bacterial family Ruminococcaceae, that harbors potential pathogen bacteria found in other intestinal inflammatory models (Wang et al., 2019). On the other hand, LPS treatment marked a decrease of the order Lactobacillales that is comprised of the lactic acid bacteria with well-known probiotic properties (Ritchie et al., 2015). The results indicated that LPS accounted for intestinal inflammation in association with the disruption of caecal microbiota. In addition, LPS treatment resulted in a significant increase of Clostridiales (at the order level) and Lachnospiraceae (at the family level) both of which are bacterial subclasses of the phylum Firmicutes. Ruminococcaceae and Lachnospiraceae are the most abundant Firmicute families observed in the intestine environments (Tap et al., 2009). Lachnospiraceae has a beneficial effect on the health of intestinal epithelial tissue as it is documented to be associated with butyrate production (Barcenilla et al., 2000). However, reduction of Ruminococcaceae and Lachnospiraceae has been reported in patients with Crohn's disease or inflammatory bowel disease, which are in agreement with our results (Frank et al., 2007; Fujimoto et al., 2013). The result changes were presumed as an immediate positive response of the caecal

microbiota to LPS induced inflammation. However, the exact explanation of these changes needs to be further elucidated.

Recently, many studies have revealed that polysaccharides, flavonoids, and polyphenols extracted from the green seaweeds such as Porphyra haitanensis and U. prolifera had potential effects on the intestinal microbiota (Zhang Z. et al., 2018; Lin et al., 2019; Yan et al., 2019). In the present study, the effects of Blidingia sp. extract on the caecal microbiota of LPStreated mice were assessed. The results displayed beneficial effects as the extract increased the bacterial diversity. Importantly, Blidingia extract increased the population of Bacilli, as well as its lower taxa; Lactobacillales, Lactobacillaceae, and Lactobacillus in LPS-treated mice. These effects were in agreement with a previous report on Enteromorpha clathrata extracts (Shang et al., 2018). However, Blidingia extract displayed no effects on Clostridiales and Campylobacterales (at the order level), as well as Lachnospiraceae and Helicobacteraceae (at the family level). Nevertheless, the results suggested a prebiotic effect of the extract on caecal microbiota of the mouse model with LPSinduced inflammation.

#### CONCLUSION

fphys-10-00763 June 22, 2019 Time: 14:12 # 8

In conclusion, the results of this study indicated that Blidingia extract has beneficial effects on the recovery of intestinal function by alleviating the inflammatory response, improving the maintenance of intestinal morphology, and decreasing cell apoptosis in a mouse model of LPS-induced intestinal inflammation. In addition, the extract also exerted positive effects on caecal microbiota diversity and composition, which may play a role in its anti-inflammatory activity. The results suggested the potential use of Blidingia extract in preventing intestinal inflammation.

#### DATA AVAILABILITY

The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the Protocol Management and Review Committee of the First Institute of Oceanography of China. The protocol was approved by the Animal Care Guidelines of the First Institute of Oceanography of China.

#### AUTHOR CONTRIBUTIONS

WS and ZW conceived and designed the research. WS, YL, and XZ performed all the protocol. WS, YL, ZW, and XZ wrote and revised the manuscript.

#### REFERENCES


#### FUNDING

This work was financially supported by the National Key R&D Program of China (2016YFC1402100), National Natural Science Foundation of China (41606190/41876140/41606140), Shandong Natural Science Foundation (ZR2016DB22), Creative Team Project of the Laboratory for Marine Ecology and Environmental Science, Qingdao National Laboratory for Marine Science and Technology (LMEES-CTSP-2018-3/LMEES-YTSP-2018-03-02), Foundation of Key Laboratory of Integrated Monitoring, and Applied Technologies for Marine Harmful Algal Blooms (MATHAB201806). This work was also supported by the Open Fund of CAS Key Laboratory of Marine Ecology and Environmental Sciences, Institute of Oceanology, Chinese Academy of Sciences (KLMEES201803).

#### ACKNOWLEDGMENTS

We thank Captain Lin Wei and the crew on the ship "SURUYUYUN-288" for their assistance in the sample collection.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphys. 2019.00763/full#supplementary-material

FIGURE S1 | Complete protein band of IκBα and pIκBα.

FIGURE S2 | Complete protein band of pNFκB.

FIGURE S3 | Complete protein band of β-actin.

FIGURE S4 | Complete protein band of NFκB.

TABLE S1 | Primer sequences for RT-PCR.

TABLE S2 | Characterization of components of Blidingia sp. extract by UHPLC-Q-Extractive-Orbitrap/MS Analysis.

ecosystem during weaning transition. Anim. Res. 53, 317–324. doi: 10.1051/ animres:2004019



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Song, Li, Zhang and Wang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# A New Isolate of Pediococcus pentosaceus (SL001) With Antibacterial Activity Against Fish Pathogens and Potency in Facilitating the Immunity and Growth Performance of Grass Carps

#### Edited by:

Jie Yin, Institute of Subtropical Agriculture (CAS), China

#### Reviewed by:

Yun-Zhang Sun, Jimei University, China Fei Ling, Northwest A&F University, China Yanbin Zhang, Leonard M. Miller School of Medicine, United States

#### \*Correspondence:

Shengbiao Hu shengbiaohu@hunnu.edu.cn Liqiu Xia xialq@hunnu.edu.cn †These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 24 February 2019 Accepted: 03 June 2019 Published: 27 June 2019

#### Citation:

Gong L, He H, Li D, Cao L, Khan TA, Li Y, Pan L, Yan L, Ding X, Sun Y, Zhang Y, Yi G, Hu S and Xia L (2019) A New Isolate of Pediococcus pentosaceus (SL001) With Antibacterial Activity Against Fish Pathogens and Potency in Facilitating the Immunity and Growth Performance of Grass Carps. Front. Microbiol. 10:1384. doi: 10.3389/fmicb.2019.01384 Liang Gong† , Haocheng He† , Dongjie Li, Lina Cao, Tahir Ali Khan, Yanping Li, Lifei Pan, Liang Yan, Xuezhi Ding, Yunjun Sun, Youming Zhang, Ganfeng Yi, Shengbiao Hu\* and Liqiu Xia\*

State Key Laboratory of Developmental Biology of Freshwater Fishes, Hunan Provincial Key Laboratory of Microbial Molecular Biology, College of Life Science, Hunan Normal University, Changsha, China

Probiotic-feeding continues to be a promising strategy to control the bacterial pathogens in aquaculture. A new Pediococcus pentosaceus strain (SL001) was isolated from 1000s of soil samples, which exhibited wide antimicrobial spectrum of against fish pathogens, involving Aeromonas hydrophila, Aeromonas veronii, Aeromonas sobria, Edwardsiella tarda, Lactococcus garvieae, and Plesiomonas shigelloide. The challenge test against A. hydrophila showed that the survival rate of SL001-supplemented group was significantly higher than that of control group (P < 0.05). Moreover, SL001 could stably colonize in gut of grass carp and increased mucus-secreting goblet cells and extended intestinal villi could be observed in SL001-supplemented group (P < 0.05). Feeding with SL001 supplemented diet could significantly enhance the growth rate (P < 0.05) and markedly affect gut microbiota structure of grass carps, resulting in reduced potential pathogens and increased potential probiotics. Furthermore, feeding grass carps with SL001 caused the up-regulated expression of insulin-like growth factor (IGF-1 and IGF-2) and down-regulated expression of myostatin (MSTN-1 and MSTN-2) (P < 0.05), which probably also account for the increased growth rate of SL001-fed group. Meanwhile, relative mRNA expression levels of immune-related genes in liver, spleen, and head kidney were analyzed in grass carps after feeding for 30 days with SL001 supplemented diets. In all three immune organs, the expression levels of immunoglobulin M (IgM) and complement 3 (C3) were significantly increased (P < 0.05), whereas the interleukin-8 (IL-8) was down-regulated (P < 0.05). Besides, whole genome sequencing revealed several probiotics properties of SL001, including organic acid synthesis, bacteriocin synthesis (coagulin), superoxide dismutase, and digestive enzymes. In conclusion, P. pentosaceus SL001 which could enhance immunity and promoter growth rate of grass carps, is prospective to be used as a dietary probiotic in freshwater fish aquaculture.

Keywords: Pediococcus pentosaceus, antibacterial activity, grass carps, gut microbiota, fish immunity, growth rate

# INTRODUCTION

fmicb-10-01384 June 26, 2019 Time: 15:43 # 2

High density of culture and increase of feeding amount have caused long-term crowding stress that aggravates fish susceptibility to pathogens, and outbreaks of fish diseases have become increasingly serious (Athanassopoulou et al., 2004; DiMaggio et al., 2014; Lin et al., 2018). Usually, fishes are frequently infected by microorganisms associated with viruses, bacteria, and parasites under extremely intensive culture conditions (Song et al., 2014; Tang et al., 2018; Zhou et al., 2018). In the past many years, control strategies on fish pathogens relied mostly on the application of chemicals, such as antibiotics and disinfectants. However, extensive use of medicines has caused antibiotic resistance of pathogenic bacteria (Chandrarathna et al., 2018). Therefore, overuse of antibiotics in aquaculture should be strictly controlled, and alternative methods of controlling fish pathogens must be developed, such as using antimicrobial peptides, vaccines, and probiotics.

Recently, probiotic-feeding has been proved to be a promising strategy to control diseases. Probiotics are beneficial microorganisms introduced by implantation or colonization in specific host's gut to reinforce the intestinal barrier, boost the immune system, or produce antimicrobial substances to suppress pathogen growth (Balcazar et al., 2006; Akhter et al., 2015). Besides, probiotics could promote digestion and enhance nutrient absorption by altering the intestinal microflora, resulting in better growth performance (Balcazar et al., 2006; Falcinelli et al., 2015). Lactic acid bacteria (LAB) and Bacillus species are among the most commonly used probiotic candidates (Cao et al., 2011; Banerjee and Ray, 2017; Alonso et al., 2018; Yi et al., 2018). LAB strains that have been used for fish pathogens control usually include the genera Lactobacillus, Leuconostoc, Streptococcus, and Pediococcus (Perez-Ramos et al., 2018). Pediococcus parvulus 2.6 produces β-glucan, which could benefit colonization of its producer in the fish gut and competition with the pathogen Vibrio anguillarum (Perez-Ramos et al., 2018). Feeding orange-spotted grouper with Pediococcus pentosaceus strain 4012 could not only enhance the growth rate of the grouper and increase the number of red blood cells, but also regulate the gene expression of the pro-/anti-inflammatory cytokines (Huang et al., 2014). Similarly, Pediococcus acidilactici-supplemented diet significantly increases the expression levels of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) in juvenile beluga (Ghiasi et al., 2018).

In aim to screen probiotics which could antagonize bacterial fish pathogens, 100s of bacterial strains have been isolated from soil samples collected from different regions of China. Among them, a novel P. pentosaceus strain (SL001) exhibited excellent antibacterial activity against several important fish pathogens has been added to fish diet and its probiotics properties have been studied. The impact of SL001 feeding on the gut microbiota and growth ability of grass carps was analyzed. Subsequently, the expression of growth-related and immune-related genes of grass carps was measured after feeding with SL001-supplemented diet. Our results indicated that P. pentosaceus SL001 could serve as potential probiotic strain in freshwater aquaculture.

# MATERIALS AND METHODS

### Bacterial Isolation, Maintenance, and Identification

Soil samples from different regions of China were collected and 10-fold serially diluted in glass tubes. 0.1 mL dilutions was spread on Man Ragosa Sharpe (MRS) agar plates and anaerobic cultured on DG250 Anaerobic workstation (DWS, United Kingdom) at 37◦C for 24 h. Single colonies appeared on MRS agar plates were purified twice and stored in 25% (v/v) glycerol at −80◦C. The P. pentosaceus SL001 was grown on MRS liquid medium at 37◦C for 12 h under anaerobic condition, and then scanning electron microscopy (SEM, Hitachi Su8010, Japan) were used to observe the morphology of the bacterial cell. Biochemical characterization of SL001 was done subsequently. Genomic DNA was extracted from a SL001 using a genomic DNA extraction kit according to the manufacturer's instructions (Sangon, China). The DNA fragments carrying 16S rRNA gene were amplified using primer pair 27F (5<sup>0</sup> to 3<sup>0</sup> : AGAGTTTGATCCTGGCTCAG) and 1492R (5<sup>0</sup> to 3<sup>0</sup> : CGGTTACCTTGTTACGACTT) (Wu et al., 2010; Yi et al., 2018). The polymerase chain reaction (PCR) products were purified and cloned into pMD18-T vector (TaKaRa, Japan). Single colonies were picked up and sent for sequencing. Phylogenetic tree was constructed on the basis of 16S rRNA genes by the neighbor-joining method using MEGA6.06 software and evolutionary distances were computed using the Maximum Composite Likelihood method (Tamura et al., 2004).

#### Antimicrobial Activity Test

SL001 was incubated in MRS liquid medium at 37◦C for 24 h under anaerobic conditions. Six fish pathogenic bacteria (**Supplementary Table S1**) were incubated in Luria-Bertani (LB) liquid medium at 30◦C for 12 h. Overnight cultures of pathogenic bacteria were diluted to 10<sup>7</sup> CFU/mL by fresh LB broth, and 200 µL dilutions were spread on LB agar plates respectively. Then, sterile oxford cups were placed on the plates dried for 30 min on vertical flow clean cabinet. 100 µL of cell-free supernatant (CFS, pH 3.48) of SL001 saturated culture was added into oxford cups. MRS and phosphate-buffered saline (PBS, need pH 3.48 with lactic acid, PL) were used as control. After 18 h of incubation at 30◦C, antimicrobial activity was evaluated by measuring inhibition zones (Wang et al., 2018; Yi et al., 2018).

#### Bacterial Safety Evaluation

Hepatic L8824 cell line derived from liver of grass carp was cultured at 37◦C with 5% CO<sup>2</sup> in DMEM (Gibico/Thermo Fisher Scientific, United States) supplemented with 10% fetal calf serum (FCS, BI, Israel) and 1% penicillin-streptomycin solution (BI, Israel). Cells were pre-incubated for 12 h in 24-well flat-bottomed plates (5 × 10<sup>5</sup> cells/well) as previous described (Matsuda et al., 2017). Both SL001 CFS (10 µL) and bacterial cells (1 × 10<sup>7</sup> CFU) were applied to L8824 cells and cell morphological changes were analyzed using an inverted light microscope (Leica Microsystems S.p.A, Italy). CFS of Aeromonas hydrophila which has been proved to be toxic to L8824 cells was used as positive control and MRS was used as negative control.

# Experimental Design

fmicb-10-01384 June 26, 2019 Time: 15:43 # 3

Healthy grass carps (32.1 ± 9 g) obtained from Wangcheng fish pond (Changsha, China) were transferred plastic tanks (volume 60 L) equipped with air pump in laboratory and fed with the basal diet for 1 week. Grass carps were randomly divided into six plastic tanks (38 fishes per tank) and fed twice with different diets per day at 1% of the body weight: three tanks fed with basal diet, three tanks fed with SL001 (1 × 10<sup>9</sup> CFU/g) supplemented diet (Huang et al., 2014; Yi et al., 2018).

Twelve grass carps were sampled randomly from every tank after 30 days of feeding. Three grass carps were taken for immune-related and growth-related genes expression analysis. Immune organs (liver, spleen, and head kidney) and muscle were collected, sliced, and immediately frozen at −80◦C until RNA isolation. Two grass carps' gut were taken for colonization identification. Two grass carps' gut were stored in 4% paraformaldehyde fix solution and sent for histological analysis. Five grass carps were taken for intestinal bacterial community diversity analysis by using high-throughput sequencing (Wuhan Nextomics Biotechnology, Co., Ltd., China). The rest 26 grass carps from every tank were used for challenge test against A. hydrophila. Liver, spleen, and head kidney from grass carps after 6 h post-infection (6 hpi) or 12 h post-infection (12 hpi) were collected for immune-related genes expression analysis.

#### High-Throughput Sequence Analysis

Polymerase chain reaction amplification of the variable region (V3–V4) of tract bacterial 16S rRNA gene was carried out using primer pair 338F (5<sup>0</sup> to 3<sup>0</sup> : ACTCCTACGGGAGGCAGCAG) and 806R (5<sup>0</sup> to 3<sup>0</sup> : GGACTACHVGGGTWTCTAAT). Amplifications were performed in ABI GeneAmp <sup>R</sup> 9700 (Applied Biosystems, United States) with initial denaturation at 95◦C for 3 min, 28 cycles of 30 s at 95◦C, 30 s at 58◦C, 45 s at 72◦C and final extension at 72◦C for 10 min. PCR products were purified by using AxyPrepDNA PCR Purification Kit (Axygen, United States) and quantified using QuantiFluorTM-ST Fluorometer (Promega, United States). Sequencing was performed by Wuhan Nextomics Biotechnology, Co., Ltd., with Illumina MiSeq system (Illumina, United States).

Paired-end (PE) reads were procured and then assembled according to the overlap relationship (Huang et al., 2016). The operational taxonomic units (OTUs) picking was completed with a minimum pairwise identity of 97% with USEARCH software (version 7.1) (Edgar, 2010) after removing chimeras. Taxonomic analyses of sequence reads were processed using Quantitative Insights Into Microbial Ecology (QIIME) pipeline (Caporaso et al., 2010). The most abundant sequence in each OTUs was selected to perform a taxonomic classification based on the Silva database (Quast et al., 2013) using the RDP classifier (Wang et al., 2007), clustering the sequences at 97% similarity with a 0.7 confidence thresholds. Alpha diversity indices were calculated with Mothur (version v.1.30.1) (Schloss et al., 2011). ACE and Chao1 were estimated to indicate the community richness, as well as Simpson and Shannon indices were reckoned to reveal the community diversity, and Good's coverage represented the sequencing depth. Rarefaction (Amato et al., 2013) and Shannon–Wiener curve (Wang et al., 2012) were constructed with R software. Two dimensional principal coordinates analysis (PcoA) characterized the similarities and differences of the community composition. The similarities and differences in overall bacterial community structure between each of intestine samples were detected using the UniFrac metric, and then phylogenetic tree was constructed using the unweighted pair group method with arithmetic mean (UPGMA).

# Colonization and Histology in the Intestine

Intestinal contents collected from foregut, midgut, and hindgut were washed with sterile saline (0.85%) and analyzed separately. Colonization of SL001 in intestine was measured by plating the various dilutions of intestinal contents on MRS agar plates. The appeared colonies were identified using 16S rRNA gene sequence.

Intestinal tissues collected for both SL001-supplemented group and control group (six fishes for each group) were preserved in 4% paraformaldehyde fix solution, cleared in xylene and dehydrated in ethanol solutions, and then embedded in paraffin wax. Then, tissues were sectioned to 4 µm and stained with hematoxylin and eosin (H&E) according to standard protocol before examined under a light microscope (Nikon Eclipse E100, Japan).

#### Activity of Non-specific Immunological Factors Measurement

The serum samples collected at 30 days post-feeding were taken for determining the acid phosphatase (ACP) and alkaline phosphatase (AKP) activity. The measurement was performed with commercial kits from Nanjing Jiancheng Institute (Nanjing, China) according to the manufacturer's instructions (Yi et al., 2018).

# RNA Isolation and Real-Time Quantitative PCR

Total RNA was isolated from harvested samples following instructions with modifications by using Trizol reagent (Sangon, China). Briefly, tissue (∼50 mg) was homogenized in 1 mL Trizol reagent. The solution was mixed with 0.2 mL chloroform and in an ice bath for 10 min. After centrifugation at 12000 rpm, 4◦C for 10 min, the aqueous phase was transferred to a new 1.5 mL tube and added equal volume of isopropanol with subsequent ice bath for 10 min. The mixture was centrifuged, and then the precipitation was washed twice with 75% ethanol and dissolved in RNase free water. The RNA quality and quantity were assessed using agarose gel (2%) electrophoresis and the ratio of OD<sup>260</sup> nm to OD<sup>280</sup> nm by NanoDrop 2000 spectrophotometer (Thermo, United States) respectively.

Total RNA (1 µg) was treated with DNase and reversed transcription to synthesize cDNA by PrimeScriptTM RT reagent Kit with gDNA Eraser (TaKaRa, Japan) according to manufacturer's instructions. For real-time quantitative PCR (qRT-PCR), the primer pair were designed with Primer Premier 5.0 according to the published sequences and listed in **Table 1**. The amplification was performed with SYBR Permix Ex TagTM


#### TABLE 1 | The sequences of primers used for qRT-PCR.

fmicb-10-01384 June 26, 2019 Time: 15:43 # 4

GC (TaKaRa, Japan) at 7500 Real-Time PCR system instruments (Applied Biosystems, United States). The PCR cycling conditions were as follows: 2 min at 50◦C and 10 min at 95◦C, followed by 40 cycles of 15 s at 95◦C and 1 min at 60◦C. Melting curves were performed from 60 to 95◦C to validate the specificity of PCRs (Palazzotto et al., 2015). The β-actin gene was used as internal control gene (Chu et al., 2018).

#### Growth Performance

To monitor the effect of SL001 on the growth of grass carps, the weight gain rate (WGR), specific growth rate (SGR), feed intake (FI), and feed conversion ratio (FCR) of grass carps were calculated as previously described (Shi et al., 2019).

WGR(%) = 100 × (W<sup>t</sup> − W0)/W<sup>0</sup> SGR(%) = 100 × (lnW<sup>t</sup> − lnW0)/30 days FI = feed consumption/[(W<sup>0</sup> + Wt)/2 × 30 days] FCR = feed consumption/(W<sup>t</sup> − W0)

W<sup>0</sup> and W<sup>t</sup> designates the average weights of grass carps at the start of the experiment and at the termination of the experiment respectively.

#### Challenge Test

As preceding description of Yi et al. (2018). A. hydrophila was inoculated into LB and shaked for 12 h. Overnight culture of A. hydrophila was diluted to 1 × 10<sup>7</sup> CFU/mL with sterile saline. Grass carps were divided into three groups: in experimental group, grass carps were fed with SL001 supplemented diet for 30 days and then intraperitoneally injected with 100 µL of A. hydrophila suspensions (10<sup>6</sup> CFU); in challenge control group, grass carps were fed with basal diet and intraperitoneally injected with 100 µL of A. hydrophila suspensions (10<sup>6</sup> CFU); in negative control group, grass carps were fed with basal diet and intraperitoneally injected with 100 µL sterile saline solution. Protective effect of SL001 against A. hydrophila was evaluated by relative percent survival (RPS) using the following formula: RPS = (1 – mortality in SL001 supplemented group/mortality in challenge control group) × 100%.

#### SL001 Genome Sequencing and Analysis

Genomic DNA of SL001 was purified using magnetic beads (0.5×) and quantified using NanoDrop 2000 spectrophotometer and Qubit 4 Fluorometer (Thermo, United States). Whole genome sequencing was performed by Wuhan Nextomics Biotechnology, Co., Ltd., using the Nanopore sequencing platform. After filtering out invalid data, the assembly was performed with Canu software (v1.3) (Koren et al., 2017) and corrected using Nanopolish 0.8.4 software (Walker et al., 2014). The assembled genome was annotated with RAST sever (Overbeek et al., 2014). Genome was further analyzed with Anti-SMASH for secondary metabolite and bacteriocin biosynthesis gene clusters. The tRNA, rRNA, and CRISPR were predicted using tRNAscan-SE 1.23, RNAmmer 1.2, and CRISPRS II respectively.

#### Statistical Analysis

All data were statistically analyzed using SPSS 21.0 software and presented as the mean ± SE of mean (SEM). Independent-Samples t-test and Least-Significant Difference test were used to calculate significant differences. P-value < 0.05 and < 0.01 were considered statistically significant and extremely significant, respectively.

#### Ethics Statement

fmicb-10-01384 June 26, 2019 Time: 15:43 # 5

All sampled fish were humanely euthanized by bath immersion using an overdose of MS222 (3-Aminobenzoic acid ethyl ester methanesulfonate, Sigma, United States). This study was carried out in accordance with the recommendations of Animal Management Regulations (Directive 1988/2/CN). This study has been reviewed and approved by the ethics committee of the Hunan Normal University.

# RESULTS

#### Isolation and Characterization of a New Isolate of P. pentosaceus

Soil samples from different regions of China were collected and subjected to screening of bacterial colonies which could inhibit the growth of bacterial fish pathogens. Among 100s of isolated bacterial strains, the isolate designated as SL001 which was isolated from soil sample collected from Dadonghai beach (located at the south coast of Sanya, China), exhibited excellent antibacterial activity against several important fish pathogens, including Aeromonas hydrophila, Aeromonas veronii, Aeromonas sobria, Edwardsiella tarda, Lactococcus garvieae, and Plesiomonas shigelloide (**Figure 1A** and **Supplementary Figure S1B**).

SL001 cells, which were spherical (0.9–1.1 µm in diameter), Gram-positive, non-spore-forming and appear in pairs or quadruples, could form opalescent and wet colonies on MRS agar plate (**Figure 1B** and **Supplementary Figure S1A**). Subsequently, SL001 was characterized on the basis of 16S rRNA gene sequence analysis, which demonstrated that SL001 clustered with P. pentosaceus strain P2 (GenBank Accession No. MH045191.1), with 99% similarity (**Figure 1C**). The results of biochemical characterization of SL001 was also consistent with that of previously reported P. pentosaceus strains (**Supplementary Table S2**) (Shukla and Goyal, 2014). Thus, SL001 was identified as a new isolate of P. pentosaceus.

inhibition zones of SL001 against fish pathogens. CFS, cell-free supernatant (pH 3.48); MRS, Man Ragosa Sharpe medium; PL, phosphate-buffered saline added lactic acid (pH 3.48). (C) The phylogenetic tree based on 16S rRNA gene sequences inferred evolutionary relationships of strain SL001 by neighbor-joining method.

#### Biosafety Evaluation of SL001 on Fish

As SL001 showed remarkable potential in controlling fish pathogens, we investigated its biosafety for fish. Cell-free supernatant (CFS) and the cell pellet of SL001 saturated culture was applied to hepatic L8824 cells. After 36 h of incubation, neither CFS nor cell pellet of SL001 exhibited cytotoxicity against L8824 cells. We then evaluated the biosafety of SL001 in vivo by supplementing SL001 (1 × 10<sup>8</sup> CFU/mL) in fish diet. After 30 days of feeding, neither mortality nor visible adverse effects was observed in grass carps (**Figure 2**).

#### Protection of SL001 Against A. hydrophila Infection

After 30 days of feeding with SL001-supplemented diet, grass carps were intraperitoneally challenged with A. hydrophila (1 × 10<sup>6</sup> CFU per fish). The cumulative mortality of SL001-supplemented group after 7 days was 51.7%, which was significantly lower than that of without SL001 supplemented group (90%) (**Figure 3**), with a RPS value of 42.6%. Thus, it indicated that SL001 could protect grass carps from A. hydrophila infection.

# Effect of SL001 on the Growth of Grass Carps

After 30 days of feeding, all the detected growth indices including weight gain rate (WGR, 15.06%), specific growth rate (SGR, 0.47%), feed intake (FI, 9.88) and feed conversion ratio (FCR, 2.35) of SL001-supplemented group were significantly increased when compared to those of control group (11.69, 0.37, 7.03, and 2.13%, respectively) (**Supplementary Table S3**). Histological analysis of intestine of grass carps demonstrated that normal villi, distinct lamina propria, enterocytes, and goblet

cells were visible in both SL001-supplemented group and without SL001 supplemented group (n = 6, **Figure 4**). However, more mucus-secreting goblet cells and longer intestinal villi could be observed in SL001-supplemented group. This observation was more pronounced in the midgut (as the main structure of nutrient absorption), with the increase in length of villi by approximately 53% (P < 0.05, **Supplementary Table S4**), indicating that SL001 promoted nutrients absorption in grass carps. Moreover, the lamina propria in control group was wider compared with that in SL001-supplemented group (**Figure 4**), implying that SL001 could reduce inflammatory response in grass carps. After anaerobic cultivated, SL001 dominated in all the dilutions of intestinal contents in SL001-supplemented group and only SL001 colonies appeared on MRS agar plate in the 10−<sup>8</sup> dilution, which was confirmed by 16S rRNA gene sequencing (**Supplementary Figure S2**).

We subsequently examined the expression of genes involved in muscle growth regulation using qRT-PCR. In SL001-supplemented group, genes involved in restraining muscle growth (myostatin, MSTN-1, and MSTN-2) were significantly down-regulated, whereas genes related to promoting muscle growth (insulin-like growth factor, IGF-1 and IGF-2) were notably up-regulated (**Figure 5**).

# Influence of SL001 on Gut Microbiota of Grass Carps

In order to evaluate the influence of SL001 on gut microbiota structure of grass carps, high-throughput sequencing was conducted to sequence 16S rRNA gene (V3–V4 region) of bacterial community from intestine of grass carps at 30 days post-feeding (dpf). A total of 431493 valid sequences and 441 OTUs were obtained. The rarefaction curves reached a saturation phase at approximately 360 OUTs, which indicated that the sequencing data were reasonable and could reliably describe the full microbial diversity (**Figure 6A**). This conclusion was further verified by the Good's coverage (**Table 2**) and Shannon–Wiener curve (**Supplementary Figure S3A**). The LDA Effect Size (LEf Se) also revealed that the abundance of gut microbiota between SL001-supplemented group and

FIGURE 4 | Photomicrographs of the different intestinal site of grass carp after 30 days of feeding with SL001-supplemented diet (n = 6, data from one fish is presented). Enterocytes (white arrow), goblet cells (red arrow), lamina propria (indicated by yellow lines) are shown in the figure. PA, SL001-supplemented group; DB, without SL001 supplemented group.

without SL001-supplemented group exhibited significant differences. Pediococcus, Lactobacillaceae, Lactobacillales and bacilli were the over-represented taxon in SL001-supplemented group (**Supplementary Figure S3B**). Alpha diversity indices showed a higher microbial diversity and lower community richness in SL001-supplemented group compared to without SL001-supplemented group (**Table 2**). PCoA plots based on weighted UniFrac distances were used to evaluate bacterial community composition, which showed a clear clustering pattern that samples were largely partitioned based on the SL001-feeding (**Figure 6D**). The existence of statistically significant differences in both abundance and community composition between SL001-supplemented group and without SL001-supplemented group (**Figure 6F**) indicated that SL001 could affect, to a great extent, gut microbiota of grass carps.

All sequences were classified into 20 phyla and 285 genera. As a whole, the phyla with relative abundance of above 0.1% were clearly observed in the bar graph (**Figure 6B**).

In general, the dominant phyla (abundance > 10%) in both groups were Firmicutes, Proteobacteria, and Fusobacteria. However, the relative abundance of Firmicutes in SL001-supplemented group was around five times higher than that in without SL001-supplemented group (**Figure 6B**). In genus level, Pediococcus, which was the dominant genus in SL001-supplemented group (46.3% of all reads), could hardly be detected in non-SL001-supplemented group (0.007% of all reads), indicating that SL001 could inhabit the gut of grass carps (P < 0.05, **Figure 6C**). Importantly, two fish pathogens, namely Aeromonas and Vibrio showed a higher abundance in without SL001-supplemented group (17.6 and 15.1%, respectively) than that in SL001-supplemented group (12.9 and 0.3%, respectively) (**Figure 6C** and **Supplementary Figure S3C**). The shared microbiome, discovered in both groups, was identified and comprised 239 OTUs; 38 OTUs were unique in SL001-supplemented group and 164 OTUs were unique in without SL001 supplemented group (**Figure 6E**).

# Expression of Immune-Related Genes After Treatment With SL001

The expression levels of genes involved in fish immunity were determined. After 30 days of feeding with SL001-supplemented diet or basal diet. In the liver, the expression levels of all detected genes but lysozyme (LSZ) gene showed significant differences between SL001-supplemented group and without SL001-supplemented group (**Figure 7A**). The gene expression levels of immunoglobulin M (IgM), complement 3 (C3), and interleukin-1β (IL-1β) in SL001-supplemented group elevated 3-, 2.6-, and 2.5-fold respectively (**Figure 7A**). The expression level of interleukin-8 (IL-8) decreased and maintained 1/3 level of without SL001-supplemented group (**Figure 7A**). In spleen, the gene expression of IgM, C3, and LSZ were significantly upregulated by SL001 after 30 days of feeding, with 1.5-, 1.4-, and 1.5-fold increases respectively (**Figure 7B**). By contrast, the gene expression levels of both genes from the interleukin (IL) family were down-regulated (**Figure 7B**), among which IL-1β gene expression showed a considerable decrease in SL001-supplemented group (**Figure 7B**). The down-regulation of two pro-inflammatory cytokine (IL-1β, IL-8) indicated the reduction of host's inflammatory response, and up-regulated expression of IgM, C3, and LSZ suggested that both specific and non-specific immunity of grass carps were improved in SL001-supplemented group.

We further analyzed the influence of 30 days-feeding with SL001-supplemented diet on the expression of immune genes in grass carps post A. hydrophila infection. After feeding grass carps with SL001-supplemented diet for 30 days, the grass carps were challenged with 1 × 10<sup>6</sup> CFU/mL A. hydrophila, and genes involved in fish immunity were determined. In the liver, the gene expression levels of LSZ, IL-1β, and IL-8 in



the SL001-supplemented group were significantly higher than those of the non-supplement group at 6 hpi (**Figure 8A**). IL-8 gene expression in the SL001-supplemented group at 12 hpi was significantly higher than that of the non-supplement group. However, the gene expression levels of IgM, C3, LSZ, and IL-1β in the SL001-fed group at 12 hpi were prominently down-regulated compared with those of the control group (**Figure 8D**). In the spleen, the C3 gene expression levels in the SL001-supplemented grass carps exhibited notable decrease at 6 hpi, and then increased remarkably at 12 hpi (**Figures 8B,E**). By contrast, the gene expression levels of LSZ, IL-1β, and IL-8 in the SL001-fed grass carps were significantly higher at 6 hpi and became considerably lower at 12 hpi (**Figures 8B,E**). In addition, the expression of IgM in the SL001-supplemented group was evidently down-regulated at 12 hpi (**Figure 8E**). In the head-kidney, the gene expression levels of C3 and IL-1β in the SL001-fed grass carps at 6 and 12 hpi were higher than those of the non-SL001-fed grass carps (**Figures 8C,F**). LSZ gene expression in SL001-fed grass carps was notably upregulated at 6 hpi and then significantly down-regulated at 12 hpi (**Figures 8C,F**). The gene expression levels of IgM in the SL001-fed grass carps were evidently lower at 12 hpi, compared to that of the control grass carps (**Figure 8F**). In the early stage of infection (6 hpi), pro-inflammatory cytokines (IL1β, IL8) and LSZ were significantly up-regulated in the immune organ of grass carp fed with SL001-diets, which was beneficial to the body against pathogens. In the late infected stage (12 hpi), the higher expression level of C3 indicated that could eliminate pathogens, because of SL001 stimulating the body to improve its immunity.

Besides, alkaline phosphatase (AKP) activity in SL001-supplemented groups was significantly increased compared with the control group after 30 dpf (P < 0.05). The activity of ACP was also increased by SL001 supplemented diets albeit not statistically significant (**Figure 9**).

#### Whole Genome of SL001

The whole genome of P. pentosaceus strain SL001 was sequenced using third-generation DNA sequencing, which generated 1,842,476 bp longest contig and 76,699 bp shortest contig, resulting in a genome assembly with total size of 1,917,175 bp (GenBank accession no. CP039378 and CP039379) and G+C content 37.4%. 1,905 coding DNA sequences (CDSs), 72 RNA (including 57 tRNAs and 15 rRNAs), and 22 CRISPRs were obtained by the method of NCBI prokaryotic genome annotation pipeline. Online software RAST was used to annotate the genome, the results of which showed that 41% of the annotated CDS were assorted to the subsystem (**Figure 10A**). Among the CDS, protein metabolism was the most-enriched metabolic category. In addition, a full-genome comparison analysis of P. pentosaceus SL001 with already-sequenced P. pentosaceus strains showed significant differences for carbohydrates and DNA metabolism (in genes 10–59, 118–181, 348–398, 608–633, 813–860, and 1794-1812) (**Figure 10B**).

The SL001 genome includes numerous genes involved in organic acid synthesis [e.g., garR (ORF540), pgk (ORF1671), glxk (ORF1746), alsD (ORF490), and ilvB (ORF491)], as well as superoxide dismutase [SOD, e.g., perR (ORF558) and npx (ORF643, ORF1158, ORF1182)], and digestive enzymes [e.g., pepE (ORF865, ORF1344, ORF1780) and IV86 (ORF1278)]. In addition, two bacteriocin synthesis gene clusters were present in SL001 genome (**Figure 10C**). DNA sequence of cluster 1 (833557 to 842637) shares 40% identity with coagulin biosynthetic gene cluster from Bacillus coagulans plasmid pI4 (**Figure 10D**), while cluster 2 (853776 to 863958) was an unknown bacteriocin that possibility synthesize new antimicrobial compound (**Figure 10C**).

#### DISCUSSION

P. pentosaceus which can antagonize against several important pathogens (Listeria, Enterococcus faecium, Pseudomonas

aeruginosa, and Klebsiella pneumoniae) has been applied in the food industry to control foodborne pathogenic bacteria (Bajpai et al., 2016; Meira et al., 2017; Sriphochanart and Skolpap, 2018), as well as probiotic in animals and humans (Todorov and Dicks, 2009; Bajpai et al., 2016; Cavicchioli et al., 2017). Although P. pentosaceus had a great potential as a probiotic in marine fishes, its probiotics property in freshwater fish has seldom been touched (Xing et al., 2013; Huang et al., 2014). In this study, a new P. pentosaceus strain SL001 was isolated from soil sample of Dadonghai beach in Hainan Province. SL001 exhibited broad-spectrum antibacterial activity against fish pathogens. qRT-PCR analysis demonstrated that feeding with SL001 could not only protect grass carps from A. hydrophila infection, but also promote growth rate. In SL001-supplemented group, more mucus-secreting goblet cells and longer intestinal villi were observed and up-regulated expression of insulin-like growth factor (IGF-1 and IGF-2) and down-regulated expression of myostatin (MSTN-1 and MSTN-2). Relative mRNA expression levels of immune-related genes in liver, spleen and head kidney were analyzed. Genome sequencing of SL001 has been performed to reveal its probiotics properties.

Fish probiotics with good colonization in the gut are an important element in growth and immune regulation (Hai, 2015; Lazado et al., 2015). Here, we demonstrated that SL001 could not only stably colonize in the intestine of grass carps but also prominently affect gut microbiota structure. The composition of gut microbiota was tightly associated with the growth of grass carps, the higher proportion relative abundance of Firmicutes over Bacteroidetes the faster growth of fish (Li et al., 2013).

The relative abundance of Firmicutes in SL001-supplemented group was around five times higher than that in control group, which resulted in the increased proportion relative abundance of Firmicutes over Bacteroidetes. The influence of SL001 on the gut microbiota structure of grass carps could be responsible for the increased growth rate. Intestine goblet cells are responsible for the production of the protective mucus layers by synthesizing and secreting mucins. Goblet cells are polarized mucus-secreting cells, which could serve as a source of several digestive enzymes, including lipase, amylase, and trypsin (O'Neill et al., 2011). Efficient digestion and absorption of nutrients by the intestine require a very large apical surface area, a feature that is enhanced by the presence of villi (Freddo et al., 2016). We inferred that elevated mucus-secreting goblet cells and elongated intestinal villi in the SL001-fed group were also key factors for increased growth rate of grass carps.

Commensal probiotics can not only promoter growth of host but also provide protection from pathogenic bacteria, by creating inhibitory compounds, competing for adhesion sites or modulating immune responses (Perez et al., 2010). It was reported that P. fluorescens strain AH2 was able to inhibit the growth of V. anguillarum and increased the survival rate of infected fish (Gram et al., 1999). Bacillus P64 showed both probiotic and immuno-stimulatory features, which could promote growth and improve immune factors activity of host (Gullian et al., 2004). According to previous reports (Isolauri et al., 2002; Ramesh et al., 2017; Meng et al., 2019), IgM, a earliest immune molecule, was produced in fish-specific immune response. C3, LSZ, IL1β, and IL-8 played an important role in fish infection prevention and inflammatory response, which could quickly counterwork against invading pathogens and participate in non-specific immunity of fish. In our study, up-regulated expression of IgM, C3, and LSZ encoding genes and down-regulated expression of pro-inflammatory cytokines (IL-1β and IL-8) in SL001-supplemented group were detected. We speculated that SL001 which could produce prebiotics (exopolysaccharides, SOD, etc.) caused the downregulation of two pro-inflammatory cytokine genes and activated the host's immune system, so that the significant upregulation of LSZ and C3 (involved in non-specific immunity) and the IgM (involved in specific immunity). Previous studies also showed that probiotic Bacillus aerophilus KADR3 significantly enhanced the levels of serum lysozyme and serum IgM and elevated the activity of the complement pathway, while P. pentosaceus, L. plantarum, and L. brevis reduced the expression levels of IL-1β and IL-8 (Isolauri et al., 2002; Huang et al., 2014; Ramesh et al., 2017). In addition, ACP localized within lysosomes, is a marker for determining whether the macrophages are activated (Song et al., 2006). AKP is an extracellular enzyme, which could hydrolyzes phosphate group within various organic compounds like proteins, nucleic acid (El-Ebiarie, 2012). ACP and AKP as important markers were usually used to evaluate the function of probiotic bacteria in aquaculture. Yi reported that the levels of ACP and AKP of Carassius auratus fed with B. velezensis JW supplemented

diets were higher compared to the control group (Yi et al., 2018), which was verified in our study. Besides, the lamina propria in control group was wider than that in SL001-supplemented group. Vasanth et al. (2015) reported that the width of lamina propria could indirectly reflect the host's health, in which a smaller width indicated no onset of inflammation and reflected healthier host. After challenged against A. hydrophila, LSZ was significantly up-regulated and then down-regulated in all three immune organs, which are consistent with Zhang's results (Zhang et al., 2018). IL-1β and IL-8 are an immune-associated cytokines that play important roles in innate immune responses (Harada et al., 1994). Feeding grass carps with SL001-supplemenged diet resulted in a significant increase in expression of IL1β and IL8 at 6 h after infection.

By analyzing the genome of SL001, we obtained some clues about its probiotics properties. Genes related to organic acids, bacteriocins, SOD, and digestive enzymes which were thought to be linked to disease-resistance, immune response, and growth were identified in the genome of SL001. It was well-known that organic acid and bacteriocin could inhibit the growth of pathogenic bacteria and regulated the micro-ecological balance of the gastrointestinal and reproductive tract of animals (Muhialdin et al., 2018). Previous report demonstrated that the SOD enhanced humoral and cellular immunity in animals and improved disease resistance (Picchietti et al., 2009). In addition, the digestive enzymes from SL001 secretion promoted the digestion and absorption of nutrients by fish, thereby promoting fish growth (Suzer et al., 2008). Besides, two putative gene clusters responsible for bacteriocins synthesis were present in SL001 genome. One gene cluster was predicated to produce coagulin, which was reported to exhibite antimicrobial activity against pathogenic bacteria (Hyronimus et al., 1998). Coagulin, a new member of the pediocin-like family protein, was firstly discovered in B. coagulans. Afterward, coagulin was also identified in P. pentosaceus MZF16 (Zommiti et al., 2018). The difference between coagulin and pediocin was only a single amino acid residue at their C terminus (Le Marrec et al., 2000).

#### CONCLUSION

Our work showed that feeding with SL001-supplemented diet can promote growth on grass carps and enhance its immune. We envisioned that this can be expanded to other freshwater fish aquaculture. It would undoubtedly provide an important strain for freshwater fish aquaculture and broaden the application

#### REFERENCES


of P. pentosaceus. Encapsulation of P. pentosaceus Li05 in an alginate-gelatin microgels significantly enhanced their viability under different conditions (Yao et al., 2018). Formulated and encapsulated probiotics can improve their survival and colonization and avoid loss of probiotics owing to the reduced probiotic activity during food storage and gastrointestinal transit during feeding. We will then focus on the microecological preparation P. pentosaceus for its practical application in future.

# DATA AVAILABILITY

This manuscript contains previously unpublished data. The name of the repository and accession number are not available.

#### AUTHOR CONTRIBUTIONS

LG and LP performed the bacterial strain isolation and identification. LC, HH, and LY performed the analysis of antimicrobial activity and determination of non-specific immunological factors activity. DL, LC, HH, and XD performed the cellular culture, challenge test, and weight analysis. YL and YS performed the collection of gut, liver, spleen, kidney, and muscle. GY performed the H&E staining. LG, HH, DL, and SH performed the Total RNA isolation and qRT-PCR analysis. LG, YZ, and LX analyzed the data of high-throughput sequence and genome. LG, HH, SH, and LX designed the study and wrote the draft of manuscript. TK, SH, and LX corrected the manuscript. All authors discussed the results and approved the final manuscript.

#### FUNDING

This present work was supported by the National Natural Science Foundation of China (Grant Nos. 31570125 and 31770106), the National Basic Research Program of China ("973" program; Grant No. 2012CB722301), and the "Hunan Province Biological Development Engineering and New Product Development Collaborative Innovation Center" project (Grant No. 20134486).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.01384/full#supplementary-material


pentosaceus 4I1 isolated from freshwater fish zacco koreanus. Front. Microbiol. 7:2037. doi: 10.3389/fmicb.2016.02037



enterotoxigenic Escherichia coli and mediates host defense. Front. Microbiol. 9:1364. doi: 10.3389/fmicb.2018.01364


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Gong, He, Li, Cao, Khan, Li, Pan, Yan, Ding, Sun, Zhang, Yi, Hu and Xia. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Probiotic Properties of Lactic Acid Bacteria Isolated From Neera: A Naturally Fermenting Coconut Palm Nectar

Rakesh Somashekaraiah<sup>1</sup> , B. Shruthi<sup>2</sup> , B. V. Deepthi<sup>1</sup> and M. Y. Sreenivasa<sup>1</sup> \*

<sup>1</sup> Department of Studies in Microbiology, University of Mysore, Mysuru, India, <sup>2</sup> Department of Biotechnology, Sahyadri Science College, Kuvempu University, Shimoga, India

Probiotic bacteria were isolated from different traditional fermented foods as there are several such foods that are not well explored for their probiotic activities. Hence, the present study was conducted to find the potential of lactic acid bacteria (LAB) as probiotics that were isolated from the sap extract of the coconut palm inflorescence – Neera, which is a naturally fermented drink consumed in various regions of India. A total of 75 isolates were selected from the Neera samples collected aseptically in the early morning (before sunrise). These isolates were initially screened for cultural, microscopic, and biochemical characteristics. The initial screening yielded 40 Grampositive, catalase-negative isolates that were further subjected to acid – bile tolerance with resistance to phenol. Among 40 isolates, 16 survived screening using analysis of cell surface hydrophobicity, auto aggregation with adhesion to epithelial cells, and gastric–pancreatic digestion for gastrointestinal colonization. The isolates were also assessed for antimicrobial, antibiotic sensitivity, and anti-oxidative potential. The safety of these isolates was evaluated by their hemolytic and deoxyribonuclease (DNase) activities. Based on these results, seven isolates with the best probiotic attributes were selected and presented in this study. These LAB isolates, with 51.91–70.34% survival at low pH, proved their resistance to gastric conditions. The cell surface hydrophobicity of 50.32–77.8% and auto aggregation of 51.02–78.95% represented the adhesion properties of these isolates. All the seven isolates exhibited good antibacterial and antifungal activity, showing hydroxyl-scavenging activity of 32.86–77.87%. The results proved that LAB isolated from Neera exhibited promising probiotic properties and seem favorable for use in functional fermented foods as preservatives.

Keywords: lactic acid bacteria, Neera, probiotics, bio-preservatives, gastrointestinal tract

# INTRODUCTION

The lactic acid bacteria (LAB) are isolated from various food matrices and those isolates with better performances and high competitiveness are used as probiotics (Bromley and Uchman, 2003). The probiotics are live organisms that confer health benefits on the host when consumed in adequate amounts, by bringing the microbial balance in the system. The use of probiotics in animal

#### Edited by:

Yuheng Luo, Sichuan Agricultural University, China

#### Reviewed by:

Atte Von Wright, University of Eastern Finland, Finland Graciela Liliana Garrote, National University of La Plata, Argentina

#### \*Correspondence:

M. Y. Sreenivasa sreenivasamy@gmail.com; mys@microbiology.uni-mysore.ac.in

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 15 January 2019 Accepted: 03 June 2019 Published: 28 June 2019

#### Citation:

Somashekaraiah R, Shruthi B, Deepthi BV and Sreenivasa MY (2019) Probiotic Properties of Lactic Acid Bacteria Isolated From Neera: A Naturally Fermenting Coconut Palm Nectar. Front. Microbiol. 10:1382. doi: 10.3389/fmicb.2019.01382

infections, especially in the gastrointestinal and vaginal tract, has been extensively studied (Pineiro and Stanton, 2007). They are proficient in inhibiting the growth of pathogenic organisms through different mechanisms such as adherence to epithelial cells, modulation of the immune system, and secretion of antimicrobial compounds. This proves their ability in the biopreservation of food and can be used as starter culture in the fermentation process under controlled conditions. Therefore, the isolation and characterization of LAB from different traditional fermented foods and products have gained research interest in recent years (Alonso et al., 2018).

Neera is a sap extract, a naturally fermented drink collected by tapping the spadix of coconut palm. This coconut palm inflorescence sap extract is consumed before sunrise in India, Sri Lanka, Malaysia, Indonesia, Myanmar, and Thailand as a sweet juice. Neera with a pH of 6.5–7.0 is rich in carbohydrates and highly nutritional and helps in digestion (Borse et al., 2006). It is naturally fermented by different microorganisms among which LAB are one of the protagonists. Neera also possesses medicinal properties because of which it is used as syrup in Indian system of Ayurveda preparations. These factors prove that LAB isolates from Neera have potential probiotic properties (DebMandal and Mandal, 2011).

Neera obtained from the spadix of coconut palm undergoes natural fermentation due to the innate presence of microorganisms like yeasts and bacteria, resulting in the production of ethyl alcohol. The amount of alcohol production or the fermentation time depends on the storage time after Neera collection, and the process of fermentation gets enhanced under sunlight (Xia, 2011). The fresh Neera collected under hygienic conditions before sunrise and transferred at low temperatures is used for analysis. The composition of Neera is influenced by the place, time, and duration of tapping. It mainly contains total sugars, reducing sugar, ethanol, volatile acids, amino acids, vitamins, and phenolic compounds (Borse et al., 2006; Chinnamma et al., 2019).

A potent probiotic isolate must possess certain characteristics like survival and colonizing ability under different environmental conditions (Palachum et al., 2018). The isolates should be able to withstand low pH of gastric juice with resistance to bile salts and also adhere to epithelial cells. They should also offer certain health benefits like antimicrobial activity, anticancer activity, toxinreducing effects, and boosting immune response. Hence, bacteria adhering to suitable surfaces and survival in the gastrointestinal tract should be confirmed by in vitro evaluation prior to using them as probiotics (Chiang and Pan, 2012; Berardi et al., 2013). The health benefits of LAB as probiotics such as lowering the risk of diseases, regulation of allergic response, and improving inhabitants of gastrointestinal tract with better immune response have been reported. The contribution of LAB in inhibiting the growth of pathogenic organisms, reducing their toxin secretions, and increasing the nutritional value and other functionality has already been studied (Bartkiene et al., 2018).

The LAB strains Enterococcus spp., Lactococcus spp., and Lactobacillus spp. that have antimicrobial properties are used in bio-control strategies such as reducing mycotoxins and enhancing bioavailability (Campagnollo et al., 2016; Deepthi et al., 2017). Recent studies have revealed that probiotic LAB strains can be used to remove mycotoxins such as aflatoxins, trichothecenes, and fumonisins from different food products during pre-harvest, production, and storage (Deepthi et al., 2016; Poornachandra Rao et al., 2017). These LAB strains, isolated from traditional fermented foods, can be used in the formulation of fermented foods with functional characteristics to manage the growth of adverse pathogenic microorganisms; this would help in the prevention and/or treatment of diseases in consumers (Batista et al., 2017).

The main objective of the current work was to report on the isolation of potential probiotic LAB from hitherto unexplored, naturally fermenting product – Neera. The isolated LAB strains were characterized by in vitro tests for their probiotic properties (antimicrobial strains surviving gastrointestinal conditions) in order for them to be used in the preservation and fermentation of food.

#### MATERIALS AND METHODS

#### Isolation of LAB

The isolation of LAB strains from the Neera (naturally fermented nectar of coconut palm) samples was based on the method described by Poornachandra Rao et al. (2015). Briefly, fresh Neera samples were collected under hygienic conditions before sunrise from different regions of Mysuru (Karnataka, India) and transferred at low temperatures for analysis; about 1 ml of each sample was enriched in a de Man, Rogosa, and Sharpe (MRS) broth and incubated under anaerobic conditions at 37◦C for 24 h. The enriched broth samples were diluted using phosphate buffer saline (PBS), plated on an MRS agar medium, and then incubated under anaerobic conditions at 37◦C for 24 h. The morphologically discrete colonies were further sub-cultured onto MRS agar plates. The viable cultures were stored in MRS slants at 4◦C. The stock cultures were maintained at −20◦C in glycerol stock for further analysis.

#### Primary Characterization of LAB Strains

Preliminary identification of the 40 LAB isolates was based on their phenotypic and biochemical characteristics that included Gram's reaction, catalase test, nitrate reduction assay, citrate utilization assay, bile salt hydrolase activity, osmotic stress (sodium chloride: 3, 5, and 7%) resistance, and sugar fermentation ability (of isolates assimilating different sugar supplements) (Boone et al., 2001; Ni et al., 2015). The growth of the isolates at different temperatures and pH levels was also tested. The cell viability of the isolates was assessed by the plate count method and the results were presented as log colonyforming units (CFU) per milliliter (Ni et al., 2015).

#### Evaluation of Probiotic Properties Tolerance to Acids and Bile Salts

The tolerance of the LAB isolates to both acidic pH value and bile salts was studied using the methodology described by Guo et al. (2009) and Ramos et al. (2013). Overnight cultures of the strains were inoculated in MRS broth, initially adjusted to pH

value 2.0 using 1 N hydrochloric acid (HCl) and MRS medium supplemented with 0.3% oxgall. The MRS broth adjusted to an initial pH of 6.5 was considered as the control for acidic pH value and the one without oxgall was considered as the control for bile salt condition. The samples were incubated anaerobically at 37◦C for time intervals 0, 2, and 4 h and retrieved for enumeration at respective end points. The biomass (CFU/ml) of each culture obtained in the assays, made in triplicate, was enumerated on MRS agar incubated anaerobically at 37◦C for 24 h. The survival rate (%) was calculated using the following formula: Survival rate (%) = Biomass at time (t)/Biomass at initial time (0) × 100.

#### Assessment of Antibacterial Activity

The antibacterial activity of the LAB isolates was determined, using the microplate assay, against Escherichia coli (ATCC 25922), Pseudomonas aeruginosa (ATCC 15422), Salmonella typhi (ATCC 27870), and Staphylococcus aureus (ATCC 6538). The overnight cultures of LAB isolates were centrifuged (at 8,000 rpm for 10 min at 4◦C), and the supernatant was filtered through a syringe filter (of 0.2-mm pore size). The prepared cell-free supernatant (CFS) was divided into two parts, one with their initial acidic pH. The rest of the CFS samples (nCFS) was neutralized to pH 6.5 using 5 M NaOH in order to eliminate the presumed effect of organic acids and both the samples stored at −20◦C for further analysis.

A sterile 96-well plate was filled with 50 µl of CFS/nCFS and 50 µl of bacterial suspension to obtain ∼10<sup>8</sup> CFU per well, which was made up to 200 µl using a Luria–Bertani (LB) broth. The LB broth with bacterial suspension was considered as the positive control and the LB broth alone was considered as the negative control. The plates were incubated at 37◦C for 24 h, and the optical density (OD) at 600 nm was measured. The total percent inhibition of bacterial growth was calculated using the following formula: [(OD of test sample - OD of control)/OD of control] × 100 (Georgieva et al., 2015).

#### Assessment of Antifungal Activity

The antifungal activity of the LAB isolates was determined, by the agar overlay method, against Fusarium graminearum (MTCC 1893), Aspergillus flavus (MTCC 2799), and Fusarium oxysporum (MTCC 1755). The MRS agar plates were used for the assay, on which LAB isolates were streaked at two different equidistant spots and incubated anaerobically at 37◦C for 24 h. Then, 20 µl of spore suspension (∼10<sup>6</sup> spores/ml) of each fungal pathogen was evenly mixed with 0.7% soft potato dextrose agar (PDA) and overlaid on the LAB-spotted MRS agar plates. The plates, after incubating aerobically at 28 ± 2 ◦C for 4 days, were examined for clear inhibitory zones around the spot area of the LAB colonies and are tabulated (Poornachandra Rao et al., 2017).

#### Molecular Identification of LAB Strains

The molecular identification of efficient LAB isolates was determined by 16S rDNA sequencing. On the basis of their potential probiotic attributes, the isolates were subjected to DNA (deoxyribonucleic acid) isolation and PCR (polymerase chain reaction) amplification using the universal primers 27F-5 <sup>0</sup>AGAGTTTGATCCTGGCTCAG3<sup>0</sup> and 1492R-50GGTTACCT TGTTACGACTT3<sup>0</sup> . The reaction was carried out in 25 µl of the reaction mixture containing dNTP (0.2 mM), 0.5 µl of DNA template (50–100 ng), forward and reverse primers (10 pmol), 1 × PCR buffer with MgCl<sup>2</sup> (magnesium chloride), and Taq polymerase (0.5 U). The optimum conditions for PCR involved an initial denaturation step for 5 min at 95◦C followed by 35 cycles of denaturation for 1 min at 95◦C, annealing for 1 min at 55◦C, extension for 5 min at 72◦C, and final extension for 7 min at 72◦C. The PCR products were confirmed on agarose gel (1%) electrophoresis. Five microliters of the PCR product was loaded with 3 µl of loading dye. The PCR products were sequenced; the sequences obtained were compared using BLAST (basic local alignment search tool) and submitted to the GenBank sequence database for accession numbers (Boubezari et al., 2018).

#### Antibiotic Sensitivity

The antibiotic susceptibility of the LAB isolates was assessed on MRS agar plates using the antibiotic disc diffusion method using the range of antibiotics suggested as per EFSA guidelines. The MRS agar medium was poured and allowed to solidify at room temperature. The overnight LAB cultures (100 µl) were spread on MRS agar plates and allowed to dry. The antibiotic discs were placed on the inoculated plates and incubated at 37◦C for 48 h. The antibiotic susceptibility pattern of the isolates was assessed using ampicillin (10 µg/disc), vancomycin (30 µg/disc), gentamicin (10 µg/disc), kanamycin (30 µg/disc), streptomycin (10 µg/disc), chloramphenicol (30 µg/disc), erythromycin (15 µg/disc), clindamycin (2 µg/disc), and tetracycline (30 µg/disc) (Singh et al., 2012). The diameter of the zone of inhibition was measured using the antibiotic zone scale (CLSI scale). The results obtained are presented in terms of susceptibility, moderate susceptibility, or resistance. These results were compared with the interpretative zone diameters as described in Performance Standards for Antimicrobial Disc Susceptibility Tests (CLSI, 2016).

#### Resistance to Phenol

The resistance of the LAB isolates to phenol was determined as described by Jena et al. (2013) with slight modifications. The MRS broth supplemented with 0.4 and 0.6% v/v phenol was inoculated with overnight-grown LAB cultures. After incubation at 37◦C for 24 h, the cultures were serially diluted and spread on MRS agar plates. The cell viability (log CFU/ml) was calculated by the plate count method.

# Auto Aggregation

The ability of the LAB isolates to auto aggregate was tested as per Zommiti et al. (2017). The overnight culture was harvested by centrifugation (at 8,000 rpm at 4◦C for 10 min) and washed with PBS twice and resuspended in PBS buffer. The sample was allowed to stand awhile, incubating anaerobically at 37◦C, and the upper suspension was checked for absorbance at 600 nm at time intervals of 0, 1, 2, 3, 4, and 5 h. The auto aggregation was measured (in percentage) using the formula auto aggregation % = [1 − (Atime/A0) × 100], where, Atime represents the absorbance at a particular time and A<sup>0</sup> represents the absorbance at time 0.

#### Cell Surface Hydrophobicity

The in vitro bacterial cell surface hydrophobicity of LAB isolates was evaluated by measuring the microbial cell adhesion to hydrocarbons according to the method described by Rokana et al. (2018). The overnight cultures in MRS broth were harvested by centrifugation (at 8,000 rpm at 4◦C for 10 min), washed twice with PBS, and resuspended in PBS buffer followed by absorbance (A0) measurement at 600 nm. A cell suspension of about 3 ml was blended with 1 ml of hydrocarbon (xylene) and incubated at 37◦C without shaking for 1 h for separation of the aqueous and organic phases. The aqueous phase (1 ml) was removed carefully and the absorbance (A1) was measured at 600 nm. The percent hydrophobicity was measured by a decrease in absorbance and calculated using the following formula: % cell surface hydrophobicity = (1 − A1/A0) × 100.

# In vitro Adhesion to Chicken Crop Epithelial Cells

The in vitro potential of the LAB isolates to adhere to epithelial cells was determined using chicken crop epithelial cells that were processed as described by Jakava-Viljanen and Palva (2007). The chicken crop was maintained in PBS at 4◦C for 30 min and washed thrice with potassium phosphate buffer (pH 7.4) to remove the surface mucus. The chicken crop tissues were gently scraped using a sterile cover slip to obtain the epithelial cells and then suspended in PBS. These epithelial cells were washed twice gently using PBS by pipetting and examined under a microscope to ensure the elimination of adhering commensal bacteria. The cells were then diluted to approximately 5 × 10<sup>6</sup> cells/ml. About 100 µl of LAB isolates (10<sup>6</sup> CFU/ml) in 400 µl of epithelial cells was mixed well followed by incubation at 37◦C for 30 min in a water bath. After incubation, the mixture was centrifuged at 3000 rpm for 3 min and then the pellet was washed twice with sterile PBS to remove non-adherent bacteria. It was then resuspended in 100 µl of PBS, stained with crystal violet, and observed under a microscope. The bacterial adhesion was examined in 10 microscopic fields and scored positive if a minimum of 10 bacteria were found adhering to each epithelial cell (Feng et al., 2018).

#### Survival in Simulated Gastric and Pancreatic Digestion

The survival of LAB isolates under gastric and pancreatic conditions was determined in vitro by simulation. The simulated gastric juice (SGJ) and simulated pancreatic juice (SPJ) were prepared as per Lo Curto et al. (2011) with slight modifications. Then, the SGJ was mixed with pepsin (0.0133 g/L) and lysozyme (0.01 g/L) prior to use. The overnight culture was harvested by centrifugation (at 8,000 rpm at 4◦C for 10 min), washed with PBS twice, and suspended with SGJ to a final absorption OD of 1.2 at 600 nm. The samples were then incubated at 37◦C for 3 h in an orbital shaker at ∼200 rpm to simulate peristaltic movement. The 0- and 3-h samples were collected, serially diluted, plated on to LB agar plates for microbial counting, and incubated at 37◦C for 24 h. The survival percentage in gastric juice was calculated using the following formula: g % = Tg3/Tg0 × 100, where Tg0 is the bacterial count at 0 h and Tg3 is the bacterial count at 3 h.

The LAB, incubated in SGJ for 3 h, were harvested by centrifugation (at 8,000 rpm at 4◦C for 10 min), washed with PBS, and resuspended in the same volume of SPJ. The sample was plated immediately for bacterial count at time 0 h (Tp0) and then incubated at 37◦C for 24 h with continuous shaking at 200 rpm. After 24 h of incubation, the sample was serially diluted and plated on LB agar plates to determine the pancreatic digestion-survived bacterial count (Tp<sup>∗</sup> ). The survival percentage in pancreatic juice was determined using the following formula: p % = Tp<sup>∗</sup> /Tp0 × 100. The overall digestion-survival ability of the LAB isolates was estimated by S % = Tp<sup>∗</sup> /Tg0 × 100.

#### Safety Evaluation of LAB Strains Hemolytic Activity

The hemolytic activity of the LAB isolates was determined using the procedure described by Yadav et al. (2016). All the isolates tested were streaked onto blood agar plates containing 5% (w/v) sheep blood and incubated at 37◦C for 48 h. After incubation, the plates were examined for β-hemolysis, α-hemolysis, and nonhemolytic activities.

#### DNase Activity

The LAB isolates were streaked onto a deoxyribonuclease (DNase) agar medium to test for production of the DNase enzyme. The plates were then incubated at 37◦C for 48 h and observed for the zone of DNase activity. A clear pinkish zone around the colonies was considered as positive DNase activity (Shuhadha et al., 2017).

#### Antioxidant Assay

The antioxidant assay was conducted as described by Wang et al. (2009). About 1.0 mL of 1,10-phenanthroline (0.75 mmol−<sup>1</sup> ), 2.0 ml of sodium phosphate buffer (pH 7.4), and 1.0 ml of ferrous sulfate (FeSO4) (0.75 mmol−<sup>1</sup> ) were mixed thoroughly. To the mixture, 1.0 ml of 10<sup>9</sup> CFU/ml culture was added; 1.0 ml of hydrogen peroxide (H2O2) (0.01% v/v) was added to initiate the reaction and the mixture was incubated at 37◦C for 90 min. After incubation, centrifugation (at 9,000 rpm for 10 min at 4 ◦C) was carried out and the absorbance of the supernatant was measured at 536 nm. The percentage of hydroxyl radical scavenging was calculated using the following formula: hydroxyl radical scavenging activity (%) = (A<sup>s</sup> − Ac)/(A<sup>B</sup> − Ac) × 100, where A<sup>s</sup> is the absorbance of the test sample; A<sup>c</sup> is the absorbance of the control including 1,10-phenanthroline, FeSO4, and H2O2; and A<sup>B</sup> is absorbance of the blank including 1,10-phenanthroline and FeSO4.

#### RESULTS

#### Preliminary Characterization of LAB Isolates

A total of 75 bacterial cultures were isolated and initially subjected to physiological and biochemical tests. Out of this, 40 isolates were Gram-positive, non-spore-forming, and



+, positive; −, negative; homo, homofermentative; hetero, heterofermentative.

catalase-negative and were considered for testing as presumptive LAB isolates. The biochemical characterization revealed that the bacterial isolates were able to ferment all the tested sugars. About 16 LAB isolates showed optimum growth and also sustained osmotic stress at different NaCl concentrations. The results are presented in **Table 1**. The growth of these LAB isolates at different temperatures and salt conditions was also tested (**Table 1**). Out of 16 isolates, the data for seven isolates (i.e., MYSN 10, MYSN 106, MYSN 43, MYSN 109, MYSN 98, MYSN 18, and MYSN 28) proving their in vitro potential as probiotics with good antimicrobial activities are further presented.

#### In vitro Testing for Probiotic Potential of LAB Isolates

#### Acid and Bile Tolerances

The acid tolerance helps in studying the survival of strains under low pH gastric juice conditions. This ability of the strains to survive the acidic pH value after 2 and 4 h of incubation at 37◦C is presented in **Figure 1A**. The bile salt tolerance helps in the in vitro evaluation of metabolic activity and colonization of isolates in the small intestine. This survival of the strains in bile salts after 2 and 4 h of incubation at 37◦C is presented in **Figure 1B**. The LAB strains were found to have a survival rate (%) above 50% at low pH and 0.3% bile salt concentration after 4-h exposure. Among all the isolates tested, the isolate MYSN 106 with a survival rate of 80.84% at low pH and MYSN 98 with 83.17% survival rate in bile proved to be the best.

#### Antagonistic Activity of LAB Isolates

The antagonistic activity of the LAB isolates against enteric bacterial pathogens was tested. The isolates proved to have significant antibacterial activity against all the enteric pathogens. The seven selected isolates (i.e., MYSN 10, MYSN 106, MYSN 43, MYSN 109, MYSN 98, MYSN 18, and MYSN 28) showed an inhibitory effect toward the tested pathogens (**Figure 2**). The CFS of the isolate MYSN 106 showed 81.84, 80.81, and 82.14% inhibition of E. coli, P. aeruginosa, and S. aureus, respectively, and MYSN 28 inhibited about 77.10% of S. typhi in comparison to the other isolates tested. The activity of the CFS after neutralization to pH 6.5 (nCFS) showed minimal activity against all the pathogens tested, proving the role of organic acids for their antimicrobial activity.

The antifungal activity of the isolates was tested against F. graminearum, A. flavus, and F. oxysporum. Among the LAB isolates tested for 7 days, MYSN 106 showed the highest zone of inhibition: 21.25 ± 0.45 and 10.45 ± 0.70 mm against F. graminearum and A. flavus, respectively. The highest inhibition against F. oxysporum was shown by MYSN 28 with a 24.95 ± 0.35 mm inhibition zone. The MYSN 109 and MYSN 18 isolates showed no inhibition after 7 days against F. graminearum and F. oxysporum, respectively. The isolates MYSN 10, MYSN 109, and MYSN 18 failed to inhibit A. flavus. The other isolates showed varied antifungal activities as reported in **Table 2**.

#### Molecular Identification by 16S rDNA Sequencing and Phylogenetic Analysis

The potential 7 from 40 isolates, showing eminent probiotic properties with highest antimicrobial activity, were identified by 16S rDNA sequencing and phylogenetic analysis as reported in **Figure 3**. The isolate MYSN 106, identified as Lactobacillus brevis with accession no. MH748630, proved to have excellent probiotic properties. The other LAB isolates were identified as Enterococcus durans (two strains, i.e., MH748609-MYSN 10 and MH748633-MYSN 109), Leuconostoc lactis (MH748629- MYSN 98), Enterococcus lactis (two strains, i.e., MH748625- MYSN 43 and MH748621-MYSN 28), and Enterococcus faecium (MH748610-MYSN 18).

#### Antibiotic Susceptibility Test

The seven selected LAB isolates were tested for their antibiotic susceptibility against different antibiotics procured from Hi Media, India. For appropriate selection of functional strains, two groups of antibiotics are generally recommended in EFSA guidelines such as inhibitors of cell wall synthesis (ampicillin and vancomycin) and inhibitors of protein synthesis (chloramphenicol, gentamycin, clindamycin, erythromycin, streptomycin, kanamycin, and tetracycline). The results obtained

were compared with the zone size interpretative chart provided in the catalog. In this study, the tested isolates were susceptible toward tetracycline and streptomycin. For most of the strains, chloramphenicol, vancomycin, and streptomycin were effective inhibitors. A variable antibiotic sensitivity was observed in all the isolates and is reported accordingly in **Table 3**.

#### Resistance to Phenol

The viable count of the LAB isolates was obtained on plating the MRS broth supplemented with phenol (0.4 and 0.6%) after 24 h of incubation. The phenol concentrations had a slight inhibitory effect in comparison to the MRS control without phenol, having a viable count more than 7.75 log CFU/ml. A viable count

TABLE 2 | Antifungal activity of LAB strains isolated from Neera samples against mycotoxigenic fungi.


ZOI, zone of inhibition; −, no effect detected; +, diameter of ZOI between 1 and 5 mm; ++, diameter of ZOI between 5 and 10 mm; +++, diameter of ZOI between 10 and 25 mm.

ranging from 7.75 to 9.28 log CFU/ml was observed with 0.6% phenol and 8.07 to 9.43 log CFU/ml with 0.4% phenol, while the viable count range was 9.23–10.26 log CFU/ml without phenol. The isolate MYSN 43 was the most tolerant to phenol with 9.43 and 9.28 log CFU/ml viable counts at 0.4 and 0.6% phenol, respectively.

#### Cell Surface Properties of LAB Isolates and Their Adhesion Ability to Chicken Epithelial Cells

The isolates were tested for their cell surface hydrophobicity to estimate their adhesion ability, using the hydrocarbon xylene. The isolates showed different hydrophobicities: MYSN 106 showed the highest hydrophobicity at 77.82% followed by MYSN 98 and MYSN 28 showing hydrophobicities of 71.59 and 66.35%, respectively. Further, the remaining isolates showed a varied degree of hydrophobicity with MYSN 43 showing the least hydrophobicity of 51.10%. The isolate MYSN 106 showed the highest auto aggregation at 78.95% in comparison to the other LAB isolates tested. Further, all the other isolates showed an auto aggregation between 50.29 and 69.28%.

When tested for adhesion to chicken epithelial cells, MYSN 106 showed the highest adhesion ability with 50–100 bacterial cells per epithelial cell. The MYSN 109 isolate exhibited least adhesion with 10–15 cells attaching to an epithelial cell. The varying adhesion properties observed in the other isolates are displayed in **Figure 4**.

FIGURE 3 | Phylogenetic tree showing the relative positions of LAB isolates (MYSN 10, MYSN 43, MYSN 98, MYSN 28, MYSN 18, MYSN 106, and MYSN 109) from Neera samples in comparison to reference strains as referred by the maximum parsimony analysis of 16S rDNA conducted in Mega X software. The percentage of replicate trees in which the associated taxa are clustered together in the bootstrap test of 500 replicates is shown next to the branches.



R, resistant; S, sensitive; MS, moderately sensitive; n.r., not recommended as per EFSA guidelines. The breakpoints for the antibiotic sensitivity/resistant in mm zone of inhibition: Amp and Van (≥17/≤14); Kan and Chl (≥18/≤12); Cli and Tet (≥19/≤14); Ste and Gen (≥15/≤12); and Ery (≥23/≤13).

cells.

fmicb-10-01382 June 27, 2019 Time: 15:14 # 8

#### Survival Under Gastric and Pancreatic Digestion

The isolates were tested for their colonization in the gastrointestinal tract by evaluation of their survival in simulated gastric and pancreatic digestion environments. All the isolates examined survived in both gastric and pancreatic digestion, which helps in colonizing the intestines. The viable cell count of the isolates showed that there was a minimal decrease in the viability of a few isolates after a 3-h incubation period. The MYSN 98 isolate showed the maximum viability at 3.35 × 10<sup>6</sup> CFU/ml. The other isolates showed intermittent survival abilities with the viable cell count ranging from 1.45 to 2.7 × 10<sup>6</sup> CFU/ml.

#### Safety Evaluation of Isolates

The safety evaluation of the isolates was determined by their hemolytic and DNAse activities, which proves the nonpathogenic status of the probiotic isolates. The results revealed no hemolytic or DNAse activities, which was confirmed by the "no zone" in the test plates inoculated with all the isolates studied.

#### Hydroxyl Radical-Scavenging Activity of Probiotic Strains

The MYSN 106 isolate showed the highest hydroxyl radicalscavenging activity at 77.87% followed by MYSN 43 and MYSN 28 with 71.51 and 68.60% hydroxyl-scavenging, respectively. The hydroxyl radical-scavenging of other isolates ranged between 28.36 and 50.47% as reported in **Figure 5**.

#### DISCUSSION

In this study, the importance of the selection of probiotic bacteria from traditional fermented food that can survive

the human gastrointestinal tract and confer health benefits is emphasized. To isolate functional probiotic bacteria, samples of a natural fermented drink – Neera – were collected and different morphotypes were identified with their probiotic characteristics (Eckburg et al., 2005). Out of 40 bacterial strains from different Neera samples, 16 isolates survived preliminary screening for LAB strains. The presumptive LAB isolates were further characterized for acid and bile tolerances to check their viability under gastrointestinal pH conditions (Gu et al., 2008). This resulted in a display of distinct tolerance to acid and bile conditions without any significant loss in cell count with good probiotic properties. A tolerance to phenol was observed in the LAB isolates; this confers their natural selection in non-debittered Neera, which is naturally fermented with different microbial and chemical processes. This phenol tolerance is important for isolates to survive the gastrointestinal conditions, where the gut bacteria have the ability to deaminate aromatic amino acids that are derived from dietary proteins and may lead to formation of phenols (Yadav et al., 2016; Divisekera et al., 2019; Singhal et al., 2019). There are many instances of phenol tolerance reported in LAB that were isolated from natural fermented food sources (Ghabbour et al., 2011). The results prove that the isolates evaluated in the present study can survive human gastrointestinal conditions.

The isolates were further tested for antimicrobial activity. The isolates MYSN 106, MYSN 98, and MYSN 28 showed the highest antibacterial activity against the enteric pathogens tested. The obtained results clearly show the role of organic acids for the antagonistic activity of the isolates tested. The increased production of the organic acids through the fermentation reduces the pH of the media, which is known to inhibit the pathogens reducing their intercellular pH, leading to disruption in vital cell functions (Kivanç and Yilmaz, 2011). The MYSN 106 isolate proved to have the best antifungal activity for the tested mycotoxigenic fungi. The inhibitory effect, due to their competitive exclusion to bind to the gastrointestinal

tract, is essential for the selection of probiotic organisms as starter cultures; this is generally associated with antimicrobial metabolites and antifungal compounds (Henning et al., 2015; Son et al., 2017). Thus, the species used as a probiotic starter culture may play an important role in providing health benefits to consumers and also in avoiding food spoilage due to colonization of mycotoxigenic fungi (Wei et al., 2006; Guimarães et al., 2018). These isolates can also be used in agricultural practices for controlling mycotoxigenic fungi in post- and pre-harvesting practices (Nagaraja et al., 2016; Deepa et al., 2018).

The LAB isolates from Neera samples are resistant to some of the antibiotics tested in the present study. The results of antibiotic susceptibility are similar to previous studies that have also reported the absence of acquired resistance in the LAB that were isolated from natural fermented samples (Casado et al., 2014). Even though certain antibiotic-resistant, infectious strains of enterococci, including E. faecium, have been identified, they very rarely present a risk of infection outside healthcare situations (Sanders et al., 2010). However, enterococci strains of plant origin generally display low levels of virulence, as they colonize GIT producing bacteriocins and by the fact that they have been used safely for years as probiotics in humans, and farm animals (Arias and Murray, 2012; Hanchi et al., 2018). Also, the LAB frequently harbor plasmids of different sizes, and some may contain antibiotic determinants. Therefore, if the LAB including enterococci isolates are used as starter cultures, they exhibit virulence only if the organism has the ability to transfer the resistance. Before using these isolates in food or feed formulations as per EFSA guidelines, the virulence and antimicrobial resistance genes will be verified to prevent the horizontal gene transfer for antibiotic resistance (Rychen et al., 2018). A pre-market safety assessment is also required where the safety of the isolates is assessed at species level (Montealegre et al., 2016; Divisekera et al., 2019).

The cell surface hydrophobicity and auto aggregation experiments help in studying the colonization and adhesion of probiotic bacteria to epithelial cells in the gastrointestinal tract, which lead to the prevention of colonization by pathogens through their interaction (Abushelaibi et al., 2017). The results obtained show that all the isolates have comparable auto aggregation ranging from 40 to 80% and hydrophobicity ranging from 50 to 75%, with MYSN 106 proving to be the best after 2 h of incubation at 37◦C. The adhesion to chicken epithelial cell experiment also showed results comparable to those of auto aggregation and hydrophobicity; good adhesion was exhibited through interaction of cell surface components, which is one of the important probiotic characteristics (Kumari et al., 2016).

The simulated gastric and pancreatic digestion was done to test the survival of LAB isolates under the harsh conditions present in the gastrointestinal tract. The isolates could sustain the simulated digestive conditions without any loss in the viable cell count. Safety evaluation through DNase test to check for the pathogenicity of bacteria producing DNase enzyme that may cause hydrolysis of the DNA molecules was performed. Therefore, the absence of DNase in antimicrobial strains tested was confirmed to support the safety of their use in fermentations (Yadav et al., 2016; Singhal et al., 2019). The hydroxyl radicalscavenging activity of the LAB isolates is due to the colonization of viable cells and their propagation in the gut. The results obtained are comparable with previous studies as they also confer that the isolates tested help in the cure of cardiovascular diseases and gastrointestinal disorders and also improve the immune response (Kaushik et al., 2009). Among all the isolates tested, the isolate MYSN 106, which exhibited the highest potential as a probiotic, was identified as L. brevis by 16S rDNA sequencing and phylogenetic analysis with 99% homology. Many studies have been conducted demonstrating the probiotic potential of L. brevis isolated from a wide variety of fermented food samples (Pennacchia et al., 2004).

The LAB strains (MYSN 10, MYSN 106, MYSN 43, MYSN 109, MYSN 98, MYSN 18, and MYSN 28) isolated from the naturally fermenting Neera demonstrated probiotic attributes in vitro with good antimicrobial properties, proving their potential to be used as a starter culture in fermented food products, food or feed preservation, scavenging pathogens, and the biocontrol of mycotoxigenic fungi.

#### CONCLUSION

The probiotic strains that are isolated from traditional fermented food have a broad spectrum of antimicrobial activity and can be used as preservatives in food products. In this work, the authors focused on screening probiotic bacteria from the naturally fermented beverage – Neera. All the seven isolates exhibited resistance to gastrointestinal conditions and good antimicrobial activity. Among the seven isolates selected on screening, L. brevis MYSN 106 proved to be best – surviving the low pH and bile conditions in the stomach, including the harsh intestinal conditions. It also possesses surface-binding properties capable of colonizing the gastrointestinal tract, which is important for antimicrobial activity and disease treatment; this makes it a potential probiotic. Finally, the LAB isolates from Neera demonstrated probiotic attributes with good antimicrobial activities in vitro, therefore exhibiting potentiality to use them as probiotics in food and feed formulations.

#### AUTHOR CONTRIBUTIONS

MYS and RS designed the research. RS, BS, and BVD carried out the research activities. RS analyzed the data and wrote the manuscript. BVD and MYS edited and submitted the final version of the manuscript. All authors gave their approval for publication.

#### ACKNOWLEDGMENTS

We are grateful to the Science and Engineering Research Board, Department of Science & Technology, Government of India for the support (REF. EEQ/2016/000273). We thank the Institution of Excellence, Manasagangotri, and University of Mysore for their continuous support.

#### REFERENCES

fmicb-10-01382 June 27, 2019 Time: 15:14 # 10


MYS6 against fumonisin producing Fusarium proliferatum associated with poultry feeds. PLoS One 11:e0155122. doi: 10.1371/journal.pone.0155122



buffalo curd samples collected from Kandy. Ceylon 159, 159–166. doi: 10.4038/ cmj.v62i3.8519


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Somashekaraiah, Shruthi, Deepthi and Sreenivasa. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

, Yu Chen<sup>5</sup>

\* and

# Pinocembrin Protects Against Dextran Sulfate Sodium-Induced Rats Colitis by Ameliorating Inflammation, Improving Barrier Function and Modulating Gut Microbiota

#### Edited by:

Lin Hu<sup>1</sup>

Kai Wang<sup>6</sup>

, Chao Wu<sup>1</sup>

\*

, Zijian Zhang<sup>2</sup>

Jie Yin, Institute of Subtropical Agriculture (CAS), China

#### Reviewed by:

Qingbiao Xu, Huazhong Agricultural University, China Qiyang Shou, Zhejiang Chinese Medical University, China Cuong Tran, Health and Biosecurity (CSIRO), Australia

> \*Correspondence: Yu Chen blow2000@163.com Kai Wang

kaiwang628@gmail.com

#### Specialty section:

This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology

Received: 08 April 2019 Accepted: 02 July 2019 Published: 19 July 2019

#### Citation:

Hu L, Wu C, Zhang Z, Liu M, Maruthi Prasad E, Chen Y and Wang K (2019) Pinocembrin Protects Against Dextran Sulfate Sodium-Induced Rats Colitis by Ameliorating Inflammation, Improving Barrier Function and Modulating Gut Microbiota. Front. Physiol. 10:908. doi: 10.3389/fphys.2019.00908 <sup>1</sup> Jiangsu Key Laboratory of Infection and Immunity, Institutes of Biology and Medical Sciences, Soochow University, Suzhou, China, <sup>2</sup> School of Biology and Basic Medical Sciences, Soochow University, Suzhou, China, <sup>3</sup> Chinese Academy of Inspection and Quarantine, Beijing, China, <sup>4</sup> Shenzhen Key Laboratory of Translational Medicine of Tumor, Department of Cell Biology and Genetics, Shenzhen University Health Sciences Center, Shenzhen, China, <sup>5</sup> Department of Experimental Animals, Zhejiang Academy of Traditional Chinese Medicine, Hangzhou, China, <sup>6</sup> Institute of Apicultural Research, Chinese Academy of Agricultural Sciences, Beijing, China

, E. Maruthi Prasad<sup>4</sup>

, Mingchang Liu<sup>3</sup>

Pinocembrin (PIN) is a natural flavonoid widely found in bee propolis with potent gastrointestinal protective effects. In consequence, PIN has great potential in preventing inflammatory bowel diseases (IBDs) while scant information is available. In this study, a dextran sulfate sodium (DSS)-induced rats ulcerative colitis model (3.5% DSS in drinking water for 7 days) was applied to explore the protective effects of PIN on macroscopic colitis symptoms, inflammation, intestinal epithelial barrier function, and gut microbiota homeostasis. While DSS-treated rats showed severe colitis clinical symptoms and histological changes (colonic pathological damages and intestinal goblet cells loss), pre-administration of PIN (5 and 10 mg/kg, p.o.) for a week alleviated these symptoms. Pre-administration of PIN also suppressed the pro-inflammatory gene expressions and improved tight junction functions of colonic epithelial cells. Additionally, PIN administration reversed DSS-induced short chain fatty acid loss, and improved the gut microbial diversity assessed by 16S rRNA phylogenetic sequencing. Overall, our results suggest a wide spectrum of protective effects of PIN in preventing IBDs.

Keywords: pinocembrin, colitis, tight junction protein, gut microbiota, IBDs

# INTRODUCTION

Ulcerative colitis (UC) and Crohn disease (CD) are major forms of the non-specific chronic inflammatory bowel diseases (IBD) with unknown etiology. Studies suggest that main disease segments of UC are the colon/rectum while those of CD are terminal ileum or colon (Neurath, 2017). The consensuses of IBD are several factors, including genetic predispositions, dysfunctional immunity, intestinal barrier dysfunction, and environmental risk factors. In addition, it has been reported widely that the intestinal microbiota plays a vital role in

mediating the detrimental factors associated with IBD at different stages (Azad et al., 2018). The intestinal microbiome plays an important role during IBD, and the literature showed characteristic dysbiosis in patients suffered from UC, CD, and pouchitis (Lane et al., 2017). Several therapies have been developed for the management of IBD, for instance using anti-inflammatory (such as sulfasalazine or corticosteroids) and immunosuppressive agents (azathioprine) showed novel biological benefits. However, the applications of these drugs have adverse effects during a long period treatment accompany with high relapse rates (Moura et al., 2015).

Recent days UC and CD have become a heavy burden for public health so there novel and alternative approaches are warranted for IBD patients (Cao et al., 2018). The epidemiological evidence suggests that increasing the diet intake of flavonoids rich fruits and vegetables is associated with low risk of IBD (Ananthakrishnan, 2015; Martin and Bolling, 2015). Pinocembrin (5,7-dihydroxy flavanone, PIN, **Figure 1A**) is one of a well-studied flavonoid widely found in many natural products, like Piperaceae family plants (more than 1950 species plants) (Rasul et al., 2013). PIN was reported to possess various pharmacological properties, including anti-oxidative, anti-inflammatory, neuro-protective and anti-carcinogenic activities (Lan et al., 2016). It is also widely recognized as a key bioactive constitute of bee propolis (Peng et al., 2012). Our research group showed recently that propolis a potent gastrointestinal protector (Wang et al., 2017, 2018). Nevertheless, the effects of PIN against IBD has not been investigated and the mechanisms underlying the activity of PIN remain unknown.

The aim of the present study was to investigate protective effect of pinocembrin (PIN) on colitis severity symptoms, antiinflammatory activities, restoring on intestinal barrier function and modulating gut microbial populations using a DSS-induced rats colitis model. Our working hypothesis is that PIN alleviated DSS-induced experimental colitis severity symptoms by its antiinflammatory activities, restoring on intestinal barrier function as well as modulating gut microbial populations.

# MATERIALS AND METHODS

### Chemical and Regents

PIN (purity >99%) was purchased from Biopurify Phytochemicals Ltd. (Chengdu, China); and DSS (M.W. 36–50 kDa) was purchased from MP Biomedicals (Irvine, CA, United States). All other reagents were obtained from Sangon Biotechnology (Shanghai, China) or as indicated in specified methods.

#### Animals and Acute Colitis Induction

Male Sprague Dawley rats (30 rats, 6 weeks old, 190–220 g) were housed in the Animal Experimental Center of the Zhejiang Institute of Traditional Chinese Medicine, in the SPF environment which following standard experimental protocols and approved from Animal Ethics Committee of Institute of Apicultural Research, Chinese Academy of Agricultural Sciences. All rats are fed by strand lab chow (Xietong Biotechnology, Nanjing, China) and water ad libitum and were maintained on a 12 h light/dark cycle (21 ± 2 ◦C with a relative humidity of 45 ± 10%). After 3 days acclimatization, experiment was started (7 days of drug pretreatment following by 7 days of DSS induction of acute colitis, concomitant with PIN or 5-ASA) and rats were divided into five groups (n = 6 each), (1) control group, (2) DSS colitis group, (3) PIN low dosage group (5 mg/kg b.w., p.o.), (4) PIN high dosage group (10 mg/kg b.w., p.o.), and (5) 5-ASA group (reference drug, 50 mg/kg b.w., p.o.). PIN and 5-ASA group rats received drugs for 7 days, prior to the DSS treatment and treated until the last day of the experiment. DSS was added in the drinking water for 7 days at 3.5% (w/v) for acute colitis induction. The colitis severity was evaluated from the DSS induction, based on the disease activity index (DAI), with the loss of body weight, stool consistency, rectal bleeding, and the overall health condition (Wang et al., 2017).

#### Histological Evaluation on the Distal Colon

All animals were anesthetized and sacrificed at the final day of the experiment. Distal colon samples were collected and fixed distal colon samples with neutral buffered formalin embedded in paraffin and stained with hematoxylin-eosin (HE) or periodic acid–Schiff (PAS) for histopathology examination. After deparaffinization and hydration, colonic sections were immersed in 1% periodic acid for 8 min, then placed in Schiff's reagent for 15 min, followed by dehydration, clearance and mounting in Canada balsam (Balaha et al., 2016). Goblet cell distribution was observed from 5 randomly selected crypts per colon using PAS-stained cross sections. Colonic tight junction proteins (ZO-1 and Occludin) were examined by immunohistochemical staining method. Paraffin sections were heated at 60◦C for 1 h in an oven. The de-paraffinization, endogenous peroxidase was blocked by 3% H2O<sup>2</sup> for 15 min, and the sections were incubated in primary antibody of ZO-1 and Occludin (Proteintech Group, Chicago, IL, United States) overnight at 4◦C in the refrigerator. After washed, the sections were incubated in secondary antibody for 15 min. Next, freshly prepared diaminobenzidine (DAB) solution was applied to visualize antibody, followed by hematoxylin staining, dehydration and mounting. The light microscope attached imaging system (Nikon Eclipse Ci, Japan) was used to visualize the stained slides and acquire images. Colonic sections were coded for blind microscopic assessment of histological changes by two pathologists. Colonic histological damage was scored from H&E-stained slides based on two subscores (cell infiltration and tissue damage), ranging from 0 to 6 (no changes to extensive cell infiltration and tissue damage). IHC score were calculated previously by using the formula: IHC score = 6(I × Pi), where I, intensity of staining and Pi, percentage of ZO-1 stained colonic cells, ranging from 0 to 300 (Yeo et al., 2015).

#### Transmission Electron Microscopy (TEM)

Distal colon samples were fixed in glutaraldehyde (2.5%) and 1% osmium tetroxide (1%) in 0.01 M phosphate

buffer (pH 7.0). Samples were then embedded in Epon 812 resin overnight and dried with gradient acetone, following the manufacturer's instructions (SPI-EM, Division of Structure Probe, Westchester, NY, United States). After obtaining ultrathin sections (70 nm), uranium acid and lead citrate were used for drying, and samples were

observed under a transmission electron microscopy (TEM, Hitachi, Tokyo, Japan).

# RNA Extraction and Quantitative Real-Time PCR (qPCR)

RNA was extracted from distal colon using a commercial kit (Carry Helix Biotechnologies Co., Ltd., Beijing, China) and synthesized to cDNA by reverse transcription (TaKaRa, Dalian, China). Quantitative real-time PCR was performed using a two-step amplification method and all information was collected by under the 7500c Real-time PCR Detection System (Applied Biosystems, Carlsbad, CA, United States) and calculated the transcriptional levels of target genes using the 2 −44Ct method based on cycle threshold (Ct) values and GAPDH was served as the housekeeping gene. Specific primers sequences can be found in our previous published literature (Wang et al., 2016a, 2018).

#### Short Chain Fatty Acids (SCFA) Analysis

Cecal digesta (∼150 mg) collected from rats were weighed and diluted at 1:3 (w/w) with deionized water containing 1.68 mM heptanoic acid/L as an internal standard (Sigma Chemical Co., St. Louis, MO, United States). Acetate, butyrate, propionate, and the total SCFA (including minor SCFA) were measured using a gas chromatography (GC) system based on our previous published methods (Wang et al., 2017).

#### Gut Microbial Community Analysis

The V3-V4 region of the 16S rRNA gene of the caecal microorganism was amplified after extracting DNA for pyrosequencing (Qiagen, Hilden, Germany). Universal primers: 319F (5<sup>0</sup> -ACTCCTACGGGAGGCAGCAG-3<sup>0</sup> ) and 806R (5<sup>0</sup> -GGACTACHVGGGTWTCTAAT-3<sup>0</sup> ) were applied for bacterial gene amplification using Premix Ex TaqTM Hot Start (Takara, Dalian, China). Then PCR products were sequenced using MiSeq Illumina sequencing (Illumina, United States) and the software package QIIME (version 1.17). Then UPARSE (version 7.1) software was applied to cluster the sequences and obtained operational taxonomic units (OTUs) with 97% similarity. Bacterial taxonomic profiles at different levels were generated within QIIME using the RDP classifier against the SILVA (SSU115) 16S rRNA database with a confidence threshold of 70% using the default settings. A Venn diagram was generated based on the relative abundances of OTUs in each sample from three groups (Normal, DSS and PVH treatment). Beta-diversity was measured by principal coordinate analysis (PCoA) at the OTU level and hierarchical clustering tree on Genus level was calculated based on UniFrac metrics dissimilarity matrix using Vegan 2.0 packages in the R (version 3.1.2) statistical environment.

#### Statistical Analysis

Data are presented as the arithmetic mean ± SD for indicated replicates. The effect of treatments was determined by oneway ANOVA and differences between treatments were analyzed post hoc by Tukey's honest significant difference test (pvalues ≤0.05 were considered statistically significant) using SPSS version 17.0.

# RESULTS

# PIN Alleviated DSS-Induced Acute Colitis in Rats

We observed the potential protective effects of PIN against colonic inflammation, DSS was used to induce the acute colitis to the rats for 7 days, followed by 1 week of oral preadministration with different doses of PIN (5 and 10 mg/kg) and 5-ASA (50 mg/kg, reference drug). As shown in **Figure 1B**, DSS-treated rats exhibited severe colitis symptoms, marked by shortened, thickened and erythematous colons, which were attenuated in PIN or 5-ASA treated rats. **Figure 1C** shows significant decrease of body weight gains in the DSS-induced colitis rats (61.8 ± 7.5 g in control group vs. 27.3 ± 3.3 in DSS group). Body weight data also suggested that rats in high dosage of PIN and 5-ASA administrations have less body weight loss compared with DSS control group. We also noticed that DSS control groups had severe colonic damages and disease activity index (DAI) among the groups (**Figure 1D**), compared with DSS control, PIN treatment regulated the DAI index (**Figure 1D**), and also prevented reduction in colon length/body weight ratio (**Figure 1E**). The protective effects in high dosage PIN group are even comparable to that of the 5- ASA group.

#### Effect of PIN Treatment on the Colonic Pathological Changes and Goblet Cells in DSS-Induced Colitis Rats

Severe histological colonic damages were observed in the DSSinduced colitis rats group due to infiltration of immune cells and muscle layer thickness (**Figure 2A**). Histopathological analysis suggested that pre-treated rats with PIN or 5-ASA showed near normal control histology of the colonic epithelium, compared with DSS induced rats (**Figure 2B**). We also observed PAS staining of the goblet cells (**Figure 2C**), found decreased goblets cells in DSS-induced colitis colon rats (19.2 ± 1.6 per crypts), and whereas PIN treated rats showed regulated goblet cells (22.8 ± 4.6, low dosage, and 27.2 ± 4.5 per crypts, high dosage) near to normal control rats (29.2 ± 2.9 per crypts **Figure 2D**).

#### Effect of PIN Treatment on Colonic mRNA Expressions in DSS-Induced Colitis Rats

As shown in **Figure 3**, pretreatment with PIN regulated the anti-inflammatory responses mediated by colonic TGF-β expression and mRNA expression of the pro-inflammatory mediators such as IL-1β, IL-6, and TNF-α in DSS-induced rats. Moreover, increased gene expressions of tight junction proteins, ZO-1 and Occludin were observed in PIN pre-administrated

were stained with hematoxylin-eosin (HE). (B) Histological scores of colon sections from each group and scores are expressed as means ± SD (n = 8 for each group). Groups with different letters differ by a statistically significant margin (p < 0.05). (C) The distal colon sections were subjected to Periodic Acid-Schiff (PAS) staining of goblet cells (blue) in the colon. (D) Averages of goblet cells/crypt in the colon are shown. The data represent the mean ± SD of six rats in each group. The stars indicate that value was significantly different (∗∗p < 0.01, ∗∗∗p < 0.001) compared with DSS control.

groups, which benefits the mucosal barrier function and epithelial restitution.

# PIN Improved the Tight Junction Functions in DSS-Induced Colitis Rats

As shown in **Figure 4A**, we observed the DSS group showed decreased ZO-1 expressions in their distal colon mucosa. PIN pre-treatment (7 days) reverted the loss of TJ compared with DSS induced rats using IHC method. TEM analysis suggested that PIN pretreatment restored the enlarged TJ and expansion of the endoplasmic reticulum compared with DSS-induced rats (**Figure 4B**).

# Effect of PIN on Cecal SCFAs and Microbiota Composition in DSS-Induced Colitis Rats

We next measured the cecal SCFA concentrations, including acetate, propionate, butyrate, and total SCFA by GC, which

indicated as hallmarks for intestinal microbial metabolic products. Total SCFA concentrations as well as acetate, propionate and butyrate were similarly decreased in DSS colitis rats (3943.8 ± 105.6 µg/g), compared with normal control rats (4666.9 ± 213.8 µg/g). High dosage of PIN pre-administration resulted in significant increase of SCFA (4521.1 ± 174.2 µg/g), indicating its beneficial modifications on the colonic microbial populations during colitis (**Figure 5A**).

As high dosage of PIN showed best gastrointestinal protective effects, we next chose this group together with control and DSS groups, to analysis cecal microbial community structure using 16S ribosomal RNA Illumina next-generation sequencing. We observed a significant decrease of the Shannon α-diversity index between DSS-induced and control rats.PIN pretreatment showed an overall impact on increasing the Shannon index (**Figure 5B**). Furthermore, we noticed a significant difference in microbial β-diversity, based on Principal coordinate analysis (PCoA) of weighed UniFrac distance (**Figure 5C**). Unweighted Pair Group Method with Arithmetic Mean (UPGMA) was further used to show Unifrac alpha and beta diversity at the genus level. We noticed that DSS has less diversity of bacteria genera, compared with the PIN and control groups. Some specific bacteria genera, including Lactobacillus spp., Alloprevotella spp., and Desulfovibrio spp. were rescued by PIN (**Figure 5D**).

# DISCUSSION

In the previous studies, we found that Chinese propolis administrations showed protective effects against acute colitis symptoms using rodent models and the intestinal mucosal barrier functions were improved in human intestinal cells (Wang et al., 2016a, 2018). As Chinese propolis contains abundant polyphenolic constitutions, like (CAPE, chrysin, galanin, and PIN)(Huang et al., 2014), we further showed that PIN might be a key contributor for the anti-colitis and modulating the gut microbiota. We used DSS-induced colitis in rats, which is a well-known experimental model and mimics clinical symptoms similar to UC (Wang et al., 2018). In agreement with previous reports, significant reduced body weight gains, a dramatic increase of DAI accompany with shortening the length of colon was observed in DSS-induced rats and suggesting the successful repentance of the model we used (Bramhall et al., 2015; Munyaka et al., 2016). We chose the dose of PIN at 5 and 10 mg/kg (∼5 and 10-fold of the human dose), p.o., based on several previous publications which support the safe usage of PIN in rodents (Gao et al., 2010; Liu et al., 2014). We showed PIN at the dose of 10 mg/kg had comparable protective effects to 5- ASA against colitis symptoms, a reference drug which was most widely used in clinical practice (Wang et al., 2016b). We also

FIGURE 4 | Effects of pinocembrin on colon tight junction. Distal colon was collected and embedded in paraffin for ZO-1 immunohistochemistry (A) or fixed in glutaraldehyde for transmission electron micrographs (TEM, B). Black arrows indicated tight junctions between the colonic epithelial cells. The immunohistochemistry scores of ZO-1 are expressed as the mean ± SD of four rats in each group. The stars indicate that values were significantly different (∗p < 0.05, ∗∗∗p < 0.001) compared with DSS control.

β-diversity distances showing microbiota taxa composition at the genus level. Bacteria represented by different colors are marked in the upper right.

coordinate analysis (PCoA) of weighted UniFrac distance. (D) Unweighted Pair Group Method with Arithmetic Mean (UPGMA) clustering of unweighed UniFrac

showed PIN pretreated rats histopathological damages were near normal control rats and the DSS induced colitis rats showed thick colonic mucus layer, which indicates that a loss of function in the physical barrier between the epithelium and lumen (Sperandio et al., 2015). Since colonic mucus is mainly secreted from the goblet cells, an increase of goblet cell numbers by PIN indicated that a restored mucous barrier in the gastrointestinal tract (Johansson and Hansson, 2016).

It has been recognized that the integrate mucosal barrier is important for maintaining the hemostasis of the gut, since the leaky gut barrier facilitates the invading of pathological bacteria into the colonic mucosa, lead to system innate and inflammatory immune responses in the host (Yang et al., 2017). Recent studies suggested that the increase of proinflammatory cytokine, TNF-α, lead to intestinal dysfunction along with the other pro-inflammatory cytokines such as IL-1β and IL-6 (Zhang et al., 2018a). Consistent with previous findings in macrophages and microglial cells (Soromou et al., 2012; Zhou et al., 2015), PIN pre-administration showed strong anti-inflammatory effects against the inflammation-related gene expressions. The TGF-β is one of an anti-inflammatory cytokine which showed alleviating the colitis and protecting the intestinal epithelial cells. It is required for intestinal mucosal healing, and lack of TGF-β strongly increases the epithelial susceptibility to injury (Beck et al., 2003; Bermudez-Humaran et al., 2015). In the present study, the PIN increased the TGF-β expression in colitis rats, which was consistent with a previous study (Zhou et al., 2018). The gut physiological barrier between the intestinal epithelial cells is formed by the TJ complex, which has key roles in restricting and modulating the intestinal permeability. Several transmembrane proteins and cytosolic adaptor proteins composed the TJ. Occludin is an integral plasma membrane protein and zona occludens (ZO) proteins-1 a peripheral membrane protein, which are indicators for tight junction assembly, stability, and barrier function (Zihni et al., 2016). Previous studies suggested that several intestinal pathologies are associated with TJ barrier disruptions, including IBD (Suzuki and Hara, 2011). A number of dietary nutrients, including flavonoids, have been demonstrated to increase TJ function as well as TJ proteins (Gil-Cardoso et al., 2016;

Yang et al., 2017). Such natural flavonoids including quercetin (Suzuki and Hara, 2009), galangin (Zhang et al., 2019), naringenin (Dou et al., 2013). Inconsistent with these studies, pre-administration of PIN regulated the gene and protein expressions of ZO-1 and Occludin in the colon. Electron microscopy images clearly showed that PIN dampened wide intercellular spaces and the damaged tight TJ structures in colitis rats (Zhang et al., 2018b). Nevertheless, the molecular mechanisms of regulating effects of TJs by PIN still need to be explored clearly. The gastrointestinal microenvironment is a complex orchestration between the host and the gut microbial populations. We showed that cecal SCFA levels were decreased significantly by DSS, which were consistent with previous findings (Wu et al., 2018). We demonstrated that PIN increased SCFA concentrations (acetate and butyrate) in colitis rats, indicative of increased microbial activity of the gut (Monk et al., 2016). Clinical evidence showed that IBD patients lead to alterations in the diversity and composition of their gut microbiota. Dietary intervention recently has been attracted much attention for the potential applications into complementary/alternative approaches of IBD prevention or treatment (Yin et al., 2018a,b).

Our studies showed that dietary intake of PIN rich propolis has great protective effects against DSS-induced rats colitis, and modulates gut microbiota composition as well as increases the intestinal microflora diversity. Although there is scant of studies investigated the modulating effects on the gut microbiota of PIN, we showed that PIN reversing the decrease on the gut microflora community induced by DSS, based on the alpha diversity of the Shannon index (Cui et al., 2018). The decrease in the probiotic bacteria, like Lactobacillus spp. was noticed from DSS-induced rats, while PIN reversed the alteration. The modulations of the gut microflora could also potentially explain the protective effects by PIN against DSS-induced rats colitis.

#### CONCLUSION

Overall, we observed that the oral pretreatment of PIN, a natural flavonoid, showed significant protective effects against acute colitis induced by DSS. These beneficial effects might be attributed to the anti-inflammatory effects, restoring on

#### REFERENCES


the intestinal barrier function and modulating on the gut microbiota. Our study provides important pre-clinical evidence for the potential application of PIN to the prevention or treatment of gastrointestinal disorders. However, further studies are needed to identify the exact underlying mechanisms of PIN consumption against IBDs.

#### DATA AVAILABILITY

The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the ARRIVE (Animal Research: Reporting of In vivo Experiments) guidelines, Animal Ethics Committee of Institute of Apicultural Research, and the Chinese Academy of Agricultural Sciences. The protocol was approved by the Animal Ethics Committee of Institute of Apicultural Research and the Chinese Academy of Agricultural Sciences.

#### AUTHOR CONTRIBUTIONS

LH and KW carried out the study and performed the statistical analysis. CW, ML, and ZZ provided help in performing the experiments. KW and YC designed the research. LH, KW, and YC prepared the draft of the manuscript. EM read and revised the final version of the manuscript.

#### FUNDING

This study was supported by the Natural Science Foundation of Jiangsu Province, China (BK20180839), National Key Research and Development Program of China (2018YFC1603606-4), and Experimental Animal Program for Application of Public Welfare Technology of Zhejiang Province, China (2018C37128).

IL-10 and TGF-beta anti-inflammatory cytokines against mouse colitis when delivered by recombinant lactococci. Microb. Cell Fact. 14:26. doi: 10.1186/ s12934-015-0198-4


Toll-like receptor 4/NF-kappaB signalling. Br. J. Nutr. 110, 599–608. doi: 10. 1017/S0007114512005594


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Hu, Wu, Zhang, Liu, Maruthi Prasad, Chen and Wang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Probiotic Properties of Lactobacillus paracasei subsp. paracasei L1 and Its Growth Performance-Promotion in Chicken by Improving the Intestinal Microflora

Yunhe Xu<sup>1</sup> , Yuan Tian<sup>1</sup> , Yunfang Cao<sup>2</sup> , Jianguo Li<sup>1</sup> , Haonan Guo<sup>1</sup> , Yuhong Su<sup>1</sup> , Yumin Tian<sup>1</sup> , Cheng Wang<sup>1</sup> , Tianqi Wang<sup>1</sup> and Lili Zhang<sup>1</sup> \*

<sup>1</sup> Department of Food Science and Engineering, Jinzhou Medical University, Jinzhou, China, <sup>2</sup> Tianwang Animal Health Supervision Institute, Jinzhou Economic and Technological Development Zone, Jinzhou, China

#### Edited by:

Yuheng Luo, Sichuan Agricultural University, China

#### Reviewed by:

Sarbjeet Makkar, Washington University in St. Louis, United States Xiaofei Cong, Eastern Virginia Medical School, United States

> \*Correspondence: Lili Zhang lilyzhang1977@163.com

#### Specialty section:

This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology

Received: 23 April 2019 Accepted: 09 July 2019 Published: 25 July 2019

#### Citation:

Xu Y, Tian Y, Cao Y, Li J, Guo H, Su Y, Tian Y, Wang C, Wang T and Zhang L (2019) Probiotic Properties of Lactobacillus paracasei subsp. paracasei L1 and Its Growth Performance-Promotion in Chicken by Improving the Intestinal Microflora. Front. Physiol. 10:937. doi: 10.3389/fphys.2019.00937 Lactobacillus paracasei subsp. paracasei L1 was previously isolated from sweet potato sour liquid. This bacterial species specifically binds onto starch granular surfaces, triggering the enzymatic hydrolysis of raw starch. We investigated the functional and safety properties of strain L1 in vitro to establish its probiotic potential, and analyzed its effect on growth performance and intestinal microflora of chicken in feeding experiments. The optimal growth conditions of strain L1 included low pH and high concentrations of bile salts and NaCl. Its 1-, 2-, and 24-h autoaggregation values were 15.8 ± 1.2%, 20.4 ± 2.3%, and 47.2 ± 0.8%, respectively, with the surface hydrophobicity value at 560 nm of 38.1 ± 2.7%. Further, its adhesion rate to Caco-2 cells was 22.37 ± 1.44%. Strain L1 was resistant to erythromycin and azithromycin, but sensitive to other antibiotics tested. For the feeding experiments, 240 chickens with similar weights were randomly divided into a control (C) group and strain L1 (L) group and fed for 8 weeks. Strain L1 promoted the weight gain of chickens in L group. A significant increase in the population size of the two phyla and 23 genera in the small intestine was observed in the presence of strain L1 (P < 0.05), with 0 phyla and 4 genera showing significant increase in the cecum (P < 0.05). In the small intestine, the abundance of six functional genes at Kyoto Encyclopedia of Genes and Genomes (KEGG) level 2 and 49 genes at KEGG level 3 was significantly increased in group L (P < 0.05), with lesser changes noted in the cecum. An increase in the metabolic pathway functions, including enzyme families and the digestive system, was observed in the intestinal microbiota in the L group compared to the C group. However, the other metabolic pathway functions, including metabolism of fatty acid biosynthesis, as well as metabolism of glycerolipids and propanoate, increased in the cecal microbiota of the L group relative to the C group. These changes are most likely related to the changes in the gut microbiota composition. Collectively, strain L1 supplementation may promote growth performance and improve the intestinal microflora in chicken although further studies are needed to confirm this.

Keywords: Lactobacillus, probiotic, gut microbiota, chicken, growth performance

# INTRODUCTION

fphys-10-00937 July 24, 2019 Time: 14:55 # 2

In 2006, the European Union prohibited the utilization of AGPs in animal production (European Commission [EC], 2001). This directive has resulted in major problems in animal production such as a significant decrease in growth performance and an increase in the prevalence of diseases that were previously prevented by the use of antibiotics (Babak and Nahashon, 2014; Wielinga et al., 2014; Zou et al., 2016; Al-Khalaifah, 2018). Hence, extensive efforts have been made to develop AGP substitutes as feed additives, which include essential oils, fermented liquid feed, organic acids, probiotics, and prebiotics. Previous studies have shown that probiotics promote growth, thereby enhancing animal production by increasing the intake and conversion rate of feeds and total body weight (Taras et al., 2007; Chaucheyras-Durand and Durand, 2010; Dittoe et al., 2018). Furthermore, probiotics have been shown to aid digestion in animals by improving the absorption of specific essential nutrients (Yu et al., 2008).

Probiotics research has lately focused on LAB, in particular, the bacterial species Lactobacillus, Lactococcus, and Bifidobacterium based on their potential health benefits (Shekh et al., 2016). The term "probiotic" refers to "live microbial species that are beneficial to the host when consumed at sufficient amounts" (FAO/WHO, 2002; O'Connell Motherway et al., 2008; Kotzamanidis et al., 2010). However, despite convincing evidence that certain lactobacilli strains are safe for human utilization as well as confer specific health benefits to the host, these positive effects cannot be applied to other strains in the absence of results from experimentation (FAO/WHO, 2002; Kotzamanidis et al., 2010). Before assessing the in vivo probiotic properties of a strain, it is essential to confirm its features relating to safety, survival in the gastrointestinal tract, colonization ability, and other probiotic characteristics.

What is the mechanism by which probiotic bacteria confer host health benefits? Macromolecules of the cell surface of bacteria are major components that interact between probiotic bacteria and its host and involve pattern recognition receptors (PRRs) in gastrointestinal mucosa of the host (Lebeer et al., 2010). In recent years, many probiotic LAB have been found to possess a variety of proteins anchored on the surface or the cell wall, and most of these were enzyme proteins related to carbohydrate metabolism and transport. These proteins play a major role in LAB adhesion onto the intestinal tract and are also responsible for sugar catabolism or degradation of various complex sugars such as lactose or starch (Zhang et al., 2017). The activity of LAB to metabolize carbohydrates is crucial for their colonization of and proliferation in the intestine (Ganzle and Follador, 2012; Duranti et al., 2014; Liu et al., 2015).

Lactobacillus paracasei subsp. paracasei L1 has been isolated from a naturally fermented sour liquid of sweet potato (Zhang et al., 2017). It is surface-anchored via glycoside hydrolase, cell wall peptidoglycan hydrolase, phosphoglycerol kinase, glyceraldehyde-3-phosphate dehydrogenase, enolase, etc., which are related to sugar metabolism and mediate the specific binding of L1 cells to a starch granule. The strain has the ability to hydrolyze raw starch to generate simple carbohydrates, including glucose and lactic acid, that are capable of altering the structural, physical, and chemical features of starch granules (Zhang et al., 2017). The ability of LAB to bind and utilize carbohydrates is important for the colonization of the intestine and promotion of carbohydrate metabolism. By relying on the characteristics of LAB starch metabolism, additional carbon sources are generated for the competitive growth of intestinal microflora, which benefits the competitive growth of LAB in the intestinal tract (Wang et al., 1999; O'Connell Motherway et al., 2008).

Chicken feed contains copious amounts of raw starch. Enhancing the utilization of starch by LAB is greatly important for improving the production performance of chicken and reducing the cost of raising chicken. Moreover, these enzyme proteins on the surface of L. paracasei subsp. paracasei L1 were also present on the surfaces of many probiotic lactic acid bacteria; therefore, we speculated that L. paracasei subsp. paracasei L1 has potential probiotic properties. Thus, the aim of the present study was to examine the functional and safety features of strain L1 in vitro to determine its potential use as a probiotic. The effect of strain L1 on the growth performance and intestinal microflora of chicken was then evaluated in feeding experiments. The study lays a theoretical foundation for the application of strain L1 in chicken production.

#### MATERIALS AND METHODS

#### Materials

Strain L1 has been previously isolated from sour liquid (Zhang et al., 2017). This strain has been deposited to the China General Microbiological Culture Collection Center (CGMCC, No. 4163). Glycerol stocks (30% glycerol, v/v) of the pure culture were stored at −80◦C until use. Bacterial cultures were aerobically prepared on a sweet potato juice medium (as described later) at 4 ◦C and then transferred to a fresh medium each month. Prior to analysis, the strain was statically cultured at 30◦C for 24 h (until the stationary growth phase) under aerobic conditions.

**Abbreviations:** AGP, antibiotic as growth promoter; cfu, colony-forming units; KEGG, kyoto encyclopedia of genes and genomes; LAB, lactic acid bacteria; OTU, operational taxonomic unit; PBS, phosphate-buffered saline; PCoA, principal coordinate analysis; PCR, polymerase chain reaction.

We employed enterocyte-like Caco-2 ECACC 86010202 cells (from colon adenocarcinoma) in simple adhesion assays. PBS, (pH 7.2) was obtained from chemical reagent company (Sigma-Aldrich, St. Louis, MO, United States). All other chemical reagents used in this study were of analytical grade.

The sweet potato juice medium was prepared according to Zhang et al. (2017). Briefly, a sweet potato infusion using 200 g of sliced (washed but unpeeled) sweet potatoes in 1 L of distilled water was boiled for 30 min, and the broth was decanted or strained through a cheesecloth. Then, distilled water was added to the infusion to a total volume of 1 L, to which 20 g of glucose, 2 g of lactose, 5 g of yeast extract, and 5 g of sodium acetate were added. The culture medium was then autoclave-sterilized at 115◦C for 15 min (Zhang et al., 2017).

#### Acid and Bile Salt Tolerance

Viability of strain L1 was examined according to de Albuquerque et al. (2018). Tolerance to various pH values and concentrations of bile salts was evaluated by inoculating 1-mL aliquots of the strain L1 suspension (grown in sweet potato juice medium for 24 h at 30◦C) in 10 mL of sweet potato juice medium at various pH levels (2.0, 2.5, 3.0, or 3.5 by using 1 M HCl) or supplemented with different bile salt concentrations [0.03, 0.3, or 0.5% (w/v)] (Sigma-Aldrich, St. Louis, MO, United States). The suspensions were then incubated at 30◦C. We collected 1-mL aliquots of the suspension at different incubation time intervals (1–4 h), and each aliquot was serially diluted in sterile peptone water (0.15 g·100 mL−<sup>1</sup> ), and streaked onto the MRS agar. The viable cells were manually counted and expressed in terms of log cfu·mL−<sup>1</sup> . For the control, strain L1 cells were cultivated in sweet potato juice medium (pH 7, adjusted with 1 M NaOH) in the absence of bile salts (de Albuquerque et al., 2018).

#### NaCl Tolerance

The strain L1 cultures (grown in sweet potato juice medium for 24 h at 30◦C) were transferred (5%, v/v) into fresh sweet potato juice medium containing 1, 2, 3, 4, or 5% (w/v) NaCl or a fresh sweet potato juice medium without NaCl (control) and incubated at 30◦C. Viable cells in the medium with and without NaCl were counted after incubating for 24 h and expressed in terms of log cfu·mL−<sup>1</sup> .

#### Autoaggregation Assay

To assess autoaggregation capacity, the strain L1 cells grown in sweet potato juice medium (for 24 h at 30◦C) were harvested by centrifugation at 3,000 × g for 10 min at 20◦C, washed with PBS twice, and then resuspended in PBS to an OD<sup>660</sup> of 0.3. After incubating at 37◦C for 60 min, the OD<sup>660</sup> value was again measured. Autoaggregation was calculated using the following equation (1):

$$\text{Autogregation} \left( \% \right) = \frac{\text{OD}\_0 - \text{OD}\_{60}}{\text{OD}\_0} \times 100\% \tag{1}$$

where OD<sup>0</sup> is the initial OD value, and OD<sup>60</sup> is the OD value after incubating for 60 min (de Albuquerque et al., 2018).

# Cell Surface Hydrophobicity

Bacterial adhesion onto hydrocarbon-like toluene was assessed as described by Iñiguez-Palomares et al. (2007). The strain L1 cells (5-mL suspension) were harvested in triplicate via centrifugation at 3,000 × g for 15 min, washed with PBS (pH 7.2) twice, and then re-suspended in the same buffer to a density of approximately 10<sup>8</sup> cfu·mL−<sup>1</sup> (OD560; A). Then, 4 mL of each suspension were mixed with 1.2 mL of toluene. After incubating for 10 min, the bacterial suspension was then thoroughly mixed with toluene by vortexing for 2 min. Then, the OD (A0) of the aqueous phase was determined at a wavelength of 560 nm. The hydrophobicity percentage (H) was estimated as the equation (2):

$$\text{H}\,(\%) = \frac{\text{A} - \text{A0}}{\text{A}} \times 100\% \tag{2}$$

where A and A<sup>0</sup> are the absorbance values that were measured before and after toluene extraction, respectively (Iñiguez-Palomares et al., 2007).

#### Bacterial Adhesion Onto Caco-2 Cells

Human colon cancer Caco-2 cells were routinely cultured in Dulbecco's modified Eagle's medium (DMEM) containing 10% (w/v) fetal bovine serum and 1% (w/v) antibiotic solution (100 µg·mL−<sup>1</sup> penicillin and 100 µg·mL−<sup>1</sup> streptomycin). The cells were cultured in flasks at 37◦C in 5% CO<sup>2</sup> atmosphere. The Caco-2 cells were inoculated into a six-well cell culture plate at a density of 10<sup>5</sup> cells per well for cell fusion, then cultured for 20 days and employed in the adhesion assay. The cell culture medium was replaced with fresh DMEM supplemented with 2% (w/v) fetal bovine serum without the antibiotics for at least 1 h prior to the adhesion assay. A Lactobacillus suspension (10<sup>8</sup> cfu·mL−<sup>1</sup> in PBS) was added to each well of the tissue culture plate and cultured for 3 h at 37◦C and 5% CO<sup>2</sup> atmosphere. The plate was then washed thrice with 1 mL of PBS to remove non-adhering bacteria. The cells were then incubated with Triton X-100 (0.05%) for 10 min. The lysate was diluted, followed by coating with the appropriate diluent on MRS agar. Adhesivity was expressed as the percentage of bacteria that adhered onto the Caco-2 cells and the initial number of bacteria (Pisano et al., 2014).

#### Safety Assessment

Strain L1 antibiotic susceptibility was assessed using the disc diffusion method as described by the Clinical and Laboratory Standards Institute (Clinical and Laboratory Standard Institute, 2009). However, the Mueller-Hinton agar was substituted with MRS agar in the assay. We tested the following antibiotics (Oxoid): amoxicillin/clavulanic acid (30 µg), azithromycin (15 µg), cefotaxime (30 µg), ciprofloxacin (5 µg), erythromycin (15 µg), gentamicin (10 µg), kanamycin (30 µg), norfloxacin (5 µg), penicillin G (10 µg), rifampicin (30 µg), streptomycin (10 µg), teicoplanin (30 µg), tetracycline (30 µg), and vancomycin (30 µg). We placed an antibiotic disc on the MRSA agar plate after spreading an overnight strain L1 culture (0.1 mL) using an antibiotic disc dispenser, and the plates incubated for 24 h at 30◦C. We measured the diameters of the bacterium-free

zones, and the results were expressed as resistance based on the interpretative criteria established by the Clinical & Laboratory Standards Institute. Plasmid extraction was done using GeneJET plasmid miniprep kit (Thermo Fisher Scientific, United States).

The amino acid decarboxylating activity of strain L1 (L-histidine, L-lysine, L-ornithine, and L-tyrosinel Sigma-Aldrich) was determined according to Bover-Cid and Holzapfel (1999). Briefly, the strain L1 culture was inoculated (2%, v/v) in a decarboxylase medium with or without amino acids (the control). After incubating at 30◦C for 72 h, we confirmed biogenic amine production based on color changes in relation to amine formation. A positive result was established when a change in the color of the medium was detected (Bover-Cid and Holzapfel, 1999).

Hemolytic activity was assessed by streaking the cells onto Columbia blood agar plates that were supplemented with 5% defibrinated sheep blood, and then incubated for 48 h at 37◦C. After incubating, the hemolytic reaction was evaluated based on the presence of a clear zone of hydrolysis surrounding the colonies (β-hemolysis), partial hydrolysis, as well as a greenish zone (α-hemolysis) or no reaction (γ-hemolysis). Positive controls were prepared using Staphylococcus aureus ATCC 25923 cells.

#### Chickens, Treatment, and Sampling

A total of 1000 1-day-old Dagu × Xianju chickens were used in this study. The chickens were maintained in plastic mesh floors (situated 80 cm aboveground) for 8 weeks. The chickens were given feed and water ad libitum. The temperature of the house was controlled at 35◦C during the 1st week and then decreased by 2◦C per week to a final temperature of 23◦C. Approximately eight weeks later, the chickens were weighed, and around 240 chickens with similar weights were randomly assigned to the control (C) and strain L1 (L) groups. Each group comprised three replicates with 40 birds (50% males and 50% females) per replicate. The chickens were raised in their cages (50 cm × 50 cm × 50 cm, situated 80 cm aboveground). The temperature of the chicken house was set at 23◦C. The chickens were given feed and water ad libitum (Xu Y. et al., 2016). The initial difference in body weight was not significant between the two groups (P > 0.05). The two groups were provided with the same basal diet (**Supplementary Table S1**) and subjected to similar environmental factors. The two groups were fed with mash diets, namely, the C group was given a basal diet and liquid medium, whereas the L group received a basal diet and 1 × 10<sup>6</sup> cfu of strain L1·g −1 . Chickens were weighed, and feed intake was recorded on the morning of days 91 and 112. The average daily gain (ADG), average daily feed intake (ADFI), and feed conversion ratio (FCR) were also calculated. When the chickens reached 16 weeks of age, these were each weighed. Three chickens that were deemed representative of the average weight were randomly picked out from each group and killed. The contents of the small intestine (the posterior duodenum and anterior jejunum) and cecum were collected, flash-frozen in liquid nitrogen, and used in DNA extraction and PCR analysis. The samples were classified into four groups as follows: the XC group (small intestines from group C), XL group (small intestine sample from group L), DC group (the cecum sample from group C), and DL group (the cecum sample from group L).

# Feed Preparation

The number of viable strain L1 cells was assessed using plate counting after culturing the cells in sweet potato juice medium at 30◦C for 24 h. We performed feed supplementation prior to each feeding as follows: approximately 10 mL of the strain L1 liquid culture (only liquid medium for C group) were thoroughly mixed with 1000 g of the corresponding diet to attain a strain L1 density of 1 × 10<sup>6</sup> cfu·g −1 after mixing (Wang et al., 2017).

# 16S rRNA Sequencing of Gut Microbes

Microbial genomic DNA was isolated from the cecal content samples with a TIANGEN DNA stool mini kit (TIANGEN, cat#DP328), following the manufacturer's recommendations. The variable region of 16S rRNA V3–V4 was PCR amplified with universal primers 338F (5<sup>0</sup> -ACTCCTACGGGAGGCAGCAG-3<sup>0</sup> ) and 806R (5<sup>0</sup> -GGACTACHVGGGTWTCTAAT-3<sup>0</sup> ) (Xu N. et al., 2016). The PCR products were purified with a QIAGEN quick gel extraction kit (QIAGEN, Cat # 28706). The PCR products from each sample were employed in the construction of a sequencing library with an Illumina TruSeq DNA sample preparation kit (the library was constructed with a TruSeq Library construction kit). For each sample, barcoded V3–V4 PCR amplicons were sequenced on an Illumina MiSeq PE300 platform.

Sequence reads were discarded when the sequence length was <150 bp, if the average Phred score was <20, if these contained ambiguous bases, if the homopolymer run >6, or if there were primer mismatches. The sequences that passed quality filtering were assembled using Flash<sup>1</sup> , which required an overlap of reads 1 and read of ≥10 bp, without any mismatches. We discarded any reads that could not be assembled. Chimera sequences were also discarded using UCHIME in MOTHUR (version 1.31.2<sup>2</sup> ). Amplification and sequencing of the 16S rRNA V3–V4 variable region was completed by Shanghai Majorbio Bio-pharm Technology Co., Ltd. (Shanghai, China).

# Operational Taxonomic Unit (OTU) Clustering

Sequence clustering was conducted with the uclust algorithm in QIIME<sup>3</sup> , and clustered into OTUs. The longest sequence in every cluster was chosen as the representative. The taxonomy of every OTU was assigned using BLAST-searching for the representative sequence in the Greengenes reference database (release 13.8<sup>4</sup> ). Unknown archaeal or eukaryotic sequences were filtered out. The Ace, Chao, Shannon, and Simpson indices were computed using the summary.single command in MOTHUR. A Venn diagram of between-group OTUs was constructed in R. We compared the relative abundance of OTUs or taxa among samples.

<sup>1</sup>http://ccb.jhu.edu/software/FLASH/

<sup>2</sup>http://www.mothur.org/

<sup>3</sup>http://qiime.org/scripts/pick\_otus.html

<sup>4</sup>http://greengenes.secondgenome.com/

#### Microbial Function Prediction

fphys-10-00937 July 24, 2019 Time: 14:55 # 5

Functional genes were predicted with PICRUSt based on the abundance at the OTU level. The OTUs were mapped to the gg13.5 database at a 97% similarity with the QIIME command "pick\_closed\_otus." OTU abundance was automatically normalized with the 16S rRNA gene copy numbers from known bacterial genomes of the Integrated Microbial Genomes database. The predicted genes and their function were annotated with the KEGG database, then differences among groups were compared to the free online platform Majorbio I-Sanger Cloud Platform<sup>5</sup> . Two-side Welch's t-test and Benjamini–Hochberg FDR correction were used in two-group analysis. The relative abundance of the KEGG metabolic pathways was designated as the metabolic profile.

#### Statistical Analysis

We statistically analyzed the diversity index data using ANOVA, and significant differences between group means were assessed using the Duncan test. Growth performance and abundance at the phylum and genus levels between groups were statistically evaluated with the t-test. Diversity indices and growth performance were expressed as the mean ± standard error (SE). We generated PCoA plots for sequence read abundance with Vegan as implemented in R. All statistical analyses were conducted using SPSS 16.0.

#### RESULTS

#### In vitro Characterization of Strain L1

Strain L1 was tolerant to acidic and biliary conditions. The viable counts slightly decrease upon exposure to low pH or high bile salt concentrations. After incubating for 3 h in a medium at pH 2, strain L1 had a survival rate of 98.73%, whereas after 4 h incubating in a medium supplemented with bile salts (0.5 g·mL−<sup>1</sup> ), it was 98.35%. These findings indicated that strain L1 can normally grow in these conditions with high viability (viability: 7.5–8 log cfu·mL−<sup>1</sup> ) (**Tables 1**, **2**). Hence, the strain could meet the concentration requirement of probiotics for use in animals.

<sup>5</sup>www.i-sanger.com

TABLE 1 | Counts of strain L1 exposed to different pH values for different time periods (n = 3).


The data are presented as the mean ± standard error. Means within a column followed by a different lowercase letter are significantly different (P < 0.05).

**Figure 1** shows the results of the NaCl tolerance test. Strain L1 exhibited good tolerance to 1–5% NaCl, with viability within the range of 10.24–10.26 log cfu·mL−<sup>1</sup> . After incubating for 24 h, the viable counts of strain L1 in an NaCl-containing medium decreased slightly compared with those in an NaCl-free medium. However, even at NaCl concentration of 5 g 100 mL−<sup>1</sup> , the cell survival rate was as high as 97.2%, with the viable counts in the NaCl-containing medium remaining above 10 log cfu mL−<sup>1</sup> . Thus, strain L1 is highly tolerant to salt, enabling it to withstand the adverse effects of high osmotic pressure in the highsalt environment of the gastrointestinal tract and maintain the relative balance of osmotic pressure under such conditions.

Strain L1 showed good autoaggregation and hydrophobicity properties. After 1, 2, and 24 h, the autoaggregation values were 15.8 ± 1.2%, 20.4 ± 2.3%, and 47.2 ± 0.8%, respectively. Further, at 560 nm, the surface hydrophobicity value was 38.1 ± 2.7%.

The adhesion of strain L1 onto Caco-2 cells was evaluated microscopically and by plate colony counting. Microscopic observation focused on observing the adhesion of cells, while the plate count was focused on quantifying the adherent cells. **Figure 2** shows that approximately 30 ± 2.8 strain L1 cells had adhered onto the surface of Caco-2 cells. In addition, plate colony

FIGURE 1 | Viable counts of strain L1 in the presence of different salt concentrations. Each column represents the mean of three replicates. The bars represent the standard deviation. Columns with same letter indicate no statistical significance (p > 0.05).

FIGURE 2 | Adhesion of strain L1 to Caco-2 cells. The cells were gram-stained and observed under a microscope imaging system (Olympus DP73) (×1000).


TABLE 2 | Counts of strain L1 exposed to different bile salt concentrations (w/v) for different time periods (n = 3).

The data are presented as the mean ± standard error. Means within a column followed by a different lowercase letter are significantly different (P < 0.05).

counts indicated showed that the adhesion rate of strain L1 onto the Caco-2 cells was 22.37 ± 1.44%.

Fourteen antibiotics from different families were investigated (**Table 3**). Strain L1 showed resistance to erythromycin and azithromycin, but sensitivity to other antibiotics tested. Importantly, it did not contain natural plasmid DNA (data not shown).

The decarboxylase activity of strain L1 was not observed as no culture medium color change was detected, which suggests that no biogenic amines were produced (data not shown). Furthermore, the strain did not show any α- and β-haemolytic activity when cultured on Columbia sheep blood agar, indicating that no hemotoxin was produced (**Figure 3**).

#### OTU Clustering and Annotation

The ability of strain L1 to act as a probiotic was next investigated through feeding experiments employing chickens fed a diet supplemented with strain L1. After 8 weeks of


<sup>∗</sup>Antibiotics: AMC, amoxicillin/clavulanic acid; P, penicillin G; CTX, cefotaxime; VA, vancomycin; TEC, teicoplanin; TE, tetracycline; ST, streptomycin; K, kanamycin; GM, gentamicin; E, erythromycin; AZM, azithromycin; CIP, ciprofloxacin; NOR, norfloxacin; RD, rifampicin. R, resistant; S, sensitive. Each experiment was repeated three times.

feeding, the intestinal contents of chickens in each group were collected. Microbial genomic DNA was extracted, followed by the amplification of the variable region of 16S rRNA V3–V4 and sequencing on an Illumina MiSeq PE300 platform. Trimmed and assembled sequences were then clustered at 97% similarity, as detailed in the Methods, and a total of 378 OTUs were identified by database alignment using BLAST-searching in QIIME. The following OTU numbers were obtained from each group: 245 in the XC group, 357 in the XL group, 247 in the DC group, and 247 in the DL group (**Figure 4**). **Figure 4A** shows 117 unique OTUs in XL group and 5 unique OTUs in XC group. The total richness in the X groups (XC and XL groups) was 364 OTUs, but it was 259 OTUs in the D groups (DC and DL groups). The number of OTUs in each group did not change in the D groups; however, this number increased in the X groups after feeding of strain L1. The microbial diversity in the X groups significantly changed. The Chao and Ace indices of the XL group significantly increased (P < 0.05) compared to the three other groups. No significant difference (P > 0.05) in the Shannon and Simpson of the small intestines was observed between the XC and XL groups; the same trend was observed in the cecum. These findings indicated that the richness of the small intestinal microbes in the XL group was greater compared with the three other groups (**Table 4**).

#### Differences in the Growth Performance and Intestinal Microbiota in Chicken Associated With the Feeding of Strain L1

In the current study, the growth performance of chickens in different groups was obviously different. At the first stage of the experiment (9–13 weeks), the ADFI of chickens in group L was relatively higher (P < 0.05) compared with group C. At the second stage of the experiment (14–16 weeks), group L show better ADG and FCR than that of group C (P < 0.05) (**Table 5**).

Twelve phyla were shared by the 12 samples. Firmicutes (>58%) were the dominant bacteria in the small intestine. Firmicutes (>43%) and Bacteroidetes (>38%) were the dominant bacteria in the cecum. Feeding strain L1 greatly impacted the composition of small intestinal microbiota. As shown in **Figure 5**, the feeding of strain L1 decreased the proportion of Proteobacteria, Actinobacteria, and Tenericutes in the small intestine (P > 0.05), increased the proportion of Firmicutes (P > 0.05) (**Figure 5A**), and significantly increased the proportion of Bacteroidetes and Synergistetes (P < 0.05) (**Figure 5B**). **Figures 5A,C** show that feeding of strain L1 resulted in minimal effects on the cecal microbiota at the phylum level.

surrounding the colonies (β-hemolysis).

At the genus level, we detected 130 genera. Lactobacillus (>50%) was the dominant bacterium in small intestine, while Bacteroides (> 18%) and Faecalibacterium (>13%) were the dominant bacteria in the cecum. Feeding strain L1 resulted in an increase in the proportion of Lactobacillus (P > 0.05), Bacteroides (P < 0.05), and Faecalibacterium (P < 0.05), and decreased the proportion of Ureaplasma, Helicobacter, and Enterococcus (P > 0.05) in the small intestine. The cecal abundance of norank\_f\_Bacteroidales\_S24-7\_group had increased (P < 0.05), and that of Bacteroides was significantly (P < 0.05) lower with strain L1 feeding (**Figure 6**).

Principal coordinate analysis indicated differences in microbial distribution among the four groups. The distribution markedly differed among groups that had or had not been fed with strain L1 (**Figure 7**). One group of microorganisms predominated in the L groups (XL and DL groups), whereas another predominated in the C groups (XC and DC group). Correlation analysis showed that the small intestine microbiota in the XC group varied from those in the XL group (0.640). However, the cecal microbiota in the DC group were the same as those in the DL group (0.912) (**Table 6**). These observations demonstrated that strain L1 greatly impacted the microbiota of the small intestine but had little effect on cecal microbiota.

Microbial functional analysis using PICTUSt was performed to assess the functions of the microbiota in the C and L groups. At KEGG level 2, in the cecum, no differences in gene abundances between the experimental and control groups were apparent (**Supplementary Figure S1**). However, in the small intestine,


The data are presented as the mean ± standard error. Means with the same superscript letter within the same row are not significantly different. Different lowercase letters indicate significance at P < 0.05.



The data are presented as the mean ± standard error. Means within the same row denoted by different lowercase letters are significantly different at P < 0.05. <sup>1</sup>ADFI, average daily feed intake; ADG, average daily gain; FCR, feed conversion ratio (ADFI/ADG).

significant differences in the abundances of six functional genes were observed (**Figure 8A**). The small intestine microbiota in the XL group exhibited a wider range of functions that are involved in metabolic pathways, which include enzyme families (P < 0.05) as well as the digestive system (P < 0.01), than those in the XC group. At KEGG level 3, in the small intestine, significant differences in the abundance of 49 genes were noted between the XL and XC groups (**Figure 8C**). The small intestine microbiota in the XL group exhibited greater abundance of functions that were involved in carbohydrate and protein metabolic pathways, including metabolism of starch and sucrose, fructose, and mannose, amino sugars, and nucleotide sugars, degradation of other glycans, insulin signaling pathway, glycosaminoglycan degradation, digestion and absorption of carbohydrates, and digestion and absorption of proteins, than those in the XC group. In the cecum, significant differences in abundance of only seven genes were noted (**Figure 8B**). Cecal microbiota in the DL group exhibited greater abundance of functions that were involved in metabolic pathways, including fatty acid biosynthesis, glycerolipid metabolism, and propanoate metabolism, than those in the DC group.

#### DISCUSSION

The present study assessed the in vitro features of a potential probiotic, L. paracasei strain L1, and evaluated its weightstimulating ability using chicken-feeding experiments. Our results showed that strain L1 may indeed be considered a good probiotic candidate.

To be qualified as a probiotic, the microbial candidate must possess specific functional and safety properties, including acid and bile salt tolerance, adhesion capacity, haemolytic activity, and susceptibility to antibiotics (Dowarah et al., 2018). Acid and bile salt tolerance is an essential criteria in identifying probiotic

strains, as these influence their survival in the gastrointestinal tract (Saarela et al., 2000). During their passage through the stomach, these probiotic microbes need to survive in as low TABLE 6 | Correlation of genus abundance between groups.


Three samples from each group were used to calculate the correlation.

as pH 3 before reaching the lower digestive tract, and also must remain viable for 4 h or more (Ouwehand et al., 1999). Consequently, strain L1 exhibited good probiotic features as it showed considerable growth at acidic pH (2.0). Du Toit et al. (1998) revealed that L. reuteri BFE1058 and L. johnsonii BFE1061 isolated from pig fecal material can grow at low pH levels for 6 h at 37◦C (Du Toit et al., 1998). Cultures using L. lactis and Enterococcus faecium have better tolerance to low pH than Lactobacillus casei and Pediococcus acidilactici, and thus they were fed as probiotics to weaned piglets (Guerra et al., 2007). For an effective probiotic culture, LAB should remain viable in the presence of 0.3% bile salts. L. plantarum ZlP001 isolate from the gastrointestinal tract of a weaned piglet exhibited 85.3, 61.4, and 9.4% tolerance to growth medium

supplemented with 0.1, 0.3, and 0.5% bile salts, respectively (Wang et al., 2011). Here, strain L1 exhibited 99.8, 99.2, and 98.3% tolerance upon respective exposure to 0.03, 0.3, and 0.5% bile salts for 4 h. The adaptation to bile salts has been shown to be related to alterations in carbohydrate fermentation and glycosidase activity (Taheri et al., 2009); exopolysaccharide production (Petsuriyawong and Khunajakr, 2011; Shazali et al., 2014); the composition of membrane proteins and fatty acids (Fairbrother et al., 2005); and enhanced adhesion to human mucus as well as inhibition of pathogen adhesion (Cho et al., 2009; Venkatesan et al., 2012). Further, strain L1 showed high tolerance to all NaCl concentrations tested (**Figure 1**), which can enable it to withstand the adverse effects of high osmotic pressure in the high-salt environment of the gastrointestinal tract and maintain a relative balance of osmotic pressure under these conditions.

Hydrophobicity is essential to adhesion to enterocytelike cells and autoagglutination. Adhesive strains exhibit high levels of hydrophobicity, and the extent of adherence apparently depends on the surface potential (Pérez et al., 1998; Juarez Tomas et al., 2005). Al Kassaa et al. (2014) revealed that CMUL57 (Lactobacillus gasseri), CMUL67 (Lactobacillus acidophilus), and CMUL140 (Lactobacillus plantarum) were most hydrophobic strains among ones screened by the authors; interestingly, these strains also exhibit the greatest autoaggregation ability (Al Kassaa et al., 2014). Adherence of lactobacilli onto epithelial cells and biofilm formation has been associated with cell autoaggregation and surface hydrophobicity (Dunne et al., 2001). In the current study, strain L1 showed good autoaggregation and hydrophobicity abilities. At 560 nm, its surface hydrophobicity value was 38.1 ± 2.7%. Further, the autoaggregation ability increased notably with incubation time. After 1, 2, and 24 h, the autoaggregation values were 15.8 ± 1.2%, 20.4 ± 2.3%, and 47.2 ± 0.8%, respectively. Shekh et al. (2016) previously described the autoaggregation of 10 LAB strains. Autoaggregation after 2 h, 4 h, and 24 h was respectively within the range of 1–19%, 5–46%, and 33–92% (Shekh et al., 2016). Das et al. (2016) reported the hydrophobicity values of L. casei SB71, SB73, and SB93 as 22.2 ± 0.8%, 22 ± 1.5%, and 25 ± 2.5%, respectively (Das et al., 2016). Hence, the hydrophobicity of

strain L1 was higher than that described in earlier studies (Pelletier et al., 1997; Vinderola and Reinheimer, 2003).

An important feature of probiotics is their ability to adhere to the intestinal epithelial layer that prevents their elimination through peristalsis. Furthermore, adhesion is a prerequisite for colonization (Forestier et al., 2001) and influences the competitive exclusion of enteropathogens (Lee et al., 2003), stimulation of the immune system (Schiffrin et al., 1995), as well as antagonistic activity against enteropathogens (Coconnier et al., 1993). The adhesion ability of Bifidobacterium and Lactobacillus strains differs with the in vitro method employed (Laparra and Sanz, 2009). Here, Caco-2 cells were employed to study the adhesion of strain L1 cells. Plate colony counting showed that the adhesion rate of strain L1 onto Caco-2 cells was 22.37 ± 1.44%. Pisano et al. (2014) revealed that these strains can adhere onto Caco-2 cells to various degrees (ranging from 3 to 20%), thus confirming that adhesion is strain-specific (Pisano et al., 2014). Although the findings of the in vitro studies are not directly utilized in in vivo situations, these support the association between the two factors (Crociani et al., 1995).

One of the properties that are crucial for identifying LAB as potential probiotics is their safety for human consumption. The antibiotic susceptibility of strain L1 to 14 antibiotics was evaluated by disc diffusion on MRS agar plates (**Table 3**). Strain L1 exhibited susceptibility to most antibiotics tested, as well as resistance to erythromycin and azithromycin. These results are with the findings involving L. plantarum and L. paracasei strains (Lavilla-Lerma et al., 2013; Solieri et al., 2014), although other research detected variations in resistance to tetracycline (Georgieva et al., 2008; Comunian et al., 2010). The observed resistance to certain antibiotics indicates that strain L1 will not be affected by therapies involving these antibiotics and thus may facilitate in maintaining the natural balance of intestinal microflora while undergoing antibiotic treatment (Salminen et al., 1998). Natural bacterial resistance to antibiotics is not deemed as a risk to the health of animals or humans (Al Kassaa et al., 2014). Abriouel et al. (2015) analyzed patterns of phenotypic and genotypic antibiotic resistance in Lactobacilli, and identified antibiotic resistance genes in the bacterial chromosome indicative of non-transferable and intrinsic resistance. Lactobacilli does not carry acquired or transmissible antibiotic resistance genes (Vizoso Pinto et al., 2006). Importantly in that context, we determined here that strain L1 does not harbor natural plasmid DNA.

The production of biogenic amines is a crucial safety criterion in the selection of probiotic strains because amines could cause health problems (Lorenzo et al., 2010). The present study has determined that strain L1 cells do not generate biogenic amines. In fact, Lactobacillus strains are actually considered as safe organisms particularly in terms of biogenic amine production (Arena and Manca de Nadra, 2001). Martin et al. (2005) assessed the ability of two lactobacillus strains, namely, L. gasseri and L. fermentum in producing biogenic amines and determined that none of these generate these compounds (Martin et al., 2005).

The community structure and activity of the gut microbiota co-evolve with the host from birth, and are exposed to various activities of the host genome, nutrition, as well as lifestyle. The gut microbiota regulates multiple host metabolic pathways, which results in interactive host-microbiota metabolic, signaling, and immunoinflammatory axes that physiologically link the gut, liver, muscle, as well as brain. Feeding probiotics such as Lactobacillus can increase the content of beneficial microorganisms (Lactobacillus and Bifidobacterium) in the intestinal tract, and inhibit the potential pathogenic microorganisms (Salmonella and Escherichia coli) to improve the intestinal microecological environment (Samuel et al., 2008; Nicholson et al., 2012). Angel et al. (2005) reported that L. acidophilus can improve the growth performance of broilers by improving the intestinal flora. Including L. acidophilus in the diet increases the Lactobacillus content in the ileum and cecum of broiler chicken, while the content of potential pathogenic bacteria such as E. coli decreases (Salarmoini and Fooladi, 2011). Further, Lactobacillus content increased in the feces of 1-day-old broilers after they had been fed a basic diet containing L. plantarum and its metabolites (Thanh et al., 2009). The present study revealed that strain L1 improves chicken growth performance and altered the composition of its intestinal microflora. Supplementation of the diet using strain L1 markedly influenced small intestinal microbial composition.

Gene products of the intestinal microflora provide enzymatic and biochemical pathways for multiple metabolic processes in the host. Turnbaugh and Gordon (2009) reported that a series of core genes encoded by the microbial genome might play a regulatory role in the host energy metabolism. For example, obese individuals carry more genes for the digestion of fats, proteins, and carbohydrates, thereby facilitating in the absorption and storage of energy from the diet compared to lean individuals (Turnbaugh and Gordon, 2009). Probiotic feeding enhances growth performance and immunity responses. The maturation of the intestinal microbiota significantly improved by probiotic feeding, yet is markedly delayed by antibiotic feeding. Probiotic feeding may thus be an intestinal health-promoting factor that may feeding efficiency during growth (Gao et al., 2017).

Gut microbiota consists of approximately 600,000 genes, i.e., 25-fold more than the number of genes in the host genome. Therefore, gut microbiota are often considered as a host organ that serves as a gut microecosystem (Lederberg, 2000; Qin et al., 2010). Such microecosystem can perform numerous metabolic functions that vary with microbiota composition. In the present study, numerous functions were determined to be involved in metabolic pathways, including the metabolism of amino acids, nucleotides, carbohydrates, energy, lipids, replication and repair, cofactors, and vitamins. All of these are probably related to the changes in the composition of gut microbiota. These findings indicate that strain L1 supplementation might promote growth performance and improve the intestinal microflora in chicken. However, additional studies are needed to confirm this.

#### CONCLUSION

Lactobacillus paracasei subsp. paracasei L1 possesses probiotic properties such as adhesion, aggregation, hydrophobicity, as well as survival upon exposure to various gastrointestinal conditions, and lack hemolytic and decarboxylation activities. Furthermore, feeding experiments revealed that strain L1 may increase the abundance of functions related to carbohydrate and protein metabolism, and fatty acid biosynthesis in the intestinal microbiota, and improve the growth performance of chicken.

#### DATA AVAILABILITY

fphys-10-00937 July 24, 2019 Time: 14:55 # 12

The datasets generated for this study can be found in the NCBI Sequence Read Archive (http://www.ncbi.nlm.nih.gov/ Traces/sra/) under the SRA Accession Number: SRP187013.

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of Guidelines for the Care and Use of Laboratory Animals, the Beijing Association for Laboratory Animal Science. The protocol was approved by the Animal Ethics Committee of the Institute of Zoology, Chinese Academy of Sciences.

#### AUTHOR CONTRIBUTIONS

LZ, YT, YX, YMT, TW, and YS performed all the experiments and wrote the manuscript. JL, HG, CW, and YC conducted the

#### REFERENCES


experiments and data analysis. All authors read and approved the final version of the manuscript.

#### FUNDING

The authors gratefully acknowledge the financial support from the Natural Science Foundation of Liaoning Province (Grant No. 20180550856), the Liaoning Province Key Research and Development Project (Grant No. 2018225027), the National Natural Science Foundation of China (Grant No. 31301499), and the First Leader of Animal Husbandry Industry Innovation Team of Jinzhou Medical University (Grant No. JYLJ201705).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphys. 2019.00937/full#supplementary-material

FIGURE S1 | Mean proportion of the cecal intestinal microbiota and the differences in predicted functional metagenomes. Comparison of the functional pathways of microbes from the DC and DL groups at KEGG level 2 is shown.

TABLE S1 | Composition of basal diets for chickens.

invasion. FEMS Microbiol. Lett. 110, 299–305. doi: 10.1111/j.1574-6968.1993. tb06339.x


its improvement with the supplementation of prebiotics. Int. J. Pl. An. Environ. Sci. 2, 94–106.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Xu, Tian, Cao, Li, Guo, Su, Tian, Wang, Wang and Zhang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Fecal Metaproteomic Analysis Reveals Unique Changes of the Gut Microbiome Functions After Consumption of Sourdough Carasau Bread

#### Edited by:

*Yuheng Luo, Sichuan Agricultural University, China*

#### Reviewed by:

*Maurizio Sanguinetti, Catholic University of the Sacred Heart, Italy Monica Di Paola, University of Florence, Italy*

\*Correspondence: *Sergio Uzzau uzzau@portocontericerche.it*

*†These authors have contributed equally to this work*

#### ‡Present address:

*Alessandro Tanca, Department of Biomedical Sciences, University of Sassari, Sassari, Italy*

#### Specialty section:

*This article was submitted to Systems Microbiology, a section of the journal Frontiers in Microbiology*

Received: *12 April 2019* Accepted: *15 July 2019* Published: *30 July 2019*

#### Citation:

*Abbondio M, Palomba A, Tanca A, Fraumene C, Pagnozzi D, Serra M, Marongiu F, Laconi E and Uzzau S (2019) Fecal Metaproteomic Analysis Reveals Unique Changes of the Gut Microbiome Functions After Consumption of Sourdough Carasau Bread. Front. Microbiol. 10:1733. doi: 10.3389/fmicb.2019.01733* Marcello Abbondio1†, Antonio Palomba2†, Alessandro Tanca2‡, Cristina Fraumene<sup>2</sup> , Daniela Pagnozzi <sup>2</sup> , Monica Serra<sup>3</sup> , Fabio Marongiu<sup>3</sup> , Ezio Laconi <sup>3</sup> and Sergio Uzzau1,2 \*

*<sup>1</sup> Department of Biomedical Sciences, University of Sassari, Sassari, Italy, <sup>2</sup> Porto Conte Ricerche, Science and Technology Park of Sardinia, Alghero, Italy, <sup>3</sup> Department of Biomedical Sciences, University of Cagliari, Cagliari, Italy*

Sourdough-leavened bread (SB) is acknowledged for its great variety of valuable effects on consumer's metabolism and health, including a low glycemic index and a reduced content of the possible carcinogen acrylamide. Here, we aimed to investigate how these effects influence the gut microbiota composition and functions. Therefore, we subjected rats to a diet supplemented with SB, baker's yeast leavened bread (BB), or unsupplemented diet (chow), and, after 4 weeks of treatment, their gut microbiota was analyzed using a metaproteogenomic approach. As a result, diet supplementation with SB led to a reduction of specific members of the intestinal microbiota previously associated to low protein diets, namely *Alistipes* and *Mucispirillum*, or known as intestinal pathobionts, i.e., *Mycoplasma*. Concerning functions, asparaginases expressed by *Bacteroides* were observed as more abundant in SB-fed rats, leading to hypothesize that in their colonic microbiota the enzyme substrate, asparagine, was available in higher amounts than in BB- and chow-fed rats. Another group of protein families, expressed by *Clostridium*, was detected as more abundant in animal fed SB-supplemented diet. Of these, manganese catalase, small acid-soluble proteins (SASP), Ser/Thr kinase PrkA, and V-ATPase proteolipid subunit have been all reported to take part in *Clostridium* sporulation, strongly suggesting that the diet supplementation with SB might promote environmental conditions inducing metabolic dormancy of *Clostridium* spp. within the gut microbiota. In conclusion, our data describe the effects of SB consumption on the intestinal microbiota taxonomy and functions in rats. Moreover, our results suggest that a metaproteogenomic approach can provide evidence of the interplay between metabolites deriving from bread digestion and microbial metabolism.

Keywords: gut microbiota, metagenomics, metaproteomics, food processes, sourdough, diet

# INTRODUCTION

Among bakery products, bread is the most abundantly consumed food worldwide, with an increase in demand for products containing wholegrain, high in fiber, or obtained through "health-promoting" processing, such as sourdough leavening. Use of sourdough has been shown to improve flavor, structure, and shelf life of baked bread, due to its differences in chemical and physical features compared to baker's yeast leavening (Gobbetti et al., 2016).

Further, cereal fermentations are widely recognized as of great potential in improving the nutritional quality of food ingredients and their healthy effects. A number of studies claimed that specific cereal matrix and/or the bakery processes used to produce bread might lead to the improvement of clinical parameters in habitual consumers (Korem et al., 2017). Sourdough leavening actively retards starch digestibility, leading to low glycemic responses, and may increase the production of non-digestible polysaccharides that escape the small intestine, together with grain fibers, eventually feeding the colonic microbiota (Maioli et al., 2008; Scazzina et al., 2009; Sanna et al., 2018). To this end, selected species of lactic acid bacteria (LAB) have been tested with the aim of improving bread quality (De Vuyst et al., 2014). Also, sourdough leavening modulates levels and bioaccessibility of bioactive compounds, and improves mineral bioavailability (Di Nunzio et al., 2018).

Different enzymatic activities in sourdough and baker's yeast fermentation might be responsible for specific hydrolysis of proteins and polysaccharides. Protein hydrolysis, in turn, may affect the absorption of bioactive compounds as well as other metabolites impacting on the host physiology. Sourdough has also been proposed to yield bread with highly degraded gluten that may be appropriate for gluten intolerant individuals (i.e., with non-celiac gluten sensitivity) (Gobbetti et al., 2018a,b).

Strikingly, sourdoughs fermentation has also been demonstrated to reduce the acrylamide content in wheat bread (Bartkiene et al., 2013). As for other food products, factors affecting acrylamide formation during bread production are acrylamide precursors (mainly asparagine), reducing sugars and specific processing conditions. These sourdough features have important practical implications, since acrylamide neurotoxicity, genotoxicity, carcinogenicity, and reproductive toxicity have been demonstrated (Keramat et al., 2011), and bakery products account for around 20% of human exposure to acrylamide.

Based on these premises, this study was designed to gain insights into the complex interplay between sourdough effects on bread preparation and the gut microbiota (GM) taxonomy and functional activities, and, in turn, to elucidate its possible impact on consumer's metabolism and health. To date, the specific contribution of sourdough bread consumption to the functional activities of microbiota has not been evaluated. Therefore, we compared the composition and the active functions of microbial intestinal communities in three groups of rats fed a diet supplemented with sourdough bread (SB), baker's yeast leavened bread (BB), or unsupplemented diet. Specifically, we choose to evaluate the impact of sourdough consumption in rats fed a calorie-restricted diet, based on low fat high fiber composition, to avoid GM modifications that have been already associated to high fat and/or high sugars obesogenic diets. With this approach, we also minimized the potential confounding effects due to individual difference in food intake, generally occurring in animals fed ad libitum (AL).

# MATERIALS AND METHODS

#### Animals and Samples

A total of 16 Fischer 344 rats (10 weeks old, male) were purchased from Charles River Laboratories Italia, SRL (Calco, Italy) together with the manufacturer's animal chow VRF1 (P) 811900 (4.5% of fat). Animals were distributed two per cage and maintained on daily cycles of alternating 12 h light-darkness (light on at 11 p.m., light off at 11 a.m.), with food and water available AL. Animal studies were reviewed and approved by the Institutional Animal Care and Use Committee of the University of Cagliari and were performed in accordance with the relevant guidelines and regulations (authorization of the Italian Health Ministry No. 840/2016-PR). After 2 weeks of acclimatization, rats were divided in four groups of four rats each and were exposed to the following feeding schedule. The first group was continued on AL chow diet ("chow-AL" group), while the other three groups were fed a calorie-restricted (CR) diet, calculated as 70% of AL food intake, as previously reported (Fraumene et al., 2018; Tanca et al., 2018). Among CR-fed rats, one group received only laboratory chow ("chow" group), while the remaining two groups were supplemented (15% w/w) with a typical Sardinian bread (carasau bread, produced by a local bakery company), leavened with BB ("BB" group) or SB ("SB" group), respectively.

Animals were weighed weekly and sacrificed after 4 weeks of treatment with their respective dietary regimens.

Glycemia was measured with Glucose Analyzer II (Beckman Coulter, Brea, CA, USA). Blood samples were taken from the tail vein 1 h prior to food delivery or 2 h after food delivery.

Stool, liver, and colonic content samples were collected from CR-fed rats after 4 weeks of diet treatment, whereas AL-fed rats were used merely as a growth control. Fecal samples were collected from all animals, apart from one rat belonging to the "chow" group. Colonic content and liver samples were collected from all animals after sacrifice. All samples were immediately stored at −80◦C until use. At the time of the analyses, stool samples were thawed at 4◦C and two portions were collected from each of them for protein and DNA extraction, respectively; colonic contents were directly processed for DNA extraction, whereas liver samples were directly processed for protein extraction.

# DNA Extraction and 16S rRNA Gene Sequencing

DNA was extracted from 11 fecal samples and 12 colon content samples, collected from rats belonging to "chow", "BB", and "SB" groups. Extraction was performed according to QIAamp Fast Stool Kit protocol (QIAGEN, Hilden, Germany). The extracted DNA was purified according to E.Z.N.A. <sup>R</sup> Soil DNA Kit (Omega Bio-Tek, Norcross, GA, USA). DNA quality and yield were evaluated via agarose gel and Qubit fluorometer (Life Technologies, CA, USA). Libraries were constructed using Illumina's recommendations as implemented in 16S Metagenomic Sequencing Library Preparation guide. To amplify the variable region 4 of the 16S rRNA gene, we used the 515F and 806R primers (GTGCCAGCMGCCGCGGTAA and GGACTACHVGGGTWTCTAAT, respectively) modified to contain adaptors for MiSeq sequencing. Two separate gene amplification reactions were performed for each sample, pooled together and cleaned up using AMPure XP (Beckman Coulter) magnetic beads. The next PCR attached dual index barcodes using the Illumina Nextera XT kit so that the PCR products may be pooled and sequenced directly. The final quality control and quantification of the libraries were conducted using a Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA). DNA sequencing was performed on the Illumina MiSeq platform, using v3 chemistry according to the manufacturer's specifications, to generate paired-end reads of 201 bases in length in each direction. Data quality control and analyses were performed using the QIIME pipeline (v.1.9.1) (Caporaso et al., 2010). The overlapping paired-end reads were merged using the script join\_paired\_ends.py inside the QIIME package. OTUs generation was done using a pipeline based on USEARCH's OTU clustering recommendations (http://www.drive5.com/usearch/ manual/otu\_clustering.html) using the closed-reference OTU picking to allow clustering of 16S rRNA gene sequences, as previously described (Tanca et al., 2017). Reads were clustered at 97% identity using UCLUST to produce OTUs (Edgar, 2010). Taxonomy was then assigned using the Greengenes 13\_8 database (DeSantis et al., 2006).

#### Protein Extraction and Proteomic Analysis

Eleven fecal samples and 12 liver samples, collected from rats belonging to "chow," "BB," and "SB" groups, were subjected to bead-beating and heating/freezing steps after resuspension in an SDS-based reducing extraction buffer, as described earlier (Tanca et al., 2014). Protein extracts were cleaned up, alkylated, and trypsin digested according to the filter-aided sample preparation procedure (Wisniewski et al., 2009), with minor modifications illustrated elsewhere (Tanca et al., 2013, 2015).

Liquid chromatography (LC)-tandem mass spectrometry (MS/MS) analyses were performed on an LTQ Orbitrap Velos mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA), operating with an EASY-spray source, interfaced with an UltiMate 3000 RSLCnano LC system (Thermo Fisher Scientific). Samples were run in a randomized order. After loading, peptide mixtures (4 µg per run) were loaded, concentrated, and desalted on a trapping pre-column (Acclaim PepMap C18, 75µm × 2 cm nanoViper, 3µm, 100 Å, Thermo Fisher Scientific), using 0.2% formic acid at a flow rate of 5 µl/min. The peptide separation was performed with a C18 EASY-spray column (PepMap RSLC C18, 75µm × 50 cm, 2µm, 100 Å, Thermo Fisher Scientific) at 35◦C with a flow rate of 250 nL/min for 247 min, using the following two-step gradient of eluent B (0.2% formic acid in 95% ACN) in eluent A (0.2% formic acid in 5% ACN): 2.5–37.5% for 242 min and 37.5–99% for 5 min.

The mass spectrometer was set up in a data dependent MS/MS mode, where a full scan spectrum (from 375 to 2,000 m/z) is followed by MS/MS spectra, under direct control of the Xcalibur software. The instrument operated in positive mode. The temperature of ion transfer capillary and the spray voltage were set to 250◦C and 1.85 kV, respectively. Full scans and MS/MS spectra were acquired in the Orbitrap with resolutions of 60,000 and 7,500 at 400 m/z, respectively. The automatic gain control was set to 1,000,000 ions, and the lock mass option enabled on a protonated polydimethylcyclosiloxane background ion as internal recalibration for accurate mass measurements (Olsen et al., 2005). Peptide ions were selected as the 10 most intense peaks of the previous scan; the signal threshold for triggering an MS/MS event was set to 500 counts, and dynamic exclusion was set to 30 s. Higher-energy collisional dissociation was used as the fragmentation method, by applying a 35% value for normalized collision energy, an isolation width of m/z 3.0, a Q-value of 0.25, and an activation time of 0.1 ms. Nitrogen was used as the collision gas.

Microbial peptide identification was carried out using the Proteome Discoverer informatic platform (version 2.0; Thermo Fisher Scientific), with Sequest-HT as search engine and Percolator for peptide validation (FDR < 1%). Search parameters were set as follows: precursor mass threshold 350–5,000 Da; minimum peak count 6; signal-to-noise threshold 2; enzyme trypsin; maximum missed cleavage sites 2; peptide length range 5–50 amino acids; precursor mass tolerance 10 ppm; fragment mass tolerance 0.02 Da; dynamic modification methionine oxidation; static modification cysteine carbamidomethylation. Two parallel processing nodes were used. The first processing node was built on a combination of three microbial sequence databases: (i) a collection of metagenomic sequences obtained in house from rat fecal samples (11,510,359 sequences in total) and processed according to previous reports (Tanca et al., 2016); (ii) a publicly available mouse metagenomic dataset (ftp://penguin. genomics.cn/pub/10.5524/100001\_101000/100114/Genecatalog/ 184sample\_2.6M.GeneSet.pep.gz) (Xiao et al., 2015) merged with a collection of metagenomic sequences obtained in house from mouse fecal samples (9,825,357 sequences in total); (iii) a pseudometagenome comprising all UniProtKB (release 2017\_11) sequences belonging to the 54 microbial genera (NCBI taxonomy IDs: 157, 270, 286, 434, 469, 475, 816, 838, 841, 872, 970, 1253, 1263, 1279, 1301, 1386, 1485, 1578, 1654, 1678, 1716, 1883, 2093, 2152, 29407, 33024, 33042, 33926, 35832, 40544, 51514, 82373, 86331, 119852, 121871, 129337, 150247, 174708, 189330, 191303, 207244, 239759, 239934, 248038, 283168, 346096, 375288, 416916, 447020, 497726, 572511, 574697, 577309, 869896) detected in the 16S rRNA analysis described in this study with an abundance > 0.1% in at least one sample (17,389,183 sequences in total). All microbial sequence databases have been deposited in PRIDE along with MS data. The second processing node was built on a database containing the protein sequences belonging to the order Rodentia and deposited in UniProtKB/SwissProt (release 2017\_11; 26,656 sequences in total). Liver samples were subjected to the second processing node only.

Taxonomic and functional annotation was performed using multiple strategies. MEGAN v.6.8.19 was used as first annotation option (Huson et al., 2016). Protein sequences were preliminary subjected to a DIAMOND (v.0.8.22) search against the NCBI-nr database (2017/09 update), using the blastp command with default parameters (Buchfink et al., 2015); then, DIAMOND outputs were loaded on MEGAN and lowest common ancestor (LCA) classification was performed using default parameters. Furthermore, the Unipept web application (v.3.3.4; https://unipept.ugent.be) was used to carry out an LCA classification of the identified peptide sequences (Mesuere et al., 2017). Functional annotation was accomplished by aligning the identified protein sequences against a database containing all bacterial sequences from UniProtKB/Swiss-Prot (release 2017\_09) using DIAMOND (blastp module, e-value threshold 10−<sup>5</sup> ); UniProtKB/Swiss-Prot accession numbers were subsequently exploited to retrieve protein family information from the UniProt website via the "retrieve" tool (Pundir et al., 2016). Metaproteomic spectral count data obtained for each sample were aggregated based on the functional and taxonomic annotation levels, generating abundance tables of family-specific and genus-specific protein families.

#### Statistical Analysis and Graph Generation

Differential analysis was performed on read (16S rRNA gene sequencing) and spectral (metaproteomics) count data using the edgeR package available in a Galaxy server (https://bioinfgalaxian.erasmusmc.nl/galaxy) (Robinson et al., 2010). The pvalue lists provided by edgeR were subsequently subjected to a multiple testing adjustment based on a sequential goodness of fit (SGoF) metatest (Carvajal-Rodriguez et al., 2009) using the SGoF+ software (v.3.8) with default parameters (Carvajal-Rodriguez and de Una-Alvarez, 2011). An adjusted p-value < 0.05 was considered as the threshold for statistical significance of differential results. Features with missing value(s) in more than one group were filtered out. Beta diversity among groups was inspected by performing principal coordinate analysis (PCoA) and permutational multivariate analysis of variance (PERMANOVA) on taxonomic data at the genus level, using the web application MicrobiomeAnalyst (http://www.microbiomeanalyst.ca) (Dhariwal et al., 2017). In addition, using GraphPad Prism (v.5.03), we performed oneway analysis of variance (ANOVA) following by Bonferroni comparison on all pairs of groups (alpha-value = 0.05) on body weight and glycemia data, and Kruskal-Wallis test followed by Dunn's multiple comparison test (alpha-value = 0.05) on alpha-diversity data (Simpson and Shannon indexes), in order to evaluate the significance of variation among groups.

Heatmaps were generated starting from relative abundance data using the web application Morpheus (https://software. broadinstitute.org/morpheus), while line graphs were generated with GraphPad Prism.

# RESULTS

#### Experimental Design and General Metrics

In this study we investigated the effect of chow diet supplementation with Sardinian typical carasau bread leavened with standard baker's yeast or sourdough on the GM of young rats. To this end, animals were fed under a CR regimen with three different diets, namely, chow only, chow plus BB or chow plus SB; in addition, a fourth control group ("chow-AL") was fed AL.

Animal weights were recorded weekly during the dietary treatment. As expected, a significant difference was observed starting from the first week between the "chow-AL" control group and the three CR-fed groups (one-way ANOVA plus Bonferroni's comparison for multiple testing, **Supplementary Figure 1**).

Glycemia was also evaluated in each group after 1, 3, and 4 weeks, and values were measured both 1 h before and 2 h after the meal. Pre-meal values were significantly higher in chow-ALfed rats compared to the CR-fed groups after 1, 3, and 4 week of dietary treatment, with the exception of "BB" group at 4 weeks (one-way ANOVA plus Bonferroni's comparison for multiple testing, **Supplementary Figure 2A**); in contrast, no significant differences were observed among groups in post-meal glycemia values (**Supplementary Figure 2B**).

We then focused on the effect of these dietary treatments on the GM of CR-fed rats, according to the following considerations: (i) overfeeding by AL food consumption is a significant uncontrolled variable that might affect the total intake of bread provided with the diet; (ii) rather, rats fed with a CR regimen consume the whole feed before the following administration; (iii) the outcomes of the study are intended to be translated to normal-weight or lean individuals. To this aim, we collected stool, colon content, and liver samples from all experimental groups after 4 weeks of dietary exposure. Stool samples were subjected to both 16S rRNA gene sequencing and metaproteomic analysis, colonic contents to 16S rRNA gene sequencing only, and liver samples to host functional profile characterization only.

A total of 554,155 reads were obtained from fecal samples (50,378 on average per sample), and 469,049 reads from colon contents (39,087 on average per sample), corresponding to 130 microbial families and 160 microbial genera/species (**Supplementary Dataset 1**). In addition, a total of 96,913 microbial peptide-spectrum matches were obtained from fecal samples (8,810 on average per sample), corresponding to 1,850 microbial family-specific protein families and 1,882 microbial genus-specific protein families. Moreover, 32,416 (2,947 on average per sample) and 304,095 (25,341 on average per sample) host peptide-spectrum matches were identified from stool and liver samples, respectively, corresponding to 1,190 and 2,870 proteins (**Supplementary Dataset 2**). Differential analysis comparing host protein expression among dietary groups in stool and liver samples did not show any significant difference in abundance (data not shown).

Alpha and beta diversity among groups were evaluated considering genera/species, according to 16S rRNA gene sequencing (stool and colonic contents) and metaproteomic (stool only) data. No significant differences were observed for alpha diversity, according to Kruskal-Wallis test followed by Dunn's multiple comparison test on both Simpson and Shannon indexes (**Supplementary Table 1**); moreover, no significant clustering was detected based on PCoAs (PERMANOVA test, **Supplementary Figure 3**).

# Taxonomic Changes Induced by Sourdough Bread in the Rat Gut Microbiota

In order to study if differences in leavening (i.e., sourdough or baker's yeast) could affect the structure of the rat GM, we compared its taxonomic composition, at family and genus/species level, in BB and SB-fed rats, based on both 16S rRNA gene sequencing (colon and feces) and metaproteomic data.

Considering 16S rRNA gene sequencing, we found 11 and 4 taxa significantly changed in colon and feces, respectively (**Table 1**). Interestingly, the genus Mycoplasma and its corresponding family Mycoplasmataceae were consistently detected as more abundant in both the colonic and fecal microbiota of "BB" animals (**Figure 1**). Furthermore, the family Cytophagaceae was found as significantly more abundant in BB-fed rat stool, while the species Alistipes indistinctus and Mucispirillum schaedleri were significantly enriched in the colonic contents of BB-fed rats. On the other hand, the SBassociated GM was observed as enriched in 7 families, including Verrucomicrobiaceae in stool, and Bacillaceae, DHVEG-1 (belonging to archeaea), Moraxellaceae, Nitrospinaceae, and Thermaceae in the colonic content.

When metaproteomic data were considered, 3 taxa exhibited a differential abundance between the two groups: Lachnospiraceae and Desulfovibrionaceae, enriched in BB-fed rats, and Dubosiella spp., more abundant in SB-fed rats (**Table 1**).

# Functional Changes Induced by Sourdough Bread in the Rat Gut Metaproteome

In addition to the taxonomic composition, we also investigated the effect of the type of leavening agent on the GM activities; hence, we focused on differences in the expression of protein functions by comparing BB- and SB-fed rats. Protein sequences were first functionally annotated and then the functional information was combined with the taxonomic classification at family and genus levels. A total of 19 and 17 differential family-(**Figure 2**) and genus-specific (**Supplementary Figure 4**) functions were identified, respectively.

Among these, two Desulfovibrio-assigned functions, namely glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and porin-like integral membrane protein (OmpA) families, and two Lachnospiraceae-assigned functions, phosphoenolpyruvate (PEP)-utilizing enzyme and ABC transporter superfamily, were detected as more abundant in BB-fed rats, in line with the taxonomic results based on metaproteomics (Taxonomic Changes Induced by Sourdough Bread in the Rat Gut Microbiota). In addition, bacterial ribosomal protein bS6 family, from Clostridium, and the Lactobacillus-specific glycosyl hydrolase 36 family, that includes α-galactosidase and α-N-acetylgalactosaminidase activities, showed the same behavior.

Interestingly, among the taxonomy-function combinations that exhibited a significant increase in the GM of SB-fed rats, we detected several protein families belonging to the amino acid metabolism and transport Cluster of Orthologous Groups (COG) category: aspartate and ornithine carbamoyltransferase (from Oscillibacter and Treponema), asparaginase 1 (Bacteroides), and type 1 argininosuccinate synthase (Clostridiaceae). Furthermore, we identified functions involved in energy production (Prevotella-specific Acetyl-CoA hydrolase/transferase and Clostridium-specific V-ATPase proteolipid subunit), translation (universal ribosomal protein uL4 and uL14, both assigned to Lactobacillus), and post-translational modification [FKBP-type peptidyl-prolyl cis-trans isomerase (PPIase) from Acinetobacter]. With reference to Clostridium, we also found manganese catalase and alpha/beta-type SASP, namely small acid-soluble spore protein; the latter binds the spore DNA and was the most abundant among the differential protein families. Finally, PEPutilizing enzyme and OmpA presented an opposite differential trend depending on the specific taxonomic assignment, a possible effect of GM taxonomic variations due to SB- or BB-based diets. Thus, PEP-utilizing enzyme molecules assigned to Dubosiella were more abundant in SB-fed rats, whereas those assigned to Lachnospiraceae were more abundant in BB-fed rats; on the other hand, Parabacteroides-specific OmpA was more abundant in SB-fed rats, whereas Desulfovibrio-specific OmpA was more abundant in BB-fed rats.

#### Taxonomic Changes Induced by Carasau Bread in the Rat Gut Microbiota Compared to Standard Chow Diet

To investigate if the supplementation with carasau bread to the common chow diet could significantly change the GM structure, we performed two separate differential analyses, i.e., "SB" vs. "chow" and "BB" vs. "chow" based on 16S rRNA gene sequencing (colon and feces) and metaproteomic data. When comparing "SB" and "chow" groups we identified 10 (stool 16S rRNA gene sequencing data), 9 (colon 16S rRNA gene sequencing data), and 11 (metaproteomics) taxa with an abundance significantly different between the two groups (**Table 2**). Most of these differences in 16S rRNA gene sequencing results were also found between "SB" and "BB" (Taxonomic Changes Induced by Sourdough Bread in the Rat Gut Microbiota). Indeed, in SB-fed rats a reduction of Mycoplasmataceae and its related genus Mycoplasma was again seen in both stool and colon (**Figure 1**), as well as a decrease of Cytophagaceae in stool, whereas Verrucomicrobiaceae (stool), Bacillaceae, DHVEG-1, Moraxellaceae, and Thermaceae (colon) were increased. Moreover, we found the GM of chow-fed group enriched in Planococcaceae and Solibacillus, both in stool and colon, while Paraprevotella was increased only in colon. Additionally, Bradyrhizobium and its related family Bradyrhizobiaceae were more abundant in SB-fed rat stool.

Metaproteomic analysis revealed an increased abundance of Marinactinospora, Phascolarctobacterium, Pseudopropionibacterium, and Turicibacter genera, as well as Acidaminococcaceae and Propionibacteriaceae families in the "SB" group; by contrast, Desulfovibrio and Dorea were significantly more abundant in chow-fed animals.

We also compared GMs from BB- and chow-fed rats, identifying 5 (stool) and 8 (colon) taxa as differentially abundant TABLE 1 | Differential taxa in gut microbiota of rats fed chow supplemented with bread leavened with baker's yeast (BB) vs. sourdough (SB).


*Families and genera/species with statistically significant differential abundance between BB- and SB-fed rats are listed. For each taxon, the adjusted p-value (edgeR test followed by SGoF adjustment) and the logarithm of the fold change (logFC), according to 16S rRNA gene sequencing (16S; stool and colonic contents) and metaproteomic (MP; stool only) data, are reported. Features are ordered based on alphabetical order, and families (up) are separated from genera/species (down).*

supplemented with baker's yeast leavened bread (light blue); SB, rats fed chow supplemented with sourdough leavened bread (orange); chow, rats fed chow only (green). Statistically significant differences between groups (according to edgeR test followed by SGoF adjustment) are indicated with asterisks (\* = adjusted *p*-value < 0.05; \*\* = adjusted *p*-value < 0.01; \*\*\* = adjusted *p*-value < 0.00001).

after 4 weeks of dietary treatment (**Table 3**). Interestingly, Planococcaceae and Solibacillus (both in stool and colon) and Paraprevotella (only in colon) were again enriched in chowfed rats. Further, Corynebacterium stationis was enriched in feces of "chow" group. In contrast, Phascolarctobacterium was lower both in stool and colon of chow-fed rats (consistent with SB vs. chow metaproteomic data), as well as [Ruminococcus] in feces and Ruminococcus bromii in the colonic contents.

No taxonomic features were differentially represented in the GMs of BB- vs. chow-fed rats based on metaproteomic analysis.

FIGURE 2 | Differential family-specific microbial functions in rats fed chow supplemented with bread leavened with baker's yeast (BB) vs. sourdough (SB). In each line, a dot represents a single animal, with its color intensity being proportional to the relative abundance of that given microbial protein in that subject, according to the scale depicted in the bottom-right corner. Missing values (function not identified in that animal) are in white; features with missing values in the most abundant group were filtered out. The upper part of the heatmap lists functions with higher abundance in the fecal microbiota of SB-fed animals, while the lower part lists those with higher abundance in the fecal microbiota of BB-fed animals. Functions are ordered based on the Cluster of Orthologous Groups (COG) category to which they belong (C, Energy production and conversion; E, Amino acid transport and metabolism; G, Carbohydrate transport and metabolism; J, Translation, ribosomal structure and biogenesis; M, Cell wall/membrane/envelop biogenesis; O, Posttranslational modification, protein turnover, chaperones; P, Inorganic ion transport and metabolism), and then in alphabetical order.

#### Functional Changes Induced by Carasau Bread in the Rat Gut Microbiota Compared to Standard Chow Diet

Finally, the taxa-specific functional changes occurring in the rat GM after diet supplementation with carasau bread were investigated. To this end we compared metaproteomic functions, based on their taxonomic annotation, of chow-fed rats with SBand BB-fed rats.

A total of 21 family-(**Figure 3**) and 22 genus-specific functions (**Supplementary Figure 5**) were detected as differentially represented between "SB" and "chow" groups, while no differential features were observed between "BB" and "chow" groups. Not surprisingly, some of the differential functions more abundant in SB, i.e., acetyl-CoA hydrolase/transferase (assigned to Prevotella), asparaginase 1 (Bacteroides), FKBP-type PPIase (Acinetobacter), and manganese catalase (Clostridium), had been already found as more abundant in the "SB" group when compared with the "BB" group. Other SB-enriched features included functions implicated in amino acid metabolism and transport, namely serine-glycine hydroxymethyltransferase (SHMT) assigned to Prevotella and phosphoserine aminotransferase SerC assigned to Porphyromonadaceae. Finally, we found as differentially expressed also AhpC/Prx1 peroxiredoxin, (assigned to Prevotella), enolase (Tannerella), glycosyl hydrolase 36 (Turicibacter), methylmalonyl-CoA mutase (Phascolarctobacterium), serineprotein kinase PrkA (Clostridium), and TonB-dependent receptor (Pseudopropionibacterium).

Several functions involved in carbohydrate metabolism (L-ribulose-5-phosphate 4-epimerase/L-fuculose phosphate aldolase from Lachnospiraceae, PEP-utilizing enzyme from Prevotella and Alloprevotella, pyruvate kinase from Turicibacter, and transketolase family from Prevotella), translation (including bacterial ribosomal protein bL9 from Parabacteroides and the elongation factors EF-Ts, EF-Tu/EF-1A, EF-G/EF-2 from Mediterranea, Desulfococcus, and Ruminococcus, respectively), and post-translational modification (heat shock protein 70 assigned to Parabacteroides and ClpC protease from Lactobacillaceae) were enriched in "chow" group. Leucine-binding protein from Lachnospiraceae was also increased.

#### DISCUSSION

Sourdough bread is recognized to possess a great variety of valuable effects on nutrition and health. The worldwide


TABLE 2 | Differential taxa in gut microbiota of rats fed chow supplemented with bread leavened with sourdough (SB) vs. unsupplemented (chow).

*Families and genera/species with statistically significant differential abundance between SB- and chow-fed rats are listed. For each taxon, the adjusted p-value (edgeR test followed by SGoF adjustment) and the logarithm of the fold change (logFC), according to 16S rRNA gene sequencing (16S; stool and colonic contents) and metaproteomic (MP; stool only) data, are reported. Features are ordered based on alphabetical order, and families (up) are separated from genera/species (down).*

interest in investigating its qualities with new and more robust methodologies is due to the existence of numerous traditional bakery products in many different countries, that are currently being re-discovered as relevant components of a well-balanced and "healthier" diet, as well as potentially useful as part of a therapeutic dietary intervention.

Indeed, compelling evidence was provided over the last few years to support nutrition and health claims of sourdough leavened bread. As extensively reviewed by Gobbetti et al., bread products obtained using LAB instead of baker's yeast are generally appreciated for the more complex and agreeable flavor and taste (Gobbetti et al., 2016). In addition, a lower glycemic index has been measured for bread leavened with LAB, when compared with the same type of bread leavened with Saccharomyces cerevisiae (Maioli et al., 2008; Poutanen et al., 2009; Stamataki et al., 2017). This feature, due to a lower amount of rapidly digestible starch in the small intestine, is accompanied by a larger amount of slowly digestible and resistant starch that reaches the colon, where it is degraded by colonic bacteria to produce short-chain fatty acids. The latter in turn provide energy to the colonic cells, reduce susceptibility to cancer, and control gut inflammation (van der Beek et al., 2017).

Hence, as different biochemical changes occur in sourdough and baker's yeast fermentations, both possibly impacting consumer's metabolism and health, the aim of our study was to compare their effects on the intestinal microbiota taxonomy and metabolism. We also aimed to evaluate for the first time the capability of metaproteomics to reveal the effects of these different bread making processes on GM structure and functions. Two kinds of flat bread (carasau bread) were obtained employing the same raw materials, manufacturing recipes, and processing conditions, but they were leavened with either yeast or sourdough fermentation.

In this study, microbiota taxonomy variation was assessed by means of 16S rRNA gene sequencing analysis. Diet supplementation with sourdough bread led to a reduction of specific members of the GM, belonging to genera as Alistipes, Mucispirillum, and Mycoplasma. Such changes might be due to differences in nutrients availability in LAB vs. baker's yeast fermented bread. Proteolysis occurring during lactic acid


TABLE 3 | Differential taxa in gut microbiota of rats fed chow supplemented with bread leavened with baker's yeast (BB) vs. unsupplemented (chow).

*Families and genera/species with statistically significant differential abundance between BB- and chow-fed rats are listed. For each taxon, the adjusted p-value (edgeR test followed by SGoF adjustment) and the logarithm of the fold change (logFC), according to 16S rRNA gene sequencing (16S; stool and colonic contents) and metaproteomic (MP; stool only) data, are reported. Features are ordered based on alphabetical order, and families (up) are separated from genera/species (down).*

FIGURE 3 | Differential family-specific microbial functions in rats fed chow supplemented with bread leavened with sourdough (SB) vs. chow only. In each line, a dot represents a single animal, with its color intensity being proportional to the relative abundance of that given microbial protein in that subject, according to the scale depicted in the bottom-right corner. Missing values (function not identified in that animal) are in white; features with missing values in the most abundant group were filtered out. The upper part of the heatmap lists functions with higher abundance in the fecal microbiota of SB-fed animals, while the lower part lists those with higher abundance in the fecal microbiota of chow-fed animals. Functions are ordered based on the Cluster of Orthologous Groups (COG) category to which they belong (C, Energy production and conversion; E, Amino acid transport and metabolism; G, Carbohydrate transport and metabolism; I, Lipid metabolism; J, Translation, ribosomal structure and biogenesis; M, Cell wall/membrane/envelop biogenesis; O, Posttranslational modification, protein turnover, chaperones; P, Inorganic ion transport and metabolism; T, Signal transduction mechanisms; V, Defense mechanisms), and then in alphabetical order.

fermentation is expected to change the protein assortments and to reduce the amount of proteins reaching the colonic mucosa (Spicher and Nierle, 1988). Consequently, free amino acids, including alanine, glutamic acid, asparagine, and arginine are more abundant in sourdough, where the bacterial metabolic activities also increase the levels of dough acidity. This may explain the reduction of Mucispirillum that has been previously associated to protein-deficient diets (Navarro et al., 2018; Zhai et al., 2018). However, in this study, we did not measure the total proteins or free amino acids amounts in SB and BB. Consistently, Alistipes was also found associated to low protein diets and to the increase of dietary fatty acids (Agans et al., 2018; Kang et al., 2018; Wei et al., 2018). Interestingly, Mycoplasma is generally acknowledged as an intestinal pathobiont and it was associated with diet-induced obesity (Turnbaugh et al., 2008). Our data suggest the possibility that a low-fat diet, supplemented with SB, might reduce its growth. On the other hand, SB induced an increase of some bacterial taxa, including Verrucomicrobiaceae, in stool. These data are not consistent with those obtained by metaproteomics, where Lachnospiraceae and Desulfovibrionaceae were found enriched in the "BB" group, while Dubosiella spp. were significantly more abundant in the "SB" group. In this context, it should be noted that the use of a different database for protein assignment and the differences in the depth of coverage, with the metagenome enabling more complete coverage than the metaproteome, can result in different taxonomic annotations. For example, Dubosiella spp. is not listed in the used metagenomic database, being a recently proposed novel genus (Cox et al., 2017).

Our study of the rat metaproteome provides unique and important insights into the variations of gut microbial taxa, their proteins, and their functions associated with CR low-fat diet supplemented with BB or SB. A very intriguing observation is the higher abundance of asparaginases expressed by Bacteroides in SB-fed rats. Normally, bacteria control their catabolic enzyme synthesis and turnover according to the abundance of the relative substrates in the environment. Hence, the differential amount of asparaginase leads to hypothesize that higher amounts of asparagine reach the colonic mucosa of SB-fed rats compared to BB-fed rats. Asparagine and, to a lesser extent, other free amino acids have been reported to represent major precursors of acrylamide in baked bread (Tareke et al., 2002). Acrylamide is a known carcinogenic agent in rodents and a probable carcinogen in human. Although most of the dietary intake of acrylamide derives from fried potatoes and coffee, attention is also directed to baked bread since it might represent a significant source of this molecule. Notably, when comparing breads obtained with different leavening processes, higher acidity reduces the acrylamide formation in sourdough vs. yeast-fermented bread, despite the higher asparagine content in the former (Nasiri Esfahani et al., 2017). Our data are in keeping with those of Bartkiene et al., that have recently demonstrated the possibility to reduce acrylamide content down to 67.2% in sourdough bread prepared with selected LAB (Bartkiene et al., 2017). Since we could not find evidence of differential variation in relative abundances of Bacteroides, the increased amount of asparaginase in SB-fed GM points toward a "turning on" of Bacteroides asparaginase as inducible enzyme (Boeck et al., 1970). Asparaginase substrate, asparagine, is converted to aspartate and ammonium, in a metabolic pathway that provides carbon and nitrogen as components of many biomolecules. Asparagine catabolism in bacteria is therefore important to compete against other bacterial members of the GM community and it is a significant virulence determinant for many enteric pathogens (Scotti et al., 2010; McLaughlin et al., 2017). At the same time, bacterial competition for asparagine, and its degradation, might be beneficial for the host, reducing the risk of colon cancer and/or its progression. Asparaginase is indeed well known as key in the treatment of acute lymphoblastic leukemia and its potential as "anti-colon cancer protein" has been recently proposed (El-Naggar et al., 2016; Miyo et al., 2016; Toda et al., 2016).

Other enzymes might be regulated in response to specific metabolite(s) produced or enhanced by SB or BB, but their limit of detection could be below that of our metaproteomic approach. Or else, a number of these enzymes might be regulated by allosteric mechanism, rather than by a change in their expression level.

Another group of five protein families, expressed by Clostridium, were observed to change their abundance in animals fed SB-supplemented diet. Of these, the bacterial ribosomal protein bS6 showed higher relative abundance in BB-fed rats. On the contrary, manganese catalase, small acid-soluble proteins (SASP), Ser/Thr kinase PrkA, and V-ATPase proteolipid subunit were higher in the GM of SB-fed rats. These proteins have all been reported to be involved in Clostridium sporulation. SASPs, in particular, play an important role in protecting DNA against damage by heat, UV radiation, or enzymic degradation in dormant bacterial endospores (Wetzel and Fischer, 2015). ATP synthase subunits were shown to be upregulated during late sporulation, possibly to meet energy demands, both in Bacillus and in Clostridium (Wang et al., 2013; Liu et al., 2016). Manganese catalase is a spore coat protein with an important role of H2O<sup>2</sup> detoxification (Permpoonpattana et al., 2013). Finally, also PrkA is involved in sporulation, although its role is not yet well understood (Pompeo et al., 2016). Taken together, these data strongly suggest that consumption of sourdough might increase the subset of metabolically dormant Clostridium spp. in the GM. Among Clostridium spp., the most studied one is the gut pathobiont C. difficile, the causative agent of pseudomembranous colitis and toxic megacolon, whose pathogenicity is potentiated by antibiotic treatment. In fact, a healthy microbiota is expected to dampen C. difficile germination, probably through C. difficilegrowth-inhibitory secondary bile acids (Shen, 2015). To this end, the GM of SB-fed rats might have a specific effect on Clostridium sporulation.

In conclusion, we provide evidence that consumption of sourdough-leavened bread has the potential to significantly change the taxonomy of the GM and the metabolic functions of some of its most important members, including Bacteroides and Clostridium. Further, the results of this study confirm that metaproteomics is able to pinpoint the impact of food processing technologies on microbial enzymes and related metabolites, which are in turn able to reach the gut mucosa and exert their potential effect on the consumer's health at intestinal and systemic level.

#### DATA AVAILABILITY

Raw read sequences were deposited in the European Nucleotide Archive under the Project Accession Number PRJEB29264.

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Vizcaino et al., 2016) partner repository with the dataset identifier PXD011441.

#### ETHICS STATEMENT

Animal studies were reviewed and approved by the Institutional Animal Care and Use Committee of the University of Cagliari and were performed in accordance with the relevant guidelines and regulations (authorization of the Italian Health Ministry No. 840/2016-PR).

#### AUTHOR CONTRIBUTIONS

MA performed 16S rRNA gene sequencing sample preparation and analysis, contributed to data interpretation, and wrote the manuscript. AP performed metaproteomics sample preparation, mass spectrometry analysis, and contributed to data interpretation and to critically revise the manuscript. AT performed metaproteomics sample preparation, supervised the global data analysis and interpretation, and contributed to

#### REFERENCES


critically revise the manuscript. CF performed 16S rRNA gene sequencing sample preparation and analysis. DP supervised mass spectrometry analysis. MS and FM performed animal experiments and sample collection. EL conceived and coordinated the study. SU conceived and coordinated the study, contributed to data interpretation, and wrote the manuscript. All authors read and approved the final version of the manuscript.

#### FUNDING

This work was supported by Sardegna Ricerche, Science and Technology Park of Sardinia, under grant program art.9 LR 20/2016 (2017) to Porto Conte Ricerche. MA was supported by a Doctoral Fellowship from the International PhD Course in Life Sciences and Biotechnologies, University of Sassari.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.01733/full#supplementary-material


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Abbondio, Palomba, Tanca, Fraumene, Pagnozzi, Serra, Marongiu, Laconi and Uzzau. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Protective Effect of Ursolic Acid on the Intestinal Mucosal Barrier in a Rat Model of Liver Fibrosis

Wang Zhang<sup>1</sup>† , Dakai Gan1,2† , Jie Jian<sup>1</sup>† , Chenkai Huang<sup>1</sup> , Fangyun Luo<sup>1</sup> , Sizhe Wan<sup>1</sup> , Meichun Jiang<sup>1</sup> , Yipeng Wan<sup>1</sup> , Anjiang Wang<sup>1</sup> , Bimin Li<sup>1</sup> \* and Xuan Zhu<sup>1</sup> \*

<sup>1</sup> Department of Gastroenterology, The First Affiliated Hospital of Nanchang University, Nanchang, China, <sup>2</sup> Department of Liver Disease, The Ninth Hospital of Nanchang, Nanchang, China

#### Edited by:

Petra Hirsova, Charles University, Czechia

#### Reviewed by:

Ana Cristina Llorente Izquierdo, University of California San Diego, United States Ekihiro Seki, Cedars-Sinai Medical Center, United States

#### \*Correspondence:

Bimin Li lbmjx@163.com Xuan Zhu jyyfyzx@163.com †These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology

Received: 04 February 2019 Accepted: 09 July 2019 Published: 30 July 2019

#### Citation:

Zhang W, Gan D, Jian J, Huang C, Luo F, Wan S, Jiang M, Wan Y, Wang A, Li B and Zhu X (2019) Protective Effect of Ursolic Acid on the Intestinal Mucosal Barrier in a Rat Model of Liver Fibrosis. Front. Physiol. 10:956. doi: 10.3389/fphys.2019.00956 Oxidative stress mediated by nicotinamide adenine dinucleotide phosphate (NADPH) oxidase (NOX) plays an important role in intestinal mucosal barrier damage in various disease states. Recent evidence suggests that intestinal mucosal barrier damage and intestinal dysbiosis occur in mice with hepatic fibrosis induced by CCl4 or bile duct ligation. Another study showed that ursolic acid (UA) attenuates experimental colitis via its anti-inflammatory and antioxidant activities. The goal of this study was to investigate the effects of UA on the intestinal mucosal barrier in CCl4-induced hepatic fibrosis in rats and identify its associated mechanisms. Male Sprague-Dawley rats were randomly divided into the following 3 groups (n = 10/group): the control, CCl4 model and UA treatment groups. Rats were sacrificed at 72 h after the hepatic fibrosis model was established and assessed for liver fibrosis, intestinal injury, enterocyte apoptosis, bacterial translocation, system inflammation, intestinal oxidative stress, and tight junction protein and NOX protein expression. The results demonstrated that UA attenuated the following: (i) liver and intestinal pathological injury; (ii) cleaved caspase-3 expression in the ileal epithelial cells; (iii) serum lipopolysaccharide and procalcitonin levels; (iv) intestinal malondialdehyde levels; and (v) the expression of the NOX protein components NOX2 and P67phox in ileal tissues. Furthermore, our results suggested that UA improved intestinal dysbiosis and the expression of the tight junction proteins Claudin 1 and Occludin in the ileum of rats. These results indicate that UA has protective effects on the intestinal mucosal barrier in rats with CCl4-induced liver fibrosis by inhibiting intestinal NOX-mediated oxidative stress. Our findings may provide a basis for further clinical studies of UA as a novel and adjuvant treatment to cure liver fibrosis.

Keywords: hepatic fibrosis, intestinal mucosal barrier function, ursolic acid, NOX, intestinal microbiota

# INTRODUCTION

Liver fibrosis is a wound-healing response to chronic liver injury that develops into liver cirrhosis or liver cancer, which is associated with significant morbidity and mortality. Progressive fibrosis can eventually result in cirrhosis, liver failure or hepatocellular carcinoma. Intestinal mucosal barrier damage and dysbiosis have been reported to occur in mice with hepatic fibrosis induced by CCl4 or bile duct ligation (Fouts et al., 2012). The increased intestinal permeability resulting from intestinal

**426**

mucosal barrier damage and intestinal dysbiosis contributes to the translocation of bacteria and/or bacterial products (Fouts et al., 2012; Hartmann et al., 2012), the latter of which induce hepatic stellate cell (HSC) activation and contribute to liver fibrosis (Seki et al., 2007; Hartmann et al., 2012). Moreover, translocated bacteria and their products are closely associated with various liver cirrhosis complications. Therefore, restoring intestinal barrier function and preventing bacterial translocation has great significance for inhibiting the progression of liver fibrosis and improving the prognosis of patients with chronic liver disease.

Various factors contribute to intestinal barrier damage, with intestinal oxidative stress having an especially important role. Nicotinamide adenine dinucleotide phosphate (NADPH) oxidase (NOX) is a multicomponent enzyme complex that generates reactive oxygen species (ROS) in response to various stimulus. NOX-derived ROS is one of the most important sources of intestinal oxidative stress. Welak et al. (2014) observed that NOX2-mediated oxidative stress plays a crucial role in the progression of necrotizing enterocolitis. In another study, the activation of the inflammasome by NOX2-derived ROS was shown to promote ileal mucositis induced by irinotecan, a chemotherapeutic agent that inhibits topoisomerase I (Arifa et al., 2014). Taken together, these results indicate that the inhibition of NOX-mediated oxidative stress can protect the intestinal mucosal barrier.

Ursolic acid (UA) is a natural pentacyclic triterpenoid with various pharmacological activities. In our previous studies, we showed that UA has unique anti-fibrotic effects, inhibiting the proliferation of activated HSCs and inducing their apoptosis but not hepatocyte apoptosis (Shen et al., 2008; Gan et al., 2018). Other researchers subsequently corroborated that UA selectively induces the apoptosis of activated HSCs without inducing liver or quiescent HSC apoptosis (Wang et al., 2011). Furthermore, we observed that UA inhibited the leptin-mediated expression of the NOX subunits NOX2, P67phox and NOX4 in an activated rat HSC cell line (HSC-T6), resulting in the accumulation of extracellular matrix (ECM) proteins (Wang et al., 2011; He et al., 2015). In addition, the results of another study suggested that UA attenuates experimental colitis in mice via its anti-inflammatory and antioxidant activities (Chun et al., 2014; Liu et al., 2016). However, because it is unknown whether UA has a protective effect on the intestinal mucosal barrier in rats with CCl4-induced liver fibrosis, the goal of the current study was to answer this question and elucidate the related mechanism.

#### MATERIALS AND METHODS

#### Reagents and Antibodies

The following reagents were used in this study: CCl4 and olive oil (Shandong Xiya Chemical Industries Co., Ltd., Shandong, China); UA (Sigma Chemical Co., St. Louis, United States); Picro Sirius Red Stain and triglyceride assay kits (Beijing Solarbio Science & Technology Co., Ltd., Beijing, China); total antioxidant capacity (TAC), malondialdehyde (MDA), total bilirubin (TBIL), hydroxyproline, alanine aminotransferase (ALT) and C-reactive protein (CRP) assay kits (Nanjing Jiancheng Bioengineering Institute, Nanjing, China); a rat tumor necrosis factor alpha (TNF-α) enzyme-linked immunosorbent assay (ELISA) kit, a rat albumin ELISA kit, a rat procalcitonin ELISA kit and a lipopolysaccharide (LPS) ELISA kit (Elabscience Biotechnology Co., Ltd., Wuhan, China); protein lysis buffer and protease inhibitor (Vazyme Biotech Co., Ltd., Nanjing, China); a Bradford Protein Assay kit (Tiangen Biotech Co., Ltd., Beijing, China); an anti-cleaved caspase-3 antibody (Cell Signaling Technology, United States); anti-NOX2/gp91phox, anti-p67phox, anticollagen I, anti-alpha smooth muscle actin (α-SMA) and anti-Occludin antibodies (Abcam, United Kingdom); an anti-Claudin 1 antibody (Thermo Fisher Scientific, United States); and mouse anti-β-actin and horseradish peroxidase-labeled goat anti-mouse IgG and goat anti-rabbit IgG antibodies (Beijing Zhongshan Golden Bridge Biotechnology, Co., Ltd., Beijing, China).

#### Animal Procedures and Treatments

All experimental procedures were approved by the Institutional Animal Care and Use Committee of the First Affiliated Hospital of Nanchang University (Nanchang, China). All animals received humane care in compliance with institutional guidelines. Male Sprague-Dawley rats (160–200 g body weight) were obtained from the Department of Laboratory Animal Science of Nanchang University and had access to water and standard chow diet ad libitum. Animals were maintained in an environment with a 12:12 h light/dark cycle, a room temperature of 22 ± 2 ◦C, and 55 ± 5% humidity. Male Sprague-Dawley rats were randomly divided into the following 3 groups (n = 10/group): the control, CCl4 model and UA treatment groups (CTR, CCl4 and UAT, respectively). The control rats were given olive oil (2 ml/kg) by gavage twice a week for 8 weeks and then were administered normal saline (40 mg/kg/day) for 4 weeks. Hepatic fibrosis was induced by gastric gavage of CCl4 (diluted 1:4 in olive oil, 2 ml/kg) twice a week for 8 weeks, after which the rats in the CCl4 model and UA treatment groups were given normal saline or UA (40 mg/kg/day) for 4 weeks, respectively. Abdominal fur was removed with a depilatory and the skin was sterilized with iodine, then laparotomies were performed under strict aseptic conditions. Blood samples were collected from the inferior vena cava, and rats were sacrificed afterward. Liver and ileal tissues adjacent to the cecum were isolated. A portion of each liver and ileum was removed for histopathological examination by fixation with 10% formalin and subsequent embedding with paraffin. The remaining tissue specimens were frozen in liquid nitrogen and stored at −80◦C.

#### Liver and Ileum Histopathology

The paraffin-embedded liver and ileum samples were used to prepare 5 µm thick slices with a microtome. The slices were stained with hematoxylin and eosin using standard methods. For Sirius red collagen staining, the liver slices were deparaffinized and stained with Picro Sirius Red for 1 h at room temperature. After washing, the slices on the slides were stained with hematoxylin and subsequently mounted in permount medium. The degree of hepatic fibrosis was evaluated

semiquantitatively based on the Metavir score (Fagan et al., 2015). The Sirius red stained area was quantified by Image-Pro Plus 6.0. The histological grade of the intestinal mucosal damage was scored according to the criteria described by Chiu et al. (1970), and microscopic scoring was performed blindly by two senior pathologists.

Immunohistochemistry was performed on serial sections of paraffin-embedded ileal tissue. After rehydrating, the sections were maintained in 0.3% H2O<sup>2</sup> for 7 min to eliminate endogenous peroxidase and then were washed with phosphate buffer saline (PBS). Next, the samples were transferred to citrate buffer (pH 7.6) and heated in a microwave oven for 20 min. After washing the sections with PBS and blocking the nonspecific binding sites with 5% bovine serum albumin (BSA), they were incubated with rabbit anti-cleaved caspase-3 (1:400), anti-Occludin (1:200), anti-Claudin 1 (1:100) polyclonal antibody overnight at 4◦C. Next, the sections were rinsed in PBS and then incubated with biotin-labeled goat anti-polyvalent for 15 min at 37◦C and horseradish peroxidase-labeled streptavidin for 20 min at 37◦C. The coloration was completed after treatment with diaminobenzidine for 10 min, after which the slides were counterstained with hematoxylin for 2 min, rinsed in tap water and dehydrated. The sections were observed under a microscope. Based on the criteria proposed by Chen and Lin (2015), immunohistochemical staining was analyzed by two pathologists in a blind manner.

#### Enzyme-Linked Immunosorbent Assay

The albumin, TNF-α, LPS, PCT, and CRP levels were determined using ELISA kits according to the manufacturer's instructions. Briefly, 100 µl samples were added to each well of a 96-well plate that was precoated with a specific antibody, after which the plate was covered with sealer and incubated for 90 min at 37◦C. After incubating with the biotinylated detection antibody working solution, 100 µl of the HRP-conjugated working solution was added to each well, and the plates were incubated for an additional 30 min at 37◦C before adding 90 µl of substrate solution. The reaction was stopped by the addition of 50 µl of stop solution to each well. The absorbance of each well was read immediately at a wavelength of 450 nm, and the values were normalized to the control.

#### Malondialdehyde Content Determination

The concentration of MDA, a reliable marker of lipid peroxidation, was determined using a thiobarbituric acid (TBA) assay kit according to the manufacturer's instructions (Nanjing Jiancheng Bioengineering Institute, China). Briefly, ileal tissue samples were homogenized using a Retsch MM400 homogenizer (Retsch, Germany) followed by centrifugation at 3000 rpm/min for 15 min. Subsequently, the supernatants were collected for MDA measurements. MDA in the sample supernatant reacts with TBA at 95◦C under acidic conditions, yielding a pink MDA-TBA conjugate. The optical density of the MDA-TBA complex at 532 nm was measured using a microplate reader (SpectraMax M5, United States). The total protein content in the ileal supernatant was analyzed using a Bradford Protein Assay kit (Tiangen Biotech Co., Ltd., Beijing, China). Finally, the MDA content (nanomoles per milligram protein) was calculated according to the formula described in the manufacturer's instructions.

# Total Antioxidant Capacity Determination

The TAC was determined using a colorimetric method according to the manufacturer's instructions (Nanjing Jiancheng Bioengineering Institute, China). The methods used for sample preparation and protein concentration determination were the same as those described in the previous section. Ferric tripyridyltriazine (Fe3+-TPTZ) is reduced to blue ferrous tripyridyltriazine (Fe2+-TPTZ) by various antioxidant components in the sample supernatant under acidic conditions. The absorbance of the blue product was measured at 520 nm using a microplate reader (SpectraMax M5, United States). Finally, the TAC (unit per milligram protein) was calculated according to the formula described in the manufacturer's instructions.

### Serum Alanine Aminotransferase, Total Bilirubin, Triglyceride, and Hydroxyproline Analysis

Blood samples without anticoagulant from the inferior vena cava were centrifuged at 3,000 rpm for 15 min to collect the serum. The serum ALT, TBIL, triglyceride and hydroxyproline were determined colorimetrically with commercial assay kits according to the manufacturer's protocols.

#### Western Blot Analyses

Total protein was prepared using radioimmunoprecipitation assay buffer supplemented with 1× protease inhibitor. The total protein content in the ileal supernatant was analyzed using a Bradford Protein Assay kit (Tiangen Biotech Co., Ltd., Beijing, China). The protein samples were loaded (30 µg/well) and separated using an SDS-polyacrylamide gel. The gel was transferred to a nitrocellulose membrane and blocked with 5% skim milk in Tris–buffered saline with Tween 20 (TBST). Next, the membrane was incubated with specific primary antibodies overnight at 4◦C followed by incubation with horseradish peroxidase-conjugated secondary antibodies for 4 h at 4◦C. The membrane was treated with chemiluminescence reagent and exposed to a luminescence image analyzer (Bio-Rad ChemiDoc MP, United States) to detect the protein bands. The relative levels of the target protein were expressed as a gray intensity ratio of the target band to the β-actin band.

#### 16S Ribosomal RNA Gene Sequencing

The V3–V4 region of the bacterial 16S ribosomal RNA (rRNA) gene was PCR amplified with indexed primers (338F and 806R) using FastPfu Polymerase. The amplicons were then purified by gel extraction and quantified using a QuantiFluor-STusing E.Z.N.A. Soil DNA Isolation kit. The purified PCR products were used for high-throughput pyrosequencing, which was carried out by Majorbio Bio-Pharm Biotechnology Co., Ltd., Shanghai, China, using an Illumina MiSeq PE250. The resulting sequences were analyzed using Quantitative Insights into Microbial Ecology. All the sequences were clustered into operational taxonomic units (OTUs) based on a 97% identity threshold by the SILVA database. A representative sequence from each OTU was selected for downstream analysis, and the community richness and diversity indices were calculated.

#### Statistical Analysis

fphys-10-00956 July 26, 2019 Time: 12:10 # 4

Quantitative data were expressed as the means ± standard deviation (SD). Normality was assessed using the single-sample Kolmogorov-Smirnov Test, and normally distributed data were analyzed by one-way analysis of variance (ANOVA) followed by the least significant difference test. Ranked data were compared by the Kruskal-Wallis H-test among the groups. If positive, multiple comparisons were carried out using the Nemenyi test. Statistical analyses were performed with IBM SPSS statistics version 23.0. Values of P < 0.05 were considered significant.

# RESULTS

#### UA Suppresses Hepatic Fibrogenesis in Rats With CCl4-Induced Liver Fibrosis

To investigate the effects of UA on liver fibrosis, a rat model of CCl4-induced liver fibrosis was established. During the modeling, the mortality rate for the control group was 0% (0/10), while that of the CCl4 model and UA treatment groups was 20% (2/10). As shown in **Figure 1A**, tissue sections from the control group showed a normal hepatic lobular structure and little collagen deposition. Heavy deposits of collagen were observed in the livers of rats from the CCl4 model group and were accompanied by disordered hepatic lobular structures and severe hepatocyte necrosis, whereas these changes were suppressed in the UA treatment group. In addition, the hepatic fibrosis scores (**Figure 1B**) and area (**Figure 1C**) in the UA treatment group were significantly lower than those of the CCl4 model group. The expression of collagen I and α-SMA were significantly elevated in the CCl4 model group compared with that observed in the control group, which was significantly inhibited by the UA treatment (**Figure 1D**). The hydroxyproline content of rat serum in the UA treatment group declined compared to the CCl4 model group (**Figure 1E**). These results confirmed that UA protected the rat liver from fibrogenesis induced by CCl4.

#### UA Ameliorates Liver Injury in Rats With CCl4-Induced Liver Fibrosis

We measured the contents of ALT, ALB, TBIL, and triglyceride in rat's serum. The results presented in **Figures 2A–D** showed decreased serum ALB and increased serum ALT, TBIL and triglyceride in the CCl4 model group compared to those in the control group, which were partly restored by UA. As shown in **Figure 2E**, the hepatic TNF-α level was significantly higher in CCl4 model group than in the control group. However, the increased level of hepatic TNF-α was reduced by UA treatment. Additionally, the UA treatment group had higher final body weights and lower liver weights than the CCl4 model group (**Table 1**).

# UA Improves Intestinal Dysbiosis in Rats With Liver Fibrosis

We assessed the abundance and diversity of microbes in the ileal mucosa of 5 randomly selected rats from each group by analyzing the 16S rRNA gene sequences of collected ileal tissue samples. The results presented in **Figure 3** showed that the abundance of intestinal flora, as measured by numbers of observed OTUs, was reduced in the UA treatment group compared to that observed in the CCl4 model group (p = 0.038, Nemenyi test; **Figure 3A**). However, the Shannon index, which measures both richness and evenness, was not significantly different between the CCl4 model and UA treatment groups (p = 0.40). An unweighted UniFrac-based principal coordinates analysis (PCoA) revealed that the overall microbial composition of the UA treatment group deviated from the CCl4 model group (PERMANOVAR, pseudo-F: 3.61, p = 0.001, **Figure 3B**). Principle component analysis (PCA) demonstrated that the intestinal bacterial communities of the three groups could be separated at the phylum abundance level (**Figure 3D**). LEfSe analysis (**Figure 3C**) and analysis of significant differences between the groups (**Figure 3E**) were performed to evaluate the relationships between the UA treatment and CCl4 model groups. Significant increases in the abundances of some bacterial phyla were observed in the UA treatment group, including Proteobacteria (p = 0.000), Firmicutes (p = 0.012), Actinobacteria (p = 0.011) and Tenericutes (p = 0.033). In general, the genera enriched in the CCl4 model group should be indirectly correlated with those enriched in the UA treatment group (**Figure 3**), suggesting an antagonistic relationship between harmful and beneficial bacteria.

#### UA Ameliorates Intestinal Mucosal Barrier Injury and Systemic Inflammation in Rats With CCl4-Induced Liver Fibrosis

We evaluated the pathological changes in the ileal tissue, and the results presented in **Figure 4A** showed that the tissue sections from rats in the control group exhibited normal intestinal structures and intact mucosa. In contrast, the ileal sections from the CCl4 model group showed disturbances in mucosal structure, including villous edema, atrophy, exfoliation and focal inflammatory cell infiltration in the lamina propria as well as mild desquamation of the mucosal epithelium lining in some areas, which was partly restored by UA treatment. Additionally, the UA treatment group had lower Chiu scores than the CCl4 model group (**Figure 4B**). The ileal TNF-α level was significantly higher in CCl4 model group than in the control group. However, the increased level of ileal TNF-α was reduced by UA treatment (**Figure 4C**).

We next assessed the expression of the apoptosis protein caspase-3 in ileal epithelial cells. The results presented in **Figure 4D** and **Table 2** showed that elevated cleaved caspase-3 immunoreactivity was detected in the CCl4 model group compared to that in the control group, which was reduced by UA treatment.

Tight junction proteins play an important role in the maintenance of intestinal barrier integrity and permeability.

TABLE 1 | Characteristics of rats in control, CCl4 model and UA treatment groups.


The data are presented as the means ± SD. <sup>∗</sup>P < 0.05 versus the control group; ∗∗P < 0.01 versus the control group; #P < 0.01 versus the CCl4 model group.

FIGURE 3 | Bacterial 16S rRNA gene sequencing of the ileal mucosa of rats with liver fibrosis (n = 5). (A) Analysis of α-diversity. The horizontal coordinate is the sample name and the vertical coordinate is the diversity index of the selected classification level. Intestinal bacterial α-diversity, indicated by the number of observed OTUs, was reduced in UA treatment groups (p = 0.045, Wilcoxon rank-sum test). (B) The results of the PCoA. Different color points represent samples from different groups, and the spatial distance of the sample points represents the differences between the samples. PCoA of unweighted UniFrac analysis demonstrated that the UA treatment group was significantly different from the CCl4 model group (pseudo-F: 3.61, p = 0.001, PERMANOVAR). (C) LEfSe analysis. Significant differences in the abundances of specific taxa were observed among the groups. The estimated effect of the abundance of each group was estimated. (D) PCA demonstrated that the intestinal bacterial communities of the UA treatment and CCl4 model groups could be separated using at the phylum level. (E) Analysis of significant differences between the groups. The vertical coordinate represents the species names at different classification levels, the horizontal coordinate represents the relative abundance of a species in a sample, and different colors represent different groups. (∗P < 0.05 and ∗∗∗P < 0.001 compared to the control group).

Therefore, we examined the expression of the tight junction proteins Claudin 1 and Occludin in ileal tissues. As shown in **Figures 4E,F** and **Tables 3**, **4**, the expressions of the tight junction proteins Claudin 1 and Occludin were significantly lower in the CCl4 model group than in the control group. However, the reduced ileal expressions of Claudin 1 and Occludin were restored by UA. Besides, the feces albumin content in the UA treatment group was significantly lower than that observed in the CCl4 model group (**Table 1**).

Intestinal barrier injury contributes to bacterial translocation and systemic inflammation, which we assessed by measuring serum LPS, PCT and CRP levels. Compared with those in the control group, increased serum LPS and PCT levels were observed in the CCl4 model group, while serum CRP levels showed no significant changes. The serum LPS and PCT levels in the UA treatment group were lower than those observed in the CCl4 model group. In addition, serum the CRP levels showed no significant differences between the UA treatment and CCl4 model groups (**Figure 4G**).

#### UA Inhibits Intestinal Oxidative Stress Mediated by NADPH Oxidase in Rats With CCl4-Induced Liver Fibrosis

To elucidate the mechanisms by which UA improves the intestinal mucosal barrier, we investigated the impact of UA on MDA levels and the TAC in the ileum of rats. Increased ileal MDA levels were observed in the CCl4 model

versus the CCl4 model group.

group compared with those observed in the control group, whereas no significant changes in TAC were observed. The ileal MDA contents decreased in response to the UA treatment, and no significant difference in the TAC was observed between the UA treatment and CCl4 model groups (**Figure 5A**).

We next assessed the expression of the NOX protein components P67phox and NOX2 in ileal tissues by Western blot assays. The results presented in **Figure 5B** showed that P67phox and NOX2 expression was significantly elevated in the CCl4 model group compared with that observed in the control group, which was significantly inhibited by the UA treatment.

TABLE 2 | Immunohistochemical score for cleaved caspase 3 in ileal epithelial cells.


∗∗P < 0.01 versus the control group; #P < 0.01 versus the CCl4 model group.

TABLE 3 | Immunohistochemical score for Claudin 1 in ileal tissue.


∗∗P < 0.01 versus the control group; #P < 0.01 versus the CCl4 model group.

TABLE 4 | Immunohistochemical score for Occludin in ileal tissue.


∗∗P < 0.01 versus the control group; #P < 0.01 versus the CCl4 model group.

#### DISCUSSION

Hepatic fibrosis is a common outcome of a variety of chronic liver diseases that is characterized by the accumulation of ECM, primarily type I collagen. The exact mechanism of liver fibrogenesis is still largely unknown, and a variety of factors contribute to fibrogenesis. The results of our previous study (Gan et al., 2018) showed that oxidative stress derived from NOX plays an important role in the pathogenesis of liver fibrosis and participates in regulating various signaling pathways involved in hepatic fibrosis. UA inhibited HSC activation by suppressing NOX activity and expression, preventing liver fibrosis (He et al., 2015). In the current study, we reevaluated the antifibrotic effects of UA, a natural pentacyclic triterpenoid carboxylic acid, using a classical animal model of liver fibrosis that causes hepatocellular necrosis and the deposition of collagen in the liver. Compared to the CCl4 model group, the fibrous septum and collagen deposition were reduced in the liver tissue of the UA treatment group. The serum ALT, TBIL, and triglyceride in the UA treatment group declined compared to the CCl4 model group. However, the serum albumin and final body weight were increased. All these results indicate that UA effectively improved liver histology and hepatocellular necrosis and inhibited collagen production in the livers of rats displaying damage and fibrogenesis caused by CCl4, which is consistent with the results of our previous studies showing that UA has unique antifibrotic effects.

The translocation of bacteria and their products across the intestinal barrier is common in liver disease, and there is evidence that experimental liver fibrosis depends on bacterial translocation (Mazagova et al., 2015). Dysbiosis can cause intestinal inflammation, disruption of the gut barrier, and bacterial translocation. Subsequently, translocated bacterial products induce hepatic inflammation, liver damage, liver fibrosis or even liver cirrhosis (Schnabl and Brenner, 2014). Fouts et al. (2012) confirmed that liver fibrosis is associated with an increase in adherent aerobic and anaerobic bacteria in the small and large intestine in a rat model of liver fibrosis induced with intraperitoneal injections of CCl4, where CCl4-induced liver injury was accompanied by intestinal bacterial overgrowth and dysbiosis. The distinctive pattern of dysbiosis observed in our model is similar to that described by Ubeda et al. (2016) in that cirrhotic rats displayed a reduced number of Firmicutes OTUs and an increased number of Proteobacterial OTUs. Intestinal dysbiosis was partly resolved in rats with liver fibrosis in response to the UA treatment, which restored the relative abundances of Proteobacteria and Firmicutes to that of the control mice. Although the composition of microbiome changed by CCl4 administration was restored by UA, it is still unknown whether intestinal dysbiosis is directly associated with liver injury and fibrosis in the present study. The cause of intestinal dysbiosis likely includes an absence or decrease in the intestinal levels of bile acids, changes in intestinal motility or feeding rates, etc. There is evidence that conjugated bile acids promote innate defense against luminal bacteria by regulating the expression of host genes (Inagaki et al., 2006). Through a similar mechanism, UA can activate the intracellular killing activity of macrophages against bacteria during infections (Podder et al., 2015). In addition, UA possesses direct antibacterial activity (Grace et al., 2016). All these factors may contribute to the effect of UA toward intestinal flora, which requires further study.

The intestinal mucosal barrier is an important modulator of intestinal homeostasis that consists of a permeable monolayer of epithelial cells. The epithelium allows for the absorption of nutrients while providing a physical barrier to prevent the translocation of potentially harmful substances, including pathogens, toxins, and antigens, from the gut lumen into the mucosal tissues and circulatory system via transcellular and paracellular pathways. The transcellular pathway is predominantly mediated by specific transporters or channels located on the apical and basolateral membranes (Suzuki, 2013). The paracellular pathway is regulated by apical junctional complexes, consisting of tight junctions, adherent junctions and desmosomes (Odenwald and Turner, 2017). Tight junctions, the primary determinants of paracellular permeability, seal the paracellular space and form a barrier that exhibits both size and charge selectivity with two distinct routes, termed the "pore" and "leak" pathways. The pore pathway refers to a high-capacity, size-selective and charge-selective route that appears to be primarily regulated by claudins, whereas the leak pathway is a low-capacity pathway that is regulated by Occludin (Odenwald and Turner, 2017). In the present study, the ileum

of rats in the CCl4 model group showed complicated mucosal structures, villous atrophy, and increased enterocyte apoptosis, which were significantly improved following UA treatment. Moreover, the results of Western blot and Immunohistochemical assays demonstrated that the expressions of the tight junction proteins Claudin 1 and Occludin were effectively upregulated in the UA treatment group compared with that in the CCl4 model group. The feces albumin content in the UA treatment group was significantly lower than that observed in the CCl4 model group. All these results indicated that an impaired intestinal mucosal barrier was present in rats with liver fibrosis induced by CCl4 and that the UA treatment could effectively ameliorate intestinal mucosal barrier injury.

An impaired intestinal mucosal barrier, including decreased expression of tight junction protein, results in increased intestinal permeability, which contributes to the translocation of bacteria and/or bacterial products (LPS, bacterial DNA, etc.) (Fouts et al., 2012; De Minicis et al., 2014). Translocated bacteria and/or bacterial products lead to intestinal endotoxemia and trigger systemic inflammatory responses. In the present study, intestinal endotoxin and a systemic inflammatory response were observed in rats with CCl4-induced liver fibrosis, as demonstrated by the increased serum LPS and procalcitonin in the CCl4 model group. However, no significant differences in the serum CRP levels were observed between the CCl4 model group and control groups. This result likely occurred because CRP is less sensitive than procalcitonin in response to systemic inflammation. The increased levels of serum LPS and procalcitonin were abolished in the UA treatment group, suggesting that UA can ameliorate intestinal endotoxemia and systemic inflammatory responses. Intestinal barrier injury promotes intestinal endotoxemia and systemic inflammatory responses. After UA treatment, we observed an improved mucosal structure, decreased inflammatory cell infiltration in the lamina propria and increased expression of tight junction proteins in the ileum of rats. These results indicated that UA reduced the serum LPS and procalcitonin contents by ameliorating the intestinal mucosal barrier injury in rats with CCl4-induced liver fibrosis.

Compromised intestinal barrier function has been shown to be associated with numerous disease states, both intestinal and systemic. Various factors contribute to intestinal barrier damage

in a pathological state, especially intestinal oxidative stress. Oxidative stress promotes intestinal barrier dysfunction through a variety of mechanisms and can impair the epithelial barrier by directly oxidizing cell components and inducing cell apoptosis. ROS impacts epithelial cells by altering mucosal glycosylation and increasing bacterial adherence, internalization and translocation (Natarajan et al., 2006; Schoultz et al., 2012). MDA is produced as a result of ROS formation from the oxidation of membrane lipids and is commonly used as a marker to indicate the level of oxidative damage in tissues (Chiva et al., 2003). In our present study, oxidative stress was observed to contribute to intestinal barrier damage as indicated by the elevated MDA levels in the intestinal tissues from the CCl4 model group. Accompanied by the decreased MDA levels, an amelioration of intestinal barrier dysfunction was observed in the UA treatment group, demonstrating that UA protected the intestinal mucosal barrier by inhibiting intestinal oxidative stress.

The antioxidant system consists of enzymatic and nonenzymatic components. The enzymatic antioxidant components include superoxide dismutase, glutathione peroxidase, catalase, etc., whereas the non-enzymatic antioxidant components include vitamins, amino acids, and metalloproteins, etc. Our results showed that ileal TAC displayed no significant differences between the control and CCl4 model groups. There are two possibilities for these outcomes: either the intestinal antioxidant system is not affected in rats with CCl4-induced liver fibrosis or the antioxidant capacity of some components of the antioxidant system is destroyed and is compensated for by other components.

Several differentially localized and expressed enzymatic systems contribute to ROS formation in the gut, including the mitochondrial respiratory chain, xanthine oxidase, nitric oxide synthase, NOX, etc. (Bhattacharyya et al., 2014). NOX is a multimeric transmembrane enzyme complex that generates ROS in response to diverse stimuli. The classical phagocytic NOX consists of a heterodimeric membrane-bound flavocytochrome b558 complex, the catalytic subunit gp91phox (renamed NOX2) and the regulatory subunit p22phox located in the membrane, as well as the cytoplasmic regulatory components P67phox, p47phox, p40phox, and Rac1 (Paik et al., 2014). Intestinal NOXmediated oxidative stress is closely associated with intestinal mucosal barrier damage in a variety of pathological conditions. Xie et al. (2014) observed that oxidation protein products trigger intestinal epithelial cell death and intestinal tissue injury via a NOX-mediated redox signaling pathway in Crohn's disease. The NOX inhibitor apocynin reduces intestinal mucosal barrier injury in a rat model of severe acute pancreatitis (Deng et al., 2016). The Western blot results from the present study showed that the expression of intestinal P67phox and NOX2 was higher in the CCl4 model group than in the UA treatment group.

#### REFERENCES

Arifa, R. D., Madeira, M. F., de Paula, T. P., Lima, R. L., Tavares, L. D., Menezes-Garcia, Z., et al. (2014). Inflammasome activation is reactive oxygen species dependent and mediates irinotecan-induced mucositis through IL-1beta and IL-18 in mice. Am. J. Pathol. 184, 2023–2034. doi: 10.1016/j.ajpath.2014.03.012

Moreover, MDA levels were also higher in the CCl4 model group. Considering the amelioration of intestinal barrier dysfunction observed in the UA treatment group, these results support that UA protects the intestinal mucosal barrier in rats with hepatic fibrosis by inhibiting the expression of intestinal NOX and oxidative stress derived from NOX.

In summary, the results of the present study demonstrated that UA has protective effects on the intestinal mucosal barrier in rats with CCl4-induced liver fibrosis by inhibiting intestinal NOX-mediated oxidative stress. Given the positive effect of UA on the intestinal mucosal barrier, we anticipate that these findings could be a stepping stone for developing UA as a novel antifibrotic agent.

#### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the humane care in compliance with institutional guidelines, Institutional Animal Care and Use Committee of the First Affiliated Hospital of Nanchang University. The protocol was approved by the Institutional Animal Care and Use Committee of the First Affiliated Hospital of Nanchang University.

#### AUTHOR CONTRIBUTIONS

WZ, DG, and JJ contributed equally to this study. WZ was responsible for experiments and manuscript writing. DG was responsible for the project design. FL and SW were responsible for molecular biology experiments. CH and MJ were responsible for the cell slide and color rendering, grading, etc. YW, AW, and BL were responsible for assisting in the data processing and picture modification. XZ was responsible for the final modification of the manuscript. WZ and DG conducted the experiments, and planned and wrote the manuscript. JJ conducted the revision of the manuscript. CH, FL, SW, MJ, and YW conducted the experiments and data analysis. AW helped to perform the experiments and wrote the manuscript. BL and XZ collaborated with the other authors to correct the manuscript.

#### FUNDING

The National Natural Science Foundation of China (Grant Numbers: 81260082 and 81660110) and the "Gan-Po Talent 555" project of Jiangxi Province [Grant Number: GCZ(2012)-1] supported this research.

Bhattacharyya, A., Chattopadhyay, R., Mitra, S., and Crowe, S. E. (2014). Oxidative stress: an essential factor in the pathogenesis of gastrointestinal mucosal diseases. Physiol. Rev. 94, 329–354. doi: 10.1152/physrev.00040.2012

Chen, Z. E., and Lin, F. (2015). Application of immunohistochemistry in gastrointestinal and liver neoplasms: new markers and evolving practice. Arch. Pathol. Lab. Med. 139, 14–23. doi: 10.5858/arpa.2014-0153-RA


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Zhang, Gan, Jian, Huang, Luo, Wan, Jiang, Wan, Wang, Li and Zhu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Gut Microbiota as an Objective Measurement for Auxiliary Diagnosis of Insomnia Disorder

Bingdong Liu1,2† , Weifeng Lin<sup>1</sup>† , Shujie Chen<sup>3</sup>† , Ting Xiang<sup>1</sup> , Yifan Yang<sup>1</sup> , Yulong Yin<sup>2</sup> , Guohuan Xu<sup>2</sup> , Zhihong Liu<sup>2</sup> , Li Liu<sup>3</sup> \*, Jiyang Pan<sup>1</sup> \* and Liwei Xie2,4 \*

<sup>1</sup> Department of Psychiatry, The First Affiliated Hospital of Jinan University, Guangzhou, China, <sup>2</sup> State Key Laboratory of Applied Microbiology Southern China, Guangdong Provincial Key Laboratory of Microbial Culture Collection and Application, Guangdong Open Laboratory of Applied Microbiology, Guangdong Institute of Microbiology, Guangdong Academy of Sciences, Guangzhou, China, <sup>3</sup> Department of Infectious Diseases, Nanfang Hospital, Southern Medical University, Guangzhou, China, <sup>4</sup> Zhujiang Hospital, Southern Medical University, Guangzhou, China

#### Edited by:

Enrica Pessione, University of Turin, Italy

#### Reviewed by:

Zongxin Ling, Zhejiang University, China Giuliana Banche, University of Turin, Italy

#### \*Correspondence:

Li Liu liuli.finnu@gmail.com Jiyang Pan jiypan@163.com Liwei Xie xielw@gdim.cn †These authors have contributed

equally to this work

#### Specialty section:

This article was submitted to Systems Microbiology, a section of the journal Frontiers in Microbiology

Received: 25 March 2019 Accepted: 17 July 2019 Published: 13 August 2019

#### Citation:

Liu B, Lin W, Chen S, Xiang T, Yang Y, Yin Y, Xu G, Liu Z, Liu L, Pan J and Xie L (2019) Gut Microbiota as an Objective Measurement for Auxiliary Diagnosis of Insomnia Disorder. Front. Microbiol. 10:1770. doi: 10.3389/fmicb.2019.01770 Insomnia is a type of sleep disorder which is associated with various diseases' development and progression, such as obesity, type II diabetes and cardiovascular diseases. Recent investigation of the gut-brain axis enhances our understanding of the role of the gut microbiota in brain-related diseases. However, whether the gut microbiota is associated with insomnia remains unknown. In the present investigation, leveraging the 16S rDNA amplicon sequencing of V3-V4 region and the novel bioinformatic analysis, it was demonstrated that between insomnia and healthy populations, the composition, diversity and metabolic function of the gut microbiota are significantly changed. Other than these, redundancy analysis, co-occurrence analysis and PICRUSt underpin the gut taxa composition, signaling pathways, and metabolic functions perturbed by insomnia disorder. Moreover, random forest together with cross-validation identified two signature bacteria, which could be used to distinguish the insomnia patients from the healthy population. Furthermore, based on the relative abundance and clinical sleep parameter, we constructed a prediction model utilizing artificial neural network (ANN) for auxiliary diagnosis of insomnia disorder. Overall, the aforementioned study provides a comprehensive understanding of the link between the gut microbiota and insomnia disorder.

Keywords: insomnia, random forest, artificial neural network, redundancy analysis, cross validation

# INTRODUCTION

Sleep disorder is associated with various diseases' development and progression, such as obesity, type II diabetes (Knutson et al., 2007) and cardiovascular diseases (Drager et al., 2017). Insomnia is the most prevalent sleep disorder, including sleep apnea, restless legs syndrome (RLS) and narcolepsy, and affects a large proportion of the population on a situational, recurrent or chronic basis and is also one of the most common complaints in medical practice (Morin et al., 2015). The signature of insomnia is that patients have difficulty falling asleep, or staying awake, despite plenty of opportunity to sleep. Like many other psychiatric disorders, insomnia is a multifactorial disorder,

though the detailed pathological aspects of insomnia remain unclear. Thus, a better understanding of the pathophysiology of insomnia may provide additional therapeutic strategies.

The gut microbiome, a key component of the intestinal environment, has been implicated as an essential modulator for human health (Tilg et al., 2016). Microbial homeostasis is critical to the host development and health. Dysbiosis perturbs the host immune system and metabolism balance, which leads to the development of various kinds of diseases (Cho and Blaser, 2012; Nieuwdorp et al., 2014; Thaiss et al., 2016). Microbial dysbiosis may also contribute to the development of neurological disorders and psychiatric disorders, such as autism spectrum disorder, anxiety disorder, depression and Alzheimer's disease (Neufeld et al., 2014; Pistollato et al., 2016; Chen et al., 2017; Lach et al., 2018). In addition, several studies have provided preliminary evidence for the involvement of the gut microbiota in sleep disorders of murine models and human patients. It was reported that after 4 weeks of sleep fragmentation in experimental mice, gut flora were dominated by Lachnospiraceae and Ruminococcaceae, with a gradually reduced relative abundance of Lactobacillaceae (Poroyko et al., 2016). In another study with partial sleep deprivation in normal-weighted young individuals, the composition of the gut microbiota was subtly affected with an increased ratio of Firmicutes/Bacteroidetes (Benedict et al., 2016). However, either sleep fragmentation or sleep deprivation refers to curtailed sleep length due to an externally imposed restriction of the opportunity to sleep, while insomnia refers to the inability to fall asleep adequately, either in length or quality. Considering the significant difference in definition between sleep fragmentation/deprivation and insomnia, to date, study investigating the relationship between insomnia and gut flora remains unexplored.

Thus, in the present investigation we combined 16S rDNA amplicon sequencing and innovative bioinformatic analysis to examine the pathological and physiological significance of the gut microbiota between healthy participants and patients suffering insomnia disorder. Leveraging these innovative analyses, such as redundancy analysis, co-occurrence analysis, PICRUSt, random forest and artificial neural networks (ANN), we demonstrated that the gut taxa composition, signaling pathways, and metabolic functions are perturbed in patients with insomnia disorder. Artificial neural networks were also incorporated by utilizing the relative abundance of the gut microbiota to establish a prediction model for an unbiased evaluation of insomnia. This study is the first to combine highthroughput sequencing and bioinformatic analysis, especially machine learning, to systemically understand the biological effect of the gut microbiota on insomnia. Comprehensive analysis indicated that gut microbiota homeostasis is a strong determinant, which is closely associated with insomnia disorder. Overall, the aforementioned study provides a comprehensive understanding of the link between gut microbiota and insomnia disorder. By utilizing the machine learning approach, we identified the signature gut microbiota, which could be utilized as novel and unbiased prediction targets, which in other aspects could provide additional interventions for clinical application.

# MATERIALS AND METHODS

# Volunteer Enrollment

The experiment was approved by the Ethics Committee of Jinan University (Approval #: GNU-20180306).

The volunteers were recruited from the public and The First Affiliated Hospital of Jinan University in Guangzhou, China. After being informed on the rights and obligation, all participants understood the benefits and risks of the experiment totally and signed an informed consent document. In compliance with strict standards for inclusion and exclusion criteria (Detailed in **Supplementary Materials**), all participants were assessed by two psychiatrists. In the event of any dispute or difference of judgment, the participant would be excluded. All participants accepted polysomnography treatment at the Sleep Medicine Center of Jinan University. Finally, twenty qualified volunteers were enrolled and separated into two groups (Insomnia group and Normal Control group). Their fecal samples were collected by sterilized instruments in the morning upon polysomnography treatment, and then stored in a freezer at −80◦C for 16S rDNA sequencing.

# 16S rDNA Amplicon Sequencing

Bacterial DNA from patients' feces was extracted by utilizing the ZR Fecal DNA Kit (Zymo Research, United States). A multiplexed amplicon library covering the V3-V4 region of 16S rDNA gene was PCR-amplified with the optimized primer sets for the Illumina HiSeq 2500 sequencing instrument. A total of 1,534,966 high-quality reads were obtained, with an average of 76,748 reads (range 66,570–84,443) per sample. All chimera sequences were removed by VSEARCH (Quince et al., 2016). Chimera-free sequences were processed using a standard QIIME 1.91 pipeline (Caporaso et al., 2010) and clustered into operational taxonomic units (OTUs) at a 97% similarity threshold using an "Open-Reference" approach. Taxonomy was assigned using the RDP classifier against the Greengenes database (May 2013 release) (McDonald et al., 2012). The raw Illumina pair-end read data for all samples have been deposited in the Short Read Archive under the Bioproject: PRJNA527914.

# Bioinformatics Analysis

Alpha rarefaction was analyzed by the Faith's phylogenetic diversity (Faith, 1992), Chao1 (Chao, 1984), Shannon and Simpson index (Chao and Shen, 2003). β-diversity was estimated by computing weighted and unweighted UniFrac distance. Principal Coordinates Analysis (PCoA), Redundancy Analysis (RDA) and heatmap of correlation were plotted by "ggplot2," "vegan," and "corrplot" packages of R (version 3.5.1). Manhattan Plot was plotted by "edgeR," "dplyr" and "ggplot2" to present the differential relative abundance between groups. These results were tested by Monte Carlo permutation and Student's t-test. Organism-level microbiome phenotype prediction was obtained by BugBase software (Riaz et al., 2017). To decipher the difference of microbiota structure between groups, LEfSe (linear discriminant analysis effect size) was performed and the cladogram was graphed with default parameter (p < 0.05

and LDA score > 2.0) (Fisher, 1936). To probe the microbial metabolism and predict metagenome functional content from the marker gene, PICRUSt was utilized to explore differences of the KEGG pathway between groups (Langille et al., 2013). To decipher the gut microbiota ecology, co-occurrence analysis was performed with the "igraph" package (Nordhausen, 2015) of R with data filtered at species level considering only those relative abundance present in at least 70% of the samples in each group. The edges were estimated by Spearman confident index (abs(r) > 0.6, p < 0.05). Communities inside two networks were determined by the fast-greedy modularity optimization algorithm (Clauset et al., 2004), which was one of the approaches to determine the dense subgraph in Graph Theory. The circle bar was plotted according to the eigenvector centrality scores (ECS) to estimate the importance and betweenness of each node (Ruhnau, 2000). To identify the key signature microbiota, five-fold cross validation together with Random Forest analysis were performed to compute importance scores (mean decrease accuracy, MDA) to estimate the importance of variables by utilizing the "randomForest" v.4.6-14 package (Breiman, 2001) in R. At species level, in order to establish a prediction model to predict the sleep-related physiological parameter, the ANN was performed on python 3.6.1 with the pyTorch, sklearn, pandas, and numpy packages. The optimized parameters, including learning rate, activation function, layers, number of neurons and dropout, were selected by grid search and cross-validation.

#### RESULTS

#### Insomnia Disorder Leads to Significant Structural and Functional Changes of Gut Microbiota

Among the twenty qualified enrolled volunteers, basic personal information including height, weight and BMI presented no significant difference between groups except for age (insomnia: 33.00 ± 6.90; normal: 26.10 ± 1.85) (**Supplementary Figure S1**). Considering previous research demonstrated the gut microbiota differed little in adults based on more than 1,000 very healthy Chinese individuals (Bian et al., 2017), only 7 mean-years of difference between groups could be tolerated. All the volunteers were accepted according to inclusion and exclusion criteria (detailed in **Supplementary Materials**). All fecal samples from participants were collected for high-throughput sequencing. 16S rDNA V3-V4 region amplicon sequencing generated 1,534,966 high-quality reads, with an average of 76,748 reads (range 66,570–84,443) per sample. All raw data were filtered by VSEARCH and processed using a standard QIIME 1.91 pipeline against the Greengenes database (May 2013 release). Rarefaction measurement of Shannon and Simpson index, Goods\_Coverage, and species accumulation curve (SAC) indicated that sequencing depth was enough to capture all bacterial species and sufficient for downstream analysis (**Supplementary Figure S2**). Rarefaction analysis of chao1 (p = 0.007) and PD whole tree (p = 0.001) index showed significant difference between the healthy and insomnia groups, suggesting that insomnia disorder may result in alteration of gut microbiota diversity (**Figure 1A**). Furthermore, β-diversity calculated with the Unweighted UniFrac (p = 0.0006) and Weighted UniFrac (p = 0.0032) algorithms indicated that the insomnia and normal groups had significant structural difference by the first dimension of space distance (**Figure 1B**). To confirm the composition of difference between two groups, a Manhattan plot was used to represent the fold change of insomnia/normal group and revealed a significant difference, especially the Firmicutes and Bacteroidetes phylum, which was confirmed by Linear Discriminant Analysis Effect Size (LEfSe) analysis and identified 87 biomarkers (**Figures 1C,D**). Meanwhile, BugBase algorithm-based prediction suggested that the insomnia group preferentially enriched with the gramnegative and potential pathogenic taxa compared with the normal group (**Figure 1E** and **Supplementary Figures S3A–F**). Other than the composition and diversity of gut microbiota, PICRUSt algorithm was performed to assess the functional difference by plotting the differential pathways against KEGG database. We identified pathways such as steroid hormone biosynthesis (ko00360), Retinol metabolism (ko00830), Vitamin B6 metabolism (ko00750), Folate biosynthesis (ko00790), Citrate cycle TCA cycle (ko00020) that were predicted to be enriched in the insomnia group (Kruskal test p < 0.05), while Arachidonic acid metabolism (ko00590), Pantothenate and CoA biosynthesis (ko00770), Lysine biosynthesis (ko00300), and Glycerolipid metabolism (ko00561) associated pathways were downregulated (Kruskal test p < 0.05) (**Figure 1F**).

#### Insomnia Disorder Disturbs the Gut Flora Interaction

Whether insomnia disorder is associated with the gut microbiota community network and the network complexity, the graph theory algorithm and Co-occurrence analysis were performed to estimate the gut microbiota ecology between groups. The radar plot computed by the graph theory analysis including the transitivity, graph density, degree centralization, number of vertices and number of edges showed that insomnia disorder did not significantly change the systemic complexity of gut bacteria, indicating that the gut microbiota in insomnia patients had already developed a mature network. With this, based on species data whose relative abundance presented at least 70% of the samples in each group, Co-occurrence analysis was used to further explore the gut microbiota interaction and sub-groups in both the normal and insomnia groups (**Supplementary Figure S4** and **Supplementary Table S1**). The gut flora interaction network was significantly altered for patients under insomnia disorder compared with that of the normal group. Furthermore, the gut microbiota was sub-divided into five and four sub-groups for the normal and insomnia groups, respectively (**Figure 2**).

# Gut Microbiota Alteration Strongly Associated With Insomnia Disorder

As demonstrated above, significant structure, composition and function of the gut microbiota as well as the bacterial interaction network were significantly changed between the normal and

FIGURE 1 | Insomnia disorder leads to significant structural and functional changes of the gut microbiota. (A) α-diversity on Chao1 and PD\_whole\_tree index between insomnia and normal group. (B) β-diversity on unweighted and weighted UniFrac PCoA1 between groups. (C,D) Manhattan plot and Linear Discriminant Analysis (LDA) Effect Size (LEfSe) plot with threshold for LDA score 2.0 showed significant structural difference and identified 87 biomarkers between the two groups. (E) BugBase algorithm predicted the microbiome phenotypes of the insomnia group differed from the normal group on gram-negative and potential pathogenicity significantly. (F) To predict the metagenome function, heatmap of PICRUSt analysis showed significant KEGG pathway between groups.

insomnia groups. To further prove whether the insomniaassociated clinical sleep parameter directly contributes to the alteration of the gut microbiota, we performed the redundancy analysis (RDA) to link the insomnia parameter with the relative abundance of gut microbiota at phylum level (**Figure 3**). These clinical sleep parameters from polysomnography (PSG) and the

psychological scale include the Pittsburgh Sleep Quality index (PSQ), Hamilton Anxiety Scale (HAMA), Hamilton Depression Scale (HAMD), Epworth Sleepiness Scale (ESS) and Insomnia Severity Index (ISI). Here, we demonstrated that 67.13% of the variance could be interpreted by twelve environmental factors (in other words: clinical sleep parameter), which means that insomnia disorder could significantly alter the population of the gut microbiota at phylum level and samples from two groups were obviously separated. In particular, according to the Monte Carlo permutation test, some clinical sleep parameters, e.g., Pittsburgh sleep quality index (PSQ, r <sup>2</sup> = 0.6074, p = 0.002) and rapid eye movement sleep (REM, r <sup>2</sup> = 0.2663, p = 0.045), play a pivotal role in clustering the distribution of flora between groups. Meanwhile, ANOSIM based on the Bray Curtis distance

fmicb-10-01770 April 1, 2020 Time: 16:37 # 5

also confirmed the observation from RDA analysis that the difference between groups was more significant than that within groups (statistic R: 0.1944, p = 0.015) (**Supplementary Figure S5**). Both RDA and ANOISM analysis clearly suggested that clinical sleep parameters associated with insomnia disorder directly contribute to the separation and clustering of the gut microbiota between groups.

#### Identification of the Signature Gut Microbiota Associated With Insomnia Disorder by Random Forest

The traditional approaches such as LEfSe by comparing the difference of relative abundance of gut flora between groups resulted in the identification of 87 biomarkers. It is difficult to utilize these markers to establish a prediction model for disease diagnosis. To improve the biomarker identification, we incorporated a robust statistical analysis and applied fivefold cross-validation together with random forest to generate ∼2 million decision trees (**Supplementary Figure S6**), leading to identification of three optimal species biomarkers with consideration of lowest error rate plus standard deviation. With further analysis to identify V68 (g\_\_Prevotella) as an outlier, we thus selected V45 (g\_\_Bacteroides) and V124 (o\_\_Clostridiales) as the most important biomarkers to distinguish the insomnia patients from healthy individuals with a ROC curve at AUC = 0.87 (**Figures 4A–D**, **Supplementary Figure S7**, and **Supplementary Table S1**). Moreover, V45 was highly correlated with HAMD (r = 0.70, p < 0.001), HAMA (r = 0.62, p < 0.01), ISI (r = 0.62, p < 0.01), sleep efficiency (r = −0.56, p < 0.05), PSQ (r = 0.63, p < 0.01) and sleep latency (r = 0.66, p < 0.01) while V124 correlated with ESS (r = −0.45, p < 0.05), ISI (r = −0.48, p < 0.05), REM latency (r = −0.49, p < 0.05), and PSQ (r = −0.51, p < 0.05) (**Figure 4E** and **Supplementary Figure S8**). Even in the Co-occurrence plot, these two key microbiotas both occupied hub-like positions with high betweenness (Insomnia: V45 0.702499227 V124 0.034447479; Normal: V45 0.467046396 V124 0.044448542) (**Supplementary Figure S9**). All above results strongly demonstrated that the key microbiota we identified via the robust statistical approaches led to the

development of an optimal and robust prediction model for insomnia diagnosis.

# Relative Abundance of Gut Microbiota-Based Prediction on the Clinical Sleep Parameter

Given that the gut microbiota tightly was correlated with the clinical sleep parameter, we sought to establish a mathematical model to utilize the relative abundance of the gut microbiota to predict the sleep-related physiological parameter. Here, we utilized a well-established regression model, LASSO regression to link the relative abundance of gut microbiota and clinical sleep parameter resulted in a poor correlation (**Supplementary Figure S10**). To overcome the shortcoming of the traditional machine learning model, we integrated an even more powerful deep learning model, called an ANN, which is considered to be able to imitate biological neural networks. By integrating the clinical sleep parameter into the ANN model, this model could result in a high coefficient of determination respective for WASO number: r <sup>2</sup> = 0.14, MAE = 4.80; WASO time: r <sup>2</sup> = 0.6, MAE = 15.77; Sleep efficiency: r <sup>2</sup> = 0.52, MAE = 5.45; ESS: r <sup>2</sup> = 0.54, MAE = 2.41; HAMA: r <sup>2</sup> = 0.66, MAE = 1.81; HAMD: r <sup>2</sup> = 0.55, MAE = 1.83; ISI: r <sup>2</sup> = 0.66, MAE = 2.96; N1: r <sup>2</sup> = 0.43, MAE = 2.88; N2 r <sup>2</sup> = 0.58, MAE = 3.28; N3: r <sup>2</sup> = 0.34, MAE = 4.97; PSQ: r <sup>2</sup> = 0.73, MAE = 1.75; REM r <sup>2</sup> = 0.40, MAE = 3.59; REM latency: r <sup>2</sup> = 0.41, MAE = 38.5; Sleep latency: r <sup>2</sup> = 0.42, MAE = 6.12; Total sleep time: r <sup>2</sup> = 0.37, MAE = 45.18 (**Figure 5**).

#### DISCUSSION

Insomnia disorder as a common clinical symptom is a critical part of sleep disorder (Panossian and Avidan, 2009). It is often accompanied by excessive arousal and sleep debt, which always lead to adverse impacts such as mental or physical fatigue. From the statistics of more than 50 epidemiological studies, the prevalence of insomnia symptoms was estimated at 10∼48% (Ohayon, 2002). Insomnia disorder is functionally linked to cardiovascular and nervous system diseases (Javaheri and Redline, 2017; Tobaldini et al., 2017). The classic hypothesis is Spielman's 3P Model including predisposing, precipitating and perpetuating factors (Spielman et al., 1987). Recently, it has been reported that the hypothalamic–pituitary–adrenal axis (HPA) may contribute to the incidence of insomnia (Levenson et al., 2015). Moreover, in 2017 a genome-wide association study (GWAS) identified risk genomic loci and genes that are associated with the incidence of insomnia, and suggested that insomnia is highly polygenic (Hammerschlag et al., 2017). However, none of these studies provided a mechanistic interpretation of the causes or even objective approaches for insomnia diagnosis. Here, our study is first to comprehensively compare the gut microbiota between insomnia patients and healthy individuals. With these data, we established a robust statistical prediction model to utilize the relative abundance of the gut microbiota to distinguish insomnia patients from the normal population and to estimate levels of sleep quality through the novel bioinformatics technology and machine learning algorithm.

The unhealthy shift of gut microbiota, also called dysbiosis, is associated with various metabolic diseases such as obesity, type II diabetes, hypertension and cardiovascular diseases (Clarke et al., 2014; Kristensen et al., 2016; Adnan et al., 2017; Sun et al., 2018). In this study, we demonstrated that α- and β-diversity of the gut microbiota in insomnia patients is significantly altered. Meanwhile, by comparing the difference of the relative abundance between insomnia and healthy individuals, we identified that Bacteroidetes are the dominant taxa in the insomnia group, while Firmicutes and Proteobacteria were enriched in the normal group, resulting in a decreased ratio of Firmicutes/Bacteroidetes. Our results are different from observations made in previous studies in individuals with sleep deprivation or restriction. In their studies, the F/B ratio shows either no change or is increased after partial sleep deprivation (Benedict et al., 2016; Bushman et al., 2017). This discrepancy with respect to the change over the F/B ratio may be due to the difference regarding the clinical definition between sleep restriction/deprivation and insomnia. Sleep deprivation or restriction is not considered to be a specific disease, but rather a result of a wide range of interruption from external environmental factors. It is worth mentioning that although subjects in previous study followed strict experimental protocol, they not only had ad libitum access to food/drink throughout the experiment, but also were allowed to read, play video or board games, watch television, and interact with laboratory staff to help remain awake (Bushman et al., 2017). These environmental factors may contribute to variation in the gut microbiota, which leads to difficulty in interpreting the results. Compared to those with sleep deprivation, patients with insomnia who do not have externally imposed restrictions on the opportunity to sleep still have trouble falling asleep, staying asleep, or waking too early, resulting in daytime impairment (Brown, 2005). Our study demonstrated that although insomnia and sleep deprivation may result in similar reductions on sleep length in most cases, they may lead to different consequences regarding the dysbiosis of the gut microbiota. In addition, this ratio change is also reported in different life stages and pathological circumstances. A study looks into the ratio of F/B between adults and elders, suggesting that a higher ratio in the adult gut is observed, while it starts to decrease in individuals undergoing aging (Doré et al., 2009). Alteration of the F/B ratio is also observed in those with metabolic diseases, such as obesity (Turnbaugh et al., 2006; Koliada et al., 2017) and type II diabetes (Vogensen et al., 2010). Furthermore, our BugBase-based phenotypical prediction also demonstrated that the insomnia group was enriched with bacterial taxa to be potentially pathogenic. This may link the insomniarelated sleep disorder population with high potential disease development and progression, as more evidence has proved that chronic sleep disorder is associated with a multitude of health conditions and even systemic metabolic disorder (Van Cauter et al., 2008).

The biological and physiological function of gut microbiota could be defined from multiple aspects, such as the taxonomic

composition and diversity, which are poorly conserved across individuals while the genetic composition and functional capacity are evolutionally conserved across individuals. Thus, to decipher the metabolic switch of gut bacteria, the PICRUSt algorithm was utilized to map the bacterial genetic pathway against the KEGG database. Compared to normal group, a wide range of pathways was altered obviously in our study. It is interesting to note that vitamin B-related pathways were significantly induced in the insomnia group, while the level of vitamins is highly associated with the clinical practice of insomnia (Lichstein et al., 2007). In our insomnia patients, the analysis suggested vitamin B6 catabolism (ko00750) in the gut microbiota is significantly enhanced, resulting in vitamin B6 deficiency for the host. It was reported that vitamin B6 is administered as a common therapeutic practice for insomnia disorder and its deficiency results in fatigue and depression (Baldewicz et al., 1998). Thus, additional vitamin B6 supplementation could ameliorate insomnia symptoms (Baldewicz et al., 1998). Moreover, the folate (also called vitamin B9) biosynthesis-related pathway (ko00790) was also increased in the insomnia group. Previous study of serum nutritional biomarkers and dietary supplementation of folate demonstrated that folate acid has a high correlation with the development of sleep disorder (Sato-Mito et al., 2011; Zonderman et al., 2014). In addition, endogenously synthesized arachidonic acid significantly facilitates the release of GABA in the striatum (Chéramy et al., 1996), while GABA could enhance the catabolism of serotonin into N-acetylserotonin (the precursor of melatonin) in rat (Balemans et al., 1983). It has long been speculated that GABA is associated with the synthesis of melatonin and thus might exert regulatory effects on sleep functions. In our study, our bioinformatic analysis demonstrates that arachidonic acid biosynthesis was lower in the insomnia group, indicating that lower production of arachidonic acid from gut microbiota may be associated with a high incidence of insomnia. However, whether arachidonic acid supplementation may improve insomnia symptoms requires further clinical investigation. These results provide a link that gut microbiota and their metabolites maybe a mediator with respect to the development of insomnia. With this information, novel therapeutic and intervention approaches could be developed for people suffering from insomnia disorder in the future.

In our study, insomnia disorder leads to the alteration of the gut microbiota composition and diversity. However, whether insomnia disorder directly contributes to the dysbiosis of the gut microbiota is still unknown. Here, our RDA analysis and ANOSIM provide strong evidence to support the role of clinical sleep parameters of insomnia individuals in dysbiosis of gut microbiota, especially PSQ (r <sup>2</sup> = 0.6074, p = 0.002) and REM (r <sup>2</sup> = 0.2663, p = 0.045), based on a Monte Carlo permutation test. Both results strongly pinpointed the importance of insomnia disorder as a key factor in separating gut microbiota from two groups. Upon establishment of the link between gut microbiota and clinical sleep parameter, taking advantage of the differential test and LEfSe algorithm, we identified 87 differential biomarkers from the normal and insomnia groups. Among the biomarkers, to further classify their importance, the machine learning approach is incorporated, such as random forest. This robust statistical method could identify the signature biomarkers with higher prediction accuracy and coefficiency, especially for the gut microbiota-based diseases prediction and diagnosis (Ren et al., 2018; Zhu et al., 2018). Here, our random forest model together with the cross-validation model identified two key bacterial taxa (g\_\_Bacteroides; o\_\_Clostridiales), which are not only tightly associated with clinical data, but also play a pivotal role in network of gut ecology and could be used as two critical biomarkers to identify patients with insomnia.

Classic diagnosis for insomnia disorder relies on either subjective or objective assessment, including the most common clinical sleep parameters such as PSQ, ESS, ISI, HAMD, and HAMA. However, most of these results are often affected by the subjectivity of individuals, especially for some patients with insomnia disorder (Åkerstedt and Gillberg, 1990; Landry et al., 2015). On the other hand, PSG, as the first choice for objective assessment, is the golden standard for insomnia diagnosis worldwide. This is restricted by the cost, equipment and space. Furthermore, the adaptation of the first-night sleep may affect the PSG results because of the temporary change of sleep environment (Tamaki et al., 2016). Thus, a convenient approach is necessary for the diagnosis of insomnia. Given the tight correlation between microbiota and disease incidence, whether there is a method to establish a regression model to predict clinical sleep parameter remains unclear. So, we introduced a LASSO regression model, which is widely used in gut microbiota-based clinical study and has been shown to effectively utilize the relative abundance to predict cancer development and progression, such as irritable bowel diseases and colorectal cancer (Tap et al., 2017; Flemer et al., 2018). Moreover, LASSO could overcome the multicollinearity problem caused by the interaction between microbes (Tibshirani, 1996), while regression models were limited in microbiology study (Alin, 2010). However, in our insomnia case, the LASSO model could not collect enough fitness for the current study (**Supplementary Figure S10**). To overcome the limitation of LASSO regression, we introduced ANN, which was originally developed to imitate the biological neural networks of the brain (McCulloch and Pitts, 1943). ANN is not only an algorithm, but also a frame for different machine learning algorithms to incorporate and work together to process complex data. As it works in the same way as the human brain, compared to traditional machining learning such as LASSO, ANN brought out stronger and more robust ability to deal with complex data, offered a good prediction model with high fitness, and thus was applied to various areas, especially in quantum chemistry (Balabin and Lomakina, 2009), general game playing (Silver et al., 2016), 3D reconstruction (Choy et al., 2016), and medical diagnosis (Kamruzzaman et al., 2004). In these areas, ANN like LASSO regression could also effectively and practically address the multicollinearity problem. Thus, we incorporated an ANN prediction model to assess sleep quality based on the relative abundance of the gut microbiota. Although based on few samples, this model could still obtain good fitness. With this, we are able to utilize the relative abundance of the gut microbiota to provide an alternative and accurate approach for insomnia diagnosis.

#### CONCLUSION

fmicb-10-01770 April 1, 2020 Time: 16:37 # 10

The model proposed in the current study utilizes the cutting edge bioinformatic algorithm to not only underpin the difference between insomnia and normal health, but also take advantage of ANN to establish the prediction model for insomnia diagnosis and sleep quality evaluation based on the relative abundance of the gut microbiota. Although all methods above are only based on bioinformatics and mathematics, we believe these approaches could validate the results and further prove that even with a small sample size. With this, we could still be able to draw a solid conclusion. Of course, more cases will be collected to provide further evidence in our future work. This will open another gate and a new perspective for the development of novel therapeutic strategies by taking advantage of the information from the gut microbiota.

#### DATA AVAILABILITY

The generated datasets for this study can be found on BioProject accession number PRJNA527914.

#### ETHICS STATEMENT

The experiment was proved by the Ethics Committee of Jinan University and recruited volunteers in public and The First Affiliated Hospital of Jinan University in Guangzhou, China (Approval #: GNU-20180306).

#### AUTHOR CONTRIBUTIONS

LX, LL, and JP designed the experiments. LX and JP collected the grant support. BL, SC, and ZL performed the data analysis. WL, TX, GX, YlY, and YfY collected participants' feces. BL and LX drafted the manuscript.

#### REFERENCES


#### FUNDING

This work was supported by the National Natural Science Foundation of China (Grant No. 8187050617), "GDAS" Project of Science and Technology Development (Grant No. 2019GDASYL-0402001), Natural Science Foundation of Guangdong Province (Grant No. 2017A030313136) and Collaborative Innovation of Industry, University and Research in Guangzhou (Grant No. 201802030013).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.01770/full#supplementary-material

FIGURE S1 | The general information including BMI (A), Height (B), Weight (C), and Age (D) between insomnia and normal group.

FIGURE S2 | Rarefaction measurement of Shannon (A) and Simpson (B) index presented a saturate platform indicated sequencing depth was enough to capture all bacterial species while Good's coverage index (C) and species accumulation curve (D) confirmed the sampling was sufficient for the experiment design on OTU taxa.

FIGURE S3 | BugBase algorithm was used to predict microbiome phenotypes including Gram positive (A) or negative (B), aerobic (C) or anaerobic (D), Potential\_Pathogenic (E), and Forms\_Biofilms (F) with Mann-Whitney U test.

FIGURE S4 | Radar plot on transitivity, number of edges, number of vertices, degree of centralization, and graph density indicated the gut microbiota in each group developed a mature network with almost same complexity.

FIGURE S5 | Analysis of similarity (ANOSIM) revealed the difference between groups was more significant than that within groups (statistic R: 0.1944, p = 0.015).

FIGURE S6 | The detailed results of random forest in ten different random seed are presented.

FIGURE S7 | Based on two key bacterial taxa (V45, V124), the random forest prediction obtained an accurate rate with the ROC curve at AUC = 0.87.

FIGURE S8 | The detailed results of correlation analysis, these key taxa and clinical sleep parameter were mapped.

FIGURE S9 | Co-occurrence network in each group plotted with "auto layout" parameter in "igraph" packages shows V45 and V124 occupied hub-like position in each network.

FIGURE S10 | LASSO regression model to utilize the relative abundance of bacterial taxa to predict clinical sleep parameter.


Firmicutes/Bacteroidetes ratio in an adult Ukrainian population. BMC Microbiol. 17:120. doi: 10.1186/s12866-017-1027-1021


fmicb-10-01770 April 1, 2020 Time: 16:37 # 11

neural networks and tree search. Nature 529, 484–489. doi: 10.1038/nature 16961


capacity for energy harvest. Nature 444, 1027–1031. doi: 10.1038/nature 05414


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Liu, Lin, Chen, Xiang, Yang, Yin, Xu, Liu, Liu, Pan and Xie. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

, Yulong Yin<sup>2</sup>

,

# Corrigendum: Gut Microbiota as an Objective Measurement for Auxiliary Diagnosis of Insomnia Disorder

#### Edited and reviewed by:

*Emilio M. Ungerfeld, Institute of Agricultural Research, Chile*

#### \*Correspondence:

*Li Liu liuli.finnu@gmail.com Jiyang Pan jiypan@163.com Liwei Xie xielw@gdim.cn*

*†These authors have contributed equally to this work*

#### Specialty section:

*This article was submitted to Systems Microbiology, a section of the journal Frontiers in Microbiology*

Received: *07 January 2020* Accepted: *09 March 2020* Published: *02 April 2020*

#### Citation:

*Liu B, Lin W, Chen S, Xiang T, Yang Y, Yin Y, Xu G, Liu Z, Liu L, Pan J and Xie L (2020) Corrigendum: Gut Microbiota as an Objective Measurement for Auxiliary Diagnosis of Insomnia Disorder. Front. Microbiol. 11:510. doi: 10.3389/fmicb.2020.00510* Guohuan Xu<sup>2</sup> , Zhihong Liu<sup>2</sup> , Li Liu<sup>3</sup> \*, Jiyang Pan<sup>1</sup> \* and Liwei Xie2,4 \* *<sup>1</sup> Department of Psychiatry, The First Affiliated Hospital of Jinan University, Guangzhou, China, <sup>2</sup> State Key Laboratory of*

, Yifan Yang<sup>1</sup>

*Applied Microbiology Southern China, Guangdong Provincial Key Laboratory of Microbial Culture Collection and Application, Guangdong Open Laboratory of Applied Microbiology, Guangdong Institute of Microbiology, Guangdong Academy of Sciences, Guangzhou, China, <sup>3</sup> Department of Infectious Diseases, Nanfang Hospital, Southern Medical University, Guangzhou, China, <sup>4</sup> Zhujiang Hospital, Southern Medical University, Guangzhou, China*

Keywords: insomnia, random forest, artificial neural network, redundancy analysis, cross validation

Bingdong Liu1,2†, Weifeng Lin1†, Shujie Chen3†, Ting Xiang<sup>1</sup>

#### **A Corrigendum on**

#### **Gut Microbiota as an Objective Measurement for Auxiliary Diagnosis of Insomnia Disorder**

by Liu, B., Lin, W., Chen, S., Xiang, T., Yang, Y., Yin, Y., et al. (2019). Front. Microbiol. 10:1770. doi: 10.3389/fmicb.2019.01770

In the original article, there was an error in the article title. There was typo in our accepted research article. The title of the article was "Gut Microbiota as a Subjective Measurement for Auxiliary Diagnosis of Insomnia Disorder."

The word "subjective" should be "objective." Thus, the correct title is "Gut Microbiota as an Objective Measurement for Auxiliary Diagnosis of Insomnia Disorder."

The authors apologize for this error and state that this does not change the scientific conclusions of the article in any way. The original article has been updated.

Copyright © 2020 Liu, Lin, Chen, Xiang, Yang, Yin, Xu, Liu, Liu, Pan and Xie. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Dietary Supplementation With Leucine or in Combination With Arginine Decreases Body Fat Weight and Alters Gut Microbiota Composition in Finishing Pigs

*Chengjun Hu1,2 , Fengna Li1 , Yehui Duan1 , Yulong Yin1,2 and Xiangfeng Kong1 \**

*1 Hunan Provincial Key Laboratory of Animal Nutritional Physiology and Metabolic Process, Key Laboratory of Agro-Ecological Processes in Subtropical Region, National Engineering Laboratory for Pollution Control and Waste Utilization in Livestock and Poultry Production, Institute of Subtropical Agriculture, Chinese Academy of Sciences, Changsha, China, 2 Guangdong Provincial Key Laboratory of Animal Nutrition Control, Institute of Subtropical Animal Nutrition and Feed, College of Animal Science, South China Agricultural University, Guangzhou, China*

#### *Edited by:*

*Yuheng Luo, Sichuan Agricultural University, China*

#### *Reviewed by:*

*Maryam Dadar, Razi Vaccine and Serum Research Institute, Iran Wenkai Ren, South China Agricultural University, China*

> *\*Correspondence: Xiangfeng Kong nnkxf@isa.ac.cn*

#### *Specialty section:*

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology*

*Received: 06 April 2019 Accepted: 17 July 2019 Published: 13 August 2019*

#### *Citation:*

*Hu C, Li F, Duan Y, Yin Y and Kong X (2019) Dietary Supplementation With Leucine or in Combination With Arginine Decreases Body Fat Weight and Alters Gut Microbiota Composition in Finishing Pigs. Front. Microbiol. 10:1767. doi: 10.3389/fmicb.2019.01767*

Obesity was associated with change in gut microbiota composition and their metabolites. We investigated the effects of dietary supplementation with leucine (Leu) in combination with arginine (Arg) or glutamic acid (Glu) on body fat weight, composition of gut microbiota, and short-chain fatty acids (SCFAs) concentration in the colon. Forty-eight Duroc × Large White × Landrace pigs with an initial body weight of 77.08 ± 1.29 kg were randomly assigned to one of the four groups (12 pigs per group). The pigs in the control group were fed a basal diet supplemented with 2.05% alanine (isonitrogenous control, BD group), and those in the three experimental groups were fed a basal diet supplemented with 1.00% Leu + 1.37% alanine (Leu group), 1.00% Leu + 1.00% Arg (Leu\_Arg group), or 1.00% Leu + 1.00% Glu (Leu\_Glu group). We found that dietary supplementation with Leu alone or in combination with Arg decreased (*p* < 0.05) body fat weight, and increased (*p* < 0.05) colonic propionate and butyrate concentrations compared to the BD group. The mRNA expression levels of genes related to lipolysis increased (*p* < 0.05) in the Leu or Leu\_Arg group compared to the BD group. Negative relationships (*p* < 0.05) were observed between body fat weight, colonic propionate, and butyrate concentrations. Compared to the BD group, the abundance of *Actinobacteria* was higher (*p* < 0.05) in the Leu group, and that of *Clostridium\_sensu\_ stricto*\_1, *Terrisporobacter*, and *Escherichia-Shigella* were higher in the Leu\_Arg group. The abundance of *Deinococcus-Thermus* was negatively correlated (*p* < 0.05) with body fat weight, and was positively correlated (*p* < 0.05) with butyrate, isovalerate, propionate, and isobutyrate concentrations, and that of *Cyanobacteria* was positively correlated (*p* < 0.05) with butyrate, propionate, and isobutyrate concentrations. In conclusion, these findings suggest that decreased body fat weight in pigs can be induced by Leu supplementation alone or in combination with Arg and is associated with increased colonic butyrate and propionate concentrations. This provides new insights for potential therapy for obesity.

Keywords: arginine, colon, dietary supplementation, glutamic acid, leucine, microbiota, short-chain fatty acid

# INTRODUCTION

Over the past few decades, obesity has increased from 16.8% in 2007–2008 to 18.5% in 2015–2016 among youth, and from 33.7% in 2007–2008 to 39.6% among adults (Hales et al., 2018). Obesity exerts a negative impact on human health, including causing insulin resistance, diabetes mellitus, cancer, inflammation, sleep apnea, and other chronic diseases (Zhang et al., 2018). It causes more than 3.4 million deaths worldwide (Lim et al., 2013). Although obesity is the one of most important public health challenge (Simmonds et al., 2016), there are few types of medication available for preventing and treating this disease.

The amino acid leucine (Leu) is a substrate for protein synthesis and is involved in the regulation of fat metabolism (Yao et al., 2016). Studies have confirmed that Leu has the potential to prevent and treat obesity. For instance, Leu supplementation with 50% food restriction results in lower body fat in rats than those subjected to the same 50% food restriction (Donato et al., 2006). Increased dietary Leu intake reduces diet-induced obesity and improves glucose metabolism (Zhang et al., 2007). In addition, Leu treatment improves mitochondrial biogenesis, fatty acid oxidation, insulin sensitivity, and glucose metabolism in dietinduced obesity in mice (Guo et al., 2010; Li et al., 2012). Arginine (Arg) and glutamic acid (Glu) also play important roles in fat metabolism. Dietary supplementation with 1% Arg reduces body fat accumulation in pigs (Tan et al., 2009), and supplementation with 0.24% L-Arg-HCl in drinking water reduces fat accretion in adult ZDF rats (Wu et al., 2007). Our previous study showed that dietary supplementation with 1.00% Glu decreased back fat thickness in finishing pigs (Hu et al., 2017a), indicating that body fat accumulation declined with Glu treatment. Although studies have demonstrated that Arg and Glu play vital roles in preventing obesity, the effects of dietary supplementation with Leu in combination with Arg or Glu on fat accumulation are still unknown.

Changes in the gut-microbiota community have been proposed as possible causes of obesity. There are distinct differences at the phylum level in the microbiota community between obese and lean subjects; obese subjects have lower bacterial diversity, and different metabolic pathways (Turnbaugh et al., 2009). In addition, inoculation of the microbiota in adult obese mice into germ free mice increases the total body fat in germ free mice (Turnbaugh et al., 2006), suggesting gut microbiota as a contributing factor to obesity. Short-chain fatty acids (SCFAs) are produced by the microbiota in the large bowel *via* fermentation of carbohydrates and amino acids (Rios-Covian et al., 2016) and have key roles in anti-obesity. For instance, butyrate protects against diet-induced obesity by reducing food intake and increasing energy expenditure (Gao et al., 2009; Lin et al., 2012), and dietary SCFA supplementation prevented and reversed high-fat diet-induced obesity in mice by decreasing peroxisome proliferator activated receptor-γ expression and activity (den Besten et al., 2015). Amino acids in the large intestine are the substrate for SCFAs. These findings motivated us to investigate whether Leu in combination with Arg or Glu alters SCFAs concentrations in the colon. The pig is one of the most commonly used model animals in biomedical studies on human obesity (Houpt et al., 1979). Therefore, our objective was to investigate the effects of dietary supplementation with Leu in combination with Arg or Glu on body fat weight, composition of the gut microbiota, and SCFA concentration in the colons in pigs.

#### MATERIALS AND METHODS

#### Ethics Statement

The protocol for this study was approved by the Committee on the Ethics of Animal Experiments of the Institute of Subtropical Agriculture, Chinese Academy of Sciences under ethic approval number ISA-2017-023, and was conducted in accordance with the recommendations of the Guide for the Care and Use of Laboratory Animals of the Institute of Subtropical Agriculture, Chinese Academy of Sciences.

#### Animals and Experimental Treatments

Forty-eight Duroc × Large White × Landrace pigs were selected and randomly assigned to one of four groups (12 pigs per group). The average body weight of the pigs used for this study was 77.09 kg. The pigs in the control group were fed a basal diet supplemented with 2.05% alanine (isonitrogenous control, BD group), and those in the three experimental groups were fed a basal diet supplemented with 1.00% Leu + 1.37% alanine (Leu group), 1.00% Leu + 1.00% Arg (Leu\_Arg group), or 1.00% Leu + 1.00% Glu (Leu\_Glu group). The basal diet was formulated on the basis of nutrient requirements established by the National Research Council (2012) (**Supplementary Table S1**). The amino acids Ala, Leu, Arg, and Glu were obtained from Wuxi Jinghai Amino Acid Co., Ltd. (Wuxi, China). The pigs were housed in cages (3.5 m × 5.0 m) and fed for 60 days. The pigs had 24 h access to feed and water. The final body weight of pigs was not affected by diet treatment.

#### Sample Collection and Body Fat Weight Determination

At the end of the trial, blood samples were obtained from the jugular vein of fasted pigs using 10 ml centrifuge tubes containing sodium heparin; samples were centrifuged at 3,000 ×*g* for 15 min to recover the plasma. After that, the pigs were slaughtered by electric shock (120 V, 200 Hz) and exsanguination. After removing the head, feet, tail, and internal organs, the carcass was cut into right and left parts longitudinally. The body fat was removed from the right side of the carcass and then weighed. The fat between the sixth and seventh ribs were immediately collected from the right side of the carcass, and snap-frozen in liquid nitrogen (approximately, 10 g per sample) and then stored at −80°C for mRNA analysis. The colon was quickly separated and the luminal contents were collected in sterile tubes and stored at −80°C for laboratory analysis.

#### RNA Extraction and Complementary DNA Synthesis

Total RNA was extracted from the fat tissue (approximately 50 mg per sample) using TRIzol reagent (Life Technologies, Carlsbad, CA, USA). The RNA concentration and 260:280 nm ratio of each sample was measured using the NanoDrop®

ND-1000 instrument (Thermo Fisher, Wilmington, DE, USA). RNA integrity was determined using 1% agarose gel electrophoresis. All RNA samples examined in this study showed the 5S, 18S, and 28S rRNA bands. Complementary DNA (cDNA) was synthesized from 1,000 ng RNA in a 20 μl reaction volume using a PrimeScript® first strand cDNA synthesis kit (Takara, Osaka, Japan), and stored at −80°C until further analysis.

#### Real-Time Polymerase Chain Reaction Analysis

Primers selected for polymerase chain reaction (PCR) analyses were designed using Primer 31 and are listed in **Supplementary Table S2**. Total reaction volumes (10 μl) comprised of 2 μl cDNA template solution, 5 μl SYBR Green PCR master mix (Thermo Fisher Scientific, Inc., Waltham, MA, USA), 2.2 μl water, and 0.4 μl of each primer. The relative expression levels of genes were determined using the ABI 7900HT system (Applied Biosystems, Carlsbad, CA, USA) and three replicates per biological sample. The RT-PCR program included a 10-min incubation at 95°C, followed by 40 cycles of denaturation for 15 s at 95°C and annealing and extension for 20 s at 60°C. A melting curve program (60–99°C with a heating rate of 0.1°C/s) and fluorescence measurement was performed to generate melting curves for each sample, check primer specificity, and ensure the purity of PCR products. The gene glyceraldehyde-3-phosphate dehydrogenase (*GAPDH*) was used to normalize the mRNA levels of the selected genes. The relative expression level of mRNA was calculated according to the following formula (Hu et al., 2017b): R = 2−∆∆Ct(sample-control), where ∆∆Ct(sample-control) = (Cttarget gene – Ct*GAPDH*) treated – (Cttarget gene – Ct*GAPDH*) control.

#### Plasma Biochemical Parameters

Plasma total cholesterol (TC), triglycerides (TG), low density lipoprotein-cholesterol (LDL-C), high density lipoproteincholesterol (HDL-C), and lipase were measured using a biochemical analytical instrument TBA-120FR (Toshiba, Otawara-shi, Japan) and respective commercial assay kits (Yonghe-Yangguang Science and Technology Co., Ltd., Changsha, China) according to the manufacturers' instructions.

#### Metabolite Concentrations in Colonic Contents

The SCFAs, including acetate, propionate, butyrate, isobutyrate, pentanoate, and isopentanoate were analyzed by gas chromatography as described previously (Ji et al., 2018). Bioamines, including putrescine, cadaverine, spermidine, spermine, and tyramine were measured by high-performance liquid chromatography as described previously (Ji et al., 2018).

#### DNA Extraction and 16S rRNA Gene Sequencing

Total microbial DNA was extracted from colonic content samples (*n* = 6 per group) using the HiPure Stool DNA kit B (Magen, Shanghai, China). The DNA concentration was measured using the NanoDrop® ND-1000 instrument (NanoDrop Technologies Inc., USA). The V3-V4 region of the 16S rRNA gene was amplified using the universal primers 515F (5′-GTGCCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) (Zeng et al., 2017). The PCR was performed in a total volume of 20 μl comprising 1 μl of DNA, 2 μl of deoxyribonucleotide triphosphate, 4 μl of 5-fold FastPfu buffer (TransGen Biotech, China), 0.4 μl of FastPfu polymerase (TransGen Biotech, China), 0.8 μl (5 μM) of each primer, and 11 μl ddH2O. The PCR program included a 3-min incubation at 95°C, followed by 27 cycles of denaturation at 95°C for 30 s, and annealing and extension at 55°C for 30 s and at 72°C for 45 s. The amplified PCR product was clearly identified using 1.2% agarose gels. Amplicons were then extracted from the agarose gels and purified using the SanPrep DNA Gel Extraction kit (Sangon Biotech, Shanghai, China). Purified amplicons were then subject to paired-end sequencing on the Illumina MiSeq platform (Illumina, Sand Diego, CA, USA) according to the manufacturer's instructions; this was performed by a commercial service provider (Shanghai Majorbio Bio-pharm Technology Co., Ltd., Shanghai, China).

#### Bioinformatics Analysis

Raw Illumina fastq files were quality-filtered, de-multiplexed, and analyzed using Trimmomatic (v.0.30) (Bolger et al., 2014) and FLASH (v.1.2.11) (Magoc and Salzberg, 2011) software packages; they were filtered to eliminate adapters and lowquality reads to obtain clean reads; overlapped paired-end reads were then merged to create tags. The main steps were as follows: (1) the bases with a trailing quality score < 20 were removed; (2) 300 bp reads at any site receiving an average quality score < 20 over a 50-bp sliding window were truncated, and truncated reads of <50 bp were removed; (3) merged reads with a mismatch ratio in the overlapping regions of <0.2 were removed, whereas the sequences that overlapped by >10 bp were assembled according to their overlap sequence; and (4) reads with primer mismatches >2 and with barcode mismatches >0 were removed. Tags were clustered into operational taxonomic units (OTUs) with sequence similarity of 97% using USEARCH (v7.0.1090) (Edgar, 2013). Representative OTU sequences were taxonomically classified using the Ribosomal Database Project (RDP) Classifier based on the Greengene (V201305) reference database. Alpha diversity values for colonic bacterial communities were estimated using the ACE, the bias-corrected Chao richness estimator, Shannon, and Simpson indices. Partial least squares discriminant analysis (PLS-DA) was used to analyze the unadjusted means of OTU-level microbial abundances. The 16S rRNA gene sequence was submitted to the NCBI Sequence Read Archive database under accession numbers SRR9672929–SRR9672950.

#### Statistical Analyses

Total fat weight, colonic metabolite, and alpha diversity indices were analyzed using one-way analysis of variance (ANOVA) and Duncan's multiple-range *post hoc* test in SPSS software (v20.0; SPPS Inc., Chicago, IL, USA). The relative

<sup>1</sup> http://bioinfo.ut.ee/primer3-0.4.0/

species abundances and overall composition (at phyla and genera level) of gut microbial communities were analyzed using the Kruskal-Wallis test. Pearson's correlation coefficient was used to assess the relationships between body fat weight, SCFA levels, and the relative abundances of phyla. LEfSe was used to identify different taxa microbes among lines using default parameters. Results were expressed as means ± SEM. Differences were considered statistically significant at *p* less than 0.05.

#### RESULTS

#### Body Fat Weight and Plasma Analysis

As shown in **Figure 1**, dietary supplementation with Leu or Leu\_Arg reduced (*p* < 0.05) body fat weight in finishing pigs, whereas supplementation with Leu\_Glu did not (*p* < 0.05) affect body fat weight, relative to the BD group.

The plasma TG and LDL-C concentrations were lower (*p* < 0.05) in the Leu and Leu\_Arg groups than in the BD group. The plasma TC concentration was lower (*p* < 0.05) in the Leu group than in the BD or Leu\_Glu group.

#### Expression of Fat Metabolism-Related Genes

As shown in **Figure 2**, no significant differences (*p* > 0.05) were observed in the mRNA expression levels of lipoprotein lipase (*LPL*), peroxisome proliferator-activated receptor γ (*PPARγ*), acetyl-coA carboxylase (*ACC*), and fatty acid synthase (*FAS*) among the four dietary groups. The mRNA expression level of hormone-sensitive lipase (*HSL*) was higher (*p* < 0.05) in the Leu and Leu\_Arg groups than in the BD group, and that of carnitine palmitoyl transferase-I (*CPT*-1) was higher (*p* < 0.05) in the Leu\_Arg group than in the BD or Leu\_Glu group.

#### Concentrations of Short-Chain Fatty Acids and Bioamines in Colonic Contents

As shown in **Figure 3**, the concentrations of acetate, valerate, isobutyrate, and isovalerate were not affected (*p* < 0.05) by diet. The concentrations of propionate and butyrate were significantly elevated (*p* < 0.05) in the Leu and Leu Arg groups relative to the BD group. Moreover, the concentration of propionate was higher (*p* < 0.05) in the Leu\_Glu group than in the BD group (*p* < 0.05). Correlations between SCFAs

concentrations and body fat weight are presented in **Figure 4**. Significant negative correlations (*p* < 0.05) were observed between body fat weight and the colonic concentrations of propionate, butyrate, and isobutyrate.

As shown in **Figure 5**, no differences (*p* < 0.05) were observed in the concentrations of putrescine, cadaverine, spermidine, spermine, and tyramine among the dietary treatment groups.

#### Diversity of Colonic Bacterial Communities

After size filtering, quality control, and chimera removal, 874,597 valid sequences were obtained, with an average of 39,754 sequences per colonic sample (**Supplementary Table S3**). These sequences were assigned to 1,098 OTUs. Overall, 971, 918, 919, and 969 OTUs were obtained from pigs in the BD, Leu, Leu\_Arg, and Leu\_Glu dietary treatments, respectively (**Supplementary Table S3**). Dietary supplementation with Leu, Leu\_Arg, or Leu\_Glu did not affect (*p* < 0.05) the ACE, Chao, Sobs, Shannon, and Simpson indices of the sampled bacterial communities (**Figures 6A–E**). According to PLS-DA, samples from the Leu and Leu\_Arg groups were clustered together (**Figure 6F**).

#### Colonic Bacterial Community Structure

The most dominant phyla in the bacterial communities (comprising >1% of the community) were *Firmicutes*, *Bacteroidetes*, *Proteobacteria*, *Spirochaetes*, *Tenericutes*, and *Actinobacteria* (**Figure 7A**), comprising >97% of the total colonic bacteria found in gut samples. *Firmicutes* was observed at highest abundance in the Leu group (73.85%). The abundance of *Bacteroidetes* was lower, while that of *Proteobacteria* was higher, in the Leu\_Arg group than in the BD group. The abundance of *Actinobacteria* was higher (*p* < 0.05) in the Leu group than in the BD group (**Figure 8A**).

**Figure 7B** shows the distribution of the abundances of the bacterial genera (>1%) among the four treatment groups. The abundances of *Clostridium\_sensu\_stricto\_*1, *Terrisporobacter*, and *Escherichia-Shigella* were highest in the Leu\_Arg group, and that of *Lactobacillus* was highest in the Leu group. Dietary supplementation with Leu or Leu\_Arg reduced (*p* < 0.05) the abundances of *Ruminiclostridium\_6* and *norank\_f\_Clostridiales\_ vadinBB60\_group* (**Figure 8B**). The abundances of *norank\_f\_ Coriobacteriaceae* and *Collinsella* were higher (*p* < 0.05) in the Leu group than in the BD group.

To further compare the taxonomic difference among the four groups, LEfSe analysis was used to assess the differential

abundance of bacterial taxa (**Figure 8C**): six bacterial biomarkers were differentially abundant among the four groups. *Actinobacteria* and *Coriobacteriales* were the dominant microbes in the Leu group.

Relationships Between Bacterial Community Composition, Body Fat Weight, and Metabolite Concentrations

As shown in **Figure 9**, the abundance of *Deinococcus-Thermus* was negatively correlated (*p* < 0.05) with body fat weight, and was positively correlated (*p* < 0.05) with butyrate, isovalerate, propionate, and isobutyrate concentrations. The abundance of *Cyanobacteria* was positively correlated (*p* < 0.05) with butyrate, propionate, and isobutyrate concentrations, and that of *Firmicutes* was positively correlated (*p* < 0.05) with butyrate concentration. In addition, the abundance of *Spirochaetae* was negatively correlated (*p* < 0.05) with propionate concentration.

#### DISCUSSION

To the best of our knowledge, the effects of supplementation with Leu in combination with Arg or Glu on fat accumulation have not been previously reported. Therefore, we investigated the effects of Leu in combination with Arg or Glu on body fat weight, and determined whether body fat weight was associated with colonic microbial composition and SCFAs. We found that dietary supplementation with Leu decreased body fat weight, consistent with Vianna et al. (2012), who reported that long-term Leu supplementation reduces fat mass gain in rats (Vianna et al., 2012). Further, Leu in

metabolites concentrations, whereas asterisk in the crimson grid represents a positive correlation. \*0.01 < *p* ≤ 0.05, \*\*0.001 < *p* ≤ 0.01,\*\*\**p* ≤ 0.001.

combination with Arg, but not with Glu, reduced body fat weight, revealing synergistic effects between Leu and Arg. Arginine can reduce fat accumulation in animals (Tan et al., 2009). Therefore, it is not surprising that body fat weight decreased following dietary supplementation with both Leu and Arg. We detected a possible antagonism between Leu and Glu. Our previous study showed that dietary supplementation with 1% Glu reduced back fat thickness in finishing pigs (Hu et al., 2017a,b), indicating that fat accumulation was attenuated with Glu treatment. Here, in contrast, we found that Glu supplementation reversed the effects of Leu on fat accumulation, as well as plasma TG and TC concentrations. Fat accumulation is related to the process of lipogenesis and lipolysis. Therefore, we analyzed genes related to lipogenesis and lipolysis. Our results suggest that the reduced body fat weight in the Leu and Leu\_Arg groups might arise from the upregulated expression of *HSL* and *CPT1* in the adipose tissue. We found that genes involved in lipogenesis (*PPAR*γ, *ACC*, and *FAS*) were not affected by diet supplementation, whereas genes associated with lipolysis, including *HSL* and *CPT-1*, were elevated in the Leu and Leu\_Arg groups. Hormone-sensitive lipase is responsible for catalyzing the hydrolysis of triacylglycerols in adipose tissue (Enevoldsen et al., 2001), and CPT-1 is mainly responsible for transferring cytosolic long-chain fatty acyl CoA into the mitochondria for oxidation. Dietary supplementation with Leu or Arg has been shown to increase fatty acid oxidation in fat (Jobgen et al., 2006; Chen et al., 2012; Zemel and Bruckbauer, 2012), suggesting that the reduced body fat weight in the Leu and Leu\_Arg groups was associated with fatty oxidation.

Obesity is associated with changes in composition, diversity and function of the gut microbiota. Therefore, gut microbiota composition in colon were determined. Our finding found that *Firmicutes* were most abundant in the Leu group, which is consistent with prior findings of elevated *Firmicutes* abundance and reduced *Bacteroidetes* abundance in obese mice (Ley et al., 2005) and in obese humans (Chakraborti, 2015). In contrast, Duncan et al. (2008) did not observe any difference in the abundance of *Bacteroidetes* and *Firmicutes* in the feces of lean and obese humans. It is clear that further research is needed to clarify the relationships between *Firmicutes*, *Bacteroidetes,* and obesity. Arg supplementation in mice is known to increase the abundance of *Bacteroidetes* and reduce that of *Firmicutes* in the jejunum and ileum (Ren et al., 2014), and Glu supplementation in pigs increases *Bacteroidetes* and *Peptostreptococcus* abundance in ileum (Feng et al., 2015). The abundance of *Actinobacteria* was highest in Leu group, in line with Pedersen et al. (2013) who reported that a higher abundance of *Actinobacteria* was observed in the cecal microbiota of lean Göttingen minipigs. However, elevated abundance of *Actinobacteria* has been demonstrated in obese humans (Turnbaugh et al., 2009). This discrepancy in the results of the present study and previous studies might be explained by differences between the species and diets used. We do not know of any possible reason to explain our finding that the abundance of *Deinococcus-Thermus* was negatively correlated with body fat weight, and was positively correlated with butyrate, isovalerate, propionate, and isobutyrate concentrations. However, the role of *Deinococcus-Thermus* in the alterations of these phenotypes remains unknown.

The SCFAs produced by the colonic microbiota provide 60–70% of the energy needs of colonic cells (Topping and Clifton, 2011). Of the SCFAs, butyrate is the major source of energy for the colonic epithelium. Colonic concentrations of SCFAs are associated with obesity: total SCFA concentration was significantly higher in obese humans than lean humans (Rahat-Rozenbloom et al., 2014), and elevated colonic propionate prevents weight gain in overweight adult humans (Chambers et al., 2015). We observed that dietary supplementation with Leu alone or in combination with Arg increased colonic propionate and butyrate concentrations, and negative correlations were observed between body fat weight and the concentrations of both propionate and butyrate. Consistent with these findings, studies in rodents have found that butyrate and propionate prevent diet-induced obesity and insulin resistance (Lin et al., 2012; Hong et al., 2016; Weitkunat et al., 2016; Wang et al., 2018). Further, elevated butyrate and propionate levels are known to reduce body fat mass mainly causing reduced intake of food or energy (Lin et al., 2012; Chambers et al., 2015). It has also been suggested that inhibition of adipose tissue accumulation is associated with elevated propionate produced by the gut microbiota; this may be because propionate increases energy expenditure (Kimura et al., 2011). In contrast, reduced body fat mass has been shown to be associated with increased cecal propionate, although no difference was observed in energy intake (Liou et al., 2013). We found that dietary supplementation with Leu alone or in combination with Arg reduced body fat weight and increased butyrate and propionate concentrations in the colon; however, in our previous study, feed intake was not affected by these amino acids (Hu et al., 2019). The discrepancy between the present study and our previous study might be explained by the fact that pigs consume more feed than mice.

# CONCLUSIONS

We found that dietary supplementation with Leu alone or in combination with Arg reduced body fat weight and increased the expression of genes involved in lipolysis in adipose tissue; it raised colonic butyrate and propionate concentrations, which were associated with reduced body fat weight. These findings provide new insight into the role of Leu in combination with Arg in preventing obesity. However, the mechanisms whereby Leu and Arg contribute to butyrate and propionate formation in the colon remain unclear. It is also unknown whether decreased body fat weight attributes to elevated colonic butyrate and propionate concentrations.

# AUTHOR CONTRIBUTIONS

CH and XK contributed to the study design, conducted the animal experiments, and wrote the manuscript. CH executed the lab analysis. YD and FL performed the statistical analysis. XK and YL revised the paper. All authors carefully read and approved the final revision of the manuscript.

# FUNDING

The present work was jointly supported by the National Key Research and Development Project (2017YFD0500503), National Natural Science Foundation of China (nos. 31572421 and 31772613), the Key Research Program of the Chinese Academy of Sciences (KFZD-SW-219-2-3), and Earmarked Fund for China Agriculture Research System (CARS-35).

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2019.01767/ full#supplementary-material

# REFERENCES


metabolite composition between days 45 and 70 of pregnancy in Huanjiang mini-pigs. *J. Anim. Sci. Biotechnol.* 9:18. doi: 10.1186/s40104-018-0233-5


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2019 Hu, Li, Duan, Yin and Kong. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Adhesive Bifidobacterium Induced Changes in Cecal Microbiome Alleviated Constipation in Mice

Linlin Wang1,2,3, Cailing Chen1,2, Shumao Cui1,2,3,4, Yuan-kun Lee<sup>5</sup> , Gang Wang1,2,3,4 \*, Jianxin Zhao1,2,4, Hao Zhang1,2,4,6,7 and Wei Chen1,2,6,8

<sup>1</sup> State Key Laboratory of Food Science and Technology, Jiangnan University, Wuxi, China, <sup>2</sup> School of Food Science and Technology, Jiangnan University, Wuxi, China, <sup>3</sup> International Joint Research Laboratory for Probiotics, Jiangnan University, Wuxi, China, <sup>4</sup> Institute of Food Biotechnology, Jiangnan University, Yangzhou, China, <sup>5</sup> Department of Microbiology and Immunology, National University of Singapore, Singapore, Singapore, <sup>6</sup> National Engineering Research Center for Functional Food, Jiangnan University, Wuxi, China, <sup>7</sup> Wuxi Translational Medicine Research Center and Jiangsu Translational Medicine Research Institute Wuxi Branch, Wuxi, China, <sup>8</sup> Beijing Advanced Innovation Center for Food Nutrition and Human Health, Beijing Technology and Business University, Beijing, China

#### Edited by:

Jie Yin, Institute of Subtropical Agriculture (CAS), China

#### Reviewed by:

Atte Von Wright, University of Eastern Finland, Finland Ana Griselda Binetti, Instituto de Lactología Industrial (INLAIN, CONICET), Argentina Jinping Chen, Guangdong Academy of Sciences (CAS), China

> \*Correspondence: Gang Wang wanggang@jiangnan.edu.cn

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 01 April 2019 Accepted: 12 July 2019 Published: 13 August 2019

#### Citation:

Wang L, Chen C, Cui S, Lee Y-k, Wang G, Zhao J, Zhang H and Chen W (2019) Adhesive Bifidobacterium Induced Changes in Cecal Microbiome Alleviated Constipation in Mice. Front. Microbiol. 10:1721. doi: 10.3389/fmicb.2019.01721 Constipation, which seriously affects living quality of people, is a common gastrointestinal disease. The engagement of the intestinal flora in the development of symptoms of constipation has been frequently hypothesized. In this study, constipated mice induced by loperamide were used to investige the alleviation of constipation by Bifidobacteria. Bifidobacteria was sorted out according to their adhesive properties into two groups. One group combined multiple strains of Bifidobacterium with adhesion property (CMB1), the other combined multiple strains of Bifidobacterium without adhesion property (CMB2). It was found that CMB1 can alleviate constipation more efficiently by improving the water, propionate and butyrate content in feces, and overall gastrointestinal transit time. Meanwhile, from the perspective of fecal microbiota, CMB1 alleviated constipation mainly by increasing the relative abundances of genera (Bifidobacterium, Lactobacillus, and Prevotella) associated with rapid bowel movement. From the perspective of cecal microbiota, CMB1 alleviated constipation mainly by increasing the relative abundances of genera Lactobacillus, Bacteroides, unclassified S24-7, Dorea, Ruminococcus, Coprococcus, and Rikenella, and decreasing the relative abundances of genera Oscillospira, Odoribacter and Unclassified F16, which are associated with methane production and colonic transit. Overall, changes of microbiota in caecum by CMB1 reflect the stage of constipation in mice more comprehensively than that in feces.

Keywords: Bifidobacterium, constipation, adhesion properties, SCFAs, gut microbiota

# INTRODUCTION

Usually, constipation is defined as infrequent or hard to pass bowel movements. It has a high incidence (2–36%) worldwide and complicated etiology (Rao et al., 2016). Its clinical symptoms include hard and dry feces, bowel movements that occur less than three times per week, rare occurrence of loose stools in the absence of laxatives and inadequate criteria for irritable bowel

syndrome (Bharucha et al., 2013). In the past few years, dietary changes and the effects of physiological, psychological, sociological and other factors have led to an increase in the prevalence of constipation, which can severely affect people's health and quality of life. Although the harmfulness of constipation is limited, it can be linked to increased risk of many related diseases such as Parkinson disease (Rossi et al., 2015) and colorectal cancer (Guérin et al., 2014). Therefore, it is very necessary to prevent and treat constipation. At present, the main drugs used to treat constipation are osmotic and secretory laxatives (Zhang et al., 2015). However, these therapies are susceptible to drug resistance to varying degrees and some lack efficacy. Many studies have shown that gut flora played a vital role in constipation. Constipation is related to the changes of gut microbiota, which may disturb cross-talk between the intestinal microflora, the intestinal endocrine system, the immune system and intestinal permeability (Caesar et al., 2015; Ge et al., 2017). Recent studies have revealed that constipation is related to imbalance in the intestinal microbiota, which mainly involve reduced levels of Bifidobacterium or Lactobacillus and an increase in pathogens (Chen et al., 2016). Supplementation with Bifidobacteria (Peter et al., 2006; Urita et al., 2015) or lactobacilli (Drouault et al., 2008; Williams et al., 2009; Yoon et al., 2015), either alone or combined, could prevent or treat constipation (Bu et al., 2007; Tabbers et al., 2011).

At present, many studies about the relationship between intestinal flora and diseases were based on fecal samples, which are easy to obtain but do not fully reflect the intestinal microbiota. In mice, microbiota in the caecum ferment carbohydrates that are unavailable in the small intestine. And it has been reported that there are significant differences in the structure of fecal, caecal and mucous membranes microbiota (Mao et al., 2016). Therefore, it is necessary to study the relationship between caecum content microbiota and constipation. In addition, our previous studies indicated that alleviation of constipation by Bifidobacterium might be related to the adhesion property of the strains (Wang et al., 2017a). It thus suggested that adhesion property may facilitate colonization of Bifidobacterium in the intestine. To date, there has been no reported research on the effect of adhesion property of Bifidobacterium on constipation. Therefore, it is of great significance to study the difference between fecal microbiota and caecum microbiota in mice after treated with Bifidobacteria with or without adhesion property.

In this study, constipated mice induced by loperamide were used to investigate the alleviation of constipation by Bifidobacteria. The Bifidobacteria were sorted by their adhesive property. It was found that strains with adhesion property can alleviate constipation more efficiently.

#### MATERIALS AND METHODS

#### Reagents

Kits used to measure the levels of gastrointestinal (GI) neurotransmitters, motilin (MTL), gastrin (Gas), substance P (SP), endothelin (ET-1), somatostatin (SS) and vasoactive intestinal peptide (VIP) were purchased from Wen LE Bioengineering Institute (Shanghai, China).

Loperamide was dissolved in sterile water and its ultimate density was 1 mg/mL. For acticarbon solution, gum arabic 100 g and water 800 mL were boiled until the solution was transparent. Acticarbon 50 g was then added and boiled three times. After cooling, the solution was diluted with water to 1000 mL and stored at 4◦C. Shake well before use.

#### Bacteria Preparation

Ten Bifidobacterium strains from five species obtained from American Type Culture Collection (ATCC) or China General Microbiological Culture Collection (CGMCC) were stored at the Culture Collection of Food Microorganisms in Jiangnan University (CCFM, Wuxi, Jiangsu Province, China). All the strains were cultured under anaerobic conditions for 24–48 h at 37◦C in modified MRS (cMRS) broth supplemented with 0.05% w/v L-cysteine-HCl (Merck). To prepare active cultures for all experiments, all strains were consecutively reactivated in an anaerobic atmosphere at least three times using 3% (v/v) inoculum in cMRS broth at 37◦C for 24–48 h before use. To use these strains in the animal experiments, the bacterial culture was centrifuged at 5000 × g for 10 min, washed twice with PBS, and centrifuged again to obtain the bacteria. The bacteria were divided into two groups (CMB1 and CMB2) of five based on their adhesion properties (**Table 1**). Adhesion properties of all these bacteria had been checked by cell adhesion assay in vitro and their adhesion characteristics can be found in our previous studies (Wang et al., 2017a). CMB1 refers to the multi-Bifidobacteria combination with the final concentration of 10<sup>10</sup> CFU/mL by mixing all Bifidobacteria with adhesion properties at the same concentration. The preparation method of CMB2 was the same as that of CMB1.

#### Experimental Design

Eight-week-old male BALB/c mice were obtained from Shanghai Laboratory Animal Center (Shanghai, China). The mice were kept in polyvinyl chloride (PVC) cages under environmentally controlled conditions with a 12-h light-dark cycle and standard commercial mouse feed and water were provided ad libitum. This study was approved by the Ethics Committee of the Jiangnan University, China (JN. No. 20150326-0110-21) and performed at the Experimental Animal Center of the Jiangnan University [License No. SYXK(SU)2016-0045].

To examine the preventive effects of CMB on constipation, 32 mice were used after a week-long adaptive period. The mice were randomly separated into four groups (n = 8): normal (healthy mice), control (constipation mice without treatment), CMB1 (constipation mice treated with CMB with adhesion properties) and CMB2 (constipation mice treated with CMB with no adhesion properties).

The mice were fasted overnight before the first experiment. The normal and control groups were given 0.25 mL normal saline (NS) using intragastric administration once a day for 17 days. The CMB1 and CMB2 groups were intragastrically administered 0.25 mL of normal saline solution containing 4 × 10<sup>10</sup> CFU/mL CMB1 or CMB2, respectively, daily for


#### TABLE 1 | Bifidobacteria used in this study.

fmicb-10-01721 August 9, 2019 Time: 16:39 # 3

CMB1, combining multiple strains of Bifidobacterium with adhesion property; CMB2, combining multiple strains of Bifidobacterium with no adherence property. CGMCC, China General Microbiological Culture Collection; ATCC, American Type Culture Collection.

2 weeks. All of the groups, except the normal group, were given loperamide (0.25 mL) intragastrically from day 15 to day 17 to induce constipation (Qian et al., 2013). Changes in food intake, water intake, and body weight were measured once per day at 9:00 AM throughout the experimental period. On the last day of the experiment, the mice were anesthetized by dipping ether with cotton in a relatively sealed space. The mouse beard was cut with surgical scissors to prevent hemolysis. Fixed the mouse, used the tweezers to clamp the eyeball, let the blood flow vertically into the centrifuge tube, put the centrifuge tube at room temperature for 2 h and then put it into the 4◦C refrigerator for 3 h and centrifuged at 3000 × g for 15 min to obtain serum. The abdomen was opened, the cecum of the mouse was cut with a surgical scissors and the contents of the cecum were aspirated into the EP tube using a syringe without a needle, then gently put the colon in sterile saline to remove the remaining contents, and finally placed the colon into the EP tube and stored at −80◦C until analysis. All operations were performed under sterile conditions. The experimental design is shown in **Supplementary Figure S1**.

#### Detection of Constipation-Related Indices

The relative indices of constipation (i.e., the water content of the feces, the small intestinal transit rate and the first black stool defecation time), stool collection and SCFAs in feces were measured as previously described (Wang et al., 2017a). Feces were collected every three days and stored at −80◦C. The water content of the feces was measured by the difference between the wet and dry weights of the feces, the small intestinal transit rate was evaluated by the distance traveled by an acticarbon solution relative to the overall length of the small intestine and the first black stool defecation time was measured by the time between the gavage of acticarbon solution and the appearance of darkened feces.

#### Detection of the GI Neurotransmitters Levels in Serum

The GI neurotransmitters levels in the serum were determined by an enzyme-linked immunosorbent assay (ELISA) instrument according to the manufacturer's instructions (Microplate Spectrophotometer Multiskan Go, Thermo Scientific, Waltham, MA, United States). The experiments mainly include the preparation of standard curve and the detection of sample absorption value (OD450) according to the standard curve in units of ng/L.

#### 16S rDNA Sequencing and Bioinformatics Analysis

The enteric microorganisms in the fecal and caecal samples were measured using a metagenomics method (Wang et al., 2017a). Microbial genomic DNA was obtained using a FastDNA Spin Kit for Soil (MP Biomedical). The V4 region of the 16S rDNA was amplified by PCR. The products were purified and quantified using Gene Clean Turbo (MP Biomedical) and the Quant-iT PicoGreen dsDNA Assay Kit (Life Technologies), respectively. Libraries were prepared using TruSeq DNA LT Sample Preparation Kits (Illumina) and sequenced by Illumina MiSeq using the MiSeq Reagent Kit.

The QIIME pipeline was used to analyze the 16S rDNA sequence data (Caporaso et al., 2010). The raw sequences were screened. The short lengths (<200 bp) were then removed, and the pair-end reads that overlapped longer than 10 bp and without any mismatch were assembled according to their overlap sequence. The sequences were then clustered into operational taxonomic units (OTUs) based on 97% identity using QIIME<sup>1</sup> . The representative sequences for each OTU were aligned to identify the species using PyNAST in QIIME. Rarefaction curves for alpha diversity were generated to assess the efficiency of the sequencing depth for representing and comparing microbial communities. Species richness was estimated using Chao-1 (Faith and Baker, 2006). The beta diversity of the microbial communities was determined by visual assessment using principle coordinate analysis (PCoA) plots and by an analysis of similarity calculated based on weighted UniFrac distances (QIIME) according to one-way non-parametric multivariate analysis of variance.

<sup>1</sup>http://qiime.sourceforge.net/

#### Statistical Analysis

The data were presented as mean ± SD and analyzed using GraphPad Prism 5 and Origin 8.5. The differences between the samples were analyzed by one-way ANOVA with Duncan's multiple range test. The results were considered significant when p < 0.05.

#### Compared with the constipation group, the symptoms of constipation (i.e., the water content of the feces, the small intestinal transit rate and the first black stool defecation time) were significantly improved in CMB1 group (**Figure 1**, p < 0.05). Meanwhile, there was no statistical difference between the CMB2 and control groups. These results demonstrate that CMB2 had

#### RESULTS

#### Bifidobacterium Combination With Adhesion Properties Significantly Improved the Symptoms of Constipation

Before constipation was induced, the water content of the feces was higher in CMB1, and CMB2 groups than in the control and normal groups (**Figure 1C**, p < 0.05). With the intake of loperamide, the fecal water content showed a downward trend in all groups, which indicated that loperamide induced constipation in mice.


CMB1 group: (combining multiple strains of Bifidobacterium with adhesion property), 1 × 10<sup>10</sup> CFU; CMB 2 group: (combining multiple strains of Bifidobacterium with no adhesion property), 1 × 10<sup>10</sup> CFU. <sup>a</sup>−<sup>c</sup> Mean values with different letters over bars are significantly different (p < 0.05) according to Duncan's multiple range test.

no effect on constipation induced by loperamide. There was no significant difference in weight gain, feed and water consumption between the mice throughout the experiment. This indicated that constipation had no effect on the weight and appetite of mice.

#### Bifidobacterium Combination With Adhesion Properties Significantly Improved the GI Neurotransmitters Levels in Serum

Constipation-related neurotransmitters, such as MTL, Gas, SP, ET-1, SS and VIP, play an important role in regulating gastrointestinal motility. In addition, the secretion of GI neurotransmitters is different in disease and normal state. Therefore, we studied these GI neurotransmitters in serum of mice after treated with CMB. The results showed that CMB1 obviously increased the levels of MTL, Gas and SP and decreased the levels of SS, VIP and ET-1. The variation tendency of GI neurotransmitters in CMB2 group were the same as that in CMB1 group, except for Gas and SP (**Figure 2**, p < 0.05). In CMB1 treated group, the Bifidobacteria combination showed preferable effect in relieving constipation. These results indicated that the difference in relieving constipation between CMB1 and CMB2 may be related to the levels of Gas and SP in serum.

#### Bifidobacterium Combination With Adhesion Properties Significantly Raised the Concentration of SCFAs in Fecal Samples

The contents of each SCFA in the feces are shown in **Tables 2**, **3**. Before constipation, acetate, propionate and butyrate levels were fairly steady in normal and constipation control groups, while the concentrations of acetate and total acids increased, and the level of butyrate decreased in CMB1 and CMB2 groups. After the induction of constipation, the concentrations of acetate showed the same level in normal and control groups. Only propionic, butyric and total acids decreased compared to normal group. In addition, acetic, propionic and butyric acid concentrations increased in CMB1 group, whereas there was no significant difference between control and CMB2 groups in the concentrations of propionic and butyric acids. These results indicated that increasing the concentration of propionic and



CMB1 group: (combining multiple strains of Bifidobacterium with adhesion property), 1 × 10<sup>10</sup> CFU; CMB2 group: (combining multiple strains of Bifidobacterium with no adhesion property), 1 × 10<sup>10</sup> CFU; <sup>a</sup>−<sup>c</sup> Mean values with different letters over bars are significantly different (p < 0.05) according to Duncan's multiple range test.

butyric acids in feces might be associated with the relief of constipation symptoms.

#### Adhesive Bifidobacterium Combination Affected Fecal Microbiota in Different Way to Non-adherent Bifidobacterium Combination

To evaluate the influences of the CMB on the intestinal flora, we analyzed the fecal samples of healthy and constipated BALB/c mice gavaged with CMB1 and CMB2 for 17 days. A dataset containing 325,440 high-quality sortable 16S rDNA gene orders was produced from 32 fecal specimens using MiSeq sequencing. The average sequence read was 10,170 per sample. Representative sequences of all of the sequences were clustered, and a 97% sequence similarity cut-off was used. The number of OTUs per sample ranged between 433 and 1952. We performed association tests based on α- and β- diversity measures. There was a remarkable difference in α-diversity between the normal and constipation control groups (**Supplementary Figure S2**). After induced constipation by loperamide, the α-diversity indices (Chao-1, observed species and phylogenetic diversity) shown a sharp declined in constipation control and CMB2 groups. In contrast, CMB1 obviously increased the α-diversity indices. It indicated that CMB1 significantly increased the taxa richness of fecal microbiota, while CMB2 had no effect on the decrease of fecal microbial diversity caused by constipation.

The β-diversity of the fecal flora in mice given CMB1 and CMB2 was revealed by using unweighted UniFrac matrixes. As shown in **Supplementary Figure S3**, the light blue symbol represents the normal group and was located at the lower section of the PCA diagram, while the red symbol represents the constipation control group and was shift from the lower section of the score plot to the left, indicating that constipation changed the fecal microbiota structure of mice. The green and dark blue symbols, respectively, represent the CMB1 and CMB2 groups and were, respectively, located at the upper middle and upper right section of the PCA diagram, indicating that CMB significantly changed the fecal microbiota structure of mice. However, CMB1 and CMB2 affected the fecal microbiota through absolutely different ways.

A total of 30,627 sequences were appointed to 1,952 OTUs, which agglomerated into 41 genera and 8 phyla using the Ribosomal Database Project (RDP) classifier. Changes in the taxa that were obviously different between treatments are shown in **Figures 3–5**.

#### Actinobacteria

Two genera (Bifidobacterium and Adlercreutzia) were included in this phylum. At the genus level, the level of Bifidobacterium in CMB groups were significantly higher than in constipation group. There was no significant difference among groups in the level of Adlercreutzia (**Figrue 3**, Bifidobacterium).

#### Bacteroidetes

Seven genera Bacteroides, Odoribacter, Parabacteroides, Barnesiella, Rikenella, Prevotella and Alistipes were included

in this phylum. Odoribacter was the lowest in CMB1 group, followed by normal and control groups and then CMB2 group. The relative abundance of Parabacteroides

(C) Ruminococcus), (D) Sporobacter, and (E) Coprobacillus.

(C) Parabacteroides, (D) Lactobacillus, and (E) Clostridium.

decreased to normal level after treated with CMB. The relative abundance of Prevotella recovered to normal level in CMB1 treatment group, whereas there was no significant

fmicb-10-01721 August 9, 2019 Time: 16:39 # 6

FIGURE 4 | Relative abundance (%) of fecal microbial taxa at the genus level in constipated mice fed CMB. (A) Unclassified Lachnospiraceae, (B) Odoribacter,

FIGURE 3 | Relative abundance (%) of fecal microbial taxa at the genus level in constipated mice fed CMB. (A) Bifidobacterium, (B) Odoribacter,

difference between CMB2 and constipation control group. Alistipes was the lowest in CMB1 treatment group and the highest in constipation control group. There was no significant difference among groups in the level of other genera (**Figrues 3**, **5**, Odoribacter, Parabacteroides, Prevotella and Alistipes).

#### Deferribacteres

This phylum included a single genus, Mucispirillum. There was no significant difference among groups in the level of Mucispirillum.

#### Firmicutes

Twenty-seven genera (such as Staphylococcus, Lactobacillus, Lactococcus, Streptococcus, Clostridium, Eubacterium, Blautia, Dorea, Oscillibacter, Ruminococcus, Sporobacter, Coprobacillus, Turicibacter and so on) were included in this phylum. At the genus level, the relative abundance of unclassified Lachnospiraceae, Clostridium, Sporobacter, Coprobacillus, Turicibacter, Oscillibacter and Ruminococcus decreased significantly in CMB groups compared to control group. Meanwhile, the relative abundance of Clostridium and Sporobacter were higher in CMB2 treatment group than that in CMB1 group. In contrast, the relative abundance of Lactobacillus was the highest in CMB1 treatment group, followed by normal and CMB2 groups and then the constipation control group (**Figures 3–5**). There was no significant difference among groups in the level of other genera.

#### Proteobacteria

Only one genus was found in this phylum. The relative abundance of unclassified Helicobacteraceae only increased in CMB1 treatment group, whereas there was no significant difference among other groups (**Figure 5**).

#### TM7

This phylum contained one genus, the unclassified TM7. There was no significant difference among groups in this genus.

#### Tenericutes

Anaeroplasma was the only genus found in this phylum. Its level was highest in CMB2 group followed by the control and normal groups and then the CMB1 group (**Figure 5**, Anaeroplasma).

#### Verrucomicrobia

Akkermansia was the only genus found in this phylum. Its level was only increased in CMB1 treatment group, whereas there was no significant difference among other groups.

In brief, in CMB1-treated group, the relative abundances of Bifidobacterium, Lactobacillus, Akkermansia, Prevotella and unclassified Helicobacteraceae significantly increased and the levels of Clostridium and Anaeroplasma in feces samples decreased compared to CMB2-treated group. Moreover, the levels of Bifidobacterium, Lactobacillus, Akkermansia and Prevotella were negatively correlated with constipation, while Clostridium and Anaeroplasma were positively correlated with constipation. Therefore, after CMB1

intervention, the gut microbiota in constipated mice tended to be beneficial to host health.

#### Adhesive Bifidobacterium Combination Affected Caecal Microbiota in Different Way to Non-adherent Bifidobacterium Combination

A dataset containing 373,216 high-quality sortable 16S rDNA gene orders was produced from 32 fecal specimens using MiSeq sequencing. The average sequence read was 11,663 per sample. Representative sequences of all of the sequences were clustered, and a 97% sequence similarity cut-off was used. The number of OTUs per sample ranged between 412 and 1330. We performed association tests based on α- and β- diversity measures. The normal and constipation control groups had obvious differences in αdiversity based on species richness, the observed species and diversity (**Supplementary Figure S4**). After CMB treatment, the chao-1 index and observed species showed significant changes in CMB1 group. The phylogenetic diversity index was obviously changed in CMB1 and CMB2 groups compared with the constipation group. The β diversity of the caecal microbiota was assessed using unweighted UniFrac matrixes. As shown in **Supplementary Figure S5**, constipation changed the caecal microbiota structure of mice greatly, while CMB significantly changed the caecal microbiota structure of mice. Different from the absolutely different ways by CMB1 and CMB2, respectively, these two Bifidobacterium combinations showed a certain overlap in the caecal microbiota regulation.

A total of 29,193 sequences were assigned to 1,396 OTUs that were clustered into 54 genera and 9 phyla using the RDP classifier. Changes in taxa are shown in **Figures 6–8**.

#### Actinobacteria

Two genera were found in this phylum, Adlercreutzia only increased in control group, whereas there was no significant difference among other groups. The relative abundance of Bifidobacterium was the highest in CMB2 treatment group, followed by CMB1 treatment group and then normal and control groups (**Figure 6**, Bifidobacterium).

#### Bacteroidetes

Twelve genera were found in this phylum. At the genus level, the level of Bacteroides was the highest in CMB1 treatment group, followed by CMB2 and then the normal and control groups. The levels of Parabacteroides only increased significantly in control group, whereas there was no significant difference among other groups. The level of Rikenella was the highest in CMB1 group and the lowest in CMB2 treatment group, whereas there was no significant difference between normal and control group. The relative abundance of unclassified S24-7 was the highest in CMB1 treatment group, followed by CMB2 and control groups and then the normal group. The relative abundance of Odoribacter recovered to normal level after treated with CMB2, whereas there was no significant difference between CMB1 and control group. There was no significant difference among groups in the level of other genera (**Figure 6**).

#### Cyanobacteria

Unclassified YS2 was the only genus found in this phylum, and there was no significant difference among groups in this genus.

#### Deferribacteres

Mucispirillum was the only genus found in this phylum, and the relative abundance of Mucispirillum recovered to normal level after treated with CMB2, whereas there was no significant difference between CMB1 and control group (**Figure 7**).

#### Firmicutes

Twenty-six genera (such as Lactobacillus, unclassified Clostridiales, Clostridium, Dehalobacterium, unclassified Lachnospiraceae, Coprococcus, Dorea, unclassified Ruminococcaceae, Oscillospira, Ruminococcus and so on) were found in this phylum. The abundances of Dehalobacterium, Dorea, Coprococcus, and Ruminococcus were significantly higher in CMB1group compared to CMB2 group, whereas the abundance of Oscillospira was obviously lower in CMB1group compared to CMB2 group. The relative abundance of Lactobacillus, and Clostridium increased significantly in CMB groups compared to control group. The relative abundance of unclassified Lachnospiraceae decreased significantly in all treatment groups, except the normal group. And there was no significant difference among groups in the level of other genera (**Figures 7**, **8**).

#### Proteobacteria

Seven genera were found in this phylum. The level of Desulfovibrio was the highest in CMB1 group and the lowest in CMB2 treatment group, whereas there was no significant difference between other groups. Meanwhile, the level of unclassified Helicobacteraceae was the lowest in CMB2 treatment group, and the highest in normal group, whereas there was no significant difference between control and CMB2 treatment group. There was no significant difference among groups in the level of other genera (**Figures 7**, **8**).

and (E) Dorea.

#### TM7

This phylum included one genus, unclassified F16. Its level was decreased to normal group in CMB1 treatment group, whereas there was no significant difference between CMB2 and control group (**Figure 8**).

#### Tenericutes

This phylum included one genus. The relative abundance of Anaeroplasma was the highest in control group, followed by CMB treatment groups and then the normal group (**Figure 8**).

#### Verrucomicrobia

Akkermansia was the only genus found in this phylum. Its level was only increased in CMB treatment group, whereas there was no significant difference between normal and control group (**Figure 8**).

In brief, after CMB intervention, the common effect of CMB1 and CMB2 on caecum microbiota was that they significantly elevated the relative abundances of Lactobacillus, Clostridium, Akkermansia and Bifidobacterium and reduced the relative abundances of Anaeroplasma. The difference was that in CMB1-treated group, the relative abundances of Bacteroides,

Rikenella, unclassified S24-7, Dorea, Coprococcus, Ruminococcus, Dehalobacterium and Desulfovibrio significantly increased and the relative abundances of Oscillospira, Odoribacter, Mucispirillum and unclassified F16 decreased, while all these changes were opposite in CMB2-treated group. Considering the result that CMB1 alleviated constipation effectively, it seemed that the differences on cecum microbiota between CMB1- and CMB2- treated groups might be closely related to the situation of constipation in mice.

Comparing the fecal flora with the cecum content flora at the genus level, we found that there were 15 genera with significant differences between these two samples. As shown in **Supplementary Figure S6**. the fecal microbiota was located at the right of the score plot, whereas the caecal microbiota was located at the left of the score plot. It indicated that there were significant differences between fecal microbiota and caecum microbiota. Differences in taxa are shown in **Figures 9–11**.

At the genus level, there were 15 genera with significant differences between fecal microbiota and caecum microbiota, and 11 of these 15 genera showed that the relative abundance of caecum samples was significantly higher than that of fecal samples. In contrary, the relative abundances of Lactobacillus, Bifidobacterium, and Odoribacter were higher in fecal samples than in cecum samples (**Figures 9**, **10**). In addition, the relative abundances of Bacteroides and Clostridium in feces sample and caecum sample presented the opposite tendency in CMB1 group, whereas the level of Anaeroplasma showed the same tendency (**Figure 9**). The relative abundances of Dorea, Rikenella and Ruminococcus in cecum sample increased obviously in CMB1 treated group, whereas the level of Oscillospira in cecum sample decreased obviously in CMB1-treated group. And the level of Odoribacter in fecal sample decreased obviously in CMB1-treated group (**Figures 9–11**).

#### DISCUSSION

Although many efforts have been made to treat constipation, it remains one of the most common diseases worldwide. Therefore, it is urgent to explore an effective, safe and low-toxicity methods to treat constipation. Clinical studies have found that gut microbiota plays an important role in the occurrence and development of constipation. Thus, micro-ecological therapy has gradually replaced the traditional treatment for constipation.

A large number of animal and human experiments have confirmed that probiotics can alleviate constipation (Bu et al., 2007; Tabbers et al., 2011; Urita et al., 2015). However, the adhesion property of strains is the premise of this study in evaluating the probiotic effect of strains. Therefore, in this study, ten strains of Bifidobacteria were divided into two groups according to their adhesion property to investigate their efficacy in alleviating constipation. The results showed that CMB1 (combination of the adhesive Bifidobacteria) significantly improved the water content of feces, the small intestinal transit rate and the first black stool defecation time, whereas the CMB2 (combination of non-adhesive Bifidobacteria) had no effect on constipation. This result was consistent with our previous study (Wang et al., 2017a). The results showed that whether a single strain or multiple strains of Bifidobacterium are used, the effect of relieving constipation is related to the adhesion property of the bacteria.

The known factors affecting intestinal motility are nervous systems (intestinal nervous system and autonomic nervous system), immune system, intestinal microbiota and their metabolites. Any disorder or dysfunction in any of these factors can lead to intestinal motility disorders (constipation or diarrhea). Therefore, in this study, GI neurotransmitters levels were measured to determine the

effect of Bifidobacteria on constipation-related intestinal nervous activities. The results showed that the reduced GI neurotransmitters in serum of constipation mice tended to be up-regulated in CMB1 intervention group, while CMB2 intervention had no effect on recovery of neurotransmitters (MTL, Gas and SP). Thus, CMB1 may alleviate constipation by up-regulating neurotransmitters (MTL, Gas and SP). Considering the adhesion properties of Bifidobacterium, we believe that the bacteria with adhesion property may have a greater chance of establishing close contact with and colonizing the intestine to exert GI neurotransmitter regulatory effects.

It has been suggested that the changes of gut flora and metabolites may be the important reasons for the pathophysiological changes in constipation. Several previous studies have reported that an increased concentration of SCFAs in the intestine is beneficial to constipation (Salminen and Salminen, 1997; Veiga et al., 2014). SCFAs are produced by the bacterial fermentation of dietary fiber. They are reported to have important physiological effects in the intestine. For example, SCFAs influence the functions of the gastrointestinal tract (Ríos-Covián et al., 2016), electrolyte balance (Vidyasagar and Ramakrishna, 2000), and ion transport (Yajima, 1988). The concentration of SCFAs in feces reflects the activity of enteric microorganisms and is important in the relief of constipation by CMB. Our study showed that CMB1 improved the symptoms of constipation and increased the levels of propionic and butyric acids. This increase corresponded to an increase in propionic and butyric acid producing bacteria. Propionic acid producing bacteria are mainly Bacteroidetes, and butyric acid levels are positively correlated with the relative abundance of Firmicutes (Guilloteau et al., 2010). In our study the observed increases in propionic and butyric acid are consistent with an increase in Prevotella and Lactobacillus in feces samples treated with CMB1, and increase in unclassified S24- 7, Rikenella, Bacteroides, Dorea, Lactobacillus, Dehalobacterium, Desulfovibrio, Ruminococcus and Coprococcus in caecal content samples treated with CMB1. It has been widely reported that either Prevotella or Bacteroides dominates the human gut and they were proposed to be antagonistic (Choi and Chang, 2015). Prevotella and Bacteroides, which are thought to have had a common ancestor, benefited their host by excluding potential pathogens from colonizing the gut (Ley, 2016). It also can be seen from the results that the ratio of Firmicutes to Bacteroidetes increased significantly in feces samples of CMB1-treated group (from 0.89 in control group to 1.55 in CMB1group). A higher ratio of Firmicutes to Bacteroidetes is associated with faster transit in the large intestine (Sadik et al., 2004; DelgadoAros et al., 2008). Meanwhile, as observed in this study, the higher abundance of Actinobacteria (Bifidobacterium) in feces sample of CMB1-treated group was also correlated with faster colonic transit and in agreement with an earlier study (Parthasarathy et al., 2016). Therefore, it seems that the fecal microbiota profile may be related to colonic transit.

After treatment with CMB1, the genera unclassified S24-7, Rikenella, Bacteroides, Dorea, Lactobacillus, Dehalobacterium, Desulfovibrio, Ruminococcus and Coprococcus increased in caecal content sample. Unclassified S24-7 is a prevalent and abundant bacterial component of the gut microbiome of mammals (Ormerod et al., 2016) and Unclassified S24-7 is more abundant following treatment-induced remission of colitis in mice (Rooks et al., 2014). The genera Firmicutes-Dorea and Bacteroidetes-Bacteroides were correlated inversely with the production of methane, while the Firmicutes-Oscillospira and Bacteroidetesodoribacter were correlated directly with the production of methane (Parthasarathy et al., 2016). In our research, the relative abundances of Dorea and Bacteroides increased significantly and the genera of Odoribacter and Oscillospira decreased obviously in caecal microbiota, thus suggest that the caecal microbiota is a better indicator for constipation and methane production than fecal microbiota. Furthermore, the genera Firmicutes-Coprococcus and Firmicutes-Lactobacillus which were correlated directly with colonic transit (Parthasarathy et al., 2016), increased significantly in caecum microbiota. These findings suggest that the caecal microbiota is also a better indicator for colonic transit. Taken together, compared with fecal microbiota, cecum microbiota reflects constipation state more comprehensively and accurately.

As mentioned above, adhesive Bifidobacterium combination affected fecal and caecal microbiota with different way to nonadherent Bifidobacterium combination. Furthermore, changes in the caecal contents microbiota and fecal microbiota were different as well whether treated with adhesive or nonadherent Bifidobacterium. It was found that there were seven genera with significant differences between fecal microbiota and caecal content microbiota. Meanwhile, six of these seven genera showed a high abundance in the caecal content compared to the relative abundance in the feces, whether in the CMB1 treatment group or the CMB2 treatment group. Such as the relative abundances of Oscillibacter, Clostridium, Anaeroplasma, Ruminococcus and so on. This might be due to the different environments in different part of intestine which resulted in different abundance of intestinal microbes. The intake of Bifidobacterium affected the microbiota in these parts of intestine. Due to the differences of microbiota species abundance, the influences by Bifidobacterium on caecal contents microbiota and fecal microbiota were different as well. We believe that adhesive Bifidobacteria can not only stay in the intestinal tract for a long time, but also produce more metabolites, so as to play a more effective role in changing gut microbiota than that by nonadhesive Bifidobacteria.

In this study, the relative abundances of Bifidobacterium (Actinobacteria), Lactobacillus (Firmicutes) and Akkermansia (Verrucomicrobia), which have been shown to improve the motility of the intestine by provoking the release of 5-HT or by promoting cholinergic pathways (Reigstad et al., 2015), were associated with faster transit in the large intestine. It is also consistent with our previous study (Wang et al., 2017a) where feeding constipated mice with CMB regulated the dysbiosis of the gut microbiota (decreases in the levels

of Bifidobacterium and Lactobacillus) caused by constipation. Recent studies in rodents indicate that Akkermansia in the gut might reduce obesity, diabetes and inflammation (Everard et al., 2013). In this study, overgrowth of Bifidobacterium, Lactobacillus and Akkermansia was found to accompany reduction in the number of other bacteria, such as Anaeroplasma, Parabacteroides, Clostridium (in fecal samples) and unclassified Helicobacteraceae (only showed slight downward trend in the caecal contents sample). Some species of Anaeroplasma, Parabacteroides and Clostridium are pathogenic to humans and are associated with colitis and gastroenteritis (Ethan et al., 2011; Michael et al., 2016). Some species of Helicobacter (unclassified Helicobacteraceae) are associated with peptic ulcers, chronic gastritis, duodenitis and stomach cancer (Alexander et al., 2014; Watari et al., 2014; Seyed et al., 2015; Sostres et al., 2015). The genera Clostridium, Dorea, Oscillibacte, Ruminococcus, Sporobacter and Turicibacter are considered opportunistic pathogens (Sokol et al., 2008; Jenq et al., 2012; Konikoff and Gophna, 2016). When probiotics are predominated in the gut, opportunistic pathogens can have adjuvant effects, but when pathogenic bacteria are predominated, opportunistic pathogens can have pathogenic effects. In our present study, the overgrowth of probiotics changed the microenvironment of the gut and helped the opportunistic pathogens to develop in the health promoting direction. Thus, the changes in the above genera had a potential regulatory role in relieving constipation. It is certain that the intestinal microbiota showed dysbiosis after constipation. However, whether dysbiosis causes constipation or represents an epiphenomenon remains unclear. Further studies are needed to determine the true cause-andeffect relationship.

#### CONCLUSION

In conclusion, this study demonstrates that strains with adhesion properties (CMB1) can alleviate constipation more efficiently. CMB1 noticeably increased the water content of feces, small intestinal transit rates, the first black stool defecation time and the concentration of propionic and butyric acids. Moreover, constipated mice treated with CMB1 had a unique profile of caecal microbiota that reflect the relief of constipation more comprehensively and accurately, in comparison with fecal microbiota which was only associated with the colonic transit.

# REFERENCES


# DATA AVAILABILITY

The datasets (gut metagenome Genome sequencing and assembly. SRA accession: PRJNA531550) for this study can be found in the [Sequence Read Archive (SRA)] (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA531550).

### ETHICS STATEMENT

This study was approved by the Ethics Committee of the Jiangnan University, China (JN. No. 20150326-0110-21) and performed at the Experimental Animal Center of the Jiangnan University [License No. SYXK(SU)2016-0045].

# AUTHOR CONTRIBUTIONS

LW and GW conceived and designed the experiments. LW, CC, and SC performed the experiments. LW, GW, Y-kL, and HZ analyzed the data. JZ, HZ, and WC contributed reagents, materials, and analysis tools. LW, GW, and Y-kL wrote the manuscript. All authors contributed to the manuscript revision, and read and approved the submitted version.

# FUNDING

This work was supported by the National Natural Science Foundation of China (Nos. 81800469, 31671839, and 31820103010), the Natural Science Foundation of Jiangsu Province (BK20180613), the Project funded by China Postdoctoral Science Foundation (2018M642164), the Postdoctoral Science Foundation of Jiangsu Province (2018K090C), the National Natural Science Foundation of China Key Program (No. 31530056), the National First-Class Discipline Program of Food Science and Technology (JUFSTR20180102), and the Program of Collaborative Innovation Center of Food Safety and Quality Control in Jiangsu Province.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.01721/full#supplementary-material

Bu, L. N., Chang, M. H., Yen-Hsuan, N. I., Chen, H. L., and Cheng, C. C. (2007). Lactobacillus casei rhamnosus Lcr35 in children with chronic constipation. Pediatr. Int. 49, 485–490. doi: 10.1111/j.1442-200X.2007.02397-x



Williams, E. A., Stimpson, J., Wang, D., Plummer, S., Garaiova, I., Barker, M. E., et al. (2009). Clinical trial: a multistrain probiotic preparation significantly reduces symptoms of irritable bowel syndrome in a double-blind placebocontrolled study. Aliment. Pharmacol. Ther. 29, 97–103. doi: 10.1111/j.1365- 2036.2008.03848.x

Yajima, T. (1988). Luminal propionate-induced secretory response in the rat distal colon in vitro. J. Physiol. 403, 559–575. doi: 10.1113/jphysiol.1988.sp017264


across lifestyles, geography and ethnicities. ISME J. 9, 1979–1990. doi: 10.1038/ ismej.2015.11

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Wang, Chen, Cui, Lee, Wang, Zhao, Zhang and Chen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Alterations in Gut Microbiota by Statin Therapy and Possible Intermediate Effects on Hyperglycemia and Hyperlipidemia

Jiyeon Kim<sup>1</sup>† , Heetae Lee<sup>1</sup>† , Jinho An<sup>1</sup> , Youngcheon Song<sup>1</sup> , Chong-Kil Lee<sup>2</sup> , Kyungjae Kim<sup>1</sup> and Hyunseok Kong<sup>3</sup> \*

<sup>1</sup> College of Pharmacy, Sahmyook University, Seoul, South Korea, <sup>2</sup> College of Pharmacy, Chungbuk National University, Cheongju, South Korea, <sup>3</sup> College of Animal Biotechnology and Resource, Sahmyook University, Seoul, South Korea

#### Edited by:

Yuheng Luo, Sichuan Agricultural University, China

#### Reviewed by:

Monica Di Paola, University of Florence, Italy Natasa Golic, University of Belgrade, Serbia

#### \*Correspondence:

Hyunseok Kong hskong0813@gmail.com †These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Systems Microbiology, a section of the journal Frontiers in Microbiology

Received: 16 March 2019 Accepted: 08 August 2019 Published: 04 September 2019

#### Citation:

Kim J, Lee H, An J, Song Y, Lee C-K, Kim K and Kong H (2019) Alterations in Gut Microbiota by Statin Therapy and Possible Intermediate Effects on Hyperglycemia and Hyperlipidemia. Front. Microbiol. 10:1947. doi: 10.3389/fmicb.2019.01947 Dysbiosis of the gut microbiota is a contributing factor for obesity-related metabolic diseases such as hyperglycemia and hyperlipidemia. Pharmacotherapy for metabolic diseases involves the modulation of gut microbiota, which is suggested to be a potential therapeutic target. In this study, the modulation of gut microbiota by statins (cholesterollowering drugs: atorvastatin and rosuvastatin) was investigated in an aged mouse model of high-fat diet-induced obesity, and the association between gut microbiota and immune responses was described. Atorvastatin and rosuvastatin significantly increased the abundance of the genera Bacteroides, Butyricimonas, and Mucispirillum. Moreover, the abundance of these genera was correlated with the inflammatory response, including levels of IL-1β and TGFβ1 in the ileum. In addition, oral fecal microbiota transplantation with fecal material collected from rosuvastatin-treated mouse groups improved hyperglycemia. From these results, the effect of statins on metabolic improvements could be explained by altered gut microbiota. Our findings suggest that the modulation of gut microbiota by statins has an important role in the therapeutic actions of these drugs.

Keywords: atorvastatin, rosuvastatin, gut microbiota, IL-1β, TGFβ1, short-chain fatty acid

#### INTRODUCTION

Obesity-related metabolic diseases including hyperlipidemia, hypertension, and type 2 diabetes (T2D) are prevalent, global health burdens (Wang et al., 2008). Lifestyle, including a lack of physical exercise and a Westernized diet, is a main contributing factor for the prevalence of metabolic diseases (Lee S.E. et al., 2018). Metabolic diseases are associated with an increased mortality risk in the elderly population; in particular, hyperlipidemia is an important risk factor for cardiovascular disease (Isomaa et al., 2001; Felix-Redondo et al., 2013).

Recently, American College of Cardiology/American Heart Association (ACC/AHA) recommended statins as a first-line treatment of hyperlipidemia, and statins are currently the most widely used treatment for hyperlipidemia (Safeer and Lacivita, 2000; Stone et al., 2014). Statins work by inhibiting 3-hydroxy-3-methylglutaryl-coenzyme A (HMG-CoA) reductase, an important enzyme involved in the cholesterol synthesis pathway (Stancu and Sima, 2001). Moreover, statins induce the expression of low-density lipoprotein (LDL) receptors and promote

**478**

the removal of LDL cholesterol from the blood (Young and Fong, 2012). Furthermore, statins have been reported to have antiinflammatory and immunomodulatory effects (Lee et al., 2016).

Gut microbiota is an important environmental factor for the regulation of energy homeostasis and immune system (Wu and Wu, 2012; Sanz et al., 2015). Modulation of gut microbiota improves the outcome of various diseases, including metabolic syndrome and inflammatory diseases, and influences pharmacotherapy (Wilson and Nicholson, 2009; Karkman et al., 2017). Recently, various interventions such as dietary supplement (Lee and Ko, 2016; Talsness et al., 2017), prebiotics or probiotics (Delzenne et al., 2011; Markowiak and Slizewska, 2017), and drugs (Shin et al., 2013; Lee and Ko, 2014) were reported to alter the composition of gut microbiota, which has an important role in the alleviation of metabolic diseases. In recent studies, statin therapy was found to affect the composition of gut microbiota (Khan et al., 2018; Liu et al., 2018). Nolan et al. (2017) showed that rosuvastatin influenced gut microbiota and significantly increased the abundance of the family Lachnospiraceae, and the genera Rikenella and Coprococcus in HFD-fed C57BL/6 mice.

The objectives of this study were (1) to identify the effects of statins on gut microbiota and metabolic improvements and (2) to investigate the correlation between significant changes in microbiota induced by statins and inflammatory regulation.

#### MATERIALS AND METHODS

#### Animals and Experimental Protocol

Male 4-week-old C57BL/6N mice were purchased from Samtako Co., Ltd (Osan, South Korea) and acclimated to laboratory conditions for 1 week, in which the animals were housed with free access to water and food in a temperature–humidity-controlled animal facility under a 12-h light–dark cycle at 22 ± 2 ◦C and 55 ± 5% humidity. Mice were fed either a 45% kcal high-fat diet (HFD) (FeedLab, Inc., Guri, South Korea) to induce metabolic disorders such as obesity, hyperglycemia, and hyperlipidemia, or a regular diet (RD) (Purina Korea, Inc., Seoul, South Korea) for 39 weeks. Atorvastatin (HFD-Ator: 10 mg/kg of body weight, n = 6) or rosuvastatin (HFD-Rosu: 3 mg/kg of body weight, n = 6) was administered every day during the HFD for the last 16 weeks. A RD-fed mice group (n = 6) and a HFD-fed mice group (n = 6) were included as normal and disease controls, respectively. All experimental procedures were performed in accordance with ethical guidelines issued by the Institutional Animal Care and Use Committee (IACUC) of Sahmyook University for the care and use of laboratory animals (SYUIACUC 2015001).

#### Metabolic Measurements

Body weight and serum glucose levels in mice were monitored every 2nd week. Mice were fasted overnight (12 h), blood samples were taken from a tail cut, and serum glucose levels were measured using the Accu-Chek Performa system (Roche, Switzerland). Intraperitoneal glucose tolerance testing (IPGTT) was performed 16 weeks after statin (Atorvastatin; Ator, Rosuvastatin; Rosu) administration. After overnight fasting (12 h), mice were intraperitoneally injected with glucose solution [2 g/kg of body weight, in phosphate-buffered saline (PBS)] and blood samples were obtained from the tail vein 0, 30, 60, 90, and 120 min after glucose injection. At the end of the experimental period, mice were sacrificed under ether anesthesia and blood was collected via cardiac puncture. Blood sample were centrifuged at 10,000 rpm for 5 min to isolate serum. Serum was prepared to determine levels of total cholesterol, LDL, apolipoprotein A-1 (ApoA-1), and apolipoprotein B (ApoB) using a biochemical analyzer (AU480, Beckman Coulter, United States).

#### Immunomodulatory Biomarkers

Ileum tissue from each group were immediately frozen in liquid nitrogen and stored at -70◦C to determine levels of gene transcripts. The expression of interleukin-1β (IL-1β; forward primer: 5<sup>0</sup> -CAGGATGAGGACATGACACC-3<sup>0</sup> , reverse primer: 5<sup>0</sup> -CTCTGCAGACTCAAACTCCAC-3<sup>0</sup> ), interleukin-6 (IL-6; forward primer: 5<sup>0</sup> -GTACTCCAGAAGACCAGAGC-3<sup>0</sup> , reverse primer: 5<sup>0</sup> -TGC TGG TGA CAA CCA CGG CC-3<sup>0</sup> ), transforming growth factor β 1 (TGFβ1; forward primer: 5 0 -GCGGACTACTATGCTAAAGAGG-3<sup>0</sup> , reverse primer: 5<sup>0</sup> - GTAGAGTTCCACATGTTGCTCC-3<sup>0</sup> ), interleukin-4 (IL-4; forward primer: 5<sup>0</sup> - GAGCCATATCCACGGATGCGACAA-3<sup>0</sup> , reverse primer: 5<sup>0</sup> - CATGGTGGCTCAGTACTACGAGTA-3 0 ), and interleukin-10 (IL-10; forward primer: 5 0 -TGGCCACACTTGAGAGCTGC-3<sup>0</sup> , reverse primer: 5<sup>0</sup> - TTCAGGGATGAAGCGGCTGG-3<sup>0</sup> ) was investigated in ileum tissue. Total RNA was extracted using a RiboExTM (GeneAll, Korea). RNA was then quantified by reading the absorbance at 260 nm. cDNA was synthesized using a HyperScriptTM RT premix (GeneAll, Korea). To quantify the level of mRNA expression, SYBR <sup>R</sup> Green PCR Master Mix (Applied Biosystems, United States) and a StepOnePlusTM real-time PCR system (Applied Biosystems, United States) were used. GAPDH was used as an internal control.

#### Gut Microbiota Analysis

Total DNA was extracted using the PowerSoil DNA Isolation Kit (MO BIO Laboratories, Inc., United States) from cecum samples, which included fecal materials. Partial sequences of 16S rRNA genes were amplified based on the 16S rRNA amplification protocol from the Earth Microbiome Project (Gilbert et al., 2010). 16S rRNA genes were amplified using the 515F/806R primer set, which includes an adapter sequence for amplification of the V4 region (515F forward primer: 5<sup>0</sup> -TCG TCG GCA GCG TCA GAT GTG TAT AAG AGA CAG **GTG CCA GCM GCC GCG GTA A**-3<sup>0</sup> ; 806R reverse primer: 5<sup>0</sup> -GTC TCG TGG GCT CGG AGA TGT GTA TAA GAG ACA **GGG ACT ACH VGG GTW TCT AAT**-3<sup>0</sup> ). To attach the dual indices and adapter to amplified polymerase chain reaction (PCR) products, an index PCR was additionally performed using an AmpONETM α-Pfu DNA polymerase (GeneAll, Korea) and Nextera <sup>R</sup> XT Index Kit v2 (Illumina, United States). PCR products after amplification and attachment were purified using the ExpinTM PCR SV (GeneAll, Korea). Partial bacterial 16S rRNA genes were amplified using the MiSeq Reagent Kit V3 (600 cycles) and MiSeq platform (Illumina, United States) at KoBioLabs, Inc.

Prior to analysis of 16S rRNA sequences, BCL files were converted into raw FASTQ files including read1, index, and read2 sequences using CASAVA-1.8.2 (Illumina). After preprocessing (quality filtering and trimming steps using FASTX-Toolkit) (Gordon and Hannon, 2010), sequences were assigned to operational taxonomic units (OTUs, 97% identity), and representative sequences were selected using QIIME 1.7.0 software (Caporaso et al., 2010). Next, taxonomic composition, alpha diversity (rarefaction curve and Shannon index of bacterial diversity) and beta diversity (PCoA of UniFrac distances) were analyzed. LDA effect size (LEfSe) was used to estimate taxonomic abundance and characterize differences between groups (Segata et al., 2011). Hierarchical clustering of the top ten most abundant bacterial genera by groups was performed using Spearman's rank correlation, and a heat map was generated using MultiExperiment Viewer (MEV) software (v4.8.1). Nextgeneration sequencing (NGS) data are available at<sup>1</sup> .

The relative abundances of four bacterial genera were confirmed using SYBR <sup>R</sup> Green PCR Master Mix (Applied Biosystems, United States) and a StepOnePlusTM real-time PCR system (Applied Biosystems, United States). Genus-specific primer sets for Eubacteria (UniF340; forward primer: 5<sup>0</sup> - ACTCCTACGGGAGGCAGCAGT-3<sup>0</sup> , UniR514; reverse primer: 5 0 -ATTACCGCGGCTGCTCCG-3<sup>0</sup> ), Akkermansia (AM1; forward primer: 5<sup>0</sup> -CAGCACGTGAAGGTGGGGAC-3<sup>0</sup> , AM2; reverse primer: 5<sup>0</sup> -CCTTGCGGTTGGCTTCAGAT-3<sup>0</sup> ), Bacteroides (AllBac296f; forward primer: 5<sup>0</sup> -GAGAGGAA GGTCCCCCAC-3<sup>0</sup> , AllBac412r; reverse primer: 5<sup>0</sup> -CGCTAC TTGGCTGGTTCAG-3<sup>0</sup> ), Butyricimonas (Buty1f; forward primer: 5<sup>0</sup> -GGTGAGTAACACGTGTGCAAC-3<sup>0</sup> , Buty1r; reverse primer: 5<sup>0</sup> -TACCCCGCCAACTACCTAATG-3<sup>0</sup> ), and Mucispirillum (MucisF; forward primer: 5<sup>0</sup> -CGTTTGCAAGAA TGAAACTCAAA-3<sup>0</sup> , MucisR; reverse primer: 5<sup>0</sup> - CACAGCATTATCTCTAACGCCTT-3<sup>0</sup> ) were used for amplification (Layton et al., 2006; Collado et al., 2007; Cahenzli et al., 2013).

#### Fecal Microbiota Transplantation (FMT)

Fecal material (0.1 g) from rosuvastatin-treated mice (HFD-Rosu group) was pooled in 1 mL of PBS. Mice were allowed to drink water containing penicillin G procaine (2,000 IU/L) and streptomycin (2.5 mg/L) for 5 days prior to FMT. After centrifugation at 2,000 g for 2 min, 500 µL of supernatant was administered to antibiotic-treated mice fed a HFD for 48 weeks (HFD-fRosu, n = 6) via oral gavage. Metabolic profiles, inflammatory cytokines and bacterial abundances were analyzed and compared between the RD (n = 5) and HFD (n = 4) groups 4 weeks after FMT.

# Statistical Analysis

Data for each group are presented as the mean ± standard error of mean (SEM). Relative abundance was analyzed using LEfSe based on the Kruskal–Wallis and Wilcoxon tests, and significance was defined at P < 0.05 (alpha value = 0.05). The logarithmic LDA score threshold was set at 3.0. To quantify in vivo mRNA levels relative to an internal control (GAPDH), the 2−1 1 Ct relative quantification method (11Ct = (Ct.Target − Ct.<sup>β</sup> <sup>−</sup>actin)Group<sup>1</sup> − (Ct.Target − Ct.<sup>β</sup> <sup>−</sup>actin)Group2) was used. Statistical significance was assessed by one-way analysis of variance (ANOVA), followed by Duncan's post hoc test with agricolae package in RStudio. All statistical analyses were performed using RStudio. A P value < 0.05 was considered to indicate statistical significance.

#### RESULTS

#### Effects of Statins on Metabolic Improvements

As expected, both atorvastatin and rosuvastatin significantly reduced total cholesterol levels (**Figure 1**). Moreover, statins significantly reduced serum glucose levels and tended to improve glucose tolerance (**Figure 1**). In addition, statins decreased levels of ApoA-1 but not ApoB. There was no significant change in body weight and levels of LDL in response to statins compared to the corresponding values in the HFD group.

#### Effects of Statins on the Gut Microbiota

A total of 334,356 sequences were generated from 24 samples. An average of 13,932 ± 6,137 sequences were recovered per sample and used for comparative analyses. **Figures 2A,B** show the differences in microbial diversity between the RD, HFD, HFD-Ator, and HFD-Rosu groups. The alpha diversities of gut microbiota analyzed using Chao 1 richness and Shannon index did not reveal any significant difference between the groups (**Figure 2A**). The Principal coordinates analysis (PCoA) of UniFrac distances showed a separation between the RD, HFD, HFD-Ator, and HFD-Rosu groups, which was clearly classified in the unweighted PCoA (**Figure 2B**).

#### Effects of Statins and Diet on the Composition of the Gut Microbiota

The phyla Bacteroidetes and Firmicutes from the RD group consisted of 40.0 ± 19.6 [average of relative abundance (%) ± standard deviation] and 38.0 ± 16.0% of gut microbiota, respectively, and the Firmicutes/Bacteroidetes (F/B) ratio was 1.3 ± 1.1%. The abundance of Bacteroidetes (10.8 ± 9.3%) decreased in the HFD group and that of Firmicutes (81.4 ± 13.4%) increased compared to the corresponding abundances in the RD group. The F/B ratio was 24.5 ± 28.6%. In contrast, the abundance of Bacteroidetes and Firmicutes from the HFD-Ator group was 11.5 ± 9.9% and 76.5 ± 10.6%, respectively; the F/B ratio was 10.6 ± 6.4%; and the abundance of Bacteroidetes and Firmicutes in the HFD-Rosu group was 33.7 ± 29.1% and 53.6 ± 24.9%, respectively. The F/B ratio was 2.8 ± 1.9%. These results indicate that rosuvastatin effectively restores the altered gut microbiota by HFD compared to atorvastatin. In addition, the abundance of Deferribacteres in the HFD-Ator group was 11.9 ± 4.4%, which was greater than that in the RD (1.2 ± 0.9%) and HFD (6.9 ± 4.4%) groups (**Figure 2C**). Relative abundances of Bacteroidetes, Firmicutes, and Deferribacteres in the HFD-Ator

<sup>1</sup>https://figshare.com/articles/NGS\_data/9104477

FIGURE 1 | Effect of statins on body weight, serum glucose, intraperitoneal glucose tolerance test (IPGTT), total cholesterol, LDL, ApoA-1, and ApoB. Five-week-old C57BL/6N mice were fed a HFD for 39 weeks to induce metabolic disorders, and then statins were administered daily for 16 weeks. RD: regular diet (n = 6); HFD: high-fat-diet (n = 6); HFD-Ator: atorvastatin administration during HFD feeding (n = 6); HFD-Rosu: rosuvastatin administration during HFD feeding (n = 6). Different superscript letters indicate significant differences (P < 0.05) according to Duncan's post hoc test following ANOVA.

FIGURE 2 | Microbial diversity according to diet and statin therapy. (A) Rarefaction curve and Shannon index of bacterial diversity. (B) PCoA of UniFrac distances. (C) Bacterial classification at the phylum level. Table value indicates relative abundance of phylum level. (D) Heatmap of average linkage hierarchical clustering of the top 10 most abundant bacterial genera using Spearman's rank correlation. A heat map was generated using MultiExperiment Viewer (MEV) software (v4.8.1). Rows represent bacterial genera, and columns represent samples.

and HFD-Rosu group were not significantly different compared to the HFD group.

Among the top 10 most abundant bacterial genera found in the four groups, Bacteroides and Butyricimonas of the phylum Bacteroidetes, Oscillospira of the phylum Firmicutes, and Mucispirillum of the phylum Deferribacteres were closely clustered (**Figure 2D**). Their abundance was enriched in the HFD-Ator and HFD-Rosu groups compared to that in the RD and HFD groups, accounting for 24.3 ± 5.6% and 27.8 ± 14.3% of the total identified bacterial OTUs, respectively (**Figure 2D**). In the LEfSe analysis, the abundances of Anaerotruncus, Bacteroides, Butyricimonas, Dorea, Mucispirillum, and Turicibacter were significantly greater in the HFD-Ator group than in the RD and HFD groups, and the abundances of Bacteroides, Butyricimonas, Clostridium, and Mucispirillum were significantly greater in the HFD-Rosu group. The abundances of Bacteroides, Butyricimonas, and Mucispirillum overlapped between the HFD-Ator and the HFD-Rosu groups (**Figures 3A,B**). Moreover, in comparison between the HFD-Ator and the HFD-Rosu groups, the abundances of Turicibacter, Dorea, and Ruminococcus were significantly greater in the HFD-Ator group than in the HFD-Rosu group, and the abundance of Clostridium was significantly greater in the HFD-Rosu group than in the HFD-Ator group (**Figure 3C**).

Among them, Odoribacteraceae, Deferribacteraceae, and Bacteroidaceae overlapped. At the genus level, Bacteroides, Butyricimonas, and Mucispirillum were increased by both atorvastatin and rosuvastatin.

#### Effects of Statins on Inflammatory Cytokines

TGFβ1 expression increased significantly in the HFD-Ator and HFD-Rosu groups compared to that in the HFD group, whereas IL-1β expression decreased with statin treatments. There was no significant change in the expression of IL-4, IL-6, and IL-10 by statins compared to that in the HFD group (**Figure 4A**). In addition, the expression of TGFβ1 and IL-1β was correlated with increased abundance of the Clostridium, Dorea, Mucispirillum, Butyricimonas, Bacteroides, Anaerotruncus, and Turicibacter by statins. Among them, the positive correlation between TGFβ1 and the abundance of Dorea and the negative correlation between IL-1β and the abundance of Dorea and Mucispirillum were significant in the HFD-Ator group (**Figure 4B**).

#### Effects of Fecal Material Transplantation on Metabolic Improvements

Both atorvastatin and rosuvastatin significantly improved the metabolic disorders, and microbial diversity of the rosuvastatintreated mice group was more different compared to the HFD group than the atorvastatin-treated mice group based on beta diversity. Therefore, fecal material transplantation (FMT) was performed using fecal material from rosuvastatin-treated mice, which significantly improved serum glucose levels and glucose tolerance with the HFD (**Figure 5A**). However, body weight and total cholesterol and LDL levels were not improved by FMT. The abundances of Bacteroides, Butyricimonas, and Mucispirillum were 2.2 ± 4.0%, 0.04 ± 0.02%, and 0.06 ± 0.03% in the HFD-fRosu group, respectively. Among them, the abundance of Bacteroides was significantly decreased and that of Butyricimonas was significantly increased after FMT compared to that in the HFD group (**Figure 5B**). TGFβ1 expression increased significantly, whereas IL-1β expression decreased significantly, in the HFD-fRosu group compared to that in the HFD group (**Figure 5C**).

### DISCUSSION

Modulation of gut microbiota has an impact on metabolic improvements in obesity and T2D (Clarke et al., 2012; Mardinoglu et al., 2016). Recent studies have revealed that certain pharmacotherapies can improve metabolic diseases through specific compositional changes in gut microbiota (Shin et al., 2013; Lee and Ko, 2014). Statins are cholesterol-lowering drugs, which also affect glucose tolerance in animal models. The characteristics of gut microbiota in response to statins have not yet been fully investigated in metabolic diseases (Caparros-Martin et al., 2017; Khan et al., 2018). In this study, changes in gut microbiota in response to atorvastatin and rosuvastatin were observed in mice with HFD-induced obesity, and the abundances of Bacteroides, Butyricimonas, and Mucispirillum were all remarkable. Moreover, FMT with fecal materials from the HFD-Rosu group has an impact on hyperglycemia improvement, which may be related to the abundance of Butyricimonas.

Therapy with atorvastatin and rosuvastatin both induced significant changes in the gut microbiota during the HFD. In a meta-analysis comparing the two statins, similar pharmacokinetics, efficacy, and side effects were observed, which may be associated with the similar changes in gut microbiota induced by atorvastatin and rosuvastatin (Abbas et al., 2012). Nevertheless, differences in gut bacterial composition were observed with atorvastatin and rosuvastatin, which may be caused by differences in chemical structure and the effects on hyperlipidemia and glucose metabolism. Atorvastatin is lipophilic and rosuvastatin is hydrophilic due to the side chain of the methane sulfonamide group, and across the dose range used, rosuvastatin was significantly more effective in decreasing cholesterol than atorvastatin. However, it is unclear how those differences influence gut microbiota.

Alterations in gut microbiota in recent studies were not consistent with the findings of the present study (Nolan et al., 2017; Khan et al., 2018), which may be due to two major differences in the experimental method used. The first is due to differences in the target region of the 16S rRNA gene. In NGS analysis, the target region of the 16S rRNA gene impacts the microbiota results (Rintala et al., 2017). Results of this study are based on PCR amplification and sequencing targeting V4 region, and a different short variable region was targeted in different previous studies. Consequently, that affected the universality of the NGS result, which is why amplification of multiple variable regions is recommended to increase the universality and resolution (Fuks et al., 2018). Second, although both studies were performed using C57BL/6 mice, the gender

and four bacterial families were identified by atorvastatin and rosuvastatin, respectively. Among them, Odoribacteraceae, Deferribacteraceae, and Bacteroidaceae

overlapped. At the genus level, Bacteroides, Butyricimonas, and Mucispirillum were increased by both atorvastatin and rosuvastatin.

and age of the mice were different. The effect of medication on the gut microbiota is influenced by gender (Lee and Ko, 2014). Liu et al. (2018) reported that variation in gut microbiota was associated with the efficacy of rosuvastatin therapy in a clinical study, with differences in gut microbiota observed by age and gender (Liu et al., 2018). In the present study, statins were administered for 16 weeks in 29-week-old aged male mice; therefore, our microbiota results may reflect the characteristics of gut microbiota in response to statin therapy. A decrease in the F/B ratio was observed with the metabolic improvement induced by statins treatment. Bacteroidetes and Firmicutes are the major bacterial phyla, which represent most bacteria in the gut, and their ratio is related with obesity-related metabolic disorders (Ley et al., 2006; Chakraborti, 2015; Louis et al., 2016). Moreover, in a recent study, statin therapy with pravastatin and atorvastatin resulted in gut dysbiosis in HFD-fed female C57BL/6 mice, which was associated with adverse effects induced by statins intolerance, including myopathy and T2D (Caparros-Martin et al., 2017). The results of recent studies suggest that statins influence changes in gut microbiota and that various factors affect the composition of gut microbiota, including age, gender, and diet. Nevertheless, recent studies reported that statin therapy increased butyric acidproducing bacteria, including those belonging to the families Lachnospiraceae, Bacteroidaceae, and Prevotellaceae. The genera Bacteroides and Butyricimonas, which were increased in this study, are butyric acid-producing bacteria, and short-chain fatty acids (SCFAs) may play key roles in metabolic improvements, including hyperlipidemia and hyperglycemia.

SCFAs are produced from dietary fiber by intestinal bacteria, and this has beneficial effects on host energy metabolism

(Den Besten et al., 2013). Among the SCFAs, acetic acid, propionic acid, and butyric acid are most abundant in the intestine, representing over 90% of the SCFAs present (Den Besten et al., 2013). A high abundance of Bacteroides spp. is characteristic of non-obese individuals and is caused by the releases of propionic acid and acetic acid from Bacteroides spp. (Chakraborti, 2015). Another SCFA, butyric acid, improves the outcome of insulin resistance and dyslipidemia (Gao et al., 2009). Several bacterial strains including Clostridium butyricum and Butyricicoccus pullicaecorum are representative butyric acid-producing anaerobic bacteria, which are involved in T2D and inflammatory bowel disease, respectively (Geirnaert et al., 2014; Jia et al., 2017). Additionally, the majority of members of the Prevotellaceae, Clostridiaceae, Ruminococcaceae, Lactobacillaceae, and Lachnospiraceae families are putative butyric acid producers and have anti-inflammatory properties (Esquivel-Elizondo et al., 2017). In the present study, Butyricimonas, a butyric acid-producing bacterium, may have played a key role in metabolic improvements induced by statins, especially in glucose metabolism. Moreover, FMT using fecal materials from the HFD-Rosu group resulted in improved glucose tolerance, and the high abundance of Butyricimonas was solely maintained. However, butyric acid was not investigated in this study, and the mechanisms through which Butyricimonas induce metabolic improvements are unknown. This limitation should be addressed in future studies.

Apart from butyric acid-producing bacteria, Mucispirillum, which inhabits the intestinal mucus layer, may have an important role in the metabolic improvements induced by statins. To date, few studies have discussed how Mucispirillum schaedleri are associated with intestinal inflammation (Loy et al., 2017). However, the role of Mucispirillum in lipid metabolism and inflammation is poorly understood. In the present study, Mucispirillum was more abundant in the HFD group than in the RD group, and was significantly increased by statins with a HFD. Moreover, the abundance of those bacteria was negatively correlated with pro-inflammatory cytokine and positively correlated with anti-inflammatory cytokines in the statin-treated HFD group. Conversely, based on genome analysis, M. schaedleri is not predicted to be a primary degrader of glycan from mucin. Instead, M. schaedleri is likely to utilize SCFAs for energy metabolism (Loy et al., 2017); those abundant in this study may be accompanied by SCFA-producing bacteria in response to statins.

Regulation of IL-1β and TGFβ expression may play a key role in the metabolic improvements associated with statin therapy, due to changes in gut microbiota. A previous study reported that statin therapy regulates inflammatory cytokines such as IL-1β, IL-6, and TNF-α (Lee et al., 2016). In the present study, IL-1β and TGFβ expression increased in the ileum, indicating that the regulation of IL-1β and TGFβ is associated with the effect of statins on the alleviation of hyperglycemia, hyperlipidemia, and inflammation. In the intestinal immune system, the proinflammatory cytokine IL-1β plays an important role in insulin resistance and hyperlipidemia (Rotter et al., 2003; Fentoglu et al., 2011). Especially, upregulation of IL-1β is a general feature underlying obesity-related insulin resistance, which increases intestinal epithelial tight junction permeability and inflammation (Al-Sadi and Ma, 2007). Intestinal permeability has been observed in dysbiosis by HFD, and changes in gut microbiota in response to statin therapy with a HFD played an important role in the downregulation of IL-1β expression and the improvement of downstream glucose control (Winer et al., 2016). Secretion of TGFβ, an anti-inflammatory mediator, occurs under normal physiological conditions in the intestine; statin therapy increased TGFβ expression (Porreca et al., 2002). TGFβ is a pleiotropic cytokine involved in various immune signaling pathways, including intestinal anti-inflammation (Letterio and Roberts, 1998). Above all, TGFβ signaling suppresses T-cell proliferation and activation through Treg differentiation, which is essential for intestinal homeostasis (Bauche and Marie, 2017). In intestinal immunity, TGFβ suppresses inflammatory responses through the canonical pathway in a mechanism mediated by the intracellular signaling proteins SMAD2/3 (Massague, 2012). Interestingly, TGFβ production is induced by gut microbiota, and Clostridium strains including Clostridium butyricum (or derived butyric acid) and Bacteroides fragilis were reported to induce TGFβ production and Treg differentiation (Round and Mazmanian, 2010; Kashiwagi et al., 2015; Hong et al., 2017; Goncalves et al., 2018). Therefore, the anti-inflammatory effects of statin therapy may be associated with the regulation of TGFβ expression in the intestine via the modulation of gut microbiota.

In this study, the effect of gut microbiota on metabolic improvements, especially anti-hyperglycemic, was shown to be induced by FMT. After FMT, the relative abundance of Butyricimonas increased significantly, compared to that in the HFD group, which may play a key role in the anti-hyperglycemic effect. In a previous study, the increased abundance of Butyricimonas and Bacteroides with metformin therapy was associated with metabolic improvements, including body weight, glucose tolerance, and lipid metabolism, and the administration of fecal material from metformin-treated mice had an effect on body weight change only, which was associated with the abundance of Bacteroides (Lee H. et al., 2018). Additionally, a recent study reported that Bacteroides prevented obesity and insulin resistance (Yang et al., 2017). Along with previous studies, the present study indicates that Butyricimonas may have a more important role in hyperglycemia than in obesity. Moreover, the

regulation of IL-1β and TGFβ1 by FMT was consistent with the effects of statin therapy during HFD. FMT is a therapeutic option for inflammatory bowel disease induced by Clostridium difficile infection, and recent studies have reported metabolic improvements, including insulin sensitivity, with FMT (Kootte et al., 2017). However, the mechanism underlying the effect of FMT on metabolism is unclear, and neither criteria for donor selection nor optimal processes were established. In this study, the anti-hyperglycemic effect persisted for 1 month after FMT in antibiotic-treated mouse model, which provides information for successful FMT therapy in metabolic diseases treatment.

#### CONCLUSION

Statin therapy with atorvastatin and rosuvastatin significantly altered the gut microbiota during HFD in an aged obese mice model. The abundance of the genera Bacteroides, Butyricimonas, and Mucispirillum was significantly increased by statins, and this was related to hyperglycemia and hyperlipidemia. In particular, Butyricimonas may be related to the anti-hyperglycemic effect of statins. Furthermore, the downregulation of IL-1β and the upregulation of TGFβ1 by statins were significantly correlated with the abundance of those bacteria. These results suggest that the alterations in gut microbiota by statins may be one of therapeutic targets for the treatment of hyperglycemia.

#### REFERENCES


# DATA AVAILABILITY

The datasets generated for this study can be found in FigShare https://figshare.com/articles/NGS\_data/9104477.

#### ETHICS STATEMENT

This study was approved by the Institutional Animal Care and Use Committee (IACUC) of Sahmyook University for the care and use of laboratory animals (SYUIACUC 2015001).

#### AUTHOR CONTRIBUTIONS

HL, HK, YS, and KK contributed to the study conception and design. HL, JK, and JA performed by the experiments. HL and JA performed the data analysis and interpretation of the data. HL and JA contributed to the manuscript drafting. All authors have approved the final version of this manuscript.

#### FUNDING

This research was supported by the Korea Institute of Planning and Evaluation for Technology in Food, Agriculture, Forestry, and Fisheries, IPET (No. 314044-3).

diet, gut microbiota, and host energy metabolism. J. Lipid Res. 54, 2325–2340. doi: 10.1194/jlr.R036012


microbiota and host gene expression profiles. Am. J. Physiol. Gastrointest. Liver Physiol. 312, G488–G497. doi: 10.1152/ajpgi.00149.2016


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Kim, Lee, An, Song, Lee, Kim and Kong. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fmicb-10-01947 August 31, 2019 Time: 18:21 # 10

# Lactobacillus brevis Alleviates DSS-Induced Colitis by Reprograming Intestinal Microbiota and Influencing Serum Metabolome in Murine Model

#### Sujuan Ding<sup>1</sup> , Yong Ma<sup>1</sup> , Gang Liu1,2 \*, Wenxin Yan<sup>1</sup> , Hongmei Jiang<sup>1</sup> and Jun Fang<sup>1</sup> \*

<sup>1</sup> College of Bioscience and Biotechnology, Hunan Agricultural University, Changsha, China, <sup>2</sup> Hunan Provincial Key Laboratory of Animal Nutritional Physiology and Metabolic Process, CAS Key Laboratory of Agro-Ecological Processes in Subtropical Region, Institute of Subtropical Agriculture, Chinese Academy of Sciences, National Engineering Laboratory for Pollution Control and Waste Utilization in Livestock and Poultry Production, Changsha, China

#### Edited by:

Stephen J. Pandol, Cedars-Sinai Medical Center, United States

#### Reviewed by:

Guoqiang Zhu, Yangzhou University, China Naoki Asano, Tohoku University, Japan

#### \*Correspondence:

Gang Liu gangle.liu@gmail.com Jun Fang fangjun1973@hunau.edu.cn

#### Specialty section:

This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology

Received: 25 March 2019 Accepted: 27 August 2019 Published: 18 September 2019

#### Citation:

Ding S, Ma Y, Liu G, Yan W, Jiang H and Fang J (2019) Lactobacillus brevis Alleviates DSS-Induced Colitis by Reprograming Intestinal Microbiota and Influencing Serum Metabolome in Murine Model. Front. Physiol. 10:1152. doi: 10.3389/fphys.2019.01152 The aim of this study was to examine the effects of Lactobacillus brevis on the microbial community and serum metabolome in colitis induced by dextran sulfate sodium (DSS). ICR mice were randomly distributed into three treatment groups: (i) L. brevis treatment alone (control), (ii) DSS administration alone, and (iii) treatment with L. brevis and DSS. Our results demonstrate that L. brevis treatment significantly alleviated DSSinduced body weight loss and colon inflammation. In addition, LC-MS analysis of serum metabolites revealed that L. brevis treatment increased the serum level of metabolites against inflammatory responses or oxidative stressors caused by DSS in the murine model. By detecting colonic microbiota, L. brevis increased colonic microbial diversity after challenging with DSS, and increased the relative abundance of Alloprevotella at genus, but Bacteroidales was reduced (P < 0.05). These result indicated that L. brevis could lower the severity of colitis induced by DSS via improving reprogramming the serum metabolome and intestinal microbiota. These findings suggest that the probiotic L. brevis may prevent tissue damage from colitis.

#### Keywords: colitis, murine, Lactobacillus brevis, intestinal microbiota, metabolome

# INTRODUCTION

The regression of mucosal inflammation promotes the recovery of the epithelial barrier and proper tissue repair. It also restores normal organ function and steady-state microenvironment conditions (Leoni et al., 2015; Lopetuso et al., 2018). The pathogenesis of inflammatory bowel disease (IBD) is associated with dysfunction and a delay in mucosal healing (Neurath and Travis, 2012). Ulcerative colitis (UC) is a chronic IBD with unknown origin (Ng et al., 2013). Its recurrence and regression can severely affect patients' quality of life. Despite efforts to reveal its pathogenesis and develop new therapies, UC remains a challenge for the medical and scientific communities (Detel et al., 2016).

Probiotics are live microorganisms that, when administered in adequate amounts, confer health benefits on the host (Hill et al., 2014). Indeed, it has long been proposed that the consumption of lactic acid bacteria in fermented products can improve health and longevity in humans

(Martin et al., 2017). The bacteria in the gut constantly interact with human cells in a variety of ways (Sekirov et al., 2010). It is widely established that the intestinal microbiota can regulate epithelial function, prevent pathogenic bacteria colonization, and affect immune responses (Martin et al., 2017). Several studies have revealed that Lactobacillus brevis KB290 can reduce the severity of colitis in murine models by increasing the ratio of CD11c<sup>+</sup> MP dendritic cells to CD103<sup>−</sup> dendritic cells in the colon (Fuke et al., 2018). Studies have also shown that L. brevis SBC8803 can regulate the expression of tumor necrosis factor α and interleukins 1b and 12 and improve the barrier function of the intestinal epithelium under oxidative stress. These results indicate that L. brevis SBC8803 contributes to the maintenance of intestinal homeostasis and can alleviate intestinal inflammation (Ueno et al., 2011). Moreover, a study has found that polyphosphate, an active substance from L. brevis, can down-regulate the expression of inflammatory and fibrosis-related molecules in intestinal epithelial cells to prevent inflammation and fibrosis (Kashima et al., 2015).

Metabolomic analysis is emerging as a powerful approach in system biology research because it provides unique insights into the understanding of organisms, disease diagnosis, pathology, and toxicology. Metabolomics can reveal the type of ongoing cellular processes in the organism by analyzing the chemical fingerprints of cells (Xu et al., 2012; Weng et al., 2015). For example, metabolomics analyses of serum samples from children with Crohn's disease (CD) and UC have revealed that most of the chemically annotated metabolites belonged to the phospholipids and that they were reduced in CD and UC patients relative to healthy individuals subjects (Daniluk et al., 2019). In addition, the partial least-squares discriminant analysis (PLS-DA) load map of murine models of dextran sulfate sodium (DSS)-induced colitis showed that succinic acid, indole-3-acetic acid, glutamic acid, and glutamine are the main metabolites that separate the stages of colitis (Shiomi et al., 2011).

Currently, most probiotic strains are Lactobacilli and other lactic acid bacteria or Bifidobacteria (de Vrese and Schrezenmeir, 2008). There is mounting evidence that Lactobacillus can counteract intestinal inflammation, but the specific mechanism for such protection is unclear (Sung and Park, 2013). DSSinduced colitis in model animals has been shown to anthropic UC in diverse disease pathologies, ranging from disordered immune responses to clinical manifestations (Detel et al., 2016). Here, using a murine model, we aimed to investigate the effects of L. brevis on the severity of DSS-induced colitis and changes in the gut microbiota and the serum metabolome and to reveal the correlation between the intestinal microbiota and serum metabolites.

# MATERIALS AND METHODS

#### Bacteria Preparation

Lactobacillus brevis stored in the Chinese Academy of Sciences Key Laboratory of Agro-ecological Processes in Subtropical Region (Changsha, China) were incubated in MRS broth (De Man, Rogosa, Sharpe) at 37◦C for 24 h. Cultured bacterial fluid was cultured in MRS agar medium for 24 h at 37◦C. Colonies were quantified, and the number of colonies per mL was determined. The bacteria samples were centrifuged at 5000 rpm for 10 min at 4◦C. The pellet was resuspended in sterile normal saline solution at 2 to 5 × 10<sup>10</sup> CFU/mL.

### Animal and Experimental Design

All animal procedures were approved by the Animal Ethics Committee of Hunan Agricultural University. Eight-week-old female ICR mice were purchased from SLAC Laboratory Animal Central (Changsha, China). DSS was purchased from Dalian Meilun Biotech., Co., Ltd. (M.W: 36000–50000, Dalian, China). All mice were housed in a pathogen-free mouse colony in a standard environment (temperature, 23◦C ± 1 ◦C; humidity, 50 ± 10%; 12-h alternating lighting cycles). After a 7-day adaptation period, the mice were randomly assigned to three groups of eight. The first group was treated with L. brevis (control), the second group was treated with 5% DSS (DSS), and the third group was treated with L. brevis and 5% DSS (LB-DSS). All mice had unlimited access to the basal diet feed and water. Our study lasted for a total of 19 days, the mice was administrated via intragastric administration with L. brevis or sterile saline, then DSS-treated group was given with 5% DSS in water from days 7 to 12, and continued to treat with L. brevis or sterile saline (12–19 days). At the end of the 19 days, the mice underwent 8 h of fasting, and blood samples were collected from the orbital blood vessels. The mice were then killed, and the colon lengths were measured. The middle part of the colon was then fixed using 4% formaldehyde. The colon contents were collected, frozen in liquid nitrogen, and stored at −80◦C. The final weight of the mice was recorded.

#### Colonic Histopathology

Colon samples were fixed using 4% formaldehyde, dehydrated using ethanol gradient, and embedded in paraffin, and 8-µm sections were stained with hematoxylin and eosin and viewed with an Olympus BX41 microscope (Münster, Germany).

Colon tissue samples were examined and graded according to the severity of inflammation as previously described (Vukelic et al., 2018). Histological examination showed that each colon tissue specimen was graded according to the severity of inflammation (no, 0; mild, 1; moderate, 2; severe, 3); inflammatory cell infiltration (normal, 0; mucosa, 1; mucosa with submucosa, 2; transmural extension of infiltration, 3); epithelial lesions (intact, 0; deformation of crypt structure, 1; erosion, 2; ulcer, 3); the degree of lesion (none, 0; point, 1; multifocal, 2; diffuse, 3); edema score (no, 0; mild edema of mucosa, 1; submucosa and mucosa, 2; whole wall of colon, 3). The final histologic injury score for the colon was the sum of the individual scores.

#### Serum Metabolomic Analyses

Serum samples were thawed slowly on ice. A 100-µL serum aliquot was added to 300 µL of methanol (Merck, Darmstadt,

Ding et al. Lactobacillus brevis and Colitis

Germany) and 10 µL of 2.8 mg/mL 2-Chloro-L-phenylalanine as internal standard (Sigma, St. Louis, MO). The mixture was briefly shaken for 30 s using a vortex and kept at −20◦C for 1 h. They were then centrifuged at 12,000 rpm for 10 min at 4 ◦C; and 200-µL aliquots of each of the sample supernatant were transferred into sampling bottles for analysis. Each sample was analyzed using a LC-MS analysis platform (Thermo Fisher Scientific, Ultimate 3000LC, Q Exactive) and a Hyper gold C<sup>18</sup> (3 µm, 100 × 4.6 mm) column (Waters, Dublin, Ireland). The column temperature was maintained at 40◦C and the flow rate at 0.35 mL/min. The automatic injector temperature was 4 ◦C and the injection volume was 10 µL. The mobile phase was composed of 5% acetonitrile and 0.1% formic acid (A) (Merck) and acetonitrile with 0.1% formic acid (B). Gradient elution of the mobile phase is shown below: 0/5, 1/5, 2/40, 7/80, 11/95, 15.01/95, 15.5/5 and 19.5/5 (min/%B). The MS parameters were as follows, ESI+: Heater Temp = 300◦C, Sheath Gas Flow rate = 45 arb, Aux Gas Flow Rate = 15 arb, Sweep Gas Flow Rate = 1 arb, spray voltage = 3.0 kV, Capillary Temp = 350◦C, S-Lens RF Level = 30%; ESI–: Heater Temp = 300◦C, Sheath Gas Flow rate = 45 arb, Aux Gas Flow Rate = 15 arb, Sweep Gas Flow Rate = 1 arb, spray voltage = 3.2 kV, Capillary Temp = 350◦C, S-Lens RF Level = 60%. Data were extracted and analyzed according to a previous study (Ding et al., 2019).

#### 16S Ribosomal RNA Amplicon Sequencing

Fecal samples were prepared for the MiSeq Library, followed by Illumina MiSeq 2 × 300 bp high throughput sequencing and bioinformatics analysis. Briefly, the cDNA library was constructed using a two-step PCR amplification method. Firstly, the target fragment was amplified using specific primers (inner primers) and recovered by glue. Then, the recovered product was used as the template for secondary PCR amplification (outer primers). The inner primer was F-(5<sup>0</sup> -TTC CCT ACA CGA CGC TCT TCC GAT CT-specific primer-3<sup>0</sup> ) and R-(5<sup>0</sup> -GAG TTC CTT GGC ACC CGA GAA TTC CA- specific primer -3<sup>0</sup> ), and the outer primers was F-(5<sup>0</sup> -AAT GAT ACG GCG ACC ACC GAG ATC TAC AC- barcode – TCT TTC CCT ACA CGA CGC TC -3<sup>0</sup> ) and R-(5<sup>0</sup> -CAA GCA GAA GAC GGC ATA CGA GAT- barcode – GTG ACT GGA GTT CCT TGG CAC CCG AGA-3<sup>0</sup> ). All PCR products were recovered using AxyPrepDNA gel recovery kit (AXYGEN, Hangzhou, China), and quantitative analysis was conducted using FTC-3000 real-time PCR. The samples were mixed according to the molar ratio, and the library was prepared for standard Illumina sequencing using HiSeq2500 PE250 (Illumina, United States).

#### Correlation Analysis Between Serum Differential Metabolites and Colonic Microbiota

To explore the relationship between serum metabolic profiles and colonic microbiota in the development of DSS-induced colitis, Pearson correlation analysis was carried out using GraphPad Prism 7.00 for Windows software. Statistical significance was defined by a P-value of less than 0.05.

#### Data Analysis

All quantifications underwent analysis of variance to test the homogeneity of variance by Levene's test and Student's t-test (SPSS 21 software). Statistical significance was defined by a P-value of less than 0.05.

# RESULTS

#### Lactobacillus brevis Counteracts DSS-Mediated Weight Loss and Colonic Injury in Mice

The mice were treated with or without L. brevis for 7 days, and then given 5% DSS for 5 days, followed by L. brevis treatment for another 7 days. The final animal weight, colon weight, and colon length were quantified (**Figure 1**). We observed that the mean weight of DSS mice was lower than those of control and LB-DSS mice (P ≤ 0.05), indicating that DSS treatment reduced overall body weight, and that L. brevis treatment counteracted DSS-mediated body weight loss. In addition, we observed no significant differences between the mean weights of control and LB-DSS mice, indicating that L. brevis did not affect the homeostatic body weight in the absence of DSS.

#### Lactobacillus brevis Retards DSS-Induced Colitis

fphys-10-01152 September 14, 2019 Time: 12:26 # 4

Histologic examination showed multiple erosive lesions and extensive inflammatory cell infiltration in the colon tissue of mice given with DSS. Infiltrated cells mainly consisted of macrophages, lymphocytes, neutrophils, and occasional eosinophils (**Figure 2B**). Intuitively, no lesions were observed in the colon tissue of control mice (**Figure 2A**). Although the colon tissue of LB-DSS mice showed some inflammatory lesions (**Figure 2C**), the severity of inflammation was less than that in the DSS mice (P ≤ 0.05; **Figure 2D**).

#### Lactobacillus brevis Affects the Serum Metabolomic Profiles During Colitis

Principal component analysis (PCA) was used to determine the intrinsic similarity of spectral profiles (**Figures 3B3,C3**). Each scatterplot displays a serum sample in positive ion model (**Figures 3A1,B1,C1**) and negative ion model (**Figures 3A2,B2,C2**). This analysis was performed by Umetrics (Sweden). To reveal significant differences in metabolite abundance, two-group comparisons of the samples were conducted and analyzed using supervised multidimensional statistical method PLS-DA (**Figures 3B2,B4,C2,C4**). The variable importance in the projection (VIP > 1.5) of PLS-DA mode and P-values were used to identify metabolites that exhibited differential abundances (**Table 1**). A total of 22 metabolites were obtained, including 20 in the control vs. DSS comparison and 14 in the DSS vs. LB-DSS comparison. We found that the abundances of 2-hydroxyglutarate, epinephrine, oxalacetic acid, pyridine, guanosine, 6-methylthioguanosine monophosphate, N1-acetylspermidine, and ascorbic acid exhibited opposite trends in the control vs. DSS (increase) and DSS vs. LB-DSS (decrease) comparisons (**Table 1**). We found that the abundances of cholesteryl acetate, tetrahydrocortisone, carnosic acid, and N-undecanoylglycine also exhibited opposite trends in the control vs. DSS (decrease) and DSS vs. LB-DSS (increase) comparisons (**Table 1**). In addition, we observed some unique metabolites, such as serotonin, 11-dehydrocorticosterone, and indole (relative increase) and vitamin A2, gamma-linolenic acid, glycochenodeoxycholate-3-sulfate, arachidonic acid, and docosahexaenoic acid (relative decrease) in the control vs. DSS comparison. Lastly, two metabolites (7α-hydroxy-3-oxo-4 cholestenoate and 25,26-dihydroxyvitamin D) exhibited relative increases in the DSS vs. LB-DSS comparison.

#### Lactobacillus brevis Affects the Alpha Diversity in the Murine Colon During Colitis

The 16S rRNA in v3-v4 region extracted from the colonic samples were analyzed. The colonic microbiota diversity was determined using the Chao index, ACE index, Shannon index, and Simpson index (**Figures 4A–D**). We found that the colonic microbiota diversity decreased in DSS mice compared with the control mice. Importantly, LB-DSS mice exhibited higher Shannon and Simpson indices than the DSS mice (P ≤ 0.05) (**Figures 4C,D**). Albeit insignificant, similar trends were observed for the Chao and ACE indices (**Figures 4A,B**). These results indicated that L. brevis prevented the loss of microbial diversity in the mouse colon after DSS challenge.

# Lactobacillus brevis Affects Microbial Abundance at the Phylum Level During Colitis

The dominant bacterial phyla observed in the mouse colon were Bacteroidetes, Firmicutes, Verrucomicrobia and Proteobacteria. They accounted for more than 97% of the microbiota. The proportion of Bacteroidetes was 41.09, 56.53, and 50.36%, that of Firmicutes was 22.75, 31.78, and 24.09%, that of Verrucomicrobia was 12.86, 16.45, and 19.02%, and that of Proteobacteria was 5.44, 7.87, and 5.04% in the control group, DSS group, and LB-DSS group, respectively (**Figure 5A**). Moreover, the relative abundance of Bacteroidetes was increased in the DSS group compared with the other two groups (**Figure 5B**), but there are no other differences (**Figures 5C–E**).

# Lactobacillus brevis Affects Microbial Abundance at the Order During Colitis

The top ten most abundant microbial orders are shown in **Figure 6A**. The three most abundant bacteria were Bacteroidales, Clostridiales, Verrucomicrobiales, and Lactobacillales. The proportion of Bacteroidales is 56.41, 41.07, and 50.07%, that of Clostridiales was 15.38, 17.58, and 19.69%, that of Verrucomicrobiales is 12.86, 16.45, and 19.02%, that of Lactobacillales is 4.93, 11.16, and 3.10% in the control group, DSS group, and LB-DSS group, respectively (**Figure 6A**). Moreover, the relative abundances of Bacteroidales (**Figure 6C**) and Lactobacillales (**Figure 6D**) were higher in the DSS group than in the other two groups (P ≤ 0.05). And there are no other differences (**Figures 6B,E–G**).

### Lactobacillus brevis Affects the Abundance of Microbial Genera During Colitis

The top ten most abundant microbial genera were selected and made into bar percentage for analysis. The top three most abundant bacterial genera were found to be Akkermansia, Bacteroides, and Lactobacillus. The proportion of Akkermansia was 12.86, 16.45, and 19.02%, that of Bacteroides was 9.16, 12.05, and 17.23%, that of Lactobacillus was 4.06, 10.89, and 2.84% in the control group, DSS group, and LB-DSS group, respectively (**Figure 7A**). Moreover, the relative abundance of Lactobacillus was lower in the DSS group than in the other two groups (**Figure 7D**, P ≤ 0.05). The abundance of Alloprevotella was higher in the control group than in the DSS group and lower in the LB-DSS group than in the DSS group (**Figure 7F**, P ≤ 0.05). And there are no other differences (**Figures 7B,C,E,G**).

#### Correlation Between Serum Metabolites and Colonic Microbiota

Pearson correlation is one of the most commonly used statistical analysis method (Hannigan and Lynch, 2013). The results of Pearson correlation (r) between metabolites and microbiota are shown in **Figure 8**. We found different levels of correlation in six groups of correlation analyses (P < 0.05): serotonin vs. Bacteroidetes (r = 0.472; P = 0.002; **Figure 8A**), serotonin vs. Bacteroidales (r = 0.47; P = 0.002; **Figure 8B**), serotonin vs. Lactobacillus (r = 0.669; P = 0.000; **Figure 8C**), arachidonic acid vs. Bacteroidetes (r = 0.648; P = 0.001; **Figure 8D**), arachidonic

TABLE 1 | Metabolomic changes in serum of control, DSS, and LB-DSS mice.


Control vs. DSS, the control group compared with the DSS group; DSS vs. LB-DSS, the DSS group compared with LB-DSS group. FC, fold change (the ratio of mean metabolite levels between the groups), (Log) FC > 0 indicated a relatively higher concentration in the control or DSS group, while (Log) FC < 0 indicated a relatively lower concentration in the control or DSS group. VIP, variable importance in the projection; ↑/↓, increase/decrease, –, no changes. Significant threshold was set at VIP > 1.50 and P < 0.05.

acid vs. Bacteroidetes (r = 0.643; P = 0.001; **Figure 8E**), and N1-acetylspermidine vs. Lactobacillus (r = −0.498; P = 0.013; **Figure 8F**).

#### DISCUSSION

Inflammatory bowel disease patients are usually treated with anti-inflammatory, immunosuppressive, or biological agents, and some may even require surgery (Reid et al., 2011; Souza et al., 2015). Although these strategies can alleviate IBD symptoms, they do not provide permanent cure due to the cyclical and lifelong nature of IBD (Gareau et al., 2010). In recent years, an increasing body of evidence suggests that probiotics can prevent IBD in both experimental models and humans. In this study, our results reveal that L. brevis counteracted DSS-mediated body weight loss and inflammatory disruption of the colon in mice. In addition, L. brevis increased microbial diversity and the relative abundance of Bacteroidetes and Bacteroidales microbials in the colon after DSS challenge. In contrast, L. brevis treatment reduced the relative abundance of Alloprevotella microbials in the colon after DSS challenge. Metabolomic analysis showed that L. brevis treatment increased antiinflammatory metabolites such as gamma-linolenic acid and carnosic acid and antioxidant metabolites such as ascorbic acid and 25,26-dihydroxyvitamin D in the serum.

The mammalian gastrointestinal tract is constantly exposed to a large number of bacteria, food, and environmental toxins (Kotula et al., 2014). The human intestinal mucosal surface is the largest body surface (∼200–300 m<sup>2</sup> ) that is in constant contact with the external environment (Lievin-Le Moal, 2013). The intestinal epithelium is lined by a single layer of columnar epithelial cells and is folded into concave or crypts. These fully differentiated epithelial cells act as physical and functional barriers to defend the body against potentially harmful microbes and viruses within the gut microenvironment (Lievin-Le Moal and Servin, 2006). Studies have shown that DSS can reduce body weight, increase intestinal permeability, ulceration, inflammatory cell infiltration, and goblet cell loss in murine models (Gadaleta et al., 2011). Another study showed that L. brevis G-101 given at a dose of 1 × 10<sup>8</sup> CFU/per mouse can prevent weight loss, colon shortening, and inflammation in ICR mouse models after trinitrobenzene sulfonic acid treatment (Jang et al., 2013). In addition, heat-killed body of L. brevis SBC8803 can alleviate DSSinduced intestinal tissue injuries and improve survival in mice (Ueno et al., 2011). Our results here suggest that L. brevis counteracted DSS-mediated body weight loss

and colon inflammation and injury without affecting colon length and weight.

Many microbes have been found to have stably colonized in our intestines for decades, but their relative abundances vary with our diets (Faith et al., 2013; David et al., 2014). In mice and humans, the abundance of the bacterial community can be significantly altered within 24 h of dietary changes (Wu et al., 2011; Holmes et al., 2017). These findings provide new insights into the treatment of IBD because the intestinal microbiota plays an important role in intestinal inflammatory responses, including changes in the relative abundance and diversity of intestinal microbiota immunity (Lepage et al., 2011; Dubin et al., 2016). In the T5KO mouse model of adherentinvasive Escherichia coli (AIEC)-induced chronic colitis, AIEC invasion of the intestine is associated with the reduction of the overall microbiota diversity (Chassaing et al., 2014). Bacteroides is the dominant phylum in infectious colitis. It is the most abundant bacterial phylum in the healthy population, and the decline of its relative abundance is related to obesity and chronic diarrhea in humans (Eckburg et al., 2005; Turnbaugh et al., 2009; Khoruts et al., 2010). Culture-based studies have reported an increase in the abundance of Bacteroidetes in the colonic mucosa of IBD patients (Wu et al., 2013). Our results showed that L. brevis protected the diversity of colonic microbial community and decreased the relative abundance of Bacteroidetes after DSS treatment. We found a lower relative abundance of Lactobacillales in DSS mice than in control mice. This may be because the length of time L. brevis colonized the colon does not necessarily include the time when we collect colon contents. Therefore, even if L. brevis was given to the mice, it would not necessarily increase the relative abundance of Lactobacillales in the colon.

Metabolomic analysis provides data on all metabolic processes in cells and organisms (Shiomi et al., 2011). This quantitative analysis of metabolites is a promising method to identify biomarkers in IBD (Shiomi et al., 2011). In the control vs. DSS comparison, our results indicate that the concentrations of 2-hydroxyglutarate, epinephrine, oxalacetic acid, pyridine, guanosine, 6-methylthioguanosine monophosphate, N1 acetylspermidine, and ascorbic acid were higher and that those of cholesteryl acetate, tetrahydrocortisone, carnosic acid, and N-undecanoylglycine were lower. However, these changes exhibited opposite trends in the DSS vs. LB-DSS comparison. The N1-acetylspermidine metabolites, which are related to amino acid metabolism and oxidative stress in cells, are polyamines derived from ornithine and methionine and play an important role in cell membrane stability, biosynthesis of signaling molecules, and cell growth and differentiation

FIGURE 5 | Analysis of the microbial composition at the phylum level. (A) Relative abundances of microbial phyla in the mouse colon. Comparisons of the relative abundances of Bacteroidetes (B), Firmicutes (C), Proteobacteria (D), and Actinobacterial (E) in the colon of control, DSS, and LB-DSS mice. <sup>∗</sup>P < 0.05.

(Ramot et al., 2010; Liu et al., 2015). Arachidonic acid is the main precursor of eicosanoid mediators, and the abundance of this metabolite is greatly increased after cell activation (Calder, 2011). Some metabolites are related to inflammation, such as ascorbic acid and guanosine (Sorice et al., 2014; Bellaver et al., 2015). Monoamine serotonin [5-hydroxytryptamine (5-HT)] is an important regulator of gastrointestinal tract and other organ systems and is a neurotransmitter in the brain (Yano et al., 2015).

In summary, our study provides insights into the mitigative role of L. brevis in colitis by regulating the colonic microbial community and altering the serum metabolome. Specifically, L. brevis retarded the manifestation of colitis, such as body weight loss and colonic tissue damage. L. brevis also

FIGURE 7 | Analysis of the microbial composition at the genus level. (A) Relative abundances of microbial genera in the mouse colon. Comparisons of relative abundance of Akkermansia (B), Bacteroides (C), Lactobacillus (D), Parasutterella (E), Alloprevotella (F), and Desulfovibrio (G) in the colon of control, DSS, and LB-DSS mice. <sup>∗</sup>P < 0.05.

improved the intestinal microorganism diversity, reduced the relative abundance of pathogenic bacteria, and altered the levels of serum metabolites. Our data indicate that the serotonin level was positively correlated with Bacteroidales and Lactobacillus abundances and negatively correlated with Lactobacillus abundance. The arachidonic acid level was negatively correlated with the abundances of Bacteroidetes and Bacteroidales, and the N1-acetylspermidine level was negatively correlated with the abundance of Lactobacillus. Serotonin plays an important role in gastrointestinal motility and intestinal dopaminergic neuron development, but its specific mechanism of influence is unknown. Taken together, our study demonstrates the feasibility and potency of using L. brevis in the treatment of IBD and provides a basis for further investigation regarding L. brevis and colitis.

#### DATA AVAILABILITY

fphys-10-01152 September 14, 2019 Time: 12:26 # 10

The data supporting this study can be found in NCBI using accession number SRP220500 (https://www.ncbi.nlm.nih.gov/ sra/SRP220500).

#### ETHICS STATEMENT

All animal procedures were approved by the Animal Ethics Committee of Hunan Agricultural University.

#### AUTHOR CONTRIBUTIONS

SD and YM performed the study and conducted data analysis. GL and JF designed the research. WY provided assistance for the

#### REFERENCES


study. SD and YM prepared the first draft of the manuscript. All authors read and revised the manuscript.

### FUNDING

This research was supported by National Natural Science Foundation of China (Nos. 31772642 and 31672457), National Key Research and Development Program of China (2016YFD0500504 and 2016YFD0501201), Ministry of Agriculture of the People's Republic of China (2015-Z64 and 2016-X47), Local Science and Technology Development Project Guided by The Central Government (YDZX20184300002303 and 2018CT5002), and Hunan Provincial Science and Technology Department (2017NK2322, 2018TP1031, 2016NK2101, 2016WK2008, 2016TP2005, and 2018WK4025), China Postdoctoral Science Foundation (2018M632963), and Double first-class construction project of Hunan Agricultural University (2019T120705 and SYL201802003).



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Ding, Ma, Liu, Yan, Jiang and Fang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Colonic Microbiota and Metabolites Response to Different Dietary Protein Sources in a Piglet Model

Rui Li 1,2, Ling Chang1,2, Gaifeng Hou1,2, Zehe Song1,2 \*, Zhiyong Fan1,2, Xi He1,2 and De-Xing Hou1,3

*<sup>1</sup> College of Animal Science and Technology, Hunan Agricultural University, Changsha, China, <sup>2</sup> Hunan Co-Innovation Center of Animal Production Safety, Changsha, China, <sup>3</sup> Department of Food Science and Biotechnology, Faculty of Agriculture, Kagoshima University, Kagoshima, Japan*

#### Edited by:

*Helieh S. Oz, University of Kentucky, United States*

#### Reviewed by:

*Wenkai Ren, South China Agricultural University, China Hein Min Tun, School of Public Health, The University of Hong Kong, Hong Kong Mrigendra Rajput, Arkansas Tech University, United States Bingkun Zhang, China Agricultural University (CAU), China*

> \*Correspondence: *Zehe Song zehesong111@163.com*

#### Specialty section:

*This article was submitted to Food Microbiology, a section of the journal Frontiers in Nutrition*

Received: *27 February 2019* Accepted: *30 August 2019* Published: *24 September 2019*

#### Citation:

*Li R, Chang L, Hou G, Song Z, Fan Z, He X and Hou D-X (2019) Colonic Microbiota and Metabolites Response to Different Dietary Protein Sources in a Piglet Model. Front. Nutr. 6:151. doi: 10.3389/fnut.2019.00151* Dietary protein sources have the potential to affect the colon microbiome of piglets that will subsequently have a large impact on metabolic capabilities and hindgut health. This study explored the effects of different protein sources on the growth performance, diarrhea rate, apparent ileal digestibility (AID) of crude protein (CP), colonic mucin chemotypes, colonic microbiome, and microbial metabolites of piglets. Twenty-four piglets were randomly divided into four groups that received isoenergetic and isonitrogenous diets containing either Palbio 50 RD (P50), Soyppt-50% (S50), concentrated degossypolized cottonseed protein (CDCP), or fish meal (FM) as the sole protein source. The experimental diets did not affect the estimated daily gain (EDG), but P50 increased fecal score compared with S50 and CDCP. CDCP increased, but P50 reduced AID of CP in comparison to FM and S50. S50 and CDCP increased the amount of mixed neutral-acidic mucins relative to P50. Venn analysis identified unique OTUs in the P50 (13), CDCP (74), FM (39), and S50 (31) groups. The protein sources did not change the colonic bacterial richness or diversity. High *Escherichia* abundance in the P50 and FM, great abundant of *Lactobacillus* in the CDCP, and high *Gemmiger* abundance in the S50 were found. The CDCP tended to elevate valeric acid and branched chain fatty acid (BCFA) concentrations compared with the other diets. The P50 and FM groups had greater ammonia nitrogen and methylamine contents than the S50 and CDCP groups. There was a positive correlation between the *Escherichia* and ammonia nitrogen, the *Lactobacillus* and short chain fatty acid (SCFA), and a negative correlation between the *Gemmige* and BCFA. These findings suggested short-term feeding of different protein sources did not affect the piglets' growth, but P50 increased the diarrhea rate. Potential pathogenic bacteria and detrimental metabolites appeared in the colons of piglets fed P50 and FM, whereas, beneficial effects were conferred upon piglets fed CDCP and S50, thus indicating that available plant proteins (cotton seed, soy) added to the diets of piglets enhanced colon health by reducing protein fermentation.

Keywords: protein sources, colon, semi-synthetic diet, bacterial communities, metabolites, piglets

# INTRODUCTION

The weaning of piglets is often accompanied by low growth performance and a high risk of morbidity (mainly diarrhea rate) and death due to immature development of the gut and immune systems, which makes the piglets vulnerable to digestive disorders and highly susceptible to pathogenic bacteria (1–3). Indigestion causes excess undigested nutrients, mainly carbohydrates and protein, to flow into the hindgut where they are fermented by microbes. Carbohydrates fermentation is generally considered beneficial to the intestinal epithelium because it generates short chain fatty acids (SCFAs, acetate, propionate, and butyrate) (4, 5), but protein fermentation is considered to have adverse impacts on intestinal health because it can produce potentially toxic metabolites such as ammonia, biogenic amines, hydrogen sulfide, indols, and phenolic compounds, which are closely associated with the proliferation of pathogenic bacteria and diarrhea (1, 5, 6). Excessive protein fermentation not only wastes protein sources, but also induces some enteric diseases. The inclusion of high-quality of protein (milk by-products, animal proteins, processed proteins, etc.) (1, 2), the reduction of dietary protein intake or protein level (2, 3, 7), the addition of dietary fiber (8–10), and the use of various feed additives (1, 8) are commonly seen in swine production. These nutritional strategies can effectively reduce the outflow of undigested protein into the hindgut and decrease substrates for protein fermentation.

The colonic microbiota plays an important role in digestive physiology and makes a significant contribution to homeostasis (11, 12). Colonic bacteria can catabolize mostly amino acids via the abundant proteases and peptidases produced during fermentation (13). The ileal digestibility of protein and amino acids varies amongst protein source and determines the outflow of substrates into the hindgut for protein fermentation and specific biomarkers (2, 8). Therefore, identification of the biomarkers of various protein sources during fermentation is useful for the selection of suitable protein ingredients for piglets' diets and to provide a reference for dietary modulation to maintain piglets' gut health.

Although enzyme-treated soybean meal (14), cottonseed protein (15), dried porcine solubles (16, 17), and fish meal (18) are widely used in piglets' diets, few studies have evaluated the microbial mechanism of these protein sources on protein fermentation in the hindgut. Therefore, this study investigated the effects of P50, S50, CDCP, and FM on the growth performance, AID of CP, colonic mucin chemotypes, and the composition and metabolites of colonic microbiota in piglets. The results provide a reference for protein sources selection in piglet diets based on hindgut micro-ecosystem regulation.

#### MATERIALS AND METHODS

The Hunan Agricultural University Animal Ethics Committee (Changsha, China) reviewed and approved all experimental protocols. The four protein sources used in the study were Palbio 50 RD (P50, a dried porcine solubles; Bioiberica, S.A, Spain), Soyppt-50% (S50, a enzyme-treated soybean meal; Jiangsu Peptid Biological Co., Ltd, Jiangsu, China), Concentrated degossypolized cottonseed protein (CDCP, a novel processed cottonseed meal; Xinrui Biotech, Hunan, China), and fish meal (FM, steam dried fish meal; Tecnológica de Alimentos S.A., Peru).

#### Animal, Diets, and Experimental Design

Twenty-four barrows (Landrace × Yorkshire; initial BW = 12.61 ± 1.45 kg) were individually housed in stainless steel metabolism chambers (0.7 × 1.4 m) equipped with a nipple drinker, a feeder, and a fully slatted plastic floor. The piglets were randomly allotted to 1 of 4 diets with 6 piglets in each group (**Figure 1**). Four cornstarch-based diets were formulated with P50, S50, CDCP, or fish meal as the sole protein (**Table 1**). Piglets were fed the experimental diets for 9 days, and each piglet was fed a daily level of 5% of the average initial BW of 24 piglets. Feed was provided three times a day, with the daily allowances averaged over the three meals. All piglets had ad libitum access to water. On the day 10, piglets were humanly slaughtered to collect the ileal and colonic digesta, and colon tissue.

concentrated degossypolized cottonseed protein; FM, fish meal.


*<sup>a</sup>S50, Soyppt-50%, a enzyme-treated soybean meal; CDCP, concentrated degossypolized cottonseed protein; P50, Palbio 50 RD, a dried porcine solubles; FM, fish meal.*

*<sup>b</sup>Provided per kg of diet: 12, 000 IU Vitamin A, 2,400 IU Vitamin D3, 45 IU Vitamin E, 3.0 mg Vitamin K, 0.40 mg Vitamin B1, 6.4 mg Vitamin B2, 0.3 mg Vitamin B6, 36* µ*g Vitamin B12, 2 mg folic acid, 40 mg nicotinic acid, 20 mg D-pantothenic acid, 0.45 mg biotin, 120 mg Fe, 6 mg Cu, 40 mg Mn, 100 mg Zn, 0.40 mg I, 0.30 mg Se.*

*<sup>c</sup>Total starch was analyzed using a commercial starch assay kit (Megazyme, Ireland) according to the method 7613.01 of American Association of Cereal Chemists (19).*

# Sample Collection and Preparation

Body weight per piglet was measured at the beginning and conclusion of the 9-day experimental period to calculate the estimated daily gain (EDG). The fecal score for each piglet was recorded daily, according to the method introduced by Hu et al. (20) (hard feces = 1; slightly soft = 2; soft and partially formed feces = 3; loose, semi-liquid = 4; watery feces = 5). After the feeding trial, all piglets were anesthetized with serazine hydrochloride (Jilin Huamu Animal Health Product Co., Ltd., Changchun, China) at a dosage of 0.5 ml/kg BW by intramuscular injection and humanely euthanatized. The digesta in ileal segment 15 cm prior to ileocecal junction was collected and lyophilized for the determination of apparent ileal digestibility (AID) of crude protein (CP). The digesta in the colonic lumen was collected and stored at −80◦C for microbiology and metabolites analyses. The colonic tissue was fixed in Bouin's solution for Histomorphological analysis.

#### Measurement of AID of CP

The methods for determining CP and titanium (Ti) contents in diets and lyophilized ileal digesta were detailedly described by Li et al. (21). The AID of CP in diets was calculated as following: AID (%) = [1–(CP<sup>i</sup> × Tid)/(CPd/Tii)] × 100, where CP<sup>i</sup> and Ti<sup>i</sup>

are the contents of CP and TiO2, respectively, in the ileal digesta and CP<sup>d</sup> and Ti<sup>d</sup> represent the concentrations of CP and TiO2, respectively, in diets.

#### AB-PAS Staining of Mucins in Colonic Tissue

To characterize the neutral and acidic mucins in colonic goblet cells, an Alcian blue (pH 2.5)-periodic acid Schiff (AB-PAS) staining procedure, based on the method of Liu et al. (22) and Henwood (23) was conducted. Briefly, the colon tissue samples were fixed in Bouin's solution and then dehydrated and embedded in paraffin blocks. A 5µm section was sliced, deparaffinized, hydrated, and then stained with AB-PAS. The neutral and acidic mucins in goblet cells were stained in magenta and blue colors, respectively. The results were visualized by computer-assisted microscopy (DT2000, Nanjing dongtu digital technology co. LTD, China) and image-analysis software (Motic Images Plus 2.0, Dongguan, China).

#### Analysis of SCFA, BCFA, and Nitrogen Metabolites

Approximately 1.0 g of colon digesta was diluted with 10 mL of ultra-pure water, vortexed for 2 min, centrifuged at 10,000× g for 10 min and filtered (0.45µm) to collect the filtrate. A 2 mL portion of the filtrate was transferred to a 10-ml centrifuge tube, and 0.2 mL of 50% sulfuric acid and 2 mL of diethyl ether were added successively. The solution was vortexed for 2 min, centrifuged at 10,000× g for 5 min and placed at 4◦C for 30 min. The supernatant was analyzed for SCFA (acetic acid, propionate acid and butyric acid) and BCFA (isobutyric acid, isovaleric acid and valeric acid) using an Agilent 7890A gas chromatographer (Agilent Technologies Inc., Palo Alto, CA) equipped with an HP-FFAP elastic quartz capillary vessel column (30 m × 0.25 mm × 0.25µm). The carrier gas was high-purity nitrogen (99.999%) with a constant flow rate of 2.0 mL/min. The column temperature procedure was from 100◦C for 1 min then increasing by 5◦C/min to reach 150◦C for 5 min. The inlet and detector (FID) temperatures were 270◦C and 280◦C, respectively. The inlet volume was 2.0 µL, as measured by splitless injection.

The contents of ammonia nitrogen (NH3-N) in samples were determined using Nessler's reagent spectrophotometry as described by Chen et al. (24). Briefly, 0.5 g of colonic digesta was mixed with 5 mL of ammonia-free water and centrifuged at 5,000× g for 15 min at 4◦C to collect the supernatants. A 1-mL portion of supernatants was transferred to a 50-mL aseptic tube filled with 19 mL of ammonia-free water and 1 mL of potassium sodium tartrate, after which 1.5 mL of Nessler's reagent was added and the mixture was left to stand for 10 min. The OD value of the mixture was measured at 420 nm against ammonia-free water on a multi-scan spectrum microplate spectrophotometer (Thermo Fisher Scientific (China), Co., Ltd, Shanghai, China).

The concentrations of biogenic amines in the colonic digesta, mainly methylamine, cadaverine, putrescine, histamine, and spermidine, were analyzed with high- performance liquid chromatography (HPLC) according to the method introduced by Fan et al. (25). In short, 0.2 g of colonic digesta was weighed into a 2-mL tube, and 1 mL of trichloroacetic acid (TCA) was added and homogenized for 10 min. Afterward, the mixture was centrifuged at 3,600× g for 10 min at 4◦C to collect the supernatants. An equal volume of n-hexane was mixed with the supernatants and vortexed for 5 min, after which the organic phase was removed and the aqueous phase was re-extracted in the same way. The extracts were transferred into a 50 mL tube, and 20 mL of internal standard, 1.5 mL of saturated sodium bicarbonate (NaHCO3), 1 mL of dansyl chloride, and

FIGURE 3 | Alcian blue (pH 2.5)-periodic acid Schiff (AB-PAS) stained section (100×) in colon tissues. Neutral mucins are marked in magenta, while acidic mucins in blue. P50, Palbio 50 RD, a dried porcine solubles; S50, Soyppt-50%, a enzyme-treated soybean meal; CDCP, concentrated degossypolized cottonseed protein; FM, fish meal.

1 mL of sodium hydroxide (NaOH) were then added successively. The mixture was heated at 60◦C for 45 min with occasional mild shaking. Next, 100 µL of termination solution (i.e., ammonia) was added to the mixture, and the solution was maintained 40◦C in a water bath to vaporize the acetone by blow-drying with N2. Finally, the sample was extracted twice with 3 mL of diethyl ether, and the collected extracts were blow-dried with N2, and the residue was then re-dissolved in acetonitrile solution for HPLC analysis. An ammonium acetateacetonitrile gradient elution program was used on an Agilent 1200 equipped with a variable wavelength detector (VWD) and a reversed–phase ZORBAX 80 A Extend–C18 (4.6 mm × 250 mm; 5µm) column (Agilent Technologies, USA). The flow rate, temperature and wavelength were set at 1.0 mL/min, 30◦C and 260 nm, respectively.

#### DNA Extraction, PCR Amplification, Library Preparation, and Sequencing

Twenty colonic digesta samples from 4 dietary groups (5 per group) were randomly selected for 16S rRNA sequencing analysis. The total genomic DNA of each sample was extracted by using the TIANamp Stool DNA kit (Tiangen Biotech (Beijing) Co., Ltd, China) according to the manufacturer's instructions. The quantity and quality of extracted DNAs were measured using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific, USA) and agarose gel electrophoresis, respectively. Nineteen acceptable DNA samples (one low-quality DNA sample from the FM group was discarded) were sent to BGI (Wuhan, China) for 16S rRNA sequencing.

The V3-V4 hypervariable region of the bacterial 16S rRNA gene was amplified with the barcoded universal primers (341F-806R). All the PCR reactions were performed in 30-µL reaction volumes containing 1.0 µL of each primer, 1.0 µL of DNA template, 15 µL of Phusion <sup>R</sup> PCR Master Mix (New England Biolabs, USA), and 12 µL of sterile water. After the PCR amplification ended under the set cycle condition, agarose gel electrophoresis and a GeneJET Gel Extraction Kit (Thermo Fisher Scientific, USA) were used to determine, and extract and purify the PCR products. Sequencing libraries were prepared using an NEB Next <sup>R</sup> UltraTM DNA Library Prep Kit for Illumina (New England Biolabs, USA) according to the manufacturer's methods. Sequencing was conducted on the Illumina HiSeq platform by HiSeq2500 PE250 (Illumina, USA), and 250 bp paired-end reads were generated.

#### Bioinformatics Analysis

Paired-end reads from the sequencing data were screened and processed by FLASH (v1.2.11) (26, 27) and QIIME (v1.8.0) (28). Chimeric sequences produced by the PCR amplification were removed by UCHIME (v4.2.40) (29). The assembled tags were clustered as OTU via USEARCH (V7.0.1090) (30). Sequences with 97% similarity were assigned to the same

operational taxonomic units (OTUs), selected by UPARSE. Meanwhile, we picked a representative sequence for each OTUs and used the RDP (v2.2) classification to assign taxonomic data to each representative sequence, with a confidence threshold of 0.6. Taxonomy classifications were set against Greengene (V201305) (31) and RDP (Release9 201203) (32). An OTU table was further generated to record the abundance of each OTU in each sample, and a profiling histogram was made using R software (v3.1.1) to represent the relative abundance of taxonomic groups from phylum to species. A Venn diagram was generated to visualize the occurrence of shared and unique OTUs among groups<sup>1</sup> Alpha diversity indices based on the OTUs in all samples were analyzed with mothur (v1.31.2) according to the calculation formula of each index<sup>2</sup> and displayed with Prism software (v7.0). Beta diversity was evaluated using the Unifrac metric, and the UPGMA tree based on the unweighted UniFrac distances between the OTUs was displayed by QIIME (v1.80). Principal component analysis (PCA) was also conducted with R software (v3.1.1) based on the OTUs among samples from different groups. A heatmap was created using R software (v3.1.1) after performing the logarithmic transformation of relative abundance in each sample. LefSe analysis was conducted based on the relative abundance of annotated taxonomic composition to

<sup>1</sup>http://www.ehbio.com/ImageGP/index.php/Home/Index/VennDiagram.html

<sup>2</sup>http://www.mothur.org/wiki/Calculators

identify biomarkers for microbial communities from phylum to genus<sup>3</sup> .

#### Statistical Analysis

EDG, fecal score, and colonic SCFAs, ammonia nitrogen, biogenic amines and alpha diversity indexes were analyzed by One-Way ANOVA using the GLM procedure of SPSS 17.0 (SPSS Inc., Chicago, IL, USA), with individual piglet as an experimental unit. Unweighted Unifrac distances were compared based on approximately-maximum-likelihood to construce UPGMA trees. The Kruskal test was used for post hoc comparison of taxonomy. For all tests, P < 0.05 was considered as significant difference, while 0.05 < P <0.10 as a tendency. Correction between metabolites and microbiota in colonic digesta was done with Spearman's rank- order correlation test in R software (v3.1.1).

#### RESULTS

#### Growth Performance, Fecal Scores, and AID of CP

No differences in EDG were observed among the four dietary groups (P = 0.173) (**Figure 2**). However, the piglets in the P50 group elevated fecal scores compared with those in the S50 and CDCP groups (P = 0.039). Protein source significantly affected the AID of CP in diets (P < 0.01) (published in our previous paper). CDCP had the highest AID of CP and P50 had the lowest AID of CP. No differences in AID of CP were found between S50 and FM (P > 0.05).

# Characterization of Mucin Chemotypes

With AB-PAS staining, the amount of mixed neutral-acidic mucins in colon tissue of piglets fed S50 and CDCP diets were increased (P < 0.05) compared with those fed P50 (**Figure 3**).

#### Microbial Community Structure in Colonic Digesta

The Venn analysis identified 13, 74, 39, and 31 unique OTUs in the P50, CDCP, FM, and S50 dietary groups, respectively (**Figure 4**). Dietary protein sources did not affect the alpha diversity indices (**Figure 5**). Firmicutes, Bacteroidetes, Proteobacteria, and Spirochaetes were the four most abundant phyla (**Figure 6A**). The Firmicutes/Bacteroidetes ratios in CDCP, S50, P50, and FM were 1.54, 1.03, 0.95, and 0.45, respectively. FM had a lower abundance of Firmicutes (P = 0.025), but a greater abundance of Spirochaetes (P = 0.040) and TM7 (P = 0.007) than the other three groups, S50 and CDCP had increased Tenericutes abundance compared with P50 and FM (P = 0.012) (**Supplementary Table S2**). Down to the genus level, P50 and FM had high abundances of Escherichia (P = 0.029) and Clostridium (P = 0.041), whereas, CDCP had abundances of Lactobacillus (P = 0.002) and Megasphaera (P = 0.033) and S50 had a high abundances of Gemmiger (P = 0.018) (**Figure 6B**).

The heatmap results showed that Phascolarctobacterium, Prevotella, and Roseburia were the most abundant genera within the four dietary groups (**Figure 7**). CDCP had greater Lactobacillus abundance than the other three dietary groups, whilst P50 and FM had increased Escherichia abundance compared with CDCP and S50.

PCA, a multivariate analysis, was used to compare the overall composition of the microbiota in the colonic samples from various dietary groups (**Figure 8A**). The first two components

<sup>3</sup>http://huttenhower.sph.harvard.edu/galaxy/

FIGURE 7 | Relative abundance of each taxonomic genus in different dietary treatments. The rows show the top 50 genera in each sample and the color scale illustrates the comparison of each genera among the samples. P50, Palbio 50 RD, a dried porcine solubles; S50, Soyppt-50%, a enzyme-treated soybean meal; CDCP, concentrated degossypolized cottonseed protein; FM, fish meal.

accounted for 52% of the variation. PC1 mainly represented the diet-induced variations, whilst PC2 explained intra-group variations in microbial structure. Greater variations were seen in P50 and FM than in CDCP and S50, whilst P50, S50, and CDCP had similar microbial communities. The UPGMA tree analysis (**Figure 8B**) revealed that remarked differences in the structure of colonic microbiota among various dietary groups, suggesting that the protein sources had a notable impact on colonic microbial communities.

LEfSe analysis showed 55 different OTUs among the 4 dietary protein groups, and the abundance of 5 OTUs was low in S50, with 12 in P50, 15 in FM, and 23 in CDCP (**Figure 9A**). There was a high abundance of Lactobacillus in CDCP, Bacteroides in FM, Clostridium in P50, and Gemmiger in S50, respectively (**Figure 9B**).

#### Fermentation Metabolites of Colonic Digesta

CDCP showed a trend of elevated valeric acid (P = 0.052) and BCFA (P = 0.064) concentrations (**Table 2**). P50 and FM had greater ammonia nitrogen and methylamine contents than S50 and CDCP (P = 0.041) (**Table 3**).

degossypolized cottonseed protein; FM, fish meal.

#### Correction Between Metabolites and Microbiota in Colonic Digesta

A spearman's rank correlation analysis was done to infer taxa associated with metabolic potentials (**Figure 10**). We found that the increased Escherichia abundance in FM and P50 group was positively correlated with ammonia nitrogen (P < 0.05) and methylamine (P < 0.05), but negatively corrected with total SCFAs (P < 0.05). Lactobacillus, contained an increased abundance in CDCP group, was positively correlated with acetic acid (P < 0.01), propionate acid (P < 0.01), butyric acid (P < 0.01), SCFA (P < 0.05), total SCFAs (P < 0.01), histamine (P < 0.01), and spermidine (P < 0.01), but negatively corrected with ammonia nitrogen (P < 0.01), methylamine (P < 0.01), cadaverine (P < 0.01), and putrescine (P < 0.05). Gemmiger, had an increased abundance in S50 group, was negatively corrected with isovaleric acid (P < 0.01), valeric acid (P < 0.01), and BCFA (P < 0.05).

#### DISCUSSION

Dietary protein is mainly digested into peptides and amino acids in the foregut via the action of proteases (13, 25). The digestion and absorption of protein is closely associated with the increased lean muscle mass (33). The weight gain of piglets is primary lean muscle mass due to their rapid growth and quick protein turnover (34, 35). In our study, equal amounts of feed with similar DE and CP were fed to each piglet daily to ensure the similar dietary protein and energy intakes. This helps to explain, along with the short experimental period (9 days), why the EDG of the piglets showed no difference among the dietary groups. However, the P50 group increased the diarrhea rate, perhaps because of the high salt content and poor carriers added to codried P50, which is the porcine mucosa hydrolysate after heparin extraction and is usually sprayed into a soybean meal carrier (16, 17, 36). Additionally, the reduced AID of CP in the P50 group in our previous study (21) could also explain the increased diarrhea rate, because more undigested protein flow into the hindgut for protein fermentation.

The porcine gastrointestinal tract harbors highly diverse and dynamic microbial communities that play a vital role in maintenance of intestinal health, which contributes to nutrient utilization and modulation of the immune system (5, 37). Diet composition, especially carbohydrates and protein, reflects the substrates available for the intestinal microbiota, which affects their composition and metabolic activity (38). Researchers and the feed industry have paid increasing attention to dietary protein manipulation of the intestinal microbiota in piglets (7, 15, 39, 40). Our experiment determined the colonic microbiome of piglets fed diets containing either P50, S50, CDCP, or FM as sole protein source via 16S rRNA high-throughput sequencing. On average, 164,709 effective tags were achieved for each sample with a high Good's coverage (>99.90%) (**Supplementary Table S1**). Meanwhile, protein sources did not affect the alpha diversity indices, suggesting that the ecological diversity of the colonic microbiota was quite similar among the samples. The alpha diversity reflects the species diversity in a single sample, including species richness indices (observed species, chao, and ace), species diversity (Shannon, Simpson, and Good's coverage) (31, 41). Our findings concurred with those of Cao et al. (15), who found no change in alpha diversity of piglets fed diets containing soybean meal, cottonseed meal, SBM and CSM or fish meal. However, Venn analysis identified 13, 74, 39, and 31 unique OTUs in the P50, CDCP, FM, and S50 groups, respectively, indicating that specific bacteria appeared in piglets fed different diets.

At the phylum level, Firmicutes, Bacteroidetes, Proteobacteria, and Spirochaetes were the major phyla in colonic digesta (**Figure 6A**), similar to the findings of a study by Cao et al. (15). The value for Firmicutes/Bacteroidetes ratio in CDCP, S50, P50, and FM was gradually reduced (**Figure 7**). This index is positively correlated with obesity (42), and interestingly, no differences in

Frontiers in Nutrition | www.frontiersin.org

indicated no difference among groups, but the other color nodes suggested the significant difference, the letters identified the taxon name with notable difference among groups. P50, Palbio 50 RD, a dried porcine solubles; S50, Soyppt-50%, a enzyme-treated soybean meal; CDCP, concentrated degossypolized cottonseed protein; FM, fish meal.

#### TABLE 2 | Colonic short chain fatty acid (SCFAs) (mg/g).


*<sup>a</sup>*,*bmeans with different superscripts in the same row differ significantly (P* < *0.05).*

TABLE 3 | Colonic nitrogen metabolites.


*<sup>a</sup>*,*bmeans with different superscripts in the same row differ significantly (P* < *0.05).*

EDG were found among the four groups (**Figure 2A**). Down to the genus level, there were high abundances of Escherichia and Clostridium in P50 and FM, of Lactobacillus and Megasphaera in CDCP, and of Gemmiger in S50 (**Figure 6B**). These observations were also confirmed by heatmap (**Figure 7**) and LEfSe analysis (**Figure 9**), which can identify unique biomarkers for high dimensional gut bacteria composition (43). Prevotella is a dominant genus in all dietary groups (**Figure 7**), and the bacteria in this genus mainly used carbohydrates as substrates (44); thus, our finding might be due to the similar total starch content among diets (**Table 1**). Bacteroides, Coliform bacteria, and Clostridium are closely related to protein fermentation and often increase the risk of diarrhea (2). The high abundance of Escherichia and Clostridium in P50 and FM can explain the increased fecal scores. The finding was similar to that of Heo et al. (45), who reported that the occurrence of pathogenic diarrhea in piglets was mainly due to the release of endotoxin by Escherichia coli, which adhered to gut epithelial cells. PCA and UPGMA tree revealed greater variations were seen in P50 and FM than in CDCP and S50, suggesting that the intestinal microbiota responds differently to animal protein.

SCFA (acetate, propionate, and butyrate) and BCFA (isobutyrate, isovalerate, and valerate) are generated from fermentation of carbohydrates and protein in the colon (1). Both carbohydrates and protein fermentation produce SCFA, whilst BCFA are coming from deamination of valine, leucine, and isoleucine (25). In piglets, considerable amounts of carbohydrates are fermented in the proximal colon, but protein fermentation takes place more distally, especially when there is a shortage of available fermentable carbohydrates (1). Our results (**Table 2**) showed that dietary protein did not affect SCFAs profiles, but the CDCP group showed a tendency of increased valeric acid and BCFA contents. Some reports have indicated that BCFA play a role in maintaining intestinal cell integrity (46, 47) and suggested that CDCP might ameliorate the colon health. This was consistent with the study by Cao et al. (15), who found a beneficial effect of cottonseed meal on the gut of piglets. However, Rist et al. (1) suggested that BCFA might be an indicators of the extent of protein fermentation and exerted detrimental effects upon the host.

The decarboxylation and deamination of amino acids generate bioamines and ammonia, respectively. Ammonia, amines, phenols, and indoles are considered as potentially toxic products, which may interfere with epithelial cell turnover and inhibit pig growth (13, 24). In this study, S50 and CDCP had lower ammonia nitrogen and methylamine contents than P50 and FM (**Table 3**), possibly because plant protein (S50 and CDCP) contains dietary fiber, which help to reduce protein fermentation. Many researchers have suggested that the addition of dietary fiber to piglets' diets could reduced protein fermentation and diarrhea rate (6, 8, 9, 37, 48). The finding that CDCP and S50 decreased Escherichia abundance compared with P50 and FM (**Figure 7**) could also explain the reduced nitrogen metabolites. Additionally, the increased Escherichia abundance

in FM and P50 was positively correlated with ammonia nitrogen and methylamine, but negatively corrected with total SCFAs (**Figure 10**), which could also provide an evidence for the beneficial effect of CDCP and S50 to colon.

Mucin layer spreads over the surface of mucosa and safeguards the intestine and its structural integrity (22). Mucins is broadly categorized into neutral and acidic chemotypes. In our study, we distinguished and detected the two types of mucins and we found that CDCP and S50 increased the amount of mixed neutral-acidic mucins in colon, which might ascribe the reduced protein fermentation in colon of piglets fed plant protein sources (**Figure 3**).

# CONCLUSION

Protein intake from 4 different diets (P50, S50, CDCP, and FM) had similar effects on piglet growth, but P50 increased diarrhea rate during short-term feeding. CDCP and S50 reduced the protein flows into hindgut and increased the amount of mixed neutral-acidic mucins in colon. Potential pathogenic bacteria (Escherichia) and detrimental metabolites (ammonia) were found in the colons of piglets fed P50 and FM, whilst beneficial effects (Lactobacillus and BCFA) were seen in pigs fed CDCP. These findings indicate that the addition of available plant protein such as CDCP to the diet of piglets enhances colon health by reducing protein fermentation.

# DATA AVAILABILITY

The datasets for this manuscript are not publicly available because we have not uploaded the datasets to the NCBI due to some secret data, temporarily. Requests to access the datasets should be directed to lirui181000@163.com.

#### ETHICS STATEMENT

The Hunan Agricultural University Animal Ethics Committee (Changsha, China) reviewed and approved all experimental protocols.

# AUTHOR CONTRIBUTIONS

RL, XH, and D-XH design the experiment and revised the manuscript. RL, ZS, and GH conducted the experiments. RL, ZS, ZF, and XH offered the experimental reagents and materials. RL analyzed the data and finished the manuscript. RL, LC, GH, ZS, and D-XH prepared the figures and edited the manuscript. All authors reviewed the manuscript.

#### FUNDING

This research was supported by the National key R & D Program of China (2017YFD0500506, 2016YFD0501200), the Double first-class construction project of Hunan Agricultural University (SYL201802015, SYL201802009) and the Hunan postgraduate research and innovation project (CX2018B399).

# REFERENCES


# ACKNOWLEDGMENTS

We would like to thank the workers in BGI (Wuhan, China) for their experiment on 16s rDNA high-throughput sequencing, and thank bioinformatics trainers in Personalbio (Shanghai, China) for their help in data preparation and analysis. The authors are also grateful to the staff at Department of Animal Nutrition and Feed Science of Hunan Agricultural University for their assistance in conducting the experiment.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnut.2019. 00151/full#supplementary-material


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Li, Chang, Hou, Song, Fan, He and Hou. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Regulation of Ruminal Short-Chain Fatty Acids on the Functions of Rumen Barriers

Hong Shen1,2† , Zhihui Xu1,2† , Zanming Shen<sup>3</sup> and Zhongyan Lu<sup>3</sup> \*

<sup>1</sup> College of Life Sciences, Nanjing Agricultural University, Nanjing, China, <sup>2</sup> Bioinformatics Center, Nanjing Agricultural University, Nanjing, China, <sup>3</sup> Key Lab of Animal Physiology and Biochemistry, College of Veterinary Medicine, Nanjing Agricultural University, Nanjing, China

#### Edited by:

Stephen J. Pandol, Cedars-Sinai Medical Center, United States

#### Reviewed by:

Shiyu Tao, China Agricultural University (CAU), China Hongbing Gui, Jiangsu Academy of Agricultural Sciences, China Cong-jun Li, Agricultural Research Service (USDA), United States Dorothee Günzel, Charité Medical University of Berlin, Germany

\*Correspondence:

Zhongyan Lu luzhongyan@njau.edu.cn †These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology

Received: 07 January 2019 Accepted: 30 September 2019 Published: 25 October 2019

#### Citation:

Shen H, Xu Z, Shen Z and Lu Z (2019) The Regulation of Ruminal Short-Chain Fatty Acids on the Functions of Rumen Barriers. Front. Physiol. 10:1305. doi: 10.3389/fphys.2019.01305 The rumen barriers, constituted by the microbial, physical and immune barrier, prevent the transmission of pathogens and toxins to the host tissue in the maintenance of host-microbe homeostasis. Ruminal short-chain fatty acids (SCFAs), which are the important signaling molecules derived from the rumen microbiota, regulate a variety of physiological functions of the rumen. So far, how the ruminal SCFAs regulate the function of rumen barriers is unclear. By the combined methods of transcriptome sequencing, 16S rRNA gene sequencing, and metagenome shotgun sequencing, we have investigated the regulatory effects of ruminal SCFAs on the functions of rumen barriers, by determining the composition and functions of epimural microbiota and on the structure and immunity of the rumen epithelium in goats receiving a 10% (LC group), 35% (MC group), or 65% concentrate diet (HC group). We found that, when the dietary concentrate shifted from 10 to 35%, the increase of total SCFA is associated with the diversification of epimural microbiota and the diversity of its gene pool. Within the microbial community, the relative abundance of genera Sphingobium, Acinetobacter, and Streptococcus increase mostly. Meantime, the signals on pathways concerning the mechanical connections and growth homeostasis in the rumen epithelium were upregulated. Under these conditions, the responses of immune components in the rumen epithelium decrease. However, when the dietary concentrate shifted from 35 to 65%, the increase of acetate and reduction of pH decrease the diversity of epimural microbiota and the diversity of its gene pool. Within the microbial community, the relative abundance of genera Sphingobium, Acinetobacter, and Streptococcus significantly decrease. Concomitantly, the signals on pathways concerning the cell growth and tight junction disruption were upregulated, while the signals on pathways concerning paracellular permeability were downregulated. Under these conditions, the signals on the pathways relating to the immune components increase. Our data thus indicates that diet-SCFA axis maintains the host-microbe homeostasis via promoting the diversification of epimural microbiota and maintaining the integrity of rumen epithelium in healthy animals, while via enhancing the activities of immune barrier in animal with lower rumen pH.

Keywords: epimural microbiota, rumen barrier, short-chain fatty acids, epithelium physiology, microbe-host interactions

# INTRODUCTION

fphys-10-01305 October 23, 2019 Time: 17:46 # 2

The rumen is the most important site for digestion in ruminant animals. On the one hand, it provides the space and nutrients for microbes to live within the rumen. On the other hand, the ruminal microbes ferment plant materials into short-chain fatty acids (SCFAs) that regulate a variety of physiological functions of the rumen (Li et al., 2016). During long-term evolution, various strategies have been developed by the animals and microbes to control their relationships. The central strategy utilized by them to maintain such homeostatic relationships is to construct barriers and, therefore, to protect the ecological niche of the commensals, limit the colonization of pathogens, and clearance the invaded microbes in the intestinal epithelium (Belkaid and Hand, 2014).

The barriers of rumen are constituted by three parts: (1) the microbial barrier, which is composed primarily of the microbes attaching to the surface of the stratified squamous epithelium (termed the epimural microbes). The commensals within epimural microbiome could inhibit the colonization of pathogens in the epithelium via the competition of shared nutritions, secretion of antimicrobial components (e.g., bacteriocins, microcins, and colicins), and alteration of environmental conditions required for the growth of pathogens (Ohland and Jobin, 2015). For example, Bifidobacterium inhibited the colonization of pathogenic Escherichia coli by decreasing the acetate concentration (Fukuda et al., 2011). Moreover, the commensals could inhibit pathogen virulence by suppressing the expression of the virulence gene. For example, Bacteroides thetaiotaomicron modulates the expression of the virulence factor ler in pathogenic E. coli by generating the fucose, a metabolite of host mucin (Pacheco et al., 2012); (2) the physical barrier, which is composed of the epithelial cells and the mechanical connections between the epithelial cells. The integrity of physical barrier is fundamental to the animal health and productivity since it prevents the translocation of toxins and pathogens from the rumen into the blood; (3) the immune barrier, which is composed of the intestinal-associated immune cells and their secretion of cytokines. Under the physiological conditions, the immune cells, especially the innate immune cells, play the key roles in the maintenance of intestinal homeostasis by preventing inappropriate adaptive immune responses. For example, the activation of the Toll-like receptor (TLR) 10 signaling enhanced the tolerance of rumen epithelium to the rumen fluid microbiota by suppressing the expressions of pro-inflammation cytokines (Shen et al., 2016). The activation of the TLR2 signaling in intestinal epithelium enhanced its barrier function via promoting the expression of zonula occludens-1 (ZO-1) (Cario et al., 2004). So far, many factors, ranging from the environmental factors of the microbial ecological niche to the physiological conditions of the host, have been reported to impact the functions of these barriers. Among them, SCFAs (mostly butyrate, acetate, and propionate) have received the greatest attentions. In the rumen, SCFAs have been reported to influence the function of the physical barrier, such as the integrity of rumen epithelium, renewal of epithelial cells and the expression of tight junction proteins (Gui and Shen, 2016; Greco et al., 2018). In the colon, SCFAs have been reported to promote the proliferation and renewal of epithelial cells (Guilloteau et al., 2010); to influence the size, and function of regulatory T cells (Tregs) via binding to the G-protein-coupled receptors (GPRs) and histone deacetylase (HDACs) expressed in the epithelium (Puertollano et al., 2014); to regulate the epithelium motility and permeability via the hormono–neuroimmune system (Perry et al., 2016). In addition, in the human intestine, SCFAs have been reported to suppress the expression of virulence genes in the opportunistic pathogens (Ohland and Jobin, 2015). However, the various parts of rumen barriers do not work separately. On the contrary, changes in environmental factors have global effects on the function of these barriers. For example, any excess amount of SCFAs generated in the rumen leads to reconstruct the epimural microbiota and the microbial barrier, damage the epithelial structure and the physical barrier, and concomitantly, express inflammatory cytokines in the rumen epithelium (Liu et al., 2013; Wetzels et al., 2015) and impair the immune barrier. Hence, the study of the global effects of ruminal SCFAs on the function of rumen barriers is valuable in order to obtain a comprehensive view of the interactions between ruminal SCFAs and rumen barriers and, furthermore, will provide potential regulation methods for animal health and growth.

The ratio of dietary concentrate is well known to influence the concentrations of ruminal SCFAs. In the current study, we have collected the ruminal epithelium and epimural microbiota from the rumen of goats receiving diet with three ratios of concentrate diet, i.e., 10% (LC), 35% (MC), and 65% (HC). Previous studies showed that MC diet promotes the performance in structural integrity, nutrient absorption and cell refreshing of rumen epithelium better than that of the LC and HC group (Yan et al., 2014; Gui and Shen, 2016), and that both of LC and HC diets had negative effects on the integrity of the rumen epithelium, indicating the strongest barrier function of MC group among three groups (Hu et al., 2018). Accordingly, in order to know the effects of the increase of ruminal SCFAs on the barrier function of rumen epithelium in the present study, we compared the barrier functions between the LC and MC group by using LC group as the control, and compared the barrier functions between the MC and HC group by using MC group as the control. Via the simultaneous measurement of the responses from three parts of the rumen barrier, we hope to gain better insights into the effects of ruminal SCFAs on the function of rumen barriers.

#### MATERIALS AND METHODS

#### Ethics Statement

This study was carried out in accordance with the recommendation of the Regulations for the Administration of Affairs Concerning Experimental Animals (No. 588 Document of the State Council of P.R. China, 2011). This protocol was approved by Nanjing Agricultural University.

#### Experiment Design

Fifteen male goats (Boer × Yangtze River Delta White, aged 4 months) were randomly allocated into three groups and

received a diet of 35% hay plus 65% concentrate (HC group, n = 5), a diet of 65% hay plus 35% concentrate (MC group, n = 5), and a diet of 90% hay plus 10% concentrate (LC group, n = 5) (**Table 1**). All goats were fed with two equal portions of the designated diet at 0800 and 1700 daily for 28 days. Water was freely available to all goats during the experimental period. On day 29, the goats were killed at a local slaughterhouse.

#### Sample Collection and Determination of SCFA Concentrations, pH and Osmolarity in the Rumen

On day 29, all goats were slaughtered at 6 h after receiving the morning feed. Ruminal content (30 mL) were strained through a 4-layer cheesecloth and immediately subjected to pH measurement by using a pH meter (Mettler-Toledo Delta 320, Halstead, United Kingdom). A 5% HgCl<sup>2</sup> solution was added to the fluid samples, which were subsequently stored at −20◦C for the determination of SCFA concentration and osmolarity. Rumen tissue from the ventral blind sac was quickly excised and washed in ice-cold phosphate-buffered saline (PBS; pH 7.4). The epithelium was subsequently separated from the muscle layers and cut into 1–2 cm<sup>2</sup> pieces. For each animal, five pieces of rumen epithelium were immediately fixed in 4% paraformaldehyde (PFA) (Sigma, St. Louis, MO, 123 United States) for morphometric analyses. Ten pieces were


Concentrate = corn + soya bean meal. DMI, dry matter intake; ME, metabolizable energy; NDF, neutral detergent fiber; NFC, non-fibrous carbohydrate; DM, dry matter. <sup>a</sup>The values are means ± SE. <sup>b</sup>NFC = 100 – (NDF + CP + crude fat + ash). <sup>c</sup>The additive was composed of calcium phosphate, limestone, trace mineral salt, and vitamin premix (vitamins A, D, and E). <sup>d</sup>ME = total digestible nutrient × 0.04409 × 1.01–0.45 (NRC, 2000).

stored at −20◦C for the later extraction of microbial DNA. Ten pieces were stored at −80◦C for the later extraction of epithelial RNA.

Ruminal SCFAs concentrations were measured by using a gas chromatograph (HP6890N, Agilent Technologies, Wilmington, DE) as described by Yang et al. (2012). 10 mL of rumen fluid was centrifuged at 18,000 g for 20 min at 4◦C (Eppendorf Centrifuge 5424 R, Eppendorf AG, Hamburg, Germany). The supernatant was collected and the osmolarity was measured by using an osmometer (Osmomat 030-D, GONOTEC Berlin, Germany).

# Morphological Analysis of Rumen Epithelium

The density, length and width of ruminal papillae were measured according to the description of Malhi et al. (2013). In brief, 1 cm<sup>2</sup> of PFA fixed ruminal epithelium was used to count for the papillae density (number/cm<sup>2</sup> ). Fifteen papillae in each of the PFA fixed epithelial sample were used to measure papillae length and width by using a sliding caliper.

# Microbial DNA Extraction and 16S rRNA Gene Sequencing

To detach the tightly attached microbes, twenty pieces of ruminal epithelium were placed in a 15 ml tube with 7 ml PBS and several plastic beads and moderately shaken on a vortex for 30 s. The ruminal epithelium was moved to a new tube and processed with the detaching step for two more times. Subsequently, the metagenomic DNA was extracted from the PBS mixture by using a Bacterial DNA Kit (Omega, Shanghai, China). The DNA concentration was determined in a NanoDrop 1000 (Thermo Fisher Scientific, Wilmington, DE, United States) and stored at −20◦C until further processing.

The amplicon library was prepared by polymerase chain reaction (PCR) amplification of the V4 region of the 16S rRNA gene. The universal primers 515F and 806R, including TruSeq adapter sequences and indices, were employed in the PCRs. All libraries were sequenced by using an Illumina HiSeq2500 platform (Illumina, San Diego, CA, United States) at Biomarker Technologies, Beijing, China.

# Composition Analysis of Epimural Microbes by Using 16S rRNA Gene Sequencing Data

Paired reads were filtered for quality (Q30) and joined by using FLASH version 1.2.11 (Magoc and Salzberg, 2011). Sequences that contained read lengths shorter than 250 bp were removed by means of PRINSEQ v0.20.4 (Schmieder and Edwards, 2011). The remaining sequences were then classified into operational taxonomic units (OTUs) by using QIIME 1.9.0 (Caporaso et al., 2010) at a 97% similarity threshold. OTUs whose counts were more than 3 in at least one of the samples were hierarchically summed at all taxonomic levels, and the counts were normalized to the relative abundance for each sample. The diversity of the microbial communities was estimated by using the R program phyloseq package (McMurdie and Holmes, 2013).

# Covariation of SCFA Concentration and Microbial Abundance

The relationships between the abundance of OTUs and the concentrations of SCFAs were explored by canonical correspondence analysis (CCA) in the R program vegan package (Oksanen et al., 2016). Subsequently, the R program ggplot2 package (Wickham, 2009) was used to generate the visual interpretation (biplot) of the gene-microbe relationships.

#### Metagenome Shotgun Sequencing and Function Comparison of Epimural Microbes

The integrity of microbial DNA was evaluated on 1.0% agarose gel. Metagenomic DNA libraries were constructed by using the TruSeq DNA Sample Prep kit (Illumina, San Diego, CA, United States). Libraries were sequenced via paired-end chemistry (PE150) on an Illumina Higseq X Ten platform (Illumina, San Diego, CA, United States) at Biomarker Technologies, Beijing, China.

Raw reads were first filtered by using FastX v0.0.13 (Gordon and Hannon, 2010), with a quality cutoff of 20, and reads shorter than 30 bp being discarded. The high-quality reads that were likely to originate from the host were removed by using DeconSeq v0.4.3 (Schmieder and Edwards, 2011), with the NCBI goat genome sequence as the reference. The remaining reads were subsequently assembled into scaftigs by using IDBA-UD (Peng et al., 2012) with the standard parameters. Genes were predicted from the scaftigs by means of FragGeneScan (Rho et al., 2010). The predicted genes of each sample were then annotated to Kyoto Encyclopedia of Genes and Genomes (KEGG) ontology (KO) databases via the KEGG Automatic Annotation Server (KAAS) (Moriya et al., 2007).

Predicted genes from all samples were gathered together to form a large gene set. BLAT v35 was used to construct the non-redundant gene set. Any two genes with more than 95% identity and more than 90% coverage of the shorter gene were picked out, and subsequently, the shorter one was removed from the large gene set. High-quality reads of each sample were mapped to the non-redundant gene set by using Bowtie2 v2.3.4 (Langmead and Salzberg, 2012) with default parameters. MarkDuplicates in the Picard toolkits version 2.0.1 was used to remove the duplicates in the reads, and then genomeCoverageBed in BEDTools 2.26.0 was used to calculate the gene coverage. The RPKM of the gene, calculated by [gene coverage × 10<sup>6</sup> /(total mapped reads × gene length)], was used to normalize the gene abundance between the treatments. The abundances of KOs were received by summing up the abundance of genes with the same KO number, and then, compared by using R program DeSeq2 package (Love et al., 2014). Since the multiple comparisons (LC vs. MC and MC vs. HC) were proceed in this study, the Bonferroni-Holm (BH) correction has been done for the p.adjust values received from DeSeq2 comparison (the corrected values were termed as p.correct in this study). Differences were considered significant when p.correct < 0.05 and the differences of KO abundance were more than two times between the groups.

In this study, the abundance of the KEGG pathway was transformed from the detected kinds of differentially abundant genes, which were located in the corresponding KEGG pathway, within the sample. Subsequently, the abundance of the microbial KEGG pathway was compared and visualized by using R program heatmap3 v1.1.4 (Zhao et al., 2014) with the complete clustering method.

# Epithelial RNA Extraction and Transcriptome Sequencing

Total RNA was extracted from the ruminal epithelium by using the RNAeasy Mini Kit (Qiagen, Shanghai, China). RNA was quantified by using the NanoDrop 1000 spectrophotometer, and its integrity was evaluated by means of the RNA 6000 Assay Kit of the Agilent Bioanalyzer 2100 system (Agilent Technologies, CA, United States). High-quality RNA (RNA integrity number > 9.0) was processed by applying the NEB Next Ultra RNA Library Prep Kit (New England Biolabs, Beijing, China). All libraries were sequenced via paired-end chemistry (PE125) on an Illumina HiSeq2500 platform (Illumina, San Diego, CA, United States) at Biomarker Technologies, Beijing, China.

#### Transcriptome Assembling, Differentially Expressed Gene Identification, and KEGG Enrichment Analysis

Low-quality reads were first removed by using PRINSEQ v0.20.4. The NCBI goat genome annotation release version 101 was applied to construct the reference genome by means of Bowtie v1.2.0 (Langmead et al., 2009). High-quality reads were mapped to the reference genome by using TopHat v2.1.0 (Kim et al., 2013) with standard parameters. for each sample, the gene expression level was estimated and normalized to the reads per kilobase of exon model per million mapped reads (RPKM) by means of Cufflinks v2.2.1 (Trapnell et al., 2010). In this study, only genes with more than 1 RPKM in at least one group of the samples were considered to be expressed. Similar with the metagenomic analysis, the DeSeq2 package was utilized to detect the differentially expressed genes between the groups. After the BH correction, the differences were considered significant when p.correct < 0.05 and the differences of genes expression were more than two times between the groups. The R program clusterProfiler package (Yu et al., 2012) was utilized to perform the KEGG enrichment analysis for the differentially expressed genes between the groups. The sets of upregulated genes and downregulated genes were used in the enrichment analysis, separately. Finally, the enrichment results were visualized by using the R program ggplot2 package.

#### SCFAs Regulatory Network Construction

The highly correlated genes of SCFAs were identified by computing the Spearman correlation coefficient (SCC) between SCFA concentration and gene expression across the groups in the R program. Only expressed genes were included in the correlation analysis. A threshold for the SCC value larger than 0.8 and a p-value less than 0.01 was employed to

identify significantly related genes. Accordingly, two SCFAs regulatory networks were constructed based on the SCCs of genes and SCFAs between the LC group and MC group (referred to as the MC-LC SCFA regulatory network) and the SCCs of those between the MC group and HC group (referred to as the HC-MC SCFA regulatory network). Subsequently, the genes that were highly correlated to the SCFAs were picked and annotated against the KEGG databases. Finally, the correlation networks were visualized by means of cytoscape 3.4.0 (Shannon et al., 2003).

#### Reverse Transcription Quantitative PCR (RT-QPCR) Verification of Target Genes

An aliquot of 2000 µg RNA, random hexamer primers (Invitrogen, Shanghai, China) and moloney murine leukemia virus (M-MLV) reverse transcriptase (Fermentas, Burlington, ON, Canada) were employed to synthesize the cDNA. RT-QPCR was performed by using the StepOne Plus real-time PCR system (Applied Biosystems, Den Ijssel, Netherlands) and SYBR-Green (Roche, Shanghai, China) for detection. The glyceraldehyde-3-phosphate dehydrogenase (GADPH) was chosen as the stably expressed reference gene. The primers of these genes were designed in this study by using Primer 5 and the available mRNA sequences in NCBI (**Supplementary Table S1**). Amplification efficiencies of the primers were determined by means of a dilution series of epithelial cDNA. All samples were run in triplicate, and the data were analyzed according to the 2−11CT method (Livak and Schmittgen, 2001). The identity and purity of the amplified products were checked by analysis of the melting curves obtained at the end of the amplification. Linear regression analysis was applied to identify the relationships between the RT-QPCR results and RNAseq results.

#### Statistical Analysis of SCFAs Concentration, pH, Osmolarity, Morphological Traits, and RT-QPCR Results

The BH corrected two-tailed t-test was used to analyze the difference of SCFAs concentration, pH, osmolarity, morphological trait and RT-QPCR results between the groups (**Tables 2**, **3**). The differences were considered significant when p.correct < 0.05. The Pearson correlation coefficient (PCC) was calculated to detect the relevance between the total SCFA concentration and ruminal osmolarity. Relevance was considered significant when PCC > 0.8 and p.correct < 0.05.

#### RESULTS

#### Morphological Analysis of the Rumen Epithelium

Compared with the LC group, the length and density of the ruminal papillae were increased in the MC group. On the contrary, the length was decreased in the HC group, compared with the MC group (**Table 2**).

# Comparisons of Ruminal SCFA Concentrations, pH, and Osmolarity Across the Groups

Compared with the LC group, the concentrations of acetate, propionate, butyrate and total SCFA were increased by 9, 16, 47, and 13% in the MC group, respectively (**Table 3**). Meantime, the concentrations of propionate, butyrate and total SCFA showed the significant changes between the LC and MC groups. Compared with the MC group, the concentrations of acetate, propionate, butyrate and total SCFA were increased by 8, 15, 37, and 12% in the HC group, respectively. Meantime, the concentrations of propionate, butyrate and total SCFA showed the significant changes between the MC and HC groups. The ruminal pH was consistently decreased with increases of dietary concentrate. The ruminal pH was significantly reduced, whereas the osmolarity was significantly raised with the increase of dietary concentrate. According to the Pearson correlation analysis, the ruminal osmolarity was significantly correlated with the total SCFA concentration (PCC = 0.865, p = 0.01).

# Composition and Diversity of Epimural Microbes

At the phylum level, a total of 20 prokaryotic phyla were identified at a 97% similarity, and 17 of them were common to all groups (**Figure 1A**). Firmicutes (46.5–50.2%), Bacteroidetes (28.1–34.5.0%), and Proteobacteria (7.3–10.1%) were most abundant among all microbial communities. Compared with the LC group, Verrucobacteria showed the most significant increase (increased by 4.3 times), whereas Actinobacteria showed the most significant reduction (decreased by 33%), in the MC group. Compared with the MC group, Actinobacteria showed the most significant increase (increased by 1.5 times), whereas Proteobacteria showed the most significant reduction (decreased by 27%), in the HC group. At the genus level, in total, 156 genera were detected in the sequences. Among them, 108 genera were common to all groups (**Figure 1B**). The abundances of all genera in the groups are shown in **Supplementary Table S2**. Butyrivibrio was the most abundant genus in both the MC and LC groups, whereas Prevotella was the most abundant genus in the HC group. Compared with the LC group, Ruminobacter and Anaerostipes showed the most significant increase (increased by 74 times and 46 times, respectively), whereas Sphingobium, Streptococcus, and Acinetobacter were the most significantly reduced (decreased by 99, 99, and 98%, respectively), in the MC group. Compared with the MC group, Sphingobium, Acinetobacter, and Streptococcus exhibited the most significant increase (increased by 135 times, 112 times, and 44 times, respectively), whereas Bifidobacterium, Microbacterium, Anaerostipes and Clostridium were the most significantly reduced (decreased by 99, 99, 98, and 96%, respectively), in the HC group.

Associated with the increase in the concentration of total SCFA, the diversity of the epimural microbiota presented the curve with a gradual increase from the LC group to the MC group, and a sharp decline from the MC group to the HC group (**Supplementary Figure S1**).



<sup>a</sup>Value is mean ± standard error (SE). <sup>b</sup>Bonferroni-Holm corrected two-tailed t-test.

TABLE 3 | Effects of LC, MC and HC diets on the concentrations of SCFAs, pH, and osmolarity in the goat rumen.


<sup>a</sup>Values are mean ± standard error (SE). <sup>b</sup>Bonferroni-Holm corrected two-tailed t-test.

#### Relationships Between SCFA Concentrations and Relative Abundance of Microbe Genera

CCA showed that the relative abundances of OTUs belonging to 25 genera were highly related to the concentrations of ruminal SCFAs. Moreover, the relative abundances of OTUs belonging to 10 genera were highly related to the ruminal pH (**Figure 2**).

#### Functional Comparisons of Epithelium Microbes Across the Groups

Metagenome shotgun sequencing generated a total of 156 G high-quality data. Among them, approximately 20 G data, which were likely to originate from the host, were excluded from the datasets. On average, 74% of the clean data was successfully assembled into the scaftigs. Subsequently, an average of 176,546, 206,490, and 211,140 open reading frames (ORFs) were detected within the LC, MC, and HC groups, which totally annotated to 6,605 KOs (**Supplementary Table S3**). Compared with the LC group, the relative abundance of 1,415 KOs was significantly upregulated, and the relative abundance of 534 KOs was significantly downregulated, in the MC group. Compared with the MC group, the relative abundance of 600 genes was significantly upregulated, and the relative abundance of 847 genes was significantly downregulated, in the HC group. Finally, these differentially abundant genes were annotated to 41 KEGG pathways of metabolism (**Figure 3**).

In the comparisons of the abundance of the metabolism pathways, we observed that galactose metabolism, starch and sucrose metabolism, three kinds of lipid metabolism, nucleotide metabolism and glycan biosynthesis and metabolism were upregulated in the MC group compared with the LC group, whereas both of them were downregulated in the HC group compared with the MC group. On the contrary, glycolysis, pyruvate metabolism, propanoate metabolism, butanoate metabolism and the metabolism of cofactors and vitamins were downregulated in the MC group compared with the LC group, whereas all of them were upregulated in the HC group compared with the MC group (**Figure 3**).

#### Enriched KEGG Pathways of Differentially Expressed Genes Related to the Immune System and Cellular Processes in the Rumen Epithelium

Transcriptome sequencing generated a total of 129 G raw data. On average, 83% of the high-quality reads were successfully mapped to the NCBI goat genome. Finally, in total, 11,131 genes were detected as being expressed in the rumen epithelium of these goats (**Supplementary Table S4**). Compared with the LC group, 374 genes were significantly upregulated, and 478 genes were significantly downregulated, in the MC group. Compared with the MC group, 267 genes were significantly upregulated, and 283 genes were significantly downregulated, in the HC group.

In the KEGG enrichment analysis, we observed that, in the MC group, all enriched immune-related pathways were downregulated in comparison with those in the LC group (**Figure 4**). The downregulated genes located on the enriched pathways, when the diets shifted from LC to MC, were listed in **Supplementary Table S5**. On the contrary, in the HC group, excepting for the IL-17 signaling pathway, all enriched immune-related pathways were upregulated in comparison with those in the MC group. The upregulated genes located on the enriched pathways, when the diets shifted from MC to HC, were listed in **Supplementary Table S6**. In the part of cellular process, in the MC group, all enriched pathways related to the cell growth and death (apoptosis, cell cycle and p53 signaling pathway), gap junction, tight junction and peroxisome were upregulated, while endocytosis, phagosome, focal adhesion and signaling pathways regulating pluripotency of stem cells were downregulated, in comparison with those in the LC group (**Figure 4A** and **Supplementary Table S5**). However, only two pathways related to the cellular community (signaling pathways

regulating pluripotency of stem cells and focal adhesion) and one pathway related to the cell motility (regulation of actin cytoskeleton) were enriched in the HC group, in comparison with those in the MC group (**Figure 4B** and **Supplementary Table S6**).

# Comparisons of MC-LC and HC-MC SCFA Regulation Networks

According to the SCCs between the MC and LC groups, the concentrations of ruminal acetate, butyrate, and propionate, together with the ruminal pH, had a high possibility of affecting the expression of genes located on 12 pathways of the cellular process. Moreover, the results suggested that genes associated with the cell cycle, tight junction, gap junction, peroxisome, and autophagy were promoted, whereas those associated with endocytosis were suppressed by the increased SCFA concentrations and the decreased pH of the MC group, compared with the LC group (**Figure 5A**). However, according to the SCC between the HC and MC groups, the further increase in the acetate concentration and the further decrease in the ruminal pH had a high possibility of suppressing the cellular functions concerning the gap junction, tight junction, and p53 signaling pathway, whereas they promoted the cellular functions concerning the lysosome, autophagy and cell cycle, through their impacts on the expressions of the corresponding genes (**Figure 5B**).

#### Confirmation of High-Throughput Sequencing Results by RT-QPCR

To validate high-throughput sequencing results, we compared the expression of 11 rumen epithelial genes which showed the significantly changed both between LC and MC groups and between MC and HC groups (**Supplementary Tables S5, S6**) by using RT-QPCR method, and the results are shown in **Supplementary Table S7**. In general, the expression analyses of these genes verified the significant difference discovered by RNAseq. Linear regression analysis showed that the RT-QPCR results were highly consistent with the RNA-seq results (R <sup>2</sup> = 0.80) (**Supplementary Figure S2**).

# DISCUSSION

Previous studies of humans have shown that an increase in the ratio of Firmicutes to Bacteroidetes is associated with the ability to extract more energy from food (Jumpertz et al., 2011). If it is also applied to the rumen epimural microbiota, the acid

environment induced the decrease on energy-extraction ability of the epimural microbiota might be part of the reason for the sharp decline of the microbial diversity in the HC group, in comparison with that of MC group. Next, the 16S rRNA sequencing data, metagenomic data and CCA results all indicated the importance of genera Sphingobium, Acinetobacter, and Streptococcus in the SCFA-microbe-host axis. So far, most of the known members in the genera Acinetobacter and Streptococcus exhibit both a pathogenic lifestyle and a commensal lifestyle in the GI tract. The study has shown that genomic plasticity enabling quick adaptation to environmental stress is a necessity for their pathogenic lifestyle, whereas the stability is a necessity for their commensal lifestyle (Kilian et al., 2014). Accordingly, we infer that, in an acid environment, the intermediate metabolites of the ruminal microbiota, such as lactate, promotes the growth of opportunistic pathogens, such as Streptococcus, in the rumen epithelium. In order to enhance the genomic plasticity of stress resistance, these opportunistic pathogens recruit microbes to form a multi-species biofilm and, subsequently, increase the ability of stress resistance in the biofilm by using the quorumsensing system (Nadell et al., 2009; Belkaid and Hand, 2014). Under these conditions, the opportunistic pathogens are prone to the pathogenic lifestyle and, thereby, promote immune responses in the rumen epithelium. On the contrary, in the MC group, the moderate increases of ruminal SCFAs promote the growth of the stratum corneum, whose keratin is an important nutrient for epimural microbes (Cheng et al., 1979) and, thereby, promote the diversification of epimural microbiota and its gene pool. According to ecological theory, the more diverse a community is, the more stable the ecosystem is. Thus, the moderate concentration of ruminal SCFAs promote the stability of the micro-ecosystem on the surface of the rumen epithelium and, thereby, promote the commensal lifestyle of the opportunistic pathogens in the rumen epithelium. Moreover, the increases of the ruminal butyrate may also help the host to suppress the expression of the virulence genes in the opportunistic pathogens (Ohland and Jobin, 2015). In short, our results indicate that, rather than having impacts on the colonization of opportunistic pathogens in the rumen epithelium, ruminal SCFAs affect the diversity of epimural microbiota and the lifestyle of opportunistic pathogens. However, further studies of meta-transcriptome and meta-metabolome are needed to reach comprehensive views concerning the interactions between ruminal SCFAs and epimural microbes.

According to the changes in the microbial metabolism pathways (**Figure 3**), our study primarily indicated that the epimural microbiome had different metabolism network responses to the dietary shifts from LC to MC, compared

FIGURE 3 | Heatmap comparing the kinds of gene, which has the significantly different abundance across the groups, on the specific KEGG metabolism pathway. M + (H0) indicates that, in this column, the abundance of genes was significantly increased in the MC group compared to the LC group, and meantime, had no significant changes between MC group and HC group; M + (H-) indicates that, in this column, the abundance of genes was significantly increased in the MC group compared to the LC group, and meantime, was significantly decreased in the HC group compared to the MC group; M0(H+) indicates that, in this column, the abundance of gene had no significant changes between LC group and MC group, and meantime, was significantly increased in the HC group compared to the MC group; M-(H +) indicates that, in this column, the abundance of genes was significantly decreased in the MC group compared to the LC group, and meantime, was significantly increased in the HC group compared to the MC group.

with the responses to the dietary shifts from MC to HC. This result is consistent with the results on the structure of microbial community. So far, there have been many reports on the kinds of microbial metabolites that modulate the barrier function of intestinal epithelium. Among them, SCFAs are the most studied metabolites, which promoted the barrier functions by promoting the production of protective mucus and IgA, regulating the differentiation of Tregs, and suppressing the production of inflammatory mediator nuclear factor κB (NFκB) (Ohland and Jobin, 2015). In addition, fose, a metabolite of Bacteroides thetaiotaomicron, inhibited the expression of the virulence factor ler in pathogenic E. coli within the colon (Pacheco et al., 2012). Trp metabolites provided the colonization resistance to the pathogenic fungus Candida albicans in the

FIGURE 4 | (A) Upregulated/downregulated KEGG pathways related to the immune system and cellular process in the rumen epithelium of the MC group, compared with those in the LC group; (B) upregulated/downregulated KEGG pathways related to the immune system and cellular process in the rumen epithelium of the HC group, compared with those in the MC group.

mice gut (Zelante et al., 2013). Lipopolysaccharides (LPS), the metabolites of gram negative bacteria, impaired the integrity of rumen epithelium in the cow with subacute ruminal acidosis (SARA) (Kleen et al., 2003; Emmanuel et al., 2007). Our study showed, when the diets shifted from the LC to the MC, the newly appeared genes and the significantly increased genes were

the pathway enzymes in (1) the metabolism of nucleotide and glycan. The nucleotide and glycan are the important components of microbes. Therefore, the increases of these metabolites could provide the materials for the growth of microbes; (2) the metabolism of lipids into fatty acids, and the metabolism of starch, sucrose and galactose into glucose. The fatty acids and glucose are the energy substrates of microbes. Therefore, the increases of these metabolites could provide the energy for the growth of microbes. When the diets shifted from the MC to the HC, the newly appeared genes and the significantly increased genes were the pathway enzymes in the fermentation of glucose and fatty acids into SCFAs and some other products, such as lactate and ethanol. The excessive increases of fermentation products impair the integrity of rumen epithelium and cause the SARA. It should be highlighted that our results, obtained from the statistical analysis, can only give the hints based on the statistical analysis. The real effects of the microbial metabolism on the function of rumen barrier, as well as the mechanisms, need further studies.

In the rumen epithelium, the results of our transcriptomic data showed that, associated with the increases on the ruminal SCFAs, the responses of immune components (including innate and adaptive immune cells, platelet, hematopoietic cell lineage and leukocyte) were decreased from the LC group to the MC group, whereas all of them were increased from the MC group to the HC group. Notably, the diversity of the gene pools related to the disease infection of the epimural microbes showed the opposite trend to the degree of immune responses in the rumen epithelium. Since most of the microbial pattern recognition receptors (MPRRs), which receive and deliver the signals from the microbes to the immune cells, are located in the basal of the rumen epithelium (Malmuthuge et al., 2012), and since, concomitantly, the changes in the thickness of the rumen epithelium are opposite to the responses of immune cells in the rumen epithelium, as revealed by the morphological measurement, we infer that the structure of the rumen epithelium and the activities of the epithelial cells play important roles in the shaping of the immune responses in the rumen epithelium through their effects on the reduction of dangerous signals from epimural microbes to the rumen epithelium. However, further evidence to support this hypothesis is required.

To date, the structure and function of the rumen epithelium is well known as being affected by the concentrations of ruminal SCFAs, the ruminal osmolarity and the ruminal pH. For example, butyrate promotes the thickening of the rumen epithelium by stimulating its growth and renewal (Malhi et al., 2013). In vitro increase of SCFA concentration at pH 6.1 upregulated the mRNA expression of tight junctions Cldn-1 and 4 (Greco et al., 2018). Ruminal osmolarity, mainly determined by the concentrations of ruminal SCFAs, influenced the paracellular resistance, apical Na+-H<sup>+</sup> exchange activity and integrity of rumen epithelium (Bennink et al., 1978; Schweigel et al., 2005; Lodemann and Martens, 2006). Increase of SCFAs concentration and decrease of ruminal pH enhanced the expressions of genes related to the cell proliferation and apoptosis in the rumen epithelium (Sun et al., 2018). In our study, in order to check the effects of ruminal SCFAs on epithelium structure, we constructed the SCFA regulatory network by using transcriptomic data. We observed that the tight junction, gap junction, and p53 signaling pathway were promoted by the increased concentrations of ruminal SCFAs and the decrease of ruminal pH from the LC group to the MC group, whereas all of them were suppressed by the further increase in the concentration of acetate and the decrease of ruminal pH from the MC group to the HC group. The upregulation of the tight junction and gap junction effectively limits the invasion of dangerous signals to the animal tissue, thereby contributing to the reduction of inflammatory responses in the rumen epithelium. The upregulation of the p53 signaling pathway, together with the upregulation of cell proliferation, promotes the cell refreshing in the MC group. These data are in agreement with the increased rumen papillae length and papillae density in MC group of current study, and consistent with the data in our previous study (Gui and Shen, 2016), which reported the moderate upregulation of ruminal SCFAs concentration induced the increase of cell apoptosis and p53 gene expression in rumen epithelium of goats. Taken together, these studies indicate the epithelial homeostasis and molecular adaptation of rumen epithelium to adequate concentrate intake is associated to moderate concentration of SCFA and pH in rumen. But, excess amount of SCFA and acidic pH, induced by high concentrate intake, may cause the damage of tight junction and gap junction in rumen epithelium and impair epithelial homeostasis (Greco et al., 2018).

Our examination of the regulatory networks additionally demonstrated that endocytosis was suppressed by the increased concentrations of ruminal SCFAs from the LC group, compared to the MC group, whereas it was not significantly changed with a further increase in the concentration of ruminal acetate and a further decrease in the ruminal pH, from the MC group to the HC group. Endocytosis is a mechanism for cells to remove ligands, including MPRRs, from the cell surface. Therefore, the upregulation of endocytosis contributes to the suppression of the immune response to non-pathogenic microbes in the rumen epithelium. Moreover, peroxisomes are exclusively promoted by increased SCFA concentrations and a decreased pH, from the LC group to the MC group. The upregulation of peroxisomes, which play an important role in lipid homeostasis and free radical detoxification, is essential for the maintenance of epithelial integrity. Whereas lysosomes are exclusively promoted by an increased acetate concentration and a decreased pH in the MC group compared to the HC group. The upregulation of lysosomes, which help the cells to kill and degrade invading bacteria, indicates the increase of dangerous signals in the rumen epithelium of the HC group. Overall, our results suggest that the regulatory effects of ruminal SCFAs on cellular activities of physical barrier play important roles in the modulation of immune responses in the rumen epithelium.

Totally, this study suggested that the medium increase of ruminal SCFAs promotes the diversity of epimural microbiota and its gene pool. Meantime, it induced the commensal lifestyle of opportunistic pathogens in the epimural microbiota. Furthermore, the moderate concentration of ruminal SCFAs suppressed the inappropriate immune responses, and promoted the tight junction and cell refreshing in the rumen epithelium. However, the further increase of ruminal SCFAs resulted in the acidification of rumen, leading to a decrease in the diversity of epimural microbiota and its gene pool. Meantime, it induced the pathogenic lifestyle of opportunistic pathogens in the epimural microbiota. Additionally, the acid environment in the rumen led to the upregulation of immune responses, the overgrowth of the epithelial cells, and the downregulation of tight junction and gap junction pathways in the rumen epithelium. Altogether, our study indicated a simultaneous regulation of ruminal SCFAs on the microbial, physical and immune barriers by their concentrations and the impacts on the ruminal pH.

#### ETHICS STATEMENT

fphys-10-01305 October 23, 2019 Time: 17:46 # 12

This study was carried out in accordance with the recommendation of the Regulations for the Administration of Affairs Concerning Experimental Animals (No. 588 Document of the State Council of the People's Republic of China, 2011). The protocol was approved by the Nanjing Agricultural University.

# AUTHOR CONTRIBUTIONS

HS wrote the manuscript. HS and ZX analyzed the data. ZL and ZS designed the research. ZL performed the experiments. All authors approved the final manuscript.

#### REFERENCES


#### FUNDING

This work was supported by the National Natural Science Foundation of China (A0201800763), Fundamental Research Funds for the Central Universities, and Science Foundation of Jiangsu Province (BK20180542).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphys. 2019.01305/full#supplementary-material

FIGURE S1 | Effects of LC, MC, and HC diets on the diversity of epimural microbiota.

FIGURE S2 | Linear regression analysis of relationships between RNA-seq results and RT-QPCR results for 11 selected genes.

TABLE S1 | Primers used for the RT-QPCR.

TABLE S2 | Genera composition and abundances of epimural microbiota.

TABLE S3 | KO compositions and abundances of epimural microbiota.

TABLE S4 | Gene compositions and abundances in the rumen epithelium.

TABLE S5 | Differentially expressed genes in the pathways related to the immune system and cellular processes when the diets shifted from LC to MC.

TABLE S6 | Differentially expressed genes in the pathways related to the immune system and cellular processes when the diets shifted from MC to HC.

TABLE S7 | RT-QPCR results of 11 selected genes that were differentially expressed in the rumen epithelium among groups.


Jumpertz, R., Le, D. S., Turnbaugh, P. J., Trinidad, C., Bogardus, C., Gordon, J. I., et al. (2011). Energy-balance studies reveal associations between gut microbes, caloric load, and nutrient absorption in humans. Am. J. Clin. Nutr. 94, 58–65. doi: 10.3945/ajcn.110.010132



**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Shen, Xu, Shen and Lu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Intestinal Microbiota-Associated Metabolites: Crucial Factors in the Effectiveness of Herbal Medicines and Diet Therapies

Yiliang Wang1,2,3† , Shurong Qin1,2,3,4† , Jiaoyan Jia1,2,3, Lianzhou Huang1,2,3,4, Feng Li1,2,3 , Fujun Jin<sup>5</sup> \*, Zhe Ren1,2,3 \* and Yifei Wang1,2,3 \*

<sup>1</sup> Guangzhou Jinan Biomedicine Research and Development Center, Institute of Biomedicine, College of Life Science and Technology, Jinan University, Guangzhou, China, <sup>2</sup> Key Laboratory of Virology of Guangzhou, Jinan University, Guangzhou, China, <sup>3</sup> Key Laboratory of Bioengineering Medicine of Guangdong Province, Jinan University, Guangzhou, China, <sup>4</sup> College of Pharmacy, Jinan University, Guangzhou, China, <sup>5</sup> Integrated Chinese and Western Medicine Postdoctoral Research Station, Jinan University, Guangzhou, China

#### Edited by:

Liwei Xie, Guangdong Academy of Sciences, China

#### Reviewed by:

Matthias J. Bahr, Sana Kliniken Lübeck, Germany Simona Bertoni, University of Parma, Italy

#### \*Correspondence:

Fujun Jin 464689341@qq.com Zhe Ren rzl62000@qq.com Yifei Wang twang-yf@163.com †These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology

Received: 12 February 2019 Accepted: 09 October 2019 Published: 29 October 2019

#### Citation:

Wang Y, Qin S, Jia J, Huang L, Li F, Jin F, Ren Z and Wang Y (2019) Intestinal Microbiota-Associated Metabolites: Crucial Factors in the Effectiveness of Herbal Medicines and Diet Therapies. Front. Physiol. 10:1343. doi: 10.3389/fphys.2019.01343 Although the efficacy of herbal medicines (HMs) and traditional Chinese medicines (TCMs) in human diseases has long been recognized, their development has been hindered in part by a lack of a comprehensive understanding of their mechanisms of action. Indeed, most of the compounds extracted from HMs can be metabolized into specific molecules by host microbiota and affect pharmacokinetics and toxicity. Moreover, HMs modulate the constitution of host intestinal microbiota to maintain a healthy gut ecology. Dietary interventions also show great efficacy in treating some refractory diseases, and the commensal microbiota potentially has significant implications for the high inter-individual differences observed in such responses. Herein, we mainly discuss the contribution of the intestinal microbiota to high interindividual differences in response to HMs and TCMs, and especially the already known metabolites of the HMs produced by the intestinal microbiota. The contribution of commensal microbiota to the inter-individual differences in response to dietary therapy is also briefly discussed. This review highlights the significance of intestinal microbiotaassociated metabolites to the efficiency of HMs and dietary interventions. Our review may help further identify the mechanisms leading to the inter-individual differences in the effectiveness of HM and dietary intervention from the perspective of their interactions with the intestinal microbiota.

Keywords: drug interventions, herbal medicines, traditional Chinese medicines, inter-individual differences, gut microbiota, metabolites

# BACKGROUND

The function of herbal medicines (HMs) and traditional Chinese medicines (TCMs) in the remedial and prophylactic management of human diseases has been recognized for a long time (Qiu, 2007; Fan et al., 2014; Wang et al., 2017; Xu et al., 2017; Nie et al., 2018; Wu and Tan, 2019), while the mechanisms of action of HMs remain largely unknown. Traditional studies focused on identifying the specific bioactive compounds in HMs, and such strategies have been successful in

**526**

developing certain compounds isolated from HMs into novel drugs (Xu et al., 2017; Feng et al., 2019). However, most components extracted from HMs exhibit poor bioactivity and bioavailability (Xu et al., 2017; Feng et al., 2019). Indeed, the pharmacological activity of HMs largely depends on intestinal microbiota-dependent biotransformation (Xu et al., 2016; Aguilar-Toalá et al., 2018). Compared to the primary drugs, metabolites produced by the intestinal microbiota often exhibit greater pharmacological activity and are more easily absorbed (Inao et al., 2004; Hussain et al., 2016). Moreover, several components of HMs can serve as nutrition for the growth of specific microbiota and hence modulate the constitution of host intestinal microbiota (Xu et al., 2017; Feng et al., 2019). Therefore, the contribution of host intestinal microbiotamediated biotransformation to the efficacy of HMs cannot be underestimated.

Indeed, the importance of the intestinal microbiota to human health and pathophysiology is indisputable. The beneficial effects of the intestinal microbiota are primarily contributed by the intrinsic constituents of the intestinal microbiota and the microbiota-associated metabolites, especially the subsets generated from beneficial bacteria (Rooks and Garrett, 2016; Bhat and Kapila, 2017; Hasegawa et al., 2017; Postler and Ghosh, 2017; Aguilar-Toalá et al., 2018; Cani, 2019; Silverman, 2019). The composition of the intestinal microbiota, and more specifically the metabolites generated through their biotransformation, has been shown to be closely associated with the large interindividual differences observed in responses to drugs and dietary interventions (Coryell et al., 2018; Gong et al., 2018; Gopalakrishnan et al., 2018; Nie et al., 2018; Olson et al., 2018; Rothhammer et al., 2018; Routy et al., 2018; Maini Rekdal et al., 2019; Zimmermann et al., 2019a). Of note, in vivo drug activity, including pharmacokinetics and toxicity, is closely associated with the gut microbiota (Coryell et al., 2018; Gong et al., 2018; Gopalakrishnan et al., 2018; Nie et al., 2018; Olson et al., 2018; Routy et al., 2018; Maini Rekdal et al., 2019; Zimmermann et al., 2019a). Accumulating evidence reveals that intestinal microbiota are crucial contributors to the high inter-individual differences in dietary intervention efficacy in treating some refractory diseases (Flint et al., 2014; Thorburn et al., 2014; Buffington et al., 2016; Rioscovián et al., 2016; Hasegawa et al., 2017; Nie et al., 2018; Requena et al., 2018), such as the anti-seizure effect of the ketogenic diet (KD) (Olson et al., 2018). However, the interaction between HMs or diet therapy and the host intestinal microbiota remains largely unknown.

Owing to a range of factors, including host-intrinsic, host-extrinsic, and environmental factors, the taxonomic composition of the intestinal microbiota varies greatly among individuals (Tsb et al., 2018). It is critical to obtain a clear understanding of the links between HMs or dietary interventions and their metabolites from commensal microbiota. Herein, we mainly discuss the metabolites produced from TCMs and HMs by the intestinal microbiota (**Figure 1**). The contribution of commensal microbiota to the high inter-individual differences in dietary intervention efficacy is also briefly discussed. Our review further suggests that the effect of microbiota should be considered while developing new dietary guidelines or drugs for clinical application.

#### INTESTINAL MICROBIOTA-ASSOCIATED METABOLITES OF THE COMPOUNDS ISOLATED FROM HMS

Herbal medicines have significantly contributed to human health through disease prophylaxis and therapy (Xu et al., 2017; Feng et al., 2019). The term HM covers raw and processed plants such as the roots, leaves, flowers, berries, and/or seeds from one or more plants (Feng et al., 2019). Materials derived from animals, fungi, and minerals are also regarded as HMs in some traditions (Xu et al., 2017; Feng et al., 2019). Although most of the supposed pharmacological effects of HMs were determined by preclinical researches or even empirical study alone, multiple traditional medicine systems, such as TCMs, Ayurveda, and Islamic medicine, are dominated by HMs (Xu et al., 2017). However, the mechanisms of action of most HMs and the reasons for the different responses of different individuals remain unclear (Xu et al., 2016, 2017; Singh et al., 2017; Nie et al., 2018; Maini Rekdal et al., 2019). Of note, most of the chemicals derived from HMs exhibit poor bioactivity and bioavailability (Xu et al., 2017). However, intestinal microbes are involved in the metabolism of drugs (Maini Rekdal et al., 2019; Zimmermann et al., 2019a,b), especially the compounds extracted from HMs (Nie et al., 2018; Tong et al., 2018). Such biotransformation may contribute to explaining the great inter-individual differences in response to HMs because the constitution of gut microbiota varies among individuals (Xu et al., 2016; Tsb et al., 2018; Maini Rekdal et al., 2019). In this section, we mainly attempt to gain a more comprehensive and detailed understanding of the interactions between HMs and the intestinal microbiota. The role of microbiota in the in vivo activity and toxicity of chemical drugs is also discussed.

The compounds extracted from HMs that can be metabolized by the intestinal microbiota are mainly classified into subsets based on their chemical skeletons and include flavonoids, glycosides, terpenoids, anthraquinones, alkaloids, and organic acids (**Table 1**). Of these compounds, flavonoids are the most thoroughly studied, and most are degraded into flavone glycosides by the microbiota once the flavonoid enters the large intestine (**Table 1**). However, the final metabolites vary according to the specific medication and particular gut bacterial composition (**Table 1**). Bifidobacteria may be the group of microorganisms that can metabolize the widest range of compounds, including soy isoflavones, puerarin, ginsenoside, and sennoside (**Table 1**). Moreover, several specific bacteria can metabolize different compounds into the same metabolites. For instance, Bifidobacterium can metabolize both soy isoflavones and puerarin into daidzein (**Table 1**). Of note, the polyphenolics of berries and pomegranate fruit, a compound in unconventional HMs, can be metabolized by Bifidobacterium pseudocatenulatum INIA p815 into urolithin A,

host. Moreover, particular components of the host diet and medicines can be metabolized by commensal microbiota to generate specific metabolites. The final metabolites may affect the toxicity and efficiency of drugs and dietary interventions, partly mediating the large inter-individual differences observed among hosts.

which has multiple activities, including combating inflammation, oxidation, and aging, and enhancing gut barrier function (Singh et al., 2019). Collectively, the metabolism of HMs may not be highly dependent on a specific bacterium. However, the specific role of intestinal microbiota in the metabolism of HMs needs to be confirmed in clinical studies in the future, as the existing studies regarding their relationship refer only to preclinical studies.

In addition to the HMs, the gut microbiota is also closely associated with the in vivo activity of chemical drugs. Given that prior influential studies have revealed the gut microbes involved in drug metabolism and their potential genes (Zimmermann et al., 2019a,b), we briefly discuss the role of microbiotamediated biotransformation in drug activity and toxicity through introducing several representative drugs (**Table 2**). For instance, gut microbes have been suggested to be crucial factors in the conversion of L-dopa to dopamine (Maini Rekdal et al., 2019). The bioconversion of L-dopa to dopamine depends on a pyridoxal phosphate-dependent tyrosine decarboxylase from Enterococcus faecalis followed by transformation of dopamine to m-tyramine by a molybdenum-dependent dehydroxylase from Eggerthella lenta (Maini Rekdal et al., 2019). In addition, the gut microbiota is responsible for varying responses to simvastatin treatment, resulting in vast differences in the hypolipidemic effect of simvastatin among patients (Krauss et al., 2013; He et al., 2017). Furthermore, although PD-1 inhibitors exhibit potent activity against cancer by blocking a "checkpoint" molecule on T cells, only 25% of patients respond well to PD-1 blockers. The gut microbiota is a crucial factor in determining the response of an individual to various treatments (Gopalakrishnan et al., 2018; Routy et al., 2018). Gut microbes are also a crucial factor affecting the in vivo drug toxicity. For example, diurnal variation in acute liver injury caused by acetaminophen is partly mediated by 1-phenyl-1,2 propanedione, a metabolite of acetaminophen generated by gut microbiota (Gong et al., 2018). Interestingly, acetaminophen hepatotoxicity can be reduced through postbiotic-induced autophagy by Lactobacillus fermentum (Dinic et al., 2017),

#### TABLE 1 | Metabolites produced by intestinal bacteria from HMs.

fphys-10-01343 October 25, 2019 Time: 17:16 # 4


which demonstrates that different bacteria play distinct roles in the toxicity of the same drug. These findings suggest that an understanding of the interaction between intestinal microbiota and drug metabolism is critical for developing new drugs that are efficacious, which is significant for the frequent emergence of drug-resistance.



#### GUT MICROBES: CRUCIAL FACTORS FOR THE FUNCTION OF TCM

It has long been known that TCM is effective for treating many human diseases, including influenza virus infection, cancer, diabetes, and cardiovascular diseases (Qiu, 2007; Fan et al., 2014; Wang et al., 2017; Xu et al., 2017; Nie et al., 2018; Wu and Tan, 2019). The fundamental functions and applications of TCM depend on the compatible application of herbal formulas (FuFang in Chinese) based on ancient empirical philosophies such as Yin-Yang (Dong et al., 2018). However, the mechanisms of action of TCM remain largely unclear or unknown. Recent insights into TCM have focused on its interactions with the gut microbiota (Xu et al., 2017; Feng et al., 2019; Wu and Tan, 2019). Firstly, the carbohydrates in HMs cannot be digested by the human body, while the human gut microbiome encodes thousands of carbohydrate-active enzymes to digest herbal carbohydrates (Xu et al., 2017; Lu et al., 2019). Secondly, the non-carbohydrate bioactive compounds in TCM, particularly triterpene glycosides, flavonoids, isoflavones, iridoid glycosides, alkaloids, and tannins, have poor lipophilicity, high hydrogenbonding capacity, and high molecular flexibility, which limit the bioavailability of TCM (Xu et al., 2017). However, these non-carbohydrate compounds can be metabolized into several metabolites by the gut microbiota, increasing the efficiency of intestinal absorption and thereby improving their bioavailability (Xu et al., 2017). Moreover, most TCM formulas can reshape the structure of commensal flora, such as by increasing the level of beneficial bacteria and reducing the abundance of harmful bacteria (**Table 3**). Of note, the enrichment of beneficial gut microbes and the reduction of harmful gut microbes is not merely a result of disease symptom improvement, because the recovery of the balance of the gut microbiota usually occurs before an improvement in the disease symptoms (Xu et al., 2015). Collectively, the efficacy of TCMs may be the comprehensive outcome of both reshaping the microbiota structure and the complex interaction between intestinal microbiota and multiple chemical substances in TCMs.

The most typical example of this is the excellent efficacy of TCMs in the management of type 2 diabetes (T2D) (Xu et al., 2015; Nie et al., 2018; Tong et al., 2018; Cheng F. R. et al., 2019; Cheng J. et al., 2019; Han et al., 2019; Li et al., 2019; Lu et al., 2019; Shi et al., 2019; Wu et al., 2019; Yuan et al., 2019). The major component of HMs, such as the polysaccharides extracted from Hirsutella sinensis, provides nutrition to specific bacteria, thereby modulating the constitution of the intestinal microbiota to improve T2D (Xu et al., 2015, 2017; Nie et al., 2018; Tong et al., 2018; Wu et al., 2019; **Table 3**). Of note, a multicenter, randomized, open-label clinical trial revealed that metformin and the Chinese herbal formula AMC (including Rhizoma Anemarrhenae, Momordica charantia, Coptis chinensis, Salvia miltiorrhiza, red yeast rice, Aloe vera, Schisandra chinensis, and dried ginger) may ameliorate T2D with hyperlipidemia by enriching beneficial bacteria, including Blautia and Faecalibacterium spp. (Tong et al., 2018). In addition, treatment of Gegen Qinlian Decoction (GQD), another TCM formula, can enrich the gut in beneficial bacteria such as Faecalibacterium spp., which is associated with the antidiabetic effect of GQD (Xu et al., 2015; **Table 3**). Indeed, under fermentation by the intestinal microbiota, HMs can be metabolized into various chemical substances with wideranging activities that improve host health (Yang et al., 2012; Nie et al., 2018; Wu et al., 2019) and jointly enhance the gut barrier, control insulin resistance, and reduce inflammation in


TABLE 3 | Effect of Traditional Chinese medicines (TCM) formulas on the constitution of commensal microbiota and host metabolisms in indicated diseases.

the host (Nie et al., 2018). Furthermore, HMs regulate many complex chemical interactions in the gut, thereby maintaining a healthy gut ecology (Nie et al., 2018), which is important in recovery from gut dysbiosis. However, whether these altered microbiotas metabolized specific components in TCMs into functional molecules remains uncertain. Metabolomics analysis is an ideal method for determining the altered microbiotaassociated metabolites of TCMs.

#### EFFECT OF INTESTINAL MICROBIOTA-ASSOCIATED METABOLITES ON THE EFFICIENCY OF DIETARY THERAPY

Dietary interventions have become an effective method for treating some refractory diseases, with the effects being associated with the commensal microbiota of the host Richards J. L. et al., 2016; Wu et al., 2016). The KD has long been known to exhibit high efficacy against refractory seizure, despite the response rate being low among tested patients (Kwan and Brodie, 2000; Olson et al., 2018). A recent influential study revealed that the gut microbiota was responsible for the high inter-individual differences observed in the anti-seizure effect of the KD (Olson et al., 2018). Ketogenic diet-associated Akkermansia and Parabacteroides confer seizure protection to mice fed a controlled diet by reducing the level of gamma-glutamyl amino acids and increasing the GABA and glutamate content in the brain (Olson et al., 2018). In addition, a Mediterranean diet, which is based on the high consumption of cereals, fruit, vegetables, and legumes, has been associated with the prevention of cardiovascular diseases and asthma (Castro-Rodriguez et al., 2008; Estruch et al., 2013; Blanco Mejía et al., 2019). The Mediterranean diet increases the abundance of Lactobacillus in the mammary gland microbiota and subsequently elevates the levels of bile acid and bacterial-modified metabolites in breast cyst fluid (Shively et al., 2018). However, the beneficial effects of the Mediterranean diet on human health also depend, in part, on non-bacterial metabolites, especially ω-3 fatty acids, which exert larger antiinflammatory effects (Thorburn et al., 2014). Further, given that the Mediterranean diet is rich in fiber, SCFAs may mediate the beneficial effect of this diet, since the administration of SCFAs is associated with significant improvements in cardiovascular diseases (Richards L. B. et al., 2016); this requires further research. Of note, in the gastrointestinal tract of human patients with type II diabetes, the administration of Bifidobacterium increases the abundance of Akkermansia muciniphila, with both microbes being able to generate SCFAs, thereby improving insulin resistance and limiting inflammation and consequently improving the symptoms of obesity (Cani, 2019). Furthermore, arsenic poisoning arising from the ingestion of contaminated

food and drinking water is a challenging disease to treat (Coryell et al., 2018). A promising finding is that gut microbes, especially Faecalibacterium, provide full protection against acute arsenic toxicity in a mouse model (Coryell et al., 2018).

However, some of the observed dietary effects have not yet been associated with specific intestinal microbes or with specific metabolites. For instance, a maternal high-fat diet negatively impacts the social behavior of offspring, resulting in a deficiency in synaptic plasticity in the ventral tegmental area and in oxytocin production, but the administration of Lactobacillus reuteri restores oxytocin levels, synaptic plasticity, and healthy social behaviors in mice (Buffington et al., 2016). It has also been recognized that a Malawian diet may induce kwashiorkor, an enigmatic form of severe acute malnutrition. In a study involving 317 Malawian twin pairs, researchers found that an altered gut microbiota constitution in response to the Malawian diet significantly contributed to the development of kwashiorkor, although the mechanism involved remains unknown (Smith et al., 2013). Notably, oligosaccharides were less abundant in the milk from mothers of severely stunted infants, and the administration of sialylated milk oligosaccharides reversed infant undernutrition in a microbiota-dependent manner (Smith et al., 2013). Such results were also confirmed in piglets that received the same diet as the human infants (Charbonneau et al., 2016), suggesting that microbiota associated-metabolites of oligosaccharides may be a crucial factor in such processes. In young children, a negative association between dietary fiber and plasma insulin levels has been observed only in those whose gut microbiota showed a high abundance of Bacteroides and Prevotella and not in those whose gut microbiota exhibited a higher proportion of Bifidobacterium (Zhong et al., 2019). This suggests a potential function for Bacteroides and Prevotella in elevating insulin levels. Indeed, convincing epidemiological studies have indicated that specific dietary components may be crucial for the pathogenesis of some diseases such as asthma and allergies (Eder et al., 2006; Graham, 2006). For example, a carnitine-rich diet induces the symptoms of atherosclerosis in a gut microbiota-dependent manner in humans and mice (Koeth et al., 2013). Specifically, the gut microbiota in humans and mice mediates the metabolism of dietary choline and phosphatidylcholine to produce trimethylamine, which is further transformed into trimethylamine-N-oxide by hepatic flavin monooxygenases, thereby promoting the development of atherosclerosis. However, the specific microbial taxa contributing to this process require further investigation.

#### CONCLUSION AND FUTURE PERSPECTIVE

The beneficial effect of HMs and dietary therapy in several refractory diseases is generally appreciated, but the underlying mechanisms involved remain obscure. However, their interaction with the host microbiota seems to be a critical factor in such processes. Indeed, a growing number of studies indicate that the commensal microbiota plays a crucial role in maintaining host health and that the constitution of the intestinal microbiota exhibits large inter-individual differences. Moreover, most components in HMs and dietary interventions can modulate the constitution of the microbiota, which may disrupt or maintain homeostasis in the host. Collectively, it is not surprising that the gut microbiota, and especially microbiota-associated metabolites, may be a crucial mediator linking HMs or dietary therapy and the physiological status of the host. Therefore, it is important to consider the effects of biotransformation by commensal microbiota when designing herbal formula dietary therapy to achieve optimal success in treating diseases, particularly in the case of precision medicine. It is also essential to determine the optimal timing of administrating HMs and specific diets, in particular given that the composition of the gut microbiota exhibits diurnal variation. Indeed, microbiota-associated metabolites have several attractive properties, including known chemical structures and long shelf lives (Aguilar-Toalá et al., 2018). In particular, these metabolites are able to mimic the health effects mediated by probiotics while avoiding the administration of live bacteria, which can produce harmful reactions such as the local inflammatory response induced by the administration of Salmonella (Tsilingiri et al., 2012). However, the importance of postbiotics does not diminish the beneficial effect of probiotics when there is stable colonization of the gut, because live bacteria undoubtedly provide more metabolites than can be provided using postbiotics. The future of next-generation probiotics lies not only in supplementation using beneficial bacteria strains but also in providing and maintaining the ecological context necessary to sustain them. The direct administration of these probioticassociated metabolites should provide a great advantage over traditional probiotics for several types of patients, including those harboring intestinal pathogens. Furthermore, since metabolites from the intestinal microbiota can also partially mediate the toxicity of some medicines in vivo, it will also be valuable to further examine these associations in order to assist in developing novel approaches to reducing the toxicity of HMs and TCMs.

# AUTHOR CONTRIBUTIONS

YlW contributed to the conception, design, collection and assembly of references, discussion, interpretation, and writing of the manuscript. SQ contributed to the collection and assembly of references, interpretation of the article, and writing of the manuscript. JJ, LH, and FL contributed to the collection and assembly of references. FJ, ZR, and YfW contributed to the conception, design, interpretation of the article, and the final article approval.

# FUNDING

This work was supported by grants from the Key Laboratory of Virology of Guangzhou, China (201705030003), the National Natural Science Foundation of China (Nos. 81573471 and 81872908), Key Projects of Biological Industry Science and Technology of Guangzhou China (Grant No. 201504291048224), Guangzhou Industry, University and Research Collaborative Innovation Major Project (No. 201704030087), and the Public

#### REFERENCES

fphys-10-01343 October 25, 2019 Time: 17:16 # 8


Service Platform of the South China Sea for R&D Marine Biomedicine Resources, Marine Biomedical Research Institute, Guangdong Medical University, Zhanjiang, China.



a Chinese herbal formula. Syst. Appl. Microbiol. 36, 188–196. doi: 10.1016/j. syapm.2012.12.009


Zimmermann, M., Zimmermann-Kogadeeva, M., Wegmann, R., and Goodman, A. L. (2019b). Separating host and microbiome contributions to drug pharmacokinetics and toxicity. Science 363:eaat9931. doi: 10.1126/science. aat9931

**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Wang, Qin, Jia, Huang, Li, Jin, Ren and Wang. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Effect of Fermented Corn-Soybean Meal on Serum Immunity, the Expression of Genes Related to Gut Immunity, Gut Microbiota, and Bacterial Metabolites in Grower-Finisher Pigs

Junfeng Lu, Xiaoyu Zhang, Yihao Liu, Haigang Cao, Qichun Han, Baocai Xie, Lujie Fan, Xiao Li, Jianhong Hu, Gongshe Yang and Xin'e Shi\*

Key Laboratory of Animal Genetics, Breeding and Reproduction of Shaanxi Province, Laboratory of Animal Fat Deposition and Muscle Development, College of Animal Science and Technology, Northwest A&F University, Yangling, China

Fermented corn-soybean meal (fermented feed, FF) is commonly used in swine production, but the effects of FF on gut health remain unclear. In this study, serum immunity, mRNA abundances of antimicrobial peptides (AMPs) and Toll-like receptors (TLR1-9), bacterial abundance in the duodenum and colon, and colonic metabolic phenotypes were determined in crossbred barrows (Duroc × Landrace × Large White) fed FF or normal feed (unfermented feed, UF) (n = 6). When compared to the UF group, the results showed that serum levels of IgG and IgM were significantly increased in FF group pigs (P < 0.05). FF significantly decreased the abundances of Bacteroides and Verrucomicrobia in the duodenum and decreased the abundances of Bacteroides, Proteobacteria, and Verrucomicrobia in the colon while it significantly increased the abundances of Firmicutes and Actinobacteria (P < 0.05). Furthermore, a Spearman's correlation analysis showed that serum immunity and the expression of genes related to gut immunity were associated with bacterial strains at the family level. Moreover, differentially abundant colonic microbiota were associated with colonic metabolites. LC-MS data analyses identified a total of 1,351 metabolites that markedly differed between the UF and FF groups. C5-Branched dibasic acid metabolism was significantly upregulated whereas the purine metabolism was significantly downregulated (P < 0.05) in the colonic digesta of pigs in the FF meal group compared to the UF meal group. Collectively, these results indicated that FF meal could influence serum immunity and the expression of genes related to gut immunity, correlating with the gut microbiota and bacterial metabolites in grower-finisher pigs. This study may provide an alternative strategy for improving the intestinal health of grower-finisher pigs.

Keywords: fermented feed, corn-soybean meal, immunity, microbiota, metabolite, pig

#### Edited by:

Yuheng Luo, Sichuan Agricultural University, China

#### Reviewed by:

Jing Zhang, Shanghai Jiao Tong University, China Christine Ann Butts, The New Zealand Institute for Plant and Food Research Limited, New Zealand

\*Correspondence:

Xin'e Shi xineshi@nwafu.edu.cn

#### Specialty section:

This article was submitted to Food Microbiology, a section of the journal Frontiers in Microbiology

Received: 05 March 2019 Accepted: 28 October 2019 Published: 20 November 2019

#### Citation:

Lu J, Zhang X, Liu Y, Cao H, Han Q, Xie B, Fan L, Li X, Hu J, Yang G and Shi X (2019) Effect of Fermented Corn-Soybean Meal on Serum Immunity, the Expression of Genes Related to Gut Immunity, Gut Microbiota, and Bacterial Metabolites in Grower-Finisher Pigs. Front. Microbiol. 10:2620. doi: 10.3389/fmicb.2019.02620

#### INTRODUCTION

fmicb-10-02620 November 18, 2019 Time: 13:39 # 2

Antibiotics have long been used in swine production to maintain health and productivity. In recent years, the overuse of antibiotics in swine production has become an important concern (van der Fels-Klerx et al., 2011) and has led to the development of alternatives to antibiotics in feeds (Dibner and Richards, 2005), such as probiotics, prebiotics, and organic acids, for use in swine husbandry (Thacker, 2013; Markowiak and Slizewska, 2018). The microbial fermentation of feed has been proposed as a potential alternatives (Verstegen and Williams, 2002).

Extensive evidence has shown that microbial fermentation can improve the nutritional quality of pig feed by increasing the bioavailability of nutrients (Borling Welin et al., 2015; Jeong et al., 2016) and reducing the content of anti-nutritional factors (Missotten et al., 2015; Shi et al., 2017; Wang et al., 2018). Moreover, the use of fermented corn-soybean meal has been shown to improve the immune function of pigs (Zhu et al., 2017), as the consumption of FF has a positive impact on the gut microbiota and can improve the fecal microbial count and intestinal morphology in grower-finisher pigs (Dowarah et al., 2017). However, so far, information on the effects of FF on the health of pigs is limited.

Many studies have highlighted the correlation between gut microbiota and intestinal morphology, regulation of immunity, digestion of carbohydrates, and body health in livestock (Richards et al., 2005; Stanley et al., 2016; Gu et al., 2019; Guevarra et al., 2019). Recent evidence suggests that the gut microbiota plays a crucial role in livestock health and disease (Tremaroli and Backhed, 2012). A growing number of studies indicate that the homeostasis and composition of the gut microbiota is dynamically formed by many factors (Ananthakrishnan, 2015; Zhao et al., 2015), including age, time, feed, and probiotics (Isaacson and Kim, 2012). For instance, in grower-finisher pigs, fermented Mao-tai lees modulate the gut microbiota and increase the abundance of the potentially beneficial bacteria Lactobacillus and Akkermansia (Wang J. et al., 2019). Moreover, 45% feed of fish meal can be replaced by fermented soybean meal without negative effects on growth performance and intestinal integrity of juvenile large yellow croaker (Wang P. et al., 2019).

It remains uncertain whether FF boosts immunity and the composition of the gut microbiota of swine. Therefore, the aim of the present study was to assess the effects of corn-soybean FF on serum immunity factors and the expression of genes related to gut immunity. The results of an intestinal microbiota analysis indicated that corn-soybean FF significantly changed the composition of the gut microbiota and that metabolites significantly differed between pigs fed FF meal versus UF meal.

#### MATERIALS AND METHODS

#### Ethics Statement

This study was carried out in accordance with the recommendations of the Animal Welfare Committee of Northwest A&F University (Yangling, Shaanxi Province, China). The protocol was approved by the Animal Welfare Committee of Northwest A&F University.

#### Preparation and Composition of FF

Corn-soybean meal was purchased from Beijing Dabeinong Technology Group Co., Ltd. (Beijing, China). Effective MicroorganismsTM used in the present study are a mixture containing 60% Lactobacillus, 20% Clostridium, and 8% Bifidobacteria and were purchased from Nongfukang Biological Technology Co., Ltd. (Zhengzhou, China) and diluted 1:30 (w/v) with sterile water. In accordance with the manufacturer's instructions, the corn-soybean meal was mixed with probiotics and incubated at 27–32◦C for 36 h. We then determined the live bacteria in the fermented product, and the final number of microorganisms was guaranteed at a concentration of 2 × 10<sup>9</sup> CFU/g. The method for CFU determination was based on the previous study (Sieuwerts et al., 2008). The 16S rRNA sequences of the bacteria in FF are shown in **Supplementary Figure S1**. After fermentation, the fermented corn-soybean meal was dried at 30–40◦C to a moisture content of 10%. The ingredients and nutrient content (%) of the UF and FF are listed in **Table 1**. Both the UF and FF met all recommended nutrient levels (NRC, 2012), and neither contained antimicrobials or growth promoters.

#### Animals, Housing, and Treatment

A total of 48 growing barrow pigs (Duroc × Landrace × Large White) (53.90 ± 1.31 kg initial body weight) were randomly allocated into one of two feeding groups; one group was fed with commercial soybean meal (UF) and a second with fermented complete commercial soybean meal (FF). Each group consisted of 24 barrow pigs that were housed in six pens, with four pigs per pen, in an environmentally controlled facility under a constant temperature of 25–28◦C and with free access to feed and clean water throughout the experimental period.


<sup>1</sup>Premix supplied per kilogram of meal: VD3, 2800 IU; VE, 26 mg; VK, 2 mg; VB1, 50 mg; VB6, 3 mg; VA, 6480 IU; VB2, 4 mg; VB12, 0.03 mg; pantothenic acid, 9 mg; nicotinic acid, 20 mg; choline chloride, 300 mg; biotin, 0.2 mg; Fe, 200 mg; Cu, 95 mg; Mn, 30 mg; folic acid, 1.2 mg; Zn 100 mg; I, 0.35 mg; Se, 0.36 mg; P 0.1%; NaCl, 0.5%; lysine, 0.1%; Ca, 0.9%. <sup>2</sup>Nutrient contents are calculated values.

#### Sample Collection and Preparation

fmicb-10-02620 November 18, 2019 Time: 13:39 # 3

Upon reaching slaughter weight (approximately 110 kg), the control group was fed UF meal for 76 days and the treatment group was fed FF meal for 56 days. One pig from each pen was randomly selected and fasted for 12 h (with free access to water) before slaughter and sampling. Blood samples were collected from the external jugular vein into serum separation tubes and centrifuged at 2,500 rpm for 15 min at 4◦C, then stored at −80◦C until analysis. After blood sampling, all 12 pigs were anesthetized and slaughtered. After recovery of the duodenum and colon, duodenal and proximal colonic tissue samples were collected, washed with 0.9% saline, quickly frozen, and then stored at −80◦C until further analysis. Finally, the digesta in the duodenum and colon was immediately collected and frozen at −80◦C.

#### Serum Biochemical Indices and Immunoglobulin Analysis

In the present study, the serum concentrations of aspartate transaminase (AST), alanine transaminase (ALT), total protein, triglycerides, glucose, albumin, globulin, and low-density lipoprotein cholesterol were measured using a Hitachi-7180 Biochemical Analyzer (Hitachi Medical Corporation, Tokyo, Japan) provided by the Yangling Demonstration Zone Hospital. Serum concentrations of immunoglobulin (Ig) A, M, and G (IgA, IgM, and IgG, respectively) were determined with a commercial Enzyme-linked immunosorbent assay kit (BIM Biosciences, San Francisco, CA, United States). All procedures were performed in accordance with the manufacturers' instructions. Each sample was tested in triplicate.

#### Total RNA Extraction and Real-Time Polymerase Chain Reaction (PCR)

Total RNA was extracted from liquid nitrogen-frozen samples of the duodenum and colon using TRIzol Regent (Takara Bio Inc., Shiga, Japan). RNA integrity and quality were determined by agarose gel electrophoresis (1%) and spectrometry (A260/A280), respectively. A commercial reverse transcription (RT) kit (Takara Bio Inc.) was used for the synthesis of cDNA. The RT products (cDNA) were stored at −20◦C for relative quantification by PCR. For a real-time quantitative polymerase chain reaction (RTqPCR), every reaction was performed in triplicate using SYBR green kits on an Applied Biosystems ABI 7500 system (Thermo Fisher Scientific, Waltham, MA, United States). The expression levels of all genes were normalized to that of glyceraldehyde 3 phosphate dehydrogenase (GAPDH) using the 2−11CT method. The sequences of primers used for RT-qPCR are listed in **Table 2**.

#### Intestinal Digesta DNA Extraction and Pyrosequencing

Total genomic DNA was extracted from the duodenal and colonic digesta samples using the cetyltrimethylammonium bromide and sodium dodecyl sulfate method. DNA concentration and purity were monitored on 1% agarose gels. DNA extracted from each sample was used as a template to amplify the V4–V5 regions of 16S rRNA genes for later pyrosequencing by Novogene TABLE 2 | Primers sequences for RT-qPCR.


F, forward; R, reverse; PBD-1, porcine beta-defensin 1; PR39, proline-arginine rich 39-amino acid peptide; TLR, Toll-like receptor; GAPDH, glyceraldehyde-3 phosphate dehydrogenase.

Biological Information Technology Co. (Beijing, China), as described previously (Tian and Zhang, 2017). The forward and reverse primer sequences for the V4–V5 rRNA gene library preparation are presented in **Supplementary Table S1**. Raw reads were submitted to the Sequence Read Archive of the National Center for Biotechnology Information (SRA, No. PRJNA524989).

#### Bioinformatics Analyses

The paired-end reads were merged using FLASH (V1.2.7)<sup>1</sup> (Magoc and Salzberg, 2011). Quality filtering on the raw tags was performed according to the QIIME 1.7.0<sup>2</sup> quality controlled process (Caporaso et al., 2010). Next, the tags were compared with the reference database (Gold database)<sup>3</sup> (Edgar et al., 2011) using an UCHIME algorithm (UCHIME Algorithm)<sup>4</sup> , to detect chimera sequences, and the chimera sequences were then removed (Haas et al., 2011). The Effective Tags were then finally obtained. High-quality clean tags were obtained and classified into the same operational taxonomic units (OTUs) at an identity threshold of 97% similarity by UPARSE software (Uparse

<sup>1</sup>http://ccb.jhu.edu/software/FLASH/

<sup>2</sup>http://qiime.org/index.html

<sup>3</sup>http://drive5.com/uchime/uchime\_download.html

<sup>4</sup>http://www.drive5.com/usearch/manual/uchime\_algo.html

v7.0.1001)<sup>5</sup> (Edgar, 2013). For each representative sequence, the Greengene Database<sup>6</sup> (DeSantis et al., 2006) was used based on a ribosomal database projects (RDPs) classifier (Version 2.2)<sup>7</sup> (Wang et al., 2007) algorithm to annotate taxonomic information. In order to study the phylogenetic relationship of different OTUs, and the difference of the dominant species in different samples (groups), multiple sequence alignments were conducted using MUSCLE software (Version 3.8.31)<sup>8</sup> (Edgar, 2004). OTUs abundance information was normalized using a standard of sequence number corresponding to the sample with the fewest number of sequences. Subsequent analyses of alpha diversity and beta diversity were all performed based on this normalized output data. Alpha diversity is applied in analyzing the complexity of species diversity for a sample through four indices, including Observed species, Chao 1, Shannon, Simpson, and ACE. All the indices in our samples were calculated with QIIME (Version 1.7.0) and displayed with R software (Version 2.15.3). The relative abundance at the phylum and genus levels was compared between the two groups, the top 30 most abundant families were defined as predominant genera and sorted for the comparison.

#### Untargeted Metabolomics Study Based on Liquid Chromatography Tandem Mass Spectrometry (LC-MS/MS)

Tissues (100 mg) were individually ground with liquid nitrogen and the homogenate was resuspended in pre-chilled 80% methanol (−20◦C) and then vortexed. The samples were incubated at −20◦C for 60 min and then centrifuged at 14,000 g and 4◦C for 20 min. The supernatants were subsequently transferred to a fresh Eppendorf tube and spun in a vacuum concentrator until dry. The dried metabolite pellets were reconstituted in 60% methanol and analyzed by LC-MS/MS.

LC-MS/MS analyses were performed using a Vanquish ultrahigh-performance liquid chromatography (UHPLC) system (Thermo Fisher Scientific) coupled with an Orbitrap Q Exactive HF-X mass spectrometer (Thermo Fisher Scientific) at Novogene Genetics, Beijing, China. Samples were injected into a Hypersil Gold column (100 × 2.1 mm, 1.9 µm; Thermo Fisher Scientific) using a 16 min linear gradient at a flow rate of 0.3 mL/min. The eluents for the positive polarity mode were eluent A (0.1% aqueous formic acid solution) and eluent B (methanol). The eluents for the negative polarity mode were eluent A (5 mmol/L ammonium acetate, pH 9.0) and eluent B (methanol). The solvent gradient was set as follows: 2% B, 1.5 min; 2–100% B, 12.0 min; 100% B, 14.0 min; 100–2% B, 14.1 min; and 2% B, 16.0 min. A Q Exactive HF-X mass spectrometer (Thermo Fisher Scientific) was operated in positive/negative polarity mode with a spray voltage of 3.2 kV, capillary temperature of 320◦C, sheath gas flow rate of 35 arb, and auxiliary gas flow rate of 10 arb.

The raw data files generated by UHPLC-MS/MS were processed using the Compound Discoverer 3.0 (Thermo Fisher Scientific) to perform peak alignment, peak picking, and quantitation of each metabolite. The main parameters were set as follows: retention time tolerance, 0.2 min; actual mass tolerance, 5 ppm; signal intensity tolerance, 30%; signal/noise ratio, 3; and minimum intensity, 100000. Afterward, peak intensities were normalized to the total spectral intensity. The normalized data were used to predict the molecular formula based on additive ions, molecular ion peaks, and fragment ions. Then, the peaks were matched with the mzCloud<sup>9</sup> and ChemSpider<sup>10</sup> databases to obtain accurate qualitative and relative quantitative results. The online KEGG database was used to identify metabolites by matching the molecular mass data. Finally, metabolites for separating the models were selected with the following requirements: variable importance in projection (VIP) > 1 and | P(corr)| ≥ 0.5 with 95% jack-knifed confidence intervals. The Student's t-test was applied to further analyze intergroup significance of the selected metabolites. Pathway analysis and enrichment analysis of differential metabolites were conducted on the MetaboAnalyst web server<sup>11</sup> .

#### Statistical Analysis

Data are presented as the mean ± standard error of the mean (SEM). All statistical analyses were performed using IBM SPSS Statistics for Windows, version 19.0 (IBM Corporation, Armonk, NY, United States) using an one-way ANOVA with Turkey's multiple comparison test or Student's test. A probability (P) value of <0.05 was considered statistically significant. The correlations between the colonic microbial composition (relative abundance of family higher than 0.1%) and serum immunity and genes related to gut immunity that were significantly affected by FF treatment were assessed by a Spearman's correlation test using GraphPad Prism version 8.00 (GraphPad Software, San Diego, CA, United States). The correlations between the colonic microbial composition (relative abundance of family higher than 0.1%) and metabolites that were significantly affected by FF meal were assessed by a Spearman's correlation analysis using GraphPad Prism version 8.00 (GraphPad Software, San Diego, CA, United States).

# RESULTS

#### FF Increased Growth Performance

To investigate the effect of FF on body weight, after reaching the similar weight (53.19 ± 2.17 kg vs. 54.60 ± 1.62 kg), the pigs fed a corn-soybean meal were given free access to FF or UF. When compared with the UF group, the average daily weight gain was significantly increased in the FF group (P < 0.05) and there was no significant difference (P > 0.05) in the average daily intake among the two groups. Furthermore, the weight gain to food ratio was significantly increased in the FF group rather than in the UF group (P < 0.05) (**Table 3**).

<sup>5</sup>http://drive5.com/uparse/

<sup>6</sup>http://greengenes.lbl.gov/Download/

<sup>7</sup>http://sourceforge.net/projects/rdp-classifier/

<sup>8</sup>http://www.drive5.com/muscle/

<sup>9</sup>https://www.mzcloud.org/

<sup>10</sup>http://www.chemspider.com/

<sup>11</sup>https://www.metaboanalyst.ca/



Data are presented as the mean ± SEM (n = 6). UF, pigs fed with unfermented corn-soybean meal; FF, pigs fed with fermented corn-soybean meal.

# FF Changed Serum Immunity

fmicb-10-02620 November 18, 2019 Time: 13:39 # 5

Next, the effects of FF on immunity were assessed. As shown in **Figures 1A–C**, the ALT and AST concentrations tended to decrease in the FF-treated group as compared to the UF group (P > 0.05). There were no significant differences in the other serum parameters between the two groups (**Supplementary Table S2**). As shown in **Figures 1D–F**, FF significantly increased the serum concentrations of IgM and IgG (P < 0.05), while there was no significant difference in IgA levels between the two meals.

### Effect of FF on the Expression of Genes Related to Gut Immunity

The effects of FF on the expression of genes related to gut immunity was determined. The expression levels of the antimicrobial peptide-encoding genes PBD-1 and PR39 revealed the potential capacity for the eradication of invading pathogens. The results of the gene expression analysis of the intestinal tissue are presented in **Figure 2**. The FF meal significantly increased the mRNA abundance of AMPs and TLRs. In the duodenum, the mRNA abundances of TLR1, TLR2, TLR3, TLR4, PBD-1, and PR-39 in the FF-treated pigs were significantly increased (P < 0.05), while those of TLR5 and TLR9 were very significantly increased (P < 0.01; **Figure 2A**). In the proximal colon, the mRNA abundances of TLR4 and TLR5 in the FF-treated pigs were significantly increased (P < 0.05), while those of TLR1, TLR2, TLR7, and PBD-1 in the FFtreated pigs were very significantly increased (P < 0.01; **Figure 2B**).

# FF Shaped the Intestinal Microbiota

To explore the cause of improved immunity in FF-treated pigs, the intestinal microbiota were quantified by 16S rRNA sequence analysis, which identified a total of 65,249 ± 16,236 V4–V5 16S rRNA sequence reads from each sample. Binning sequences using a pairwise identity threshold of 97% obtained an average of 800 ± 231 operational taxonomic units per sample. MOTHUR plotting for the construction of a sparse curve between the number of reads and the number of operational taxonomic units revealed a tendency for plateau saturation (**Supplementary Figure S2**). Statistical estimates of species richness for 5,000 subsets of each sample at a genetic distance of 3% showed

that there was no difference in richness estimators (ACE and Chao 1) and the diversity indices (Shannon and Simpson diversity) (**Table 4**).

At the phylum level, Firmicutes and Bacteroidetes were the most predominant phyla in the duodenum and colon (**Tables 5**, **6**). FF meal significantly decreased the proportion of Bacteroidetes and Verrucomicrobia (P < 0.05) and tended to decrease the relative abundance of Spirochaetes in the duodenum (P = 0.069) (**Table 5**). FF meal significantly increased the proportion of Firmicutes and Actinobacteria (P < 0.01), whereas the proportion of Proteobacteria, Bacteroidetes, and Verrucomicrobia was decreased in the colon (P < 0.05) (**Table 6**).

A family-level analysis of the top 30 most abundant families revealed that Acidaminococcaceae and Clostridiales\_vadinBB60\_group were significantly decreased in relative abundance by the FF meal treatment in the duodenum (P < 0.05) (**Table 7**). A family-level analysis of the top 30 most abundant families revealed that Ruminococcaceae, Rikenellaceae, Christensenellaceae, Lactobacillaceae, and Family XIII were significantly increased in relative abundance by the FF meal treatment, whereas the abundance of Prevotellaceae, Lachnospiraceae, Clostridiaceae\_1, Bacteroidales\_RF16\_group, Streptococcaceae, Veillonellaceae, Erysipelotrichaceae, Peptostreptococcaceae, Acidaminococcaceae, Bacteroidaceae,


TABLE 4 | Diversity estimation of the 16S rRNA gene libraries from microbiota in the duodenum and colon of pigs fed the control UF and FF meals.

UF, pigs fed with unfermented corn-soybean meal; FF, pigs fed with fermented corn-soybean meal. Data are presented as the mean ± SEM (n = 6).

Succinivibrionaceae, Clostridiales\_vadinBB60\_group, Anaeroplasmataceae, Alcaligenaceae, Fibrobacteraceae, Rhodospirillaceae, and Unidentified\_Thermoplasmatales was decreased in the colon (P < 0.05) (**Table 8**).

#### Correlation Between Colonic Microbial Composition and Serum Immunity and Genes Related to Gut Immunity

To comprehensively analyze the relations between host serum immunity, genes related to gut immunity, and gut microbiota, a correlation matrix was generated by

TABLE 5 | Relative abundance of microbial phylum (percentage) in the duodenum of pigs in the FF and control groups.


UF, pigs fed with unfermented corn-soybean meal; FF, pigs fed with fermented corn-soybean meal. Data are presented as the mean ± SEM (n = 6).

TABLE 6 | Relative abundance of microbial phylum (percentage) in the colon of pigs in the FF and control groups.


UF, pigs fed with normal commercial feed; FF, pigs fed with fermented meal. Data are presented as the mean ± SEM (n = 6).

calculating the Spearman's correlation coefficient. As shown in **Figure 3**, Spearman's correlation analysis showed that the serum IgM was positively associated with the abundance of Enterobacteriaceae but negatively related to the proportion of Lachnospiraceae in the colon. The gene expression of TLR1 was negatively correlated with the abundance of Ruminococcaceae and Erysipelotrichaceae. The gene expression of TLR2 was positively associated with the abundance of Veillonellaceae. The gene expression of TLR3 was negatively correlated with the abundance of Peptostreptococcaceae. The gene expression of TLR4 was positively correlated with the abundance of Enterobacteriaceae, Clostridiaceae\_1, and Bacteroidales\_S24-7\_group. The gene expression of TLR5 and PBD-1 were positively associated with the abundance Lachnospiraceae but negatively related to the proportion of Bacteroidales\_S24-7\_group. The gene expression of TLR7 was positively associated with the abundance Bacteroidales\_S24-7\_group but negatively related to the proportion of Lachnospiraceae. Additionally, the gene expression of TLR5 was also negatively related to the proportion of Veilonellaceae.

#### Effect of FF on Fermentation Metabolites in Colonic Digesta

To further study the effect of FF on the gut, the colonic digesta metabolic profiles of the two meals were acquired by LC-MS. As shown by the PCA score plots presented in **Figures 4A,B** which distinguished metabolic communities based on meal of colonic sampling, the metabolic communities were clustered. The partial least squares discriminant analysis (PLS-DA) score plots also reflected that FF led to significant biochemical changes (**Figures 4C,D**). Furthermore, LC-MS/MS (ESI−) and LC-MS/MS (ESI+) detected a total of 398 and 953 biomarker metabolites, respectively (**Figures 4E,F**).

In summary, there were significant differences in the 1,351 biomarker metabolites between the UF and FF groups (detailed information is presented in **Supplementary Data S1**). Next, the Kyoto Encyclopedia of Genes and Genomes (KEGG) was used to analyze the pathways of the metabolites that differed between the two meal groups (**Figure 5**). The effects of C5-branched dibasic acid metabolism were significantly upregulated while the purine metabolism was significantly downregulated in the FF group, as compared to the control group (P < 0.05). Meanwhile, the effects of dioxin degradation, phenylpropanoid biosynthesis, and aminobenzoate degradation were downregulated in the FF group (0.05 < P < 0.1).


TABLE 7 | Relative abundance (percentage) for the top 30 most abundant family in the duodenum of pigs in the fermented feed (FF) and control (UF) groups.

TABLE 8 | Relative abundance (percentage) for the top 30 most abundant family in the colon of pigs in the fermented feed (FF) and control (UF) groups.


UF, pigs fed with unfermented corn-soybean meal; FF, pigs fed with fermented corn-soybean meal. Data are presented as the mean ± SEM (n = 6).

#### Correlation Between Microbiota Communities and Related Metabolites

Metabolomics has been shown to be an important tool to reveal the potential crosstalk of host and gut microbiota. Therefore, correlations between metabolites and familylevel microbiota with significant differences between two meals were obtained via spearman's correlation analysis (**Figure 6**). As shown in **Figure 6**, ten bacterial strains (family Enterobacteriaceae, Lactobacillaceae, Prevotellaceae, Clostridiaceae\_1, Ruminococcaceae, Veillonellaceae, Peptostreptococcaceae, Erysipelotrichaceae, Lachnospiraceae, and Bacteroidales\_S24-7\_group) were most closely related to the metabolites in the FF group, as compared to the control group (P < 0.05).

# DISCUSSION

The gut microbiota are critical to metabolism, nutrient absorption, and host immunity (El Aidy et al., 2013), and the pig


UF, pigs fed with normal commercial feed; FF, pigs fed with fermented meal. Data are presented as the mean ± SEM (n = 6).

microbiota has become the focus of much attention (Frese et al., 2015; Ji et al., 2018; Wang W. et al., 2019). FF, as an available feed alternative, has great potential to improve gut health and maintain gastrointestinal tract microbial homeostasis (Jin et al., 2017) and could modulate the host gut microbiota through dietary manipulation (Wang et al., 2018). In this study, the effects of FF versus normal feed on serum immunity, expression of genes related to gut immunity, gut microbiota composition, and bacterial metabolites were investigated in grower-finisher pigs. The results reflected that FF regulated the microbiota composition in the duodenum and colon of pigs, and it also selectively changed the metabolomics profiles.

Corn-soybean meal is the most frequently used meal for livestock production in China. Solid-state fermentation can improve the nutritional value of plant materials and has it been suggested to increase the use of FF in livestock feeds (Shi et al., 2017). In the animal production, pigs are often slaughtered at a constant body weight (market weight, around 110 kg) to maintain uniformity of pork products and maximize profits (Kim et al., 2005; Frederick et al., 2006; Vermeer et al., 2014). In the present

study, we performed the study at an average initial body weight (53.19 ± 2.17 vs. 54.60 ± 1.62 kg), and an interesting finding in this study was that FF meal greatly reduced the time to market (76 vs. 56 days). As compared with the UF group, the average daily weight gain and weight gain: food ratio in the FF group were significant increased (P < 0.05), which reflects improved growth performance and feed conversion efficiency, in accordance with the findings of previous studies (Canibe and Jensen, 2003). The fermentation process is believed to promote functional activities, such as antimicrobial and antioxidant activities, and increases the production of growth factors, hormones, and amino acids (Ng et al., 2011; Laskowska et al., 2017).

Proper function of the immune system is important for grower-finisher pigs. Immunoglobulins are immune-active molecules that play important roles in the humoral immune response (Kong et al., 2007). In the present study, the concentrations of IgG and IgM were significantly greater in the FF group (P < 0.05). The levels of IgG reflected immune status (Machado-Neto et al., 1987). IgM is associated with antiinflammation and a higher concentration reflects better immune status (Vaschetto et al., 2017). Our results were similar to those of Zhu et al. (2017), who reported that weaned piglets fed fermented soybean meal had higher serum concentrations of IgG and IgM, as compared with the controls. Some probiotic strains can be used as immunomodulators to enhance the concentrations of serum immunoglobulins (Vitini et al., 2000). Laskowska et al. (2017) found that dietary supplementation of Effective Microorganisms activated and enhanced the humoral and cell-mediated immune responses and protected against infection. Although high concentrations of immunoglobulins were observed, there was no intestinal inflammation, suggesting that the immunoglobulins were in the normal ranges. Moreover, FF meal decreased the concentrations of serum ALT and AST (P > 0.5). Reportedly, ALT is a liver-specific enzyme and ALT concentrations increased in response to acute liver injury (Robertson et al., 2016). Here, the lower levels of ALT and AST indicated that FF boosted overall health. In short, FF meal enhanced immune performance.

Toll-like receptors are the earliest discovered pattern recognition receptors and play critical roles in innate immunity (Newburg and Walker, 2007; Werling and Coffey, 2007), and TLR expression levels may indicate disease resistance in pigs (Uenishi and Shinkai, 2009; Cheng et al., 2015). Our data showed that the mRNA abundances of TLRs in the FF group were significantly greater than in the control group (P < 0.05). In this study, in the duodenum of FF-fed pigs, the mRNA abundances of PBD-1 and PR39 were significantly higher than that of the UF control pigs (P < 0.05). Meanwhile, the gene expression of PBD-1 in the colon was also higher than that of the UF meal pigs (P < 0.01). These results indicate that FF may benefits the pig gut immunity and in order to fully understand the regulating mechanism of FF on gut immunity, in vivo pathogenic challenge model is needed in further studies.

We then used a high-throughput sequencing method based on the 16S rRNA genes to demonstrate the effects of FF on the intestinal microbiota of grower-finisher pigs. In this study, FF had no effect on gut microbial community evenness (ace, Shannon H) and richness (Simpson and Chao 1). Previous studies have shown that probiotic-supplemented FF decreased microbial diversity, which may be linked with improved resistance to gastrointestinal disorders (Ott et al., 2004; He et al., 2017). Similar to previous studies (Kim et al., 2012; Looft et al., 2012;

FIGURE 4 | Multivariate statistical analysis of untargeted metabolomics data obtained using the LC-MS/MS approach. PCA score plot of colonic metabolomics data for treatment (blue) and control (red) pigs obtained by (A) LC-MS (ESI−) and (B) LC-MS (ESI+) (n = 6). (C) PLS-DA score plot of colonic metabolomics data obtained by LC-MS (ESI−); R2Y = 0.99; Q<sup>2</sup> = 0.95. (D) PLS-DA score plot of colonic metabolomics data obtained by LC-MS (ESI+) data; R2Y = 1.00; Q<sup>2</sup> = 0.97. (E) Score plot of LC-MS (ESI−) data with 2,395 metabolite signals detected. (F) Score plot of LC-MS (ESI+) data with 5,708 metabolite signals detected. Red circles in volcano plots are model-separated metabolites following the conditions of VIP > 1 and | P(corr)| ≥ 0.5 with 95% jack-knifed confidence intervals. Red or green rectangles indicate the numbers and tendency of metabolites to separate in the model when FF group pigs are compared with UF group.

Niu et al., 2015; Li et al., 2017), Firmicutes and Bacteroidetes were the most dominant phyla in the present study. Studies have shown that in obese animals, the ratio of Firmicutes to Bacteroides is increased (Ley et al., 2005, 2006b). In the present study, the ratio of the Firmicutes to Bacteroides was increased in both the duodenum and colon of the treatment group, which indicated that the use of FF changed the proportion of the microbiota and is beneficial to the weight gain of finisher pigs. In addition, a microbiome enriched in Firmicutes has been associated with an increased capacity for energy harvest and obesity (Ley et al., 2006a; Turnbaugh et al., 2006), and an increase of this phylum could therefore increase the amount of calories extracted from the diet. Compared with the control group, the relative abundance of Verrucomicrobia was significantly decreased in the FF group in the duodenum. Verrucomicrobia usually represents a minor population of intestinal microbiota in response to dietary shifts in mice (Pantoja-Feliciano et al., 2013). A relatively smaller proportion of Proteobacteria in the FF group was detected in our current study. It's reported that the members of the phylum Proteobacteria have a low abundance in the gut of healthy humans (Shin et al., 2015) and the increased levels of Proteobacteria may be indicative of a diseased state that commonly occurs during enteric infection or following perturbation of the microbiota (Singh et al., 2015). Moreover, FF significantly changed the gut microbiota composition, as indicated by decreased proportions, and significantly increased the proportions of Actinobacteria (P < 0.05). Meanwhile, we found there were five bacterial strains were significantly increased and 17 bacterial strains were significantly decreased in the colon in response to FF meal at the family level (P < 0.05); further studies are needed to investigate the roles of these gut bacteria in regulating the swine gut development.

The gut microbiota is important to host health and physiology status (Lalles, 2016). It is reported that microbes in the large intestine undertake more metabolism tasks (Zhao et al., 2015). Using a KEGG pathway analysis, we found that C5-Branched dibasic acid metabolism was significantly upregulated whereas the purine metabolism was significantly downregulated (P < 0.05). Similarly, Gomez et al. (2017) measured the fecal microbiota and calculated the functional potential of the microbial communities in healthy and diarrheic calves. They found that C5-branched dibasic acid metabolism was enriched in healthy calves, suggesting that C5-branched dibasic acid metabolism is related to energy generation (Turnbaugh et al., 2006). The purine content of animal feed is a concern because excess intake of purines may increase the risk of hyperuricemia and gout (Zheng et al., 2018). It is necessary to conduct further research beyond the scope of the present study.

Fermented feed meal altered the metabolic functions and phenotypes of gut microbiota in pigs. The relative abundances of bacteria at the family levels were closely associated with the concentration of specific microbial metabolites in the colonic digesta. For instance, Lachnospiraceae is a bacterial family known to be abundant in the intestinal ecosystem (Sagheddu et al., 2016), and it is reported that Lachnospiraceae was positively correlated with carbohydrate metabolism (Morgan et al., 2012). Moreover, we found that Lachnospiraceae was also positively correlated with Sorbitan monooleate, N-Acetyl-L-aspartic acid, N-Acetylcadaverine, and Mucronine B, whereas it negatively correlated with Caffeic acid phenethyl ester, Dodecanedioic acid, Torsemide, N-Acetyl-L-leucine, and Viloxazine. The mechanism underlying the dietary regulation of host phenotypes should be explored with a particular focus on metabolism and related receptors.

Further studies are needed to explore the benefits of FF benefits on the gut microbiota and health of pigs. To this end, fecal microbiota transplantation (FMT) may be an effective strategy. FMT refers to the process of transplantation of fecal bacteria from healthy individuals into a recipient (Drekonja et al., 2015). Growing evidence has shown that the host phenotypes,

such as obesity (Lee et al., 2019), Clostridium difficile infection (Khoruts and Sadowsky, 2016), and anti-seizure (Olson et al., 2018) and anti-tumor immunity (Sivan et al., 2015), can be altered by FMT in mammals, demonstrating the critical roles of gut microbiota in host health. Thus, FMT also helped evaluate the potential links between intestinal microbiota and host phenotypes. Recently, Hu et al. (2018) proposed the standardized preparation for FMT for use in pig production and to identify host microbiota-derived bacteriocin targets to determine diarrhea resistance in early-weaned piglets (Hu et al., 2018). In the current study, we determined the FF could benefits the pig's gut microbiota; it would be interesting to explore the mechanism and we would like to focus on this in our next study.

#### CONCLUSION

fmicb-10-02620 November 18, 2019 Time: 13:39 # 13

In conclusion, we found that long-term consumption of FF meal increased the serum concentrations of IgG and IgM and changed the expression of genes related to gut immunity, which may be associated with alterations in the microbiota community and microbial metabolites and benefits to the pig's health. In addition, the results suggest that FF altered the microbial composition and modulated the metabolic pathway of microbial metabolism in pig colons. These alterations provide an alternative strategy for improving the intestinal health of pigs.

#### DATA AVAILABILITY STATEMENT

The datasets generated for this study can be found in NCBI, No. PRJNA524989.

#### ETHICS STATEMENT

The use of animals and the performance of all experimental protocols were approved by the Northwest A&F University Animal Welfare Committee (Yangling, Shaanxi Province, China).

#### REFERENCES


The processing of animal experiments and sample collection strictly followed the relevant guidelines.

#### AUTHOR CONTRIBUTIONS

XS and JL designed the study, and wrote and revised the manuscript. JL, XZ, HC, QH, BX, and LF helped took samples and performed the experiments and analyses. JL, YL, XL, JH, GY, and XS edited the manuscript. All authors reviewed the final manuscript.

#### FUNDING

This study was supported by the Agricultural Special Fund Project in Shaanxi Province (NYKJ-2018-YL01) and the Technical Innovation Guidance Special (Foundation) Project of Shaanxi Province (2017ZKC07-114).

#### ACKNOWLEDGMENTS

We thank others in Li's team and Shi's team for their excellent support during this experiment.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.02620/full#supplementary-material

microbiota and gut health of grower-finisher crossbred pigs. Livestock Sci. 195, 74–79. doi: 10.1016/j.livsci.2016.11.006



**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Lu, Zhang, Liu, Cao, Han, Xie, Fan, Li, Hu, Yang and Shi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Dietary Supplementation With Citrus Extract Altered the Intestinal Microbiota and Microbial Metabolite Profiles and Enhanced the Mucosal Immune Homeostasis in Yellow-Feathered Broilers

Miao Yu1,2,3,4,5, Zhenming Li1,2,3,4,5, Weidong Chen1,2,3,4,5, Gang Wang1,2,3,4,5, Yiyan Cui1,2,3,4,5 and Xianyong Ma1,2,3,4,5 \*

1 Institute of Animal Science, Guangdong Academy of Agricultural Sciences, Guangzhou, China, <sup>2</sup> State Key Laboratory of Livestock and Poultry Breeding, Guangzhou, China, <sup>3</sup> Key Laboratory of Animal Nutrition and Feed Science in South China, Ministry of Agriculture, Guangzhou, China, <sup>4</sup> Guangdong Key Laboratory of Animal Breeding and Nutrition, Guangzhou, China, <sup>5</sup> Guangdong Engineering Technology Research Center of Animal Meat Quality and Safety Control and Evaluation, Guangzhou, China

#### Edited by:

Jie Yin, Institute of Subtropical Agriculture (CAS), China

#### Reviewed by:

Yong Su, Nanjing Agricultural University, China Jingrang Lu, United States Environmental Protection Agency, United States Wei Chen, Jiangnan University, China

> \*Correspondence: Xianyong Ma maxianyong@gdaas.cn

#### Specialty section:

This article was submitted to Systems Microbiology, a section of the journal Frontiers in Microbiology

Received: 13 August 2019 Accepted: 01 November 2019 Published: 26 November 2019

#### Citation:

Yu M, Li Z, Chen W, Wang G, Cui Y and Ma X (2019) Dietary Supplementation With Citrus Extract Altered the Intestinal Microbiota and Microbial Metabolite Profiles and Enhanced the Mucosal Immune Homeostasis in Yellow-Feathered Broilers. Front. Microbiol. 10:2662. doi: 10.3389/fmicb.2019.02662 The present study aimed to investigate the effects of citrus extract (CE) on intestinal microbiota, microbial metabolite profiles, and the mucosal immune status in broilers. A total of 540 one-day-old yellow-feathered broilers were randomly allotted into three groups and fed a basal diet (control group), or a basal diet containing 10 mg/kg of zinc bacitracin (antibiotic group), or 10 mg/kg of CE (CE group). Each treatment consisted of six replicates, with 30 broilers per replicate. After 63-day feeding, two broilers per replicate were randomly selected and slaughtered, and their ileal and cecal digesta and ileal tissue were collected for microbial composition, microbial metabolites, and gene expression analysis. The results showed that CE significantly increased the abundance of Barnesiella and Blautia than did the antibiotic group (adjusted P < 0.05), whereas it decreased the abundance of Alistipes and Bacteroides (adjusted P < 0.05). Meanwhile, the CE group also increased the numbers of Bifidobacterium and Lactobacillus than did the control and antibiotic groups (P < 0.05), whereas it decreased the number of Escherichia coli (P < 0.05). For microbial metabolites, dietary supplementation with CE increased the concentrations of lactate, total short-chain fatty acids, acetate, and butyrate in the cecum than did the control and antibiotic groups (P < 0.05), whereas it decreased the concentrations of amino acid fermentation products (ammonia, amines, p-cresol, and indole) (P < 0.05). Additionally, supplementation with CE up-regulated (P < 0.05) the mRNA expression of intestinal barrier genes (ZO-1 and Claudin) in the ileum than did both the control and antibiotic groups. However, antibiotic treatment induced gut microbiota dysbiosis, altered the microbial metabolism, and disturbed the innate immune homeostasis. In summary, these results provide evidence that dietary supplementation with CE can improve the intestinal barrier function by changing microbial composition and metabolites, likely toward a host-friendly gut environment. This suggests that CE may possibly act as an efficient antibiotic alternative for yellow-feathered broiler production.

Keywords: citrus extract, immune homeostasis, intestinal microbial community, microbial metabolites, yellowfeathered broilers

#### INTRODUCTION

fmicb-10-02662 November 22, 2019 Time: 16:28 # 2

In-feed antibiotics have been extensively used as growth promoters in livestock production to maintain health and to improve feed conversion efficiency, utilization, and growth performance (Castanon, 2007). However, the continuous and excessive use of in-feed antibiotics for animals' production has led to the development of antibiotic-resistant microbes and a number of residual antibiotics in animal products, both of which pose a potential threat to human health (Gadde et al., 2017). In China, the use of antibiotics in poultry feeds is still a common practice, although it has increasingly caused safety concerns and increased the consumer demand for antibiotic-free animal products; thus, the use of antibiotics for growth promotion in feed will be banned in the future. Consequently, it is necessary to develop novel feed alternatives that offer both security and efficiency with the potential to replace antibiotics while improving poultry health and product quality.

As a safe and efficient alternative to antibiotics, many plant extracts have been used as promising feed additives for livestock production. Among these potential alternatives, citrus extract (CE) is often used as one of promising candidates in poultry (Akbarian et al., 2013; Ebrahimi et al., 2015; Pourhossein et al., 2015). CE is a rich source of many important bioactive ingredients including vitamins, minerals, phenolic compounds, nobiletin, and flavonoids (Li et al., 2006; Parmar and Kar, 2010; Bermejo et al., 2011). Thus, CE may provide numerous health benefits, including as an antimicrobial agent against Escherichia coli and Salmonella typhimurium with the ability to selectively inhibit the growth of potentially pathogenic bacteria (Nannapaneni et al., 2008, 2009; O'Bryan et al., 2008) and enhance immune system activities (Chen et al., 2012). Indeed, the beneficial effects of CE were extensively investigated in poultry production. Several previous studies reported that the dietary supplementation with citrus products in broiler feed could enhance growth performance (Seidavi et al., 2015), stimulate IgG and IgM antibody production in serum (Pourhossein et al., 2015), and decrease the number of E. coli in the cecum digesta by using a culture-based approach (Ebrahimi et al., 2015; Alefzadeh et al., 2016). These results indicated that CE can modulate the intestinal microbiota and immune system activities. However, the effects of CE on the intestinal microbial community and epithelial immune status remain limited and require further investigation. Additionally, alterations in the microbiota by dietary treatment can also induce changes in the metabolic end-products of microbial degradation (Fouhse et al., 2015). However, whether dietary supplementation with CE affects the intestinal microbial metabolites in broilers remains unclear.

To test the hypothesis that CE as an antibiotic alternative may positively alter the microbial community and its metabolites, and that these alterations can also modulate the mucosa immune response in yellow-feathered broilers, the current study investigated the effects of dietary supplementation with CE on the microbial community, microbial metabolite profiles, and expression of immune-related genes in the intestine.

#### MATERIALS AND METHODS

#### Ethics Approval and Consent to Participate

The experimental proposals and procedures for the care and treatment of the broilers were approved by the Animal Care and Use Committee of Guangdong Academy of Agricultural Sciences (authorization number GAASIAS-2016-017).

#### Animals, Experimental Design, and Sampling

A total of 540 one-day-old yellow-feathered male broilers were randomly allotted into three groups. Each treatment consisted of six replicates, and each replicate had 30 broilers. There was no difference of statistics in initial average body weight of broilers among the three group (41.37 ± 0.35 g). The broilers of the control group were fed a basal diet without any antibiotics (control group); the antibiotic and CE groups were fed the same basal diet with 10 mg/kg of zinc bacitracin (antibiotic group) and 10 mg/kg of CE (CE group) during the whole trial period, respectively. The CE used in the current study was provided by the Guangdong Runsen Health and Environmental Technology Development Co., Ltd., Guangdong, China. The contents of total flavone, polysaccharide, citric acid, and chlorogenic acid in the CE were measured as previously described (Kong et al., 2009; Wan et al., 2016) and were 2.48, 1.20, 1.30, and 0.68%, respectively. The basal diets were formulated to either meet or exceed the nutrient requirements of Chinese yellow-feathered broilers (Ministry of Agriculture of P. R. China, 2004). The dietary composition and nutrient contents for the starting (1–21 days), growing (22–42 days), and finishing (43–63 days) phases are shown in **Table 1**. All broilers were housed in battery cages (3.0 m × 3.0 m × 0.9 m) in an environmentally controlled room with a continuous light regimen throughout the 63 day experimental period. The environment temperature was maintained at 33◦C for the first week and then decreased by 3 ◦C every week until a final temperature of 24◦C. All broilers



ME, metabolized energy; TP, total phosphorus; AP, available phosphate. <sup>1</sup>Provided per kilogram of complete diet: pantothenic acid, 10.9 mg; nicotinic acid, 30 mg; folic acid, 0.95 mg; biotin, 0.16 mg; vitamin A, 8,000 IU; vitamin D, 2,800 IU; vitamin E, 19 mg; vitamin K3, 3.32 mg; vitamin B1, 1.7 mg; vitamin B2, 8.2 mg; vitamin B6, 2.78 mg; vitamin B12, 0.015 mg; Mn (MnSO4·H2O), 82 mg; Zn (ZnSO4·H2O e), 68 mg; Fe (FeSO4·H2O), 81 mg; Cu (CuSO4·5H2O), 9 mg; I (KI), 0.50 mg; Se (Na2SeO3), 0.27 mg. <sup>2</sup>Values were calculated from data provided by Feed Database in China (2016) except that crude protein was analyzed.

were provided with diets and water ad libitum throughout the whole trial period.

At the end of the feeding period (day 63), two broilers per replicate were randomly selected and slaughtered after being fasted for approximately 12 h. The ileal and cecal digesta were collected, homogenized, and stored at –80◦C for later determination of the microbial communities and metabolites analyses. In addition, a segment of mid-ileum tissue was also rapidly removed and washed with phosphate-buffered saline (PBS, pH 7.0), then immediately frozen in liquid nitrogen, and stored at –80◦C for later gene expression.

#### Intestinal Microbial DNA Extraction, Illumina MiSeq Sequencing, and Bioinformatics Analysis

The microbial total genomic DNA extraction was conducted from 250 mg of ileal and cecal digesta samples, using the QIAamp PowerFecal DNA Kit (Qiagen, Hilden, Germany), according to the manufacturer's instructions. The concentrations of every DNA sample were quantified using a NanoDrop 2000 spectrophotometer (Thermo Scientific, Wilmington, DE, United States). To balance the cost of the experiment and the number of replicates necessary, two samples from each replicate were pooled in equal proportion and were selected for 16S rRNA MiSeq sequencing. The V3–V4 region of the bacterial 16S rRNA gene was amplified using primer pairs 338F (5<sup>0</sup> -ACTCCTRCGGGAGGCAGCAG-3<sup>0</sup> ) and 806R (5<sup>0</sup> - GGACTACCVGGGTATCTA AT-3<sup>0</sup> ) (Mao et al., 2015). PCR products were purified using an AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, United States) to remove excess primer dimers and dNTPs. After purification, the amplicons were pooled in equimolar and performed on an Illumina MiSeq 2 × 250 platform (Illumina, San Diego, CA, United States) at Majorbio Bio-Pharm Technology (Shanghai, China) according to standard protocols (Caporaso et al., 2012). The raw reads in this study were submitted to the National Center of Biotechnology Information (NCBI) Sequence Read Archive (SRA) database under accession number SRR9074899–SRR9074916.

QIIME (version 1.17) software package was used to demultiplex and quality filter the obtained sequences from 18 samples according to a previous study (Mu et al., 2016). Chimeric sequences were identified and removed using UCHIME (Edgar, 2010), and the high-quality sequences were clustered into operational taxonomic units (OTUs) with a cutoff of 97% similarity using UPARSE (version 7.1)<sup>1</sup> . Each OTU was annotated using the Ribosomal Database Project (RDP) classifier against the Silva (SSU119) 16S rRNA database at a confidence threshold of 90%. The diversity of cecal microbiota (such as rarefaction analysis, ACE and Chao1 richness estimators, and Shannon and Simpson diversity indices) was performed using Mothur (version 1.36.1) according to a previous study (Schloss et al., 2009). Beta diversity was evaluated by principal coordinate analysis (PCoA) based on the Bray–Curtis distance. An unweighted distance-based analysis of molecular variance (AMOVA) was conducted to compare the significant differences between samples by using Mothur (Schloss et al., 2009). Linear discriminant analysis effect size (LEfSe) analysis was employed to explore the significantly different bacteria at the OTU level among the three groups (Segata et al., 2011).

#### Quantification of Microbes by Real-Time PCR

To identify the quantitative changes in bacterial groups, several key bacteria groups, such as total bacteria, Firmicutes, Bacteroidetes, Clostridium cluster IV, Clostridium cluster XIVa, E. coli, Bifidobacterium, Lactobacillus, Prevotella, and Ruminococcus, in the ileal and cecal digesta were quantified by real-time quantitative PCR using specific primers (**Supplementary Table S1**). qPCR was performed by using TB GreenTM Premix Ex TaqTM (Takara Biotechnology, Dalian, China) on the CFX96 Real-Time PCR Detection System (Bio-Rad, Hercules, CA, United States). The real-time PCR mixtures and conditions were set according to previously described methods (Yu et al., 2017b; Song et al., 2019a).

<sup>1</sup>http://drive5.com/uparse/

#### Analysis of Cecal Microbial Metabolites

fmicb-10-02662 November 22, 2019 Time: 16:28 # 4

Cecal digesta samples were analyzed for microbial metabolites. Short-chain fatty acids (SCFAs) concentrations were measured by gas chromatography (GC) according to a method described previously (Yu et al., 2017b). Lactate concentration was determined using a commercial assay kit (Nanjing Jiancheng Biological Engineering Institute, Nanjing, China) according to the manufacturer's instructions. Ammonia concentration was analyzed using a UV spectrophotometer (Shimadzu, Tokyo, Japan) according to the description of previous study (Chaney and Marbach, 1962). Biogenic amine concentrations were measured using high-performance liquid chromatography (HPLC) with precolumn dansylation method as previously described (Yang et al., 2014). The concentrations of phenolic and indolic compounds were determined by HPLC according to a method that has been previously described (Schüssler and Nitschke, 1999).

#### RNA Extraction and qPCR for Ileal Gene Expression

The total RNA of ileal tissue was extracted using TRIzol reagent (Takara Biotechnology, Dalian, China) according to the method described by manufacturer's instructions. The RNA concentration and purity were quantified using NanoDrop 2000 spectrophotometer (Thermo Scientific, Wilmington, DE, United States), and the absorption ratio (OD260:OD280) of all samples ranged from 1.8 to 2.0. One microgram total RNA was reverse-transcribed using the PrimeScriptTM RT reagent Kit with gDNA Eraser (Takara Biotechnology, Dalian, China). The specific primer sequences used in the current study are presented in **Supplementary Table S2**. The target genes and GAPDH were determined by quantitative real-time PCR with TB GreenTM Premix Ex TaqTM (Takara Biotechnology, Dalian, China), and fluorescence was detected on CFX96 Real-Time PCR Detection System (Bio-Rad, Hercules, CA, United States). The reaction condition and real-time PCR condition were previously described (Yu et al., 2017a). GAPDH mRNA expression levels were used as a housekeeping gene. The results of target genes mRNA expression level calculated using the 2(−11Ct) method, where 11Ct = (Cttarget − CtGAPDH)treatment − (Cttarget − CtGAPDH)control.

#### Statistical Analysis

Experimental data were analyzed using the IBM SPSS statistics V20.0.0 software package (IBM Corp., Armonk, NY, United States). The Shapiro–Wilk test was used to assess whether all variables exhibited a normal distribution before assessing differences among the three groups. Then, the variables that showed a non-normal distribution were analyzed by Kruskal-Wallis one-way analysis of variance (ANOVA). To avoid type I errors, the resulting P values of bacterial abundance were adjusted via the Benjamini–Hochberg false discovery rate (FDR) multiple-testing correction (Benjamini and Hochberg, 1995). Bacterial abundance data are expressed as medians, and an FDR-adjusted P value < 0.05 was regarded as significant. The data of bacterial gene copy, microbial metabolites (SCFAs, lactate, and amino-acid derived metabolites), and ileal gene expression were analyzed via one-way ANOVA with a Tukey post hoc test. Differences were regarded significant at P < 0.05.

# RESULTS

### Growth Performance

During the whole trial, the outward appearance of broilers was healthy, and no mortality was observed. In the current study, dietary supplementation with CE significantly increased (P < 0.05) the average daily gain (ADG) of broilers (means ± SEM: 32.12 ± 0.35 and 34.59 ± 0.71 g/day in the control and CE groups, respectively) and decreased (P < 0.05) the feed conversion rate (F:G, means ± SEM: 2.57 ± 0.05 and 2.39 ± 0.04 in the control and CE groups, respectively) than did the control group. The ADG and F:G ratio in the antibiotic group showed no difference (P > 0.05) than those in the control and CE groups. Furthermore, no difference (P > 0.05) in the average daily feed intake (ADFI) was observed among the control group (means ± SEM: 82.61 ± 1.11 g/day), antibiotic group (means ± SEM: 83.30 ± 1.67 g/day), and CE group (means ± SEM: 82.43 ± 1.56 g/day).

#### Microbial Composition of Ileal and Cecal Digesta

To profile the microbial composition, the cecal microbiota of yellow-feathered broilers was analyzed by 16S rRNA MiSeq sequencing. In the current study, a total of 673,959 sequences from the 18 samples (with an average of 37,442 sequences per sample) after Illumina sequencing was revealed for the subsequent analyses. Mean-based rarefaction curves showed that the sampling of each group provided sufficient sequences to reflect the diversity and abundance of bacteria (**Supplementary Figure S1**). The ACE, Chao1 index, Shannon index, and Simpson index did not differ among the three different groups (**Supplementary Figure S2**). The PCoA with the Bray–Curtis distance indicated that the samples of antibiotic group gathered together and clearly separated from the samples of the control and CE groups (**Figure 1A**), suggesting that the microbial composition of yellow-feathered broilers in the antibiotic group differs from that of the control and CE groups. AMOVA also showed significant dissimilarities among the antibiotic, control, and CE groups (Fs = 2.84, Fs = 2.84, P < 0.001, among the control, antibiotic, and CE groups; Fs = 4.07, P < 0.01, antibiotic vs. CE; Fs = 3.07, P < 0.001, antibiotic vs. control; Fs = 1.27, P > 0.05, control vs. CE).

At the phylum level (**Figure 1B**), eight phyla were identified: Bacteroidetes, Firmicutes, Proteobacteria, Actinobacteria, Tenericutes, Verrucomicrobia, TM7, and Elusimicrobia in the cecal digesta of yellow-feathered broilers. Among these phyla, Bacteroidetes, Firmicutes, and Proteobacteria formed the three dominant phyla, contributing 59.61, 27.11, and 10.84% in the control group; 56.46, 30.86, and 11.01% in the antibiotic group; and 53.78, 33.47, and 10.07% in the CE group, respectively. There were no significant changes in the abundances of any phyla among the three groups (P > 0.05).

At the genus level, the 30 most predominant genera of the cecal digesta are presented in a heat map (**Supplementary Figure S3**). The nine dominant genera (those with a relative abundance ≥ 2% in at least one group) included Barnesiella, Alistipes, Staphylococcus, Bacteroides, Salmonella, unclassified Ruminococcaceae, unclassified vadinBB60, unclassified Enterobacteriaceae, and unclassified Lachnospiraceae. Compared with that in the control group, the antibiotic treatment significantly decreased the abundance of Barnesiella, Blautia, and unclassified S24-7 in the cecum of broilers (adjusted P < 0.05) (**Figure 2**). Compared with that in the CE group, antibiotic treatment decreased the abundance of Barnesiella, Blautia, and unclassified S24-7 (adjusted P < 0.05), whereas it increased the abundance of Alistipes, Bacteroides, and unclassified ML635J-40 (adjusted P < 0.05).

At the species level, 2,477 effective OTUs were identified in the cecal samples. To confirm specific bacteria that are characteristic for dietary treatment, LEfSe analysis was also performed at the species level (**Figure 3**). A total of 33 OTUs were different among the three groups. Among these different OTUs, seven OTUs were higher in the control group, 13 OTUs were higher in the antibiotic treatment group, and 13 OTUs were higher in the CE treatment group (adjusted P < 0.05).

Real-time PCR was used to identify the number changes of several key bacteria groups in the ileal and cecal digesta

of broilers following treatment with antibiotic and CE. In the ileum (**Figure 4A**), antibiotic treatment significantly decreased the number of total bacteria, Firmicutes, Lactobacillus, Ruminococcus, and Prevotella compared with those in the control group (P < 0.05), whereas it increased the number of E. coli (P < 0.05). Antibiotic treatment also decreased the numbers of Lactobacillus, Ruminococcus, Prevotella, and Bifidobacterium (P < 0.05) compared with those in CE group, whereas it increased the number of E. coli (P < 0.05). Additionally, CE treatment increased the number of Bifidobacterium compared with those in control group (P < 0.05), whereas it decreased the number of E. coli (P < 0.05). The numbers of Bacteroidetes, Clostridium cluster IV, and Clostridium cluster XIV has no significant difference among the three groups (P > 0.05). Compared with the control and CE groups, antibiotic treatment significantly decreased the number of total bacteria and Lactobacillus in the cecum (**Figure 4B**) (P < 0.05), whereas it increased the number of E. coli (P < 0.05). Moreover, CE treatment increased the number of Lactobacillus compared with that in the control group (P < 0.05). However, there were no significant differences (P > 0.05) observed in the numbers of Firmicutes, Bacteroidetes, Ruminococcus, Prevotella, Clostridium cluster IV, Clostridium cluster XIV, and Bifidobacterium among the different dietary treatments. Overall, these results indicated that antibiotic and CE treatments significantly altered the intestinal bacterial community structure and the numbers of individual microbes.

#### Microbial Metabolites in the Ileal and Cecal Digesta

Lactate and SCFAs are the major carbohydrate fermentation products of gut microbes and serve as indicators of microbial activity. For lactate, dietary CE supplementation significantly increased the concentration of lactate compared with that in control and antibiotic groups in the ileum (**Figure 5A**) and cecum (**Figure 5C**, P < 0.05). Antibiotic treatment significantly decreased the lactate concentration in the cecum compared

with that in the control group (P < 0.05). For SCFAs, dietary supplementation with CE significantly increased total SCFA, acetate, and butyrate concentrations in the ileum (**Figure 5B**) and cecum (**Figure 5D**) compared with those in control and antibiotic groups (P < 0.05), whereas it decreased branchedchain fatty acids (BCFAs) and isovalerate concentrations in the cecum compared with those in the antibiotic group (P < 0.05). Furthermore, antibiotic treatment increased the concentrations of BCFA and isovalerate compared with those in the control group (P < 0.05). However, the concentrations of propionate, valerate, and isobutyrate were not affected by different dietary treatments (P > 0.05).

Ammonia and biogenic amines are amino acid deamination and decarboxylation products of gut microbes, respectively, and phenolic and indolic compounds are a product of aromatic amino acid degradation by gut microbes. For ammonia (**Figure 6A**), dietary CE supplementation decreased the concentration of ammonia compared with that in the control group (P < 0.05), whereas no significant differences were observed between the control and antibiotic groups (P > 0.05). For phenolic and indolic compounds (**Figure 6B**), CE supplementation decreased the concentration of p-cresol compared with that in the control and antibiotic groups (P < 0.05) and decreased the concentration of indole compared with that in the control group (P < 0.05). Dietary treatment showed no effect on the concentrations of phenol and skatole (P > 0.05). For amines (**Figure 6C**), cadaverine and spermidine were the major amines in the cecal digesta of broilers. Dietary supplementation with CE significantly decreased total amines, spermidine, methylamine, and tyramine concentrations compared with those in the control group (P < 0.05) and decreased total amines, spermidine, methylamine, putrescine, and tryptamine concentrations compared with those in the antibiotic group (P < 0.05). Furthermore, antibiotic treatment increased putrescine and tryptamine concentrations

compared with those in the control group (P < 0.05). However, the concentrations of cadaverine and spermine were not changed by different dietary treatments (P > 0.05). Overall, the results of the current study indicate that dietary supplementation with CE markedly enhanced the bacterial fermentation of carbohydrates and decreased the bacterial fermentation of protein in the cecum, whereas antibiotic treatment increased protein fermentation.

#### Gene Expression in the Ileal Tissue

The alteration of the gut microbial community and their metabolites could regulate intestinal epithelial gene expression. Thus, the mRNA expressions of genes involved in toll-like receptor (TLR), cytokines production, and mucosal defense were analyzed in the ileum. As shown in **Figure 7**, antibiotic treatment significantly up-regulated the relative mRNA expression of TLR4 and its downstream signal response genes (MyD88 and NF-κB) compared with those in the control and CE groups (P < 0.05). Dietary supplementation with CE down-regulated the relative mRNA expression of TNFα compared with that in the antibiotic group (P < 0.05), whereas it up-regulated IL-10, ZO-1, and Claudin expression (P < 0.05). Moreover, CE treatment also upregulated ZO-1 and Claudin expression compared with that in the control group (P < 0.05). There was no significant difference in the expression of IL-8, IL-1β, IFN-γ, occludin, and MUC2 among the three groups (P > 0.05).

#### Correlation Analysis Between Mucosal Gene Expression With Ileal Microbes or Their Metabolites

A Pearson correlation analysis was carried out to determine whether there was any relationship among mucosal gene

expression and main microbial numbers and the concentrations of metabolites (**Figure 8**). Correlation analysis revealed that the mRNA expression level of TLR-4 and NF-κB were positively correlated with the number of E. coli (P < 0.05), whereas it negatively correlated with the concentrations of acetate and total SCFAs, and the number of Lactobacillus, Bifidobacterium, and Prevotella (P < 0.05). The MyD88 mRNA expression level was positively correlated with the number of E. coli (P < 0.05), whereas it negatively correlated with the concentrations of acetate and total SCFAs, and the number of Lactobacillus (P < 0.05). The mRNA expression level of TNFα was negatively correlated with the concentrations of acetate, total SCFAs, and lactate, and the number of Bifidobacterium (P < 0.05). The mRNA expression level of IL-10 and ZO-1 was positively correlated with the concentrations of acetate, total SCFAs, and lactate, and the number of Bifidobacterium (P < 0.05), and the ZO-1 expression level also positively correlated with the butyrate concentration (P < 0.05), whereas the IL-10 expression level negatively correlated with the number of E. coli (P < 0.05). Meanwhile, the Claudin mRNA expression level was positively correlated with the acetate and butyrate concentrations (P < 0.05), whereas it negatively correlated with the number of E. coli (P < 0.05). Additionally, the correlation between microbes and

the concentrations of metabolites in the cecum was also analyzed and is shown in **Supplementary Figure S4**. Overall, these results indicated that the alteration in the ileal digesta microbiota and metabolites was correlated with the changes of epithelial gene expression in broilers.

#### DISCUSSION

The present study employed a whole growth stage continuous feed strategy to evaluate the effects of dietary supplementation with CE on the microbial community, microbial metabolite profiles, and expression of immune-related genes in the intestine of yellow-feathered broilers. The results showed that dietary supplementation with CE dramatically increased the number of Bifidobacterium and up-regulated the mRNA expression of intestinal barrier genes (ZO-1 and Claudin) in the ileum, whereas it decreased the number of E. coli. Meanwhile, CE supplementation increased the number of Lactobacillus and the concentrations of lactate and SCFAs in the cecum, whereas it decreased the concentrations of protein fermentation products (ammonia, p-cresol, indole, total amines, spermidine, methylamine, and tyramine). These findings further highlight

the key role of CE in regulating intestinal microbiota, metabolic profiles, and mucosal immune system, suggesting that CE may act as an efficient alternative of antibiotics for yellow-feathered broiler production.

# Citrus Extract Altered the Intestinal Microbiota of Yellow-Feathered Broilers

The gastrointestinal microbes of mammals play an important role in prevention of infectious diseases, regulation of nutrient digestion and metabolism, maintenance of intestinal morphology, and immune homeostasis of the host (Nicholson et al., 2005; Hooper et al., 2012; Song et al., 2019b). Applying functional substances to the diet of animals is an advantageous strategy for the modulation gastrointestinal microbiota and maintenance of host health. Citrus products have antimicrobial activity against E coli and S. typhimurium in poultry production and have a positive effect on intestinal health (Nannapaneni et al., 2008; O'Bryan et al., 2008). In the current study, we found that dietary supplementation with CE selectively regulated intestinal

short-chain fatty acid.

fmicb-10-02662 November 22, 2019 Time: 16:28 # 10

microbiota, including stimulating bacterial species with beneficial function (Bifidobacterium and Lactobacillus) and inhibiting the number of potential pathogen (E. coli). Bifidobacterium is recognized as a beneficial bacterium and potential probiotic, and a number of species can produce acetate and lactate (Gibson et al., 2004) and have a positive effect on intestinal health of both humans and animals and offer the ability to normalize the ratio of pro-inflammatory and anti-inflammatory cytokines (O'Mahony et al., 2005). Lactobacillus is well known as a potentially beneficial species in the intestine, which can inhibit the colonization of potential pathogenic groups by competing with the epithelial binding sites and nutrients, and the productions of antimicrobial factors including lactate and bacteriocins, thus maintaining the homeostasis of the intestinal environment of host (Yu et al., 2018). E. coli is an opportunistic pathogenic bacteria that has been shown to be positively associated with numerous infections and the development of diseases, including bacillary dysentery or colitis disease (Barnich et al., 2007). Thus, these results indicate that dietary inclusion of CE may be beneficial for the intestinal health for broilers.

To date, the research for the effect of CE on intestinal microbiota is still limited. The beneficial functions of plant extract mainly depended on their specific bioactive components (such as organic acids, polysaccharide, and flavone), which can synthesize as antimicrobial agents against microbial infection and alter their composition (Lillehoj et al., 2018). In our study, CE contained plentiful contents of total flavone (about 2.48%), which can target and modulate microbiota composition. Dietary flavonoid intake significantly increased the abundance of Bifidobacterium in humans (Klinder et al., 2016) or mice (Espley et al., 2013). CE also has a high concentration of polysaccharide (about 1.2% in the current study), which has the capability to change the composition and diversity of intestinal microbiota, such as increase the abundance of Lactobacillus in mice (Li et al., 2019). Additionally, CE is also a potential source of organic acids (including 1.3% citric acid and 0.68% chlorogenic acid), which can decrease the pH in the gastrointestinal tract and then inhibit the growth of some pathogenic bacteria (such as E. coli) owing to their susceptibility to low pH (Feng et al., 2018; Zhang Y. et al., 2018). Thus, the intestinal microbiota alteration of broilers in response to administering CE may be attributed to the flavonoid, polysaccharide, and/or organic acid contents, whereas the underlying mechanisms should be further studied.

# Citrus Extract Increased Microbial Fermentation of Carbohydrate but Decreased Fermentation of Protein

The microbial metabolite profiles in the intestinal digesta can reflect the microbial activity and intestinal health. SCFAs and lactate are primarily fermentation products of the carbohydrate metabolism by bacteria in the gut. In the current study, CE significantly increased the concentrations of lactate, total SCFA, acetate, and butyrate in the cecum, which suggests that CE increased the carbohydrate fermentation by bacteria. Bifidobacterium and Lactobacillus are the main acetate, butyrate, and lactate producers in the gut (Gao et al., 2018). Increase in SCFAs and lactate concentrations was accompanied by an increase of the number of Bifidobacterium and Lactobacillus. Thus, one potential explanation for increased cecal digesta SCFAs and lactate concentrations may be due to increase the abundance or numbers of SCFA- and lactate-producing bacteria. SCFAs and lactate can exert many beneficial effects for host health. Acetate can be used as energy substrate of peripheral tissues, butyrate exerts an anti-inflammatory function and is the main energy source of colonic epithelial cells (Yu et al., 2018), and lactate can reduce the pH value in the gastrointestinal tract and inhibit the multiplication of pathogens that invade the gut (such as E. coli) (Gao et al., 2018; Yu et al., 2018). Thus, the increase of total SCFAs, acetate, butyrate, and lactate concentrations

in the current study suggests a host-friendly gut environment after CE treatment.

Nitrogenous compounds, such as undigested protein and amino acids, can also be fermented by intestinal bacteria and form putrefactive catabolites, such as ammonia, biogenic amines, phenol, and indole compounds. In the current study, the decrease in cecal ammonia, biogenic amines, phenol, and indole levels suggests that bacterial deamination and decarboxylation of amino acids were affected after CE administration. A high ammonia level has been shown to exert a negative effect on the health of the hosts, such as inhibiting the growth and differentiation of intestinal epithelial cells and increasing the incidence of diarrhea in the host (Dong et al., 1996). Additionally, high concentrations of biogenic amines (tyramine), indole, and p-cresol also exert adverse impacts on gut health and are regarded as co-carcinogens and colon cancer promoters (Nowak and Libudzisz, 2006; Davila et al., 2013). Thus, the decrease in the concentrations of ammonia, biogenic amines, indole, and phenolic compounds via CE supplementation may have a beneficial effect on gut health. Overall, together with the increase in lactate and SCFAs, these findings indicate that CE changed the microbial metabolic activity, increased microbial fermentation of carbohydrate, and decreased microbial protein catabolism.

#### Citrus Extract Affected the Ileal Mucosal Response Involved in the Intestinal Barrier Function

The intestinal barrier performs the essential function of defense against the passage of pathogenic agents and luminal antigens into the gastrointestinal epithelium while enabling the acquisition of dietary nutrients (Broom, 2018). In the present study, CE supplementation up-regulated the mRNA expression of intestinal barrier genes in the ileum compared with that in control and antibiotic groups, such as ZO-1 and Claudin. This suggests that CE may improve the integrity of the intestinal epithelium, consequently generating a hostfriendly gut environment, which could help defend against pathogen infection. The changes in intestinal microbes and associated metabolites can both positively and negatively affect the histological function of the gastrointestinal epithelium, such as intestinal barrier permeability. Previous studies have shown that a number of pathogens, such as the Enterotoxigenic E. coli K88, can impair intestinal barrier functions by downregulating ZO-1 and occludin gene and protein expression (Wang et al., 2017), whereas Lactobacillus reuteri LR1 up-regulated the expression of ZO-1 and occludin gene in piglets (Yi et al., 2018). Moreover, some bacterial fermentation product, such as butyrate, can maintain gut integrity. As described above, dietary inclusion of CE increased the numbers of Bifidobacterium and Lactobacillus and the concentrations of total SCFAs, acetate, and butyrate and decreased the number of E. coli. Our previous study also revealed that the Bifidobacterium and Lactobacillus numbers and butyrate concentration were positively correlated with mRNA levels of ZO-1 and occludin (Yu et al., 2019). Therefore, we might speculate that the high number of Bifidobacterium and Lactobacillus, SCFAs, and butyrate concentrations in the CE group may be the factors accounting to the up-regulation of intestinal barrier gene expression, although the underlying mechanism required further clarification.

#### Antibiotic Treatment Induced Intestinal Microbiota Dysbiosis, Altered Fermentation Profiles of Microbial Metabolism, and Disturbed the Innate Immune Homeostasis

Antibiotics can cause gut microbiota dysbiosis, inhibit the innate immune defenses, and lead to increased pathogen colonization and disease susceptibility (Zhang C.J. et al., 2018). In the current study, antibiotic treatment increased the abundance and the number of potential pathogens, while decreasing those with beneficial function, which was consistent with previous studies (Yu et al., 2017a, 2018). Among the affected bacterial groups, many bacterial species have previously been related to increased disease risk in humans or animals. For example, antibiotic treatment decreased the abundances of Barnesiella and Blautia, and the numbers of Lactobacillus, Ruminococcus, Prevotella, and Bifidobacterium, and increased the abundance of Alistipes and the number of E. coli. Barnesiella and Blautia are SCFA producers (Zhang et al., 2012) and have the ability to protect the integrity of the intestinal barrier function and alleviate dextran sulfate sodium-induced inflammation (Lu et al., 2018). Ruminococcus and Prevotella are key bacterial groups for the degradation of dietary fibers and polysaccharide to regulate the host metabolism (Wang et al., 2018). Alistipes can directly elevate inflammation levels and induce mucosal injuries in goats (Ye et al., 2016). The potentially beneficial function of Lactobacillus and Bifidobacterium and the potential adverse function of E. coli have been mentioned above. Taken together, these results point to a potentially detrimental impact of antibiotic treatment on the composition of intestinal microbiota in broilers.

Corresponding to the alteration of intestinal microbiota, antibiotics showed a marked impact on the microbial metabolism of carbohydrate and amino acids in the cecum, as indicated by the decrease in the concentrations of lactate and most SCFAs, and the increase in the concentrations of BCFA, most amines, and p-cresol. This is consistent with previous studies (Gao et al., 2018; Pi et al., 2018; Yu et al., 2018), which observed that antibiotic treatment led to a decrease in most SCFA concentrations, whereas it increased amine concentrations in pigs. Overall, the findings of this study provide clear evidence for a shift of the microbial metabolic activity, with lower microbial carbohydrate fermentation and higher microbial catabolism of amino acids after antibiotic treatment.

The alteration of intestinal microbes and their metabolites could regulate the immune homeostasis of intestinal mucosa. In the present study, we found that antibiotic treatment upregulated the mRNA expression of TLR4, NF-κB, MyD88, and pro-inflammatory cytokines TNFα. TLR4 can transfer signals to NF-κB via MyD88-dependent pathway and then induce the activation of pro-inflammatory cytokines (Zhang et al., 2017).

Thus, the increased gene expression levels of pro-inflammatory cytokines after antibiotic treatment may be attributed to an up-regulation of the TLR4-MyD88-NF-κB signaling pathway of broilers. Collectively, together with the alteration of intestinal microbiota, fermentation profiles of microbial metabolism, and the mucosal gene expression levels, our results indicate that antibiotic treatment induced the gut microbiota dysbiosis, altered the microbial metabolism, and inhibited the innate immune defenses, likely toward a host-adverse gut environment.

#### CONCLUSION

In conclusion, this study demonstrated that dietary supplementation with CE changed the intestinal microbial composition, microbial metabolite profiles, and immune status of broilers, likely toward a host-friendly gut environment. Intestinal microbes, such as the numbers of Bifidobacterium and Lactobacillus, increased after CE supplementation, whereas the number of E. coli decreased. Meanwhile, CE markedly increased the concentrations of lactate and SCFAs, whereas it decreased the concentrations of amino acid fermentation products. The expression of intestinal barrier genes (ZO-1 and Claudin) was increased. However, antibiotic treatment induced gut microbiota dysbiosis, altered the microbial metabolism, and disturbed the innate immune homeostasis, likely toward an unhealthy gut environment. These findings suggest that CE may act as an efficient antibiotic alternative for yellow-feathered broiler production.

#### DATA AVAILABILITY STATEMENT

The datasets generated for this study can be found in the following repository. The raw reads in this study were submitted to the National Center of Biotechnology Information (NCBI) Sequence Read Archive (SRA) database under accession numbers SRR9074899–SRR9074916.

#### REFERENCES


# ETHICS STATEMENT

The experimental proposals and procedures for the care and treatment of the broilers were approved by the Animal Care and Use Committee of the Guangdong Academy of Agricultural Sciences (authorization number GAASIAS-2016-017).

#### AUTHOR CONTRIBUTIONS

MY, WC, and XM conceived and designed the whole trial. ZL and YC conducted the broiler trials. MY and ZL conducted laboratory analyses. MY, XM, and GW wrote the manuscript.

#### FUNDING

This work was supported by the National Key Research and Development Program of China (2016YFD0501210), the Presidential Foundation of the Guangdong Academy of Agricultural Sciences (201802B), Talent Project of the Guangdong Academy of Agricultural Sciences (201803), Guangdong Modern Agro-Industry Technology Research System (2019KJ115), and Science and Technology Planning Project of Guangdong Province (2019A050505007).

# ACKNOWLEDGMENTS

The authors would like to thank Dr. Dun Deng, Dr. Zhichang Liu, Mr. Ting Rong, Ms. Zhimei Tian, and Ms. Huijie Lu for their help in the successful completion of this study.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.02662/full#supplementary-material


the gut microbiota and consequences for the host". Pharmacol. Res. 68, 95–107. doi: 10.1016/j.phrs.2012.11.005


diet compared with a normal-protein diet. J. Nutr. 146, 474–483. doi: 10.3945/ jn.115.223990


fmicb-10-02662 November 22, 2019 Time: 16:28 # 13


different dietary protein levels. J. Anim. Sci. Biotechnol. 8:60. doi: 10.1186/ s40104-017-0192-2


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Yu, Li, Chen, Wang, Cui and Ma. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fmicb-10-02662 November 22, 2019 Time: 16:28 # 14

# Effect of Diet on the Enteric Microbiome of the Wood-Eating Catfish *Panaque nigrolineatus*

*Ryan C. McDonald1 , Joy E. M. Watts2 and Harold J. Schreier1,3 \**

*1Department of Biological Sciences, University of Maryland, Baltimore County, Baltimore, MD, United States, 2Department of Biological Sciences, University of Portsmouth, Portsmouth, United Kingdom, 3Department of Marine Biotechnology, University of Maryland, Baltimore County, Baltimore, MD, United States*

Wood is consistently found in high levels in the gastrointestinal tract of the Amazonian catfish *Panaque nigrolineatus*, which, depending on environmental conditions, can switch between xylivorous and detritivorous dietary strategies. This is highly unusual among primary wood consumers and provides a unique system to examine the effect of dietary change in a xylivorous system. In this study, microbiome and predictive metagenomic analyses were performed for *P. nigrolineatus* fed either wood alone or a less refractory mixed diet containing wood and plant nutrition. While diet had an impact on enteric bacterial community composition, there was a high degree of interindividual variability. Members of the Proteobacteria and Planctomycetes were ubiquitous and dominated most communities; Bacteroidetes, Fusobacteria, Actinobacteria, and Verrucomicrobia also contributed in a tissue and diet-specific manner. Although predictive metagenomics revealed functional differences between communities, the relative abundance of predicted lignocellulose-active enzymes remained similar across diets. The microbiomes from both diets appeared highly adapted for hemicellulose hydrolysis as the predicted metagenomes contained several classes of hemicellulases and lignin-modifying enzymes. Enteric communities from both diets appeared to lack the necessary cellobiohydrolases for efficient cellulose hydrolysis, suggesting that cellobiose is not the primary source of dietary carbon for the fish. Our findings suggest that the *P. nigrolineatus* gut environment selects for an enteric community based on function, rather than a vertically transferred symbiotic relationship. This functional selection strategy may provide an advantage to an organism that switches between dietary strategies to survive a highly variable environment.

Keywords: lignocellulose digestion, microbiome, 16S rRNA gene amplicon sequencing, predictive metagenomics, Amazonian catfish

#### INTRODUCTION

The Amazonian catfish *Panaque nigrolineatus* consumes large quantities of wood as part of its diet, a trait shared among a limited number of related fish species (Schaefer and Stewart, 1993). Depending on environmental conditions, *P. nigrolineatus* can adjust its feeding behavior, switching between xylivorous ("wood-eating"), detritivorous, and herbivorous feeding behaviors.

#### *Edited by:*

*Jie Yin, Institute of Subtropical Agriculture (CAS), China*

#### *Reviewed by:*

*Weida Gong, University of North Carolina at Chapel Hill, United States Woo Jun Sul, Chung-Ang University, South Korea*

> *\*Correspondence: Harold J. Schreier schreier@umbc.edu*

#### *Specialty section:*

*This article was submitted to Microbial Symbioses, a section of the journal Frontiers in Microbiology*

*Received: 14 August 2019 Accepted: 05 November 2019 Published: 29 November 2019*

#### *Citation:*

*McDonald RC, Watts JEM and Schreier HJ (2019) Effect of Diet on the Enteric Microbiome of the Wood-Eating Catfish Panaque nigrolineatus. Front. Microbiol. 10:2687. doi: 10.3389/fmicb.2019.02687*

**566**

While jaw and tooth morphologies appear to be well adapted for wood consumption (Geerinckx et al., 2007; Adriaens et al., 2008), other features such as GI tract length, microvilli surface area, and gut retention times are inconsistent with a diet dependent primarily on wood consumption (German, 2009; German et al., 2010). It has been hypothesized that wood intake may serve a selective advantage during the Amazonian dry season (Araujo-Lima et al., 1986), or a consequence of epiphyte or fungal hyphae consumption. The nutritional benefits of wood feeding and subsequent digestion by *P. nigrolineatus* are, as yet, unknown.

Consuming wood as a primary food source poses many challenges. The physical and chemical properties of most woods make them highly recalcitrant and of poor nutritive value (Pu et al., 2011). Although rich in carbohydrates, these compounds are largely inaccessible due to their incorporation into structural polysaccharides, such as cellulose and hemicelluloses (Cosgrove, 2005). The crystalline nature of cellulose and presence of other structural plant cell wall polymers, such as lignin, limit microbial infiltration and exclude water, making the environment non-conducive to enzymatic hydrolysis (Cosgrove, 2005; Grabber, 2005). In addition to these challenges, wood is also nitrogen deficient; the nitrogen content of Amazonian woods is regularly below 0.5% (Martius, 1992). To overcome this deficiency, many wood-feeding organisms will also consume non-wood nutrient sources or rely on the activity of enteric microorganisms to supply reduced nitrogen compounds (Raychoudhury et al., 2013; Gruninger et al., 2016; Mikaelyan et al., 2017). With a few exceptions, nutrient acquisition by wood-eaters is mediated by their enteric microbial community, which liberates assimilatory sugars and generates nitrogenous compounds (Tokuda et al., 2014; Cragg et al., 2015).

The enteric bacterial community of *P. nigrolineatus* has been shown to possess a unique microbiome, with the potential of assisting lignocellulose degradation and conducting biological nitrogen fixation (McDonald et al., 2012). Included in these communities are several species of Rhizobiales, Flavobacteriales, Legionellales, Burkholderiales, and Clostridiales. Distinct communities have been identified in the fore, mid, and hindguts of the fish despite any well-defined anatomical features (e.g., sphincters or cecum) demarcating these regions. A diverse and distinct fungal community also resides and is associated with cellulose degradation in the GI tract (Marden et al., 2017). Culture-based analyses and biochemical tests confirmed the presence of a lignocellulolytic and diazotrophic community (Watts et al., 2013; McDonald et al., 2015). These analyses suggest that a considerable amount of microbial metabolic cross-feeding may be occurring within the fish GI tract, where carbohydrates are liberated by cellulolytic species and consumed by non-cellulose utilizers. In comparison to the cellulolytic community, the enteric diazotrophic community was less diverse and was comprised of known nitrogen-fixing Rhizobiales and *Clostridium* (McDonald et al., 2015).

Diet has been shown to have a major impact on enteric microbial communities of animals (David et al., 2014; Ringø et al., 2016). Many xylivores are highly adapted for wood consumption, but retain some capacity to survive on less refractory diets (Tanaka et al., 2006; Miyata et al., 2014). However, the enteric communities are specialized and necessary for the processing of digesta; their manipulation often has deleterious effects on the host (Rosengaus et al., 2011). For some wood-feeding organisms, the enteric communities appear highly stable, where dietary changes have minimal impact on community composition (Boucias et al., 2013; Wang et al., 2016). Dual feeding behavior in adult fish is not unique to *P. nigrolineatus*, but far less common to the ontogenetic diet shifts seen in other fish species (Bledsoe et al., 2016; Sánchez-Hernández et al., 2019). To examine the effects of diet on a primarily wood-feeding fish species, we characterized the enteric microbiomes of animals fed either a wood or mixed diet.

#### MATERIALS AND METHODS

#### Animals, Experimental Design, and Tissue Sampling

*Panaque nigrolineatus* (L190) were acquired through aquarium wholesalers. Fishes were randomly assigned to aerated, filtered tanks (29 ± 1°C), and acclimated for 3 weeks on an *ad libitum* mixed diet of red palm (*Cocos nucifera*), hearts of palm (*Euterpe precatoria*), and sinking algae wafers (Hikari, Hayward, CA). After acclimation, fishes were randomly assigned to either a mixed or wood diet for 6 weeks. Mixed diet fishes were provided algae wafers and hearts of palm daily with continuous access to wood. Wood-fed fishes were provided with wood exclusively. All wood was autoclaved twice prior to being placed in the tanks. To inhibit algae growth, all fishes were reared under low/no light conditions. Lights were turned on approximately 30 min each day for tank maintenance and feeding. Two independent feeding studies were carried out, designated feeding study 1 (two fishes, designated 1X and 1W, for mixed (X)-and wood (W)-fed diets, respectively) and feeding study 2 (six fishes, designated 2X, 2W, 3X, 3W, 4X, and 4W). Both feeding studies were conducted under identical experimental conditions.

At the conclusion of the feeding experiments, fishes were euthanized *via* an overdose of the anesthetic 3-aminobenzoic acid ethyl ester (MS-222, 50 mg/L) and were immediately transferred to a chilled dissecting tray where the ventral body plate was removed. The body cavity was filled with cold, sterile phosphate buffered saline (PBS) to facilitate removal of the GI tract. Once uncoiled, the intestines were disconnected by cutting at the anus and just posterior to the stomach, and measured for length. The intestine was divided into three equal lengths demarcating the fore (F), mid (M), and hindgut (H).

#### DNA Extraction, Amplification, and Sequencing

GI tract samples were processed for fish from each diet using the Qiagen (Germantown, MD, USA) DNeasy Blood and Tissue Kit with pre-treatments for Gram-positive and Gramnegative bacteria according to the manufacturer's instructions. DNA was extracted from three samples of each GI tract region for each fish and was pooled and processed for PCR amplification. To profile the bacterial community, 16S rRNA gene sequencing libraries were prepared according to the manufacturer's instructions (Illumina, San Diego, CA). Briefly, the V3-V4 region of the 16S rRNA gene was amplified using the primer pair evaluated previously (Klindworth et al., 2017) using the 2X KAPA HiFi HotStart ReadyMix (Sigma-Aldrich, St. Louis, MO) with the following polymerase chain reaction (PCR) program parameters: an initial denaturation step of 3 min at 95°C followed by 25 cycles of denaturation for 30 s at 95°C, annealing for 30 s at 55°C, and elongation for 30 s at 72°C, followed by a final elongation for 5 min at 72°C. Index PCR was performed using the Nextera XT Index Kit according to the manufacturer's instructions (Illumina, San Diego, CA). PCR products purified using AMPure XP beads (Beckman Coulter, Brea, CA), pooled in equimolar amounts, and sequenced using the Illumina MiSeq platform (250 bp paired-end reads). For the second feeding study, the standard Illumina primers for V3-V4 were modified to include a unique trinucleotide sequence (**Table 1**) between the overhang adapter and the 16S rRNA primer, which allowed for double-dual indexing of the samples (Fadrosh et al., 2014). A total of 10 samples were pooled in equimolar amounts prior to the standard index PCR reaction.

#### DNA Sequence Processing and Community Analysis

Raw reads were preprocessed using CLC Workbench (version 9) (Qiagen). Adapter sequences were removed and read pairs were quality trimmed (qual. limit = 0.05; ambiguous nucleotide maximum = 2; minimum sequence length = 100 bp) and merged (mismatch cost = 2; gap cost = 3; maximum unaligned = 0; minimum score = 8). Sequences were analyzed using the Quantitative Insights Into Microbial Ecology (QIIME) bioinformatics pipeline (Caporaso et al., 2010). Operational taxonomic units (OTUs) were picked using the open reference method against the Silva\_132 database (minimum OTU cluster size = 2; OTU similarity = 0.97) (Desantis et al., 2006). Taxonomies were summarized at multiple levels (L2-L6) using the *summarize\_ taxa.py* script. Rarefaction plots and alpha diversity measures were calculated using the *alpha\_rarefaction.py* and *alpha\_diversity. py* scripts, respectively. OTU matrices were normalized using DESeq2 variance stabilizing transformation prior to Bray-Curtis distance matrix generation. Principle coordinate analysis (PCoA) plots were generated using *principal\_cordinates.py*. OTU networks were generated for all tissue samples using the *make\_otu\_network. py* script. For this analysis, OTUs were re-picked according to the above method; however, the minimum cluster size was increased to 25 in order to reduce the number of nodes. The


TABLE 1 | List of modified Illumina V3-V4 primers and resulting read counts from the second feeding study 16S rRNA gene survey.

*The primers were modified to include a unique trinucleotide sequence between the 16S rRNA primer and overhang adapter. Successful amplification from the sample is indicated by (+).*

network was visualized using Cytoscape v3.2.1 using an edgeweighted spring-embedded layout.

#### Predictive Functional Profiling of Microbial Communities

Predictive functional profiling was performed using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) (Langille et al., 2013). For the PICRUSt analysis, all samples were normalized to 80,000 reads/sample prior to OTU picking. OTUs were picked using the closed reference method against gg\_13\_8 (minimum OTU cluster size = 2; OTU similarity = 0.94). The resulting BIOM table was normalized for 16S rRNA gene copy number prior to predicting functions for metagenomes. For this analysis, KEGG orthologs (KO) were recorded. KOs were collapsed into pathways (L1-L3) using the *categorize\_by\_function.py* script. The results were visualized in Rstudio (v 0.98.1083) using the heatmap.2 function of the gplots package. To determine which OTUs and samples were contributing particular functions, *metagenome\_contributions.py* was also run. To determine if functional profiles differed across tissue region or diet, principal component analysis (PCA) was performed on the predicted metagenome (not collapsed within the KEGG hierarchy) using the ggplot2 package in Rstudio.

#### RESULTS

#### Microbial Community Characterization

Microbial communities were analyzed from GI tract regions of wood-fed and mixed diet-fed fish by next generation 16S rRNA amplicon sequencing, and functional gene profiles were then extrapolated using *in silico* methods. Except for foregut samples from two wood-grown and one mixed-diet fish, which could not be amplified by PCR, more than 1.2 × 106 high quality reads were generated across all samples and redistributions are shown in **Table 1**. Rarefaction analysis (**Supplementary Figure S1**) and alpha diversity measures (**Table 2**) showed that the bacterial communities were sufficiently sampled, and further sequencing would be unlikely to significantly increase the observed microbial diversity detected. Species richness varied greatly between feeding study replicates, with a marked reduction in diversity observed in the second feeding experiment. Despite this, Good's coverage estimates remained high (>0.95 for all samples).

#### Effect of Diet on Enteric Microbial Community Composition

All microbial communities analyzed were dominated by a few microbial phyla (**Figure 1**). These included Proteobacteria (19–99%), Planctomycetes (<1–66%), Fusobacteria (0–43%), Bacteroidetes (<1–26%), and Actinobacteria (<1–23%). While the relative abundance of these major phyla varied greatly across diet, tissue region, and feeding study replicate, the constituent members of these phyla largely remained unchanged (**Supplementary Table S1**). Planctomycetes were observed in high abundance in all the first experimental study samples, as well as the foreguts of mixed diet fish from the second feeding study; the most predominant classes were Pirellulales, Gemmatales, Planctomycetales, and Isosphaerales. Sequences with the highest sequence similarity with an unidentified *Cetobacterium* species in the class Fusobacteria were observed in the mid and hindgut of a single mixed (X) diet fish (2XM and 2XH) as well as the hindgut of a wood-fed (W) fish (1WH).

TABLE 2 | Observed OTUs, Chao1, and Good's coverage values were calculated to compare bacterial diversity among diets and tissue regions.


*All calculations were performed using QIIME with 0.97 OTU identity threshold.*

In the first experiment, Bacteroidetes were predominant in all tissue regions of the mixed diet fish, as well as the hindgut of the wood-fed fish. While the diversity of Bacteroidetes was high, most sequences had high sequence similarity to the endosymbiont *Candidatus* Cardinium. However, in both experiments, Actinobacteria was the most common phyla, found in nearly all tissue regions across feeding regimens and feeding study replicates. The highest proportions of Actinobacteria were identified in the mixed diet fish of the second feeding experiment and had high sequence similarity to the Corynebacteriales genera *Mycobacterium* and *Gordonia*.

Both the Proteobacteria and Verrucomicrobia populations showed marked composition differences between the two feeding study replicates (**Figure 1**). The Proteobacteria identified in the first experiment were predominantly Alphaproteobacteria, consisting of Rhizobiales and Rhodobacterales. The Proteobacteria from the remaining enteric samples of the second experiment were almost exclusively Gammaproteobacteria and consisted of Aeromonadales. Distinct differences in Proteobacteria were also identified in the tank water; sequences most closely aligned with Rhodospirillales were foremost in mixed diet tank water, while Sphingomonadales-like sequences dominated wood diet tank water. Verrucomicrobia was identified predominantly in the mixed diet-fed fish of both feeding studies. In the first experiment, the sequences had highest similarity to an uncultivated Verrucomicrobiales and *Luteolibacter* species, contrasting to the second experiment where the sequences were most similar to either a species of *Chthoniobacter* or *Prosthecobacter*.

Trends in community composition were detected; however, diet did not appear to select for specific microbial communities in any tissue region of the *P. nigrolineatus* GI tract. PCoA and UPGMA trees based on Bray-Curtis distances showed samples largely clustered based on experimental study and diet (**Figures 1, 2**). Feeding experiment 1 and 2 were distinguished along PC1 of the PCoA. UPGMA trees revealed that within experimental study groups, wood and mixed diet samples largely formed monophyletic clades. The exception to this is sample 3XH, which formed a paraphyletic group with the other mixed diet samples from the second experimental feeding group.

*P. nigrolineatus* possesses a taxonomically restricted core microbiome. Fifty-five OTUs were identified in >70% of samples across both feeding regimens (**Supplementary Table S2**). The majority of these OTUs had high sequence similarity to species of Pirellulales, Rhizobiales, Rhodobacterales, and Aeromonadales, and their relative abundance was highly variable. For this analysis, OTU relative abundance was not used to define the core microbiome (e.g., only including OTUs that represent >1% of all reads). Distinct core microbiomes were observed between wood and mixed diet-fed fish (**Supplementary Tables S3, S4**). Wood-fed fish had an expanded core microbiome (115 OTUs) relative to mixed diet-fed fish (72 OTUs) (**Supplementary Figure S2**), with minimal overlap between the two groups (~21%) (**Supplementary Figure S3**). Despite,

abundance of predominant bacterial phyla in the tank water (TW, diamond) as well as fore (F, triangle), mid (M, circle), and hindguts (H, square) of *P. nigrolineatus* fed either a wood (W, open symbols) or mixed diet (X, closed symbols). OTUs were picked using the SILVA\_132 16S database (OTU ID =0.97).

minimal OTU overlap, the core microbiomes were taxonomically compositionally similar (**Supplementary Tables S2, S3**).

Network analysis revealed that a small number of predominant core microbiome OTUs with high sequence similarity to members of the genera *Aeromonas* were primarily responsible for the observed shift in Proteobacteria abundance (**Figures 3B,C**). Most OTUs were shared among multiple gut regions, within and across feeding regimens. The reduced OTU network (**Figure 3A**) represents the most abundant OTUs (minimum OTU cluster size of 1 × 103 reads after rarefaction) and are identical with those identified in the core microbiome analysis. Consistent with the PCoA analysis, more OTUs were shared among samples within a feeding study. Sample node degree distributions also suggest increased biodiversity within the first feeding study (**Supplementary Figure S4**). This is consistent with the alpha diversity measures. Because the network analysis is constructed from a rarefied OTU table, the reduction in average node degree distributions likely represents a true reduction in biodiversity in these samples and is not a result of reduced sampling depth.

#### Predictive Metagenome Profiling

To gain insight into the metabolic capacity of the enteric microbiome, a PICRUSt analysis was performed to generate a predictive functional profile. To compare across feeding regimens, the predicted gene profiles from the mid and hindguts were averaged and collapsed at a higher level (L2) within the KEGG hierarchy (**Figure 4**). Based on this analysis, the relative abundance of several pathways was statistically different between feeding regimens. Wood-fed fish had a higher abundance of genes involved in transcription and enzyme families that include peptidases, cytochrome P450, and protein kinases, while mixed diet fishes were enriched for genes involved in the metabolism of amino acids, terpenoids, and polyketides, as well as the metabolism of other amino acids. These differences were reflected in the PCA analysis of the predicted microbiome, which shows clustering of wood and mixed diet samples along PC2 (**Supplementary Figure S5**). No significant differences were observed for the relative abundance of genes involved in carbohydrate metabolism or xenobiotic degradation. There were

no significant differences in pathway relative abundances between the mid and hindgut regions within feeding regimens suggesting the microbiome within these tissues may function similarly in regard to lignocellulose degradation. This finding was supported by PCA analysis (**Supplementary Figure S6**), which showed no distinction between any tissue regions.

Despite high similarity between wood and mixed diet-fed fish predicted metagenome at higher levels in the KEGG hierarchy, comparisons at lower levels (L3) revealed many differences (**Supplementary Table S5**). The majority of these KEGG pathways were associated with the metabolism of amino acids and carbohydrates. Mixed diet fish had an increase in pathways associated with amino acid transformation and included the metabolism of glycine, serine, threonine, histidine, lysine, tryptophan, and tyrosine, as well as the degradation of lysine valine, leucine, and isoleucine. However, pathways associated with carbohydrate utilization were found in high abundance in both feeding regimens. Mixed-diet fish had a higher relative abundance of pathways involved in the metabolism of butanoate, glyoxylate, dicarboxylate, propanoate, and pyruvate. In comparison, wood-fed fish had higher abundance of pathways involved with metabolism of amino and nucleotide sugars, fructose, mannose, galactose, starch, and sucrose, as well as the interconversion of pentose and glucuronate and a higher abundance of genes associated with degradation of other glycans. Additionally, pathways were found to be differentially abundant between feeding regimens and included processes involved with the metabolism of lipids, cofactors, and vitamins, as well as the degradation of xenobiotics.

Because wood degradation requires a diverse suite of enzymes for complete hydrolysis, the relative abundance of genes for several lignocellulose-active enzymes were also examined (**Figure 5**). Included in the analysis were enzymes active against cellulose, hemicellulose, lignin, and cello-oligosaccharides. Despite statistically significant differences in starch metabolism pathways at lower KEGG classifications, there were very few differences in the relative abundance of lignocellulose active enzymes. For most enzymes, neither diet nor tissue region appeared to influence relative abundances with the only significant differences between wood and mixed diet-fed fish seen in the relative abundances of lysophospholipase (EC 3.1.1.5) and carboxylesterase (EC 3.1.1.1), which both act upon hemicellulose. The vast majority of carbohydrate active enzyme genes were observed infrequently (~1 × 10<sup>−</sup><sup>6</sup> to 1 × 10<sup>−</sup><sup>5</sup> ); however, a limited number were predicted to be in higher abundance (>5.0 × 10<sup>−</sup><sup>4</sup> ). These were primarily limited to activities likely associated with lignin degradation (e.g., cytochrome C peroxidase, catalase-peroxidase, and glycolate oxidase), but also included glyceraldehyde 3-phosphate dehydrogenase.

Endocellulases (EC 3.2.1.4), exocellulases (EC 3.2.1.91), and ß-glucosidases (EC 3.2.1.21) were detected in both feeding regimens (**Figure 5** and **Table 3**). The relative abundance of the individual cellulases varied (~1 × 10<sup>−</sup><sup>7</sup> to 4 × 10<sup>−</sup><sup>4</sup> ) but remained largely the same across tissue regions and diets. All tissue regions across both diets had a higher relative abundance of endoglucanases and ß-glucosidases than exocellulases (**Table 3**). Exocellulases were exceedingly rare in all samples and completely absent from several samples of the second feeding study. Distributions were calculated for each of the three cellulases in the predicted metagenome (**Figure 6**) and, like the relative abundance analysis, differences were observed

TABLE 3 | Relative abundance of the three classes of cellulose degrading enzymes based on predictive metagenomics.


*Abundances were calculated using PICRUSt (see methods) and compared across diet type and tissue region.*

between various classes of cellulases. Endocellulases were primarily attributed to Proteobacteria and Planctomycetes in both feeding regimens, while Bacteroidetes, Verrucomicrobia, Armatimonadetes, and Actinobacteria provided functions in specific samples. ß-Glucosidases were provided primarily by Proteobacteria and Actinobacteria. While the relative abundance of Actinobacteria and Proteobacteria varied between samples and diets, the taxonomic makeup was highly consistent and included the Proteobacteria genera *Enterobacter*, *Aeromonas*, *Citrobacter*, *Novospirillium*, *Cronobacter*, and *Rhodobacter*, as well as the Actinobacteria genus *Gordonia* and an unidentified Microbacteriacea. Exocellulases were provided by a single phylum of bacteria in nearly all enteric samples. In the first study, exocellulases were found to be exclusively represented by the Firmicutes and included members of the orders OPB54 and Clostridiales, while in the second study, they originated either Proteobacteria or Actinobacteria and included Rhizobiales and Actinomycetales.

#### DISCUSSION

Diet has been shown to have a major impact on enteric bacterial communities of a variety of animals. *P. nigrolineatus* provides a unique opportunity to examine the effects of diet change on a wood-consuming organism. Unlike other xylivorous animals, *P. nigrolineatus* is capable of shifting between diets with seemingly minimal deleterious health effects. In this study, enteric microbial communities were examined for *P. nigrolineatus* provided either a wood or mixed diet. Attention was given to abundance and distribution of putative lignocellulolytic microorganisms that are essential for deriving nutrition from a wood diet. Predictive metagenomics were also used to determine whether a wood diet enriched for pathways involved in lignocellulose degradation.

#### Enteric Bacterial Community Composition

Difficulties in obtaining large numbers of fish at one time necessitated two feeding experiments. Any differences between community composition for the two feeding studies might

the PICRUSt analysis.

be attributed to the environment where the animals were caught and their initial handling. While our study would have benefitted from examining all animals at the same time, we found that enteric bacterial community composition for both feeding studies were largely consistent with those previously established by previous 16S rRNA metagenomic library analyses (McDonald et al., 2012). Therefore, we believe that communities associated with either mixed diet or wood-fed fish are representative of their dietary regimen and differences observed for community structure are significant and diet influences the *P. nigrolineatus* enteric bacterial community. These results diverge from studies in termites where non-wood diets had minimal impact on bacterial community composition but did cause changes in relative bacterial abundance (Huang et al., 2013; Wang et al., 2016; Su et al., 2017). There was no obvious correlation between diet and OTU richness in any tissue region. Wood-fed fish in the first feeding experiment had a less diverse enteric community than mixed diet-fed fish, a trend that was reversed in the second feeding experiment. These results were inconsistent with studies showing that fishes that are dietary specialists have higher microbial diversity than dietary generalists (Bolnick et al., 2014).

Fish from the first feeding experiment had a higher relative abundance of Planctomycetes (~10–66%) than fish from the second feeding study (<1–35%). Most Planctomycetes were members of the class Pirellulales but also included Gemmatales, Isosphaerales, and Planctomycetales. Planctomycetes have been identified in variety of environments including the GI tracts of fish and termites, soils, and bioreactors (Van Kessel et al., 2011; Wang et al., 2011; Abdul Rahman et al., 2015). Although some Planctomycetes appear to be responsible for heteropolysaccharide degradation in diverse environments, this activity has not been described in Pirellulaceae (Wilhelm et al., 2019).

Proteobacteria were ubiquitous in all samples and were represented by several classes including alpha, beta, gamma, and delta. A higher proportion was observed in the second feeding experiment (~33–99%) than the first (~19–50%). Several proteobacterial families were distributed in a feeding study-, tissue-, and diet-specific manner. In the first feeding study, these families consisted primarily of Rhizobiales, Rhodobacterales, and Aeromonadales, while those from the second feeding study consisted of Rhodobacterales, Rhodospirillales, Aeromonadales, Betaproteobacteriales, Enterobacteriales, Legionellales, and Pseudomonadales. The diversity of Proteobacteria in mixed diet-fed fish of the second feeding study was higher than the wood-fed fish community, which was dominated by Aeromonas and Enterobacterales. Many of these families of Proteobacteria as well as Planctomycetes have been associated with fasting fish, but their role in the wood-fed fish is unclear (Kohl et al., 2014; Xia et al., 2014). Aeromonas has been identified as a major contributor of cellulases in the GI tract of herbivorous grass carp *Ctenopharyngodon idella*, but is also a known pathogen of fresh water fishes. Its role in *P. nigrolineatus* is unclear (Jiang et al., 2011; Ran et al., 2018).

Several bacterial phyla were highly represented in limited number of samples from both feeding studies. These included Actinobacteria, Fusobacteria, Bacteroidetes, and Verrucomicrobia. Actinobacteria were predominantly identified in the second feeding study mixed diet fish and were comprised of species of *Gordonia* and *Mycobacterium*. Both genera have been identified as major components of fish microbiomes where they may play a role in xenobiotic and cellulose degradation and enhance the growth rates of the host (Arenskötter et al., 2004; Medie et al., 2010; Sheikhzadeh et al., 2017). The Bacteroidetes had high sequence similarity to *Candidatus* Cardinium, which are known obligate intracellular pathogens of arthropods and can regulate host health, development, and reproduction (Giorgini et al., 2009). Organisms related to *Candidatus* Cardinium have been identified in plant pathogenic nematodes (Noel et al., 2006) as well as intracellular symbionts of several plant-feeding arthropods (Zchori-Fein and Perlman, 2004; Zhang et al., 2013). Fusobacteria sequences had high sequence similarity to species of *Cetobacterium*. Members of this genus have been identified in high abundance in the GI tracts of a variety of fish species where they may play a role in vitamin synthesis (Liu et al., 2016; Ramírez et al., 2018). Verrucomicrobia were identified as members of the genera *Prosthecobacter* and *Chthoniobacter* in the first and second feeding studies, respectively. They are often plant-associated and exist as endophytes or members of the rhizospheric community (Rascovan et al., 2016; Dong et al., 2018).

#### Predicted Metagenome Reconstruction

Any changes in diet that may have influenced the metabolic capacity of the enteric microbial community was examined by predictive metagenomics, which revealed functional differences at higher KEGG classifications. Relative to the mixed diet, the microbiomes of the wood-fed *P. nigrolineatus* did not appear to be enriched for genes involved in lignocellulose degradation. Several studies have demonstrated the ability of microbial communities to shift in response to changes in resource availability (Muegge et al., 2011; Gomez et al., 2016). The relative lack of a reduced metabolic response in *P. nigrolineatus* suggests either top down regulation by the host, i.e., the GI tract environment selects for microbes with specific functional capacities independent of diet, or that the mixed and wood-only diets were not different enough to drive divergence in gut microbiome function. However, studies have shown gut environments to select for specific microbial functions independent of the taxonomy of the microorganisms (Lozupone et al., 2008; Sanders et al., 2015).

Diet had minimal impact on the relative abundance of microbial carbohydrate-active enzymes. Of the 36 investigated genes involved in cellulose, hemicellulose, and cello-oligosaccharide metabolism, a single gene, encoding lysophospholipase (E.C. 3.1.1.5), was significantly more abundant in the wood-fed fish. Conversely, a single gene, encoding carboxylesterase (E.C. 3.1.1.1), was predicted to be in higher abundance in mixed diet-fed fish. Despite this, the microbiomes of both diets appeared to be adapted to metabolize plant polysaccharides. The relative abundance of the three major cellulases was similar to those reported for nitrogen amended green plant waste bioreactors using the same PICRUSt method (Yu et al., 2017). Consistent with our findings, the study also reported large variations in the relative abundances of the different cellulases, with endocellulases and β-glucosidases estimated to be much more abundant (~100×) than exocellulases. The relative proportion of β-glucosidases to cellulases in bacterial genomes are generally lineage specific (Berlemont and Martiny, 2013), but there are typically fewer exocellulases than endocellulases and β-glucosidases in the genomes of true cellulose utilizers. However, these genes are not known to exist at ratios approaching 1:100, which we have identified here. Similar ratios of cellulases have been found in the predicted metagenomes of other environmental and enteric microbiomes known to hydrolyze plant polysaccharides (Zheng et al., 2018; Gao et al., 2019). This suggests either a limitation of PICRUSt to accurately predict the abundance of one or more classes of cellulases. Or, alternatively, the skewed ratios are indicative of a large population of microbial opportunists who do not hydrolyze cellulose directly, but are capable of exploiting the disaccharides and short oligosaccharides released by cellulolytic species.

The presence of a large, opportunistic cellulose-utilizing community was confirmed by analysis of cellulase contributions. Distinct taxonomic lineages contributed each of three major cellulases. Endocellulases were contributed primarily by Planctomycetes, Proteobacteria, Bacteroidetes, Actinobacteria, and Verrucomicrobia, while β-glucosidases were provided by Proteobacteria and Actinobacteria. This contrasts sharply to exocellulases, which were contributed almost exclusively by either Firmicutes or Actinobacteria in the first or second feeding studies, respectively. Many of the bacterial orders identified as contributing the endocellulases and β-glucosidases were also identified as major constituents of the microbial community. These included Pirellulales, Gemmatales, and Planctomycetales of the Planctomycetes as well as Rhodobacterales, Rhizobiales, Aeromonadales, Enterobacterales, Pseudomonadales, Verrucomicrobiales, and Cytophagales of the Proteobacteria. Endocellulases and β-glucosidases are common among non-cellulose utilizers. While there are true lignocellulose utilizing members of these orders, the majority utilize these enzymes to degrade microbial derived exopolysaccharides, modify their cell walls, or as a means of infiltrating plant hosts and not as method of obtaining fixed carbon. Exocellulases appear to be provided by a taxonomically narrow group of organisms including Ruminococcus, OPB54 Rhizobiales, and Actinomycetales. These bacteria have been identified as lignocellulose degrading organisms in a variety of environments and likely represent the true cellulolytic consortium of the enteric microbial community despite comprising a relatively small fraction of the overall community (<1–17%) (Ben David et al., 2015; Joshi et al., 2018; Šimůnek et al., 2018).

Carbohydrate active enzyme profiles suggest that *P. nigrolineatus* derives most of its nutrition from the hydrolysis of hemicellulose and not cellulose. At L3, the majority (9/15) of functional categories were differentially represented across the feeding diets. More abundant in the wood diet were genes involved in the metabolism of amino and nucleotide sugars, fructose, mannose, galactose, sucrose, and starches, as well as the interconversion of pentose sugars and glucuronic acid. Most of these carbohydrates have been identified as major components of xyloglucan, glucomannan, mannan, xylan, and arabinoxylan, which, together, form hemicellulose (Fry, 1989; Moreira and Filho, 2008; Rennie and Scheller, 2014). A similar pattern of gene enrichment was observed in the GI tract of giant pandas whose microbiome is replete with amylases and hemicellulases (Zhang et al., 2018). Also similar to *P. nigrolineatus*, the giant panda microbiome lacks the abundance of cellulases observed in other herbivores (Zhang et al., 2018). Coupled with a short gut transit time, it is unlikely that *P. nigrolineatus* is capable of sufficient cellulose hydrolysis. Rather, it is more likely that the limited numbers of cellulases provided by the microbiome are used to liberate the more easily hydrolysable and assimilable hemicellulose from the lignocellulose matrix.

Consumption of the higher protein mixed diet by *P. nigrolineatus* selected for a microbiome capable of amino acid catabolism and fermentation. The microbiome of mixed diet-fed fish was enriched for genes involved in the metabolism and degradation of several amino acids including branched-chain amino acids (valine, leucine, and isoleucine), tyrosine, tryptophan, lysine, histidine, glycine, serine, and threonine. Amino acid fermentation typically occurs in the distal intestines where bacterial densities are high and carbohydrate concentrations are minimal. The substrates and products of amino acid fermentation vary depending on diet, gut environment, and microbial consortium (Dai et al., 2011; Neis et al., 2015). However, the preferred amino acids include glutamine, asparagine, lysine, serine, threonine, arginine, glycine, histidine, and the branched-chain amino acids. Catabolism usually includes both deamination and decarboxylation and results in the formation of various products including ammonia, as well as short-chained/branched fatty acids and organic acids (Dai et al., 2011). Examination of the predicted metagenome (KEGG L3) showed that the mixed-diet fed fishes were enriched for genes involved in the metabolism of fatty acids, butyrate, propionate, and pyruvate. Products from the fermentation of amino acids are likely used as precursors for gluconeogenesis in mixed diet-fed fish as there is also an increased abundance of genes associated with the glyoxylate cycle.

Intestinal bacteria play a major role in host nutrition by serving as a source of essential vitamins and nutrients. Plant-based diets typically lack several compounds including sterols, B vitamins, and nitrogenous compounds, many of which cannot be synthesized by the host. Genomic and culture-based studies have identified several vitamin-producing bacteria from the GI tracts of woodfeeding invertebrates as well as herbivorous and omnivorous fish (Sugita et al., 1991; Rosenthal et al., 2011; Abhishek et al., 2015). Administering antibiotics to these wood-eating organisms often results in reduced rates of intestinal vitamin biosynthesis, strongly implicating the gut microbiome as the primary source (Lovell and Limsuwan, 1982). Dietary factors such as feed type, nitrogen content, and age have also been shown to effect vitamin synthesis in ruminant animals (Schwab et al., 2006; Beaudet et al., 2016). In this study, there was no significant increase in the relative abundance of genes associated with cofactors and vitamin metabolism in the wood-fed fish relative to the mixed-diet fish (KEGG L2). However, closer examination revealed that the mixed diet-fed fish microbiome was enriched with genes involved in ubiquinone and other terpenoid-quinone biosynthetic pathways. Included in this collection of genes are the biosynthetic pathways for α-tocopherol (vitamin E), menaquinone (vitamin K2), and phylloquinone (vitamin K1). Ubiquinone, in addition to its role in electron transport, may play a role in dissimilatory lignin degradation (DeAngelis et al., 2013).

Genes associated with xenobiotic degradation were more abundant in the mixed diet-fed fish. These findings differ from metatranscriptomic studies in termites, which showed increased expression of several detoxifying enzymes when fed bulk wood instead of less refractory foods such as paper (Raychoudhury et al., 2013). Lignin, a structural component of plant cell walls, is a major barrier to the liberation and saccharification of cellulose and hemicellulose. It is a heterogeneous, highly recalcitrant, polymer of radically coupled aromatic compounds. Degradation of lignin releases several cytotoxic and anti-nutritive compounds including organic acids, phenolic compounds, and reactive oxygen species that disrupt cellular processes and damage cell components (Abhishek et al., 2015). While relatively little is known about bacterial lignin-active enzymes, bacteria have been shown to play a major role in the degradation of lignin in several environments (Bugg et al., 2011; Abhishek et al., 2015). Of the selected lignocellulose-active enzymes examined in this study, those with activity against lignin were the most abundant, including cytochrome c peroxidases, glycolate oxidases, peroxiredoxin, and the recently described catalase-peroxidases. However, there were no significance differences in the relative abundance of these genes across diets or tissue regions, although the metagenomic analysis was limited to the bacterial fraction of the microbiome. In termite systems, the largest increases in expression of xenobiotic degrading enzymes were observed in the gut flagellate and host transcript pools (Brown and Chang, 2014). While there is no evidence of flagellates in the *P. nigrolineatus* GI tract, a diverse fungal community has been detected in all regions of the gut (Marden et al., 2017) and fungi have tremendous capacity for lignocellulose degradation, with the potential of playing an active role in lignin processing (Cragg et al., 2015).

Previous studies have intimated that wood consumption by *P. nigrolineatus* occurs simply as a means of accessing bacterial and fungal hyphae present below the wood surface and that wood consumption causes starvation in the fish (Tartar et al., 2009). In this study, fungal and photosynthetic bacterial growth was inhibited by rearing fish in the dark and autoclaved, and sterile wood was provided as the sole food source, thereby alleviating the likelihood of obtaining nutrients from wood-related microorganisms. Loricariids have a remarkable capacity to survive extended periods without food and could have easily survived the duration of the feeding experiment without consuming any food, doing so through a combination of reduced metabolic rates and reduced GI tract mucosal and microvilli surface areas (Lujan et al., 2011). Although we did not address fish dietary performance, predictive metagenomic analysis suggested that the fish did not appear to be exhibiting a "starved" phenotype. Bacterial metagenomes of starved fish have been shown to have lower abundances of genes involved in transcription, cell division, and DNA replication/ repair, while genes involved in membrane transport and protein turnover are enriched (Xia et al., 2014), patterns of relative gene abundances that were not observed between the wood- and mixeddiet fed in fish. However, these measures of gene abundance are predictive and calculated from the relative abundance of different taxonomic lineages. Understanding the metabolic potential of the enteric bacterial community would benefit from a metatranscriptomic analysis of a larger number of fish.

Results demonstrated the functional resiliency of the *P. nigrolineatus* enteric bacterial community. Despite large, dietinduced, shifts in community composition, little change was observed in the predicted relative abundance of genes related to lignocellulose degradation. Our findings suggest that the enteric community composition is altered by the metabolic capacity of the microorganisms, with the GI tract environment selecting for overall community function and not specific microbial lineages. Selection based on function may serve as an advantage for wood-feeding organisms like *P. nigrolineatus* that switch between feeding habits as it insures the presence of essential metabolic pathways even after prolonged feeding on less refractory foods. How the *P. nigrolineatus* GI tract selects for lignocellulose-degrading microorganisms despite changing diets is unclear and the focus of future investigations.

#### DATA AVAILABILITY STATEMENT

The datasets generated during and/or analyzed during the current study are available from the NCBI Sequence Read Archive database under accession numbers PRJNA407967 (feeding study 1) and PRJNA549277 (feeding study 2).

#### ETHICS STATEMENT

Fish growth conditions and all experimental protocols were approved by the Towson University Institutional Animal Care

#### REFERENCES


and Use Committee (IACUC 071509JW-01) and the University of Maryland School of Medicine Office of Animal Welfare Assurance (IACUC #0618005). All methods were performed in accordance with IACUC guidelines and regulations.

#### AUTHOR CONTRIBUTIONS

RM, JW, and HS conceived the project and designed research. HS and JW supervised the study. RM and HS performed research. RM, JW, and HS guided the analysis. RM, JW, and HS analyzed data. RM prepared figures. RM, JW, and HS wrote, reviewed and edited the paper. HS provided lab space. HS and JW provided funding. All authors reviewed and approved the final manuscript.

#### FUNDING

This work was supported, in part, by the National Science Foundation (award 0801830 to JW and HS). RM was supported by the NIH Chemistry/Biology Interface Program (T32GM066706).

#### ACKNOWLEDGMENTS

We thank Tsvetan Bachvaroff for helpful discussions and Sabeena Nazar, Bio Analytical Services Lab, University of Maryland Center for Environmental Sciences, for sequencing.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2019.02687/ full#supplementary-material


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2019 McDonald, Watts and Schreier. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Corrigendum: Effect of Diet on the Enteric Microbiome of the Wood-Eating Catfish Panaque nigrolineatus

#### Ryan C. McDonald<sup>1</sup> , Joy E. M. Watts <sup>2</sup> and Harold J. Schreier 1,3 \*

*<sup>1</sup> Department of Biological Sciences, University of Maryland, Baltimore County, Baltimore, MD, United States, <sup>2</sup> Department of Biological Sciences, University of Portsmouth, Portsmouth, United Kingdom, <sup>3</sup> Department of Marine Biotechnology, University of Maryland, Baltimore County, Baltimore, MD, United States*

Keywords: lignocellulose digestion, microbiome, 16S rRNA gene amplicon sequencing, predictive metagenomics, Amazonian catfish

#### **A Corrigendum on**

Edited and reviewed by:

*Martin G. Klotz, Washington State University, United States*

#### \*Correspondence:

*Harold J. Schreier schreier@umbc.edu*

#### Specialty section:

*This article was submitted to Microbial Symbioses, a section of the journal Frontiers in Microbiology*

Received: *18 December 2019* Accepted: *14 February 2020* Published: *27 February 2020*

#### Citation:

*McDonald RC, Watts JEM and Schreier HJ (2020) Corrigendum: Effect of Diet on the Enteric Microbiome of the Wood-Eating Catfish Panaque nigrolineatus. Front. Microbiol. 11:331. doi: 10.3389/fmicb.2020.00331* **Effect of Diet on the Enteric Microbiome of the Wood-Eating Catfish Panaque nigrolineatus** by McDonald, R. C., Watts, J. E. M., and Schreier, H. J. (2019). Front. Microbiol. 10:2687. doi: 10.3389/fmicb.2019.02687

In the original article, there was a mistake in **Table 3** as published. All exponents were incorrectly shown as positive values when they should have been negative. The corrected **Table 3** appears below.

TABLE 3 | Relative abundance of the three classes of cellulose degrading enzymes based on predictive metagenomics.


*Abundances were calculated using PICRUSt (see methods) and compared across diet type and tissue region.*

The authors apologize for this error and state that this does not change the scientific conclusions of the article in any way. The original article has been updated.

Copyright © 2020 McDonald, Watts and Schreier. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Xylooligosaccharide Modulates Gut Microbiota and Alleviates Colonic Inflammation Caused by High Fat Diet Induced Obesity

Yanquan Fei\*, Yan Wang, Yilin Pang, Wenyan Wang, Dan Zhu, Meigui Xie, Shile Lan\* and Zheng Wang\*

Obesity leads to colonic inflammation and may increase the risk of colorectal

College of Bioscience and Biotechnology, Hunan Agricultural University, Changsha, China

#### Edited by:

Liwei Xie, Guangdong Academy of Science, China

#### Reviewed by:

Savneet Kaur, Institute of Liver and Biliary Sciences, India Supriyo Bhattacharya, City of Hope National Medical Center, United States

#### \*Correspondence:

Yanquan Fei fyq0614@stu.hunau.edu.cn Shile Lan hulanshl@126.com Zheng Wang wz8918@163.com

#### Specialty section:

This article was submitted to Gastrointestinal Sciences, a section of the journal Frontiers in Physiology

Received: 23 April 2019 Accepted: 20 December 2019 Published: 22 January 2020

#### Citation:

Fei Y, Wang Y, Pang Y, Wang W, Zhu D, Xie M, Lan S and Wang Z (2020) Xylooligosaccharide Modulates Gut Microbiota and Alleviates Colonic Inflammation Caused by High Fat Diet Induced Obesity. Front. Physiol. 10:1601. doi: 10.3389/fphys.2019.01601 cancer. Xylooligosaccharide (XOS) exhibits strong antioxidant and excellent antibacterial properties, and can be utilized by gut microbes to maintain the ecological balance of the intestinal tract. In this study, we explored how XOS modulates the microbiota and regulates high fat diet (HFD) induced inflammation. We measured the changes in body weight and visceral coefficients in rats fed a high-fat diet. We also measured the expression levels of inflammatory factors in the plasma and colonic tissues of the rats using the enzyme-linked immunosorbent assay and real-time quantitative polymerase chain reaction. We analyzed the composition of fecal microorganisms and short chain fatty acid (SCFA) content using 16S rDNA and GC-MS. We found that XOS significantly counteracted HFD induced weight gain. Moreover, the plasma levels of monocyte chemoattractant protein-1, tumor necrosis factor (TNF-α) and lipopolysaccharide decreased in the XOS-treated rats. XOS treatment decreased TNF-α mRNA expression and increased occludin mRNA expression in the rat colon. We observed a reduction in the overall microbial abundance in the feces of the XOStreated rats, although the proportion of Bacteroidetes/Firmicutes increased significantly and the number of beneficial bacteria increased in the form of dominant microbes. We found that both SCFA-producing bacteria and SCFA content increased in the gut of the XOS-treated rats. We identified a correlation between the abundance of Prevotella and Paraprevotella and SCFA content. Taken together, we propose that XOS can alleviate colonic inflammation by regulating gut microbial composition and enhancing SCFA content in the gut.

Keywords: obesity, high-fat diet, colon inflammation, gut microbiota, SCFA

# INTRODUCTION

Obesity is a global health problem (Cox et al., 2015), affecting nearly two billion people worldwide. Obesity is associated with high blood pressure, diabetes, and other chronic diseases. Recent studies have shown that obesity is also closely related to chronic inflammation (Liu et al., 2012). Excess fat deposits in the body induce intestinal structural and morphological changes, including damage to the colonic crypt and colonic epithelium, a reduced number of goblet cells and an increased number of intestinal epithelial cells (Paturi et al., 2010). Obesity also changes the composition of intestinal

**582**

microbiota. The intestine is a complex micro-ecological system. The micro-ecological balance of the intestinal microflora and its metabolites are inextricably linked to human health (Lynch and Pedersen, 2016). The gut microbiota and certain intestinal microbial metabolites maintain the integrity of the intestinal barrier and epithelium, which are key to preventing high fat diet (HFD) induced colonic damage (Peng et al., 2009). Importantly, these microbes can digest carbohydrates to produce short chain fatty acids (SCFAs), which maintain intestinal health and relieve colonic inflammation. Moreover, SCFAs can supply energy to the intestinal cells and regulate their cell cycle progression (Pryde et al., 2002; Di Sabatino et al., 2005).

Xylooligosaccharide (XOS) is a sugar oligomer of 2–7 xylose single molecules linked by β-1,4 (Jordan and Kurt, 2010), a hydrolysate of dietary fiber commonly found in corn stover, wheat bran and rice bran (Ma et al., 2017). XOS shows strong antioxidant and antibacterial activity in vitro (Gao et al., 2012). Although XOS cannot be digested by humans, it can be metabolized by intestinal microbes (Gao et al., 2017). Indeed, XOS promotes the growth of Bifidobacterium and Lactobacillus in the gut and increases the SCFA content in these intestinal microbes, in turn enhancing the intestinal barrier function (Li et al., 2015). In pre-diabetic patients, XOS can enhance insulin sensitivity and reverse the changes in microbial composition caused by insulin insensitivity (Yang et al., 2015). Specifically, dietary supplementation of XOS can increase the relative abundance of Lactobacillus spp. and Bifidobacterium spp. in the gut, and the expression of tight junction protein occludin (OCLN) in the cecal tissues (Christensen et al., 2014). Both XOS and Bifidobacterium can enhance the immune function of the host (Childs et al., 2014). Studies have shown that XOS promotes the glycolysis of bifidobacteria throughout the gut, leading to an increase in SCFA concentration (Hansen et al., 2013). Here, we hypothesized that XOS can regulate the intestinal flora, increase SCFA levels and alleviate HFD-induced colonic inflammation in obese rats.

# MATERIALS AND METHODS

#### Animal Experimental Design and Sample Collection

All of the animal procedures were performed in accordance with the Guidelines for Care and Use of Laboratory Animals of Hunan Agricultural University. The protocol was approved by the Animal Care and Use Committee of Hunan Agricultural University. Thirty male Sprague Dawley rats (weighing 280 ± 20 g, aged 8 weeks, n = 30) were purchased from Hunan Silaike Jingda Co. (Changsha, China), with a certificate number of HNASLKJ2016-0002. After 1 week of acclimatization, the rats were randomly divided into three groups: normal control (NC) (n = 10), HFD (n = 10), and HFD plus xylooligosaccharide (HFD + XOS) (n = 10). The NC group was fed a normal diet (total calorie rate of 3.6 kcal/g and 72.3% of the caloric percentage of carbohydrates), and the HFD groups were fed a high fat diet (total calorie rate of 4.6 kcal/g, 46.4% of the caloric percentage of carbohydrates). After 7 weeks on the specified diets, the NC and HFD groups were given purified water by gavage, while the HFD + XOS group was given 2 g/kg XOS solution by gavage. XOS (CNS: Lu XK13-217-00581) of 95% purity was purchased from Shandong Longli Biotechnology Co. (Shandong, China). The animals were housed under cyclical illumination conditions (12-h light/dark cycle) with free access to food and water. They were weighed once a week. After 7 weeks of gavage, the rats were fasted for 12 h and sacrificed using pentobarbital sodium. The liver, heart, and spleen of each mouse were completely cut and weighed. At the same time, the white fat wrapped around the kidneys and the white fat attached to the testicles were cut and weighed. The colonic tissue was harvested and weighed. The middle part of the colon was fixed in 10% formalin, and the rest was preserved in liquid nitrogen. Heparin anticoagulant was added to the whole blood samples. The mixture was incubated for 30 min and spun at 12,000 rpm for 15 min at 4◦C. The supernatant was then collected and stored.

# Histopathological Analysis

Colon and liver tissues from three group (n = 6) were removed from the fixative solution and slowly flushed with water. The tissue mass was then soaked in ethanol of different concentrations and dehydrated at 37–45◦C for 2–4 h. Next, the tissue was embedded in paraffin wax (SVA, Uppsala, Sweden) and sectioned at a slice thickness of 5 µm. The sections were stained with hematoxylin and eosin, and imaged using a microscope (ML31, MSHOT, Guangzhou, China).

# Enzyme-Linked Immune Sorbent Assay

The content of lipopolysaccharide (LPS), interleukin 6 (IL-6), interleukin 10 (IL-10), tumor necrosis factor (TNF-α), and monocyte chemoattractant protein-1 (MCP-1) in the plasma from three group (n = 10) was measured using CUSABIO kits (CSB-E14247r/E04640r/E04595r/E11987r/E07429r, Wuhan, China) and a microplate reader (14041717, VT, United States).

#### Real-Time Fluorescence-Based Quantitative PCR

Total RNA was extracted from the colonic tissues from three group (n = 10) using TriQuick Reagent (Solarbio, Beijing, China), and quantified and purified using an ultramicro UV visible spectrophotometer (NanoDrop 2000, Thermo, United States). The RNA was reverse-transcribed into cDNA using a PrimeScriptTM RT reagent kit with gDNA Eraser (TaKaRa, Japan) and stored at −80◦C. The resulting cDNA was analyzed by conducting the real-time quantitative polymerase chain reaction with a SuperReal PreMix Plus (SYBR Green) reagent kit (TIANGEN, Beijing, China) with various primers (**Supplementary Table S1**). The relative expressions were determined using the 2-11Ct method.

# 16S rDNA and Illumina MiSeq Sequencing

The colonic contents from three group (n = 5) were collected in a sterile sampling tube, frozen in liquid nitrogen and extracted

using a DNA extraction kit (Majorbio Bio-Pharm Technology, Shanghai, China). The extracted DNA was detected using 1% agarose gel electrophoresis to ensure the purity and integrity of the DNA. Qualifying samples were used to construct the library. The Amplicon fragment of interest was recovered, and the sticky ends formed by the disruption were repaired into blunt ends using T4 DNA Polymerase, Klenow DNA Polymerase and T4 PNK. By adding the base "A" at the 3<sup>0</sup> end, either the DNA fragment was ligated to a specific linker with a "T" base at the 3<sup>0</sup> end, or a double-index fusion primer containing a sequencing linker was designed and synthesized using genomic DNA as a template. Fusion primer PCR, magnetic bead screening purpose Amplicon tablets Section, finally, cluster preparation and sequencing were performed using a qualified library. The data obtained were used for the corresponding biological information analysis. All of the offline data were analyzed by the Beijing Genomics Institute).

#### SCFA Content in Feces

Sample preparation: Rat feces from three group (n = 5) were collected after gavage. ddH2O was added and the samples were mixed, incubated at 4◦C overnight and centrifuged. The supernatant was mixed with 25% metaphosphoric acid (Sinopharm, Shanghai, China) at a volume ratio of 1:1, incubated at room temperature for 4 h, and centrifuged at 12,000 rpm for 15 min. The supernatant was filtered using a 45-µm microporous filtration membrane, and the SCFA content was determined using gas chromatography–mass spectrometry (GC-MS) (Agilent 7890-5975C, Santa Clara, CA, United States).

Standard curve: A stock solution of SCFAs (Sigma, St. Louis, MO, United States) was prepared and preserved at 4◦C (avoiding light). The stock solution was prepared as a standard solution based on the sample concentration before the measurement.

Chromatographic condition: Chromatographic analysis was conducted using DB-FFAP of 30 m × 250 µm × 0.25 µm equipped with a flame ionization detector (FID). The flow rate of high-purity nitrogen was 0.8 mL/min. The auxiliary gas was hydrogen with a high purity. The injection port and FID detector temperature were 250 and 280◦C, respectively. The sample injection volume in the GC-MS analysis was 1 µL. The initial temperature was 60◦C, increasing by 20◦C/min to 220◦C for 1 min.

#### Statistical Analysis

fphys-10-01601 January 17, 2020 Time: 16:21 # 4

All of the data were generated using SPSS 16.0 software and are represented here as mean ± standard deviation (SD). Differences between the mean values were evaluated using oneway analysis of variance and Tukey's multiple comparisons test (if applicable). A P value < 0.05 was considered to be statistically significant.

#### RESULTS

#### Physical Characteristics

The body weights of the HFD and HFD + XOS rats were significantly greater than those of the NC rats from week 3 onward (**Figure 1**). The difference in body weights between the HFD and NC rats was 20% by the end of week 7. XOS was administered to the HFD + XOS rats from week 8 onward. Compared with the HFD rats, the HFD + XOS rats weighed significantly less from week 10 onward, despite being significantly heavier than the NC rats (**Figure 1**). The effects of XOS treatment on the body weight, organ weight and organ/body weight ratios of the rats are shown in **Table 1**. The body weight of the HFD rats was significantly greater than that of the NC rats. XOS treatment partially inhibited body weight gain in the HFD rats. These results indicated that XOS counteracted the effects of the HFD. The weight of the liver, perirenal fat and epididymal fat, and their corresponding organ/body weight ratios, were significantly higher in the HFD rats than the NC rats. However, these effects were largely inhibited in the HFD + XOS rats. Moreover, the HFD + XOS rats and the NC rats exhibited similar liver weight and liver/body weight ratios. There were no significant weight differences in the heart and spleen between the three groups of rats, and their

TABLE 1 | Organ/body weight ratio.


Organ/body weight ratio (%) is the percentage of the weight of the corresponding organ. Data are given as mean ± SD (n = 10), a, b, cmean values with different letters are significantly different from each other (P < 0.05).


corresponding organ/body weight ratios were also similar. We performed a morphological analysis of the middle segment of each group (n = 6). We used the histologic scoring system to analyze the morphology of the colon (n = 6) (**Table 2**). In the morphological analysis (**Figure 2**), we observed intact crypt structures and no significant inflammatory infiltration in the intestinal epithelial surface of the colonic tissues of the NC rats. In contrast, intestinal edema of the colonic mucosa (**Figure 2Bb**), severe crypt injury (**Figure 2Ba**), fracture on the surface of the colon (**Figure 2Bc**), loss of goblet cells and significant inflammatory infiltration were found in the HFD rats. Importantly, the HFD + XOS rats exhibited less severe colonic tissue damage, more intact crypt structures and reduced inflammatory infiltration compared with the HFD rats. In the liver morphology analysis, we can find that compared with the NC group, the liver of the HFD group showed obvious fat accumulation vacuoles, and macrophages and neutrophils aggregated, and there was obvious inflammatory infiltration. However, after treatment with XOS, these levels of inflammatory infiltration have been alleviated.

#### Plasma Inflammatory Cytokines

We suspected that long-term HFD could lead to colonic inflammation in rats. We therefore determined the levels of inflammatory cytokines in the plasma of our three groups of rats. As shown in **Figure 3**, the levels of TNF-α, MCP-1, IL-6 and LPS were significantly higher in the HFD rats than the NC rats. In contrast, the levels of TNF-α, MCP-1 and LPS were similar between the HFD + XOS rats and the NC rats. IL-6 levels were similar between the three groups of rats. IL-10 levels were significantly lower in the HFD rats than in the NC rats, and XOS supplementation reversed this trend.

#### Colonic Inflammatory Cytokines

We next measured the levels of expression of inflammatory factors and tight junction proteins in the colonic tissues of the rats. As shown in **Figure 4**, we found significantly higher levels of colonic TNF-α expression and IL-10 mRNA expression in the HFD rats than the NC rats. In contrast, colonic TNF-α expression and IL-10 mRNA expression were similar between the HFD + XOS and NC rats. Moreover, colonic OCLN mRNA expression was significantly lower in the HFD rats than the NC rats. However, we found no significant differences in the colonic MCP-1 expression and IL-6 mRNA expression between the three groups.

# Rat Fecal Microbe Composition

The intestinal microflora plays an important role in obesity and lipid metabolism. We thus examined the intestinal microbial components in the three groups of rats by measuring fecal 16srRNAs. We obtained 103,259 optimized sequences in each sample (n = 5). We then clustered these optimized sequences to obtain the number of operational taxonomic units (OTUs) in each sample. As shown in **Table 3**, the HFD rats had significantly

fewer OTUs in the feces than the NC rats. Interestingly, the number of OTUs in the feces was lower for the HFD + XOS rats than the other two groups. First, we performed the determination of α-diversity. Fecal microbial diversity was significantly lower in the HFD rats than in the NC rats, as indicated by the Sobs, Chao, Ace and Shannon indices. Surprisingly, fecal microbial diversity was significantly lower in the HFD + XOS rats than in the other two groups. This may have been due to the favorable effects of XOS on certain species of microbes associated with lipid metabolism or inflammatory responses. The addition of XOS resulted in a significant increase in the percentage of SCFA-producing bacteria such as Prevotella, while the percentage of pathogens such as Oscillospira was significantly reduced. At the same time, Prevotella became the absolute dominant flora in the gut. To analyze the overall difference in β-diversity, principal component analysis (PCA) was performed on all of the samples (**Figure 5**). We observed significant microbial community structure clustering in our three groups of rats.

At the phylum level (**Figure 6A**), Bacteroidetes, Firmicutes, Spirochaetes, and Proteobacteria had the highest phase abundances and were the dominant bacteria in the three groups of rats. The relative abundances of Bacteroidetes, Firmicutes, Spirochaetes, and Proteobacteria in the NC rats were 56.53, 34.21, 4.57, and 3.33%, respectively. The relative abundances of Bacteroidetes, Firmicutes, Spirochaetes, and Proteobacteria in the HFD rats were 44.81, 42.32, 2.59, and 6.54%, respectively. The relative abundances of Bacteroidetes,

Firmicutes, Spirochaetes, and Proteobacteria in the HFD + XOS rats were 64.56, 23.41, 3.52, and 6.01%, respectively. Compared with the HFD rats, HFD + XOS rats exhibited an increased percentage of Bacteroidetes (P < 0.05) and a decreased percentage of Firmicutes (P < 0.05).

At the genus level (**Figure 6B**), Prevotella, Treponema, and Oscillospira were the most abundant microorganisms in the three groups of rats. The relative abundances of Prevotella, Treponema, and Oscillospira in the NC rats were 25.91, 7.35, and 4.92%, respectively. The relative abundances of Prevotella, Treponema, and Oscillospira in the HFD rats were 23.88, 2.31, and 4.63%, respectively. The relative abundances of Prevotella, Treponema, and Oscillospira in the HFD + XOS rats were 32.46, 4.23, and 2.56%, respectively. The relative abundances of Prevotella and Paraprevotella were significantly higher in the HFD + XOS rats than the NC rats (P < 0.05).

#### Fecal SCFA Content and Its Association With Microbial Composition

To verify whether the anti-inflammatory effects were a direct consequence of compositional changes in the intestinal microbiota or an indirect effect involving metabolites, we examined the fecal SCFA content in the three groups of rats

#### TABLE 3 | Alpha-diversity analysis.

fphys-10-01601 January 17, 2020 Time: 16:21 # 8


Results of experiments using the mixed linear regression model for analysis of independent effects of sample type on baseline sample alpha-diversity. Data are given as mean ± SD (n = 5), a, b, cmean values with different letters are significantly different from each other (P < 0.05).

(**Figures 7A–G**). The levels of acetic acid, propionic acid, butyric acid, isobutyric acid, valeric acid, and total SCFAs were found to be significantly lower in the HFD rats than in the NC rats. The levels of acetic acid, propionic acid, isobutyric acid, and valeric acid were similar between the three groups. The levels of butyric acid and total SCFAs in the HFD + XOS rats were significantly higher than those in the HFD rats and similar to those in the NC rats. The levels of fecal isovaleric acid were similar between the three groups. Previous reports have suggested that the SCFA content is regulated by microorganisms. Therefore, we analyzed the correlation between the percentage of intestinal microbes and the SCFA contents of all samples (**Figures 7H,I**). We found a positive correlation between butyric acid and the relative abundances of Prevotella and Paraprevotella.

#### DISCUSSION

Dietary supplementation of XOS has been found to alleviate intestinal inflammation (Yoshino et al., 2006). XOS can modulate inflammatory cytokines such as TNF-α and IL-1β in mice (Hansen et al., 2013). Here, we successfully modeled obesity and systemic low-grade inflammation in rats using HFD. Specifically, the rats fed a HFD exhibited significant increases in body weight, organ fat deposition, organ weight and the organ/body weight ratio. The HFD rats also exhibited structural damage to the intestine. We found that XOS effectively alleviated many of the adverse effects of HFD, including body and organ weight gain, as well as some systemic chronic low-grade inflammation. Longterm HFD can also lead to colonic damage, which is characterized by the shortening of the colon, marked damage to the crypts and the disappearance of goblet cells (Deol et al., 2015).

We speculate that inflammatory damages are due to the high level of LPS. LPS is a lipopolysaccharide produced by Gram-negative bacteria, and can cause inflammation via TLR4 mediated NF-kB activation (Hoshino et al., 1999) and the production of various inflammatory factors, such as TNF-α, IL-6, and IL-1β (Dinarello, 1991). In our experiments, the morphological damage to the colon was relieved by XOS, and the plasma LPS content and expression of TNF-α and MCP-1 were reduced. Our results showed that plasma and colonic IL-10 levels exhibited similar trends in the NC and HFD-XOS rats. In contrast, the HFD rats exhibited increased colonic IL-10 levels and decreased plasma IL-10 levels. There are two possible reasons for this reciprocal behavior. First, IL-10 is produced by a variety of immune cell types, including monocytes and macrophages, which are potent inhibitors of pro-inflammatory cytokines and chemokines, and can prevent diet-induced insulin resistance (Akdis and Blaser, 2001; Hong et al., 2009). Plasma IL-10 can prevent insulin resistance, and the level of blood glucose in the blood is constantly changing. We measured the level of plasma IL-10 at one time point. As the blood glucose level decreases after fasting, and the role of IL-10 in preventing diet-induced insulin resistance weakens, the IL-10 level is low. Second, we measured IL-10 protein levels in the plasma and IL-10 mRNA levels in the colon. In addition, previous studies have shown that XOS can improve intestinal health by regulating intestinal microbes (Makelainen et al., 2010), which may also contribute to the reduction of colonic inflammation observed in the HFD-XOS rats compared with the HFD rats.

Intestinal microbes are intimately linked with the immune system of the host (Zhu et al., 2017; Metzger et al., 2018). There is a mutually beneficial relationship between the microorganisms and the host. The metabolites of some microorganisms are beneficial to the physiological functions of the host, while the host provides energy for the growth and activity of the microorganisms (Kamada et al., 2013; Zhou et al., 2017). The disruption of the balance between the host and gut microbes is associated with many diseases (Littman and Pamer, 2011). Intestinal microbes can increase intestinal permeability and thus control obesity-induced inflammation (Cani et al., 2008). We thus analyzed the composition of intestinal microbes in the three groups of rats. Previous

studies have shown that HFD causes tremendous changes in the gut microbial composition in mice (Turnbaugh and Gordon, 2009). Indeed, an increase in Firmicutes and a decrease in Bacteroidetes are common trends in both obese humans and obese mice (Ley et al., 2005; Moschen et al., 2012). Our results showed that the Firmicutes/Bacteroidetes ratio nearly doubled in the HFD rats compared with the NC rats, and that this effect was inhibited by XOS treatment. At the genus level, XOS treatment led to a significant increase in the percentages of Prevotella and Paraprevotella. Prevotella is a Gram-negative anaerobic bacterium that helps break down proteins and carbohydrates. Prevotella abundance is thought to be associated with polysaccharides and fiberrich plant-based diets (Gorvitovskaia et al., 2016; Ley, 2016). One study showed that metformin improves inflammation in type 2 diabetic rats by selectively acting on Prevotella and reducing IL-6 and TNF-α levels (Liu et al., 2018). In patients with clinical inflammatory bowel disease (IBD), Prevotella abundance at the site of inflammation is significantly reduced (Hirano et al., 2018).

Related studies have shown that Prevotella can synthesize SCFAs using pyruvate as a substrate via the acetyl-CoA pathway (Rey et al., 2010; Louis et al., 2014). We suspected that XOS treatment altered the composition of gut microbes and promoted SCFAs production in rats. SCFAs are thought to be involved in lipid metabolism and transport (Marcil et al., 2002). Although the specific mechanism by which SCFAs are involved in lipid metabolism is unclear, SCFAs promote lipid oxidation in tissues and reduce the storage of white fat (den Besten et al., 2013). SCFAs increase the expression of PGC-1α and AMPK phosphorylation to promote lipid oxidation in tissues (Donohoe et al., 2011). Additionally, SCFAs are metabolites that can improve lipid metabolism and enhance immune functions. Studies have shown that SCFA treatment increases the expression of antimicrobial peptides such as LL-37 and CAP-18 in human intestinal epithelial cells (Raqib et al., 2006); promotes the expression of IL-18, which is a cytokine that maintains the internal stability of the intestinal epithelium; and acts on the epithelial barrier itself (Kelly et al., 2015).

Both Prevotella and Paraprevotella are butyrate-producing bacteria (Gao et al., 2018; Kong et al., 2019). We showed that the total levels of SCFAs and butyric acid were significantly increased after XOS treatment in rats. Moreover, we found that the abundance of Prevotella and Paraprevotella positively correlated with the levels of butyric acid. These results indicated

that XOS regulated the composition of intestinal microbes and increased the abundance of butyric acid-producing bacteria such as Prevotella and Paraprevotella. In the MetS, CRC, Colitis, and other model experiments, the increase of Fecal butyrate content will reduce the expression of TNF-α, IL-2, NF-κB, and other pro-inflammatory factors (Rodriguez-Cabezas et al., 2002; Hijova et al., 2013; Hald et al., 2016). Studies have shown that butyric acid can inhibit the expression of PPARγ, thereby alleviating many PPARγ-related diseases including IBD (Kinoshita et al., 2002; Alex et al., 2013). As butyric acid is a known to inhibit inflammation (Manrique Vergara and Gonzalez Sanchez, 2017), increased butyric acid content can be assumed to alleviate obesityinduced inflammation.

Studies have shown that the composition of intestinal microflora in humans and rats is similar, and it is mainly composed of Firmicutes and Bacteroidetes at the level of the phylum. However, the proportion of Prevotellaceae and Prevotella in the human intestinal flora is reduced compared with rats. On the other hand, the individual differences of human intestinal microbes are also significantly higher than that of rats, which may be related to the diet and lifestyle of human subjects. As we know, diet is one of the most important elements in shaping gut microbes. Therefore, we are convinced that the dietary supplement of XOS can change the intestinal microbial structure of humans and increase the bacteria which can produce butyric acid to improve human intestinal immunity.

# CONCLUSION

Our study revealed that XOS alleviated obesity-led intestinal inflammation. We also found that XOS treatment increased the proportion of butyric acid producing bacteria in the rat intestine. Taken together, these results indicate that XOS protects against obesity-induced colonic inflammation by improving the intestinal microbial structure and increasing the abundance of SCFA-producing microbes.

#### DATA AVAILABILITY STATEMENT

The datasets generated for this study can be found in NCBI SRA https://www.ncbi.nlm.nih.gov/sra/?term=PRJNA593411.

#### ETHICS STATEMENT

fphys-10-01601 January 17, 2020 Time: 16:21 # 11

All animal procedures were performed in accordance with the Guidelines for Care and Use of Laboratory Animals of Hunan Agricultural University.

#### AUTHOR CONTRIBUTIONS

YF and YW performed the study and conducted the data analysis. YP, SL, and ZW designed the research. WW, DZ, and MX provided the assistance for the study. YF and ZW prepared the first draft of the manuscript. All authors read and revised the manuscript.

#### REFERENCES


#### FUNDING

This study was supported by the National Natural Science Foundation of China (No. 31071531), Hunan Provincial Natural Science Foundation (2019JJ40134), and Hunan Provincial Graduate Research and Innovation Project (CX2018B404).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphys. 2019.01601/full#supplementary-material

dogs. Regul. Toxicol. Pharmacol. 86, 379–385. doi: 10.1016/j.yrtph.2017. 04.009


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Fei, Wang, Pang, Wang, Zhu, Xie, Lan and Wang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fphys-10-01601 January 17, 2020 Time: 16:21 # 12