# MICROBIOME INTERPLAY AND CONTROL

EDITED BY : Christine Moissl-Eichinger, Gabriele Berg and Martin Grube PUBLISHED IN : Frontiers in Microbiology

#### Frontiers Copyright Statement

© Copyright 2007-2018 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.

The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.

Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.

Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.

As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.

All copyright, and all rights therein, are protected by national and international copyright laws.

The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use. ISSN 1664-8714 ISBN 978-2-88945-505-8 DOI 10.3389/978-2-88945-505-8

#### 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

# MICROBIOME INTERPLAY AND CONTROL

Topic Editors:

Christine Moissl-Eichinger, Medical University of Graz, Austria Gabriele Berg, Graz University of Technology, Austria Martin Grube, University of Graz, Austria

Root colonization by PSEUDOMONAS bacteria. Confocal laser scanning micrographs using fluorescence in situ hybridization with probes targeting bacterial cells of GAMMAPROTEOBACTERIA in the rhizosphere of PSEUDOMONAS-treated oilseed rape plants. Images courtesy of Birgit Wassermann, Institute of Environmental Biotechnology, Graz University of Technology, Austria.

In complex systems, such as our body or a plant, the host is living together with thousands of microbes, which support the entire system in function and health. The stability of a microbiome is influenced by environmental changes, introduction of microbes and microbial communities, or other factors. As learned in the past, microbial diversity is the key and low-diverse microbiomes often mirror out-ofcontrol situations or disease.

It is now our task to understand the molecular principles behind the complex interaction of microbes in, on and around us in order to optimize and control the function of the microbial community – by changing the environment or the addition of the right microorganisms.

This Research Topic focuses on studies (including e.g. original research, perspectives, mini reviews, and opinion papers) that investigate and discuss:

1) The role of the microbiome for the host/environmental system


The articles span the areas: human health and disease, animal and plant microbiomes, microbial interplay and control, methodology and the built environment microbiome.

Citation: Moissl-Eichinger, C., Berg, G., Grube, M., eds. (2018). Microbiome Interplay and Control. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-505-8

# Table of Contents

#### MICROBIOME IN HUMAN HEALTH AND DISEASE

*08 Molecular Characterization and Meta-Analysis of Gut Microbial Communities Illustrate Enrichment of Prevotella and Megasphaera in Indian Subjects*

Shrikant Bhute, Pranav Pande, Sudarshan A. Shetty, Rahul Shelar, Sachin Mane, Shreyas V. Kumbhare, Ashwini Gawali, Hemal Makhani, Mohit Navandar, Dhiraj Dhotre, Himangi Lubree, Dhiraj Agarwal, Rutuja Patil, Shantanu Ozarkar, Saroj Ghaskadbi, Chittaranjan Yajnik, Sanjay Juvekar, Govind K. Makharia and Yogesh S. Shouche

*22 Fungal ITS1 Deep-Sequencing Strategies to Reconstruct the Composition of a 26-Species Community and Evaluation of the Gut Mycobiota of Healthy Japanese Individuals*

Daisuke Motooka, Kosuke Fujimoto, Reiko Tanaka, Takashi Yaguchi, Kazuyoshi Gotoh, Yuichi Maeda, Yoki Furuta, Takashi Kurakawa, Naohisa Goto, Teruo Yasunaga, Masashi Narazaki, Atsushi Kumanogoh, Toshihiro Horii, Tetsuya Iida, Kiyoshi Takeda and Shota Nakamura


Derrick R. Samuelson, David A. Welsh and Judd E. Shellito


Shantelle Claassen-Weitz, Charles S. Wiysonge, Shingai Machingaidze, Lehana Thabane, William G. C. Horsnell, Heather J. Zar, Mark P. Nicol and Mamadou Kaba

*75* Bacteroides Fragilis *Lipopolysaccharide and Inflammatory Signaling in Alzheimer's Disease*

Walter J. Lukiw


#### ANIMAL MICROBIOME


Yue Sun, Liping Zhou, Lingdong Fang, Yong Su and Weiyun Zhu

*122 Insights into Abundant Rumen Ureolytic Bacterial Community Using Rumen Simulation System*

Di Jin, Shengguo Zhao, Pengpeng Wang, Nan Zheng, Dengpan Bu, Yves Beckers and Jiaqi Wang


Beng-Soon Teh, Johanna Apel, Yongqi Shao and Wilhelm Boland

*155 Bacterial Community and PHB-Accumulating Bacteria Associated With the Wall and Specialized Niches of the Hindgut of the Forest Cockchafer (*Melolontha Hippocastani*)*

Pol Alonso-Pernas, Erika Arias-Cordero, Alexey Novoselov, Christina Ebert, Jürgen Rybak, Martin Kaltenpoth, Martin Westermann, Ute Neugebauer and Wilhelm Boland

*168 Lower Termite Associations With Microbes: Synergy, Protection, and Interplay*

Brittany F. Peterson and Michael E. Scharf


Ana Lokmer, M. Anouk Goedknegt, David W. Thieltges, Dario Fiorentino, Sven Kuenzel, John F. Baines and K. Mathias Wegner

*204 Microbiomes of* Muricea Californica *and* M. Fruticosa*: Comparative Analyses of Two Co-Occurring Eastern Pacific Octocorals* Johanna B. Holm and Karla B. Heidelberg

#### PLANT MICROBIOME

*215 Advances in Research on* Epichloë *Endophytes in Chinese Native Grasses* Hui Song, Zhibiao Nan, Qiuyan Song, Chao Xia, Xiuzhang Li, Xiang Yao, Wenbo Xu, Yu Kuang, Pei Tian and Qingping Zhang


Song-Mei Shi, Ke Chen, Yuan Gao, Bei Liu, Xiao-Hong Yang, Xian-Zhi Huang, Gui-Xi Liu, Li-Quan Zhu and Xin-Hua He

*291 Microbiota Influences Morphology and Reproduction of the Brown Alga*  Ectocarpus *sp.*

Javier E. Tapia, Bernardo González, Sophie Goulitquer, Philippe Potin and Juan A. Correa

#### MICROBIAL INTERPLAY AND CONTROL


### METHODOLOGY

#### *325 Critical Issues in Mycobiota Analysis*

Bettina Halwachs, Nandhitha Madhusudhan, Robert Krause, R. Henrik Nilsson, Christine Moissl-Eichinger, Christoph Högenauer, Gerhard G. Thallinger and Gregor Gorkiewicz

	- Elisabeth Santigli, Slave Trajanoski, Katharina Eberhard and Barbara Klug

### BUILT ENVIRONMENT MICROBIOME

*377 Microbiome Interplay: Plants Alter Microbial Abundance and Diversity Within the Built Environment*

Alexander Mahnert, Christine Moissl-Eichinger and Gabriele Berg

*388 Microorganisms in Confined Habitats: Microbial Monitoring and Control of Intensive Care Units, Operating Rooms, Cleanrooms and the International Space Station*

Maximilian Mora, Alexander Mahnert, Kaisa Koskinen, Manuela R. Pausan, Lisa Oberauner-Wappis, Robert Krause, Alexandra K. Perras, Gregor Gorkiewicz, Gabriele Berg and Christine Moissl-Eichinger

*408 Functional Metagenomics of Spacecraft Assembly Cleanrooms: Presence of Virulence Factors Associated With Human Pathogens* Mina Bashir, Mahjabeen Ahmed, Thomas Weinmaier, Doina Ciobanu, Natalia Ivanova, Thomas R. Pieber and Parag A. Vaishampayan

## Molecular Characterization and Meta-Analysis of Gut Microbial Communities Illustrate Enrichment of Prevotella and Megasphaera in Indian Subjects

Shrikant Bhute1‡, Pranav Pande2‡, Sudarshan A. Shetty 2 †, Rahul Shelar <sup>2</sup> , Sachin Mane<sup>2</sup> , Shreyas V. Kumbhare<sup>2</sup> , Ashwini Gawali <sup>2</sup> , Hemal Makhani <sup>2</sup> , Mohit Navandar <sup>2</sup> , Dhiraj Dhotre<sup>2</sup> , Himangi Lubree<sup>3</sup> , Dhiraj Agarwal <sup>4</sup> , Rutuja Patil <sup>4</sup> , Shantanu Ozarkar <sup>5</sup> , Saroj Ghaskadbi <sup>1</sup> , Chittaranjan Yajnik <sup>3</sup> , Sanjay Juvekar <sup>4</sup> , Govind K. Makharia<sup>6</sup> and Yogesh S. Shouche<sup>2</sup> \*

<sup>1</sup> Department of Zoology, Savitribai Phule Pune University, Pune, India, <sup>2</sup> Microbial Culture Collection, National Centre for Cell Sciences, Savitribai Phule Pune University campus, Pune, India, <sup>3</sup> Diabetes Unit, KEM Hospital Research Centre, Pune, India, <sup>4</sup> Vadu Rural Health Program, KEM Hospital Research Centre, Pune, India, <sup>5</sup> Department of Anthropology, Savitribai Phule Pune University, Pune, India, <sup>6</sup> Department of Gastroenterology and Human Nutrition, All India Institute of Medical Sciences, New Delhi, India

The gut microbiome has varied impact on the wellbeing of humans. It is influenced by different factors such as age, dietary habits, socio-economic status, geographic location, and genetic makeup of individuals. For devising microbiome-based therapies, it is crucial to identify population specific features of the gut microbiome. Indian population is one of the most ethnically, culturally, and geographically diverse, but the gut microbiome features remain largely unknown. The present study describes gut microbial communities of healthy Indian subjects and compares it with the microbiota from other populations. Based on large differences in alpha diversity indices, abundance of 11 bacterial phyla and individual specific OTUs, we report inter-individual variations in gut microbial communities of these subjects. While the gut microbiome of Indians is different from that of Americans, it shared high similarity to individuals from the Indian subcontinent i.e., Bangladeshi. Distinctive feature of Indian gut microbiota is the predominance of genus Prevotella and Megasphaera. Further, when compared with other non-human primates, it appears that Indians share more OTUs with omnivorous mammals. Our metagenomic imputation indicates higher potential for glycan biosynthesis and xenobiotic metabolism in these subjects. Our study indicates urgent need of identification of population specific microbiome biomarkers of Indian subpopulations to have more holistic view of the Indian gut microbiome and its health implications.

Keywords: Indian subjects, 16S rRNA amplicon, qPCR, Prevotella and Megasphaera

#### Edited by:

Nicole Webster, Australian Institute of Marine Science, Australia

#### Reviewed by:

David William Waite, Ministry for Primary Industries, New Zealand Anne Schöler, Helmholtz Zentrum München, Germany

#### \*Correspondence:

Yogesh S. Shouche yogesh@nccs.res.in

#### †Present Address:

Sudarshan Shetty, Laboratory of Microbiology, Wageningen University, Netherlands

‡

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: 23 November 2015 Accepted: 21 April 2016 Published: 09 May 2016

#### Citation:

Bhute S, Pande P, Shetty SA, Shelar R, Mane S, Kumbhare SV, Gawali A, Makhani H, Navandar M, Dhotre D, Lubree H, Agarwal D, Patil R, Ozarkar S, Ghaskadbi S, Yajnik C, Juvekar S, Makharia GK and Shouche YS (2016) Molecular Characterization and Meta-Analysis of Gut Microbial Communities Illustrate Enrichment of Prevotella and Megasphaera in Indian Subjects. Front. Microbiol. 7:660. doi: 10.3389/fmicb.2016.00660

### INTRODUCTION

The gut microbial ecosystem is known to be governed by ecological and evolutionary forces (Ley et al., 2006) and is often controlled by secretions from the host at intestinal epithelium-microbiota interface such that beneficial microbes are maintained (Schluter and Foster, 2012). The physiological diversity of gut microbiota and its role in human health has been an inspiration for the initiation of elite projects such as Human Microbiome Project (HMP; Turnbaugh et al., 2007) and Metagenomics of the Human Intestinal Tract (MetaHIT) project (Qin et al., 2010). These projects and other related studies have generated wealth of information suggesting a link between gut microbiota and their genomic capabilities in maintenance of general wellbeing (Cho and Blaser, 2012) and also in highly specialized functions such as development of the immune system (Chung et al., 2012), neurodevelopmental disorders (Hsiao et al., 2013), and xenobiotic metabolism (Maurice et al., 2013).

Studies in past few years have highlighted discernible patterns of gut microbiota and microbiome in geographically separated populations (Mueller et al., 2006; De Filippo et al., 2010; Nam et al., 2011; Yatsunenko et al., 2012). Such studies are important in light of possible role of gut microbiota in the modulation of efficacy of oral vaccines (Valdez et al., 2014). In addition, action of pre and probiotics varies based on type of prebiotic, strain of probiotics used and possibly host gut environment (Boyle et al., 2006). Population specific microbiota studies such as American Gut, Canadian Microbiome, Brazilian Microbiome project and others are likely to yield valuable information about the gut microbiota as a target for medical interventions, may be in the form of fecal microbial transplantation to restore the healthy state (Borody et al., 2014).

Indian population is a unique conglomeration of genetically diverse groups having varied dietary habits and residing in vast geographic locations (Basu et al., 2003; Xing et al., 2010). In addition to these genetic differences, Indians have distinctive metabolic (Shukla et al., 2002) and anthropometric features (Yajnik et al., 2003; Prasad et al., 2011). Moreover, Indians are also confronted with the double burden of underand over-nutrition primarily due to the income inequalities (Subramanian et al., 2007). In this study, we provide detailed account of prominent attributes of the Indian gut microbial composition and its functions from 34 healthy Indian subjects. We carried out 16S rRNA gene amplicon sequencing using different sequencing platforms viz. Ion Torrent PGM and Illumina HiSeq. We then combined the 16S rRNA amplicon data of Indian subjects together with American (Muegge et al., 2011), Korean (Nam et al., 2011), Spanish (Peris-Bondia et al., 2011), and Bangladeshi (Lin et al., 2013) to compare it with gut microbiota of these populations. In addition, considering the response of gut microbiota to different types of diets; we compared Indian gut microbiota with non-human primates including hind-gut-fermenters, fore-gut-fermenters, herbivorous, and carnivorous organisms (Ley et al., 2008).

### MATERIALS AND METHODS

### Study Population, Sample Collection, and DNA Extraction

We included 34 healthy Indian subjects from two urban cities: Delhi and Pune (one from Northern and one from Western part) of India and nearby rural regions of these cities. These cities are characterized by diverse groups of individuals from different parts of the country. Institutional Ethical Committee of National Centre for Cell Science approved the study and informed consent was obtained form all the participants. Although, this was not a clinical trial, we followed all good clinical practices as per Indian Council of Medical Research guidelines while recruiting the subjects and throughout the study. Fecal samples were collected from all of the subjects and stored at −80◦C until DNA extraction. Total community DNA was extracted from each fecal sample using QIAmp DNA Stool Mini kit (Qiagen, Madison USA) as per manufacturer's instructions.

#### 16S rRNA Gene Amplicon Sequencing

16S rRNA amplicon sequencing of samples from Western region was performed using Ion Torrent PGM and that from Northern region using Illumina Hiseq2000 sequencing technology. For Ion Torrent PGM sequencing, samples were processed as follows: PCR was set up in 50µl reaction using AmpliTaq Gold PCR Master Mix (Life Technologies, USA) and with 16S rRNA V3 region specific bacterial universal primers: forward primer 341F (5′ -CCTACGGGAGGCAGCAG-3 ′ ) and reverse primer 518R (5′ -ATTACCGCGGCTGCTGG-3′ ; Bartram et al., 2011). Following conditions were used for PCR: initial denaturation at 95◦C for 4 min, followed by 20 cycles of 95◦C for 1 min, 56◦C for 30 s, and 72◦C for 30 s with a final extension at 72◦C for 10 min. PCR products were purified using Agencourt AMPure XP DNA purification Bead (Beckman Coulter, USA), end repaired and ligated with specific barcode adaptor as explained in Ion XpressTM Plus gDNA Fragment Library Preparation user guide. Fragment size distribution and molar concentrations of amplicon were assessed on a Bioanalyzer 2100 (Agilent Technologies, USA) using High Sensitivity DNA Analysis Kit as per manufacturer's instructions. Emulsion PCR was carried out on diluted and pooled amplicon (10 samples in each pool) using the Ion OneTouchTM 200 Template Kit v2 DL (Life Technologies). Sequencing of the amplicon libraries was carried out on 316 chips using the Ion Torrent PGM system and Ion Sequencing 200 kit (Life Technologies). For Illumina sequencing, samples were processed as follows: A PCR reaction of 50µl was set up using AmpliTaq Gold high fidelity polymerase (Life Technologies, USA) and PCR conditions were as follows: initial denaturation at 95◦C for 10 min; followed by 30 cycles of 95◦C for 30 s; 56◦C for 30 s; and 72◦C for 30 s. The final extension was set at 72◦C for 7 min. The PCR products were purified using gel elution and the eluted products were used for library preparation. The libraries were quantified on Bioanalyzer using the DNA high sensitivity LabChip kit (Agilent Technologies, USA) and sequenced using Illumina HiSeq2000 (2x150 PE).

### Sequence Processing and Bioinformatics Analysis

All PGM and Illumina HiSeq reads were pre-processed using Mothur pipeline (Schloss et al., 2009) with following conditions: minimum 150 bp to maximum 200 bp, maximum homopolymer–5, maximum ambiguity–0, and average quality score–20. This way we derived total of ∼17 million high quality amplicon reads from 34 samples, which we pooled into single FASTA file for further analysis in QIIME: Quantitative Insights Into Microbial Ecology (Caporaso et al., 2010). Closed reference based OTU picking approach was used to cluster reads into Operational Taxonomic Units (OTUs) at 97% sequence similarity using UCLUST algorithm (Edgar, 2010) and a representative sequence from each OTU was selected for downstream analysis. All OTUs were assigned to the lowest possible taxonomic rank by utilizing RDP Classifier 2.2 (Wang et al., 2007) and Greengenes database 13.8 with a confidence score of at least 80%. Estimations of Core OTUs were done as described previously (Huse et al., 2012). Various estimates of alpha diversity such as Chao1, PD whole tree, Simpson, and Shannon were applied on rarefied sequence count (1181 sequence per sample) and UniFrac was used as beta diversity measures to understand the microbial communities in Indian individuals. UniFrac analysis is known to be affected by sequencing depth and evenness, therefore, we performed jackknifing in which samples are subjected to even subsampling for n replicates and UniFrac distance matrix is calculated for each replicate (Lozupone and Knight, 2005). This way we generated 1000 replicates of PCoA coordinates and Procrustes analysis was applied to each PCoA replicate to plot average position of individuals on PCoA plot. The interquartile range of the distribution of points among the replicates was represented as an eclipse around the point (Lozupone et al., 2011).

### qPCR Based Quantification of Dominant OTUs

The abundance of intestinal bacterial groups belonging to genus Prevotella, Faecalibacterium, and Megasphaera were measured by absolute quantification of 16S rRNA gene copy number by using primers listed in Supplementary Table 1. Template concentration for each sample was initially adjusted to 50 ng/µl. qPCR amplification and detection were performed in 10µl reaction (consisting of 5µl Power SYBR Green PCR Master Mix, 0.1µM of each specific primer and 1µl template) in triplicate using 7300 Real time PCR system (Applied Biosystems Inc., USA). Following conditions were used for qPCR assays: one cycle of 95◦C 10 min followed by 40 cycles of 95◦C for 15 s and 60◦C for 1 min. Group specific standard curves were generated from 10 fold serial dilutions of a known concentration of PCR products for each group. Average values of the triplicate were used for enumerations of 16S rRNA gene copy numbers for each group using standard curves generated (Marathe et al., 2012). Percent abundance of each genus was obtained by calculating ratio of copy number of that genus to that of total bacteria. Throughout the qPCR experiments efficiency was maintained above 90% with a correlation coefficient >0.99.

### Imputation of Metagenome Using PICRUSt

The metagenome imputation was done using method as described earlier (Langille et al., 2013). Briefly, closed reference based OTU picking approach was utilized to bin the amplicon sequences using latest Greengenes database 13.5 at 97% sequence similarity cut-off. The normalization for 16S rRNA gene copy number was carried out before prediction of the metagenome. This OTU table was used for predicting metagenome at three different KEGG levels (L1 to L3). Metagenomic differences between Indians-Americans as well as Indian-non-human primates were analyzed using linear discriminant analysis (LDA) effect size (LEfSe; Segata et al., 2011). PICRUSt and LEfSe analysis were performed with available parameters at http://huttenhower.sph.harvard.edu/galaxy/.

### Publically Available Data Used

We did a PubMed search restricted only to publically available 16S rRNA amplicon data. Upon further narrowing down our search, we obtained raw sequence data of Korean subjects (DDJB project ID 60507; Nam et al., 2011), Bangladeshi subjects (SRA-SRA057705; Lin et al., 2013), data of 18 American individuals and 33 non-human primates (MG-RAST qiime625 and qiime626; Muegge et al., 2011) and data of Spanish individuals (SRA-SRP005393; Peris-Bondia et al., 2011). The list of primers, variable region of 16S rRNA gene and sequencing technology for each of the study is listed in Supplementary Table 2. Any previously reported sequence data for Indian population was not available. To avoid biases introduced due to respective studies describing microbiota of these populations and inter-individual variations, sequence data of all individuals from a study was merged and considered as a representative microbiota of that country. The raw data from all these samples was processes along with the Indian sequence data (Ion Torrent and Illumina amplicons) in the same way as explained earlier.

### Additional Statistical Tests

We applied Good's coverage to have a sense of understanding that the sequencing we have performed was enough to cover microbial diversity in the samples studied (Good, 1953). We also applied Welch's t-test with Benjamini-Hochberg FDR correction to examine the significantly differing bacterial families between Indians and Americans and Kruskal-Wallis test (a non-parametric measure of variance) to examine the population specific OTUs. Similar comparisons were made to evaluate the differential OTUs among non-human primates and Indians. Random Forest, a supervised machinelearning approach was applied to our data sets to identify taxa that were indicators for community differences in Indians-Americans as well as Indian-non-human primates (Knights et al., 2011; Yatsunenko et al., 2012). An OTU was given importance scores by estimating amount of error introduced if that OTU is removed from the set of indicator taxa.

### RESULTS

#### Key Features of Indian Gut Microbiota

We obtained over 17 million good quality reads which were clustered into 3782 OTUs from the 34 healthy Indian individuals, for further analysis the sequences were normalized to 1181 per sample (Supplementary Table 3). We first employed Good's coverage in order to estimate that enough sequencing has been performed to address the gut microbial diversity; with mean Good's coverage of 94% ±0.03, we were convinced of capturing dominant OTUs in all study subjects and to comment on gut microbial features of them.

We used alpha diversity indices to understand community composition of gut microbiota, some of which were based on species richness and species abundance and some on phylogenetic distance between them. Alpha diversity indices

FIGURE 1 | (A) Variation in alpha diversity indices in Indian Subjects. (B) Abundance of dominant bacterial phyla in Indian subjects. Subjects are separated and shown according to sequencing platform used. Samples with prefix PMS are from rural region and rest are from urban region. (C) Unweighted and (D) weighted UniFrac PCoA bi-plots; the gray colored sphere represent a taxonomic group that influence clustering of samples in particular area of the PCoA plot and its size demonstrate abundance of that taxonomic group (Rural samples are encircled). Colors indicate the sequencing technology used. Red: Illumina, Green: Ion Torrent PGM.

#### TABLE 1 | Showing the core OTUs found in Indian subjects.


such as Chao1, Shannon, Simpson, and PD\_Whole tree revealed that there were large differences in the community composition in study subjects under consideration (**Figure 1A**). Upon comparison of alpha diversity indices between rural and urban population, it was observed to be higher in urban subjects, however, no significant differences were noted for alpha diversity indices with respect to sequencing technology used. Overall, we could detect 201 bacterial genera belonging to 11 bacterial phyla in Indian subjects (**Figure 1B**). Upon closer examination of the OTU table we were able to detect 50 OTUs that were present across the samples, such OTUs are commonly termed as core OTUs (**Table 1**). Presence of just 50 core OTUs suggest that the gut microbiome of Indians is very diverse. This was further confirmed by performing beta diversity analysis using unweighted (sensitive to presence of unique OTUs) and weighted (sensitive to the abundance) UniFrac distance matrices. In each case, jackknifed PCoA biplots were produced to illustrate the compositional variation in gut microbiota between the samples; position of each sample is the average of jackknifed replicate shown with ellipses representing the IQR in each axis. Presence of large ellipses around each sample sphere in unweighted PCoA plot (**Figure 1C**) is indicative of variations on beta diversity measures due to random subsampling and thus the presence of unique OTUs particular to each individual. Interestingly, we also noted that the samples that were happened to be collected from rural areas (eight samples on the right side of **Figure 1C**) clustered separately from the urban samples on unweighted PCoA plot indicating the contribution of lifestyle associated factors on sample segregation. However, on weighted PCoA plot (**Figure 1D**), all samples found scattered indicating the abundance of taxa influencing the segregation of samples on weighted PCoA plot was not different among the samples. Further, from the taxa contributing sample segregation of PCoA plots and from core OTUs, it was noticed that the gut microbiota of Indians is highly enriched with the OTUs belonging to bacterial genera Prevotella and Megasphaera and bacterial families such as Lachnospiraceae, Ruminococcaceae, and Veillonellaceae.

To confirm that Indian gut microbiota is enriched with Prevotella and Megasphaera OTUs, we carried out qPCR assays for absolute quantification of 16S rRNA copy number of these genera in the study subjects. Mean count of Prevotella and Megasphaera was found to be 4.45% and 8.45%, respectively of total bacterial count. On the contrary, Faecalibacterium mean count was as low as 0.63% of total bacterial count (**Figure 2**). Interestingly, based on absolute count of Prevotella and Megasphaera Indian subjects were demarcated into two groups, one with moderate and other with high copy number of these genera. These results confirmed the 16S rRNA gene amplicon analysis and signify the dominance of Prevotella and Megasphaera in Indians.

#### Quantitative Differences between Gut Microbiota of Indians and Americans

The mean abundance of bacterial phyla and families between Indians and Americans was compared using t-test. Significant

expressed as percent abundance of Prevotella, Faecalibacterium and Megasphaera to that of the total bacteria for each of the sample.

differences were observed in four dominant phyla in these populations: Actinobacteria (P = 0.0003), Bacteroidetes (P = 0.029), and Proteobacteria (P = 0.0015) being significantly more abundant in Indians and Firmicutes (P = 0.0004) in Americans (**Figure 3A**). At family level, 11 families were observed to be significantly different in the two populations (**Figure 3B** and Supplementary Table 4). Prevotellaceae, Lactobacillaceae, Veillonellaceae, Bifidobacteriaceae, Enterobacteriaceae, Streptococcaceae, Peptostreptococcaceae, Leuconostocaceae, Micrococcaceae, Carnobacteriaceae, and Gemellaceae were more dominant in Indians (P < 0.05) whereas Lachnospiraceae, Ruminococcaceae, Bacteroidaceae, Coprobacillaceae, Porphyromonadaceae, Rikenellaceae, Erysipelotrichaceae, Desulphovibrionaceae, and Christensenellaceae were more dominant in Americans (P < 0.05). Kruskal-Wallis test revealed total of 127 OTUs differed significantly (P < 0.01) between the two populations (Supplementary Table 5) of which 50 were unique to Indians and 475 to Americans. Principal coordinate analysis (PCoA) based on weighted and unweighted UniFrac distance matrices demonstrated that the Americans clustered along coordinate 1 (**Figures 4A,B**) as against the Indians who were found dispersed along the coordinate 2. For Random Forest analysis, we considered an OTU to be highly predictive if its importance score was at least 0.001, this revealed 76 highly predictive OTUs between the two populations (Supplementary Table 6). Among these 76 highly predictive OTUs, 6 were overrepresented in Indians while rest were overrepresented in Americans. The OTUs overrepresented in Indians belonged to genus Prevotella, Lactobacillus, Lachnospira and Roseburia. Our results highlight profound differences at various taxonomic levels in gut microbial community structure of the two populations.

We further analyzed the differences in gut microbiota of Indian, Bangladeshi, American, Korean and Spanish populations in terms of unique and shared bacterial families plus the

OTUs among these populations. For this, we normalized the sequence data to 4389 sequences per sample which contributed to 1807 OTUs. At bacterial family level, Indians shared more families with Bangladeshis; while fewer with Americans, Koreans and Spanish (**Figure 5A**). With 460 unique OTUs (Supplementary Table 7), Indians shared maximum of 25 OTUs with Bangladeshi, 15 with Americans and Spanish while 7 with Koreans (**Figure 5B**). Most of the shared OTUs between Indians and Bangladeshis belonged to families Lachnospiraceae, Ruminococcaceae and Enterobacteriaceae, and genus Prevotella (Supplementary Table 8). Interestingly, only 3 OTUs were common in all populations, which were contributed by Streptococcaceae and Enterobacteriaceae families.

### Indians Share Microbiota with Omnivorous Mammals

For the comparison of gut microbiota of Indians with nonhuman primates, we normalized the sequences to 1181 sequences per sample, which constituted 6189 OTUs. We observed that Indians share maximum 68 of 236 bacterial families (**Figure 6A**) and 112 OTUs with omnivorous mammals (Supplementary Table 9) while minimum of 32 OTUs with carnivorous mammals. Interestingly, only 2 OTUs were common in all non-human primates and Indians (**Figure 6B**). Further, principal coordinate analysis (PCoA) based on unweighted and weighted UniFrac distance matrices showed scattered distribution of omnivorous samples (**Figures 4C,D**). On the contrary, herbivorous (hind gut and fore gut fermenters), and carnivorous clustered separately

(**Figure 4C**). Random Forest analysis of Indians and non-human primates revealed 652 highly discriminants OTUs. Of the 341 OTUs, 122 and 174 OTUs were overrepresented in Indians and omnivorous mammals, respectively (Supplementary Table 10).

#### Imputed Metagenome

For comparing functional potential of the microbial communities in Indians and Americans, we used PICRUSt (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States). PICRUSt uses extended ancestralstate reconstruction algorithm to estimate which gene families are present and then combines gene families to give complete metagenome of the samples. From the data of functional capabilities, we focused primarily on those, which are associated with the microbial metabolism. We noticed significant differences in all major metabolic functions in gut microbiome of Indians and Americans. Broadly, gene families associated with xenobiotic biodegradation, nucleotide metabolism, enzyme families, metabolism of terpenoids and polyketides, glycan biosynthesis, and metabolism were overrepresented in Indians, whereas, metabolic functions associated with energy metabolism, carbohydrate metabolism, amino acid metabolism, and biosynthesis of other secondary metabolites were overrepresented in

FIGURE 5 | (A) Family level distribution of bacterial taxa among: Indian, Bangladeshi, American, Korean, and Spanish population. (B) Venn diagram demonstrating overlap of OTUs at 97% sequence similarity cut off among these populations.

Americans (**Figure 7A**). Further, the metagenomic comparison between Indians and non-human primates revealed that gene families linked to energy harvesting potential such as carbohydrate metabolism, glycolysis-gluconeogenesis, and fatty acid biosynthesis were enriched in omnivorous mammals (**Figure 7B**).

#### DISCUSSION

Studies concerning population specific microbiota have revealed peculiar patterns in distribution of specific microbial communities in their gut. Surprisingly, till date, no efforts have been made to understand specific features of the microbiota of healthy Indian subjects. Based on 16S rRNA data from 34 individuals and 3782 OTUs, in this work, we first systematically describe gut microbiota features in Indian subjects. We suggest vast inter-individual variation in gut microbial communities in these subjects, characterized by dominance of Prevotella and Megasphaera. We further demonstrate the graded difference in microbial communities of these subjects from neighboring country (Bangladeshi) to distant population (Americans) as well as show that they indeed share most of the microbiota with omnivorous animals.

Our observation of compositional and phylogenetic variation within Indian gut microbiota as revealed by alpha diversity indices, could be a result of different variables like biogeographic separations of individuals (like rural-urban setting) and associated life-style factors. Further, we noted large variation in alpha diversity indices in urban individuals. Thus, to check whether there are unique taxa responsible for this, we performed UniFrac based beta diversity analysis. Indeed, the separation between rural and urban subjects observed on unweighted UniFrac PCoA which is influenced by less abundant unique OTUs was lost on weighted UniFrac PCoA because of abundance of dominant OTUs. The distinct separation was also not evident at phyla level abundance. On PCoA bi-plots, we further showed the contribution of dominant taxonomic groups influencing the segregation of samples. Thus, our results are robust and proves the presence of individual specific OTUs; at the same time it confirms that Indian subjects could not be separated into two or more groups based on presence and abundance of dominant taxonomic groups.

Knowing the fact that gut microbiota is influenced by diet and geography, we extended our analysis and compared gut microbiota of Indian subjects with American gut microbiota. Based on composition of microbiota, Americans were closely clustered while Indians were found dispersed on PCoA-biplots. This distinctive clustering could be partly because of genetic make-up and largely due the calorie restricted diet that these subjects were following. Interestingly, in an another study Americans from metropolitan areas which were not on any specific diet, segregated distantly from those of Malawians and Amerindians and were clustered closely (Yatsunenko et al., 2012). This provides the clue that though the cohort was calorie restricted, gut microbiota of Americans is indeed different from gut microbiota of other communities. Thus, diet can be one of many factors which influence the gut microbial communities and other factors such as genetic make-up and other current practices could also have a major influence on gut microbial composition. On broader scale, Indian population which originated from first wave of modern humans Out-of-Africa following the coastal route; and American population, which is effectively descendants of post-Columbian European migrants (Lazaridis et al., 2014), are genetically different hosts

with varied dispersal histories (Macaulay et al., 2005; Mellars et al., 2013). The lack of cohesive Indian population cluster may be due to the heterogeneous representation of Indian samples from different endogamous groups experiencing diverse dietary patterns, prescriptions-proscriptions for food and food taboos that vary culturally.

Upon analysing the differentiating bacterial lineages and contributors in PCoA-biplots, we discovered that the OTUs belonging to genus Prevotella, Lactobacillus, Bifidobacterium, and Megasphaera were discriminately abundant in Indians. Members of genus Prevotella are known for their ability to degrade complex plant polysaccharides (De Filippo et al., 2010), thus its high abundance in Indian gut microbiota could be a result of the nature of Indian diet, which is primarily rich in plant derived preparations (Vecchio et al., 2014). Predominance of members of Lactobacillus and Bifidobacterium could be explained by the fact that fermented foods are another major components in Indian diet; these fermented foods are good source of lactic acid bacteria (Satish Kumar et al., 2013). Members of genus Megasphaera, a normal inhabitant of ruminant gut, have been isolated by us from gut microbiota of Indians (Shetty et al., 2013). The genome analysis and physiological characterization of these Megasphaera isolates highlighted their ability to produce short chain fatty acids viz. propionate, acetate, and butyrate and vitamins like of cyanocobalamin. One of the interesting observations of our study is the demarcation of Indian individuals into two groups (moderate and high copy number of Prevotella and Megasphaera). Recently, bimodal bacteria (with low and high abundance groups) in more than 1000 western individuals were reported and were predicted to be key bacterial groups associated with host health (Lahti et al., 2014). Considering the metabolic features of Prevotella and Megasphaera explained earlier and effect of different environmental factors on microbiota, they can be represented as tipping elements in Indian gut microbiota and are possibly linked with general well-being of these subjects as all the participants were healthy. However, further analysis would be needed to confirm the bimodal nature of these groups. Further we obtained the evidence for variations in gut microbiota of Indians by comparing it with gut microbiota of Spanish, Korean, Bangladeshi and American population, which are unique with respect to their dietary patterns and biogeographic locations. Indians shared maximum taxonomic groups with next-door neighbor Bangladeshi, which became progressively less with American, Spanish, and Koreans. High similarity shared between gut microbiota of Indian and Bangladeshi population is a reflection of shared ethnicity and other life-style factors between these populations. Interestingly, Indians shared least OTUs with Korean, which in turn shared maximum OTUs with Americans is in accordance with observations of previous study (Nam et al., 2011). The most intriguing finding of this analysis however, was the presence of only three common OTUs amongst all the populations, strengthening the fact that gut microbiome of geographically separated population is indeed unique and very few OTUs may contributes to core microbiome of the global population (Huse et al., 2012).

In the meta-analysis of microbial studies often comparisons are made between the data generated using different

97% sequence similarity cut off among Indians and non-human primates.

experimental protocols, hence a critical question is whether the principal conclusions derived are because of the technical differences or they are indeed biologically meaningful? Taking into account the effect of different experimental protocols including method of DNA extraction, use of specific primers and sequencing technologies, it cannot be denied that these factors could introduce some bias in the observed results (Lozupone et al., 2013). However, by the use of more stringent approach during bioinformatics analysis of amplicon (as presented in the current manuscript), it is possible to reduce such biases. The results presented in **Figure 1C** indicate that the segregation of samples is not due to the sequencing technologies used, but are indeed due to the large compositional differences in microbiota. Thus, such comparisons are required to identify the influence of these factors on the observed results and will bring into light the ways of optimizing the analysis protocol in order to minimize the effect of such confounding factors.

One of the major life-style factors, which characterize a population, is its dietary habits. There are abundant evidences in the literature suggesting effect of diet on microbiota (David et al., 2014; Xu and Knight, 2014). We therefore hypothesized, that gut microbiota of Indians who typically display mixed vegetarian and non-vegetarian dietary habits may be alike omnivorous mammals. The observation of the present study regarding similarities of gut microbiome of Indians and omnivorous mammals are in congruent to previous study findings (Ley et al., 2008; Muegge et al., 2011). In a study, Ley et al. showed that indigenous gut microbial communities co-diversify with their hosts and the microbial diversity increases from carnivory to omnivory to herbivory. Moreover, presence of only two common OTUs amongst all the types of dietary patterns, hint toward subtle differences and rapid trade-offs in gut microbial communities shaped by evolutionary forces in response to animal and plant diets.

Metagenomic studies of gut microbiome suggests that microbes residing in the gut have enormous genetic potential to code for functions essential for them to thrive in the gut environment and maintain homeostasis of gut ecosystem (Qin et al., 2010). To the best of our knowledge, report on experimentally derived human gut metagenomic data from adult Indian individuals is unavailable. In this context, our metagenomic imputations become minimum essential to have first glimpse at the functional capabilities of Indian individuals. Our metagenomic imputations using PICRUSt followed by LEfSe analysis reveals vast diversity in metabolic functions in these subjects. Although, the findings of differences in metabolic capabilities among the Indians-Americans and Indians-Nonhuman primates are based on imputed metagenome and has some limitations as explained earlier (Langille et al., 2013), we were able to capture broader functional features in gut microbiota and correlate it with the taxonomic features. Higher abundance of Bacteroidetes are generally attributed to ability to degrade xenobiotics like antibiotics (Maurice et al., 2013) and metabolism of complex glycans (Martens et al., 2009) Whereas, the Firmicutes are related to increased energy harvest through excessive carbohydrate metabolism and production of SCFAs (Turnbaugh et al., 2006). High Bacteroidetes and low Firmicutes found in Indian subjects and their correlation with metabolic abilities, indeed suggests that their gut microbiota not only differ at taxonomic level but also at the functional levels from that of Americans.

### CONCLUSION

Our study raises the exciting possibility that the difference in microbiota may contribute to differences in health and disease characteristics of Indian population that could be different compared to the observations in the western population. Findings of the present study will serve as a basis for large-cohort studies in near future on Indian Gut Microbiome to address the questions such as if there are specific bacterial taxa or microbial functions which can be treated as a potential target for medical intervention studies.

### AUTHOR CONTRIBUTIONS

YS and SJ conceptualized and designed the study whereas, YS also coordinated it. SB, SM, RS performed Ion torrent PGM sequencing. SS performed Illumina Hiseq sequencing. SK, AG, HM downloaded all relevant 16S rRNA sequence data. SB, PP carried out the detailed bioinformatics analysis. MN wrote the specific PERL script for bioinformatics analysis. DD coordinated the bioinformatics analysis. SO provided the anthropological insight on Indian context. SB, SG, HL, CY, DA, RP, SJ, and GM were involved in sample collection. SB and PP wrote the manuscript and all authors edited and approved the manuscript.

## AVAILABILITY OF SEQUENCE DATA

Ion Torrent PGM runs were deposited to NCBI SRA under the accession numbers SRP041693, SRP055407 and Illumina raw reads to DDBJ under the accession number DRA002238.

### ACKNOWLEDGMENTS

Authors are grateful to Diabetes Unit and Vadu Unit of KEM Hospital Research Centre, Pune; Department of Gastroenterology and Human Nutrition, AIIMS, New Delhi, India for their help in subject recruitment and sample collection. Authors are thankful to Dr. Surendra Ghaskadbi (Head, Animal Sciences Division, Agharkar Research Institute, Pune), for critically reviewing the manuscript. The study was funded by Department of Biotechnology, Government of India, under the grant: Microbial Culture Collection Project (BT/PR10054/NDB/52/94/2007). SB is thankful to the University Grant Commission (UGC) for providing financial assistance for carrying out the research work.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.00660

#### REFERENCES


**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 © 2016 Bhute, Pande, Shetty, Shelar, Mane, Kumbhare, Gawali, Makhani, Navandar, Dhotre, Lubree, Agarwal, Patil, Ozarkar, Ghaskadbi, Yajnik, Juvekar, Makharia and Shouche. 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) or licensor 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.

# Fungal ITS1 Deep-Sequencing Strategies to Reconstruct the Composition of a 26-Species Community and Evaluation of the Gut Mycobiota of Healthy Japanese Individuals

Daisuke Motooka<sup>1</sup>† , Kosuke Fujimoto2,3† , Reiko Tanaka<sup>4</sup> , Takashi Yaguchi<sup>4</sup> , Kazuyoshi Gotoh1,5, Yuichi Maeda2,3, Yoki Furuta<sup>2</sup> , Takashi Kurakawa<sup>2</sup> , Naohisa Goto<sup>1</sup> , Teruo Yasunaga<sup>1</sup> , Masashi Narazaki<sup>3</sup> , Atsushi Kumanogoh<sup>3</sup> , Toshihiro Horii<sup>1</sup> , Tetsuya Iida<sup>1</sup> , Kiyoshi Takeda<sup>2</sup> and Shota Nakamura<sup>1</sup> \*

#### Edited by:

Martin Grube, University of Graz, Austria

#### Reviewed by:

Nicola Segata, University of Trento, Italy Gerhard G. Thallinger, Graz University of Technology, Austria

#### \*Correspondence:

Shota Nakamura nshota@gen-info.osaka-u.ac.jp

†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: 21 January 2016 Accepted: 03 February 2017 Published: 15 February 2017

#### Citation:

Motooka D, Fujimoto K, Tanaka R, Yaguchi T, Gotoh K, Maeda Y, Furuta Y, Kurakawa T, Goto N, Yasunaga T, Narazaki M, Kumanogoh A, Horii T, Iida T, Takeda K and Nakamura S (2017) Fungal ITS1 Deep-Sequencing Strategies to Reconstruct the Composition of a 26-Species Community and Evaluation of the Gut Mycobiota of Healthy Japanese Individuals. Front. Microbiol. 8:238. doi: 10.3389/fmicb.2017.00238 <sup>1</sup> Department of Infection Metagenomics, Genome Information Research Center, Research Institute for Microbial Diseases, Osaka University, Suita, Japan, <sup>2</sup> Laboratory of Immune Regulation, Department of Microbiology and Immunology, Graduate School of Medicine, WPI Immunology Frontier Research Center, Osaka University, Suita, Japan, <sup>3</sup> Department of Respiratory Medicine, Allergy and Rheumatic Diseases, Graduate School of Medicine, Osaka University, Suita, Japan, <sup>4</sup> Division of Bio-resources, Medical Mycology Research Center, Chiba University, Chiba, Japan, <sup>5</sup> Department of Bacteriology, Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, Okayama University, Okayama, Japan

The study of mycobiota remains relatively unexplored due to the lack of sufficient available reference strains and databases compared to those of bacterial microbiome studies. Deep sequencing of Internal Transcribed Spacer (ITS) regions is the de facto standard for fungal diversity analysis. However, results are often biased because of the wide variety of sequence lengths in the ITS regions and the complexity of high-throughput sequencing (HTS) technologies. In this study, a curated ITS database, ntF-ITS1, was constructed. This database can be utilized for the taxonomic assignment of fungal community members. We evaluated the efficacy of strategies for mycobiome analysis by using this database and characterizing a mock fungal community consisting of 26 species representing 15 genera using ITS1 sequencing with three HTS platforms: Illumina MiSeq (MiSeq), Ion Torrent Personal Genome Machine (IonPGM), and Pacific Biosciences (PacBio). Our evaluation demonstrated that PacBio's circular consensus sequencing with greater than 8 full-passes most accurately reconstructed the composition of the mock community. Using this strategy for deepsequencing analysis of the gut mycobiota in healthy Japanese individuals revealed two major mycobiota types: a single-species type composed of Candida albicans or Saccharomyces cerevisiae and a multi-species type. In this study, we proposed the best possible processing strategies for the three sequencing platforms, of which, the PacBio platform allowed for the most accurate estimation of the fungal community. The database and methodology described here provide critical tools for the emerging field of mycobiome studies.

Keywords: mycobiota, fungi, internal transcribed spacer (ITS), high-throughtput sequencing (HTS), the gut mycobiota in Japanese

## INTRODUCTION

fmicb-08-00238 February 13, 2017 Time: 11:53 # 2

Comprehensive analysis of bacterial communities (microbiota) has been made possible with the advent of high-throughput sequencing (HTS) technologies. Particularly, this has led to understanding the composition of the gut microbiome as influenced by factors, such as genetic background, diet, and immune function. Rapid progress has been made in discovery of associations between gut microbiota, the host's immune system, and various diseases (Arumugam et al., 2011; Wu et al., 2011; Gevers et al., 2012; Ursell et al., 2012).

In addition to bacteria, various types of fungi are also present in the intestines (Schulze and Sonnenborn, 2009; Chen et al., 2011; Dollive et al., 2012). Although the absolute number of commensal fungi accounts for as little as approximately 0.1% of total microorganisms in the intestines (Underhill and Iliev, 2014), they may have an important role in human health (Chang et al., 2011; Gaitanis et al., 2012; Cui et al., 2013; Huffnagle and Noverr, 2013; Wang et al., 2014). For instance, Candida albicans is a species of fungi persistently present in model mice that causes the exacerbation of allergies and autoimmune diseases (Sonoyama et al., 2011). In addition, fungi in the intestines induce colitis through the activity of host Dectin-1, a C-type lectin receptor (Sonoyama et al., 2011; Iliev et al., 2012). Fungal communities (mycobiota) in the intestines clearly play an important role in the health of the host.

Various genera of fungi have been detected in mycobiota analyses of the healthy human gut, and the genera Saccharomyces, Candida, and Cladosporium have been detected in particularly high percentages (Schulze and Sonnenborn, 2009; Chen et al., 2011; Dollive et al., 2012; Hoffmann et al., 2013). Moreover, analyses of the mycobiota in the oral cavity (Kleinegger et al., 1996; Ghannoum et al., 2010) and skin (Paulino et al., 2006, 2008; Zhang et al., 2011; Park et al., 2012; Findley et al., 2013) have demonstrated that mycobiota populations vary according to sites on the body (Cui et al., 2013; Underhill and Iliev, 2014). Furthermore, there have recently been a number of discussions on the relationships between both fungi and host and among bacteria, fungi, and host (Romani and Luigina, 2011; Mason et al., 2012a,b; Huffnagle and Noverr, 2013; Underhill and Iliev, 2014; Wang et al., 2014; Romani et al., 2015). However, unlike studies analyzing the bacterial component of the host microbiome, fewer studies are available on mycobiomes. In addition, databases of fungal sequences and methodologies for mycobiota analysis are not yet fully developed. Therefore, the potential impact of the mycobiome on pathogenesis and the development of the host immune system remains unknown (Romani and Luigina, 2011; Cui et al., 2013; Huffnagle and Noverr, 2013; Underhill and Iliev, 2014; Wang et al., 2014).

Characterizations of mycobiomes mainly rely on analysis of one portion of the internal transcribed spacer (ITS) region, ITS1, which is located between the 18S and 5.8S subunits of the fungal rRNA genes (Huffnagle and Noverr, 2013; Underhill and Iliev, 2014). This region is universally present in fungi and the mutation rate has been used to analyze evolutionary relationships. For these reasons, it is widely used for identification of species and genetic analyses. However, while the 16S rRNA gene used for bacterial microbiota analysis has a virtually identical length regardless of the species, the length of the ITS region in fungi varies between species. MiSeq and IonPGM sequencing platforms commonly used in mycobiota analysis are able to produce read lengths of several hundred base pairs (bp), however, shorter DNA sequences are favored on these platforms, and thus, overrepresented in the sequencing results. For example, the ITS1 sequence of the well-known fungi Saccharomyces cerevisiae has a length of approximately 460 bp, however, IonPGM do not completely read the ITS1 region with sufficient margins. As a result, it is impossible to have an accurate census of the composition of some fungi in the mycobiota using this strategy. A previous study aimed at solving this issue made attempts to evaluate the accuracy of the analysis by using the ITS1 sequence together with additional gene regions (Tonge et al., 2014). In addition to the problem of short reads, there is also a challenge in establishing a consensus method for mycobiome assessment. There are studies approached with the analytical strategy using automated analysis platforms such as QIIME (Caporaso et al., 2010) and PIPITS (Gweon et al., 2015), and using some curated databases of fungal ITS sequences (Findley et al., 2013; Kõljalg et al., 2013; Schoch et al., 2014; Tang et al., 2015). However, due to the complexity in evaluating the combination of both experimental and analytical strategies, not many attempts have been made to establish a consensus method for mycobiome assessment.

In this study, combinations of experimental and analytical strategies using three sequencing technologies, Illumina MiSeq (MiSeq), Life Technologies Ion Torrent PGM (IonPGM), and Pacific Biosciences RS II system (PacBio), were evaluated using a newly constructed database, ntF-ITS1, and a mock community containing mixtures of 26 fungal species representing 15 genera. Because reports of the analysis of human mycobiomes have been relatively rare and the gut mycobiota is particularly poorly understood (Cui et al., 2013; Huffnagle and Noverr, 2013; Tonge et al., 2014; Wang et al., 2014), we analyzed the mycobiota of feces from healthy Japanese subjects. In this paper, we propose the best strategies to assess the fungal communities and mycobiota of healthy Japanese individuals that is a potential consensus method for mycobiome assessment that may overcome the problem of short reads.

### MATERIALS AND METHODS

#### Database Construction

Sequences classified as fungi (Taxonomy ID:4751) were extracted from the NCBI-nt database. Additional sequences with ambiguous taxonomic information (those whose taxonomic name included the terms "uncultured," "environmental," "unclassified," or "mixed") were excluded. From these sequences, the ITS1 region was identified between the ITS1-F (5<sup>0</sup> -CTTGGTCATTTAGAGGAAGTAA-3<sup>0</sup> ) and ITS2 (5<sup>0</sup> -GCATCGATGAAGAACGCAGC-3<sup>0</sup> ) sequences using cutadapt-1.7.1 with default parameters (Martin and Marcel, 2011). Because no ITS1 sequences were identified for Rhodosporidium babjevae, Acremonium alternatum, Penicillium

digitatum, Mucor ramosissimus, or Rhizopus oryzae in the curated NCBI-nt database, we manually added these sequences to our database. The resulting sequences were converted into a format that was suitable for the Ribosomal Database Project (RDP) Classifier (Wang et al., 2007), and this was used as the database nt-Fungi-ITS1 (ntF-ITS1) for mycobiota analysis. Sequence extraction and format conversion were conducted with a Ruby script. The ntF-ITS1 database can be downloaded from our website<sup>1</sup> .

### Culture and DNA Extraction from 26 Known Fungal Species

The 26 examined strains listed in **Table 1** were cultivated on potato-dextrose agar (PDA) slants at 25◦C for 3 – 7 days. DNA was extracted from each of cultivated mycelium using a modified benzyl bromide extraction method reported previously in Zhu et al. (1993). Briefly, 500 µL of extraction buffer (100 mM Tris-HCl, 40 mM EDTA, pH 9.0), 100 µL of 10% SDS, and 300 µL of benzyl chloride were added to approximately 200 mg of mycelium. This suspension was then mixed by vortexing and incubated at 65◦C for 25 min. The mixture was kept on ice for 5 min and centrifuged at 6,000 × g for 10 min at 4 ◦C. DNA was collected from the supernatant by isopropanol precipitation.

#### Amplification of ITS1 Regions and Sanger Sequencing of the Fungal Mock Community

PCR was performed using a primer set (ITS1-F: 5<sup>0</sup> -CTTGGTC ATTTAGAGGAAGTAA-3<sup>0</sup> ) and (ITS2: 5<sup>0</sup> -GCATCGATGAA GAACGCAGC-3<sup>0</sup> ) targeted the ITS1 region of the ITS. To amplify the targeted region, 5 ng of extracted DNA from 26 known individual fungal species served as the template in 50 µL reactions using KAPA HiFi HotStart Ready Mix (KAPA Biosystems, Woburn, MA, USA). DNA was amplified with an initial denaturation step at 95◦C for 3 min, followed by 15 cycles of denaturation at 98◦C for 20 s, annealing at 56◦C for 15 s, and elongation at 72◦C for 30 s. Products were purified using DNA clean and Concentrator-5 (Zymo Research, Orange, CA, USA). Sequencing reactions were performed using 3130xI Genetic Analyzer (Applied Biosystems, Foster City, CA, USA).

#### Library Preparation, Sequencing, and Read Trimming for the Fungal Mock Community

The concentrations of the ITS1 amplicons from all 26 fungal species were measured using a Qubit Fluorometer (Invitrogen, Carlsbad, CA, USA). Amplicons were mixed to yield equal amounts of DNA (50 ng each), and the mixture was used as a fungal mock community for the preparation of libraries for each sequencing system. Since different fungal species has different ITS1 length, percent composition of each fungi in the mock community based on their ITS1 length is calculated as shown in the **Table 1**. The raw sequencing data have been deposited in the DDBJ Sequence Read Archive (DRA) under the accession code DRA004340.

#### IonPGM

A library was prepared from 5 ng of mixed amplicon using the Ion Fragment Library kit (Life Technologies, Gaithersburg, MD, USA) according to the manufacturer's instructions. Sequencing was performed using a 318 chip and Ion PGM Sequencing Hi-Q Kit (Life Technologies). Raw sequences were trimmed using BBtrim software<sup>2</sup> with mean quality value of 10–30. Sequences without full length ITS1 (i.e., the sequences amplified with primers by PCR) were removed using the FASTX-Toolkit<sup>3</sup> for subsequent analysis.

#### MiSeq

A library was prepared from 10 ng of mixed amplicon using KAPA Library Preparation kits (KAPA Biosystems) according to the manufacturer's instructions. Paired-end sequencing of 251 bp was performed using a MiSeq v2 500 cycle kit (Illumina, San Diego, CA, USA). Paired-end sequences were merged using PEAR software<sup>4</sup> . The merged reads were then quality filtered with the same condition as those of the IonPGM data.

### PacBio

A library was prepared from 300 ng of mixed amplicon using the DNA Template Prep kit 2.0 (Pacific Biosciences, Menlo Park, CA, USA) according to the manufacturer's instructions. Sequencing was performed with the PacBio RS II system using the DNA Sequencing Kit C2 (Pacific Biosciences) with P4 polymerase. CCS constructed from more than three full-pass subreads were produced using PacBio SMRT Analysis.

### Fungal ITS1 Deep Sequencing of Healthy Japanese Feces

Feces were collected in tubes containing RNAlater (Ambion, Austin, TX, USA). Samples were weighed, and RNAlater added to make 10-fold dilutions of homogenates. Homogenates (200 mg) of feces were washed twice with 1 mL PBS and fecal DNA was extracted with the PowerSoil DNA isolation kit (MO BIO Laboratories, Solana Beach, CA, USA) according to the manufacturer's protocol. Samples were stored at −20◦C. DNA was amplified with PCR using the following protocol: Initial denaturation at 95◦C for 2 min, followed by 40 cycles of denaturation at 95◦C for 20 s, annealing at 56◦C for 30 s, elongation at 72◦C for 30 min, followed by a final elongation step at 72◦C for 10 min. Barcoded PacBio libraries were prepared using the DNA Template Prep kit 2.0 (Pacific Biosciences, Menlo Park, CA, USA) according to the manufacturer's instructions. Three libraries per SMRT Cell were pooled and subjected to Single Molecule Real-Time (SMRT) sequencing using the PacBio RS II system (Pacific Biosciences). All fecal samples were collected and analyzed at Osaka University. The Osaka University ethics

<sup>1</sup>https://bitbucket.org/daisukem/fungi/wiki/Home

<sup>2</sup>http://sourceforge.net/projects/bbmap/

<sup>3</sup>http://hannonlab.cshl.edu/fastx\_toolkit/

<sup>4</sup>http://sco.h-its.org/exelixis/web/software/pear/

#### TABLE 1 | Genera, species, ITS1 length, and the percentage of fungi used in the mock community.


a IFM is the abbreviation for Institute of Food Microbiology. <sup>b</sup>Each percentage is the theoretical mock community abundance of each fungal species. The amplicons were mixed to yield equal amount of DNA (each was 50 ng) of each sample.

committee approved this study and written informed consent was obtained from all study subjects (12237-3).

was performed using the "stats" package from CRAN. A heat map visualization of OTUs was generated with the heatmap.2 function of the "gplots" package from CRAN.

#### Bioinformatic Analysis and Taxonomic Assignment

Sequences were clustered into operational taxonomic units (OTUs), defined at 90 – 100% similarity cutoff for the mock community and 95% for fecal samples using UCLUST version 1.2.22q (Edgar, 2010) using the script (pick\_otus.py) in QIIME 1.9.1 (pick\_otus.py). Representative sequences for each OTU were classified taxonomically using RDP Classifier version 2.2 using the script (assign\_taxonomy.py) in QIIME 1.9.1 with our ntF-ITS1 database and the minimum confidence value is 0.55, and using blastn 2.3.30+ using the script (assign\_taxonomy.py) in QIIME 1.9.1 with the default parameters. For the analysis of mock community, we also used the databases; UNITE (Kõljalg et al., 2013), Findley (Findley et al., 2013), and THF (Tang et al., 2015). The characterization of these ITS reference database were summarized in Supplementary Table S1. For principal component analysis (PCA), we visualized data sets using the statistical programming language R 3.1.3 (R Core Team, 2016). Hierarchical clusters of mycobiota were calculated from the relative abundances and the ranked abundances of genera. PCA

## RESULTS

#### Construction of a Fungal ITS1 Database

For the database construction to identify fungal species, at first, 2,444,619 records of fungi-derived sequences were retrieved from the NCBI-nt database. Of those records, 1,849,386 sequences remained after excluding the records that included ambiguous taxonomic information (e.g., "environmental samples"). Finally, from these, 13,943 sequences remained by excluding the data records with no ITS1 region information, and the database has been constructed. The constructed database, termed ntF-ITS1, represented 1,218 genera and 6,525 species. The average length of ntF-ITS1 sequences was 257 bp, and was widely distributed from approximately 100–800 bp (**Figure 1**). The length of ITS1 for each sequence in THF and UNITE showed the similar distribution to ntF-ITS1 as shown in Supplementary Figure S1. The number of fungal genera and species included in each database were summarized in Supplementary Table S1.

### Taxonomic Assignment of a Mock Community

sequences of each length.

A mock community composed of 26 fungal species representing 15 genera (**Table 1**) was used to evaluate the various methods of taxonomic assessment. To characterize this community, we sequenced the ITS1 regions of 26 species. The average length of ITS1 in the mock community was 298 bp; the shortest sequence measured was 228 bp, and the longest 520 bp. Thus, the distribution of ITS1 sequence lengths in the mock community were of similar diversity to ntF-ITS1. We next examined the percent identity that allowed for accurate OTU clustering of the mock community. Using a similarity threshold of 98% or higher for clustering, the 26 fungal species were classified as 25 individual OTUs. Trichoderma viride and Trichoderma koningii were clustered in the same OTU due to the fact that their sequences had 100% similarity. Using a threshold of 97% similarity, Penicillium digitatum and Penicillium chrysogenum were classified in the same OTU. Using a threshold of 96% similarity, Cladosporium herbarum and Cladosporium cladosporioides were classified in the same OTU. At a 95% similarity threshold, C. albicans and C. dubliniensis were classified in the same OTU. If the similarity threshold were reduced to 94% and below, fungal species from differing genera (Penicillium chrysogenum and Aspergillus fumigatus) were clustered in the same OTU. These clustering results at each similarity level are summarized in Supplementary Table S2. We also classified each OTU using blastn and UCLUST. As shown in Supplementary Table S3, blastn and UCLUST mis-assigned some OTU's to other genera. Our evaluation suggested that to conduct a clustering of fungal genera on the basis of the ITS1 region in an accurate manner, the sequence similarity required for clustering must be 95% or higher. We employed the 95% sequence similarity threshold for taxonomic assignment with our ntF-ITS1 using RDP Classifier, resulting in the accurate assignments of all 15 fungal genera in our mock community (Supplementary Table S4). We further compared our ntF-ITS1 to existing fungal ITS database, UNITE, Findley, and THF. As shown in the Supplementary Table S5, while some sequences were mis-assigned to wrong genera using UNITE and Findley, all sequences were assigned to correct genera using ntF-ITS1 and THF. While THF has sequences for only 1,816 species, ntF-ITS1 for 6,525 species. Therefore we used ntF-ITS1 for analysis of fungal ITS1 deep sequence.

### Mock Community Analysis by IonPGM Sequencing

The Ion PGM Hi-Q sequencing of the mock community yielded 219,756 reads. We evaluated the composition of the community with trimming conditions at varying quality levels (**Figure 2A**). The estimated compositions of the mock community varied widely at different quality scores. It is noteworthy that, when high quality trimming with a mean quality value (MQV) of 30 was employed, the genus Filobasidiella (dark green) accounted for almost 90% of the total community. The compositions for the genera Acremonium (blue), Candida (green), Fusarium (dark gray), and Cryptococcus (light blue) were more highly represented in the middle quality ranges, around MQV20, and decreased toward MQV30. The compositions of the genera Saccharomyces and Nakaseomyces were barely detected in this analysis, as its ITS1 regions consist of sequences that are longer than others (**Figure 2B**). Plotting MQV at each sequence position clearly revealed that the quality scores rapidly dropped to MQV20 at 50 bp, and decreased to MQV15 at 275 bp, the average length of all ITS1 regions (**Figure 2C**). When we performed hierarchical clustering analysis of the compositions at different quality scores (**Figure 2D**), the population compositions at MQV11 clustered with the original mock population. However, the original composition could not be reconstructed accurately when the MQV was set over 21. Mycobiota analysis using IonPGM depended largely on the length of the ITS1 regions. This shows that the quality of the reads must be carefully considered when using IonPGM. In our case, the quality trimming score of MQV 11 yielded the most accurate results.

## Mock Community Analysis by MiSeq Sequencing

MiSeq paired-end sequencing of 251 bp yielded 181,436 reads. These paired reads were merged, trimmed, and subjected to taxonomic assignment with various trimming conditions. Unlike the case of IonPGM, the estimated compositions at each MQV score were not influenced by the trimming quality (**Figure 3A**). However, the genus Nakaseomyces was not identified at any quality score. The composition of the genus Saccharomyces was estimated to be as low as approximately half of the original population. As stated above, the ITS1 regions of these specific genera have long sequences, around 500 bp. In particular, the ITS1 region of Candida glabrata, which was used as a representative of the genus Nakaseomyces, had an ITS1 length of 520 bp. This long ITS1 region could not be sequenced by the

251 bp paired-end sequencing method. Similar to the IonPGM data, the MQV plot of MiSeq also tended to decrease according to the length of the reads (**Figure 3B**). The overall MQVs of the read2 region are lower than those of the read1 region. However, because the overall MQVs were higher than 30, quality trimming had no impact. The hierarchical clustering analysis revealed that there are no large differences between MQVs. The cluster most closely resembling the mock population was MQV26-30 (**Figure 3C**). The analysis with MiSeq sequencing reconstructed nearly the entire population of the mock community, except the genus Nakaseomyces which was not identified.

### Mock Community Analysis by PacBio Sequencing

PacBio sequencing using one cell yielded 2,189,947 subreads. The MQV of these subreads was 9.5, which was extremely low. To increase sequence accuracy, circular consensus sequencing (CCS) was employed. As the consensus sequence was generated, the number of full passes used in the analysis ranged from 2 to 12. At two full-passes, 72,406 reads were obtained. At 12 full-passes, 40,934 reads were obtained. We evaluated the composition change according to the numbers of full-passes (**Figure 4A**). The PacBio sequencing reconstructed the composition of the original mock community, including the genera Nakaseomyces and Saccharomyces. Among the data associated with a small number of full-passes, sequences that could not be identified, such as those labeled as "fungi" or "other," were confirmed to account for less than 10% of sequences, and were not present in data sets with a higher number of full-passes. The MQV plot of PacBio showed that MQVs of reads longer than 300 bp increased according to the number of full-passes (**Figure 4B**). At the 520 bp position, consensus sequences with four and

eight full-passes achieved MQV30 and MQV40, respectively. The hierarchical clustering analysis revealed that there are no large differences related to the number of full-passes (**Figure 4C**). Therefore, the consensus sequence with eight full-passes, which could identify 98% or more of all 15 fungal genera, was considered to have been the optimal analytical parameters for PacBio sequencing, and those parameters were used for subsequent analysis of the gut mycobiota of healthy Japanese individuals.

### Gut Mycobiota in Healthy Japanese Individuals

Mycobiota analysis was performed on fecal samples from 14 healthy subjects. PacBio consensus sequences with eight full-passes yielded an average of 2,984 reads per sample. We performed the PCA on mycobiota at the genus level for 14 individuals (**Figure 5A**), and characterized two major mycobiota types, Genera Candida-dominant and Saccharomyces dominant type. A third group composed of other fungi was also identified. Relationships between mycobiota and gender or age differences were analyzed, but no apparent differences were found (Supplementary Figure S2). The heat map of fungal OTUs revealed a low diversity among the major fungi present in each individual (**Figure 5B**). The mycobiota of seven individuals was composed of either only a single species or was >90% of C. albicans or S. cerevisiae (**Figure 5C**). For the remaining seven individuals, the mycobiota were composed of multiple species, including C. glabrata, C. dubliniensis, Ganoderma lingzhi, Aspergillus oryzae, and unidentified fungal sequences. All fungi found in feces are listed in Supplementary Table S6. We determined that the mycobiota of healthy Japanese individuals is comprised of a simple assemblage consisting of only one or a few species.

## DISCUSSION

#### Comparison of IonPGM, MiSeq, and PacBio ITS1 Sequencing for Characterization of a Fungal Community

We evaluated the performance of three sequencing technologies, IonPGM, MiSeq, and PacBio, for an ITS1 deep sequencing analysis of a 26-species fungal community. **Figure 6A**

summarizes the results obtained from surveys of each sequencing technology. For IonPGM sequencing, relative abundances of genera Nakaseomyces and Saccharomyces were evaluated to be less than 10% of the original mock while the genus Filobasidiella was estimated to be twice as much. For MiSeq sequencing, the genus Nakaseomyces was not identified, and the genus Saccharomyces was estimated at less than half of the original mock population. In addition, the abundances of genera Aspergillus, Cryptococcus, and Rhizopus deviated by approximately 30% from the original mock. For PacBio sequencing, the relative abundances of genera Nakaseomyces and Saccharomyces were underestimated, with results suggesting less than half of the concentration of the original mock population, while the relative abundances of the other genera were estimated accurately. While second-generation sequencing, such as IonPGM and MiSeq, tends to read shorter DNA sequences, PacBio has less of a length-dependent sequencing bias, however, the length dependence cannot be eliminated completely (Fichot and Norman, 2013).

We generated a heat map of the relative abundances for each sequencing technology according to the length of the ITS1 region of each fungal genus (**Figure 6B**). For all sequencing technologies, the longer ITS1 sequences, shown in red, were underestimated from the original mock sample, and the shorter ITS1, which are shown in green, were overestimated. The ITS1 regions of C. glabrata and S. cerevisiae are approximately 500 bp. Therefore, it is impossible to sequence them in their entirety using the current IonPGM sequencing kit, which allows for reading DNA sequences up to a maximum length of 400 bp. C. glabrata is a fungal species commonly present in humans (Underhill and Iliev, 2014) and S. cerevisiae is a very well-known species included in common food products. Therefore, these two species of fungi are highly likely to be present in the human mycobiome.

Despite the importance of the two major species, C. glabrata was not found with MiSeq or IonPGM sequencing. Pyrosequencing technology such as IonPGM sequencing and 454 sequencing (Roche) are currently the most widely used methods for the analysis of mycobiota. However, pyrosequencing results contain many homopolymer errors, and all services pertaining to 454 sequencing will be unavailable after 2016. In addition, IonPGM sequencing is fundamentally dependent on the sequence length. Even though the Hi-Q sequencing chemistry, which reduces insert-deletion error compared with previous version, was applied in this study, sequence quality decay was observed (Supplementary Figure S3). Compared with

Table S4. (C) Bar charts of the mycobiota found in the feces of each of the Japanese subjects. The upper graph represents the "single species type" the bottom one represents the "multi-species type."

IonPGM, MiSeq was found to be more suitable for mycobiota analysis. These results were consistent with recently reported findings that compared MiSeq with IonPGM analysis (Tang et al., 2015).

Our study using PacBio revealed that all fungi from our mock community could be accurately assigned. The hierarchical clustering analysis of results from each sequencing analysis revealed that PacBio provided the most accurate estimation of the

mock community population (**Figure 6C**). This was particularly true for some species of fungi with ITS1 regions as long as 800 bp. In the further studies aiming to assign all OTUs at the species level, expanding the fungal sequence database and conducting an analysis of the entire length of the ITS region by selecting different regions will be required. With our approach, the PacBio method was the only one that allowed for analysis of long sequence lengths with sufficient margins, covering the entire length of the ITS region. However, it must be noted that some fungi cannot be differentiated by the ITS region alone, which is a limitation of this analysis (Mello et al., 2011; Schoch et al., 2012; Tonge et al., 2014).

#### Mycobiota Analysis of Healthy Japanese Individuals

Our study is the first report analyzing the mycobiota of the feces of Japanese individuals using deep ITS1 sequencing. We found that the mycobiota of most subjects were composed of only a few species, mainly consisting of genera Candida and Saccharomyces. These results were consistent with previous reports exploring the composition of fungi in human feces (Dollive et al., 2012; Hoffmann et al., 2013). In the genus Candida, we found the following species: C. albicans, C. glabrata, and C. dubliniensis, which have been reported as the most commonly found fungal species in humans (Underhill and Iliev, 2014; Romani et al., 2015). Our study revealed that while the two major genera in the mycobiota were Candida and Saccharomyces, some people were carriers of other populations of mycobiota. These mycobiomes were composed of identifiable fungi such as Tricosporon spp., G. lingzhi, Hypsizygus sublateritium, and A. oryzae, as well as a large number of sequences that could not be assigned to a known taxa.

Although the mycobiota of each individual participant did not show much diversity, the identification of those unknown minor fungi in the mycobiota would be the next targets for research on human commensal microorganisms. Further development of fungal genome databases will be essential for the analysis of various mycobiota and diseases associated with specific commensal microorganisms.

#### REFERENCES


### AUTHOR CONTRIBUTIONS

DM and SN designed the study, performed experiments and data analysis, interpreted the analyzed results, and coauthored the manuscript. KF and RT performed experiments and coauthored the manuscript. KG, YM, YF, TK, NG, TeY, MN, AK, and TH contributed valuable advice on the analyzed results. TaY, TI, and KT designed the study, coordinated research and helped to author the manuscript. All authors have read and approved the final manuscript.

### FUNDING

This work was supported in part by the program of the Japan Initiative for Global Research Network on Infectious Diseases (J-GRID) by the Ministry of Education, Culture, Sports, Science and Technology of Japan. DM is supported in part by JSPS KAKENHI Grant Number 24890103. SN is supported in part by JSPS KAKENHI Grant Number 25670271.

### ACKNOWLEDGMENTS

We thank S. Yoshiki and T. Yoshimura for help with the experiments. We would like to thank S. Ito, N. Jung, and J. Rozewicki. This work was supported in part by the National BioResource Project – Pathogenic Microbes in Japan (http://www.nbrp.jp/).

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.00238/full#supplementary-material



and molecular data for Fungi. Database 2014:bau061. doi: 10.1093/database/ bau061


**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 © 2017 Motooka, Fujimoto, Tanaka, Yaguchi, Gotoh, Maeda, Furuta, Kurakawa, Goto, Yasunaga, Narazaki, Kumanogoh, Horii, Iida, Takeda and Nakamura. 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) or licensor 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.

# Disruptions of the intestinal microbiome in necrotizing enterocolitis, short bowel syndrome, and Hirschsprung's associated enterocolitis

*Holger Till1\*, Christoph Castellani1, Christine Moissl-Eichinger2, Gregor Gorkiewicz3 and Georg Singer1*

*<sup>1</sup> Department of Paediatric and Adolescent Surgery, Medical University of Graz, Graz, Austria, <sup>2</sup> Department of Internal Medicine, Medical University of Graz, Graz, Austria, <sup>3</sup> Institute for Pathology, Medical University of Graz, Graz, Austria*

#### *Edited by:*

*Zhongtang Yu, The Ohio State University, USA*

### *Reviewed by:*

*M. Andrea Azcarate-Peril, University of North Carolina at Chapel Hill, USA Miguel Gueimonde, Consejo Superior de Investigaciones Científicas, Spain*

> *\*Correspondence: Holger Till holger.till@medunigraz.at*

#### *Specialty section:*

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

*Received: 16 August 2015 Accepted: 05 October 2015 Published: 16 October 2015*

#### *Citation:*

*Till H, Castellani C, Moissl-Eichinger C, Gorkiewicz G and Singer G (2015) Disruptions of the intestinal microbiome in necrotizing enterocolitis, short bowel syndrome, and Hirschsprung's associated enterocolitis. Front. Microbiol. 6:1154. doi: 10.3389/fmicb.2015.01154* Next generation sequencing techniques are currently revealing novel insight into the microbiome of the human gut. This new area of research seems especially relevant for neonatal diseases, because the development of the intestinal microbiome already starts in the perinatal period and preterm infants with a still immature gut associated immune system may be harmed by a dysproportional microbial colonization. For most gastrointestinal diseases requiring pediatric surgery there is very limited information about the role of the intestinal microbiome. This review aims to summarize the current knowledge and outline future perspectives for important pathologies like necrotizing enterocolitis (NEC) of the newborn, short bowel syndrome (SBS), and Hirschsprung's disease associated enterocolitis (HAEC). Only studies applying next generation sequencing techniques to analyze the diversity of the intestinal microbiome were included. In NEC patients intestinal dysbiosis could already be detected prior to any clinical evidence of the disease resulting in a reduction of the bacterial diversity. In SBS patients the diversity seems to be reduced compared to controls. In children with Hirschsprung's disease the intestinal microbiome differs between those with and without episodes of enterocolitis. One common finding for all three diseases seems to be an overabundance of Proteobacteria. However, most human studies are based on fecal samples and experimental data question whether fecal samples actually represent the microbiome at the site of the diseased bowel and whether the luminal (transient) microbiome compares to the mucosal (resident) microbiome. In conclusion current studies already allow a preliminary understanding of the potential role of the intestinal microbiome in pediatric surgical diseases. Future investigations could clarify the interface between the intestinal epithelium, its immunological competence and mucosal microbiome. Advances in this field may have an impact on the understanding and non-operative treatment of such diseases in infancy.

Keywords: microbiome, necrotizing enterocolitis, short bowel syndrome, Hirschsprung's disease, pediatric surgery

## INTRODUCTION

The development of the human microbiome seems to start in the prenatal period as fetuses may have contact to microorganisms already *in utero* (Jimenez et al., 2008; Donnet-Hughes et al., 2010; Ardissone et al., 2014). However, the major microbial colonization of the human body starts at birth: During delivery, by passing the mother's birth canal, the babies acquire a possibly optimized set of microorganisms. Mother's milk, in addition, provides maternal microbes and prebiotics, which help the gut microbiome to settle and stabilize (Bode, 2009; Hunt et al., 2011). Within the first years of life, the intestinal microbiome of infants undergoes abrupt changes until it reaches a similar status like in adults at approximately three years of age (Palmer et al., 2007; Yatsunenko et al., 2012). The gut microbiome plays an important role during the physiological development: It is extremely important for the differentiation of gut epithelium, interacts closely with the gut-associated lymphoid tissue and influences activity and morphology of the gastrointestinal tract (Cebra, 1999; Cho and Blaser, 2012; Guaraldi and Salvatori, 2012; Hooper et al., 2012; Putignani et al., 2014). Thus, the human development until the age of three is characterized by a highly sensitive interplay between microorganisms and the body. Any disturbances like prematurity, formula feeding, treatment with antibiotics or mode of delivery (Dominguez-Bello et al., 2011; Cho et al., 2012) may have a direct impact on microbial abundance and diversity.

The understanding of the microbiome development seems especially relevant for preterm babies with a gestational age below the 30th week or a very low birth weight (VLBW, *<*1500 g) (Groer et al., 2014). At this age the physiological interaction between microbes and structural, metabolic, or immunological functions of the gut remains poorly understood. Moreover, there is still a very limited knowledge about the influence of the microbiome on the development of gastrointestinal diseases in early infancy (Groer et al., 2014). Thus, the focus of this review is to provide an in-depth summary of the current knowledge about the gastrointestinal microbiome in infancy with special respect to relevant pediatric surgical diseases such as necrotizing enterocolitis (NEC) of the newborn, short bowel syndrome (SBS), and Hirschsprung's disease associated enterocolitis (HAEC).

#### Necrotizing Enterocolitis

Necrotizing enterocolitis represents a devastating disease primarily affecting premature infants weighing less than 1,500 g. It has been shown that 7% of these infants develop NEC, which carries a mortality rate of up to 30% (Neu and Walker, 2011). Many surviving children subsequently suffer from serious long-term morbidity including intestinal adhesions, bowel resections and even SBS (Pike et al., 2012) making this disease highly relevant for the pediatric surgeon. Despite intense research performed in the field, the exact pathogenesis of NEC has not been revealed yet. Risk factors include gestational age, birth weight, and formula feeding (Stewart and Cummings, 2015). Since NEC does not occur in germ free mice, bacterial colonization seems to play a major role in the development of this disease (Afrazi et al., 2011). In neonates, the intestinal microbiome plays a pivotal role in the development of the epithelial barrier function, integrity and the local and system immune function. Disturbances of the cross-talk between the intestinal microbial community and the immune system may initiate an exaggerated inflammatory response ultimately resulting in NEC (Berrington et al., 2013). The association between bacterial colonization and NEC has been recognized already some decades ago (Neu, 2013). Since then, numerous different bacteria and also viruses have been related to the development of NEC. Until recently, examinations of the intestinal bacterial colonization have been restricted to culture dependent methodologies.

The advent of culture independent technologies, however, has driven research and further deepened our understanding of NEC. There is an increasing number of publications applying molecular sequencing methods comparing intestinal microbial profiles of infants with and without NEC. Many of the available reports have demonstrated disturbances of the intestinal microbiome in infants with NEC. However, the specific findings differ significantly among those studies. Mai et al. (2011) have used high throughput 16S rRNA gene sequencing of stool samples to compare the diversity of microbiota and the prevalence of specific bacterial sequences in nine infants with NEC and in nine matched controls. Interestingly, microbiota composition differed in the matched samples collected one week but not *<*72 h prior to NEC diagnosis. An increase of Proteobacteria and a decrease in Firmicutes in NEC cases collected one week and *<*72 h prior to NEC diagnosis were found. Another study with a higher number of patients (11 infants with NEC, 21 matched controls) has demonstrated a tendency toward a lower alpha-proteobacterial diversity in infants, who later developed NEC (Morrow et al., 2013). Furthermore, NEC preceded by Firmicutes dysbiosis occurred earlier (onset days 7–21) than NEC preceded by Proteobacteria dysbiosis (onset days 19–39). The lower bacterial diversity of NEC cases versus controls was confirmed recently (McMurtry et al., 2015). Microbial diversity and *Clostridia* abundance and prevalence even decreased with increasing severity of NEC. One recently published study has investigated genes regulating structural proteins such as tight junctions and cell adhesion in a neonatal rat model of NEC applying a transcriptomic approach (Hogberg et al., 2013). Several tight junction genes such as claudins 1, 8, 14, and 15 and gap junction proteins were found to be involved in the pathogenesis of NEC.

The disruptions of the intestinal microbial profile prior to the development of NEC might open the doors for an early detection and a focused intervention of infants at risk for developing NEC. However, the currently available results are contradictory and inconclusive and support the need of future studies.

Moreover, the microbial composition of fecal samples does not necessarily reflect the situation on the mucosa (Haange et al., 2012). Thus, there is a further need to assess both the physiological development of the microbiome in the different parts of the intestinal tract and potential disruptions in preterm infants developing NEC. A slowly growing body of evidence suggests that in NEC cases there is also a shift of microbes within

the intestinal mucosa (Carlisle and Morowitz, 2013). However, no studies have been performed applying deep sequencing techniques on operative specimen of children suffering from NEC compared to control groups. Therefore, further research has to be performed to unravel the influence of the microbiome on the distinct pathogenesis of NEC. Another focus would be to deepen the understanding of microbiome changes prior to NEC development. Such knowledge could foster potential therapeutic strategies including treatment with pre- or probiotics and stool transplantation in order to optimize treatment of infants affected by this devastating disease.

#### Short Bowel Syndrome

Short bowel syndrome represents the most common cause for intestinal failure in children. SBS occurs as a congenital disorder or results from surgical removal of diseased gut segments affected by NEC, abdominal wall defects (gastroschisis, omphalocele), midgut volvulus, intestinal atresia, Hirschsprung's disease, or abnormalities of the superior mesenteric artery (Reddy et al., 2013). The disease-associated loss of absorption capacity of the intestine leads to an inability to maintain fluid, electrolyte, nutrients, or micronutrient balances resulting in a frequent dependency on parenteral nutrition. In a recently published report from a children's hospital in Canada the incidence of SBS was found to be 22 per 1,000 admissions to the neonatal intensive care unit, which increased further to 43 per 1,000 admissions in premature infants (Wales et al., 2005). The health burden of SBS is significant. A case fatality rate of 27.5–37.5% has been reported within 1.5–5 year follow-up periods in several retrospective studies (Reddy et al., 2013). Failure to thrive (body weight *<*5th percentile) is seen in more than 70% of patients at 6 months and in almost half of the patients at 2.5 years (Sukhotnik et al., 2002).

The gastrointestinal microbiome of patients suffering from SBS has only been addressed in a very limited number of clinical studies so far (**Table 1**). Using stool and colonic biopsy samples of adult patients with SBS, Joly et al. (2010) have demonstrated that the microbiome of SBS patients is altered (compared to controls) with overabundance of *Lactobacillus* along with a reduced diversity of *Clostridium leptum*, *Clostridium coccoides*, and Bacteroidetes comparing 11 adult patients with SBS to 8 control patients without intestinal pathologies. However, there was no statistically significant microbiome change comparing both groups. SBS-associated microbiota with its high prevalence of *Lactobacillus* is enriched in facultative anaerobic carbohydrate fermenting bacteria (Dethlefsen et al., 2006). Since a considerable amount of fermentable sugars can be consumed by lactobacilli its high prevalence can be interpreted as an expression of the adaptive response of the bowel to SBS (Joly et al., 2009).

Engstrand Lilja et al. (2015) analyzed the microbial profile of 11 children with SBS (5 of them not weaned from parenteral nutrition) and compared the results to seven healthy siblings. One of the major findings was that again diversity (Shannon index) was significantly reduced in children with SBS still on parenteral nutrition compared to weaned children. Additionally, there was a significant overabundance of *Enterobacteriaceae* (belonging to the phylum Proteobacteria) in four out of the five children with SBS on parenteral nutrition. The results demonstrate that intestinal dysbiosis is related to parenteral nutrition in children suffering from SBS. However, it is not clear, whether the observed changes were a cause or a consequence of the disease state. Moreover, four out of the five children with SBS still on parenteral nutrition were treated with antibiotics due to suspected episodes of small bowel bacterial overgrowth (SBBO) at the time of stool sampling. The antibiotic treatment may be a confounding factor since antibiotics have been shown to lower the colonization resistance against *Enterobacteriaceae* by increasing the inflammatory status of the intestinal mucosa (Spees et al., 2013).

Another study also performed an in-depth analysis of the intestinal microbiota in children with intestinal failure using culture-independent phylogenetic microarray analysis (Korpela et al., 2015). An overabundance of lactobacilli, Proteobacteria, and Actinobacteria was observed and the overall diversity and richness were reduced. Proteobacteria, a major phylum of Gram-negative bacteria, were associated with liver steatosis and fibrosis, prolonged parenteral nutrition, and liver and intestinal inflammation in SBS. The lipopolysaccharides (LPS) produced by this Gram-negative strain may explain these results. An experimental approach using TLR-4 knockout mice (LPS insensitive) following small bowel resection could represent an approach to strengthen this assumption.

Further insights into the microbial community changes in SBS may be generated experimentally. Lapthorne et al. (2013) have demonstrated in a piglet model of SBS (75% small bowel resection) a colonic dysbiosis both two and six weeks postresection. While the total colonic bacterial number (as assessed by absolute quantification using qPCR) showed no significant differences either two weeks or six weeks following small bowel resection, bacterial diversity in the colon was significantly decreased in the resection group at six weeks. Bacteroidetes were decreased and Fusobacteria increased in the resection groups compared to their controls. The majority of differences were observed at family-level within the Firmicutes phylum and a general Firmicutes overabundance leading to an increased Firmicutes/Bacteroidetes ratio at both time points, i.e., 2 and 6 weeks, following small bowel resection. In patients suffering from SBS this ratio and its association with a pro-inflammatory state have not been studied in detail. Moreover little is known about the characterization of the 'relative microbiota maturity index' and the 'microbiota-for-age *Z*-score', two indices that have been recently described by Subramanian et al. (2014). These two metrics are based on the age-discriminatory bacterial species and a comparison of the observed maturity of a child's fecal microbiota to healthy children of his/her chronologic age. Therefore, these additional parameters facilitate the classification of undernourished states in children and a possible monitoring tool for applied therapies. The authors have defined a healthy development of the gut microbiome by applying a machinelearning-based approach to 16S rRNA datasets generated from monthly fecal samples. These samples were obtained from a birth-cohort of children living in an urban slum of Dhaka, Bangladesh who exhibited consistently healthy growth. These


TABLE 1 | Number of patients, specimen taken, and main findings of human studies investigating the intestinal microbiome in short bowel syndrome (SBS).

*SBS, short bowel syndrome; PN, parenteral nutrition; IF, intestinal failure.*

age-discriminatory bacterial species were incorporated into a model computing the above mentioned indices that compare postnatal development of a child's fecal microbiota relative to healthy children of similar chronologic age. The model was applied to children with malnutrition of different severities. The obtained results indicated that severe acute malnourishment is associated with significant relative microbiota immaturity that is only partially ameliorated following two widely used nutritional interventions (Khichuri-Halwa and a peanut based ready-touse therapeutic food). The authors concluded that microbiota maturity indices provide a microbial measure of human postnatal development, a way of classifying malnourished states, and a parameter for judging therapeutic efficacy (Subramanian et al., 2014). Therefore, assessing these indices would be of interest in children suffering from SBS.

Data conflicting the findings of Lapthorne et al. (2013; colonic dysbiosis two and six weeks post-resection) were recently described by Sommovilla et al. (2015) using a murine model. The authors performed a 50% small bowel resection in C57BL6 mice and collected enteric contents from the small bowel, cecum and stool at 7 and 90 days postoperatively for subsequent 16S rRNA gene analysis. No significant changes in bacterial diversity scores of stool, cecal, and ileal samples were found when comparing SBS mice 7 and 90 days following resection to their corresponding sham group. Additionally, no significant community differences at the phylum level at any site of the sampled gastrointestinal tract in the short arm of the study was found. However, at 90 days following small bowel resection the ileal contents significantly differed driven by a decrease in Proteobacteria and Actinobacteria when compared to the respective sham group. Additionally, a comparison of preand postoperative (90 days) stool and ileal samples revealed increases of *Lactobacillus* genera. These changes most likely reflect an appropriately adapted community of organisms in response to bowel resection. The beneficial effects of *Lactobacillus* on the gastrointestinal tract include promotion of the innate immunity and its administration has been demonstrated to decrease bacterial translocation (Eizaguirre et al., 2011) and to promote intestinal adaptation (Tolga Muftuoglu et al., 2011). On the other hand, there are cases of complications associated with *Lactobacillus* treatment of children suffering from SBS. Reported complications include bacteremia, sepsis, and D-lactic acidosis (Kunz et al., 2004; De Groote et al., 2005; Munakata et al., 2010). Taken together, there is still insufficient evidence on the effects of probiotics in children with SBS. The safety and efficacy of probiotic supplementation in this high-risk cohort needs to be evaluated in larger trials. The authors of the abovementioned study explained the absence of differences in the overall bacterial diversity throughout the gastrointestinal tract by the fact that unlike other studies perioperative antibiotics were not used and antibiotics lead to a reduced microbial diversity (Dethlefsen and Relman, 2011). Additionally, a 50% small bowel resection might retain sufficient intestinal length in this murine model. In contrast to the findings of Sommovilla et al. (2015), a decreased bacterial diversity in the colon in a murine model of ileocecal resection (removal of 12 cm ileum, cecum, and proximal right colon) and accompanying antibiotic treatment has been described recently (Devine et al., 2013).

In patients suffering from SBS, parenteral nutrition, the frequent use of antibiotics due to recurrent infections and the accelerated intestinal transit time may all lead to substantial clinical problems (Kaneko et al., 1997). The altered microbiomegut homeostasis in SBS leads to a disrupted gastrointestinal barrier function and capacity of the microbiota to provide vitamins and their precursors contributing to the frequent complications seen in SBS patients such as sepsis, vitamin deficiencies and failure to wean from parenteral nutrition (Sommovilla et al., 2015). Moreover, SBBO, defined as an increase of the total number of bacteria per ml content, may develop in children with SBS. SBBO is caused by stasis, poor peristalsis, and intestinal dilatation. Clinically SBBO significantly increases the risk for recurrent blood stream infections and the systemic proinflammatory response decreases with increasing enteral feeding and weaning parenteral nutrition (Cole et al., 2010).

Taken together, microbiome research of SBS is still in its infancy. A marked dysbiosis of the gastrointestinal tract seems to be characteristic for patients suffering from SBS. The presently available data based on next-generation sequencing indicate that the disruptions of the gastrointestinal microbiome may be a reason for the pro-inflammatory state associated with SBS. Nevertheless, the precise alterations of the microbiome associated with SBS, i.e., changes of the transient/resident microbiome during intestinal adaptation processes, the physiological and pathophysiological roles of the altered microbes and safe possibilities of therapeutic manipulations have still to be unraveled.

#### Hirschsprung's Associated Enterocolitis

Hirschsprung's disease describes a congenital segmental absence of the enteral nervous system (ENS) in the myenteric and submucosal plexuses with variable proximal expression due to a failure of migration of neural crest cells during embryonic development (Sasselli et al., 2012). The resulting intestinal obstruction is usually treated by surgical removal of the aganglionic bowel and a pull-through of unaffected ganglionic bowel.

Despite correct removal of the aganglionic bowel, up to 40% of the patients continue to suffer from Hirschsprung's associated enterocolitis (HAEC; Frykman and Short, 2012). HAEC is defined as a clinical condition with explosive diarrhea, abdominal distension, fever, and subsequent septic shock (Elhalaby et al., 1995). Immediate treatment of HAEC patients with bowel rest, rectal washouts, adequate resuscitation, and antibiotic treatment has decreased its mortality to about 1%. Even though a variety of different hypotheses have been formulated, the exact pathophysiology of HAEC is still unclear. Recent studies have shown that a disruption of the intestinal mucosal barrier, an abnormal immune response of the intestinal tract and infection due to specific pathogens like *Clostridium difficile* may play pivotal roles in the development of HAEC (Hong and Poroyko, 2014). Considering the interrelation between the epithelium, the immune system and the microbiome of the intestine, disturbances of the intestinal microbial composition may predispose a patient to develop HAEC independent of correct surgical treatment.

Both, experimental and clinical studies recently have given a first insight into an altered intestinal microbiome in Hirschsprung's disease and HAEC. Applying 16S rRNA gene pyrosequencing Ward et al. (2012) assessed the intestinal microbiome in a murine model of HAEC for the first time. The authors used endothelin receptor B knockout mice (Ednrb –/–) as an established experimental model of intestinal aganglionosis. Colonic and fecal samples were analyzed at different time points and compared to wild type mice (WT). WT exhibited increasing species diversity with age, while mutant mice possessed an even greater diversity. On the phylum level, mutant mice contained more Bacteroidetes and less Firmicutes compared to WT mice. Based on these results the ENS can be added to the list of regulatory host factors influencing microbial composition. These findings were – at least partially – confirmed by a study comparing bacterial DNA of cecal contents between Ednrb knockout and heterozygous mice (Pierre et al., 2014). The heterozygous mice demonstrated decreased levels of Bacteroidetes and Proteobacteria with increased Firmicutes compared to the knockout mice. Additionally, mutant mice exhibited a reduced ileal expression and activity of secretory phospholipase A2 (sPLA2), an antimicrobial molecule secreted by Paneth cells at the base of the intestinal crypts. These results suggest that Hirschsprung's disease is not limited to the absence of ganglia within the gastrointestinal tract, but also affects the mucosal immune system and subsequently the microbial composition of the GI tract.

De Filippo et al. (2010) assessed 15 stool samples of a 3-years old Hirschsprung patient harvested during four HAEC episodes and remission phases, respectively. The samples were analyzed using amplified ribosomal DNA restriction analysis (ARDRA). This analysis revealed that HAEC episodes clustered together suggesting a sort of predisposing bacterial community for HAEC development. The available human studies on HAEC applying high-throughput sequencing are summarized in **Table 2**. Yan et al. (2014) assessed the microbial signature of intestinal contents taken during surgery from different sections along the intestinal tract in a study population consisting of four patients (two patients with HEAC and two patients with Hirschsprung's disease). Bacteroidetes and Proteobacteria accounted for the highest proportion among the intestinal flora in Hirschsprung's patients. In contrary, Proteobacteria and Firmicutes were the most common microbes in HAEC patients. At the genus level, marked differences comparing HAEC and Hirschsprung's patients were observed. The altered Firmicutes/Bacteroidetes ratio found in this study confirms the above-mentioned changes described in experimental settings and fuels the speculation that the increased ratio is associated with a pro-inflammatory state of the intestinal microbiomes. The alterations of the dominating bacterial phylum reduce the endogenous production of GLP-2, an intestinal peptide enhancing tight junctions of the cells and thereby preventing LPS from entering plasma (Cani et al., 2009). In a larger group of patients suffering from Hirschsprung's disease consisting of 19 children, 9 with a history of HAEC and 9 without, fecal DNA was isolated and the bacterial and fungal microbiome was analyzed subsequently (Frykman et al., 2015). Even though bacterial microbiome analysis revealed some differences between the two groups at the phylum level, the changes did not reach statistical significance. In contrast, the fecal fungal composition (the mycobiome) of the HAEC group showed a marked reduction in diversity with increased *Candida* sp. and reduced *Malassezia* and *Saccharomyces* sp. compared with the group of patients without HAEC. These results therefore identified *Candida* sp. as a potential player in HAEC, either as an expanded commensal species as a consequence of enterocolitis or its treatment or even possibly contributing to the pathogenesis of HAEC (Frykman et al., 2015). Once more it should be emphasized that fecal samples may not reveal the microbial diversity of the diseased



*HAEC, Hirschsprung's associated enterocolitis; HD, Hirschsprung's disease.*

bowel and even luminal samples may not represent the mucosal microbiome (Haange et al., 2012). Thus, further research seems to be mandatory.

Taken together, there is a limited but growing body of evidence that a shift in the intestinal microbiome with respect to the colonization with specific intestinal bacteria may affect the intestinal immune responses causing a susceptibility to recurrent life-threatening episodes of HAEC. Additionally, it has already been shown that an altered transit time due to disruptions of the intestinal motility is an important factor for shaping the intestinal microbiome (Gorkiewicz et al., 2013). Nevertheless, it still seems too early to recommend possibilities to alter the intestinal microbiome in a therapeutic way. Further studies have to be performed to reveal the crosstalk between the intestinal microbiota and the immune system.

#### DISCUSSION AND FUTURE PERSPECTIVES

Especially in neonates or preterm babies the immunological, structural, and metabolic interaction between the microbiome and the gastrointestinal tract is not completely understood yet. Latest (and future) 16S rRNA gene-based and metagenomic analyses may provide novel insights into the development of the gastrointestinal microbiome in such patients. In addition to genome-based information, functional data (assessed via transcriptomics or metabolomics) of the infant microbiome need to be retrieved allowing the vision of an improved diagnosis and treatment of infectious diseases in the future.

At present we already have scientific evidence for diseases like NEC, SBS, and HAEC that dysbiosis of the microbiome may influence the course of the diseases. Further insights into the molecular interaction between the microbes and the gut, as well as high-resolution visualization of the interplay seem mandatory as present data are not conclusive.

One challenge associated with studies of the microbiome in pediatric surgical diseases like NEC, SBS, and HAEC is the low incidence of these diseases. The resulting low numbers of individuals combined with the presence of potential confounding factors make interpretation of "microbiome data" challenging. Confounding factors include mode of delivery and feeding (breast milk vs. formula; enteral vs. parenteral). For instance, in the case of caesarian delivery a different set of environmental bacteria (similar to the skin communities of the mothers) form the basis for the infant's microbiome compared to the microbiome of infants delivered vaginally (Biasucci et al., 2008; Dominguez-Bello et al., 2010). Moreover, while the microbiome of breast-fed infants is dominated by bifidobacteria the counts of *Escherichia coli*, *Clostridium difficile*, *Bacteroides fragilis*, and lactobacilli are higher in exclusively formula-fed infants (Penders et al., 2006). In a recently published study it has also been shown that immaturity and perinatal antibiotics strongly affected the infant's microbiome (Arboleya et al., 2015).

Nutrition as well as environmental conditions of the neonatal intensive care unit can also strongly influence the microbial development. For example, it has been shown that human milk carries pre- and probiotics, which are necessary for an ideal colonization of the gut and could thus, help pre-term and lowweight infants to develop optimally (Sela and Mills, 2010).

Moreover, the environment seems to matter. Recent studies indicate that (neonatal) intensive care units harbor hot spots of possibly pathogenic microorganisms located on the medical personnel and equipment (Gastmeier et al., 2007; Oberauner et al., 2013). Due to the frequent and harsh cleaning processes in such areas, the natural microbial community is strongly influenced and shifted toward a potentially more resistant microbiome, as it has also been observed for highly cleaned and monitored clean rooms (Moissl-Eichinger et al., 2013). An overview summary of neonatal intensive care unit outbreaks has identified *Klebsiella* sp. to be responsible for most outbreaks. This bacterium, which is widely distributed in various habitats including the sinks of patient rooms (Leitner et al., 2015), possesses a polysaccharide capsule, which makes this microbe more robust against desiccation – a critical feature to survive on surfaces or on skin.

Finally the babies age matters. Brooks et al. (2014) have shown that VLBW infants adopt microorganisms from their hospital environment. Microbial reservoirs in the rooms inoculate the infant's intestinal tract and thus impact the development of the microbiome. From there, the microorganisms are distributed again into the environment, creating a cycle of inoculation. In this study, the most probable reservoirs for different microorganisms were found to be tubings and surfaces, whereas hands or skin contributed to a lesser extent.

#### CONCLUSION

The interplay of the microorganisms with VLBW infants is extremely complex, and understanding the processes and finding possibilities to control this sensitive interaction requires

#### REFERENCES


are large (international) collaboration. With the joint goal to understand the impact of the microbiome on infant development also geographical differences between regions for the same disease can be studied, with the chance to improve health and development for generations of VLBW- and hospitalized newborns. However, collecting fecal samples may not be sufficient to define the microbiome of the diseased bowel, because Haange et al. (2012) have shown in an experimental setting that the microbial diversity differed considerably along the intestinal tract and even the luminal (transient) microbiome displayed a different diversity compared to the mucosal (resident) microbiome at the same portion of the gut. Finally, epidemiological studies beyond infancy seem promising to observe the "normal development" of the microbial diversity (e.g., in the appendix) and ensure a protective and beneficial symbiosis.

dynamics in Hirschsprung's disease-associated enterocolitis: a pilot study. *Pediatr. Surg. Int.* 26, 465–471. doi: 10.1007/s00383-010-2586-5


**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 © 2015 Till, Castellani, Moissl-Eichinger, Gorkiewicz and Singer. 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) or licensor 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.*

# Regulation of lung immunity and host defense by the intestinal microbiota

#### Derrick R. Samuelson, David A. Welsh and Judd E. Shellito\*

*Section of Pulmonary/Critical Care and Allergy/Immunology, Department of Medicine, Louisiana State University Health Sciences Center, New Orleans, LA, USA*

Every year in the United States approximately 200,000 people die from pulmonary infections, such as influenza and pneumonia, or from lung disease that is exacerbated by pulmonary infection. In addition, respiratory diseases such as, asthma, affect 300 million people worldwide. Therefore, understanding the mechanistic basis for host defense against infection and regulation of immune processes involved in asthma are crucial for the development of novel therapeutic strategies. The identification, characterization, and manipulation of immune regulatory networks in the lung represents one of the biggest challenges in treatment of lung associated disease. Recent evidence suggests that the gastrointestinal (GI) microbiota plays a key role in immune adaptation and initiation in the GI tract as well as at other distal mucosal sites, such as the lung. This review explores the current research describing the role of the GI microbiota in the regulation of pulmonary immune responses. Specific focus is given to understanding how intestinal "dysbiosis" affects lung health.

Keywords: Gut-Lung Axis, intestinal microbiota, immunology, pulmonary infections, pulmonary immunology, dysbiosis

#### Introduction

Respiratory tract infectious diseases, such as influenza and pneumonia, result in the death of 3·2 million people annually worldwide (WHO, 2014). Most of the current therapies used in the treatment and management of these diseases are suboptimal as antibiotic resistance, efficacy, and toxicity have been difficult to overcome (Keely et al., 2011). Infection of the respiratory tract represents a breakdown of the host's immune defenses. In addition, non-infections respiratory diseases are the third and fifth (infections respiratory diseases are the fourth) leading causes of death worldwide (WHO, 2014). Understanding the mechanisms that mediate cross-talk between the gastrointestinal (GI) tract and lung defenses and how this interaction facilitates optimal lung health is of growing interest. More specifically, the role of the GI microbiota in mediating, maintaining, and regulating this cross-talk represents an exciting area of research that is poised to aid in the development of novel treatment and management strategies for lung disease.

### The Human "Super-Organism": The Role of the Gastrointestinal Microbiota in Health

The importance of the homeostatic maintenance of human health by the intestinal microbiota has become a topic of great interest (Noverr and Huffnagle, 2004; Lupp et al., 2007; Maslowski et al., 2009; Garrett et al., 2010; Hooper et al., 2012; Bollrath and Powrie, 2013;

#### Edited by:

*Gabriele Berg, Graz University of Technology, Austria*

#### Reviewed by:

*Jan S. Suchodolski, Texas A&M University, USA Marius Vital, Michigan State University, USA*

#### \*Correspondence:

*Judd E. Shellito, Section of Pulmonary/Critical Care and Allergy/Immunology, Department of Medicine, Louisiana State University Health Sciences Center, 1901 Perdido Street, Suite 3205, New Orleans, LA 70112, USA jshell@lsuhsc.edu*

#### Specialty section:

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

Received: *29 June 2015* Accepted: *22 September 2015* Published: *07 October 2015*

#### Citation:

*Samuelson DR, Welsh DA and Shellito JE (2015) Regulation of lung immunity and host defense by the intestinal microbiota. Front. Microbiol. 6:1085. doi: 10.3389/fmicb.2015.01085*

**43**

Sansonetti, 2013). Evolution of an individual's microbiota begins shortly after birth cumulating in a stable adult microbiota by the age of two (Foxx-Orenstein and Chey, 2012). This microbial community includes autochthonous (permanent inhabitants) and allochthonous (transient inhabitants) microorganisms. Microbiota of the human GI tract contains bacterial (microbiota), viral (virome), and fungal (mycobiota) species. Surprisingly, approximately 60% of these organisms cannot be grown in traditional culture (Noverr and Huffnagle, 2004; Hooper et al., 2012). However, new methods known as "microbial culturomics," which utilize 212 different culture conditions have allowed for a significant advancement in our ability to culture intestinal microorganisms (Lagier et al., 2012). The average human adult intestinal microbiota is composed of approximately 400– 1000 species (Noverr and Huffnagle, 2004; McLoughlin and Mills, 2011; Hooper et al., 2012). However, it is estimated that roughly 30-40 species dominate this niche, with bacteria from the genera Bacteroides, Bifidobacterium, Eubacterium, Fusobacterium, Clostridium, and Lactobacillus highly represented (McLoughlin and Mills, 2011). In addition, intestinal microbial diversity and composition changes not only along the length of the intestinal tract but is spatially distributed between the mucosa and the lumen of the intestinal tract within each region (Hill et al., 2010; Macpherson and McCoy, 2013). Many environmental factors will drastically alter the normal intestinal microbiota (Noverr and Huffnagle, 2004). Changes in diet, the use of antibiotics, chemotherapy, GI tract infection, and host immune status significantly alter, either transiently or permanently, the intestinal ecosystem (Round and Mazmanian, 2009; Hooper and Macpherson, 2010; Hooper et al., 2012). Alterations of the microbiota that lead to intestinal dysbiosis (a microbial imbalance within the intestinal tract) are characterized by a loss or significant decrease in the amount of beneficial bacterial species and/or an outgrowth or population shift of other species. Intestinal dysbiosis can affect overall health in multiple ways such as growth of opportunistic bacterial pathogens, alterations in host's metabolic profiles, and/or increased inflammation. This review will focus on the microbiota as it affects pulmonary immunity.

#### Maintenance of the Intestinal Microbiota

Alterations of the intestinal microbiota not only affect the growth of opportunistic pathogens but can have a broad impact on immune status and function within the host (Hooper et al., 2012). The impact of the GI microbiota on host mucosal immunity has been studied extensively in germ-free mice (mice without any intestinal microbiota). Germ-free mice exhibit impaired GI development characterized by smaller Peyer's patches, fewer CD8αβ intraepithelial lymphocytes, underdeveloped isolated lymphoid follicles, and lower levels of the mucosal IgA antibodies (Hooper et al., 2012). The specific microbial molecules or components that inform host immune development are still being discovered and characterized. These interactions are crucial for the maintenance of host-microbial homeostasis. This topic has been reviewed in several recent articles (Round and Mazmanian, 2009; Hooper and Macpherson, 2010; Hooper et al., 2012). **Figure 1** highlights a current overview of understanding of how the GI microbiota shape immune responses and how the host immune system shapes the GI microbiota.

### The Intestinal Microbiota and Systemic Immunity

Commensal microorganisms modulate host immunity not only in the intestinal tract but at distal sites as well (Kieper et al., 2005). The intestinal microbiota affects systemic immune responses by modulation of several key pathways; expansion of extra-intestinal T cell populations, production of short-chain fatty acids, development of oral tolerance, and control of inflammation.

#### Regulation of T Cell Populations

Expansion and differentiation of extra-intestinal T cell populations are meditated by the intestinal microbiota (Kieper et al., 2005). Several recent studies have shown that the intestinal microbiota is critical for maintenance of T cell subsets that are important for systemic immunity. The intestinal microbiota is required for expansion of CD4+ T cells, regulatory T cells, Th1 or Th2 responses, and Th17 T cells. For example, colonization of germ-free mice with Bacteroides fragilis that synthesize PSA results in a higher number of circulating CD4+ T cells and levels of circulating Th1 cells compared to mice colonized with B. fragilis unable to produce PSA (Mazmanian et al., 2005). While, colonization of gnotobiotic mice with a cocktail of mouse derived Clostridial strains enhances anti-inflammatory signaling by directing the expansion of lamina propria and systemic regulatory T cells (Treg) with an associated increase in IL-10 secretion (Atarashi et al., 2011). The specific Clostridial strain that drives this regulatory affect is not known. Further, mice with high levels of Bacteroides vulgatus colonization exhibit a biased T cell differentiation favoring a Th2 over a Th1 phenotype, as characterized by increased levels of IgE, IgG1, IL-4 and decreased IFNγ (Sudo et al., 2002) Finally, colonization of germ-free mice with segmented filamentous bacteria (SFB) induces expansion of the Th17 cell population and a slight increase in Th1 cells (Abraham and Cho, 2009; Wu et al., 2010; Lee et al., 2011).

#### Oral Tolerance

Development of oral tolerance occurs following oral administration of antigen and represents a local and systemic immunological state of immune unresponsiveness to a subsequent antigen challenge. Low doses of antigen favor active suppression, whereas higher doses favor clonal deletion of antigen-specific T cells. Ingestion of oral antigen induces expansion of Th2 and Th3 T cells and CD4+ CD25+ regulatory cells and latency-associated peptide+ T cells (Faria and Weiner, 2005). Further, individuals with impaired intestinal permeability often have dysfunctional oral tolerance. Impaired intestinal permeability also leads to inadequate production of IgE and recruitment of mast cells in the GI mucosa. Individuals suffering from these conditions exhibit enhanced IgE-CD23-mediated transport across the mucosa and increased levels of inflammatory mediators, such as proteases and cytokines, which further affect intestinal permeability. This leads to an increase in the leakage of

allergens and hence contributes to perpetuate the inflammatory reaction (Perrier and Corthésy, 2011). Further, Cassani and colleagues recently observed defective oral tolerance in CCR9 deficient mice (CCR9 targets T cells to the small intestine) and that defective oral tolerance in CCR9-deficieint mice could be restored by transfer of wild-type T cells (Cassani et al., 2011). However, Pabst and co-workers found that CCR9-deficient mice developed normal oral tolerance to ovalbumin. Pabst and co-workers suggest that these differences may be due to the differences in individual strains of CCR9-deficient mice, or that differences in the composition of the microbiota may influence the impact of CCR9 on oral tolerance (Pabst and Mowat, 2012). There are controversial reports on the capacity to induce oral tolerance in germ-free mice devoid of live intestinal bacteria (Walton et al., 2006; Ishikawa et al., 2008). Yet, numerous studies have demonstrated that the intestinal microbiota has a profound effects on the immune system. Therefore, it is conceivable that differences in microbiota composition may also affect oral tolerance.

#### Production of Short-chain Fatty Acids

Dietary fermentable fiber content changes the composition of the GI microbiota, in particular by altering the ratio of Firmicutes to Bacteroidetes. Alteration of the ratio of Firmicutes to Bacteroidetes directly affect how the gut microbiota metabolize fiber, consequently increasing or decreasing the concentration of circulating short-chain fatty acids (SCFAs). Intestinal microbiotameditated production of various SCFAs have also been shown to be important for host systemic immunity (Meijer et al., 2010; den Besten et al., 2013; Trompette et al., 2014). More precisely, SCFAs, especially butyrate, seem to exert broad antiinflammatory activities by affecting immune cell migration, adhesion, cytokine expression, as well as, cellular proliferation, activation, and apoptosis through the activation of signaling pathways (NF-κB) and inhibition of histone deacetylase. In addition, histone deacetylase inhibitors enhance the numbers and function of Treg cells (Meijer et al., 2010). Two recent studies demonstrated that short-chain fatty acids directly regulate/prime the size and function of Treg cell pool in the colon. Both studies showed that mice are protected from colitis through butyrate induced differentiation of Treg cells in a Ffar2-dependent manner (Furusawa et al., 2013; Smith et al., 2013). Furusawa and colleges also demonstrated that treatment of naive T cells under the Treg-cell-polarizing conditions with butyrate enhanced histone H3 acetylation in the promoter and conserved non-coding sequence regions of the Foxp3 locus, which they proposed may be the possible mechanism for how microbial-derived butyrate regulates the differentiation of Treg cells (Walton et al., 2006). Increased levels of butyrate also induce the expression of IL-10, which influence the balance between Th1, cytotoxic CD8+ T cells and Treg cells. Finally, SCFAs are also important in the control of allergic inflammation. Trompette and colleagues found that mice fed a high-fiber diet had increased circulating levels of SCFAs and were protected against allergic inflammation in the lung, whereas a low-fiber diet decreased levels of SCFAs and increased allergic airway disease. Specifically, increased levels of SFAs lead to enhanced generation of dendritic cell precursors and subsequent seeding of the lungs by DCs with high phagocytic capacity, which was accompanied by an impaired ability to promote Th2 cell effector function (Trompette et al., 2014).

#### Regulation of Systemic Inflammation

Several recent studies have provided insight into the role that the commensal microbiota has on influencing systemic inflammation (Noverr et al., 2004, 2005; Ichinohe et al., 2011). Disease severity in animal models of colitis are all dependent on the animal's intestinal microbial communities (Abraham and Cho, 2009; Wu et al., 2010; Lee et al., 2011). For example, germ-free mice with chemically induced colitis exhibit markedly attenuated pathological signs of colitis and restoration of the intestinal microbiota prevents the attenuation. This suggests that the intestinal microbiota is crucial for modulating the host's ability to control inflammation. Further, Verdam et al., found that obese humans exhibit a reduced bacterial diversity, a decreased Bacteroidetes/Firmicutes ratio, and an increased abundance of potential proinflammatory Proteobacteria (Verdam et al., 2013). The shifts in the intestinal microbiota populations were also accompanied by increased levels of fecal calprotectin and plasma C-reactive protein, which suggest that the intestinal microbiota alterations found in obese humans are associated with local and systemic inflammation and that the obesity-related microbiota has a proinflammatory effect (Verdam et al., 2013). Finally, Biagi and co-workers found that by evaluating the correlation between systemic inflammation and the fecal microbiota that about 9% of the variable microbiota was related to the increased levels of pro-inflammatory cytokines IL-6 and IL-8 (Biagi et al., 2010). All of the taxa that showed a slightly positive correlation with either IL-6 or IL-8 belonged to the phylum Proteobacteria (Biagi et al., 2010). The intestinal microbiota also has many inflammation-suppressing fractions, which function to; counteract some of the inflammatory bacteria, decrease the inflammatory tone of the system, improve the barrier function of the GI mucosa, and prevent inflammation-inducing components from translocating into the body (Hakansson and Molin, 2011). Furthermore, Clostridium cluster XIVa and Faecalibacterium prausnitzii has been demonstrated to possess anti-inflammatory effects by inhibiting NF-κB activation and IL-8 secretion and stimulation of peripheral blood mononuclear cells, which ultimately led to an IL-10/IL-12 ratio that favors antiinflammatory conditions (Sokol et al., 2008). By far, the most studied inflammation-suppressing taxa of the GI microbiota are from the genera of Lactobacillus and Bifidobacterium. Lactobacillus and Bifidobacterium will be discussed further in the probiotics section.

While the all of the direct mechanistic contributions of the GI microbiota on systemic immunity beyond the intestinal mucosa remain to be determined, these studies demonstrate that commensal bacteria can impact host immunity beyond the GI tract. **Table 1** summarizes our current understanding of the effect of the intestinal microbiota on systemic immunity.

### Effects of the GI Microbiota on Pulmonary Health

#### Priming of Intestinal and Lung Mucosal Immunity

It is important to understand the cross-talk and collaboration between the GI tract and the respiratory tract at both an immune and microbial level. Numerous studies have shown that fluids, particles, or even microorganisms deposited into the nasal cavity of mice can also be found in the GI tract a short time later (Southam et al., 2002). In fact, as little as 2·5µl of inoculum into the nasal cavity can later be detected in the GI tract (Southam et al., 2002). Therefore, the GI tract will ultimately be exposed to any pathogen or antigen that is introduced into the respiratory system. This also suggests that the mucosal immune system of the GI tract may serve as a primary sensor of foreign antigens and organisms from the environment. Importantly, disturbances in the intestinal homeostasis by either alterations in the host's genetics or alterations in the microflora could have drastic effects on systemic (e.g., lung) immune responses (see **Figure 2**). We have also provided a summary our current understanding of the effect of the intestinal microbiota on pulmonary health in **Table 2**.

#### Allergies and Asthma

Allergies are often associated with an abnormal Th2 T cell response. Th2 cells are characterized by their ability to produce IL-4, IL-5, IL-9, and IL-13 (McLoughlin and Mills, 2011). The notion that the alterations of immune responses in the gut can directly affect the development of allergic disease in the lung is now widely accepted due to strong epidemiologic (Björkstén et al., 2001) and experimental evidence (Noverr et al., 2004, 2005; Maizels, 2009). A pivotal study reported by Noverr et al. (2004) demonstrated that allergies can develop as a consequence of an altered intestinal microbiota (Noverr et al., 2004). Antibiotic treated mice were given a single oral dose of Candida albicans. This significantly altered the composition of the intestinal microbiota (Noverr et al., 2004). Treated animals had more CD4 cell–mediated inflammation in the lung following aerosol introduction of an allergen compared to mice with a normal GI flora (Noverr et al., 2004). This suggests that alterations in the GI flora can facilitate an immunological state that is predisposed to respiratory allergies. There is a growing interest in understanding other T cell subsets in the development of allergy and asthma, specifically the role of Th17 cells and Th9 cells, which may be impacted by GI microbes (Forsythe, 2014). In addition, Vital and colleagues examined the associations between the intestinal microbiota and allergic airway disease in both young and old mice

#### TABLE 1 | Current understanding of the effects of the intestinal microbiota on systemic immunity.


that were sensitized and challenged with house dust mite. They found that the microbial community structure changed with age and allergy development and interestingly that the alterations in the intestinal microbiota from young to old mice resembled the microbial structure of mice after house dust mite challenge. The changes in the intestinal microbial communities were also associated with increased levels of serum IL-17A. Further, old mice developed a greater allergic airway response compared to young mice. Vital and colleagues also suggest the composition of the gut microbiota changes with pulmonary allergy, indicating bidirectional gut-lung communications (Vital et al., 2015).

#### Infectious Diseases

It is evident that the intestinal microbiota plays a crucial role in the regulation and immune response to respiratory viral infections such as influenza (Ichinohe et al., 2011). A recent study from Ichinohe and co-workers demonstrated that the GI microbiota directly influenced virus-specific CD4 and CD8 T cell subsets in experimentally infected mice (Ichinohe et al., 2011). Treatment of mice with different antibiotic regimens revealed a population of neomycin-sensitive commensal organisms associated with a protective immune response in the lung following influenza infection. Furthermore, injection of TLR ligands, either locally in the lung or at distal sites, rescued the immune impairment in the antibiotic-treated mice. In addition, an intact GI microbiota was required for expression of the proinflammatory chemokines pro–IL-1β and pro–IL-18, which are necessary for influenza clearance (Ichinohe et al., 2010). This suggests that the intestinal microbiota provides microbial signals or determinants that are critical for immune priming and shaping the response to viral pneumonia.

Similar observations regarding the critical role of the intestinal microbiota in the regulation and immune response to respiratory bacterial infections have also been made (Fagundes et al., 2012). Fagundes et al. (2012) employed germ-free mice to analyze the ability of the host to resist bacterial infection. Germ-free mice were highly susceptible to pulmonary infection with the bacterial pathogen Klebsiella pneumonia. The enhanced susceptibility to K. pneumoniae was associated with increased levels of IL-10, which suppresses neutrophil recruitment, and permits pathogen growth and dissemination. The administration of a TLR agonist followed by LPS inoculation prevented pulmonary K. pneumoniae infection, reduced IL-10 secretion, normalized TNF-α and CXCL1 levels, and neutrophil mobilization to the lungs (Fagundes et al., 2012). Germ-free mice that were conventionalized (normal mouse intestinal flora has been restored) had significantly less K. pneumoniae in the lungs and blood (Fagundes et al., 2012). Conventionalization also restored neutrophil influx, CXCL-1, TNF-α, and IL-10 to levels found in wild-type mice (Fagundes et al., 2012). These findings suggest that the commensal microbiota maintain host defenses to infectious agents by facilitating a normal inflammatory response to pulmonary pathogens.

#### Gut-derived Sepsis and Acute Respiratory Distress Syndrome

Gut-derived sepsis is the process during which gut-derived proinflammatory microbial and non-microbial factors induce or enhance a systemic inflammatory response syndrome (SIRS), acute respiratory distress syndrome (ARDS), or multiple organ dysfunction syndrome (MODS). Several mechanistic theories of gut-derived sepsis leading to SIRS, ARDS, or MODS, have been postulated (Deitch, 2002, 2012; Senthil et al., 2006; Clark and Coopersmith, 2007; Deitch and Ulloa, 2010). The "gutlymph" theory proposes that macrophages and other immune cells in the intestinal submucosa or the mesenteric lymph nodes are sufficient to contain the majority of translocating bacteria. However, any surviving bacteria, cell wall fragments, or protein components of the dead bacteria that escape macrophage containment together with cytokines and chemokines produced in the gut, travel along the mesenteric lymphatics to the cisterna chyli. These products then enter into the systemic circulation through the left subclavian vein, via the thoracic duct. Access to the pulmonary circulation leads to uncontrolled activation of alveolar macrophages leading to acute lung injury or ARDS and then MODS (Senthil et al., 2006). Several experimental models support this theory. For example, experimental models of endotoxinemia (Watkins et al., 2008) trauma-hemorrhagic shock (Senthil et al., 2007) or burn injury (Lee et al., 2008) all support this theory. An additional theory put forward by Clark and Coopersmith is the "intestinal crosstalk" theory (Clark and Coopersmith, 2007). This theory assumes a three-way partnership among the intestinal epithelium, immune tissues, and the endogenous microflora of the gut. Within this three dimensional relationship, each factor modifies the others through crosstalk. During normal homeostasis all three components interact normally, which facilitates intestinal crosstalk with extraintestinal tissues. However, in critically ill patients, loss of the balance between these highly interrelated systems results in the development of systemic manifestations of disease, specifically SIRS, ARDS, or MODS (Clark and Coopersmith, 2007). The mechanisms governing gut-derived sepsis and ARDS are poorly understood and are actively being investigated.

#### Chronic Obstructive Pulmonary Disease

COPD is an inflammatory disorder characterized by incomplete reversible airflow obstruction leading to increased mortality and morbidity (Keely et al., 2012; Hui et al., 2013). It has been known for several years that individuals who suffer from COPD have an altered lung microbiome compared to healthy individuals. (Keely et al., 2012; Hui et al., 2013). There is evidence that components of the gastrointestinal microflora, specifically Gram negative bacilli, may also make up a component of the lung microflora and may be increased in individuals with COPD (Keely et al.,

#### TABLE 2 | Current understanding of the effects of the intestinal microbiota on pulmonary immunity.


2012; Hui et al., 2013). These bacteria are resistant to cigarette smoke and may contribute to severe exacerbations of COPD (Keely et al., 2012; Hui et al., 2013). While no definitive studies on the effect of smoking on the respiratory or intestinal microbiome have been performed, it is possible that smoke-induced changes to the intestinal microbiome may exacerbate COPD symptoms (Keely et al., 2012; Hui et al., 2013).

#### Potential Mechanisms of GI Mediated Lung Immunity

The mechanisms by which the intestinal microbiota exert a systemic immunomodulatory effect are not fully understood, but several potential pathways may be involved. Highlighted below are several potential mechanisms regulating gut-mediated systemic immunity. **Figure 3** provides a conceptual figure of our current understanding to the potential mechanisms involved in the immune regulation along the gut-lung axis.

#### Toll Like Receptor (TLR) Activation

The intestinal immune system initiates immune signaling events via the interactions of the gut microbiota with pattern recognition receptors of the innate immune system (i.e., TLRs). TLRs recognize microbial components and trigger inflammatory responses (Abreu, 2010). Different bacterial products, such as lipopolysaccharide, lipoteichoic acid, CpG, peptidoglycan, and polyinosinic:polycytidylic acid, stimulate TLR signaling (Abreu, 2010). One downstream effect of TLR signaling is the activation of the transcription factor NF-κB, which is required for expression of many genes regulating innate immunity and inflammation (Abreu, 2010). The intestinal microbiota is crucial for maintaining normal TLR signaling (Round et al., 2011; Fagundes et al., 2012). Microbiota-mediated activation of antigen-specific CD4 and CD8 T cells, pathogen specificantibodies, steady state expression of mRNA for pro–IL-1β and pro–IL-18, inflammasome activation, and migration of dendritic cells (DCs) from the tissue to the draining lymph node, which leads to normal T cell priming, all occur in a TLR dependent manner (Ichinohe et al., 2010). Further, intestinal initiated TLR signaling has been shown to induce lung immune responses by Ichinohe and colleagues, who showed that a single dose of LPS delivered intrarectally, restored an immune response in the lung of influenza infected mice (Ichinohe et al., 2010). The elucidation of the major bacterial species or bacterial products necessary to maintain normal microbiota-TLR signaling is an area primed for investigation.

#### T and B Cell Homing

Tissue specific homing of lymphocytes is crucial for an effective immune response and clearance of infection. The localization and homing of lymphocytes is determined by expression of integrin and chemokine receptors, such as CCchemokine receptor 9 (CCR9) (Christensen et al., 2002; Niess and Reinecker, 2005; Meijerink and Wells, 2010). Specific adhesion and chemokine receptors expressed by lymphocytes allow these immune cells to target tissues that express their cognate ligands

and home to areas of high chemokine secretion (Christensen et al., 2002). T cells acquire the capacity to home to non-lymphoid tissues by direct interaction with mucosal DCs at the sites of antigen acquisition. DC-mediated imprinting of T cells confers selectivity for specific non-lymphoid tissues, such as the gut, skin, and lung (Sigmundsdottir and Butcher, 2008; Hart et al., 2010; Mikhak et al., 2013). Dendritic cells sense antigen in tissues before migrating to draining lymph nodes, where they have the ability to activate and influence the differentiation of naïve T cells. Gut DCs promote the expression of α4β7 and CCR9 on T cells, and, in doing so, enable T cells to migrate to the small intestine by homing to the intestinal ligand MAdCAM-1 and chemokine CCL25 (Sigmundsdottir and Butcher, 2008). Alternatively, lung DCs promote the expression of CCR4 on T cells, which allows for activated T cells to traffic into the lung via increased levels of CCL17 (Mikhak et al., 2013).

The ability to induce tissue-specific homing of lymphocytes suggests that mucosal surfaces can not only induce homing back to the site of antigen acquisition but may be able to target distal mucosal sites as well via altered lymphocyte priming. In support of this, Ruane et al. (2013), found that lung DCs can specifically up-regulate the expression of the gut-homing integrin α4β7 on T cells, which guided migration to the GI tract. Consistent with this, intranasal immunization with Salmonella induced protective immunity against enteric challenge with Salmonella and was dependent on lung DCs (Ruane et al., 2013).

Additional mechanisms allow for lymphocytes to traffic to multiple mucosal sites to combat infection at mucosal surfaces. For example, CCR6 is expressed on immature DCs, most B cells, subsets of CD4+ and CD8+ T cells, and Natural Killer T cells (Ito et al., 2011). The cognate ligand of CCR6, CCL20, is expressed by a variety of epithelial cell types including keratinocytes, pulmonary epithelial cells, and intestinal epithelial cells (Ito et al., 2011). The expression of CCL20 in these tissues remains at a low steady-state level, but is strongly induced by pro-inflammatory signals and TLR agonists originating from bacterial species (Ito et al., 2011). These data suggest that pathogens/antigens encountered in the gut or the lung can prime central and effector memory T cells, which then home to sites of new infection (i.e., sites of inflammation). Thus, CCR6 may be critical for an effective immune response within the gut-pulmonary axis.

### Targeting the Intestinal Microbiota in the Prevention of Lung Related Infectious Diseases

#### Prevention of Respiratory Infections with Probiotics

The use of probiotics for the treatment of various diseases and for the maintenance of health in general has become an intensely studied area, with broad appeal. Probiotics are now used to treat a variety of aliments including diarrhea, gas, cramping, vaginal yeast infections, urinary tract infections, and to help control inflammatory bowel disease (IBD). Studies are also underway evaluating the benefits of probiotics in the treatment of colon cancer, skin infections, irritable bowel syndrome (IBS), liver disease, rhinoconjunctivitis/rhinosinusitis, and lung health (Forsythe, 2011; Eslamparast et al., 2013; Kumar et al., 2013; Baquerizo Nole et al., 2014; Kramer and Heath, 2014; Sandhu and Paul, 2014). **Table 3** describes the current research targeting the intestinal microbiota for the prevention of lung related infectious diseases.

The benefits of probiotics in the maintenance and regulation of lung health has been described in several studies (Forsythe, 2011, 2014; Yoda et al., 2012; West, 2014). One of the first studies indicating that intestinal microflora may influence lung health came from the study of obese mice. Researchers found that in obese mice the GI microbiota plays an important role in controlling inflammation in the lungs (Yoda et al., 2012). Mice fed heat-killed Lactobacillus gasseri had significant increases in pulmonary mRNA expression of cytokines and other immune molecules accompanying the changes in their GI bacterial profiles (Yoda et al., 2012). These results suggest that Lactobacilli may stimulate the respiratory immune responses of mice to enhance host defenses against respiratory infection by increasing inflammatory signaling.

Additionally, the probiotic bacteria Lactobacillus rhamnosus displays immune stimulatory effects associated with increased resistance to infection (Salva et al., 2010). L. rhamnosus feeding not only attenuated infection with Salmonella Typhimurium, an intestinal pathogen, but also conferred resistance to infection with Streptococcus pneumoniae, a respiratory pathogen. Probiotic treatment decreased the burden of S. pneumoniae in the lung, prevented dissemination to the blood, and increased INFγ, IL-6, IL-4, and IL-10 in bronchoalveolar lavage (BAL) fluid (Salva et al., 2010). Interestingly, while both of the tested L. rhamnosus strains improved resistance to intestinal S. Typhimurium, only one strain provided a beneficial protective effect against pulmonary infection with S. pneumoniae (Salva et al., 2010). This suggests differential probiotic effects and that different infectious diseases may benefit from a specific and unique probiotic treatment regime.

Furthermore, Alvarez et al. (2001) found that mice administrated Lactobacillus casei prior to pulmonary challenge with Pseudomonas aeruginosa exhibited increased pathogen clearance, phagocytic activity of alveolar macrophages, and IgA in BAL (Alvarez et al., 2001). Similarly, Hori et al. (2001) observed parallel results in murine viral infection models. Feeding mice L. casei for 4 months prior to challenge reduced influenza viral titers in nasal washings, which was accompanied by a significant increase in natural killer (NK) activity and IFNγ and TNF-α production (Hori et al., 2001).

#### Prevention of Respiratory Infections with Oral Vaccines

Over the past two centuries the study and use of vaccines has become increasingly sophisticated. However, even with technological advances many infectious diseases have thwarted the development of effective vaccines. While many routes of vaccination exist, one emerging area of vaccinology is mucosal immunization. This approach provides several benefits over

TABLE 3 | The current research targeting the intestinal microbiota for the prevention of lung related infectious diseases.


conventional systemic vaccination, including higher levels of antibodies and protection at the mucosal surface. In addition, mucosal vaccines can target specific mucosal surfaces such as the respiratory, genital, or intestinal mucosa. Noteworthy is that vaccination at one mucosal surface often confers resistance at other sites (Ryan et al., 2001; Pasetti et al., 2011). Thus, it is possible to immunize against respiratory pathogens using a gut vaccination strategy.

Several studies have addressed the potential of gut-mediated lung immunity (Doherty et al., 2002; Izadjoo et al., 2004; Aldwell et al., 2006; KuoLee et al., 2007). For example, Doherty et al. (2002) sought to protect the lung against infection with Mycobacterium tuberculosis by targeting the gut mucosa. Comparing subcutaneous immunization of mice with oral immunization using a subunit vaccine carrying two M. tuberculosis immunodominant proteins, they found that oral vaccination was relatively less effective. However, by using a heterologous priming and boosting strategy, oral immunization induced significant systemic type 1 responses, which were comparable to, or better than, those obtained following standard subcutaneous immunization protocols. Moreover, the immune responses correlated with protection against subsequent aerosol infection with virulent M. tuberculosis similar to or greater than that obtained by repeated subcutaneous vaccinations (Doherty et al., 2002).

Evaluating the ability of orally administered live attenuated Brucella melitensis to elicit cellular and humoral immune responses and to protect mice against intranasal challenge with virulent B. melitensis, Izadjoo and colleges found increases in serum antibodies directed against lipopolysaccharide and non-O-polysaccharide antigens. Additionally, orally delivered B. melitensis elicited a systemic response as characterized by increased production of IL-2 and IFNγ from splenocytes. Oral immunization of mice with live attenuated B. melitensis protected against disseminated infection and enhanced clearance of the challenge inoculum from the lungs. Optimal protection following inoculation with live bacteria was dose dependent and enhanced by a booster vaccine inoculation. These data suggest that oral immunization may provide protection from pneumonia due to Brucella (Izadjoo et al., 2004).

Two additional studies utilizing oral vaccination demonstrated protective immune responses against Mycobacterium spp. and Francisella tularensis (Aldwell et al., 2006; KuoLee et al., 2007). Both studies showed an induction of antigen-specific antibody responses in the serum and bronchoalveolar lavage fluids, proliferation of antigenspecific IFN-γ producing cells, and an overall reduction in pathogen burden in the lung (Aldwell et al., 2006; KuoLee et al., 2007). These studies demonstrate the potential of GI mucosal vaccination to prevent lung infections. Furthermore, these data suggest that oral vaccination may represent an attractive alternative strategy for the prevention of infection with these two pathogens.

#### Prevention of Respiratory Infections with Probiotic Coupled Vaccines

Preclinical data suggests that vaccine responses may be improved by modulation of the gut microbiota. Several animal studies and human clinical trials have been performed. Mice orally immunized with a recombinant Lactococcus lactis modified to express the pneumococcal protective protein A (PppA) in the cell wall were protected from subsequent lung infection with Streptococcus pneumoniae. Oral immunization with L. lactis expressing PppA increased anti-PppA IgM, IgG, and IgA antibodies in serum, in bronchoalveolar, and in intestinal lavage fluids. A mixture of Th1 and Th2 responses were observed, characterized by the presence of both IgG1 and IgG2a anti-PppA antibodies in serum and BAL and by the production of both IFNγ and IL-4. Furthermore, oral immunization was enhanced by a prime-boost strategy, which induced cross-protective immunity to all S. pneumoniae serotypes tested (Villena et al., 2008). Another animal study evaluated a recombinant oral L. lactis vaccine against the influenza H5N1 strain. Immunization with L. lactis expressing the H5N1 HA antigen induced HA-specific serum IgG and fecal IgA antibody production and mice were completely protected against a lethal challenge of the H5N1 virus (Lei et al., 2010).

Several human clinical trials have been completed in this area. Highlighted below are probiotic-coupled respiratory vaccine human clinical trials. Lactobacillus rhamnosus GG pretreatment for 28 days enhanced seroprotection against H3N2 influenza following nasal vaccination with the trivalent live attenuated influenza vaccine containing H1N1-like, H3N2-like, and B-like strains. However, no improvement in protection against the H1N1 and B strains was observed at 28 days and no seroprotection was found at 56 days for any strain (Davidson et al., 2011). An additional study examined the effects of 6 wks. of daily Bifidobacterium animalis and Lactobacillus paracasei probiotics administration as an immune primer prior to intramuscular influenza vaccination containing H1N1-like, H3N1-like, and B strains. Individuals given probiotic-coupled vaccination had significantly higher vaccine-specific IgG (IgG1 and IgG3), as well as, significantly higher salivary IgA (Rizzardini et al., 2012). However, another study reported no significant difference between the probiotics and placebo groups when evaluating innate immune response to infection (Akatsu et al., 2013). Taken together animal studies and human clinical trial data suggest that coupling vaccines with probiotics may increase the effectiveness of vaccines. However, the data are conflicting. It is clear that optimization of the dosing strategy, as well as, the type of probiotic used will be critical to achieve maximal vaccine effectiveness.

## Conclusions

While we are now beginning to understand the effect of the GI microbiota on lung immunity, characterization of the lung microbiota may also provide valuable insight into lung-mediated immune regulation in response to influenza and pneumonia, as well as, cystic fibrosis, COPD, allergies, and asthma. This review has focused on the mechanisms that govern the effects of the GI microbiota on immune effector and regulatory functions. While there have been significant advancements made in the past 10 years in our understanding of the global impacts of the GI microbiota on human immune function, many questions

#### TABLE 4 | Current understanding of the GI microbiota and future questions.


still remain. **Table 4** describes our current understanding of GI regulation of the immune system and areas that remain to be explored.

#### Search Strategy and Selection Criteria

References for this review were identified through searches of PubMed and Google Scholar for articles published from January, 1990, to January, 2015, by use of the terms "Gut-Lung Axis," "microbiota," "microbiome," "oral vaccines," "oral vaccines + microbiota," and "microbiota + respiratory infection." Articles resulting from these searches and relevant references cited in those articles were reviewed. Articles published in English were included.

#### References


### Author Contributions

DS, DW, and JS conceived of and designed the review. DS and DW did the literature searches. DS, DW, and JS designed and compiled the figures. DS wrote the review.

#### Acknowledgments

The authors thank Nick de la Rua, Tysheena Charles, and Sanbao Ruan for review of this manuscript. This work was supported by The National Institutes of Health Public Health Service (PHS) Grant #P01-HL076100, The National Institute of General Medical Sciences grant #UG54-GM104940, and The National Institute on Alcohol Abuse and Alcoholism grant numbers P60 AA009803 and R24 AA019661.


autoimmune encephalomyelitis. Proc. Natl. Acad. Sci. U.S.A. 108(Suppl. 1), 4615–4622. doi: 10.1073/pnas.1000082107


gastrointestinal tract. J. Exp. Med. 210, 1871–1888. doi: 10.1084/jem.201 22762


WHO. (2014). World Health Organization. Available online at: http://www.who. int/en/


**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 © 2015 Samuelson, Welsh and Shellito. 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) or licensor 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.

# Mycobiome in the Lower Respiratory Tract – A Clinical Perspective

Robert Krause<sup>1</sup> \*, Christine Moissl-Eichinger1,2, Bettina Halwachs<sup>3</sup> , Gregor Gorkiewicz2,3 , Gabriele Berg<sup>4</sup> , Thomas Valentin<sup>1</sup> , Jürgen Prattes<sup>1</sup> , Christoph Högenauer<sup>5</sup> and Ines Zollner-Schwetz<sup>1</sup>

<sup>1</sup> Section of Infectious Diseases and Tropical Medicine, Department of Internal Medicine, Medical University of Graz, Graz, Austria, <sup>2</sup> BioTechMed, Medical University of Graz, Graz, Austria, <sup>3</sup> Institute of Pathology, Medical University of Graz, Graz, Austria, <sup>4</sup> Institute of Environmental Biotechnology, Graz University of Technology, Graz, Austria, <sup>5</sup> Division of Gastroenterology and Hepatology, Department of Internal Medicine, Medical University of Graz, Graz, Austria

Recently the paradigm that the healthy lung is sterile was challenged and it is now believed that the lungs harbor a diverse microbiota also contributing to the pathogenesis of various diseases. Most of the research studies targeting the respiratory microbiome have focused on bacteria and their impact on lung health and lung diseases. Recently, also the mycobiome has gained attention. Lower respiratory tract (LRT) diseases (e.g., cystic fibrosis) and other diseases or conditions (e.g., HIV infection, lung transplantation, and treatment at intensive care units) have been investigated with regard to possible involvement of mycobiome in development or progression of diseases. It has been shown that diversities of mycobiome in the LRT vary in different populations and conditions. It has been proposed that the mycobiome diversity associated with LRT can vary with different stages of diseases. Overall, Candida was the dominant fungal genus in LRT samples. In this review, we summarize the recent findings regarding the human LRT mycobiome from a clinical perspective focussing on characterization of investigated patient groups and healthy controls as well as sampling techniques. From these data, clinical implications for further studies or routine practice are drawn. To obtain clinically relevant answers efforts should be enhanced to collect well characterized and described patient groups as well as healthy individuals for comparative data analysis and to apply thorough sampling techniques. We need to proceed with elucidation of the role of mycobiota in healthy LRT and LRT diseases to hopefully improve patient care.

Keywords: mycobiome, lower respiratory tract, Candida, cystic fibrosis, intensive care unit

### INTRODUCTION

The lower respiratory tract (LRT) includes airways distal of the vocal cords, i.e., trachea, bronchi, bronchioles and anatomical structures of the lung, i.e., respiratory bronchioles, alveolar ducts, alveolar sacs, and alveoli (Schönbach, 2013). Fungal diseases affecting the LRT comprise invasive fungal infections (IFI) and diseases triggered by fungal LRT colonization or inhalation (e.g., allergic bronchopulmonary aspergillosis). Some respiratory diseases are possibly associated with fungal colonization, i.e., exacerbation of chronic obstructive pulmonary disease (COPD) or deterioration of lung function and structure in cystic fibrosis (CF) patients (Nguyen et al., 2015). Candida is often

#### Edited by:

Suhelen Egan, University of New South Wales, Australia

#### Reviewed by:

Irene Wagner-Doebler, Helmholtz Centre for Infection Research (HZ), Germany Paolo Puccetti, University of Perugia, Italy

\*Correspondence: Robert Krause robert.krause@medunigraz.at

#### Specialty section:

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

Received: 21 July 2016 Accepted: 23 December 2016 Published: 10 January 2017

#### Citation:

Krause R, Moissl-Eichinger C, Halwachs B, Gorkiewicz G, Berg G, Valentin T, Prattes J, Högenauer C and Zollner-Schwetz I (2017) Mycobiome in the Lower Respiratory Tract – A Clinical Perspective. Front. Microbiol. 7:2169. doi: 10.3389/fmicb.2016.02169

cultured from LRT samples but the clinical significance is still unknown. Baum (1960) started a discussion regarding the significance of Candida in the LRT by presenting data from conventional microbiological investigations of LRT samples. He described that the source of Candida present in the sputum samples was the mouth and that little significance could be attached to the finding of Candida within the sputum in the diagnosis of pulmonary candidiasis. Since then subsequent culture based studies investigating Candida in the LRT have given conflicting results (Corley et al., 1997; el-Ebiary et al., 1997; Eggimann et al., 2003; Meersseman et al., 2009).

In general, cultivation-independent molecular assays are more efficient in describing microbial communities than culturebased approaches which fail to detect the whole diversity of a microbiota and, therefore, hinder complete understanding of the interactions between the host and the microbiome (Ghannoum et al., 2010; Bittinger et al., 2014; Nguyen et al., 2015). However, compared to bacterial microbiota investigation surveys on the mycobiota of the LRT are still rare. Performing a literature search via PubMed 86 hits are generated by using "mycobiome" as a search term, 9 using "lung mycobiome" and 10 using "respiratory mycobiome" (four of the publications appear also in "lung mycobiome" search<sup>1</sup> ; accessed November 21, 2016).

An overview of the current state of knowledge with respect to concepts of LRT mycobiota, the association of LRT mycobiota and fungal communities in other anatomical regions of the human body as well as the role of LRT mycobiota in some LRT diseases has been published recently (Nguyen et al., 2015). In comparison, the LRT mycobiota review presented here primarily addresses the characterization of investigated patient groups and healthy controls as well as utilized sampling techniques. From a clinical point of view, thorough selection of patient groups together with clear description of underlying diseases as well as antiinfective and non-antiinfective treatments that might influence bacterial and fungal microbiota are essential to detect relevant differences in fungal communities. From that informations it might be possible to draw clinically relevant conclusions and diagnostic as well as therapeutic implications. Beside differences in patient selection and molecular approaches, sampling techniques are also relevant for interpretation of results. In this review we therefore summarize recent literature addressing LRT mycobiome and its potential role in respiratory disease in certain patient groups and discuss the results from a clinical perspective including the potential influences by sampling techniques applied.

#### ORAL MYCOBIOME

Since LRT samples are commonly collected by procedures wherein the oral cavity is passed, the composition of the oral mycobiota is relevant to interpret LRT mycobiota studies. Although fungal colonization of the oral cavity is not imperatively connected to fungal colonization of other anatomical regions, e.g., the intestinal tract (Zollner-Schwetz et al., 2008), the contact of true LRT samples with saliva and/or mucosal surfaces occurring during expectoration of e.g., sputum samples might alter mycobiota composition in LRT samples. Oral mycobiota were investigated in healthy individuals to provide a basis for a detailed characterization of the oral mycobiome in health and disease. With results of that study the term mycobiome was introduced in medical literature as confirmed by our mycobiome literature search in PubMed starting with Ghannoum's paper published in 2010 (Ghannoum et al., 2010). In that study, oral rinse using 15 ml of phosphate buffered saline was obtained from 20 healthy subjects (21–60 years of age, eight females and 12 males; non-smoking, no recent antifungal use, and no clinical signs of oral mucosal disease). Exclusion criteria were a history of receiving medication or treatment with topical or systemic steroids, pregnancy, and a history of insulin-dependent diabetes mellitus. Internal transcribed spacer (ITS) based sequencing revealed that the oral cavity contained 74 cultivable (most abundant genera: Candida, Cladosporium, Aureobasidium, Saccharomycetales) and 11 non-cultivable fungal genera. The low-abundance genera were considered to represent environmental fungi that have been transferred into the oral cavity by inhalation of spores or by ingesting fungal contaminated food (Ghannoum et al., 2010). As only healthy individuals were investigated the influence of antibacterial and antifungal therapy and certain immunosuppressive conditions predisposing for fungal infections (e.g., underlying diseases like hematooncological malignancies, diabetes mellitus, intensive care unit treatment, immunosuppressive treatments like corticosteroids, cyclophosphamide, antineoplastic chemotherapy, biological therapy) on oral mycobiota remain unknown. It should be mentioned, however, that this study lacked appropriate contamination controls and extraction blanks, thus the possible introduction of fungal signatures by the used reagents cannot be excluded.

### MYCOBIOME OF CYSTIC FIBROSIS PATIENTS

In one of the first papers presenting LRT mycobiome data, eight sputum samples from four (two male and two female) adult CF patients (range 19–39 years) were investigated by pyrosequencing of ITS2 amplicons (Delhaes et al., 2012). For assessment of fungal and bacterial interaction 16S rRNA gene sequencing was included. Two temporal sputum samples were collected from each clinically stable patient with a sampling interval of 1 year, but in one patient the sampling interval was only 3 months. Samples were collected by expectoration into a sterile cup after a water rinse. Two patients received azithromycin due to Pseudomonas colonization and other unspecified antibiotic therapies (two courses in one patient and up to seven courses in the second patient previous to the first sputum sample). Three patients had inhalative and one patient had additional systemic corticosteroid therapy. One patient was treated with

<sup>1</sup>http://www.ncbi.nlm.nih.gov/pubmed/?term=mycobiome

itraconazole 600 mg for 183 days due to bronchopulmonary allergic aspergillosis prior to sampling. Results showed high molecular diversities for the main fungal and bacterial taxa and high intra-individual differences as mycobiota changed between the sampling time points (e.g., 202 Candida reads in sample 1 and 8688 Candida reads in sample 2 in patient 1). Anaerobes were isolated together with Pseudomonas aeruginosa, and the latter was found more likely co-associated with Candida albicans than with Aspergillus fumigatus. Aside of the bacterial findings the authors concluded that the diversity and species richness of fungal communities was significantly lower in patients with decreased lung function and poor clinical status. It was further mentioned that C. albicans and C. parapsilosis, as part of the oral mycobiome, could migrate from the oral niche, and colonize and persist within the lower airways of CF patients. However, since Candida is the major genus found in the oral cavity as shown in the oral mycobiota study investigating healthy subjects mentioned above and since oral mycobiota have not been investigated in this CF study these conclusions must be interpreted with caution. It is not clear from that CF study whether "sputum" samples represented adequate sputum produced from the LRT and how the oral mycobiota impacted the LRT sputum mycobiota. In addition, the present CF cohort was small (four patients), included CF patients with different treatments that might affect LRT mycobiota (corticosteroids, antibiotics, and antimycotics), and lacked a specific control group. The association of reduced diversity and richness of fungal and bacterial microbiota with poor clinical status and lung function (reduced FEV1 and FVC) was mainly derived from two patients with preceding antibiotic therapy and antifungal therapy. Therefore, influence of antiinfective therapy on reduced microbial diversity and richness cannot be ruled out and the association with lung function remains questionable.

Another study encompassed six CF patients, which were admitted for antibiotic therapy due to respiratory disease exacerbation and Pseudomonas sp. positive sputum cultures. ITS based mycobiota analysis was performed with sputum samples collected from this cohort prior to and after systemic antibiotic treatment (Willger et al., 2014). One patient received tobramycin and meropenem; one tobramycin and ceftazidime; one tobramycin, ceftazidime and linezolid; two tobramycin, doripenem and vancomycin; and one tobramycin and doripenem. Inhalative antibacterial therapy (as applied in other CF cohorts as described below) was not mentioned in the studied patients as well as information on preceding antibacterial therapy was lacking. As CF patients repeatedly receive antibacterial (and sometimes antifungal) therapy for infectious exacerbation of underlying disease previous therapies might have impacted the observed LRT mycobiota results. In mycobiota analysis Candida represented the dominant genus with 11 different species (top three were C. albicans, C. dubliniensis, and C. parapsilosis) and accounted for 74–99.9% of all ITS1 reads in all samples making these species both the most abundant and also the most prevalent. However, there was a broad distribution of C. albicans relative abundance ranging from 0.1% of reads (post-treatment sample from subject number 2) to 98.8% of reads (post-treatment sample from subject number 6). Besides C. albicans only Malassezia species occurred in all samples with 10- to 50-fold lower reads compared to Candida species. Also non-fumigatus Aspergillus species were detected at low levels. The relative abundance of Candida species was stable during antibiotic treatment and antibiotic treatment did not establish a specific fungal community. Fungal density varied after completion of antibiotic treatment with increase or decrease in individual patients but no common trend could be observed. Overall, no specific trend in fungal burden associated with antibacterial therapy could be detected. In contrast to decreased bacterial richness after antibacterial treatment no significant difference was determined for fungal richness although a trend for a decrease was observed. Candida and Malassezia species accounted for more than 80% of all assigned reads per sample and persisted throughout antibacterial treatment without significant change in relative abundance suggesting stability within the mycobiome. As with the other CF study, it is not clear if "sputum" samples represented adequate sputum produced from the LRT and if the oral mycobiota contaminated and impacted presumed LRT sputum mycobiota. For further clarification of the role of mycobiota in LRT the authors also suggested methods to differentiate between dead or alive microorganisms to assess active microbiota in the LRT.

In another study 72 sputum samples were collected from 56 CF patients recruited in an ambulatory center including a subcohort of 13 patients who provided sputum two (10 patients) or three (three patients) times within a 2-years sampling period (Kramer et al., 2015a). Sputum samples were collected in duplicate, stored at −20◦C and subsequently analyzed by ITS primer based PCR, preparation of single stranded DNA, single-strand conformation polymorphism electrophoresis and sequencing of individual bands. C. albicans and C. dubliniensis were the most dominant fungal species in the CF cohort, with 44.4% and 23.6% positive samples, respectively, followed by Saccharomyces cerevisiae, with 19.4%, and C. parapsilosis, with 13.8% positive samples. Several other fungi were detected at lower rates including Aspergillus sp., and Cladosporium sp., Scedosporium and Exophiala sp. In patients investigated at two time points different fungi were detected in the same patient indicating dynamic mycobiota composition. In concordance with the other CF study mentioned above it is not clear from this particular CF study if "sputum" samples represented adequate sputum produced from the LRT and how oral mycobiota impacted presumed LRT sputum mycobiota (as sputum passes the oral cavity). In another study published by the same group the antibiotic therapy of the same CF cohort was described in more detail. Investigated CF patients were under continuous treatment of antibiotics by inhalation applied twice a day at home (mostly colistin or tobramycin according to clinical needs). Interestingly, none of the patients received antibiotics intravenously during the 2-years time frame of study (Kramer et al., 2015b). Information on antimycotic therapy during or prior to sampling was not provided. Even if the CF cohort

represented routine clinical cases conclusions from mycobiota (and microbiota) analysis can hardly be drawn based on the missing information regarding antifungal therapy and the incomplete description of inhalative antibiotic therapy (how many patients received which inhalative antibacterial medication?).

### MYCOBIOME OF PATIENTS WITH ASTHMA BRONCHIALE

Lower respiratory tract mycobiota were investigated in 30 asthma patients and compared to 13 controls (van Woerden et al., 2011, 2013). Patients were free of oral steroid use ≥2 weeks prior to induced sputum sampling but most of them were on inhaled corticosteroids. No information regarding previous or current antibiotic or antifungal therapy was provided. After nebulizing isotonic saline induced sputum was coughed into petri dishes and spread on microscopic slides for microscopic examination. Afterward a 5 mm<sup>2</sup> area was subsequently excised from the slide for DNA extraction and PCR. The excised samples from each study group were combined to yield two pooled samples, one from all asthma patient sputum samples and one from all healthy control subject sputum samples. Unfortunately, as described by the authors, a sample from one asthma patient was inadvertently included in the control set. Primer pairs Euk1a and Euk516r were used for PCR, sequenced using a 454 pyrosequencer and DNA sequences compared to SILVA database (van Woerden et al., 2013). In total 136 fungal species were identified, with Psathyrella and Malassezia spp. representing each approximately 25% of all reads in asthma patients and Eremothecium spp. representing approximately 40% of all reads in healthy controls. As the authors noted in their discussion pooling sputum samples prevented clear interpretation of results as fungi identified could origin from one individual heavily colonized or from several individuals colonized with identical or nearly similar fungal species. In addition, oral mycobiota as well as material used for sampling (petri dishes, slides, and saline) and its impact on mycobiota results by potential contamination was not investigated. Considering fungal species found in asthma patients the pathophysiological role of Psathyrella candolleana, a mushroom found on lawns or pastures in Europe and North America, is unknown as well as the role of Eremothecium sinecaudum, a yeast with pathogenic properties for plants, in healthy subjects. The duration for collection of pooled sputum and time intervals between sample collection and storage was not provided as well as storing conditions. It has been shown that prolonged time between sample collection and sample storage (more than 12 h) as well as more than four freeze thaw cycles distorted microbiota profiles (Cuthbertson et al., 2014, 2015). As mentioned above, fungi found in sputum from asthma patients and healthy controls mainly contained environmental fungi present in plants and grassland (van Woerden et al., 2013). Thus contamination of the pooled sputum from both study groups during sampling and storage cannot be ruled out. As sputum samples were mixed up from individual patients and the pooled control sample included one asthma sample interpretation of results and considerations of differences between study populations is very limited.

#### MYCOBIOME OF PATIENTS WITH LUNG TRANSPLANTS

Mycobiota based on ITS-sequencing of 21 lung transplant patients were investigated in BAL and oropharyngeal washes and compared to healthy controls, the latter undergoing a two-bronchoscope sampling technique or conventional bronchoscopy (number of healthy controls undergoing each of the sampling techniques was not provided in the text; in figures within the manuscript data from six healthy controls were shown). Two patients undergoing routine bronchoscopy for evaluation of sarcoidosis and a lung nodule representing adenocarcinoma (but both without lung transplantation) were also included, but they were not further characterized (Charlson et al., 2012). All lung transplant patients had immunosuppressive and antibiotic therapy, three received voriconazole and six nystatin (one patient had both antifungals). Prior to bronchoscopy instruments were flushed with 10 ml saline to investigate microbial DNA present in bronchoscopes and saline. Fungal populations in lung transplant patients were typically dominated by Candida in oral washes and BAL and by Aspergillus in BAL but not in oral washes. Oral washes of lung transplant patients revealed 13 samples with high fungal sequences, all but one were dominated by Candida. Patients without Candida in BAL had no (six patients) or low (two patients) Candida sequences in oral washes. As eight lung transplant patients had nystatin and/or voriconazole therapy the presence or absence of Candida and other fungi (targeted by nystatin and voriconazole) in oral washes and BAL is hard to assess. Nystatin did not consistently decrease or eliminate Candida from oral washes but patients on voriconazole had no Candida sequences in BAL (patient no. 34 was only presented in Figure E2, supplementary material of the article; Charlson et al., 2012). However, in two of voriconazole treated patients Aspergillus sequences were detected (patient no. 34, also with positive Aspergillus culture, and patient no. 43). One patient (no. 36) had Aspergillus in culture and in mycobiota analysis but in a second BAL 2 months later Aspergillus was no longer detected by both methods while the patient was treated with voriconazole; voriconazole obviously eliminated Aspergillus in that particular patient. Candida was absent in BAL of healthy volunteers although Candida was present in oral washes. Unfortunately no information regarding a potential pathogenetic role of fungi detected within mycobiota was provided for the presented lung transplant patients. The presence or absence of invasive fungal infection or fungal breakthrough infection in voriconazole treated patients (based on clinical data, radiological examination, biomarkers like galactomannan or 1,3-beta-Dglucan testing, voriconazole trough level) and outcome was not mentioned.

### MYCOBIOTA IN MISCELLANEOUS PULMONARY DISEASES AND HIV POSITIVE PATIENTS

In a study published by the same authors as mentioned above in the lung transplant patients, data from an almost similar study cohort included mycobiota from 42 lung transplant patients, 19 HIV positive patients, 13 patients with various pulmonary diseases and 12 healthy controls (Bittinger et al., 2014). Four control patients were counted twice as they were investigated by a double bronchoscopy and single bronchoscopy technique at least 1 year apart. Healthy and HIV positive subjects were described not to have respiratory symptoms and antibiotic therapy. However, in the **Table 1** one HIV patient did in fact have antibiotic therapy (clindamycin within 3 months prior to sampling) and another one had inhalative corticosteroid treatment. Subjects within the healthy and HIV positive study groups where smokers and non-smokers. In patients with various pulmonary diseases two patients had immunosuppressive therapy, two had systemic antibiotic therapy but two patients with pneumonia surprisingly had no antibiotic therapy. Two out of 24 lung transplant patients refused collection of clinical data. Thirteen of the remaining 22 lung transplant patients had antifungal therapy (11 nystatin, two voriconazole treatments). Twenty received antibiotic therapy (mainly trimethoprim and sulfamethoxazole), two had atovaquone therapy and all received immunosuppressants (presented in Supplementary Table S3 of Bittinger et al., 2014). The authors also focused on analysis of possible confounding factors as 132 samples from laboratory water, bronchoscope canal flushes, saline used for oral washes and BAL, sterile swabs and lab tabletop surface were investigated [supplemental report in (Bittinger et al., 2014)]. Interestingly, the composition of fungi in BAL samples was generally similar to that in contamination controls. After correction of abundance by subtraction of reads found in the control samples (e.g., saline) Candida was mainly found in BAL and oral washes of the study groups (Figure 37–43 in supplemental report of Bittinger et al., 2014) as well as Aspergillus and Cladosporium in lung transplant patients and those with miscellaneous pulmonary diseases (Figure 37 and 38 in supplementary report of Bittinger et al., 2014). Clinical significance of fungal detection in presented patient groups is unclear from that study as clinical data including radiological examination, biomarkers, introduction of antifungal treatment and outcome have not been reported.

Another study provided mycobiota data from 32 HIV positive (10 with COPD) and 24 HIV-uninfected patients extracted from a cohort comprising 396 HIV-infected and HIV-uninfected individuals with smokers and marijuana abusers in both study groups (Cui et al., 2015). Prior to sampling study subjects had stable respiratory symptoms in previous 4 weeks and did not receive antibiotic or immunosuppressive therapy in previous 6 months. Oral wash, induced sputum (<30% squamous cell count) and BAL passing the oral cavity were collected


as well as control samples by aspirating saline through the bronchoscope channel. Fungal microbiota were analyzed by sequencing amplicons generated by PCR using ITS- and 18Stargeting primers. 18S data were used for operational taxonomic unit (OTU)-type analysis and ITS data were used for specieslevel analysis. The authors mentioned that the 18S rRNA gene gave higher PCR-positive rates, but fewer fungal reads compared with the ITS, because the 18S primers cross-reacted with host (=human); consequently, approximately 90% of the raw 18S sequencing reads were human. By using 18S data for comparison of different sampling techniques, the authors interestingly found that induced sputum shared little similarity with BAL. In principle coordinate analysis of 18S data of healthy subjects (HIV negative, normal lung function) using only 30% of all data, BAL samples clustered together with oral wash, induced sputum overlapped in part with oral wash samples but not with BAL. A similar distribution pattern was seen with the entire cohort including HIV-infected and HIV-uninfected individuals with or without normal lung function. Candida represented more than 90% of the ITS reads and was clearly dominant in oral washes compared with BAL. There was also a higher prevalence of Candida in induced sputum compared with BAL. Ceriporia lacerata, Saccharomyces cerevisiae, and Penicillium brevicompactum were predominant in BAL compared to oral wash. None of these species were detected in the corresponding bronchoscopic control samples, indicating that environmental contamination was unlikely. In HIV infected individuals nine species including Pneumocystis jirovecii (confirmed by nested PCR) were overrepresented in BAL compared to HIV-uninfected individuals. No species was associated with COPD in the HIVinfected cohort as compared with HIV-infected individuals with normal lung function. In marijuana smokers, with and without HIV infection, four species (including two that are known plant pathogens, Phialocephala virens and Taphrina tormentillae) were overrepresented as compared with individuals who had not smoked marijuana in the year prior to sampling. The authors mentioned that smoking marijuana without filters might have contributed to inhalation of this certain environmental/plant fungi. Considering all subjects investigated, fungal communities in the LRT resembled those in the mouth. This result could be caused by microaspiration of oral content, contamination during bronchoscopy when passing the oral cavity, contamination of induced sputum with oral microbiota or contamination of oral microbiota by coughing or exhalation. The latter is unlikely as oral wash was collected prior to sputum sampling or bronchoscopy.

#### MYCOBIOME IN LRT OF ICU PATIENTS

Mycobiota in LRT of ICU patients were investigated in two studies. In the first study 18S rRNA gene sequencing was performed to analyze the etiology of pneumonia in ICU patients with community acquired pneumonia (CAP, 32 episodes), ventilator-associated pneumonia (VAP, 106 episodes), nonventilator ICU pneumonia (NV-ICU P, 22 episodes), aspiration pneumonia (AP, 25 episodes) and compared to 25 ICU patients without pneumonia (CS; Bousbia et al., 2012). As pneumonia cases were counted as episodes (185 pneumonia episodes) the total number of patients included in each of the pneumonia subgroups is not clear. However, in some tables the number of pneumonia patients were 185 enabling the conclusion that one pneumonia episode obviously corresponded to one patient. Clinical data regarding underlying diseases beside pneumonia, (previous) antibiotic and antimycotic therapy or initiation of new antiinfective therapy, presence or absence of tracheal tubes, invasive or non-invasive mechanical ventilation and underlying diseases were not provided. Detailed sampling procedures, e.g., time of BAL sampling after admission or diagnosis or pneumonia, BAL by bronchoscopy passing the oral cavity or through oropharyngeal ventilation tubes, transport and storage of BAL cultures, investigation of material used for BAL, investigation of pre-BAL bronchoscopy flushes were not mentioned. Interestingly, ICU patients used as control patients (=no pneumonia) also had ARDS but with comparable or even higher numbers compared to some pneumonia patients (38% in CAP, 33% in VAP, 41% in NV-ICU P, 12% in AP; 31% in all pneumonia patients and 28% in patients without pneumonia considered as controls). The proportion of patients with immunosuppressive therapy in each pneumonia group was reported but the type of immunosuppressive therapy was not provided (38% in CAP, 40% in VAP, 50% in NV-ICU P, 36% in AP; 40% in all pneumonia patients and 16% in controls). C. albicans signatures were most abundant in all patients groups and controls. C. utilisshowed higher abundance in controls compared to pneumonia subgroups (p = 0.01; Supplementary Table S5 in supplementary material of Bousbia et al., 2012). Aspergillus, Penicillium and Cladophialophora genera were dominant in the CAP cohort but did not reach significant differences compared to other cohorts. Tremellomycetes (represented by Cryptococcus) was identified in NV-ICU P, whereas Agaricomycetes and unclassified Ascomycota were only identified in VAP patients. Sordariomycetes were only identified in controls. However, maybe based on small numbers, no significant difference could be calculated for these classes. Beside bacterial and fungal microbiome analysis the authors included a lot of other microbiological methods to detect microorganisms causing pneumonia. For many of detected microorganisms including those found in mycobiota analysis the authors raised the question on the actual role of these microorganisms in pneumonia. The authors concluded that the respiratory microbiota was more complex than expected (Bousbia et al., 2012).

In another study, LRT mycobiota in non-neutropenic intubated and mechanically ventilated ICU patients with antibiotic therapy to treat pneumonia were investigated and compared to several other groups: healthy controls without preceding antibiotic therapy, patients without pulmonary diseases with antibiotic therapy, non-neutropenic intubated and mechanically ventilated ICU patients without preceding or current antibiotic therapy, non-neutropenic intubated and mechanically ventilated ICU patients without pulmonary disease but with antibiotic therapy for treatment of extrapulmonary infection. Smoking behavior of study patients was not evaluated. All study subjects were without antimycotic treatment ≥8 weeks

prior to sampling (Krause et al., 2016). LRT in all patients and healthy study subjects were sampled through endotracheal tubes avoiding contact to the oral cavity. Non-ICU patients and healthy controls were sampled by aspiration of endobronchial secretion and ICU patients were sampled by bronchoscopic guided BAL. Saline used for BAL and collected after flushing the sampling channel of the bronchoscope immediately before bronchoscopy was also investigated. Sampling techniques (endobronchial secretion versus BAL) showed similar results and saline did not influence mycobiota results. The results showed that Candida was part of the fungal microbiota of various intubated and mechanically ventilated ICU patients with and without antibiotic therapy and with and without pneumonia (59–73% Candida signatures within 5–8 fungal genera; Krause et al., 2016). Candida sequences were not present in mycobiota of non-ICU patients with antibiotic therapy and healthy controls. Admission to and treatment on ICUs shifted LRT fungal microbiota to Candida spp. dominated fungal profiles but antibiotic therapy did not. Mycobiota analysis was not done with samples from the oral cavity but in conventional cultures antibiotic therapy increased the prevalence of Candida spp. in this compartment. Conventional cultures also showed that admission to the ICU increased the prevalence of Candida spp. in the LRT, but other fungi were not detected by culture. None of the ICU patients had candidemia or other invasive fungal disease as determined by clinical, imaging, laboratory and microbiological investigations.

#### DISCUSSION

The mycobiota in the LRT is diverse and varies in different populations and LRT diseases. Links to certain stages of diseases or LRT dysfunction have been proposed. In one study the diversity and species richness of fungal and bacterial communities were significantly lower in patients with decreased lung function as assessed by forced expiratory pressure in 1 s in lung function tests (Delhaes et al., 2012). However, the association was mainly derived from two patients with preceding antibiotic therapy and antifungal therapy and the contribution of fungal or bacterial microbiota to lung function could not be distinguished. Despite variable fungal composition in LRT, Candida overall dominated mycobiota in most of the studies including CF patients, lung transplant patients, HIV infected patients and ICU patients (Bousbia et al., 2012; Charlson et al., 2012; Bittinger et al., 2014; Willger et al., 2014; Cui et al., 2015; Kramer et al., 2015a; Krause et al., 2016). In remaining studies with dominating fungi other than Candida methodologies were too weak to draw profound conclusions. However, Candida was also the most abundant genus in oral mycobiome analysis (Ghannoum et al., 2010) questioning the results of studies that presented sputum samples without quality control (e.g., <10 squamous cells in high power field sputum examination). Unfortunately, even in bronchoscopy guided BAL contact to oral mucosa or saliva during insertion of the bronchoscope might alter microbiota composition of BAL obtained through orally contaminated bronchoscopy tips. But LRT mycobiota similar or close to oral mycobiota does not necessarily represent contamination of bronchoscopes or translocation of mycobiota by bronchoscopy, as upper and LRT might be colonized by identical or closely related microorganisms. Similar or close mycobiota in upper and LRT analysis could also be caused by microaspiration of oral content, contamination of induced sputum with oral microbiota or contamination of oral microbiota by coughing or exhalation. To overcome limitations in interpreting mycobiota data sampling techniques bypassing the oral cavity (e.g., via endotracheal tubes) or two-bronchoscopy sampling techniques should therefore be preferred to obtain true LRT samples. By applying such techniques and therefore exclusion of contamination, mycobiota in upper and LRT could be assessed more reliable.

In two previous studies reagents (e.g., saline, lab water, lab surfaces) have been analyzed to assess its influence on microbiota results from BAL (Bittinger et al., 2014; Krause et al., 2016); however, in other studies this was not done (Erb-Downward et al., 2011; van Woerden et al., 2011, 2013). Nevertheless, none of the studies investigating reagents showed that Candida abundance reached >50% as has been found in BAL samples from ICU patients (Krause et al., 2016). The reason for the elevated Candida abundance in LRT of ICU patients is not clear. The role of ICU environment or translocation of microorganisms from intestinal mycobiota (via ICU personnel or retrograde translocation up to the oral cavity followed by microaspiration passing the cuff or the orotracheal tube) in elevated LRT Candida abundance or in shifts of LRT mycobiota has not been investigated yet.

Three studies included sequential sampling for assessment of dynamics in mycobiota composition over time. Observation periods varied between days (ICU patients; Krause et al., 2016), weeks (CF patients; Willger et al., 2014), 3 months to 1 year (CF patients; Delhaes et al., 2012), and months up to 2 years (CF patients; Kramer et al., 2015a). Conclusions are difficult to draw based on low numbers of individuals investigated, different patient groups and differences in time frames elapsing between sampling. Whereas high intraindividual differences with changes in Candida abundance were found in one CF study (Delhaes et al., 2012) another CF study found stable results (Willger et al., 2014). The third CF study found differences in number of fungal genera in some patients providing two sputum samples (Kramer et al., 2015a). In ICU patients fungal microbiota were stable in 3 of 4 patients as shown by analysis of first and second BAL sample 4 to 7 days apart (Krause et al., 2016). Two studies compared different LRT sampling techniques in individual patients, i.e., induced sputum and BAL in one study (Cui et al., 2015) and endotracheal aspiration and BAL in another study (Krause et al., 2016). Whereas differences were detected between sputum and BAL based on 18S rRNA sequencing no difference was found between endotracheal aspirate and BAL based on ITS sequencing. As the authors mentioned in 18S based sequencing human 18S rRNA reads reduced the sequencing depth of fungal 18S rRNA gene reads, which might have influenced comparative data analysis of sputum and BAL. Although some studies included conventional cultures for assessment of microbial composition all studies mentioned above did not account for viability of microbiota explored by sequencing methods. Therefore it is almost unknown if detected mycobiota represent true alive residential or inhaled dead microorganisms or both. Future studies need to include this

aspect by e.g., dividing samples and treating of one half with propidium monoazide (PMA) to select for dead cells (Nocker et al., 2007).

The role of Candida in LRT in above mentioned diseases and conditions is still largely unknown. Patient characteristics provided in some mycobiota studies prohibit clinical conclusions as relevant data regarding underlying diseases, presence or absence of LRT infectious disease, initiation of antibiotic or antimycotic therapy, and outcome (e.g., invasive fungal infection) are missing. In the study investigating asthma bronchiale patients no information regarding previous or current antibiotic or antifungal therapy was provided (van Woerden et al., 2013). As shown in one of the studies in ICU patients (Krause et al., 2016) antibiotic therapy surprisingly did not influence mycobiota. Interestingly, antibiotic therapy in CF patients overall also had no impact on mycobiota composition and fungal burden although fungal burden varied between individual patients with decreasing and increasing levels (Willger et al., 2014). The influence of antimycotic therapy on LRT mycobiota is still unclear but some antimycotics might alter LRT mycobiota as detected in one patient in the lung transplant patient group discussed above (Charlson et al., 2012).

#### CONCLUSION

Although some studies suggest renewal of our pathophysiological understanding of mycobiota in LRT, shortcomings in clinically relevant methodologies in some studies call for further investigations to compensate previous deficiencies. Improved study settings will allow to draw better clinically relevant conclusions regarding diagnostic or therapeutic procedures in certain LRT diseases. As previously suggested, future mycobiota

#### REFERENCES


(and bacterial microbiota) studies should enable to answer one or more of the following questions: Can experiments detect differences that matter (in the context of disease and treatment)? Does the study show causation or just correlation? What is the mechanism? How much do experiments reflect reality? Could anything else explain the results? (Hanage, 2014). To obtain clinically relevant answers efforts should be enhanced to collect well characterized and described patient groups as well as healthy individuals for comparative data analysis and to apply thorough sampling techniques. For clinical routine we learn from current stage of knowledge that conventional microbiology obviously does not reflect true composition of LRT mycobiome. However, therapeutic studies based on new sequencing techniques are missing. We need to proceed with elucidation of the role of mycobiota in healthy LRT and LRT diseases to hopefully improve patient care. This might include diagnostic procedures with cultivation-independent molecular assays and prophylactic as well as therapeutic interventions in patients potentially suffering from mycobiota alterations (e.g., CF patients, lung transplant patients, ICU patients).

#### AUTHOR CONTRIBUTIONS

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

#### ACKNOWLEDGMENTS

This work was supported by the Austrian Science Fund, grant number P 23037-B18 and KLI 561-B31.

respiratory infections. J. Clin. Microbiol. 52, 3011–3016. doi: 10.1128/JCM. 00764-14



**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 © 2017 Krause, Moissl-Eichinger, Halwachs, Gorkiewicz, Berg, Valentin, Prattes, Högenauer and Zollner-Schwetz. 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) or licensor 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.

# Current Knowledge and Future Research Directions on Fecal Bacterial Patterns and Their Association with Asthma

Shantelle Claassen-Weitz <sup>1</sup> , Charles S. Wiysonge2, 3, Shingai Machingaidze<sup>4</sup> , Lehana Thabane5, 6, William G. C. Horsnell 7, 8, 9, Heather J. Zar 10, 11, 12, Mark P. Nicol 1, 8, 13 and Mamadou Kaba1, 8 \*

*<sup>1</sup> Division of Medical Microbiology, Department of Pathology, Faculty of Health Sciences, University of Cape Town, Cape Town, South Africa, <sup>2</sup> Centre for Evidence-based Health Care, Faculty of Medicine and Health Sciences, Stellenbosch University, Cape Town, South Africa, <sup>3</sup> Cochrane South Africa, South African Medical Research Council, Cape Town, South Africa, <sup>4</sup> Vaccines for Africa Initiative, Institute of Infectious Disease and Molecular Medicine, University of Cape Town, Cape Town, South Africa, <sup>5</sup> Department of Clinical Epidemiology and Biostatistics, McMaster University, Ontario, Canada, <sup>6</sup> Biostatistics Unit, Father Sean O'SulliVan Research Centre, Ontario, Canada, <sup>7</sup> Division of Immunology, Department of Pathology, Faculty of Health Sciences, University of Cape Town, Cape Town, South Africa, <sup>8</sup> Institute of Infectious Diseases and Molecular Medicine, University of Cape Town, Cape Town, South Africa, <sup>9</sup> International Centre for Genetic Engineering and Biotechnology, University of Cape Town, Cape Town, South Africa, <sup>10</sup> Department of Paediatrics and Child Health, University of Cape Town, Cape Town, South Africa, <sup>11</sup> Red Cross War Memorial Children's Hospital, Cape Town, South Africa, <sup>12</sup> Medical Research Council Unit on Child and Adolescent Health, University of Cape Town, Cape Town, South Africa, <sup>13</sup> National Health Laboratory Service, Groote Schuur Hospital, Cape Town, South Africa*

#### Edited by:

*Christine Moissl-Eichinger, Medical University of Graz, Austria*

#### Reviewed by:

*Geanncarlo Lugo-Villarino, Institut de Pharmacologie et de Biologie Structurale/Centre National de la Recherche Scientifique, France Jan S. Suchodolski, Texas A&M University, USA*

#### \*Correspondence:

*Mamadou Kaba mamadou.kaba@hotmail.com*

#### Specialty section:

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

Received: *15 January 2016* Accepted: *18 May 2016* Published: *29 June 2016*

#### Citation:

*Claassen-Weitz S, Wiysonge CS, Machingaidze S, Thabane L, Horsnell WGC, Zar HJ, Nicol MP and Kaba M (2016) Current Knowledge and Future Research Directions on Fecal Bacterial Patterns and Their Association with Asthma. Front. Microbiol. 7:838. doi: 10.3389/fmicb.2016.00838* Keywords: asthma, fecal bacteria, mechanisms, microbiome, systematic review

### INTRODUCTION

Asthma is a complex respiratory condition that involves interplay between genetic predisposition, environmental, and immunological factors (Edwards et al., 2012). It is considered to be one of the most common chronic diseases, affecting ∼300 million people (Masoli et al., 2004), and causing an estimated 250,000 deaths annually (Bateman et al., 2008). Furthermore, because of an increased Westernized lifestyle and urbanization in developing countries, it is estimated that by 2025 the global burden of asthma will increase by 100 million people (Masoli et al., 2004).

An increase in the occurrence of allergic diseases, including asthma, was initially attributed to the "hygiene hypothesis," suggesting that a reduced exposure to microbes during the first years of life plays a role in the development of allergic diseases (Strachan, 1989, 2000). Although this hypothesis is widely accepted, studies showed that reduced microbial exposure cannot fully account for the increased prevalence of asthma, rhinitis, or neurodermitis (Mallol, 2008; Brooks et al., 2013; Kramer et al., 2013). Alternative hypotheses or reformulations of the "hygiene hypothesis" (Hunter, 2012), such as the "microbiota hypothesis" (Wold, 1998), the "old friends hypothesis" (Rook, 2012), the "microbial deprivation hypothesis" (Bloomfield et al., 2006), the "biodiversity hypothesis" (Hanski et al., 2012) and the "disappearing microbiota hypothesis" (Blaser and Falkow, 2009; Taube and Müller, 2012) soon followed; all of which mainly postulate that dysbiosis of the human gastro-intestinal tract (GIT) microbiome may contribute to intra- and extra-intestinal immune-mediated diseases (Penders et al., 2007; Štšepetova et al., 2007; Sekirov et al., 2010; Clemente et al., 2012; Russell and Finlay, 2012). Understanding which bacteria from our GITs contribute to the development or prevention of allergic asthma may result in further research to discern the mechanisms behind bacterial-host interactions and potentially facilitate treatment strategies.

### METHODS USED TO STUDY THE ROLE OF HUMAN FECAL BACTERIA IN ASTHMA

Although culture-independent techniques have revolutionized the world of microbiology (Suau et al., 1999; Zoetendal et al., 2006; Rajilic-Stojanovi ´ c et al., 2007 ´ ); conventional culturedependent techniques have been the method widely used to study the role of human fecal bacteria in asthma (Mansson and Colldahl, 1965; Stockert, 2001; Nambu et al., 2008; Vael et al., 2008; Bisgaard et al., 2011). To date, the culture-independent techniques used to characterize fecal bacteria from patients with asthma include quantitative real-time polymerase chain reaction (qPCR) (Van Nimwegen et al., 2011), denaturing gradient gel electrophoresis (DGGE) (Bisgaard et al., 2011; Vael et al., 2011), fluorescent in situ hybridization (FISH) (Salminen et al., 2004), and massively parallel high-throughput sequencing of the 16S ribosomal RNA (rRNA) gene (Arrieta et al., 2015). Despite the advantage of detecting uncultivable bacteria, these cultureindependent techniques are not without limitations. Among others, they do not allow for whole community analysis of the microbial population (Sekirov et al., 2010; Fraher et al., 2012; Sankar et al., 2015), which is considered key in determining the patterns of fecal bacteria associated with health and disease states (Schippa and Conte, 2014). For example, qPCR and FISH do not provide identification of novel organisms as they are used to characterize and quantify targeted groups of bacteria (Sekirov et al., 2010; Fraher et al., 2012). DGGE, a band-based method for determining bacterial diversity, does not enable direct identification of bacteria (Sekirov et al., 2010; Fraher et al., 2012). Furthermore, DGGE has low bacterial detection limits and limited phylogenetic resolution (Sekirov et al., 2010). Although massively parallel high-throughput sequencing of the 16S rRNA gene provides an almost comprehensive view of bacterial communities; it does not provide classification at species-level (Gosalbes et al., 2012; Arrieta et al., 2015). The importance of species-level characterization in health and disease states has been demonstrated in murine models of allergic diseases (Karimi et al., 2009; Russell et al., 2012; Kim et al., 2013). An overgrowth of the genus Lactobacillus has been associated with an increased risk of allergic asthma (Russell et al., 2012), while the species L. reuteri and L. rhamnosus provide a protective role in allergic airway disease (Karimi et al., 2009; Kim et al., 2013). In comparison to 16S rRNA gene sequencing techniques; whole genome shotgun (WGS) sequencing offers a higher and more reliable resolution of microbiota profiles at lower taxonomic levels (Morgan and Huttenhower, 2014; Van Dijk et al., 2014; Ranjan et al., 2015). For example, WGS sequencing is able to improve the issue related to the Bifidobacterium amplification bias by certain primer sets (Kurokawa et al., 2007; Sim et al., 2012; Walker et al., 2015). In addition, it allows for determination of the metabolic and functional properties of fecal bacteria which may greatly contribute to our understanding of the role of fecal bacteria in health and disease (Qin et al., 2012; Arrieta et al., 2015; Quince et al., 2015). However, despite the number of advantages that WGS sequencing provides, it has not been incorporated by any of the studies investigating the importance of fecal bacteria in the development of asthma. Furthermore, a causal link between fecal bacterial profiles and asthma in humans has recently been confirmed using murine models (Arrieta et al., 2015). To the best of our knowledge, this is the only report of its kind where the causal role of the fecal bacteria (Lachnospira, Veillonella, Faecalibacterium, and Rothia), which potentially confers protection against the development of asthma in humans, was demonstrated using germ-free mice models (Arrieta et al., 2015).

#### WHAT DO STUDIES IN HUMANS REVEAL ABOUT THE ROLE OF FECAL BACTERIA IN ASTHMA?

Although prospective longitudinal studies are key in demonstrating the role of fecal bacteria in disease development (Zhao, 2013); we identified only two studies which assessed whether fecal bacterial profiles sampled over time preceded the occurrence of asthma at later stages in life (Bisgaard et al., 2011; Arrieta et al., 2015). Bisgaard et al. (2011) did not report a significant association (Bisgaard et al., 2011). In contrast, Arrieta et al. (2015) found significantly reduced abundances of the bacterial genera Lachnospira, Veillonella, Rothia, and Faecalibacterium in infants at risk for asthma, as evidenced using the Asthma Predictive Index (API) (Arrieta et al., 2015). Moreover, in a prospective birth cohort study conducted in Belgium (using fecal specimens sampled at 3 weeks of age); the detection of Bacteroides (B. fragilis, B. finegoldii, and B. thetaiotaomicron), Ruminococcus (R. productus and R. hansenii), and Clostridium spp. was associated with an increased risk for asthma development (as based on the API) (Vael et al., 2011). At species-level, the prospective birth cohort study by Van Nimwegen et al. (2011) conducted in the Netherlands reported a two-fold increased risk of asthma at 6–7 years in infants colonized with Clostridium difficile at 1 month of life (OR = 2.06; 95% CI 1.16–3.64) (Van Nimwegen et al., 2011).

All prospective birth cohort studies, except for the study by Nambu et al. (2008), made use of the API when assessing asthma as an outcome at <5 years of age (Vael et al., 2008, 2011; Arrieta et al., 2015). The API, incorporated by three studies cited in this review, is an example of a predictive assessment for asthma development later in life, recommended for young children experiencing recurrent wheeze (Castro-Rodriguez, 2010). Considering that asthma diagnosis in children <5 years of age is challenging and often based on symptom patterns, clinical assessment of the family history and the presence of atopy (Pedersen, 2007; Sly et al., 2008; Pedersen et al., 2011); the use of predictive assessments, such as the API, is essential. However, despite its success in developed countries, the API should be used with caution in infants from low and middle income countries (LMICs) (Zar and Levin, 2012). This may be explained by the fact that young children from LMICs are more commonly affected by viral lower respiratory tract infections (LRTIs) or pulmonary tuberculosis. Furthermore, it has been suggested that atopy may be less strongly associated with asthma in these settings compared to the more developed countries (Zar and Levin, 2012). This suggests that non-atopic wheeze may be Claassen-Weitz et al. Fecal Bacteria and Asthma

the primary form of asthma in these children, making the API, which relies primarily on the presence of atopy for assessing the risk of asthma, a less reliable predictive assessment tool in LMICs (Zar and Levin, 2012).

### FACTORS INFLUENCING FECAL BACTERIAL PROFILES AND POTENTIALLY ASTHMA

Both murine models and human studies have provided evidence that early life changes in the GIT microbiome are most influential in the development of allergic asthma (Russell et al., 2013; Arrieta et al., 2015). Some of the well described factors responsible for these early life changes in fecal bacterial profiles, which have also been associated with childhood asthma, are mode of delivery, feeding practices, and antibiotic use (Kozyrskyj et al., 2011).

### Mode of Delivery

A number of childhood studies have reported that infants delivered via cesarean section are at an increased risk for the development of asthma (Thavagnanam et al., 2008). However, these studies do not account for confounding factors that may be associated independently with asthma, as well as changes in fecal bacterial profiles, which will allow for determining the true effect of external factors on fecal microbiota and the resultant health outcome. To date, only a single study using mediation analysis (Van Nimwegen et al., 2011) supported the role of mode and place of delivery (independent variable) in C. difficile colonization (mediator variable), together with its consequent impact on asthma development via modulation of C. difficile profiles (dependent variable).

### Feeding Practices

Although, it has been reported that breastfeeding has the potential to protect against allergic airway disease (Dogaru et al., 2014), no studies have used mediation analysis (as performed by Van Nimwegen et al. (2011)) to determine whether bacteria from breast milk protect against asthma via the modulation of infant fecal bacteria.

## Antibiotic Use

In humans, a modest increased risk of asthma development, associated with antibiotic use, has been reported (Marra et al., 2009; Risnes et al., 2010; Murk et al., 2011; Penders et al., 2011). To date, only fecal C. difficile colonization has been associated with the occurrence of asthma (Van Nimwegen et al., 2011), which might be explained (among other factors) by a loss of intestinal commensal microbes through the use of antibiotics (Azad and Kozyrskyj, 2012).

### POTENTIAL MECHANISMS SUPPORTING THE ROLE OF GASTROINTESTINAL BACTERIA IN ASTHMA

The exact mechanisms by which GIT bacteria may influence the development of respiratory diseases are unclear; however recent work has demonstrated that crosstalk between host mucosal immune cells and resident microbes significantly influences the risk for respiratory disease (Forsythe, 2011; Samuelson et al., 2015; Vital et al., 2015). A central player in this regulation of pulmonary immunity by the GIT microbiome are dendritic cells (DCs) (McLoughlin and Mills, 2011) (**Figure 1**). Intestinal DCs encounter bacterial antigens presented in organized GIT immune tissue (i.e., lamina propria and Peyer's patches) and also directly sample lumen residing bacteria in the GIT by extending their dendrites into the intestinal lumen (Salzman, 2011). This sampling of intestinal bacterial antigens results in DCs co-ordinating B and T cell subset expansion both locally (Peyer's patches and lamina propria) as well as systemically (e.g mesenteric lymph nodes) (Hill and Artis, 2010) (**Figure 1**). This results in DC-guided local and systemic immune education driven by microbiota associated antigens which has profound effects not just in the intestine but throughout the body (Hill and Artis, 2010; Russell and Finlay, 2012) (**Figure 1**). An important consequence of this effect of GIT bacteria is manifested in subsequent host T-cell immune responses in the lungs and has been particularly well demonstrated in murine models of asthma (Herbst et al., 2011; Navarro et al., 2011; Konieczna et al., 2012; Oertli et al., 2012). For example, B. fragilis and Clostridium species (cluster IV and XIVa), both intestinally restricted bacteria, can drive induction of T regulatory (Treg) cells and associated elevated secretion of the regulatory cytokine IL-10 in mesenteric lymph nodes to mediate protection against allergic T-helper cell (Th-) 2 airway inflammation (Round et al., 2011) (**Figure 1**). Other studies have demonstrated that early life depletion of Bacteroidetes species using vancomycin abrogates the ability of mice to launch Treg protection from allergic asthma (Atarashi et al., 2011; Russell et al., 2012). In addition, raised levels of Helicobacter pylori has also been shown to elicit protection against the development of asthma, again, via the induction of Treg cells (Arnold et al., 2011). Interestingly this effect may, in part at least, also be due to de novo production of IL-10 orthologs by H. pylori driving this Treg induction. Moreover, oral administration of probiotics (L. reuteri, L. rhamnosus GG, Bifidobacterium breve, or B. lactis) can impair the onset of ovalbumin induced allergic airway inflammation; again related to the reduced induction of Treg cells (Feleszko et al., 2007; Karimi et al., 2009). Taken together, these and other studies are generating an important profile of the microbial species driving Treg dependent protection against allergic airway inflammation. Other studies have also identified bacteria which may drive the onset of allergic pathology. Segmented filamentous bacteria (SFB), non-cultivable Clostridia-related host-specific species (Gaboriau-Routhiau et al., 2009), and members of the cytophaga-flavobacter-bacteroides (CFB) phylum, for example, have been shown to promote differentiation of pro-inflammatory Th17 cells associated with airway inflammation (Ivanov et al., 2008; Atarashi et al., 2011) (**Figure 1**).

Although the studies described here provide insight into the potential role of GIT bacteria in the development of asthma in murine models; more studies are needed to explore the manner in which whole GIT bacterial communities, from asthmatic

enhanced generation of DC precursors in bone marrow, followed by seeding of the lungs with DCs with high phagocytic capacity and limited ability to promote Th2 cell effector function. 8. Localization of inflammatory GIT bacteria in the GIT mucus layer may induce strong IgA responses and chronic local inflammation. An influx of inflammatory Th17, Th1, and neutrophil cells in the GIT could potentially circulate to the lungs where they may contribute to asthma pathogenesis. This hypothesis may be supported by the strong associations found between irritable bowel disease (IBD) and asthma.

and non-asthmatic participants, interact with the innate and adaptive immune cells of the GIT, as well as their subsequent immune effects in the lungs. Besides, studies should also investigate a broader scope of mechanisms to explain the role of GIT bacteria in asthma pathogenesis. For example, a potential mechanism in need of further investigation is the tenable role of IgA-coated inflammatory GIT bacteria in the development of asthmatic responses in the lungs (**Figure 1**). To date, no clear link between host GIT microbiota-idiopathic intestinal inflammation and allergic lung disease has been demonstrated. However, our recent understanding of the involvement of IgAcoated bacteria in intestinal inflammation (Van der Waaij et al., 2004; Palm et al., 2014), as well as the number of clinical studies denoting an association between inflammatory lung disease and intestinal inflammation (Tulic et al., 2016); provides rationale for investigating the systemic immune effect of IgAcoated GIT bacteria. For example, it is suspected that around 50% of patients suffering from ulcerative colitis and Crohn's disease have subclinical pulmonary abnormalities with low-grade airway inflammation (Kuzela et al., 1999; Mohamed-Hussein et al., 2007). Moreover, a large cohort study, investigating 5260 IBD patients together with 21,040 non-IBD participants, recently provided strong evidence for the association between IBD and an increased risk for asthma (Peng et al., 2015). In support of this, Palm et al. (2014) clearly showed microbial localization of IgA positive bacteria from IBD patients in the normally sterile GIT mucosa of germ-free mice, which was not observed for IgA negative bacteria from IBD patients (Palm et al., 2014). We therefore hypothesize that GIT bacteria characterized by high levels of IgA coating may enter the GIT mucosa (Palm et al., 2014) where they may elicit systemic inflammatory responses at extra-intestinal mucosal sites such as the lungs.

In addition to assessing the immuno-regulatory effect of the composition of GIT bacteria in asthmatic and non-asthmatic participants; studies should also investigate the functional characteristics of GIT bacteria in the occurrence of asthma. In support of this, Trompette et al. (2014) reported the role of circulating short-chain fatty acids (SCFAs) in the protection against allergic airway inflammation (Trompette et al., 2014) (**Figure 1**). Here, microbiome metabolism in a high fiber diet setting resulted in enhanced SCFA metabolism leading to the generation of myeloid bone marrow precursors that gave rise to populations of pulmonary DCs that protected against Th2 driven allergic airway disease (Trompette et al., 2014). In support of this, Zaiss et al. (2015) demonstrated attenuated allergic airway inflammation via a GPR41 (SCFA receptor) dependent manner, as well as the effect of changes in GIT bacteria on SCFA production (Zaiss et al., 2015). Furthermore, microbial vitamin B<sup>2</sup> (riboflavin) metabolites have been shown to activate a subset of innate-like T cells, the mucosa-associated invariant T (MAIT) cells, which are highly abundant in peripheral blood, mucosal tissues, as well as the liver (Treiner et al., 2003; Le Bourhis et al., 2013). Vitamin B<sup>2</sup> from a wide range of bacteria and fungi are presented to MAIT cells by MR1 molecules (Kjer-Nielsen et al., 2012; Patel et al., 2013), followed by the rapid production of proinflammatory Th1/Th17 cytokines such as interferon-gamma (IFNγ) and IL-17 (Le Bourhis et al., 2011) (**Figure 1**). MAIT cells' pro-inflammatory responses in reaction to bacterial metabolites, together with their preferential location in the GIT lamina propria and mesenteric lymph nodes (Treiner et al., 2003), may support the "gut-lung axis" theory in a similar manner to what has been proposed for DCs. Therefore, functional properties of GIT bacteria such as dietary fiber metabolism and the production of vitamin B<sup>2</sup> may be an important aspect of host microbe crosstalk.

### THE POTENTIAL OF MODULATING GASTROINTESTINAL MICROBIOTA TO PROTECT AGAINST ASTHMA

The mechanistic insights into how GIT bacteria may protect or contribute to the development of asthma have provided great potential for the development of intervention studies. For example, the administration of probiotics (beneficial live bacterial species) (De Kivit et al., 2014), prebiotics (non-digestible food ingredients) (Jeurink et al., 2013) or symbiotics (synergistic nutritional supplements combining probiotics and prebiotics) (Van de Pol et al., 2011; Van der Aa et al., 2011) have demonstrated immune-modulatory potential via the restoration of an altered intestinal microbiota. The efficacy of probiotic administration (mainly Lactobacillus or Bifidobacterium spp.) in treatment or prevention of asthma has clearly been demonstrated in animal models (Feleszko et al., 2007; Karimi et al., 2009; MacSharry et al., 2012; Kim et al., 2013); however data in humans are not conclusive (Vliagoftis et al., 2008; Elazab et al., 2013). Nevertheless, probiotic administration needs to be carefully considered as we do not fully understand its effect on GIT bacteria. It may also hold more complex effects for the host (Shenderov, 2013), such as infections (Fijan, 2014) and allergic sensitization (Viljanen et al., 2005; Taylor et al., 2007). Various factors therefore need to be taken into account in the development of probiotics. These include the immunological pathways behind immune responses elicited by live bacteria and bacterial molecules (Caselli et al., 2011); the ongoing research around what a "healthy" GIT profile should look like (Koren et al., 2013; Knights et al., 2014) (prior to considering modulation thereof); what the effect of probiotics are on these "healthy" GIT profiles (Eloe-fadrosh et al., 2015); the inter-individual variability of the human GIT microbiome (De Filippo et al., 2010; Grze´skowiak et al., 2012; Yatsunenko et al., 2012; Lin et al., 2013; Ou et al., 2013; Suzuki and Worobey, 2014); the effect of probiotics on host metabolic and signaling pathways (Shenderov, 2013); and whether diversity within specific bacterial taxa is of importance in immunological tolerance (West, 2014). In addition, studies are needed to assess the period, dose and duration of probiotic supplementation. As for probiotics, prebiotic supplementation was not significantly associated with the prevention of asthma in humans (Arslanoglu et al., 2012; Osborn and Sinn, 2013); however, administration of oligosaccharides in mice has been associated with decreased parameters of allergic asthma (Vos et al., 2007).

It is important to also highlight the potential role of vitamin D in modulation of the GIT bacterial community and consequent immune responses such as asthma (Arshi et al., 2014). Vitamin D not only acts on a number of immune cells and processes involved in immune regulation of asthma (Brehm et al., 2009, 2010; Mann et al., 2014), but also has the potential to modulate GIT bacterial profiles and their functions (Mai et al., 2009; Jin et al., 2015). Thus, further exploring the therapeutic potential of vitamin D supplementation, together with pro-, pre- and synbiotic interventions, in modulating the host's GIT microbiota and its subsequent effect on allergic airway diseases such as asthma has merit.

Moreover, understanding the effects of other GIT microbiota, such as fungi and helminths, on the composition of GIT bacteria is also likely to be extremely informative. For example, an overgrowth of commensal fungal Candida species in the GIT, as a result of antibiotic treatment, has been shown to promote M2 macrophage activation in the lungs, as well as increased allergic airway inflammation (Kim et al., 2014). In addition, changes in the GIT bacterial composition of mice following chronic infection with the murine helminth Heligmosomoides polygyrus bakeri have been elegantly shown to protect against house dust mite induced airway inflammation (Zaiss et al., 2015). Importantly this study shows that these changes resulted in elevated SCFA production that actually underlies the protective phenotype (Zaiss et al., 2015). This and work discussed above from the Marsland laboratory provide strong evidence for dietary modulation of the microbiome protecting against allergy.

#### CONCLUSION AND PERSPECTIVES

A systematic search of the literature revealed that studies investigating fecal bacteria from humans and their relationship with asthma have been increasingly published since the beginning of the 21st century. However, reports on the role of fecal bacteria in the development of asthma in humans are limited, and primarily investigate the role of select GIT bacteria in asthma pathogenesis. Large longitudinal prospective cohort studies, with clear definitions of asthmatic outcomes, incorporating high-resolution methods (such as massively parallel 16S rRNA gene sequencing, whole-genome shotgun sequencing or culturomics), are therefore needed to determine the role of fecal bacteria in the development of asthma in both developed and developing countries. Studies also need to assess the impact of covariates (such as mode of delivery and intestinal microbes other than bacteria) on both fecal bacterial profiles and the outcome of interest using rigorous statistical analyses. Furthermore, studies should aim to test the causal link between human fecal bacteria and asthma development using murine models. Finally, the role of GIT bacteria in asthma should be investigated alongside the airway microbiome in order not to mask the importance of the local respiratory microbial-host interactions. In addition, it would be interesting to assess whether GIT bacteria impacts on corticosteroid responsiveness in asthma (Goleva et al., 2013), as well as asthma severity and phenotypes (Zhang et al., 2016).

#### Literature Search Strategy and Selection Criteria

We systematically searched peer-reviewed articles published on bacteria detected from feces and their association with asthma from six electronic databases (Medline via Pubmed, Scopus via SciVerse, Academic Search Premier, Africa-Wide Information and CINAHL via EBSCOHost and Web of Science via Web of Knowledge), using a combination of keywords [(microbiota<sup>∗</sup> OR metagenome OR microbiome<sup>∗</sup> OR "human microbiota<sup>∗</sup> "

#### REFERENCES


OR "human microbiome<sup>∗</sup> " OR "gut microbiota<sup>∗</sup> " OR "gut microbiome<sup>∗</sup> " OR "intestinal flora" OR "digestive flora" OR "gut flora" OR feces OR stool OR faeces OR fecal OR faecal) AND (asthma OR "bronchial asthma" OR "bronchial disease<sup>∗</sup> " OR "respiratory sound<sup>∗</sup> " OR "lung sound<sup>∗</sup> " OR wheez<sup>∗</sup> )]. The last literature search was 19 November 2015. All articles published in English and French were assessed for inclusion in the review. Original research articles investigating bacteria from fecal specimens and their relation to asthma in humans unexposed to antibiotic, pre- or probiotic treatments were included. In addition, we cross-checked the reference lists of all eligible studies included in this review for any additional articles.

#### AUTHOR CONTRIBUTIONS

MK, SC, and MN initiated the project. SC extracted the data and reviewed the articles with MK. SC, CW, SM, LT, and MK performed the statistical analysis and interpreted the results. SC, LT, WH, HZ, MN, and MK wrote the manuscript.

#### FUNDING

This work was supported by the Bill and Melinda Gates Foundation Global Health Grant (OPP1017641), the National Research Foundation (South Africa), the Carnegie Corporation of New York (United States of America), the US National Institutes of Health (1U01AI110466-01A1), and the Wellcome Trust, United Kingdom (102429/Z/13/Z). The first (SC) and the corresponding author (MK) had full access to the study data. All authors had final responsibility for the decision to submit the article for publication.

#### ACKNOWLEDGMENTS

SC is supported by the National Research Foundation and the Drakenstein Child Health Study, University of Cape Town (South Africa), a birth cohort study funded by Bill and Melinda Gates Foundation (OPP1017641). MK was a recipient of Carnegie Corporation of New York (USA) fellowship, and he is currently supported by Wellcome Trust, United Kingdom (102429/Z/13/Z).


rural Africans and African Americans. Am. J. Clin. Nutr. 98, 111–120. doi: 10.3945/ajcn.112.056689


**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 © 2016 Claassen-Weitz, Wiysonge, Machingaidze, Thabane, Horsnell, Zar, Nicol and Kaba. 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) or licensor 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.

# Bacteroides fragilis Lipopolysaccharide and Inflammatory Signaling in Alzheimer's Disease

Walter J. Lukiw\*

Bollinger Professor of Alzheimer's disease (AD), Neuroscience Center and Departments of Neurology and Ophthalmology, Louisiana State University Health Sciences Center, New Orleans, LA, USA

The human microbiome consists of ∼3.8 × 10<sup>13</sup> symbiotic microorganisms that form a highly complex and dynamic ecosystem: the gastrointestinal (GI) tract constitutes the largest repository of the human microbiome by far, and its impact on human neurological health and disease is becoming increasingly appreciated. Bacteroidetes, the largest phylum of Gram-negative bacteria in the GI tract microbiome, while generally beneficial to the host when confined to the GI tract, have potential to secrete a remarkably complex array of pro-inflammatory neurotoxins that include surface lipopolysaccharides (LPSs) and toxic proteolytic peptides. The deleterious effects of these bacterial exudates appear to become more important as GI tract and blood-brain barriers alter or increase their permeability with aging and disease. For example, presence of the unique LPSs of the abundant Bacteroidetes species Bacteroides fragilis (BF-LPS) in the serum represents a major contributing factor to systemic inflammation. BF-LPS is further recognized by TLR2, TLR4, and/or CD14 microglial cell receptors as are the proinflammatory 42 amino acid amyloid-beta (Aβ42) peptides that characterize Alzheimer's disease (AD) brain. Here we provide the first evidence that BF-LPS exposure to human primary brain cells is an exceptionally potent inducer of the pro-inflammatory transcription factor NF-kB (p50/p65) complex, a known trigger in the expression of pathogenic pathways involved in inflammatory neurodegeneration. This 'Perspectives communication' will in addition highlight work from recent studies that advance novel and emerging concepts on the potential contribution of microbiome-generated factors, such as BF-LPS, in driving pro-inflammatory degenerative neuropathology in the AD brain.

Keywords: 42 amino acid amyloid-beta (Aβ42) peptides, Alzheimer's disease, bacteroidetes, DAMPs and PAMPs, lipopolysaccharides, microbiome, NF-kB, systemic inflammation

### THE HUMAN MICROBIOME AND Bacteroides fragilis

Homo sapiens and their complex microbiome, consisting chiefly of bacteria with microbial eukaryotes, archaea, fungi, protozoa, viruses, and other microorganisms making up the balance, together compromise the entire metaorganism whose symbiotic associations and host interactions are critical to human health and disease (Hugon et al., 2015; Seksik and Landman, 2015; Youssef et al., 2015). There are approximately 52 recognized divisions of bacteria, however, humans, have

#### Edited by:

Christine Moissl-Eichinger, Medical University of Graz, Austria

#### Reviewed by:

Benjamin P. Willing, University of Alberta, Canada Catherine Maree Burke, University of Technology, Sydney, Australia

#### \*Correspondence:

Walter J. Lukiw wlukiw@lsuhsc.edu

#### Specialty section:

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

Received: 06 June 2016 Accepted: 15 September 2016 Published: 26 September 2016

#### Citation:

Lukiw WJ (2016) Bacteroides fragilis Lipopolysaccharide and Inflammatory Signaling in Alzheimer's Disease. Front. Microbiol. 7:1544. doi: 10.3389/fmicb.2016.01544

**75**

co-evolved with – 2 dominant phyla: Bacteroidetes [about 20% of all gastrointestinal (GI) tract bacteria] and Firmicutes (about 80%); with Actinobacteria (about 3%), Proteobacteria (∼1%), and Verrucomicrobia (∼0.1%) making up significantly smaller fractions. These four major bacterial divisions represent the 'bacterial-core' of the human microbiome (Hakansson and Molin, 2011; Zhao et al., 2015; Hug et al., 2016; Sender et al., 2016). About 98% of GI tract microbiota consists of anaerobic bacteria, and Bacteroidetes species, constituting ∼30% of all GI tract bacteria are the most abundant Gram-negative bacteria found in the GI tract outnumbering Escherichia coli abundance by at least 100 to 1 (Fathi and Wu, 2016; Foster et al., 2016; Hug et al., 2016; Rogers and Aronoff, 2016; Sampson and Mazmanian, 2016; Sender et al., 2016).

Bacteroidetes species such as Bacteroides fragilis, normal commensals of the GI tract, are thought to be generally beneficial to human health via their production of polysaccharides, volatile fatty acids, cleavage of dietary fibers into digestible short-chain fatty acids, and other nutrients, however, when they escape this environment they can cause substantial inflammatory pathology with significant morbidity and mortality (Hofer, 2014; Khanna and Tosh, 2014; Fathi and Wu, 2016). Dietary intake may have a role in regulating the composition and stoichiometry of the GI tract microbiome; for example Bacteroidetes species have been observed to proliferate in porcine models fed high-fat diets deprived of sufficient dietary fiber (Heinritz et al., 2016). In addition to their lipopolysaccharide (LPS) generation, B. fragilis endotoxins are a leading cause of anaerobic bacteremia and sepsis/systemic inflammatory distress through their generation of the highly pro-inflammatory zinc metalloprotease metalloproteinase B. fragilis toxin (BFT) fragilysin (Choi et al., 2016; Fathi and Wu, 2016). BFT has recently been shown to disrupt epithelial cells of GI tract barriers via cleavage of the synaptic adhesion zonula adherens protein E-cadherin (Seong et al., 2015; Choi et al., 2016; Zhan and Davies, 2016). It is currently not well understood if GI tract barrier-disrupting proteolytic endotoxins such as BFT are able to propagate pathogenic actions via the systemic circulation to further disrupt the blood-brain barrier and transfer LPS, BFT, and other endotoxins into the cerebrovascular circulation to the neural cells and synaptic circuitry of the CNS. B. fragilis has been shown to play a pathological role in neurodevelopment including autism spectrum disorder (ASD) via circulating metabolites (Hofer, 2014). It has also recently been reported that along with BFTs amyloid peptide-dependent changes in synaptic adhesion affect both the function and integrity of synapses, suggesting that the observed deficits in synaptic adhesion in Alzheimer's disease (AD) play key roles in the progressive disruption of functional signaling throughout neuronal networks (Lin et al., 2014; Seong et al., 2015; Leshchyns'ka and Sytnyk, 2016).

### INFLAMMATORY SIGNALING IN ALZHEIMER'S DISEASE (AD)

Multiple and highly interactive aspects of increased inflammatory signaling is a consistent and recurrent feature of AD and the major pathological lesions that define AD, including insoluble Aβ42-enriched peptide deposits, neurofibrillary tangles, apoptotic, damaged, and dying neurons, and activated microglia are potent neuropathological stimulants that maintain the brain in a chronic and self-reinforcing inflammatory state (Hill and Lukiw, 2015; Calsolaro and Edison, 2016; Minter et al., 2016; Richards et al., 2016; Varatharaj and Galea, 2016). These progressive and ultimately fatal pro-inflammatory and neurodegenerative processes appear to be further stimulated by aberrant or excessive deregulation of the innate-immune response, and an increasing focus has been placed on pathological contributions by the human microbiome including dietary effects on microbial composition that appear to support pathological functions (Hill and Lukiw, 2015; Zhao et al., 2015; Foster et al., 2016). There is a wealth of accumulating evidence (a) that specific types of microbial LPS and endotoxins (such as BF-LPS and BFT) from enterotoxigenic microbes specifically impact microglialmediated innate-immune responses, phagocytic, and detoxifying mechanisms and amyloidogenesis characteristic of inflammatory neurodegeneration (Bhattacharjee and Lukiw, 2013; Hill et al., 2014; Clark and Vissel, 2015; Hill and Lukiw, 2015; Lim et al., 2015); and (b) a resurgence of our interest in the importance of GI tract and blood-brain barrier systems that normally exclude these microbiome-sourced toxins from the systemic circulation and CNS but become leaky with age and disease (Montagne et al., 2015; Choi et al., 2016; Minter et al., 2016; Richards et al., 2016; Soenen et al., 2016; van de Haar et al., 2016; Varatharaj and Galea, 2016; Zhan and Davies, 2016; Zhao et al., 2016).

Interestingly, while secreted LPS, proteolytic endotoxins, and amyloid monomers are generally soluble as monomers over time they form into highly insoluble fibrous protein aggregates that are implicated in the progressive degenerative neuropathology of several common, age-related disorders of the human systemic circulation and CNS including systemic inflammation response syndrome, multiple sclerosis (MS), prion disease, and AD (Asti and Gioglio, 2014; Clark and Vissel, 2015; Devier et al., 2015; Richards et al., 2016; Zhao et al., 2016). At the genetic level, virtually all of this inducible inflammatory signaling within the CNS involves NF-kB activation and NFkB-recognition and binding to target NF-kB DNA sequences as a prelude to the up-regulation of pro-inflammatory gene expression pathways, including the up-regulation of discrete families of pro-inflammatory pathogenic microRNAs that selectively down-regulate their mRNA targets (Lukiw and Bazan, 1998; Lukiw et al., 2008; Devier et al., 2015; Zhao et al., 2016).

### LIPOPOLYSACCHARIDE (LPS) AND LPS SIGNALING

Lipopolysaccharides are characteristic components of the outer leaflet of the outer membrane of Gram-negative bacteria shed into the extracellular space that play key roles in host– pathogen interactions of the innate-immune system (Hill and Lukiw, 2015; Zhao et al., 2015; Jiang et al., 2016; Maldonado

et al., 2016). While LPSs contain large and hypervariable polysaccharide/oligosaccharide regions, the relatively conserved lipid region (lipid A) is the endotoxic and biologically active moiety that is responsible for septic shock (Jiang et al., 2016; Maldonado et al., 2016). A "canonical" LPS structure is represented by that of LPS from E. coli, that contains one of the most potent neurotoxic lipid A species known, consisting of a 1,4<sup>0</sup> -biphosphorylated glucosamine disaccharide bearing six fatty acids which are unbranched chains 12–14 methyl(ene) units in length. Other 'lipid A' species show variability in the number, length, and composition of the attached fatty acids, as well as variability in the degree of phosphorylation and number and types of substituted phosphate ligands. For instance, BF-LPS lipid A is penta-acylated and mono-phosphorylated and contains branched fatty acids 15–17 methyl(ene) units in length; deviations from the canonical lipid A structure are known to have a profound impact on the host innateimmune responses. LPS activates Toll-like receptors (TLRs), membrane-spanning protein receptors expressed in microglial cells of the innate immune system which recognize common damage- or pathogen-associated molecular-patterns (DAMPS, PAMPs; Gustot et al., 2013; Land, 2015; Maldonado et al., 2016; Minter et al., 2016; Richards et al., 2016; Varatharaj and Galea, 2016). TLRs play key roles in host protection from microbial invasion via the activation of the innate-immune system by sensing structurally conserved DAMPS or PAMPs from microbes or microbial exudates that are distinguishable from, and not innate to, the host organism (Land, 2015; Yu and Ye, 2015). Interestingly, of the 13 currently characterized TLRs the microglial TLR2 and TLR4 are activated by amyloid, LPS, lipoglycans and/or other microbial triggers that subsequently induce cytokine production, inflammation, phagocytosis and innate immune defense responses that directly impact the development of CNS pathology. More specifically, the TLR2 complex can recognize biofilm-associated LPS and amyloids produced by both Bacteroidetes and Firmicutes (Nishimori et al., 2012; Bhattacharjee and Lukiw, 2013; Asti and Gioglio, 2014; Hill and Lukiw, 2015). Interestingly, Aβ42 peptides overproduced in AD that are associated with microglia-mediated inflammatory responses have very recently been shown to activate TLR2 and/or TLR4 (Hill and Lukiw, 2015; Yu and Ye, 2015; Minter et al., 2016). Of further interest is: (a) that microbial amyloids induce pro-inflammatory interleukin IL-17A and IL-22, triggers for NF-kB activation/signaling and cyclooxygenase 2 activation via direct TLR2 activation (Nishimori et al., 2012); and (b) that increased levels of both IL-17A and IL-22 are associated with age-related inflammatory neurodegenerative diseases such as AD (Zhang et al., 2013; Calsolaro and Edison, 2016; Maldonado et al., 2016; Richards et al., 2016). Bacterial endotoxin- and LPS-induced neuro-inflammation has been known for some

FIGURE 1 | Relative induction of NF-kB (p50/p65)-DNA binding in pro-inflammatory factor and lipopolysaccharides (LPS)-treated primary human neuronal-glial (HNG) co-cultures. (A) HNG cells in primary co-culture for 1.5 weeks; HNG cells were stained with a neuron-specific β-tubulin III (red fluorescence λmax∼650 nm; anti-βTUBIII antibody, Sigma-Aldrich, St Louis, MO, USA); an antibody against glial fibrillary acidic protein (GFAP; green fluorescence; λmax ∼510 nm; Santa Cruz Biotechnology, Santa Cruz, CA, USA) and DAPI nuclear stain (blue fluorescence; λmax∼470 nm; Thermo Fisher Scientific, Waltham, MA, USA); 20×; (B) induction of the pro-inflammatory transcription factor NF-kB (p50/p65 activation complex) by various physiologically relevant, pro-inflammatory factors all at equal dosage (25 nM); NF-kB abundance was measured by NF-kB-DNA binding assay (a measure of NF-kB activation and binding to NF-kB-DNA recognition sequences) onto a 36 nucleotide end-labeled double stranded DNA fragment containing the canonical human NF-kB (p50/p65) recognition sequence 5 0 -GGGGACTTTCCC-3<sup>0</sup> as previously described (Lukiw and Bazan, 1998; Lukiw et al., 2008; Devier et al., 2015; Clement et al., 2016); a scrambled control nucleotide containing no such NF-kB recognition sequence showed NF-kB-DNA binding activity (data not shown); (C) data from gel bands in panel (B) quantified in bar graph format; note robust induction of the NF-kB (p50/p65 complex) by Escherichia coli lipopolysaccharide (EC-LPS; LPS from Escherichia coli 0111:B4; Sigma L3012, St Louis, MO, USA) or B. fragilis lipopolysaccharide (BF-LPS; prepared by methods previously published; Eidhin and Mouton, 1993) that was ∼45- to ∼55-fold higher than that of the control human serum albumen (HSA) protein, and was fivefold to sevenfold higher than the combination of the pro-inflammatory Aβ42 peptide and IL-1β together (at 25 nM each); complex mixtures of microbiome bacterial LPS on NF-kB induction might be expected to be additive or synergistic; HSA = (control) human serum albumen; CFH = complement factor H; TNFα = tumor necrosis factor alpha (cachectin); IL-1β = interleukin 1-beta; Aβ40, Aβ42 = amyloid beta peptide, 40 and 42 amino acids in length; EC-LPS, BF-LPS = E. coli, Bacteroides fragilis lipopolysaccharide; error bars represent one standard error of the mean; N = 4; <sup>∗</sup>p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001, ANOVA.

time to be important in driving the generation of Aβ42 (Lee et al., 2008; Asti and Gioglio, 2014; Hill and Lukiw, 2015; Zhao et al., 2015). In addition to TLR2 and TLR4 at least one additional microglial transmembrane LPS receptor CD14 mediates phagocytosis of both bacterial components and Aβ42 peptides, hence expanding roles for microglia and microglial LPS receptors in the pathophysiology of AD (Lee et al., 2008; Halmer et al., 2015; Jiang et al., 2016). Of further interest is that gram negative bacterial exudates such as BF-LPS are hypervariable in composition, and different Bacteroidetes species appear to generate unique temporal patterns of LPS that exhibit rapid adaptive changes – these include the modulation of LPS synthesis and structure and alterations in DAMP/PAMP recognition features as strategies for host immune system evasion (Land, 2015; Maldonado et al., 2016; Richards et al., 2016).

#### CONCLUDING REMARKS

fmicb-07-01544 September 22, 2016 Time: 14:46 # 4

We hope that this 'Perspectives' article has adequately highlighted some recent findings on microbial-derived LPS and proteolytic endotoxins and has engendered interest in the potential contribution of these neurotoxic and proinflammatory microbial exudates to amyloidogenesis and age-related inflammatory neurodegeneration. Taken together, these current observations advance seven key areas in our understanding of the role of the microbiome in progressive, age-related inflammatory neurodegeneration: (a) that relatively low, nanomolar amounts of bacterial LPS are extremely potent inducers of the pro-inflammatory transcription factor NF-kB (p50/p65 complex) in human primary human neuronal-glial (HNG) co-cultures of brain cells (**Figure 1**); (b) that different LPS preparations from different bacterial species appear to exhibit slightly different trends in the induction of an inflammatory response, as quantified by the extent (mean values) of NF-kB activation and DNA-binding (**Figure 1**); (c) that the proliferation of microbial species such as the BF-LPS and BFT generating Bacteroidetes may be regulated by diet, environment, and lifestyle factors such as dietary fiber intake can impact neurological health, CNS inflammation, and degenerative disease (Hofer, 2014; Khanna and Tosh, 2014; Zhan and Davies, 2016); (d) that LPS transit across compromised GI tract and blood-brain barriers underscore the critical roles of cellular adhesion structures in allowing passage of noxious molecules from the GI tract into the systemic circulation and CNS (Montagne et al., 2015; Soenen et al., 2016; van de Haar et al., 2016); (e) that biophysical, gastrointestinal, and neurobiological barriers which may become more 'leaky' with aging again underscore the important role of tight junctions in moderating systemic and CNS inflammation and immune-mediated inflammatory disease (Hill and Lukiw, 2015; Varatharaj and Galea, 2016); (f) that DAMPS and/or

#### REFERENCES

Asti, A., and Gioglio, L. (2014). Can a bacterial endotoxin be a key factor in the kinetics of amyloid fibril formation? J. Alzheimers Dis. 39, 169–179. doi: 10.3233/JAD-131394

PAMPs common to both LPS, endotoxins, and/or amyloid may trigger TRL2, TRL4 and/or CD14 microglial receptors to propagate and sustain inducible AD-relevant inflammatory responses within the CNS (Land, 2015; Zhao and Lukiw, 2015; Varatharaj and Galea, 2016); and (g) that LPS abundance, speciation and complexity in the CSF and/or blood serum may be useful diagnostically for the onset of mild cognitive impairment (MCI), the clinical precursor for the development of AD. Clearly, more research into the intriguing realm of human microbiomehost interaction is warranted, and the study of the complex interactions between each biological niche is certain to shed new light on the still evolving concepts and mechanisms of microbiome interplay and control in human neurological health and disease.

#### AUTHOR CONTRIBUTIONS

WL and the late Dr. James Hill performed experiments, researched and wrote this paper; the author is sincerely grateful to colleagues and collaborators for helpful discussions, medical artwork and unpublished data; the contributions of these researchers have been recognized in the 'Acknowledgements' section.

#### ACKNOWLEDGMENTS

This research work was presented in part at the Society for Neuroscience (SFN) Annual Meeting 17–21 October 2015, Chicago IL, USA and at the Alzheimer Association International Congress 2016 (AAIC 2016) Annual conference 21–27 July 2016 in Toronto CANADA. These studies utilized total nucleic acid and/or cytoplasmic fractions extracted from primary human neuronal-glial (HNG) co-cultures; sincere thanks are extended to Drs. P. Alexandrov, J. G. Cui, F. Culicchia, W. Poon, and Y. Zhao for short post-mortem interval (PMI) human brain tissues or extracts, HNG tissue culture and NF-kB-DNA binding assay and initial data interpretation, and to D. Guillot and A. I. Pogue for expert technical assistance and medical artwork. Thanks are also extended to the many neuropathologists, physicians, and researchers of the US, Canada, and Europe who have provided high quality, short PMI human CNS, or extracted tissue fractions for scientific study. Research on the human microbiome, proinflammatory, and pathogenic signaling in the Lukiw laboratory involving the innate-immune response, neuroinflammation and amyloidogenesis in AD and in other neurological diseases was supported through an unrestricted grant to the LSU Eye Center from Research to Prevent Blindness (RPB); the Louisiana Biotechnology Research Network (LBRN) and NIH grants NEI EY006311, NIA AG18031, and NIA AG038834.

Bhattacharjee, S., and Lukiw, W. J. (2013). Alzheimer's disease and the microbiome. Front. Cell. Neurosci. 7:153. doi: 10.3389/fncel.2013.00153

Calsolaro, V., and Edison, P. (2016). Neuroinflammation in Alzheimer's disease: current evidence and future directions. Alzheimers Dement. 12, 719–732. doi: 10.1016/j.jalz.2016.02.010



**Conflict of Interest Statement:** The author declares 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 © 2016 Lukiw. 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) or licensor 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 Skin Microbiome: Is It Affected by UV-induced Immune Suppression?

#### VijayKumar Patra1,2, Scott N. Byrne<sup>3</sup> and Peter Wolf<sup>1</sup> \*

<sup>1</sup> Research Unit for Photodermatology, Department of Dermatology, Medical University of Graz, Graz, Austria, <sup>2</sup> Center for Medical Research, Medical University of Graz, Graz, Austria, <sup>3</sup> Cellular Photoimmunology Group, Infectious Diseases and Immunology, Sydney Medical School, The Charles Perkins Center Hub at The University of Sydney, Sydney, NSW, Australia

Human skin apart from functioning as a physical barricade to stop the entry of pathogens, also hosts innumerable commensal organisms. The skin cells and the immune system constantly interact with microbes, to maintain cutaneous homeostasis, despite the challenges offered by various environmental factors. A major environmental factor affecting the skin is ultraviolet radiation (UV-R) from sunlight. UV-R is well known to modulate the immune system, which can be both beneficial and deleterious. By targeting the cells and molecules within skin, UV-R can trigger the production and release of antimicrobial peptides, affect the innate immune system and ultimately suppress the adaptive cellular immune response. This can contribute to skin carcinogenesis and the promotion of infectious agents such as herpes simplex virus and possibly others. On the other hand, a UV-established immunosuppressive environment may protect against the induction of immunologically mediated skin diseases including some of photodermatoses such as polymorphic light eruption. In this article, we share our perspective about the possibility that UV-induced immune suppression may alter the landscape of the skin's microbiome and its components. Alternatively, or in concert with this, direct UV-induced DNA and membrane damage to the microbiome may result in pathogen associated molecular patterns (PAMPs) that interfere with UV-induced immune suppression.

#### Edited by:

Martin Grube, University of Graz, Austria

#### Reviewed by:

Juris A. Grasis, San Diego State University, USA David William Waite, University of Queensland, Australia

#### \*Correspondence:

Peter Wolf peter.wolf@medunigraz.at

#### Specialty section:

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

Received: 14 March 2016 Accepted: 25 July 2016 Published: 10 August 2016

#### Citation:

Patra V, Byrne SN and Wolf P (2016) The Skin Microbiome: Is It Affected by UV-induced Immune Suppression? Front. Microbiol. 7:1235. doi: 10.3389/fmicb.2016.01235 Keywords: skin microbiome, ultraviolet radiation, immune suppression, innate immunity, environmental factors

### SKIN MICROBIOME

### Introduction

The human skin is the largest organ of the body with a diverse physical, chemical, and biological ecosystem over a surface of 1.8 m<sup>2</sup> . The epithelial surface engages itself in a mutualistic relationship with a wide range of composite microorganisms, including bacteria, fungi, viruses, and mites, all of them residing on or within the skin. Around 1 million bacteria reside per square centimeter of skin surface, making up an estimated total of 10<sup>10</sup> bacterial cells growing on the entire skin of the human body (Grice et al., 2008; Hannigan and Grice, 2013). The majority of those bacteria are commensals or transients. The diversity of the skin's microbiome is due to the diverse environments present on the skin, resulting from a divergent physical nature of the skin, humidity, temperature, pH, lipid and sebum content, as well as antimicrobial peptide (AMP) expression (Nakatsuji et al., 2013). Considering the capacity for a robust cutaneous immune system to rapidly detect and eliminate foreign invaders, it is intriguing that innumerable numbers of microorganisms reside on the skin surface and also extend to sub-epidermal compartments, associated appendages such as, hair follicles and sebaceous glands (Nakatsuji et al., 2013; Belkaid and Segre, 2014).

### Bacteria

fmicb-07-01235 August 10, 2016 Time: 10:59 # 2

Sequencing the bacterial 16S small-subunit ribosomal RNA gene reveals that the skin surface is dominated by Proteobacteria, Actinobacteria, Bacteriodetes, and Firmicutes (Grice et al., 2008). Colonization varies topographically; for example, Proteobacteria and Staphylococcus spp. are abundantly present on the skin surface and are deeply intertwined between themselves and other microorganisms. Recent studies indicate that specific bacterial communities are associated with moist, dry or sebaceous microenvironments of the skin. Moist areas such as umbilicus, the axillary vault, inguinal crease, gluteal crease, foot, popliteal fossa, and the antecubital fossa have an abundance of Staphylococcus and Corynebacterium spp. and these organisms are well known to prefer skin sites with high humidity. The dry areas such as buttocks, forearms and certain other parts of the hand have the most intertwined collection of phyla Actinobacteria, Proteobacteria, Firmicutes, and Bacteriodetes (Grice et al., 2008). Interestingly, these dry sites are known to have an abundance of gram-negative bacteria which were thought to colonize very rarely on the skin. When compared to the gut or the oral cavity, dry skin sites are home to a much greater phylogenetic diversity of bacteria (Gao et al., 2007; Costello et al., 2009; Grice et al., 2009). The lowest bacterial diversity is seen around the sebaceous sites, which suggests that only few bacterial communities can flourish under those conditions (Costello et al., 2009). The forehead, retroauricular crease (behind the ears), the back and the alar crease (sides of the nose) are few of the sebaceous sites containing low phylotype richness. Propionibacterium spp. dominate the sebaceous skin areas like the hair follicle, hair shaft, and the sebaceous gland (Costello et al., 2009).

### Fungi

The skin microbiome is not only limited to bacteria, but extends to fungal species as well, which have major roles in health, and disease. Genomic methods to characterize fungal species are very limited when compared to that available for the bacteria. Studies done by Findley et al. (2013) described fungal communities residing in the skin by sequencing 18s rRNA (phylogenetic marker within ribosomal RNA region) and ITS1 (internal transcribed spacer 1 region). They observed that the genus Malassezia predominated across the 11 core body and arm sites of their study (back, occiput, external auditory canal, inguinal crease, retroauricular crease, glabella, manubrium, nare, anticubital fossa, volar forearm, hypothenar palm, plantar heel, toe nail, and toe-web space). Plantar heel had the highest diversity of fungal species, with a mixed representation of Malassezia, Aspergillus, Cryptococcus, Rhodotorula, Epicoccum, and few others (Findley et al., 2013). ITS1 sequencing also revealed that Candida species like tropicalis, parapsilosis, and orthopsilosis, and Cryptococcus species flavus, dimennae, and diffluens were observed on and within the skin. This may be important as these species are thought to be potential pathogens in wounds of immunocompetent subjects or immunocompromised patients in general (Larone, 2002; Findley et al., 2013).

### Archaea

Probst et al. (2013) reported about the less known archaea found on the human skin based on the 16S rRNA sequencing. They observed around 4% of the overall microbial genes was found to be archaeal 16s rRNA genes. They also found that around 88% of all the observed operational taxonomical units (OTUs) consisted of phyla Thaumarchaeota with the rest being Euryarchaeota (Probst et al., 2013). As bacteria constitute the bulk of the microbiome, archaea have received little research attention. Advances in detection technologies now allow to detect all the archaeal taxa and study its potential prevalence on the skin and understand its impact on health and diseases. The prevalence of archaea within the human microbiome has been addressed in a recent review by Horz (2015).

#### Viruses

Skin viral microbiota is one of the most rarely investigated subsets of the human microbiome, as the skin related viruses cannot be cultivated and do not portray consensus sequences to be detected by molecular methods. In recent years, there has been increasing evidence suggesting that healthy skin harbors resident or short-lived viruses, such as, human alpha, beta, and gamma papillomaviruses (α-HPV, β-HPV, and γ-HPV) present on and within the upper layers of human skin (Chen et al., 2008; Antonsson et al., 2003a,b). By applying the functional metagenomic methods (Hannigan et al., 2015) to skin samples have also led to observing new viral species such as Polyomoaviridae family (Foulongne et al., 2012). Foulongne et al. (2012) reported that eukaryotic DNA viruses on human skin consisted of Papillomaviridae, Polyomoaviridae, and Circoviridae. They also observed 13 new γ-HPV strains present on the healthy skin (Foulongne et al., 2012). These reports indicate the presence of cutaneous viral microbiota and their possible involvement in various proliferative skin diseases.

#### Mites

Small arthropods such as Demodex mites (Demodex folliculorum and Demodex brevis) have long been associated with rosacea and other skin conditions such as chronic blepharitis (Georgala et al., 2001; Lacey et al., 2007; Elston, 2010). These mites are usually found on the facial skin, around the pilosebaceous glands and the hair follicles and are thought to be a part of commensal skin microbiota, as they have prevalence rates between 23 and 100% in healthy individuals (Norn, 1971; Rufli and Mumcuoglu, 1981). Furthermore, Demodex mites are also associated with cutaneous conditions of rosacea-like appearance, commonly clubbed as demodicosis or demodicidosis, though their presence on the skin most commonly remains asymptomatic.

#### Factors Affecting Skin

The skin is one the most exposed organs of our body, which provides a physical defense from the external environment and most often reacts appropriately to the wide range of hazards encountered. To list only a few of the environmental hazards to the skin microbiome, exposure to UV-R, pollutants, climatic elements at certain geographic locations, occupation,

and exaggerated hygiene have to be noted. Despite all these environmental assaults, the microbiome and the host must maintain cutaneous homeostasis to sustain normal physiological behavior.

### UV-R AND IMMUNE SUPPRESSION

Among the above mentioned factors, UV-R is one of the most prominent factors linked to skin hazards. UV-R, especially UV-B (280–315 nm) and UV-A (315–400 nm) are known to be involved in skin freckling, wrinkling, photo allergic, and phototoxic responses and tumor induction and progression (Krutmann, 2000; Lee et al., 2013). One of the widely known aspects of UV-R on the skin is the ability to induce photoproducts such as cyclobutane pyrimidine dimers and subsequently occurring mutations, which are linked to carcinogenesis (Lee et al., 2013). Immune suppression is also induced by UV-R and is considered to be another of its harmful impacts (Phan et al., 2006). Kripke et al. (1977) first discovered that UV-R exposure and immune suppression were linked to UV-induced carcinogenesis (Schwarz, 2010). Since then, the immunomodulating properties of UV-R were confirmed by employing contact hypersensitivity (CHS) models in mice (Toews et al., 1980; Elmets et al., 1983) as well as in humans (Cooper et al., 1992; Kelly et al., 2000; Wolf et al., 2003). UV-R-induced immune suppression is known be mediated through T cells (Elmets et al., 1983). The relation of immune suppression linked to various subtypes of regulatory immune cells such as regulatory T cells (Tregs) (Schwarz, 2008; Schweintzger et al., 2015; Schweintzger et al., 2016) and regulatory B cells (Bregs) (Byrne and Halliday, 2005) depend on UV-R doses, antigens and type of immune response (i.e., CHS vs. DTH). Some of the key events observed in the skin after the UV-R exposure are shown in **Figure 1**. The most studied photochemical reactions triggered by UV-R are DNA damage (Applegate et al., 1989), isomerization of urocanic acid (De Fabo and Noonan, 1983), and formation of reactive biophospholipids such as platelet activating factor (PAF) (Wolf et al., 2006) in the skin. Cytokines such as tumor necrosis factor (TNF)-α (Wolf et al., 2000), interleukin (IL)-4, IL-10, IL-33 (Byrne et al., 2011), and prostaglandin E<sup>2</sup> (Shreedhar et al., 1998) are expressed and upregulated after UV-R exposure. In healthy skin, the expression of these cytokines by UV-R leads to infiltration of suppressor macrophage and neutrophils (Cooper et al., 1985, 1986, 1993. Furthermore, UV-R also induces the emigration of Langerhans cells (LC) from the epidermis into the draining lymph nodes (Toews et al., 1980; Noonan et al., 1984) and affects mast cells which are known to be involved in immune suppression (Hart et al., 2001). In addition to DNA damage,cis-urocanic acid (UCA) is a UV-induced immunosuppressive molecule (De Fabo and Noonan, 1983), acting via the 5-HT2A receptor (Walterscheid et al., 2006; Wolf et al., 2016). Upon exposure to UV-R, transurocanic acid (trans-UCA) is converted to the cis isoform, which accumulates on the stratum corneum and epidermis. Kubica et al. (2014) observed temperate changes in the skin microbiome of capase-14 deficient mice, which are generally characterized by reduced overall levels of UCA (Kubica et al., 2014). Interestingly, it is known that caspase-14 controls the proteolysis of fillagrin and patients with fillagrin mutations are more likely to develop atopic dermatitis (AD), which in turn is associated with alterations of microbial load on the skin (McAleer and Irvine, 2013). Another immunosuppressive factor released upon UV exposure is PAF that binds to the PAF receptor, resulting in a cascade of downstream events that lead to the release of IL-10, and ultimately immune suppression (Walterscheid et al., 2002; Wolf et al., 2006).

Since UV-R suppresses the immune system and previous research has shown in experimental models that UV-R can suppress the immune response to infectious microorganisms (Chapman et al., 1995), one can speculate that exposure to UV-R could enhance susceptibility to microbial infections and/or it could worsen infectious diseases. However, clinical evidence of the increased infections after UV-R exposure remains very low, with a few exceptions. It has been known for a long time that UV exposure can trigger and/or exacerbate herpes simplex virus (HSV) manifestations (Norval, 2006). A UV dose dependent induction was observed on the lips of patients harboring the latent infections of HSV-1 (type 1). This was also reported in cases of HSV-2 (type 2). Local modifications of the immune system are thought to be the involved in the reactivation of the latent HSV infection (Wheeler, 1975; Tesini et al., 1977; Spruance, 1985). In immunocompetent humans, Candida albicans is known to cause (minor) infections of the skin and mucosa of the genital and gastrointestinal tract, however, in immunosuppressed patients it can cause life-threatening systemic infections. Importantly to note, suppression of the response to candida albicans antigen is used in one of the standard assays to quantify immune suppression induced by UV-R (Wolf et al., 2016). In mice, it is also known that UV-R decreases the delayed type hypersensitivity (DTH) responses and significantly modifies the course of the infection caused by Mycobacterium bovis bacillus Calmette-Guérin (BCG), which is closely related to the organism causing tuberculosis in humans (Jeevan and Kripke, 1989). Higher number of viable bacilli were observed in the peripheral lymph nodes of mice irradiated with UV, compared to that of unirradiated mice (Jeevan and Kripke, 1989). This phenomenon was similarly observed in mice infected with Mycobacterium lepraemurium, a pathogen known to cause infections which to some extent resemble leprosy in humans (Jeevan et al., 1992). In humans, experimental exposure to UV-R reduced the granulomatous reaction in individuals sensitized with lepromin (antigens of Mycobacterium lepraemurium) (Cestari et al., 1995). Brown et al. (2001) reported that exposure to UV-R can alter the immune responses in mice infected by Borrelia burgdorferi and intensify the associated arthritic component (that resembles Lyme disease in humans) (Brown et al., 2001).

#### Antimicrobial Peptides

AMPs are small proteins, known to have microbicidal activity. Besides being microbicidal, they possess immunomodulating properties. AMPs are mostly produced by the cells in constant exposure to microorganisms. In the skin there are two main classes of AMPs, i.e., β-defensins and cathelicidins. In humans LL-37 is the only cathelicidin found, which is expressed by

various epithelial cells such as keratinocytes in inflamed skin (Niyonsaba et al., 2007). Apart from defensins and cathelicidins, skin also constitutively expresses a wide range of AMPs like RNase 7 (ribonuclease 7), S100A7 (Psoriasin, calcium binding protein), and dermcidin (that is sweat gland derived). AMPs production can be increased during inflammation or in cases of infection. The production of these AMPs, including beta defensins, cathelicidins, ribonucleases, and S100 proteins, are triggered by various pathogen or damage associated molecular patterns (PAMPs/DAMPs), exogenous microbial danger signals like Toll-like receptor (TLR) agonists, or endogenous mediators of inflammation such as TNF-α, IL-1, IFN-γ, and IL-17 (Biragyn et al., 2002; Chadebech et al., 2003; Joly et al., 2005; Kolls et al., 2008). It is also known that UV-R induces productio of AMPs, as essential components and triggers of the innate immune system. Studies have shown that UV-R induces human beta defensin 2 (hBD2), hBD3, ribonuclease 7 (RNase7), S100A7 (psoriasin), S100A12 and elafin by keratinocytes in vitro and in vivo (Yang et al., 2002; Hong et al., 2008; Gläser et al., 2009; Felton et al., 2013; Kennedy Crispin et al., 2013). One mechanism of UV-induction of AMPs could involve production of the active form of Vitamin D (Wang et al., 2004). Alternatively, or in concert with this, Vitamin D<sup>3</sup> itself could be suppressing adaptive immune responses (Damian et al., 2010) and/or tempering inflammatory events in UV-exposed skin (Mason et al., 2010).

Apart from participating in innate immune responses, AMPs are also involved in activating and mediating adaptive immune responses (Yang et al., 1999, 2002; Biragyn et al., 2002; Niyonsaba et al., 2007; Navid et al., 2012). The much abundant skin's physiologically beneficial microbiome vastly depends upon AMPs to be kept and maintained in homeostasis in order to release and allow the immune system to mount an immune response, when needed, and protecting against invading pathogens.

#### UV-R, SKIN MICROBIOME, AND IMMUNE SYSTEM INTERACTION

In the last years the cellular components of the acquired (i.e., adaptive) immune response and to a lesser extent of the innate immune system and its changes after UV-R were characterized extensively. To date, relatively little is known about the effect of UV-R on the microbiome of the skin and how this affects the immune response after UV-R. Since the skin microbiome is established all over the skin surface and reaches deep down into appendages, logic dictates that it experiences similar impacts from UV-R as mammalian skin cells do (**Figure 2**). This exposure to UV-R can alter/damage the microbial community, possibly resulting in disruption of microbial components and/or formation of bacterial antigens, some of which may become immunogenic.

#### UV-R and the Skin Microbiome

fmicb-07-01235 August 10, 2016 Time: 10:59 # 5

Much of the skin microbiome at sun-exposed body sites is directly exposed to solar UV-R, either completely or during much of its life cycle. UV-R can impose an intense change in the microbial species and its genotypic composition at exposed sites, depending upon the exposure time and intensity of UV-R. Some bacteria and fungi show a selective tolerance to UV-R for part of their life cycles, and are often vulnerable to the effects of UV during sporulation, diffusion and other processes such as infection. One of the major effects of UV-R on microbes is DNA damage, which can result in an increase in genetic variation or can alter the landscape of the microbial communities, thus disrupting the healthy microbiome (Rothschild, 1999). However, not all microbes are susceptible to the damaging effects of UV-R. For instance, many fungi show photomorphogenic effects upon exposure to UV-R. Wang et al. (2012) showed that the production of porphyrins by Propionibacterium acnes (P. acnes) was decreased with increased doses of UV-R, which provides evidence that facial bacteria are responsive to UV-R. They also observed the decrease of porphyrin production by P. acnes, at doses lower than 20 mJ/cm<sup>2</sup> of UV-B. This indicates that P. acnes responds to UV-B even before a significant skin injury can be detected (Wang et al., 2012).

Malassezia spp. which also belong to the commensal microflora, is commonly causing pityriasis versicolor, a common skin disease condition particularly occurring in tropical regions but also moderate latitudes. At the site of the typical brownishwhite scaly skin lesions of manifested disease, a sunburn response can hardly be provoked. Pityriacitrin, a UV-filtering compound which is produced by Malassezia furfur is believed to be protective. On the other hand, UV-R is well known to inhibit the cellular growth of Malassezia furfur (Wikler et al., 1990). Machowinski et al. (2006) looked into the effects of pityriacitrin on the other commensals residing on skin and observed that Malassezia/Pityrosporum was inhibited by UV-R and was much more sensitive than other commensals. It is hypothesized that fungi developed this UV-filter to reduce UV sensitivity which helps to grow and survive by competing with other commensals. However, Machowinski et al. (2006) they did not find any negative effect of pityriacitrin on other common skin commensals such as Staphyloccoccus aureus, Staphylococcus epidermidis, or Candida albicans.

Among viral populations, UV-R is known to be a stimulus for HSV. UV-R has been found to activate herpes virus promotor(s) and transcription factors such as c-jun (Stein et al., 1992; Loiacono et al., 2003). However, the complete mechanisms by which UV-R triggers clinical HSV manifestations have not been understood so far, but it is thought that UV effects on the immune system may contribute. In addition, there is some evidence that UV exposure may be a reactivation trigger of latent Varicella zoster virus infection, resulting in herpes zoster (Zak-Prelich et al., 2002).

Human papillomaviruses (HPVs), especially which come under the group epidermodysplasia verruciformis (EV) types, are known to be widespread on human skin (Norval, 2006). The EV HPV types have been linked to skin carcinogenesis in the genetic disease of EV (Lewandowsky-Lutz dysplasia). However, EV HPV types are also often found in non-melanoma skin cancers of the normal population though their pathologic significance in the later population remains uncertain. HPV appears to be part of the commensal flora of the skin, making it a challenge identifying causal relationships between HPV presence and skin conditions, including cancer (Neale et al., 2013). For instance, EV HPV was also found in hair follicles of psoriasis patients, in particular treated with psoralen + UV-A radiation, whereas its expression was not observed in patients who had received no treatment (Wolf et al., 2004). Together, however, there is increasing evidence suggesting that UV-R could affect the homeostasis between certain EV HPV types and the host, thus driving the infection from a latent stage to potentially oncogenic (Norval, 2006).

Patients suffering from AD were shown to have increased Staphylococcus aureus colonization on the skin, which was reduced after treatment with UV-B. Interestingly, UV-B had no effect on Staphylococcus epidermidis, possibly due to the fact that Staphylococcus epidermidis is mostly located around hair follicles, whereas Staphylococcus aureus is present on superficial skin layers more accessible for UV-B (Silva et al., 2006; Dotterud et al., 2008). Another study covered the effect of antimicrobial photodynamic therapy (APDT) which involved killing of microbes using light along with a photosensitizer. Almost all species used in the study were susceptible to APDT in vitro. Killing of Staphylococcus epidermidis and Staphylococcus aureus was significantly higher with natural sunlight than with polychromatic visible light produced by a standard slide projector (Zeina et al., 2001). Intriguingly, PDT was found to be effective in antibiotic-resistant folliculitis (Horn and Wolf, 2007). Blue light treatment (Charakida et al., 2004; Noborio et al., 2007) and conventional UV phototherapy (Rassai et al., 2014) may act beneficially in acne vulgaris by altering the skin microbiome and reducing Propionibacterium acnes density. Indeed, UV-R is known to be bactericidal and can break lipopolysaccharides (LPSs), lipoteichoic acids (LTAs) and other bacterial metabolites which have immunomodulatory properties (Liu et al., 2010; Weill et al., 2013). Moreover, PDT light treatment may not only work by direct effects on microorganisms but also by modulating the immune response directed against them (Reginato et al., 2014). However, the exact effects of UV-R on the skin microbiome are largely unexplored and its effects on archaea or skin mites is least known. In depth studies are warranted to understand the effects of UV-R, as with the emerging evidence of the wide arrangement of skin microbiome with the host immune system.

#### Skin Microbiome-Immunity Dialog

The skin immune system and the microbiome have to be in constant communication in order to establish an equilibrium with each other. For this reason, it is important that the immune response is tailored to the appropriate threat, as any immune reaction toward commensals could lead to inflammatory

to microbial killing. At sub toxic levels, UVR may initiate a pathogen/damage-associated molecular pattern (PAMP/DAMP) response. Such a response may result in the expression of various microbial signals such as oleic acid, LPS and/or porphyrins, affecting the overall immune signaling cascade, leading to inflammation, and altered immune response. Microbial metabolites can also exert an effect on dendritic cells which can recognize or can be involved in direct capture of microbes. Moreover, microorganisms can produce natural AMPs directly or can control AMP production by keratinocytes and their production can be increased by exposure to UV-R. UVR-induced cis-UCA could not only contribute to induction of altered immune response but also indirectly change the microbial load, by affecting the microenvironment through yet unknown pathways. In addition, the microbiome can also induce complement and IL-1 set under stress by UV-R and together with directly induced microbial signals influence the skin immunity by induction of various cytokines such as that of Th17 pathway. Said so, keratinocyte effector function could be influenced by production of IL-17, leading in circle to altered AMP production, and release, in turn affecting the microbiome.

responses and subsequent disease (SanMiguel and Grice, 2015). To execute these functions, skin is equipped with specialized immune cells of innate and adaptive immune system. An acute damage to skin will lead to production of certain ligands, which can activate keratinocytes and eventually results in the release of inflammatory mediators. Under these conditions LPA, a product of Staphylococcus epidermidis can reduce inflammation and take part in wound healing by binding to Toll-like receptor-2 (TLR-2), which is one of the innate immune receptors (Lai et al., 2009). Recently it has been reported that microbial LPS immunogenicity can lead to autoimmunity in humans (Vatanen et al., 2016). Skin microbiome has the capacity to control the expression of a wide range of innate immune sensors, such as AMPs (Gallo and Hooper, 2012). Most of the AMPs are known be constitutively expressed in the skin, and they can also be triggered by defined microbiota like Propionibacterium species and other gram-positive bacteria (Nagy et al., 2006; Lee et al., 2008; Gallo and Hooper, 2012). The mechanisms by which AMPs shape the microbial landscape remain to be clarified, as AMPs are induced by both UV-R and certain microbes. The multidirectional interaction between AMPs, microbes, and UV-R could be a key factor for the ecosystem of the skin microbiome, and could provide insight into the physiology of healthy skin as well as pathophysiology various acute and chronic inflammatory disorders.

The skin microbiome also induces the expression of other conserved pathways of the host immune system, such as components of the complement systems which contain large amounts of proteins. These proteins can react with each other and can take part in opsonization of a pathogen and induce other inflammatory responses to clear the pathogens (Belkaid and Segre, 2014). Germ-free mice, which are raised in absence of microbes, are known that they have a decreased expression of the C5aR, a component of the complement system, which results in reduced expression of AMPs and other pro-inflammatory factors. These are the changes which are commonly associated with dysbiosis of the skin microbiome (Naik et al., 2012; Chehoud et al., 2013). Exposure of mice to UV-R is known to activate both the classical (Hammerberg et al., 1998) and alternative (Stapelberg et al., 2009) complement pathways in skin, an event that precipitates immune suppression. However, the impact UVactivation of complement has on commensals remains to be investigated.

The skin microbiome itself can also control the expression of IL-1, which is actively involved in initiating and amplifying immune response (Naik et al., 2012). Upregulation of innate

immunity by the skin microbiome in this way may lead to subsequent activation of the adaptive immune system. Indeed, the skin microbiome is known to functionally modulate T-cells by adjusting the local innate immune setting, especially via IL-1 production. This ultimately leads to increased production of proinflammatory cytokines such as IL-17A and interferon-γ (IFN- γ) by T cells.

Commensals may have evolved by distinctly controlling the network of the immune system depending upon environmental conditions. Since the skin has one of the largest pools of immune cells, and there is an enormous amount of pressure exerted by the skin microbiome, the cells of the innate immune system of the skin may commonly recognize the skin microbial antigens and prevent spread (Belkaid and Segre, 2014). Moreover, mice lacking adaptive immunity fail to recognize and restrain their commensal skin microbiota which leads to microbial diffusion to the local lymph nodes (Shen et al., 2014). In perspective with inflammation, various changes within the skin such as barrier permeability or increased contact with the resident microbiome can increase the local immune responses and with the ability of the skin microbiome to co-control both innate and adaptive immune response. Thus, the skin microbiome is likely the main driver and amplifier of various skin pathologies (**Figure 2**). Finally, DNA repair enzymes produced by evolutionary UVresistant microbes such us Micrococcus luteus colonizing the human skin (Tomlin et al., 1978) may be capable of not only repairing UV-induced damage to their own DNA but potentially transfer this help to other commensals and possibly even cells of the human skin. For instance, endonuclease T4N5 from the bacteriophage T4 encapsulated in multilamellar liposomes can penetrate human cells, is delivered to the nucleus and has been shown repairing UV-induced DNA damage in cell culture and human skin explants (Ceccoli et al., 1989; Yarosh et al., 1991; Gilchrest et al., 1993). Preparations with liposomes containing T4N5 have been extensively investigated in mice (Wolf et al., 1993, 1995; Yarosh et al., 1994) and have been employed in prevention studies of skin cancer in patients with the genetic disease Xeroderma pigmentosum and other individuals prone to multiple skin cancers (Wolf et al., 2000; Yarosh et al., 2001). Similarly, photolyase from Cyanobacteria and enzymes with DNA repair activity from Micrococcus luteus have been formulated into liposomes for use on humans' skin (Stege, 2001; Hofer et al., 2011). Such preparations have been administered in clinical studies and helped to modulate the UV-induced immune response, as described by its experimental use in patients with polymorphic light eruption (Hofer et al., 2011) and in turn may lead to changes in microbiome colonization.

#### PERSPECTIVE

It is likely that UV-R leads to various changes within the landscape of microbial communities of the skin. We hypothesize that exposure to UV-R leads to alterations in skin commensals leading to change in quantity and spread of certain, defined bacteria or growth of opportunistic microorganisms. This could disrupt the equilibrium with the host immune system and trigger a local immune response. For example, when skin is colonized with Staphylococcus aureus, this can lead to induction of a local response by producing δ-toxin, which is involved in mast cell degranulation and thus promotes innate and adaptive type 2 response, as seen in allergic skin disease (Nakamura et al., 2013). However, on the other side it is known that UV-R can reduce the growth of Staphylococcus aureus in vivo and in vitro (Jekler et al., 1992; Yoshimura et al., 1996; Thyssen et al., 2015) and also decrease the production of super antigens, which are known to be potential triggers of immune responses (Yoshimura-Mishima et al., 1999; Thyssen et al., 2015). UV-R can also directly breakdown cell structures of microbes, leading to production of various microbial signals (**Figure 2**). They can then be recognized by the immune system and reacted against appropriately. UV-R can induce pyrimidine dimers in DNA and can have a drastic effect on the microbial communities. Furthermore, microbial DNA photoproducts could even be a potential trigger of immune responses (Rothschild, 1999). New technologies such as high-throughput sequencing and the availability of germ-free animal models, will allow us to extend our knowledge of the relationship between UV-R, the skin microbiome and immunity. One could directly study the effects of UV-R on the microbiome by taking skin swabs from the human body and directly extract DNA for further analysis but also cultivating skin microbes obtained from skin swabs and exposing them in vitro to different doses of UV-R in order to study the response, damage, mutations etc. This will lead to a better understanding of various photosensitive conditions, give further insights into the mechanisms of medical phototherapy and elucidate the role of the skin microbiome and its potential modulation by UV-induced immune suppression (Bouslimani et al., 2015).

#### CONCLUSION

The skin microbiome plays an important role in developing and maintaining homeostasis and regulation of the host immune system. With increasing incidence of UV-induced skin conditions, the importance of the skin immune system in maintaining tolerance toward the resident microbiome may be crucial. Since the skin microbiome can interact with and co-control both the innate and adaptive immune system, it is of great importance to understand its relationship with the host immune system.

#### AUTHOR CONTRIBUTIONS

VP drafted the manuscript. PW and SB contributed to the draft. All authors revised and approved the final version of the manuscript.

#### ACKNOWLEDGMENTS

PhD student VP received funding from the Austrian Science Fund FWF (W1241) and the Medical University of Graz through the PhD Program Molecular Fundamentals of Inflammation (DK-MOLIN).

### REFERENCES

fmicb-07-01235 August 10, 2016 Time: 10:59 # 8



anti-bacterial effects. World J. Immunol. 4, 1–11. doi: 10.5411/wji. v4.i1.1



**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 the authors VP and PW, and states that the process nevertheless met the standards of a fair and objective review.

Copyright © 2016 Patra, Byrne and Wolf. 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) or licensor 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 Salivary Microbiome in Polycystic Ovary Syndrome (PCOS) and Its Association with Disease-Related Parameters: A Pilot Study

Lisa Lindheim<sup>1</sup> \*, Mina Bashir <sup>1</sup> , Julia Münzker <sup>1</sup> , Christian Trummer <sup>1</sup> , Verena Zachhuber <sup>1</sup> , Thomas R. Pieber 1, 2, Gregor Gorkiewicz 3, 4 and Barbara Obermayer-Pietsch1, 2

<sup>1</sup> Division of Endocrinology and Diabetology, Department of Internal Medicine, Medical University Graz, Graz, Austria, <sup>2</sup> Center for Biomarker Research in Medicine, Graz, Austria, <sup>3</sup> Institute of Pathology, Medical University Graz, Graz, Austria, <sup>4</sup> BioTechMed, Interuniversity Cooperation, Graz, Austria

#### Edited by:

Gabriele Berg, Graz University of Technology, Austria

#### Reviewed by:

Irene Wagner-Doebler, Helmholtz Centre for Infection Research, Germany Catherine Maree Burke, University of Technology, Sydney, Australia

> \*Correspondence: Lisa Lindheim lisa.lindheim@medunigraz.at

#### Specialty section:

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

Received: 02 April 2016 Accepted: 02 August 2016 Published: 25 August 2016

#### Citation:

Lindheim L, Bashir M, Münzker J, Trummer C, Zachhuber V, Pieber TR, Gorkiewicz G and Obermayer-Pietsch B (2016) The Salivary Microbiome in Polycystic Ovary Syndrome (PCOS) and Its Association with Disease-Related Parameters: A Pilot Study. Front. Microbiol. 7:1270. doi: 10.3389/fmicb.2016.01270 Background: Polycystic ovary syndrome (PCOS) is a common female endocrine condition of unclear etiology characterized by hyperandrogenism, oligo/amenorrhoea, and polycystic ovarian morphology. PCOS is often complicated by infertility, overweight/obesity, insulin resistance, and low-grade inflammation. The gut microbiome is known to contribute to several of these conditions. Recently, an association between stool and saliva microbiome community profiles was shown, making saliva a possible convenient, non-invasive sample type for detecting gut microbiome changes in systemic disease. In this study, we describe the saliva microbiome of PCOS patients and the association of microbiome features with PCOS-related parameters.

Methods: 16S rRNA gene amplicon sequencing was performed on saliva samples from 24 PCOS patients and 20 healthy controls. Data processing and microbiome analyses were conducted in mothur and QIIME. All study subjects were characterized regarding reproductive, metabolic, and inflammatory parameters.

Results: PCOS patients showed a decrease in bacteria from the phylum Actinobacteria and a borderline significant shift in bacterial community composition in unweighted UniFrac analysis. No differences between patients and controls were found in alpha diversity, weighted UniFrac analysis, or on other taxonomic levels. We found no association of saliva alpha diversity, beta diversity, or taxonomic composition with serum testosterone, oligo/amenorrhoea, overweight, insulin resistance, inflammatory markers, age, or diet.

Conclusions: In this pilot study, patients with PCOS showed a reduced salivary relative abundance of Actinobacteria. Reproductive and metabolic components of the syndrome were not associated with saliva microbiome parameters, indicating that the majority of between-subject variation in saliva microbiome profiles remains to be explained.

Keywords: polycystic ovary syndrome, sex steroids, human oral microbiome, next-generation sequencing, 16S rRNA, obesity, inflammation

## INTRODUCTION

Polycystic ovary syndrome (PCOS) is a common female endocrine condition affecting 6–18% of reproductive-age women and comprising the three primary symptoms hyperandrogenism, oligo/amenorrhoea, and polycystic ovarian morphology (Diamanti-Kandarakis et al., 1999; Asuncion et al., 2000; Azziz et al., 2004; March et al., 2010). In addition to reduced fertility, pregnancy complications, and cosmetic problems, women with PCOS are at risk to develop disorders of glucose and lipid metabolism, chronic low-grade inflammation, and the associated long-term complications (Barry et al., 2011; Escobar-Morreale et al., 2011; Lerchbaum et al., 2011, 2013; Wehr et al., 2011; Dumesic et al., 2015; Kollmann et al., 2015). Different criteria for the diagnosis of PCOS exist, leading to a wide variety of phenotypes ranging from mild to severe. Currently, the Endocrine Society recommends the use of the Rotterdam Criteria for the diagnosis of PCOS (Legro et al., 2013).

The etiology of PCOS is still unclear, although a multifactorial pathogenesis including genetic, lifestyle, and intrauterine factors has been suggested (Dumesic et al., 2015). Recent research in rodents and humans has implicated the gut microbiome in the pathogenesis of numerous diseases, including obesity, insulin resistance, and type 2 diabetes (Bäckhed et al., 2004; Turnbaugh et al., 2006, 2009; Vrieze et al., 2012). The majority of gut microbiome studies have investigated the distal digestive tract (e.g., fecal or cecal samples); however, cooling, transport, and DNA extraction methods from these sample types are non-standardized and known to cause substantial variation in sequencing results (Goodrich et al., 2014).

We have investigated the microbiome of the proximal digestive tract as a possible indicator of disease in PCOS. Saliva offers several advantages over stool as a sample material for microbiome studies. These are the non-invasive on-site collection with little or no discomfort to the patient, the possibility for immediate processing and/or freezing following collection to conserve bacterial community structures, and the use of defined, reproducible sample volumes for DNA extraction. It has recently been shown that saliva microbiome profiles correlate with those in the stool, despite the fact that the bacterial communities in the two locations differ greatly (Ding and Schloss, 2014). Therefore, saliva may be a useful alternative to stool as an indicator of bacterial dysbiosis in systemic disease.

To our knowledge, there are no published studies of either the fecal or saliva microbiome in patients with PCOS using a next-generation sequencing approach. PCOS patients exhibit an increased prevalence of gingivitis, which was found to be accompanied by changes in certain oral bacterial species, assessed by qPCR (Akcali et al., 2014). However, data on the global saliva microbiome in periodontally healthy PCOS patients compared to control women is lacking.

We performed a pilot study to describe the salivary microbiome in PCOS and to investigate the potential of specific taxa and measures of bacterial diversity to distinguish between women with PCOS and healthy women. Additionally, we investigated the association of diagnostic (serum testosterone, oligo/amenorrhoea) and common co-occurring (overweight, insulin resistance, inflammation) features of PCOS with saliva microbiome parameters. Finally, we addressed the role of age and diet as possible confounding factors in saliva microbiome studies.

### MATERIALS AND METHODS

### Study Cohort

Twenty-five women with PCOS and 25 hormonally healthy controls were recruited from the endocrinological outpatient clinic at the University Hospital Graz. PCOS was diagnosed according to the Rotterdam Criteria, requiring the presence of two out of three of the following criteria: clinical/biochemical hyperandrogenism, oligo-/anovulation, and polycystic ovaries (Rotterdam ESHRE/ASRM-Sponsored PCOS consensus Workshop Group, 2004). Clinical hyperandrogenism, based on the presence of hirsutism, was defined as a score of eight or higher in the modified Ferriman-Gallwey (FG) assessment (Yildiz et al., 2010). Biochemical hyperandrogenism was defined as above-normal values of one or several serum androgens. Oligo-/anovulation was defined as prolonged menstrual cycles (>35 days) or the absence of menstruation for at least 3 months. Polycystic ovarian morphology in a gynecological ultrasound was assessed based on medical history. Thyroid disorder, congenital adrenal hyperplasia, Cushing's syndrome, hyperprolactinemia, androgen-secreting tumors, and pregnancy were excluded by appropriate laboratory tests and clinical examination. Healthy controls did not meet any of the Rotterdam Criteria, with the following exceptions: isolated elevation of dehydroepiandrosterone sulfate (DHEAS) or androstenedione without other signs of PCOS (6 subjects) and long-standing hirsutism without hyperandrogenemia (1 subject). Exclusion criteria for both groups were pregnancy/lactation, menopause, use of antibiotics, hormonal contraceptives, or antidiabetic medication within the preceding 3 months, gastrointestinal or periodontal disease, active infections of any kind, a body mass index (BMI) <18, and smoking. All study participants were at least 18 years old and provided written informed consent. The study protocol was approved by the Ethics Committee at the Medical University Graz.

#### Sampling

Study visits took place in the morning after an overnight fast. Study subjects were instructed not to brush their teeth and to drink only water prior to saliva sampling. Saliva was collected in the mouth for several minutes and then voided into Sali-Tubes (DRG Diagnostics, Marburg, Germany). This process was repeated until the desired volume of 1–2 ml was reached. Saliva samples were immediately cooled on ice, flash-frozen in liquid nitrogen, and stored at −70◦C until further processing.

Anthropometric data were recorded and a baseline hormonal and metabolic assessment performed. Following the baseline blood sampling, a 2-h, 75 g oral glucose tolerance test (oGTT; Glucoral 75 Citron, Germania Pharmazeutika, Vienna, Austria) was performed and glucose and insulin were measured after 30, 60, and 120 min.

#### Laboratory Measurements

Estrone (E1), 17-estradiol (E2), total testosterone, androstenedione, dehydroepiandrosterone (DHEA), DHEAS, and dihydrotestosterone (DHT) were measured by liquid chromatography-tandem mass spectrometry at the Department of Clinical Chemistry at the University Hospital of South Manchester, Manchester, United Kingdom, as described by Keevil et al. (Chadwick et al., 2005; Owen et al., 2014, 2015; Münzker et al., 2015).

Insulin was measured by chemiluminescence immunoassay on the ADVIA Centaur XP (Roche, Rotkreuz, Switzerland). Anti-Muellerian hormone (AMH) was measured by chemiluminescence immunoassay on the Access2 (Beckman Coulter, Brea, USA). Luteinizing hormone (LH) and folliclestimulating hormone (FSH) were measured by ELISA (both DiaSource, Louvain-la-Neuve, Belgium). Sex hormone-binding globulin (SHBG) was measured by chemiluminescence immunoassay on the Cobas e411 (Roche). Total cholesterol, high-density lipoprotein-cholesterol (HDL), triglycerides, and glucose were measured by enzymatic colorimetric assay on the Cobas c module (Roche). Serum high-sensitivity C-reactive protein (hs-CRP) was measured by ELISA (BioVendor, Brno, Czech Republic). A total and differential blood count was performed on the XE-5000 Hematology Analyzer (Sysmex, Vienna, Austria).

#### Calculations and Definition of Terms

BMI was calculated as Weight (kg) (Height (m))<sup>2</sup> . Overweight was defined as a BMI ≥ 25. The homeostasis model assessment for insulin resistance (HOMA2-IR) index was calculated using the opensource software HOMA calculator V2.2.3 provided by the Diabetes Trial Unit, University of Oxford, UK (www.dtu.ox. ac.uk/homacalculator/, last accessed Dec 17, 2015). Insulin resistance was defined as a HOMA2-IR ≥ 2. The area under the curve (AUC) for glucose and insulin was calculated from the oGTT using the trapezoidal method in GraphPad Prism 5. Free androgen index (FAI) was calculated according to the formula 100× Total testosterone (nmol/l) SHBG (nmol/l) . Free testosterone and free DHT were calculated from total testosterone/DHT and SHBG according to Mazer et al. assuming a blood albumin concentration of 6.2µmol/l (Mazer, 2009).

A food frequency questionnaire designed by dieticians of the Clinical Medical Nutrition Therapy Unit, University Clinic Graz, was administered to assess the intake of major food groups. Based on the results of the questionnaire, study participants were categorized as consuming a high carbohydrate or high animal protein diet.

#### Next-Generation Sequencing

Total DNA was extracted from saliva samples using the MagNAPure LC DNA Isolation Kit III (Bacteria, Fungi) on the MagNA Pure Instrument (Roche, Rotkreuz, Switzerland). Saliva was thawed, vortexed, and 250µl saliva was added to 250µl bacteria lysis buffer in a sample tube containing MagNALyser Green Beads (1.4 mm diameter ceramic beads, Roche). Samples were homogenized in a MagNALyser Instrument (2 × 6000 rpm for 30 s, separated by 1 min cooling), treated with 25µl lysozyme (Roth, Karlsruhe, Germany) at 37◦C for 30 min, and then with 43.3µl proteinase K (Roche) at 60◦C for 1 h. Lysates were incubated at 95◦C for 10 min, cooled on ice for 5 min, and centrifuged for 5 min at full speed. DNA was isolated from 200µl lysate supernatant by the MagNAPure Instrument using the manufacturer's software and eluted in 100µl elution buffer. A PCR reaction was performed to amplify the V1-2 region of the bacterial 16S rRNA gene using the primers F27 (AGAGTTTGATCCTGGCTCAG) and R357 (CTGCTGCCTYCCGTA; Eurofins Genomics, Ebersberg, Germany) and the FastStart High Fidelity PCR System, dNTPack (Roche) with initial denaturation at 95◦C for 3 min followed by 28 cycles of denaturation at 95◦C for 45 s, annealing at 55◦C for 45 s, and extension at 72◦C for 1 min, one cycle of final extension at 72◦C for 7 min, and a final cooling step to 10◦C. Triplicates were pooled, checked on a 1% agarose gel, and 15µl of pooled PCR product were normalized according to manufacturer's instructions on a SequalPrep Normalization Plate (Life Technologies, Vienna, Austria). Fifteen microliters of the normalized PCR product were used as template for indexing PCR in a 50µl single reaction to introduce barcode sequences to each sample according to Kozich et al. (2013). Cycling conditions were initial denaturation at 95◦C for 3 min followed by eight cycles of denaturation at 95◦C for 45 s, annealing at 55◦C for 45 s, and extension at 72◦C for 1 min, one cycle of final extension at 72◦C for 7 min, and a final cooling step to 4◦C. After indexing, 5µl of each sample were pooled and 50µl of the unpurified library were loaded on a 1% agarose gel and purified from the gel with the Qiaquick Gel Extraction Kit (Qiagen, Hilden, Germany) according to manufacturer's instructions. The pool was quantified using QuantiFluor ONE dsDNA dye on a Quantus Fluorometer (Promega, Mannheim, Germany) according to manufacturer's instructions and visualized for size validation on a 2100 Bioanalyzer Instrument (Agilent Technologies, Santa Clara, USA) using a high sensitivity DNA assay according to manufacturer's instructions. The final 6 pM library containing all pooled samples was run with 20% PhiX and version 3, 600 cycles chemistry according to manufacturer's instructions on a MiSeq desktop sequencer (Illumina, Eindhoven, Netherlands). One negative control (250µl sterile PBS instead of saliva) was included in each MagNAPure run and subjected to the same procedures as samples. A mock community containing genomic DNA from 20 bacterial species was included to estimate PCR and sequencing errors (HM-782D, BEI Resources, Manassas, USA).

#### Sequencing Data Analysis

Raw reads were processed using the open-source software mothur V1.35.0 according to the protocol by Kozich et al. (April 2015), with the following adaptations: no maxlength was defined during the screening step, start (1046) and end (6426) positions were adapted to the V1-V2 region, and a difference of 3 bases was permitted during the precluster step (based on the recommendation by the authors to allow one mismatch per 100 bp; Kozich et al., 2013). Chimeric sequences were removed by UCHIME (Edgar et al., 2011). After removal of non-bacterial sequences, classified using the SILVA119 database

(www.arb-silva.de), the remaining sequences were degapped, deuniqued, split into individual samples, and formatted for use with the open-source software QIIME 1.8.0 (Caporaso et al., 2010). Open reference operational taxonomic unit (OTU) picking was performed in QIIME using UCLUST against the Greengenes 13.8 database (DeSantis et al., 2006). An OTU was defined as a group of sequences with a similarity of 97% or more. Based on the mock community sequencing results, a relative abundance cutoff of 0.1% was applied for subsequent analyses. Faith's phylogenetic diversity and the number of observed OTUs were used as metrics for alpha rarefaction, which was performed in QIIME. Principal coordinate analyses (PCoA) were based on unweighted and weighted UniFrac distances and calculated in QIIME (Lozupone and Knight, 2005). Taxa summaries were performed in QIIME. All samples were normalized to the sample with the lowest read count for alpha and beta diversity comparisons. For taxa comparisons, relative abundances based on all obtained reads were used.

Raw sequencing data are available in NCBI's short read archive (SRA) under the accession number SRP077213.

#### Statistical Analysis

Nonparametric student's t-tests using Monte Carlo permutations were used for alpha diversity comparisons, Mann–Whitney U-tests for taxa comparisons, and Adonis for category comparisons of distance matrices, all calculated in QIIME. Benjamini–Hochberg false discovery rate (FDR) correction was used to correct for multiple hypothesis testing where applicable.

All remaining statistical calculations were performed in IBM SPSS Statistics Version 22. Depending on the statistical distribution of the variable, unpaired t-tests or Mann–Whitney U-tests were used to compare groups. Fisher's Exact tests were used to compare categorical parameters. All data are expressed as median and interquartile range (IQR).

## RESULTS

### Study Subject Characteristics

All 50 subjects included in the study provided saliva samples. Three subjects were excluded from the control group due to previously undetected hyperandrogenemia (elevation of two or more androgens in fasted blood sample), two subjects were excluded due to smoking prior to sampling, and one subject was excluded due to a BMI < 18. The final analyses were performed with 20 healthy controls and 24 PCOS patients.

Laboratory, anthropometric, and patient history data are summarized in **Table 1**. Patients with PCOS had significantly higher total testosterone, androstenedione, and DHEA (p = 0.002, <0.001, and 0.015, respectively) and lower E2 (p < 0.001) levels than healthy controls, while no difference was found for DHEAS, DHT, and E1 (p = 0.073, 0.096, and 0.138, respectively). Calculated free DHT, free testosterone, and FAI were higher in the PCOS group (p < 0.001 for all). PCOS patients showed a characteristic dysregulation of FSH and LH secretion, with increased LH levels compared to controls (p = 0.035). Hirsutism and oligo/anmenorrhoea were more prevalent in the PCOS group (p = 0.003 and <0.001, respectively). Nearly all PCOS patients reported a history of polycystic ovaries (p < 0.001), which was corroborated by elevated AMH levels at the time of sampling (p < 0.001). An increased basal insulin secretion and AUCinsulin in the oGTT, elevated total triglycerides, and reduced HDL-cholesterol were observed in the PCOS group (p = 0.022, 0.009, 0.010, and 0.006, respectively). The studied cohort included lean as well as obese PCOS patients. Overall, BMI did not differ between PCOS patients and controls (p = 0.147). Total blood leukocytes were significantly higher in PCOS patients compared to healthy controls (p = 0.040), while hsCRP was not significantly different between the two groups (p = 0.078).

### Assessment of Sequencing Error and Bias Using a Mock Community

A mock community containing genomic DNA from twenty bacterial species, representing 17 genera, was included in the 16S rRNA PCR and sequencing to estimate OTU inflation and classification bias due to sequencing errors. After removal of singleton OTUs, we detected 214 OTUs from 29 genera in the mock community sample, indicating an overestimation of the number of OTUs due to sequencing errors (**Table 2**). After filtering the mock community and our dataset to 1, 0.1, and 0.01% relative abundance, we determined that a cutoff of 0.1% best represented the mock community, detecting 31 OTUs from 19 genera (**Supplementary Data Sheet 1**). We therefore performed the subsequent analysis using this abundance filter.

Using the 0.1% cutoff, all bacteria in the mock community were correctly classified at the family level, 15/17 at the genus level, and 7/20 at the species level (**Supplementary Data Sheet 1**). The observed relative abundance of most genera was within 50% of the expected value (**Table 2**). Bacteria from the genera Bacteroides and Helicobacter were more than two-fold overestimated, while bacteria from the family Gammaproteobacteria were more than two-fold underestimated (**Table 2**).

### The Saliva Microbiome Composition in PCOS and Its Association with Metabolic Dysfunction and Inflammation

16S rRNA amplicon-based microbiome analysis was performed on saliva samples from 20 healthy controls and 24 PCOS patients, using an OTU relative abundance cutoff of 0.1%. A median of 80,555 (IQR 18,509) and 72,284 (IQR 20,330) paired-end Illumina reads were analyzed per sample in the control and PCOS groups, respectively (p = 0.131). A total number of 131 OTUs [median(IQR) = 119.5(9.0) for controls and 116(8.5) for PCOS] from 35 genera [median(IQR) = 33(1.0) for controls and PCOS] were identified. As PCOS is often accompanied by overweight/obesity, insulin resistance, and chronic low-grade inflammation, we investigated the association of these features with saliva microbiome profiles. In addition, we performed analysis with samples grouped by diet and age, as these factors have been shown to influence gut microbiome composition (33–35).

The saliva microbiome was dominated by bacteria from the phylum Bacteroidetes (median relative abundance 45%) and

#### TABLE 1 | Study subject characteristics.


PCOS, polycystic ovary syndrome; IQR, interquartile range; AUC, area under the curve; HOMA2-IR, homeostasis model assessment for insulin resistance; HDL, high density lipoprotein; hsCRP, high-sensitivity C-reactive protein. Groups were compared using unpaired t-tests, Mann–Whitney U-tests, and Fisher's Exact tests. #according to the World Health Organization, † according to the American Diabetes Association, ‡ depending on menstrual cycle stage, § reference range not available. \*p < 0.05, \*\*p < 0.01, \*\*\*p < 0.001.

Firmicutes (26%), while bacteria from the phyla Proteobacteria, Fusobacteria, Actinobacteria, and TM7 contributed <10% each to total bacterial content (**Table 3**). On genus level, Prevotella was the single most abundant genus (median relative abundance 31%), followed by Streptococcus (11%), with other genera contributing <10% each to total bacterial content (**Table 3**). Saliva samples from PCOS patients showed a significant decrease in the relative abundance of bacteria from the phylum Actinobacteria (FDR p = 0.024). On class, order, family, genus, and OTU level, no differences were observed between patients and controls (**Table 3**, **Supplementary Data Sheet 2**).

The cumulative curve of observed genus level abundances followed a long-tailed distribution, with the ten most abundant genera accounting for 86% of all identified bacteria (**Figure 1**).

Faith's phylogenetic diversity and the number of observed OTUs did not differ between PCOS patients and controls (**Figure 2**). Additionally, serum testosterone and the presence of oligo-/amenorrhoea were not associated with a change in these parameters (**Figure 3**). Alpha diversity curves of individual samples showed excellent saturation both unrarefied and at the selected rarefaction level of 45,949 reads (**Supplementary Image 1**).

TABLE 2 | Expected and observed relative abundances of bacterial genera in a mock community.


Only sequences with a relative abundance >0.1% were included in the analysis. #Proteobacteria, † Actinobacteria, ‡ Firmicutes, §Bacteroidetes, ||[Thermi]. RA, relative abundance; unclass., unclassified; spp., species.

In beta diversity analyses, saliva samples showed a tendency toward a statistically significant clustering in unweighted UniFrac analysis (**Figure 2D**, p = 0.050). No clustering was observed when comparing weighted UniFrac distance matrices based on PCOS status, serum testosterone, or the presence of oligo-/amenorrhoea (**Figures 2C**, **3**).

Grouping samples based on overweight, insulin resistance, hsCRP, blood leukocytes, age, and diet did not affect alpha diversity, beta diversity, or taxonomic composition (data not shown).

#### DISCUSSION

To our knowledge, this is the first study reporting a nextgeneration sequencing-based profile of the saliva microbiome in PCOS patients. The phyla and genera that were found to dominate saliva microbiome profiles in our study cohort correspond to those previously reported for healthy adults (Keijser et al., 2008; De Filippis et al., 2014; Ding and Schloss, 2014). We show that PCOS is associated with a decreased relative abundance of salivary Actinobacteria and a borderline significant clustering of bacterial profiles in unweighted UniFrac analysis. This observation was not explained by individual components of the syndrome, namely hyperandrogenemia and oligo/amenorrhoea, or by associated features such as overweight, insulin resistance, and low-grade inflammation.

Actinobacteria, a phylum of gram-positive bacteria, are common members of the skin and oral microbiota and have been reported to be reduced in periodontal disease (Liu et al., TABLE 3 | Relative abundances of bacterial genera and phyla with a median relative abundance >1%.


Groups were compared using Mann–Whitney U-tests followed by Benjamini-Hochberg FDR correction. PCOS, polycystic ovary syndrome; IQR, interquartile range. #Bacteroidetes, † Firmicutes, ‡ Proteobacteria, §Fusobacteria, ||Actinobacteria, ¶TM7. Square brackets indicate a Greengenes suggested taxonomic assignment. \*p < 0.05.

2012; Wang et al., 2013). Akcali et al. used quantitative realtime polymerase chain reaction to show changes in several bacterial species in women with PCOS and gingivitis compared to healthy women with gingivitis (Akcali et al., 2014). However, the authors observed no change between periodontally healthy women with and without PCOS. As our PCOS patients were periodontally healthy and we did not observe an association between the saliva microbiome and markers of inflammation, we hypothesize that the reduction in the relative abundance of Actinobacteria within the context of PCOS does not itself cause disease, but rather provides a more favorable environment for pathology-associated bacteria, which can result in periodontal disease in the presence of other permissive factors. This hypothesis is supported by the fact that the prevalence of periodontal disease is higher in PCOS patients than in the general population (Porwal et al., 2014; Rahiminejad et al., 2015).

Salivary microbiome profiles of PCOS patients showed a borderline significant clustering in unweighted UniFrac analysis, while weighted UniFrac distance matrices and alpha

diversity metrics were not significantly different to controls. Several explanations exist for this apparent lack of pronounced differences. On average, the patients in our study displayed mild phenotypes of PCOS, in that serum androgens were only slightly elevated compared to controls, parameters related to glucose and lipid metabolism were within the normal range for most patients, and BMI was not significantly different from controls. We did not specifically recruit only lean or obese PCOS patients, as we were aiming for a broadly representative cohort of PCOS phenotypes. Shifts in the saliva microbiome may parallel the clinical phenotype, becoming more pronounced in the presence of severe hyperandrogenism and anovulation, either alone or in combination with obesity and/or manifest type 2 diabetes. Furthermore, the approach of this study provides only information on the presence of bacterial DNA, but not on bacterial function. Salivary bacteria may have altered gene expression patterns either in response or as a contributing factor to the biochemical changes observed in PCOS.

As it is known that the microbiome can be affected by many exogenous and endogenous factors (Goodrich et al., 2014), we addressed these possible confounders by either defining them as an exclusion criterion (such as smoking, antibiotics

FIGURE 2 | Alpha and beta diversity of saliva samples from PCOS patients and controls. (A,B) Alpha rarefaction curves of Faith's phylogenetic diversity (A) and the number of observed OTUs (B). Samples were rarefied to the smallest observed number of reads (45,949). Median and IQR are plotted. (C,D) Principal coordinate analysis (PCoA) plots of weighted (C) and unweighted (D) UniFrac distances. Each dot represents the bacterial community composition of one individual saliva sample. Groups were compared using Monte Carlo permutations for alpha diversity and Adonis for beta diversity. PCOS, polycystic ovary syndrome.

use, and periodontal disease) or by performing separate analyses for these variables (as for BMI, insulin resistance, inflammation, diet, and age). We did not find an association of any of these factors with alpha diversity, beta diversity, or changes in bacterial relative abundance on any taxonomic level.

Recent research has indicated that the effect of microbiome "confounders" may be less significant than previously assumed. Studies by De Filippis et al. and Belstrøm et al. have reported no effect of age and diet on saliva microbiome profiles (Belstrøm et al., 2014; De Filippis et al., 2014), while another study showed no effect of gender and BMI (Stahringer et al., 2012). Chen et al. investigated the effect of race, BMI, alcohol intake, sex, tobacco use, and age on the stool microbiome and found that each factor explained <1% of variability in stool microbiome profiles (Chen et al., 2016). These studies, together with our results, illustrate the large knowledge gaps that still exist about the factors shaping the microbiome, currently termed "inter-individual variation."

The Illumina approach which we selected is among those with the highest sequencing depth (Sims et al., 2014). Therefore, we do not expect a great improvement of taxonomic resolution with an even higher coverage. This could be achieved by increasing the length of the sequenced 16S rRNA gene fragment. Short read lengths are a limitation of Illumina paired-end sequencing. However, lower sequencing errors compared to pyrosequencing and IonTorrent led us to prefer this approach over one employing longer read lengths but lower quality. By including a mock community containing genomic DNA from twenty bacterial species in equal concentrations, we were able to assess PCR and sequencing bias. We found that the relative abundances of the genera Bacteroides and Helicobacter were overestimated, while the relative abundance of the family Gammaproteobacteria was underestimated by our employed sequencing approach. It should be noted that the bacterial community representation of saliva samples may deviate from this pattern, as DNA extraction method is also a known source of bias which was not assessed by the mock community (Goodrich et al., 2014).

The main strength of our study is the thorough characterization of our study cohort, which included an assessment of ovarian and adrenal androgens, lipid metabolism, and glucose tolerance. Furthermore, we applied strict exclusion criteria to ensure that no subjects with an undiagnosed mild form of PCOS or other hormonal imbalance were included in the control group and to eliminate factors which may influence the saliva microbiome, such as smoking, the use of antibiotics, and periodontal disease. A second strength is our sampling approach. We collected saliva samples after an overnight fast, avoiding a disturbance of the oral microbiome due to brushing teeth or using mouthwash. Samples were immediately frozen in liquid nitrogen to optimally conserve the bacterial community structure at the time of sampling. Finally, we used a mock community to evaluate the quality of the sequencing methodology and found that a large percentage of OTUs were most likely the result of sequencing errors. By removing these OTUs, we greatly improved the validity of our results. Since we thereby also removed a proportion of true sequences, we applied several filters at different relative abundance levels, as well as performing an analysis on unfiltered data, to attain a balanced interpretation of the bacterial composition of our samples. The significant result was obtained only when using the 0.1% filter, therefore it should be interpreted with caution until it can be replicated in a larger cohort of patients.

Weaknesses of our study are the small sample size, which precluded stratification of PCOS subtypes, and the paucity of extreme phenotypes. However, this pilot study was designed to represent a spectrum of typical Austrian PCOS phenotypes, allowing a first glimpse at the saliva microbiome in this common condition. Future studies should aim to recruit large patient and control groups and stratify based on different PCOS phenotypes, either based on Rotterdam vs. NIH diagnostic criteria or as described by Jamil et al. (Dumesic et al., 2007; Jamil et al., 2015). The addition of a "positive control," such as patients with periodontal disease, would underscore the presence or lack of differences due to PCOS alone and confirm that our result was not obtained due to technical shortcomings of the employed method.

While the saliva microbiome appears to be only minorly changed in PCOS, the microbiome of other body areas may play a more significant role in this pathology. Two research groups have recently shown an alteration in the fecal microbiome of two different rodent models of PCOS (Guo et al., 2016; Kelley et al., 2016). Furthermore, bacterial colonization of the vagina and ovarian follicles was found to affect the outcome of assisted reproductive treatment in women with infertility of various etiologies, including PCOS (Pelzer et al., 2011). Next-generation sequencing of samples from these body sites in PCOS patients presents an interesting approach for future studies.

In conclusion, we present a first report of the saliva microbiome composition in PCOS. In our cohort, PCOS patients showed a reduced relative abundance of bacteria from the Actinobacteria phylum, while bacterial community composition and diversity seems to be independent of the reproductive and metabolic abnormalities observed in these patients. Larger studies with stratification of PCOS phenotypes are needed to clarify the presence or absence of microbiome changes due to different components of the syndrome. Bacterial functionality, assessed by metagenomics and metatranscriptomics, can provide further insights into the role of salivary bacteria in this condition.

## AUTHOR CONTRIBUTIONS

LL: study design, patient recruitment, data collection, laboratory analyses, data analysis, interpretation of results, drafting of manuscript; MB: sequencing data analysis, interpretation of results; JM: coordination of steroid hormone measurements, laboratory analyses, interpretation of results; CT: patient recruitment; VZ: laboratory analyses; TP: interpretation of results; GG: study design, interpretation of results; BO: study design, interpretation of results, study supervision. All authors were involved in the revision and approved the final version of the manuscript.

### FUNDING

This work was funded by the DK-MOLIN (Austrian Science Fund (FWF) W1241) and the Medical University of Graz. The funding source was not involved in the study design, data collection, analysis, or interpretation, drafting of, and decision to publish the manuscript.

#### ACKNOWLEDGMENTS

The authors would like to thank Roswitha Gumpold, Cornelia Missbrenner, and Hannelore Pock for contributions to sample collection and management, Ingeborg Klymiuk and the CF-MB at the ZMF Graz for advice on molecular biology techniques, Slave Trajanoski and Andrea Groselj-Strele for bioinformatics and statistics, Daniela Hofer, Matthias Ulbing, Olivia Trummer, and Christine Moissl-Eichinger for scientific discussion, and all study participants for their generous donations. The following reagent was obtained through BEI Resources, NIAID, NIH as part of the Human Microbiome Project: Genomic DNA from Microbial Mock Community B (Even,

### REFERENCES


Low Concentration), v5.1L, for 16S rRNA Gene Sequencing, HM-782D.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01270

Supplementary Data Sheet 1 | Mock community sequencing results.

Expected and observed classification and relative abundances of a mock community at genus and OTU levels using singleton removal and relative abundance cutoffs at 0.1 and 0.01%. Square brackets indicate a Greengenes suggested taxonomic assignment.

Supplementary Data Sheet 2 | Taxa comparisons of saliva samples from PCOS patients and healthy controls. Detected taxa and group comparisons on phylum, class, order, family, genus, and OTU level. FDR, Benjamini-Hochberg false discovery rate correction for multiple testing. Mean relative abundances are shown. Square brackets indicate a Greengenes suggested taxonomic assignment.

Supplementary Image 1 | Alpha diversity of individual saliva samples.

Faith's phylogenetic diversity (A, PD\_whole\_tree) and the number of observed OTUs (B, observed\_species) of individual saliva samples plotted against the number of reads analyzed.

workbench compatible with ARB. Appl. Environ. Microbiol. 72, 5069–5072. doi: 10.1128/AEM.03006-05


with PCOS: comparison of different diagnostic definitions. Hum. Reprod. 30, 2396–2403. doi: 10.1093/humrep/dev187


ovary syndrome with and without medical treatment. J. Periodontol. 85, 1380–1389. doi: 10.1902/jop.2014.130756


**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 © 2016 Lindheim, Bashir, Münzker, Trummer, Zachhuber, Pieber, Gorkiewicz and Obermayer-Pietsch. 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) or licensor 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 Shifts May Trigger Dysbiosis and Mucous Stools in Giant Pandas (Ailuropoda melanoleuca)

Candace L. Williams1,2†‡, Kimberly A. Dill-McFarland<sup>3</sup>‡ , Michael W. Vandewege<sup>1</sup> , Darrell L. Sparks1,4, Scott T. Willard<sup>1</sup> , Andrew J. Kouba<sup>5</sup>† , Garret Suen<sup>3</sup> \* and Ashli E. Brown1,2,4 \*

#### Edited by:

Nicole Webster, Australian Institute of Marine Science, Australia

#### Reviewed by:

John Everett Parkinson, University of the Ryukyus, Japan Jeffrey David Galley, Baylor College of Medicine, USA

\*Correspondence:

Garret Suen gsuen@wisc.edu; Ashli E. Brown abrown@mscl.msstate.edu

#### †Present address:

Candace L. Williams, Institute for Conservation Research, San Diego Zoo Global, Escondido, CA, USA; Andrew J. Kouba, Department of Wildlife, Fisheries and Aquaculture, Mississippi State University, Mississippi State, MS, USA

> ‡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: 07 December 2015 Accepted: 21 April 2016 Published: 06 May 2016

#### Citation:

Williams CL, Dill-McFarland KA, Vandewege MW, Sparks DL, Willard ST, Kouba AJ, Suen G and Brown AE (2016) Dietary Shifts May Trigger Dysbiosis and Mucous Stools in Giant Pandas (Ailuropoda melanoleuca). Front. Microbiol. 7:661. doi: 10.3389/fmicb.2016.00661 <sup>1</sup> Department of Biochemistry, Molecular Biology, Entomology and Plant Pathology, Mississippi State University, Mississippi State, Mississippi, MS, USA, <sup>2</sup> Institute for Genomics, Biocomputing and Biotechnology, Mississippi State University, Mississippi State, Mississippi, MS, USA, <sup>3</sup> Department of Bacteriology, University of Wisconsin-Madison, Madison, WI, USA, <sup>4</sup> Mississippi State Chemical Laboratory, Mississippi State, Mississippi, MS, USA, <sup>5</sup> Department of Conservation and Research, Memphis Zoological Society, Memphis, TN, USA

Dietary shifts can result in changes to the gastrointestinal tract (GIT) microbiota, leading to negative outcomes for the host, including inflammation. Giant pandas (Ailuropoda melanoleuca) are physiologically classified as carnivores; however, they consume an herbivorous diet with dramatic seasonal dietary shifts and episodes of chronic GIT distress with symptoms including abdominal pain, loss of appetite and the excretion of mucous stools (mucoids). These episodes adversely affect the overall nutritional and health status of giant pandas. Here, we examined the fecal microbiota of two giant pandas' non-mucoid and mucoid stools and compared these to samples from a previous winter season that had historically few mucoid episodes. To identify the microbiota present, we isolated and sequenced the 16S rRNA using next-generation sequencing. Mucoids occurred following a seasonal feeding switch from predominately bamboo culm (stalk) to leaves. All fecal samples displayed low diversity and were dominated by bacteria in the phyla Firmicutes and to a lesser extent, Proteobacteria. Fecal samples immediately prior to mucoid episodes had lower microbial diversity as compared to mucoids. Mucoids were mostly comprised of common mucosalassociated taxa including Streptococcus and Leuconostoc species, and exhibited increased abundance for bacteria in the family Pasteurellaceae. Taken together, these findings indicate that mucoids may represent an expulsion of the mucosal lining that is driven by changes in diet. We suggest that these occurrences serve to reset their GIT microbiota following changes in bamboo part preference, as giant pandas have retained a carnivorous GIT anatomy while shifting to an herbivorous diet.

Keywords: 16S rRNA sequencing, mucoid, fecal microbiota, mucosal microbiota, bamboo part preference

### INTRODUCTION

The host-symbiont relationship within the gastrointestinal tract (GIT) of animals is critical, as these symbionts play a fundamental role in fiber digestion, modulation of the host immune system, and maintenance of host-bacterial homeostasis (Hooper et al., 2002; Flint et al., 2012). In particular, microorganisms associate with the gastrointestinal lymphoid tissue to exclude pathogens and produce short chain fatty acids (SCFAs) that serve as an energy source for the host

(Johansson et al., 2011; Flint et al., 2012). Also, SCFAs cause the intestinal epithelial cells (IEC) to increase expression of tight junction proteins, thus further increasing the barrier to pathogens (Brown et al., 2003; Louis and Flint, 2009).

The GIT biology of giant pandas (Ailuropoda melanoleuca) is peculiar because they are evolutionarily related to carnivores and possess the GIT morphology of a carnivore, yet they consume an exclusively herbivorous diet. This feature is surprising, given that the switch from an omnivorous to an herbivorous diet occurred approximately 2 to 2.4 million years ago (Davis, 1964; Schaller et al., 1985; Jin et al., 2007), yet giant pandas have not evolved adaptations seen in traditional herbivores, like a rumen or an enlarged cecum, to aid in fiber degradation. It remains unclear how pandas persist solely on bamboo, as they consume large amounts of the fibrous plant (Schaller et al., 1985), relative to other herbivores of their size. However, it has been suggested that they rely on bamboo's hemicellulose content, rather than more difficult to digest cell wall components such as lignin and cellulose (Dierenfeld et al., 1982).

Both wild and captive pandas annually undergo dramatic shifts in bamboo part preference between culm (stalk) and leaves (Schaller et al., 1985; Tarou et al., 2005; Hansen et al., 2010; Williams et al., 2012) resulting in significant changes in fecal consistency and GIT microbial communities (Nickley, 2001; Williams et al., 2012; Xue et al., 2015). These dietary shifts have been attributed to changes in bamboo composition, as Schaller et al. (1985) found levels of silica to increase in the leaf portion of bamboo during times when pandas preferred the culm portion. This increase in silica content has been associated with anti-herbivory defense pathways in plants, which may explain why pandas undergo such a dramatic change in diet preference (Schaller et al., 1985; Ito and Sakai, 2009).

These endangered bears also suffer greatly from GIT disorders both ex situ and in situ (Qiu and Mainka, 1993; Loeffler et al., 2006). In humans, when the host-gut microbe relationship is severely disturbed, a condition termed dysbiosis can occur, and the host can experience an inflammatory response; if unchecked, this can develop into a chronic condition (Fava and Danese, 2011). Similarly, captive giant pandas undergo chronic GIT distress, with bouts of abdominal discomfort and loss of appetite, resulting in the excretion of a mucous-like stool (mucoid), although no investigation into their composition has occurred to date (Edwards et al., 2006; Loeffler et al., 2006). Necropsies from pandas that chronically suffer from this condition often show evidence of ulcerative and necrotizing suppurative colitis (Loeffler et al., 2006).

While mucoid occurrence has been associated with the presence of some pathogenic microorganisms (Loeffler et al., 2006), a direct link between specific pathogens and mucoids has not been found. Increases in dietary protein are known to result in greater occurrences of mucoids (Edwards et al., 2006; Janssen et al., 2006), suggesting that diet may be the underlying cause. However, captive giant pandas fed a highfiber bamboo diet still commonly experience mucoids, so the cause and means to prevent these episodes remains unclear. The timing of mucoids is also critical, as they typically occur during a seasonal dietary shift directly following the breeding season, and any decreased nutritional status during gestation or lactation may affect offspring (Steinman et al., 2006; Zhang et al., 2006). Here, we used next-generation sequencing to characterize the fecal- and mucoid-associated microbiota in two giant pandas and to determine if drastic changes in diet correlate to a concomitant shift in the GIT microbiota and the expulsion of mucoids. Comparison of the bacterial communities associated with mucoid episodes (mucoid) to fecal samples, both within (non-mucoid) and outside the sample season (winter), provides insights into possible microbial contributions to this important chronic ailment in giant pandas.

### MATERIALS AND METHODS

#### Study Animals

The two giant pandas ("YaYa," female, studbook number: 507, and "LeLe," male, studbook number: 466) used in this study were housed at the Memphis Zoological Society, Memphis, TN, USA. Samples were collected under a signed biomaterials request form, and no IACUC protocol was needed as this project was viewed as non-invasive by the institution.

### Behavior Analysis of Bamboo Consumption

The study of bamboo consumption behavior at the Memphis Zoo has been ongoing since the fall of 2003 and was conducted as previously described (Hansen et al., 2010; Williams et al., 2012). In brief, behavior data were collected in 20-min periods in 30 s increments while the bear was feeding on bamboo using an ethogram focusing on foraging behaviors. These behaviors were divided into three consumption categories: leaf, culm (stalk), and other (shoot or branch). For each month, the total consumption behaviors were quantified by time spent consuming specific parts and each individual's behavior was expressed as a percentage of the total consumption behaviors.

### Sample Collection

Fresh fecal (n = 5 female, 13 male) and mucous excretion (n = 1, 5) samples were collected. All samples were transported on dry ice, and stored at −80◦C prior to processing. Samples were classified as "winter," "non-mucoid," or "mucoid" (Supplementary Table S1). Winter control samples (n = 5,5) were collected on 02/12/13, during a season with historically low mucoid occurrence and prior to first mucoid excretion in this study. Additional sample collection occurred between 6/29/14 and 8/22/14. During this period, collected male and female stool samples were categorized into non-mucoid or mucoid movements. The date of the movement was also recorded to study temporal changes. Of note, the male produced a mucoid sample on 07/17 (day 14) that was not successfully sequenced but a non-mucoid fecal sample on this date was.

#### DNA Extraction

Total genomic DNA from fecal samples was extracted via mechanical disruption and hot/cold phenol extraction following

the protocol described by Stevenson and Weimer (2007) with the following modification: 25:24:1 phenol:chloroform:isoamyl alcohol was used in place of phenol:chloroform at all steps. DNA was quantified using a Qubit Fluorometer (Invitrogen) and stored at −20◦C following extraction.

#### Library Preparation and Sequencing

Library preparation was carried out following manufacturer's recommendations (Illumina, 2013) with some modifications. In brief, an amplicon PCR targeted the V3–V4 region of the 16S rRNA gene using a forward (V3-4F, TCGTCGGCAGCGT CAGATGT GTATAAGAGACAGCCTACGGGNGGCWGCAG) and reverse (V3-4R, GTCTCGTGGGCTCGGAGATGTGT ATAAGAGACAGGC TACHVGGGTATCTAATCC) primer (Klindworth et al., 2013) in a 25-µL reaction with 1X KAPA HiFi Hot Start Ready Mix (Kapa Biosystems), 0.2 mM each primer, and 1–10 ng DNA. Amplification conditions were as follows: 95◦C for 3 min, 25 cycles of 95◦C for 30 s, 55◦C for 30 s, 72◦C for 30 s, and a final elongation of 72◦C for 5 min. PCR products were purified via gel extraction (Zymo Gel DNA Recovery Kit; Zymo, Irvine, CA, USA, USA) using a 1% low melt agarose gel (National Diagnostics, Atlanta, GA, USA). Purified products underwent an indexing 25 µL-PCR reaction (1x KAPA HiFi Hot Start Ready Mix, 0.2 mM indices, and 5 µL of purified product) with the same reaction conditions as amplicon PCR with the exception of a reduction in the number of cycles to 8.

The final index PCR product underwent gel extraction (Zymo Gel DNA Recovery Kit; Zymo, Irvine, CA, USA), and the resulting purified product concentration was determined by a Qubit Fluorometer (Invitrogen). Samples were combined to yield an equimolar 4 nM pool. Following manufacturer's protocol, sequencing was conducted on an Illumina MiSeq using reagent kit V3 (2 x 300 bp cycles), as described previously (Illumina, 2013). All sequences were deposited into the National Center for Biotechnological Information's Short Read Archive under Accession Number SRP065974.

### Data Analysis

Sequence analysis was carried out using mothur v.1.34.1 following the MiSeq SOP (Kozich et al., 2013). In brief, contigs were formed from 16S rRNA reads, and poor quality sequences were removed. Sequences were trimmed and filtered based on quality (maxambig = 0, minlength = 250, maxlength = 600). Unique sequences were aligned against the SILVA 16S rRNA gene alignment database (Pruesse et al., 2007) and classified with a bootstrap value cutoff of 80, and operational taxonomic units (OTUs) found with <2 sequences in the total dataset were removed. Chimeras (chimera.uchime) and sequences identified as members of Eukaryota, Archaea, and Cyanobacteria lineages were also removed.

#### Statistical Analyses

Sequence coverage was assessed in mothur by rarefaction curves and Good's coverage (Good, 1953). Samples were then iteratively subsampled 10 times to 600 sequences per sample, and OTU abundances were calculated as the whole-number means across iterations. Differences in bacterial community were visualized by non-metric dimensional scaling plots (nMDS, iters = 10,000; Shepard, 1966) of Bray-Curtis (Bray and Curtis, 1957) and Jaccard (Jaccard, 1912) similarity (beta-diversity) indices, also calculated in mothur.

All other statistical analyses were carried out in R [vegan package (Oksanen et al., 2015; R Core Team, 2015)] or SAS 9.3 software (Cary, NC, USA), and data were expressed as the mean ± SEM and considered significant if P < 0.05. In R, differences in taxonomic profiles were assessed at the phyla, family, and OTU levels. Due to uneven sampling, analysis of similarity (ANOSIM) was used to compare community structure (Bray–Curtis; Bray and Curtis, 1957) and community composition (Jaccard; Jaccard, 1912) of winter, non-mucoid, and mucoid sample types. Samples were randomized with respect to sample type and tested to ensure true significance. Similarity percentages (SIMPER) analyses were then used to determine the contributions of taxonomic groups to differences observed in the ANOSIM. In SAS, the general linearized model (PROC GLM) was used to determine if diversity differed with respect to sample type.

### RESULTS

### Bamboo Consumption Behavior

Dramatic shifts in eating behavior were observed in both pandas (**Figure 1**). In general, the bears consumed more culm than leaf throughout the year, but shifted to higher proportions of leaf consumption for the months of August and September. The pandas consumed negligible amounts of leaf material in May (0.88%) and increased their leaf consumption to its highest relative proportion in August, (59%) around the time of mucoid sampling in this study. Following this peak, leaf consumption steadily declined through December (**Figure 1**).

#### Sequence Coverage and Taxonomy

For all samples (n = 34), a total of 457,358 raw and 375,406 high-quality (11,376 ± 2,170 sequences per sample) 16S rRNA sequences were generated using Illumina MiSeq paired-end sequencing (Supplementary Table S2). A Good's coverage value of >0.99 (Supplementary Table S2) and a leveling off of rarefaction curves (Supplementary Figure S1) indicated that sequencing was adequate to detect the majority of bacterial diversity present in all samples. A 97% OTU analysis corresponding to specieslevel classification (Schloss and Handelsman, 2005) identified 118 unique OTUs across all samples with 14 to 84 OTUs per sample type (Supplementary Table S2).

Sequences from 15 phyla were found across all samples, with 70 ± 5.8% belonging to the Firmicutes and 28 ± 5.7% belonging to the Proteobacteria (Supplementary Figure S2). All other phyla represented less than 1.0% relative sequence abundance. Bacterial classes with >1.0% included the Clostridia (40 ± 4.9%), Gammaproteobacteria (27 ± 5.7%), Erysipelotrichia (16 ± 2.9%), and Bacilli (15 ± 3.2, Supplementary Figure S2). Orders with >1.0% representation corresponded to the Clostridiales (30 ± 4.9%), Enterobacteriales (25 ± 5.7%), Erysipelotrichales (16 ± 2.9%), Lactobacillales (15 ± 3.2%), and Pasteurellales (1.3 ± 0.65%, Supplementary Figure S2). At the family and genus levels, 99 and 98% of the sequences were annotated, respectively.

### Sample Type Affects Overall Bacterial Diversity

Mucous stools (mucoid) and fecal samples (non-mucoid) were obtained from the same season and compared to fecal samples from a historically low-mucoid season (winter). Sample diversity varied over the sampling period for both male and female giant pandas (**Figure 2**). In particular, both male and female displayed higher Shannon's diversity than winter samples. At the beginning of mucoid season sampling diversity decreased dramatically prior to the appearance of the first mucoid (**Figure 2**). Overall, mucoid samples from both pandas displayed higher diversity, as measured by indices taking into account both presence and abundance of all taxa in the sample (Shannon: 1.7 ± 0.26, inverse-Simpson: 4.0 ± 1.0), than winter and non-mucoid fecal samples (Shannon, ANOVA, P = 0.0166). Although not significant, these samples were less dominated by single OTUs (inverse Berger-Parker: 2.6 ± 0.56, P > 0.05, **Table 1**). Non-mucoid fecal samples displayed the lowest average diversity with the highest variation (Shannon: 1.1 ± 0.13, inverse-Simpson: 2.6 ± 0.32) of all sample types, and were dominated by a single OTU (inverse Berger-Parker: 1.9 ± 0.20; Supplementary Table S2).

#### Overall Fecal Communities Differ According to Sample Type

The male and female samples were grouped by sample type (winter, non-mucoid, and mucoid), and total bacterial community structure (Bray–Curtis) and composition (Jaccard) within the winter and non-mucoid groups, with sample types tested both individually and combined, did not significantly differ by animal (ANOSIM, P > 0.05, Supplementary Tables S3 and S5). The mucoid group could not be tested as the female

had only one mucoid sample. Fecal communities were found to differ by sample type, as statistical analysis revealed differences in community structure (Bray–Curtis) at the phyla (ANOSIM, P = 0.035), family (P = 0.00030) and OTU levels (P = 0.00040) (Supplementary Table S4). Community composition (Jaccard) was also found to vary significantly with respect to sample type across all three taxonomic levels (ANOSIM, P = 0.040, 0.0007, and 0.00040, respectively; Supplementary Table S4). These differences in overall bacterial community composition and structure were visualized by non-metric dimensional scaling (nMDS; **Figure 3**; Supplementary Figure S3). No significant differences with respect to sample type were observed when randomized (ANOSIM, P > 0.05; Supplementary Table S4).

are shaded in gray. The male panda experienced a mucoid on day 14 but this

sample was not successfully sequenced and is not included here.

### Few Taxonomic Groups Shape Overall Bacterial Community

To determine which taxonomic groups contributed to the significant differences observed between sample types, analyses at the phyla, family and OTU levels were conducted. Only two phyla, the Proteobacteria and the Firmicutes, were found to drive differences between the three sample types [SIMPER, contribution to overall dissimilarity: winternon-mucoid comparison (WN): 50 and 49%, respectively; winter-mucoid comparison (WM): 40 and 49%, respectively; non-mucoid-mucoid comparison (NM): 45 and 46%)] (**Table 1**). Family members of these phyla also contributed to the differences


observed, with five families found to be important in sample comparisons. For WN, the Enterobacteriaceae, Clostridiaceae, and Erysipelotrichaceae, were found to be important drivers (SIMPER, contribution to overall dissimilarity: 31, 28, and 18%, respectively; **Table 1**). These three families, with the addition of the Streptococcaceae, were found to significantly shape differences in WM comparisons (**Table 1**). An additional family, the Leuconostocaceae, was also observed to significantly drive differences in NM bacterial communities (**Table 1**).

Of the 118 OTUs observed in the samples, only six were found to significantly contribute to the differences seen in the sample types (**Figure 4**). Three OTUs, an Escherichia-Shigella species (OTU 2), a Clostridium species (OTU 1), and a Turicibacter species (OTU 3), were influential in shaping differences in the WN comparison (SIMPER, contribution to overall dissimilarity: 31, 28, and 18%, respectively; **Table 1**). These, as well as an additional two OTUs, contributed to differences observed in BM: a Streptococcus species (OTU 12) and an unclassified member of the Pasteurellaceae (OTU 11; **Table 1**). For the NM analysis, all of the previously observed OTUs, except OTU 11, contributed to differences as well as a Streptococcus species (OTU 4) and a Leuconostoc species (OTU 6) (**Table 1**).

#### DISCUSSION

Here, we characterized the bacterial microbiota associated with fecal and mucous stools in giant pandas and correlated these communities to feeding shifts to determine the dietary and microbial contributions to mucoid episodes, a chronic and detrimental condition among these herbivorous carnivores. Fecal samples in this study were grouped as feces from a non-mucoid season (winter), feces immediately preceding or following a mucoid episode (non-mucoid), or mucus stools (mucoid).

Consistent with previous reports (Zhu et al., 2011; Xue et al., 2015), all fecal samples had low diversity (**Table 1**) and were dominated by bacteria in the phyla Firmicutes and Proteobacteria, with substantial contributions from the genera Clostridium, Escherichia-Shigella, Streptococcus and Turicibacter. Mucoid season samples (non-mucoid and mucoid) were obtained from a predominately leaf-eating season and tended to have less abundant Clostridium species and more abundant members of the family Enterobacteriaceae, particularly Escherichia-Shigella. Similar seasonal trends were observed previously with the same animals using culture techniques (Williams et al., 2012). However, the opposite trends were previously reported using next-generation sequencing of samples across a dietary change to more leaf for both wild and captive giant pandas (Xue et al., 2015). This discrepancy could be due to dietary differences as the pandas in the Xue et al. (2015) study consumed a different species of bamboo, as well as steamed bread, throughout the study. Importantly, it was not noted that consumption of bread varied with bamboo portion preference. Also, a wide range of methodological differences between the two studies, such as fecal sample processing (blending vs. vortexing), DNA extraction, or primer bias, may account for the observed differences (Brooks et al., 2015; Wagner Mackenzie et al., 2015).

FIGURE 3 | Three-dimensional non-metric multidimensional scaling analysis showing differences in (A) community structure (Bray–Curtis, lowest stress: 0.0810, R-square: 0.965) and (B) community composition (Jaccard, lowest stress: 0.192, R-square: 0.766) of winter, non-mucoid, and mucoid samples in giant pandas.

When examining phyla-level contributions to sample type differences, there is a shift in the Gram-positive to Gramnegative ratio due to changes in the relative abundances of the phyla Firmicutes and Proteobacteria. The Firmicutes (Grampositive) are commonly considered protective commensals (Craven et al., 2012) and dominated winter samples, but were less abundant in non-mucoid and mucoid samples. Inversely, the Proteobacteria (Gram-negative) include aggressive, pathogenic species (Baumgart et al., 2007; Craven et al., 2012), and they were found to be increased in the non-mucoid and mucoid samples, relative to winter. The largest changes in both phyla were observed between the winter and non-mucoid samples, indicating that these giant pandas experiences significant changes within their microbiota between seasons, possibly as a result of

differences in diet throughout the year. Such extreme changes, as well as the higher variability seen among non-mucoid fecal samples during the mucoid season (**Figure 2**), are often indicative of a dysbiosis between the host and its microbiota (Frank et al., 2007) and a lack of this dysbiosis could explain why the winter season has historically low mucoids.

We also found that the Proteobacteria Escherichia-Shigella species (OTU 2) underwent a dramatic increase from winter (0.22 ± 0.12%) to non-mucoid samples (43 ± 8.2%), and the phylum Actinobacteria were absent in winter but present in mucoids. Increases in the abundances of these bacteria are indicative of a dysbiotic event (Darfeuille-Michaud et al., 1998; Baumgart et al., 2007; Krogius-Kurikka et al., 2009; Craven et al., 2012), particularly in humans with inflammatory bowel disease (IBD), a condition that most similarly reflects the chronic mucoids suffered by giant pandas. Therefore, we hypothesize that diet-induced dysbiosis between the giant panda and its gut microbiota, may trigger mucoid episodes similar to dysbiosistriggering IBD symptoms in humans.

In addition to differences between winter and non-mucoid samples, mucoids were characterized by a number of unique taxa, indicating their divergence from other seasonal changes in the giant panda's fecal microbiota. Specifically, mucoids contained intermediate abundances of the Firmicutes and Proteobacteria, specifically Clostridium (OTU 1), Turicibacter (OTU 3), and Escherichia-Shigella (OTU 2). This presence indicates that mucoids differ from feces produced in the days surrounding mucoid episodes, and these bacteria may be members of the giant panda GIT microbiota that are shed during these events. Additionally, some differences between the mucoid and both non-mucoid and winter fecal samples were similar to differences between the mucosa- and fecal-associated microbial communities found in other animals (Zoetendal et al., 2002; Malmuthuge et al., 2013). Though not significant, mucoids had a higher abundance of the phyla Bacteroidetes, particularly the class Flavobacteria, and are these taxa are more commonly found associated with the mammalian mucosal lining than with fecal material (Jakobsson et al., 2015). Furthermore, a single unclassified OTU in the family Pasteurellaceae (OTU 11) and a Streptococcus (OTU 12) were found only in mucoids, and these taxa are known mucosa-associated bacteria in other animals (Kuhnert and Christensen, 2008; Kaci et al., 2014). Taken together, we speculate that mucoids are a combination of excreted mucosa, along with continued excretion of fecal material following the panda's switch to leaves.

Although mucoids have been observed throughout the year in captive pandas, they are more frequently observed in the summer months following decreases in dietary fiber (Nickley, 2001; Williams, 2011). A similar relationship has been observed in goats, where decreased fiber intake resulted in changes in the bacterial community and fermentation, leading to a decreased caecal pH and an increased lipopolysaccharide concentration. These alterations to mucosal morphology were associated with intense epithelial damage and local inflammation as a result of dietary change (Metzler-Zebeli et al., 2013; Liu et al., 2014). Given that diet also changed for the giant pandas in our study prior to mucoid collection (**Figure 1**), we speculate that this may also lead to similar mucosal injury, subsequent changes in GIT microbiota, and the need for mucoid excretion.

Under this model, the giant panda's dramatic shift to the less fibrous leaf portion of bamboo might also cause a change in its mutualistic GIT microbiota. Diet is known to be a major driving force of the GIT microbiota (Flint et al., 2012), and a seasonal shift is evident when comparing winter and non-mucoid samples (**Figures 3** and **4**) both in our study, as well as in previous reports (Williams et al., 2012; Xue et al., 2015). Decreases in dietary fiber and the altered microbiota may then result in inflammation and damage to the mucosal barrier by processes similar to those seen in goats (Metzler-Zebeli et al., 2013; Liu et al., 2014). Without an intact mucosal barrier, the panda experiences further dysbiosis, and fails to maintain its GIT microbiota, as seen by the higher variability within the non-mucoid samples (**Figures 2**–**4**). This dysbiosis, and possibly mucosal damage, reaches a critical level characterized by very low diversity (**Figure 2**). In order to "reset" the system, the mucosal layer along with the altered GIT microbiota is shed, resulting in increased diversity in mucoids, as the mucosa is generally more diverse than feces (Zoetendal et al., 2002; Malmuthuge et al., 2013). Shedding would allow for the reestablishment of the mucosal barrier and a healthier GIT microbiota, as shown by the more winter-like composition and diversity of non-mucoid samples immediately following a mucoid episode (**Figures 2** and **4**). This cycle continues in the weeks following the panda's sudden change from a culm-rich to a leaf-rich diet until such as time as a stable microbial community and mucosal barrier establish for the new diet. However, this dysbiosis hypothesis remains to be tested in giant pandas.

Interestingly, one mucoid sample from the male panda (day 35) did not fit our proposed model as it had diversity (**Figure 2**) and composition (**Figure 4**) more similar to winter fecal samples. Non-mucoid samples from the same day, both before and after the mucoid episode, had higher diversity and characteristic non-mucoid communities. Importantly, another mucoid with characteristic mucoid diversity and composition occurred the following day (day 36). Thus, we speculate that the day 35 mucoid was a failed shedding event and thus contained more fecal material than mucosa, thereby skewing its observed diversity toward a lower value. Since this shedding failed to remove sufficient mucosa, another episode (day 36) was needed in order to achieve a "reset" of the system.

Giant panda mucoid episodes can have several negative nutritional and health impacts, as giant pandas typically do not feed during these periods. Moreover, these mucoids are more prevalent following the typical breeding season, and reduced nutritional status can be transferred to offspring both during gestation and lactation, potentially impacting cub development. Further work in this area is needed to assess the mucosal injury and dysbiosis hypothesis proposed here. Moreover, investigating possible practices like dietary supplements, might help alleviate GIT distress and the subsequent decline in nutritional status in giant pandas. This work is the first characterization of the mucoid-associated microbiota in giant pandas and serves as an initial step toward elucidating the mechanism behind this phenomenon that affects the overall health of this critically endangered species.

### AUTHOR CONTRIBUTIONS

fmicb-07-00661 May 4, 2016 Time: 17:2 # 8

CW designed the experiment, and with KD-M, conducted experimentation and data analysis, and wrote the manuscript. MV customized the computing environment that facilitated bioinformatic analyses. AB, GS, DS, SW, and AK aided in data interpretation. AB, GS, DS, and MV provided editorial assistance with the manuscript.

#### FUNDING

This work was supported by the Memphis Zoological Society, the United States Forest Service International Programs (Asia-Pacific Office), the United States Department of Agriculture National Institute of Food and Agriculture (grant numbers MIS-409030 to AB and WIS-01729 to GS), the Special Research Initiative, Mississippi Agricultural and Forestry Experiment Station (MIS-409030 to DS), and the Leo M. Seal Family Foundation. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of

#### REFERENCES


the author(s), and do not necessarily reflect the views of our supporters.

### ACKNOWLEDGMENTS

The authors would like to thank the Memphis Zoological Society staff (Suzie Zaledzieski, Josephine Fields Falcone, Lexi Yang, and Sierra Browning) for sample collection and Olivia Crowe (Mississippi State University) for assistance in sample collection and processing. We would like to thank all members of the Suen lab (University of Wisconsin-Madison) for their support, insightful discussions, and careful reading of the manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.00661


and next-generation sequencing-based diversity studies. Nucleic Acids Res. 41:e1. doi: 10.1093/nar/gks808


**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 © 2016 Williams, Dill-McFarland, Vandewege, Sparks, Willard, Kouba, Suen and Brown. 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) or licensor 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.

# Responses in colonic microbial community and gene expression of pigs to a long-term high resistant starch diet

#### Yue Sun, Liping Zhou, Lingdong Fang, Yong Su\* and Weiyun Zhu

*Jiangsu Key Laboratory of Gastrointestinal Nutrition and Animal Health, College of Animal Science and Technology, Nanjing Agricultural University, Nanjing, China*

Intake of raw potato starch (RPS) has been associated with various intestinal health benefits, but knowledge of its mechanism in a long-term is limited. The aim of this study was to investigate the effects of long-term intake of RPS on microbial composition, genes expression profiles in the colon of pigs. Thirty-six Duroc × Landrace × Large White growing barrows were randomly allocated to corn starch (CS) and RPS groups with a randomized block design. Each group consisted of six replicates (pens), with three pigs per pen. Pigs in the CS group were offered a corn/soybean-based diet, while pigs in the RPS group were put on a diet in which 230 g/kg (growing period) or 280 g/kg (finishing period) purified CS was replaced with purified RPS during a 100-day trial. Real-time PCR assay showed that RPS significantly decreased the number of total bacteria in the colonic digesta. MiSeq sequencing of the V3-V4 region of the 16S rRNA genes showed that RPS significantly decreased the relative abundance of *Clostridium*, *Treponema*, *Oscillospira*, *Phascolarctobacterium*, RC9 gut group, and S24-7-related operational taxonomic units (OTUs), and increased the relative abundance of *Turicibacter*, *Blautia*, *Ruminococcus*, *Coprococcus*, *Marvinbryantia*, and *Ruminococcus bromii-*related OTUs in colonic digesta and mucosa. Analysis of the colonic transcriptome profiles revealed that the RPS diet changed the colonic expression profile of the host genes mainly involved in immune response pathways. RPS significantly increased proinflammartory cytokine IL-1β gene expression and suppressed genes involved in lysosome. Our findings suggest that long-term intake of high resistant starch (RS) diet may result in both positive and negative roles in gut health.

Keywords: raw potato starch, pig, transcriptional profiling, colon, mucosa-associated microbiota

#### Introduction

It has been known that starches are one of the major carbohydrates available in the human colon (Anderson et al., 1981). Starch that escapes digestion in the small intestine (human and animal) can enter the large intestine, where it is used as a substrate for bacterial fermentation (Topping and Clifton, 2001). The effects and potential health benefits of resistant starch (RS) have been extensively studied in recent years (Martinez-Puig et al., 2003; Fung et al., 2012). Bacterial fermentation of RS results in the production of short-chain fatty acids (SCFAs), mainly acetate, propionate and

#### Edited by:

*Gabriele Berg, Graz University of Technology, Austria*

#### Reviewed by:

*Jan S. Suchodolski, Texas A&M University, USA Douglas Morrison, Univeristy of Glasgow, UK*

#### \*Correspondence:

*Yong Su, College of Animal Science and Technology, Nanjing Agricultural University, Weigang 1#, Nanjing 210095, China yong.su@njau.edu.cn*

#### Specialty section:

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

Received: *15 May 2015* Accepted: *10 August 2015* Published: *25 August 2015*

#### Citation:

*Sun Y, Zhou L, Fang L, Su Y and Zhu W (2015) Responses in colonic microbial community and gene expression of pigs to a long-term high resistant starch diet. Front. Microbiol. 6:877. doi: 10.3389/fmicb.2015.00877* butyrate, which are considered to have beneficial effects upon hindgut health (Young et al., 2005; Le Leu et al., 2009). In addition, RS is considered to have a role in regulating the adipose metabolism of humans and animals, and has positive effects on the health of the body (So et al., 2007). The addition of raw potato starch (RPS) to the diet of pigs could increase the amount of starch entering the large bowel because the native granular structure of this starch, which became an available source for hindgut microflora (Martinez-Puig et al., 2003; Fang et al., 2014).

A study by Haenen et al. (2013b) showed that diets high in RS affected the microbiota composition in the colon of pigs, where butyrate-producing bacteria including Faecalibacterium prausnitzii and Ruminococcus bromii were stimulated in abundances. In vitro studies also found that populations related to R. bromii were the primary starch degrader, while bacteria related to Prevotella spp., Bifidobacterium adolescentis, and Eubacterium rectale might be further involved in the trophic chain (Kovatcheva-Datchary et al., 2009). In a recent study, R. bromii was confirmed to possess an exceptional ability to colonize and degrade starch particles when compared with previously studies (Ze et al., 2012). While most of the study focused on the microbiota in the gut digesta, mucosa-associated microbiota were believed to be more closely related to the gut function of the host. Despite the importance of the mucosal community for bacterial intestinal colonization, pathogen resistance, and hostmicrobiota cross-talk (Leser and Mølbak, 2009), the porcine mucosa-associated microbiota and their diet-related changes were not well investigated at this degree until now.

Recently, genome-wide transcriptional profiling is extensively used to investigate how animals or humans respond to their diets, which contributes to our understanding of the mechanism of a healthy diet. Microarray assay showed that long-term ingestion of a rapidly digestible starch (high amylopectin ratio) significantly elevated hepatic lipogenesis, which is associated with a higher serum insulin concentration and more lipogenic genes being expressed in the liver (Jun et al., 2010). It has been recently reported that consumption of a diet high in RS for 2 weeks induced catabolic pathway but suppressed immune and cell division pathways in the proximal colon of male pigs (Haenen et al., 2013a). Information on the long-term effects of RS on the gene expression profiles of pigs, however, remains limited.

In this study, we postulated that a long-term intake diet high in RS could change the microbial composition in both colonic digesta and mucosa, modulate mucosal transcriptomes, and eventually influence host health. Because of its similar homology to human, the pig has been recognized as one of the ideal models for the study of human nutrition (Guilloteau et al., 2010). Thus, by using RPS to substitute about half of the corn starch (CS) in the diet, the present study aimed to investigate the effects of longterm intake of RPS on digesta- and mucosa-associated microbial composition and gene expression profiles in the colon of pigs.

#### Materials and Methods

#### Animals, Housing, and Diets

This study was approved by the Nanjing Agricultural University Animal Care and Use Committee. The treatment, housing, husbandry and slaughtering conditions conformed to the Experimental Animal Care and Use guidelines of China (Chinese Science and Technology Committee, 1988). All pigs were raised on a commercial farm in the Jiangsu Province of China. Thirtysix Duroc × Landrace × Large White growing barrows (70 days of age, 23.78 ± 1.87 kg) were randomly allocated to two groups, each group consisting of three pigs per pen, and six replicates. Pigs in the control group were offered a corn/soybean-based diet, while 230 g/kg purified CS was replaced with purified RPS in the RPS diet group. Diets were formulated (**Table 1**) according to the nutrient requirements of the National Research Council (1998). When animals reached the age of 120 days, diets were adapted to the nutrient requirements of the animals (finishing diet) and the amount of purified starch increased to 280 g of CS or RPS per kilogram of feed. Pigs had unlimited access to feed and water throughout the experimental period, which consisted of two 50 day trials in which the pigs consumed the growing diet (days 0–50) and finishing diet (days 51–100), respectively.

#### Sampling

On day 100, one pig from each replicate was slaughtered when it approached target slaughter weight (105–110 kg). Feed was withheld from the pigs 12 h before slaughter. Pigs were weighed and transported (10 km) to a local commercial slaughterhouse, and slaughtered via electrical

TABLE 1 | Composition and nutrient analysis of experimental diets (as-fed basis).


*<sup>a</sup>This mineral and vitamin premix (1%) supplies per kg diet as follows: VA 11 000 IU, VD3 1 000 IU, VE 16 IU, VK1 1 mg, VB1 0.6 mg, VB2 0.6 mg, d-pantothenic acid 6 mg, nicotinic acid 10 mg, VB12 0.03 mg, folic acid 0.8 mg, VB6 1.5 mg, choline 800 mg, Fe 165 mg, Zn 165 mg, Cu 16.5 mg, Mn 30 mg, Co 0.15 mg, I 0.25 mg, Se 0.25 mg.*

stunning followed by exsanguination. The animals were bled and opened immediately, and the proximal colon was excised for the collection of digesta and mucosa samples. The pH value of digesta was determined using a handheld pH meter (HI 9024C; HANNA Instruments, Woonsocket, RI). The colon tissues were washed with sterile phosphate-buffered saline (PBS) (pH 7.0). Mucosal samples were collected by scraping the luminal surface with a sterile glass slide. All samples were kept in liquid nitrogen for further gene expression and microbial community analysis.

#### DNA Extraction, PCR Amplification and Illumina MiSeq Sequencing

The total genomic DNA was isolated from colonic digesta and mucosa using a commercially available stool DNA extraction kit according to the manufacturer's instructions of (QIAamp DNA Stool Mini Kit, Qiagen, Hilden, Germany). The concentration of extracted DNA was determined by using a Nano-Drop 1000 spectrophotometer (Thermo Scientific Inc., Wilmington, DE, USA). The V4-V5 region of the bacterial 16S ribosomal RNA gene was amplified by polymerase chain reaction (PCR) using bacterial universal primers 515F and 907R with an eight-base sequence unique to each sample as a barcode (Lane et al., 1985; Kroes et al., 1999). The amplification program consisted of an initial denaturation step at 95◦C for 2 min, followed by 25 cycles at 95◦C for 30 s, 55◦C for 30 s, and 72◦C for 30 s and a final extension at 72◦C for 5 min. PCR reactions were performed in triplicate 20µl mixture containing 4µl of 5 × FastPfu Buffer, 2µl of 2.5 mM dNTPs, 0.8µl of each primer (5µM), 0.4µl of FastPfu Polymerase, and 10 ng of template DNA.

Because one mucosal DNA sample from RPS group failed to be amplified by PCR, amplicons from other 23 samples were extracted from 2% agarose gels and purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA) according to the manufacturer's instructions and quantified using QuantiFluor™-ST (Promega, USA). Purified amplicons were pooled in equimolar and paired-end sequenced (2 × 250) on an Illumina MiSeq platform according to the standard protocols.

#### Bioinformatics Analysis

Raw fastq files were demultiplexed and quality-filtered using QIIME (version 1.17) with the following criteria: the 250 bp reads were truncated at any site receiving an average quality score <20 over a 10 bp sliding window, discarding the truncated reads that were shorter than 50 bp; exact barcode matching, 2 nucleotide mismatch in primer matching, reads containing ambiguous characters were removed; and only sequences that overlap longer than 10 bp were assembled according to their overlap sequence. Reads that could not be assembled were discarded.

Operational taxonomic units (OTUs) were clustered with 97% similarity cutoff using UPARSE (version 7.1 http://drive5.com/ uparse/) and chimeric sequences were identified and removed using UCHIME. To assess bacterial diversity among samples in a comparable manner, a randomly selected, 43274-sequence (the lowest number of sequences in the 23 samples) subset from each sample was compared for the phylogenetic affiliation by RDP Classifier (http://rdp.cme.msu.edu/) against the Silva (SSU115) 16S rRNA database using a confidence threshold of 70% (Amato et al., 2013). We also calculated the coverage percentage using Good's method (Good, 1953), the abundance-based coverage estimator (ACE), the bias-corrected Chao richness estimator, and the Shannon and Simpson diversity indices using the MOTHUR program (http://www.mothur.org) (Schloss et al., 2009). The raw pyrosequencing reads were submitted to Sequencing Read Archive (SRA) database under the accession id: SRA061866. Genera (OTUs) with relative abundances higher than 0.05% within total bacteria were defined as predominant genera (OTUs), and sorted for comparing the difference among different groups.

#### Real-time PCR

Primer set Bact1369/Prok1492 was used for the quantification of total bacteria by real-time PCR on an Applied Biosystems 7300 Real-Time PCR System (Applied Biosystems, USA) using SybrGreen as the fluorescent dye (Suzuki et al., 2000). A reaction mixture (25µl) consisted of 12.5µl of IQ SYBR Green Supermix (Bio-Rad), 0.2µM of each primer set and 5µl of the template DNA. The amount of DNA in each sample was determined in triplicate, and the mean values were calculated.

#### RNA Extraction and Purification

The total RNA of colonic mucosa was extracted using Trizol reagent (Life Technologies, Carlsbad, CA, USA) following the manufacturer's instructions, and were checked for RNA integrity and purity by an Agilent Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA). The qualified total RNA was further purified by an RNeasy mini kit (Qiagen, Gmbh, Hilden, Germany) and an RNase-Free DNase Set (Qiagen, Gmbh, Hilden, Germany). Only those samples that had an OD260/OD280 ratio of approximately 2.0 and showed no degradation (RNA integrity number ≥ 7.0) were used to generate labeled targets.

#### Microarray Hybridization and Data Analysis

The RNA sample from each group was hybridized to an Agilent Porcine 4 × 44 K one-color gene expression microarray (catalog number 026440; Agilent Technologies, Santa Clara, CA, US) containing 43,603 probe sets, as described in the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/). Because there were six replicates from each group, four biological replicates were randomly selected for the microarrays to reduce the costs of the experiment. Total RNA was amplified and labeled using a Low Input Quick Amp Labeling Kit (Agilent Technologies, Santa Clara, CA, USA), following the manufacturer's instructions. Each slide was hybridized with 1.65µg Cy3-labeled cRNA using a Gene Expression Hybridization Kit (Agilent Technologies, Santa Clara, CA, USA) in hybridization oven. After 17 h of hybridization, slides were washed with a Gene Expression Wash Buffer Kit (Agilent technologies, Santa Clara, CA, USA), following the manufacturer's instructions. The slides were then scanned on an Agilent Microarray Scanner (Agilent Technologies, Santa Clara, CA, USA), and data were extracted with Feature Extraction Software 10.7 (Agilent Technologies, Santa Clara, CA, USA). Raw data were normalized using a quantile algorithm, Gene Spring Software 11.0 (Agilent

Technologies, Santa Clara, CA, USA). Array data have been submitted to the Gene Expression Omnibus under accession number GSE71305. Microarrays were performed by Shanghai Biochip Co., Ltd (Shanghai, China).

#### Pathway and Network Analysis

The SBC Analysis System (SAS, http://sas.ebioservice.com) was used to further identify the differentially expressed genes between the two dietary groups, and genes with values of P < 0.05 were extracted. To further clarify the function of the differentially expressed genes in this study, transcripts were first annotated to pig (ssc) genes, and then other significant transcripts were annotated against human (hsa), mouse (mmu) and rat (rno) genes. Highly expressed genes in pigs fed the RPS diet that showed at least a 1.5-fold higher or lower expression level than in pigs fed the CS diet were selected for further study. In addition, the differentially expressed genes identified between the two groups were mapped to Gene Ontology (GO, http://www. geneontology.org/) terms and the Kyoto Encyclopedia of Genes and Genomes (KEGG, http://www.genome.jp/kegg/) pathways to identify potential pathways associated with dietary treatment. Pvalues < 0.05 and a false discovery rate < 0.05 were considered to be statistically significant. The gene network analysis of the differentially expressed genes involved in significant pathways was carried out using the KEGG database.

#### Statistical Analysis

Data were analyzed by SPSS 17.0 as a randomized block design, considering the diet as the main effect and the replicate as a block. The effects of diet and niche compartments on the colonic microbial community were tested for significance using a Two-Way ANOVA program. The effects of diet on gene expression in the colon were tested for significance using the Student's ttest. P values were corrected for multiple testing by using a false-discovery rate (Q-value) method (Benjamini and Hochberg, 1995). Significant differences were declared when P < 0.05.

### Results

#### Microbiota Analysis

Across all 23 samples, 1,397,445 quality sequences were classified as being bacteria with a read length higher than 250 bp. The average length of the quality sequences was 396 bp. The statistical estimates of species richness for 43,274-sequence subsets from each sample at a genetic distance of 3% are presented in **Table 2**. The richness estimators (ACE and Chao) of colonic microbiota were significantly affected by niche compartment and dietary treatment, while the diversity indices (Shannon and Simpson) were not affected by dietary treatment. The rarefaction curves generated by MOTHUR plotting the number of reads by the number of OTUs tended to approach the saturation plateau (**Figure 1**). The curves also showed that ACE and Chao indices from colonic mucosa were significantly higher than those from colonic digesta, and the indices from the CS diet group were significantly higher than those from the RPS group (P < 0.05).

At the phylum level, Firmicutes was the predominant phylum with the abundance higher than 73% in both colonic digesta and mucosa, followed by the phyla Bacteroidetes and Spirochaetae (**Figure 2**). Compared to the colonic digesta, the relative abundances of Bacteroidetes and Synergistetes from colonic mucosa increased significantly, while the abundance of Firmicutes decreased significantly (P < 0.05). However, dietary treatment did not have an effect on the relative abundance of any phylum.

Genus-level analysis revealed that Streptococcus, uncultured Ruminococcaceae and Lactobacillus were the three most predominant genera in the colonic digesta and mucosa of pigs.

FIGURE 1 | Rarefaction curves comparing the number of sequences with the number of phylotypes found in the 16S rRNA gene libraries from microbiota in the colonic digesta (D) and mucosa (M) of pigs fed corn starch (CS) and raw potato starch (RPS) diets.

TABLE 2 | Diversity estimation of the 16S rRNA gene libraries from microbiota in the colon of pigs fed corn starch (CS) and raw potato starch (RPS) diets.


*<sup>a</sup>SEM, standard error of means, n* = *5 or 6 (one mucosal sample in RPS group was missing because it failed to be amplified by PCR).*

As shown in **Table 3**, compared with colonic digesta, genera Oscillospira, Parabacteroides, Phascolarctobacterium, S24-7, Alloprevotalla, Anaerotruncus, RF16, Bacteroides, Spirochaeta, Mucispirillum, Oscillibacter, dgA-11 gut group, Sutterella, and Mogibacterium increased in relative abundance from colonic mucosa, whereas the abundances of Turicibacter, Anaerotruncu, and Peptostreptococcaceae incertae\_sedis decreased (P < 0.05). The consumption of RPS significantly increased the abundances of Turicibacter, Ruminococcus, Blautia, Coprococcus, Marvinbryantia, and Lachnospiraceae incertae sedis, and decreased the abundances of Clostridium, S24-7, RC9 gut group, Parabacteroides, Phascolarctobacterium, Oscillospira, Oscillibacter, and Mogibacterium (P < 0.05).

At the OTU level, consumption of RPS diet significantly decreased the relative abundance of Clostridium, Treponema, Oscillospira, Phascolarctobacterium, RC9 gut group, and S24-7-related OTUs, and increased the relative abundance of Turicibacter, Lachnospiraceae, Blautia, Ruminococcus, Coprococcus, Marvinbryantia, and Ruminococcus bromii-related OTUs (Supplementary Table 1). Compared with the colonic digesta, Phascolarctobacterium, Bacteroidales, Parabacteroides, Oscillospira, Prevotellaceae, Bacteroides, and Spirochaeta-related OTUs increased significantly in the relative abundance in the colonic mucosa of pigs (P < 0.05).

Because MiSeq sequencing analysis can only reflect the relative abundance of bacteria, quantitative real-time PCR was used to determine the completed 16S rRNA gene copies of bacteria in the colon of pigs (**Figure 3**). The consumption of RPS significantly decreased the total number of bacteria in the colonic digesta of pigs (P < 0.05), while there was no difference in colonic mucosa between the two dietary groups. In addition, the pH in the colon of pigs fed RPS diet was significantly lower than that in pigs fed CS diet (P < 0.05) (**Figure 4**).

#### Transcriptome Analysis

Microarrays were used to identify genes that were differentially expressed due to different dietary starch treatments. The analysis of gene expression data from colonic mucosal samples identified 781 differentially expressed genes with the fold change higher than 1.5 at a nominal P-value of 0.05. Of these 781 genes, 465 genes were up-regulated by the RPS diet, including 311 genes with functional annotation and 154 genes without functional annotation, whereas 316 genes were down-regulated, including 238 genes with functional annotation and 78 genes without functional annotation. The most induced gene was coagulation factor VII (F7), showing a 4.05-fold increase with the RPS treatment, whereas the most suppressed gene was dual oxidase 1 (DUOX1), with a 4.17-fold decrease (Supplementary Table 2).

The differently expressed genes with functional annotation were included in a follow-up gene ontology (GO) and pathway analysis. GO category analysis showed that these genes were involved in a wide variety of physiological and biological events, such as cell adhesion, transferase activity, coagulation, regulation of body fluid levels, ion binding, and odorant binding (**Figure 5**). As is shown in Supplementary Table 3, by using the KEGG database, the significant pathways were found to mainly contain immune response (hematopoietic cell lineage, antigen processing and presentation, complement and coagulation cascades, cytosolic DNA-sensing pathway, and Toll-like receptor signaling pathway); signaling molecules and interaction (ECM-receptor interaction, cytokine-cytokine receptor interaction, and neuroactive ligand-receptor interaction); signal transduction (TGF-beta signaling pathway and MAPK signaling pathway); cardiovascular diseases (dilated cardiomyopathy and hypertrophic cardiomyopathy); transport and catabolism (lysosome); cell communication (focal adhesion); and several pathways associated with the biosynthesis of other secondary metabolites, parasitic infectious diseases, and nervous system and nucleotide metabolism.

#### Discussion

We investigated in the present study the effects of longterm consumption of RPS on the mucosa transcriptome, digesta- and mucosa-associated microbiota composition in the proximal colon of pigs. We found that compared with the control diet, the RPS diet changed the colonic expression profile of the host gene involved in immune response pathways; decreased the number of total bacteria and the relative abundance of Clostridium, Treponema, Oscillospira, Phascolarctobacterium, RC9 gut group, and S24-7-related OTUs; and increased the relative abundance of Turicibacter, Blautia, Ruminococcus, Coprococcus, Marvinbryantia, and Ruminococcus bromii-related OTUs.

Most studies today focus on the microbiota in the digesta of the hindgut or feces, whereas research on the effect of RS on mucosa-associated microbiota is limited. Our previous study showed that the substitution of about half of CS with RPS in pig diets increased the amount of RS in the diet, and the amount of starch in the cecum and colon of pigs (Fang et al., 2014). The present study is the first to our knowledge to use a highthroughput technique to investigate the RS effect on colonic mucosa-associated microbiota in pigs. In the present study, TABLE 3 | Relative abundances of microbial genera (percentage) that were significantly affected by the dietary treatment or the niche compartment in the colon of pigsa.


*<sup>a</sup>Genera with relative abundances higher than 0.05% within total bacteria were sorted and showed in the table.*

*<sup>b</sup>SEM, standard error of means, n* = *5 or 6.*

consumption of an RPS diet didn't affect the abundances of phyla, both in the digestive and mucosal microbiota. Martínez et al. (2010) reported that type RS4 but not RS2 significantly induced Bacteroidets and Actinobacteria while decreasing Firmicutes at the phylum level. Young et al. (2012), however, found blooms of Bacteroidets and Actinobacteria in colonic digesta when feeding type 2 RS to rats. The inconsistent results from the different studies may be due to different type of RS being used in different animal models.

The abundance of genera Oscillospira, Parabacteroides, Phascolarctobacterium, S24-7, Alloprevotalla, Anaerotruncus, RF16, Bacteroides, Spirochaeta, Mucispirillum, Oscillibacter, dgA-11 gut group, Sutterella, and Mogibacterium in colonic mucosa was significantly higher than that in digesta, while genera Turicibacter, Anaerotruncu, and Peptostreptococcaceae incertae\_sediss showed the inverse tendency. In addition, we found that the richness estimators (ACE and Chao) of colonic mucosa-associated microbiota were higher than colonic lumen, which is in agreement with the results of a recent study where the mucosa-associated ileal microbiota harbored greater bacterial diversity than the lumen but similar membership to the mucosa of the large intestine in pigs (Looft et al., 2014). Meanwhile, we found that the RPS had the similar modulation effects on either relative abundance or diversity of mucosa-associated microbiota to the lumen in the colon of pigs, which suggests that the RS can affect the pigs through the interaction between

mucosa-associated microbiota and the host cell, not just its fermentation products.

The genus Ruminococcus, which belong to clostridial cluster IV are numerically abundant in the large intestine of humans and pigs, typically accounting for 10–40% of total bacterial 16S rRNA sequences, but are under-represented by cultured isolates (Lay et al., 2005). There are strong indications, however, that this genus plays a primary role in the degradation of particulate substrates such as fiber and RS (Flint et al., 2008; Chassard et al., 2012). Recently, some Ruminococcus species were found to play a primary role in the degradation of dietary RS. Ze et al. (2012) found that Ruminococcus bromii is a keystone species for the degradation of RS in the human colon. Recent dietary intervention studies have reported increased populations of R. bromii and other gram-positive bacteria in fecal or colonic samples during consumption of diets high in RS (Abell et al., 2008; Martínez et al., 2010; Walker et al., 2011; Haenen et al., 2013b). Similarly, the present study also showed that the abundance of this species increased significantly in both colonic

digesta and mucosa of pigs fed an RPS diet compared to the CS diet. In addition, we also found that the long-term intake of RPS increased the abundance of genus Coprococcus in the colon of pigs. Strains related to Coprococcus that belonged to the Clostridium coccoides cluster were also listed in major butyrateproducing bacteria isolated from the human colon in a previous review (Louis and Flint, 2009). The increase in the abundance of butyrate-producing bacteria is in agreement with the fact that an RPS diet significantly increased the concentrations of butyrate and total SCFA, as was shown in our preliminary study (Fang et al., 2014).

In the present study, we found that RPS diet decreased the pH value in the colon contents of pigs, which may be due to the increase of SCFAs produced by microbial fermentation of starch (Fang et al., 2014). In addition, the lactate concentration in colon of the RPS group was also higher than the CS group (data no shown). The array data showed that gene expression of TLR-7 in the colon was down-regulated by the consumption of RPS. TLR7 recognizes single-stranded RNA in endosomes, which is a common feature of viral genomes that are internalized by macrophages and dendritic cells (Hipp et al., 2013). The decline in the pH value may result in the decrease of bacteria or virus populations in the colon of pigs. This was confirmed by the results of real-time PCR quantification of the complete 16S rRNA gene copies of total bacteria and the rarefaction curves of the 16S rRNA gene sequences. The array data in the present study also showed that RPS significantly suppressed genes involved in lysosome (**Figure 6**). Lysosomes are cellular organelles that contain acid hydrolase enzymes that break down waste materials and cellular debris. One of the major functions of lysosomes is the digestion of material taken up from outside the cell by endocytosis including excess or worn-out organelles, food particles, and engulfed viruses or bacteria. Similarly, Haenen et al.

(2013b) found that short-term (2 weeks) intake of RS could also decrease the relative abundance of potential pathogens in the bacterial community, and could suppressed genes involved in both the innate and the adaptive immune response.

Recent studies have shown that the microbial metabolites, SCFA, can regulate colonic regulatory T cell homeostasis and can control gut inflammatory responses (Arpaia and Rudensky, 2014). Butyrate can suppress pro-inflammatory cytokine production by intestinal macrophages (Arpaia et al., 2013; Chang et al., 2014). The suppression role in immune response of a high RS diet may be due to the increase in SCFA. The array data in the current study showed, however, that longterm intake of RPS significantly increased pro-inflammatory cytokine IL-1β gene expression, which is not in line with the observation that the RPS increased the total SCFA and butyrate production in the colon in our previous study (Fang et al., 2014). The possible reason is that the significant decrease of pH value in the colon caused by long-term intake of PRS may result in the death of bacteria, and the accumulation of lipopolysaccharides (LPS). Similarly, the LPS have received increasing attention in rumen acidosis caused by a high concentration (starch) diet because of their role in the inflammatory response of the body (Tao et al., 2014). Although their benefit roles such as improving colonic mucosal integrity and reducing gut apoptosis are well known in humans and animals consuming a high RS diet (Nofrarías et al., 2007), the results of this study suggest that longterm intake of RPS may result in a negative effect on the health of pigs because of the low pH in the hindgut. Further studies are needed, however, to know whether this negative role will happen when using other type of RS diet or in humans and other animals in the long-term.

In conclusion, we have shown that long-term consumption of a diet high in RS modulated digesta and mucosa-associated microbial composition, and gene expression in the colon of pigs. While our findings suggest that a high RS diet may result in both positive and negative effects on gut health, which should not be ignored in further studies, additional investigation is required to further elucidate the underlying mechanisms of balancing the roles of RS on animal health.

#### Author Contributions

Conceived and designed the experiments: YS, WZ. Performed the experiments: YS, YS, LF, LZ. Analyzed the data: YS, LF, YS. Wrote the paper: YS, YS.

#### References


#### Acknowledgments

This research has received funding from the National Basic Research Program of China (2012CB124705 and 2013CB127603) and the Fundamental Research Funds for the Central Universities (KYZ201153).

#### Supplementary Material

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2015.00877

SCFA concentrations, and gene expression in pig intestine. J. Nutr. 143, 274–283. doi: 10.3945/jn.112.169672


**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 © 2015 Sun, Zhou, Fang, Su 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) or licensor 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.

# Insights into Abundant Rumen Ureolytic Bacterial Community Using Rumen Simulation System

Di Jin1, 2 †, Shengguo Zhao1 †, Pengpeng Wang<sup>1</sup> , Nan Zheng<sup>1</sup> , Dengpan Bu<sup>1</sup> , Yves Beckers <sup>2</sup> and Jiaqi Wang<sup>1</sup> \*

<sup>1</sup> State Key Laboratory of Animal Nutrition, Institute of Animal Sciences, Chinese Academy of Agricultural Sciences, Beijing, China, <sup>2</sup> Animal Science Unit, Gembloux Agro-Bio Tech, University of Liège, Gembloux, Belgium

Urea, a non-protein nitrogen for dairy cows, is rapidly hydrolyzed to ammonia by urease produced by ureolytic bacteria in the rumen, and the ammonia is used as nitrogen for rumen bacterial growth. However, there is limited knowledge with regard to the ureolytic bacteria community in the rumen. To explore the ruminal ureolytic bacterial community, urea, or acetohydroxamic acid (AHA, an inhibitor of urea hydrolysis) were supplemented into the rumen simulation systems. The bacterial 16S rRNA genes were sequenced by Miseq high-throughput sequencing and used to reveal the ureoltyic bacteria by comparing different treatments. The results revealed that urea supplementation significantly increased the ammonia concentration, and AHA addition inhibited urea hydrolysis. Urea supplementation significantly increased the richness of bacterial community and the proportion of ureC genes. The composition of bacterial community following urea or AHA supplementation showed no significant difference compared to the groups without supplementation. The abundance of Bacillus and unclassified Succinivibrionaceae increased significantly following urea supplementation. Pseudomonas, Haemophilus, Neisseria, Streptococcus, and Actinomyces exhibited a positive response to urea supplementation and a negative response to AHA addition. Results retrieved from the NCBI protein database and publications confirmed that the representative bacteria in these genera mentioned above had urease genes or urease activities. Therefore, the rumen ureolytic bacteria were abundant in the genera of Pseudomonas, Haemophilus, Neisseria, Streptococcus, Actinomyces, Bacillus, and unclassified Succinivibrionaceae. Insights into abundant rumen ureolytic bacteria provide the regulation targets to mitigate urea hydrolysis and increase efficiency of urea nitrogen utilization in ruminants.

Keywords: rumen, ureolytic bacteria, urea, acetohydroxamic acid, high-throughput sequencing

## INTRODUCTION

The use of urea in feeds of ruminants is increasing to reduce the supplementation of true protein and the costs of rations. The recommendations of urea would be for no more than 1% in the concentrate, ∼135 g/cow daily (Kertz, 2010). In the rumen, ureolytic bacteria produce urease to hydrolyze urea to ammonia, which is subsequently used for the synthesis of amino acids and

#### Edited by:

Christine Moissl-Eichinger, Medical University Graz, Austria

#### Reviewed by:

Simon M. Dittami, Station Biologique de Roscoff, France Sudeep Perumbakkam, Michigan State University, USA

> \*Correspondence: Jiaqi Wang jiaqiwang@vip.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: 19 January 2016 Accepted: 13 June 2016 Published: 28 June 2016

#### Citation:

Jin D, Zhao S, Wang P, Zheng N, Bu D, Beckers Y and Wang J (2016) Insights into Abundant Rumen Ureolytic Bacterial Community Using Rumen Simulation System. Front. Microbiol. 7:1006. doi: 10.3389/fmicb.2016.01006 microbial protein. Normally, the rate of urea hydrolysis exceeds the rate of ammonia utilization, which leads to poor efficiency of urea utilization in the rumen and explosion of toxic ammonia in the blood (Patra, 2015). Acetohydroxamic acid (AHA), an inhibitor of urease activity that prevents the rapid hydrolysis of urea and consequent explosion of ammonia in rumen, is commonly applied in the rations of ruminants (Upadhyay, 2012).

Ureolytic bacteria play an important role in the hydrolysis of urea in the rumen. Previous studies have isolated some ureolytic bacteria from the rumen including Succinovibrio dextrinosolvens, Treponema sp., Ruminococcus bromii, Butyrivibrio sp., Bifidobacterium sp., Prevotella ruminicola, and Peptostreptococcus productus (Wozny et al., 1977). However, due to the difficulty in cultivating the rumen bacteria, those that have been isolated represent only 6.5% of the community (Kim et al., 2011). Thus, sequencing and phylogenetic analysis of 16S rRNA genes and functional genes have been extensively used in studies focused on members of the uncultured bacteria. By sequencing, ureolytic bacterial diversity has been observed in the environment including open oceans (Collier et al., 2009), groundwater (Gresham et al., 2007), sponges (Su et al., 2013), and soil (Singh et al., 2009). We have previously studied rumen ureolytic bacteria using a urease gene clone library, and found that ureolytic bacterial composition in the rumen was distinct from that in the environment (Zhao et al., 2015). Therefore, it is interesting and meaningful to explore the rumen ureolytic bacterial communities further.

Rumen simulation systems have been developed and used in the evaluation of feeds nutrients degradation and rumen fermentation manipulation in order to avoid the use of animals or decrease study costs (Hristov et al., 2012). We invented a dual-flow continuous rumen simulation system with real-time monitoring of pH, temperature, gas production, methane, and carbon dioxide concentration (Figure S1). We demonstrated that the conditions of microbial fermentation in the system were similar to those in the rumen of dairy cows (Shen et al., 2012), making it a powerful and practical tool for the study of rumen microbes or fermentation.

The objective of this study was to reveal abundant ureolytic bacterial community by high-throughput sequencing in a rumen simulation system when treated with an activator (urea) or inhibitor (AHA) of ureolytic bacteria.

#### MATERIALS AND METHODS

#### Experimental Design and Continuous Cultivation

The rumen simulation system with eight fermenters were used in two replicated periods of 10 d each (7 d for adaptation and 3 d for sampling; Shen et al., 2012). The basic total mixed ration (TMR) was ground down to 1 mm for subsequent use. Fermenters were assigned to four treatments: U0\_A0 (basic diet only), U0\_A0.45 [basic diet plus AHA of 0.45 g/kg dry matter (DM)], U5\_A0 (basic diet plus urea of 5 g/kg DM), U5\_A0.45 (basic diet plus urea of 5 g/kg DM and AHA of 0.45 g/kg DM). Two fermenters were randomly assigned to each treatment in each period. A total of 40 g feed (DM based) was placed into each fermenter daily in two equal portions at 09:00 and 21:00. Urea and AHA were dissolved in artificial saliva (Weller and Pilgrim, 1974) and were added directly into the fermenters after each feeding. The basic diet (DM based) primarily consisted of alfalfa hay (17.72%), corn silage (17.50%), oaten hay (5.09%), cotton seed (5.61%), apple pulp (3.74%), sugar beet pulp (6.71%), and compound packet (40.95%). The compound packet provided the following per kg of diets: steam corn 180.39 g, soybean skin 55.84 g, soybean meal 64.43 g, extruded soybean 38.66 g, distillers dried grains with soluble (DDGS) 24.48 g, double-low rapeseed meal 25.77 g, Ca(HCO3)<sup>2</sup> 2.58 g, CaCO<sup>3</sup> 2.58 g, NaCl 3.44 g, and NaHCO<sup>3</sup> 6.01 g (Table S1).

On the first day of each period, all fermenters were inoculated with ruminal fluid obtained from three rumen-fistulated cows fed the same TMR diet as used in the in vitro study. Animals involved in this study were cared for according to the principles of the Chinese Academy of Agricultural Sciences Animal Care and Use Committee (Beijing, China). Ruminal fluid was strained through four layers of cheesecloth and transferred to the laboratory in a sealed container. A total 500 mL of the strained ruminal fluid was added to each of the eight fermenters, which also contained 500 mL of artificial saliva. Anaerobic conditions were established by flushing the headspace of the fermenters with N<sup>2</sup> at a rate of 20 mL min−<sup>1</sup> . The artificial saliva was continuously infused into the flasks. The temperature of the fermenters was maintained at 39◦C by circulating water, and the fermenter content was stirred continuously at 25 rpm.

#### Rumen Fluid Sampling and DNA Extraction

During the last 3 days of each period, 3 mL of fermenter liquid was collected from each fermenter at 0, 2, 4, 6, 8, and 10 h after morning feeding. Collected samples were stored at <sup>−</sup>80◦C for detection of ammonia nitrogen (NH3-N) and urea nitrogen (urea-N) concentrations. The NH3-N concentration was determined using the method based on the Berthelot (phenol–hypochlorite) reaction (Broderick and Kang, 1980). Urea nitrogen (urea-N) concentration was determined using the diacetyl monoxime method with a commercial kit (Nanjing Jiancheng Co., Nanjing, China). Rumen fluid collected at 2 h was used to extract microbial DNA with a cetyl trimethylammonium bromide (CTAB) plus bead beating method (Minas et al., 2011). Extracted DNA was assessed by agarose gel (1%) electrophoresis and quantified using a NanodropTM spectrometer (Thermo Scientific, Waltham, MA, USA).

### Quantitative PCR of Urease and 16S rRNA Genes

The urease alpha subunit encoding gene (ureC) primers UreC-F (5′ -TGGGCCTTAAAATHCAYGARGAYTGGG-3′ ) and UreC-R (5′ -SGGTGGTGGCACACCATNANCATRTC-3) were used to quantify the ureC gene copies (Reed, 2001). 16S rRNA gene of total bacteria were quantified using 338- F (5′ -ACTCCTACGGGAGGCAGCAG-3′ ) and 533-R (5′ -TTA CCGCGGCTGCTGGCAC -3′ ) as primers (Huse et al., 2008). The assays were performed in an iQTM5 Multicolor Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA) using SYBR <sup>R</sup> Premix Ex TaqTM II (Takara, Dalian, China). Standard curves were generated using plasmids DNA cloned with ureC gene or 16S rRNA gene (Figure S2). Copy number of ureC gene or 16S rRNA gene in per ng of DNA was determined by relating the CT value to the standard curves. The proportion of ureC gene copies was calculated as the ratio of ureC gene copies to total 16S rRNA gene copies. The detailed qPCR protocols were provided in the Supplementary Material. The proportion of ureC gene copies in each treatment were shown in a boxplot constructed using R (R Core Team, 2013).

### Bacterial 16S rRNA Genes Amplification and Illumina Sequencing

Microbial DNA was used as a template for amplification of partial 16S rDNA sequence using the universal bacterial primers 515F (5′ -GTGCCAGCMGCCGCGGTAA-3′ ) and 806R (5′ - GGACTACHVGGGTWTCTAAT-3′ ; Nelson et al., 2014) with both primers tagged with unique barcode sequences for each sample. All polymerase chain reactions (PCRs) were carried out in 50µL reactions with 0.5µL of PrimeSTAR <sup>R</sup> HS DNA Polymerase (TaKaRa, Dalian, China), 10 µL 5 × PrimeSTAR Buffer (plus Mg2+) (TaKaRa), 0.2µM of the forward and reverse primers, 200µM dNTP (TaKaRa), and 100 ng microbial DNA. Thermal cycling consisted of initial denaturation at 98◦C for 1 min, followed by 30 cycles of denaturation at 98◦C for 10 s, annealing at 50◦C for 30 s, and elongation at 72◦C for 60 s, and a final elongation at 72◦C for 5 min. Unique bands were identified using agarose gel (2%) electrophoresis of PCR amplicons (Figure S3). The bands were cut and purified with a QIAGEN MinElute PCR Purification Kit (Qiagen, Valencia, CA, USA). Amplicon libraries were generated using NEB Next <sup>R</sup> UltraTM DNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA, USA) following the manufacturer's recommendations, with the addition of index codes. Library quality was assessed on the Qubit <sup>R</sup> 2.0 Fluorometer (Thermo Scientific) and Agilent Bioanalyzer 2100 system. The library was sequenced on an Illumina MiSeq platform (2 × 250 bp).

#### Sequencing Data Processing and Analysis

Paired-end reads were merged using FLASH (Magoè and Salzberg, 2011). Merged reads were assigned to each sample based on the unique barcode, after which the barcodes and primers were removed. The quality of raw reads was checked, and reads were truncated at any site of >3 sequential bases receiving a quality score of <Q20, and reads with <75% (of total read length) consecutive high quality base calls were removed (Caporaso et al., 2010; Bokulich et al., 2013). Chimeric sequences were detected and removed using UCHIME (Haas et al., 2011). Operational taxonomic units (OTU) were generated by aligning the reads to the GreenGenes database released in May 2013 (DeSantis et al., 2006) and clustered at 97% sequence identity using the PyNAST tool (Caporaso et al., 2010) and the UCLUST algorithm (Edgar et al., 2011). The OTUs were filtered based on the total observation count of an OTU <10 and the number of samples in an OTU <2 in QIIME (Caporaso et al., 2010). The OTUs were further assigned to taxa using the RDP classifier (Wang et al., 2007). The OTU table was rarified for alpha diversity analysis. Simpson, Shannon, Chao1, and the PD\_whole\_tree index were calculated for each sample. Good's coverage was used to estimate the percentage of the total species that were sequenced in each sample (Caporaso et al., 2010). QIIME was used to calculate the weighted UniFrac distances, which are phylogenetic measures of beta diversity. The weighted UniFrac distance was used for Principal Coordinate Analysis (PCoA; Lozupone et al., 2007). The significance of grouping in the PCoA plot was tested by analysis of similarity (ANOSIM) in QIIME with 999 permutations (R Core Team, 2013; Mahnert et al., 2015). The relative abundance of bacteria was expressed as the percentage. The potential ureolytic bacteria were selected using the criterion that their abundance increased with urea treatment and decreased with AHA treatment. The urease alpha subunit sequences of representative species from potential ureolytic bacteria were checked against the NCBI protein database and the urease activities of these bacteria were verified by published studies.

#### Statistical Analysis

Urea-N, ammonia, proportion of ureC gene copies, bacterial abundance, and diversity index were statistical analyzed using the SAS MIXED procedure (SAS Institute, Inc, Cary, NC) as shown in the following model: Yijk = µ + a<sup>i</sup> + b<sup>j</sup> + abi<sup>j</sup> + eijk, where Yijk is the dependent variable, µ is the overall mean, a<sup>i</sup> is the effect of urea treatment i, b<sup>j</sup> is the effect of AHA treatment j, abij is the interaction between a<sup>i</sup> and b<sup>j</sup> (Both factors and their interaction are considered fixed effects), and eijk is the residual, assumed to be normally distributed. Data of bacterial abundance were transformed to log<sup>10</sup> (n+1) if necessary to ensure normal distribution. Mean separation was conducted by using Fisher's least significant difference test. Differences were declared significant at P < 0.05. Tukey's test was used to determine where the differences occurred.

#### Nucleotide Sequence Accession Number

All the raw sequences after assembling and filtering were submitted to the NCBI Sequence Read Archive (SRA; http:// www.ncbi.nlm.nih.gov/Traces/sra/), under accession number SRP074113.

## RESULTS

#### Changes of Urea, Ammonia Concentrations, and Proportion of ureC Genes

The urea-N concentrations in the two urea treated groups were higher (P < 0.01) than the other two groups at 2 h after morning feeding (**Figure 1**). In the two urea treated groups, Group U5\_A0.45 exhibited a higher (P < 0.01) urea concentration than group U5\_A0, indicating a decreased urea hydrolysis rate with AHA inhibition (**Figure 1**). The NH3-N concentrations of all four treatments showed a peak value after fermentation for 2 h. Urea supplementation significantly increased (P < 0.01) NH3-N concentration during whole sampling period, while in the two urea-treated groups, AHA addition also decreased NH3- N concentration significantly (P < 0.01). Two hours after the

urea of 5 g/kg DM and AHA of 0.45 g/kg DM. \*Means values in group U5\_A0 was significantly different from that in group U5\_A0.45 (P < 0.05).

morning feeding, the proportion of ureC genes was higher (P < 0.05) in urea-treated groups than in non-urea treated groups. The addition of AHA did not have a significant effect on the

### Changes of Ureolytic Bacterial Diversity

proportion of ureC genes (**Figure 2**).

A total of 2,105,448 merged sequences were acquired from 16 samples, and 1,672,529 high-quality sequences, with an average read length of 253 bases were obtained. After removing chimeric sequences, the remaining 1,603,997 sequences were used to generate OTUs with 97% sequence similarity across all samples. The OTU table was filtered, leaving 5075 OTUs for subsequent analysis. Collectively, 24 bacterial phyla were identified. Bacteroidetes, Firmicutes, and Proteobacteria were the three predominant phyla, representing 35, 28, and 23% of all sequences, respectively (**Figure 3**). Genera that were each represented by ≥ 0.1% of the total sequences in at least 1 of the 16 samples were selected for further analysis. The 10 predominant genera were Prevotella, Treponema, YRC22, Succinivibrio, Porphyromonas, Oscillospira, Roseburia, Bacteroides, Butyrivibrio, and Coprococcus (**Figure 4**).

After rarefaction, 9000 sequences per sample were used for diversity analysis. Alpha bacterial diversity was presented in **Table 1**. Group U5\_A0 had the highest Chao 1 and PD\_whole\_tree estimates, followed by groups U5\_A0.45, U0\_A0.45, and U0\_A0. No significant differences were observed among the four groups based on the results of the Simpson and Shannon diversity index. PCoA analysis of overall diversity based on the unweighted UniFrac metrics was performed to compare the four treatments (**Figure 5**). ANOSIM (cutoff = 0.01) showed no significant differences in bacterial community composition between treatments U0\_A0 and U0\_A0.45 (R = −0.198, P = 0.925) or between treatments U5\_A0 and U5\_A0.45 (R = −0.135, P = 0.888). A tendency of difference was found between treatments U0\_A0 and U5\_A0 (R = 0.323, P = 0.091). Principal Coordinate 1 and 2 accounted for 44.19 and 25.14% of the total variation, respectively.

urea and AHA supplementation. The proportion of ureC gene copies was calculated as the ratio of ureC gene copies to total 16S rRNA gene copies. U0\_A0, basic diet only; U0\_A0.45, basic diet plus AHA of 0.45 g/kg DM; U5\_A0, basic diet plus urea of 5 g/kg DM; U5\_A0.45, basic diet plus urea of 5 g/kg DM and AHA of 0.45 g/kg DM. <sup>a</sup>,bDifferent letters for different treatments indicate statistically significant differences (P < 0.05).

### Changes of the Relative Abundance of Ureolytic Bacteria

At the phylum level, the group treated with urea only had the highest proportion of Proteobacteria and Actinobacteria, and the lowest proportion of Bacteroidetes compared with the other three groups (**Figure 3**). Both of the two urea-treated groups had relatively high proportions of Acidobacteria and low proportions of Spirochaetes compared with the other two groups. In addition, the two urea-treated groups had higher percentages of unclassified bacteria than the other two groups.



<sup>a</sup>−cMean values within a row with different letters differ significantly (P < 0.05). SEM, standard error of the mean. U0, basic diet without urea; U5, basic diet plus urea of 5 g/kg DM; A0, basic diet without AHA; A0.45, basic diet plus AHA of 0.45 g/kg DM.

At the genus level, the relative abundance represented by ≥0.1% of the total sequences in at least one of the whole samples were further analyzed (**Table 2**). Pseudomonas (1.25%) from Proteobacteria and Streptococcus (1.00%) from Firmicutes were more predominant in group U5\_A0 compared to the other three groups (P < 0.01). Haemophilus and Neisseria from Proteobacteria, and Actinomyces from Actinobacteria were the most abundant in the U5\_A0 group compared with the other three groups (P < 0.05). The relative abundance of Bacillus from Firmicutes and unclassified Succinivibrionaceae were higher in the two urea-treated groups compared with the other two groups (P < 0.01). According to the results retrieved from the NCBI protein database and reported in previous studies, the representative species from Pseudomonas, Haemophilus, Streptococcus, Neisseria, Bacillus, Actinomyces, and unclassified Succinivibrionaceae were identified as containing urease genes and having urease activity (**Table 3**).

#### DISCUSSION

In the rumen, urea is a source of nitrogen for the growth of ureolytic bacteria. AHA, an inhibitor of urease, inhibits urea

usage by ureolytic bacteria, and results in insufficient nitrogen source for bacterial growth. In this study, we used urea and AHA to promote or inhibit the growth of rumen ureolytic bacteria, respectively. We observed that AHA is a useful inhibitor for slowing down the hydrolysis of urea within the rumen fluid. This is consistent with previously published studies in vivo (Jones and Milligan, 1975; Makkar et al., 1981).

Urea supplementation significantly increased bacterial community richness and the number of bacterial species. AHA supplementation resulted in no changes of richness and diversity of bacterial community. The proportion of urease gene copies was served as a proxy to observe changes in the proportion of ureolytic bacteria. Urea supplementation significantly increased the proportion of ureolytic bacteria, which suggested that urea stimulated the growth of rumen ureolytic bacteria. In addition, ANOSIM revealed that the composition of the entire bacterial community in urea-treated groups showed a trend of difference from those in non-urea treated groups (P < 0.10). Changes of the bacterial community in response to urea treatment were possibly related to urease activity and the production of ammonia. Kim

et al. (2014) found that urease genes and enzyme activities were regulated by the level of ammonia in ruminal cellulytic bacteria Ruminococcus albus 8. The lack of a significant effect by AHA on the diversity of the rumen bacterial community may be due to microbial adaption of AHA. Previous studies found that rumen microbe could adapt to chronic AHA supplementation, while AHA was capable of short-term inhibition of urease activity in the rumen (Zhang et al., 2001).

Across the four groups, three phyla (Bacteroidetes, Firmicutes, and Proteobacteria) were predominant. Similar to our results previously published studies have reported that the distribution of phylotypes of rumen bacterial communities fell predominantly into these three phyla (Hook et al., 2011; Wu et al., 2012; Zhang et al., 2014). The bacterial community from our in vitro simulation system was thus similar to the communities observed in vivo. The group treated with urea only had the highest proportion of Proteobacteria and the lowest proportion of Bacteroidetes. In accordance, Collier et al. (2009) investigated the diversity of ureolytic microorganisms in open ocean and estuarine planktonic communities, and found that ureolytic microorganisms were most commonly found in Proteobacteria and rare in Bacteroidetes.

Bacillus was in higher abundance in the two groups supplemented with urea, indicating it was more responsive to urea. Bacillus spp. in the rumen is able to degrade hemicellulose, and produce polysaccharidases and glycoside hydrolases to utilize polysaccharide (Williams and Withers, 1983). B. pasteurii, B. lentus, and B. cereus have proven to be ureolytic bacteria (Benini et al., 2000; Rasko et al., 2004; Sarda et al., 2009), and the urease activity of B. pasteurii is inhibited by AHA (Benini et al., 2000). The unclassified Succinivibrionaceae was also observed at a higher relative abundance in the two urea-treated groups. In the rumen, Succinivibrionaceae is very common and important for degradation of starch, pectin, and dextrin to succinate and propionate (Santos and Thompson, 2014). Succinivibrionaceae WG-1 isolated from the foregut of tammar wallaby produced urease for urea catabolism (Pope et al., 2011). Several isolates of S. dextrinosolvens from the rumen were also shown to have urease activity (Wozny et al., 1977).

Pseudomonas and Streptococcus were both relatively more abundant in the group treated with urea only, but these bacteria had lower abundance in AHA-treated groups. These results confirmed the urea stimulating and AHA inhibiting effects on the microbial community. Several species of Pseudomonas and Streptococcus are able to hydrolyze cellulose (Lynd et al., 2002; Oyeleke and Okusanmi, 2008). In the genus Pseudomonas, species such as P. fluorescens (isolated from soil) and P. aeruginosa (isolated from ocean) possess urease activity (Jyothi and Umamahe, 2013; Goswami et al., 2015). In addition, two Streptococcal species, S. thermophiles and S. salivarius, also produce urease (Chen et al., 2000; Zotta et al., 2008). Kakimoto et al. (1989) assayed about 16,000 isolates from animal feces and intestines for production of acid urease, and found 370 urease-positive strains belonging to the genus Streptococcus. This is consistent with the results of our study in which Streptococcus were found in higher abundance in response to urea supplementation.

The relative abundance of genera Haemophilus, Neisseria, and Actinomyces increased in response to urea and decrease in response to AHA supplementation. The members of Haemophilus ferment glucose (Kilian, 2015), and H. haemolyticus and H. influenzae Rd have urease activity (McCrea et al., 2008). The H. somnus strains of ruminants have varying urea hydrolysis ability (Garcia-Delgado et al., 1977). Neisseria, a Gram-negative aerobic cocci, produces acid from different types of sugars, and some species are disease-causing (Marri et al., 2010). N. sicca strains SB and SC isolated from soil have proven to be urease positive (Sakai et al., 1996). Neisseria had a higher proportion in groups treated with urea, suggesting the potential of bacterial species in the rumen to have urea hydrolysis activity. Actinobacteria, a group of Gram-positive bacteria, represent up to 3.00% of the total rumen bacteria (Pandya et al., 2010; Šul'ák et al., 2012). Some strains of A. meyeri, A. radicidentis, and A. johnsonii are known to have urease activity (Schaal and Yassin, 2015), and A. naeslundii had urease gene and activity (Morou-Bermudez and Burne, 1999, 2000). However, An et al. (2006) described a novel species, Actinomyces ruminicola sp., from cattle rumen, was unable to hydrolyze urea. So it needs to be verified for ureolytic activity of different Actinomyces species.

#### CONCLUSION

The composition of bacterial community following urea or AHA supplementation treatment showed no significant difference TABLE 2 | Bacterial genera that accounted for ≥0.1% of the total sequences in at least one of the samples with significant variation under different treatments (abundance of the genera was expressed as %).


<sup>a</sup>−cMeans values within a row with different letters differ significantly (P < 0.05). SEM, standard error of the mean. U0, basic diet without urea; U5, basic diet plus urea of 5 g/kg DM; A0, basic diet without AHA; A0.45, basic diet plus AHA of 0.45 g/kg DM.

TABLE 3 | Urease gene and enzyme activity of selected genera containing ureolytic bacteria in rumen.


+Positive urease genes or enzyme activity.

compared to the groups without supplementation. In the rumen, the ureolytic bacteria were abundant in the genera including Pseudomonas, Streptococcus, Haemophilus, Bacillus, Neisseria, Actinomyces, and unclassified Succinivibrionaceae. The insights into abundant ureolytic bacteria provide the basis for designing strategies to efficiently manipulate the bacterial community or function and improve urea utilization in ruminant production.

#### AUTHOR CONTRIBUTIONS

JW, DB, and SZ designed the experiments. DJ and PW performed the experiments. SZ and DJ analyzed the data. DJ wrote the paper. SZ, NZ, and YB revised the paper. All authors agree to be accountable for all aspects of the work.

#### ACKNOWLEDGMENTS

This research was supported by the funds from National Natural Science Foundation of China (31430081 and 31261140365), the Agricultural Science and Technology Innovation Program

#### REFERENCES


(ASTIP-IAS12) and Modern Agro-Industry Technology Research System of the PR China (nycytx-04-01). We thank the University of Liège-Gembloux Agro-Bio Tech and more specifically the research platform Agriculture Is Life for the funding of the scientific stay in Belgium that made this paper possible.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01006


to hemolytic and nonhemolytic Haemophilus haemolyticus strains. J. Clin. Microbiol. 46, 406–416. doi: 10.1128/JCM.01832-07


**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 © 2016 Jin, Zhao, Wang, Zheng, Bu, Beckers 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) or licensor 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 Dietary Forage and Calf Starter Diet on Ruminal pH and Bacteria in Holstein Calves during Weaning Transition

Yo-Han Kim<sup>1</sup> , Rie Nagata<sup>1</sup> , Natsuki Ohtani <sup>2</sup> , Toshihiro Ichijo<sup>2</sup> , Kentaro Ikuta<sup>3</sup> and Shigeru Sato1, 2 \*

*<sup>1</sup> United Graduate School of Veterinary Science, Gifu University, Gifu, Japan, <sup>2</sup> Cooperative Department of Veterinary Medicine, Faculty of Agriculture, Iwate University, Morioka, Japan, <sup>3</sup> Awaji Agricultural Technology Center, Minami-Awaji, Japan*

#### Edited by:

*Christine Moissl-Eichinger, Medical University of Graz, Austria*

#### Reviewed by:

*Charles James Newbold, Aberystwyth University, UK Amlan Kumar Patra, West Bengal University of Animal and Fishery Sciences, India*

> \*Correspondence: *Shigeru Sato sshigeru@iwate-u.ac.jp*

#### Specialty section:

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

Received: *31 March 2016* Accepted: *21 September 2016* Published: *21 October 2016*

#### Citation:

*Kim Y-H, Nagata R, Ohtani N, Ichijo T, Ikuta K and Sato S (2016) Effects of Dietary Forage and Calf Starter Diet on Ruminal pH and Bacteria in Holstein Calves during Weaning Transition. Front. Microbiol. 7:1575. doi: 10.3389/fmicb.2016.01575* We investigated the relationship between ruminal pH and bacteria in calves fed calf starter with and without forage during weaning transition. First, 16 Holstein bull calves were obtained from dairy farms and equipped with rumen cannulas by cannulation surgery. Then, calves (73.5 ± 4.2 kg; mean ± SE) were assigned to groups fed calf starter either with forage (HAY, *n* = 8) or without forage (CON, *n* = 8), and all calves were weaned at 8 weeks of age. Ruminal pH was measured continuously, and rumen fluid samples were collected at 7, 8, 9, and 11 weeks of age, namely −1, 0, 1, and 3 weeks after weaning, respectively, to assess volatile fatty acid concentrations and bacterial DNA. The 24-h mean ruminal pH was significantly (*P* < 0.05) different between the two groups. Diurnal changes in the 1-h mean ruminal pH were observed throughout the study in the HAY group; however, they were not observed at 0 and 1 weeks after weaning in the CON group. Moreover, the HAY group had significantly (*P* < 0.05) higher proportions of acetate and butyrate and lower proportion of propionate, and significantly (*P* < 0.05) lower ruminal acetate-to-propionate ratios were observed in the CON group. The ruminal bacterial diversity indices decreased after −1 week in both groups and increased at 0 and 1 weeks after weaning in the HAY and CON groups, respectively. From the 454 pyrosequencing analysis, significant differences (*P* < 0.05) were observed in the relative abundance of several phyla (*Bacteroidetes*, *Actinobacteria*, and *Tenericutes*) and one genus (*Prevotella*) between the two groups. From quantitative real-time PCR analysis, the HAY group had the higher copy numbers of cellulolytic bacteria (*Ruminococcus flavefaciens* and *Ruminococcus albus*) compared with the CON group. This study demonstrated that feeding of dietary forage alleviates subacute ruminal acidosis due to diurnal changes in ruminal pH. Furthermore, changes in ruminal pH affect the ruminal bacterial diversity and relative abundance, and these changes might have influenced the establishment of fermentative ruminal functions during weaning transition.

Keywords: ruminal pH, ruminal bacteria, calf weaning, 454 pyrosequencing, forage

## INTRODUCTION

Weaning transition is defined as the period of transition from liquid to solid feed consumption, which is critical for the development of an active and functional rumen (National Research Council, 2001). For example, calves fed starch sources during weaning transition exhibit increased volatile fatty acid (VFA) and lactic acid production, which decreases ruminal pH (Laarman and Oba, 2011). Increased VFA production, especially butyrate, via solid feed fermentation in the developing rumen is responsible for functional ruminal epithelial tissue development (Sander et al., 1959). Conversely, excessive amounts of rapidly fermentable carbohydrates in feed can cause a sudden decrease in ruminal pH, which is associated with immunosuppression and inflammation (Kleen et al., 2003; Gozho et al., 2005). Under low ruminal pH conditions, increased amounts of free ruminal lipopolysaccharides (LPS) translocate into the blood, activating an inflammatory response (Gozho et al., 2005). While subacute ruminal acidosis (SARA) does not adversely affect calf performance during weaning transition, decreasing dietary calf starter consumption does not alleviate ruminal acidosis in calves (Laarman et al., 2012). In contrast, hay consumption might be important in mitigating ruminal acidosis in dairy calves during weaning transition (Laarman and Oba, 2011).

Ruminal bacteria can adapt adequately to dietary changes. For example, bacterial diversity and population size in the ruminal epithelium are affected by dietary changes (Liu et al., 2015), and epimural bacterial communities differ among cattle fed highgrain and forage diets (Petri et al., 2013a). In goats, a high-grain diet not only decreased ruminal pH but also caused a strong shift in the epimural bacterial community, which was associated with alterations in the relative expression of toll-like receptors in the ruminal epithelium (Liu et al., 2015). Fructose feeding increased the relative abundance of Firmicutes and decreased that of Proteobacteria after a short period, and Streptococcus bovis was specifically observed in fructose-fed heifers, which was identified by a false discovery rate analysis (Golder et al., 2014). Furthermore, rumen bacterial microbiota differed in bacterial diversity, richness, and composition between dairy cattle fed a control diet vs. a SARA-inducing diet (Mao et al., 2013). Among cattle with induced SARA, reductions in bacterial diversity and abundance of Gram-negative bacteria were observed, which were directly correlated with an increase in ruminal LPS levels (Mao et al., 2013).

Recent studies have found that rumen microbial communities are established soon after birth, before solid feed consumption (Jami et al., 2013; Rey et al., 2014). Rumen bacteria can be transmitted from the dam via the birth canal, teat surface, skin, or saliva; therefore, calf rumen bacterial communities depend greatly on maternal interactions (Chaucheyras-Durand and Ossa, 2014). Moreover, a diet of milk, compared with milk plus solid feed, fed to calves during the first 3 weeks of life differentially affected microbial communities in the gastrointestinal tract and feces, as well as body weight and ruminal pH and weight (Guzman et al., 2015). Some bacteria essential for mature rumen function can be detected as early as 1 day after birth, long before the ruminal bacterial community is established (Jami et al., 2013). Prevotella was observed to be the predominant genus upon a rapid increase in solid feed intake from 15 to 83 days of age (Rey et al., 2014) as well as in animals fed high-fiber diets (Jami et al., 2013). Although it has been demonstrated that rumen ecology is influenced by maternal dams, feeding materials, and rumen conditions from very soon after birth into adulthood, the correlation between ruminal pH and bacteria during weaning transition remains unclear.

The objective of this study was to investigate the relationship between ruminal pH and bacteria in calves fed calf starter with and without a forage diet during weaning transition. We hypothesized that forage consumption would mitigate adverse changes in ruminal pH and that bacteria would change due to these alterations.

### MATERIALS AND METHODS

### Animals and Experimental Design

All animals were cared for according to protocols approved by Iwate University Laboratory Animal Care and Use Committee. In total, 16 4-week-old Holstein bull calves were obtained from dairy farms and equipped with rumen cannulas by cannulation surgery 4 days after arrival. Calves were housed individually in 2.0 × 1.2-m pens with rubber mats and had free access to water throughout the study period. For the experiment, the calves were subjected to 3 weeks of adaptation (4–6 weeks of age), a preweaning phase (7 weeks of age), weaning transition (8 weeks of age), and a post-weaning phase (9–11 weeks of age). The diet was supplied in two equal portions at 08:00 and 16:30 daily, and daily total dry matter intake (DMI) was recorded individually for each calf throughout the study. The chemical compositions of the milk replacer, calf starter concentrate, and mixed forage (orchard and timothy hay) fed to calves are shown in **Supplementary Table 1**.

During the adaptation period, calves were fed 300 g/1.8 L commercial milk replacer (Meiji Feed Co., Kashima, Japan), 600 g calf starter (Meiji Feed Co., Kashima, Japan), and 200 g forage. At 6 weeks of age, calves (73.5 ± 4.2 kg; mean ± SE) were divided into two groups, the HAY (n = 8) and CON groups (n = 8). At 7 weeks of age, weaning was started in both groups by reducing the milk replacer to 150 g/0.9 L, and forage was restricted in the CON group until the end of the study. All calves were weaned at 8 weeks of age. From 7 to 11 weeks of age, the amount of calf starter was gradually increased from 800 to 1,200 g in the HAY group and 800 to 1,600 g in the CON group, while the amount of forage was concurrently increased from 200 to 400 g in the HAY group. Both groups had the same total DMI from weaning transition to a post-weaning phase. The amount of feeding was based on the Japanese Feeding Standard for Dairy Cattle.

#### Sampling and Measurements

Ruminal pH was measured continuously every 10 min throughout the experiment using a radio transmission system (YCOW-S; DKK-TOA Yamagata, Yamagata, Japan) as reported previously (Sato et al., 2012). The pH sensor was placed in the ventral sac of the rumen, and rumen fluid samples were collected from the ventral sac of the rumen, adjacent to the pH sensor, in the morning before feeding at 7, 8, 9, and 11 weeks of age (−1, 0, 1, and 3 weeks after weaning, respectively). Rumen fluid samples were immediately filtered through two layers of cheesecloth after sampling. For the VFA analysis, 2 mL 25% HO3P in 3N H2SO<sup>4</sup> were added to 10 mL rumen fluid. Total VFA and VFA components (i.e., acetic acid, propionic acid, and butyric acid) were separated and quantified by gas chromatography (Model 135, Hitachi, Tokyo, Japan) using a packed glass column (Thermon-3000, 3%) on a Shimalite TPA 60–80 mesh support (Shinwa Chemical Industries Ltd., Kyoto, Japan). Filtered rumen fluid samples were stored at −80◦C until further analysis.

### DNA Isolation

For DNA isolation, rumen fluid samples were thawed, and 250 µL aliquots were centrifuged at 9,700 × g for 30 min, after which the supernatant was discarded. For each sample, the pellet was re-suspended in 300 µL TE buffer, and total bacterial DNA was extracted as described previously (Morita et al., 2007) with minor modifications. The mixture was incubated with 750 µg/mL lysozyme (Sigma-Aldrich, St. Louis, MO, USA) at 37◦C for 90 min. Then, 10 µL purified achromopeptidase (Wako Pure Chemical Industries, Ltd., Osaka, Japan) were added at a concentration of 10,000 U/mL and incubated at 37◦C for 30 min. The suspension was treated with 60 µL 1% sodium dodecyl sulfate and 1 mg/mL proteinase K (Merck Japan, Tokyo, Japan), and incubated at 55◦C for 5 min. The lysate was treated with phenol/chloroform/isoamyl alcohol (Wako Pure Chemical Industries, Ltd.) and chloroform (Life Technologies Japan, Ltd., Tokyo, Japan). DNA was precipitated by adding 5M NaCl and 100% ethanol and centrifuged at 21,900 × g for 15 min. The DNA pellet was rinsed with 70% ethanol, dried, and dissolved in TE buffer. The purified DNA was quantified using a Biospecnano (Shimadzu, Kyoto, Japan) and stored at −80◦C until further analysis.

### DNA Pyrosequencing

The V1/V2 region of the 16S rRNA gene was amplified using a forward primer (5′ -CCATCTCATCCCTGCGTGTCTCCGACT CAGNNNNNNNNNNAGRGTTTGATYMTGGCTCAG-3′ ) containing 454 primer A, a unique 10-bp barcode sequence for each sample (indicated as N), and 27Fmod (5′ -AGRGTT TGATYMTGGCTCAG) in which the third base, A, in the original primer 27F was changed to R, as well as the reverse primer (5′ -CCTATCCCCTGTGTGCCTTGGCAGTCTCAG TGCTGCCTCCCGTAGGAGT-3′ ) containing 454 primer B and reverse primer 338R (5′ -TGCTGCCTCCCGTAGGAG T). Amplified products of ∼370 bp were confirmed using agarose gel electrophoresis, purified using AMPure XP magnetic purification beads (Beckman Coulter, Inc., Brea, CA, USA), and quantified using the Quant-iT PicoGreen dsDNA Assay Kit (Life Technologies Japan). Mixed samples were prepared by pooling approximately equal amounts of PCR amplicons from each sample and then subjected to 454 GS Junior (Roche Applied Science, Indianapolis, IN, USA) sequencing following the manufacturer's instructions. The sequencing data were deposited into the Sequence Read Archive (SRA) of NCBI1 and can be accessed via accession number SRP076881.

### Pyrosequencing Data Analysis

All pyrosequencing reads were filtered according to the procedure of Kim et al. (2013), who developed an analysis pipeline for barcoded 454 pyrosequencing of PCR amplicons in V1/V2, the region amplified by the 27Fmod/338R primers. Pyrosequencing reads were assigned to each sample based on the barcode sequence information. Resulting sequences that did not have PCR primer sequences at both sequence termini and those with an average quality value >25 were filtered out. Chimeras containing BLAST match lengths of >90% similarity with reference sequences in the database were removed. Finally, filter-passed reads were obtained for further analysis by trimming off both primer sequences. For the operational taxonomic unit (OTU) analysis, 16S reads were clustered using a 96% pairwise identity cutoff in the UCLUST program (www.drive5.com). Representative sequences for each OTU were assigned to bacterial species based on a BLAST search, with a 96% pairwise identity cutoff, against the 16S rRNA gene sequence database, constructed using Ribosomal Database Project tools (ver. 10.27; http://rdp.cme.msu.edu/), and against a reference genome database constructed from genome sequences collected from GenBank (ftp://ftp.ncbi.nih. gov/genbank/, November 2013).

A total of 506,717 filter-passed sequences were obtained from the analysis pipeline (Kim et al., 2013). Filter-passed reads were processed using MOTHUR (ver. 1.35, University of Michigan; http://www.mothur.org/wiki/; Schloss et al., 2009) and all samples were standardized by random subsampling to 3420 sequences per sample using the "sub.sample" command to generate rarefaction curves and calculate the abundancebased coverage estimator (ACE), Chao1 richness estimator, and Shannon diversity index, according to the Illumina MiSeq protocol described previously (Kozich et al., 2013). In order to obtain a non-redundant set of sequences, unique sequences were determined, and used to align against the SILVA reference alignment database (Pruesse et al., 2007); chimera were removed using chimera.uchime (http://drive5.com/uchime); sequences identified as being of eukaryotic origin were removed; the candidate sequences were screened and preclustered to eliminate outliers; and a distance matrix was generated from the resulting sequences. Sequences were clustered into OTU with a cutoff of 97% similarity. Rarefaction curve was generated at the level of 97% similarity level, which was calculated by the distancebased OTU (Schloss et al., 2011). For calculation of the nonparametric species richness estimators Chao 1 and ACE, the diversity index Shannon, the "summary.single" command was used. The unweighted UniFrac distance method (Lozupone and Knight, 2005) was used to perform a principal coordinates analysis (PCoA) with all OTU.

#### Real-Time Quantitative PCR

Quantitative real-time PCR (qRT-PCR) was performed to evaluate the copy number of methyl-coenzyme M reductase α-subunit (mcrA) from total methanogens and 16S rRNA genes from Fibrobacter succinogenes, Megasphaera elsdenii, Ruminococcus albus, Ruminococcus flavefaciens, Streptococcus bovis, and Selenomonas ruminantium using SYBR green (iQ SYBR Green Supermix, Bio-Rad, Hercules, CA, USA) with the MiniOpticon Real-Time PCR system (Bio-Rad). Primer pairs (**Supplementary Table 2**) were selected to detect bacterial species closely associated with dietary changes and other bacterial species. Each sample contained 10 ng DNA, 2× SYBR green, and 0.6 µM each primer in a final volume of 20 µL. Amplification conditions were as follows: 95◦C for 3 min, 40 cycles of 10 s at 95◦C, 20 s at 63◦C (for total methanogens), 60◦C (for F. succinogenes), 58◦C (for M. elsdenii), 57◦C (for S. bovis and S. ruminantium), or 55◦C (for R. albus and R. flavefaciens), and 30 s at 72◦C. The fluorescence signal was collected at the end of each cycle. To obtain melting curve data, the temperature was increased in 0.5◦C increments from 65 to 94◦C. A standard curve for each primer pair was constructed from recombinant plasmid DNA containing 16S rRNA inserts of DNA purified from a pure culture of the target species. The strains used for plasmid preparation were as followed: Methanobrevibacter ruminantium JCM13430 (DSM1093), F. succinogenes ATCC19169, M. elsdenii ATCC25940, R. albus ATCC27210, R. flavefaciens ATCC19208, S. bovis ATCC33317, and S. ruminantium ATCC12561. Plasmid DNA was quantified and subjected to seven sequential 10-fold dilutions. Data were collected and processed using CFX Manager software ver. 1.5 (Bio-Rad).

#### Statistical Analysis

Ruminal pH data were summarized as 24- and 1-h means. VFA, relative abundance of bacterial phyla and genera, ruminal bacterial diversity indices, and bacterial species copy numbers were summarized at −1, 0, 1, and 3 weeks after weaning. All numerical data were expressed as means ± standard error (SE) and analyzed using Prism ver. 7.01 (GraphPad Software, Inc., La Jolla, CA, USA). The normality of the distribution of variables was tested using Shapiro-Wilk test, and non-normal data were root-square transformed before analysis. Total DMI at 6 weeks of age, 24- and 1-h mean ruminal pH, total VFA concentration, proportions of individual VFA, relative abundance of each OTU, ruminal bacterial diversity indices, and bacterial species copy number were compared between the HAY and CON groups using two-way repeated measures ANOVA, and multiple testing false discovery rate (FDR) p-value was determined (Benjamini and Hochberg, 1995). The statistical model included the main effects of diet, time, and their interaction, plus the random effect of animal. Differences were considered to be significant at P < 0.05, and trends suggesting possible significance were determined at P < 0.10.

#### RESULTS

#### Daily Total DMI, Ruminal pH, and VFAs

No significant difference (P = 0.283) in total DMI was observed at 6 weeks of age between the two groups. The 24-h mean ruminal pH was significantly (P < 0.01) different between the HAY and CON groups (**Figure 1**). The interaction of diet × time

of sampling and the effect of time were significant (P < 0.01) for the 24-h mean ruminal pH between the groups. The ruminal pH in both groups decreased after the morning feeding and then began to increase up to 3 h after feeding at −1 week after weaning (**Figure 2**). Diurnal changes in the 1-h mean ruminal pH were observed at −1, 0, 1, and 3 weeks after weaning in the HAY group (**Figure 2**). However, they were not observed at 0 and 1 weeks in the CON group and were weak at 3 weeks after weaning. The interaction of diet × time of sampling and the effect of time were significant (P < 0.01) for the 1-h mean ruminal pH between the two groups throughout the experiment.

Significant differences (P < 0.01) were observed in the proportions of acetate and propionate between the HAY and CON groups (**Table 1**). The proportion of acetate was significantly (P < 0.05) higher at 1 and 3 weeks in the HAY group, and the proportion of propionate was significantly (P < 0.05) higher at 1 and 3 weeks after weaning in CON group. The proportion of butyrate was significantly (P < 0.05) higher at 3 weeks after weaning in the HAY group. Others that VFA components exclude acetate, propionate, and butyrate from total VFA was significantly (P < 0.05) higher at 0 week after weaning in the HAY group. In addition, a significant (P < 0.05) difference in the ruminal acetate-to-propionate ratio was observed between the two groups. Furthermore, the ruminal acetate-to-propionate ratios were significantly (P < 0.05) lower at 1 and 3 weeks after weaning in the CON group. The effect of time was not significant for the ruminal acetate-to-propionate ratio, total VFA concentration, and individual VFA proportions. The interaction of diet × time of sampling was significant for the proportion of acetate, propionate, butyrate, and the ruminal acetate-to-propionate ratio, whereas others VFA component was differed slightly (P = 0.065) between the HAY and CON groups.

FIGURE 2 | Diurnal changes in the 1-h mean ruminal pH in Holstein bull calves fed calf starter with (HAY group, n = 8) and without (CON group, n = 8) forage. Week −1, Week 0, Week 1, and Week 3 represent calves at − 1, 0, 1, and 3 weeks after weaning, respectively. Values represent the means ± SE. Arrows indicate feeding times (08:00 and 16:30). <sup>a</sup> and <sup>b</sup> denote significant differences (*P* < 0.05 and *P* < 0.01, respectively) between the HAY and CON groups at the same time point.

TABLE 1 | Total VFA concentration and individual VFA proportions of rumen fluid in Holstein bull calves fed calf starter with (HAY group) and without (CON group) forage.


*<sup>1</sup>VFA components excluded acetate, propionate, and butyrate from total VFA.*

*<sup>a</sup>*,*bMeans within a row, different superscripts differ (P* < *0.05) between the HAY and CON groups at the same week point.*

### Bacterial Diversity Analysis

The rarefaction curve of ruminal bacterial microbiota calculated at a 97% similarity level indicated that the HAY group had higher bacterial diversity than that of the CON group (**Supplementary Figure S1**). An unweighted UniFrac distance analysis in MOTHUR was used to evaluate β-diversity across the samples. The PCoA results indicated that the HAY group was distinctly separate from the CON group in the plot from weaning transition to post-weaning phase (**Figure 3**; PC1 + PC2 = 36.1%). The HAY and CON groups of the plot displayed close distance at −1 week, whereas those of the two groups were separated by age from 0 to 3 weeks after weaning. The plots of each group showed marked similarities between 1 and 3 weeks after weaning.

Bacterial diversity was estimated in the HAY and CON groups each week using OTUs, ACE, Chao1, and the Shannon index (**Table 2**). The OTUs, ACE, and Chao1 results differed significantly (P < 0.05) between the two groups and decreased after −1 week after weaning. However, they rebounded at 0 and 1 weeks after weaning in the HAY and CON group, respectively. The Shannon index analysis showed a tendency (P = 0.055) toward different bacterial diversity between the two groups. The interaction of diet × time of sampling and the effect of time were not significant for the OTUs, ACE, Chao1, and Shannon index between the HAY and CON groups.

after weaning in the CON group (*n* = 8). Blue circles represent the HAY group

#### Bacterial Abundance

A total of 16 bacterial phyla and 1 candidate phylum were identified within the ruminal bacteria. Of the major phyla, Firmicutes, Bacteroidetes, Actinobacteria, and Proteobacteria were the most abundant in both groups, accounting for 88.56% of the total ruminal bacteria (**Figure 4**). The remaining phyla had low relative abundances of <1%, reporting threshold of the relative abundance. A total of 341 bacterial genera were identified, and the relative abundances of 327 genera comprised <1% of the total sequences. There were 72 bacterial genera specific to the HAY group, 47 bacterial genera specific to the CON group, and 222 bacterial genera common to both groups. Prevotella was the predominant genus in both groups. Prevotella (22.84%), Lactobacillus (9.48%), and Ruminococcus (5.41%) were the most abundant bacterial genera in the HAY group, while Prevotella (18.16%), Olsenella (10.50%), and Lactobacillus (7.38%) were the most abundant genera in the CON group. Ruminal bacterial phylum and genus which had the relative abundance of below our reporting threshold (<1%) were excluded from the report.

The relative abundances of bacterial phyla (P < 0.05), including Bacteroidetes, Actinobacteria, and Tenericutes, and the Firmicutes-to-Bacteroidetes ratio (P < 0.01) differed significantly between the HAY and CON groups (**Table 3**). The genus Prevotella, which belongs to Bacteroidetes, was the only bacterial genus that differed significantly (P < 0.05) between the two groups, and there were no significant differences in other bacterial genera. The diet × time of sampling interaction was significant for Dialister (P < 0.01) and unclassified Lachnospiraceae (P < 0.05) and showed a tendency toward significance (P = 0.072) for Syntrophococcus. The effect of time was significant for Bulleidia (P < 0.05) between the two groups.

#### Copy Number of Bacterial 16S rRNA

The copy numbers of R. albus differed significantly (P < 0.01), while those of R. flavefaciens differed slightly (P = 0.087) between the HAY and CON groups (**Table 4**). The higher copy numbers of R. flavefaciens, R. albus, and S. bovis in the HAY group were observed compared with the CON group. Total methanogens (P < 0.05) and M. elsdenii (P < 0.01) copy numbers were affected significantly by the effect of time. The diet × time of sampling interaction was not significant for changes in the bacterial species copy numbers.

TABLE 2 | Ruminal bacterial diversity calculated from 454 pyrosequencing data at a 97% similarity level in Holstein bull calves fed calf starter with (HAY group) and without (CON group) forage.


*<sup>a</sup>Operational taxonomic units.*

*<sup>b</sup>Abundance-based coverage estimator.*

and red triangles the CON group.

#### DISCUSSION

This study aimed to identify the long-term relationship between ruminal pH and bacteria during weaning transition, and rumen fluid samples were collected in the morning before feeding to minimize the short-term effects of diet on ruminal bacteria that observed in grain-, fructose-, and histidine-fed dairy heifers (Golder et al., 2014). Total DMI at 2 weeks before weaning and the analyses of total VFA, individual proportion of VFA, PCoA, bacterial diversity indices (OTUs, Chao1, and ACE), and the copy number of ruminal bacteria at 1 week before weaning indicate that there was no difference in DMI and the rumen environment between the two groups. Moreover, ruminal bacteria was affected by low ruminal pH regardless of sample type, such as rumen fluid, rumen contents, and rumen epithelium (Mao et al., 2013; Liu et al., 2015; Sato, 2016). In our study, the feeding of calf starter with forage alleviates the depression of 24-h mean ruminal pH compared with the feeding of only calf starter, and the lower ruminal pH in the CON group was associated with greater consumption of calf starter. The diurnal changes in the 1-h mean ruminal pH observed in the HAY group play an important role in increasing the 24-h mean ruminal pH. However, the relationship between ruminal pH regulation and VFA absorption could not be explained in this study due to insufficient sampling throughout the day. Production of acetate, butyrate, and propionate and ruminal acetate-to-propionate ratios in the CON group were consistent with general feature of starch source feeding in calves (Laarman et al., 2012; Castells et al., 2013).


TABLE 3 | Relative abundances (% of total sequences) of the major bacterial phyla and genera identified by 454 pyrosequencing in Holstein bull calves fed calf starter with (HAY group) and without (CON group) forage.

*<sup>a</sup>*,*bMeans within a row, different superscripts differ (P* < *0.05) between the HAY and CON groups at the same week point.*

TABLE 4 | Copy number of 16S rRNA genes identified from qRT-PCR in Holstein bull calves fed calf starter with (HAY group) and without (CON group) forage.


*Values are expressed as log.*

Khan et al. (2011) found that providing chopped hay to calves fed a high volume of milk at an early age improved their total solid feed intake and was beneficial for rumen development. Moreover, calves fed diets supplemented with oat hay had increased ruminal pH than that of calves offered no forage (Castells et al., 2013), and providing chopped hay was necessary soon after weaning to improve calf performance (Terré et al., 2013). In our study, calves in both groups had a ruminal pH of <5.8 at −1 week after weaning (**Figure 2**), which is applied as a diagnostic of SARA in dairy calves (Laarman et al., 2012). However, the ruminal pH of <5.8 was mitigated in calves fed calf starter with forage, while that of calves fed only calf starter was maintained throughout the experiment. The higher ruminal pH in the HAY group compared with the CON group could have been caused by hay intake, stimulating chewing and salivary buffer flow (Laarman et al., 2012), and by part of the concentrate, the fermentation source, being replaced by forage during weaning transition.

The impact of dietary changes on rumen microbial composition has been investigated in several ruminants using a variety of molecular techniques. For example, terminal restriction fragment length polymorphism analysis indicated that the predominant rumen bacterial shift during SARA was a decline in the Gram-negative Bacteroidetes, induced by either grain or alfalfa pellets (Khafipour et al., 2009). In addition, 454 pyrosequencing analysis showed that the relative abundances of the phyla Bacteroidetes and Proteobacteria were reduced by consumption of concentrated feed in cattle with repeatedly induced SARA (Sato, 2016). In our study, lower relative abundances of the phyla Bacteroidetes and Tenericutes and Gram-negative bacteria in the CON group could be partly explained by a low rumen pH, which can lead to death and lysis of Gram-negative bacteria (Nagaraja and Titgemeyer, 2007). The higher Firmicutes-to-Bacteroidetes ratio in the CON group, due to a lower relative abundance of Bacteroidetes, was consistent with a previous study (Golder et al., 2014). In both groups, Firmicutes was the most relatively abundant Gram-positive bacteria, suggesting an increase in bacterial species that were metabolically capable of consuming newly available fermentable carbohydrates (Mao et al., 2013).

Liu et al. (2015) reported that a high-grain diet decreased the ruminal pH and resulted in lower bacterial diversity of the rumen epithelial community than hay diet. In this study, the ruminal bacterial diversity was also affected by diet, and feeding calf starter decreased 24-h mean ruminal pH and modified the composition of the ruminal bacterial microbiota at a 97% similarity level. Late increase in OTUs, ACE, and Chao1 observed in the CON group indicated that feeding only calf starter may signify the decline in rumen bacterial diversity during weaning transition. Furthermore, the higher ruminal bacterial diversity indices identified in the HAY group at a postweaning phase suggested that feeding calf starter with forage enhances the increase in rumen bacterial diversity after weaning. Therefore, dietary forage supplementation that increases 24- and 1-h mean ruminal pH would increase rumen bacterial diversity and promote rapidly recovery from damage during weaning transition.

The most common genera of bacteria detected in the rumen of cattle linked to SARA and acidic challenge are Lactobacillus and Streptococcus (Petri et al., 2013b). Moreover, the number of Lactobacillus and S. bovis in the rumen contents of dairy cows during the transition period increased by switching from a low- to a high-grain diet (Wang et al., 2012). In our study, the most abundant lactate-producing genera were the relative abundance of Lactobacillus (9.48%) and Olsenella (10.50%) in the HAY and CON groups, respectively. The growth rate of Lactobacillus decreased linearly with increases in the concentration of sugars, mostly due to the osmotic stress exerted by the sugars (Narendranath and Power, 2005). Moreover, Olsenella, Atopobium, and Bifidobacterium, which constituted the lactic acid-producing bacteria sensu lato (Inês et al., 2008), accounted for up to 97.5% of the total relative abundance of Actinobacteria genera in the CON group. Therefore, it can be assumed that feeding relatively higher amount of starch source in the CON group affects the composition of lactate-producing bacteria and their growth might be enhanced by increasing the amount of calf starter feed in both groups.

Streptococcus and S. bovis were not affected by diet in this study. The relative abundance of Streptococcus identified by 454 pyrosequencing was below our reporting threshold, and the qRT-PCR results showed that S. bovis copy numbers did not differ significantly between the two groups, although they were higher in the HAY group at 3 weeks after weaning compared with the CON group. Because S. bovis is not always the main cause of rumen acidity (Hungate, 1966), not all studies have observed increases in or even identified S. bovis in grain-fed cattle (Tajima et al., 2000; Klieve et al., 2003). Ruminococcus bacteria, known as cellulolytic bacteria, was higher in the CON group at a preweaning phase and weaning transition in the 454 pyrosequencing results, while the qRT-PCR analysis revealed the higher R. albus and R. flavefaciens copy numbers in the HAY group at a post-weaning phase. Because lower amounts of Ruminococcus were likely due to the decrease in forage supplementation and not specifically a result of acidosis (Petri et al., 2013b), the higher R. albus and R. flavefaciens copy numbers in the HAY group at a post-weaning phase might have been induced by forage supplementation in this study. Although Ruminococcus consists largely of cellulolytic bacteria, there are also Ruminococcus species, such as Rumincoccus bromii, that can utilize starch (Klieve et al., 2007). Therefore, a major inconsistency was observed in the analysis results between the two methods for Ruminococcus due to the increase in starch fermentable Ruminococcus during weaning transition.

Lactate-metabolizing species such as M. elsdenii increase proportionately as the bacterial community adapts to more readily fermentable carbohydrates (Huber, 1976), which also leads to an increase in the prevalence of starch-fermenting bacteria such as S. bovis (Hungate, 1966). Ruminal methanogens contribute to eliminating reducing equivalents produced by carbohydrate-fermenting bacteria and protozoa by removing hydrogen generated during fermentation (Whitman et al., 2006). Aschenbach et al. (2011) reported that M. elsdenii populations were synchronized with S. bovis populations, indicating that they could assist in preventing lactic acid acidosis (Russell et al., 1981). However, according to our qRT-PCR analysis, the M. elsdenii copy number was changed with a low relative abundance of Streptococcus and S. bovis copy number in both groups, indicating that M. elsdenii might be associated with relatively more associated with Lactobacillus in the HAY group and Olsenella in the CON group than other lactate-producing species. Notably, increasing the amount of total DMI could increase the rate of passage from the rumen, thereby leaving less time for microbial fermentation (Huhtanen and Kukkonen, 1995). Increased passage rates could shift methanogenesis to the hindgut and manure (Hindrichsen et al., 2006), which could explain the gradualreduction in ruminal total methanogens observed in both groups, concurrent with the gradual increase in total DMI during weaning transition.

Due to their capacity to use a large variety of substrates, including starches, other non-cellulosic polysaccharides, and simple sugars, as energy sources to produce succinate as the major fermentation end product (Purushe et al., 2010), Prevotella bacteria can dominate and thrive under a range of diets (Stevenson and Weimer, 2007; Bekele et al., 2010). Moreover, Prevotella ruminicola was found in 1-day-old calves, and increased in number by day 3 (Jami et al., 2013). Although most Prevotella strains in the rumen represent species other than the classical ruminal Prevotella spp., recent studies have observed a clear predominance of Prevotella in bacterial populations. For example, the relative abundance of this genus accounted for up to 19.7% of total bacteria, whereas the representative species Prevotella bryantii and P. ruminicola accounted for only 0.6 and 3.8%, respectively (Bekele et al., 2010). Rey et al. (2014) suggested that Prevotella is associated with diets containing solid food; calves that start eating solid foods earlier tended to develop rumen bacterial communities similar to those of adults earlier. In our study, Prevotella had the greatest relative abundance in the both groups, and the positive correlation between the relative abundance of Prevotella and solid feed consumption was consistent with previous research (Rey et al., 2014). These results indicate that Prevotella could constitute one of the most crucial members of the ruminal bacteria in dairy calves during weaning transition.

To our knowledge, little is known regarding how and when ruminal bacteria establish a stable rumen microbiome in dairy calves. In humans, at ∼1–2 years of age, the infant gut microbiome undergoes its second shift, and a stable adult microbiome begins to emerge, consistent with the establishment of a varied solid food diet (Bergström et al., 2014). In ruminant studies, changes in the rumen bacterial community caused by a transition from liquid to solid feed consumption have been mostly consistent with human studies (Jami et al., 2013; Rey et al., 2014). Therefore, the microbial shifts caused by dietary changes during weaning transition in this study were closely associated with the establishment of a mature rumen microbiome. However, dietary factors such as the composition of calf starter and forage type and timing of introduction should be considered carefully, although the causes and effects of the relationship between these factors have not been established (Yáñez-Ruiz et al., 2015). While this study contributes to a greater understanding of the response of the ruminal bacteria to dietary factors, further studies are needed to clarify the effects of diet on the establishment of ruminal bacteria from immature to mature animals.

#### REFERENCES

Aschenbach, J. R., Penner, G. B., Stumpff, F., and Gäbel, G. (2011). Ruminant nutrition symposium: role of fermentation acid absorption in the regulation of ruminal pH. J. Anim. Sci. 89, 1092–1107. doi: 10.2527/jas. 2010-3301

#### CONCLUSIONS

We investigated the relationship between ruminal pH and bacteria in calves fed calf starter with and without forage during weaning transition. The results supported our hypothesis that feeding calf starter with and without forage differentially affected the rumen environment. Feeding calf starter with forage mitigates the depression of 24-h mean ruminal pH due to diurnal changes in ruminal pH, in particular, a rebound from a rapid decrease in ruminal pH after feeding. Bacterial diversity was greater and recovered more rapidly from damage during weaning transition in the HAY group. Changes in the relative abundance and copy number observed in phyla, genus, and species might have affected the establishment of fermentative ruminal functions during weaning transition. This study increased understanding of the response of ruminal pH and bacteria to dietary factors in calves.

#### AUTHOR CONTRIBUTIONS

YK carried out majority of the experiment including animal care, DNA isolation, real-time PCR, and pyrosequencing data analysis and interpretation; TI, RN, and NO were responsible for animal care, VFA analysis and DNA isolation; KI, and SS contributed to the conception of the project; The manuscript was prepared by YK and SS.

#### ACKNOWLEDGMENTS

The author thanks Dr. Yasuo Kobayashi, Research Faculty of Agriculture, Hokkaido University, who kindly provided the recombinant plasmid DNA for qRT-PCR analysis. This research was financially supported by a KAKENHI Grant-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (No. 26292156).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01575

Supplementary Figure S1 | Rarefaction curves generated from the 454 pyrosequencing data. HW-1, HW0, HW1, and HW3 represent calves at −1, 0, 1, and 3 weeks, respectively, after weaning in the HAY group (*n* = 8), and CW-1, CW0, CW1, and CW3 represent calves at −1, 0, 1, and 3 weeks, respectively, after weaning in the CON group (*n* = 8).

Supplementary Table S1 | Chemical composition of milk replacer, calf starter concentrate, and mixed hay fed to calves.

Supplementary Table S2 | Primer sequences used for qRT-PCR.


**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 © 2016 Kim, Nagata, Ohtani, Ichijo, Ikuta and Sato. 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) or licensor 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.

# Colonization of the Intestinal Tract of the Polyphagous Pest *Spodoptera littoralis* with the GFP-Tagged Indigenous Gut Bacterium *Enterococcus mundtii*

#### Beng-Soon Teh<sup>1</sup> , Johanna Apel <sup>2</sup> , Yongqi Shao<sup>3</sup> and Wilhelm Boland<sup>1</sup> \*

*<sup>1</sup> Department of Bioorganic Chemistry, Max Planck Institute for Chemical Ecology, Jena, Germany, <sup>2</sup> Clinic for Internal Medicine II, Department of Haematology and Medical Oncology, University Hospital Jena, Germany, <sup>3</sup> Laboratory of Invertebrate Pathology, College of Animal Sciences, Zhejiang University, Hangzhou, China*

The alkaline gut of Lepidopterans plays a crucial role in shaping communities of bacteria. *Enterococcus mundtii* has emerged as one of the predominant gut microorganisms in the gastrointestinal tract of the major agricultural pest, *Spodoptera littoralis*. Therefore, it was selected as a model bacterium to study its adaptation to harsh alkaline gut conditions in its host insect throughout different stages of development (larvae, pupae, adults, and eggs). To date, the mechanism of bacterial survival in insects' intestinal tract has been unknown. Therefore, we have engineered a GFP-tagged species of bacteria, *E. mundtii,* to track how it colonizes the intestine of *S. littoralis*. Three promoters of different strengths were used to control the expression of GFP in *E. mundtii*. The promoter *ermB* was the most effective, exhibiting the highest GFP fluorescence intensity, and hence was chosen as our main construct. Our data show that the engineered fluorescent bacteria survived and proliferated in the intestinal tract of the insect at all life stages for up to the second generation following ingestion.

Keywords: *Spodoptera littoralis*, green fluorescent protein, promoter, lactic acid bacteria, *Enterococcus mundtii*, intestinal tract

## INTRODUCTION

Insects' guts harbor a wide range of microbial communities. Intestinal gut microbes contribute significantly to the development of their insect hosts by providing essential nutrients, aiding in food digestion, and protecting against other harmful pathogens. However, the functions of these microbes in the insect gut are still largely unknown due to the complexity and diversity of the microbes. In recent years, the agricultural pest, Spodoptera littoralis (Lepidoptera, Noctuidae) has been used as an experimental model insect to study gut microbiomes. The microbial composition in the gut of S. littoralis has been well characterized (Tang et al., 2012), yet the factors controlling its colonization are unknown.

Insect guts contain multiple compartments with different physicochemical conditions such as pH and oxygen availability which enrich for certain species of bacteria. The gut of certain lepidopteran, coleopteran, and dipteran is highly alkaline due to specific dietary preferences (Brune and Kühl, 1996; Harrison, 2001). The lepidopteran insects which feed on tannin-rich leaves have

#### *Edited by:*

*Martin Grube, Karl-Franzens-University Graz, Austria*

#### *Reviewed by:*

*Seungha Kang, CSIRO, Australia Morten Schiøtt, University of Copenhagen, Denmark*

> *\*Correspondence: Wilhelm Boland boland@ice.mpg.de*

#### *Specialty section:*

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

*Received: 05 March 2016 Accepted: 31 May 2016 Published: 14 June 2016*

#### *Citation:*

*Teh B-S, Apel J, Shao Y and Boland W (2016) Colonization of the Intestinal Tract of the Polyphagous Pest Spodoptera littoralis with the GFP-Tagged Indigenous Gut Bacterium Enterococcus mundtii. Front. Microbiol. 7:928. doi: 10.3389/fmicb.2016.00928* alkaline midguts with pH as high as 11–12 (Appel and Martin, 1990; Harrison, 2001). A clear pH gradient occurs along the lepidopteran midgut from highly alkaline (pH ∼ 10) anterior end to almost neutral posterior ends (Funke et al., 2008). The microbial community of S. littoralis (cotton leafworm) is dominated by Pantoea and Citrobacter from the phylum Proteobacteria in early-instar larvae (Shao et al., 2014). Bacteria in this phylum have the ability to degrade polysaccharide in insects (Anand et al., 2010; Adams et al., 2011; Engel et al., 2012). As insects aged toward late-instar, more than 97% of the total bacterial community shifted to Firmicutes, dominated mostly by Enterococcus and Clostridium sp. (Tang et al., 2012; Shao et al., 2014). The proliferation of Clostridia is linked to its role in cellulose digestion and fermentation of sugars (Watanabe and Tokuda, 2010). Interestingly, the alkaline midgut of gypsy moth larva also harbors Enterococcus (Broderick et al., 2004) while the Firmicutes dominates the midgut of the beetle Pachnoda ephippiata (Egert et al., 2003). Some insects harbor similar bacterial lineages in their alkaline guts.

To date, the genus Enterococcus is known to include more than 33 species (Kohler, 2007). The members of this genus are typically found in the intestinal tracts of humans and animals, in dairy products, and also in the environment: for example, in plant material, soil, and surface water (Giraffa, 2003; Ogier and Serror, 2008). E. mundtii is part of this genus. It is a non-motile, Gram-positive, facultative anaerobic organism that belongs to the group of lactic acid bacteria (LAB). It forms either cocci or rods, and is capable of producing lactic acid as a by-product of the fermentation of carbohydrates. The biological role of E. mundtii is still poorly understood, as most studies have focused on the model bacteria Enterococcus faecalis and Enterococcus faecium, which often cause human systemic infection (Arias and Murray, 2012).

Green fluorescent protein (GFP) originally isolated from Aequorea victoria has been extensively used as a reporter for gene expression in bacterial and mammalian cells (Yang et al., 1996; Valdivia and Falkow, 1997; Hazelrigg et al., 1998; Rolls et al., 1999). GFP is advantageous as it requires neither cofactors nor a substrate to be expressed in its host cells. Different variants of GFP, such as EGFP (enhanced green fluorescent protein) have been developed to improve fluorescent intensity (Cormack et al., 1996). The expression of GFP in several LAB has been successfully demonstrated (Scott et al., 2000; Hansen et al., 2001; Lun and Willson, 2004). In recent years GFP has been mostly used to investigate Gram-negative bacteria, and, less often, Gram-positive bacteria (Bubert et al., 1999; Freitag and Jacobs, 1999; Lewis and Marston, 1999; Fernandez de Palencia et al., 2000). Increasingly, GFP has been used to track how and where target bacterial species colonize the guts of several host insects (Thimm et al., 1998; Mumcuoglu et al., 2001; Husseneder and Grace, 2005; Kounatidis et al., 2009; McGaughey and Nayduch, 2009; Doud and Zurek, 2012).

In this work, we determine the fate of GFP-tagged E. mundtii within the digestive tract of S. littoralis when administered in vivo. In addition, we track the transmission route of the bacteria through all stages of the life cycle of S. littoralis. In fact, the incorporation of GFP-tagged E. mundtii provides a non-invasive monitoring of its survival in the insect gut, but still far from addressing the relationship between the insect and the bacterial symbiont. We are interested to further explore the underlying factors that drive this complex relationship by analyzing the bacterial and insect gut epithelial transcriptomes in future work. The transcriptome data will significantly expand our understanding of the functional roles of indigenous bacteria toward the development of the insect and other microbes. This can easily be done by identifying the insect- or microbe-derived compounds from the metabolic pathways resulted from the transcriptome data.

## MATERIALS AND METHODS

## Maintenance of Egg and Larvae

The eggs of S. littoralis were purchased from Syngenta Crop Protection Münchwilen AG (Münchwilen, Switzerland). Eggs were hatched at 14◦C. Larvae were maintained at room temperature (24◦C). Larvae were provided with sterile artificial diet made of white bean and essential nutrients without antibiotics and prepared based on Spiteller et al. (2000).

### Bacterial Strains and Growth Conditions

**Table 1** lists the bacterial strains and plasmids used in this study. Escherichia coli strain DH5α was used to maintain all GFP-containing plasmids. The plasmid pTRKH3-ermGFP (Addgene plasmid # 27169), pTRKH3-slpGFP (Addgene plasmid # 27168), and pTRKH3-ldhGFP (Addgene plasmid # 27167) were gifts from Michela Lizier. E. mundtii strain KD251 (isolated from the gut of S. littoralis at the Department of Bioorganic Chemistry, Max Planck Institute for Chemical Ecology) was used as the recipient of all plasmids (Shao unpublished). E. coli DH5α and E. mundtii were grown at 37◦C with agitation (220 rpm) in Luria-Bertani (LB) and Todd-Hewitt Bouillon, THB (Roth, Karlsruhe, Germany) medium for both broth and agar, respectively. Antibiotics were used at the following concentrations: erythromycin, 50µg ml−<sup>1</sup> (for E. coli) or 5µg ml−<sup>1</sup> (for E. mundtii). All plasmids were extracted from E. coli using the GeneJet plasmid miniprep kit (Thermo Scientific, Vilnius, Lithuania). All strains were kept in glycerol stocks at −80◦C for preservation and long-term storage.

### Plasmids

All GFP expression vectors were derived from pTRKH3, a backbone shuttle vector for E. coli and various species of LAB, including Streptococcus, Lactococcus, Enterococcus, and Lactobacillus (O'Sullivan and Klaenhammer, 1993). The vector carries a gene for erythromycin resistance which is highly suitable for expression in Enterococcus. In addition, the vector possesses a modified GFP 5 (mGFP5) that is controlled by three constitutive promoters of different strengths. pTRKH3 ermGFP harbors EGFP that is controlled by a strong enterococcal erythromycin ribosomal methylase (ermB) promoter (Swinfield et al., 1990). The Lactobacillus acidophilus lactate dehydrogenase (ldhL) promoter (Kim et al., 1991) and surface layer protein (slp) promoter (Boot and Pouwels, 1996) constitutively control pTRKH3-ldhGFP and pTRKH3-slpGFP, respectively.

#### TABLE 1 | Bacterial strains and plasmids used in this study.


*Em<sup>r</sup> , erythromycin resistant; Tet<sup>r</sup> , tetracycline resistant.*

#### Electroporation of *Enterococcus mundtii*

Electroporation was carried out based on the modified protocol of E. coli (Dower et al., 1988). A single colony of E. mundtii was grown at 37◦C in THB broth on a rotary shaker (Certomat BS-1 Sartorius, Goettingen, Germany) with agitation (220 rpm). An overnight culture was diluted 1:1000 in 100 ml of THB medium before being harvested by centrifugation at 4000 × g for 10 min (Sigma 3K18, Sigma, Germany) at 4◦C when growth reached the exponential phase (A600 nm approximately 2.2). The cells were washed with 100 ml of ice-cold distilled water, centrifuged as above and washed again with 50 ml of ice-cold water before being centrifuged again. The cells were then washed with 20 ml of 10% glycerol, centrifuged and finally suspended in 2 ml of 10% glycerol. The suspension was divided into 50µl aliquots and stored at -80◦C. Prior to electroporation, the frozen cells were thawed on ice and mixed with plasmid for 15 min before being transferred into a chilled 0.2 cm gap cuvette. Electroporation was performed by a single pulse at 1.8 kV (E = 9 kV/cm), 600 and 10 µF, with a pulse length of 3.6 ms in an electroporator 2510 (Eppendorf, Hamburg, Germany). The concentration of purified plasmids used during electroporation was between 0.15 and 0.2µg. The pulsed cells were immediately suspended with 950µl of THB broth and further incubated for 2 h at 37◦C with agitation (220 rpm) before 100µl was plated on THB agar containing 5µg ml−<sup>1</sup> of erythromycin. The plates were incubated at 37◦C for 48 h. Bacterial transformants containing target plasmids were verified by PCR screening.

### Verification of Bacterial Identity by 16S rRNA Sequencing

All bacterial transformants were checked for identity by PCR to prevent contamination. Total DNA was extracted from GFP-tagged bacteria of three different constructs from overnight culture by using a MasterPure Complete DNA and RNA purification kit (Epicentre, Madison, WI, USA) according to the manufacturer's protocol. The bacterial 16S rRNA genes were amplified using universal primers, 27f (5′ - AGAGTTTGATCCTGGCTCAG-3′ ) and 1492r (5′ - GGTTACCTTGTTACGACTT-3′ ). PCR was performed in a final volume of 50µl using 10µM of each primer, 10 mM concentration of deoxynucleoside triphosphates, 50 mM MgCl2, 1 U of Taq polymerase and buffer (Invitrogen, Carlsbad, CA, USA). Denaturation was performed at 95◦C for 2 min, followed by 30 cycles of 95◦C for 30 s, annealing at 54◦C for 30 s, and 72◦C at 1 min 30 s. The final extension was at 72◦C for 7 min. PCR products were purified using the PureLink Quick Gel Extraction and PCR Purification Combo Kit (Invitrogen, Carlsbad, CA, USA). The purified PCR products were sent for Sanger sequencing. DNA sequences were assembled with DNA baser sequence assembly software (http://www.dnabaser.com). The assembled sequences were used for blast searches at the National Center for Biotechnology Information (http://www. ncbi.nlm.nih.gov).

#### Feeding of *S. littoralis* Larvae with GFP Bacteria

A total of 50 first-instar larvae were fed artificial diet supplemented with antibiotics for 3 days in the following final concentration: ampicillin (5.75µg ml−<sup>1</sup> ) and erythromycin (9.6µg ml−<sup>1</sup> ). Each larva was fed small cubes (1 g) of artificial diet inoculated with E. mundtii-harboring pTRKH3-ermGFP for 1 day starting at day 6, followed by food without bacteria starting at day 7 until pupation. Control larvae were fed food without bacteria continuously. A single colony of bacteria was grown overnight in THB broth containing erythromycin (5µg ml−<sup>1</sup> ) and diluted 1:10 in the same broth before being fed. A total of 100µl from the 1:10 dilution broth (A600 nm approximately 0.65) containing ∼4.7 × 10<sup>7</sup> CFUs of GFP bacteria was applied to the food of the larvae. Every day, feces were removed to avoid re-inoculating the GFP bacteria.

### Quantification of Bacteria from the Intestinal Tract

Larvae (n = 6 for each stage), adults or pupae (n = 3 for each stage), and a control (n = 1 for each stage) were killed by freezing at −20◦C for 15 min. Each individual was surface sterilized in 70% ethanol and immediately rinsed in sterile distilled water. Guts were dissected in sterile 1 × PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4.7H20, and 2 mM KH2PO<sup>4</sup> [pH 7.4]), with sterile forceps under a stereomicroscope (Stemi 2000-C, Zeiss, Jena, Germany). Larval guts were excised into three sections: foregut, midgut, and hindgut. Gut tissues were aseptically homogenized in 100µl of PBS. A serial dilution of 10 fold was performed by transferring 100µl of the homogenized sample into 900µl sterile PBS, vortexing vigorously, and spreadplating 100µl of each dilution onto THB agar supplemented with erythromycin (5µg ml−<sup>1</sup> ). All plates were incubated at 37◦C for 48 h. Total bacterial cells were counted as colony forming units (CFUs) for each intestinal tract region. Erythromycin resistance was used as selection marker for picking bacterial colonies. In addition, to verify the presence of plasmid-containing GFP, PCR screening was performed.

#### Tissue Cross-Sectioning

The fresh gut tissues were cut into sections (foregut, midgut, and hindgut) and frozen at −24◦C in mounting medium for cryotomy (OCT compound, VWR, Leuven, Belgium) for 30 min. They were then cut with cryomicrotome (Microm Cryo-Star HM560 Cryostat, Walldorf, Germany) into 14–100µm sections.

#### Fluorescence Microscopy

The cultures containing GFP-producing bacteria were harvested, and the pellets were suspended in 1 × PBS. Bacterial suspensions of 20µl or slices of cross-sectioned tissue were mounted on microscope slides (Superfrost Plus, Thermo Scientific). Live cells were observed under an Axio Imager Z1 fluorescent microscope equipped with an AxioCam MRm camera (Zeiss, Jena, Germany). The GFP signal was detected using the filter set 10 (Cy2/GFP). All images were captured with a 63X magnification oil objective with an aperture of 1.4. The images were analyzed using the Axio Vision Rel 4.8 software (Zeiss, Jena, Germany). ImageJ, Fiji (Schindelin et al., 2012), an open-source software, was used to process all fluorescent images.

### DNA Extraction and PCR Amplification of *gfp* Gene

Total DNA was extracted from larvae and pupae at all instars, and from adults, by using a DNA kit as mentioned above. The 735 bp of gfp gene was amplified using a set of primers consisting of GFP3fw (5′ -TCGGAATTCATGAGTAAAGGAGAAGAA-3′ ) and GFP3rev (5′ - TCAGGATCCTTATTTGTATAGTTCATCC-3 ′ ) (Lizier et al., 2010). An EcoRI and a BamHI site (underlined) were introduced for forward and reverse primers, respectively. The PCRs were performed in a final volume of 20µl using 10µM of each primer, 10 mM concentration of deoxynucleoside triphosphates, 50 mM MgCl2, 1 U of Taq polymerase and buffer (Invitrogen, CA, USA). The following PCR conditions were used: 3 min at 94◦C, followed by 35 cycles of 45 s at 94◦C, 30 s at 60◦C, and 2 min at 72◦C, and final extension of 10 min at 72◦C.

### Western Blot

Bacterial cells were harvested from exponentially growing cultures. The cells were suspended in TE buffer (10 mM Tris-HCl, pH 8.0; 1 mM EDTA) containing 20% sucrose, lysozyme (1 mg ml−<sup>1</sup> ), RNase (1µg ml−<sup>1</sup> ), and DNase (1µg ml−<sup>1</sup> ) and further disrupted by repeating a freeze-thaw cycle. The protein extracts were subjected to sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) on a 4–12% gel (Laemmli, 1970). The proteins were transferred onto a PVDF transfer membrane, pore size 0.45µm (Thermo Scientific, Schwerte) with an electroblotter (Trans-blot Turbo Transfer System, BIO-RAD, Munich, Germany). The blots were blocked with 5% nonfat skimmed milk (NFDM) in TBS-T (Tris-buffer saline with Tween-20) for 1 h at room temperature. The membrane was then incubated for 16 h at 4◦C with the mouse primary antibody anti-GFP (Roche Applied Science, Rotkreuz, Switzerland) diluted 1:2000 in blocking buffer. After three washes in blocking buffer, the membrane was incubated for 1 h at room temperature with Anti-mouse lgG, HRP-linked Antibody (Cell Signaling Technology, Cambridge, UK) diluted 1:5000 in blocking buffer. The membrane was washed three times in blocking buffer followed by incubation with the chemiluminescent reagent for 1 min. In the dark room, the membrane was transferred onto a foil, and X-ray film (CL-XPosure Film, Thermo Scientific, Schwerte, Germany) was placed on top of it. The film was developed after different exposure times (3 s–10 min).

#### Flow Cytometry

Bacteria from overnight cultures were re-suspended and diluted 1:10,000 in 1 × PBS. Fluorescence was determined in a CyFlow Space (Sysmex Partec, Görlitz, Germany). The data were analyzed using the CyFlow Space Operating Software FloMax. A blue laser (488 nm) was used for GFP fluorescence detection.

#### Statistical Analysis

Bacterial plate counts between 30 and 300 colonies were included in the calculation. Samples with colonies above 300 may not be distinguishable from one another on a plate count, whereas those below 30 may not be representative of the sample (Madigan et al., 2009). The total number of fluorescent E. mundtiirecovered from each intestinal tract (foregut, midgut, and hindgut) across different larval stages was analyzed using JMP <sup>R</sup> 12.1.0<sup>1</sup> . Counts were analyzed using a one-way ANOVA test (P < 0.05). To further understand the different survival rates of GFP-E. mundtii at different larval stages, we compared the means of the combined three gut parts (foregut, midgut, and hindgut) as well as the means of individual gut regions of each larva using the Tukey– Kramer test (P < 0.05).

### RESULTS

### Comparison of Different GFP Constructs

Three different promoters, ermB, ldhL, and slp were used to control the expression of GFP, using pTRKH3 as a backbone shuttle vector. The strength of these GFP constructs was tested in E. mundtii by electroporation. This method was able to yield transformed colonies for all constructs. The recombinant bacterial colonies were picked and grown in THB at 37◦C overnight before GFP fluorescence was visualized by epifluorescence microscopy. The highest fluorescence intensity was detected for E. mundtii transformed with pTRKH3-ermGFP (**Figure 1A**), followed by pTRKH3-ldhGFP (**Figure 1B**) and no signal for pTRKH3-slpGFP (**Figure 1C**) as well as wild-type E. mundtii (**Figure 1D**). The bacterial cultures were all grown simultaneously for 24 h, equivalent to stationary phase. The GFP content represents the same amount of cells which was measured as OD600 nm.

We analyzed the total expressed GFP in E. coli DH5α and E. mundtii. Western blot results showed that the GFP gene

<sup>1</sup> JMP <sup>R</sup> Version 12.1.0. SAS Institute Inc., Cary, NC, USA, 2015.

was expressed in large quantities in E. mundtii and E. coli DH5α cells when the cells were transformed with pTRKH3 ermGFP. A thick band associated with the production of large amounts of GFP protein was observed in the immuno-blotting gel for both bacterial cells transformed with pTRKH3-ldhGFPexpressing plasmid. Low quantities of protein were produced by the slp promoter controlling the GFP expression in both bacteria. The wild-type bacteria did not express GFP protein as expected (**Figure 2**).

Flow cytometry analysis confirmed the results obtained by epifluorescence microscopy and western blot. As expected, overnight cultures of E. mundtti cells with pTRKH3-ermGFP were highly fluorescent (38.5%), whereas cultures of pTRKH3 ldhGFP (21.7%) were slightly fluorescent and those of pTRKH3 slpGFP (0.65%) showed almost no fluorescence (**Figures 3A–C**). Due to the efficiency of pTRKH3-ermGFP, this construct was chosen to transform E. mundtii and used for feeding experiments.

### Colonization of the Intestinal Tract of *S. littoralis* with Genetically Tagged Bacteria

GFP-tagged E. mundtii were fed to larvae of S. littoralis to visually monitor the persistence and fate of the bacteria within the digestive tract of different stages in the life cycle. We observed that fluorescent bacteria multiplied in the foregut, midgut, and hindgut regions after early ingestion of fluorescent bacteria. A high concentration of green fluorescent bacterial cells could be visualized in the foregut and midgut, but decreasing

amounts toward the hindgut region of the third-instar larvae (data not shown). During this early stage of ingestion, the density of bacteria was high in most parts of the gut tissues. GFP bacteria were seen scattered in the foregut of the fourth-instar larvae, around the peritrophic membrane as well as entering the epithelium, adjacent to the hemocoel and fat body (**Figure 4A**).

In the early stages of ingestion the fluorescent bacteria could be seen accumulating in the midgut of fifth-instar larvae (**Figure 4B**) where they clumped together in the region of the peritrophic membrane. GFP-tagged bacteria were seen starting to double in larvae from fourth to fifth instars (data not shown). In sixth-instar larvae, strikingly, bacteria were trapped in the nodules of granular hemocytes, suggesting the occurrence of phagocytosis (**Figures 4C,D**). The number of bacteria was reduced significantly in sixth-instar larvae in most parts of the gut. A sharp reduction in the density of fluorescent enterococci occurred during pupation (**Figure 4E**). Bacteria went from clusters to free-standing groups by attaching to the fat body of pupae; at this stage there was no inner gut in the pupae to keep them from moving around.

Viable fluorescent cells of E. mundtii were detected in the tracheole of the adult insect (**Figure 4F**). This shows bacteria were successfully transmitted from pupal to new adult gut tissue, although there were few or no other bacteria found in the tracheole. We also tested the transmission route of recombinant

bacteria by allowing individual adults to mate. Remarkably, we observed that a few bacteria were detected in the oocyte (Knorr et al., 2015) and none in the chorion of the eggs (**Figure 4G**), which proves GFP-tagged bacteria can survive after almost 30 days at various stages of the entire life cycle of S. littoralis. We also showed that, after hatching, fluorescent bacteria were detected in the muscular tissue of the first-instar larvae of second generation offspring (**Figure 4H**).

## Viable GFP Bacterial Cell Counts

Total fluorescent E. mundtii were recovered and counted from individual gut regions (foregut, midgut, and hindgut) on selective THB agar containing erythromycin. The mean of CFUs of bacteria recovered from each gut region showed significant difference between larval stages (F = 15.38; df = 2; P < 0.0001) by one-way ANOVA test. Further pairwise comparison revealed significant differences between the mean number of CFUs of combined gut parts between larvae in the fourth and fifth instars (likelihood ratio: 4.062; P < 0.0001) as well as the fifth and sixth instars (likelihood ratio: 3.048; P = 0.0006) but not between the fourth and sixth instars (likelihood ratio: 1.014; P = 0.3853; **Figure 5A**).

The number of E. mundtii in the foregut region was relatively low until to the fourth instar and transiently raised to 3.2 ± 1.9× 10<sup>6</sup> cells (P < 0.0001) during the fifth instar, followed by a decrease to 9.2 ± 8.6 × 10<sup>4</sup> cells (P < 0.0053) during the sixth instar. A sharp decrease by 97.1% occurred toward the

FIGURE 4 | Colonization of GFP-expressing *E. mundtii* in the *Spodoptera* intestinal tract. (A) Fluorescent image of the foregut region of fourth-instar larvae: bacteria are immobilized around the peritrophic membrane and gut epithelium (arrowheads) located adjacent to the hemocoel and fat body, scale bar = 20 µm. (B) In fifth-instar larvae, large clumps of bacteria are attached to the peritrophic membrane of the midgut tissue, scale bar = 10 µm. (C,D) Histological sections show fluorescent bacteria (arrowheads) are trapped within granular hemocytes containing nodules (black arrows) in the hindgut and midgut, leading to phagocytosis at the end of larval life, the sixth instar, scale bars = 10 µm. (E) A few bacteria are attached (arrowheads) to the fat body of pupae showing bacterial lysis occurs. (F) A single viable bacterial cell is immobilized in the tracheole of the adult, scale bar = 5 µm. (G) The fluorescent *E. mundtii* (arrowhead) is detectable in the oocyte of the eggs, scale bar = 10 µm. (H) Clusters of fluorescent bacteria (arrowheads) are scattered in the muscular tissue of the second generation first-instar offspring after hatching, scale bar = 10 µm. ch, chorion; ep, epithelium; fb, fat body; gh, granular hemocyte; hc, hemocoel; lu, gut lumen; mu, musculature; oc, oocyte; pm, peritrophic membrane; tr, tracheole. Magnification, (A–G), 630X; (H), 400X.

late-instar larval stage. The midgut CFU count rose during the fourth instar from a mean of 1.0 ± 0.6 × 10<sup>5</sup> to 2.7 ± 2.1 × 10<sup>7</sup> at the end of the fifth instar, representing a significant difference (P = 0.0425). In the sixth instar, bacterial counts fell to 4.0 ± 3.7 × 10<sup>6</sup> and showed no significant difference to larvae in the fifth and sixth instars (P = 0.1576). Also in the hindgut region there was a transient increase of bacterial counts from 2.2 ± 1.0 × 10<sup>5</sup> (fourth instar) to 1.5 ± 1.2 × 10<sup>7</sup> at the end of the fifth instar; this number fell by 94.5% to 2.2 ± 1.3 × 10<sup>5</sup> at the end of the sixth instar (P = 0.1087; **Figure 5B**). The ± values represent the standard error (SE). For some larvae, the hindgut region did not show any CFUs, possibly due to the high variation in the feeding behavior of individual larva. Overall, the number of bacteria in the intestinal tissues steadily increased from fourth- to fifth-instar larvae, but decreased tremendously during the sixth larval instar. The number of mean CFUs remained low during early pupation and slightly increased in the late pupation and adult stages (data not shown). No CFUs of fluorescent E. mundtii were detected from control larvae.

#### Tracking of Ingested GFP Bacteria by Colony PCR

The gut content of different stages of development of the insect was enumerated on selective agar plates. The bacterial colonies grown on agar were picked for colony-PCR experiments to verify the presence of GFP-containing plasmid. We were able to amplify the gfp gene of around ∼735 bp from bacterial colonies at all

stages of development (**Figure 6**). In addition, we could detect the GFP amplicon in fecal samples of all stages (data not shown), which confirmed that transgenic bacteria were present and could colonize the intestinal tract of S. littoralis.

### DISCUSSION

The GFP-expressing plasmids used in this study were derived from a common backbone E. coli-enterococcal shuttle vector (pTRKH3), which was controlled by three constitutive promoters of different strengths. This vector contained moderate copy numbers (30–40) in E. coli and a high copy number (45– 85) in Streptococcus and Lactococcus species (O'Sullivan and Klaenhammer, 1993; Papagianni et al., 2007). Moreover, it was stably maintained up to 25 generations without erythromycin and lost <4% after transformation into Lactococcus lactis (Papagianni et al., 2007). Our results showed that the strongest GFP expression signal was derived from the ermB promoter, which displayed the high fluorescence of recombinant bacteria upon detection by epifluorescence microscopy, western blotting and flow cytometry. This promoter is likely to be highly effective in many Enterococcus species, as it is derived from the broadhost range plasmid pAMß1 of E. faecalis (Swinfield et al., 1990). In Staphylococcus aureus, erythromycin resistance is caused by ribosome methylases encoded by ermA, ermB, and ermC genes which are involved in the methylation of 23S rRNA (Leclercq,

FIGURE 6 | Colony PCR-amplification of enumerated colonies of *E. mundtii* harboring GFP recovered from the intestinal tracts of larvae at different life stages. The *gfp* gene was amplified from fourth-, fifth,- and sixth-instar pupae and from adult insects. M, molecular weight marker (1-kb Plus DNA ladder, Invitrogen); Lane 1, fourth instar; Lane 2, fifth instar; Lane 3, sixth instar; Lane 4, pupa; Lane 5, adult; Lane 6, positive control (plasmid pTRKH3-*erm*GFP); Lane 7, negative control. The size of *gfp* gene is ∼735 bp.

2002). The addition of erythromycin antibiotic as substrate may increase the expression of ermB gene in E. mundtii, thus activates the GFP expression. In addition, the pTRKH3 vector which originates from pAMß1 may be suitable for replication in Grampositive bacteria. The strength of gfp gene expression controlled by these promoters was similar to that reported in Lactobacillus reuteri strains (Lizier et al., 2010).

The strength of promoter used to drive successful expression of heterologous proteins depends on strain and vary within LAB (McCracken and Timms, 1999). The constitutive ldh promoter is highly efficient in Lactobacillus casei (Pouwels et al., 2001) as well as in E. mundtii. It has been shown that the ldh gene is highly active in the logarithmic phase, but its expression decreases in the stationary phase in Lactobacillus helveticus (Savijoki and Palva, 1997). The low GFP expression signal from the slp promoter in our study may be due to different rate of transcription and translation of S-protein genes. In two Lactobacillus species, similar genes can be expressed with different regulatory mechanism (Pouwels et al., 1998). In some species of bacteria, the S-protein genes are controlled by multiple promoters (Vidgren et al., 1992), and some are preceded by a single promoter. The yield of mRNA controlled by multiple promoters might be higher than the yield directed by a single promoter. The regulation of S-protein gene expression is still not very well-known and may be growth-dependent. One of the five promoters upstream of S-protein gene in Brevibacillus brevis is active during all growth phases, while another promoter is only active during exponential growth (Adachi et al., 1989). It has been shown that the half-life of the S-protein mRNAs is different between bacterial species, Aeromonas salmonicida (22 min; Chu et al., 1993), Caulobacter crescentus (10–15 min; Fisher et al., 1988), and L. acidophilus (15 min; Boot et al., 1996). Another possible explanation of low GFP expression directed by a single slp promoter might be that E. mundtii do not synthesize S-layer protein which was also reported in L. casei as well (Masuda and Kawata, 1983).

Expression of gfp has effect on the physiology and fitness of the bacteria (Rang et al., 2003; Allison and Sattenstall, 2007). It was reported that the growth of Salmonella was suppressed due to constitutive expression of gfp (Oscar, 2003). In contrast, two case studies using E. coli and other pathogenic bacteria showed that the gfp expression did not affect bacterial survival (Leff and Leff, 1996; Valdivia et al., 1996). The use of erythromycin as an antibiotic selective marker has a number of drawbacks. It may cause toxic effects on the host insect and other gut microbes. The excessive use of antibiotics causes its spread in the environment and thus produces many antibiotic resistant pathogenic bacteria (Hamer and Gill, 2002; Livermore, 2007; Walsh and Fanning, 2008). Under laboratory conditions, it has been shown that the antibiotic resistance genes via a plasmid can be transferred into foodborne pathogenic bacteria by turning antibiotic sensitive strains into resistant ones (Van Meervenne et al., 2012).

Researchers have found that indigenous bacteria derived from the host insect could be reintroduced and could survive in the native gut environment (Chapco and Kelln, 1994; Dillon and Charnley, 1996; Martinez-Sanudo et al., 2011). In previous experiments, we introduced GFP-tagged E. coli into the gut of S. littoralis and were able to monitor the bacteria for up to 4 days after which they disappeared (Wallstein, 2014). Our observation was independently confirmed by others (Thimm et al., 1998), who found that genetically modified non-indigenous E. coli vanished within 1 day after introduction into the gut of collembola. In contrast, the indigenous Alcaligenes faecalis was able to colonize the intestinal tract of Folsomia candida (Collembola) for about 2 months (Thimm et al., 1998). In other studies, Husseneder and Grace failed to produce a persistent population of transgenic E. coli in the guts of termites (Husseneder and Grace, 2005). However, they successfully introduced the genetically modified indigenous Enterobacter cloacae in termite guts, where the bacteria persisted for almost 3 months. The failure of nonindigenous E. coli to establish a stable population in the guts of termites may be due to resistance by the indigenous gut bacteria (Dillon and Dillon, 2004), and the fact that the non-indigenous bacteria might be outclassed by the natural microbial flora (Chao and Feng, 1990; Leff and Leff, 1996).

Since it is of interest to study the mode of transmission of GFP-labeled E. mundtii to the next generation, we also analyzed the occurrence of GFP-labeled cells in pupae, and oocytes of S. littoralis. The fluorescent bacteria were found, indeed, in the oocytes and were transmitted to the second-generation larvae. Recent examinations using fluorescent bacteria have found the bacteria to be transmitted from the gut into the eggs in T. castaneum (Knorr et al., 2015). In one hypothesis, the egg-smearing mode of vertical transmission, the surface of the eggs is contaminated with the environmental bacterial symbionts, which the freshly hatched larvae acquire by feeding on the eggshell (Douglas and Beard, 1997; de Vries et al., 2001). Bakula showed that the methylene blue dye used to stain the embryos of Drosophila was detected in the intestine of hatched first-instar larvae, suggesting that the larvae had ingested the embryos (Bakula, 1969). Transmission through parents also occurs, either from the mother to the offspring or from the father to mother and then to the offspring (Moran and Dunbar, 2006; Damiani et al., 2008). Damiani et al. also demonstrated that male-borne symbionts of the bacteria of the genus Asaia were transferred to females during mating of Anopheles stephensi mosquitoes (Damiani et al., 2008). These bacteria were then further transmitted from the mother to the offspring during sexual reproduction. In separate studies, Moran and Dunbar also showed that it is possible, though rare, for symbionts to be transferred from the father to the offspring (paternal transfer) in aphids (Moran and Dunbar, 2006).

Several factors—for instance, pH and oxygen availability can shape microbial colonization in different gut niches. It is known that the pH inside the gut of Lepidoptera such as S. littoralis is highly alkaline (pH ∼ 8.5–10) in the foregut and midgut, and neutral (pH 7.0) in the hindgut (Funke et al., 2008). It has been reported that there is relatively low diversity of bacteria in the extremely alkaline guts of gypsy moth larvae, Lymantria dispar (Broderick et al., 2004) and high bacterial density in the low-alkaline guts of larvae in beetles (Coleoptera), flies (Diptera), or bees (Apoidea) (Kadavy et al., 1999; Egert et al., 2003; Mohr and Tebbe, 2006). The survival of E. mundtii in alkaline environment shows that it has developed adaptation mechanisms.

It has been shown through FISH analyses that enterococci can form a biofilm-like structure by attaching themselves to the mucus layer of the gut epithelium (Koch and Schmid-Hempel, 2011; Engel et al., 2012; Shao et al., 2014). In our study, interestingly, most of the GFP-tagged bacteria did not spread throughout the whole gut content but were confined within the mucus layer of the peritrophic membrane. This membrane prevents the bacteria from gut lumen from entering the epithelium, as reported in the study of Bactrocera oleae (Mazzon et al., 2008). The peritrophic membrane was shown to have a defensive role against pathogens in Drosophila melanogaster (Kuraishi et al., 2011) and to act as a barrier against food particles and digestive enzymes (Lehane, 1997; Hegedus et al., 2009). In addition, the membrane was able to protect the bacteria from unfavorable gut conditions such as alkaline and acidic pH (Crotti et al., 2009). In our case, we observed that the fluorescent bacteria crossed the peritrophic membrane and invaded the gut epithelium of the fourth-instar larvae.

The composition and density of microorganisms changed as insects aged, for example in the case of the fruit fly, D. melanogaster (Ren et al., 2007; Buchon et al., 2009; Storelli et al., 2011; Wong et al., 2011). In our study, the number of fluorescent bacteria increased throughout the larval stage, from fourth- to fifth-instar larvae. This number was significantly higher in tissues from the midgut than from those in the foregut and hindgut, supporting the hypothesis that the midgut is a crucial region for digestion. Beneficial bacteria may be needed to aid in the metabolic activity of host insect. A strong decline of recombinant bacteria was observed in the sixth-instar larvae. This reduction prior to the pupal stage may be associated with the enhanced expression of antimicrobial peptide genes which has been shown in a few previous studies (Samakovlis et al., 1990; Tryselius et al., 1992; Tzou et al., 2000).

Humoral responses, such as the production of antimicrobial peptides, reactive oxygen species, and lysozymes, as well as activation of the prophenoloxidase system, are noticed when microorganisms invade (Jiang, 2008; Tsakas, 2010). Antimicrobial peptides can be repressed by transcription factors, including the homeobox gene caudal, in order to retain beneficial gut bacteria in the host insect (Ryu et al., 2008). Remarkably, we detected the encapsulation of fluorescent E. mundtii within

#### REFERENCES


the nodules of granule hemocytes which leads to bacterial lysis. The two most abundant hemocytes present in the larvae of Lepidoptera are granular cells (granulocytes) and plasmatocytes (Ratcliffe, 1993; Strand and Pech, 1995). The processes of phagocytosis, nodulation and encapsulation are hemocyte-mediated immune responses (Strand, 2008; Tsakas, 2010). Hemocytes encapsulate various cells ranging from bacteria to yeast and even synthetic beads and particles of India ink (Yokoo et al., 1995; Hernandez et al., 1999; Da Silva et al., 2000). We observed that the fluorescent enterococci survived pupation and became transmitted to the adults. Accordingly, the bacteria can be successfully transmitted during metamorphosis escaping the removal or reduction of midgut bacteria by colonizing sites farther away from the meconium. Another explanation may be that bacteria are resistant to the antimicrobial exuvial fluids that are consumed as part of the ecdysial process (Moll et al., 2001).

### CONCLUSION

We have succeeded in tagging E. mundtii (strain KD251) with the gfp gene. The recombinant strain that harbors the pTRKH3 ermGFP plasmid was chosen to be reintroduced into S. littoralis. Interestingly, the fluorescent bacterial cells were able to colonize the intestinal tract of the host insect for nearly 30 days. These bacteria were efficiently transmitted from larval stages to the adult stage, where they survived up to the second generation. Increased knowledge of the distribution and transmission route of indigenous gut symbionts may lead us to better understand their biological role in the host insect.

### AUTHOR CONTRIBUTIONS

Conceived and designed the experiments: BT, YS, WB. Performed the experiments: BT, JA. Analyzed the data: BT. Wrote the paper: BT and WB.

### ACKNOWLEDGMENTS

We thank Angelika Berg for laboratory assistance, Ewald Grosse-Wilde for the help in cryomicrotomy, Domenica Schnabelrauch for sequencing, and Grit Kunert for help with the statistics. We gratefully acknowledge editoral assistance by Emily Wheeler. This work was supported by the DFG (ChemBioSys, CRC 1127), the excellence graduate school Jena School for Microbial Communication (JSMC), and the Max Planck Society.

Allison, D. G., and Sattenstall, M. A. (2007). The influence of green fluorescent protein incorporation on bacterial physiology: a note of caution. J. Appl. Microbiol. 103, 318–324. doi: 10.1111/j.1365-2672.2006.03243.x

Anand, A. A., Vennison, S. J., Sankar, S. G., Prabhu, D. I., Vasan, P. T., Raghuraman, T., et al. (2010). Isolation and characterization of bacteria from the gut of Bombyx mori that degrade cellulose, xylan, pectin and starch and their impact on digestion. J. Insect Sci. 10, 107. doi: 10.1673/031.010.10701

Appel, H. M., and Martin, M. M. (1990). Gut redox conditions in herbivorous lepidopteran larvae. J. Chem. Ecol. 16, 3277–3290. doi: 10.1007/BF00982098


Pachnoda ephippiata (Coleoptera: Scarabaeidae). Appl. Environ. Microbiol. 69, 6659–6668. doi: 10.1128/AEM.69.11.6659-6668.2003


lithium acetate and dithiothreitol. BMC Biotechnol. 7:15. doi: 10.1186/1472- 6750-7-15


**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 © 2016 Teh, Apel, Shao and Boland. 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) or licensor 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.

# Bacterial Community and PHB-Accumulating Bacteria Associated with the Wall and Specialized Niches of the Hindgut of the Forest Cockchafer (Melolontha hippocastani)

#### Edited by:

Martin Grube, University of Graz, Austria

#### Reviewed by:

Armin Erlacher, Graz University of Technology, Austria Tomislav Cernava, Austrian Centre of Industrial Biotechnology, Austria

#### \*Correspondence:

Pol Alonso-Pernas palonso@ice.mpg.de Wilhelm Boland boland@ice.mpg.de

†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: 01 May 2016 Accepted: 13 February 2017 Published: 28 February 2017

#### Citation:

Alonso-Pernas P, Arias-Cordero E, Novoselov A, Ebert C, Rybak J, Kaltenpoth M, Westermann M, Neugebauer U and Boland W (2017) Bacterial Community and PHB-Accumulating Bacteria Associated with the Wall and Specialized Niches of the Hindgut of the Forest Cockchafer (Melolontha hippocastani). Front. Microbiol. 8:291. doi: 10.3389/fmicb.2017.00291 Pol Alonso-Pernas<sup>1</sup> \* † , Erika Arias-Cordero<sup>1</sup>† , Alexey Novoselov<sup>1</sup> , Christina Ebert2,3 , Jürgen Rybak<sup>4</sup> , Martin Kaltenpoth<sup>5</sup> , Martin Westermann<sup>6</sup> , Ute Neugebauer2,3 and Wilhelm Boland<sup>1</sup> \*

<sup>1</sup> Department of Bioorganic Chemistry, Max Planck Institute for Chemical Ecology, Jena, Germany, <sup>2</sup> Center for Sepsis Control and Care, Jena University Hospital, Jena, Germany, <sup>3</sup> Leibniz Institute of Photonic Technology, Jena, Germany, <sup>4</sup> Department of Evolutionary Neuroethology, Max Planck Institute for Chemical Ecology, Jena, Germany, <sup>5</sup> Department of Evolutionary Ecology, Institute of Zoology, Johannes Gutenberg University Mainz, Mainz, Germany, <sup>6</sup> Electron Microscopy Center, Jena University Hospital, Jena, Germany

A characterization of the bacterial community of the hindgut wall of two larval and the adult stages of the forest cockchafer (Melolontha hippocastani) was carried out using amplicon sequencing of the 16S rRNA gene fragment. We found that, in second-instar larvae, Caulobacteraceae and Pseudomonadaceae showed the highest relative abundances, while in third-instar larvae, the dominant families were Porphyromonadaceae and Bacteroidales-related. In adults, an increase of the relative abundance of Bacteroidetes, Proteobacteria (γ- and δ- classes) and the family Enterococcaceae (Firmicutes) was observed. This suggests that the composition of the hindgut wall community may depend on the insect's life stage. Additionally, specialized bacterial niches hitherto very poorly described in the literature were spotted at both sides of the distal part of the hindgut chamber. We named these structures "pockets." Amplicon sequencing of the 16S rRNA gene fragment revealed that the pockets contained a different bacterial community than the surrounding hindgut wall, dominated by Alcaligenaceae and Micrococcaceae-related families. Poly-β-hydroxybutyrate (PHB) accumulation in the pocket was suggested in isolated Achromobacter sp. by Nile Blue staining, and confirmed by gas chromatography–mass spectrometry analysis (GC-MS) on cultured bacterial mass and whole pocket tissue. Raman micro-spectroscopy allowed to visualize the spatial distribution of PHB accumulating bacteria within the pocket tissue. The presence of this polymer might play a role in the colonization of these specialized niches.

Keywords: hindgut, Melolontha hippocastani, gut bacteria, poly-β-hydroxybutyrate, PHB, Achromobacter, Raman microscopy

## INTRODUCTION

fmicb-08-00291 February 25, 2017 Time: 15:46 # 2

Bacteria not only thrive as free-living organisms in the environment, they also engage in complex symbiotic relationships with higher organisms (Wells and Varel, 2011). Insects, in particular, are associated with a large diversity of microorganisms that play important roles for their host's physiology, ecology, and evolution. The insect gut is colonized by a wide range of bacterial phylotypes that interact with the host and allow it to subsist on nutritionally imbalanced diets. The recycling of nitrogen, the provisioning of essential amino acids and cofactors, and the digesting of recalcitrant polymers in the host's diet are among the functions for which symbiotic microorganisms play an integral role (Potrikus and Breznak, 1981; Douglas, 2009; Watanabe and Tokuda, 2010), increasing the overall fitness of the insect host.

Typically, the insect gut is divided into three regions, i.e., foregut, midgut and hindgut. The symbiotic bacteria are either attached to the gut wall or colonize the gut as freeliving organisms, usually mostly in the mid- and hindgut regions. The structure of these communities differs among insect species, influenced by the host's diet and taxon (Egert et al., 2003, 2005; Colman et al., 2012; Jones et al., 2013). In the Scarabaeidae family, the hindgut region is of special importance. It is anatomically modified to serve as fermentation chamber. This chamber, in addition to its original function, namely, absorbing water and salts from the gut content, is also devoted to aiding digestion, probably with the help of the fermentative bacteria that colonize it (Egert et al., 2005; Huang et al., 2010; Arias-Cordero et al., 2012; Engel and Moran, 2013). These microbial associates are transmitted either vertically, directly from mother to offspring, or horizontally, that is, being taken anew from the environment by each host generation (Bright and Bulgheresi, 2010). In horizontally transmitted symbiosis, the host usually ingests the symbiont along with unwanted microbes that may compete for the colonization of the gut. The selection of the right symbiont may depend on its phenotypic traits. Kim et al. (2013) showed that poly-β-hydroxybutyrate (PHB) accumulation by the symbiont is crucial for the maintenance of host–microbe relationship.

In this study, we investigate the forest cockchafer (Melolontha hippocastani). This scarabaeid constitutes an interesting model due to its particular life cycle, consisting in two well-differentiated stages: the rhizophagous larvae spend up to 4 years underground, while the adults, after pupation, emerge from the soil and shift to a diet based exclusively on foliage. To date, there is a lack of comparative studies on the variation of the gut bacterial community associated with the transition from larva to adult. Only one study addressed this question, a study conducted by Arias-Cordero et al. (2012), focused on the midgut of M. hippocastani. Surprisingly, they found a group of bacterial phylotypes that seems to always be stable. This core community is maintained through metamorphosis and is unaffected by the radical change of the host diet from roots to leaves, when the shift occurs from a below-ground (larval) to an above-ground (adult) stage (Arias-Cordero et al., 2012).

In view of this unexpected stability of the gut microbial community, we considered appropiate to characterize the bacterial communities inhabiting the hindgut wall of both below- and above-ground stages of the forest cockchafer, thus complementing the above-mentioned midgut-based study (Arias-Cordero et al., 2012). We put our focus on the hindgut wall itself, and also on particular bacterial niches attached and connected to it, at both sides of the distal part of the larval hindgut. These small structures, called from now on "pockets," have been hitherto only once described in the literature (Wildbolz, 1954). They consist of several tubular poles connected to the hindgut chamber, which contain bacterial phylotypes that are minor or not detected in the hindgut wall. We detected the presence of PHB within the pockets, and Achromobacter sp., one of the major pocket bacterial species, is able to accumulate PHB in pure culture. This suggests that some of the pocket symbionts may be horizontally transmitted, as previous studies found this type of inclusions in symbiotic Burkholderia of environmental origin harbored in the midgut crypts of the midgut of Riptortus pedestris (Kim et al., 2013). The question of whether PHB plays a role in host nutrition remains unknown.

### MATERIALS AND METHODS

#### Sample Collection and DNA Extraction

Second-instar (L2) and third-instar (L3) larvae of M. hippocastani and actively flying adults were collected in forests of red oak in Mannheim (49◦ 290 2000N 8◦ 280 9 <sup>00</sup>E), and Graben-Neudorf (49◦ 9 0 5500N 8◦ 290 2100E), respectively, between December 2010 and May 2014. Beetles were collected at the same sites. The insects were transported alive in boxes with soil or tree leaves. Before dissection, the insects were kept at −20◦C for 20 min to kill them, and then rinsed three times alternately with sterile distilled water and 70% ethanol. Dissection was performed on ice in a phosphate-buffered saline (PBS) solution. Hindguts, as shown between dotted lines in **Figure 1D** (top for larva and bottom for adult), were excised, cut open, and carefully washed three times with sterile PBS in order to remove any unattached bacteria. The pockets were separated from the hindgut wall, and as much of the surrounding epithelium was removed as possible. Samples were stored at −20◦C before DNA extraction. The day of the extraction, frozen samples were thawed on ice and dried at 45◦C for 90 min in a Speedvac (Concentrator 5301, Eppendorf), then crushed in a 1.5 ml tube with a sterile pestle. For 454 pyrosequencing, DNA extractions of the tissue were carried out using the PowerSoilTM DNA Isolation Kit (MO BIO Laboratories Inc., Carlsbad, CA, USA) according to the protocol provided by the manufacturer. Final DNA concentrations were determined using a Nanovue device (GE Healthcare, Little Chalfont, UK). In order to test for the quality of the extracted DNA and confirm the presence of DNA from bacteria, a diagnostic PCR reaction was carried out as described (Arias-Cordero et al., 2012).

### Transmission Electron Microscopy (TEM)

Dissected hindguts and pockets of larvae were fixed in a solution of 2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 M

sodium cacodylate buffer (pH 7.2). Immediately afterward, the tissue was transferred to the same solution for overnight fixation. Next day, the fixative was removed, and the tissue was post fixed with 1% osmium tetroxide in cacodylate buffer for 2 h. During the following ascending ethanol series samples were stained with 2% uranyl acetate. The samples were embedded in Araldite CY212 epoxy resin (Agar Scientific Ltd, Stansted, United Kingdom) according to manufacturer's instruction. Semithin sections (1 µm thickness) were stained with Richardson's methylene blue in order to localize the right position for the examination. Hindgut areas were further trimmed down to 500 µ × 500 µm. Ultra-thin sections of 80 nm thickness were cut using an ultramicrotome Ultracut E (Reichert–Jung, Vienna, Austria) and mounted on Formvar-carbon coated grids (100 meshes, Quantifoil GmbH, Großlöbichau, Germany). Finally, sections were contrasted with lead citrate for 4 min and analyzed in a transmission electron microscope EM900 (Zeiss AG, Oberkochen, Germany).

#### Light Microscopy, Richardson Staining, and Autofluorescence Visualization

In all cases the tissue was fixed as described above for transmission electron microscopy (TEM). The tissues employed were larvae hindgut walls and pockets. For the Richardson staining, semi-thin sections of 0.3–0.6 µm (embedded as for TEM) were immersed in a 60◦C staining solution for 3–5 min. Afterward, the tissue was washed twice with sterile water. Finally, the sections were placed on a glass slide, dried and mounted for microscopic observation. For the autofluorescence visualization, a excised complete hindgut pocket was placed onto a glass slide and covered with PBS. Visualization was carried out using a LeicaTCS-SP2 confocal microscope using a 10× dry or 40× oil Leica objective (HC PL APO 10×/0.4, Leica, Bensheim, Germany) in both cases. For autofluorescence, laser line employed was 488 nm.

### Bacterial Tag-Encoded FLX Amplicon Pyrosequencing (bTEFAP) and Data Analysis

For pyrosequencing, a sample was composed of the extracted DNA of six insects collected during the same year, pooled together in equal amounts for a single run. A total of four samples were sequenced (L2 pocket, L2 hindgut wall, L3 hindgut wall, and adult hindgut wall). DNA was sent to an external service provider (Research and Testing Laboratories, Lubbock, TX, USA) for bTEFAP with 16S rRNA primers Gray28F (5<sup>0</sup> -GAGTTTGATCNTGGCTCA-3<sup>0</sup> ) and Gray519R (50 -GTNTTACNGCGGCKGCTG -3<sup>0</sup> ) (Ishak et al., 2011). A sequencing library was generated through one-step PCR with 30 cycles, using a mixture of HotStar and HotStar HiFidelity Taq polymerases (Qiagen, Hilden, Germany). Sequencing extended from Gray28F, using a Roche 454 FLX instrument with Titanium reagents and procedures at Research and Testing Laboratory (RTL, Lubbock, TX, USA<sup>1</sup> ). Quality control and analysis of 454 reads, including calculation of rarefaction curves and community richness and diversity indexes, was done in QIIME version 1.8.0 (Caporaso et al., 2011). Low-quality ends of the sequences were trimmed with a sliding window size of 50 and an average quality cut-off of 25. Subsequently, all lowquality reads (quality cut-off = 25) and sequences <200 bp were removed, and the remaining reads were denoised using the "denoiser" algorithm as implemented in QIIME (Reeder and Knight, 2010). Denoised high-quality reads were clustered into operational taxonomic units (OTUs) using a multiple OTU picking strategy with cdhit (Li and Godzik, 2006) and uclust (Edgar, 2010), with 97% similarity cut-offs, respectively. For each OTU, the most abundant sequence was chosen as a representative sequence and aligned to the Greengenes core set<sup>2</sup> using PyNast (Caporaso et al., 2010). RDP classifier was used for taxonomy assignment (Wang et al., 2007). An OTU table was generated describing the occurrence of bacterial phylotypes within the samples.

#### qPCR Analysis of Pocket and Hindgut Wall Tissue

For the quantitative real-time PCR (qPCR) analysis, thirdinstar larvae were used. A sample was composed of the pooled DNA from hindgut wall, or pockets, of three different larval individuals. Three samples from each tissue (hindgut wall and pockets) were considered, and each one was analyzed per triplicate. Specific primers were designed using Geneious 6.0.5<sup>3</sup> for the five most consistently found bacterial taxa in the pocket (Achromobacter, Citrobacter, Bosea, Brevundimonas, and Pseudomonas), based on the alignment of the representative set of sequence data for all OTUs available from the 454 pyrosequencing. PCR conditions for each primer pair were optimized using gradient PCRs (Salem et al., 2013). Their specificity was verified in silico against the SILVA ribosomal RNA database<sup>4</sup> and in vitro by sequencing. Briefly, PCR products from pocket DNA were analyzed on 1% agarose gels (150 V, 30 min). The products were purified from the gel with Invisorb Fragment CleanUp kit (Stratec Molecular, Berlin, Germany) and cloned in pCR 2.1 vector using the Original TA Cloning kit (Invitrogen, Carlsbad, CA, USA). Ninety clones with positive inserts were selected according to the manufacturer's protocol and sequenced on a 3730 XL DNA Analyzer (Applied Biosystems, Foster City, CA, USA) with BD 3.1 chemistry. If the sequence matched the expected OTU, the primer pair was assumed to specifically amplify the target OTU within the gut and pocket. The sequences of the primers are listed in Supplementary Table S2. Quantitative PCRs for individual bacterial taxa were performed on a CFX96 Real Time System (Bio-Rad, Munich, Germany), in final reaction volumes of 10 µL containing 1 µL of template DNA (usually a 1:10 dilution of the original DNA extract), 0.6 µL of each primer (10 pM) and 5 µL of SYBR Green Mix (Rotor-Gene SYBR Green kit, Qiagen, Hilden, Germany). Standard curves were established using 10−6–10−<sup>2</sup> ng of specific

<sup>1</sup>www.researchandtesting.com/

<sup>2</sup>http://greengenes.lbl.gov/

<sup>3</sup>http://www.geneious.com

<sup>4</sup>http://www.arb-silva.de

PCR product as templates for the qPCR. A NanoDrop ND-1000 spectrophotometer (Peqlab Biotechnology Limited, Darmstadt, Germany) was used to measure template DNA concentration for the standard curve. Five different replicates of the standard concentrations for each bacterial taxon were used to calculate a correction factor and determine equitation parameters. PCR conditions were as follows: 95◦C for 3 min, followed by 40 cycles of denaturation at 95◦C for 10 s, annealing for 30 s and elongation at 72◦C for 10 s. Then, a melting curve analysis was performed to ensure that amplicons were the same across samples for each primer assay, by increasing the temperature from 65 to 95◦C within 5 min. The annealing temperature was specific for each primer pair: for Achromobacter and Citrobacter, 60◦C; for Bosea, 63◦C; for Brevundimonas, 55◦C; for Pseudomonas, 68◦C. Based on the standard curves, the 16S copy number could be calculated for each individual sample from the qPCR threshold values (Ct) by the absolute quantification (Lee et al., 2006, 2008), taking the dilution factor and the absolute volume of DNA extract into account. The quantitative differences in the microbial community abundances of the pocket were tested using SPSS 17.0 (Tukey HSD test, confidence interval of 0.05).

### Isolation and Identification of Pocket Bacteria

Four second-instar pockets from different larvae were dissected as mentioned above and incubated together in a 0.8% NaOCl aqueous solution for 3 min on ice for surface sterilization. Then, the tissue was transferred in Ringer+ppi buffer (Cazemier et al., 1997) and sonicated using a Sonorex Super RK 102h sonicator (Bandelin, Germany) for 7 min at RT. After sonication, the tubes were incubated 15 min on ice and gently tapped from time to time. Ten-fold dilutions of the supernatant were plated on LB agar (Carl Roth, Germany) and ATCC agar in order to enrich for Achromobacter sp. The ATCC agar contained (per liter): 7.32 g K2HPO4, 4.6 g ammonium tartrate, 1.09 g KH2PO4, 0.04 g MgSO<sup>4</sup> 7H2O, 0.04 g FeSO<sup>4</sup> 7H2O, 0.014 g CaCl<sup>2</sup> 2H2O, and 35 g agar. Plates were incubated at 30◦C for 48 h. Morphologically different colonies were subcultured three times before identification. Colony PCR targeting the small ribosomal subunit gene was performed on a GeneAmp 9700 Thermocycler (Applied BioSystems) using the general bacterial primers 27f and 1492r (Arias-Cordero et al., 2012). The 50 µL reaction mixture contained 1x buffer, 1.5 mM MgCl2, 10 mM of the four deoxynucleotide triphosphates (dNTPs), 2.5 U Taq DNA polymerase (Invitrogen) and 0.5 mM of each primer. The PCR program was as follows: initial denaturation at 94◦C for 3 min followed by 32 cycles of denaturation at 94◦C for 45 s, annealing at 55◦C for 30 s and elongation at 72◦C for 1 min, and a final elongation step at 72◦C for 10 min. Amplicon size was confirmed in a 1% agarose gel; then the PCR product was purified using the Invisorb Fragment CleanUp kit (STRATEC Molecular GmbH, Berlin, Germany). Sequencing was performed at Macrogen Europe (Amsterdam, The Netherlands), and the taxonomy of resulting sequences was assigned using Basic Local Alignment Search Tool (BLAST) (Tatusova and Madden, 1999).

### Metabolic Testing of Bacterial Isolates

Nile Blue agar was prepared as described (Luellen and Schroth, 1994). A representative of each bacterial isolate was plated and incubated for 48 to 72 h at 30◦C. The plates were then viewed under UV light to detect putative PHB production based on the fluorescence of the colonies. Nitrate reduction test was purchased from Sigma and conducted following the instructions provided by the manufacturer. A representative of each bacterial isolate was inoculated at high density, and tubes were sealed with liquid paraffin to create oxygen-poor conditions and incubated at 30◦C up to 5 days.

### Gas Chromatography – Mass Spectrometry

Twenty-five third-instar larvae were dissected as described, and their 50 pockets were analyzed as one single sample. Achromobacter sp. isolated from the pocket was cultured for 3 days in PHB inducing broth at 30◦C for 72 h. The composition of PHB inducing broth is the same as Nile Blue agar (Luellen and Schroth, 1994) without Nile Blue or agar. The bacterial mass was recovered by centrifugation and washed twice with sterile distilled water prior to drying [45◦C for 90 min in a Speedvac (Concentrator 5301, Eppendorf)]. 5 mg (dry weight) of bacterial mass was used for the analysis. Poly[(R)-3-hydroxybutyric acid] standard was obtained from Sigma (Germany), and 1 mg was used for the analysis. Derivatization was performed as described (Riis and Mai, 1988), using methanol instead of propanol for the esterification. GC analysis was performed in a ThermoQuest, Finnigan Trace GC-MS 2000 series (Egelsbach, Germany), equipped with a fused-silica capillary Phenomenex ZB-5 column (15 m × 0.25 mm, film thickness 0.25 µm) with a split ratio of 10:1. Helium was used as carrier gas at a flow rate of 1.5 ml/min. The oven temperature was programmed as follows: the initial temperature of 60◦C was held for 3 min, then increased to 230◦C at 30◦C/min and held for 2 min. The inlet temperature was 250◦C and the injection volume 1 µL. Mass spectra were measured in electron impact (EI) at 70 eV under full scan mode (m/z 35–575). Acquired data were further processed using the software Xcalibur (Thermo Scientific). 3-hydroxybutyric acid methyl esters were identified by comparison of the mass spectrum and retention time with poly[(R)-3-hydroxybutyric acid] standard.

### Raman Micro-Spectroscopy

One pocket was used for each Raman measurement. CaF<sup>2</sup> slides suitable for Raman spectroscopy were poly-L–lysine coated by being soaked overnight in 0.1% poly-L–lysine solution (Sigma) at 4◦C prior to measurement. The pocket tissue was fixed overnight with 4% paraformaldehyde solution in 0.9% NaCl at 4 ◦C. After fixation, the paraformaldehyde was removed and the tissue was washed three times for 10 min with 0.9% NaCl solution under mild agitation. Then the pocket tissue was embedded in a mounting medium for cryotomy, OCT compound (VWR Chemicals, Radnor, PA, USA) and sliced in 12-µm thick sections using a Microm HM 560 cryomicrotome (Thermo Scientific, Waltham, MA, USA). The tissue slices were put onto the

poly-L–lysine coated CaF<sup>2</sup> slide, washed carefully with 0.9% NaCl to remove the remains of the mounting medium and viewed under a bright-field microscope to check for the characteristic round-shaped cross-sections of the pocket poles. The Raman spectra were acquired with a confocal Raman microscope alpha 300R (WITec, Ulm, Germany) using a 532 nm Nd:YAG solid laser with a power of 15 mW for excitation. The samples were measured in 0.9% NaCl using a 60× water immersion objective with NA 1.0 (Nikon, Tokyo, Japan). Collection of backscattered photons occurred through a back-illuminated CCD camera (DV401-BV-352, Andor, Belfast, UK). For spectral grating, 600 lines/mm were used for 532 nm. A multimode fiber of 25 µm diameter served as pinhole for confocal imaging. The Raman spectra were recorded by using 1 s integration time. Characteristic spectra and compartments in the pocket poles were detected by analyzing the Raman scans with the N-FINDR unmixing algorithm (Winter, 1999; Hedegaard et al., 2011) using Matlab software (MathWorks). The PHB was detected by identifying specific peaks through comparison with measured reference spectrum of pure PHB compound.

## Nucleotide Sequence Accession Numbers

The 16S RNA gene sequences obtained by colony PCR have been deposited at the NCBI GenBank under accession numbers from KY178280 to KY178284 (**Table 1**). Pyrosequencing data from L2 hindgut wall, adult hindgut wall, L2 pocket and L3 hindgut wall have been deposited under accession numbers SRR5059348, SRR5059349, SRR5059340, and SRR5059351, respectively.

### RESULTS

### Localization and Morphology of the Pockets

During the dissection of larval individuals (**Figure 1A**), two small structures ["pockets," colored either white or black (**Figures 1F,G**)] attached outside the terminal point of the hindgut chamber (**Figures 1B,C,D,E, 2A**) were spotted. The pockets have a diameter of around 500 µm, and showed high autofluorescence when illuminated with a 488 nm laser (**Figure 2B**). They are covered by a fine layer of muscle tissue (Supplementary Figure S1). Their anatomy is composed by poles connected to the hindgut lumen (Supplementary Figure S2). Further anatomical investigation by TEM revealed that each pole

FIGURE 1 | Gut anatomy of larvae and adults of Melolontha hippocastani. (A) L3 larval instar living in the soil. (B) Hindgut fermentation chamber. White arrowheads point to the position of the pockets. (C) Close-up of a hindgut lobe. (D) Whole gut preparation of an L3 larval instar (top image) and an adult beetle (bottom image). The hindgut section used for microscopy and pyrosequencing is between the dashed lines. (E) The fermentation chamber and the pocket position (pointed with arrows). (F) Close-up of the M. melolontha pocket and (G) close-up of the M. hippocastani pocket. Scale bars: green 5 mm., white 100 µm.

was surrounded by a thick acellular tissue layer (possibly mucouslike, **Figure 2C**). Additionally, it was observed that each pole was lined with large numbers of bacterial cells (**Figure 2C**). These cells showed a high number of cytoplasmatic inclusions (**Figure 2D**).

#### TABLE 1 | Bacterial isolates from Melolontha hippocastani's pockets with their metabolic capabilities.


NT, not tested.

### Pyrosequencing of the Bacterial Community from the Hindgut Wall of Adult Insects and Larvae, and Pockets

To establish the dynamics of the hindgut wall community across different host's life stages, the bacterial communities of the hindgut wall of L2 and L3 larvae and adults were compared. DNA from six different insects of each life stage was used, pooled together in a single pyrosequencing run. In the final output, 110,772 high quality reads were obtained (Supplementary Table S1). It was found that, in the L2 hindgut wall, the main bacterial phyla were Pseudomonadaceae, Caulobacteraceae and Micrococcaceae, while in L3 hindgut wall, those were Bacteroidetes phylum and Clostridia, with a large proportion of unknown bacteria. In the adults, an increase of the relative abundance of the Bacteroidales order, Proteobacteria (γ- and δ- classes) and the family Enterococcaceae (Firmicutes) was observed (**Figure 3**). Estimation of alpha-diversity in these samples was done using rarefaction methods, and richness and diversity indexes were also calculated (Supplementary Figure S3 and Table S1).

Amplicon sequencing revealed considerable differences in microbial communities between the L3 and L2 hindgut walls. In L2, approximately 47% of the sequences obtained belong to the family Pseudomonadaceae and 30% to the family Caulobacteraceae, taxa that were not detected in the L3 hindgut wall; the L3 hindgut wall, in turn, had families at high abundances which were not or only at low abundances detected in L2 (e.g., Porphyromonadaceae, Bacteroidales, and Ruminococcaceae) (**Figure 3**). This may reflect the changes that the bacterial community undergoes throughout the different stages of the insect's life, suggesting that the hindgut wall is a dynamic environment.

A pooled sample of DNA extracted from 12 excised pockets (from 6 L2 larvae) was also sequenced, in order to compare their bacterial communities with the surrounding hindgut wall. It was found that the main bacterial phyla of the pocket tissue were Actinobacteria and Proteobacteria (α- and β- classes). Within the β-Proteobacteria, Achromobacter sp., which accounted for 85% of sequences from the family Alcaligenaceae, was the genus with the overall highest relative abundance in the pockets. The classification at genus level of the family Micrococcaceae was not achieved. These two families were present in low abundance in the L2 and L3 hindgut wall, as well as in the hindgut wall of adult beetles (**Figure 3**).

FIGURE 3 | Bacterial community composition in different life stages of M. hippocastani. Relative abundance of bacterial taxa, in percentage of total sample sequences, from 454 pyrosequencing data (110,772 sequences in total) is displayed as a heat map based on the log-transformed values. Warm colors indicate higher and cold colors lower abundances. Families with total relative abundance lower than 0.4% were considered "low abundance families" and listed in Supplementary Table S3. unk., unknown; fam., family.

### Estimation of Absolute Abundances of Main Bacterial Genera in the Pockets and the Hindgut Wall

In order to compare the absolute abundances of key genera inhabiting the pocket and the hindgut wall of L3 larvae, namely Achromobacter (family Alcaligenaceae), Bosea (family Bradyrhizobiaceae), Brevundimonas (family Caulobacteraceae), Citrobacter (family Enterobacteriaceae) and Pseudomonas (family Pseudomonadaceae), qPCR with genus-specific primers was performed. In the pocket, Achromobacter was the most dominant of the genera, with an abundance about 10 times greater than that of Pseudomonas (**Figure 4**). Citrobacter, Brevundimonas, and Bosea showed lower abundances, with that of Bosea being three orders of magnitude lower than that of Achromobacter. The abundances of all four lowerabundant genera in the pockets differed significantly from that of Achromobacter (ANOVA, Tukey HSD test, p < 0.05). This is in line with the outcome of the 454-pyrosequencing, in which Achromobacter sp. (85% of family Alcaligenaceae sequences) was the most dominant of the identified genera in the pocket (**Figure 3**). However, since it was not possible to classify the family Micrococcaceae at the genus level, it must be taken into account that Achromobacter sp. may be overcome by a Micrococcaceae-related genus.

In the hindgut wall, the abundances of Pseudomonas, Brevundimonas, and Bosea spp. (Pseudomonas > Brevundimonas > Bosea) were in good agreement with their respective family abundances showed by the 454 pyrosequencing approach. The occurrences of Citrobacter and Achromobacter spp., respectively, the first and second most ubiquitous genera according to the qPCR outcome, matched their respective abundances in the L3 hindgut wall pyrosequencing (families Enterobacteriaceae and Alcaligenaceae, respectively), but were significantly higher than their abundances in L2 hindgut wall pyrosequencing (**Figure 3**). This outcome fits with the abovementioned idea that the relative abundances of the gut bacterial community members are dynamic depending on the larval instar.

### PHB Detection in Pocket Isolates and Pocket Tissue by Nile Blue Staining and GC-MS

Considering the relatively close phylogenetic relationship between the major genus in M. hippocastani pockets, Achromobacter sp., and the PHB-accumulating bacterium that colonizes the R. pedestris midguts crypts, Burkholderia sp. (Kim et al., 2013), we speculated that PHB accumulation could also take place in the pocket symbionts. To test this hypothesis, pocket symbionts were isolated in selective media. The bacterial species that were retrieved are listed in **Table 1**. PHB accumulation was suggested in Achromobacter marplatensis, Stenotrophomonas maltophilia, and Phyllobacterium myrsinacearum by its positive fluorescence under UV light when cultured in Nile Blue agar (**Table 1**) (Ostle and Holt, 1982).

Gas chromatography coupled with mass spectrometry (GC-MS) of pocket tissue as well as isolated A. marplatensis was conducted in order to confirm PHB presence. For the analysis, pockets and bacterial mass were derivatized through trans-esterification with methanol in the presence of acid (see Materials and Methods) prior to injection into the gas chromatograph. The resulting chromatograms (**Figure 5**) showed a peak corresponding to 3-hydroxybutyric acid methyl ester, the derivatized 3-hydroxybutyric acid monomeric unit of PHB, with a retention time of 2.21 min (±0.01 min). Its identification was carried out by comparing the obtained mass spectrum and the retention time with the commercially available reference compound.

### Raman Micro-Spectroscopy of the Pocket Tissue

In order to determine the spatial distribution of PHBaccumulating bacteria within the pocket pole, Raman microspectroscopy was performed. The Raman spectroscopic scans and spectra obtained are shown in **Figure 6**. Spectral unmixing using the N-FINDR algorithm revealed false-color images that showed different constituents by identifying different Raman spectral signatures. In the pocket poles containing

area marked with a black square is enlarged above. The other peaks correspond to a variety of fatty acids of different chain lengths, common in both eukaryotic and prokaryotic cells, and to artifacts created by the method (peaks in the PHB standard chromatogram).

PHB-accumulating bacteria, they were distributed uniformly throughout the inner area of the pole as dots of approximately 1 µm diameter (spectrum 2 of **Figure 6B**, green area in false-color image). Within a typical bacterial Raman spectrum (Ciobota et al., 2010 ˇ ; Majed and Gu, 2010), the presence of PHB granules was indicated by the bands at 837 and 1058 cm−<sup>1</sup> (C-C stretching), and especially by the highly significant band at 1741 cm−<sup>1</sup> (C=O stretching; compare PHB reference spectrum in **Figure 6C** with spectrum 2 in **Figure 6B**). The spectrum showing mainly C-C stretching (1067, 1131 cm−<sup>1</sup> ), CH<sup>2</sup> twisting (1299 cm−<sup>1</sup> ), and CH<sup>2</sup> bending (1444 cm−<sup>1</sup> ) vibrations (spectrum 1 in **Figure 6B**, blue area in false-color image), were likely derived from fatty acids, probably of a saturated nature as the bands that provide evidence of unsaturation were missing (1260, 1650, and 3023 cm−<sup>1</sup> ), whereas the bands that support saturation were strong (1299, 1444, CH stretch region at 2800 – 3000 cm−<sup>1</sup> ) (Wu et al., 2011). Finally, the spectrum of the mucus-like layer (**Figure 2C**) surrounding the inner part of the pole (spectrum 3 of **Figure 6B**, red area in false-color image) revealed a complex composition, consisting mainly of proteins with disulphide bridges (S-S stretch, band 500 and 505 cm−<sup>1</sup> , respectively), high tyrosine (Tyr) content with bands at 650 cm−<sup>1</sup> (C-C twist Tyr), 854 and 859 cm−<sup>1</sup> (ring vibration Tyr), 1270 cm−<sup>1</sup> (protein amide

III), 1456 cm−<sup>1</sup> (CH<sup>2</sup> deformation), 1622 and 1626 cm−<sup>1</sup> , respectively (C=C stretching Tyr and Trp), 1670 cm−<sup>1</sup> (protein amide I or C=C stretching) (Tuma, 2005), and lipids (band 1270 cm−<sup>1</sup> CH bend), and 1456 cm−<sup>1</sup> (CH<sup>2</sup> deformation). For more detailed band assignment information, see Supplementary Table S4.

#### DISCUSSION

### Bacterial Communities of the Hindgut Wall and the Pockets

Four hundred and fifty-four-pyrosequencing revealed that, in L2 larvae, the bacterial community of the hindgut wall was dominated by the families Pseudomonadaceae and Caulobacteraceae. These families, however, were overgrown in L3 by representatives of the family Porphyromonadaceae and the orders Bacteroidales and Clostridiales. Since these taxa are anaerobic, their proliferation in late larval instars may reflect a thickening of the bacterial layer attached to the hindgut wall, allowing the symbionts to reach more anaerobic areas toward the hindgut lumen, or a pronounced decrease in oxygen concentration due to high bacterial density. Similar shifts in bacterial abundances depending on the maturity of the larvae have been previously reported by Zheng et al. (2012) in Holotrichia parallela larvae. The hindgut wall of these larvae is populated by a reduced amount of coccoid cells in the L1 stage, although in the L3 stage, the density of bacteria is largely increased, with bacteroid cells dominating (Zheng et al., 2012).

In the adults, the relative abundances of Bacteroidetes, Proteobacteria (γ- and δ- classes) and the family Enterococcaceae (Firmicutes) were increased. Nevertheless, the overall composition of the adult hindgut wall community remained fairly constant compared to L3. This is in line with previous observations on M. hippocastani (Arias-Cordero et al., 2012). It was noted that the similarity between larval and adult bacterial communities becomes more evident in the later larval instars, suggesting that L3 larvae possess a community that is more closely related to that of the adults than to the L2 larvae. In addition, they noticed that the abundance of Enterobacteriaceae in the midgut increased continuously throughout the L2, L3, and adult stages (L2 < L3 < adult). In line with these findings is the presently observed increase of the genus Citrobacter from L2 to L3 (**Figures 3**, **4**). Such increase in abundance of Citrobacter representatives toward latter larval instars may be related to the increasing amount of ingested food as the larvae grow, as previously isolated Citrobacter sp. from the gut of M. hippocastani showed the ability to degrade xylan and starch in pure culture (Arias-Cordero et al., 2012). Furthermore, in adults, the high abundance of Enterococcaceae and Enterobacteriaceae representatives might be related to the shift to leaf-based diet, as these families showed resistance to tannins, an ubiquitous plant defense compound (Smith and Mackie, 2004; Singh et al., 2011).

The abundances of the bacterial genera in the L3 hindgut wall showed by qPCR (**Figure 4**) are in good agreement with the pyrosequencing result, being Citrobacter sp. dominant over Achromobacter sp., just as the Enterobacteriaceae family is more abundant than Alcaligenaceae in **Figure 3**. Contrary, Achromobacter sp. dominates in the pocket. This is also in line with the 454-pyrosequencing, where the sequences obtained clustered mainly within Actinobacteria and α- and β-Proteobacteria, taxa that showed very low abundances in the hindgut wall. This result highlights the singularity of the pocket bacterial community and suggest that they function as specialized symbiotic niches, analogously to previously described structures in other insects (Kikuchi et al., 2005; Grünwald et al., 2010).

#### Significance of the PHB Inclusions

Transmission electron microscopy unveiled a number of white cytoplasmatic inclusions in the pocket bacteria [potential poly-3 hydroxybutyrate (PHB)]. By GC-MS analyses, it was possible to confirm the presence of poly-3-hydroxybutyrate. Raman microspectroscopy revealed that PHB-accumulating bacteria are widely distributed throughout the lumen of the pocket pole. PHB is commonly accumulated by Eubacteria and Archaea and serves as a carbon reserve, stored in the form of water insoluble droplets in the cytoplasm (Rehm, 2003). Its presence is probably linked to the white cytoplasmatic inclusions observed in TEM. Likewise, PHB inclusions also are present in the endosymbiont Burkholderia sp. colonizing the midgut crypts of the bean bug R. pedestris. Each generation of this insect orally acquire the Burkholderia bacterium de novo from the environment, and the accumulation of PHB by the symbiont is crucial to ensure proper colonization of the crypts and correct development of the insect host (Kim et al., 2013). The colonization success by the PHB-accumulating symbiont could be related to its enhanced ability to cope with stress, as previous studies linked PHB accumulation to an increase of bacterial colonization efficiency and to tolerance to a variety of stresses such heat, reactive oxygen species, osmotic imbalance and nutritional depletion, among others (Kadouri et al., 2003; Kim et al., 2013). The high lipidic content within the pocket pole revealed by Raman micro-spectroscopy (**Figure 6**), suggests that the pockets are a nutritionally imbalanced habitat with a high C:N ratio that may favor the colonization by bacteria with the ability of accumulate PHB (Rehm, 2003). Also, oxygen limitation might contribute on selecting PHB-accumulating bacterial species over non-accumulating ones (Trainer and Charles, 2006). Symbiont sorting mechanisms in order to discard potentially pathogenic bacteria from the soil have been reported in the bean bug R. pedestris (Kim et al., 2013; Ohbayashi et al., 2015). However, in M. hippocastani, this putative discriminative process would not be as specific as in R. pedestris, since more than one bacterial phylotype are established in the pockets.

The presence of PHB is uncommon in vertically transmitted bacterial symbionts. Its accumulation is displayed mainly by freeliving microorganisms, or by symbionts of environmental origin (Kim et al., 2013). This suggests that the PHB-accumulating pocket symbionts (A. marplatensis and possibly S. maltophilia and P. myrsinacearum; see **Table 1**) might be acquired from the environment. These genera, along with Ochrobactrum thiophenivorans (which is not likely to accumulate PHB; see **Table 1**), have been previously detected in the rhizosphere (Bertrand et al., 2000; Kämpfer et al., 2008; Ryan et al., 2009). Moreover, the BLAST alignments of the pocket isolates belonging to these taxa matched those of bacteria previously isolated from roots and soil (data not shown). Considering that, an environmental origin for these pocket symbionts is more plausible than a vertical transmission from mother to offspring. This latter possibility, nevertheless, cannot be totally discarded (Engel and Moran, 2013).

#### Physiological Role of the Pockets

The pockets in M. hippocastani have been only once described in literature (Wildbolz, 1954). Nonetheless, symbioses between insect and bacteria is a common and disparate phenomenon in nature (Douglas, 2009; Hansen and Moran, 2014) and analogous structures harboring symbiotic microorganisms have been found in other insects. Bugs belonging to the family Alydidae are associated with ectosymbiotic bacteria of the genus Burkholderia. It this case, the bacterium colonizes the crypts located in the distal section of the midgut (Kikuchi et al., 2005). Similarly, stinkbugs of the families Pentatomidae and Cydnidae harbor Gammaproteobacteria related bacteria in crypts located in the same region of the midgut (Prado and Almeida, 2009; Hosokawa et al., 2012). Other structures containing endosymbiotic bacteria and yeasts have been characterized in the proximal midgut of cerambycid beetles (Grünwald et al., 2010). The role of these symbionts within the insect gut and their involvement in host's nutrition, however, remains largely unknown.

In M. hippocastani, the pockets might be sites for denitrification processes. A. marplatensis isolated from these

small structures showed full denitrifying capabilities in a commercial nitrate reduction assay (**Table 1**). Moreover, the abundance of lipids within the pocket pole unveiled by Raman micro-spectroscopy (**Figure 6**) makes possible that these compounds are used by the pocket symbionts as electron donors for respiratory processes using nitrate as an electron acceptor (NO<sup>3</sup> <sup>−</sup>). Denitrification has already been reported in other rhizophagous white grubs (Majeed and Miambi, 2014). The presence of pockets could be also related to the rhizophagous diet of the larvae, as they were spotted in the rhizophagous larvae of M. melolontha as well (**Figure 1F**), but no similar structure was found in Pachnoda marginata (Supplementary Figure S4), whose grub-like larvae thrive not on roots but on humic acids. Host's diet and taxonomy have been pointed as key determinants of the composition of the gut symbiotic community by previous studies (Egert et al., 2003, 2005; Colman et al., 2012; Jones et al., 2013). Either way, it is possible that the pocket symbionts produce some kind of beneficial compound for the insect host. This hypothesis, however, remains for future research.

#### CONCLUSION

Our data revealed a complex and dynamic microbial community attached to the hindgut wall of the forest cockchafer. The composition of this community may be dependent on host's life stage. L3 larvae showed a more close community to the adults than L2 larvae. In addition, the presence of particular bacterial niches attached to the larval hindgut (pockets) is reported. Regarding the surrounding hindgut wall, these niches harbored a differentiated bacterial community in which the families Micrococcaceae and Alcaligenaceae were dominant. These structures could be related to denitrification processes. Furthermore, the presence of poly-β-hydroxybutyrate (PHB) granules among pocket bacteria is demonstrated. Further research is needed to fully understand the function of the pockets, and especially to determine the role(s) of the cytoplasmatic inclusions.

#### AUTHOR CONTRIBUTIONS

PA-P performed DNA extraction, light microscopy, pyrosequencing data analysis, gas chromatography measurements,

#### REFERENCES


isolation, identification and metabolic testing of symbiotic bacteria and prepared the samples for Raman analysis. Also wrote the manuscript. EA-C spotted the pockets in the gut. Performed light microscopy, DNA extraction, fluorescence in situ hybridization, pyrosequencing data analysis and TEM data analysis. Also contributed in writing the manuscript. AN designed and performed qPCR experiments and analyzed the data. Also contributed in writing the manuscript. CE performed the Raman micro-spectroscopy analysis and analyzed the data. Also contributed in writing the manuscript. JR contributed in the design of TEM experiments and in the analysis of the data. MK contributed in the analysis of pyrosequencing data and calculated richness indexes and rarefaction curves. Also contributed in writing the manuscript. MW prepared samples for TEM, performed analysis and contributed in the analysis of the data. Spotted PHB inclusions in TEM images. UN contributed in the analysis of the Raman data. Also contributed in writing the manuscript. WB had the main idea of the project and supervised it. Also contributed in writing the manuscript.

#### FUNDING

This work was supported by the International Leibniz Research School for Microbial and Biomolecular Interactions (ILRS Jena), the Deutsche Forschungsgemeinschaft (DFG) and the Max Planck Society.

#### ACKNOWLEDGMENTS

The authors want to thank Eiko Wagenhoff and Peter Gawehn for supplying the insects, Angelika Berg for rearing them, and Anja David, Maritta Kunert, Henry Jahn, and Kerstin Ploß for their help during the experiments. Special thanks to Lorena Halty for her help during insect collection.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.00291/full#supplementary-material


by means of Raman spectroscopy. Anal. Bioanal. Chem. 397, 2929–2937. doi: 10.1007/s00216-010-3895-1


Escherichia coli. Appl. Microbiol. Biotechnol. 78, 371–376. doi: 10.1007/s00253- 007-1300-6



**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 © 2017 Alonso-Pernas, Arias-Cordero, Novoselov, Ebert, Rybak, Kaltenpoth, Westermann, Neugebauer and Boland. 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) or licensor 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.

# Lower Termite Associations with Microbes: Synergy, Protection, and Interplay

Brittany F. Peterson\* and Michael E. Scharf

Department of Entomology, Purdue University, West Lafayette, IN, USA

Lower-termites are one of the best studied symbiotic systems in insects. Their ability to feed on a nitrogen-poor, wood-based diet with help from symbiotic microbes has been under investigation for almost a century. A unique microbial consortium living in the guts of lower termites is essential for wood-feeding. Host and symbiont cellulolytic enzymes synergize each other in the termite gut to increase digestive efficiency. Because of their critical role in digestion, gut microbiota are driving forces in all aspects of termite biology. Social living also comes with risks for termites. The combination of group living and a microbe-rich habitat makes termites potentially vulnerable to pathogenic infections. However, the use of entomopathogens for termite control has been largely unsuccessful. One mechanism for this failure may be symbiotic collaboration; i.e., one of the very reasons termites have thrived in the first place. Symbiont contributions are thought to neutralize fungal spores as they pass through the termite gut. Also, when the symbiont community is disrupted pathogen susceptibility increases. These recent discoveries have shed light on novel interactions for symbiotic microbes both within the termite host and with pathogenic invaders. Lower termite biology is therefore tightly linked to symbiotic associations and their resulting physiological collaborations.

#### Edited by:

Christine Moissl-Eichinger, Medical University Graz, Austria

#### Reviewed by:

Saria Otani, Copenhagen University, Denmark Rebeca B. Rosengaus, Northeastern University, USA

\*Correspondence:

Brittany F. Peterson bfpeterson@email.arizona.edu

#### Specialty section:

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

Received: 08 January 2016 Accepted: 16 March 2016 Published: 08 April 2016

#### Citation:

Peterson BF and Scharf ME (2016) Lower Termite Associations with Microbes: Synergy, Protection, and Interplay. Front. Microbiol. 7:422. doi: 10.3389/fmicb.2016.00422 Keywords: termite, digestion, immunity, insect–microbe interactions, social insect, symbiosis

### INTRODUCTION

The close association of lower termites with microbes is fundamental to their biology. For the last century, understanding the intricacies of the relationship between termites and their gut symbionts, i.e., the termite holobiont, has been a major focus of termite research. The majority of this work emphasizes both the complexity and novelty of functions carried out to process lignocellulose within the termite gut (reviewed in Brune, 2014). For decades, termite wood digestion has been a quintessential example of symbiotic collaboration; however, symbionts have also been associated with a myriad of other functions in this system (reviewed in Ohkuma, 2008). For example, in addition to synergistic digestive collaboration, symbionts of lower termites have also been shown to play protective roles against pathogens both in vivo and ex vivo (Rosengaus et al., 1998, 2014; Chouvenc et al., 2009, 2013). This interaction between the termite symbiotic consortium and potential pathogens adds a layer of interplay within this already-complex microbial community. Here we summarize the diversity and roles symbionts play in lower termites, highlight the broad implications of both topics for understanding termite biology and symbiotic evolution, and emphasize how a holistic approach to studying termite biology is necessary to encompass the impact of this obligate symbiotic association.

Peterson and Scharf Lower Termite Associated Microbes

Lower termites are distinct from higher-termites in that they form relationships with both eukaryotic and prokaryotic symbionts within their digestive tracts (Eutick et al., 1978). While the diversity, abundance, and functionality of these symbionts fluctuates from species to species, an association with symbionts is ubiquitous and connected with much of the biology of termites. Fundamental defining aspects of lower termites, from eusociality to niche occupation, are impacted by their obligate association with microbes. Disruption of this community impacts termite physiological function, fitness, and survivorship (Cleveland, 1924; Thorne, 1997; Rosengaus et al., 2011b, 2014; Peterson et al., 2015; Sen et al., 2015). Lower termites house protists (unicellular eukaryotes), bacteria, and archaea all within the one-microliter environment of their hindgut, many of which are never found outside of this association. Restricted to their association with termites, these symbionts are exposed to and must tolerate a variety of chemical and biological stressors in the termite gut microenvironment. As the host termite feeds, forages, grows, and encounter pathogens, its symbiota are impacted. Thus, termites cannot be studied without also considering their symbionts. Characterizing and cataloging these microbes poses many challenges because most are unable to be cultured with traditional techniques due to their fastidious nature. This gut microenvironment boasts organismal and metabolic diversity which rivals some of the better studied macro-ecosystems. Approaching the termite holobiont as a fully functional, multifaceted ecosystem allows for concentration on individual species or processes and on the larger collaborative nature of the gut microenvironment.

### CHARACTERIZING THE LOWER TERMITE GUT CONSORTIUM

The key division between lower and higher termite species is the respective nature of their symbiotic partners. While both retain prokaryotic symbionts, lower termites also have flagellated protists living in their guts which is an ancestral trait shared with wood-feeding cockroaches, Cryptocercus sp. (Stingl and Brune, 2003; Lo and Eggleton, 2011; Brune and Dietrich, 2015). These protists belong to two groups: the oxymonads and the parabasalids. Originally described as parasites, protists were first found associated with termites over a century ago (Leidy, 1877). Since this original observation, roughly 500 termite-associated protist species have been described (reviewed in Ohkuma and Brune, 2011). As technology advances we are continually able to improve our understanding of the players and complexity of the termite gut community. In fact, new species of protistan symbionts are continually described from lower termite guts (Brugerolle and Bordereau, 2004; Gile et al., 2012; James et al., 2013; Tai et al., 2013; Radek et al., 2014), and the breadth of their diversity is thought to be drastically underestimated in general (Harper et al., 2009; Tai and Keeling, 2013). That being said, lower termites are thought to possess anywhere from a few to a dozen protist species as symbionts that maintain tight phylogenetic associations with their hosts (Tai et al., 2015).

As has happened with protist symbionts, our understanding of the bacterial consortium composition in lower termites is constantly evolving as methodologies and analyses improve. Early estimates from the eastern subterranean termite, Reticulitermes flavipes, numbered bacteria per gut in the millions, which seems to be a conservative approximation at best (Schultz and Breznak, 1978). Using culture-independent, cloning based methods, several groups have estimated the guts of lower termite species to contain anywhere from 222–1,318 ribotypes of bacteria (Hongoh et al., 2003a,b; Shinzato et al., 2005; Yang et al., 2005; Fisher et al., 2007). With the onset of nextgeneration sequence technologies this number has only grown. More recently, the gut lumen content of R. flavipes workers was described to contain over 4,761 species-level phylotypes of prokaryotic symbionts, with over 99% being bacteria (Boucias et al., 2013). The majority of these identified phylotypes are unique to the termite gut, having never been reported elsewhere and not having close-relative sequences available in databases. Coptotermes gestroi has been estimated to house 1,460 species of bacteria using Illumina technology (Do et al., 2014). These estimates vary for a variety of possible reasons, including local environment, study locus, methodological limitations/caveats, sampling strategy, diet, genetic background, and termite species. While identifying the microbial players within this system is an important step, describing the functions and interplay between them will be equally necessary for understanding termite biology and evolution.

### SYMBIOTIC COLLABORATION IN TERMITE DIGESTION AND NUTRITION

Apart from cataloging symbiont diversity, much of termite research has focused on their associations with the symbiotic microbes which aid in wood digestion. Feeding on this ligninrich, nitrogen-poor diet requires a suite of enzymes both to catalyze its breakdown and supplement its nutritional deficiencies. Termites and their symbionts complement each other's capabilities in this way. Termites contribute several highly active enzymes important to this process including endogenous cellulases (β-1, 4-endoglucanase, β-glucosidase) and lignin/phenolic detoxifiers (aldo-keto reductase, laccase, catalase, cytochrome p450s) (Scharf et al., 2010; Zhou et al., 2010; Raychoudhury et al., 2013; Sethi et al., 2013b). Protists in the hindgut of lower termites have been credited with the contribution of several important glycosyl hydrolases (GHFs 5, 7, 45) which aid in cellulolytic activity (Ohtoko et al., 2000; Todaka et al., 2010; Sethi et al., 2013a) and are important in hydrogen cycling (Inoue et al., 2005, 2007). Based on transcriptomic studies, protists possess many more potentially important cellulases (Todaka et al., 2007; Tartar et al., 2009). Also, both the termite host and protist symbionts possess proteases which may be important for utilizing bacteria as sources of nitrogenous compounds (Sethi et al., 2011; Tokuda et al., 2014). Although both protists and bacteria possess many hemicellulases (Inoue et al., 1997; Tartar et al., 2009; Tsukagoshi et al., 2014), termite endogenous cellulases have

been shown to have hemicellulase activity as well (Scharf et al., 2010, 2011; Karl and Scharf, 2015). However, despite this apparent hemicellulolytic redundancy, protists, bacteria, and archaea in the hindgut paunch clearly all contribute significantly to the overall efficiency of wood digestion (Peterson et al., 2015).

While protists are mainly responsible for lignocellulolytic activity, the prokaryotic community provides a more diverse subset of services in the termite gut. Spirochetes, the most conspicuous bacterial group in lower termite guts, are capable of diverse metabolic processes including acetogenesis, nitrogen fixation, and degradation of lignin phenolics (Lilburn et al., 2001; Graber and Breznak, 2004; Lucey and Leadbetter, 2014). The isolation and maintenance of pure cultures of several species of spirochetes from lower termite guts has been a powerful tool for describing their metabolic capabilities and collaborative potential within the community as a whole (Leadbetter et al., 1999; Lilburn et al., 2001; Salmassi and Leadbetter, 2003; Graber and Breznak, 2004, 2005; Graber et al., 2004; Dröge et al., 2006; Rosenthal et al., 2011).

Another major component of lower termite microbiota are the bacteria which are intimately associated with gut flagellates as intracellular endosymbionts (Stingl et al., 2005; Noda et al., 2009). There are four phyla of bacterial endosymbionts found within protist cells: Elusimicrobia, Bacteroidetes, Proteobacteria, and Actinobacteria (Hara et al., 2004; Noda et al., 2005; Stingl et al., 2005; Strassert et al., 2012). These groups have been found to ferment glucose, synthesize amino acids, produce cofactors, fix nitrogen, and recycle nitrogenous wastes (Noda et al., 2007; Hongoh et al., 2008a,b; Ohkuma and Brune, 2011; Strassert et al., 2012; Zheng et al., 2015). Methanobrevibacter, a methanogenic archaeal genus common across termite-associated flagellates, contribute methane to the gut environment using hydrogen that is present in copious amounts in the gut lumen as a product of cellulose metabolism (Shinzato et al., 1999; Tokura et al., 2000; Hara et al., 2004; Hongoh and Ohkuma, 2011). This adds another level of complexity to termite gut ecology by creating a tripartite symbiosis: prokaryotes within protozoa within termites.

Apart from archaea associated with termite gut flagellates, representative Methanobacteriaceae are also associated with the microaerobic termite gut lining (Leadbetter and Breznak, 1996; Ohkuma et al., 1999; Brune, 2011). Together with the flagellate endosymbiota, the large amount of methane created by termite digestion can be attributed to archaea which are typically associated with the hindgut lining (Brune, 2011; Hongoh and Ohkuma, 2011). In sum, the microbes present in lower termite guts comprise a diverse ecosystem capable of nitrogen cycling, carbohydrate metabolism, methanogenesis, amino acid biosynthesis, hydrogen turnover, and consequently, complementing deficiencies of the host.

In addition to the contributions of individual organisms, the host fraction (foregut, midgut, and salivary glands) and the symbiont fraction (hindgut) of the termite digestive system have been shown to work synergistically (Scharf et al., 2011). While both fractions have lignocellulolytic activity, combining protein extracts from both the host and symbiont fractions results in more sugar release in vitro than the sum of the parts. Additionally, recombinant host and symbiont enzymes have been shown to work efficiently in vitro to liberate glucose and pentose sugars from wood (Sethi et al., 2013a). Hence, wood digestion is truly the result of successful collaboration between termites and their hindgut symbionts. This collaborative physiological functionality is a driver in termite success and niche occupation, and it should continue to be a major focus to understand termite holobiont biology and ecology as we go forward.

### SYMBIONT–PATHOGEN INTERPLAY

Social living and foraging in microbe-rich environments puts termite workers at risk to encounter pathogens and creates the potential for epizootic events within termite colonies. Though the relationship between termites and their symbionts is often perceived to be purely nutritional, there is growing evidence that gut microbiota have infection-buffering potential. However, termites also have evolved complex hygienic behaviors to mitigate the spread and persistence of pathogenic agents (i.e., fungal conidia) within colonies (Rosengaus et al., 1998; Rosengaus et al., 2011a; Gao et al., 2012). Termites have been frequently observed to auto- and allogroom conidia from the bodies of themselves and nestmates. Passage through the alimentary canal and symbiont-filled hindgut effectively neutralizes fungal conidia (Chouvenc et al., 2009). Termites with perturbed gut microbiota, by oxygenation or chemical means, display a marked increase in susceptibility to fungal pathogens such as Metarhizium anisopliae and Beauveria bassiana (Boucias et al., 1996; Ramakrishnan et al., 1999; Rosengaus et al., 2014; Sen et al., 2015). One biochemical mechanism has been linked to this anti-fungal gut phenomenon in the form of symbiont-derived β-1, 3-glucanase activity (most likely protist) that is able to act on fungi and prevent their germination (Rosengaus et al., 2014). Similarly, the inhibition of this antifungal enzyme activity, β-1, 3-glucanase, results in a marked increase in termite susceptibility to a variety of pathogens (Bulmer et al., 2009) and is conserved evolutionarily from woodroaches to termites (Bulmer et al., 2012).

As mentioned above, grooming and hygienic behavior play an important role in termite immunity. Termites also participate in proctodeal trophallaxis as a means to replenish symbionts, nutrients, and chemical signals amongst individuals in the colony (Suarez and Thorne, 2000; Machida et al., 2001). This is another means by which symbionts and potential pathogens may interact, but it does not seem to play an important role in immune priming (Mirabito and Rosengaus, 2016).

Lastly, outside of the termite body, symbiotic bacteria provide additional protection. Termites build elaborate nest structures from fecal material to house their colonies. As with hindgut populations, these nest materials contain varying degrees of microbial abundance and richness dependent upon the species of termite (Rosengaus et al., 2003). This material contains diverse kinds of bacteria but has comparatively less fungus (Rosengaus et al., 2003; Manjula et al., 2015). The nests of one species of subterranean termite, Coptotermes formosanus, are commonly laden with symbiotic Actinobacteria demonstrated to

have antifungal activity ex vivo in nest walls (Chouvenc et al., 2013). This finding extends symbiont-mediated protection from the termite gut outside into the nest material, in at least one species.

### CONCLUDING REMARKS

Lower termite symbioses with microorganisms are unmistakably integral to termite biology. Hindgut microbial communities are tightly linked with termite digestion of wood and play important roles in supplementing this nutrient-poor food source. Symbionts catalyze reactions involved in the breakdown of all three major components of wood (cellulose, hemicellulose, and lignin phenolics) and supplement this diet by synthesizing other important nutrients. However, outside of the classic role for termite symbionts in digestion and nutrition, there is increasing recognition that they buffer the impacts of environmental stressors to their hosts. In particular, both protists and bacteria have been found to provide anti-fungal defenses in lower termites (Chouvenc et al., 2013; Rosengaus et al., 2014). Even fitness is impacted by the interconnectivity between termites and their symbionts (Rosengaus et al., 2011b). Recent discoveries emphasize that despite nearly a century of studying the obligate relationships between lower termites and microbes, there are still many facets of this complex association which are yet to be understood. Lower termites provide an important model for studying persistent, multi-layer symbioses.

It is also important to consider the role that symbiota play in other animal systems for the purpose of formulating relevant questions to probe, interrogate and eventually understand the termite holobiont. Recent discoveries in other models highlight microbiota as playing more active roles in host physiology, development, and behavior. These roles extend further than the bounds of the intestinal walls, affecting a range of processes from immune system development/maturation to mood and pain perception (Sommer and Backhed, 2013). The broad influence of gut microbiota found in these other systems can serve as an excellent guide to generate hypotheses for testing in the termite system.

Moving forward, based on recent and emerging trends, it will be imperative to consider all components of the termite holobiont when investigating aspects of termite biology. Understanding the role of symbiotic microbes in the physiological processes of digestion and immunity represent some of the first steps toward a better understanding the broader functionality of the lower termite consortium. Viewing any of these interactions within

#### REFERENCES


the termite holobiont as discrete may be an oversimplification. However, as methodologies and analyses advance, our ability to understand the functions of the consortium as a whole will continue to improve, as will our understanding of the roles of individual taxa in the system, and collaborations between host and symbiota. Efforts to characterize the holobiont in the presence and absence of stressors, both biotic and abiotic, using comprehensive omics-based approaches are likely to be major hypothesis-generating endeavors. However, the key to doing this successfully will involve careful sample preparation and carefully constructed analysis pipelines to limit taxonomic biases whenever possible. These big data approaches will in turn become a springboard into understanding symbiotic association, trends and commonalities, which may help to begin building models for the compartmentalization, complementation, and collaboration between lower termites and their symbiota.

Understanding the extent, bounds, and ramifications of these associations will be necessary to move toward a fuller appreciation of lower termite biology. Ultimately, studying the collective function and interplay between all members of this symbiosis in response to environmental challenges and in periods of stasis will shed light both on the micro-ecosystem that is a termite gut and the super-organism that is a termite colony.

### AUTHOR CONTRIBUTIONS

BP and MS developed, wrote, and revised the ideas and content presented in this manuscript. Both BP and MS approve the publishing of this manuscript and take responsibility for all of its contents.

#### FUNDING

Funding was provided by the U.S. National Science Foundation (grant no. 1233484CBET), Indiana Academy of Sciences (grant no. 2014-13), the 2014 Entomological Society of America Monsanto Graduate Research Grant, and the O. Wayne Rollins-Orkin Endowment at Purdue University.

#### ACKNOWLEDGMENT

The authors acknowledge Dr. S. P. Rajarapu, Dr. M. N. Fardisi, and our reviewers for helpful review, discussion, and comments during revision of this manuscript.



Rhinotermitidae). FEMS Microbiol. Ecol. 44, 231–242. doi: 10.1016/S0168- 6496(03)00026-6




**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 © 2016 Peterson and Scharf. 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) or licensor 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.

, Travis Ptacek 2, 3

,

# An abundance of Epsilonproteobacteria revealed in the gut microbiome of the laboratory cultured sea urchin, *Lytechinus variegatus*

#### *Edited by:*

*Christine Moissl-Eichinger, Medical University Graz, Austria*

#### *Reviewed by:*

*Jillian Petersen, Max Planck Institute for Marine Microbiology, Germany Irene Newton, Indiana University, USA*

#### *\*Correspondence:*

*Stephen A. Watts, Department of Biology, University of Alabama at Birmingham, 1300 University Blvd., CH464, Birmingham, AL 35294-1170, USA sawatts@uab.edu*

> *† 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: 21 May 2015 Accepted: 14 September 2015 Published: 13 October 2015*

#### *Citation:*

*Hakim JA, Koo H, Dennis LN, Kumar R, Ptacek T, Morrow CD, Lefkowitz EJ, Powell ML, Bej AK and Watts SA (2015) An abundance of Epsilonproteobacteria revealed in the gut microbiome of the laboratory cultured sea urchin, Lytechinus variegatus. Front. Microbiol. 6:1047. doi: 10.3389/fmicb.2015.01047* Casey D. Morrow<sup>4</sup> , Elliot J. Lefkowitz 2, 3, Mickie L. Powell <sup>1</sup> , Asim K. Bej <sup>1</sup> and Stephen A. Watts <sup>1</sup> \* *<sup>1</sup> Department of Biology, University of Alabama at Birmingham, Birmingham, AL, USA, <sup>2</sup> Biomedical Informatics, Center for*

, Ranjit Kumar <sup>2</sup>

Joseph A. Hakim1 †, Hyunmin Koo1 †, Lacey N. Dennis <sup>1</sup>

*Clinical and Translational Sciences, University of Alabama at Birmingham, Birmingham, AL, USA, <sup>3</sup> Department of Microbiology, University of Alabama at Birmingham, Birmingham, AL, USA, <sup>4</sup> Cell, Developmental, and Integrative Biology, University of Alabama at Birmingham, Birmingham, AL, USA*

In this study, we have examined the bacterial community composition of the laboratory cultured sea urchin *Lytechinus variegatus* gut microbiome and its culture environment using NextGen amplicon sequencing of the V4 segment of the 16S rRNA gene, and downstream bioinformatics tools. Overall, the gut and tank water was dominated by Proteobacteria, whereas the feed consisted of a co-occurrence of Proteobacteria and Firmicutes at a high abundance. The gut tissue represented Epsilonproteobacteria as dominant, with order Campylobacterales at the highest relative abundance (>95%). However, the pharynx tissue was dominated by class Alphaproteobacteria. The gut digesta and egested fecal pellets had a high abundance of class Gammaproteobacteria, from which *Vibrio* was found to be the primary genus, and Epsilonproteobacteria, with genus *Arcobacter* occurring at a moderate level. At the class level, the tank water was dominated by Gammaproteobacteria, and the feed by Alphaproteobacteria. Multi-Dimensional Scaling analysis showed that the microbial community of the gut tissue clustered together, as did the pharynx tissue to the feed. The gut digesta and egested fecal pellets showed a similarity relationship to the tank water. Further analysis of Campylobacterales at a lower taxonomic level using the oligotyping method revealed 37 unique types across the 10 samples, where Oligotype 1 was primarily represented in the gut tissue. BLAST analysis identified Oligotype 1 to be *Arcobacter* sp., *Sulfuricurvum* sp., and *Arcobacter bivalviorum* at an identity level >90%. This study showed that although distinct microbial communities are evident across multiple components of the sea urchin gut ecosystem, there is a noticeable correlation between the overall microbial communities of the gut with the sea urchin *L. variegatus* culture environment.

Keywords: egested fecal pellet, illumina MiSeq, 16S rRNA, QIIME, Gulf of Mexico, microbiome

## Introduction

Recent advancements in the discovery of gut microbial communities in the animal kingdom has offered a glimpse into the supportive role of various microbial taxa in growth, development, metabolism, and digestive physiology of the host, as well as protection from predators, and adaptive fitness to the environment they inhabit (Shin et al., 2011; Gomez et al., 2012; Nguyen and Clarke, 2012; Guinane and Cotter, 2013; Kostic et al., 2013; Heintz and Mair, 2014). Conventional microbiological culture-based methods, and more recently the advent of the culture-independent NextGen sequencing approach, has enhanced our capability to understand the gut microbial composition of many animals with the highest coverage, and in particular, a number of invertebrates such as Crustacea, Mollusca, and some Echinodermata (Harris, 1993; King et al., 2012; Gerdts et al., 2013; Kostic et al., 2013; Chauhan et al., 2014). Besides determining the microbial community profile of these invertebrates, the predictive roles of various microbial taxa in both the digestive health of the host, as well as the ecological importance of those bacteria to the host's community has been proposed. Among many ecologically and commercially important invertebrates, the sea urchin has received attention for its importance in the seafood industry (Muraoka, 1990; Andrew et al., 2002), as a model organism for developmental biology (McClay, 2011), and its role in nutrient cycling effecting the community structure and dynamics in the ecosystem they inhabit (Sauchyn and Scheibling, 2009a,b; Sauchyn et al., 2011). Yet, relatively little attention has been given to the sea urchin gut microbial ecology, and the potential role of those microbes in host health and other facets of its natural community (Becker et al., 2007, 2008, 2009; Lawrence et al., 2013).

Lasker and Giese (1954) first proposed a role of microbiota in nutrient digestion and absorption in sea urchins (Lasker and Giese, 1954), and in fact, most of the previous microbial analysis work on the sea urchin has focused on a generalized role of microbes in digestive support (Lawrence et al., 2013), or in disease progression (Becker et al., 2007, 2008, 2009). Later examinations would suggest involvement of the sea urchin gut egesta bacteria in nutrient transfer among trophic levels in their communities (Sauchyn and Scheibling, 2009a,b). Nevertheless, as the microbial ecosystems of the sea urchin gut continue to foretell a relationship between the microbial community and nutrient intake, determining the bacterial composition within the gut of the sea urchin fed a formulated diet in an aquaculture environment would provide valuable insights into sea urchin digestive physiology and health.

The variegated sea urchin, Lytechinus variegatus is often found in nearshore seagrass communities in the Gulf of Mexico, and consumes a wide variety of plant and animal material (Watts et al., 2013). In the laboratory culture environment, L. variegatus can process formulated diets containing macronutrients from a variety of sources (Hammer et al., 2012). Since gut microbiota has previously been implicated in the digestive process of sea urchins (Lasker and Giese, 1954; Fong and Mann, 1980; Sawabe et al., 1995), understanding the microbial composition of the sea urchin digestive system may elucidate the role of the gut microbiome in conferring host health through formulated diet. In this study, we describe the microbiome composition in the lumen of the digestive tract and gut digesta, along with egested fecal pellets, feeds, and the culture environment with high taxonomic coverage using a culture-independent method of NextGen sequencing technology and bioinformatics tools. The results from this study will help establish the microbial population that is conferred onto the sea urchin through the aquaculture conditions, as well as the trends of distribution and selective enrichment of the microbial community associated with the sea urchin, L. variegatus.

#### Materials and Methods

#### Collection and Culture of *L. variegatus*

Adult sea urchins were collected on April 2013, from Port Saint Joseph, Florida (29.80◦ N 85.36◦ W), and transported in seawater to a recirculating salt water system within the laboratory at the University of Alabama at Birmingham. Water conditions were maintained at 22 ± 2 ◦C, with a pH of 8.2 ± 0.2 and a salinity of 32 ± 1 ppt. using synthetic sea salt (Instant Ocean; Spectrum Brands, Inc., Blacksburg, VA) added to treated municipal water. Prior to use, municipal water was filtered by 5 micron sediment, charcoal, and reverse osmosis membranes, followed by an ion exchange resin, with the final addition of Instant Ocean sea salts to achieve the desired salinity of 32 ppt. Water was replaced in the recirculating seawater culture system at a rate of ca. 5% water exchange per day. Water quality was maintained using a dolomite mechanical gravel filter, followed by biological filtration using Bioballs biological media (Foster and Smith, Inc., Rhinelander, WI), and UV sterilization of water exiting the recirculating filter. The sea urchins were fed a formulated feed (Hammer et al., 2006) ad libitum, consisting of a relative percentage of 6% lipid, 28% protein, and 36% carbohydrate, once every 24–48 h for a 6 month period prior to analysis.

#### Sample and DNA Preparation

Two laboratory-cultivated sea urchins were used for the study (UR1 d = 50 mm, wet wt = 60.3 g, and UR2 d = 49 mm, wet wt = 63.2 g during the time described in the previous section). Sample collection from each sea urchin began 22 ± 1 h after feeding. Prior to dissection, the sea urchins were relocated to a temporary container containing sterile (autoclaved at 121◦C for 20 min at 103.42 kPa) sea water, from which the egested fecal pellets from each sea urchin were collected. After fecal pellet collection, the sea urchins were then removed from the water and dissected immediately. Briefly, an incision was made with sterilized scissors into the test surrounding the peristomial membrane, and a dissection was performed circumnavigating the mouth. The peristomial membrane, along with the nested mouth (the Aristotle's lantern) (Sodergren et al., 2006), was lifted from the sea urchin, while still maintaining the integrity of the digestive tract (Watts et al., 2013).

The pharynx enclosed by the lantern was separated from the digestive tract, collected intact, and rinsed with autoclaved sea water. The remaining segment of the digestive tract (gut tissue), which included the esophagus, stomach, and intestine (Holland, 2013), was then removed from the sea urchin. The gut was rinsed with autoclaved sea water, and voided of gut food pellets by gentle shaking. The gut tissue was collected separately from the gut food pellets and both were rinsed with autoclaved sea water. The microbiota obtained from the seawater within the closed recirculating system where the sea urchins were maintained was collected via vacuum filtration through Millipore 0.22µm filtration paper (EMD Millipore Corporation, Danvers, MA), and feeds were collected from the stock sea urchin food source (Hammer et al., 2006). All samples were divided into 3 separate sub-samples, flash frozen in liquid nitrogen, and preserved at −80◦C until used for DNA purification and preparation for sequencing of the 16S rRNA gene. Food samples and whole filter paper containing water system microbes were also divided into three subsamples, frozen in liquid nitrogen, and preserved at −80◦C until used.

#### Metacommunity DNA Purification and Generation of 16S rRNA Amplicon Library

Microbial community DNA was isolated using the Fecal DNA isolation kit from Zymo Research (Irvine, CA; catalog # D6010) following the manufacturer's instructions. Once the sample DNA was prepared, PCR was used with unique bar coded primers to amplify the hyper variable region 4 (V4) of the 16S rRNA gene, to create an amplicon library from metacommunity DNA samples (Kozich et al., 2013; Kumar et al., 2014). The oligonucleotide primers used for the PCR amplification of the V4 region of the 16S rRNA gene were as follows: Forward primer V4: 5′ -AATGAT ACGGCGACCACCGAGATCTACACTATGGTAATTGTGTGC CAGCMGCCGCGGTAA-3′ ; and Reverse primer V4: 5′ -CAA GAGAAGACGGCATACGAGATNNNNNNAGTCAGTCAGC CGGACTACHVGGGTWTCTAAT-3′ (Eurofins Genomics, Inc., Huntsville, AL) (Kumar et al., 2014). The individual PCR reactions were set up as follows: 10µL of 5X Reaction Buffer; 1.5µL (200µM) of each of the dNTPs; 2µL (1.5µM) of each of the oligonucleotide primers; 1.5µL (5 U) of the "LongAmp" enzyme kit (New England Biolabs, Ipswich, MA; cat # E5200S); 30µL (2–5 ng/µl) of the template DNA; and 3µL of sterile H2O to a total reaction volume of 50µL. The PCR cycling parameters were as follows: initial denature 94◦C for 1 min; 32 cycles of amplification in which each cycle consisted of 94◦C for 30 s, 50◦C for 1 min, 65◦C for 1 min; followed by final extension of 65◦C for 3 min; then a final hold at 4◦C. Following PCR amplification of the targeted gene, the entire PCR reaction was electrophoresed on a 1.0% (w/v) Tris-borate-EDTA/agarose gel. The PCR product (approximately 380 bp predicted product size) was visualized by UV illumination. The amplified DNA band was excised with a sterile scalpel, and purified from the agarose matrix using QIAquick Gel Extraction Kit according to manufacturer's instructions (Qiagen, Inc., Venlo, Limburg; cat # 28704).

#### Nextgen Sequencing and Bioinformatics Tools

The PCR products were sequenced using the NextGen sequencing Illumina MiSeq™ platform (Caporaso et al., 2012; Kozich et al., 2013; Kumar et al., 2014). We used a 250 bp paired-end kit from Illumina for the microbiome analysis. The samples were first quantified using Pico Green dye (Life Technologies, Grand Island, NY), adjusted to a concentration of 4 nM, then used for sequencing on the Illumina MiSeq (Kumar et al., 2014). The raw sequence data was then de-multiplexed and converted to FASTQ format (http://maq.sourceforge.net/fastq. shtml). The FASTQ files were subjected to quality assessment using FASTQC (http://www.bioinformatics.babraham.ac.uk/ projects/fastqc/), prior to merging and trimming of the raw sequence data, which was followed by quality filtering using the FASTX toolkit (http://hannonlab.cshl.edu/fastx\_toolkit/). Since the overlap between the paired reads from each 16S fragment was approximately 245 bases, the overlapping paired end regions were merged to generate a single high quality read, using the "fastq\_mergepairs" module of USEARCH (Edgar, 2010). Read pairs with an overlap of less than 50 bases or with mismatches (>20) in the overlapping region were discarded. The sequences were again checked for quality using FASTQC, which was followed by chimeric filtering using the "identify\_chimeric\_seqs.py" module of USEARCH (Edgar, 2010). The remainder of the steps were performed with the Quantitative Insights into Microbial Ecology microbiome analysis package (QIIME, v1.7.0) (http://qiime.org/) (Lozupone et al., 2007; Caporaso et al., 2010b; Navas-Molina et al., 2013; Kumar et al., 2014). Sequences were grouped into Operational Taxonomic Units (OTUs) using the clustering program UCLUST at a similarity threshold of 97% (Edgar, 2010). The Ribosomal Database Program (RDP) classifier was used to make taxonomic assignments (to the species level wherever possible) for all OTUs at a confidence threshold of 80% (0.8) (Wang et al., 2007). The RDP classifier (http://rdp.cme.msu.edu/) was trained using the Greengenes (v13.8) 16S rRNA database (http://greengenes. lbl.gov/cgi-bin/nph-index.cgi) (McDonald et al., 2011). The resulting OTU table included all OTUs, their taxonomic identification and abundance information. Additionally, OTUs whose average abundance was less than 0.0005% were filtered out. Remaining OTUs were then grouped together to summarize taxon abundance at different hierarchical levels of taxonomic classification (e.g. phylum, class, order, family, and genus). These taxonomy tables were also used to generate stacked column bar charts of taxon abundance using Microsoft Excel software (Microsoft, Seattle, WA). Multiple sequence alignment of OTUs was performed with PyNAST (Caporaso et al., 2010a). Subsampling was performed using the "single\_rarefaction.py" module of QIIME (v1.7.0), to account for variation in read depth across samples, (Gotelli and Colwell, 2011), at an even sampling depth of 77,194 reads per sample. The subsampled OTU table was used for downstream Beta and Alpha diversity analyses. A heatmap with the top 25 most highly abundant (>1% in any sample) taxa at the order level was generated using the "heatmap.2" function in R package (available at http://CRAN. R-project.org/package=gplots). The raw sequence files from this study are deposited in the NCBI SRA (http://www.ncbi.nlm.nih. gov/sra), under the accession number SRP062365.

#### Oligotyping of the V4 Hypervariable Region of the Campylobacterales 16S rRNA Gene

Oligotyping utilizes informative nucleotide variations between similarly clustered reads to designate an oligotype identity (Eren et al., 2013, 2014; Schmidt et al., 2014). After assignment of taxonomy for the total 1,137,478 quality reads, 296,777 sequences from the 10 samples were aligned using MUSCLE, which was implemented in MEGA software (Tamura et al., 2013). The aligned sequences were then used for oligotyping (Eren et al., 2013). After the initial Shannon entropy analysis, 29 variable sites were identified for oligotyping. The parameters required that each oligotype must (1) appear in at least one sample and (2) have a minimum abundance of 100 sequences for each unique oligotype. After elimination of oligotypes not meeting these parameters, 275,566 reads (92.85%) were retained. Each oligotype representative sequence was aligned to the NCBI non-redundant (nr) database using BLAST (http://blast.ncbi. nlm.nih.gov/Blast.cgi).

#### Statistical Analyses of Bacterial Diversity

The alpha diversity (diversity within the samples) of the sea urchin microbiome and the culture environment was determined using QIIME (v1.7.0). The alpha-diversity was estimated using observed OTUs, Shannon diversity index (Shannon, 1948; Hill et al., 2003; Marcon et al., 2014), and Simpson diversity index (Simpson, 1949; Hill et al., 2003). In order to estimate the beta diversity (differences between the samples), the OTUs of the bacterial communities were analyzed using Primer-6 analytical software (Primer-E Ltd., Plymouth Marine Laboratory, Plymouth U.K., v6.1.2) (www.primer-e.com). Discrete OTU counts per sample were standardized, and then transformed to the square root values (Clarke and Gorley, 2001). Multidimensional scale plots (Kruskal and Wish, 1978; Clarke, 1993; Clarke and Gorley, 2001), were generated according to Bray–Curtis similarity values (Bray and Curtis, 1957; Clarke and Gorley, 2001).

### Results

#### Total Illumina Sequence Reads, Quality Trimming, and OTU Designation

A total of 1,481,476 raw sequence reads of the V4 segment of the 16S rRNA gene from 10 samples of the two sea urchin (UR1 and UR2) gastrointestinal tracts, feeds, and tank water, were generated on an Illumina Miseq sequencing platform (**Table 1**). The sea urchin microbiome samples consisted of the gut tissues, pharynx tissues, gut digesta, and egested fecal pellets. After high stringent quality-based trimming, 1,137,478 quality sequence reads were used for further bioinformatics analyses. Within these reads, 181,169 sequences clustered into 609 OTUs from the gut tissue; 221,150 sequences clustered into 2,455 OTUs from the pharynx tissue; 219,512 sequences clustered into 926 OTUs from the egested fecal pellets; 204,048 sequences clustered into 1,562 OTUs from the gut digesta; 164,930 sequences clustered into 1,654 distinct OTUs from the sea urchin feed; and lastly 146,669 reads clustered into 1,511 OTUs from the tank water (**Table 1**). All OTUs were clustered at a 97% sequence similarity from the trimmed sequences of the respective samples using UCLUST (Edgar, 2010; Koo et al., 2014).

#### Microbial Diversity across Different Samples

The relative abundances of taxa identified to the most resolvable taxa (phylum, class, order, family, and genus) across all 10

TABLE 1 | Sample statistics following NextGen sequencing and the diversity values, as determined by QIIME (v1.7.0), are listed.


*Included are the number of raw sequences, trimmed sequences, and unique OTUs. Shannon and Simpson diversity indices are also presented. UR1, sea urchin 1; UR2, sea urchin 2.*

samples are elaborated in **Figure 1**. In the gut tissue samples of the sea urchins, microorganisms belonging to phylum Proteobacteria represented the highest relative abundance. Further analysis revealed class Epsilonproteobacteria to be dominant, and from within this class, order Campylobacterales was the most abundant taxon. Resolution to the genus level could not be achieved in the gut tissue samples. The pharynx tissue of the sea urchins was also dominated by Proteobacteria, and at the class level, Alpha-, Beta-, Epsilon-, and Gammaproteobacteria were presented. Arcobacter, Mycoplana, and Vibrio appeared as the highly represented genera from phylum Proteobacteria. Phylum Firmicutes was represented by a high relative abundance of the genera Bacillus and Allobaculum.

The gut digesta consisted mainly of bacteria belonging to phylum Proteobacteria, with class Gammaproteobacteria being distinguishably elevated. The dominant genera were Agarivorans, Arcobacter, Shewanella, and Vibrio, all of which belonging to phylum Proteobacteria. The bacterial composition in the egested fecal pellets consisted of many of the same taxa observed in the gut digesta. In the egested fecal pellets, Proteobacteria accounted for the highest abundance, and at the class level, Gammaproteobacteria was dominant. At the genus level, Agarivorans, Arcobacter, Shewanella, and Vibrio were detected as dominant taxa.

The microbiota of the sea urchin feed consisted of phylum Proteobacteria, as well as Firmicutes at the highest abundance. Classes Alpha- and Betaproteobacteria were dominant in the feed, and at the genus level, Agrobacterium, Acinetobacter, Limnohabitans, and Mycoplana were observed. From phylum Firmicutes, order Lactobacillales dominated in the feed, and at the genus level, Lactobacillus, Lactococcus, Leuconostoc, and

Proteobacteria. At the highest resolution, order Campylobacterales was determined to be the most abundant taxa in the gut tissue. In the gut digesta and egested fecal pellets, *Vibrio*, *Arcobacter*, and *Agarivorans* were observed. Relative abundances were performed through QIIME (v1.7.0), and graphs were generated using Microsoft Excel software (Microsoft, Seattle, WA). UR1, sea urchin 1; UR2, sea urchin 2.

Streptococcus were observed. The microbial composition of the tank water was found to be more diverse as compared to the other samples. Of the represented phyla, Proteobacteria was found to be dominant, followed by Chloroflexi, and to a lesser extent Bacteroidetes. Classes Gamma- and Alphaproteobacteria were dominant, and at the order level, Alteromonadales and Vibrionales were represented at relatively high abundances. In addition, significant abundances of genera Arcobacter, Agarivorans, Shewanella, Pseudoalteromonas, and Vibrio were identified within phylum Proteobacteria. For all samples, the taxonomic groups identified at the genus level have been elaborated in Supplementary Table 1.

#### Differentiation of Distinct Taxa using Oligotyping Methods and Blast

Oligotyping analysis of those sequences corresponding to order Campylobacterales in the 10 samples of this study revealed 37 different oligotypes (**Figure 2**; UR1, sea urchin 1, UR2, sea urchin 2). Of these oligotypes, 21 were found in the UR1 and 11 in the UR2 gut tissues; 21 in the UR1 and 30 in the UR2 pharynx tissues; 17 in the UR1 and 26 in the UR2 gut digesta; 18 in the UR1 and 17 in UR2 egested fecal pellets. The tank water and feed contained 18 and 6 oligotypes, respectively. Of all the identified oligotypes, Oligotype 1 was found to be overrepresented in the gut tissues of the sea urchins, with a relative abundance of 92.7% for UR1 and 91% for UR2. This oligotype was detected in the tank water at 0.3%, and the sea urchin feed at 22.8% (**Figure 2**). Across all samples, Oligotype 2 (which ranged from 8.5% to 88.36%) and Oligotype 3 (2.3% to 60%) were highly abundant, except for the gut tissues (**Figure 2**). A MEGABLAST search of the representative sequence of Oligotype 1 displayed a close match to an uncultured Arcobacter sp. clone (Identity: 91%, Evalue: 1.82E–87), Arcobacter bivalviorum (Identity: 91%, E-value: 2.00e–89), Sulfuricurvum sp. (Identity: 90%, E-value: 4.00E–86),

and an uncultured bacterium clone (Identity: 90%, E-value: 2.00e-89; Supplementary Table 2). A MEGABLAST search was performed on the other 36 identified oligotypes, revealing most to be closely related to uncultured Arcobacter sp., or uncultured bacterium clones (Supplementary Table 2).

#### Statistical Analysis

Rarefaction curves representing the number of unique OTUs from the normalized 16S rRNA sequences obtained from two sea urchins and their environments (total of 10 samples) reached or approached a plateau, indicating that a sufficient sequencing depth was used to assess community diversity (Supplementary Figure 1). Shannon (Shannon, 1948; Hill et al., 2003; Marcon et al., 2014) and Simpson diversity indices (Simpson, 1949; Hill et al., 2003) displayed relatively low diversity within the gut tissue samples, whereas moderate diversity within egested fecal pellet and gut digesta samples; and high diversity within pharynx tissue, sea urchin feeds, and tank water samples (**Table 1**). The multidimensional-scaling (MDS) plot (Kruskal and Wish, 1978; Clarke, 1993; Clarke and Gorley, 2001) revealed three distinct clusters of similarity among corresponding samples from the two sea urchins (**Figure 3**). In the MDS plot, the first dimension of gut tissues were differentiated from all other samples, and the second dimension separated the pharynges and feeds from the rest of the samples, i.e., the egested fecal pellet, gut digesta, and tank water (**Figure 3**). Subsampling of OTUs showed no significant differences in the cluster patterns of microbial communities in the respective samples.

Inter-sample microbial community compositions showed a similarity between samples (**Figure 4**). The gut tissue revealed a significant abundance of members from order Campylobacterales. The presence of Campylobacterales was also observed to be highly abundant in the gut digesta and egested fecal pellets, along with a significant presence of order Vibrionales. In the pharynx tissue, orders Burkholderiales and Caulobacterales were found to be abundant, whereas the tank water had high representation of order Alteromonadales, and the feed had a significant presence of Lactobacillales. The feed also presented orders Burkholderiales and Caulobacterales (**Figure 4**).

#### Discussion

Our study revealed that, although the sea urchin L. variegatus has a primitive gut as compared to the highly compartmentalized digestive systems in higher order deuterostomes (Sauchyn et al., 2011; Holland, 2013), distinct microbial compositions and abundances were noticed in the gut tissue, pharynx and the gut digesta, which shared a striking similarity with the food and culture environments. Additionally, it appears that the

microbiota of the sea urchin consisted of a high abundance of Proteobacteria, which is comparable to observations of previously examined marine invertebrate gut microbiota (Van Horn et al., 2011). For example, in the sea slug, members of Alpha-, Beta-, and Gammaproteobacteria have been observed as overrepresented (Devine et al., 2012), and in the gut of the sea cucumber Apostichopus japonicus, an echinoderm, it was shown that members of Delta- and Gammaproteobacteria are dominant (Gao et al., 2014).

The luminal surface of the gut contained a low overall bacterial diversity, but a high relative abundance of order Campylobacterales of class Epsilonproteobacteria (**Figure 1**). It has been reported that representatives from this class have been found to inhabit many ecological niches, both terrestrial and marine, performing a diversity of metabolic functions (Eppinger et al., 2004; Gupta, 2006). In the marine environment, members of Epsilonproteobacteria have been associated as gill symbionts of hydrothermal vent dwellers such as the bivalve Bathymodiolus azoricus (On, 2001) and gastropod Cyathermia naticoides (Zbinden et al., 2014); as residents of other bivalves such as mussels Brachidontes sp. of marine lakes (Cleary et al., 2015) and the Chilean oyster Tiostrea chilensis (Romero et al., 2002); as epibionts of crustaceans such as Kiwa puravida (Goffredi et al., 2014); and lastly, as gut microbial inhabitants of the aquacultured Norway lobster Nephrops norvegicus (Meziti et al., 2012) and hydrothermal vent dwelling shrimp, Rimicaris exoculata (Durand et al., 2010). Therefore the commonality of the occurrence of Epsilonproteobacteria in marine invertebrates and the sea urchins in our study may indicate a mutual benefit between the bacterial taxa and the host, perhaps at the physiological and nutritional level.

Further analysis of the lower level of taxonomic groups within Campylobacterales showed 37 oligotypes across all ten samples, with Oligotype 1 displaying a dominant presence in the gut tissue (**Figure 2**). This suggests that Oligotype 1 is the preferred bacterial group in the sea urchin gut. Additionally, a MEGABLAST search of the representative sequence of the highly abundant gut tissue Oligotype 1 revealed an uncultured species of Arcobacter sp., as well as Sulfuricurvum sp., and Arcobacter bivalviorum (Identities >90%). In a previous study, Epsilonproteobacteria clones identified as Arcobacter sp. were found to be associated with marine organisms, including shrimp species (Rimicaris exoculata) and the Chilean oyster (Tiostrea chilensis) (Romero et al., 2002; Durand et al., 2010). Taxonomic groups similar to Oligotype 1 were also found in the sea urchin feed and water samples, although to a much lesser extent, suggesting that the culture environment may have contributed to the high abundance of Oligotype 1 in the gut tissue microbial ecosystem following proliferation (**Figure 2**).

As food enters the digestive tract of sea urchins, it is enveloped in a mucosal film that remains intact even after egestion, as a microbial-enriched fecal pellet (Sauchyn et al., 2011; Holland, 2013). The microbiota of the gut digesta and egested fecal pellets both contained a high abundance of Gammaproteobacteria, specifically Vibrio of family Vibrionaceae (**Figure 1**). In as early as 1954, Lasker and Geise reported colonization of bacteria in the gut digesta through microscopic observation (Lasker and Giese, 1954). Similarly in our study, a preliminary examination of the egested fecal pellets using transmission electron microscopy showed comma, round, and rod shaped structures, which appeared to be bacteria resembling Vibrio, Arcobacter and Agarivorans, genera later determined by NextGen sequencing using the Illumina MiSeq sequencing platform (Supplementary Figure 2). Besides morphological studies, much attention has been allotted to the bacteria colonizing the ingested feed of the sea urchin, with many investigations implicating those bacteria as both crucial to the digestive physiology of the sea urchin, as well as an enriched source of nutrients to organisms at various trophic levels in the hydrosphere (Johannes and Satomi, 1966; Koike et al., 1987; Sauchyn et al., 2011). Previous studies on the gut related microbiota of sea urchins have described the potential symbiotic support of certain strains of Vibrio to the sea urchin Strongylocentrotus droebachiensis, specifically nitrogenase activity, which is necessary for nitrogen fixing in the assimilation of proteins in sea urchin gonad (Fong and Mann, 1980; Guerinot et al., 1982).

Trends of microbial ecology in the sea urchin have been suggested by Guerinot and Patriquin (1981), who proposed a possibility of an endemic microbiota that will not dissociate

from the gut wall of the sea urchin as food transits through the digestive tract (Guerinot and Patriquin, 1981; Lawrence et al., 2013). Evidence of this can be observed in the current study, as the gut digesta and egested fecal pellets were heavily dominated by Vibrio species, which were not observed to be significant in the gut tissue (**Figure 1**). Moreover, a unique oligotype (Oligotype 1) was observed in the gut tissue, which did not appear to be as significant in the gut digesta and egested fecal pellets. This indicates that there is a preference by the host to select specific microbial taxa, perhaps necessary for their nutrition and health (Thorsen, 1998). Moreover, the pharynx tissue shared many of the bacterial taxa of the sea urchin feed (**Figure 1**), suggesting a likely influence and transmittance of microbes from the food source, which is supported through oligotype analysis (**Figure 2**), a trend also observed by Meziti et al. (2007) in P. lividus (Meziti et al., 2007). The outcome of this study has established for the first time the microbial community composition in the sea urchin L. variegatus gut ecosystem, as well as its culture environments, using NextGen sequencing and bioinformatics to achieve taxonomic coverage at the highest level. Future evaluation of the functional metagenomics of

#### References


the gut microbiome of L. variegatus is warranted to establish the role of the microbial community associated with the digestive physiology, nutritional and other health benefits of this animal.

#### Acknowledgments

The following are acknowledged for their support of the Microbiome Resource at the University of Alabama at Birmingham: School of Medicine, Comprehensive Cancer Center (P30AR050948), Center for AIDS Research (5P30AI027767), Center for Clinical Translational Science (UL1TR000165) and Heflin Center. Animal husbandry supported in part by NIH P30DK056336. Animal care and use was approved by the UAB Institutional Animal Care and Use Committee.

#### Supplementary Material

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2015.01047


stress resistance in Caenorhabditis elegans fed respiratory deficient E. coli. BMC Microbiol. 12:300. doi: 10.1186/1471-2180-12-300


Kozich, J. J., Westcott, S. L., Baxter, N. T., Highlander, S. K., and Schloss, P. D. (2013). Development of a dual-index sequencing strategy and curation pipeline for analyzing amplicon sequence data on the MiSeq Illumina sequencing platform. Appl. Environ. Microbiol. 79, 5112–5120. doi: 10.1128/AEM.01043-13


Echinodermata). Mar. Biol. 132, 423–433. doi: 10.1007/s0022700 50408


gastropod Cyathermia naticoides (North East-Pacific Rise). Mar. Biol. 162, 435–448. doi: 10.1007/s00227-014-2591-7

**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 © 2015 Hakim, Koo, Dennis, Kumar, Ptacek, Morrow, Lefkowitz, Powell, Bej and Watts. 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) or licensor 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.

# Spatial and Temporal Dynamics of Pacific Oyster Hemolymph Microbiota across Multiple Scales

Ana Lokmer <sup>1</sup> \*, M. Anouk Goedknegt <sup>2</sup> , David W. Thieltges <sup>2</sup> , Dario Fiorentino<sup>1</sup> , Sven Kuenzel <sup>3</sup> , John F. Baines 3, 4 and K. Mathias Wegner <sup>1</sup>

<sup>1</sup> Coastal Ecology, Wadden Sea Station Sylt, Alfred Wegener Institute - Helmholtz Centre for Polar and Marine Research, List auf Sylt, Germany, <sup>2</sup> Department of Coastal Systems, Royal Netherlands Institute for Sea Research, Utrecht University, Texel, Netherlands, <sup>3</sup> Max Planck Institute for Evolutionary Biology, Plön, Germany, <sup>4</sup> Institute for Experimental Medicine, Christian-Albrechts-Universität zu Kiel, Kiel, Germany

#### Edited by:

Christine Moissl-Eichinger, Medical University of Graz, Austria

#### Reviewed by:

Colleen A. Burge, University of Maryland, Baltimore Country, USA Elisabeth Margaretha Bik, Stanford University, USA

> \*Correspondence: Ana Lokmer alokmer@awi.de

#### Specialty section:

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

Received: 27 April 2016 Accepted: 18 August 2016 Published: 31 August 2016

#### Citation:

Lokmer A, Goedknegt MA, Thieltges DW, Fiorentino D, Kuenzel S, Baines JF and Wegner KM (2016) Spatial and Temporal Dynamics of Pacific Oyster Hemolymph Microbiota across Multiple Scales. Front. Microbiol. 7:1367. doi: 10.3389/fmicb.2016.01367 Unveiling the factors and processes that shape the dynamics of host associated microbial communities (microbiota) under natural conditions is an important part of understanding and predicting an organism's response to a changing environment. The microbiota is shaped by host (i.e., genetic) factors as well as by the biotic and abiotic environment. Studying natural variation of microbial community composition in multiple host genetic backgrounds across spatial as well as temporal scales represents a means to untangle this complex interplay. Here, we combined a spatially-stratified with a longitudinal sampling scheme within differentiated host genetic backgrounds by reciprocally transplanting Pacific oysters between two sites in the Wadden Sea (Sylt and Texel). To further differentiate contingent site from host genetic effects, we repeatedly sampled the same individuals over a summer season to examine structure, diversity and dynamics of individual hemolymph microbiota following experimental removal of resident microbiota by antibiotic treatment. While a large proportion of microbiome variation could be attributed to immediate environmental conditions, we observed persistent effects of antibiotic treatment and translocation suggesting that hemolymph microbial community dynamics is subject to within-microbiome interactions and host population specific factors. In addition, the analysis of spatial variation revealed that the within-site microenvironmental heterogeneity resulted in high small-scale variability, as opposed to large-scale (between-site) stability. Similarly, considerable within-individual temporal variability was in contrast with the overall temporal stability at the site level. Overall, our longitudinal, spatially-stratified sampling design revealed that variation in hemolymph microbiota is strongly influenced by site and immediate environmental conditions, whereas internal microbiome dynamics and oyster-related factors add to their long-term stability. The combination of small and large scale resolution of spatial and temporal observations therefore represents a crucial but underused tool to study host-associated microbiome dynamics.

Keywords: host-associated communities, Crassostrea gigas, distance-decay relationship, spatiotemporal patterns, spatiotemporal dynamics, marine invertebrate microbiota, amplicon analysis

## INTRODUCTION

Assessing the temporal and spatial stability of microbial communities is vital for understanding and predicting their response to disturbances (Shade et al., 2012) and thus their functioning in a changing environment. This requires knowledge of the underlying disturbance-free community dynamics (Hunt and Ward, 2015). More precisely, it is crucial to identify regular (e.g., daily, seasonal) patterns, normal range of variation in community dynamics, as well as the processes and factors affecting community assembly and structure (Mutshinda et al., 2009; Costello et al., 2012; Nemergut et al., 2013) to establish a baseline against which to measure disturbance effects (Stenuit and Agathos, 2015).

Although controlled repeated-measures experiments in the laboratory (e.g., Shade et al., 2011; Berga et al., 2012; Lokmer and Wegner, 2015) are indispensable for a mechanistic understanding of how environmental factors affect microbial community dynamics, such results may not directly translate to natural conditions, which represent a blend of abiotic and biotic disturbances varying in their intensity, predictability, spatial scale and duration (Bender et al., 1984; Sousa, 1984; Paine et al., 1998; Berga et al., 2012). Studying natural temporal variability of microbial communities represents a valuable complement to controlled experiments, as it provides an opportunity to estimate the effects of known environmental factors and disturbances, as well as to uncover yet unknown and potentially important determinants of community structure and dynamics (Shade et al., 2013; David et al., 2014; Faust et al., 2015). So far, longitudinal studies and time series have helped to elucidate the dynamics of free-living microbial communities ranging from marine sediments (Gobet et al., 2012) and coastal ocean (Gilbert et al., 2012) to soil (Kato et al., 2015) and freshwater habitats (Peura et al., 2015). Similar studies regarding hostassociated microbiota are almost exclusively limited to humans, (e.g., Caporaso et al., 2011; David et al., 2014; DiGiulio et al., 2015), and a handful of model organisms (e.g., Fink et al., 2013; Marino et al., 2014). Whereas temporal patterns have been studied in non-model organisms, the focus remains on the population level and the sampling resolution usually coincides with significant host/environment-related shifts: developmental (e.g., Trabal et al., 2012; Hroncova et al., 2015), seasonal (e.g., Zurel et al., 2011; Bjork et al., 2013; Ransome et al., 2014) or abiotic disturbances (e.g., Vega Thurber et al., 2009; Wegner et al., 2013; Tracy et al., 2015). Longitudinal, individual-based, repeated-measure studies remain scarce (but see Pratte et al., 2015; Glasl et al., 2016).

Analogous to longitudinal studies, examining spatial variability and biogeographical patterns over multiple spatial scales, especially if combined with knowledge of environmental gradients, can shed light on relative importance of stochastic and deterministic processes for shaping microbial communities (Green and Bohannan, 2006; Nakaoka et al., 2006; Caruso et al., 2011; Hanson et al., 2012; Borer et al., 2013). One example is the distance-decay relationship, i.e., decreasing similarity between communities with increasing distance, a universal biogeographical pattern that arises through spatially-correlated environmental conditions or through dispersal limitation and has been demonstrated for microbial communities in both marine and terrestrial habitats (Bell, 2010; Martiny et al., 2011; Zinger et al., 2014; Nguyen and Landfald, 2015). Martiny et al. (2011) found that dispersal limitation affected community similarity within salt marshes, whereas environmental factors played more prominent role at regional or continental scale. Conversely, dispersal-related effects in arctic heathland soils were apparent only at larger scales (Hill et al., 2015), illustrating the importance of a particular context for interpretation of the observed patterns.

Spatially stratified sampling strategies can reveal drivers behind the structure and dynamics of free-living microbial communities (e.g., Caruso et al., 2011; Ristova et al., 2015) as well as of host-associated microbiota (Mihaljevic, 2012). In addition to environmental factors (Moro et al., 2011), examining spatial variation can disentangle geographic influences from those of host life stage (Hroncova et al., 2015), genotype and/or diet (Sudakaran et al., 2012; Yatsunenko et al., 2012; Linnenbrink et al., 2013). However, spatial and geographic information has so far primarily served to delineate core microbiomes (e.g., King et al., 2012; Wong et al., 2013; Dishaw et al., 2014) or to differentiate between the microbiomes of closely related species (Zouache et al., 2011; Phillips et al., 2012). This applies especially to marine hosts (e.g., Morrow et al., 2012; Reveillaud et al., 2014; Trabal Fernandez et al., 2014). Studies considering aspects of within-species spatial variation are less common and focused on large-scale differences between environmentally distinct sites (e.g., Trabal et al., 2012; Pierce et al., 2016; Ziegler et al., 2016). Although marine sedentary organisms offer a good opportunity to examine factors and processes shaping dynamics of their associated microbiota over multiple spatial scales using spatially nested designs, this has not been done yet. Including a temporal component into such studies would further improve our understanding of natural microbial community dynamics (e.g., Fortunato et al., 2012; Ransome et al., 2014; Pierce et al., 2016) and thus refine the reference framework for evaluating disturbance effects.

The Pacific oyster (Crassostrea gigas) is such a sedentary organism, highly suitable for the combined estimation of spatial and temporal patterns of microbiome assembly. However, sitespecific differences in host-associated microbial communities cannot be separated from host factors by studying natural spatial variability only, as hosts at different sites can be adapted or acclimated to their abiotic and biotic environment (Wendling and Wegner, 2015) or differ due to historical reasons, e.g., invasion histories (Moehler et al., 2011). In contrast to vertebrates, microbiota of most other organisms are closely related to environmental microbial communities (Ley et al., 2008) and translocation experiments with algae (Campbell et al., 2015), and sponges (Burgsdorf et al., 2014) indicate that site is a major determinant of microbiome composition. However, similar experiments with oysters suggest that the influence of site and host factors differs between the tissues (Lokmer et al., 2016). Altogether, combining translocation with a survey of spatial and temporal variation represents a relatively simple means to improve our understanding of the dynamics and function of host-associated microbiota in marine sedentary organisms.

One important function of microbiota that directly contributes to host fitness is defense against pathogens (McFall-Ngai et al., 2013). For example, some of the bacteria inhabiting the oyster hemolymph (a tissue with immune function analogous to vertebrate blood) produce antimicrobial compounds, thus preventing colonization by external pathogens and disease (Defer et al., 2013; Desriac et al., 2014). Hemolymph microbiota can also play part in oyster interactions with abiotic environment (i.e., temperature) by quick (hours to days) adjustments in community composition (Lokmer and Wegner, 2015; Lokmer et al., 2016). Despite the openness of oyster circulatory system and high oyster filtration activity, some bacteria such as Vibrio spp. persist in the hemolymph in the absence of an environmental source population (e.g., if held in sterile seawater) over a range of environmental conditions and could thus be considered resident (Vasconcelos and Lee, 1972; Lokmer et al., 2016). Presence of other, transient bacteria is strictly dependent on the external source community and thus reflects immediate environmental conditions (Lokmer et al., 2016). Dynamics of resident and transient components of the hemolymph microbiome are thus likely shaped by different processes and factors. However, despite its significance for oyster fitness, our knowledge about the variability and dynamics of hemolymph microbiota under natural conditions is almost exclusively limited to a subset of cultivable and potentially pathogenic bacteria, mostly of the genus Vibrio (Garnier et al., 2007; Wendling et al., 2014; Lemire et al., 2015).

In order to examine how site and host genotype affect diversity, composition and dynamics of oyster hemolymph microbiota, we performed a reciprocal translocation experiment with two genetically differentiated oyster populations from two sites in the Wadden Sea (Texel, Netherlands and Sylt, Germany, Moehler et al., 2011), and repeatedly sampled hemolymph from the same individuals over one summer season (**Figure 1**). Prior to the field deployment, we administered antibiotics to half of the oysters in order to remove a large portion of resident microbiota and account for priority effects. In addition, our field deployment (**Figure 1**) allowed us to examine spatial variation of complete and resident hemolymph microbiota over small (<1 m) and medium scales (101–10<sup>2</sup> m, within site). With such spatially and temporally stratified design we can now try to disentangle the relative contribution of different processes (immigration, within-microbiome interactions), and factors (host genetics, geography, environmental conditions) that shape the oyster hemolymph microbiota under natural conditions.

#### MATERIALS AND METHODS

The experimental setup and the experiment timeline are shown in **Figure 1**. The pretreatment, the sampling of the hemolymph and the seawater as well as wet-lab procedures are described in detail in (Lokmer et al., 2016) and therefore only briefly outlined here.

### Oyster Collection, Laboratory Pretreatment and Sampling

Oysters from the northern and southern Wadden Sea populations (Moehler et al., 2011) were collected at intertidal mixed oyster/mussel beds at Oddewatt, Sylt, Germany (55◦ 1 ′ N, 8◦ 26′ E) and at de Cocksdorp, Texel, Netherlands (53◦ 0 ′ N, 4◦ 54′ E), respectively. After the removal of epibionts (mainly barnacles) by scrubbing, half of the animals were transported from Sylt to Texel and vice-versa (**Figure 1**). The laboratory pretreatment was then conducted at the AWI Wadden Sea Station Sylt for the Sylt experiment (i.e., for the oysters later deployed on Sylt) and at the NIOZ Texel for the Texel experiment (i.e., for the oysters later deployed on Texel). For the pretreatment, oysters from each population were divided into two groups: the control group was kept in local 0.2µm filtered (sterile) seawater, whereas a mix of antibiotics with different mode of action (ampicillin, tetracycline, gentamicin, and kanamycine, Sigma-Aldrich, Hamburg, Germany, final concentration 400µg/l seawater) was added to the second one to remove as wide range of resident bacteria as possible. After 4 days and prior to the field deployment, hemolymph samples for the analysis of the associated microbial communities were taken with a syringe.

### Field Deployment and Sampling

Mesh bags with four oysters each (one per treatment combination: origin × antibiotic) were deployed in groups of two (Sylt) or three (Texel) and fixed with iron rods at original sites of collection on Sylt or Texel (**Figure 1**). In this way, we could estimate how the spatial scale - within the bag groups/sampling spots (<1 m) and between the sampling spots (101–10<sup>2</sup> m) - affects hemolymph microbiota. Hemolymph and seawater samples for the analysis of microbiota were taken directly in the field, placed on ice and immediately frozen upon return to the laboratory. Sampling was performed biweekly on Sylt and monthly on Texel. This, along with some other differences between Sylt and Texel (total number of oysters: 120 on Sylt, 96 on Texel; experiment duration: 1 June—23 August 2012 on Sylt, 14 June—24 August 2012 on Texel) was due to logistic reasons.

### DNA Extraction, PCR and Sequence Quality Control and Preprocessing

DNA was extracted from 200 ± 20 µl of hemolymph with Wizard SV 96 Genomic DNA Purification System (Promega, Manheim Germany) and from the rententate obtained by filtering (0.2 µl) of 100 ml seawater with the DNeasy Blood and Tissue Kit, Qiagen, Hilden, Germany. To check for bacterial contamination of reagents, we included additional blank extractions.

PCR (25 µl, 30 cycles, 1 min annealing at 55◦C) of the 16S rRNA gene V1-V2 regions, including positive and negative (ddH2O) controls, was performed with equal concentrations of uniquely barcoded 27f and 338r PCR primers (Wang et al., 2015), using 0.5 unit of Phusion Hot Start II High-Fidelity DNA Polymerase per reaction. Equal amounts (estimated by gel electrophoresis and determined fluorometrically) of PCR

Texel) are marked by an asterisk (\*). The arrows on the map link the oyster collection sites with the deployment sites (i.e., half of the oysters were returned to their collection site while the other half was transplanted). The inserted plots show within-site deployment scheme with all sampling spots. Within each sampling spot there are two (Sylt) or three (Texel) bags with four oysters each. Each oyster in the bag belongs to one of the four treatment groups (i.e., antibiotic treatment/control × Sylt/Texel oyster origin). Note that the colors (orange = Sylt, blue = Texel) can denote site or the oyster origin, depending on the context. Tables show the total and the per treatment sample sizes for each site. The box below the map shows the sampling timeline for both sites.

products were mixed together, purified and sequenced on a Illumina MiSeq platform at the Max Planck Institute for Evolutionary Biology in Plön, Germany.

Sequence quality control and preprocessing was performed as described in Mothur MiSeq SOP (Schloss et al., 2009; Kozich et al., 2013). We defined OTUs (Operational Taxonomic Units) based on a 97% identity threshold. Based on rarefaction curves (not shown), we decided to subsample the dataset to 10,000 reads per sample (6 samples with less than 10,000 were also included, Supplementary Table S1). Due to some oyster mortality, the final dataset comprised of 713 samples in total: 10 seawater, and 703 hemolymph (166 laboratory and 537 field).

Raw demultiplexed reads are deposited at European Nucleotide Archive under the study accession number PRJEB9624.

#### Statistical Analysis

Statistical analyses were performed in R (R Core Team, 2013). The analysis of α-diversity was based on the Shannon's H index, calculated from the complete subsampled dataset (10,000 reads per sample). We first tested for differences between the seawater and oysters within a site and then between the oysters at the two sites using Asymptotic Wilcoxon Mann-Whitney Rank Sum Test (Wilcoxon RS Test). To assess the effects of oyster origin and antibiotic treatment on α-diversity, as well as its temporal dynamics, we fitted a separate linear mixed model for each site with oyster origin, antibiotic treatment, time and their interactions as fixed factors, and with oyster as a random factor. To test if bag and sampling spot influenced α-diversity, we fitted an additional model for each site excluding the laboratory samples, with oyster, bag and sampling spot as random factors. The models were fitted and tested using the packages lme4 (Bates et al., 2014), lmerTest (Kuznetsova et al., 2014), and MuMIn (Barton, 2014 ´ ).

In our previous association network analysis of hemolymph microbiota (Lokmer et al., 2016), we identified a cluster consisting of the OTUs abundant in the seawater that could be defined as transient. We analyzed all the hemolymph samples in this study in the same way using the igraph package (Csardi and Nepusz, 2006) and again found this transient OTU assemblage (Supplementary Figure S1). To examine how transient microbiota affect β-diversity and distance-decay relationship, we performed the analyses on the complete dataset and excluding the transient OTUs.

For β-diversity, we removed low abundance OTUs (<10 sequences in the sample) to reduce the dataset complexity (Gobet et al., 2010). The analysis was based on Bray-Curtis distances calculated from hellinger-transformed OTU tables. We used non-metric multidimensional scaling (NMDS) to visualize the overall variability in the dataset (including the seawater samples), and large-scale temporal variability (between sampling points) of hemolymph microbiota. We then analyzed hemolymph communities by constrained analysis of principal coordinates (CAP, Anderson and Willis, 2003), which takes into account only the variability associated with tested predictors, and by Permanova (non-parametric permutational multivariate analysis of variance, Anderson, 2001), using the functions capscale resp. adonis, both implemented in the vegan package (Oksanen et al., 2013). In order to examine how oyster origin, antibiotic treatment and distance affected the β-diversity throughout the summer, we analyzed the hemolymph communities separately at each of the four time-points: before deployment, and once in June, July and August. Although the sampling on Sylt and Texel was not simultaneous, the time difference was at most 10 days and the samples were analyzed together. Two additional sampling points on Sylt were analyzed as well. Variability explained by distance was partialled out prior to plotting CAP results in order to more clearly represent the effects of experimental treatments within sites. To explicitly identify the taxa (from phylum to genus level) contributing to the observed variability, we calculated multivariate generalized (negative binomial) mixed models (GLMs) for each date and for the whole dataset using the mvabund package (Wang et al., 2012).

In order to assess bacterial turnover at a large spatiotemporal scale, we calculated average Bray-Curtis distances between the composite communities at different sampling dates (i.e., all samples from a site at a given date were combined into a single sample) as well as the average individual dissimilarity within sampling dates. To estimate how autocorrelation within oyster individuals influenced community structure and dynamics, we compared the average Bray-Curtis distances between all the samples from the same oyster and between the corresponding number of randomly chosen samples from different oysters. To examine within-individual temporal dynamics, we calculated bacterial turnover within oysters as a proportion of OTUs shared between the initial and subsequent sampling points (Gobet et al., 2012).

The distance-decay relationship was analyzed as described in (Martiny et al., 2011). Briefly, we used 1- Bray-Curtis distance as a measure of similarity and calculated all pairwise distances between the samples from the same sampling date. We then calculated linear models for distance-decay relationship including all samples, as well as for within-spot (up to 1 m) and between-spots (tens of meters) distance ranges separately. In order to estimate how this relationship was affected by transient OTUs, we performed the analysis excluding the seawater OTUs and compared the resulting slopes to the original ones.

Temperature is an important determinant of oyster hemolymph microbiota (Lokmer and Wegner, 2015). The mean temperature experienced by the oysters throughout the experiment was estimated from the Sylt seawater temperature time-series (courtesy of Tatyana Romanova, Wadden Sea Station Sylt, Germany) and from NIOZ Jetty, Texel, Netherlands (van Aken, 2008). On Sylt, on-spot fine-scale temperature measurements were taken during sampling to compare microenvironmental and overall temperature variability.

#### RESULTS

### Hemolymph Microbiota under Laboratory and Field Conditions

During the pre-treatment, the oysters were kept in the laboratory. As laboratory conditions differ substantially from those in the oyster natural environment, we first examined their effect on the hemolymph microbiota. Whereas we found no systematic difference between α-diversity in the laboratory and in the field (**Figure 2**), NMDS ordination revealed that the laboratory conditions consistently affected hemolymph community composition at both sites, resulting in a clear separation of laboratory and field samples along the first NMDS axis (**Figure 3**). Laboratory samples were characterized by higher relative abundances of Fusobacteria, ε- and γ-Proteobacteria (mainly Arcobacter and Vibrionaceae), whereas α-Proteobacteria, Tenericutes and an unclassified bacterium related to Spirochaetes were more common in the field (Supplementary Figures S2–S4, Supplementary Table S2). In addition, laboratory communities were clearly separated from the seawater samples, which formed a small group within the cluster of field hemolymph communities (**Figure 3**). Permanova confirmed these results,

period. The dotted line represents mean daily temperature at the sites. Legend explanation: Sy and Tx refer to the oyster origin, A and C to antibiotic treatment and control, respectively.

showing that 13.9% of the compositional variability could be explained by sample type (hemolymph or seawater) for laboratory communities [F(1, 710) = 115.241, p = 0.001] and only 0.5% for the ones in the field [F(1, 710) = 4.503, p = 0.001]. This pattern reflects the absence of seawater OTUs in laboratory conditions (i.e., 0.2 µm filtered seawater) thus providing support for their transient character (Supplementary Figure S1), while indicating that these OTUs represent a common and important component of hemolymph microbiota under field conditions. Despite the resemblance between the seawater and hemolymph microbiota in the field, significantly higher heterogeneity of oyster-associated communities [Levene's test for homogeneity of multivariate variances, average distance to median: oyster <sup>=</sup> 0.588, seawater <sup>=</sup> 0.407, <sup>F</sup>(1, 546) <sup>=</sup> 55.9, <sup>p</sup> <sup>&</sup>lt; <sup>10</sup>−<sup>6</sup> , effect size = 0.147, **Figure 3**], implies that the hemolymph microbiota are not a simple reflection of the microbiota in the surrounding environment, but are shaped by other factors as well.

### Effects of Experimental Treatments (Translocation and Antibiotics) on Diversity, Composition, and Dynamics of Hemolymph Microbiota

To examine the temporal dynamics of α-diversity and quantify the effects of oyster origin and antibiotic treatment, we analyzed each site separately. On Sylt, a strong initial effect of antibiotic treatment persisted for at least 3 weeks (**Figure 2A**, main effect of treatment and time x treatment interaction of the Sylt model in **Table 1**). The initial sorting by antibiotic treatment was reversed toward the beginning of July, when diversity in control translocated oysters increased to match the diversity in their antibiotic-treated counterparts (**Figure 2A**, origin × treatment × time interaction of Sylt model in **Table 1**). In local Sylt oysters, the difference between the treatment and control remained visible until late July, when the diversity decreased in the antibiotictreated animals. On Texel, the initial reduction of community



diversity following antibiotic treatment was reversed over time in the field (**Figure 2B**, time × treatment interaction of the Texel model in **Table 1**). Although **Figure 2B** indicates a similar trend for Texel as observed on Sylt (i.e., the tendency of translocated oysters to group according to origin), the interactions between origin, antibiotic treatment and time were not significant in the Texel model.

To explicitly quantify the effects of antibiotics and oyster origin on hemolymph microbial community composition, we analyzed each sampling point (**Figure 1**) separately. We observed strong initial effects of oyster origin and antibiotic treatment at both locations (**Figure 4A**, **Table 2**). Despite the tendency of the samples to separate according to origin along the first CAP axis and according to antibiotic treatment along the second CAP axis, both **Figure 4A** and significant two- and threeway interactions between the main factors in the Permanova (**Table 2**) and multivariate GLMs (Supplementary Table S3) imply that the effects of our treatments depended at least partially on the initial community composition and conditions. Two weeks after deployment, the signature of oyster origin was still apparent, but it disappeared soon afterwards (**Table 2**, **Figures 4B–F**). Similarly to α-diversity, the effect of antibiotic treatment persisted for a longer time (i.e., until the end of July). However, the variability explained by oyster origin and antibiotic treatment was generally small (1–2%, **Table 2**), indicating that the hemolymph community structure was largely determined by other factors (e.g., individual and/or microenvironmental variability). As expected, the exclusion of transient OTUs had virtually no influence on the variability explained by experimental treatments, since their presence and abundance should depend on immediate environmental conditions only (**Table 2**).

#### Spatial Patterns and Dynamics of Hemolymph Microbiota across Scales

At a large scale, α-diversity was higher on Texel than on Sylt, in the hemolymph (Shannon's H median: Texel = 4.422 (N = 289), Sylt = 3.816 (N = 414); Wilcox RS Test: Z = −7.032, p < 0.001, effect size = −0.265) as well as in the seawater (Shannon's H median: Texel = 6.186 (N = 4), Sylt = 4.468 (N = 6); Wilcox RS Test: Z = −2.559, p = 0.011, effect size = −0.810). Within-site analysis revealed lower diversity in the hemolymph compared to the seawater on Texel (Wilcoxon RS Test: Z = −2.335, p = 0.020, effect size = −0.127), and no differences on Sylt (Wilcoxon RS Test: Z = −0.759, p = 0.448). Similarly, we found a positive correlation between the seawater temperature and the diversity of the hemolymph microbiota on Texel (Kendall's τ = 0.140 ± 0.032, p = 0.01), but not on Sylt (p = 0.388). These discrepancies between the two sites suggest that αdiversity may be influenced by different biotic and abiotic factors at each site. To test for fine-scale spatial influence, we included the bag and the sampling spot as random factors in the α-diversity

FIGURE 4 | Constrained analysis of principal coordinates (CAP) of β-diversity of hemolymph communities on Sylt and Texel showing the effects of oyster origin and antibiotic treatment after partialling out the effect of distance (variation explained by distance is given in the plots, "CondV"). (A) Pre-deployment, (B) June, (C) July Sylt only, (D) July, (E) August Sylt only, (F) August. The percentages in parentheses represent the variability explained by significant axes. \*\*\*p < 0.001, \*\*p < 0.01, \*p < 0.05, 'p < 0.1. Hulls enclose all samples from the same site (Sylt = orange, Texel = blue).

TABLE 2 | Permanova (adonis) showing the effects of oyster origin, antibiotic treatment and distance on hemolymph communities at monthly sampling time points during the experiment.


(Continued)

#### TABLE 2 | Continued


Samples from Sylt and Texel on comparable sampling points in June, July and August are analyzed together. Two additional sampling points on Sylt, one in July and one in August are analyzed separately.

field-only models (Supplementary Table S4). However, we found no evidence that the spatial proximity resulted in more similar diversity values (Supplementary Table S4, random effects for spot and bag).

Regarding β-diversity, a relatively high amount of variability was explained by distance in all except Sylt-only models (**Table 2** and conditional variability in **Figure 4**), indicating that this distance-related variability primarily reflected the differences between the sites. Multivariate GLMs (Supplementary Tables S5–S8) confirmed the considerable influence of site on community composition. However, the majority of significantly differing taxa were not very abundant (except for the unclassified Spirochaetesrelated bacteria; compare univariate significant scores in Supplementary Tables S5–S8 with the taxa depicted in Supplementary Figures S2–S4, where only the taxa with >0.1 relative abundance in at least one oyster group are shown). These rare taxa may thus represent transient microbiota that reflect site-specific environmental conditions, as consistently higher abundance of Cyanobacteria on Sylt or Oceanospirillaceae on Texel might suggest (Supplementary Tables S5–S7).

Despite the differences between Texel and Sylt, the average dissimilarity between the individuals within a site largely exceeded overall dissimilarity between the sites (pairwise Bray-Curtis distance - mean ± SD - between individuals within a site on the same sampling date: Sylt = 0.779 ± 0.105, Texel = 0.751 ± 0.110; between Sylt and Texel composite communities on the same date = 0.490 ± 0.047). This high between-individual variability might suggest the important role of factors such as oyster genotype, physiology, condition and microenvironmental heterogeneity.

Although the community composition varied widely at each spatial scale, we detected a negative correlation between community similarity and geographic distance (**Figure 5**, overall distance-decay slope: b = −0.024, p < 0.001). The relationship was significantly stronger over small (up to 1 m, within spot: b = −0.102, p < 0.001) and intermediate (between spots, up to 186 m: b = −0.113, p < 0.001) spatial scales. Exclusion of transient OTUs affected neither overall (**Figure 6**, b = −0.027, p < 0.001) nor the small-scale distance-decay relationship (b = −0.082, p < 0.001). On the other hand, it flattened the distancedecay slope at intermediate spatial scale (**Figure 6**, b = −0.034, p < 0.001), suggesting that the transient OTUs could indeed reflect the higher probability of adjacent spots to experience similar environmental conditions during immersion. However, the same analysis performed for each month separately revealed that the distance-decay relationship varied over time (Supplementary Figure S4). The overall results were driven by patterns observed in June and July, whereas in August both large- and intermediatescale slopes were steeper if transient OTUs were excluded. In fact, in August we found no significant distance-decay relationship at intermediate scale with the transient OTUs included (p = 0.065) or at the small scale regardless of the microbial community portion considered (complete: p = 0.167, without transient: p = 0.169).

### Temporal Dynamics of Hemolymph Microbiota: Sites and Individuals

Overall, α-diversity increased over the course of the experiment, with a clearer and more pronounced effect on Texel (**Figure 2**, **Table 1**, main effect of time in both Sylt and Texel model). Including the oyster individual as a random effect into the αdiversity models substantially increased the amount of explained

given spatial scale.

variability, as illustrated by the difference between marginal (fixed effects only) and conditional (including random factors) R 2 in the legend to **Table 1**. Thus, the factors like oyster genotype, but also historical contingency might play an important role in shaping the within-diversity of hemolymph microbiota.

Similar to spatial patterns, the analysis of large-scale temporal dynamics of β-diversity revealed relatively high degree of stability over the examined period (mean ± SD Bray-Curtis distance between composite communities on individual sampling dates: Sylt = 0.425 ± 0.060, Texel = 0.447 ± 0.068), with NMDS plot showing a slight tendency of hemolymph communities sampled in August to cluster separately from the rest at both sites (Supplementary Figure S5). However, the August samples were indistinguishable from the June and July samples down to the genus level (Supplementary Figure S3), indicating that the shift occurred at the OTU level. This overall temporal stability likely resulted from relatively stable overall environmental conditions throughout the sampling period (Supplementary Figure S6). On the other hand, temporal variability at the individual scale was substantial (mean ± SD Bray-Curtis distance = 0.818 ± 0.140), probably reflecting highly dynamic microenvironmental conditions experienced by the oysters at short timescales. Still, albeit high, the within-oyster temporal variability was smaller than the among-oyster variation (mean ± SD Bray-Curtis distance among oysters = 0.876 ± 0.105, Wilcox RS test: Z= −9.283, p < 10−<sup>6</sup> ), implying that the oyster-related factors such as genotype as well as internal community dynamics and priority effects shape the composition of the hemolymph microbiota. In addition, high but constant within-individual turnover rate between consecutive sampling times suggests non-directional fluctuations in community composition during the sampling period, i.e., repeated occurrence of at least some taxa (**Figure 7**).

#### DISCUSSION

Animal fitness is inextricably linked to the stability of its associated microbiota. An essential prerequisite for assessing the stability is a thorough understanding of factors and processes shaping the microbial community structure and dynamics. To identify these and their relative importance under natural conditions can be challenging, especially in a highly

variable environment, such as the intertidal. Here, we combined experimental manipulation with a field survey of temporal and spatial patterns in Pacific oyster hemolymph microbiota over multiple scales to improve our understanding of their dynamics in complex natural environments. Our observations revealed high small-scale spatial variability in the field. This, together with pronounced differences between microbial community composition in the laboratory and in the field, implies a quick response of the hemolymph microbiota to large shifts in environmental conditions. Although environmentallydependent acquisition/loss of transient microbiota indicate an important role of colonization by exogenous microbes, the persistent effects of translocation and antibiotic treatment together with recognizable individual temporal dynamics suggest that the community structure and dynamics are also influenced by host-related factors as well as by biotic interactions within the microbiome.

### Resident and Transient Hemolymph Microbiota under Laboratory and Natural Conditions

Although common in oyster hemolymph under field conditions, seawater OTUs were virtually absent from the hemolymphassociated communities in the laboratory, where the oysters were kept in the filtered seawater. Dependence on the environmental source population implies that these OTUs could be considered transient (Vasconcelos and Lee, 1972). As such, they are expected to strongly depend on the immediate environmental conditions, in this case immersion during the tidal cycle in the field (Lokmer et al., 2016) and thus should not (and did not) reflect the long-term effects of our experimental treatments (**Table 2**, compare the variability explained by the treatments including/excluding transient microbiota). In addition, the strong effect of transient microbiota on the distance-decay slope over microenvironmentally variable scales, i.e., between the sampling spots (**Figure 5**), provides further support for the link between the immersion and the dynamics of transient bacteria. Explicitly, whereas the abiotic conditions experienced by oysters are very similar within the sampling spots (<1 m), the factors such as immersion time and tidal currents likely differ between the spots in a distance-dependent manner, subsequently affecting the dynamics of the transient microbiota and the distance-decay slope.

Whereas our results clearly support the classification of seawater OTUs as transient, it is less clear which bacteria should be considered resident. For example, an unindentified bacterium close to Spirochaetes and a Tenericutes OTU were both abundant in the field and rare in the laboratory, indicating they might be transient. However, these bacteria have been previously found in Pacific oysters in Tasmania (Fernandez-Piquer et al., 2012), suggesting their affinity to form associations with oysters. Both Spirochaetes and Tenericutes are commonly isolated from various oyster tissues (Prieur et al., 1990; Green and Barnes, 2010; Husmann et al., 2010; King et al., 2012; Trabal et al., 2012; Wegner et al., 2013; Trabal Fernandez et al., 2014; Lokmer et al., 2016). However, Spirochaetes and Tenericutes are rare in the hemolymph in the laboratory (Lokmer and Wegner, 2015; Lokmer et al., 2016), where high abundance of Tenericutes is mainly linked to stress or even mortality (Lokmer and Wegner, 2015). Increased abundance of Tenericutes in the field observed here could thus have been due to a secondary infection of the injection site on the adductor muscle caused by repeated hemolymph sampling (Ayling et al., 2011). On the other hand, potential benefits that Tenericutes and Spirochaetes provide to their hosts would likely be nutrition-related (Prieur et al., 1990; Tanaka et al., 2004; Fraune and Zimmer, 2008) and the decrease in their abundance in the laboratory might have been linked to starvation, as oysters were not fed during the pretreatment period (Green and Barnes, 2010). Similarly, high abundance of Vibrio and Arcobacter commonly found in the laboratory (Lokmer and Wegner, 2015; Lokmer et al., 2016) could result from stationary conditions and represent a pre-disease state (Petton et al., 2015a). However, these bacteria are also commonly isolated in the field (Garnier et al., 2007; Wendling et al., 2014; Lokmer et al., 2016) and may play role in pathogen defense and acclimation (Defer et al., 2013; Lokmer and Wegner, 2015). Due to geographical (Petton et al., 2015b) and seasonal (Wegner et al., 2013; Wendling et al., 2014; Pierce et al., 2016) differences in dynamics of these potential hemolymph residents, further studies addressing largescale spatial and temporal variation of hemolymph microbiota, their function as well as the factors affecting their interactions with oysters are needed (Pruzzo et al., 2005; Aagesen et al., 2013).

### Variation of Oyster Microbiota Related to Origin and Antibiotic Treatment

Weeks-lasting effects of antibiotic treatment and oyster origin imply gradual turnover of resident microbiota and demonstrate the existence of internal community dynamics and importance of historical contingencies (Nemergut et al., 2013). Although antibiotics can have long-term negative effects on diversity in some cases (Stein et al., 2013), they increase diversity in others, probably due to the increased invasion susceptibility of the affected communities (Shea et al., 2004; Robinson et al., 2010). In relatively stable and isolated environments, such as mammalian gut, antibiotic-induced changes may induce permanent shifts resulting in alternative stable states (Stein et al., 2013). Oyster hemolymph, on the other hand, is a highly variable habitat closely connected with the external environment, and thus the establishment of such stable states in the associated microbial community is highly unlikely.

Genetic differentiation between oysters from Texel and Sylt (Moehler et al., 2011) could have contributed to the observed differences in β-diversity since oyster microbiota can assemble according to host genotype (Wegner et al., 2013). In addition, many Vibrio spp. are pathogenic and oyster populations rapidly adapt to their local Vibrio species (Rosa et al., 2012; Wendling and Wegner, 2015), suggesting that at least parts of hemolymph microbiota are affected by host genotype. However, the gradually decreasing difference between translocated and local oysters at both sites implies that the divergence in community composition on the host origin level was in the long run mainly affected by site, a common pattern found in marine sedentary animals (Burgsdorf et al., 2014; Lear et al., 2014; Campbell et al., 2015). Gradual turnover of resident bacteria following translocation has previously been demonstrated for Vibrio spp. populations in the oyster hemolymph (Wendling et al., 2014). Interestingly, the diversity of hemolymph microbial communities in control translocated oysters did not initially differ from their local counterparts, but increased during July at both sites, matching the timespan reported for the Vibrio spp. turnover (Wendling et al., 2014). This effect was only marginally significant, but it is tempting to speculate that this trend might have been linked to inability of translocated-oyster immune system to control the resident microbiota acquired from the new environment, resulting in more bacteria evading the oyster immune defenses and establishing in the hemolymph.

Close contact of the hemolymph with the external environment might dampen genotype-specific community assembly as opposed to other tissues more shielded from the environment (Wegner et al., 2013; Lokmer et al., 2016) as well as prevent the evolution of specialist hemolymph symbionts (Preheim et al., 2011). Nevertheless, the seawater and the coastal sediments are characterized by seasonally recurring bacterial populations (Gilbert et al., 2012; Gobet et al., 2012), likely resulting in predictable encounters between the oysters, their resident bacteria and external microbiota. Unpredictable disturbances to any component of that system, such as translocation or antibiotic treatment here, might influence the community structure and dynamics and subsequently affect oyster fitness (Lokmer et al., 2016).

### Hemolymph Microbiota across Temporal and Spatial Scales

At coarse temporal and spatial resolution, the hemolymph microbiota appeared relatively stable throughout the sampling period and we found little variation in dominant taxa (Supplemetary Figures S2, S3, Supplementary Tables S2, S5–S7). Previous studies have demonstrated a strong effect of temperature on the hemolymph microbiota (Lokmer and Wegner, 2015), but have also shown that the community structure in the natural conditions exhibits a seasonal pattern and does not respond to quick temperature shifts (Wendling et al., 2014). In addition, we can show that hemolymph microbiota are also affected by the immediate external microbial environment (Lokmer et al., 2016). Therefore, high large-scale spatiotemporal stability can be explained by stable mean temperature throughout the sampling period at both sites (**Figure 2**, Supplementary Figure S6) as well as by seasonally stable microbial communities in oyster surroundings, namely in the sediments and the seawater (Campbell et al., 2011; Gilbert et al., 2012; Gobet et al., 2012).

Nevertheless, we detected differences, albeit rather small and mostly constrained to less abundant phylotypes (Supplementary Tables S5–S7), between the sites in both diversity and composition. Apart from dispersal limitation, regarding primarily OTUs within the dominant lineages, these differences might have been related to environmental factors such as smaller sediment grain size at Texel, which could affect the structure of associated microbial communities (Jackson and Weeks, 2008; Legg et al., 2012) in the environment and also result in higher number of suspended particles in the seawater with consequences for the oyster filtering activity (Riisgard, 1988; Frechette et al., 2016) and microbiota.

However, it is important to remember that, although the 16S rRNA gene and its fragments represent an important tool for understanding of microbial communities, they lack resolution power. In addition, taxonomy is only partially consistent with ecology (Koeppel and Wu, 2012) and allows solely for distinction between broad habitat types (Schmidt et al., 2014). 16S rRNA defined OTUs may consist of variety of ecotypes, and, in case of host-associated bacteria, they can significantly differ in crucial traits such as virulence (Koeppel and Wu, 2013; Lemire et al., 2015; Wendling and Wegner, 2015). Moreover, closely related bacteria exhibit adaptation at very small spatial scales (Belotte et al., 2003). Therefore, although the communities at both sites and throughout the summer appear similar through the lens of 16S rDNA based taxonomy, they can actually consist of ecologically different bacteria with important consequences for hemolymph microbiota dynamics and their oyster hosts.

Interestingly, in August we observed both a slight shift in community composition as well as a change of the distancedecay relationship. It remains unclear whether there is a link between the both, but the observed changes might have been related to spawning, which occurred during August at both sites (according to observations of spat size in autumn). Spawning represents a stressful period in the oyster lifecycle, increasing the susceptibility to pathogens and affecting the composition of associated Vibrio communities (Wendling et al., 2014) and likely of other oyster-associated bacterial populations.

In contrast to large-scale stability, the withinindividual temporal variability and the between-individual variability at small spatial scale were high, likely reflecting microenvironmental spatiotemporal heterogeneity. Virtual lack of directionality in large-scale dynamics in combination with this high small-scale variability suggests that the latter could be related to extreme but periodic environmental fluctuations in the intertidal. Namely, the quick response of the hemolymph microbiota to such fluctuations (Lokmer and Wegner, 2015; Lokmer et al., 2016) may result in pronounced but predictable (cyclic) dynamics, as bacteria disappear from the hemolymph or fall below the detection limit (Caporaso et al., 2012; Shade et al., 2013, 2014), and re-colonize the oyster or increase in abundance when the conditions are right again. High but constant turnover rate (**Figure 6**) indeed suggests that bacterial populations may disappear or become very rare, but reappear at a later stage (Gobet et al., 2012). In addition to strong influence of the environment, factors such as oyster genotype, physiology and health condition are also likely to affect the structure of the hemolymph microbiome (Wegner et al., 2013; Lokmer and Wegner, 2015), accounting for the pronounced differences between the oysters and at the same time for the consistent temporal dynamics within individuals (i.e., large portion of α-diversity variability explained by individual).

Overall, our results confirm that temporal (Shade et al., 2013) and spatial (Martiny et al., 2011; Borer et al., 2013; O'Brien et al., 2016) scale strongly affect inference about community stability and dynamics (Faust et al., 2015). High perceived temporal stability of microbiota associated with the subtidal sessile marine invertebrates (Erwin et al., 2012; Bjork et al., 2013; Pita et al., 2013; Hardoim and Costa, 2014) is often related be low temporal and/or taxonomic resolution. While the analysis of composite communities reveals influences of large-scale environmental and host factors (e.g., site, season, tissue type), focusing on small-scale individual dynamics is necessary for deciphering host-microbiota interactions and thus for understanding their role for animal survival.

## CONCLUSION

Our study aimed to examine the dynamics of Pacific oyster hemolymph microbiota under natural conditions. By switching the focus between large- and small-scale temporal and spatial variation, we identified potentially important factors and processes shaping the hemolymph microbiome. High smallscale variability (within-site or within-individual) likely reflects microenvironmental heterogeneity as well as host genetic differences, with the range of variability determined by large-scale mean abiotic conditions and internal microbiome interactions.

As drivers of host-associated community dynamics are numerous and act in a scale-dependent manner, the appropriate scale for investigations depends on the questions that one aims to address. Spatially stratified sampling designs and the analysis of individual and population temporal dynamics provide useful hints for choosing the adequate resolution. In this way one can also design experiments that will more closely mimic characteristics of the natural environment crucial for dynamics and assembly of host-associated microbiota and thus contribute to elucidating their role for animal fitness.

## AUTHOR CONTRIBUTIONS

AL planned and conducted the experiment, collected and analyzed the data and wrote the manuscript. AG and DT collected the data and critically revised the manuscript. DF took part in the analysis and interpretation of the data and critically revised the manuscript, SK collected the data, JB critically revised the manuscript, KW planned the experiment and wrote the manuscript.

## FUNDING

This study was supported by the DFG (Deutsche Forschungsgemeinschaft) Emmy Noether Programme (We4641/1-3), Excellence Cluster 306 "Inflammation at Interfaces," the Netherlands Organization for Scientific Research (NWO) and the German Federal Ministry of Education and Research (BMBF, NWO-ZKO project 839.11.002) and the International Max Research School for Evolutionary Biology. All procedures were performed according to national and European law and experiments were approved by the local authorities.

#### ACKNOWLEDGMENTS

We would like to thank Mirjana Markovic, Jennifer Welsh, Anne-Karin Schuster and Cátia Carreira for their help with hemolymph sampling at the site on Texel, Silke Vollbrecht und Franziska Schade for assisting with laboratory work on Sylt, Tobias Mayr and Kaibil Escobar Wolf for cleaning the oysters

### REFERENCES


and, Katja Cloppenborg-Schmidt at the Evolutionary Genomics group at Kiel University for assistance with the preparation of MiSeq libraries, Tatyana Romanova for Sylt temperature data and WaLTER for the NIOZ Jetty data.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01367


of the Pacific oyster Crassostrea gigas. Microb. Ecol. 53, 187–196. doi: 10.1007/s00248-006-9061-9


to temporal changes in microbial diversity. MBio 5, e01371–e01314. doi: 10.1128/mBio.01371-14


Aedes albopictus and Aedes aegypti, from different geographic regions of Madagascar. FEMS Microbiol. Ecol. 75, 377–389. doi: 10.1111/j.1574- 6941.2010.01012.x

Zurel, D., Benayahu, Y., Or, A., Kovacs, A., and Gophna, U. (2011). Composition and dynamics of the gill microbiota of an invasive Indo-Pacific oyster in the eastern Mediterranean Sea. Environ. Microbiol. 13, 1467–1476. doi: 10.1111/j.1462-2920.2011.02448.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 © 2016 Lokmer, Goedknegt, Thieltges, Fiorentino, Kuenzel, Baines and Wegner. 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) or licensor 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.

# Microbiomes of Muricea californica and M. fruticosa: Comparative Analyses of Two Co-occurring Eastern Pacific Octocorals

Johanna B. Holm\* and Karla B. Heidelberg

Division of Marine Environmental Biology, Department of Biological Science, University of Southern California, Los Angeles, CA, USA

Octocorals are sources of novel but understudied microbial diversity. Conversely, scleractinian or reef-building coral microbiomes have been heavily examined in light of the threats of climate change. Muricea californica and Muricea fruticosa are two co-occurring species of gorgonian octocoral abundantly found in the kelp forests of southern California, and thus provide an excellent basis to determine if octocoral microbiomes are host specific. Using Illumina MiSeq amplicon sequencing and replicate samples, we evaluated the microbiomes collected from multiple colonies of both species of Muricea to measure both inter- and intra-colony microbiome variabilities. In addition, microbiomes from overlying sea water and nearby zoanthids (another benthic invertebrate) were also included in the analysis to evaluate whether bacterial taxa specifically associate with octocorals. This is also the first report of microbiomes from these species of Muricea. We show that microbiomes isolated from each sample type are distinct, and specifically, that octocoral species type had the greatest effect on predicting the composition of the Muricea microbiome. Bacterial taxa contributing to compositional differences include distinct strains of Mycoplasma associated with either M. californica or M. fruticosa, an abundance of Spirochaetes observed on M. californica, and a greater diversity of γ-Proteobacteria associated with M. fruticosa. Many of the bacterial taxa contributing to these differences are known for their presence in photosymbiont-containing invertebrate microbiomes.

#### Edited by: Christine Moissl-Eichinger,

Medical University Graz, Austria

#### Reviewed by:

Julie L. Meyer, University of Florida, USA Devin Coleman-Derr, USDA-ARS and University of California, Berkeley, USA Erik Cordes, Temple University, USA

\*Correspondence: Johanna B. Holm jholm@som.umaryland.edu

#### Specialty section:

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

Received: 21 February 2016 Accepted: 27 May 2016 Published: 21 June 2016

#### Citation:

Holm JB and Heidelberg KB (2016) Microbiomes of Muricea californica and M. fruticosa: Comparative Analyses of Two Co-occurring Eastern Pacific Octocorals. Front. Microbiol. 7:917. doi: 10.3389/fmicb.2016.00917 Keywords: microbiome, octocoral, zoanthid, amplicon sequencing, microbial diversity, MiSeq

### INTRODUCTION

Corals and their microbiomes (including microbial eukaryotes, bacteria, archaea, and viruses) together comprise the entire metaorganism (Bosch and McFall-Ngai, 2011), and these symbiotic associations are critical to host survival (Rosenberg et al., 2007; Bourne and Webster, 2013). Members of these microbiomes contribute to shared metabolic functions, such as nutrient acquisition, environmental sensing, and protection from disease (Rohwer et al., 2002; Reshef et al., 2006; Vega Thurber et al., 2009; Thomas et al., 2010). Despite the perceived importance, clear factors shaping a coral's microbial composition have yet to be fully discerned (Bourne and Webster, 2013).

Most coral microbiome studies have focused on reef-building scleractinian corals. Few studies have examined the diversity and composition of gorgonian-associated microbial communities (Brück et al., 2007; Webster and Bourne, 2007; Gray et al., 2011; Bourne et al., 2013; Correa et al., 2013) and fewer have examined those of temperate gorgonians (La Rivière et al., 2013; Vezzulli et al., 2013). The gorgonian microbiomes described to date were targeted using culture-based, fingerprinting, and/or clone library analyses [except Bourne et al., 2013, which uses 454 pyrosequencing]. Such strategies are capable of providing taxonomic resolution but capture only a small portion of total microbial diversity. Additionally, there are no studies to our knowledge that statistically compare gorgonian-associated prokaryotic diversity to the surrounding sea water, other benthic organisms, or co-occurring sister species, which would highlight unique niches provided by each host type from the same environment.

The previous studies described above show gorgonianassociated microbial communities from various environments, including the Great Barrier Reef, the deep-sea, and the Mediterranean Sea, were dominated by Gammaproteobacteria (i.e., Endozoicomonas) or Tenericutes (especially Mycoplasma). However, it is intriguing that similar bacterial classes dominate the observed microbiomes of these gorgonian genera located in vastly different (and, in some cases, extreme) marine environments, and raises the question of how much the environment influences microbiome composition. This highlights a need to more deeply examine the gorgonianassociated prokaryotic community with high sampling effort and biological replication, in order to more fully characterize prokaryotic diversity.

We evaluated the microbiomes of two species of Muricea, a genus of azooxanthellate (Van Oppen et al., 2005) gorgonian octocorals found throughout the tropical and temperate eastern Pacific and western Atlantic oceans. Muricea californica and M. fruticosa co-exist in the temperate kelp forests of southern California and are easily distinguished from each other by the colors of their polyps, golden-orange or white, respectively (Grigg, 1970). Their overlapping habitats, similar abundances, colony structures, and general life histories make these species of Muricea suitable for comparison. Multiple colonies of both species were sampled with biological replication to determine the mean microbiome compositions for each colony. To test the specificity of microbial associations, we attempted to maximize intra-colony and inter-colony microbiome variations by purposely sampling colonies from different depths. To further evaluate the presence of gorgonian-specific associations, we also examined the microbiomes of nearby Parazoanthus lucificum (zoanthid, suborder Macrocnemia) colonies in addition to the surrounding sea water. Azooxanthellate P. lucificum (also referred to as Savalia lucifica, Sinniger et al., 2013), named for the brilliant bioluminescence it emits, was specifically chosen as a comparative organism because it occupies similar space as Muricea colonies in both the water column and the benthos due to its life-history trait of infecting and overgrowing M. californica colonies (Cutress and Pequegnat, 1960). To our knowledge, only one other zoanthid (suborder: Brachycnemina) microbiome has been described (Sun et al., 2014). Herein, we characterize and compare deeply sequenced microbiomes of M. californica, M. fruticosa, and P. lucificum, and in addition, examine gorgonian-associated microbes via light and fluorescence microscopy from mucus and polyp tissue to better understand bacterial micro-niches.

### MATERIALS AND METHODS

### Sample Collection

All sample collections were made in accordance with CA-DFW Scientific Collecting Permit #12734, issued to J. Holm. M. californica (Mc) and M. fruticosa (Mf) samples were collected in replicate in addition to nearby samples of the zoanthid, P. lucificum (Pl). Samples were collected from a rocky wall of Santa Catalina Island, CA (33◦ 260 53.9<sup>00</sup> N, 118◦ 280 42.3<sup>00</sup> W) midday on October 14, 2013. Muricea species were distinguished using morphological characters previously described (Grigg, 1970).

For each Mc and Mf species, subsamples from three colonies from different depths were sampled in situ (range: 8–16 m depth). Collection depth is indicated in the sample's name, following the replicate branch number. Muricea colonies are <1 m wide so, the distances between branch replicates from the same colony were no more than 1 m (i.e., colonies were no more than 1 m wide). Colonies from similar depths were >1 m but no more than 3–4 m apart. Sampling techniques to reduce contamination were employed; samples were captured in 50 mL conical tubes without handling. Additionally, single branches from two Pl colonies from depths 9 and 16 m growing adjacent to the 11–12 m gorgonian colonies, were collected. The zoanthid samples were cut and collected in a similar manner as Muricea samples. One liter of SW was collected from 12 m depth ca. 2 m away from the rocky wall. Upon returning to the boat, samples were immediately processed as follows: ethanol-wiped forceps were used to remove a collected sample from the conical tube and the sample was dipped multiple times in 0.02 µm-filtered sea water to remove unattached debris and contaminating overlying sea water. Samples were immediately placed in Ambion RNAlater as per manufacturer's instructions (Thermo Fisher Scientific, Waltham, MA, USA) and stored at 4◦C for 3 weeks until DNA extraction.

### Total DNA Extraction and PCR Amplification of 16S rRNA

Prior to extraction, samples were processed to remove RNAlater as per the manufacturer instructions. Briefly, each sample was aseptically removed from RNAlater, weighed, and a 50 mg subsample was placed in 450–500 mL 1× PBS, pH 8.0. Samples were centrifuged for 1 min at 4000 × g, and the supernatant containing residual RNAlater was carefully removed. Pellets were subsequently processed using the PowerPlant Pro DNA Isolation kit (MO BIO Laboratories, CA, USA) according to the manufacturer instructions using a TissueLyser II (Qiagen, Valencia, CA, USA) at 30 Hz for 10 min. Due to large amounts

of mucous, the Sj sample and 50 mg of each Pl sample were first ground using liquid nitrogen and a sterilized mortar and pestle prior to DNA isolation.

Bacterial and Archaeal V4–V6 regions of the 16S rRNA gene were amplified using primers A519F (CAGCMGCCGCGGTAA; Wang and Qian, 2009) and 1061R (CRRCACGAGCTGACGAC; Andersson et al., 2008) from probeBase (Loy et al., 2007). Final amplification reaction volumes were 25 µL and contained 1× Q5 High-Fidelity 2× Master Mix (New England Biolabs, Ipswich, MA, USA), 100 ng template, and 1 µM of each primer. Reactions were run with a single denaturation step at 98◦C for 30 s followed by 30 cycles at 98◦C for 30 s, 59◦C for 15 s, and 72◦C for 30 s and completed with a final extension step of 72◦C for 2 min. DNA from a previously collected, typical water sample was also amplified using this protocol and running the PCR for 35 cycles. Amplified DNA was visualized on a 1% agarose gel using SYBR Gold Nucleic Acid Stain, purified using DNA Clean & Concentrator-5 kits (Zymo Research, Irvine, CA, USA), and eluted with 6 µL nuclease-free water (Thermo Fisher Scientific, Waltham, MA, USA). Amplicons were first quantified on a Nanodrop (Thermo Fisher Scientific, Waltham, MA, USA) to ensure the 260/280 absorbance ratio was near 1.8 and then quantified again using the Quant-iT dsDNA HS Assay kit (Invitrogen, Carlsbad, CA, USA) and measured on a Qubit fluorometer (Invitrogen).

### Illumina Library Preparation and High-Throughput Sequencing

Amplicon libraries were prepared for Illumina MiSeq multiplex paired-end sequencing using NEBNext Ultra DNA Library Prep kit and the NEBNext Multiplex Oligos for Illumina Index Primers Sets 1 and 2 (New England Biolabs, Ipswich, MA, USA). Twenty nanograms of amplicon DNA were used for each library preparation reaction. Ampure Beads (Beckman Coulter, Indianapolis, IN, USA) were used for all DNA purification steps following the manufacturer instructions. Adaptor-ligated and indexed samples were visualized for purity and quantification using an Agilent 2100 Bioanalyzer (Agilent, Santa Clara, CA, USA). Twenty samples with final molar concentrations >1 nM were submitted to the UC Davis Genome Center (Davis, CA, USA) for paired-end, multiplex sequencing on the Illumina MiSeq platform using the MiSeq Reagent Kit v3 (Illumina, San Diego, CA, USA). Cluster generation, sequencing (600 cycles), image processing, demultiplexing, and quality score calculations were all performed on the MiSeq 500 platform (Illumina). Raw read data have been submitted to the NCBI Sequence Read Archive under BioSample Accession numbers SAMN03203155-SAMN03203174 within BioProject Accession number PRJNA268033. We further filtered reads for quality using the IlluminaClip (default settings), and Sliding Window (4 bases, average quality score of >25) options in Trimmomatic (Bolger et al., 2014).

Sequence assembly and quality control described here were performed using mothur (v 1.32.1-v.1.33.0; Schloss et al., 2009). The MiSeq v3 reagent kit (Illumina) produced read lengths ca. 300 bp. Paired-end reads that included the V4–V6 variable regions and overlapped in the 16S ribosomal RNA gene C45 region were processed. Sequence contigs shorter than 501 bp with >10 ambiguous bases and homopolymers >10 bases were removed. Sequences were aligned to the SILVA SSU Ref Nonredundant (NR) 119 database (Quast et al., 2013) and trimmed to equal alignment length (608 bp with gaps, 552 bp mean sequence length). Chimeric sequences were removed (39,076 sequences or 6.4%, UCHIME (Edgar et al., 2011), and the remaining sequences (606,679 unique) were taxonomically classified as described in (Wang and Qian, 2009) using the SILVA SSU Ref NR 119 formatted for mothur.

#### Sequence Binning and Phylogenetic Analyses

Sequencing and assembly generated an average of 22,685 assembled 16S rRNA gene fragments per sample (range: 11,063– 33,481). Sequences were binned into operational taxonomic units (OTUs) of 97% identity using the Average Neighbor method (Schloss and Westcott, 2011) and resulted in 5594 OTUs. Of these, 3736 were singletons and an additional 70 OTUs were characterized as chloroplasts and were culled from further analyses. Samples were randomly normalized to contain an equal number of sequences for comparative purposes (determined by the sample containing the fewest sequences, Mc\_1\_12m, 10,448 sequences). Alpha-diversity statistics including Good's Coverage Estimator and the mean number of observed OTUs were then calculated using mothur (v. 1.33.3) and compared using Student's t-test.

Remaining Bray–Curtis Dissimilarity Indices of the abundances of remaining OTUs were calculated to assess betadiversity, and visualized by hierarchical clustering (hclust) using the average neighbor method and nMDS analyses (metaMDS), with R and the Vegan package (Oksanen et al., 2013). Statistical differences in microbial community composition between different samples (gorgonian vs. non-gorgonian), species (Mc vs. Mf), colonies, and depth (8–9 m vs. 10–12 vs.16 m) were tested for using the adonis function (PERMANOVA test) with 999 permutations, also in Vegan. The OTUs contributing the most to the observed clusters were determined using the Vegan function simper, and verified using the same function on Primer-E (Clarke, 1993). OTU consensus taxonomies were obtained using the SILVA SSU Ref NR 119 database, and OTU sequence representatives were extracted in mothur. OTUs that were unclassifiable beyond Phylum or Class level were examined more closely using the NCBI GenBank NR and 16S ribosomal RNA reference (Bacteria and Archaea) databases (October 2014) and the BLASTN algorithm (Altschul et al., 1990). Reference sequences with the highest percent identity and lowest e-values were used to construct a phylogenetic tree (the number of top matches, highest percent identities, and e-values varied across OTU representatives). Maximum-likelihood phylogenetic relationships of the OTU representative sequences were assessed using ClustalW (Thompson et al., 2002) to align sequences, and maximum-likelihood trees calculated using the Jukes–Cantor model of substitution (Jukes and Cantor, 1969) with 1000 bootstrap replicates (Geneious v.5.6.4).

Relative abundances of normalized microbiomes were compared across samples bacterial classes using a Bubble Plot in MS Excel. Pie charts of the same data were composed excluding major OTUs (determined from the SIMPER analyses) for each species of Muricea to examine underlying diversity. The OTU representative sequence data have been submitted to the GenBank database under accession numbers KP174126-KP174134.

#### Microscopic Observations and Mucus Production

Endozoic cells were imaged in vivo using a BX51 epifluorescent microscope equipped with a DP70 digital camera (Olympus) using the "chlorophyll" filter set (excitation: 480 nm, emission: 660 nm).

Four colonies of Mc and two colonies of Mf were collected and maintained in tanks with unfiltered, flow through sea water, for four weeks during June 2012. Colonies were exposed to natural light conditions. All colonies were examined for the percentage of the total colony that was visibly covered with mucus. When visible, samples of mucus were collected with a sterile syringe, incubated with SYBR Green nucleic acid stain as per the manufacturer's instructions (Bio-Rad), and examined using an Olympus BX5100 epi-fluorescent microscope.

#### RESULTS

Good's coverage estimated a mean of 96.3% for all organismal samples (range: 92.3–98.3%), indicating adequate sampling effort (Supplemental Table S1). SW had the lowest coverage (83.4%) but the greatest number of observed OTUs, while all organismassociated microbiomes exhibited less diversity (**Figure 1**). Normalized Mf samples consistently had more observed OTUs than Mc (t<sup>15</sup> = 2.44, p = 0.027).

The microbiomes produced distinct sample-specific clusters whereby taxonomic assemblages from each sample were 20% similar to the other samples within the cluster (**Figure 2**). OTUs primarily contributing to the observed clusters in **Figure 2** were OTUs 1, 3, 4, and 7 according to the SIMPER analysis.

Of the tested variables, host species type had the greatest effect in determining octocoral microbiome compositions accounting for 58.2% of the observed variability (PERMANOVA F1,<sup>15</sup> = 19.5, p = 0.001). Muricea microbiomes all contained an average of 15% unclassified Bacteria compared to the SW and Pl samples where 2–10% of sequences were unclassifiable (**Figure 3**). Sequences clustered into OTU1 composed 6–64% of Mc microbiomes, compared to <0.001% of any other sampled microbiome, and were noticeably abundant in the Mc\_12m and Mc\_3\_16m microbiomes. Phylogenetic analyses indicated OTU1 was at least 96% identical to OTU7 and 85% identical to NR\_044756, Spirochaeta halophila (**Figure 4**). This observation is corroborated by our microscopic examinations of Spirochaetes in Mc mucus (**Figure 5**). OTU4 represented 6–61% of Mc microbiome sequences, vs. <0.3% in all other samples, while OTU3 composed 12–62% of Mf microbial communities, compared to <0.05% of all other sample microbiomes (**Figure 3**).

microbiome. Orange: Mc 9 m ( ), 12 m( ), 16 m( ). Purple: Mf 8 m( ), 10 m( ), 11 m( ). Blue: SW ( ), and Green: Pl (stars). Dotted circles indicate significant clusters at 20% similarity. Mc: Muricea californica, Mf: Muricea fruticosa, SW: sea water, Pl: Parazoanthus lucificum.

OTUs 3 and 4 were 87.5% identical to each other, and representative sequences formed a monophyletic group with multiple Mycoplasma sequences isolated from Muricea elongata, a sister gorgonian species found in the coastal Gulf of Mexico and

the Caribbean (NCBI PopSet: 134140623, Ranzer et al., 2007). OTU4, found in the Mc samples, formed a monophyletic group with Mycoplasma sequences isolated from healthy M. elongata colonies, while OTU3 grouped with those from bleached, diseased M. elongata colonies (**Figure 4**).

Overall, Muricea microbiomes had on average more γ-Proteobacteria sequences than α-Proteobacteria (t = 2.03, df = 80, p = 0.05; **Figure 5**), with the exception of Mf\_3\_11m, and Mf\_2\_10m, due to a large number of sequences from α-Proteobacteria OTU75. Other taxa showing Muricea spp. specificity included α-Proteobacteria such as Thalassospira (observed in 5 of 7 Mc samples, and no Mf samples), Nitratireductor (all Mc samples, no Mf sample), and γ-Proteobacteria like Endozoicomonas, Caedibacter and Francisella (observed in all Muricea samples, but higher abundances in Mc samples; Caedibacter and Francisella observed in 6 of 7 Mc samples; **Figure 5**). Also, Vibrio spp. (γ-Proteobacteria) sequences were observed in all Muricea samples, but more so in Mf samples (1–7% of Mf communities, <0.4% of Mc communities). Candidatus nitrosopumilus and Sulfuricurvum were observed in all Mf samples and no Mc samples (1–2% of mean Mf community composition). Overall, Muricea microbiomes had on average more γ-Proteobacteria sequences than α-Proteobacteria (t = 2.03, df = 80, p = 0.05; **Figure 5**), with the exception of Mf\_3\_11m, and Mf\_2\_10m, due to a large number of sequences from α-Proteobacteria OTU75. OTU75 was 98% identical to uncultured Sinorhizobium, a clonal sequence isolated from bleached M. elongata (**Figure 4**, NCBI PopSet: 134140623, Ranzer et al., 2007).

Variation between Mc colony microbiomes was significant (PERMANOVA: F1,<sup>6</sup> = 7.45, R <sup>2</sup> = 0.60, p = 0.025). Colony Mc\_9m was distinct from other Mc colonies because more Francisella (γ-Proteobacteria) sequences were observed in each of the replicate branch communities (**Figure 3**, dendrogram). As previously stated, the high number of OTU1 sequences defined the Mc\_12m microbiomes. Depth was not a significant factor in contributing to these observed differences.

Mf colony microbiomes were also significantly different from each other (PERMANOVA: F1,<sup>6</sup> = 4.33, R <sup>2</sup> = 0.38, p = 0.007). Colonies from 8 m and 11 m clustered together and separate from the Mf\_10m colony. Approximately 80% of the 8-m and 11-m replicate branch microbiomes were dominated by OTU9 (6–17% of colony replicate branch microbiomes) and previously described OTU4 (**Figure 3**). OTU9 was 94% identical to OTU1, and >80% identical to NR\_044576, S. halophila. OTU4 contributed to only 37% of Mf\_10m microbiomes; OTU9 was minimally observed. This, and the relatively high abundance of Family NB1d sequences (γ-Proteobacteria OTU54,

99% identical to AB930131, Psychrobium conchae, **Figure 4**), and Tenacibaculum (Flavobacteria) sequences produced the clustering of branch replicate samples 1 and 2 from colony Mf\_10m. Colony Mf\_8m replicate branch microbiomes were distinct from Mf\_11m because of the relatively high abundance of Sphingobacteria Saprospiracea (Bacteroidetes) and the Thaumarchaea C. nitrosopumilus (Marine Group I). Again, depth did contribute significantly in explaining this variation.

The sea water microbiome was different from all organismassociated microbiomes, dominated by α-Proteobacteria, namely Candidatus Pelagibacter (53% of microbiome) and γ-Proteobacteria SAR86 clade (12% of microbiome; **Figure 3**). These specific taxa minimally contributed (<0.13%) to all other microbiomes.

Pl microbiomes were also distinct. The relative abundance of OTU7 primarily distinguished the Pl microbiomes from all other samples (**Figure 3**; 21% of Pl\_16m, 84% in Pl\_9m, and less than 0.03% of any other sample communities), and most closely related to NR\_044756, Spirochaeta halophila (85% identical, **Figure 4**). Pl\_16m harbored a large number of Flavobacteriaceae sequences (**Figure 3**, 48% of community), specifically OTU40 (47.8% of community), which was 97% identical to NR\_116269, Maritimimonas rapanae (**Figure 4**). OTU40 contributed <0.04% to any other microbiome, including Pl\_9m.

Many features of the octocoral microbiomes were suggestive of photosymbiont-containing invertebrates including the observed strains of Mycoplasma, the presence of Endozoicomonas, and a higher γ-:α-Proteobacteria ratio in Mc samples. Thus, we examined the polyps of both species of

Muricea using transmitted light and epifluorescence microscopy to determine if photosynthesizing algae were present. Indeed, 5-15 µm brown cells were observed in abundance within Mc polyps and fluoresced red when excited with blue light, indicating that the cells contain chlorophyll (**Figure 6**). Pigmented cells were not observed in Mf polyps (data not shown).

Mc colonies produced mucus daily and, in general, mucus production appeared to be greater during daylight hours. Mucus was almost never observed on Mf colonies (**Figure 7**). Spirillumshaped bacteria were observed in mucus from Mc (**Supplemental Figure S1**).

## DISCUSSION

The microbiomes of two co-occurring and ecologically important temperate gorgonian octocoral species, M. californica and M. fruticosa, were compared to overlying sea water and nearby zoanthid associated microbiomes. We confirmed that host-associated microbial assemblages exist and are distinct from those in surrounding sea water (**Figure 3**) and other nearby benthic organisms (P. lucificum). Using inter- and intracolony replication, we also observed that the microbiomes of each species of Muricea each have specific and predictable compositions.

Our results revealed specific relationships between hosts and members of their associated microbial communities. Both species of Muricea had an abundance of sequences from the bacterial phylum, Tenericutes, specifically from the genus, Mycoplasma. This relationship has been observed in deep-sea gorgonians (Gray et al., 2011), the cold water coral, Lophelia pertusa (Kellogg et al., 2009), and a species of gorgonian from the Atlantic Ocean, Muricea elongata (Ranzer et al., 2007). The data of Ranzer et al. (2007) also showed two strains of Mycoplasma associating separately with different colonies of M. elongata, but the distinction between those colonies was healthy and bleached.

FIGURE 6 | A polyp of M. californica contains brown algae visualized with transmitted light (top) and fluorescence from blue light excitation (bottom, see Materials and Methods). Scale bar: 100 µm.

We saw a similar relationship: Mycoplasma strains associated with healthy M. elongata were also found in photosymbiontcontaining M. californica microbiomes, and Mycoplasma strains

associated with bleached M. elongata were observed in white polyp M. fruticosa microbiomes.

Mycoplasma requires secondary metabolite sterols and fatty acids (FAs) for growth (Ludwig et al., 2010). Such compounds are produced by Muricea (Popov et al., 1983; Gutiérrez et al., 2006), but the types of FA produced are highly dependent on the coral diet: FA produced by and transferred from symbiotic photosynthetic algae to the host coral are compositionally different than those produced by the coral's own biosynthesis pathways (Imbs et al., 2007). This and our observation of chlorophyll-containing cells in the polyps of M. californica leads us to hypothesize that different strains of Mycoplasma consistently associate with either M. californica or M. fruticosa because of the different FA each produce.

Another major distinction between the microbial communities of co-occurring Muricea species was the greater abundance of sequences from the phylum Spirochaetes in M. californica samples. The genera Spirochaeta are chemoheterotrophic and can thrive in a variety of environments (Ludwig et al., 2010). Spirochaetes have been observed in other cnidarian groups including the cold-water coral Lophelia (Kellogg et al., 2009), hydra Hydra attenuata (Hufnagel and Myhal, 1977), and the mucous of sea pens Pennatula phospherea and Pteroides spinosum (Porporato et al., 2013). We observed Spirochaetes-like bacteria in the mucus of M. californica (**Supplemental Figure S1**) and hypothesize that the greater relative abundance of sequences in M. californica samples may be due to the daily mucus production and sloughing performed by this species, which was not been observed in M. fruticosa.

Muricea microbiomes consisted of relatively more γ-Proteobacteria sequences than α-Proteobacteria. Recently, Bourne et al. (2013) showed that photosymbiontcontaining marine invertebrates had a greater abundance of γ-Proteobacteria. Those without photosymbionts generally maintained relatively more α-Proteobacteria. The photosymbionts observed in M. californica polyps may contribute to the abundance of γ-Proteobacteria sequences observed. The γ-Proteobacteria genera from M. californica samples were limited to Endozoicomonas, Caedibacter, and Francisella. Endozoicomonas sequences have been isolated from photosymbiont-containing corals and gorgonians (Bayer et al., 2013; Bourne et al., 2013) and metabolize the organic sulfur compound, dimethylsulfoniopropionate (DMSP), a byproduct of photosynthetic algae (Raina et al., 2009). The observed photosymbionts may be producing DMSP and therefore influencing the presence of Endozoicomonas. Interestingly, M. fruticosa had similar γ-Proteobacteria genera but in lower relative abundances. Vibrio sequences were the most abundant γ-Proteobacteria sequences found. Members of the genus Vibrio are among the known pathogens of corals (Bally and Garrabou, 2007; Mouchka et al., 2010; Vezzulli et al., 2010), and diseased corals tend to harbor greater microbial diversity

compared to healthy conspecifics (Bourne et al., 2008; Sunagawa et al., 2009). While M. fruticosa colonies may simply harbor more microbial diversity naturally, the greater OTU diversity and number of Vibrio sequences, in addition to the particular strain of Mycoplasma observed could be indicative of microbial community instability and possibly a diseased state.

The sequences of nitrifying microbes found uniquely amongst the host-associated microbiomes of the Muricea were of particular interest. M. fruticosa samples contained the Archaeon C. nitrosopumilus, a known ammonia oxidizer (Francis et al., 2007), and M. californica had an abundance of Nitratireductor, a nitrate-reducing bacteria (Labbé et al., 2004). Nitrogen cycling in corals has been observed (Shashar et al., 1994), and a model of the role Nitrosopumilus may play in ammonia oxidation has been presented (Siboni et al., 2008). Nitrogenous waste (dissolved inorganic nitrogen, DIN) from coral colonies is either released into the surrounding sea water, in the case of azooxanthellate corals, or transferred to photosymbionts, if present (Dubinsky and Jokiel, 1994). In fact, the density, chlorophyll a content, and rate of photosynthesis of coral photosymbionts is tightly coupled to the availability of DIN (Hoegh-Guldberg and Smith, 1989). Thus, Muricea metabolic waste products may offer a distinct niche for the observed nitrifying microbes, and the presence of the chlorophyll-containing cells may impact this diversity by altering the amounts and types of DIN released.

Interestingly, the gorgonian samples both contained a high number of novel sequences, those that were unclassifiable beyond the domain level, because there were no reference sequences greater than 80% identical. This proportion of sequences implies that these corals may be important sources of novel biological diversity. Gorgonians have been targeted as promising sources of novel secondary metabolites or marine natural products (MNPs; Rocha et al., 2011). But in most cases, an associated microbe, and not the invertebrate host, is producing the compound(s) of interest (Rath et al., 2011) and our study offers novel information about the microbial organisms associated with gorgonians that may be relevant targets for bio-discovery.

Gorgonian microbiomes, in general, are extremely understudied, especially compared to their scleractinian relatives. While a 16S rRNA gene diversity study is not definitive of functional or metabolic interactions in a metaorganism, a comparative approach to microbial diversity, as we have

#### REFERENCES


described here, can highlight and substantiate significant differences between the available niches provided by various host organisms.

### AUTHOR CONTRIBUTIONS

JH and KH designed the experiment. JH collected, prepared, and analyzed the samples. JH produced the figures and wrote the manuscript. KH advised on the figure production as well as the manuscript preparation.

#### FUNDING

The USC-Wrigley Rose Hills Summer Graduate Research Fellowship supported this project.

#### ACKNOWLEDGMENTS

We are grateful to Kellie Spafford and Lorraine Sadler and Wrigley Marine Science Center Staff for assistance in sample collections. Undergraduate summer fellows, Nicole McNabb and Samantha Wright Leigh, were integral to gorgonian data collection. They thank David Caron, Benjamin Tully, Rohan Sachdeva, Jay Liu, Jacob Cram, Ella Sieradzki, and three anonymous reviewers for their invaluable advice on sequence analysis and manuscript review. Support was provided by a USC Wrigley Institute for Environmental Studies Rose Hills Summer Fellowship. These data contributed to a dissertation thesis (Holm, 2015).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.00917

FIGURE S1 | Muricea californica mucous contains Spirochaetes-like cells. Mucus 30 from M. californica stained with SYBR Gold and examined using fluorescence microscopy.

Mediterranean gorgonian coral Eunicella cavolini. Mar. Ecol. Prog. Ser. 479, 75–84. doi: 10.3354/meps10197




anthropogenic influence and disease outbreaks. PLoS ONE 8:e66745. doi: 10.1371/journal.pone.0067745


**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 © 2016 Holm and Heidelberg. 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) or licensor 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.

# Advances in Research on Epichloë endophytes in Chinese Native Grasses

Hui Song, Zhibiao Nan\*, Qiuyan Song, Chao Xia, Xiuzhang Li, Xiang Yao, Wenbo Xu, Yu Kuang, Pei Tian and Qingping Zhang

State Key Laboratory of Grassland Agro-Ecosystems, College of Pastoral Agriculture Science and Technology, Lanzhou University, Lanzhou, China

Epichloë fungal endophytes are broadly found in cool-season grasses. The symbiosis between these grasses and Epichloë may improve the abiotic and biotic resistance of the grass plant, but some Epichloë species produce alkaloids that are toxic for livestock. Therefore, it is important to understand the characteristics of the grass-Epichloë s symbiosis so that the beneficial aspects can be preserved and the toxic effects to livestock can be avoided. Since the 1990s, Chinese researchers have conducted a series of studies on grass-Epichloë symbiosis. In this review, we describe the current state of Epichloë endophyte research in Chinese native grasses. We found that more than 77 species of native grasses in China are associated with Epichloë endophytes. In addition, we review the effects of various Epichloë species on native grass responses to abiotic and biotic stress, phylogeny, and alkaloid production. We provide an overview of the study of Epichloë species on native grasses in China and directions for future research.

#### Edited by:

Martin Grube, University of Graz, Austria

#### Reviewed by:

Elisabeth Margaretha Bik, Stanford University, USA Martina Oberhofer, University of Vienna, Austria

> \*Correspondence: Zhibiao Nan zhibiao@lzu.edu.cn

#### Specialty section:

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

Received: 27 February 2016 Accepted: 24 August 2016 Published: 07 September 2016

#### Citation:

Song H, Nan Z, Song Q, Xia C, Li X, Yao X, Xu W, Kuang Y, Tian P and Zhang Q (2016) Advances in Research on Epichloë endophytes in Chinese Native Grasses. Front. Microbiol. 7:1399. doi: 10.3389/fmicb.2016.01399 Keywords: alkaloid, Chinese native grass, Epichloë endophyte, grass-Epichloë symbiosis, grass stress response, Epichloë phylogeny

### INTRODUCTION

Fungi of the genus Epichloë (Clavicipitaceae, Ascomycota) and their asexual state (Neotyphodium) are common endophytes of cool-season grasses in the subfamily Pooideae (Leuchtmann et al., 2014). Most previous research has indicated that asexual Epichloë species (29 species) are efficiently transmitted through host seeds (vertical transmission) (White et al., 1993; Leuchtmann et al., 2014). However, some recent studies have suggested that strictly asexual Epichloë endophytes are occasionally transmitted horizontally among plants in close proximity via frequent mowing, trampling, and grazing (Cheplick and Faeth, 2009; Iannone et al., 2009; Wiewióra et al., 2015; Saikkonen et al., 2016), and by conidia from epiphytic mycelia (Tadych et al., 2007, 2012; Oberhofer and Leuchtmann, 2014) via splashing water and possibly wind. Sexual Epichloë species (12 species) are transmitted to new hosts with filamentuos ascospores in addition to condia (horizontal transmission) (Leuchtmann et al., 2014; Saikkonen et al., 2016). Leuchtmann et al. (2014) renamed the anamorphs of Neotyphodium as the asexual endophyte genus Epichloë and examined the classification of sexual and asexual Epichloë species and varieties based on β-tubulin (tubB) sequences.

Epichloë species often provide numerous benefits to their hosts, such as increased tolerance to drought (Malinowski and Belesky, 2000; Kannadan and Rudgers, 2008; Gundel et al., 2013b), disease resistance (Vignale et al., 2013), resistance to herbivory and parasitism (Bush et al., 1997; Schardl et al., 2007; Gundel et al., 2013a), and enhanced aboveground and belowground vegetative and reproductive growth (Marks et al., 1991; Clay and Holah, 1999; Yue et al., 2000; Gundel et al., 2013b; Tadych et al., 2014). Previous studies have confirmed that certain alkaloids play a crucial role in a plant's pasture persistence. For example, lolines and peramine may confer significant toxicity against insect pests (Johnson et al., 1985, 2013; Schardl et al., 2013; Philippe, 2016). However, conflicting results have been reported. When Lolium perenne was grown under conditions of extremely poor nutrient availability, Epichloë festucae var. lolii infection led to a reduced root: shoot ratio and reduced photosynthetic shoot fraction (Cheplick, 2007). Some symbiont combinations, such as Schedonorus arundinaceus with Epichloë coenophiala and Lolium perenne with E. festucae var. lolii, accumulate alkaloids that are toxic to grazing animals (Di Menna et al., 2012; Schardl et al., 2013; Philippe, 2016). On the other hand, sexual Epichloë species could result in "choke disease" in host plants, in which sexual Epichloë species produce stromata that envelop the inflorescences and upper leaf sheaths of flowering culms; this leads to a reduced number of offspring (Lembicz et al., 2010).

Various Epichloë species have been discovered in China, but have not been formally taxonomically described. There are two reasons for this lack of taxonomic data: (i) the limited number of researchers in this field and (ii) insufficient knowledge on the identification and classification of Epichloë species. To address the latter issue, Chinese researchers are establishing collaborations with international institutes. The topic of hybrid occurrence in Chinese Epichloë species is not discussed in-depth in this manuscript because few Epichloë species are confirmed to be of hybrid origin. However, known hybrid species from native grasses appear to have the same two ancestors, for two main reasons. First, researchers have only confirmed some Epichloë species crosses for the Epichloë bromicola × Epichloë typhina complex. Second, hybrid species are distributed in the same and similar natural and geographic environments. Accordingly, these hybrid species underwent the same hybridization process, but are hosted by different grasses. This topic will be discussed in future reviews when more data are available on hybrid endophytes.

We have built a long-term collaboration with Prof. Christopher L. Schardl from the University of Kentucky and Prof. German Spangenberg from the Australian Academy of Technological Sciences and Engineering. With their help, two kinds of Epichloë endophytes in drunken horse grass were confirmed. The whole genome sequencing of an Epichloë endophyte in Festuca sinensis is near completion. These studies will push Epichloë research to a new level in China. We firmly believe that the research prospects with respect to Epichloë species are bright in our country.

### THE DISTRIBUTION AND DIVERSITY OF GRASS-EPICHLOË SYMBIOSIS

More than 77 species of native grasses in China have been documented as infected with Epichloë species (Nan and Li, 2000; Li et al., 2004, 2006b, 2009, 2012b; Wang et al., 2005; Wei et al., 2006; Moon et al., 2007; Chen et al., 2009; Ji et al., 2009, 2011, 2012; Kang et al., 2009, 2011a; Zhan et al., 2009; Zhang et al., 2009, 2011a, 2013; Han et al., 2012; Zhu et al., 2013; Card et al., 2014; Leuchtmann et al., 2014). The endophytes have been found in the following grass genera: Achnatherum, Agropyron, Agrostis, Brachypodium, Bromus, Calamagrostis, Cleistogenes, Dactylis, Deschampsia, Elymus, Elytrigia, Eragrostis, Festuca, Hordeum, Koeleria, Leymus, Melica, Poa, Polypopon, Roegneria, and Stipa (**Table 1**). Among these, many species of Triticeae, Stipeae, and Poeae have been reported as infected and some new Epichloë species have been described from these tribes (Li et al., 2004, 2006b; Wei et al., 2006; Chen et al., 2009; Kang et al., 2009, 2011a; Zhu et al., 2013). To date, nine Epichloë species have been identified from Chinese native grasses (Li et al., 2009; Leuchtmann et al., 2014). Unfortunately, many isolates from Chinese native grasses have not been identified to the species level based on morphology and DNA data (**Table 1**). For example, an Epichloë endophyte was isolated from Festuca sinensis (**Figure 1**). We found that this Epichloë endophyte is likely a new species, based on phylogenetic trees constructed using many markers. However, this research is still in progress. We posit that many Epichloë species new to science could be infecting Chinese native grasses.

Most Epichloë species are asexual endophytes without external symptoms in their Chinese host grasses (Leuchtmann et al., 2014), such as E. bromicola, E. gansuensis, E. gansuensis var. inebrians, E. sibirica, and E. sinica. However, Dactylis glomerata (Li et al., 2009), Roegneria kamoji (Li et al., 2006b), and Poa pratensis ssp. pratensis (Kang et al., 2011a) can also be infected with sexual Epichloë species. Although most Chinese Epichloë endophytes are not hybrids, E. sinofestucae (from F. parvigluma) (Chen et al., 2009), E. sinica (from Roegneria spp.) (Kang et al., 2009), E. liyangensis (from P. pratensis ssp. pratensis) (Kang et al., 2011a), and E. sp. (from F. myuros) (Han et al., 2012) are hybrids of E. bromicola and E. typhina (**Table 1**). E. bromicola is abundant in its host genera Elymus, Hordeum, and Roegneria, including some of the most widely distributed grass species native to China. The hybrid species E. liyangensis, E. sinica, E. sinofestucae, and other Epichloë spp. have a common ancestor, e.g., the sexual E. bromicola from R. kamoji in China.

Interestingly, Epichloë endophytes in natural grasses are morphologically diverse, e.g., the species that infect Achnatherum sibiricum (Wei et al., 2007) and Elymus species (Song et al., 2015b). Ren et al. (2009) isolated 484 Epichloë endophytes from seven populations of A. sibiricum in Inner Mongolia, China and detected five morphotypes that also exhibited different magnitudes of inhibition of Rhizoctonia solani, Fusarium oxysporum, Curvularia lunata, Cladosporium cucumerium, and Phomopsis vexans. Researchers have also detected morphological differences along an altitudinal gradient. Epichloë isolates from populations of Elymus above 3000 m present similar morphological traits, while Epichloë populations below 3000 m are morphologically variable (Song et al., 2015b). Asexual Epichloë endophytes below 3000 m tend to grow faster on potato dextrose agar than asexual Epichloë endophytes above 3000 m (Song et al., 2015b). In addition, a phylogenetic analysis showed that Epichloë endophytes above 3000 m form a clade, but isolates from regions below 3000 m belong to several clades (Song et al., 2015b).

#### TABLE 1 | Summary of Epichloë endophytes in Chinese native grasses.


(Continued)

#### TABLE 1 | Continued


### EFFECTS OF EPICHLOË SPECIES ON ABIOTIC AND BIOTIC STRESS IN GRASSES

#### Salt Stress

Plant cells are harmed by salt stress and do not intake sodium as an essential element for their physiology. Although plants have evolved several strategies to adapt to salt stress (Zhu, 2003; Dinneny, 2015), only a few studies have confirmed that Epichloë endophytes can increase salt tolerance in a grass host (Reza Sabzalian and Mirlohi, 2010). When Hordeum brevisubulatum was infected with Epichloë (EI), the grass exhibited significantly increased N, P, and K<sup>+</sup> concentrations, which led to an increase in total biomass. The Epichloë infection also reduced Na<sup>+</sup> accumulation in the EI plants compared to Epichloë-free plants (EF) (Song et al., 2015d). Based on this work, we inferred that salt tolerance could be further increased in grass-Epichloë symbiosis, which potentially provide a valuable resource for improved salt tolerance in crops.

### Drought Stress

Compared to salt stress, crop plants are inclined to suffer from drought stress (Boyer, 1982). Studies have confirmed Epichloë endophytes play a vital role in increasing drought tolerance in EI grasses (Richardson et al., 1992; Clay and Schardl, 2002; Schardl et al., 2004). A relationship between increased drought tolerance and EI has been well documented in five EI grasses that are native to China. Under drought stress, EI Leymus chinensis had significantly more total biomass than EF L. chinensis, regardless of fertilizer levels (Ren et al., 2014). Peng et al. (2013) found that seed hydropriming treatment is an effective strategy to improve seed germination and plant growth in EI F. sinensis. Epichloë infection also increased the germination of Elymus dahuricus under different osmotic potential pressures, but germination success was variable among populations (Zhang and Nan, 2010). Several studies have shown that Epichloë infection can improve the relative fitness of grasses under drought stress (Faeth, 2002; Faeth et al., 2004; Iannone et al., 2012). Zhang and Nan (2007b) showed that EI E. dahuricus produced more biomass, more tillers, and taller plants under low water treatment, but EI had no influence on plant biomass in the high water treatment. However, in a study of EI A. sibiricum, the addition of fertilizer resulted in greater plant growth, but this advantage decreased under reduced water and/or nutrient availability (Ren et al., 2011). Moreover, Song et al. (2015e) demonstrated that asexual Epichloë endophyte infection can increase resistance to waterlogging stress in H. brevisubulatum. The effect of EI on drought tolerance seems to differ among grass species. It remains to be determined whether these effects are caused by the species of infectious Epichloë, the grass species, or other factors.

#### Other Abiotic Stress

Epichloë endophytes confer stress tolerance to native grasses in China and play a significant role in the survival of some plants in high-stress environments, such as cadmium (Cd)-contaminated soils and nutrient-depleted soils. Epichloë-infected A. inebrians (Zhang et al., 2010a,b) and E. dahuricus (Zhang et al., 2012a) had higher germination rates, more tillers, longer shoots and roots, and more biomass compared to EF plants in high Cd2<sup>+</sup> concentrations. There was no significant difference between EI and EF plants under low Cd2<sup>+</sup> concentrations, indicating that Epichloë infection was only beneficial to the growth and development of A. inebrians and E. dahuricus exposed to high Cd2<sup>+</sup> concentrations.

Studies of nutrient acquisition in EI grasses have focused on the influence of nitrogen (N), since this element is a constituent of alkaloids in infected plants and is also one of the most important limiting resources for plant growth in general (Li et al., 2012a). It has been documented that increased N availability may change the relative availability of other nutrients, such as phosphorus (P) (Van Der Wouder et al., 1994). Li et al. (2012a) found that A. sibiricum–Epichloë associations are conditional on both N and P availability, but are more conditional on N than P. Changes in N allocation increase the photosynthetic ability of EI plants and also significantly increase their biomass. In addition, the benefits of Epichloë infection decline when nutrient availability decreases (Ren et al., 2011). Epichloë infection tends to reduce overall nitrogen concentration in A. sibiricum leaves, but causes the host to allocate significantly higher fractions of N to the photosynthetic machinery (Ren et al., 2011). Thus, EI plants have higher photosynthetic N use efficiency and shoot biomass than that of EF plants when fertilizer is limited (Ren et al., 2014). Song et al. (2015d) confirmed that EI H. brevisubulatum has lower ratios of C:N, C:P, Na+:K<sup>+</sup> and a higher ratio of N:P than EF plants under salt stress. According to Jia et al. (2014), the effects of EI on A. sibiricum suggest that the A. sibiricum host genotype has a stronger influence on the response to stress than the influence of Epichloë. They found that Epichloë infection did not positively affect general growth, physiology, or nutrient content of A. sibiricum, before or after clipping.

#### Pest Resistance

The grass-Epichloë symbiosis provides the grass host protection from herbivorous insects by producing alkaloids in the form of secondary metabolites (García Parisi et al., 2014; Thom et al., 2014). Aphid populations exhibit slow growth when feeding on grass infected with Epichloë species (Hartley and Gange, 2009; Saikkonen et al., 2010). However, Börschig et al. (2014) concluded that the effect of Epichloë endophytes on herbivores is generally weak and depends on the regional environmental context. They posit that more field research is necessary to detect the relative importance of Epichloë endophytes and environmental context on biotic interactions in grasslands (Börschig et al., 2014).

To date, insect resistance has only been reported for L. chinensis–E. bromicola, A. sibiricum–E. sibirica and A. inebrians–E. gansuensis associations in China. Jia et al. (2013) concluded that L. chinensis–E. bromicola and A. sibiricum– E. sibirica symbioses could diminish the negative effects of infection by Meloidogyne incognita. The researchers used a 72-h exposure to undiluted culture filtrates of the two endophytes and found L. chinensis infected with E. bromicola had an especially strong antagonistic effect on Meloidogyne infection. Similarly, Zhang et al. (2012b) found that A. inebrians infected with E. gansuensis reduced the survival of the aphids Rhopalosiphum padi, Tetranychus cinnabarinus, Oedaleus decorus, and Messor aciculatus under laboratory and field conditions. Additionally, they demonstrate that EI had an anti-herbivore effect on a wide range of arthropod groups (Zhang et al., 2012b).

#### Pathogen Resistance

Reports that EI grasses are resistant to diseases and pathogens are limited compared to evidence that EI increases pest resistance. Epichloë endophytes negatively impact the in vitro growth of plant fungal pathogens (White and Cole, 1985; Siegel and Latch, 1991). However, Sabzalian et al. (2012) found that EI tall fescue was not more resistant to powdery mildew (Blumeria graminis) than EF tall fescue. Yue et al. (2000) demonstrated that extracts from a wide range of Epichloë endophytes exhibited various degrees of antifungal activity and the greatest antifungal activity was detected from extracts of E. festucae and E. tembladerae.

Li et al. (2007) confirmed that the fungi Bipolaris sorokiniana, Curvularia lunata, Fusarium acuminatum, and Alternaria alternate cause lesions on detached A. inebrians leaves, regardless of their status as EI or EF. When leaves were EF, the number and size of lesions caused by all pathogens were reduced compared to those on EI leaves. In addition, Xia et al. (2015) demonstrated that, in greenhouse conditions, EI reduced the ability of Blumeria graminis to colonize A. inebrians and enhanced the photosynthetic performance of host plants under pathogen stress or ameliorated host plant damage, to some degree (Xia et al., 2016). Zhou et al. (2015b) found that EI F. sinensis produced secondary metabolites that inhibited fungal pathogens, including Alternaria alternata, Bipolaris sorokiniana, Curvularia lunata, and Fusarium acuminatum. They found significant reductions in disease incidence and lesion size on EI detached leaves compared to EF leaves (Zhou et al., 2015b). Song et al. (2015f) found that E. bromicola from Elymus tangutorum exhibits antifungal activities against Alternaria alternata, Fusarium avenaceum, Bipolaris sorokiniana, and Curvularia lunata.

### MOLECULAR IDENTIFICATION OF CHINESE EPICHLOË SPECIES

In the past, taxonomic identification of Epichloë endophytes relied on morphological features, e.g., colony morphology, colony growth rate, and spore type and size. Currently, allozyme profiles and molecular methods have been applied to Epichloë research and greatly aid in identification. Recent research combines morphological features and molecular data to identify Epichloë endophytes.

Epichloë endophytes are typically analyzed using β-tubulin (tubB) (Tsai et al., 1994), translation elongation factor 1-α (tefA) (Moon et al., 2002), actin (actG) (Moon et al., 2007; Zhang et al., 2009), simple sequence repeats (SSR) (Moon et al., 1999; Schirrmann et al., 2015), amplified fragment length polymorphisms (AFLP) (Karimi et al., 2012), internal transcribed spacers of the nuclear ribosomal RNA (ITS) (Moon et al., 2000), calmodulin M (calM) (McCargo et al., 2014), and so on. The most common markers for taxon identification and determining phylogenetic relationships are tubB, tefA, and actG (Clay and Schardl, 2002). These studies have shown that asexual Epichloë endophytes evolved from sexual Epichloë species and subsequently lost the ability to sexually reproduce (Moon et al., 2000).

Although new Epichloë endophytes have been identified based on traditional morphology, this method has limitations when determining whether the Epichloë endophytes experienced hybridization events. Fortunately, DNA sequencing can help resolve this problem. To date, all putative Epichloë hybrids contain more than one copy of tubB and can be detected by allozyme analysis (Moon et al., 2004; Oberhofer and Leuchtmann, 2012; Leuchtmann et al., 2014; Iannone et al., 2015). For example, E. chisosa and E. coenophiala each have three copies (Leuchtmann et al., 2014), indicating they experienced multiple ancient hybridization events or subsequent gene duplication. Oberhofer and Leuchtmann (2012) found four new Epichloë species in Hordelymus europaeus using five enzymes; two were interspecific hybrids and the others were of nonhybrid origin.

Molecular markers can be used to identify new species and to estimate evolutionary relationships with phylogenetic trees. Molecular studies on Chinese Epichloë species have mainly been applied to identify new species. Various Epichloë species, e.g., E. stromatolonga (Li et al., 2006b; Ji et al., 2009), E. sinica (Kang et al., 2009), E. sinofestucae (Chen et al., 2009), E. liyangensis (Kang et al., 2011a), and E. sp. (Han et al., 2012), have been described and exhibit natural symbioses with R. kamoji, Calamagrostis epigeios, Roegneria spp., F. parvigluma, P. pratensis ssp. pratensis, and F. myuros. These Epichloë species are native to China and were described based on host specificity, morphology, and molecular phylogenetic evidence. Zhang et al. (2009) identified a new Epichloë endophyte, E. sibirica (A. sibiricum), and three morphotypes based on morphological and phylogenetic analyses. They found that its ancestor was probably derived from E. sibirica (Zhang et al., 2009). Zhu et al. (2013) analyzed L. chinensis and found that its Epichloë associate is E. bromicola, which was classified into three morphotypes based on morphological features and phylogenetic analyses of tubB, tefA, and actG sequences. Additionally, a molecular phylogenetic study showed that E. gansuensis var. inebrians from Chinese A. inebrians is a unique and novel non-hybrid species (Moon et al., 2007).

Although some studies have examined the evolutionary relationships among Epichloë species, few have examined the phylogeny or co-evolution of Chinese Epichloë species and hosts. In the southern hemisphere, most asexual Epichloë species are the result of hybridization events between two sexual species (e.g., E. festucae and E. typhina) from the northern hemisphere (Gentile et al., 2005). These studies have looked at the extent of Epichloë gene flow between the Northern and Southern Hemispheres based on molecular data (Moon et al., 2002). Iannone et al. (2009) studied South American Epichloë endophytes from Bromus auleticus and found that E. tembladerae was a hybrid of the Northern E. festucae and E. typhina, but the ancestral E. typhina genotype was distinguished based on tubB and tefA. Schirrmann et al. (2015) used 15 microsatellites to assess the population structure of sympatric species in the E. typhina complex and found that host specificity and maladaptation of Epichloë hybrids to host grasses may act as reproductive isolation barriers in asexual Epichloë and therefore promote their speciation.

Notably, Kang et al. (2011b) analyzed the asymptomatic symbiosis between Roegneria and E. sinica and found no relationship between phylogeny and morphology in the E. sinica isolates. They concluded that E. sinica is a species complex that resulted from multiple, independent hybridization events (Kang et al., 2011b). In a comparison of genetic diversity in Epichloë species and their host plants, Zhang et al. (2010c) found approximately 4–7-fold greater diversity among Epichloë endophytes than among host plants based on SSR markers. This indicates more gene flow of Epichloë endophytes than hosts. The authors also state that Epichloë infection might confer selective advantages to A. sibiricum under certain conditions, which could help to maintain high-EI frequencies, even when their population structure would not suggest selection for EI (Zhang et al., 2010c).

Song et al. found that Epichloë species likely originated in Eurasia, and Epichloë gene flow between the Western and Eastern hemispheres is common based on phylogenetic and network analyses (Song and Nan, 2015; Song et al., 2015a). They suspect that migratory birds or humans might have aided the dispersal of Epichloë endophytes from Eurasia to other continents (Song and Nan, 2015). Furthermore, Song et al. (2015c) analyzed Hordeumendophytes and Elymus-endophytes and found that Chinese Hordeum species likely contain two Epichloë endophyte species. One is also found in North American Elymus species and the other endophyte is found in Chinese Elymus species, indicating that Epichloë endophytes isolated from Chinese Hordeum are not host-specific. They proposed that Epichloë endophytes spread among different grass hosts by plant hybridization, and this could likely transform the hybrid offspring from EF status to EI status (Song et al., 2015c). This needs to be tested in future studies, but it would add further evidence to the hypothesis that asexual Epichloë endophytes are horizontally transmitted (Tadych et al., 2012; Wiewióra et al., 2015). Moreover, molecular phylogenetic studies based on tubB and tefA intron sequences have confirmed that E. gansuensis infected A. sibiricum and A. inebrians in China, indicating the potential of conidia cultures to mediate horizontal transmission (Li et al., 2015).

### ALKALOIDS

From an agronomic point of view, a negative aspect of the grass-Epichloë symbiosis is that some Epichloë produce ergot and indole-diterpene fungal alkaloids that are highly toxic for livestock (Clay and Schardl, 2002). Variability in the profile and level of alkaloids has allowed researchers to inoculate grass cultivars with selected Epichloë endophytes that are not toxic to livestock and still confer benefits to host plants. This has become a key strategy for breeding drought-, salt-, and pestresistant forage grasses (Gundel et al., 2013b; Johnson et al., 2013). A. inebrians is widely distributed in northern China and is commonly known as drunken horse grass because of its longrecognized toxic and narcotic effects on livestock, especially horses. Additionally, owing to the toxicity to livestock, recent research has shown that A. inebrians can protect biodiversity (Yao et al., 2015). These toxins are apparently caused by E. gansuensis (Li et al., 2004; Zhang et al., 2014). Epichloë-infected drunken horse grass contains high levels of the ergot alkaloids, ergine, and ergonovine (Miles et al., 1996; Li et al., 2006a). High alkaloid levels have also been confirmed in EI A. inebrians under salt or drought stress, with higher levels of ergonovine than ergine (Zhang et al., 2011b). Cytotoxic effects to animal muscle tissue have been described after the consumption of ergonovine and ergine (Zhang et al., 2014). The EI E. dahuricus only produces the alkaloid peramine. Production is seasonal; the concentration of peramine are highest in October and below detectable levels in June (Zhang and Nan, 2007a). Recently, Zhou et al. (2015a) evaluated the effects of temperature on ergot alkaloid production in three F. sinensis ecotypes and found that concentrations of ergine and ergonovine differed considerably in the three endophyte-infected ecotypes. They also found the ecotypes varied in their production of secondary metabolites, the bioprotective alkaloids ergine and ergonovine, in response to short-term cold stress. However, compared to recent research abroad (Schardl et al., 2012), little is known about alkaloid production in Chinese native grasses using molecular methods. We hope to increase research in this area in the future.

#### CONCLUSIONS AND PERSPECTIVES

In this review, we briefly summarized progress in Epichloë endophyte research in China in the past 25 years. We found that more than 77 species of native grasses in China were infected with Epichloë species. To date, nine Epichloë species have been identified from Chinese native grasses. Additionally, seven have

#### REFERENCES


been confirmed as new Epichloë endophytes. Epichloë species originated in Eurasia based on the high species diversity in the area (Song and Nan, 2015). Unfortunately, many isolates from Chinese native grasses have not been identified to the species level. Therefore, to apply this precious resource, Chinese research should focus on taxonomical evaluations of Epichloë species from Chinese native grasses. In addition, Chinese studies have extensively examined abiotic and biotic resistance using Epichloë endophytes. However, little is known about Epichloë evolution, functional genomics, and comparative genomics. Nevertheless, we believe that Chinese researchers will intensify their efforts in these areas in the future.

#### AUTHOR CONTRIBUTIONS

HS wrote the article. ZN served as the principal investigator, facilitated the project, and assisted in manuscript preparation. QS and CX wrote and revised the paper. XL, XY, WX, YK, PT, and QZ explored literature and modified the article.

#### ACKNOWLEDGMENTS

We apologize to colleagues whose work could not be cited owing to space limitations. We sincerely thank Dr. Leopoldo J. Iannone for his critical reviews and comments on an earlier version of this paper. This study was supported by the National Basic Research Program of China (2014CB138702), the National Natural Science Foundation of China (31502001), and the Fundamental Research Funds for the Central Universities (lzujbky-2014-76, lzujbky-2014-81).

animal and plant health, with particular emphasis on ryegrass staggers. N. Z. Vet J. 60, 315–328. doi: 10.1080/00480169.2012.697429


inebrians," in China APS, CPS and MSA Joint Meeting Abstracts (Quebec, QC), 203–204.


water than on nutrient availability. Environ. Exp. Bot. 108, 71–78. doi: 10.1016/j.envexpbot.2013.11.019


**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 © 2016 Song, Nan, Song, Xia, Li, Yao, Xu, Kuang, Tian 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) or licensor 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 Control of Lettuce Drop and Host Plant Colonization by Rhizospheric and Endophytic Streptomycetes

Xiaoyulong Chen<sup>1</sup> † , Cristina Pizzatti 1 †, Maria Bonaldi <sup>1</sup> , Marco Saracchi <sup>1</sup> , Armin Erlacher <sup>2</sup> , Andrea Kunova<sup>1</sup> , Gabriele Berg<sup>2</sup> and Paolo Cortesi <sup>1</sup> \*

<sup>1</sup> Department of Food, Environmental and Nutritional Sciences, University of Milan, Milan, Italy, <sup>2</sup> Institute of Environmental Biotechnology, Graz University of Technology, Graz, Austria

#### *Edited by:*

Michael Thomas-Poulsen, University of Copenhagen, Denmark

#### *Reviewed by:*

Akifumi Sugiyama, Kyoto University, Japan Mika Tapio Tarkka, Helmholtz Centre for Environmental Research - UFZ, Germany Alessio Mengoni, Università degli Studi di Firenze, Italy

#### *\*Correspondence:*

Paolo Cortesi paolo.cortesi@unimi.it † 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:* 08 March 2016 *Accepted:* 29 April 2016 *Published:* 20 May 2016

#### *Citation:*

Chen X, Pizzatti C, Bonaldi M, Saracchi M, Erlacher A, Kunova A, Berg G and Cortesi P (2016) Biological Control of Lettuce Drop and Host Plant Colonization by Rhizospheric and Endophytic Streptomycetes. Front. Microbiol. 7:714. doi: 10.3389/fmicb.2016.00714 Lettuce drop, caused by the soil borne pathogen Sclerotinia sclerotiorum, is one of the most common and serious diseases of lettuce worldwide. Increased concerns about the side effects of chemical pesticides have resulted in greater interest in developing biocontrol strategies against S. sclerotiorum. However, relatively little is known about the mechanisms of Streptomyces spp. as biological control agents against S. sclerotiorum on lettuce. Two Streptomyces isolates, S. exfoliatus FT05W and S. cyaneus ZEA17I, inhibit mycelial growth of Sclerotinia sclerotiorum by more than 75% in vitro. We evaluated their biocontrol activity against S. sclerotiorum in vivo, and compared them to Streptomyces lydicus WYEC 108, isolated from Actinovate®. When Streptomyces spp. (10<sup>6</sup> CFU/mL) were applied to S. sclerotiorum inoculated substrate in a growth chamber 1 week prior lettuce sowing, they significantly reduced the risk of lettuce drop disease, compared to the inoculated control. Interestingly, under field conditions, S. exfoliatus FT05W and S. cyaneus ZEA17I protected lettuce from drop by 40 and 10% respectively, whereas S. lydicus WYEC 108 did not show any protection. We further labeled S. exfoliatus FT05W and S. cyaneus ZEA17I with the enhanced GFP (EGFP) marker to investigate their rhizosphere competence and ability to colonize lettuce roots using confocal laser scanning microscopy (CLSM). The abundant colonization of young lettuce seedlings by both strains demonstrated Streptomyces' capability to interact with the host from early stages of seed germination and root development. Moreover, the two strains were detected also on 2-week-old roots, indicating their potential of long-term interactions with lettuce. Additionally, scanning electron microscopy (SEM) observations showed EGFP-S. exfoliatus FT05W endophytic colonization of lettuce root cortex tissues. Finally, we determined its viability and persistence in the rhizosphere and endorhiza up to 3 weeks by quantifying its concentration in these compartments. Based on these results we conclude that S. exfoliatus FT05W has high potential to be exploited in agriculture for managing soil borne diseases barely controlled by available plant protection products.

Keywords: biocontrol, hazard ratio, lettuce, *Sclerotinia sclerotiorum*, *Streptomyces*, rhizosphere competence, endophytes

## INTRODUCTION

The world population will continue to grow until at least 2050, and possibly increase from 7 to 11 billion people (Van Den Bergh and Rietveld, 2004). For this reason, food security has become one of the main challenges to human development, and therefore any plant pathogen causing substantial crop yield losses needs to be minimized. Drop, caused by Sclerotinia species, is globally one of the most destructive soil borne diseases of important horticultural crops. Three are the possible Sclerotinia species involved in lettuce drop, S. sclerotiorum, S. minor, and S. nivalis (Van Beneden et al., 2009). On lettuce, the pathogens can survive in the soil as sclerotia for years, or as mycelium on dead plants. Sclerotinia can infect the lettuce crown, roots, and leaves at any stage of plant development (Rabeendran et al., 2006). The hyphae arising from sclerotia penetrate lettuce directly through senescent leaves and root tissues, and can cause wilting and complete plant collapse in less than 2 days (Subbarao, 1998). In Lombardy, northern Italy, commercial lettuce cultivation is threatened by S. sclerotiorum infections (Bonaldi et al., 2014) and different strategies and methods are being used to prevent and manage lettuce drop epidemics. So far, fungicides have been extensively used, however, the adverse side effects of chemicals represent a serious threat to living organisms including human and the environment (Kohler and Triebskorn, 2013; Lamberth et al., 2013). In addition, for many plant pathogens, fungicide resistant populations have made many molecules ineffective. Therefore, there is an increasing demand for alternative and sustainable methods of disease management (Spadaro and Gullino, 2004; Ishii, 2006). An up-and-coming alternative to chemicals is the use of biological control agents (BCAs). Coniothyrium, Trichodema, Bacillus, and Pseudomonas spp. have been used for the management of numerous diseases (Walsh et al., 2001; Howell, 2003; Jacobsen et al., 2004). In comparison to these wellknown BCAs, there is only limited application of Streptomyces in agriculture, contrary to its exploitation in pharmaceutical industry.

Streptomyces are Gram-positive bacteria ubiquitously found in soil, where they significantly contribute to the turnover of organic matter. They are the largest genus of Streptomycetaceae family (order Actinomycetales), comprising more than 500 species (Labeda et al., 2012). Very few species are pathogenic to human or plants. S. scabies and S. turgidiscabies cause scab disease on tuber and taproot crops, such as potatoes, sweet potatoes, carrots or beet (Lehtonen et al., 2004; Loria et al., 2006). On the contrary, many species produce a variety of bioactive secondary metabolites and enzymes, which gives them potential in biocontrol and plant growth promotion. It has been hypothesized that high levels of antagonistic Streptomyces in naturally-occurring or induced suppressive soils significantly contribute to disease suppression (Kinkel et al., 2012). Similarly, organic soil amendments resulted in shift and increase of the density of indigenous Streptomyces populations and led to disease suppression (Cohen et al., 2005; Mazzola and Zhao, 2010). The current research, however, focused mainly on evaluating biocontrol activity of individual antagonistic Streptomyces spp.: S. globisporus JK-1 inhibited Pyricularia oryzae, reducing thus rice blast severity (Li et al., 2011); S. rochei ACTA1551 protected tomato seeds from F. oxysporum infection (Kanini et al., 2013); the metabolites of S. bikiniensis HD-087 effectively suppressed F. oxysporum and induced resistance in cucumber (Zhao et al., 2012); three endophytic Streptomyces isolates significantly promoted tomato plant growth by producing auxins and siderophores (Verma et al., 2011). Until now, only few commercial Streptomyces-based biocontrol products have been developed for the market, e.g., Mycostop <sup>R</sup> based on S. griseoviridis strain K61, or Actinovate <sup>R</sup> and Micro108 <sup>R</sup> based on S. lydicus strain WYEC 108 (Palaniyandi et al., 2013). They showed moderate protection of different plants against various pathogens (Paulitz and Belanger, 2001; Zeng et al., 2012; Tian and Zheng, 2013). Although vast array of secondary metabolites have been assumed to act in the biocontrol and plant growth promoting activity of streptomycetes (Trejo-Estrada et al., 1998; Prapagdee et al., 2008; Schrey and Tarkka, 2008; Tarkka and Hampp, 2008), only in few cases the exact mechanism was elucidated, e.g., disruption of geldanamycin production in recombinant S. melanosporofaciens strain FP-60 resulted in the loss of its activity against S. scabies (Agbessi et al., 2003), or the involvement of siderophores in rice growth promotion by Streptomyces sp. GMKU 3100 (Rungin et al., 2012). Moreover, priming by streptomycetes to activate plant defense responses through induced and/or acquired systemic resistance pathways could be an additional mechanism of action involved in disease suppression (Conn et al., 2008; Lehr et al., 2008; Kurth et al., 2014; Salla et al., 2016).

Plant roots are colonized by vast amount of microbes, some of which contribute to biological control (Whipps, 2001; Hardoim et al., 2015). The complex community of microbes produces a variety of compounds and develops interactions, including the competition between BCAs and plant pathogens (Raaijmakers et al., 2009). The rhizosphere—a layer of the soil surrounding the root surface including rhizoplane—harbors an array of microorganisms, whose composition is influenced by root exudates (Hiltner, 1904; Lugtenberg and Kamilova, 2009). Rhizosphere competence is a prerequisite for a BCA to establish beneficial relationship with the host. In fact, some rhizobacteria successfully colonizing rhizosphere protected the host from soil borne fungal pathogens (Kloepper et al., 2004; Haas and Defago, 2005; Weller, 2007). Nowadays, several genetic markers are available for the identification and quantification of microorganisms in the rhizosphere as well as in the endorhiza the plant inner root area. Among these, antibiotic resistance has been used as a marker to quantify the colonization dynamics of microbes in the plant root system (Gamalero et al., 2003; Adesina et al., 2009; Angelopoulou et al., 2014; Schreiter et al., 2014; Bonaldi et al., 2015). At the same time, fluorescent proteins, such as the green fluorescent protein (GFP), provide appropriate tool to monitor the colonization patterns of BCAs on plants. Enhanced GFP (EGFP), a modified version of GFP, has numerous silent nucleotide substitutions to maximize its expression in mammalian cells (Haas et al., 1996), and is also suitable for use in Streptomyces spp. because of a similar codon usage (Sun et al., 1999). GFP tagging was frequently used to determine colonization of host by beneficial Bacillus and Pseudomonas species (Krzyzanowska et al., 2012; Li et al., 2013; Subramanian et al., 2015). However, up to now, very few studies have addressed the plant colonization by EGFP-tagged Streptomyces spp. The strain EN 27 colonized the inner seed area of wheat at early stage of development (Coombs and Franco, 2003) and the pathogenic strain S. turgidiscabies Car8 colonized severalday-old radish seedlings (Joshi et al., 2007). For BCAs, the viability and persistence in rhizosphere and endorhiza are prerequisites for their application against soil borne pathogens. In fact, certain biocontrol rhizobacteria showed stable and longterm colonization of the root surface, as well as endophytic colonization (Compant et al., 2005; Berg, 2009). Therefore, determining the rhizosphere competence and endophytic colonization of the host by tagged Streptomyces will unravel part of the mechanisms involved in Streptomyces-mediated biocontrol. Moreover, the evidence of disease suppression by beneficial microbes in vivo encourages their development into bio-products for large-scale applications. However, the inconsistency between the biocontrol performance of BCAs in laboratory and in field occurred frequently and is considered one of the restraining factors of the biocontrol products (Velivellil et al., 2014). In addition, the application timing and method, as well as the concentration of BCAs play crucial roles in their biocontrol efficacy in vivo (Bonaterra et al., 2003; Fravel, 2005; Fernando et al., 2007; Müller and Berg, 2008).

In our previous study, two Streptomyces strains, S. exfoliatus FT05W and S. cyaneus ZEA17I, showed high in vitro inhibition of S. sclerotiorum (Bonaldi et al., 2015). The objective of this work was to evaluate their in vivo biological control activity against S. sclerotiorum on lettuce, assessing two different cell concentrations and two application timings in growth chamber, and subsequently their activity in field. Their performance in greenhouse and in field experiments was compared to S. lydicus WYEC 108, isolated from the commercial product Actinovate <sup>R</sup> . Simultaneously, we determined the colonization patterns of the EGFP-tagged Streptomyces on lettuce rhizoplane, using confocal laser scanning microscopy (CLSM) and we performed scanning electron microscopy (SEM) observations to verify the endophytic colonization of lettuce roots by EGFP- S. exfoliatus FT05W, the most promising strain. Finally, we determined the colonization dynamics by quantifying its concentration in lettuce rhizosphere and endorhiza at different times after lettuce inoculation.

### MATERIALS AND METHODS

### *Sclerotinia Sclerotiorum* Inoculum Preparation

Sclerotinia sclerotiorum strain FW598 from the Plant Pathology Laboratory fungi repository, Department of Food, Environmental and Nutritional Sciences (DeFENS), University of Milan, was grown for 3 days at 20◦C on Malt Extract Agar (MEA) medium (30 g/L Malt Extract, Difco, 15 g/L agar, Applichem). Then, ten agar-mycelium discs (6 mm diameter) were taken from the edge of an actively growing colony and transferred into a 300 mL flask containing 25 g of sterilized wheat kernels and 50 mL distilled water (Budge and Whipps, 2001). The flask was incubated for 3 weeks at 20◦C and was regularly shaken. Afterwards, the pathogen-colonized wheat kernels were blended with 100 mL of sterilized water to obtain the "S. sclerotiorum slurry". One gram of S. sclerotiorum slurry was diluted in an adequate volume of water to facilitate the distribution and added to 100 g of non-sterile Irish and Baltic peat-based growing substrate (Vigorplant, Piacenza, Italy). The inoculum density of S. sclerotiorum was estimated by plating serial dilutions on MEA medium. The plates were incubated at 20◦C for 2 days, the number of colonies was counted and the inoculum density was calculated as CFU/g of slurry.

### *Streptomyces* Biological Control of Lettuce Drop in Growth Chamber Experiment

Biological activity of the two Streptomyces strains, S. exfoliatus FT05W and S. cyaneus ZEA17I against S. sclerotiorum was first investigated in vivo in a growth chamber (24◦C, 55% relative humidity and 15 h photoperiod) using plastic pots (Sterivent, Duchefa, Italy), 10 × 10 × 10 cm, filled with 200 g of inoculated growing substrate as mentioned above. S. sclerotiorum inoculum was ca. 3 × 10<sup>4</sup> CFU/g of slurry. One mL of each Streptomyces strain spore suspensions (10<sup>4</sup> CFU/mL or 10<sup>6</sup> CFU/mL) was sprayed on the growing substrate immediately after the pathogen inoculation. Lettuce seeds, Lactuca sativa var. capitata, "Regina dei ghiacci", (Semeurop, Italy) were sterilized in 2 mL of 0.7% sodium hypochlorite (NaOCl) for 5 min and were rinsed three times with sterilized water. Thirty seeds were sown in three rows in each pot at two different times. In the experiment A, lettuce was sown on the same day of substrate inoculation with Streptomyces strains and the pathogen. In the experiment B, lettuce was sown 7 days after the inoculation of Streptomyces strains and pathogen inoculation. Streptomyces lydicus WYEC 108, isolated from commercial product Actinovate <sup>R</sup> (Natural Industries, Inc. Houston), was used as the reference strain. The pot inoculated only with S. sclerotiorum was used as the inoculated control. The pot inoculated neither with S. sclerotiorum nor Streptomyces was used as the non-inoculated control. For experiments A and B, eight trials were prepared in three replicates: (1) non-inoculated control; (2) S. sclerotiorum inoculated control; (3) S. exfoliatus FT05W-10<sup>4</sup> CFU/mL; (4) S. exfoliatus FT05W-10<sup>6</sup> CFU/mL; (5) S. cyaneus ZEA17I-10<sup>4</sup> CFU/mL; (6) S. cyaneus ZEA17I-10<sup>6</sup> CFU/mL; (7) S. lydicus WYEC 108-10<sup>4</sup> CFU/mL; (8) S. lydicus WYEC 108-10<sup>6</sup> CFU/mL. Dead plants were counted from the emergence up to 18 days for the experiment A, and up to 25 days for the experiment B. Disease incidence was calculated as the percent of dead plants over the plants germinated in the non-inoculated control.

### *Streptomyces* Biological Control of Lettuce Drop in Field Experiment

Field experiment was carried out in Travacò Siccomario (Pavia, Italy), characterized by loamy soil. Lettuce, Lactuca sativa var. capitata, "Regina dei ghiacci" was grown in polystyrene seed trays (84 cells—48 cm<sup>3</sup> each), filled with the non-sterile Irish and Baltic peat-based growing substrate described above. One seed was sown in each cell added with 0.5 mL of Streptomyces spore suspension (10<sup>4</sup> CFU/mL) uniformly distributed on the growing substrate. Each tray was first covered with a thin layer of the growing substrate and then with coarse perlite. Two weeks later, the same amount of Streptomyces spore suspension was added to each cell. Three weeks after sowing, each cell was inoculated with 1 mL of S. sclerotiorum slurry (ca. 3 × 10<sup>4</sup> CFU/g of slurry) prepared as described above. One day after the pathogen inoculation, the lettuce plants were transplanted into the field under plastic tunnel (width 1.2 m), at a density of 5.5 plants/m<sup>2</sup> . Five trials were prepared following a completely randomized block design in four replicates: (1) non-inoculated control; (2) S. sclerotiorum inoculated control; (3) S. exfoliatus FT05W; (4) S. cyaneus ZEA17I; (5) S. lydicus WYEC 108. Each trial consisted of 20 plants. Dead plants were counted at 3-week intervals, from the day the disease symptoms appeared until the end of the experiment. Disease incidence was calculated as the percent of dead plants over the number of transplanted plants.

### CLSM Observations of Lettuce Root Colonization by EGFP-*Streptomyces* Strains

A Leica TCS SPE Confocal Laser Scanning Microscope (Leica Microsystems, Mannheim, Germany) equipped with solid state lasers for excitation was used to unravel lettuce root colonization patterns by the two Streptomyces strains, EGFP- S. exfoliatus FT05W and EGFP- S. cyaneus ZEA17I. Plant colonization assays were carried out at the Institute of Environmental Biotechnology, Graz University of Technology, Austria. The lettuce seeds were sterilized and bacterized with EGFP-tagged Streptomyces as previously described (Bonaldi et al., 2015). Subsequently, nine bacterized seeds were sown in three rows in a seed tray filled with 640 g of a mixture of autoclaved quartz sand (Scherf GmbH & Co. KG, Austria) and peat soil ("Gramoflor Profi-Substrat-Topfpikier M+Ton+Fe" GBC, Kalsdorf, Austria) in 1:3 ratio (w/w), and 200 mL of sterilized tap water were added. In two trays, no seeds were planted to monitor the soil moisture (≥25%) by a moisture analyzer (MB35 Halogen, Ohaus, USA). Nine surface sterilized non-bacterized seeds were sown in seed trays prepared in the same way and were used as non-inoculated control. After sowing, the seed trays were incubated in a growth chamber (24◦C, 55% relative humidity and 14 h photoperiod). Two- and three-day-old seedlings and 2-week-old plants were used to verify the ability of EGFP-Streptomyces to colonize lettuce. At each interval, the roots of three bacterized plants, taken from a different seed tray, and one non-bacterized plant (negative control) were cleaned in sterile water and cut into 0.5 cm long sections for CLSM observation. Filter settings were adjusted to achieve the maximum signal from EGFP and low background autofluorescence of the plant tissues. The EGFP was excited with a 488 nm laser beam and the detection window was optimized for every field of view, in order to gain a better discrimination between the signal and the noise. Plant tissues were excited with a 635 nm laser beam and the autofluorescence emitted in the range 650–690 nm was recorded. The fluorescence signals from EGFP and from plant tissues were acquired sequentially. For each field of view, maximum projections of an appropriate number optical slices were acquired with a Z-step of 0.15–0.5µm ("confocal stacks") and the software Imaris 7.3 (Bitplane, Zurich, Switzerland) was used for post-processing (Erlacher et al., 2015).

### SEM Observations of *Streptomyces* Endophytic Colonization

To further verify the endophytic Streptomyces colonization of lettuce roots, we carried out SEM observations (Leo Electron Microscopy, Cambridge, UK) at DeFENS, University of Milan, Italy, using the representative strain EGFP- S. exfoliatus FT05W, whose wild-type strain showed promising biocontrol potential. Inoculated and non-inoculated control lettuce seeds were grown in sterile conditions as described for CLSM observations with minor modification. Each seed was sown individually in a closed 200 mL box containing 80 g of a mixture of autoclaved sandy substrate ("Sabbia Vagliata" Gras Calce s.p.a., Italy) and peat soil (Vigorplant, Piacenza, Italy) in 1:3 ratio (w/w), and 20 mL sterilized tap water. Root samples from 1-, 2- and 3-week-old plants were harvested from two inoculated plants and one control plant. Root fragments, 1 cm long, were cut: in the proximity of soil surface, in the middle, and at the root apex. The fragments were rapidly frozen in liquid nitrogen, broken into pieces with the aid of two forceps (cryo-fractured) in order to expose the internal tissues, and prepared for SEM observations (Sardi et al., 1992; Rocchi et al., 2010). In total, 22 samples from 6 inoculated plants and 11 samples from three control plants were observed.

### Colonization Dynamics in Lettuce Rhizosphere and Endorhiza

To understand the competence of EGFP- S. exfoliatus FT05W to colonize lettuce rhizosphere and endorhiza, we exploited the introduced apramycin resistance marker to quantify the amount (as colony forming units, CFU) in sterile conditions as described by Bonaldi et al. (2015) for non-sterile conditions. Briefly: lettuce plants, obtained as described above for SEM observations, were collected at 1, 2, and 3 weeks after sowing. Seedlings with the whole root system were carefully extracted from the growth substrate and the bulk soil was removed by gently shaking the plants. Excised roots were immersed in 50 mL Falcon tubes containing 8–18 mL (volume varying according to plant age) of sterilized washing solution containing 0.9% NaCl (Sigma-Aldrich, United States) and 0.02% Silwet L-77 (Chemtura Manufacturing, Italy) and vortexed two-times for 15 s. The roots were removed and kept for inner root tissue analysis. The rhizosphere suspension was filtered through a 100µm nylon mesh placed on the top of a Falcon tube, and centrifuged for 60 s to remove any remaining washing solution from the nylon mesh. The rhizosphere soil retained on the nylon mesh was collected and its dry weight was determined. The suspension was centrifuged at 10,600 g for 10 min and the pellet was resuspended in 2.5 mL of washing solution and plated in serial dilutions on Water Agar (WA) medium (15 g/L agar) added with 50 mg/L apramycin, 50 mg/L cycloheximide, and 50 mg/L nystatin. The plates were incubated at 24◦C for 7 days. Streptomyces colonies were counted and the concentration was expressed as CFU/g of rhizosphere dry weight. For inner root tissues analysis, the roots were surface sterilized with propylene oxide for 1 h. Afterwards, they were washed in 2–3 mL of washing solution, depending on plant age, and 0.5 mL of the total volume of washing solution was plated on WA medium to verify the absence of contaminants. Subsequently, the roots were finely homogenized in the washing solution, left to macerate for 1 h and plated in serial dilutions on WA medium. The Streptomyces concentration was determined as described above and expressed as CFU/g of roots dry weight.

#### Statistical Analyses

All analyses were done using R software, version R3.0.2. (R\_Core\_Team, 2013). The data of the in vivo biological control experiments, concerning the activity of Streptomyces strains against S. sclerotiorum, were submitted to survival analysis by the survival package (Therneau, 2014). First, the time-to-death of lettuce plants untreated and treated with the streptomycetes was computed using the Kaplan-Meier method. Then, the estimated survivor curve of each Streptomyces-inoculated group was compared to the inoculated control via log-rank test (P = 0.05). Finally, the effect of each strain was quantified using the Cox proportional hazard model (Kleinbaum and Klein, 2012). This model computes the hazard h at time t, as follows: h (t, X) = h0(t)e Pp i=1 <sup>β</sup>iX<sup>i</sup> where Xi are the explanatory variables and ßi are the coefficients for each variable included in the model. The effect of each treatment was quantified as Hazard Ratio (HR) expressed as HR<sup>c</sup> <sup>=</sup> exp hP<sup>p</sup> i=1 βi(<sup>X</sup> ∗ <sup>i</sup> <sup>−</sup>Xi) i where X ∗ is the covariate for one group, generally the one with the larger hazard, and X for the group with the smaller hazard. The HR values equal to 1 were interpreted as no effect of Streptomyces-treated trial over Streptomyces non-treated control, HR > 1 means that the Streptomyces non-treated plants have a higher risk of lettuce drop and HR < 1 the opposite. The rhizosphere and endorhiza colonization dynamics data were submitted to ANOVA, followed by a Tukey post hoc test for multiple comparison (P = 0.05), using the TukeyC package (Faria et al., 2013).

### RESULTS

### *Streptomyces* Biological Control of Lettuce Drop in Growth Chamber Experiment

The germination rate of lettuce, calculated from the noninoculated control, was 86.7%. When lettuce was sown the same day of the pathogen and Streptomyces inoculation (experiment A), the number of dead plants was recorded from the 4th day after sowing to the 18th day after sowing (**Supplementary Table 1**). The disease incidence of S. sclerotiorum inoculated control at the end of the experiment was 85% and none of the Streptomyces strains showed significant protection against lettuce drop according to both log-rank test and Cox model analysis (**Table 1**).

When lettuce was sown 7 days after the pathogen and Streptomyces inoculation (experiment B), the dead plants were recorded from the 4th day after sowing to the 25th day after sowing (**Supplementary Table 2**). Disease incidence of the S. sclerotiorum inoculated control was 74.4% at the end of the experiment. The Streptomyces strains showed 25.7– 51.7% protection of lettuce against S. sclerotiorum, which was statistically significant based on survival curves analyzed by logrank test, except for S. lydicus WYEC 108 applied at the lower dose (10<sup>4</sup> CFU/mL; P = 0.175). According to Cox regression model, S. exfoliatus FT05W, at both spore concentrations, significantly reduced the risk of lettuce drop disease, compared to the S. sclerotiorum inoculated control (HR = 2.078 and HR = 2.172, respectively). S. cyaneus ZEA17I was less effective than S. exfoliatus FT05W at both spore concentrations (HR = 1.595 and HR = 1.784, respectively). S. lydicus WYEC 108 applied at 10<sup>6</sup> CFU/mL reduced the most the risk of lettuce drop (HR = 2.462), whereas when applied at 10<sup>4</sup> CFU/mL, it was ineffective, which was in accordance with log-rank test analysis (HR = 1.261, P = 0.24, **Table 2**).

#### *Streptomyces* Biological Control of Lettuce Drop in Field Experiment

Under field conditions, the number of dead plants was recorded from the 10th to the 142nd day after transplanting (**Supplementary Table 3**). At the end of the experiment, drop incidence of the S. sclerotiorum inoculated control was 50.0% and treatments with S. exfoliatus FT05W and S. cyaneus ZEA17I showed respectively 40.0% and 10.0% protection against lettuce drop (**Supplementary Figure 1**). Survival curves of lettuce treated with S. exfoliatus FT05W and S. cyaneus ZEA17I were not significantly different from the S. sclerotiorum inoculated control according to the log-rank test (**Table 3**). However, the HR used to estimate the effect of S. exfoliatus FT05W was 2.178, therefore the model estimated a risk of lettuce drop about two-times lower than that of the S. sclerotiorum inoculated control (P = 0.120). The survival curve of lettuce treated with S. lydicus WYEC 108 was significantly different from the S. sclerotiorum inoculated control (P = 0.0305), but with a negative protection of 30%. The HR of S. lydicus WYEC 108 was 0.448, confirming that plants inoculated only with S. sclerotiorum had significantly lower risk of drop compared to those treated with the S. lydicus WYEC 108 (P = 0.0309, **Table 3**).

### CLSM Observations of Lettuce Root Colonization by EGFP-*Streptomyces* Strains

Filamentous growth of EGFP-Streptomyces was frequently observed on the surface of 2- and 3-day-old lettuce roots (**Figure 1**) and the mycelium of EGFP-S. cyaneus ZEA17I colonized abundantly the lettuce rhizoplane (**Figure 1A**). The colonization by EGFP- S. cyaneus ZEA17I was observed mostly in the zone of cellular maturation of the main and lateral roots, and particularly on or in the proximity of root hairs (**Figure 1B**). Moreover, germinated spores grouped together in an area close to the root hair zone (**Figure 1C**). Interestingly, a piece of soil substrate that remained attached to the lettuce root tissue showed that EGFP- S. exfoliatus FT05W colonized more extensively the lettuce root surface than the soil particle (**Figure 1D**). We also observed EGFP-Streptomyces colonization on 2-week-old lettuce roots. In general, Streptomyces at different stages of their life cycle appeared concurrently at some sites of lettuce roots. Spores, single hyphae, spore chains, and mycelium of EGFP- S. cyaneus ZEA17I were observed on the root surface (**Figure 2**). We only


TABLE 1 | Biological control of *Streptomyces* strains against lettuce drop, when *Lactuca sativa* var. *capitata*, "Regina dei ghiacci" was sown the same day of *S. sclerotiorum* and *Streptomyces* co-inoculation.

<sup>a</sup> No value.

<sup>b</sup>P-value of the log-rank test.

<sup>c</sup>β is the coefficient for the treatment covariate in the Cox model.

<sup>d</sup>Hazard Ratio (95% confidence interval).

<sup>e</sup>P-value of the Cox model.

TABLE 2 | Biological control of *Streptomyces* strains against lettuce drop, when *Lactuca sativa* var. *capitata*, "Regina dei ghiacci" was sown one week after *S. sclerotiorum* and *Streptomyces* co-inoculation.


<sup>a</sup> No value.

<sup>b</sup>P-value of the log-rank test.

<sup>c</sup>β is the coefficient for the treatment covariate in the Cox model.

<sup>d</sup>Hazard Ratio (95% confidence interval).

<sup>e</sup>P-value of the Cox model.

rarely detected colonization on the root cap and elongation zone of the roots.

### SEM Observations of Lettuce Root Endophytic Colonization by *Streptomyces* Strains

Following sample cryo-fracturation, 88 sections were obtained and observed. Mycelium of EGFP- S. exfoliatus FT05W was frequently observed on the root surface of inoculated plants (micrograph not shown). Endophytic colonization of lettuce roots by EGFP- S. exfoliatus FT05W was observed in 99% of root sections from all samples from 1- to 3-week-old roots. Generally, several cells were colonized in each section (**Figure 3**). Along the entire length of the root, both close to the collar and near the apex, single hyphae were frequently detected inside cortical cells in 1-week-old (**Figure 3A**), and 2-week-old roots (**Figure 3B**), but not inside the vascular cylinder. In a few cases, mainly in 3 week-old roots, the hyphae grew abundantly inside cortical cells forming a tangled structure (**Figure 3C**). Hyphae growing inside cortical cells had a diameter of about 0.2µm, half the size the ones grown on the root surface or in vitro cultures. EGFP- S. exfoliatus FT05W mainly colonized the endorhiza of lettuce as vegetative hyphae and rarely short spore chains were found (**Figure 3D**).

### *Streptomyces* Colonization Dynamics in Lettuce Rhizosphere and Endorhiza

EGFP- S. exfoliatus FT05W showed stable concentration up to three weeks, both in lettuce rhizosphere and endorhiza, ranging from 1.72 × 10<sup>6</sup> to 5.49 × 10<sup>6</sup> CFU/g rhizosphere dry weight and from 1.10×10<sup>5</sup> to 7.36×10<sup>6</sup> CFU/g root dry weight, respectively (**Table 4**). There were no statistically significant differences in its concentration based on plant age both in rhizosphere and in endorhiza.

#### DISCUSSION

Biological control strategies are gaining popularity in agriculture as a way to address some of the concerns about food


TABLE 3 | Biological control of *Streptomyces* strains against lettuce drop of *Lactuca sativa* var. *capitata*, "Regina dei ghiacci" under field conditions, Travacò Siccomario (Pavia, Italy).

<sup>a</sup> No value.

<sup>b</sup>P-value of the log-rank test.

<sup>c</sup>β is the coefficient for the treatment covariate in the Cox model.

<sup>d</sup>Hazard Ratio (95% confidence interval).

<sup>e</sup>P-value of the Cox model.

points to a group of germinating spores, (D) EGFP-S. exfoliatus FT05W colonizing lettuce root tissue with a soil particle attached (orange arrow). The white arrow points to the mycelium on the root surface, which is more abundant than that on the soil particle. Scale bar equals to 30µm, for Figures (A–D).

security. Studies exploring novel biocontrol microorganisms and investigating their mechanisms of action have consistently increased. However, we are still facing significant fluctuations in the efficiency of biocontrol microorganisms, which represent a critical limitation to a more general and broader use in agriculture as plant protection products. The variable performance of BCAs could be due to the limited knowledge about their mode of action and about their ability to survive and establish stable relation with plant host (Compant et al., 2005; Cuppels et al., 2013). Even when they successfully establish symbiosis with the host, e.g., in the rhizosphere, another challenge is whether the beneficial microbes can compete and suppress the pathogens. Nowadays, tremendous efforts are increasingly done to investigate the colonization patterns of BCAs on plants and their biocontrol against pathogens in vivo (Chen et al., 2013; Xue et al., 2013; Maldonado-Gonzalez et al., 2015a,b; Santiago et al., 2015).

In this work we have studied in vivo biocontrol activity of two promising Streptomyces strains against S. sclerotiorum as a follow up of a previous study, in which we obtained excellent in vitro activity against this pathogen (Bonaldi et al., 2015). S. lydicus WYEC 108, reisolated from the commercial product Actinovate <sup>R</sup> , recommended for the management of several soil borne fungal pathogens including S. sclerotiorum (Yuan and Crawford, 1995; Leisso et al., 2009; Zeng et al., 2012), was incorporated to our growth chamber and field experiments

as the reference strain. In growth chamber experiments, we tested two different timings of Streptomyces application and two different spore concentrations. Both microorganisms, the pathogen and the Streptomyces strains, were inoculated at the same time of lettuce sowing, or 7 days before the lettuce sowing. The obtained results clearly showed that the timing of Streptomyces application has a significant impact on their biocontrol activity. In particular, both S. exfoliatus FT05W and S. cyaneus ZEA17I significantly reduced S. sclerotiorum lettuce drop when they were applied 7 days before the plant sowing. In contrast, when application of S. sclerotiorum and Streptomyces was postponed until sowing, neither strain was able to reduce disease incidence and improve lettuce survival. The importance of application timing on biocontrol activity of BCAs against apple diseases was also reported for blue mold, bitter rot, and

FIGURE 2 | CLSM observations of lettuce root colonization by EGFP-*Streptomyces* two weeks after lettuce sowing. Root surface colonization by EGFP-S. cyaneus ZEA17I. Scale bar equals to 30µm.

apple scab, as well as for brown rot of cherries and plums (Teixido et al., 1999; Poleatewich et al., 2012; Rungjindamai et al., 2014). For Streptomyces, the production of biocontrol related secondary metabolites is induced or increased when the aerial hyphae appear and sporulation starts (Hopwood, 1988; Chater, 1996; Pope et al., 1996). Therefore, we hypothesize that Streptomyces need time after application to perform biocontrol activity. We also observed that the efficacy of some strains was positively correlated to the application rate, as for the reference strain S. lydicus WYEC 108, whose protection reached 51.7% when applied at 10<sup>6</sup> CFU/mL. In the field experiment, the three Streptomyces strains showed different biological control activity against S. sclerotiorum and the results were not always consistent with those observed in the greenhouse. S. exfoliatus FT05W was able to reduce drop incidence by 40%, which can be considered a promising biocontrol performance compared to other studies. For instance, application of S. padanus SS-07 resulted in 17% reduction of Rhizoctonia damping-off on Chinese cabbage, while four Streptomyces spp. strains showed 2 to 9.8% reduction of Verticillium wilt on eggplant (Chung et al., 2005; Bubici et al., 2013). On the contrary, S. lydicus WYEC 108 had a surprisingly negative effect on lettuce drop incidence in the field, opposite to the 51.7% protection obtained in the growth chamber experiment. One possible explanation could be that in the rhizosphere, the microbes respond to the many metabolites released by plant roots, as well as to the natural microflora producing a variety of compounds (Morgan et al., 2005). Such complex interactions, especially under field conditions, may result in positive, neutral, and negative effects on plant growth, health, and survival (Bouwmeester et al., 2007; Berg, 2009). Negative effect of S. lydicus WYEC 108 was previously reported for tomato bacterial spot as well as for tomato early blight in Canada (Cuppels et al., 2013). However, in the same study, a

and 3 µm (B).



<sup>a</sup>ANOVA analysis, means in a row were not significantly different (P = 0.05).

combined application of S. lydicus WYEC108 and P. fluorescens A506 resulted in a good protection against the two diseases. Similarly, S. lydicus WYEC 108 applied to control Fusarium wilt of watermelon resulted in increased disease severity in American soils, whereas the combination of green manure and S. lydicus WYEC 108 mitigated the negative effect. S. lydicus inefficacy was probably due to its lack of survival on watermelon roots in those specific conditions (Himmelstein et al., 2014). Another hypothesis might be that under certain environmental conditions S. lydicus WYEC 108 produces fungal growth promoting secondary metabolites, which enhance the pathogen growth and promote the infection of the host plant. Fungal growth promotion was shown for Streptomyces sp. strain AcH 505 producing auxofuran, a molecule which improved mycelial growth of the ectomycorrhizal fungus, Amanita muscaria, and its interaction with spruce (Schrey et al., 2005; Riedlinger et al., 2006).

The beneficial plant-microbe interactions occurring at specific sites usually require the microbe competence for host colonization (Berg et al., 2015; Hardoim et al., 2015). It has been hypothesized that the Streptomyces-mediated disease suppression is linked to the production of active secondary metabolites and their ability to colonize plant roots (Tokala et al., 2002; Franco et al., 2007). In this study, we investigated Streptomyces lettuce colonization as one of the characters underlying Streptomyces-mediated biocontrol. The use of fluorescent proteins to study plant colonization by BCAs such as Bacillus and Pseudomonas spp. has been widely reported (Buddrus-Schiemann et al., 2010; De-Bashan et al., 2010; Krzyzanowska et al., 2012; Sun et al., 2014). However, very few studies investigated Streptomyces colonization patterns on plants using fluorescent proteins in combination with CLSM. Coombs and Franco (2003) demonstrated that the EGFP-tagged endophytic Streptomyces sp. strain EN27 rapidly colonized the wheat embryo, as it was detected in developing seeds as early as 24 h after inoculation, but long-term rhizosphere competence and root colonization were not investigated. Similarly, Joshi et al. (2007) labeled a pathogenic strain of S. turgidiscabies with EGFP, and it was detected mainly on the surface of several-dayold radish seedlings, without any further monitoring. In our study, both EGFP-S. exfoliatus FT05W and EGFP-S. cyaneus ZEA17I were able to rapidly colonize the lettuce root system, and establish interactions with the host from early stages of seed germination and root development. Although it is not known if the localization of Streptomyces regulates their activity for biological control of pathogens, it has been hypothesized that endophytic bacteria form more stable interactions with plants than rhizospheric or epiphytic bacteria (Ryan et al., 2008; Compant et al., 2010; Malfanova et al., 2011). Using CLSM, we were able to detect EGFP-Streptomyces extensively colonizing the rhizoplane, and the SEM analyses confirmed the presence of EGFP-S. exfoliatus FT05W on the root surface and revealed the endophytic colonization in the root cortex. To our knowledge, this is the first study, which describes the observation of lettuce epiphytic and endophytic colonization by EGFP-tagged Streptomyces up to three weeks. In addition, we consistently recovered high concentration of EGFP-S. exfoliatus FT05W (10<sup>5</sup> -10<sup>6</sup> CFU/g dry weight) from both, lettuce rhizosphere and endorhiza, up to three weeks after seed inoculation. This evidence allows us to conclude that S. exfoliatus FT05W is both rhizospheric and endophytic in lettuce roots.

The ability of microorganisms to colonize plant roots enables them to establish long-term beneficial interactions including biocontrol against plant pathogens (Adesina et al., 2009; Schreiter et al., 2014). The ability of S. exfoliatus FT05W to produce chitinases, to solubilize phosphates and to synthesize IAA (Bonaldi et al., 2015) coupled with its stable rhizosphere competence and endophytic colonization of lettuce roots determined in this study, could explain its biocontrol activity against S. sclerotiorum. When S. exfoliatus FT05W was applied 1 week before plant sowing, it showed significant protection against lettuce drop in growth chamber, data that have been confirmed in field. Studying the colonization patterns of Streptomyces on lettuce in the presence of the pathogen will give us insight into whether and how Streptomyces spp. compete with plant pathogens, leading to better understanding of Streptomycesmediated biocontrol. In addition, studies evaluating S. exfoliatus FT05W activity against other soil borne fungal pathogens (e.g., Fusarium, Pythium, Rhizoctonia, or Verticillium spp.) and its ability to establish stable interactions with other hosts are needed to make it more attractive for its development into a commercial biocontrol product.

### AUTHOR CONTRIBUTIONS

XC performed the CLSM observations, promoted the SEM and colonization dynamics studies, and drafted the manuscript. CP evaluated the colonization dynamics data and conducted the statistical analyses. MB performed the biocontrol experiments in greenhouse and in field. MS performed SEM sample preparation, observations and acquired SEM pictures. AE contributed to the CLSM observations and post-processed the CLSM photos. AK participated to the discussions of each section of experiments, and improved the manuscript. GB hosted and supported XC in her lab to perform the CLSM observations with assistance of AE. PC designed the outline of the study and partly supported the research. All authors read and approved the final manuscript.

### ACKNOWLEDGMENTS

The authors thank Prof. Flavia Marinelli (University of Insubria, Italy), for providing the donor strain E. coli ET12567 and the reference strain S. coelicolor A3(2) and Prof. Mervyn Bibb (John Innes Centre, UK), for kindly providing the plasmid pIJ8641. The authors also thank Dr. Tomislav Cernava and Mr. Tobija Glawogger, from Graz University of Technology, Austria, for their technical assistance during the acquisition of CLSM photos, and their helpful advice and discussion regarding the experiments done in Graz. This research was supported in part by research program "Dote ricerca applicata" funded by Lombardy Region and Sipcam Italia Spa.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.00714

#### REFERENCES


Supplementary Figure 1 | Survival of lettuce plants (*Lactuca sativa* var. *capitata*, "Regina dei ghiacci") in the field experiment (ca. 60 days after transplanting) inoculated with (A) *Sclerotinia sclerotiorum* (inoculated control); (B) *S. sclerotiorum* + *S. exfoliatus* FT05W; (C) *S. sclerotiorum* + *S. cyaneus* ZEA17I; and (D) *S. sclerotiorum* + *S. lydicus* WYEC 108.

Supplementary Table S1 | Number of lettuce dead plants recorded for the experiment A, when *Lactuca sativa* var. *capitata*, "Regina dei ghiacci" was sown the same day of *S. sclerotiorum* and *Streptomyces* co-inoculation.

Supplementary Table S2 | Number of lettuce dead plants recorded for the experiment B, when *Lactuca sativa* var. *capitata*, "Regina dei ghiacci" was sown one week after *S. sclerotiorum* and *Streptomyces* co-inoculation.

Supplementary Table S3 | Number of *Lactuca sativa* var. *capitata*, "Regina dei ghiacci" dead plants recorded for field experiment, Travacò Siccomario (Pavia, Italy).


performance. Australas. Plant Pathol. 36, 524–531. doi: 10.1071/ AP07067


the family Streptomycetaceae. Antonie Van Leeuwenhoek 101, 73–104. doi: 10.1007/s10482-011-9656-0


Computing).htigung der Gründüngung und Brache. Arb. Dtsch. Landwirtsch. Ges 98.


**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 © 2016 Chen, Pizzatti, Bonaldi, Saracchi, Erlacher, Kunova, Berg and Cortesi. 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) or licensor 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.

# Biochar Treatment Resulted in a Combined Effect on Soybean Growth Promotion and a Shift in Plant Growth Promoting Rhizobacteria

Dilfuza Egamberdieva<sup>1</sup> \*, Stephan Wirth<sup>1</sup> , Undine Behrendt <sup>1</sup> , Elsayed F. Abd\_Allah<sup>2</sup> and Gabriele Berg<sup>3</sup>

1 Institute for Landscape Biogeochemistry, Leibniz Centre for Agricultural Landscape Research, Müncheberg, Germany, <sup>2</sup> Plant Production Department, Faculty of Food and Agricultural Sciences, King Saud University, Riyadh, Saudi Arabia, 3 Institute of Environmental Biotechnology, Graz University of Technology, Graz, Austria

#### *Edited by:*

Suhelen Egan, The University of New South Wales, Australia

#### *Reviewed by:*

Hovik Panosyan, Yerevan State University, Armenia Adriano Reis Lucheta, Netherlands Institute of Ecology, Netherlands

> *\*Correspondence:* Dilfuza Egamberdieva egamberdieva@yahoo.com

#### *Specialty section:*

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

*Received:* 10 December 2015 *Accepted:* 08 February 2016 *Published:* 25 February 2016

#### *Citation:*

Egamberdieva D, Wirth S, Behrendt U, Abd\_Allah EF and Berg G (2016) Biochar Treatment Resulted in a Combined Effect on Soybean Growth Promotion and a Shift in Plant Growth Promoting Rhizobacteria. Front. Microbiol. 7:209. doi: 10.3389/fmicb.2016.00209 The application of biochar to soil is considered to have the potential for long-term soil carbon sequestration, as well as for improving plant growth and suppressing soil pathogens. In our study we evaluated the effect of biochar on the plant growth of soybeans, as well as on the community composition of root-associated bacteria with plant growth promoting traits. Two types of biochar, namely, maize biochar (MBC), wood biochar (WBC), and hydrochar (HTC) were used for pot experiments to monitor plant growth. Soybean plants grown in soil amended with HTC char (2%) showed the best performance and were collected for isolation and further characterization of rootassociated bacteria for multiple plant growth promoting traits. Only HTC char amendment resulted in a statistically significant increase in the root and shoot dry weight of soybeans. Interestingly, rhizosphere isolates from HTC char amended soil showed higher diversity than the rhizosphere isolates from the control soil. In addition, a higher proportion of isolates from HTC char amended soil compared with control soil was found to express plant growth promoting properties and showed antagonistic activity against one or more phytopathogenic fungi. Our study provided evidence that improved plant growth by biochar incorporation into soil results from the combination of a direct effect that is dependent on the type of char and a microbiome shift in root-associated beneficial bacteria.

Keywords: soybean, rhizosphere, plant growth promoting rhizobacteria, biochar

## INTRODUCTION

Biochar is a fine-grained charcoal that is rich in organic carbon, produced by pyrolysis or by heating biomass in a low oxygen environment and has been used worldwide as a soil amendment to increase soil fertility (Lehmann and Joseph, 2009; Schomberg et al., 2012). However, biochar application is a very old method of improving soil quality and plant growth, as reported by the Amazonian Dark Earths (ADE) or Terra Preta de Índio formed in the past by pre-Columbian populations (Barbosa Lima et al., 2015). Presently, there are extensive literature reviews about the use of biochar and hydrochar to mitigate climate change by increasing carbon storage in soils (Lehmann et al., 2011). Other topics are about improving soil nutrient availability and the growth and development of agriculturally important crops, inducing systemic resistance in plants against soil borne fungal pathogens (Elad et al., 2010). Improvements in plant growth and yield following biochar application have been reported under field and greenhouse conditions for a variety of crops, including legumes such as soybean (Glycine max L.; Tagoe et al., 2008) and common bean (Phaseolus vulgaris; Rondon et al., 2007). Suppadit et al. (2012) reported an increased number of nodules, plant height, dry weight, yield and nutrient uptake in soybeans by quail litter biochar. Reibe et al. (2015a) observed that plant growth and development were affected by the type of char and rates of application, e.g., increasing amounts of fermented hydrochar (HTC) increased shoot biomass and the shoot/root ratio in case of spring wheat. Whereas the agricultural benefits of incorporating biochar into soils are frequently reported, there is little and incomplete evidence concerning the mechanisms of plant growth stimulation or the protection of plants from fungal pathogens by biochar. There are several studies explaining an indirect effect of biochar on root growth and development by altering soil properties, such as porosity and pore size distribution, water holding capacity, mechanical stability, sorption properties and the bioavailability of nutrients and trace elements (Laird et al., 2010; Spokas et al., 2010), but the functional response of soil microbial populations after biochar amendments are not well-understood (Lehmann et al., 2011). Anders et al. (2013) stated that the change in the structure of the microbial community by biochar application is an indirect effect and depends on soil nutrient status. Barbosa Lima et al. (2015) revealed that soil type contributes to the composition of bacterial communities in studies of forest sites (Mimosa debilis) and open areas (Senna alata) in the Amazon region. However, despite numerous reports on microbial changes induced by biochar application in various cropping systems, there have been no studies on how biochar affects the diversity and physiological activity of plant growth stimulating rhizobacteria, especially in legumes.

Most members of root-associated microbes are capable to promote plant growth and are commonly studied for their ability to stimulate plant yield, nutrient uptake, stress tolerance, and biological control of soil borne disease (Egamberdieva et al., 2008, 2011; Argaw, 2012; Berg et al., 2013a). The composition of rhizosphere bacteria is influenced not only by the plant species but also by the soil type (Berg and Smalla, 2009). The mechanisms involved in the beneficial effects conferred to plants include the production of phytohormones (Spaepen, 2015), the solubilization of insoluble phosphorus into solution available for plant use (Oteino et al., 2015), ACC deaminase enzymes, which effectively reduce plant ethylene levels in the root system (Glick, 2014), siderophores to competitively acquire ferric iron (Solanki et al., 2014), antifungal activity against a variety of plant-pathogenic fungi (Köberl et al., 2013), cell wall degrading enzymes and competition for nutrients and niches (Egamberdieva et al., 2011). Recently, a microbiome shift induced by rhizobacteria was identified as a novel mode of action for biocontrol agents (Schmidt et al., 2012; Erlacher et al., 2014).

In our study, we focused on soybean (Glycine max L.) as an important grain legume because it is a source of protein, oil, animal feed, and biodiesel in many countries worldwide, with an annual production of 276.4 Mio t<sup>1</sup> . Improved growth and production of soybeans after biochar application have been reported by Suppadit et al. (2012) and Mete et al. (2015) but mechanisms remain mostly unresolved. We hypothesized that improved growth induced by biochar amendment is strongly linked to interactions with root-associated soil microbes because biochar would promote favorable conditions for microbial proliferation in the rhizosphere. Thus, the main objectives of our study were (i) to evaluate the growth of soybean plants in response to the application of different concentrations of biochar and hydochar, and (ii) to reveal whether char incorporation into soil influences interactions between plants and root-associated microbes that are linked with plant fitness.

### MATERIALS AND METHODS

#### Plant Growth Under Greenhouse Conditions

The soil used for pot experiments was from an experimental arable field under irrigation (V4) operated by the Experimental Field Station of Leibniz Centre for Agricultural Landscape Research (ZALF), Müncheberg, Germany. The selected chemical and physical properties of soil are as follows: clay and fine silt, 7%; coarse and medium silt, 19%; sand, 74%; Corg – 570 mg 100 g −1 ; pH, 6.2; organic C content, 0.55%; total N content, 0.07%; P content, 32.0 mg (100 g soil)−<sup>1</sup> ; K content, 1.25 g (100 g soil)−<sup>1</sup> ; and Mg content, 0.18 g (100 g soil−<sup>1</sup> ).

The three types of char were supplied from the Leibniz-Institut for Agrartechnik Potsdam-Bornim e.V. (ATB) and used for pot experiments (Reibe et al., 2015a,b): (i) pyrolysis biochar from maize (MBC, 600◦C for 30 min), (ii) pyrolysis biochar from wood (WBC, 850◦C for 30 min), and (iii) hydrochar from maize silage (HTC char, processed by batch-wise hydrothermal carbonization at 210◦C and 23 bar for 8 h). The chemical composition of the chars is presented in **Table 1**.

The soil was mixed with crushed chars (particle size <3 mm) at increasing rates of 1, 2, and 3% (w/v) just before planting pre-germinated soybean seeds. All pots were arranged in a randomized block design. The soybean seeds (Glycine max. cv. Sultana, Naturland Markt, Berlin, Germany) were surface-sterilized using 10% v/v NaOCl for 5 min and 70% ethanol for 5 min, and then rinsed five times with sterile distilled water. Surface-sterilized seeds were transferred on paper tissue soaked in 0.5 mM CaSO<sup>4</sup> and germinated for 5 days in a dark room at 25◦C. The germinated seeds were transferred to pots containing 800 g of soil with four replicates. The treatments were control plants without biochar, soil amended with biochar (MBC, WBC) and HTC char at rates of 1, 2, and 3% (w/v). The plants were grown under greenhouse conditions (day/night temperature 24◦C/16◦C; humidity 50–60%; day length 12 h) and were watered when necessary. After 6 weeks, the plants were harvested, the roots were separated from shoots and the dry weight was determined.

<sup>1</sup>FAOstat (2013). Available online at: http://faostat.fao.org/ (2013-02-25).



FM, fresh matter; DM, dry matter; HTC, hydrochar; MBC, maize biochar; WBC, wood biochar; EC, electrical conductivity.

#### Isolation of Rhizosphere Bacteria

Among the biochar types under study, HTC char showed stimulatory effects on soybean plants in previous experiments and thus was used for further study. Three plants from each treatment, soil without biochar and soil amended with HTC char (2%) were collected for bacterial isolation. Excess soil was removed from the root by shaking, and only tightly adherent soil remained for study. The root samples (10 g for each treatment) were added to 100 ml of PBS buffer (PBS; 20 mM sodium phosphate, 150 mM NaCl, pH 7.0) supplemented with cycloheximide (Sigma, St. Louis, USA) at a final concentration of 100µg ml <sup>−</sup><sup>1</sup> and were shaken for 1 h. Serial dilutions (up to 10−<sup>3</sup> ) were prepared, and 100µl from appropriate dilutions was dispensed on Tryptic Soy Agar (TSA, Difco Laboratories, Detroit, USA) for bacterial culture and Peptone dextrose agar (PDA, Difco Laboratories, Detroit, USA) for fungal culture. The plates were incubated at 28◦C for 2 days, and the total numbers of bacteria and fungi were counted. The colonies of bacteria that displayed differentiable colony morphologies were picked from plates and were re-streaked on fresh agar plates for purification. One hundred bacterial cultures were selected from each treatment and maintained at 4◦C for further study.

### Plant Growth Stimulation

To test whether bacterial isolates were capable of stimulating plant growth, a pot experiment was conducted in the greenhouse using loamy sand. The seeds were surface-sterilized and inoculated with bacterial strains as described above. The sterility of the seeds was previously tested on TSA agar by incubating the plates for 3 days at 28◦C. No contaminants were found, indicating that the surface-sterilization was effective. Two hundred bacterial strains isolated from the rhizosphere of soybeans were grown overnight in Tryptic Soy Broth (TSB), and one milliliter of each culture was pelleted by centrifugation (10,000 × g for 10 min); the supernatant was discarded. Non-inoculated plants were used as negative controls. Cell pellets were washed with 1 ml of PBS and re-suspended in PBS. Cell suspensions corresponded to a cell density of 10<sup>7</sup> cells/ml. Germinated seeds were placed in the bacterial suspension with sterile forceps and shaken gently. After approximately 10 min, the inoculated seeds were aseptically planted into the potting soil. Three seeds were sown per plastic pot (12 cm diameter; 10 cm deep) to a depth of approximately 1.5 cm. After germination, plants were thinned to one per pot, and pots were set-up in a randomized design with six replications. The plants were grown under greenhouse conditions (day temperature 24◦C/night 16◦C; humidity 50–60%, day length 12 h) for 1 month. At harvest, the plants were removed from the pots, and the dry weights of roots and shoots were determined. A total 32 strains from control soil and 43 strains from HTC char amended soil were selected based on their plant growth promoting abilities and were further identified and characterized.

### Identification of Beneficial Plant Strains

The identification of isolated strains was performed using wholecell matrix assisted laser desorption/ionization (MALDI)–time of flight (TOF) mass spectrometry. Sample preparation was carried out according to the ethanol/formic acid extraction protocol recommended by Bruker Daltonics (Bremen, Germany). The isolates were cultured on tryptic soy agar (TSA, Difco Laboratories, Detroit, Michigan, USA) for 24 h, and approximately 10 mg of cell mass was suspended in 300µL of water and vortexed to generate a homogenous suspension. The suspension was mixed with 900µL of ethanol and centrifuged. The pellet was resuspended in 50µL of 70% formic acid and subsequently carefully mixed with 50µL of acetonitrile. After centrifugation, aliquots of 1µL of supernatant were placed immediately on spots of a MALDI target. Each spot was allowed to dry and subsequently overlaid with 1µL of matrix (α-ciano-4-hydroxycinnamic acid in 50% aqueous acetonitrile containing 2.5% trifluoroacetic acid). Mass spectra were acquired using a MALDI-TOF MS spectrometer in a linear positive mode (Microflex™LT, Bruker Daltonics, Bermen, Germany) in a mass range of 2–20 kDa. A bacterial test standard (BTS, Bruker Daltonics, Bremen, Germany) was used for instrument calibration. The raw spectra were imported into MALDI Biotyper™ software (Bruker Daltonics, Germany) and then processed and analyzed using standard pattern matching against the reference spectra in the MALDI Biotyper™ reference database (version 3.0, Bruker Daltonics, Germany).

### *In vitro* Screening of Bacterial Isolates for their PGP Activities

#### Indole 3-acetic Acid Production

Production of IAA (indole 3-acetic acid) was determined as described by Bano and Musarrat (2003). Briefly, bacterial strains were grown in TSB medium. After 3 days, 1 ml of each culture was pelleted by centrifugation, and the supernatant was discarded. Cell pellets were washed with 1 ml of PBS and re-suspended in PBS. One milliliter of cell suspension (corresponding to a cell density of 10<sup>7</sup> cells/ml) was added to 10 ml of TSB amended with tryptophan (100µg/ml). After 3 days of cultivation, 2 ml aliquots of bacterial cultures were centrifuged at 13.000 × g for 10 min. One milliliter of supernatant was transferred to a fresh tube to which 100µg/ml of 10 mM orthophosphoric acid and 2 ml of reagent (1 ml of 0.5 M FeCl<sup>3</sup> in 50 ml of 35% HClO4) were added. After 25 min, the absorbance of the developed pink color was read at 530 nm. The IAA concentration in culture was calculated using a calibration curve of pure IAA as a standard.

#### Phosphate solubilization

The phosphate-solubilizing activity of the bacterial strains was determined on Pikovskaya agar (Pikovskaya, 1948) containing precipitated tricalcium phosphate. The bacterial culture grown in TSA medium for 2 days was streaked on the surface of Pikovskaya agar plates and incubated for 3 days. The presence of a clearing zone around bacterial colonies was considered to be an indicator of positive P-solubilization.

#### Production of Cell Wall Degrading Enzymes

The cellulose-degrading ability of bacterial isolates was analyzed by streaking inocula on cellulose Congo-Red agar media, as described by Gupta et al. (2012). Zones of clearance around and beneath the colony were detected, indicating enzymatic degradation of cellulose. Lipase activity of the bacterial strains was determined by the Tween lipase indicator assay. Bacterial strains were grown in LA (Luria Agar) containing 2% Tween 80 at 28◦C (Howe and Ward, 1976). Protease activity was determined using 5% skimmed milk agar (Brown and Foster, 1970), and pectinase activity was determined using 0.5% pectine amended in M9 medium agar (Kumar et al., 2005).

#### HCN Production

For testing HCN production by bacterial strains, the isolates were grown in Kings' B agar medium (KB). A sterilized filter paper saturated with a 1% solution of picric acid and 2% sodium carbonate was placed in the upper lid of the Petri plate. The Petri plate was sealed with Parafilm <sup>R</sup> M and incubated at 28◦C for 3 days. The change in the paper color from yellow to dark blue was recorded as an index of HCN production (Castric, 1975).

#### In vitro Antibiosis Assay

The bacterial isolates were tested in vitro for their antagonistic activities against the pathogenic fungi Fusarium solani, F. culmorum, F. graminearum, Alternaria infectoria, and A. teniussima. The bacterial isolates were grown in TSB broth for 3 days and 50µL of bacterial culture was dropped into the hole of a PDA plate (4 mm in diameter). Fungal strains were grown in peptone dextrose agar (PDA) plates at 28◦C for 5 days, and disks of fresh fungal culture (5 mm diameter) were cut out and placed 2 cm from the hole filled with bacterial filtrate. The plates were sealed with Parafilm <sup>R</sup> M and incubated at 28◦C in darkness until the fungi had grown over the control plates without bacteria. Antifungal activity was recorded as the width of the zone of growth inhibition between the fungus and the test bacterium.

#### Statistical Analyses

Data were tested for statistical significance using the analysis of variance package included in Microsoft Excel 2007. Comparisons were performed using Student's t-test. Mean comparisons were conducted using a least significant difference (LSD) test (P = 0.05).

### RESULTS

#### Response of Soybeans to the Type and Concentration of Biochar

The response of the soybeans to the type of biochar and to different concentrations was investigated under greenhouse conditions. Our study showed that shoot and root biomass of soybeans were not significantly affected by either MBC or WBC amendments in all concentrations (1, 2, and 3%; **Figures 1A,B**). However, there was a slight but not significant increase in shoot and root growth in the soybeans grown in soil amended with MBC at 1 and 2% concentrations compared with control plants (**Figure 1A**). In contrast, the root dry weight of soybeans was significantly increased up to 34–41%, and the shoot dry weight was increased up to 24–28% by HTC char amendment at 1 and 2% concentrations, respectively (**Figure 1C**).

#### Enumeration of Microbes and Isolation of Root-Associated Bacteria

The results of the pot experiments showed that HTC char at a concentration of 2% stimulated the growth of soybeans and thus was used for the characterization of root-associated plant growth promoting bacteria. The bacteria were enumerated after 48 h in the plate count agar and fungi after 5 days in PDA medium. The total numbers of cultivable bacteria isolated from the rhizosphere of plants grown in soil without biochar were 1.5 × 10<sup>7</sup> CFU (colony-forming units, per gram fresh weight) and 5.3×10<sup>7</sup> CFU (per gram fresh weight) in soil with 2% HTC char. Furthermore, a remarkably greater number of fungi (1.8 × 10<sup>4</sup> CFU per gram fresh weight) were observed in the rhizosphere of the plants grown in soil without biochar compared with the plants grown in soil amended with 2% HTC char (0.9 × 10<sup>4</sup> CFU per gram fresh weight).

In total, 200 bacterial strains were isolated from the rhizosphere of soybeans. Among these, 90 isolates were selected from plants grown in control soil, and 110 isolates, from plants grown in HTC char amended soil. All strains were tested for their abilities to stimulate root and shoot growth of soybeans under greenhouse conditions in loamy sand soil. Root and shoot growth stimulating abilities (>20%) were observed in 27–32% of isolates from plants grown in soil without biochar and in 45–57% of isolates from soil amended with 2% HTC char, respectively. A total of 32 isolates from the control plants and 43 isolates from the HTC amended soil induced stimulatory effects on plant growth compared with the non-treated control plants.

### Identification of Plant Growth Promoting Bacteria by MALDI-TOF MS

A total of 35 pure isolates from the rhizosphere of control plants and 43 isolates from the rhizosphere of soybeans grown in HTC char amended soil showing plant growth stimulation ability were taxonomically analyzed by MALDI-TOF MS. As shown in **Tables 2A,B** and **Figure 2**, there are considerable differences in the diversity of strains isolated from the HTC char amended soil and the control soil. In the rhizosphere of soybeans grown in the control soil, isolates were affiliated with seven genera, whereas 24 isolates were identified at the species

level. Bacillus was the predominant genus, which was followed by the genera Arthrobacter and Rhizobium. Furthermore, isolates affiliated with the genera Cellulosimicrobium, Enterobacter and Pseudomonas were also found. The most abundant species were identified as Rhizobium radiobacter (C14, C53, C19, C87), followed by the species Arthrobacter globiformis (C3, C16), Bacillus megaterium (C32, C38), Cellulosimicrobium cellulans (C29, C42), Enterobacter asburiae (C46, C50), and Pseudomonas chlororaphis (C28, C44). Only one isolate was identified as Burkholderia terricola (**Figure 2**).

A total of 13 bacterial genera were isolated from the rhizosphere of soybeans grown in HTC char amended soil, whereas 12 isolates were identified at the species level (**Table 2B**). The isolates from biochar amended soil showed a greater diversity compared with the isolates originating from the plant rhizosphere of the control soil. The most abundant isolates were identified as Cellulosimicrobium cellulans (H4, H90, H12, H20), Ochrobactrum intermedium (H7, H26, H86, H65), Pseudoxanthomonas kaohsiungensis (H31, H37, H55, H79, H100), and Stenotrophomonas sp. (H69, H93, H92, H75). Members of the genera Achromobacter, Brevibacillus, Chryseobacterium, Microbacterium, Ochrobactrum, Paenibacillus, Pseudoxanthomonas, Sphingobacterium and Stenotrophomonas were not found among isolates from the control soil.

#### *In vitro* Plant Growth Promoting Traits

All bacterial strains isolated from the rhizosphere of soybeans grown in HTC char amended soil and without biochar were screened for multiple plant growth promoting traits. Most of the bacterial isolates exhibited one or more plant growth-promoting activities (**Tables 2A,B**).

The production of the phytohormone IAA by bacterial isolates is shown in **Table 2A**. A large amount of the rhizosphere isolates (48%) from HTC char amended soil produced IAA, whereas only 28% of the isolates from control soil showed IAA production. Most of the IAA producing isolates from control soil belonged to the genera Arthrobacter (C99, C16, C71) and Bacillus (C21, C32, C90). Three isolates belonging to the genus Stenotrophomonas (H93, H75, H66) from HTC char amended soil showed IAA activity, followed by the genera Cellulosimicrobium (H90, H20), Pseudomonas (H70, H73) and Rhizobium (H8, H76).

Positive P-solubilization was observed in 7 strains from 4 genera (20%) originating from plants grown in control soil and 16 strains from 11 genera (37%) originating from HTC char amended soil. All bacterial isolates were screened for their ability to suppress plant pathogenic fungi, such as Fusarium solani, F. culmorum, F. graminearum, Alternaria infectoria, and A. teniussima. The proportions of isolates with antagonistic activity to one or more pathogens was higher for the HTC char amended soil (51%) than for the control soil (28%). As shown in **Table 2A**, two Pseudomonas chlororaphis strains, C27 and C28, from control soil and six isolates, Stenotrophomonas maltophilia H66, Stenotrophomonas sp. H92, Cellulosimicrobium cellulans H90, Bacillus megaterium H82, Paenibacillus polymyxa C44, and Pseudomonas putida 73, from HTC char amended soil exerted the highest inhibition of mycelial growth of the genus Fusarium.

The ability of isolates to produce cell wall degrading enzymes, as well as proteases and lipases, was also determined. The isolates from the HTC char amended soil exhibited a higher proportion of enzyme producers than the control soil, where lipase, protease, pectinase and cellulase activity were detected in 14, 33, 40, and 26% of the isolates, respectively. The percentage of enzyme producing bacteria isolated from control soil was lower, where only 6% of isolates exhibited lipase, 11% protease, 20% pectinase, and 29% cellulase activity. Out of isolates that exhibited plant growth-promoting activities in vitro, eight isolates (H66, H75, H72, H76, H73, H44, H22, and H90) originating from HTC char amended soil and six isolates (C99, C28, C46, C78, C30, and C87) originating from control soil were selected for plant growth stimulation under greenhouse conditions.



\* + ++, highly probable species identification; ++, secure genus identification; +, probable genus identification.

#### Plant Growth Stimulation

All 14 selected bacterial strains were screened for plant growth stimulating abilities in pots under greenhouse conditions. The results showed that six strains isolated from plants grown in the control soil without biochar significantly increased root or shoot dry weight compared with the untreated controls (**Figures 3A,B**). The root dry weight increased up to 51% after inoculation with Pseudomonas chlororaphis (C28) and the shoot dry weight increased up to 44% with Enterobacter asburiae (C46; **Figure 3A**). Significant increases (between 28 and 63%) in plant dry weight relative to non-inoculated controls were observed with isolates from HTC char amended soil. The isolates Cellulosimicrobium cellulans (H90), Pseudomonas putida (H73), Stenotrophomonas maltophilia (H66) and Stenotrophomonas sp. (H75) showed significantly higher plant growth stimulation, from 40 to 63% (**Figure 3B**).

#### DISCUSSION

Biochar incorporation into soil has been shown to enhance plant growth, to sequester carbon and to improve soil fertility, and moreover to protect plants from various soil borne pathogens (Lehmann and Joseph, 2009; Zimmerman, 2010). The increase in plant growth with biochar application has been reported for various species such as pine and alder (Robertson et al., 2012), peanut (Agegnehu et al., 2015), tomato (Vaccari et al., 2015), wheat (Akhtar et al., 2015) and also soybean (Sanvong and Nathewet, 2014)—however, several other studies reported no significant effect on plant growth (Chan et al., 2007; Van Zwieten

#### TABLE 2B | Plant growth promoting traits of strains isolated from the rhizosphere of soybeans grown in soil amended with 2% HTC char.


\* + ++, highly probable species identification; ++, secure genus identification; +, probable genus identification.

et al., 2010). In summary, these observations indicate that effects of biochar on plant growth depend on the type of biochar, the application rate, and soil properties (Alburquerque et al., 2014) but mechanisms behind effects mostly remain unresolved. In our study, we confirmed a positive impact of HTC char treatment on the growth of soybean, but not in case of either MBC or WBC amendments. Similar observations were reported by Reibe et al. (2015b), when Pyro-char (MBC) and HTC-char applications resulted in significantly higher dry matter yields of wheat after 6 weeks of growth in rhizoboxes, as compared to Pyreq-char (WBC) or a control. There are several possible reasons why hydrochar might increase plant growth and enhance nutrient acquisition. Hyrdochar contains a higher amount of labile carbon fractions (Cao et al., 2010), which may stimulate microbial activity and thereby improve soil nutrient cycling (Kolb et al., 2009). Furthermore, hydrochars were found to reduce nitrogen losses from soil by immobilization and may provide nitrogen in plant-available form (Libra et al., 2011), whereas

Pyro-chars contain less nitrogen with a decreased availability to plants (Gaskin et al., 2010).

Furthermore, HTC biochar amendment showed an impact on root associated microbes and on microbial interactions with plants, which were previously rarely studied in this context. Our findings are confirmed by the results of Kolton et al. (2011), who showed a clear shift in the total root-associated microbial community composition of mature sweet pepper (Capsicum annuum L.) after amendment with biochar from citrus wood. In our study, the analysis of cultivable root associated bacteria demonstrated that HTC char amendments increased bacterial populations in the rhizosphere of soybeans compared with control plants, whereas fungal growth was decreased over the control, in agreement with Chen et al. (2013). An increased microbial activity in the rhizosphere after the addition of hydrochar could be explained as a result of changes in soil chemical and physical properties in the root surface area. Prendergast-Miller et al. (2014) observed that biochar-amended soils had larger rhizosphere zones than the control. Moreover, the rhizosphere contained biochar particles providing additional labile carbon, nitrogen and phosphorus sources and also habitat niches, supporting bacterial proliferation and persistence in the rhizosphere.

In all rhizosphere samples from soybeans, we found a high diversity of potential plant growth promoting rhizobacteria. However, the species composition in the treated and nontreated plants was different. The most abundant species isolated from soybeans grown in control soil were Rhizobium radiobacter, Arthrobacter globiformis, and Bacillus megaterium, whereas in HTC char amended soil, Cellulosimicrobium cellulans, Ochrobactrum intermedium, Pseudoxanthomonas kaohsiungensis, and Stenotrophomonas sp. were dominant. The species identified in our study are already known for their plant growth promoting abilities, e.g., R. radiobacter stimulated growth of barley (Hordeum vulgare; Humphry et al., 2007), and B. megaterium stimulated growth of bean (Phaseolus vulgaris; Ortíz-Castro et al., 2005). Furthermore, a strain of the species C. cellulans (KUGr3) is able to form IAA, solubilize phosphate and stimulated growth of chili plants (Capsicum annuam; Chatterjee et al., 2009). O. intermedium increased seed germination, root and shoot length, and grain yield in lentil (Lens esculenta; Faisal, 2013). Several Stenotrophomonas sp. strains increased root and shoot growth and the nutrient uptake of soybean (Glycine max), cucumber (Cucumis sativus), and tomato (Solanum lycopersicum; Egamberdieva et al., 2011; Berg and Martinez, 2015).

In addition, the beneficial properties of species in treated and non-treated plants were different. Compared with control soil, a higher proportion of isolates from the HTC char amended soil was found to produce IAA, HCN and cell wall degrading enzymes. Furthermore, a higher proportion of bacterial isolates was capable of hydrolyzing organic and inorganic phosphorus from insoluble compounds and showed antagonistic activity to one or more pathogens. In the present study, a decrease in fungal populations (∼50% reduction) was observed after HTC char addition. The increased proportion of bacteria capable of inhibiting fungal pathogens following amendment of HTC char suggests that the observed suppression of the fungal population was due to antagonistic interactions of microbes. The phytohormone IAA is a naturally occurring auxin which has a major role in the regulation of plant growth. The stimulation of the growth of various plants by inoculation with PGPR and IAA producing ability is well-documented (Egamberdieva, 2009, 2012; Berg et al., 2010). Phytohormones produced by rootassociated bacteria will be taken up by plant cells, stimulate cell proliferation, and enlarge the root system so that nutrients and water can be taken up more efficiently. For example, IAA

producing Stenotrophomonas rhizophila significantly affected plant growth, N and P uptake and the number of nodules in soybean (Egamberdieva et al., 2015). Similarly, multiple isolates from the rhizosphere that suppress fungal growth by the production of HCN, cell wall degrading enzymes or antifungal compounds were used to prevent and control fungal diseases (Berg et al., 2013b; Maurer et al., 2013). Seed coating with Pseudomonas strains antagonistic to soilborne pathogens, such as Sclerotium rolfsii, Fusarium oxysporum, and Rhizoctonia solani, produced siderophores, chitinase, and HCN and were therefore able to suppress infections in soybean seedlings by fungal pathogens (Susilowati et al., 2011). In another study, the charcoal root rot of soybean caused by Macrophomina phaseolina was attenuated by the antagonistic bacterial strains P. agglomerans and Bacillus sp. under greenhouse conditions (Vasebi et al., 2013). The mechanisms involved in plant growth stimulation and the biological control of plant pathogens were also observed for bacterial isolates in our study and were thus further evaluated for their impact on plant growth promotion of soybeans under greenhouse conditions. Indeed, inoculation of soybeans with these isolates led to significant increases in plant growth and development. In previous studies, PGPR Stenotrophomonas rhizophila was able to stimulate root and shoot growth, nodulation and nutrient uptake of soybeans under greenhouse conditions (Egamberdieva et al., 2015). Similarly, Aung et al. (2013) found a significant increase in shoot and root biomass, as well as nodulation in soybeans inoculated with Azospirillum sp., compared to non-inoculated controls under pot conditions.

From our study, we conclude that increased plant growth in response to soil amendment with biochar is based on the type of char, i.e., HTC application increased growth of soybean but not in case of either MBC or WBC. Moreover, HTC application was shown to alter the community composition of root associated microbes exhibiting plant growth-promoting activities in vitro such as phytohormone production and suppression of fungal pathogens. Thus, we provided evidence that improved plant growth by hydrochar incorporation into soil is mostly an indirect rather than a direct effect that depends on the type of char and the activity of plant-associated beneficial soil bacteria. The stimulation of certain plant-beneficial bacteria by biochar also suggests the possibility of developing combined approaches of biochar treatment and biological control solutions (Berg et al., 2013b).

#### AUTHOR CONTRIBUTIONS

DE, SW, and GB did experimental design work. DE and UB conducted experiments. EA analyzed the data. DE, SW, and

#### REFERENCES


GB wrote the manuscript. All authors read and approved the manuscript.

#### ACKNOWLEDGMENTS

This research was supported by the Georg Forster Research Fellowship (HERMES), the Alexander von Humboldt Foundation for DE. 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 group no. (RG-1435-014).


**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 © 2016 Egamberdieva, Wirth, Behrendt, Abd\_Allah and Berg. 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) or licensor 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 Effect of Seed-Borne Fungi and Epichloë Endophyte on Seed Germination and Biomass of Elymus sibiricus

Xiu-Zhang Li <sup>1</sup> , Mei-Ling Song<sup>1</sup> , Xiang Yao<sup>1</sup> , Qing Chai <sup>1</sup> , Wayne R. Simpson<sup>2</sup> , Chun-Jie Li <sup>1</sup> \* and Zhi-Biao Nan<sup>1</sup>

<sup>1</sup> State Key Laboratory of Grassland Agro-Ecosystems, Key Laboratory of Grassland Livestock Industry Innovation, Ministry of Agriculture, College of Pastoral Agriculture Science and Technology, Lanzhou University, Lanzhou, China, <sup>2</sup> AgResearch Grasslands Research Centre, Tennent Drive, Palmerston North, New Zealand

Edited by: Martin Grube, University of Graz, Austria

#### Reviewed by:

Shengguo Zhao, Institute of Animal Science, Chinese Academy of Agricultural Sciences, China Daria Rybakova, Graz University of Technology, Austria

> \*Correspondence: Chun-Jie Li chunjie@lzu.edu.cn

#### Specialty section:

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

Received: 07 April 2016 Accepted: 29 November 2017 Published: 15 December 2017

#### Citation:

Li X-Z, Song M-L, Yao X, Chai Q, Simpson WR, Li C-J and Nan Z-B (2017) The Effect of Seed-Borne Fungi and Epichloë Endophyte on Seed Germination and Biomass of Elymus sibiricus. Front. Microbiol. 8:2488. doi: 10.3389/fmicb.2017.02488 The interactive effects of asexual Epichloë (formerly known as Neotyphodium) endophytes isolated from Hordeum brevisubulatum, Elymus tangutorum and Achnatherum inebrians, and seed-borne fungi on Elymus sibiricus seeds, were determined by an in vitro study using supernatants from liquid cultures of the endophyte strains. In an 8 week greenhouse study, the effects on the seedlings growth was measured. The in vitro study was carried out with the seed-borne fungi Alternaria alternata, Bipolaris sorokiniana, Fusarium avenaceum, and a Fusarium sp. isolated from E. sibiricus. Different concentrations and combinations of the liquid cultures of endophytic fungi enhanced the interim germination, germination rate, length of coleoptile and radicle, and seedling dry weight of E. sibiricus under stress from seed-borne fungi. In the greenhouse study, different concentrations of the supernatant of the endophytes from H. brevisubulatum and E. tangutorum but not A. inebrians, signficantly (P < 0.05) enhanced E. sibiricus seed germination. There was no significant (P > 0.05) increase of the tiller numbers after 2 weeks. However, later on, there were significant (P < 0.05) increases in tiller number (4–8 weeks), seedling height (2–8 weeks) and dry weight (2–8 weeks). The application of Epichloë endophyte culture supernatants was an effective strategy to improve seed germination and growth under greenhouse conditions.

Keywords: Epichloë endophyte, Elymus sibiricus, seed-borne fungi, seed germination, Elymus tangutorum, Achnatherum inebrians, seedling, dry weight

## INTRODUCTION

Endophytic fungal associations with grasses are very common, and the most intensively studied are those between ascomycete fungi and temperate grasses, in particular those involving asexual endophytes of the genus Epichloë (Schardl, 2001; Schardl et al., 2004). Asexual or anamorph-typified Epichloë have a common origin with the sexual Epichloë or teleomorphtypified species (Kuldau et al., 1997; van Zijll de Jong et al., 2011; Leuchtmann et al., 2014). Teleomorph-typified Epichloë species are sexually reproducing and cause a condition known as "choke" in grasses, whereby the fungal stromata formed during sexual reproduction leads to reduced flower and seed production (Schardl et al., 2004). The host range of symbiotic fungal endophytes has been described in cool-season grasses (Leuchtmann, 1993; Scott, 2001). Fungal endophytes are of increasing interest due to a growing list of benefits that they can confer on their hosts, including both abiotic and biotic factors such as tolerance to drought (Malinowski and Belesky, 2000; Clay and Schardl, 2002; Hahn et al., 2008), resistance to insects, nematodes and other herbivorous attacks (Omacini et al., 2001; Schardl et al., 2004; Schardl, 2009; Zhang et al., 2012) including bird deterrence (Pennell et al., 2010; Pennell and Rolston, 2012).

Besides that, Epichloë endophytes can increase tolerance to pathogenic fungi, although the deployment of Epichloë as agents for the biological control of diseases has shown mixed results (Kuldau and Bacon, 2008). in vitro suppression of plant pathogens by endophytic fungi has been demonstrated (White and Cole, 1985; Holzmann-Wirth et al., 2000), there is some evidence showing that colony growth of plant-pathogenic fungi is inhibited by Epichloë endophytes (Christensen and Latch, 1991; Christensen, 1996; Wäli et al., 2006) and that disease tolerance or resistance can be imparted by Epichloë species (Li et al., 2007b; Tian et al., 2008; Porras-Alfaro and Bayman, 2010). The name Epichloë gansuensis (Neotyphodium gansuense) was proposed by Chunjie Li and Zhibiao Nan (Li et al., 2004; Leuchtmann et al., 2014) for an endophytic fungus symbiotic with A. inebrians from Gansu, China. Dual-culture testing and inoculation of detached leaves have shown that E. gansuensis can inhibit growth and disease lesion development of some fungal pathogens (Li et al., 2007a).

At present, although many of the mechanisms of the interaction between Epichloë and fungal pathogens are not clear, it is reported that several pathogenic fungi are controlled to some level by endophyte infection in vitro: Alternaria alternata, A. triticina, Bipolaris sorokiniana, Cladosporium spp. including C. cladosporioides, Curvularia spp. including C. lunata, Drechslera erythrospila, Fusarium acuminatum, Phomopsis spp., Rhizoctonia cerealis, and R. zeae (White and Cole, 1985; Gwinn and Bernard, 1988; Holzmann-Wirth et al., 2000; Li et al., 2007b; Xie et al., 2008). Compared with un-infected grasses of Agropyron cristatum, Elymus cylindricus, and Festuca rubra, Epichloë endophyte can reduce the numbers of Alternaria, Cladosporium, and Fusarium species on leaves of host grasses (Nan and Li, 2000). Vignale et al. report that E. pampeana (N. pampeanum) and E. tembladerae (N. tembladerae) can protect their host plant Bromus auleticus against the pathogenic fungus Ustilago bullata (Vignale et al., 2013). Other studies have demonstrated inhibitory effects in vivo against Ascochyta leptospora, F. avenaceum, F. chlamydosporum, F. culmorum, F. oxysporum, F. solani, Gliocladium roseum, Laetisaria fuciformis, and Sclerotinia homeocarpa (Bonos et al., 2005; Clarke et al., 2006; Li et al., 2007b; Tian et al., 2008).

Elymus sibiricus (Siberian wildrye) is a perennial, caespitose grass, widely distributed around the world (Ma et al., 2012). It usually grows on arid or semiarid mountain or valley grasslands at altitudes from 1,000 to 4,000 m in northwestern China. It has also played an important role in native grassland restoration on the Qinghai-Tibet Plateau of China as a pioneer grass species (Ma et al., 2012). E. sibiricus usually serves as an important forage grass, and has been widely employed in establishing sown grasslands to develop stock raising, due to its strong adaptability, excellent tolerance to drought and cold, high crude protein content, and good palatability (Yan et al., 2007). However, the pathogenic fungus of seed-borne is important factor to limit the host E. sibiricus germination and seedling growth (Li et al., 2007b).

Presently much of the research involving Epichloë is concentrated on the relationship with the host grass, secondary metabolites, interaction mechanisms, taxonomy and ecology. Here we examine Epichloë endophytes isolated from three species of grass and seed-borne fungi isolated from E. sibiricus. The effect of endophytic liquid medium exudate on E. sibiricus germination under a seed-borne fungus burden is examined to provide a theoretical basis for the rational use of Epichloë endophytes in the field.

## MATERIALS AND METHODS

### Epichloë Endophyte Biological Material

Epichloë gansuensis (N. gansuense) (CBS 119808, ATCC-MYA-3669) was isolated from stems of A. inebrians from Sunan, Gansu Province, China (Li et al., 2007b). Epichloë/Neotyphodium spp. isolated from H. brevisubulatum and E. tangutorum were marked as Eb and Et, and the Epichloë gansuensis was marked as Eg. The H. brevisubulatum and E. tangutorum samples were collected from Linze (E:102◦ 54′ , N:37◦ 29′ ; 1,450 m) and Lanzhou (E103◦ 56′ , N36◦ 01′ ; 1,714 m), Gansu Province, China in 2012 (Song and Nan, 2015; Song et al., 2015).

A 4 mm diam plug of 1 week-old endophytic fungus grown on potato dextrose agar (PDA) was used to inoculate 150 mL flasks of potato dextrose broth (PDB) nutrient medium, with 3 repetitions of each strain. Four weeks later, the broth was filtered and centrifuged. The filtrate was diluted to 50 and 25%. The sample of endophytic fungi was marked as Eb01, Eb02, Eb03, Et01, Et02, Et03, Eg01, Eg02 and Eg03, 01 is 25% diluted, 02 is 50% and 03 is undiluted.

### E. sibiricus Sample and the Seed-Borne Fungi

The E. sibiricus seed sample was collected from from Sunan (E99◦ 38′ , N38◦ 50′ ; 2,233 m), Gansu Province, China. Ten seeds were inoculated to each of 10 Petri plates (9 cm diameter) containing PDA, then incubated at 22 ± 1 ◦C in the dark. Observation was made of fungal colony growth on the seed, colonies were picked off to clean PDA plates, and subsequently identified. The 4 fungal strains use in this test were identified as A. alternata, B. sorokiniana, F. avenaceum along with an unidentified Fusarium species. Ten replicates of each strain were cultured for 1 week on PDA.

Sterile water was added to the cultures and a spore suspension produced. The spore suspension was micropore filtered. Conidial concentration was determined using a blood count plate and suspensions adjusted to spore concentrations no <10<sup>6</sup> /mL. Adjusted spore suspensions were maintained at 4◦C in conical flasks. All isolates of Epichloë spp. and fungal pathogens were deposited at the Mycological Herbarium of Lanzhou University.

### Effects of the Endophytic Fungi Supernatant on Germination and Seedling Growth under Seed-Borne Fungi Stress

Seeds were surface sterilized in 75% ethanol (v/v) for 5 min and 5% sodium hypochlorite for 10 min, then washed with sterile water 3 times. The seeds were then placed into a centrifuge tube with 15 ml of the endophytic fungi supernatant and imbibed for 12 h under axenic conditions. The control was a centrifuge tube containing the same amount of sterilized water. Seed was incubated in the dark with ventilation. Incubate seed is dry, then sow on two layers of sterile filter paper in a 90 mm glass Petri dish with 5 ml of spore suspension of seed-borne fungi. Each treatment consisted of 4 replicates, for a total of 200 seeds. According to the method of ISTA (1996), all of the seeds were incubated in a growth chamber (25 ± 1 ◦C, 24 h illumination), with sterile distilled water (5 mL) added to each Petri plate every 2 days using a disposable syringe. Germination potential was noted on day 6, percentage of germination on day 12. After 12 day, the radicle length (RL), coleoptile length (CL) and dry weight were measured. The dry weight was obtained after drying at 75◦C until a constant weight was recorded (to 0.0001 g) with an electronic balance. The radicle length (RL) and coleoptile length (CL) were measured by Vernier calipers. The above characteristics were calculated by the formula: Interim germination = (the number of germinated seeds at 6 day/total number of seed) ×100%; Germination rate (GR) = (the number of germinated seeds at 12 day/total number of seed) ×100%; Vigor index (VI) = (the number of germinated seeds at 6 day/6<sup>∗</sup> length of radicle) ×100%.

### Effects of the Endophytic Fungi Supernatant on Plant Growth under Greenhouse Conditions

Uniform, plump seeds were selected and soaked in the Eb03, Et03, and Et02 for 12 h under axenic condition, from this, three supernatant samples were chosen because of they have the effect with high germination rate of the seed. The control consisted of a centrifuge tube containing the same amount of sterilized water. The seed was incubated in the dark with ventilation. Seed was dried and sown into plastic pots (12 cm diameter × 10 cm depth) containing sterile soil (Junzilan, Lanzhou, China), watered as needed, and grown under controlled greenhouse conditions (22/18◦C, day/night; sunlight; 65% relative humidity. Each treatment was repeated 5 times independently. Plant height and tiller number were recorded every 2 weeks for 8 weeks. Seedling biomass was harvested after 8 weeks, dried at 75◦C and weighted until a constant was recorded.

#### Statistical Analysis

The data were analyzed for variance (ANOVA) and least significant difference (LSD) using SPSS. 19.0 software (SPSS Inc., Chicago, IL, USA).

#### RESULTS

### Effects of the Endophytic Fungi Supernatant on Germination and Seedling Growth under Seed-Borne Fungi Stress

Exudates of the 9 endophytic fungi in supernatant when applied to seed were able to enhance interim germination under seed-borne fungi stress, the seed germination was differs under different endophytic fungi sample (**Table 1**). The seed germination is significantly (P < 005) higher when treated with Et supernatant compared to the treatments with Eb and Eg which resulted in higher germination than the (CK) control. For different species of seed-borne fungi, the effect of the application of endophytic fungal supernatant on interim seed germination differed. Across the range of Epichloë supernatant examined interim germination of seed exposed to seed-borne fungi A. alternata, B. sorokiniana, F. avenaceum, and Fusarium sp. is 21.47–39.13% (mean 29.30%), 18.07–39.93% (mean 27.80%), 13.03–28.60% (mean 19.44%), and 13.77–33.70% (mean 21.47%), respectively. For most strains, as the endophytic fungi supernatant dilution was reduced, the interim germination of seed under seed-borne fungi stress increased. However, the seed germination rate decreased with the increase of concentration of endophytic fungi treated by Eg under all of the seed-borne fungi and Eb03 under Fusarium sp.

The germination rate of the E. sibiricus seeds imbibed in endophytic fungi supernatant was significantly (P < 0.05) higher than the control (**Table 2**). Germination rate of the seed of E. sibiricus under seed-borne fungi stress increased with increasing endophytic fungi supernatant concentration except for Eb and Eg. With the increase of concentration of endophytic fungi supernatant of Eb, the germination rate under seedborne fungi stress of B. sorokiniana and F. avenaceum was higher at first then descended as concentration increased. Eg also showed a higher germination rate across all the seedborne fungi treatments, but as concentration of the supernatant

TABLE 1 | Interim Germination (%) of Elymus sibiricus under seed-borne fungi stress resulting from effects of the Epichloë endophyte.


Values are mean (±SE) of four independent replications with 50 seeds for each replication at 6 days. Significant differences at the 0.05 level in the same column are indicated by different letters A, B, C, D, E, and in the same row with a, b, c. 01 is 25% diluted of Epichloë endophyte, 02 is 50% and 03 is undiluted.

TABLE 2 | Germination rate (%) of Elymus sibiricus under seed-borne fungi stress resulting from effects of the Epichloë endophyte.


Values are mean (±SE) of four independent replications with 50 seeds for each replication at 12 days. Significant differences at the 0.05 level in the same column are indicated by different letters A, B, C, D, E, F, G, and in the same row with a, b, c. 01 is 25% diluted of Epichloë endophyte, 02 is 50% and 03 is undiluted.

increased germination rate decreased. However, they had no significant difference (P > 0.05). The germination rate of the E. sibiricus seeds imbibed in Et supernatant was higher than those imbibed in the other supernatants. The germination rate of the E. sibiricus seeds imbibed in the exudates of 3 endophytic fungi, Eb, Et, and Eg is 62.00–78.93% (mean 71.53%), 66.67–85.07% (mean 77.79%), and 61.63–76.80% (mean 69.05%), respectively.

Coleoptile length of E. sibiricus under seed-borne fungi stress increased with increasing concentration of most endophytic fungi supernatant with the exception of the Eg exudate (**Table 3**). The coleoptile length of E. sibiricus seedlings generated from seed imbibed in Eg supernatant and exposed to A. alternata and Fusarium sp. increased as the concentration of Eg liquid medium increased but then decreased with increasing concentration 3.97, 4.07, 4.03, 3.92, 4.04, and 4.01 cm, however there was no significant difference (P > 0.05). The coleoptile length of seedlings generated from seed exposed to the supernatant of the 3 fungal strains were significantly (P < 0.05) higher than that of the control. The increase in coleoptile length observed for seedlings generated from endophytesupernatant imbibed seed under seedborne fungi burden varied. The coleoptile length of seedlings treated by Et02 and Et03 were significantly (P < 0.05) higher than the others, with coleoptile lengths of 4.47–4.97 cm (mean 4.70 cm), 4.50–4.91 cm (4.79 cm), respectively.

Radicle length of E. sibiricus under seed-borne fungi stress increased with increasing concentration of endophytic fungi liquid medium except with Et exudate (**Table 4**). The radicle length of E. sibiricus under A. alternata pressure increased as the concentration of Eg exudate increased but then decreased with increasing exudate concentration however there was no significant difference (P > 0.05). The radicle length of E. sibiricus treated with Et02 and Et03 were significantly (P < 0.05) higher 3.41–3.92 cm (mean 3.71 cm) and 3.86–4.08 cm (mean 3.99 cm), than the others. The radicle length of endophyte exudate treated E. sibiricus under seed-borne fungi stress involving A. alternata, B. sorokiniana, F. avenaceum, and F. sp. was 2.15–3.33 cm (mean 2.84 cm), 2.42–4.08 cm (mean 3.49 cm) and 2.39–2.76 cm (mean

TABLE 3 | Coleoptile length (cm) of Epichloë endophyte-infected Elymus sibiricus under seed-borne fungi stress.


Values are mean (±SE) of four independent replications with 50 seeds for each replication at 12 days. Significant differences at the 0.05 level in the same column are indicated by different letters A, B, C, D, E, and in the same row with a, b, c. 01 is 25% diluted of Epichloë endophyte, 02 is 50% and 03 is undiluted.

TABLE 4 | Radicle length (cm) of Elymus sibiricus under seed-borne fungi stress resulting from effects of the Epichloë endophyte.


Values are mean (±SE) of four independent replications with 50 seeds for each replication at 12 days. Significant differences at the 0.05 level in the same column are indicated by different letters A, B, C, D, E, F, and in the same row with a, b, c. 01 is 25% diluted of Epichloë endophyte, 02 is 50% and 03 is undiluted.

2.57 cm), respectively, which was higher than the control at 1.92–2.14 cm (mean 2.02 cm).

Dry weight of E. sibiricus seedlings under seed-borne fungi stress increased with increasing concentration of endophytic fungi exudate except for Eg (**Table 5**). The dry weight of seedlings treated with A. alternata and F. avenaceum initially increased with increasing concentration of Eg exudate then decreased. Treatment of B. sorokiniana stressed seedlings with Eg resulted in a relative seedling dry weight decrease, initially, followed by an increase with increasing exudate concentration, but with no significant difference (P > 0.05).

#### Effects of the Endophytic Fungi Liquid Medium on Plant Growth under Greenhouse Conditions

Plant height of E. sibiricus treated by the optimal concentration of Eb03, Et03 and Et02 was significantly (P < 0.05) higher than the TABLE 5 | Dry weight (10−2g) of Elymus sibiricus under seed-borne fungi stress resulting from effects of the Epichloë endophyte.


Values are mean (±SE) of four independent replications with 50 seeds for each replication at 12 days. Significant differences at the 0.05 level in the same column are indicated by different letters A, B, C, D, E, F, and in the same row with a, b, c. 01 is 25% diluted of Epichloë endophyte, 02 is 50% and 03 is undiluted.

control during 2–8 weeks (**Figure 1**). The mean plant height of the treatment is 5.4, 12.8, 17.4, and 21.4 cm at 2, 4, 6, and 8 week, respectively. The plant height difference among the 3 treatments was not significant (P > 0.05) during 2–4 weeks, from 4 to 8 weeks plants treated with Et03 and Et02 showed no significant (P > 0.05) difference, but the difference between plants treated with Et03 and Eb03 was significant (P < 0.05). The height of plants treated with Et03 is higher than that of plants subjected to other treatments during 2–8 weeks, and significantly (P < 0.05) higher than the control of 4.9, 11.0, 14.7, and 18.2 cm.

After 2 weeks, the tiller numbers of the treatment and control showed no significant (P > 0.05) difference (**Figure 2**). The tiller number of treated plants was significantly (P < 0.05) higher than the control plants during 4–8 weeks. Plants treated with Et03 had more tillers than the other treatments, however this was not significant (P > 0.05) with respect to plants treated with Eb03 and Et02. Treatment with endophyte supernatant resulted in large effects on the formation of tillers during the 4–6 week period. The tiller number of treated plants was significantly higher than that of the control plants by 29.17, 58.33, and 37.50%, respectively.

At the final harvest after 8 weeks, treatment with endophyte supernatant from Eb03, Et03, and Et02 resulted in a significant (P < 0.05) increase of total dry weight, higher than the control seedlings by 37.20, 53.63, and 42.43%, respectively (**Figure 3**). Among the three different treatments, that treated by Et03 was higher than the others, the effect of the other endophytes was not significant (P > 0.05).

#### DISCUSSION

Biological control of plant diseases is an important effect observed on grasses infected by Epichloë endophytes although there are mixed results. It is observed that epiphyllous mycelial nets in some endophyte-grass associations may play a role in defense against pathogens by niche exclusion (Moy et al., 2000). The research methods of endophytic fungi interaction with pathogenic fungi are mainly concentrated in dual-culture on plate, PDB trials, in vitro inoculation, vital inoculation and field trial. The present work, to our knowledge, is the first report of the protective effect of Epichloë endophytes against seed-borne fungi, and resulting in increased seed germination and plant growth.

Previous research has shown that Epichloë can have inhibition activity against pathogenic fungi such as Alternaria, Bipolaris, Fusarium, Cladosporium, Pythium, Curvularia, Drechslera, Rhizoctonia, and so on (Holzmann-Wirth et al., 2000; Nan and Li, 2004; Li et al., 2007a; Tian et al., 2008). In this work we found that, compared with controls, the Epichloë endophyte increases interim germination, germination rate, coleoptile length, radicle length and dry weight of E. sibiricus to varying degrees when plants are under stress from seed-borne fungi such as A. alternata, B. sorokiniana, F. avenaceum and Fusarium sp.

Similar results were obtained by New Zealand scientists, which ranged from nil to strong inhibition on growth and conidial germination of grass-pathogenic fungi by E. festucae var. lolii (N. lolii) and endophytes in in vitro tests (Christensen and Latch, 1991; Christensen et al., 1991). A previous study with E. coenophiala (N. coenophialum) showed effective inhibition of growth of A. alternata (White and Cole, 1985); E. festucae var. lolii (N. lolii) effectively inhibited Drechslera spp. (Holzmann-Wirth et al., 2000). Nan and Li (2000) showed that detached leaves of E+ Elymus cylindricus had fewer and smaller lesions than those on E– plants inoculated with A. alternata, F. avenaceum. These studies addressed antifungal activities of endophyte in vitro or on detached leaves of other grass species. That the total lengths of lesions on detached leaves were greater (P < 0.05) on E– plants than on E+ plants when inoculated with the plant pathogens A. alternata, B. sorokiniana, C. lunata, F. acuminatum, F. avenaceum, and 10 other species of pathogenic fungi. Although differences between E+ and E– were not consistently significant at all sample times (days after inoculation) in detached leaves. The numbers of lesions were greater (P < 0.05) and the lesions were larger (P < 0.05) on intact E– plants than on intact E+ plants for the pathogens of four pathogens (A. alternata, B. sorokiniana, Curvularia lunata and F. avenaceum) when living plants were studied (Tian et al., 2008).

This study found that compared with the seeds without soaking with endophytic fungi supernatants, the plant height, tiller and biomass all have different degrees of increase. This is similar to the Epichloë endophyte induced improvments in plant height, tillering and biomass of grasses, including A. inebrians (Nan and Li, 2000), F. arundinacea (Joost, 1995) and L. perenne (Clay, 1987). Latch et al. (1985) reported that L. perenne infected with E. festucae var. lolii (N. lolii) resulted in dry weights nearly 38% higher than the non-infected L. perenne, and the leaf area, number of branches and root dry weight were also significantly (P < 0.05) higher than the control. Compared with the un-infected F. arundinacea, E. coenophiala (N. coenophialum), endophyte significantly (P < 0.05) increases the production, performance and the grass coverage by 20– 30% 4 months after sowing (Joost, 1995). The endophytic fungi can increase forage yield of 22–55%, the number of tillers 20–45% and the seed weight 26–41% (Clay et al., 1993). Nan and Li (2000) founds that endophytic fungi infection

FIGURE 1 | Effects of Epichloë endophyte on plant height of Elymus sibiricus. Significant differences at the 0.05 level are indicated within a time by different letters above bars. Values are means of five replicates ±SE.

of H. bogdanii significantly (P < 0.05) promotes the growth of host plants, the plant tiller number increased by 136.8%, herbage yield increased by 33.3% and root dry weight increased by 30%. In filed experiments examining E. cylindricus, the endophyte increased the tiller number by 84.5%, the above ground dry weight increased by 278.7% and the tiller weight increased by 105.3%. A potted plant experiment showed that the extracts from E+ A. inebrians remarkably enhanced the growth of F. arundacea, L. perenne and P. pratensis (Yang et al., 2010).

Tillering of grasses is controlled, amongst other things, by indole acetic acid and other plant hormones. Endophytic fungi have the ability to produce indole acetic acid, this might be one of the reasons for the promotion of tillering in infected plants (West and Gwinn, 1993). However, in this study, there were no significant (P > 0.05) differences of plant tiller number between plants that were treated with endophyte exudate and untreated controls in the 2 weeks under greenhouse conditions. This absence of effect at the P < 0.05 level might be due to low levels of indole acetic acid being produced by the seeding at this early stage of growth.

Based on the results of the germination test presented here, the seed germination decreased with increased culture concentration of the endophytic fungi, similar to the results of Huang et al. (2010) research results. They reported that the water extraction of the E+ A. inebrians had significant inhibitory effects on seed germination and seedling growth of S. capillata and P. sphondylodes using a Petri dish-paper germination method. Inhibition is also seen on the seed germination rate of L. perenne and bud length of P. pratensis by the same method. But the potted experiment showed that the grass powder of E+ A. inebrians accelerated plant growth and initial seedling emergence rates of L. perenne and P. pratensis. It is speculated that the ergonovine and ergine

concentration from A. inebrians/E. gansuensis (N. gansuense) may be responsible for this effect (Yang et al., 2010). Petroski et al. (1990) reported that loline has strong allelopathy to annual ryegrass and alfalfa. The loline produced by F. arundinacea has higher allelopathy to its competitors (Malinowski and Belesky, 2000). The suggestion that the inhibitory effects of Eg and Eb on seed germination due to their heigher alkaloid concentrations requires further study.

Here we have demonstrated that fungal culture supernatants can greatly promote germination and subsequent plant growth under seed-borne fungi stress. However, high concentrations can suppress growth, this action appears to be complex. The interactive mechanisms of Epichloë endophyte and seedborne fungi require further study, and the effective range of concentrations of different endophyte exudates needs to be determined. This may eventually provide insight into strategies for the improvement of field performance and stress tolerance in grasses of forage and turf.

#### REFERENCES


### AUTHOR CONTRIBUTIONS

X-ZL designed and performed experiments, analyzed the data and wrote the manuscript. C-JL designed experiments, polished manuscript, provided reagents and experimental equipment. M-LS, XY, and QC performed experiments and analyzed the data. WS analyzed the data and polished manuscript. Z-BN provided reagents and experimental equipment. All authors reviewed the manuscript.

#### ACKNOWLEDGMENTS

This study is supported by National Basic Research Program of China (2014CB138702), the Natural Science Foundation of China (31372366), Program for Changjiang Scholars and Innovative Research Team in University of China (IRT17R50), the 111 project (B12002) and Fundamental Research Funds for the Central Universities (lzujbky-2016-bt10, lzujbky-2017 -kb11).


accompanying species of Stipa capillata and Poa sphondylodes. Acta Pratacult. Sin. 19, 87–93. doi: 10.3724/SP.J.1105.2010.00087


**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 © 2017 Li, Song, Yao, Chai, Simpson, Li and Nan. 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) or licensor 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.

# Microbiome Selection Could Spur Next-Generation Plant Breeding Strategies

#### Murali Gopal\* and Alka Gupta

Microbiology Section, ICAR-Central Plantation Crops Research Institute, Kasaragod, India

#### "No plant is an island too. . ."

Plants, though sessile, have developed a unique strategy to counter biotic and abiotic stresses by symbiotically co-evolving with microorganisms and tapping into their genome for this purpose. Soil is the bank of microbial diversity from which a plant selectively sources its microbiome to suit its needs. Besides soil, seeds, which carry the genetic blueprint of plants during trans-generational propagation, are home to diverse microbiota that acts as the principal source of microbial inoculum in crop cultivation. Overall, a plant is ensconced both on the outside and inside with a diverse assemblage of microbiota. Together, the plant genome and the genes of the microbiota that the plant harbors in different plant tissues, i.e., the 'plant microbiome,' form the holobiome which is now considered as unit of selection: 'the holobiont.' The 'plant microbiome' not only helps plants to remain fit but also offers critical genetic variability, hitherto, not employed in the breeding strategy by plant breeders, who traditionally have exploited the genetic variability of the host for developing high yielding or disease tolerant or drought resistant varieties. This fresh knowledge of the microbiome, particularly of the rhizosphere, offering genetic variability to plants, opens up new horizons for breeding that could usher in cultivation of next-generation crops depending less on inorganic inputs, resistant to insect pest and diseases and resilient to climatic perturbations. We surmise, from ever increasing evidences, that plants and their microbial symbionts need to be co-propagated as life-long partners in future strategies for plant breeding. In this perspective, we propose bottom–up approach to co-propagate the co-evolved, the plant along with the target microbiome, through – (i) reciprocal soil transplantation method, or (ii) artificial ecosystem selection method of synthetic microbiome inocula, or (iii) by exploration of microRNA transfer method – for realizing this next-generation plant breeding approach. Our aim, thus, is to bring closer the information accrued through the advanced nucleotide sequencing and bioinformatics in conjunction with conventional culture-dependent isolation method for practical application in plant breeding and overall agriculture.

#### Edited by:

Christine Moissl-Eichinger, Medical University of Graz, Austria

#### Reviewed by:

Juris A. Grasis, San Diego State University, USA Pablo Rodrigo Hardoim, University of Algarve, Portugal

> \*Correspondence: Murali Gopal mgcpcri@yahoo.co.in

#### Specialty section:

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

Received: 14 May 2016 Accepted: 24 November 2016 Published: 07 December 2016

#### Citation:

Gopal M and Gupta A (2016) Microbiome Selection Could Spur Next-Generation Plant Breeding Strategies. Front. Microbiol. 7:1971. doi: 10.3389/fmicb.2016.01971

Keywords: microbiome, holobiont, artificial ecosystem selection, plant breeding, synthetic microbiota

### THE 'HOLOBIONT' AS HERITABLE UNIT OF SELECTION

In the age of new ecology, the understanding of a plant as no more an individual at its genomic level but a larger genetic entity comprising of its associated microbial genome, 'the microbiome,' has given rise to the 'holobiont' concept (Zilber-Rosenberg and Rosenberg, 2008; Rosenberg and Zilber-Rosenberg, 2016). A 'holobiont' is thus an assemblage of the individual and its symbionts living and functioning as a unit of biological organization (Bordenstein and Theis, 2015; Theis et al., 2016), having the capacity to replicate and pass on its genetic composition; therefore, a unit of selection (Zilber-Rosenberg and Rosenberg, 2008; Booth, 2014; van Opstal and Bordenstein, 2015). The genomic reflection of complex symbiotic interactions of the plant holobiont is governed by its holobiome or hologenome comprising of the host and its microbial genome (Guerrero et al., 2013; Bordenstein and Theis, 2015). In fact, the collective genome of the rhizosphere microbiome is much larger than that of the plant and therefore referred to as the plant's second genome or pan-genome (Berendsen et al., 2012; Turner et al., 2013). The 'holobiont' concept has its roots in the hypothesis that the complex eukaryotic cells have evolved from simple prokaryotes (Embley and Martin, 2006; Douglas, 2014; Koonin and Yutin, 2014). The recent finding of 'Lokiarchaeota,' a complex archaeabacteria clade that appears to be a missing link between prokaryotes and eukaryotes (Spang et al., 2015), strengthens the presence of prokaryote-to-eukaryote genomic continuum in the plant holobiont (Turner et al., 2013).

### THE MICROBIOME REGULATES HOLOBIONT FITNESS

The plant microbiome is compartmentalized into its rhizosphere, endosphere, phyllosphere, and endophytic microbiota (**Figure 1**) with soil largely being the original source of the microbial diversity as observed in Arabidopsis, maize, rice, grapevine, cannabis and cucurbits (Bulgarelli et al., 2012; Lundberg et al., 2012; Schlaeppi et al., 2014; Winston et al., 2014; Edwards et al., 2015; Glassner et al., 2015; Zarraonaindia et al., 2015). It has also been reported that the diversity of above ground phyllosphere microbiota includes many taxa that are encountered in soil and water (Vorholt, 2012; Kembel et al., 2014). The selection of the microbes from the soil pool into the plant microbiome is driven by the host (Berg and Smalla, 2009; Hartmann et al., 2009; Hirsch and Mauchline, 2012), modulated by salicylic acid production (Lebeis et al., 2015) as well as phenols (Badri et al., 2013a) released from the roots, and the plant's evolutionary history (Bouffaud et al., 2014). In the ecological perspective, the plant holobiont and not the plant as an individual, is now known to respond to the various biotic and abiotic perturbations in a given environment. A significant proportion of the plant holobiont's response is contributed by the microbial symbionts via their ecological services of nutrient mineralization and delivery (Terrazas et al., 2016), protection from pests and diseases, and tolerance to abiotic stress. Therefore, the overall fitness of the plant is governed by the self and its microbiota (Vandenkoornhuyse et al., 2015). Several examples where the plant microbiome, particularly of the root and endophytic compartments, has been used to suppress diseases of field and horticultural crops (Mendes et al., 2011; Spence et al., 2014; Cha et al., 2016), improve drought resistance in desert crops (Lau and Lennon, 2012; Marasco et al., 2012) and grapevine (Rolli et al., 2015) and alter above-ground herbivory (Hol et al., 2010; Badri et al., 2013b) have unequivocally proved that the host microbiome indeed impact the fitness of the plants. Next-generation sequencing technologies, advanced bioinformatic analyses coupled with meticulous culture-dependent isolations had been employed in all the above studies to decode the plant microbiome and get to the important bacterial species involved in regulating the phenotypic expression of the plants.

### MICROBES WORK IN NETWORK MODE TO REGULATE PLANT FITNESS

A 'microbiome' includes bacteria, fungi, actinomycetes, viruses, and protists. However, current information pertaining to the plant microbiome is mostly in reference to the bacterial community. Fungal and virus microbiome research have just begun. Several exciting new studies are unveiling the way in which the plant microbiome performs its duties. They indicate that, like any other species, microorganisms – operate in interlocked networks (van der Heijden and Hartmann, 2016) possessing microbial hubs. Within the networks reside certain keystone species that are critical for the plant-microbe interactions (Agler et al., 2016). It has been found that bacterial communities having high connectence and low nestedness afford them a stabilizing configuration which are able to prevent pathogen attack on some plants (Wei et al., 2015). Before these basic findings became apparent, several works clearly indicated more efficiency when bacteria were applied in a consortium mode for controlling soil borne pathogens (Stockwell et al., 2011; Sarma et al., 2015). Mendes et al. (2011) reported that control of Rhizoctonia solani of sugar beet in disease suppressive soil was because of a suite of 111 Pseudomonas spp. representing the bulk of antagonistic bacteria isolated from the soil, confirming the results obtained by metagenomic analysis of the disease suppressive soil. Similarly, the work of Koberl et al. (2013) in managing Ralstonia disease in medicinal crops in arid ecosystem of Egypt, using a combination of 45 Bacillus spp. with Streptomyces, highlight the phenomenon that microbes act in network mode. A core consortium of five bacteria was found to rescue tobacco (Nicotiana attenuate) from the sudden-wilt caused by Fusarium–Alternaria like complex in continuous cropping system (Santhanam et al., 2015). Consortia level application had also helped in improving drought tolerance in grape vine (Rolli et al., 2015) and date palm (Cherif et al., 2015). A combination of Pseudomonas spp. altered the post-embryonic root development in Arabidopsis that stimulated production of more lateral roots and root hairs and helped the plants perform better under waterand nutrient-limited conditions (Zamioudis et al., 2013). These results indicate that there is better performance of bacteria when

they are applied in a consortium underlining their network mode of activity (Hays et al., 2015) in regulating plant fitness. Now, the plant-microbiome relationship via 'holobiont' concept is not only restricted to production and protection applications in plants but is also expanding into the realm of plant breeding.

#### CONVENTIONAL SELECTION BREEDING FOCUSED ON PLANT GENOME

Wild plants have evolved over time by selectively assembling plant-beneficial microbiota from soil as their partners. This association was disrupted with the development of agriculture through domestication of important crops. Further disruption entailed as conventional plant breeding and modern genomicsassisted methods focused only on the plant genome, not the hologenome, for developing crops with higher yield, resistance to insect pest and fungal pathogens, tolerance to abiotic stresses such as drought and salinity and characteristics of superior quality for many other desirable attributes. Plant breeding has greatly helped in the food security of the global population. However, domestication of such genetically homogenous crops, cultivated in different ecological conditions, has led to not only the erosion of genetic diversity of the plants; but also extinction of huge microbial diversity in soil that would have been the source of several plant-beneficial microbiota (Perez-Jaramillo et al., 2016).

Domestication and intensive cultivation of a single crop has led to appearance of several qualitative issues such as reduced nutrient use efficiency, increased susceptibility to pests and diseases, inability to overcome abiotic stresses, etc. Domestication also could have removed those traits from plants that were needed to assemble host-specific microbiome affording the plants a very high adaptability to biotic and abiotic stresses (Bulgarelli et al., 2013; Chen et al., 2015). This necessitated application of high quantities of inorganic fertilizer, spraying of insecticides and growth hormones, etc. to maintain the required output (Matson et al., 1997) and on the flip side, drastically losing the soil microbial diversity to a great extent (Weese et al., 2015). Integration of plant-beneficial microorganisms such as nitrogenfixing bacteria, phosphate solubilizing microbes, plant growth promoting rhizobacteria (PGPR) and arbuscular mycorrhizae were included as agronomic components of crop husbandry and became an environmentally benign alternative to supplement the inorganic inputs. From individual inoculations in the beginning, either bacteria or fungi, to mixed inoculations having both bacteria and fungi yielded desirable results in some crops grown under certain soil and environmental conditions. However, the microbial applications did not always perform to expected levels

under different ecological conditions even if the host was the same (Ambrosini et al., 2016). Perhaps singular or combination of two microbes were not able to establish in the soil resulting in below par effectiveness of the bioinoculants. One of the possible reasons could be that the introduced microbes were not able to find their interdependent groups in the foreign soil as in the native soils from which they were originally isolated, which would have helped them to share and exchange critical metabolites like amino acids and sugars to promote their survival under challenging microenvironments. In short, microorganisms are dependent upon their groups for key metabolites to cooccur in an environment having diverse microbial communities (Zelezniak et al., 2015). This again highlights the fact that microorganisms work in network mode and their networking offers a broad base of microbial genomic diversity that could impact plant genetic variability.

### MICROBIOME OFFERS GENETIC VARIABILITY TO PLANTS

Genetic variability in plants, in the form of landraces and wild relatives, is a key factor that conventional plant breeders focused on to produce new varieties and hybrids. This approach, as mentioned earlier, completely focused on the plant genome for the variability. Though, it has yielded splendid results in developing better crops in terms of yield, selection and domestication has led to erosion of plant genetic diversity making plant breeders look for newer sources of variability in plants. With advancement in cutting edge technologies, another new source of variability in plant genetic material viz. 'epigenetics,' has become a focus in crop improvement programs in recent years (Varshney et al., 2005; Tsaftaris et al., 2008). Epigenetics refers to the different phenotypic manifestations by plants arising from altered expression of genes without any actual changes in the base pairs. Mechanisms driving epigenesis include: DNA methylation, modifications in chromatin via modifications in the histones and DNA, and RNA interference. It is considered heritable too. Epigenetics pathways are, therefore, reported to produce phenotypic plasticity in plants which enables them to overcome and reproduce in erratic ecosystems (Pikaard and Scheid, 2014). A report on the recently concluded meeting of Epigenetics of Plants International Consortium in the USA highlighted several themes including basic mechanisms of gene regulation, nucleolar dominance, histone dynamics, DNA methylation, and small RNA functions in plant epigenetics and how they could be used for crop improvement as well as stress and defense response by plants (Slotkin, 2016).

Apart from these, the development of holobiont theory is now unveiling a new basis of genetic variation, which is heritable and offered by the plant microbiome, particularly from the endophytic compartment (Nogales et al., 2016). The dependence of plant on its microbiome is to such a great extent that many plants failed to be cultured as transplants in the absence of bacterial and fungal endophytes (Hardoim et al., 2008). Among the endophytes, seed endophytes are of great importance because seeds not only carry the genetic blueprint of plants during transgenerational propagation, but are home to diverse microbiota too. Advancements in the knowledge of microbiome associated with seeds has, therefore, become critical as it forms the basis of vertical transmission of the microorganisms and hence, acts as a closely linked reservoir of plant endophytic microbiome having many positive impacts on plant germination and growth (Hardoim et al., 2015; Truyens et al., 2015). The transmission of endophytic bacteria can take place from parent plant to seed and then to the seedlings (proper vertical transmission), as in rice, or as in wheat, where bacteria are present in the seed coat, crease tissue and endosperm (Robinson et al., 2016). Studies performed to track the seed microbiome diversity indicated that a core-microbiota of endophytes was conserved during the domestication of wild maize (teosinte) to 10 different varieties of modern cultivated maize (Johnston-Monje and Raizada, 2011). In rice too, about 45% of the bacterial endophytes present in first seed generation were found to be transmitted to the second generation, in a study carried out using PCR-DGGE method with surface sterilized seeds (Hardoim et al., 2012). Bacterial endophytes, such as Bacillus spp. transmitted vertically in quinoa, helped in priming of the seeds to counter external reactive oxygen species during germination, thereby, helping the plants to overcome saline and dry soil pressures and improve their stress resistance (Pitzschke, 2016). While terroir was considered as the main source of seed microbial communities (Klaedtke et al., 2015), it was observed that a flux also existed between the rhizosphere and seeds with regard to endophytes. Johnston-Monje and Raizada (2011) have reported such a flux where a seed bacterium, Enterobacter asburiae, was found to egress out of the root and colonize the maize rhizosphere, thereby, indicating that seeds can also modulate the rhizosphere microbiome (Johnston-Monje et al., 2016). Thus, in plants like maize, seeds are known to propagate a set of core-microbiome from generation to generation even when grown in ecologically different soil conditions (Johnston-Monje et al., 2014). Seed microbiome, therefore, form an important source of variability in plants.

Next to seeds, the rhizosphere microbiome introduces heterogeneity in plants by affecting their health and productivity (Berendsen et al., 2012; Berg et al., 2014; Pieterse et al., 2016), improving stress tolerance (Rodriguez et al., 2008), and providing an overall adaptive advantage (Haney et al., 2015). The works of Lau and Lennon (2011, 2012), Panke-Buisse et al. (2014), and Wagner et al. (2014) bring to light the role of soil or rhizosphere microbiome in altering the flowering time, indicating the depth of variability microbiomes offer to plant genome. Microbiomes that help plants develop early or late flowering could be used as breeding strategies to escape drought or salinity or heat or cold stress as plants are known to adopt altered flowering time in response to the above abiotic stresses (Kazan and Lyons, 2016). Therefore, sufficient evidence has accrued to show that the microbiome mediates several critical plant functional traits (Friesen et al., 2011), has a great significance on plant phenotypic plasticity (Goh et al., 2013), and can become a new trajectory for plant neodomestication (Duhamel and Vandenkoornhuyse, 2013). In addition to the variability proffered to plants by the microbiome diversity harbored in various plant tissues, another

Gopal and Gupta Microbiome for Plant Breeding

layer of variability is also added by the epigenetic occurrences in the microbiome similar to epigenetic occurrences in plants. DNA methylation in bacteria and archaebacteria not only saves their DNA from self cleavage by its restriction enzymes through restriction modification but is also involved in gene regulation and introduces genetic variability (Casadesus and Low, 2006). Studies using the single molecule real-time (SMRT) sequencing technology in 230 bacterial and archaeal species showed pervasive occurrence of DNA methylation in 93% of the observed species, stressing the incidence of epigenetic events in prokaryotes. The study unraveled twice as many hitherto known DNA binding specificities of methytransferases (MTases) and more than 800 distinct reproducible methylated motifs (Blow et al., 2016). The role of epigenetic events becomes more relevant to our perspective when it is reported to drive the phase change of freeliving bacteria such as Bradyrhizobium diazoefficiens to symbiotic bacteria because of methylation of specific motifs during the process of symbiosis (Davis-Richardson et al., 2016). Yet another basis of variability in the microbiome is the phenomenon termed as 'horizontal gene transfer' (HGT) (de la Cruz and Davies, 2000) that predominantly occurs in rhizosphere environment. This becomes an additional derivative for heterogeneity to the plants. HGT is brought about by the mobile elements such as gene cassettes, plasmids, transposons, and bacteriophages. Thus, it is evident that the microbiome is able to offer important genetic variability to plants that can be considered for future plant breeding strategies, particularly, when an experimental technique such as artificial ecosystem selection is now available to transfer the complete microbial community.

### ARTIFICIAL ECOSYSTEM SELECTION OF PLANT MICROBIOME

Application of individual microorganism (bacteria or fungi) for improving plant growth, health and overall fitness is comparatively an easy task. But its success in an open system is challenging. Whereas, the application at the microbiome or core-microbiome level has shown to be more successful for the reasons explained elsewhere. However, getting to the relevant bacterial species and preparing their appropriate consortia is the main challenge here because of the complex nature of the microbe-plant interactions. By adopting artificial ecosystem selection method of microbiome transfer (Swenson et al., 2000; De Roy et al., 2014; Voss et al., 2015), strong evidence of heritable changes in drought tolerance in Arabidopsis thaliana (Zolla et al., 2013), alteration of flowering time in Arabidopsis thaliana genotypes, Brassica rapa (Panke-Buisse et al., 2014) and Boechera stricta (Wagner et al., 2014) have been reported. The findings of overlapping core-microbiome in sugarcane (Yeoh et al., 2015) and rice (Edwards et al., 2015) with those of Arabidopsis (Lundberg et al., 2012) give more hope for cross-compatibility of microbiome transfer with phylogenetically unrelated plant species. Not only bacterial but fungal communities are also shared between different plant compartments, with soil being the main source (Coleman-Derr et al., 2016). Even the important biocontrol fungus Trichoderma has been found to have a global core community in endemic plants such as Aeonium, Diospyros, Hebe, Rhododendron in comparison with cosmopolitan plants like maize (Zachow et al., 2016).

Interestingly, this new area of synthetic ecology, in which ecologists and medical professionals design beneficial microbial communities, has its origins in almost century-old field ecological studies (Inouye, 2015), such as the one carried out by Henry (1931), wherein control of Helminthosporium foot rot disease of wheat was achieved by transplantation of soils suppressive to the pathogen. More recently, using a similar soil inoculation technique, it has been shown that plant communities can be restored quickly on degraded or disturbed land with soil communities such as microbes, nematodes and microarthropods being some of the main drivers (Wubs et al., 2016).

### HOST GENOME AND ITS MICROBIOME: STRANGE, THEY ARE NOT BED FELLOWS YET IN THE STRATEGY FOR PLANT BREEDING

As an integral part of the plant hologenome, the plant microbiome is a tool that can be selected together with the plant genome to develop next-generation plant breeding approach. Though some critical views on studies of the microbiome (Hanage, 2014) and hologenome concept (Moran and Sloan, 2015; Douglas and Werren, 2016) exist, it is possible to develop a new plant breeding strategy in which the plant microbiome from a desired field can be developed into a synthetic inoculum and reared with the plant progeny to produce nextgeneration crops. Challenges for developing large quantities of the microbiome inoculum can be surmounted with the help of next-generation sequencing technologies combined with bioinformatic analyses for determining the pan-microbiome, at different hierarchical scales, on which the plant depends for its fitness (Vandenkoornhuyse et al., 2015) and identifying candidate organisms whose abundance in soil correlates with the plant function (Wagner et al., 2014). Systematic isolations that capture the species present in a community (Bai et al., 2015) which produce the desired phenotypic effect will be able to help kick-start this effort. The proposed new plant breeding strategy is an extension of the bespoke microbiome therapy where the possibility of transfer of core-microbiome from pathogen suppressive soils to pathogen prevalent soils was suggested for managing plant diseases (Gopal et al., 2013). It also draws upon from the 'neodomestication' of plants along with its full complement of mutualist theme put forth by Sessitsch and Mitter (2015) as the concept for current century's agriculture for attaining food security. Berg et al. (2016) advocated integration of plant-associated microbiome in research dealing with plant physiological experiments and breeding approaches for the reason that plant microbiome is known to respond ahead of its host plant to any environmental perturbation, which influences the hormonal activity of the plant and thereby its physiology. This integration would lead to

improved understanding of the plant–microbiome interactions and would help in unraveling the functions of the holobiont. They considered it necessary to include cultivar-specific microbiomes in plant breeding studies in view of the high-specificity observed between the symbionts and its host, thus, providing relevant inputs to our proposed perspective on use of microbiome for plant breeding. In another elaborate report, Mueller and Sachs (2015) professed a top–down approach for artificially selecting upon plant and animal microbiomes for improving their health. They described co-evolution as an evolutionary adjustment occurring between two interdependent populations of species in such a way that changes in one population brings about reciprocal changes in the other, and co-propagation as the continuous transmission of host and its microbiome across several generations linking them together in each round of replication. The approaches envisaged by them to establish the functions of microbiome, techniques to manipulate the microbiome through host-mediated selection and to develop starter microbiome culture also form basis of our bottom–up perspective of co-propagating the co-evolved.

### CO-PROPAGATING THE CO-EVOLVED

The approach in our proposed perspective is to co-propagate the co-evolved, i.e., the plant genome and its microbiome. It aims to propel the development in the current knowledge of the microbiome to more practical use in plant breeding, particularly in consideration of disease and drought management, two areas in urgent need of attention to improve agricultural production for food security (Lakshmanan et al., 2014; Haney et al., 2015) in the climate change scenario (Hamilton et al., 2016). Drought and extreme heat, in particular, have been the reason for up to 10% decline in yield of cereals around the world making it the top challenge to crop production (Lesky et al., 2016). Scope for tackling drought using PGPR, i.e., rhizosphere microbiome, is a good option (Ngumbi and Kloepper, 2016). With the current knowledge on the plant microbiome, which is mainly concentrated on bacterial communities, we suggest to co-propagate the microbiome with the plant offspring in the new cultivation with a starter microbiome culture of keystone plant-beneficial microbiota from the target soils. This approach will provide an opportunity to the plants to easily recognize the suite of microbiota with which it had co-evolved and, therefore, preferably recruit them in the new environment. It is also possible that the offspring may have a set of microbiota transferred vertically from the parent, which will enable them to function efficiently in the new environment, if their microbiota are able to interact with the known set of rhizosphere microbiome that was available in the original soil environment in which the parents of the offspring grew. It is now known that the roots attract 2–10 times more types of bacteria than leaves and that the root microbiome is regulated by soil factors such as pH, moisture, and temperature in addition to plant genotype and age (Wagner et al., 2016). Our strategy, therefore, tries to provide the missing microbiome as starter rhizosphere microbiome culture that the plant may require to perform in new environments (**Figure 2**). Providing the starter microbiome culture can be attained either by direct approaches of (i) reciprocal soil transplantation/inoculation from the original soil in which the desired plant had been grown, and (ii) development of synthetic microbiome containing keystone microbiota (plantbeneficial bacteria, arbuscular mycorrhizae, and actinomycetes) or by indirect approach of (iii) transferring microRNA from rhizosphere of target soils to recipient soils. Experiments showing reciprocal soil inoculation or soil transplantation capable of surmounting disease in wheat (Henry, 1931), restore degraded land and giving direction to the type of vegetation grown based on soil inocula (Wubs et al., 2016) and degrade crude oil (Bell et al., 2016) lend credence to the first approach. The work of Calderon et al. (2016) on the restoration of the microbial communities responsible for N-cycling in degraded soil using reciprocal soil inoculum suggests that having an understanding of the priority effects along with the relatedness of the established microbial community and the introduced microbial communities could help in better microbial assemblage and successful restoration of target areas. In a recent work of Bai et al. (2015), it has been shown that, with some meticulous,

systematic and exhaustive isolation of bacteria from phyllosphere and rhizosphere, it is possible to capture majority of the species found reproducibly in their respective natural communities. Studies with synthetic communities of bacteria prepared from the isolations could replicate the gnotobiotic reconstitution system allowing for bacterial community establishment. Current research approach for isolation of 'unculturable' microbiota from the human gut using cutting edge genomics and bioinformatics tools (Browne et al., 2016) can be followed to isolate keystone microbiota from target soils. More support to the second strategy comes from the work of Panke-Buisse et al. (2016) wherein inoculation of a subset of whole microbiome, associated with early flowering in Arabidopsis thaliana cultivated on four different types of solid media, was able to reproduce the same flowering timing in Arabidopsis. The third strategy mentioned using transfer of rhizosphere microRNA is a possibility of adopting the recent development in human gut microbiology where it has been shown that incorporation of microRNAs harvested from feces is able to restore the disturbed gut microbiome to healthy status (Liu et al., 2016). One recent report by Zhang et al. (2016) highlighting export of microRNAs (miRNA166 and miRNA 159), accumulated in root-hypocotyl junction, cotyledon vasculatures, root tissues, etc. of cotton plants, to the hyphae of pathogenic fungus Verticillium dahliae to suppress its virulence, suggests that the third strategy is also feasible.

The ultimate aim of the perspective is to take the research out of the lab and apply it to practical farming techniques using a matching microbiome inoculum to cultivate a given crop. Our perspective reflects the opinion of Denison (2014) who suggested that the key to past and future agriculture depended on increasing the cooperation among plants, their symbionts and the farmers. To make this happen, awareness

#### REFERENCES


amongst farmers about the beneficial role of microorganisms in plant production and protection will need to be strengthened through innovative extension programs and communications (Shugart and Racaniello, 2015). Mass-production of the starter microbiome inoculum can be thought of with improvements in the additive printing technology (3D printing technology) of microscopic bacterial communities (Connell et al., 2013). Though the plant microbiome research is in its growing stage, with increased understanding of the mechanisms by which community coalescence takes place vis-a-vis the microbial assemblage (Rillig et al., 2016) and several new methods available for studying the rhizosphere environment (Oburger and Schmidt, 2016) including nano-scale tools (Biteen et al., 2016), the challenge can be surmounted with improvement in the knowledge of the microbe-to-microbe and microbe-to-plant interactions by the end of the decade (Mitter et al., 2016) to be able to provide solutions for 21st century crises (Blaser et al., 2016).

#### AUTHOR CONTRIBUTIONS

MG has originally thought about this concept. MG and AG have written the manuscript.

#### ACKNOWLEDGMENTS

The authors are extremely grateful to the reviewers for several rounds of reviewing and suggesting critical changes to improve the manuscript. The graphical representations were hand drawn by AG and then photoshopped by Mr. C. H. Amarnath, Technical Officer (Retd.), for which authors express their sincere gratitude.





biocontrol of phytopathogens. J. Biotechnol. 235, 162–170. doi: 10.1016/j. jbiotec.2016.03.049


**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 © 2016 Gopal and Gupta. 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) or licensor 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.

# Temporally Variable Geographical Distance Effects Contribute to the Assembly of Root-Associated Fungal Communities

Christopher J. Barnes1,2 \*, Christopher J. van der Gast<sup>3</sup> , Caitlin A. Burns<sup>1</sup> , Niall P. McNamara<sup>4</sup> and Gary D. Bending<sup>1</sup>

<sup>1</sup> School of Life Sciences, Gibbet Hill Campus, University of Warwick, Coventry, UK, <sup>2</sup> Section of Evolutionary Genomics, National History Museum of Denmark, University of Copenhagen, Copenhagen, Denmark, <sup>3</sup> Natural Environment Research Council Centre for Ecology and Hydrology, Wallingford, UK, <sup>4</sup> Natural Environment Research Council Centre for Ecology and Hydrology – Lancaster Environment Centre, Lancaster, UK

Root-associated fungi are key contributors to ecosystem functioning, however, the factors which determine community assembly are still relatively poorly understood. This study simultaneously quantified the roles of geographical distance, environmental heterogeneity and time in determining root-associated fungal community composition at the local scale within a short rotation coppice (SRC) willow plantation. Culture independent molecular analyses of the root-associated fungal community suggested a strong but temporally variable effect of geographical distance among fungal communities in terms of composition at the local geographical level. Whilst these distance effects were most prevalent on October communities, soil pH had an effect on structuring of the communities throughout the sampling period. Given the temporal variation in the effects of geographical distance and the environment for shaping rootassociated fungal communities, there is clearly need for a temporal component to sampling strategies in future investigations of fungal ecology.

#### Edited by:

Martin Grube, Karl-Franzens-Universität Graz, Austria

#### Reviewed by:

Rebecca Case, University of Alberta, Canada Henrik R. Nilsson, University of Gothenburg, Sweden

> \*Correspondence: Christopher J. Barnes c.barnes@snm.ku.dk

#### Specialty section:

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

Received: 18 October 2015 Accepted: 05 December 2015 Published: 25 February 2016

#### Citation:

Barnes CJ, van der Gast CJ, Burns CA, McNamara NP and Bending GD (2016) Temporally Variable Geographical Distance Effects Contribute to the Assembly of Root-Associated Fungal Communities. Front. Microbiol. 7:195. doi: 10.3389/fmicb.2016.00195 Keywords: fungal ecology, mycorrhizal fungi, root-associated fungi, soil fungi, temporal variation in microbial communities

## INTRODUCTION

Root-associated fungi are functionally and genetically diverse, and can play an integral role in connecting the aboveground biomass to the belowground ecosystem (Berendsen et al., 2012). Rootassociated fungi can be both beneficial and detrimental for plant growth. Mycorrhizal fungi are obligate root symbionts, which can dominate the root-associated microbial biomass (Högberg and Högberg, 2002) and are associated with enhanced plant nutrient uptake (Garcia et al., 2014) and disease resistance (Chakravarty and Unestam, 1987; Liu et al., 2007). Similarly, saprophytic fungi decompose organic matter, thereby promoting nutrient availability to plants (Niklaus et al., 2001). Conversely, phytopathogenic fungal species also reside within the rhizosphere, causing disease and reducing plant growth (Espinoza et al., 2008). Given the strong and variable effect the rootassociated fungal community can have on aboveground biomass, and the potential for shifts in root-associated fungal composition to influence ecosystem function (Berg and Smalla, 2009; Smith and Read, 2010), the understanding of how these communities assemble is of significant ecological importance yet remains relatively poorly understood.

Plant hosts have been shown to play a major role in determining rhizosphere fungal communities. The main types of mycorrhizal fungal associations are generally confined to specific groups of plant (Smith and Read, 2010), and individual mycorrhizal fungal species can associate with broad or narrow ranges of plant species (Allen, 1992; Lang et al., 2010), and even genotypes within a species (Leski et al., 2010). There is also a large body of literature linking root-associated fungal community assembly with edaphic properties, with the composition of ectomycorrhizal (ECM), arbuscular mycorrhizal (AM) and nonmycorrhizal fungi all shown to be associated with specific soil parameters (Berg and Smalla, 2009). Whilst soil pH in particular is almost ubiquitously important for determining soil microbial community composition and function (Högberg et al., 2001; Toljander et al., 2006; Griffiths et al., 2011), soil nutrients such as P, K, N, and Ca have also been shown to influence root-associated fungal community structure (Jones et al., 1990; Lilleskov et al., 2002; Högberg et al., 2010; Põlme et al., 2013). The high nutrient concentrations and reduced pH that often occur following regular fertilization within agricultural soils can have particularly profound effects on root-associated fungi, leading to reduced richness and biomass (Phillips and Fahey, 2007; Verbruggen et al., 2012; Gosling et al., 2013).

The composition of root-associated fungal communities has also been shown to vary temporally, with seasonal changes in the composition of ECM (Jumpponen et al., 2010), AM fungi (Dumbrell et al., 2011), saprophytes and pathogens (Hilton et al., 2013). However, there is much uncertainty about the factors which drive temporal changes in root-associated fungal composition, and furthermore the extent to which environment and time interact to affect community composition remains to be elucidated (Last et al., 1984; Pickles et al., 2010). Abiotic factors such as climatic variables (Swaty et al., 1998; Bahram et al., 2012), and biotic factors such as plant growth and vegetation composition vary throughout the year, and these have been hypothesized to drive changes in root-associated fungal composition over time (Dumbrell et al., 2011). However, ecological drift, in which stochastic processes contribute to heterogeneous distributions of taxa in time, could also play a role in shaping changes in root-associated communities over time (Martiny et al., 2011). A major area of uncertainty is the extent to which root-associated fungi, and microbial communities in general, show true seasonality through the formation of distinct and predictable seasonal assemblages, or simply undergo ecological drift, since comprehensive studies over long-time periods have scarcely been performed.

The way in which geographical distances between sampling locations affects root-associated community assembly has become an area of increasing interest in recent years, with the magnitude of effects varying greatly between geographical scales and systems that have been sampled. For AM fungi, strong evidence of dispersal limitation was found at the continent scale, however, this disappeared at environmental extremes (Green et al., 2004; Kivlin et al., 2011). Other studies have found limited effects of geographical distance on AM community composition at the regional level (An et al., 2008; Peay et al., 2010; van der Gast et al., 2011; Hazard et al., 2013), although spatial scaling effects at the local level may have been overlooked through 'pooling' of samples from the same sampling site. Whilst the role of distance in determining ECM communities has been studied less than AM communities, previous investigations found have found evidence of spatial scaling effects within ECM communities at the local level (Lilleskov et al., 2004), with effects also shown to be much greater at larger spatial scales, with distributions of ECM fungi reflecting barriers to dispersal at the regional and global levels (Peay et al., 2010; Bahram et al., 2012; Põlme et al., 2013). Given the repeated findings of a breakdown in community similarity within fungal communities at the regional and global spatial scales (Kivlin et al., 2011), and less frequently at the local level, the ubiquitous dispersal hypothesis (Finlay and Clarke, 1999) which proposes that microbes have a cosmopolitan distribution seem unlikely to hold true. More recently Foissner described a model of 'moderate endemicity' for protists, where some species have cosmopolitan distributions, and others are dispersal limited (Foissner, 2006, 1999), a model which could prove useful in describing the spatial distribution of other microbial communities, including root-associated fungi.

In the current study we quantified the influence of geographical distance between root-associated fungal communities and soil characteristics on structure of rootassociated fungal communities over a 13-month period. Using a commercial short rotation coppice (SRC) willow plantation in the UK, roots were collected over line transects within the site, at four time points between October 2010 and October 2011. Terminal restriction length polymorphism (TRFLP) was supplemented with 454-pyrosequencing to assess variation of fungal community composition in space and time.

#### MATERIALS AND METHODS

### Study Site and Experimental Design

Biomass energy crops are monodominant-cropping systems that remain untilled throughout their lifetime, thus allowing microbial communities to develop over time in a relatively undisturbed soil matrix in a low-diversity aboveground system. This study used a SRC willow plantation at a field site near Lincoln, Lincolnshire, UK. The soil was a fine loam over clay, with approximately 15% clay, 49% sand, and 36% silt. The 30 year mean air temperature was 9.9◦C. Soil at the site has a pH gradient that ranged from 5.2 in the southern edge of the site to 6.8 in the northernmost, and has a mean total C and N content of 1.81 and 0.28% respectively. The willow was planted in 2000 at a density of 15,000 stools ha−<sup>1</sup> and covered approximately 9.44 ha. Previously the land was rotated between wheat and oilseed rape for at least 20 years before conversion to willow. Trees were planted in paired rows, with trees within rows planted roughly 0.75 m apart, whilst neighboring trees between paired rows were 1.5 m apart. The willow consisted of 6 different varieties of Salix viminalis that were planted according to commercial practice to prevent disease spread, with Tora (60%) being the most abundant (the others being Bjorn (10%), Bowles Hybrid (10%), Jorr (10%) and Jorunn (10%)). The crop was coppiced first in 2001, then in 2004, 2007, and 2010 with an average annual yield of 6.72 t ha−<sup>1</sup> .

The growing season is between March and September for SRC willow in the UK. The site received concentrated PK fertilizer (Fibrophos, UK) at 660 kg ha−<sup>1</sup> at establishment, with 20 tones of compost and lime applied to the field site in February 2010. No further exogenous nutrient input was applied during the course of the experiment.

A single row of the willow was used for line transects, which started from the southern edge of the field heading north. In order to avoid edge effects, transects started 25 m along the row, and eight locations were sampled every 20 m across the row, spanning 160 m (**Figure 1**). Subsequent transects shifted sequentially 3 m north of the previous sampling in order to avoid repeat sampling of disturbed soil. At each location, four subsamples were taken in a cross shape 1 m around a central point. These consisted of 4.5 cm diameter soil cores that were 15 cm deep (van Walt, Netherlands). Transects were taken in October 2010, and July, August, and October 2011.

#### Soil Nutrient Analysis

around the central point.

For mineral analysis, 100 g of soil was air dried for 1 week before being ground and sieved to <2 mm particle size. To measure pH, 10 ml of soil was added to 25 ml deionized water and the mixture shaken for 15 min before measurement using a pH meter (Accumet AB15, Fischer Scientific, UK). In order to measure available P, 5 g of finely ground soil was shaken for 30 min in 0.5 M NaHCO3, before analysis on an inductively coupled plasma optical emission spectrometer (ICP) (Jobin-Yvon ICP-OES, HORIBA, UK; Olsen, 1954). Available K and Mg were extracted by shaking 10 g of soil with 1 M NH4NO<sup>3</sup> for 30 min, followed by centrifugation. Concentrations of K and Mg in the supernatant were measured using ICP (Bremner and Keeney, 1965). NH4, and NO<sup>3</sup> were extracted by shaking 20 g of soil with 10.5% K2SO<sup>4</sup> for 2 h, and following centrifugation, measured in the supernatant using a FIAstar 500 flow injection analyser system (FOSS, Denmark) (Bremner and Keeney, 1965; Henriksen and Selmer-Olsen, 1970).

#### Sample Preparation

Soil was softened by soaking at 19◦C for 1 h in deionized water, before roots were hand extracted using forceps. Nonsenescent fine roots were selected by morphology (lighter color and branched structure with the presence of fine root tips). Roots less than 2 mm diameter were washed on a 6 mm sieve in order to remove adhering soil. Roots were then cut into 1 cm lengths and mixed thoroughly, before 0.5 g was taken for DNA extraction. Using the lysis tubes provided with the PowerSoil DNA isolation kit (MP Biomedicals, Cambridge, UK), roots were mechanically lysed in a TissueLyser (QAIGEN, UK) using three separate 30 s pulses at 30 Hz, before undergoing the remainder of the extraction process exactly as per manufacturers instructions.

#### TRFLP Protocol

Labeled ITS1F (5-CTTGGTCATTTAGAGGAAGTAA-3)-6FAM and ITS4 (5-TCCTCCGCTTATTGATATGC-3<sup>0</sup> )-TET primers were used to investigate root-associated fungal communities (Gardes and Bruns, 1993). PCR was performed using 25 ng of DNA from each individual subsample in a total volume of 50 µl, which included 47 µl of Megamix (Microzone, Haywards Heath, UK), 1 µl of forward and reverse primers, and 1 µl of 25 ng µl −1 template DNA from samples. The program for PCR consisted of: 5 min at 92◦C; followed by 25 cycles of 30 s at 92◦C, 90 s at 56◦C, followed by 30 s at 72◦C; a final extension of 5 min at 72◦C.

ITS amplicons from each subsample underwent digestion with Hpy8I (Fermentas, UK). 20 µl reactions contained 200 ng of DNA, 2 units of enzyme and 2 µl of x10 manufacturer's buffer before reactions were equilibrated to 20 µl using molecular grade water (MO BIO, Carlsbad, CA, USA) and incubated at 37◦C overnight. Following digestion, samples were run through sephadex columns for further purification (Sigma– Aldrich, Germany). Four micro liter of digested and purified samples were then loaded on a capillary sequencer (ABI 3010, Applied Biosystems, UK). Samples were run with GeneScan 1200 LIZ ladder (Applied Biosystems, UK).

Genemarker v1.50 (Softgenetics, State College, PA, USA) was used to quantify the number and area of the resulting TRFLP peaks, which were exported to MS Excel (Microsoft, USA) for further analysis. Baseline noise was considered to be 50 fluorescence units, with peaks lower than this removed from the analysis. Samples were normalized by the conversion of peak areas (independently for the 5<sup>0</sup> and 3<sup>0</sup> ends) to percentage relative abundance of the total fluorescence area (Hilton et al., 2013).

#### Barcoded Pyrosequencing

fmicb-07-00195 February 25, 2016 Time: 12:11 # 4

The 4 subsamples collected in October 2010 at each sampling location were equilibrated to 25 ng µl −1 and 10 µl of each was pooled to make a DNA template from each of the eight sampling locations for pyrosequencing. In order to improve comparability with the data produced via TRFLP, unlabeled ITS1F and ITS4 primers were used to generate pyrosequencing data. However, it should be noted that ITS1F and ITS4 primers show some selectivity (De Beeck et al., 2014) and produce relatively long fragments that can increase bias in community analyses (Ihrmark et al., 2012) when used in high-throughput sequencing. The initial PCR with the ITS primers was performed with MyTaq HS DNA polymerase (Bioline, USA) and consisted of: 1 µl of DNA template, 2 mM dNTPs and 10 pmol of each primer. Thermocycler conditions were 95◦C for 5 min; 40 cycles of 95◦C for 30 s, 55◦C for 30 s, 72◦C for 0.5 min; and 72◦C for 5 min, using a Biometra TJ3000 thermocycler (Biometra, Germany). A secondary semi-nested PCR reaction was performed to add the fusion primer necessary for pyrosequencing. Fusion primers consisted of: GS FLX LR70-specific adapter A, a multiplex identifier (MID), and a new forward primer, a modified version of the universal M13 primer. The fusion primers used in the secondary reaction were: forward 5<sup>0</sup> - GTGTGAAATTGTTACGCT (10 bp MID) CTTGGTCATTTAGAGGAAGTAA-3 and reverse 5 <sup>0</sup> TCCTCCGCTTATTGATATGC-3<sup>0</sup> . The forward primer comprised of the A adapter (in italic type) for the pyrosequencing reaction, the 10-bp MID is part of Roche's extended MID set (www.454.com) and the final part is the modified M13 primer. The reverse primer consists of the fusion adapter B only. The secondary PCR was performed in a volume of 25 µl, and consisted of: 1 µl of DNA template, 2 mM dNTPs and 10 pmol of each primer. Thermocycler conditions were: 94◦C for 1 min 40 s; 40 cycles of 95◦C for 20 s, 55◦C for 20 s and 72◦C for 20 s; and 95◦C for 10 min. Sample concentrations were calculated by SYBR gold based quantitation (Shmidazu, Japan), before two plates of 20 equimolar concentrations of MID tagged samples were loaded onto a Roche 454 GS Junior pyrosequencer (454 Life Sciences/Roche Applied Biosystems, Nurley, NJ, USA) and sequenced at Micropathology Ltd (Coventry, UK). Raw sequence reads were deposited at the NCBI Sequence Read Archive under the accession number **SRR1951015**.

#### Processing of Pyrosequencing Data

Sequencing data underwent denoising with Acacia-1.52 software (Bragg et al., 2012). The software package 'Quantitative insights into microbial ecology' (QIIME v1.7.0, USA) was used to perform the majority of the remaining sample processing (Caporaso et al., 2010). OTUs were picked de novo using the UCLUST algorithm at 97% similarity and from these reads a representative sequence set was created from across all samples (Edgar, 2010). Chimeric sequences were removed using the UCHIME algorithm in de novo mode (as part of USEARCH8.0; Edgar et al., 2011). Taxonomy was assigned at the 97% level using the 27.08.2013 release of the ITS fungal database for QIIME, from the UNITE project, before an OTU table was created with the taxonomic assignment and relative amplicon abundance for each sample (Wang et al., 2007). An initial 28,125 sequences formed 2,138 OTUs. Samples were subsampled to the lowest sample number, 890, before the removal of singletons, leaving approximately 592 reads per sample. In total, there were 4,738 reads spanning 320 OTUs. A second round of taxonomic assignments were performed on OTUs of greater than 1% relative abundance in which sequences underwent manual BLAST searches against the constantly updated online version of the UNITE database to increase taxonomic resolution (**Table 3**). Taxonomy was reassigned when sequence identity was 99% or greater and a DOI was added to improve data reproducibility.

#### Statistical Analyses

Relative abundance data produced via TRFLP underwent arcsine transformation in order to homogenize variation before statistical analyses were performed. To compare differences in alphadiversity, the number of TRFs [terminal restriction fragment(s)] between sampling locations and sampling time points were analyzed using repeated measures analysis of variance (ANOVA). Beta diversity was assessed by calculating distance decay-rates (DDR) across each transect as previously described (Green et al., 2004). Briefly, Bray–Curtis similarity matrices were created from the community data and Euclidean distance matrices were created from the distance between individual subsamples, using the vegdist package within R (Bray and Curtis, 1957). An exponential gradient was calculated by plotting the similarity values of the community against geographical distance, giving the distance-decay rate (using the formula S = cDddr). Differences between the rates of decay between sampling points were assessed using the t-distribution method (Fowler et al., 1998).

For direct ordination, community data underwent arcsine transformations before undergoing canonical correspondence analysis (CCA) with integrated forward selection (Canoco v5.0, Wageningen, Netherlands; Braak and ter Šmilauer, 2002). In order to incorporate distance between sampling locations within the CCA, principle coordinates of neighbor matrices (PCNM) were calculated from grid coordinates for each subsample and the initial first PCNM was used as an explanatory variable (using the Vegan package of R). Forward interactive selection was performed to obtain significantly correlating explanatory variables, including soil properties (pH, C, N, NO3, NH4, Mg, P, and K) and geographical separation (as PCNM1) against the ordination analysis of the community, whilst limiting the effects of multicolinearity (van den Wollenberg, 1977). The total variation explained in the ordination analysis as well as the variation explained by each explanatory variable was calculated during the analysis.

To investigate temporal variation in community composition, the TRFLP datasets were used to generate a single Bray–Curtis matrix. A non-metric multidimensional scaling (nMDS) analysis was performed and used to visualize community similarity between the different seasons. In order to partition and quantify the temporal variance explained by and time (in weeks after

sampling started), the ADONIS function in the Vegan package of R, using the combined community Bray–Curtis matrix and sampling time points as the main factor (either October 2010, July 2011, August 2011, or October 2011) was used. As this analysis is sensitive to the order explanatory variables are analyzed, individual ADONIS analyses were performed for each month (either October, July, August) and time (in weeks after sampling started) and placed in order of largest proportion of variation explained before a combined analysis was performed.

To compare community profiles produced using TRFLP and 454 pyrosequencing approaches, using the October 2010 dataset a paired t-test was performed between the average number of TRF at each sampling location and the number of OTU determined using pyrosequencing. In addition a DDR was calculated for the pyrosequencing dataset as described above, and this was compared to the DDR curve produced using TRFLP data, using the t-distribution method, as used previously.

#### RESULTS

#### Edaphic Properties

There were strong gradients in pH, available P and available K across transects, which were conserved across all seasons (**Figures 2A–C**), with P (r = 0.687, P < 0.001) and K (r = 0.573, P < 0.001) inversely correlating with pH. pH ranged from 5.2 to 6.8 across the 160 m transect in all seasons, while available P varied from 80 mg kg−<sup>1</sup> to 35 mg kg−<sup>1</sup> across the transect in October 2010, and from 40 to 20 mg kg−<sup>1</sup> in the following seasons. Similarly available K was present at over 300 mg kg−<sup>1</sup> in transect sites 1–3 and at approximately 200 mg kg−<sup>1</sup> in sites 5–8 across all time points. The available Mg concentration did not significantly change over the length of transects or over time (**Figure 2D**). Both NO<sup>3</sup> (P < 0.001) and NH<sup>4</sup> demonstrated significant (P < 0.001) variation over time, with highest concentrations in the October 2010 and July 2011 transects respectively (**Figures 2E,F** respectively). Whilst NH<sup>4</sup> ranged from 1.59 to 9.59 mg kg−<sup>1</sup> , no significant differences were found between sampling locations, NO<sup>3</sup> ranged from 0.54 to 8.60 and significantly varied between sampling locations (P < 0.001).

#### Fungal Community Analysis

From the root-associated fungal TRFLP analysis, 273 different TRFs were identified over the four seasons, with TRF richness ranging between 22.0 and 83.8 across sampling locations. TRF richness significantly varied between sampling locations (P = 0.023), with site 4 (60 m into the transect) consistently having a lower TRF richness than the other sites. The number of TRFs did not differ between sampling times (P = 0.058).

Distance decay-rates were calculated to quantify the breakdown in community similarity of the root-associated fungi across each transect. All seasons showed some decline in similarity over transects, with the DDR varying throughout the sampling period, from −0.136 (P = 0.001) in October 2010, to −0.171 (P = 0.001), −0.017 (P = 0.001) and −0.115 (P = 0.001) in July 2011, August 2011 and October 2011 respectively (**Figure 3**). The August 2011 transect had a significantly lower rate of decay across the transect than all other sampling points (P = 0.001, P = 0.001 and P = 0.001 for October 2010, July 2011, and October 2011 respectively), whilst none of these others were significantly different to each other (Supplementary Table S1).

#### Investigating Distance and Edaphic Effects on the Root-Associated Fungal Community

Canonical correspondence analyses were performed on the TRFLP data with integrated forward selection of explanatory variables (**Table 1**). Soil pH was associated with composition in every season except October 2011 (October 2010 P = 0.002, July 2011 and P = 0.002, August 2011 P = 0.01), accounting for between 7.1 and 9.2% of the variation in the root-associated fungal community. The geographical separation (as PCNM1) across the transect significantly correlated with community variation in both October 2010 (P = 0.002) and October 2011 (P = 0.002), accounting for 4.8 and 8.4% of community variation.

### Investigating Temporal Changes in the Structure of the Root-Associated Fungal Community

An initial nMDS was performed in order to visualize changes in community composition over time (**Figure 4**). In order to test for significance and quantify temporal changes in community variation, ADONIS analyses were performed to analyse the variance between month of sampling (July, August, and October) and time (in weeks after sampling). Both the month of sampling and time independently accounted for change within the community (**Table 2**), with month of sampling accounting for 7.16% of change with the root-associated fungi (P = 0.014) and time 1.53% (P = 0.014).

#### 454-Pyrosequencing Assessment of the October 2010 Community

The pooled DNA stocks for each of the eight sampling locations of the October 2010 transect underwent pyrosequencing to assess the diversity of root-associated fungi. All reads that could be assigned to a phylum were from the Basidiomycota and Ascomycota, with Basidiomycota the dominant fungal phylum, averaging 68.02% of reads and Ascomycota 6.90% of reads from October 2010. Basidiomycota were also more diverse than the Ascomycota, with an average of 41.75 and 9.38 OTUs respectively across the October 2010 community.

OTUs were assigned to the highest taxonomic classification and investigated further. The OTUs consisted of a few dominant and many rare OTUs, with only 15 OTUs having an average abundance greater than 1% (**Table 3**). Reads assigned to a Sebacinales OTU were the most abundant, accounting for 15.00% average abundance, whilst a Cortinarius diasemospermus OTU was the second most abundant accounting for a further 10.75% average community abundance. Whilst 4 of the 10 most abundant OTUs were assigned within the Cortinarius genus, a further 2 were assigned within the Sebacinales order and another 2 to the Thelephoraceae. Literature suggests

FIGURE 2 | Average of the four subsamples per sampling location of (A) pH, (B) available P (mg kg−<sup>1</sup> ), (C) available K (mg kg−<sup>1</sup> ), (D) available Mg (mg kg−<sup>1</sup> ), (E) NO<sup>3</sup> (mg kg−<sup>1</sup> ), and (F) NH4(mg kg−<sup>1</sup> ) across the transects. Error bars are ± 1 SEM.

FIGURE 3 | Distance-decay of fungal community similarity over the (A) October 2010 (B) July 2011, (C) August 2011 and (D) October 2011 transects. Plotted are the Bray–Curtis similarity values against geographical distance for all paired sampling combinations, with the DDR calculated for each transect using the formula S = cDddr .

TABLE 1 | Canonical correspondence analysis determining the variation of the rhizosphere fungal communities from TRFLP of each transect explained by metadata parameters.


Abbreviations: Var. Exp. (%), the percentage of rhizosphere fungal community variation explained by a parameter given by the canonical correspondence analysis.

that up to 10 of the 15 highly abundant OTUs could be ectomycorrhizal association forming species whilst the Thelephoraceae may also fulfill saprotrophic roles (Matheny et al., 2006; Hibbett, 2007). Reads were also assigned to other potential saprotrophs including Exophiala equina and Didymella exigua, whilst D. exigua may also be pathogenic along with the Seimatosporium obtusum OTU (Tanaka et al., 2011; Tedersoo et al., 2014).

There was clear evidence of spatial variation within the rootassociated fungal communities. Whilst Basidiomycota was the largest phylum within all samples, relative abundance ranged from a high of 87.1% within the community at 60 m across the transect, to a low of 40.1% within the site 100 m across (**Figure 5**). Ascomycota remained stable in relative abundance, ranging from 1.7 to 5.8% between sites 0–120 m. However, the Ascomycota accounted for 25.7% of the community at 140 m

TABLE 2 | ADONIS analysis of intra-annual (seasonal) and inter-annual (time in weeks after sampling started) variation within the root-associated fungi of the SRC willow between October 2010 and October 2011.


across the transect, largely driven by a 10.8% increase in reads assigned to a putative pathogen, Seimatosporium, which was absent in nearly all other samples. OTU richness also varied across transects, with the Basidiomycota ranging from 66 OTUs at the site 40m across the transect to 27 OTUs at 0m. Additionally, the Ascomycota ranged from just 4 OTUs at 40 m across the transect to 14 at 140 m. OTU richness within the Basidiomycota strongly strongly correlated (R <sup>2</sup> = 0.736, P = 0.037) with relative abundance but the Ascomycota did not (R <sup>2</sup> = 0.625, P = 0.097).

Average taxa richness produced via TRFLP for each sampling location was compared against OTU richness produced via 454 pyrosequencing (**Table 4**). Whilst 454-pyrosequencing produced significantly greater alpha diversity estimates for each sampling location (t = −5.161, P = 0.001), the trend in relative fungal richness was maintained between the two techniques, with a significant correlation of 0.767 (P < 0.05). Additionally, a DDR of −0.290 (P = 0.005) was calculated for the October 2010 rootassociated fungal community analyzed with 454-pyrosequencing, and whilst this was higher than the DDR of −0.136 produced via TRFLP, there was no significant difference between the rates produced by differing techniques (P = 0.090).

### DISCUSSION

In this work, we show that both geographical distance and environmental variability (pH) simultaneously affect rootassociated fungal community composition, with the magnitude of their influences being variable over time. A number of studies have found evidence for independent geographical distance effects on the composition of root-associated fungal communities (Green et al., 2004; Bahram et al., 2012; Põlme et al., 2013; Davison et al., 2015), whilst other studies have found no such effects (Tedersoo et al., 2012; Hazard et al., 2013; Mundra et al., 2015). However, here we suggest that distance effects plays a variable role in role in shaping the root-associated fungal community at the local level throughout time, whilst more temporal replicates are required to confirm if these effects are consistently greater later in the growing season within willow SRC plantations. CCAs revealed that 4.8% of community variation in October 2010 and 8.4% in October 2011 could be explained by geographical separation. Whilst CCAs revealed that the geographical separation between samples had a significant effect on the community compositions of the October 2010 and October 2011 transects, beta-diversity was significantly higher in October 2010, July 2011, and October 2011 than August 2011. Although the gradient in pH and soil nutrients affected the turnover of species over transects, these properties remained stable throughout the sampling period, suggesting factors other than edaphic properties affected the spatial turnover of species. Whilst this work was performed within a managed ecosystem, these findings are in agreement with a growing number of studies performed within unmanaged systems that suggest that the ubiquitous dispersal hypothesis does not hold true for fungi (Green et al., 2004; Põlme et al., 2013; Horn et al., 2014). Interestingly, studies which have not found distance effects on root-associated fungal community composition at the local and regional level have investigated this relationship within single sampling time points, and consequently may have missed the 'window' in time that geographical separation has an effect (Hazard et al., 2013; Põlme et al., 2013).

Our results suggest that the root-associated fungal community and the parameters that regulate community assembly vary throughout the year. Whilst effects of geographical distance on community assembly were variable throughout the year, in contrast pH played a near ubiquitous role in affecting community composition throughout the sampling period. When the temporal variation in the root-associated fungal community was analyzed, intra-annual variation (month of sampling) rather than inter-annual variation (time in weeks after sampling started) had the largest temporal effect on the community variation, suggesting the two October sampling points shared similarity to each other, and were distinct from the July and August assemblages. Time did, however, also independently affect root-associated fungal composition,


For ease of reading only those that have a greater than 1.00% relative abundance are listed with their taxonomic assignment, accession numbers and DOI.

indicating that longer-term processes such as ecological drift may also be occurring. Whilst there has been a number of previous studies demonstrating seasonal community dynamics, these have mostly been performed over single growing seasons and have not enabled year-on-year comparisons, therefore have been unable to isolate seasonal effects from other more longterm temporal processes (Jumpponen et al., 2010; Dumbrell et al., 2011; Martiny et al., 2011). In this work there was only 1 yearon-year comparison, therefore results should be interpreted cautiously. However, since the communities of October 2010 and 2011 shared the most similarity, yet were still significantly different, this suggests considerable complexity in the temporal variation of root-associated fungal communities, even within a relatively simple aboveground system. Whilst community shifts have been shown between sites differing in age by multiple years (Husband et al., 2002; Fujiyoshi et al., 2011; Blaalid et al., 2012), there is a need for further studies which define the relationships among inter-annual and intra-annual variations in fungal community assembly.

As within this study, ECM communities were found to vary globally with pH, but also with mean annual precipitation (Tedersoo et al., 2014). Changes in the composition of rootassociated fungal communities over time could reflect either a direct response to annual variation in environmental parameters (such as precipitation) or indirect effects such as changing patterns of C supply from the plant. For example mature Pinus sylvestris was shown to increase plant derived carbon belowground by 500% later in the growing season compared to early growing season (Högberg et al., 2010), whilst the number of fine root tips of Salix viminalis has specifically been

TABLE 4 | Comparison of TRF richness produced via TRFLP and OTU richness produced by 454-pyrosequencing for the October 2010 root-associated fungal communities.


Brackets indicate alpha diversity site rank from highest to lowest fungal OTU richness.

shown to increase throughout the growing season (Rytter and Hansson, 1996). Compositional shifts have also been observed in belowground microbial communities when belowground carbon allocation is increased (Phillips et al., 2002), suggesting changes in belowground plant growth dynamics throughout the year, in response to changing environmental conditions, may drive relatively short-term differences in fungal assemblages and their regulation found within this study. However, there are also a number of mechanisms that may not of been easily detected within the relatively short 13-month sampling period that may effect the root-associated fungal community. Longterm transitions in abiotic conditions, host age (Last et al., 1984; Visser, 1995; Husband et al., 2002), changing aboveground biomass composition (Blaalid et al., 2012) or even ecological succession series have been hypothesized to effect microbial assemblages over time (Deacon and Fleming, 1992; Bergemann and Miller, 2002) and there is a clear need for more studies contrasting inter and intra-annual variations within microbial communities.

454-pyrosequencing provides insightful taxonomic information about fungal composition and was used to confirm that a diverse root-associated fungal community resides within the willow monoculture. The PCR amplification with ITS1F and ITS4 primers successfully amplified a broad range of fungi and has been effectively used to detect Glomeromycota, Zygomycota, and Chytridiomycota in addition to Ascomycota and Basidiomycota within root-associated fungal communities (Yu et al., 2012), although only Ascomycota and Basidiomycota were detected within this study. These did, however, include ectomycorrhizal fungi, and a range of saprotrophs, endophytes and pathogenic fungi. Whilst many of the most abundant reads were assigned to probable ECM species, interestingly no reads were assigned to the Glomeromycota. Salix sp. can form ECM or AM associations depending on genotype and soil environment (Becklin et al., 2012; Corredor et al., 2012, 2014). Furthermore under some circumstances, AM inhibition by ECM fungi has been shown in laboratory experiments using willow, and this maybe replicated within willow plantations (Becklin et al., 2012).

TRFLP suffers from some well-documented limitations (Avis et al., 2006), therefore the pyrosequencing data from the October 2010 transect was also used to provide insight into the effectiveness of TRFLP to profile fungal communities. Whilst the fungal taxa richness produced via TRFLP was significantly lower than OTU richness produced via pyrosequencing, the trend in fungal richness and DDR across sampling locations with the two technologies was well conserved. This is in agreement with other studies that suggest microbial community profiles produced between TRFLP and pyrosequencing are reproducible (Pilloni et al., 2012). Pyrosequencing too suffers a number of limitations and the subsequent bioinformatics analyses can strongly impact taxa assignment, thus there is a great need to standardize high-throughput sequencing analyses in order to improve comparability between studies using the same as well as different technologies that profile soil communities (Tedersoo et al., 2015).

Given the importance of root-associated fungi in ecosystem functioning, understanding of the factors that regulate community assembly in environmental systems remain relatively unknown. Here we suggest that the spatial scaling of root-associated fungal communities does not follow the ubiquitous dispersal model, even at the local level, and the magnitude of this spatial scaling effect varies throughout the year. Future studies of root-associated fungal community assembly should therefore not underestimate the potential of distance effects occurring at the local geographic scale, and would also benefit from multiple sampling time points to fully characterize variation within ecosystems.

#### AUTHOR CONTRIBUTIONS

CB, main author of the manuscript, data producer and analysis. CvdG, assisted with statistical analyses and manuscript production. CAB, assisted in laboratory work and bioinformatics. NM, assisted in logistics, access to field sites and biogeochemical analyses. GB, main supervisor, obtained funding and initial experimental design.

#### ACKNOWLEDGMENTS

We thank Cahyo Prayogo and Lawrence Davies for assistance during sample collection and preparation. We'd also like to thank Dr. Steven Hanley (Rothamsted Research) for willow genotype analysis. This project was funded by the Natural Environment Research Council (NERC) as part of the wider Carbo-Biocrop project. For the Lincoln commercial plantation we acknowledge NERC Centre for Ecology and Hydrology National Capability funding through project NEC03487, Jon Finch and the landowner, Jonathan Wright.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb.2016. 00195

#### REFERENCES

fmicb-07-00195 February 25, 2016 Time: 12:11 # 11



practice. Environ. Microbiol. 13, 241–249. doi: 10.1111/j.1462-2920.2010. 02326.x


by 454 pyrosequencing. Plant Soil 358, 225–233. doi: 10.1007/s11104-012- 1188-5

**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 © 2016 Barnes, van der Gast, Burns, McNamara and Bending. 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) or licensor 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-07-01030 June 25, 2016 Time: 12:42 # 1

# Arbuscular Mycorrhizal Fungus Species Dependency Governs Better Plant Physiological Characteristics and Leaf Quality of Mulberry (Morus alba L.) Seedlings

Song-Mei Shi<sup>1</sup>† , Ke Chen<sup>1</sup>† , Yuan Gao<sup>1</sup> , Bei Liu<sup>1</sup> , Xiao-Hong Yang1,2 \*, Xian-Zhi Huang<sup>3</sup> , Gui-Xi Liu<sup>1</sup> , Li-Quan Zhu<sup>1</sup> and Xin-Hua He2,4 \*

<sup>1</sup> Key Laboratory of Horticulture Science for Southern Mountainous Regions, Ministry of Education/College of Horticulture and Landscape Architecture, Southwest University, Chongqing, China, <sup>2</sup> Centre of Excellence for Soil Biology, College of Resources and Environment, Southwest University, Chongqing, China, <sup>3</sup> State Key Laboratory of Silkworm Genome Biology, Southwest University, Chongqing, China, <sup>4</sup> School of Plant Biology, University of Western Australia, Crawley, WA, Australia

#### Edited by:

Suhelen Egan, The University of New South Wales, Australia

Reviewed by:

Ilana Kolodkin-Gal, Weizmann Institute of Science, Israel Shengguo Zhao, Chinese Academy of Agricultural Sciences, China

#### \*Correspondence:

Xiao-Hong Yang yangxh2@swu.edu.cn; Xin-Hua He xinhua.he@uwa.edu.au †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: 15 December 2015 Accepted: 17 June 2016 Published: 28 June 2016

#### Citation:

Shi S-M, Chen K, Gao Y, Liu B, Yang X-H, Huang X-Z, Liu G-X, Zhu L-Q and He X-H (2016) Arbuscular Mycorrhizal Fungus Species Dependency Governs Better Plant Physiological Characteristics and Leaf Quality of Mulberry (Morus alba L.) Seedlings. Front. Microbiol. 7:1030. doi: 10.3389/fmicb.2016.01030 Understanding the synergic interactions between arbuscular mycorrhizal fungi (AMF) and its host mulberry (Morus alba L.), an important perennial multipurpose plant, has theoretical and practical significance in mulberry plantation, silkworm cultivation, and relevant textile industry. In a greenhouse study, we compared functional distinctions of three genetically different AMF species (Acaulospora scrobiculata, Funneliformis mosseae, and Rhizophagus intraradices) on physiological and growth characteristics as well as leaf quality of 6-month-old mulberry seedlings. Results showed that mulberry was AMF-species dependent, and AMF colonization significantly increased shoot height and taproot length, stem base and taproot diameter, leaf and fibrous root numbers, and shoot and root biomass production. Meanwhile, leaf chlorophyll a or b and carotenoid concentrations, net photosynthetic rate, transpiration rate and stomatal conductance were generally significantly greater, while intercellular CO<sup>2</sup> concentration was significantly lower in AMF-inoculated seedlings than in non-AMFinoculated counterparts. These trends were also generally true for leaf moisture, total nitrogen, all essential amino acids, histidine, proline, soluble protein, sugar, and fatty acid as they were significantly increased under mycorrhization. Among these three tested AMFs, significantly greater effects of AMF on above-mentioned mulberry physiological and growth characteristics ranked as F. mosseae > A. scrobiculata > R. intraradices, whilst on mulberry leaf quality (e.g., nutraceutical values) for better silkworm growth as F. mosseae ≈ A. scrobiculata > R. intraradices. In conclusion, our results showed that greater mulberry biomass production, and nutritional quality varied with AMF species or was AMF-species dependent. Such improvements were mainly attributed to AMF-induced positive alterations of mulberry leaf photosynthetic pigments, net photosynthetic rate, transpiration rate, and N-containing compounds (methionine, threonine, histidine, and proline). As a result, application of Funneliformis mosseae or A. scrobiculata in mulberry plantation could be a promising management strategy to promote silkworm cultivation and relevant textile industry.

Keywords: Acaulospra scrobiculata, amino acids, Funneliformis mosseae, Rhizophagus intraradices, net leaf photosynthesis rate, transpiration

### INTRODUCTION

fmicb-07-01030 June 25, 2016 Time: 12:42 # 2

As a multipurpose plant, mulberry (Morus alba L.) has become an increasingly attractive plant for its economic value and restoration function. It has been extensively cultivated in China for the main purpose of rearing silkworm (Bombyx mori L.), an economically important insect as the primary producer of silk, and thus contributing to local farming and textile industries. The quality of mulberry leaves strongly associates with the quality and quantity of cocoon, especially their high contents of carbon, nitrogen (N), amino acids and proteins that contribute greatly to the silkworm development (Machii and Katagiri, 1991; Sujathamma and Dandin, 2000; Rao et al., 2007). Although mulberry can survive in poor soil (Kumar et al., 2003; Huang et al., 2013), but its growth is retarded under infertile soil. A number of biotic or abiotic factors, such as soil beneficial organisms and fertilization, etc., do enhance the growth and leaf quality of mulberry (Jeffries et al., 2003; Setua et al., 2007; Chowdhury et al., 2013; Fernandez et al., 2014).

The symbiotic associations formed by arbuscular mycorrhizal fungi (AMF) are widely associated with >80% vascular plants including mulberry in terrestrial ecosystems (Smith and Read, 2008). Rajagopal et al. (1989) was probably the first to report that mulberry roots were highly colonized by AMF in natural conditions, and phosphorus (P) uptake was hence increased to lead a better leaf growth and biomass production of mulberry (Katiyar et al., 1995; Setua et al., 1999a,b). Both the growth and yield of 10-year-old mulberry plants were similar between fertilizations under 100% P and 50% P + Glomus (now Funneliformis) mosseae inoculation (Mamatha et al., 2002). Recently, for 90-day-old AMF inoculated mulberry seedlings, either Funneliformis mosseae or Glomus (now Rhizophagus) intraradices had enhanced leaf numbers, plant height, chlorophyll, N and P, both aboveground and belowground biomass production, root length and root TTC (2,3,5-triphenyl tetrazolium chloride) deoxidizing ability, although there was no such growth benefits by their dual fungal inoculation (Lu et al., 2015). Meanwhile, the combined application of mycorrhizal fungi and other beneficial microorganisms could also promote the growth of mulberry. For instance, the co-inoculation of two AMF species (Glomus fasciculatum and F. mosseae) with other three beneficial microorganisms (N2-fixing Azotobacter sp., phosphate solubilizing bacterium Bacillus megaterium and fungus Aspergillus awamori) did increase the uptake of N, P, K (potassium) and leaf chlorophyll in 1-year-old mulberry plants in India (Baqual and Das, 2006). In addition, a combined positive effect of F. mosseae and bacterial biofertilizer (Azotobacter chrococcum) on mulberry leaf quality and cocoon characters was also found under semiarid fields in India (Rao et al., 2007). However, limited information is available about the influence of different AMF species either within the same fungal genus or different genera on the quantity and quality of mulberry leaves, which in turn will directly affect the growth and development of silkworm, the production of cocoon and the quality of silk products. Therefore, further studies will determine if different AMF species in the same genus or different genera have distinctive roles in the growth and leaf quality of mulberry.

We have found that a high diversity of AMF, including Acaulospra scrobiculata, Funneliformis mosseae, and Rhizophagus intraradices, associates with field-growing mulberry plants, and that such an individual AM fungus promotes the growth of mulberry (Shu et al., 2011; Shi et al., 2013) and soil fertility (Chen et al., 2014) in southwest China. In this study, we hypothesized that the growth of mulberry and quality of mulberry leaves could have varied responses to different AMF species or their physiological characteristics of mulberry might be AMFspecies dependent. Our objectives were to address the following questions: (1) If AM fungus could successfully colonize the host mulberry, and what physiological alterations could the host have after colonization? (2) Whether the growth and performance of mulberry depend on its symbiont fungal species characteristics or not? In doing so, 3-month-old mulberry seedlings grown in a greenhouse were inoculated with or without an individual AMF species (A. scrobiculata, F. mosseae, and R. intraradices) from three different AM genera, and some of their basic physiological characteristics were then compared after a further 3-month-old growth. Answers to these questions could improve our understanding of the physiological responses of mulberry to different AM fungi in order to employ better functionally effective AM species for promoting sustainable mulberry plantation and silkworm industry in China.

#### MATERIALS AND METHODS

#### Experimental Design

The greenhouse experiment set-up was a randomized complete block design that had three AMF inoculation treatments (Acaulospra scrobiculata, Funneliformis mosseae, or Rhizophagus intraradices) and a non-AMF control with four replicates for each treatment.

### Plants, Mycorrhizal Inoculums, and Plant Growth Media

Seeds of mulberry (Morus alba var. Gui-you-sang 12, supplied by the Institute of Sericulture and Systems Biology, Southwest University, Chongqing, China) were disinfected with 0.1% HgCl<sup>2</sup> for 5 min, rinsed thoroughly with sterilized water and then germinated on sterilized moistened filter paper. The germinated seeds were cultivated for 1 month in plastic trays, which contained autoclaved (121◦C, 0.1 MPa, 120 min) sand growth media (peat: decomposed rice chaff: sand = 2:2:1, v/v/v.

Three AMF inoculums [A. scrobiculata (BGC HK02A), Funneliformis mosseae (BGC NM04A) and R. intraradices (BGC AH01)] were purchased from the Bank of Glomales in China (BGC), locating in the Institute of Plant Nutrition and Resources, Beijing Academy of Agriculture and Forestry. Inoculums consisted of spores (40–50 spores per 10 g dry soil), mycorrhizal mycelia, root segments, and sand. The growth medium was a mixture of soil (Eutric Regosol, FAO Soil Taxonomic Classification), peat and rice chaff (2:1:1, v/v/v). With a pH 7.28, this soil growth medium had 71.17 g organic matter/kg, 373.12 mg available nitrogen/kg, 14.72 mg available phosphorus/kg, 132.57 mg available potassium/kg.

#### Plant Growth and Harvest

fmicb-07-01030 June 25, 2016 Time: 12:42 # 3

Four seedbeds containing the above-mentioned soil growth medium were used to grow mycorrhizal mulberry seedlings. Each seedbed (28 m × 1.2 m) was fumigated with 0.5% formaldehyde and divided into four blocks or replicates (6 m × 1.2 m with 1 m interblock space) for each treatment. A total of 600 1 month-old non-mycorrhizal mulberry seedlings from the sand growth media were transplanted and grown in seedbeds with the autoclaved soil growth media. After 2 months, about 400 uniform seedlings (100 for each block or replicate) were chosen for mycorrhizal inoculation with 3 g AMF inoculum (autoclaved for the control) close to each root rhizosphere. Plants were then grown in a greenhouse under 26/15◦C (day/night), 70–80% relative humidity and natural light intensity. No-fertilization was employed for a further 3-month period till harvest (i.e., the total growth period were 6 months).

After 3 months of mycorrhizal inoculation, 20 uniform seedlings out of the 100 seedlings from each replicate were randomly harvested and a total of eighty seedlings were thus harvested from each treatment to determine growth and physiological parameters. At harvest, plant height, stem diameter, leaf numbers, taproot length and diameter, and fibrous root numbers were recorded. Leaves, shoots and roots were then ovendried at 70◦C and weighed. The fresh fifth fully expanded leaves from the top to the bottom of each seedling were harvested, and a total of 80 leaves from each treatment were divided into two sample groups. One sample group of leaves was cut into pieces and mixed for the determination of leaf chlorophyll and carotenoid, soluble protein and sugar (see below). Another sample group of leaves was oven-dried at 70◦C till constant weight, pulverized to fine powder for the determination of amino acid, total nitrogen, and fatty acid (see below). The leaf moisture was recorded according to the method described by Gao (2000).

#### Determination of Mycorrhizal Colonization and Dependency

Root mycorrhizal colonization was measured according to Phillips and Hayman (1970). The fresh roots were cut into 2 cm segments and cleared with 10% (w/v) KOH at 98◦C for 30 min, rinsed in water three times, and bleached with 10% H2O<sup>2</sup> until root clear. The clear root segments were then soaked into 0.2 M HCl for 3 min and stained with 0.05% trypan blue. Five stained segments were mounted on one slide and a total of 50 root segments from each treatment were examined under microscope. The root mycorrhizal colonization was determined as the percent (%) of infected root segments out of the total observed root segments. Mycorrhizal dependency (MD) was calculated as the dry weight of mycorrhizal seedling out of the dry weight of non-mycorrhizal seedling (Menge and Johnson, 1978).

#### Photosynthesis Measurements

The fifth leaf from the top to the bottom of each plant growing in the greenhouse was selected for the measurements of photosynthesis variables. Gas exchange parameters (including net photosynthetic rate, stomatal conductance, intercellular CO<sup>2</sup> concentration, and transpiration rate) were determined with 20 seedlings from each replicate using a portable infrared gas analyzer LI-COR 6400 (LI-COR, Inc., Lincoln, NE, USA) from 09:00 to 11:00 am on a sunny day after 3 months of mycorrhizal inoculation. The photosynthetically active radiation was 1000 ± 12 µmol m−<sup>2</sup> s −1 , CO<sup>2</sup> concentration 350 ± 2 cm<sup>3</sup> m−<sup>3</sup> , leaf temperature 28.0 ± 0.8◦C, and flow rate of atmosphere 0.5 dm<sup>3</sup> min−<sup>1</sup> . Meanwhile, the chlorophyll and carotenoid in the fresh leaves were extracted with 80% acetone and the absorbance was read spectrophotometrically at 663 and 645 nm. Chlorophyll and carotenoid concentration was estimated by the method of Hiscox and Israelstam (1979).

#### Determination of Leaf Soluble Sugar, Soluble Protein, Amino Acid, Fatty Acid, and Total Nitrogen

Determination of soluble sugar (mg g−<sup>1</sup> FW) in fresh leaves was followed by the anthrone method (Yemm and Willis, 1954). Briefly, 0.5 g fresh leaves were grinded with 80% ethanol and then centrifuged at 3,500 × g for 10 min. The supernatant was collected. The total soluble sugar was determined by reacting 0.1 ml supernatant with 5 ml freshly prepared anthrone sulphuric acid solution (75%, v/v), and incubated in boiling water for 10 min. After cooled, the absorbance of the incubated supernatant was spectrophotometrically read at 620 nm.

Soluble proteins (mg g−<sup>1</sup> FW) of fresh leaves were extracted with 50 mM potassium phosphate buffer (pH 7.5) on ice. The extracts were centrifuged at 12,000 × g for 10 min at 4◦C. The supernatant was then collected for soluble protein determination according to the protein dye-binding method of Bradford (1976). For each extract, the absorbance was spectrophotometrically read at 595 nm and the bovine serum albumin (BSA) protein was used as the standard.

For the measurement of leaf total amino acids (mg g−<sup>1</sup> DW), 230 mg leaf powder was put into a beaker containing 8 ml 6 M HCl, and hydrolyzed at 110◦C for 22 h in thermostat drier. After cooled, the solution was filtered and the filtrate was transferred into a volumetric flask and dried at 60◦C. The dried hydrolysate was dissolved in 0.02 M HCl and then analyzed for total amino acids with the ninhydrin method (Lee and Takahashi, 1966) by an automatic amino acid analyzer (L-8900; Hitachi, Tokyo, Japan).

Leaf fatty acid (mg g−<sup>1</sup> DW) was based on the Chinese Standard Method (GB5009.6-85). Briefly, 3 g dried leaf powder was placed in a filter cartridge that was dried at 105◦C for 2 h. The cartridge was placed in a 50 mL anhydrous ether and refluxed 12 h using a Soxhlet apparatus. After the ether was fully evaporated from the extract, the cartridge containing the extract was oven-dried at 105◦C, cooled in a desiccator and then weighed.

The leaf nitrogen was determined with the Kjeldahl procedure. Briefly, 0.5 g leaf powder was placed in a digestion tube containing 10 ml digestion reagent and then dissolved in concentrated H2SO4. After gently mixed, the digestion tube was placed in the heating block preheated at 300◦C. A small glass funnel was inserted in the mouth of the digestion tube for ensuring efficient refluxing of the digestion mixture and preventing loss of H2SO4. The sample was digested at the boiling point of the mixture for 2.5 h, removed from the heating block, and cooled to room temperature. The digest was diluted with distilled water to 100 ml and performed a steam distillation analysis as described by Nelson and Sommers (1973).

All relevant variables were analyzed with a UV1000 spectrophotometer and all analyses were performed three times.

#### Statistical Analyses

fmicb-07-01030 June 25, 2016 Time: 12:42 # 4

Data were subjected to one-way ANOVA and significant differences among treatments were tested by the Duncan's multiple range test at P < 0.05 using the SPSS 19.0 (SPSS Inc., Chicago, IL, USA). To test correlations, a polynomial regression analysis was performed using the OriginPro 8.0 (OriginLab Corp., Northampton, MA, USA).

#### RESULTS

#### Effects of AMF Colonization on Plant Growth

Root AMF colonization existed only in AMF inoculated mulberry plants, and neither AMF colonization nor AMF structures were microscopically observed in the control seedlings (**Figure 1**). Percentages of root colonization in mulberry seedlings were over 50% and significantly highest with F. mosseae, greater with A. scrobiculata and least with R. intraradices (**Figure 1**). The mycelial numbers were similar among the three mycorrhizal inoculations, while the vesicle numbers were lower under the R. intraradices treatment (**Figure 1**). Correspondingly, the growth performance of mulberry seedlings was affected by mycorrhizal fungal colonization (**Figure 2**). Compared with non-mycorrhizal seedlings, all tested plant growth variables, including aboveground height and taproot length, stem base and taproot diameter, shoot and root biomass, leaf and fibrous root numbers, were significantly increased by a range of 31 to 121% in AMF inoculated plants (**Figure 3**). In short, significantly greater plant growth performance, as well as MD, generally ranked as F. mosseae > A. scrobiculata > R. intraradices > non-AMF control (**Figure 3**).

### Effects of AMF Colonization on Photosynthetic Variables

Concentrations of leaf photosynthetic pigments (chlorophyll a and b, and carotenoid) were significantly higher in all mycorrhizal plants than in non-mycorrhizal plants (**Figure 4**). Among AMF plants, the higher concentration order of tested photosynthetic pigments ranked as F. mosseae > A. scrobiculata ≈ R. intraradices for leaf chlorophyll a, leaf chlorophyll b and carotenoid, but under F. mosseae > A. scrobiculata > R. intraradices for leaf chlorophyll (a+b) (**Figure 4**).

Meanwhile, AMF colonization significantly enhanced leaf net photosynthetic rate by 52–99%, stomatal conductance by 24–56%, transpiration rate by 66–104%, while significantly decreased intercellular CO<sup>2</sup> concentration by 5–15%, compared with non-AMF mulberry plants (**Figure 5**). Among

FIGURE 1 | Root arbuscular mycorrhizal fungal colonization in 6-month-old mulberry seedlings. Fifty mycorrhizal root segments from each treatment were examined under a microscope: (a) non-AMF treatment; (b) A. scrobiculata, Acaulospora scrobiculata; (c) F. mosseae, Funneliformis mosseae; (d) R. intraradices, Rhizophagus intraradices.

AMF colonized mulberry plants, significantly greater leaf photosynthetic variables, ranked as F. mosseae > A. scrobiculata > R. intraradices for the net photosynthetic rate, R. intraradices > A. scrobiculata > F. mosseae for the stomatal conductance, F. mosseae ≈ R. intraradices > A. scrobiculata for the transpiration rate, and R. intraradices > A. scrobiculata > F. mosseae for the intercellular CO<sup>2</sup> concentrations (**Figure 5**). fmicb-07-01030 June 25, 2016 Time: 12:42 # 5

FIGURE 3 | Effects of mycorrhizal fungal species on growth performance of mulberry plants. Data (means ± SE, n = 4) followed by different letters indicate significant differences among AMF treatments at P < 0.05. Eighty plants from each treatment were randomly selected to detect growth performance. Abbreviations: Control, non-AMF treatment; A. scrobiculata, Acaulospora scrobiculata; F. mosseae, Funneliformis mosseae; R. intraradices, Rhizophagus intraradices.

treatment; A. scrobiculata, Acaulospora scrobiculata; F. mosseae, Funneliformis mosseae; R. intraradices, Rhizophagus intraradices.

fmicb-07-01030 June 25, 2016 Time: 12:42 # 6

scrobiculata; F. mosseae, Funneliformis mosseae; R. intraradices, Rhizophagus intraradices.

### Effects of AMF Colonization on Leaf Quality Variables

Leaf quality including amino acids, total N, soluble sugar and protein, fatty acid of mulberry plants were significantly improved by AMF colonization (**Table 1**; **Figure 6**). Significantly higher concentrations of leaf individual or subtotal essential amino acids, histidine, proline, and total amino acids ranked as F. mosseae ≥ A. scrobiculata > R. intraradices (**Table 1**). Significantly higher leaf total N concentrations and soluble protein among AMF colonized mulberry plants ranked as F. mosseae > A. scrobiculata > R. intraradices (**Figure 6**). Significantly higher leaf soluble sugar among AMF colonized mulberry plants ranked as A. scrobiculata > R. intraradices ≈ F. mosseae, while significantly higher leaf fatty acid concentrations ranked as A. scrobiculata > F. mosseae > R. intraradices (**Figure 6**). In addition, significantly higher leaf moisture among AMF colonized mulberry plants patterned as F. mosseae (80.56 ± 0.36%) > A. scrobiculata (79.11 ± 0.62%) > R. intraradices (76.19 ± 0.41%).

Correlations between leaf biomass production and leaf quality were showed in **Figure 7**. Leaf biomass production was significantly related to leaf total nitrogen, soluble protein or soluble sugar (R <sup>2</sup> = 0.94–0.98, P < 0.05) (**Figures 7A,C,D**). Meanwhile, leaf total nitrogen concentrations were significantly related to soluble protein or soluble sugar (R <sup>2</sup> = 0.78–0.93, P < 0.05) (**Figures 7E,G**). In contrast, either leaf biomass or leaf total nitrogen was not significantly related to total amino acids (R <sup>2</sup> = 0.75–0.78, P > 0.05, **Figures 7B,F**).

### DISCUSSION

### AMF Significantly Promotes Growth of Mulberry Plants

In this study, the three AMF species showed different effects on growth of shoot and root in mulberry. Overall F. mosseae colonized mulberry seedlings showed the best potential and consistently performed better in respect to plant growth characteristics than other two AMF species of A. scrobiculata and R. intraradices (**Figures 2** and **3**). However, Lu et al. (2015) found that R. intraradices was more efficient than F. mosseae in colonizing mulberry roots under greenhouse conditions. Meanwhile, R. fasciculatus was more efficient than Glomus etunicatum (now Claroideoglomus etunicatum) and F. mosseae


 (means SE, 4) by significant by Abbreviations; His, Histidine; Ile, Isoleucine; Leu, Leucine; Lys, Met, Methionine; Phe, Phenylalanine; Pro, Proline; Thr, Threonine; Subtotal, total essential amino acid; Val, Valine.

to infect mulberry roots under rainfed lateritic soil conditions (Setua et al., 1999b). As a result, such differences of AMF functioning might be related to AMF species and abiotic factors. In general, the successfully competitive establishment of AMF colonization not only depended on the indigenous mycorrhizal species, but also on the introduced species (Sharma et al., 2005) and its placement in the soil (Hepper et al., 1988). It is wellknown that AMF could facilitated mulberry P uptake (Setua et al., 1999a,b) and other macro- and micro-elements uptake (Baqual and Das, 2006; Lu et al., 2015). The uptake function could reduce 75% cost of phosphate fertilization under low soil P levels (Katiyar et al., 1995). Among nine AMF tested (Acaulospora laevis, G. mosseae, Gigaspora margarita, Glomus caledonicum, G. fasciculatum, G. leptotichum, G. macrocarpum, R. intraradices, and Scutellospora calospora), G. leptotichum is the best AM symbiont for 50-day-old nursery teak (Tectona grandis) seedlings in terms of P, Zn (zinc), and Cu (copper) nutrition and plant growth (Rajan et al., 2000). The improved nutritional status in mycorrhizal plants contributed to an enhanced plant biomass production (Treseder, 2013; Taylor et al., 2014), resistances of root disease (Al-Askar and Rashad, 2010), salt (Kashyap et al., 2004) and drought (Tang et al., 2013). Moreover, an enhanced root system by mycorrhization could greatly increase the absorption surface and thus nutrient uptake capacity (**Figure 3**). As a result, both AMF species and cultivation management should be taken into account in mulberry plantations.

#### AMF Significantly Promotes the Photosynthetic Capacity of Mulberry Leaf Photosynthetic Pigments Are Increased by AMF and Vary with AMF Species

Significantly higher leaf chlorophyll and carotenoid, and photosynthetic rate displayed in the three AMF mulberry plants, and F. mosseae was the best symbiotic associate with the host mulberry (**Figures 1** and **4**). Higher leaf chlorophyll with higher photosynthetic rates under AM associations made it possible to have higher carbon fixation and carbohydrate accumulation (**Figures 4–6**). Concurrently, AMF colonization also enhanced significantly higher carotenoid accumulations, leading to better leaf nutraceutical values of mulberry plants. Leaf carotenoid of mulberry not only acted as a growth-promoting factor for the silkworm larvae (Shimizu et al., 1981), but also played an important role in the larvae's and adult's phototactic response and visual sensitivity (Shimizu and Kato, 1978). Meanwhile, the pigmentation development and brilliant color of cocoons were mainly depended on the carotenoid concentration in the mulberry leaves (Tabunoki et al., 2004). Studies also showed that AMF colonization could result in higher concentrations of both leaf chlorophyll and carotenoid in lettuces (Baslam et al., 2013). As a result, the application of AMF species could benefit not only the growth of mulberry, but also the growth and quality of silkworms.

#### Leaf Gas Exchange Capacity Is Enhanced by AMF

The increment of photosynthetic pigments might be due to an effective synergism between transpiration and photosynthesis under mycorrhization (Sheng et al., 2008; Borde et al., 2011).

fmicb-07-01030 June 25, 2016 Time: 12:42 # 7

fmicb-07-01030 June 25, 2016 Time: 12:42 # 8

fmicb-07-01030 June 25, 2016 Time: 12:42 # 9

In the present study, photosynthesis, transpiration, and stomatal conductance were significantly greater but intercellular CO<sup>2</sup> was significantly lower under AMF-inoculated than under non-AMF inoculated mulberry plants. These results from mulberry were consistent with those from maize and alfalfa (Medicago sativa L.), regardless of water and salt stress conditions (Zhu et al., 2012; Campanelli et al., 2013), and even under karst rocky desertification area (Chen et al., 2014). The greater stomatal conductance of AM plants implied a lower leaf resistance to moisture diffusion, and ultimately leading to a faster water transportation (Talaat and Shawky, 2014). In addition, an increased leaf area of AMF colonized plants was concomitant with an increase in photosynthetic rate and transpiration rate (**Figure 2B** vs. **Figure 5**). Meanwhile, a potential increase of leaf surfaces with an increase of leaf numbers under mycorrhization could promote plant sunlight capture and thus better photosynthetic production (Adolfsson et al., 2015).

#### AMF Significantly Enhances Leaf Quality of Mulberry

#### AMF Inoculation Accelerates Carbohydrate Formation of Mulberry Plants

Bombyx mori L. requires essential amino acids, proteins, sugars, and fatty acid from mulberry leaves for its normal growth and silk production (Sengupta et al., 1972). We observed that an increase of leaf soluble sugar in AMF inoculated mulberry plants with higher photosynthetic rates. This observation was consistent with results from other studies, i.e., the inoculation with F. mosseae and R. intraradices did stimulate soluble sugar production in Poncirus trifoliate L. (Wu et al., 2011), Cyclobalanopsis glauca (Zhang et al., 2014), and Jatropha curcas L. (Kumar et al., 2015). Meanwhile, when leaf moisture was improved and stomas opened wider in mycorrhizal mulberry leaves, greater CO<sup>2</sup> fixation was then enhanced and carbohydrate accumulation was ultimately increased (Augé et al., 1987). In addition, soluble sugar is a precursor of carotenoids, and an enhanced soluble sugar production could hence increase the carotenoid production (**Figures 4** and **6**; Baslam et al., 2011). Moreover, significantly greater ingestion, digestibility and consumption index were observed in silkworms fed with 80–85% high moisture leaves than with 55–60% low moisture leaves (Rahmathulla et al., 2004). As a result, a higher range of 76–81% leaf moisture under AMF inoculation certified that AMF colonization could improve leaf nutritional quality of mulberry plants. Consequently, the palatability of silkworm to mulberry leaves could be also improved.

#### AMF Inoculation Accelerates Nitrogen Metabolism of Mulberry Plants

The quantity and quality of cocoon shell closely related with concentrations of mulberry leaf N and amino acids, especially methionine, histidine, and threonine. For instance, methionine and histidine are essential for the growth of silkworms, and threonine is crucial for the synthesis of silk protein (Machii and Katagiri, 1991). Our results showed that AMF colonization generally significantly increased total nitrogen, soluble protein, total amino acid, and all seven essential amino acids including methionine, histidine, and threonine, particularly under F. mosseae and A. scrobiculata (**Figure 6**; **Table 1**). Pentón et al. (2014) reported that inoculation with Glomus cubense improved soil N extraction capacity of mulberry and complemented with chemical fertilization. Singh et al. (2010) showed that leaf protein and total amino acid in tea tree were increased by AMF inoculation with an indigenous AMF consortia containing nine AMF species from three genera of Acaulospora, Funneliformis, and Glomus. Govindarajulu et al. (2005) revealed that inorganic N taken up by the AM fungus was incorporated into amino acids and N transportation was from extraradical mycelium to intraradical mycelium and then to plants, which would be a great boost to the nutrition composition and growth of host plants (Azcón et al., 1992). Azcón et al. (1996) also claimed that mycorrhizal colonization increased activity of nitrate reductase and glutamine synthetase involving in N assimilation of the host plant. Therefore, further studies on nitrogen, perhaps also phosphorus nutrition, are guaranteed if mulberry plants could positively response to genetic different AMF species by improving the growth and quality of mulberry hosts.

### CONCLUSION

Our results showed that inoculation with AM fungi had improved mulberry leaf biomass production and nutritional quality through an enhanced photosynthesis and growth performance. A significantly higher rank of such positive responses of mulberry to AMF species was as follows: Funneliformis mosseae > A. scrobiculata > R. intraradices for plant's physiological and growth characteristics, and F. mosseae ≈ A. scrobiculata > R. intraradices for leaf quality. Such improvements were relevant to the AMF-induced alterations of leaf carbohydrates and N-containing compounds, essential amino acids and soluble protein in particular. As a result, application of Funneliformis mosseae or A. scrobiculata in mulberry plantation could be a promising management strategy to promote silkworm cultivation and relevant textile industry in southwest China.

### AUTHOR CONTRIBUTIONS

X-HY and X-HH conceived and designed the experiments; S-MS, KC, YG, and BL performed the experiments. X-ZH was responsible for the field experiment and provided fund support. G-XL analyzed amino acid data with the help of L-QZ. S-MS and KC wrote the paper. All authors approved the final manuscript.

## ACKNOWLEDGMENTS

This study was supported by the earmarked fund for Modern Agro-industry Technology Research System (CARS-22-ZJ0503) and a Postgraduate Education Teaching Reform Project in Chongqing (yjg143028), China.

## REFERENCES

fmicb-07-01030 June 25, 2016 Time: 12:42 # 10


the hydro-fluctuation belt of Three Gorges Reservoir Areas. Environ. Sci. Pollut. Res. int. 20, 7103–7111. doi: 10.1007/s11356-012-1395-x


alba L.) under rainfed, lateritic soil conditions. Biol. Fertil. Soils 29, 98–103. doi: 10.1007/s003740050531


**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 © 2016 Shi, Chen, Gao, Liu, Yang, Huang, Liu, Zhu and He. 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) or licensor 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-07-01030 June 25, 2016 Time: 12:42 # 11

# Microbiota Influences Morphology and Reproduction of the Brown Alga Ectocarpus sp.

Javier E. Tapia1,2, Bernardo González<sup>3</sup> , Sophie Goulitquer<sup>4</sup> , Philippe Potin<sup>5</sup> and Juan A. Correa1,2 \*

<sup>1</sup> CNRS, Université Pierre-et-Marie-Curie, UMI 3614, Biology and Ecology of Algae, Station Biologique de Roscoff, Roscoff, France, <sup>2</sup> Departamento de Ecología, Facultad de Ciencias Biológicas, Pontificia Universidad Católica de Chile, Santiago, Chile, <sup>3</sup> Facultad de Ingeniería y Ciencias, Universidad Adolfo Ibáñez – Center of Applied Ecology and Sustainability, Santiago, Chile, <sup>4</sup> MetaboMer Mass Spectrometry Core Facility, Université Pierre-et-Marie-Curie, CNRS, FR2424, Station Biologique de Roscoff, Roscoff, France, <sup>5</sup> Université Pierre-et-Marie-Curie, CNRS UMR 8227, Integrative Biology of Marine Models, Station Biologique de Roscoff, Roscoff, France

Associated microbiota play crucial roles in health and disease of higher organisms. For macroalgae, some associated bacteria exert beneficial effects on nutrition, morphogenesis and growth. However, current knowledge on macroalgae–microbiota interactions is mostly based on studies on green and red seaweeds. In this study, we report that when cultured under axenic conditions, the filamentous brown algal model Ectocarpus sp. loses its branched morphology and grows with a small ball-like appearance. Nine strains of periphytic bacteria isolated from Ectocarpus sp. unialgal cultures were identified by 16S rRNA sequencing, and assessed for their effect on morphology, reproduction and the metabolites secreted by axenic Ectocarpus sp. Six of these isolates restored morphology and reproduction features of axenic Ectocarpus sp. Bacteria-algae co-culture supernatants, but not the supernatant of the corresponding bacterium growing alone, also recovered morphology and reproduction of the alga. Furthermore, colonization of axenic Ectocarpus sp. with a single bacterial isolate impacted significantly the metabolites released by the alga. These results show that the branched typical morphology and the individuals produced by Ectocarpus sp. are strongly dependent on the presence of bacteria, while the bacterial effect on the algal exometabolome profile reflects the impact of bacteria on the whole physiology of this alga.

Keywords: microbiota, bacteria–algae interaction, Ectocarpus, bacterial isolate, algal morphology, exometabolome

## INTRODUCTION

Plants and animals are associated with their microbiota, a complex assortment of microorganisms. As example, in the human gut, bacteria play a major role in stimulating immune system development (Lee and Mazmanian, 2010; Littman and Pamer, 2011). Recently, the communication between gut microbiota and the central nervous system has been established (Mayer, 2011), along with the emerging concept of a microbiota-gut-brain axis (Cryan and Dinan, 2012). Similarly, plant roots are colonized by a large diversity of soil microorganisms which are capable of producing

#### Edited by:

Martin Grube, Karl-Franzens-Universität Graz, Austria

#### Reviewed by:

Ilana Kolodkin-Gal, Weizmann Institute of Science, Israel Benoit Chassaing, Georgia State University, USA

> \*Correspondence: Juan A. Correa jcorrea@bio.puc.cl

#### Specialty section:

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

Received: 16 October 2015 Accepted: 05 February 2016 Published: 24 February 2016

#### Citation:

Tapia JE, González B, Goulitquer S, Potin P and Correa JA (2016) Microbiota Influences Morphology and Reproduction of the Brown Alga Ectocarpus sp. Front. Microbiol. 7:197. doi: 10.3389/fmicb.2016.00197

**291**

beneficial (although sometimes negative, pathogenic, and presumably mostly neutral) effects on the plant (Newton et al., 2010). Plant growth-promoting bacteria (PGPB) stimulate growth by increasing photosynthetic capacity (Zhang et al., 2008), increasing tolerance to abiotic stress (Yang et al., 2009), by suppressing plant diseases (Choudhary and Johri, 2009; Van der Ent et al., 2009) and herbivory by insects (Van Oosten et al., 2008), among several other relatively poorly understood mechanisms/functions/processes.

On comparative grounds, plant and animal-bacteria interactions have received more attention than other macroorganisms-microorganisms interactions. In aquatic environments, microorganisms are quite abundant. It is estimated that, on average, one milliliter of seawater contains more than 10<sup>6</sup> bacteria (Harder, 2009). In addition, marine environments favor formation of biofilms on diverse surfaces, including those of macroalgae (Weinberger, 2007), and other marine macroorganisms (Qian et al., 2007).

In this context, it is known that seaweeds interact with marine microorganisms throughout their life cycle (Goecke et al., 2012). The microbial communities inhabiting macroalgae are highly complex, dynamic and are constituted by a variety of microorganisms where bacteria are better described in terms of their diversity and function (Corre and Prieur, 1990; Burke et al., 2011a,b). In this interaction, macroalgae represent an excellent environment for bacterial colonization and reproduction by providing nutrients and a suitable surface for attachment (Armstrong et al., 2000; Singh and Reddy, 2014). The advantages for the algal host have been also described during recent years. Bacteria can mineralize organic substrates giving the algae carbon dioxide, minerals and growth factors (Matsuo et al., 2005). Other studies have shown that marine bacteria produce nitrogen compounds that are a source of nutrients for algae. For example, the nitrogen supply of Caulerpa taxifolia is provided by an endophytic bacteria from the Agrobacterium-Rhizobium group, which lives in the rhizoids of this algae (Chisholm et al., 1996).

In addition to the nutritional benefits, it has been shown that the presence of certain bacteria is needed for normal morphological development and growth of some green (Matsuo et al., 2003; Marshall et al., 2006; Spoerner et al., 2012) and red macroalgae (Singh et al., 2011; Fukui et al., 2014). Moreover, associated bacteria are known to induce settlement of zoospores of Ulva species and release of spores from Acrochaetium sp. (Joint et al., 2007; Weinberger et al., 2007).

The above information has been obtained mainly based on studies using green and red algal species, leaving aside the important group of brown algae. The Phaeophycean taxa is one of the more diverse groups of macroalgae (Andersen, 2004) and possesses significant ecological roles in coastal ecosystems (Cock et al., 2011).

Brown algae are phylogenetically distant not only from terrestrial plants, animals and fungi, but also from red and green algae (Baldauf, 2003). Indeed, they differ in many aspects of their biology with respect to the other algal groups. Some of these differences correspond to: composition and pathways of cell wall synthesis (Nyvall et al., 2003), their ability to synthesize C18 and C20 oxylipins (Ritter et al., 2008), in their ability to accumulate iodine (Kupper et al., 2008), among several others. Bacteria have been described living in association with brown algae (Hengst et al., 2010; Lachnit et al., 2011), and there are some early observations linking bacterial presence with normal development and growth of these organisms (Pedersen, 1968).

In order to elucidate basic aspects of the biology of brown algae, a small species with a filamentous structure, Ectocarpus siliculosus, has been chosen as a model (Peters et al., 2004). Several molecular tools and databases are now available for this algae including its complete genome sequence (Cock et al., 2010), genetic maps (Heesch et al., 2010), transcriptomics (Le Bail et al., 2008a; Dittami et al., 2009) and proteomics (Contreras et al., 2008) approaches. Despite a diverse array of studies addressing its life cycle (Coelho et al., 2011a,b; Arun et al., 2013), acclimation to biotic and abiotic stress (Dittami et al., 2011; Grenville-Briggs et al., 2011), morphological development (Le Bail et al., 2010, 2011) and genetic diversity on the field (Peters et al., 2010), to date there is limited knowledge on the interactions between Ectocarpus and its associated microbiota. Recently, Dittami et al. (2015) described how Ectocarpus associated bacteria are essential for acclimation to salinity gradients, showing the importance of these microorganisms to the alga under stress conditions. More than 40 years ago, Pedersen (1968) also reported a potential role for bacteria in the development of members of this algal genus. She described that axenic cultures of E. fasciculatus showed slow growth and atypical development when kept under sterile, axenic conditions, suggesting an influence of bacteria for the normal growth and development of these algae.

A more detailed evaluation of the role of bacteria on brown algae development and physiology is clearly required in order to establish and understand the influence of these microorganisms and the mechanisms involved in this interaction. The present study describes the isolation of bacteria and the evaluation of their role as regulators of morphology and reproduction of the brown algal model Ectocarpus sp. [strain Ec32 formerly referred as E. siliculosus (Peters et al., 2010)]. The effects of bacterial inoculation and bacterial exudates were determined, and proved to be essential in shaping the development and reproduction of this algal model. The impact of bacterial presence on the metabolites secreted by the alga, as an approach to understand the bacterial influence on the general metabolism of the host (Macel et al., 2010; Goulitquer et al., 2012), was also assessed. The result of this approach revealed that colonization of axenic Ectocarpus sp. with single bacterial species drives a major impact in the algal exometabolome profile, highlighting the effect of bacteria on the whole physiology of this alga.

### MATERIALS AND METHODS

#### Culture of Axenic Ectocarpus sp.

The experiments were carried out using the axenic laboratory cultures of haploid Ectocarpus sp. parthenosporophyte isolate Ec 32 (Culture Collection of Algae and Protozoa accession no. 1310/4; origin, San Juan de Marcona, Perú), which was produced by germination of unfertilized gametes (Le Bail et al., 2008b). Axenization of algal individuals was carried out according to

Müller et al. (2008). Briefly, small Ectocarpus fragments were placed around antibiotic disks on Zobell medium. Four weeks later, algal fragments from bacteria-free areas were taken and put into Petri dishes with sterilized natural seawater. After another 4 weeks, some of the fragments were put on Zobell medium to check for bacterial growth while others were checked for bacterial presence by microscopy. Fragments from bacterial-free algal material were then transferred to Petri dishes with SFC culture medium (Correa and McLachlan, 1991) for growth and experimentation. Individuals were grown in 12-mL Petri dishes in sterile-pasteurized SFC medium in a controlled-environment cabinet at 13◦C with a 12:12-h light:dark cycle (light intensity of 30 µmol photons m−<sup>2</sup> s −1 ). All growth treatments were performed according to these conditions.

#### Axenicity Controls

In order to check for axenicity and cross-contamination, the following approaches were used:

(1) Visualization of bacteria on Ectocarpus surface at the beginning and the end of each treatment. Algal individuals were washed twice with sterile seawater and then exposed for 10 min to sterile seawater containing 0.22 µm filter-sterilized SYBR Green II. Observations were performed with an Olympus BX60 (Tokyo, Japan) epifluorescence microscope. See results of this approach (**Figures 1C,D** and **4D** and Supplementary Figure S1).

(2) DNA extraction from the treatment supernatants, PCR amplification of the 16S rRNA gene and AluIII restriction of the amplicons obtained. With this method it was possible to check bacterial presence (positive amplification) and also if the bacterial treatments were contaminated with other bacteria, by looking at the digestion profiles of the amplicons (see results of this approach in Supplementary Figure S1). 16S rRNA gen amplification and amplicon digestion procedures were also performed with DNAs from the bacterial isolates in order to compare the digestion profiles with those obtained from supernatants at the beginning and at the end of each treatment.

(3) To add supernatant of the treatments or Ectocarpus individuals that were exposed to bacterial isolates, to bacterial culture media (Zobell broth) and observe the growth of microorganisms after 2 weeks. See results of this approach in (Supplementary Figure S2).

### Seawater Microorganisms Effect on Axenic Ectocarpus sp.

In order to evaluate the influence of microorganisms on Ectocarpus sp. SFC medium using natural seawater (SW) coming from two different places: Caleta Maitencillo (32◦ 390 S, 71◦ 29<sup>0</sup> W) and Las Cruces (33◦ 300 S, 71◦ 37<sup>0</sup> W) were prepared. Seawater from Las Cruces was obtained in two different seasons, summer and winter. The SW was filtered using a 3 µm pore size filter (Merck Millipore, Darmstadt, Germany) so bacteria, some unicellular fungi and fungal spores were still present. To check for the presence of bacteria, 10 µL of filtered SW were plated on Zobell agar and after 3 days of incubation at 20◦C microbial growth was clearly observed. To assess the effect of microbes containing SFC medium, four axenic Ectocarpus sp. individuals per plate (three plates) were exposed to 12 mL of this medium. Each experiment was performed in triplicate using these three different media. The spores produced that were settled and germinated after 7 days, were counted (30 random observations in a 1 mm<sup>2</sup> area each). After 3 weeks, the percentage of individuals with upright filaments was determined (50 individuals per replicate). Observations were made in a Nikon Optiphot-2 microscope. To test the presence of Ectocarpus spores, running controls of filtered SW culture medium without the alga were performed along with each experiment. No Ectocarpus individuals were detected in any of these controls.

### Isolation of Bacteria from Ectocarpus Individuals Maintained in Unialgal Cultures

Bacteria were isolated from the surfaces of Ectocarpus unialgal strains Ec 32 (mentioned above) and Ec 524 (Culture Collection of Algae and Protozoa accession 1310/333, origin Caleta Palito, Chile 26◦ 150 S, 70◦ 140W). Both strains were maintained under laboratory conditions as described above and always displayed a filamentous morphology. To isolate bacteria, small algal pieces were gently washed twice in sterile seawater, then grinded and spread on three different marine agar media: marine broth (Zobell) supplemented with 1.5% agar; sterile natural seawater, obtained by filtration and pasteurization, supplemented with 1.5% agar; and seawater R2A agar (Suzuki et al., 1997). The dishes were incubated at 20◦C for 10 days and individual colonies were picked off and streaked onto the agar from which they were isolated in order to obtain single colonies. Bacterial isolates were maintained at 4◦C while they were used, stocks were passed to −70◦C in glycerol to conserve them.

### Identification of Bacterial Isolates by 16S rRNA Genes Sequencing

DNA from bacterial isolates was obtained using the PureLink <sup>R</sup> Genomic DNA Mini Kit (Life Technologies, Carlsbad, CA, USA) according to the manufacturer's instructions. PCR amplification of partial 16S rRNA gene sequences were carried out using the forward primer 8f (5<sup>0</sup> -AGATTTGATCCTGGCTCAG-3<sup>0</sup> ) and the reverse primer 1492r (5<sup>0</sup> -GGTTACCTTGTTACGACTT-3<sup>0</sup> ) (Weisburg et al., 1991). Sequencing was carried out at Macrogen Inc. (Seoul, Korea). A search for 16S rRNA similarities of sequences from isolated bacteria was made with the BLAST tool available online<sup>1</sup> . 16S rRNA gene sequences of bacterial isolates have been deposited at GenBank under accession numbers provided in Supplementary Table S1.

### Screening the Effects of Bacteria on Axenic Ectocarpus sp. Morphology and Reproduction

Axenic Ectocarpus sp. individuals were exposed to 12 mL of pasteurized SFC medium (four individuals per plate). Pasteurized medium, 95◦C for 30 min followed by 90 min at 72◦C, was preferred over autoclaved medium to avoid some salt precipitation during the sterilization process. Bacterial isolates

<sup>1</sup>http://blast.ncbi.nlm.nih.gov/Blast.cgi

were used in a density of approximately 10<sup>7</sup> cells per milliliter in pasteurized SFC medium. Control (individuals without bacteria) and treatment plates were incubated according to the conditions mentioned above. Growth medium was replaced by sterile fresh material every 7 days. To determine the effect on morphology, individuals grown for 21 days after germination were evaluated according to the presence or absence of upright filaments (50 individuals per analysis, the analysis was repeated three times). To evaluate reproduction, the number of individuals produced 6 weeks after germination was counted in 30 random observations in a 1 mm<sup>2</sup> area each (10 observations per plate). Observations were made in a Nikon Optiphot-2 microscope. Experiments were performed in three replicates for each of the nine isolates tested.

When evaluating the presence or absence of upright filaments, 20 random individuals were chosen to analyze filament and elongated cells sizes. Three cells per individual were evaluated. Cell sizes were measured with the ImageJ software<sup>2</sup> .

### Effect of Bacterial Growth Culture Supernatants and Bacteria-Ectocarpus Co-culture Supernatants on Axenic Ectocarpus Morphology and Reproduction

To obtain bacterial growth culture supernatants, each bacterial isolate was cultivated in 50 mL of sterile SFC medium in a 500 mL flask supplemented with 1% (w/v) glucose until they reached a density of approximately 10<sup>7</sup> cells per milliliter in a shaker at 15◦C. Bacterial cultures were centrifuged (30 min, 5000 × g) and the supernatant was filtered twice through 0.22 µm pore size filters (Merck Millipore, Darmstadt, Germany). The supernatants were used immediately. The experimental cultures media were refreshed every week using fresh bacterial supernatant.

To obtain bacteria-Ectocarpus co-culture supernatants, 1 week old media from direct bacterium inoculation treatments were used. Media from bacteria-Ectocarpus co-cultures were centrifuged and filtered the same way as bacterial supernatants. The obtained co-culture supernatants (approximately 12 mL) were directly exposed to axenic Ectocarpus, as previously mentioned. The effect on morphology and reproduction was evaluated as indicated in the corresponding section above. The experimental cultures were refreshed every week using 1-week old co-cultures supernatants. Supernatants from 1-week old axenic Ectocarpus cultures along with bacterial and algal culture media were used as controls. In order to check that bacterial supernatants and 1-week old co-cultures supernatants were not depleted of essential nutrients to sustain Ectocarpus growth, we placed individuals from unialgal and axenic cultures under these conditions and we compared them with their growth under starvation stressing conditions: natural seawater (NSW) without addition of any supplementary nutrient. While individuals from unialgal cultures developed normally, axenic Ectocarpus had an arrested growth and did not develop upright filaments and did not produce any new individuals (Supplementary Figure S3).

#### Analysis of the Exometabolome

Exudate extracts were obtained by Solid Phase Extraction. Triplicates of 200 mL culture medium from axenic Ectocarpus sp. and bacterial isolate Z3 growing together plus exudate from both but growing alone were slowly passed through C18 cartridges (Sep Pak 6 mL, 1 g, Waters, Saint-Quentin en Yvelines, France) using an automated Dionex AUTO Trace 280 instrument (Thermo Fisher Scientific, Bremen, Germany). After washing with 5 mL of deionized water, the Sep Pak cartridges were dried under a nitrogen flux and then eluted in glass vials with 4 mL dichloromethane, followed with 4 mL methanol.

Ultra-high pressure liquid chromatography analysis of these extracts was performed using an RSLC Ultimate 3000 from Dionex (Thermo Fisher Scientific, Bremen, Germany) equipped with a quaternary pump and autosampler. Separations were achieved using an Acclaim RSLC 120 C18 1.9 µm (2.1 mm × 100 mm) column (Dionex) operated at 20◦C, using 5 µL injection volume and a flow-rate of 250 µl min−<sup>1</sup> . Mobile phase A was composed of 0.1% acetic acid in MiliQ H2O, and mobile phase B was 0.1% acetic acid in acetonitrile. The gradient consisted of an initial hold at 20% mobile phase B for 2 min, followed by a linear gradient to 100% B in 8 min and a hold for 14 min, followed by re-equilibration for 6 min at 20% B, in a total run time of 30 min.

Mass spectrometry was performed using a LTQ-Orbitrap DiscoveryTM mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). Scans were collected in both positive and negative ESI mode over a range of m/z 50–1000. Ionization parameters were set as follows: sheath gas 5 psi, auxiliary gas 5 (arbitrary units), sweep gas 0 (arbitrary units), spray voltage 2.7 kV, capillary temperature 300◦C, capillary voltage 60 V, tube lens voltage 127 V and heater temperature 300◦C. The Xcalibur 2.1 software (Thermo Fisher Scientific) was used for instrument control and data acquisition. Following their acquisition, metabolomic fingerprints were deconvoluted to allow the conversion of the three-dimensional raw data (m/z, retention time, ion current) to time- and mass-aligned chromatographic peaks with associated peak areas. Massmatrix File Conversion tools were used to transform the original Xcalibur data files (<sup>∗</sup> .raw) to a more exchangeable format ( ∗ .mzXML). Raw files were converted to the mzXML format using MassMatrix File Conversion Tools (Version 3.9, April 2011). Data were processed by the open-source XCMS software (Smith et al., 2006) running under R or on the online version, and further annotated by CAMERA<sup>3</sup> .

#### Statistical Analysis

Data for number of spores produced and percentage of germinated individuals in sterile and non-sterile tests were compared using a two-sample t-test run on Minitab software version 16.1. Asterisk (<sup>∗</sup> ) indicates differences on at least 5% level of significance (p < 0.05). Data for number of spores produced in alga-bacteria co-cultures were compared using a two-sample t-test run on Minitab software version 16.1. Different letters were used to indicate means that differ

<sup>2</sup>http://rsb.info.nih.gov/ij/

<sup>3</sup>http://camera.calit2.net/

significantly (p < 0.05). For experiments addressing the effect of bacterial growth culture supernatants and bacteria-Ectocarpus co-culture supernatants on axenic Ectocarpus morphology and reproduction, univariate and multivariate analyses with a Tukey's post hoc test, run on Minitab software version 16.1, were used for testing differences in individuals produced between treatments. Different letters were used to indicate means that differ significantly (p < 0.05). Multivariate statistical analyses of metabolite data were carried out using SIMCA-P (12.0.1, Umetrics, Umeå, Sweden). Data were log10-transformed and normalized using Pareto scaling. Principal Component Analysis (PCA) was carried to compare the intensity of mass/retention time pairs between the chromatograms.

#### RESULTS

#### Differences in Ectocarpus sp. Morphology Growing in Unialgal or Axenic Culture Conditions

Ectocarpus sp. strain Ec32 in unialgal culture conditions displays its typical branched morphology (**Figure 1A**). When cultured under axenic conditions, Ectocarpus sp. shows a small ball-like appearance (**Figure 1B**), which differs from its branched natural morphology. The only difference between these two culture conditions is that

FIGURE 1 | Morphological differences between Ectocarpus sp. individuals growing under axenic and unialgal culture conditions. (A,B) Examples of 2 month-old Ectocarpus sp. individuals growing in sterilized SFC medium. (A) Ectocarpus sp. individual showing its characteristic regular branched morphology in a unialgal culture. (B) Axenic Ectocarpus sp. individuals showing the atypical "small ball-like" appearance. (C,D) Detection of bacteria on Ectocarpus sp. surface (red, chloroplasts autofluorescence), by epifluorescence microscopy visualization using green-yellow, SYBR green II staining. (C) Bacteria detected on the filament surface of an Ectocarpus sp. individual grown in a unialgal culture. (D) Absence of bacteria on the surface of an axenic Ectocarpus sp. individual. All bars, 30 µm.

the individuals in unialgal culture conditions still possess normally associated bacteria (**Figure 1C**) while under axenic conditions, individuals were previously treated with antibiotics to remove the associated microbiota (**Figure 1D**). Thus, bacterial absence in Ectocarpus cultures produces abnormal algal development.

### Effect of Seawater Microorganisms on Morphology and Reproduction of Ectocarpus sp.

In order to test whether microorganisms affect morphology and reproduction of Ectocarpus sp. a first approach was to expose axenic individuals to the presence or absence of microorganisms. Culture media prepared with surface seawater samples taken from two different coastal places were evaluated. First, these seawater samples were filtered, first, using a 3 µm pore size filter to prepare culture media still containing microorganisms, to perform the non-sterile tests. Then, the same seawater samples were filtered again using a 0.22 µm pore size filter and pasteurized in order to obtain the appropriate culture media to perform the sterile tests. After 1 week, axenic individuals proliferating in culture media containing microorganisms produced more algal spores than those individuals grown on sterile culture media (**Figure 2A**). The settled spores average counted in non-sterile tests was 23 per cm<sup>2</sup> in contrast to an average of four settled spores per cm<sup>2</sup> found under sterile conditions. Considering the spores already germinated (more than two cells), differences were also significant as in non-sterile tests the number of germinated spores per cm<sup>2</sup> was seven times higher than those found in sterile tests. The percentage of germinated spores relative to the number of observed spores was also significantly higher when seawater microbes were present (**Figure 2B**). This indicates that bacterial presence notably improves Ectocarpus spore production and germination.

A morphological trait of Ectocarpus sp. (presence of upright filaments, the basis for the branched morphology) was also determined. The percentage of individuals with upright filaments after 3 weeks of growth under non-sterile or sterile conditions was calculated. In this case, 54% of the individuals grown under non-sterile conditions had already developed upright filaments whereas none individuals had developed these structures in sterile conditions (data not shown). This observation stresses the importance of bacteria for proper Ectocarpus development.

#### Effect of Bacterial Isolates on Ectocarpus sp. Morphology and Reproduction

Bacteria have a remarkable influence on the algal morphogenesis. After 21 days, upright filaments became visible in unialgal cultures (**Figure 3A**) but not in axenic Ectocarpussp. (**Figure 3B**). In order to check for the presence of bacteria, we obtained DNA from unialgal culture supernatants and amplify 16S rRNA gene sequences by PCR. Then, we analyzed the digestion profile of the PCR amplicons. The presence of several electrophoretic bands demonstrated the presence of bacterial species in these

Ectocarpus culture samples (Supplementary Figure S1B, right), and prompted us to isolate some of them. Nine different bacterial isolates were obtained from the surfaces of two different Ectocarpus unialgal strains, i.e., Ec 32 and Ec 524, and each of them was screened for its effects on Ectocarpus morphology. Although these cultivable bacteria do not reflect the entire Ectocarpus microbiota, they represent bacteria that are indeed associated to the alga. Thus, testing these isolates allowed description of the effects of at least part of the algal associated bacteria, but it should kept in mind that there might be a lot of missing bacteria that are uncultivable under the conditions used.

Seven of the nine isolates belonged to the Proteobacteria phylum while the other two were cataloged as Actinobacteria according to their 16S rRNA gene sequences (**Table 1**). Six out of these bacterial isolates triggered the development of upright filaments, being all members of the Proteobacteria phylum. Isolate 869\_1 has no effect on such morphological

trait (**Figure 3C**), whereas isolates Z3 and Z8a\_1 were examples of the six isolates producing upright filaments (**Figures 3D,E**). After 6-weeks cultivation, bacterial inoculated Ectocarpus (**Figure 3H**) resembled the branched morphology of unialgal cultures (**Figure 3F**). On the other hand, the lack of upright filaments on axenic individuals gave them a "small ball"-like appearance (**Figure 3G**), which was far different to the other morphologies observed in conditions where bacteria were present. These findings confirm the initial observations that pointed out to the necessity of bacterial presence for Ectocarpus to develop its upright filaments and also show that a single bacterium can be enough to achieve this goal.

Bacterial effects directly influenced the cell types present in Ectocarpus sp. individuals. Under sterile conditions, Ectocarpus was composed of just two types of cells, elongated (E; **Figure 4A**) and round (R; **Figure 4B**). Ectocarpus in the presence of bacteria, in addition to contain R and E cells, displayed other types of cells that compose the upright filaments (**Figure 4C**). The cells in these filaments were very different from the R and E cells since they were larger, with an average of 55 and 17 µm of length and width, respectively; while E cells had an average of 30 and 6 µm for the same dimensions (**Figures 4A,C,E**). The effect of the presence of bacteria (in this case isolate Z3) in the formation of upright filaments was corroborated by epifluorescence microscopy (**Figure 4D**). Thus, by affecting Ectocarpus morphology, bacteria are also involved in cell differentiation processes crucial for algal development.

The effect of the presence or absence of bacterial isolates on Ectocarpus sp. reproduction was also addressed. In axenic algal cultures the number of individuals produced was around five per square centimeter versus the 25 to 70 counted when filament-producing bacteria (either in unialgal cultures, or as individual isolates) were present (**Figure 5**, third and last column in **Table 1**). While some bacterial isolates, -e.g., Z8a\_1, Z7, and R1-, generated similar levels of individuals produced in unialgal cultures, Z3 increased significantly the number of individuals produced (**Figure 5** and **Table 1**). A clear correlation was found for the isolates that were capable to recover upright filaments development and their ability to trigger production of new individuals (**Table 1**). These results emphasize that, besides affecting morphology, bacteria are also relevant for Ectocarpus production of new germlings.

### Effects of Bacterial and Bacterial-Algal Co-cultures Supernatants on Ectocarpus sp. Morphology and Reproduction

Growth culture supernatants from the nine isolated bacteria were obtained and tested for their ability to induce filaments development. None of the bacterial supernatants was capable of inducing growth of filaments (**Table 1**, **Figures 6B,C**), producing an algal morphology as that found for axenic cultures (**Figure 6A**), although bacterial supernatant from isolate Z3 modified Ectocarpus early development (Supplementary Figure S4). The effect of supernatants obtained from co-cultures of bacterial isolates and Ectocarpus sp. were then tested. Some of these co-culture supernatants did recover upright filament development, but this effect was only achieved for those cocultures of bacterial isolates that were able to induce filaments presence (**Figures 6D,E**).

Concerning reproduction of Ectocarpus individuals, bacterial supernatants did not increase the number of individuals, except for isolate Z3 supernatant which slightly increased the individuals produced with respect to the control (**Figure 6F**). Co-culture supernatants had the same effect of the isolates from which these supernatants were produced, but with a lower impact


The number of individuals per cm<sup>2</sup> is given with the standard deviation (±). <sup>∗</sup>For axenic cultures 5.3 ± 1.5 and for unialgal cultures 39 ± 5.2.

FIGURE 4 | Effect of bacterial isolates on cells types of Ectocarpus sp. (A) Elongated cells (E cells) and (B) round cells (R cells) present in axenic and non-axenic individuals, respectively, as part of the prostrate body of the alga. (C) Typical upright filament cells in non-axenic individuals, in this case inoculated with isolate Z3. (D) Representative epifluorescence microscopy image of Ectocarpus individual inoculated with strain Z3. Arrows indicate bacterial presence on Ectocarpus sp. filaments determined by SYBR Green II staining. All bars correspond to 10 µm. (E) Comparison between length and width of elongated cells and filament cells. Cell lengths and widths of 60 individuals were measured. The <sup>∗</sup> indicates means statistically different at p < 0.05.

compared to direct exposure to bacteria (**Figure 6F**). According to these results, bacterial effects on Ectocarpus morphology and reproduction are accomplished by active interaction with the alga, needing both organisms to be in the same culture. Furthermore, the compound (s) responsible of algal upright filaments emergence and stimulation of reproduction is (are) produced during bacterium-alga co-cultures and is (are) released to the media.

### Effect of Bacteria on Ectocarpus sp. Released Metabolites, i.e., Exometabolome

In order to get some insight about the bacterial effect on the metabolism of the alga, the metabolite profiles of exudates from axenic Ectocarpus sp. alone, a bacterial isolate alone, and the combination of both organisms were determined by ultrahigh pressure liquid chromatography (UPLC) coupled to mass spectrometry (MS). In this case, bacterial isolate Z3 was chosen to perform the evaluation because it had filament-inducing activity and it was the one with the greatest effect on reproduction (**Table 1**). The samples were taken after 3 weeks of co-culturing since at this time point filaments were already developed. A global metabolite profiling by LC-MS-MS in positive ion mode provided the more informative set of data, with 320 signals, and was used for further analysis. Multivariate analysis of these exometabolome profiles revealed specific clustering for the three conditions analyzed, with a clear separation of a three distinct groups along the two axis, explaining 76.4 and 14.2% of the variance as it is shown by PCA plot (**Figure 7**). These data indicate that Ectocarpus sp. plus this bacterium released a set of metabolites that is distinct from those generated by the same bacterium and alga growing alone.

using 320 monoisotopic peaks quantified by UPLC-MS in positive ionization mode. All metabolites were considered for PCA (p-value < 0.05) generated by SIMCA-P v12.0.

### DISCUSSION

Algae provide an advantageous environment for proliferation of bacteria, some of which have already been shown to have positive effects on their hosts (reviewed in Singh and Reddy, 2014). For brown algae, putative beneficial effects of bacteria on development still remain to be experimentally tested and fully established. Regarding to Ectocarpus, microbiota relevance under abiotic stress was recently investigated (Dittami et al., 2015). In this context, the present study goes deeper on previous observations about the importance of bacteria for these marine organisms. This work demonstrates that bacteria influence morphology and reproduction of the brown algal model Ectocarpus sp. Typical branched morphology of this alga is clearly dependent on the presence of bacteria. This finding is consistent with previous studies on green algae, as members of Ulvaceae lose their typical morphology when cultured under axenic conditions (Provasoli and Pintner, 1980) but recovered it when inoculated with appropriate morphogenesis-inducing bacterial isolates (Matsuo et al., 2003; Marshall et al., 2006). For red algae, the role of bacteria on morphological development had been also demonstrated (Singh et al., 2011; Fukui et al., 2014). The fact that the three major groups of multicellular algae are influenced in their morphology by bacteria, strongly suggest that this type of interaction has been relevant for these organisms during their evolution. The effect of microorganism communities contained in natural seawater on axenic Ectocarpus sp. (**Figure 1**) resembled the effect of isolated bacteria. This is significant because it validates the use of single bacterial isolates as a proxy of what the alga could found in the field.

In the present study, we evaluated the effect of nine bacterial isolates obtained from unialgal laboratory cultures of Ectocarpus sp. This rather low number of bacterial isolates may be explained by the constraints imposed to this alga under laboratory conditions. The Ectocarpus strains used to isolate bacteria has been kept under laboratory conditions for long time (years) and they have been exposed to conditions (including antibiotic treatments), which decreased bacterial diversity and abundance at an extent difficult to determine.

It might appear that there could be some specificity in the ability of these bacterial isolates to have an effect on Ectocarpus morphology because only proteobacterial isolates showed effects on this alga. Although Proteobacteria has been shown as a dominant phylum in other studies describing bacterial communities associated with algae (Hengst et al., 2010; Burke et al., 2011a; Hollants et al., 2013) we cannot discard a possible bias in the bacterial isolation procedure, which led to preferential selection of these microorganisms. In order to clarify this issue,

we did a gross survey on bacterial diversity associated to field and laboratory Ectocarpus. The majority (72 and 56% for field and laboratory samples, respectively) of the sequences analyzed were affiliated to Proteobacteria (Supplementary Figure S5A), which is consistent with the dominance of this phylum between the bacterial isolates reported here. Remarkably, the recent study of Dittami et al. (2015) also reports the dominance of Proteobacteria associated to laboratory strains of Ectocarpus. The similarity in abundances at phylum level between field and laboratory samples supports the idea that what we observed in laboratory specimens could be applied to the field. Interestingly, most of the bacterial strains isolated in this work were detected using this culture-independent approach in both field and laboratory algae (Supplementary Figure S5B). Again, the abundances of these bacteria in field and laboratory Ectocarpus were quite similar. In general, the bacterial isolates correspond to 11% of total microbiota. Most isolates are low-abundance bacteria (less than 1%) except for the Roseobacter representative, which is very abundant when consider all samples together (Supplementary Figure S5B), although its abundance is rather low in several samples (Supplementary Figure S5C).

It should be kept in mind that this report evaluated the role of bacteria using a culture-dependent approach. There are studies that have established that only a small proportion of bacteria can be cultivated using conventional methodologies (Whitman et al., 1998; Fry, 2000; Handelsman, 2004). In this context, the results showed in this work might apply to a small part of bacteria thriving on the surface of Ectocarpus, although they reflect bacteria indeed associated with this alga.

A deeper exploration of the taxonomic affiliation of bacterial isolates capable to induce filaments development showed that, apart of belonging to the Proteobacteria phylum, there is no further taxon specificity in the effects observed. Bacterial isolates producing morphology/reproduction effects are distributed among several families and genera. This observation has been also reported for green algae (Nakanishi et al., 1996; Marshall et al., 2006). These studies reported that several bacterial genera are capable to influence morphology of Ulva pertusa and U. linza, including genus Vibrio, Pseudomonas, Halomonas, Escherichia and some Gram-positive bacterial genera. In our study, we also found a Halomonas isolate (strain Z3) having a strong impact on morphology. In contrast, isolates belonging to the genera Antarctobacter (R6a), Marinobacter (Z8a\_1, R8), and Methylophaga (R1), are for the first time described to influence macroalgal development. On the other hand, the two Actinobacteria isolates studied here did not have any effect on Ectocarpus development, although the impact of member of this phylum on green algal morphology has been reported (Nakanishi et al., 1999; Marshall et al., 2006).

Ectocarpus sp. early sporophyte development has been already described. Le Bail et al. (2008b) reported that sporophytes grow as prostrate filaments composed of two cell types, E and R. These cells form the prostrate body of the alga. If the growth conditions are favorable, upright filaments emerge after a few days, contributing to the establishment of an overall filamentous architecture (Ravanko, 1970). In the present study, the upright filaments appearance was found to be a bacterial modulated process. When bacteria were not present in the culture medium, Ectocarpus sp. developed its prostrate body without any upright filaments (**Figure 3**) producing only E and R cells (**Figures 4A–E**). In contrast, when axenic Ectocarpus sp. was cultivated in culture medium containing microorganisms, or with bacterial isolates (**Figure 3**), it developed upright filaments and recovered most of its filamentous morphology (**Figure 5**). Although the influence of bacteria on algal morphology had been reported, it is relevant to stress that the effect of bacteria on the appearance is not only on the filaments per se, but also in the new cell types required to form these structures. The cells composing the filaments are very different from those of the prostrate body (**Figure 4**), which means that bacteria are capable of triggering cell differentiation mechanisms in the alga. In this regard, plant hormones represent very good candidates to produce these kinds of effects. These compounds control plant growth by affecting the spatial and temporal expression of genes involved in cell division, elongation, and differentiation. Pedersen (1968), early suggested that E. fasciculatus, a sister species of Ectocarpus sp. needs cytokinins in order to grow normally under culture conditions. In Ectocarpussp. it had been suggested that auxins could be involved on upright filaments appearance by repressing its emergence (Le Bail et al., 2010). Although phytohormones presence on macroalgae have been reported (Stirk et al., 2003), to date there is no evidence of bacterial phytohormones production having a direct effect on algal development, despite it is already known that marine bacteria can produce these compounds (Maruyama et al., 1986, 1990).

A possible explanation to the results obtained with the supernatant essays is that all the bacterial isolates capable of inducing filament appearance secrete filament-inducing factor(s) (e.g., phytohormones) into the culture supernatant only when Ectocarpus sp. is also present (co-cultures). When bacterial isolates were grown alone their supernatants did not have an effect in morphology or reproduction (**Figure 6**). In this context, it has been proposed that Ectocarpus could manage to produce phytohormones in association with bacteria (Dittami et al., 2014), and the same has been predicted recently for diatoms and their interaction with bacteria (Amin et al., 2015). The production and exchange of chemicals cues between algae and bacteria seems to be critical for the wellbeing of these organisms in natural conditions. Nevertheless, in other studies bacterial supernatants have been shown to be sufficient in modulating algal development. In the green alga Monostroma oxyspermun, supernatants of bacterial cultures recover the normal morphology of the alga (Matsuo et al., 2003). Matsuo et al. (2005) identified this exogenous growth factor as thallusin, produced by bacteria belonging to Bacteroidetes phylum. Because of their evolutionary distance, it is not surprising that the mechanisms involved in the effect of bacteria on green and brown algae could be different.

It is not clear if contact between bacteria and Ectocarpus sp. is required for morphology and reproduction to be affected. The reported observations do not rule out the possibility that bacteria need not to be in contact with the alga but just closely enough to communicate with each other and produce the compound(s) responsible for the described effects. On this regard, some

PGPB have been shown to exert their effect by production of volatile organic compounds (VOCs) without requiring direct contact with the plant (Gutiérrez-Luna et al., 2010; Meldau et al., 2013). These evidences suggest that a similar mechanism could be involved in the described effects of bacteria on Ectocarpus development. What is clear is that the presence of both organisms in the same culture is needed in order to produce filaments development, which implies that some interaction exists between bacteria and alga. The effect accomplished by the co-culture supernatant in morphology and reproduction means that the compound(s) responsible for this phenomenon is (are) secreted and stable in the culture medium, at least for some time. When comparing algal individuals produced by axenic Ectocarpus sp. exposed to direct bacterium inoculation versus the exposure to co-culture supernatants, direct inoculations have an stronger impact than co-culture supernatants. This suggests that the compound(s) responsible for the effects was (were) not stable for a long time in the culture medium, so the permanent presence of bacterium (and concomitant continuous production) seems to be required to produce more pronounced effects.

The influence of bacteria on Ectocarpus physiology was reflected by the results of the metabolomic approach shown in this work. The recorded data demonstrated that Ectocarpus sp. associated with a single bacterium produces a different metabolite profile compared to those of axenic alga. However, which compounds cause the effects described was not assessed. Other reports combined metabolomics with bioassays thus narrowing down the metabolome to one biologically active compound (Matsuo et al., 2005; Schroeder et al., 2006). The untargeted analysis performed here does not yield functional information, unless it is combined with a bioassay as well. For the majority of metabolites that were detected, both the identity and the function in the Ectocarpus-bacterium interaction, is largely unknown. However, it is quite clear that the impact of bacteria on Ectocarpus metabolomic profile shows that bacterial influence is exerted at several levels of algal physiology.

In summary, this article supports the importance of bacteria for reproduction, growth and development of the brown algal model Ectocarpus sp. The range of bacteria that affect development on Ectocarpus sp. could confer ecological flexibility to the alga. This may be important since this alga inhabits worldwide along temperate coastlines, where it can grow on either rocky and/or artificial substrates or epiphytically on other algae thus being challenged by very different bacterial communities. The mechanisms involved in this interaction are presently unknown, but at least some communication is required to display the effects described. Bacterial impacts on physiology were also highlighted since one bacterial isolate could drive major changes in the algal exometabolomic profile. Altogether, the data

#### REFERENCES

Amin, S. A., Hmelo, L. R., van Tol, H. M., Durham, B. P., Carlson, L. T., Heal, K. R., et al. (2015). Interaction and signalling between a cosmopolitan phytoplankton and associated bacteria. Nature 522, 98–101. doi: 10.1038/nature14488

Andersen, R. A. (2004). Biology and systematics of heterokont and haptophyte algae. Am. J. Bot. 91, 1508–1522. doi: 10.3732/ajb.91.10.1508

reported in this study along with the molecular tools already available for Ectocarpus sp. open a new window in the study of algal host–microbes interactions.

#### AUTHOR CONTRIBUTION

JT performing experiments, conducting the work, design of the work, analysis, interpretation of data for the work, responsible for the integrity of the work as a whole, final approval of the version to be published. BG design of the work, interpretation of data for the work, critically revising the final approval of the version to be published. SG performing experiments, analysis, interpretation of data for the work. PP design of the work, analysis, interpretation of data for the work, ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved, critically revising the final approval of the version to be published, responsible for the integrity of the work as a whole. JC design of the work, analysis, interpretation of data for the work, ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved, critically revising the final approval of the version to be published, responsible for the integrity of the work as a whole.

### FUNDING

This work benefited from the support of the French Government via the National Research Agency in investment expenditure program IDEALG (ANR-10-BTBR-04). JT was supported by a CONICYT PhD scholarship (21110796).

#### ACKNOWLEDGMENTS

We would like to thank Jessica Beltrán, Verónica Flores, and Laurence Dartevelle for their technical support during this work, along with the Algal Team at the Pontificia Universidad Católica de Chile and the Algal Defense Team, UMR 7139 CNRS-UPMC at the Station Biologique de Roscoff, for their assistance and helpful advice on experimental setup and analysis. We also thank Simon Dittami for critical reading of the manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.00197


analysis of Ectocarpus siliculosus infected with the basal oomycete Eurychasma dicksonii. PLoS ONE 6:e24500. doi: 10.1371/journal.pone.0024500


fmicb-07-00197 February 23, 2016 Time: 19:25 # 13


important red alga Gracilaria dura. FEMS Microbiol. Ecol. 76, 381–392. doi: 10.1111/j.1574-6941.2011.01057.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 © 2016 Tapia, González, Goulitquer, Potin and Correa. 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) or licensor 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.

# Cross-Kingdom Similarities in Microbiome Ecology and Biocontrol of Pathogens

Gabriele Berg<sup>1</sup> \*, Robert Krause<sup>2</sup> and Rodrigo Mendes <sup>3</sup>

1 Institute of Environmental Biotechnology, Graz University of Technology and ACIB Austrian Centre of Industrial Biotechnology, Graz, Austria, <sup>2</sup> Department of Internal Medicine, Medical University of Graz, Graz, Austria, <sup>3</sup> Laboratory of Environmental Microbiology, Embrapa Environment, Jaguariuna, Brazil

Keywords: microbiome, biocontrol agent, rhizsophere, pathogens, ecological theories

"Imagination is more important than knowledge." -Albert Einstein

### INTRODUCTION

#### Edited by:

Suhelen Egan, The University of New South Wales, Australia

#### Reviewed by:

Stéphane Hacquard, Max Planck Institute for Plant Breeding Research, Germany Catherine Leblanc, Centre National de la Recherche Scientifique, France

#### \*Correspondence:

Gabriele Berg gabriele.berg@tugraz.at

#### Specialty section:

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

Received: 06 September 2015 Accepted: 09 November 2015 Published: 25 November 2015

#### Citation:

Berg G, Krause R and Mendes R (2015) Cross-Kingdom Similarities in Microbiome Ecology and Biocontrol of Pathogens. Front. Microbiol. 6:1311. doi: 10.3389/fmicb.2015.01311 The concept of "gut and root microbiota commonalities" was already presented by Ramírez-Puebla et al. (2013); they discussed a lot of similar functional traits, host-bacteria interactions as well as evolutionary trends but also several differences. Based on deeper insights obtained by omics technologies, Mendes and Raaijmakers (2015) recently presented their concept that the structure and function of rhizosphere and gut microbiomes show cross-kingdom similarities. In parallel, Hacquard et al. (2015) analyzed similarities of the microbiota composition across plant and animal kingdoms and found only little overlap comparing fish gut and plant root communities. They explained the differences by various start inoculants and abiotic, niche-specific factors. In this context, to establish concepts is pivotal in microbial ecology for the critical evaluation of the immense amount of data obtained by omics technologies, not only for conceptual work in microbial ecological theories (Prosser et al., 2007), but also for translational fields such as biocontrol of pathogens (Berg et al., 2013). Therefore, we would like to extend the concept of "cross-kingdom similarities" presented by Mendes and Raaijmakers (2015) to an ecological context, which is shared for host-associated microbiomes beyond the boundaries of their respective kingdoms. Finally, we discuss the impact and implications of microbiome ecology on biocontrol of pathogens in plants and in humans.

### SIMILARITIES IN HOST-ASSOCIATED MICROBIOME ECOLOGY

1. **Each host provides microhabitats with different abiotic conditions, which shape the structure of microbial communities. Despite their specific composition, communities are connected to each other and share microbial populations.** The different microhabitats of plants carrying their individual names, e.g., rhizosphere, phyllosphere, endosphere, have been well-studied for decades (Philippot et al., 2013; Berg et al., 2015; Hardoim et al., 2015). In parallel, human microenvironments and their specific microbiomes have been thoroughly studied over the past years (Blaser et al., 2013). In addition, microbial exchanges between different host's compartments or niches were analyzed. For plants, the microbial transfer from soil to the rhizo- and endo-sphere was analyzed in particular (Edwards et al., 2015). In humans, analysis of dent-associated microbial communities was shown to be important for the health of the whole body including the placenta during pregnancy periods (Aagaard et al., 2014).


to their role as degrader (Berg et al., 2015). Strikingly, members of the Archaea domain showed the same ecological behavior; they colonize old/senescent plants and humans in high abundances (Probst et al., 2013; Müller et al., 2015). In addition, there is an impact of physiological rhythms of the host, e.g., the activity of all organisms is regulated by diverse molecular clock mechanisms that synchronize physiological processes to diurnal environmental fluctuations. Recently, Thaiss et al. (2014) showed that the intestinal microbiota in humans exhibits diurnal oscillations that are influenced by feeding rhythms, leading to time-specific compositional and functional profiles of the microbiome over the course of a day. Ablation of host molecular clock components or induction of jet lag leads to aberrant diurnal microbiota fluctuations and dysbiosis, driven by impaired feeding rhythmicity. Plants, as photosynthetically active organisms, show strong daily rhythms, influencing the release of root exudates. We therefore assume that this rhythm in turn affects the rhizosphere microbiome.


systems (Köberl et al., 2011). In parallel to human diseases, it is well known that long-term agricultural monoculture potentially results in disease outbreaks, which are often followed by establishment of disease suppressive microbiomes (Mendes et al., 2011; Kwak and Weller, 2013). This was also experimentally evidenced in a field plot by Santhanam et al. (2015). In the future, hopefully biocontrol approaches can be extended to microbiome control and design strategies to prevent diseases.

#### CROSS-KINGDOM SIMILARITIES IN BIOCONTROL OF PATHOGENS

The human microbiome has been predicted to become one of the most important tools for personalized health and targeted medicine (De Vrieze, 2013). In addition, to understand the plant microbiome is crucial to find solutions for environmentally friendly agriculture especially under climate change condition and a growing human population (Berg et al., 2013). Altogether, the modulation of microbiota is currently a growing area of research as it just might hold the key to treatment. Mueller and Sachs (2015) call this an engineering approach for host-mediated microbiome selection. They proposed designed microbiomes, which enhance host functions, contributing to host health and fitness. Biological control exists much longer as an environmentally sound and effective means of reducing pathogens and their symptoms through the use of natural antagonists. All ecological rules or patterns presented above have an impact on the development of novel biocontrol approaches and offer an enormous potential for biotechnology. Summarizing, a microbiome approach for biological control should consider the following: (i) the specific composition of microbiomes at different developmental stages and for different species/cultivars/ecotypes, (ii) core microbiomes as an important source for biologicals, (iii) the microbiome view should include members from the three domains of life, that is, Bacteria, Archaea, and Eukaryotes, (iv) functional diversity within a microbiome is often more important than structural diversity, and (v) the loss of diversity caused by human intervention should be compensated.

While in the past mainly single organisms were used as biocontrol agents (BCAs) in human medicine as well as for agricultural purposes, e.g., Bacillus strains for plants and Lactobacillus strains for humans, it is now possible to develop predictable microbiome-based biocontrol strategies and to avoid inconsistent effects of the first generation of biologicals (**Figure 1**). These novel biocontrol strategies can not only be used to suppress pathogens, but they can also be effectively used to establish microbiomes in a desirable beneficial composition for particular purposes in the future (Berg et al., 2013). Diversity vs. pathogenicity should be an important criterion for microbiome design (van Elsas et al., 2012). This was shown by Ratner (2015) combining fecal transplants with microbial cocktails against inflammatory bowel disease. In parallel, suppressive soils were used with biologicals to supress plant diseases as Panama disease in bananas (Xue et al., 2015) and dampingoff in sugar beet (Mendes et al., 2011). Stress protection agents such as Stenotrohomonas rhizophila are able to protect maize against drought but they also shift the whole plant-associated community and overgrow or eliminate latent fungal pathogens (Alavi et al., 2013). In addition to current developments in probiotics, prebiotics, synbiotics, and psychobiotics as defined by Wasilewski et al. (2015), many more translations are however possible, e.g., combining biologicals for plant and human health.

Functional diversity is an important aspect for health but also for biocontrol. Therefore, the mode of action plays a crucial role for biocontrol. In addition to the well-studied interaction

with hosts and pathogens, the interaction with the indigenous microbiome has to be studied (**Figure 1**). Biologicals cause a microbiome shift, in parallel to antibiotics, but the overall goal is that they (i) enhance indigenous microbial diversity, (ii) eliminate (minor) pathogens or avoid pathogen overgrowth, and (iii) promote indigenous beneficials. Recent studies have shown promising results, including the use of biologicals in plants to enhance structural microbial diversity (Erlacher et al., 2014) as well as those applied to elderly people to improve functional diversity (Eloe-Fadrosh et al., 2015).

Altogether, there are a lot of similarities in biocontrol approaches for plants and humans. Although, single strain-based approach for biological control has begun more than 100 years ago, inconsistent control results and the fact that only a limited number of success cases exists made its use and acceptance difficult. Now we have the tools to move from a single isolate- to a community-based biocontrol approach and develop predictable biocontrol strategies on the basis of the microbiome ecology.

#### REFERENCES


However, there are still several hurdles in this field, including for example the formulation and shelf life of microbial communities. Furthermore, the occurrence of potential pathogens and diverse resistomes—the sum of antibiotic resistance genes—in all microbiomes needs a conceptual framework in biocontrol and microbial ecology theories.

#### AUTHOR CONTRIBUTIONS

All three authors contribute to the opinion paper.

#### ACKNOWLEDGMENTS

We would like to thank Timothy Mark (Graz) for helpful revision. This work has been supported by a grant to GB within the frame of the COMET-Funding Program managed by the Austrian Research Promotion Agency FFG.


Mendes, R., and Raaijmakers, J. M. (2015). Cross-kingdom similarities in microbiome functions. ISME J. 9, 1905–1907. doi: 10.1038/ismej.2015.7


**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 © 2015 Berg, Krause and Mendes. 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) or licensor 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.

# Controlling the Microbiome: Microhabitat Adjustments for Successful Biocontrol Strategies in Soil and Human Gut

Eveline Adam<sup>1</sup> † , Anneloes E. Groenenboom2 †, Viola Kurm3 †, Magdalena Rajewska4 † , Ruth Schmidt 5, 6 †, Olaf Tyc5, 6 †, Simone Weidner 7 †, Gabriele Berg1‡, Wietse de Boer 5, 6 \* ‡ and Joana Falcão Salles 8 ‡

1 Institute of Environmental Biotechnology, Graz University of Technology, Graz, Austria, <sup>2</sup> Laboratory of Genetics, Wageningen University, Wageningen, Netherlands, <sup>3</sup> Department of Terrestrial Ecology, Netherlands Institute of Ecology, The Royal Netherlands Academy of Arts and Sciences, Wageningen, Netherlands, <sup>4</sup> Laboratory of Biological Plant Protection, Intercollegiate Faculty of Biotechnology, University of Gdansk and Medical University of Gda ´ nsk, Gda ´ nsk, Poland, ´ <sup>5</sup> Department of Microbial Ecology, Netherlands Institute of Ecology, The Royal Netherlands Academy of Arts and Sciences, Wageningen, Netherlands, <sup>6</sup> Department of Soil Quality, Wageningen University and Research Centre, Wageningen, Netherlands, <sup>7</sup> Department of Biology, Institute of Environmental Biology, Utrecht University, Utrecht, Netherlands, <sup>8</sup> Institute of Evolutionary Life sciences, Groningen University, Groningen, Netherlands

Keywords: host beneficial bacteria, microbiome control, minor disturbances, major disturbances, synbiotics

#### Edited by:

INTRODUCTION

Joerg Graf, University of Connecticut, USA

#### Reviewed by:

Carl James Yeoman, Montana State University, USA Anton Hartmann, German Research Center for Environmental Health, Germany

> \*Correspondence: Wietse de Boer w.deboer@nioo.knaw.nl

> > † Junior authors. ‡ Senior authors.

#### Specialty section:

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

Received: 15 March 2016 Accepted: 27 June 2016 Published: 13 July 2016

#### Citation:

Adam E, Groenenboom AE, Kurm V, Rajewska M, Schmidt R, Tyc O, Weidner S, Berg G, de Boer W and Falcão Salles J (2016) Controlling the Microbiome: Microhabitat Adjustments for Successful Biocontrol Strategies in Soil and Human Gut. Front. Microbiol. 7:1079. doi: 10.3389/fmicb.2016.01079 The human gut and the rhizosphere are environments colonized by highly diverse communities of microbes, which perform complex interactions with their host and carry out important functions including enhanced disease resistance and nutrient uptake. In humans they are involved in energy harvest and storage as well as in interactions with the immune system (Clemente et al., 2012). In plants they have profound effects on seed germination, seedling vigor, nutrition, plant health, and development of the innate immune system (Mendes et al., 2013; Berg et al., 2014a; Schikora et al., 2016). The composition of the microbial communities is host-specific and related to its health status (Smalla et al., 2001; Kinross et al., 2011; Berg et al., 2014a). Imbalances caused by disturbance-induced shifts in microbial species abundances can lead to disease outbreaks in both environments (Berendsen et al., 2012; Robles Alonso and Guarner, 2013; Berg et al., 2014b) and further to probable proliferation of pathogenic species (Van Elsas et al., 2012; Van Agtmaal et al., 2015).

To restore or maintain the health of the host, bio-based solutions supporting the pathogensuppressing ability of the hosts' native microbiome can be applied, including probiotics, synbiotics and biocontrol agents (de Vrese and Schrezenmeir, 2008). Such methods aim to increase the abundance and activity of host beneficial bacteria (HBB). However, addition of HBB does not always result in the desired pathogen suppression due to insufficient establishment, i.e., lower survival and/or poor colonization rates of the HBB (Bashan et al., 2014).

Concepts from invasion ecology suggest that survival rates of invaders are inversely related to the diversity of the native microbiome. This can be explained by higher resource uptake and consequent reduction in niche availability (Mallon et al., 2015). In addition, prevailing physical and chemical parameters in the respective environment like texture, pore size distribution, and moisture content might not favor the establishment of the introduced HBB. For a long-term establishment of the HBB in the soil these abiotic factors have to be considered. In the gut, the colonization resistance determined by the commensal microbiome is linked to its capacity to exploit the available niches and to prevent the establishment of invaders via niche occupation (reviewed in Stecher et al., 2013). The knowledge on mechanisms of microbial invasions (Mallon et al., 2015) can be used to improve the survival of HBB in both environments.

Given that similar mechanisms drive microbial colonization and establishment in the gut and rhizosphere microbiomes, we suggest that biocontrol strategies could be similar for both environments (Ramírez-Puebla et al., 2013; Berg et al., 2015; Mendes and Raaijmakers, 2015). Here we develop possible strategies to ensure long-term establishment of HBB by manipulating niche availability.

### CREATING MICROHABITATS FOR HOST BENEFICIAL BACTERIA BY INTRODUCING MINOR DISTURBANCES

Several studies have shown that soils harboring low microbial biomass or low microbial diversity are more susceptible to colonization by other organisms (Fließbach et al., 2009; Van Elsas et al., 2012). Certain agricultural practices can result in major disturbances of the **rhizosphere microbiome**. Examples include disinfestation with chemical pesticides, heat treatment (Stapleton, 2000), radiation or anaerobic disinfestation (Van Agtmaal et al., 2015). Moreover, tillage systems may have major effects on the established community by reducing certain soil microbial populations, particularly fungi (Ventorino et al., 2012). Analogous events, leading to changes in the **human gut microbiome**, are the application of broad spectrum antibiotics, fecal transplantations (Landy et al., 2011; de Vos, 2013) or considerable changes in diet (Turnbaugh et al., 2009). Whilst major disturbances are frequently used to eliminate pathogens, those methods possibly also disrupt beneficial functions of the indigenous microbial community (Altieri, 1999; Geiger et al., 2010).

An alternative strategy is to introduce **minor disturbances** to create free niches for HBB's in both the rhizosphere and the human gut microbiome. This strategy aims to selectively empty niches in the existing community.

In the rhizosphere the introduction of accessory bacterial predators such as protozoa (e.g., flagellates, ciliates) or nematodes (Jousset et al., 2006; Abada et al., 2009; Pedersen et al., 2009; Freyth et al., 2010; Neidig et al., 2011; Müller et al., 2013) could foster biocontrol strains via enhanced selective predation when the biocontrol strain protects itself through production of antibiotics. The increase in predation pressure might also stimulate biocontrol strategies by direct predation on pathogens as well as nutrient turnover and bacterial activity in soil. Likewise, specific bacteriophages could be applied to selectively eliminate target bacterial species or strains. This strategy has been effectively shown as part of disease management for Rhizobium sp., Bacillus sp., Burkholderia sp., Xanthomonas sp., Pectobacterium sp. and Dickeya sp. (Evans et al., 1979; Sharp et al., 1986; Lynch et al., 2012; Chae et al., 2014; Santamaría et al., 2014; Czajkowski et al., 2015). For this approach, elimination of pathogens and reduction of soil bacterial species that directly compete with the biocontrol agents (i.e., those sharing similar metabolic capacities) are desirable. Due to their specificity, bacteriophages have also been used to treat gastrointestinal infections of bacterial origin in humans (Sulakvelidze et al., 2001; Abedon et al., 2011). Moreover, they were successfully used together with bifidobacteria to treat antibiotic-associated dysbacteriosis in infants (Litvinova et al., 1978). Therefore, bacteriophages represent an alternative to selectively wipe out bacteria (either pathogens or strong competitors) in the gut and to form a niche for potential HBBs to thrive. In the rhizosphere the use of bacterial helper strains, an application of targeted specific antibiotics or enzymes (e.g., chitinases; Herrera-Estrella and Chet, 1999) might affect the microbiome composition sufficiently to form free niches for HBB. Another possibility is to introduce minor changes in physical properties like pH value (Rousk et al., 2010), temperature (Van Veen et al., 1997; Haas and Défago, 2005), moisture dynamics or salinity (Canfora et al., 2014; Dini-Andreote et al., 2014).

Most of the methods described here apply to the rhizosphere (e.g., substantial temperature or salinity changes), but due to ethical concerns can only be considered in a limited manner for the human gut. Thus, the direct applicability to the human gut remains to be investigated. The concept of freeing/forming a microhabitat for HBB by minor disturbances in the rhizosphere or the human gut should be developed and optimized for different situations.

Apart from making an existing niche available for the HBB by removal of at least a part of the adapted community, creation of a new niche could also be taken into consideration.

### IMPROVEMENT OF THE ENVIRONMENT—THE HUMAN GUT AS A PARAGON FOR CONCEPTS IN BIOCONTROL

To alleviate competition and increase the chance of establishment of HBB in an environment that harbors a highly diverse microbial community utilizing all available resources can be enabled by adding specific energy resources, for example prebiotics. Prebiotics selectively stimulate growth and/or activity of the beneficial bacteria and facilitate their establishment in the heavily colonized gut (Teitelbaum and Walker, 2002; Tuohy et al., 2003). Moreover, administration of synbiotics, a combination of a probiotic (i.e., the HBB) and a prebiotic, has recently attracted attention (Schrezenmeir and de Vrese, 2001). The prebiotic provides a selective food source for the HBB enhancing its growth and establishment (Teitelbaum and Walker, 2002; Saulnier et al., 2008). The success of synbiotics has been demonstrated in vitro as well as in vivo (Bartosch et al., 2005; Saulnier et al., 2008). We suggest that the use of synbiotics in the human gut can serve as a paragon to enhance the establishment of HBB in the soil. In the rhizosphere the addition of a selective food source e.g., rhizopins (Oger et al., 2004) could be used to stimulate specific bacteria in the rhizosphere community.

## SYNBIOTICS FOR THE SOIL

Parallels with prebiotics can be seen in the application of general resources to the soil, such as composts and green manures. These strategies have shown to be effective in the control of soilborne diseases as they combine the introduction of biocontrol microorganisms with organic matter after the thermophilic phase low in competition and free nutrients. This substrate favors the growth of beneficial microbes and suppresses the growth of saprophytic pathogens (Hoitink et al., 1997; Hoitink and Boehm, 1999). A disadvantage of using this method, however, is varying compost quality, which results in inconsistent colonization by biocontrol agents and subsequent effects on disease-suppression (Sturz and Christie, 2003). To ensure the presence of the desired HBB, composts can be fortified with specific beneficial microorganisms or amended with substrates that stimulate growth and activity of a selected group of microorganisms (Haggag and Abo-Sedera, 2005; Chae et al., 2006; Dukare et al., 2011).

In addition, specific substrates and HBB can be combined to complement each other. Several studies have shown that certain carbon sources and minerals increase the activity of biocontrol bacteria (Duffy and Défago, 1999; Shaukat and Siddiqui, 2003; Kim et al., 2008). Moreover, plants are able to select for specific bacteria by exudation of sugars, polysaccharides, amino acids, and a variety of secondary metabolites (Teplitski et al., 2000; Badri et al., 2009). These compounds are comparable to mucosal glycans in the human gut. As a soil synbiotic, these compounds could be artificially applied in combination with the respective HBB. Not only nutrient sources, but also signaling molecules and chemo-attractants should be taken into account, which are often highly specific for certain bacterial species or even strains. Ultimately, engineering beneficial microbes or genetically modified plants that are capable of synthesizing certain enzymes quenching bacterial signal particles might allow for shaping microbial communities against plant host pathogens (Dong et al., 2001; Ryan et al., 2009).

To support a long shelf life and stability of the product, these compounds can be formulated with specific carrier materials, membrane stabilizers and buffering agents in finetuned quantities (Paau, 1998; Bashan et al., 2014). An example for such soil inoculum carrier is biochar (charcoal used as soil amendment), known to have positive effects on soil properties such as pH (Saxena et al., 2013; Hale et al., 2014) and potentially be amended with extra HBB-specific resources.

#### UTILIZING THE SPECIFICITY OF HOST-BACTERIUM INTERACTIONS

Selection of the appropriate crop plant or a particular bacterial genotype can significantly influence the growth and establishment of HBB in soil (Mazzola, 2004) as interactions between plant and bacterial genotypes are assumed to be highly specific. This specificity could also be used in the human gut hosting defined beneficial strains (Tap et al., 2009). This selection could counteract down-sides of synbiotics, in which the presence of the HBB usually decreases dramatically once the consumption of the prebiotic stops (Bezkorovainy, 2001). For the same reasons, enhancing indigenous soil bacteria should be considered as an alternative to introducing new strains as they are likely to be better adapted to the respective environment (Chaparro et al., 2012).

#### OUTLOOK AND CONCLUSIONS

In our opinion, the future of the HBB application lies in milder treatment of soils by using case-specific nutrient-microbe combinations as well as individualized treatments of patchy field sites after field structure analysis. As summarized in **Figure 1**, we suggest considering approaches such as the use of minor disturbance combined with timely application of HBB to improve their establishment in the soil. Soil treatments could be selected in analogy to therapies chosen for human guts. A new term "**synbiotics for the rhizosphere**" could reflect such intention.

It is assumed, that modern crop plants lost beneficial traits due to breeding programs conducted under conditions with high nutrient supply and the use of chemical pesticides. Consequently, breeding plants for beneficial plant-microbe interactions is an emerging research topic that might give birth to cultivars, which

#### REFERENCES


interact more efficiently with beneficial indigenous strains or with the applied HBB.

We see a sustainable future for agriculture by comparing methods for restoring or retaining the human gut microbiome and those altering the rhizosphere microbiome. Therefore, we suggest a paradigm shift in agricultural practices toward specialized treatment of the rhizosphere microbiomes as described in this work. We invite researchers of agricultural and human health related research areas to compare the methods of both fields and take into consideration findings of the other for their own future work.

#### AUTHOR CONTRIBUTIONS

The authors EA, AG, VK, MR, RS, OT, and SW contributed equally to this opinion paper. GB, WD, and JF shared senior contribution.

#### ACKNOWLEDGMENTS

We thank the graduate schools PE&RC, Ecology & Evolution, SENSE, and the organizers of the post graduate course "Microbial ecology." This course gave the authors the platform for creating the outline of this article. MR participation in the course was possible through the Mobi4Health programme, within the European Union Seventh Framework Programme FP7/2007- 2013 under grant agreement no. 316094. SW contribution was funded by the Netherlands Organisation for Scientific Research NWO (project no. 870.15.050). EA participation through fundings of the Austrian Research Promotion Agency FFG (project no. 836466). AG participation through fundings of NOW-WOTRO. The authors thank Monique Beijaert (NIOOKNAW) who contributed to the design of the figure. This is publication 6097 of the NIOO-KNAW.

practical perspectives (1998–2013). Plant Soil 378, 1–33. doi: 10.1007/s11104- 013-1956-x


compost containing multitude of chitinase-producing bacteria. BioControl 51, 339–351. doi: 10.1007/s10526-005-2934-x


grazing. Appl. Environ. Microbiol. 72, 7083–7090. doi: 10.1128/AEM.00 557-06


**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 © 2016 Adam, Groenenboom, Kurm, Rajewska, Schmidt, Tyc, Weidner, Berg, de Boer and Falcão Salles. 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) or licensor 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.

# Understanding Microbial Multi-Species Symbioses

#### Ines A. Aschenbrenner1,2, Tomislav Cernava1,3, Gabriele Berg<sup>1</sup> and Martin Grube<sup>2</sup> \*

1 Institute of Environmental Biotechnology, Graz University of Technology, Petersgasse, Graz, Austria, <sup>2</sup> Institute of Plant Sciences, University of Graz, Graz, Austria, <sup>3</sup> Austrian Centre of Industrial Biotechnology – Gesellschaft mit beschränkter Haftung, Graz, Austria

Lichens are commonly recognized as a symbiotic association of a fungus and a chlorophyll containing partner, either green algae or cyanobacteria, or both. The fungus provides a suitable habitat for the partner, which provides photosynthetically fixed carbon as energy source for the system. The evolutionary result of the self-sustaining partnership is a unique joint structure, the lichen thallus, which is indispensable for fungal sexual reproduction. The classical view of a dual symbiosis has been challenged by recent microbiome research, which revealed host-specific bacterial microbiomes. The recent results about bacterial associations with lichens symbioses corroborate their notion as a multi-species symbiosis. Multi-omics approaches have provided evidence for functional contribution by the bacterial microbiome to the entire lichen meta-organism while various abiotic and biotic factors can additionally influence the bacterial community structure. Results of current research also suggest that neighboring ecological niches influence the composition of the lichen bacterial microbiome. Specificity and functions are here reviewed based on these recent findings, converging to a holistic view of bacterial roles in lichens. Finally we propose that the lichen thallus has also evolved to function as a smart harvester of bacterial symbionts. We suggest that lichens represent an ideal model to study multi-species symbiosis, using the recently available omics tools and other cutting edge methods.

#### Keywords: lichens, symbiosis, microbiome, Alphaproteobacteria, host-associated bacteria

## INTRODUCTION

Twenty years after the theory of evolution by natural selection started to revolutionize biology, the German mycologist Anton de Bary introduced the term symbiosis to the broader scientific community as a living together of dissimilar organisms (de Bary, 1879). One of his prominent examples were lichens, even though the symbiotic nature – revealed earlier by Schwendener (1869) – was hardly accepted at that time. Scientific peers still considered them as an independent group of organisms with a unique morphology. Meanwhile every biology textbook includes lichens as an obligate association between a fungal (mycobiont) and a photosynthetic partner (photobiont), which can be either cyanobacteria and/or green algae (Nash, 2008). By this association, the photobiont's production of energy via carbon dioxide fixation is enhanced by the sheltering structures of the exhabitant fungal partner. The joint structure, also known as the lichen thallus, is unique and one of the most complex vegetative structures in the entire fungal kingdom. The lichen thallus evolved as early as terrestrial plant life, as the first ancestors of lichens with characteristic morphology can be traced back to the Devonian 400 million years

#### Edited by:

M. Pilar Francino, Foundation for the Promotion of Health and Biomedical Research in the Valencian Region – Public Health, Spain

#### Reviewed by:

Gwenael Piganeau, Centre National pour la Recherche Scientifique, France Devin Coleman-Derr, United States Department of Agriculture – Agricultural Research Service/University of California, Berkeley, USA

#### \*Correspondence:

Martin Grube martin.grube@uni-graz.at

#### Specialty section:

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

Received: 04 November 2015 Accepted: 02 February 2016 Published: 18 February 2016

#### Citation:

Aschenbrenner IA, Cernava T, Berg G and Grube M (2016) Understanding Microbial Multi-Species Symbioses. Front. Microbiol. 7:180. doi: 10.3389/fmicb.2016.00180

ago (Remy et al., 1994; Honegger et al., 2013). In this paper, we will show that lichens are not merely a partnership involving two unrelated organismal groups, but include a so far largely neglected bacterial component, which contributes to the biology of the holobiont. We will start with some general aspects of the lichen ecology and will then continue with an outline how modern analytical tools are used to understand lichens as a fascinating case of a multisymbiosis.

The successful fungal symbiosis, which comprises more than 18,000 named species of fungi is characterized by a poikilohydric lifestyle, which enables lichens to colonize almost all terrestrial environments, ranging from tropical to polar climatic zones, and coastal to high altitude habitats. In addition, lichens grow on the surface of almost every kind of substrate including bare soils, rocks, and plants, but they can be also found in freshwater streams and in marine intertidal zones (Nash, 2008), and various man-made material surfaces. The vegetative bodies vary in color, size (a few millimeters to meters) and growth forms, and some may persist for several 1000 of years (Denton and Karlén, 1973). The wide variety of lichen thallus structures, which are primarily determined by the fungal partner, can be roughly divided into three most common morphological types: crustose, foliose, and fruticose growth forms. Other types exist, but are less frequent (Grube and Hawksworth, 2007). Internally, the vegetative body is either homoiomerous (without stratification), where the mycobiont and photobiont are evenly distributed in the lichen thallus, or heteromerous (with stratification), where at least a fungal upper layer and an algal layer underneath can be distinguished. Crustose lichens are characterized by the attachment of the entire lower surface to the substrate, whereas foliose and fruticose lichens are only partially attached (Büdel and Scheidegger, 2008), and usually have a more or less dense lower fungal layer. Sexual reproduction of the fungal partner requires the development of the species-specific thallus with appropriate algae, since fungal fruit-bodies directly arise in the mature lichen thallus and often incorporate thallus structures. Nevertheless, lichens also evolved various means of asexual reproduction to disperse symbiotic partners together in diverse and specific joint propagules (Büdel and Scheidegger, 2008).

Even though the literature continues to report on antibacterial or antifungal compounds from lichens (reviewed in Boustie and Grube, 2005), the long-lived thalli provide interesting microhabitats for other eukaryotic and prokaryotic (both bacteria and archaea) microorganisms (Lawrey and Diederich, 2003; Grube and Berg, 2009; Bjelland et al., 2011; Bates et al., 2012). In previous years attention was increasingly paid to lichenassociated bacteria that were not recognized as being an integral part of the symbiosis.

In this review we discuss recent literature on lichen-associated microbiota with focus on diversity, functions, dispersal, habitat specificity, and inter-microbiome relations of the Lobaria pulmonaria-associated bacterial community and conclude with an outline to promote a holistic view on lichen-bacteria interactions. In the first part we review historic aspects and then discuss recent results to develop a more holistic lichen model.

### UNRAVELING THE LICHEN-ASSOCIATED MICROBIOME – THEN AND NOW

Bacteria associated with lichens were initially mentioned in the first half of the 20th century (Uphof, 1925; Henkel and Yuzhakova, 1936; Iskina, 1938). During these early studies various bacterial genera were reported to be associated with lichens such as Azotobacter, Pseudomonas (Gammaproteobacteria), Beijerinckia (Alphaproteobacteria), and the Firmicutes genera Bacillus and Clostridium (Iskina, 1938; Panosyan and Nikogosyan, 1966; Henkel and Plotnikova, 1973). At that time descriptions of bacteria underlay solely phenotypical and physiological characterizations indicating a possible role in nitrogen fixation for some of these bacteria. Nevertheless, Lenova and Blum (1983) already estimated that millions of bacterial cells per gram could colonize a lichen thallus. Several decades passed before the first molecular analyses started using bacterial isolates, (e.g., González et al., 2005; Cardinale et al., 2006, Liba et al., 2006, or Selbmann et al., 2009). While González et al. (2005) only focused on culturable Actinomycetes (with Micromonospora and Streptomyces as predominant genera) of various lichen species from tropical and cold areas, Cardinale et al. (2006) attempted to describe the overall bacterial community composition associated with seven different lichen species from temperate habitats. The latter enabled the identification of several genera affiliated to Firmicutes, Actinobacteria, and Proteobacteria, highlighting Paenibacillus, and Burkholderia to be ubiquitous genera in lichens. However, culture-dependent methods capture only 0.001–15% of the bacterial diversity in environmental samples (Amann et al., 1995), whereas the majority remains unobserved (Rappé and Giovannoni, 2003). To overcome the limitations of selective bacterial isolation from environmental samples and to obtain a more unbiased and less restricted view on the microbial communities, new techniques were employed to complement the traditional methods.

First culture-independent investigations on lichen-associated microbiota were assessed with different fingerprinting methods (Cardinale et al., 2006; Grube et al., 2009; Bjelland et al., 2011; Mushegian et al., 2011; Cardinale et al., 2012a) and molecular cloning approaches (Hodkinson and Lutzoni, 2009). Such techniques (e.g., DGGE: Muyzer and Smalla, 1998; T-RFLP: Liu et al., 1997; SSCP: Schwieger and Tebbe, 1998) were used to generate microbial community profiles by amplifying genetic markers (e.g., 16S ribosomal DNA) with universal primers. Based on sequence or length polymorphisms PCR products are separated and the degree of sample similarity according to the specific band patterns can be characterized (Smalla et al., 2007). Although many samples can be analyzed in parallel and their profiles can be compared with each other easily, the identification of the bacterial community members in detail is tedious and limited. Margulies et al. (2005) introduced a new time reduced and cost efficient technology to study community compositions and diversity of environmental samples in depth by large-scale high throughput sequencing. Bates et al. (2011) described lichenassociated bacteria for the first time based on this next generation

pyrosequencing technology, followed by Grube et al. (2012), Hodkinson et al. (2012), and Aschenbrenner et al. (2014).

With the improvement of sequencing technologies and bioinformatics tools the focus in microbial ecology research shifted from the basic taxonomical descriptions to a more detailed and holistic view on microbial communities. Metagenomic, transcriptomic, and proteomic analyses can now shed light on the questions "Who is there?", "What are they capable of?", and "Who is actively doing what?" (Schneider et al., 2011; Aschenbrenner, 2015; Grube et al., 2015). To address these questions, the lung lichen L. pulmonaria (L.) Hoffm. was used as model system due to its relatively fast growth and other facilitative characteristics, e.g., epiphytic growth on tree bark and a low number of secondary metabolites, which could interfere with the conducted analyses. L. pulmonaria is characterized by a leaf-like structure (foliose lichen) and mainly found in old-growth forests with unpolluted air. Its sensitivity to air pollution can be employed for indirect evaluations of air quality and ecosystem integrity (Scheidegger and Werth, 2009). It harbors two photosynthetic partners, a phenomenon observed for approximately 4% of all described lichens (Honegger, 1991). However, only the green alga Dictyochloropsis reticulata forms a continuous layer, whereas cyanobacterial Nostoc strains are maintained in spaced, nodule-like internal compartments (cephalodia).

#### COMPOSITION AND DIVERSITY OF THE LICHEN-ASSOCIATED MICROBIOME DRIVEN BY VARIOUS ABIOTIC AND BIOTIC FACTORS

The amount of bacteria found on lichens is surprisingly high in relation to surfaces of higher plant foliage. While a leaf surface comprises only 10<sup>5</sup> cells/cm<sup>2</sup> , some lichen species analyzed for bacterial abundance exceed this value dramatically (Saleem, 2015). For example, Cladonia rangiferina, is colonized by approximately 107–10<sup>8</sup> bacteria per gram of lichen thallus (Cardinale et al., 2008; Grube et al., 2009). Moreover, Alpha diversity indices (Shannon index) of bacterial communities were shown to vary between different lichens, e.g., from on average 4.5 (Solorina crocea) to 7.0 (L. pulmonaria) at a genetic distance of 3% among the microbial OTUs based on 16S rRNA gene sequence dissimilarity (Grube et al., 2012; Aschenbrenner et al., 2014).

L. pulmonaria is mainly colonized by Alphaproteobacteria with Sphingomonadales as the predominant order, followed by Sphingobacteria, Actinobacteria, and Spartobacteria (Aschenbrenner et al., 2014). Contrarily, shotgun sequencingbased studies suggested Rhizobiales as the main order within Alphaproteobacteria (Erlacher et al., 2015; Grube et al., 2015). These results were additionally confirmed with adapted visualizing techniques. Thereby, the predominance of Alphaproteobacteria and Rhizobiales on lichen surfaces were shown with a combined approach of fluorescence in situ hybridization (FISH) and confocal laser scanning microscopy (CLSM). Related to these findings, the lichen-associated Rhizobiales group (LAR1) was reported to be a lichen-specific lineage of Alphaproteobacteria, which can be found among many examined species (Hodkinson and Lutzoni, 2009; Bates et al., 2011; Hodkinson et al., 2012). However, this lineage could not be detected in L. pulmonaria (Aschenbrenner et al., 2014). The observed compositional differences within the same lichen species can be attributed to various reasons such as metagenomic sequencing approach (amplicon vs. shotgun sequencing), utilized databases, or activity of the bacteria in case of metatranscriptomic analysis (Aschenbrenner, 2015) since less than 10% of a microbial community is metabolically active at one time (Locey, 2010).

While the predominance of Alphaproteobacteria was also reported in other studies (Bates et al., 2011; Hodkinson et al., 2012), bacterial community composition in general differed among lichen species. These variations are supposed to be driven by various biotic and abiotic factors. Hodkinson et al. (2012) who thoroughly studied the bacterial communities associated with various lichen species comprising 24 mycobiont types with all photobiont combinations of different sampling locations (tropical and arctic regions) highlighted the photobiont type (chlorolichens vs. cyanolichens) and large-scale geography as the main driving forces.

Hodkinson et al. (2012) argued that the differences in community composition could be ascribed to both the availability of fixed nitrogen and the type of fixed carbon. Regarding the first one, bacteria associated with cyanolichens have access to fixed atmospheric nitrogen due to the cyanobacterial photobiont, whereas those of chlorolichens lack this benefit in nitrogenrestricted environments. According to that, chlorolichens would preferably enrich species capable of nitrogen fixation rather than cyanolichens. Another suggestion was that green algae release different types of fixed carbon (sugar alcohols: ribitol, erythritol, or sorbitol) than cyanobacteria (glucose; Elix and Stocker-Wörgötter, 2008), thereby shaping the bacterial community with respect to carbon utilization. Both explanations can only partly explain community differences based on taxonomic descriptions as bacteria can exchange and share genes encoding for certain functions via horizontal gene transfer. This agrees with Burke et al. (2011) who argued that ecological niches are colonized randomly by bacteria equipped with suitable functions rather than following bacterial taxonomy. The attempt to explain observed community compositions gets more complicated with regard to tripartite lichens as they carry both types of photobionts as it is the case in L. pulmonaria.

Species-specificity for bacterial communities associated with chlorolichens was already indicated in previous studies (Grube et al., 2009; Bates et al., 2011). Lichenized fungi are able to produce secondary metabolites, which are unique to lichens and comprise several 100 compounds which can be deposited on the extracellular surface of the fungal hyphae (Elix and Stocker-Wörgötter, 2008). As already suggested by Hodkinson et al. (2012) the considerable fraction of secondary metabolites with antimicrobial activities (Kosanic and Rankovi ´ c, 2015 ´ ) might cause a selective pressure on lichen-colonizing bacteria as well. However, as L. pulmonaria contains only low concentrations of lichen-specific substances like many other lichens of the suborder

Peltigerineae (Beckett et al., 2003), secondary metabolites might play only a minor role in shaping the community structure of Lobaria-associated bacteria.

Differences in bacterial community composition might be also due to the lichen growth type as for instance previous studies reported that the bacterial community compositions of crustose lichens differed from those of foliose or fruticose lichens (Grube et al., 2009; Hodkinson et al., 2012). While the foliose lichens were mainly colonized by Alphaproteobacteria, the crustose lichen Ophioparma sp. was dominated by Acidobacteria (Hodkinson et al., 2012). Another rock-inhabiting crustose lichen Hydropunctaria sp. was mainly colonized by Cyanobacteria, Actinobacteria, and Deinococcus (Bjelland et al., 2011). But growth type on its own does not explain the predominance of certain taxa since the foliose lichen Solorina sp. was also dominated by Acidobacteria (Grube et al., 2012). This agrees with previous results of Cardinale et al. (2012b) who showed that growth types do not affect the main bacterial community structure.

### BACTERIA ARE SPATIALLY STRUCTURED ON LICHENS

Thallus sub-compartments of varying age as well as external and internal surfaces offer chemically and physiologically distinct micro-niches and facilitate the formation of various distinct bacterial communities. Based on FISH and CLSM the lichenassociated eubacteria as well as specific bacterial taxa therein were demonstrated to colonize distinct lichen thallus parts in different abundances and patterns (Cardinale et al., 2008). Confocal laser scanning microscopy of the L. pulmonaria surfaces showed that both the upper and the lower cortexes were evenly colonized by Alphaproteobacteria among other eubacteria (Cardinale et al., 2012a; Grube et al., 2015). This was also demonstrated for other dorsiventrally organized lichen thalli such as the leafy Umbilicaria sp. (Grube et al., 2009). In the case of the shrubby species Cladonia the outer cortex of the radially organized hollow thallus (podetium) was merely colonized by single cell colonies and smaller colony clusters, while the highest bacterial density examined on this lichen was found on the internal layer of the podetia forming a biofilm-like coat (Cardinale et al., 2008, 2012b). Contrarily, bacterial colonization on crustose lichens such as Lecanora sp. was distinctly higher in the cracks between the areoles of the thalli (Grube et al., 2009). There were also first indications for endobiotic bacteria within the cell walls of fungal hyphae (Cardinale et al., 2008). Erlacher et al. (2015) previously reported in L. pulmonaria endosymbiotic Rhizobiales, localized in varying depths of the interhyphal gelatinous matrix of the upper cortex and seldom in the interior of fungal hyphae. So far, there is no documentation of bacterial growth in other compartments of L. pulmonaria such as the internal thalline tissue (medulla) or the photobiont layer.

The age states in a mature lichen thallus might influence and shape bacterial community structure, which resembles the community succession found, e.g., in the apple flower microbiome (Shade et al., 2013). A recent study has shown that the vegetative propagules of L. pulmonaria were colonized by a more distinct bacterial community than the mature lichen thallus (Aschenbrenner et al., 2014) indicating that the community structure might change over time. In detail, only 37% of thallus-associated bacterial OTUs were shared with the vegetative propagules, conversely, shared OTUs associated with the propagules comprised 55%. While both lichen parts were mainly colonized by Alphaproteobacteria, the lichen thallus was additionally dominated by Deltaproteobacteria, whereas the juvenile vegetative propagules were also colonized in higher abundances by Spartobacteria and Sphingobacteria. Previously, Cardinale et al. (2012b) reported that older thallus parts hosted significantly higher amounts of bacteria than the younger thallus structures including a change of the predominant Alphaproteobacteria to other taxa such as Actinobacteria, Gamma-, and Betaproteobacteria. Also Mushegian et al. (2011) observed a spatial diversification of the bacterial compositions between the more diverse and consistent thallus centers (older parts) and those of the more variable and species poor edges (younger parts). Cardinale et al. (2012b) referred to this bacterial distribution patterns as anabolic centers in the growing and catabolic sinks in the senescing parts of the lichen thallus, respectively. The hypothesis of recycling nutrients in the decaying lichen parts by bacteria can be also underpinned by the presence of specific taxa known for their degradation potential. Sphingomonas sp., which are known to degrade organic matter and xenobiotic substances, were previously isolated from lichens sampled in Arctic and Antarctic regions (Lee et al., 2014), but also reported in other studies (Grube et al., 2009, 2012; Hodkinson et al., 2012; Aschenbrenner, 2015). However, also other genera such as Paenibacillus and Streptomyces were mentioned for their functions (e.g., chitinolytic activity) in the degradation of lichen tissues (Cardinale et al., 2006).

### DISTRIBUTION AND TRANSFER OF HOST-ASSOCIATED BACTERIA

Analyses of lichen-associated bacteria revealed differences in community composition and diversity among geographically distant habitats (Printzen et al., 2012; Aschenbrenner et al., 2014). Printzen et al. (2012) analyzed the geographic structure of lichenassociated Alphaproteobacteria in Antarctic regions indicating that this group is affected by environmental parameters since thalli from sub-polar habitats had more similar communities than those from extrapolar regions. Hodkinson et al. (2012) explained these large-scale geographical effects by the dispersal efficiency of the lichen hosts, where the dispersal happens on small spatial scales rather than on large-scale distances resulting in a geographic differentiation of the community composition. Aschenbrenner et al. (2014) visualized and described the bacterial colonization of lichen propagules. Their results demonstrate that at least a certain proportion of the lichen microbiome is transferred vertically via these symbiotic structures. These bacterial communities were dominated by Alphaproteobacteria, as was already found by Cardinale et al. (2012a). Interestingly, the bacterial consortia of the lichen propagules were more

than only a subset of the parental thallus microbiome and also comprised unique species, not shared by the mature thallus. Thus, Aschenbrenner et al. (2014) suggested that the vegetative propagules are equipped with a bacterial starter community. Such bacteria colonizing juvenile structures might influence the subsequent recruitment of new bacteria (Fukami, 2010), thereby shaping the community composition. The importance of the lichen-associated bacteria during the establishment of the lichen symbiosis was already suggested (Hodkinson and Lutzoni, 2009), as the growth of stratified lichen thalli was so far only successful in cultures based on lichen fragments, which apparently include bacteria.

Although vertical transmission of lichen-associated bacteria was only shown in a single lichen species, it is very likely that this strategy of microbiome transfer is also common in other species utilizing vegetative diaspores for reproduction, and definitely in other symbioses. There are various examples reporting on a transmission of host-associated bacteria (Bright and Bulgheresi, 2010), e.g., in marine sponges (Wilkinson, 1984; Li et al., 1998). Bacteria associated with terrestrial invertebrates such as insects are known to assist in nutrient uptake and provision of essential amino acids and vitamins (Douglas, 1998; Feldhaar and Gross, 2009), but their vertical transmission strategies vary among distinct species (Sacchi et al., 1988; Attardo et al., 2008; Prado and Zucchi, 2012). In vertebrates including humans the transfer of maternal microbes to the child through natural birth and breast feeding as first inoculum was reported to be important for the baby's health, in particular by shaping the microbiome structure with beneficial microbes (Funkhouser and Bordenstein, 2013). But also in the plant kingdom transfer of plant-associated bacteria, in particular of seeds, from the mother plant was reported (van Overbeek et al., 2011), even though it is common for higher plants to recruit their substantial rhizosphere communities from the surrounding soil (Berg and Smalla, 2009). Vertical transmission was previously shown for the oldest group of land plants, mosses, which belong together with lichens to the group of poikilohydric cryptogams; associated bacteria, especially specific Burkholderia strains, are transferred from the sporophyte to the gametophyte via spores (Bragina et al., 2012, 2013).

#### Lichens as Bacterial Hubs

Lichens are pioneers in the colonization of hostile environments with extreme temperatures, desiccation, and high salinity, but they may also become very old, either as individuals or as associations (it is assumed that some non-glaciated sites were colonized by lichens since the tertiary). Colonized habitats include arid and semi-arid regions where bare soil can be colonized by, e.g., cryptogamic soil crusts (an association comprising soil particles, lichens, cyanobacteria, algae, fungi, and bryophytes; Beckett et al., 2008), but also more extreme regions such as deserts, where lichens are one of the few successful colonizers. In particular, their capability to become hydrated without contact to liquid water (Printzen et al., 2012) only by fog, dew or high air humidity (Beckett et al., 2008) ensures survival in these dry areas. This suggests that lichens as slow-growing and long-living host organisms might serve as bacterial hubs in these environments facilitating their survival by nutrient and water supply, offering a habitat with various micro-niches and ensuring their distribution over short distances by the dispersal strategies of the lichen host. Thereby the lichens could be important sources/reservoirs of beneficial bacterial strains for other habitats in an environment as well.

#### Habitat Specificity

Host specificity for cryptogams (i.e., lichens and mosses) was already reported in previous independent studies (Grube et al., 2009; Bragina et al., 2012). However, bacterial communities were so far described almost always without a view of adjacent habitats and potential inter-microbiome relationships. Previously bacterial specificity was reported in studies of lichen thalli and their underlying rock substrate (Bjelland et al., 2011). A recent study within the doctoral thesis of Aschenbrenner (2015) focusing on this topic unraveled the specificity of the lichen-associated microbiome compared with the neighboring habitats, i.e., moss and bare bark. This comparative analysis highlighted potential habitat specialists and generalists. In this survey, members of the genus Sphingomonas were identified as generalists in all the three habitats, whereas members of Mucilaginibacter were described as potential specialists of lichens. The lung lichen frequently establishes on mosses, and the sharing of Nostoc strains between both cryptogams suggests a previously undescribed form of ecological facilitation that is mediated by the shared microbiome fraction (Aschenbrenner, 2015). The lung lichen takes up Nostoc strains during growth and incorporates them in the thallus as distinct clusters (known as internal cephalodia in the literature). As Nostoc is enriched on mosses rather than on bark, the growth promoting effect of nitrogen-fixing Nostoc apparently facilitates the efficient development of the lichen thallus, which mostly emerges from moss patches.

### THE LICHEN-ASSOCIATED MICROBIOME PLAYS A CENTRAL FUNCTIONAL ROLE IN THE LICHEN HOLOBIONT

While the host-specific bacterial colonization of various lichen species was demonstrated over the past years, the roles of the bacteria remained largely unknown. This is mainly due to inherent problems to study lichens by experimental approaches (especially re-synthesis of the symbiosis in culture). Metaomics meanwhile emerged as a set of suitable technologies to globally identify potentially beneficial contributions of the bacterial population. Recently, the L. pulmonaria associated microbiome was investigated with an integrated metagenomics and metaproteomics approach to screen for potential functions encoded in genomes and to verify their expression at the protein level (Grube et al., 2015), based on a previous pioneering proteomics study (Schneider et al., 2011). The results of Grube et al. (2015) provided strong evidence that the bacterial

microbiome is involved in nutrient provision and degradation of older lichen thallus parts, biosynthesis of vitamins and hormones, detoxification processes, and the protection against biotic as well as abiotic stress. Additionally, the high prevalence of bacterial nitrogen fixation was confirmed with –omic data and quantitative RT-PCR. Moreover, a comparison of the whole Lobaria-associated metagenome with a representative set of publicly available metagenomes highlighted its uniqueness. The most closely related metagenomes were found to be those obtained from plant-associated habitats.

In particular, Rhizobiales (Alphaproteobacteria) were previously shown to be remarkably abundant in the L. pulmonaria microbiome mainly represented by the families: Methylobacteriaceae, Bradyrhizobiaceae, and Rhizobiaceae. Although they are well known for their beneficial interactions with many higher plants, less is known about their specific roles in terms of the lichens. According to Erlacher et al. (2015) functional assignments based on hierarchical SEED classification indicated an involvement of Rhizobiales in various beneficial functions (e.g., auxin, folate, and vitamin B12 biosynthesis). A further breakdown demonstrated that the predominant Methylobacteriaceae were also the most potent producers of the examined metabolites. These findings suggest the potential for various biotechnological applications of this group.

### Stress Amelioration and Pathogen Defense Functions are Supported by Metagenomic Data and Culturable Members of the Microbiome

Recently, it was shown that the L. pulmonaria associated microbiome includes also various bacteria with antagonistic potential (Cernava et al., 2015a). The most abundant antagonists were assigned to Stenotrophomonas, Pseudomonas, Micrococcus, and Burkholderia. These genera accounted for 67% of all identified antagonistic bacteria. Metagenomic screening revealed the presence of genes involved in the biosynthesis of stress-reducing metabolites. Complementary high-performance liquid chromatography-mass spectrometry (HPLC-MS) analyses enabled the detection of Stenotrophomonas-produced spermidine which is known to reduce desiccation- and high-salinity-induced stress in plants. It was also tested if these protective effects can be transferred to non-lichen hosts such as primed tomato (Solanum lycopersicum) seeds. Results indicated a significant increase in the root and stem lengths under water-limited conditions. The application of lichen-associated bacteria in plant protection and growth promotion may prove to be a useful alternative to conventional approaches. However, further studies are required to evaluate the host range and to elucidate the overall applicability (Cernava et al., 2015a).

Furthermore, volatile organic compounds (VOCs) profiles from bacterial isolates showed that lichen-associated bacteria are emitting a broad range of volatile substances. These molecules are most likely involved in various interactions (e.g., communication between microorganisms and the host) and might also increase the overall resistance against various pathogens (Cernava et al., 2015b).

### The Microbiome Provides Complementary Detoxification Mechanisms

Besides the evidence for mechanisms conferring enhanced resistance against biotic as well as abiotic stress, the microbiome provided a first evidence for the involvement in the detoxification of inorganic substances (e.g., As, Cu, Zn), the detailed mechanisms remaining unknown. A deeper insight into these beneficial contributions was possible with samples exposed to elevated arsenic concentration (Cernava, 2015). Metagenomic analyses revealed that the overall microbial community structures from different lichens were similar, irrespective of the arsenic concentrations at the sampling locations, whereas the spectrum of functions related to arsenic metabolism was extended. These functions include bioconversion mechanisms that are involved in the methylation of inorganic arsenic and consequently generate less toxic substances. Furthermore, the abundance of numerous detoxification related genes was enhanced in arsenic-polluted samples. Supplementary qPCR approaches have shown that the arsM gene copy number is not strictly related to the determined arsenic concentrations. Additionally, a culture collection of bacterial isolates obtained from three lichen species was screened for the arsM gene. Detected carriers of arsM were later identified as members of the genera Leifsonia, Micrococcus, Pedobacter, Staphylococcus, and Streptomyces. The overall results underscored the important role of the microbiome in host protection and they provided more detailed insights into the taxonomic structure of involved microorganisms.

### BACTERIAL MICROBIOME ASSEMBLY ON A SYMBIOTIC FUNGAL STRUCTURE

The lichen thallus with its various micro-niches represents a miniature ecosystem for microorganisms. While lichenassociated bacteria were previously neglected and often recognized as contamination of lichen thalli, recent research considers them – with increasing evidence – as important and crucial component of the lichen meta-organism. By their microbiomes lichens are ecologically linked with their surrounding environment (**Figure 1**). Even though a fraction of their microbiome can be transmitted by local dispersal of vegetative propagules, further recruitment of strains occurs from the local resources in the environment. This finally leads to a specific community structure of mature lichen thalli, which shares a core microbiome over larger distance (Aschenbrenner et al., 2014). Lichen thalli, already present on Earth since the lower Devonian, and representing the most complex vegetative structures in the fungal kingdom, may have evolved as bacterial enrichment structures. The exposed surfaces of lichens are ideally suited to benefit from functions of adapted and enriched bacteria, or from degradation of spurious non-adapted bacteria caught from the environment. The bacterial harvest may readily be dissipated to the symbiotic corporates via the fungal textures. It is this new perspective of the lichen symbiosis, which offers a wide range of new research questions in the near future.

### CONCLUSIONS – LICHENS AS A CASE MODEL TO UNDERSTAND MULTI-SPECIES SYMBIOSES

Undoubtedly, there exist other cases of symbioses involving multiple organismal groups in terrestrial ecosystems. Similar to lichens, these were originally recognized as dual eukaryotic partnerships, but later shown to involve specific bacterial associations as well (e.g., fungi/leaf-cutter ants, Little and Currie,

contributes a variety of beneficial functions for the host symbiosis (purple circle).

2007; mycorrhiza, Garbaye, 1994). Modern tools now overcome the difficulties to re-establish complex symbioses under axenic laboratory conditions, and moreover, they allow us to precisely study symbioses in their environmental context. We consider lichens as ideal research objects for this purpose, because in contrast to many other symbiotic systems, they have an unsurpassed ecological range in general, but with rather specific adaptation of each species to their ecological niches. It will thus clearly be a novel and highly interesting theme in symbiotic research to establish the role of the microbiome in ecological adaptation and evolution of the lichen multi-species symbiosis.

#### AUTHOR CONTRIBUTIONS

fmicb-07-00180 February 16, 2016 Time: 16:27 # 8

IA, TC, GB, and MG wrote the manuscript. IA and TC contributed with results from their Ph. D. studies. GB and MG

#### REFERENCES


complemented the manuscript with profound experience in the fields of microbiome and lichen research.

#### ACKNOWLEDGMENTS

This work was supported by a grant of the Austrian Science Fund (FWF) to GB and MG (FWF Project I882).


de Bary, A. (1879). Die Erscheinung der Symbiose. Strasburg: Triibner.



Schwendener, S. (1869). Die Algentypen der Flechtengonidien. Basel: Schultze.


**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 © 2016 Aschenbrenner, Cernava, Berg and Grube. 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) or licensor 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.

# Critical Issues in Mycobiota Analysis

Bettina Halwachs 1, 2, 3 \*, Nandhitha Madhusudhan1, 2, Robert Krause<sup>4</sup> , R. Henrik Nilsson<sup>5</sup> , Christine Moissl-Eichinger 3, 4, Christoph Högenauer 2, 3, 6, Gerhard G. Thallinger 3, 7 and Gregor Gorkiewicz 1, 2, 3 \*

1 Institute of Pathology, Medical University of Graz, Graz, Austria, <sup>2</sup> Theodor Escherich Laboratory for Medical Microbiome Research, Medical University of Graz, Graz, Austria, <sup>3</sup> BioTechMed-Graz, Interuniversity Cooperation, Graz, Austria, <sup>4</sup> Section of Infectious Diseases and Tropical Medicine, Department of Internal Medicine, Medical University of Graz, Graz, Austria, <sup>5</sup> Department of Biological and Environmental Sciences, University of Gothenburg, Gothenburg, Sweden, <sup>6</sup> Division of Gastroenterology and Hepatology, Department of Internal Medicine, Medical University of Graz, Graz, Austria, <sup>7</sup> Institute of Molecular Biotechnology, Graz University of Technology, Graz, Austria

Fungi constitute an important part of the human microbiota and they play a significant role for health and disease development. Advancements made in the culture-independent analysis of microbial communities have broadened our understanding of the mycobiota, however, microbiota analysis tools have been mainly developed for bacteria (e.g., targeting the 16S rRNA gene) and they often fall short if applied to fungal marker-gene based investigations (i.e., internal transcribed spacers, ITS). In the current paper we discuss all major steps of a fungal amplicon analysis starting with DNA extraction from specimens up to bioinformatics analyses of next-generation sequencing data. Specific points are discussed at each step and special emphasis is placed on the bioinformatics challenges emerging during operational taxonomic unit (OTU) picking, a critical step in mycobiota analysis. By using an in silico ITS1 mock community we demonstrate that standard analysis pipelines fall short if used with default settings showing erroneous fungal community representations. We highlight that switching OTU picking to a closed reference approach greatly enhances performance. Finally, recommendations are given on how to perform ITS based mycobiota analysis with the currently available measures.

#### Edited by:

David Berry, University of Vienna, Austria

#### Reviewed by:

Carlotta De Filippo, National Research Council, Italy Micah Egge Dunthorn, Kaiserslautern University of Technology, Germany

#### \*Correspondence:

Bettina Halwachs bettina.halwachs@medunigraz.at Gregor Gorkiewicz gregor.gorkiewicz@medunigraz.at

#### Specialty section:

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

Received: 20 May 2016 Accepted: 24 January 2017 Published: 14 February 2017

#### Citation:

Halwachs B, Madhusudhan N, Krause R, Nilsson RH, Moissl-Eichinger C, Högenauer C, Thallinger GG and Gorkiewicz G (2017) Critical Issues in Mycobiota Analysis. Front. Microbiol. 8:180. doi: 10.3389/fmicb.2017.00180 Keywords: microbiota, mycobiota, internal transcribed spacer (ITS), 16S rRNA gene, multiple sequence alignment (MSA), OTU picking, formalin-fixed paraffin-embedded tissue (FFPE), DNA isolation

#### INTRODUCTION

It is now well-established that the microbiota contributes significantly to human health and disease. So far, microbiota investigations have been mainly focused on bacteria, but also archea, viruses, and micro-eukaryotes such as protozoa and fungi are part of human-associated microbial communities. Fungi are prevalent in all microbially colonized body habitats including skin, the gastrointestinal (GI)-, urogenital-, and respiratory tract (Charlson et al., 2012; Findley et al., 2013; Hallen-Adams et al., 2015). Up to now more than 390 fungal species have been described in humans (Oever and Netea, 2014; Gouba and Drancourt, 2015). Depending on the habitat the abundance of fungal cells varies from <0.1% of microorganisms in the GI tract to up to 10% on skin (Belkaid and Naik, 2013). An average fungal cell is about 100-fold larger than an average bacterial cell, which translates into a significant fungal biomass, providing abundant bioactive molecules to the host and shaping its physiology (Underhill and Iliev, 2014). The GI mycobiota actively interacts with the immune system, for instance through the human innate immune receptor Dectin-1 able to dampen GI inflammation (Iliev et al., 2012). A balanced mycobiota prevents from hyperinflammation of the GI tract and alterations in fungal community composition due to antifungal drugs exacerbate colitis in mice (Wheeler et al., 2016). In humans genetic defects in certain immuneregulatory genes (e.g., STAT1, CARD9, etc.) or Il-17 and Il-22 signaling pathways lead to severe fungal syndromatic infections, such as chronic mucocutaneous candidiasis or the APECED (Autoimmune Polyendocrinopathy, Candidiasis, Ectodermal Dystrophy) syndrome (Oh et al., 2013; Underhill and Iliev, 2014). Compositional mycobiota shifts are reported in various diseases (Cui et al., 2013) and also interdependencies between the fungal and bacterial component of the microbiota exist. They are exemplified by disease-specific inter-kingdom alterations, reported for instance in inflammatory bowel disease (IBD, Ott et al., 2008; Hoarau et al., 2016; Sokol et al., 2016) or in the lung microbiome of cystic fibrosis patients (Kim et al., 2015). Importantly, fungi contribute significantly to human infections, especially in immune-compromised, chronically ill and intensive care patients wherein the respiratory or GI tract are often the origins of fungal systemic infections (Brown et al., 2012; Krause et al., 2016).

#### INTERNAL TRANSCRIBED SPACERS (ITS) AS FUNGAL MOLECULAR BARCODES

Currently, amplicon-based next generation sequencing is the standard measure for the culture-independent assessment of the mycobiota. Also metagenomic approaches are increasingly used, providing functional insights into the mycobiota. However, their broad application is still too costly due to the required sequencing effort to capture the relatively rare fungal biosphere and the special needs for bioinformatics analysis paired with underdeveloped fungal reference genome databases make metagenomics approaches still cumbersome (Tang et al., 2015). Early culture-independent mycobiota investigations used the eukaryotic 18S ribosomal RNA gene, in analogy to the prokaryotic 16S rRNA gene, as molecular target enabling PCR amplification of fungal DNA and subsequent taxonomic profiling via sequence analysis (Simon et al., 1992; Kappe et al., 1996; Smit et al., 1999; Hunt et al., 2004). The 18S rRNA gene, however, is less discriminatory for fungi compared to its prokaryotic equivalent often failing to discriminate fungi at lower taxonomic levels, such as genus or species (Hartmann et al., 2010; Lindahl et al., 2013).

The prokaryotic and the eukaryotic rRNA operons exhibit different genetic architectures (**Figures 1A,B**). The eukaryotic rRNA cistron consists of the 18S (small subunit, SSU), 5.8S, and 28S (large subunit, LSU) rRNA genes transcribed as a unit by RNA polymerase I, including two internal transcribed spacer regions, ITS1 and ITS2, flanking the 5.8S rRNA gene. The two ITS regions are post-transcriptionally removed and are absent in the mature ribosome. Since they are dispensable for ribosome function, they experience a lower evolutionary pressure leading to higher sequence variability (**Figures 1C–E**). The increased level of sequence variability enables discrimination of even closely related taxa (e.g., at species level). In addition ITS sequences seem to represent superior molecular targets for fungal PCR amplification compared to SSU and LSU sequences, signified by higher positive PCR amplification rates (Schoch et al., 2012). Based on these observations, the Fungal Barcoding Consortium recently denoted the ITS region as the universal barcode for fungi superior to other molecular markers (Schoch et al., 2012).

In the following sections, we discuss the main steps of amplicon-based mycobiota analyses with special emphasis on the bioinformatics challenges emerging if standard bioinformatics analysis pipelines such as mothur, QIIME, or MICCA are employed (Schloss et al., 2009; Caporaso et al., 2010; Albanese et al., 2015).

### FUNGAL DNA ISOLATION

A variety of studies have shown that DNA isolation methods and oligonucleotide primer choice significantly influence the outcome of molecular phylogenetic surveys (Gorkiewicz et al., 2013; Tedersoo et al., 2014; Hallen-Adams et al., 2015). Numerous protocols and kits are available for isolation of fungal DNA and they follow similar basic principles with slight modifications dependent on the specimen type used (Paulino et al., 2006; Ghannoum et al., 2010; Findley et al., 2013; Lindahl et al., 2013; Gosiewski et al., 2014; Oh et al., 2014). The basic protocol involves mechanical cell disruption using bead beating, followed by enzymatic cell lysis. Especially the addition of lyticase, and endoglucanase hydrolyzing the covalent bounds between β-(1-3)-D-glucose molecules in the fungal cell-wall glycan, is an essential step to enable complete fungal cell lysis (Muñoz-Cadavid et al., 2010; Goldschmidt et al., 2014). The final DNA purification step is often performed by using membranebased procedures (van Burik et al., 1998; Lindahl et al., 2013).

Aside of typically sampled native material (e.g., swabs, etc.) also other resources for mycobiota investigations exist. Formalinfixed paraffin-embedded (FFPE) tissue samples play an important role in the clinical context. Biopsies or surgically removed tissues are typically fixed in formalin (10%) immediately after they are collected from the patient, thus they represent a well-preserved resource for the analysis of biomolecules including nucleic acids (Sangoi et al., 2009; Kocjan et al., 2015). FFPE specimens are typically used for diagnostic purposes (e.g., histopathology) but are also amenable for molecular scientific investigations. Their prevalence in biological repositories such as biobanks make them ideal specimens to study the mycobiota in the context of human disease (Yuille et al., 2008). About 70 commercially kits are available for DNA extraction out of FFPE material (Kocjan et al., 2015), however, nucleic acid isolation from FFPE material is challenging. Biomolecules are typically cross-linked and fragmented due to formalin, and factors such as the pH of the fixative, duration of fixation, and importantly the DNA extraction method applied greatly influence the quality of the extracted DNA (Bonin and Stanta, 2013; Kocjan et al., 2015). Factors such as residual formalin inhibiting proteinase K activity and omitting complete cell lysis, as well as the presence of PCR inhibitors in the DNA extract might altogether interfere with successful fungal DNA amplification (Coura et al., 2005; Muñoz-Cadavid et al., 2010).

#### FIGURE 1 | Continued

sequence alignment (MSA) of the entire 16S rRNA operon of five different bacterial species (encompassing five different phyla). Variable regions (V1–V9) are highlighted in blue, conserved regions in yellow, positions according to the E. coli 16S rRNA (GenBank acc. no.: J01695.2). (D) MSA of the complete internal transcribed spacer region of five different fungal species of the same genus (Hydnum sp.). (E) MSA of the complete ITS region of seven fungal taxa representing different phyla. Information about sequences used for MSA generation (C,D) is given as Supplementary Tables S3–S5.

FIGURE 2 | DNA isolation from human FFPE skin samples and ITS PCR amplification influenced by beat beating. (A) Significant difference in overall DNA yield from FFPE skin samples (n = 10) with and without bead beating (\*\*p < 0.005 by Mann Whitney test; data are mean + SEM). (B) Significantly increased detection of fungal DNA isolated without bead beating by ITS2 qPCR (n = 10; \*p < 0.05, \*\*\*p < 0.005, Kruskal-Wallis test; data are mean + SEM). NTC, negative control.



\*ns, not specified.

These difficulties make a thorough review of the (pre-) analytical process of mycobiota studies mandatory. To highlight the influence of pre-analytics on ITS based mycobiota investigations we assessed the performance of DNA extraction from human skin FFPE samples (see Supplementary Table S1 for sample information) with a commercially available kit (QIAamp DNA FFPE tissue kit, Qiagen) reported to be efficient for fungal DNA extraction out of FFPE material (Muñoz-Cadavid et al., 2010). We added a mechanical cell disruption step (bead-beating) to the procedure (MagnaLyser, Roche), since this step was shown to be crucial for complete lysis of microbial cells in specimens, significantly influencing correct community representation (de Boer et al., 2010; Reck et al., 2015). A detailed description of the applied method is given in the Data Sheet S1. Interestingly, we observed that bead-beating significantly lead to lower DNA yields and a significantly decreased signal-to-noise ratio in ITS PCR, impairing efficient fungal PCR amplification (**Figure 2**). Thus, mechanical lysis of specimens could also counteract reliable mycobiota investigations especially if low-biomass samples such as skin are used.

### ITS AMPLIFICATION VIA PCR

For amplification of fungal DNA various primers have been designed targeting different regions of the rRNA operon or other marker genes encoding translation elongation factor 1 α, RNA polymerase II, β-tubulin, and the minichromosome maintenance complex component 7 (MCM7) protein (White et al., 1990; Tanabe et al., 2002; McLaughlin et al., 2009; O'Donnell et al., 2010; Schoch et al., 2012; Toju et al., 2012; Lindahl et al., 2013). Of these, the ITS regions are considered the formal barcode for fungal taxonomy (Schoch et al., 2012; Lindahl et al., 2013). As noted above, ITS1 and ITS2 sequences are highly variable and can be used to discriminate fungi even down to species level (Martin and Rygiewicz, 2005; Porras-Alfaro et al., 2014). However, each ITS primer combination fails to amplify certain species, a situation similar to bacterial 16S rRNA gene based analysis (Bellemain et al., 2010). Thus the use of multiple primer combinations and/or primers with degenerated nucleotide positions is recommended to capture the entire fungal community (Ihrmark et al., 2012; Toju et al., 2012). **Table 1** summarizes commonly used ITS1 and ITS2 oligonucleotide primers. Of note, the ITS2 region was reported to perform better for fungal DNA amplification out of FFPE material (Muñoz-Cadavid et al., 2010; Flury et al., 2014). We also observed increased PCR performance using ITS2 primers and human skin FFPE samples (**Figure 2B**). However, other reports obtained similar amplification rates with ITS1 and ITS2 oligonucleotides (Mello et al., 2011; Bazzicalupo et al., 2013; Blaalid et al., 2013; Lindahl et al., 2013).

### BIOINFORMATICS CHALLENGES IN MYCOBIOTA ANALYSES

The bioinformatics analysis workflow of amplicon data can be summarized into four main steps: (i) pre-processing, (ii) OTU picking, (iii) taxonomic classification, and (iv) visualization and statistical analysis (**Figure 3**; Kuczynski et al., 2012). So far dedicated bioinformatics tools for mycobiota analyses are sparse. Measures originally developed for 16S rRNA gene data, like QIIME (Caporaso et al., 2010) and mothur (Schloss et al., 2009) are often employed to investigate ITS amplicons. However, these

tools pose several shortcomings when applied to ITS sequences, especially when standard protocols are used. In the following the main analytical steps and potential hurdles of ITS based amplicon data analyses are discussed with special emphasis on OTU clustering (OTU picking) and classification. We also highlight the effect of different OTU picking strategies on taxonomic classification of ITS data by comparative analysis of an ITS1 in silico mock community.

### PRE-PROCESSING OF AMPLICON RAW DATA

Current pre-processing recommendations include rigorous length filtering of reads, noise reduction (detection, correction, and removal of sequencing errors and artifacts), quality filtering (removal of reads with quality scores below a defined threshold; average > 25), chimera removal (detection and removal of artificially created reads, produced different targets during PCR), as well as removal of singletons/doubletons (Bokulich et al., 2013). The latter could emerge due to sequencing errors (e.g., within homopolymers) leading to OTU inflation of data, which is dependent also on the sequencing technology used (Schirmer et al., 2015). Choice of pre-processing methods and used parameters heavily influence the number of created OTUs, which could lead to underestimation of species diversity if too stringent filtering is applied (Flynn et al., 2015; Kopylova et al., 2016). However, adequate pre-processing of raw reads is mandatory

#### TABLE 2 | List of commonly used clustering algorithms.


independent of the used maker gene, leading to a reduced number of assigned OTUs and less noise in the data. Basically we refer to the suggestions of Schloss et al. (2011), but as there are no general rules for pre-processing we strongly recommend looking carefully into what is happening during filtering rather than just applying default parameters.

#### OTU PICKING—CLUSTERING INTO OPERATIONAL TAXONOMIC UNITS (OTUs)

Numerous approaches and tools are available for clustering sequences into OTUs. Current algorithms developed primarily for 16S rRNA gene amplicons are summarized in **Table 2**. In general OTU clustering and annotation could be achieved by using three different strategies (i) de novo-, (ii) closed reference-, and (iii) open reference-based clustering. Briefly, a closed reference approach calculates for each input sequence the best pairwise alignment to a pre-defined reference database collection. Sequences with the same best match are binned into the same cluster (i.e., OTU). In contrast, de novo based strategies cluster sequences within a pre-defined distance (commonly 3%). For each of these clusters a representative sequence is selected and taxonomically classified. Open-reference OTU picking is a mixture of both. Reads are first clustered using a closed reference approach and all reads which fail in this first step are subsequently clustered using a de novo strategy (Rideout et al., 2014; Westcott and Schloss, 2015). A recent comparison of the three different clustering strategies revealed the de novo approach based on a global distance matrix (implemented by default by mothur) as the optimal method for clustering 16S rRNA gene sequences into OTUs (Westcott and Schloss,

2015). Such benchmark comparisons are unfortunately missing for ITS amplicons. Importantly, the use of multiple sequence alignments (MSA) for clustering ITS sequences in a de novo approach poses a significant problem. ITS sequences show a high degree of intraspecific variation (**Figures 1D,E**), which leads to the introduction of gaps during the alignment process and subsequently to erroneous multiple sequence alignments exhibiting wrong phylogenetic resolution (**Figure 4**). In addition, there is no commonly accepted genus or species level cutoff for the formation of ITS clusters, such as 5% variation for genus- and 3% for species-level clustering applied to 16S rRNA gene data (Stackebrandt and Goebel, 1994). Often 3% variation is used and this cut-off seems to perform reasonable for fungal ITS sequences, although taxonomic resolution is clearly impaired within certain taxa. Both, ITS1 and ITS2, show a highly congruent fungal taxonomic resolution (Blaalid et al., 2013).

## TAXONOMIC CLASSIFICATION OF OTUs

If a closed reference-based approach is used, taxonomic classification is achieved already during the OTU picking step, wherein OTUs represent clusters of identical matches to the reference database. If a de novo strategy is employed a proxy sequence from each cluster is chosen and taxonomically classified either by calculating sequence similarities between the proxy sequence and a reference database or by estimating the classification confidence using a pre-trained classifier, such as the RDP classifier (Wang et al., 2007). The latter one offers training sets for ITS (Porras-Alfaro et al., 2014) as well as for LSU (Liu et al., 2012) sequences. Accurate taxonomic classification of sequences requires reference databases of high quality. The UNITE (Unified system for the DNA based fungal species linked to the classification, https://unite.ut.ee) database for ITS fragments represents a curated full-length ITS sequence repository devoid of ambiguous sequences (Nilsson et al., 2014). Several factors lead to misannotated ITS sequences in repositories, such as GenBank, EMBL, or DDJB. For instance many fungi have sexual (teleomorph) and asexual (anamorph) forms and they are often classified as different taxa assigned even to different families (Mahé et al., 2012; Underhill and Iliev, 2014). UNITE represents currently the most comprehensive taxonomic ITS classification resource, providing ready-to-use application files for mothur, QIIME, and MICCA. Although still some fungal lineages are uncovered it comprises 536,881 sequence entries (as of January 2016, UNITE version 7.0). Recently, the hand curated ISHAM-ITS reference DNA barcoding database, with 3,200 sequences covering about 415 fungal species (as of December 2015) maintained by the Society for Human and Animal Mycology (ISHAM) was incorporated into UNITE (Irinyi et al., 2015). Noteworthy is UNITE's key concept, the so-called species hypotheses (SH). A SH represents an operational taxonomic unit at approximately species level (Kõljalg et al., 2013). Each SH is represented by the most homologous high quality sequence within a respective sequence cluster linked to a unique, permanent digital object identifier (DOI), which allows for unambiguous identification even in absence of a full formal taxonomic name or when a fungal OTU remains taxonomically unassigned. Of note, the global fungome is estimated to comprise 1.5–6 million different species (Hawksworth, 1991; Blackwell, 2011; Taylor et al.,


TABLE 3 | Correct classification of the in silico ITS1 mock community with different analysis pipelines and OTU picking strategies (% in parenthesis).

2014), wherein currently 130,000 species are represented in the public sequence repositories (http://www.speciesfungorum.org/, accessed March 2016). These counts give already an idea about

phylogenetic treeing methods based on distance matrices.

the "completeness" of the current fungal reference databases (Tedersoo et al., 2014).

#### THE EFFECT OF DIFFERENT OTU PICKING STRATEGIES ON TAXONOMIC CLASSIFICATION OF ITS DATA

To demonstrate the influence of different OTU picking strategies on phylogenetic resolution of fungal communities we compared three commonly used analysis pipelines mothur, QIIME, and MICCA, employing an in silico created fungal ITS1 mock community. Therefore, 582,779 ITS1 fragments were extracted by ITSx (Bengtsson-Palme et al., 2013) from the public UNITE sequence collection (version 7, comprising 656,899 sequences). Amplicons were filtered for ambiguous lineage definitions, resulting in 345,201 sequences. These amplicons were quality filtered yielding finally 56,451 unique ITS1 fragments (accession numbers and taxonomic annotations are given in Supplementary Table S2). ITS1 fragments were subsequently clustered into OTUs by the default de novo strategies employed by mothur, QIIME, and MICCA, according to the standard protocol of each pipeline (for details see Data Sheet S1). For analyses with QIIME and MICCA, sequences were additionally binned into OTUs according to their taxonomic classification using a closed reference OTU picking strategy employing the UNITE database (version 7, 22.08.2016). The database was used for classification of representative sequences either directly for similarity-based comparisons or indirectly for training the RDP classifier. Finally, the assigned taxonomic classifications were compared to the true annotation of the ITS1 mock community. A scheme highlighting the experimental design and used parameters for comparison of pipelines is shown in **Figure 5**. **Table 3** summarizes the comparison results, which clearly indicates that choice of the OTU picking strategy severely impacts the phylogenetic resolution of the ITS mock community. All pipelines used with default parameters failed to accurately classify the mock community down to species level. All approaches classified ITS1 reads with a reasonable accuracy only to the order level (range 87.61–97.34% correct assignment), except QIIME with default settings (de novo), which behaved poor (classifying only 33.53% of sequences correctly at phylum level and 0.07% at species level). A high number of singletons emerged by using all three de novo approaches, leading to OTU inflation, and wrongly clustered OTUs. Importantly, changing the default OTU picking approach of QIIME (de novo) to a closed reference approach increased the amount of correctly classified species to 71.62% (**Table 3**). Taken together these data indicate that closed reference based strategies should be preferred if ITS amplicons are analyzed. Nevertheless, a relatively large fraction of wrongly annotated OTUs might still persist, thus manual correction of taxonomic assignments (i.e., by individual blast analysis of sequences) might still improve classification (Iliev et al., 2012).

#### VISUALIZATION AND STATISTICAL ANALYSIS OF ITS DATA

Visualization and statistical analyses of mycobiota data typically enable measures for community structure, such as alpha-diversity metrics (e.g., richness, evenness, Shannon index), as well as taxonomic turnover (i.e., changes in microbial composition between conditions or groups) called beta-diversity, which can be calculated with different distance measurements (Bray Curtis, Andernberg, UniFrac, etc.). Principle coordinates analysis (PCoA) plots based on these distance matrices enable simplified visualization of the structural resemblance of mycobiota profiles. Statistical identification of differential abundant taxa between groups could be achieved using tools such as LEfSe (Segata et al., 2011) or linear modeling approaches, such as DESeq (Paulino et al., 2006) or edgeR (Robinson et al., 2010). Measures for alphaand beta-diversity are readily provided by tools such as mothur and QIIME and operate on the created OTU tables. Caution must be taken if measures derive phylogenetic information based on diversity matrices emerging from MSAs of ITS reads, such as UniFrac (Lozupone et al., 2011). Such methods lead to erroneous results because of the bad performance of aligning ITS reads as shown above (**Figure 4**).

#### CONCLUSION

Fungal amplicon studies benefit greatly from the advancements made in the analysis of bacterial communities, nonetheless, many hurdles need still to be solved and standards are waiting to be defined. Although numerous protocols and kits are available for fungal DNA isolation out of complex specimens such as human tissue, protocols need to be adapted to the special study needs. Recommendations on how to perform ITS analyses using mothur and QIIME with non-phylogenetic diversity metrics have been recently released (e.g., https://mothur.org/wiki/ Analysis\_examples#Sanger\_16S-ITS\_rRNA\_sequence\_analysis,

#### REFERENCES


accessed February 2017, http://qiime.org/1.7.0/tutorials/fungal\_ its\_analysis.html, accessed April 2016). Based on our experience, pre-processing, and quality filtering of ITS sequencing data, as well as chimera filtering could be done with standard 16S rRNA gene based procedures. We use the default workflow of mothur for ITS data pre-processing, assembling of paired reads, length-, quality-, and chimera filtering, as well as noise reduction as described in the MiSeq 16S SOP of Kuczynski et al. (2012, accessed May 2016). Since mothur employs pairwise distance matrices, which require the creation of multiple sequence alignments, we recommend switching to tools such as QIIME or MIICA for further analyses, which allow for closed reference-based approaches. Subsequently QIIME can be used for visualization of mycobiota data. The crucial step within QIIME is to suppress tree generation within the OTU picking step and to use closed reference OTU picking instead of the default de novo strategy. The pre-formatted version of the UNITE ITS reference database which is provided directly by UNITE works perfectly with one of the reference-based OTU picking scripts of QIIME and MICCA. Alternatively sequences can be also classified and binned based on the information gained by the RDP classifier trained for ITS fragments or simply by an individual blast approach. A final summary of the recommended analysis steps for ITS based mycobiota analysis is given in **Figure 6**.

### AUTHOR CONTRIBUTIONS

Conceptualization: BH, RN, GT, and GG. Data analysis: BH and NM. Manuscript draft: BH and GG. Final manuscript and approval: All authors.

### FUNDING

This work was supported by BioTechMed-Graz and the Austrian Science Fund (FWF W1241-B18).

#### ACKNOWLEDGMENTS

Jürgen C. Becker, Karl Kashofer, and Andrea Thüringer are acknowledged for their input regarding mycobiota analysis.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2017.00180/full#supplementary-material

Belkaid, Y., and Naik, S. (2013). Compartmentalized and systemic control of tissue immunity by commensals. Nat. Immunol. 14, 646–653. doi: 10.1038/ni.2604


ITS2 from ribosomal ITS sequences of fungi and other eukaryotes for analysis of environmental sequencing data. Methods Ecol. Evol. 4, 914–919. doi: 10.1111/2041-210x.12073


sequencing of amplified markers - a user's guide. New Phytol. 199, 288–299. doi: 10.1111/nph.12243


**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 © 2017 Halwachs, Madhusudhan, Krause, Nilsson, Moissl-Eichinger, Högenauer, Thallinger and Gorkiewicz. 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) or licensor 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.

# 16S Based Microbiome Analysis from Healthy Subjects' Skin Swabs Stored for Different Storage Periods Reveal Phylum to Genus Level Changes

Ingeborg Klymiuk<sup>1</sup> \*, Isabella Bambach<sup>2</sup> , Vijaykumar Patra<sup>2</sup> , Slave Trajanoski<sup>1</sup> and Peter Wolf<sup>2</sup>

<sup>1</sup> Center for Medical Research, Medical University of Graz, Graz, Austria, <sup>2</sup> Research Unit for Photodermatology, Department of Dermatology, Medical University of Graz, Graz, Austria

#### Edited by:

Martin Grube, University of Graz, Austria

#### Reviewed by:

Tomislav Cernava, Austrian Centre of Industrial Biotechnology, Austria Stefanie Maier, Max-Planck-Institut für Chemie, Germany

\*Correspondence: Ingeborg Klymiuk ingeborg.klymiuk@medunigraz.at

#### Specialty section:

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

Received: 03 May 2016 Accepted: 30 November 2016 Published: 20 December 2016

#### Citation:

Klymiuk I, Bambach I, Patra V, Trajanoski S and Wolf P (2016) 16S Based Microbiome Analysis from Healthy Subjects' Skin Swabs Stored for Different Storage Periods Reveal Phylum to Genus Level Changes. Front. Microbiol. 7:2012. doi: 10.3389/fmicb.2016.02012 Microbiome research and improvements in high throughput sequencing technologies revolutionize our current scientific viewpoint. The human associated microbiome is a prominent focus of clinical research. Large cohort studies are often required to investigate the human microbiome composition and its changes in a multitude of human diseases. Reproducible analyses of large cohort samples require standardized protocols in study design, sampling, storage, processing, and data analysis. In particular, the effect of sample storage on actual results is critical for reproducibility. So far, the effect of storage conditions on the results of microbial analysis has been examined for only a few human biological materials (e.g., stool samples). There is a lack of data and information on appropriate storage conditions on other human derived samples, such as skin. Here, we analyzed skin swab samples collected from three different body locations (forearm, V of the chest and back) of eight healthy volunteers. The skin swabs were soaked in sterile buffer and total DNA was isolated after freezing at −80◦C for 24 h, 90 or 365 days. Hypervariable regions V1-2 were amplified from total DNA and libraries were sequenced on an Illumina MiSeq desktop sequencer in paired end mode. Data were analyzed using Qiime 1.9.1. Summarizing all body locations per time point, we found no significant differences in alpha diversity and multivariate community analysis among the three time points. Considering body locations separately significant differences in the richness of forearm samples were found between d0 vs. d90 and d90 vs. d365. Significant differences in the relative abundance of major skin genera (Propionibacterium, Streptococcus, Bacteroides, Corynebacterium, and Staphylococcus) were detected in our samples in Bacteroides only among all time points in forearm samples and between d0 vs. d90 and d90 vs. d365 in V of the chest and back samples. Accordingly, significant differences were detected in the ratios of the main phyla Actinobacteria, Firmicutes, and Bacteroidetes: Actinobacteria vs. Bacteroidetes at d0 vs. d90 (p-value = 0.0234), at d0 vs. d365 (p-value = 0.0234) and d90 vs. d365 (p-value 0.0234) in forearm samples and at d90 vs. d365 in V of the chest =

(p-value = 0.0234) and back samples (p-value = 0.0234). The ratios of Firmicutes vs. Bacteroidetes showed no significant changes in any of the body locations as well as the ratios of Actinobacteria vs. Firmicutes at any time point. Studies with larger sample sizes are required to verify our results and determine long term storage effects with regard to specific biological questions.

Keywords: storage, skin-swabs, microbiome, standardization, large cohort studies, stability

### INTRODUCTION

Published studies on the human derived microbiome, as the entity of all microbial genomes in and on the human body, have increased tremendously in the last decade from 257 publications in 2005 to 5849 in 2015 [retrieved from PubMed by the search term 'human microbiome' and updated from Toh and Allen-Vercoe (2015)]. New cost efficient and high throughput next generation sequencing technologies have spurred this scientific development to add a new field with immense significance to medical research (Metzker, 2010). We know that the human microbiome and alterations in the bacterial composition are associated with a wide range of human diseases from neurological [multiple sclerosis (Miyake et al., 2015)], intestinal [Crohn's disease (Raes, 2016)] and skin (Yu et al., 2015) disorders to infertility (Franasiak and Scott, 2015). Microbiota (the entity of all microorganisms living in and on the human body) may influence our physiology directly by stimulating our immune system, occupying and affecting habitats on the human body defending us against pathogens or influence us through their metabolites (Integrative Hmp (iHMP) Research Network Consortium, 2014). Although the significance of microbiota for human health and physiology is recognized, many studies lack statistical significance due to inter-individual differences of microbiomes and insufficient significant sample sizes for statistical power and calculation of the biological traceability. A prerequisite for processing and analysis of large cohort samples is on the one hand the coordinated collection and on the other hand the reproducible laboratory processing and bioinformatics analysis of samples. For comparison and reproduction of these studies on the human microbiome, standardization of the workflow steps is paramount and already requested from scientific community (Sinha et al., 2015). Future needs to implement microbiome analysis in daily clinical procedures are already drafted (Fricke and Rasko, 2014; Klymiuk et al., 2014). This started process of standardization to increase reproducibility, efficiency and quality of data output on microbiome research projects will not only affect the processing in the wet lab during analyses of sample material. The storage of hundreds to thousands of samples needed for a large cohort study is not only a logistical challenge but requires strict standardization criteria as used for collection, processing, and data analysis. 'Microbiome' samples imply a variety of sample materials ranging from stool, swabs, and different body fluids to tissues and biopsies. This variety of samples requires evaluation of their storage conditions and storage time to prevent contamination of biological results from

artifacts caused by the experimental procedure (Meisel et al., 2016) and to assure that storage will not alter or destruct valuable information of invaluable samples. Since DNA can degrade through oxidation, hydrolysis, or enzymatic degradation (Gorzelak et al., 2015), we must consider that sampling methods and storage conditions can be a main parameter for microbiome studies and on the data output. These processes must be optimized to reduce DNA degradation and ultimately minimize the variability observed in microbiome analyses. Evaluating the effects of storage temperature, condition (e.g., buffer) and time on the microbiome composition is an important prerequisite for long term storage of microbiome samples in bio banks to utilize this information for personalized medicine approaches (Kinkorova, 2015). The effect of sampling and storage of human derived microbiome samples has already been studied in stool (Roesch et al., 2009; Lauber et al., 2010; Bahl et al., 2012; Carroll et al., 2012; Choo et al., 2015; Gorzelak et al., 2015; Tedjo et al., 2015), vaginal (Bai et al., 2012), sputum (Zhao et al., 2011), and skin (Lauber et al., 2010) specimens. Some previous studies conclude that there were no significant differences in the bacterial community or the richness due to sample storage, although various experimental procedures were used (Zhao et al., 2011; Bai et al., 2012; Carroll et al., 2012; Choo et al., 2015; Tedjo et al., 2015), and most studies conclude that storage at room temperature for several hours or at 4, −20, or −80◦C for durations from hours to months did not alter the main biological information (overall community composition and relative abundance of major taxa) on the main habitat specific phyla (Lauber et al., 2010; Carroll et al., 2012; Choo et al., 2015; Tedjo et al., 2015). Lauber et al. (2010) found no significant changes in the phylogenetic diversity of skin samples even after sample storage for up to 14 days at various temperatures, ranging from 20 to −80◦C, before DNA isolation. In most setups, statistical separation of samples occurred by test subjects rather than storage conditions. Nevertheless, Bahl et al. (2012) investigated the effect of freezing fecal samples prior to DNA isolation and detected changes in the ratios of some predominant and prevalent phyla. They found that storage of samples at −20◦C for about 2 months did not alter DNA yield, but did significantly alter the ratio between Firmicutes and Bacteroides. Roesch et al. (2009) also found significant differences in the community composition in individual samples following storage at room temperature for 12–72 h before freezing samples at −80◦C. Choo et al. (2015) described significant changes due to preliminary treatment and storage at various conditions before final storage at −80◦C, though they provide no data on possible long term storage effects. Cardona et al. (2012) found

a negative effect on DNA integrity during storage of samples at room temperature or after freezing and defrosting samples before final storage at −80◦C. Other studies demonstrated that microbial diversity remains relatively stable among various storage conditions, whereas the relative abundance of main taxa can change dramatically if samples are stored at room temperature for 2 weeks (Cardona et al., 2012). However, all these studies lack information on long term storage effects on microbiome analyses and specifically on human derived skin samples.

In this study, we investigated the bacterial microbial pattern derived from skin swab samples stored for various time periods to provide recommendations for the standardization of storage in long term projects, as already performed for other sample materials (Peakman and Elliott, 2008). Most previous studies demonstrated an influence of storage on the microbial pattern as a function of different freezing conditions prior to long term sample storage due to home self-sampling in most stool based analyses. Definitive conclusions on the influence of long term storage on the results of microbial community in skin samples are still missing. Here, we describe the sampling and analysis of skin swab samples from eight healthy volunteers at three body locations with DNA isolation performed after overnight freezing (d0), 90 days (d90), and 365 days (d365) of storage at −80◦C to analyze long term storage effects on the results of bacterial microbial composition. We also discuss other possible sources for alterations that may change or bias the biological results like technical artifacts, such as variability in different lots of nucleic acid isolations kits that should be considered in large scale cohort studies. Our study offers a trendsetting approach for the handling and long term storage of skin microbiome samples and provides valuable information to plan large scale analyses.

## MATERIALS AND METHODS

#### Study Set-up

Skin swab samples were collected from eight healthy volunteers (seven women and one man) between 25 and 60 years in age. None of the volunteers had received antibiotic treatment for 3 months prior to sampling. All sampling procedures were employed for a pilot study of an explorative microbiome project (approved by the Ethics Committee of the Medical University Graz; protocol no. 27-263 ex 14/15). All participants provided informed consent and the study was conducted in accordance with the Declaration of Helsinki.

### Sampling Procedure

Three equivalent samples were taken from each volunteer at three different body locations (forearm exterior left side, V of the chest, and back). Subjects were instructed not to wash or to use any cosmetics on the day of sample collection using skin swabs. Three adjacent quads of 5 cm side length were sampled with a BD Culture SwabsTM EZ Collection and Transport system soaked with sterilized 0.15 M NaCl and 0.1% Tween-20 (Gao et al., 2008). The swabs were cut under sterile conditions into a sterile 1.5 ml reaction tube and were frozen at −80◦C immediately after sampling. DNA was extracted after storage at −80◦C the day after sampling or after 90 or 365 days of storage before extraction. Unused swabs soaked in the sterile buffer were cut to the collection tubes and used as negative controls for each time point.

## Total DNA Isolation, 16S Library Preparation and Sequencing

Total DNA was isolated from frozen swab samples with a combination of mechanical and enzymatic lysis with the MagnaPure LC DNA Isolation Kit III (Bacteria, Fungi; Roche, Mannheim, Germany) according to manufacturer's instructions. Three hundred and eighty microliter of bacterial lysis buffer (Roche, Mannheim, Germany) were added directly to the frozen sample and vortexed vigorously for 60 s to ensure bacterial transfer from swabs into solution. Unused swabs and unused buffer tubes without swabs served as negative controls for sampling and DNA isolation. The swabs were removed and the solutions were transferred to Magna Lyser green bead tubes (Roche, Mannheim, Germany), and bead beated for mechanical lysis at 6500 rpm for 30 s two times in a MagNA Lyser Instrument (Roche, Mannheim, Germany). Samples were incubated with 20 µl lysozyme at 37◦C for 30 min followed by 30 µl Proteinase K for 1.5 h at 65◦C. Enzymes were inactivated at 95◦C for 10 min. The remaining steps were performed according to instructions from the Magna Pure DNA isolation kit III (Bacteria, Fungi). Two hundred microliter of each sample were used for DNA purification in a MagnaPure instrument. Total DNA was eluted in 100 µl and stored at −20◦C until PCR amplification. For target specific PCR amplification of hypervariable regions the primers 27f (AGAGTTTGATCCTGGCTCAG) and 357r (CTGCTGCCTYCCGTA) were used according to Baker et al. (2003) and synthesized at Eurofins (MWG, Ebersberg, Germany). Five microliter of total DNA extract were used for a 25 µl PCR reaction in triplicates containing 1 x Fast Start High Fidelity Buffer (Roche, Mannheim, Germany), 1.25 U High Fidelity Enzyme (Roche, Mannheim, Germany), 200 µM dNTPs (Roche, Mannheim, Germany), 0.4 µM barcoded primers and PCR-grade water (Roche, Mannheim, Germany). Thermal Cycling was of initial denaturation at 95◦C for 3 min, followed by 30 cycles of denaturation at 95◦C for 45 s, annealing at 55◦C for 45 s and extension at 72◦C for 1 min, one cycle of final extension at 72◦C for 7 min and a final cooling step to 4◦C. Triplicates were pooled, amplification was verified using a 1% agarose gel and 15 µl of pooled PCR product were normalized according to manufacturer's instructions on a SequalPrep Normalization Plate (Life Technologies, Vienna, Austria). Fifteen microliter of the normalized PCR product were used as the template for indexing PCR in a 50 µl single reaction composed as described for targeted PCR to introduce barcode sequences for each sample according to Kozich et al. (2013). Cycling conditions were the same as for the targeted PCR with only eight cycles for amplification. After indexing, 5 µl of each sample were pooled, 50 µl of the unpurified library were loaded to a 1% agarose gel (Sigma–Aldrich, St. Louis, MO, USA) and then purified from the gel with a Qiaquick Gel Extraction Kit (Qiagen, Hilden, Germany) according to

manufacturer's instructions. The pool was quantified using PicoGreen dsDNA reagent (Life Technologies, Vienna, Austria) according to manufacturer's instructions and visualized for size validation on an Agilent 2100 Bioanalyzer (Agilent Technologies, Waldbronn, Germany) using a high sensitivity DNA assay according to manufacturer's instructions. The sequencing library pool was diluted to 4 nM until run on a MiSeqII desktop sequencer (Illumina, Eindhoven, Netherlands). Version 3 600 cycles chemistry (Illumina, Eindhoven, Netherlands) was used according to manufacturer's instructions to run the 6 pM library with 20% PhiX (Illumina, Eindhoven, Netherlands) and FASTQ files were used for data analysis.

#### Data Analysis

In the first data analysis step MiSeq paired-end raw sequence forward and reverse reads were merged using ea-utils v1.1.2 with standard settings, followed by a split library step from the Quantitative Insights Into Microbial Ecology (QIIME, v1.9.1) software. During this step a quality control step removed sequence reads shorter than 200 nucleotides, reads that contained ambiguous bases or reads with an average quality score of <30. Chimera were removed with USEARCH v6.1 method in QIIME against 97% clustered GreenGenes reference 16S rRNA database (v13.8). In the second step, operational taxonomic units (OTUs) picking utilized QIIME open reference pipeline to perform clustering steps at 97% sequence similarity, the taxonomy assignment with UCLUST algorithm, alignment of reference sequences with pyNAST and generation of a phylogenetic tree with FastTree. The OTU table was reduced by removing all OTUs present in only one sample with <10 reads. Prior to rarefaction and subsequent data analysis, the median absolute number of reads was evenly distributed over the storage groups between 138,081 at d0, 142,152 at d90, and 154,087 at d365 (Supplementary Figure 1A). To even all samples, we performed a rarefaction to 65,000 sequence reads per sample for further analysis. Downstream data analysis for alpha and beta diversity as well as statistical calculations were performed in R statistical programming language (v3.2.3) using vegan (2.4-0), GUniFrac (1.0), phytools (0.5–38), and phangorn (2.0.4) packages. For alpha diversity analyses, we calculated and compared richness and Shannon diversity index. The effects of different storage conditions were tested with the non-parametric Friedman test, where the normality assumption was violated, followed by a pairwise Wilcoxon signed-rank test, or student's t-test. In case of multiple testing p-values were corrected with Benjamini and Hochberg method. Multivariate data analysis of microbiota community dataset was based on two distance measurements Bray–Curtis and weighted UniFrac. UniFrac distances were calculated on a phylogenetic tree and provided a phylogenetic estimate of community similarity, whereas Bray Curtis dissimilarity provided an abundance-weighted measure. Using the generated distance matrices, we visualized the data with the average-linkage agglomerative hierarchical clustering method and Kruskal's non-metric multidimensional scaling (NMDS). Statistical differences in the overall community composition of samples were assessed using the "permutational manova" test (Adonis in vegan).

## RESULTS

### Skin Swab Samples Overall Analysis

Skin swab samples from eight healthy volunteers were analyzed from three different body locations and after three sample storage time periods at −80◦C. FastQ raw data can be accessed through the SRA accession number SRPO74170 at NCBI Trace Archive. From these 72 specimens and six controls (one for each sample collection date and one for each MagnaPure isolation batch), we analyzed a dataset of 18,136,666 passed filter paired end raw sequence reads (for details in reads distribution see Supplementary Figures 1A– C). Depending on the body location, a total number of 5,646 OTUs were detected in forearm samples, 5,058 OTUs in back and 5,901 OTUs in V of the chest specimen. Under reference conditions with DNA isolation after overnight freezing analyzing all samples together, specimen were dominated by the phyla Actinobacteria (M = 42.5%, SD = 19.9), Firmicutes (M = 30.9%, SD = 17.0), and Bacteroidetes (M = 12.8%, SD = 6.9). The most abundant genera in forearm samples were Propionibacterium (M = 24.9%, SD = 10.9), Bacteroides (M = 12.5%, SD = 7.2), and Corynebacterium (M = 11.2%, SD = 14.0). The V of the chest samples contained Propionibacterium (M = 31.2%, SD = 18.8), Streptococcus (M = 9.9%, SD = 7.4), and Bacteroides (M = 7.3%, SD = 3.4). In back samples Propionibacterium (M = 41.3%, SD = 26.5), Bacteroides (M = 9.2%, SD = 7.2), and Streptococcus (M = 8.6%, SD = 8.0) were most abundant (**Figure 1**). The interpersonal differences and the changes among sampled body locations, we observed are well known from large cohort human studies (Turnbaugh et al., 2007; Oh et al., 2013). Non-metric multidimensional scaling (NMDS) of Bray– Curtis distances revealed a clustering of samples based on volunteer (Rˆ2 49%, p-value = 0.001, **Figure 2D**) followed by body location (Rˆ2 4.7%, p-value = 0.023, **Figure 2D**). No clustering, however, was found for sample storage duration at −80◦C (Rˆ2 4.0%, p-value = 0.07) calculated for each individual location (**Figures 2A–C**) and for all body locations together (**Figure 2D**). Tree based hierarchical agglomerative clustering with average linkage dendrogram analysis on Bray– Curtis distances accordingly revealed no significant clustering of samples based on the sample storage duration at −80◦C but our results show a clustering by volunteer followed by the clustering per body location (Supplementary Figure 2).

### Sample Storage Effects on the Results of Microbial Diversity and Richness

Under reference conditions, microbial richness of the eight volunteers per body location was between 311 and 1763 OTUs (Supplementary Table 1) and the Shannon diversity index was between 0.98 and 4.62 (**Figures 3A,B** and Supplementary Table 1). No statistically significant differences were observed in richness and Shannon diversity between the sample groups at different storage durations at −80◦C analyzed over all body locations without grouping according to the body location (**Figures 3A,B**). A non-significant trend toward an increase in richness but not in Shannon diversity was observed in the data

derived from samples stored at −80◦C for 90 days from all body locations. Analyzing different freezing durations per body location separated, the only significant differences in richness were observed between d0/d90 and d90/d365 in forearm samples but not between d0/d365 (**Figure 3C**). No significant differences in Shannon diversity were observed among different freezing periods per body location (Supplementary Table 1).

### Sample Storage Effects on the Results of the Ratios of Most Abundant Phyla

Some former studies discovered differences in the ratios among the main phyla Actinobacteria, Firmicutes, Bacteroidetes, Proteobacteria, or Cyanobacteria as a function of storage or biological alterations (Stadlbauer et al., 2015; Compare et al., 2016). Accordingly, we analyzed all ratios of the dominant phyla, Actinobacteria, Firmicutes, and Bacteroidetes, found in all samples. Calculating the ratios of Actinobacteria vs. Bacteroidetes for each body location, we found significant differences (pvalue < 0.05) in forearm samples across all freezing periods (d0/d90 = 0.0234, d90/d365 = 0.0234, d0/d365 = 0.0234), in V of the chest samples between d90 and d365 (p-value = 0.0234) and in back samples comparing d90 and d365 (p-value = 0.0234) (**Figure 4A**). For the ratios of Actinobacteria vs. Firmicutes and Firmicutes vs. Bacteroidetes, none of the body locations revealed significant differences among any of the three storage periods (**Figures 4B,C**).

### Sample Storage Effects on the Results of Relative Abundance of Main Phyla

Summarizing over all analyzed subjects, body locations and storage durations, the most abundant phyla found were Actinobacteria, Firmicutes, and Bacteroidetes. Significant differences (p-value < 0.05) in the relative abundance of the phylum Actinobacteria was found between DNA isolation after 90 days (d90) and after 365 days (d365; p-value = 0.0468) in back samples only (**Figure 5A**). Storage of samples at −80◦C from the forearm or V of the chest for 90 or 365 days did not result in significant differences in the relative abundance of this phylum compared to the reference method (DNA isolation after overnight freezing, d0) (**Figure 5A**). The relative abundance of Bacteroidetes differed significantly between d0 and d90 in samples from all body locations and between d90 and d365, respectively (**Figure 5B**). Only forearm samples revealed significant differences in the relative abundance of Bacteroidetes between d0 and d365 of sample storage (**Figure 5B**). No significant differences were found in the relative abundance of

Firmicutes in any sampled body location across any storage times (**Figure 5C**).

Further, we analyzed location specific differences across storage periods in a class, order, family, and genus level analysis with Lda Effective Size (LEfSe) (Supplementary Table 2). Considering only taxa with a relative abundance of at least 1% in at least 50% of all samples analyzed, we found the genus Bacteroides significantly differed in relative abundance in all body locations. LEfSe analysis revealed significant changes in all hierarchical levels the genus Bacteroides belong to Supplementary Table 2. Further analysis on Bacteroides revealed the genus significantly altered between d0 and d90 and between d90 and d365 among all body locations (**Figures 6A–C**). However, between d0 and d365 significant differences for Bacteroides were only detected in forearm samples but not in V of the chest and back samples (**Figures 6A–C**).

### DISCUSSION

We analyzed the effects of long term storage at −80◦C from skin derived dermal microbiome swab samples on the results of microbial composition. Previous studies investigated the effect of sample storage at different temperatures before final storage at −80◦C for weeks primarily in human stool samples, which found minimal to no effects on the analyzed microbial patterns. Our study on eight healthy volunteers investigated for the first time the effect of long term storage on skin derived microbiome samples for up to 1 year at recommended conditions on the data outcome. The three most dominant genera found in our forearm (Propionibacterium, Bacteroides, Corynebacterium), V of the Chest (Propionibacterium, Streptococcus, Bacteroides) and back samples (Propionibacterium, Streptococcus, Bacteroides) under reference conditions at d0 correspond to genera found in former studies on the human skin microbiome (Grice et al., 2008). The estimated microbial richness and the Shannon diversity index were within expected values compared to other skin studies (Grice et al., 2008; Zeeuwen et al., 2012; Meisel et al., 2016). Analyzing our storage groups over all body locations together, no significant differences were found in the richness and Shannon diversity index among sample groups of different storage periods or in the number of observed OTUs. There was an insignificant tendency for increased richness, but not Shannon diversity, at d90 (**Figures 3A,B**). These results correspond to former studies on different storage conditions of stool samples that do not find statistically significant differences in microbial richness due to different storage preconditions (Bai et al., 2012; Flores et al., 2015; Tedjo et al., 2015). The high stability of DNA even after microbial organism death and the robustness of the 16S rRNA PCR based microbiome analysis method may account for this persistent finding. Nevertheless, analyzing the storage groups separated per body locations a significant change in the richness of forearm

statistically significant. Only the change in forearm sample richness between d0 and d90 (p-value = 0.0234) and d90 and d365 (p-value = 0.0234) is statistically significant (C), but no difference is observed between d0 and d365 using Wilcoxon test. Significant differences (p-value below 0.05) are marked with a line between the affected sample groups and an asterisk.

samples was detected between d0 vs. d90 and d90 vs. d365 but in no other body location. In our study the main biological parameters remained valid between storage groups and samples clustered according to volunteer and body location rather than storage time (**Figures 2A–D**). However, some of the dominant phyla and genera detected in the different skin locations of the eight volunteers showed significant changes over storage periods. The phylum Actinobacteria changed significantly between d90 vs.

FIGURE 4 | Boxplot diagram of the ratios of the three main phyla Actinobacteria vs. Bacteroidetes (A), Firmicutes vs. Bacteroidetes (B), and Actinobacteria vs. Firmicutes (C) in forearm, V of the chest and back samples: significant differences in the phyla ratios between storage periods at −80◦C (p-value below 0.05) are marked with a line between the affected sample groups and an asterisk.

d365 in back samples only and the phylum Bacteroidetes between d0 vs. d90 and d90 vs. d365 in all body locations (**Figure 5**). Between d0 and d365 the phylum changed only significantly in forearm samples. Additionally the genus Bacteroides showed significant changes in all body locations (**Figure 6**) between d0 vs. d90 and d90 vs. d365. Changes between d0 and d365 are only statistically significant in forearm samples.

Analyzing the ratios of the most abundant phyla detected in the skin swab samples (Actinobacteria, Bacteroidetes, and Firmicutes) revealed statistically significant differences in

Actinobacteria vs. Bacteroidetes across storage durations. In all body locations, differences were found between d90 and d365 and also in forearm samples between d0 and d90 and d0 and d365, respectively (**Figure 4A**). These results may indicate biological changes but also possible technical artifacts caused by different DNA isolation kit batches used at d0, d90, and d365. As such, differences in the relative abundance of Actinobacteria between d90 and d365 were only detected in back samples and of Bacteroidetes were observed between d0 and d90 and between d90 and d365 in all body locations but the change between d0 and d365 only in forearm samples (**Figures 5A,B**). No differences in the relative abundance of Firmicutes were observed among the three time points at all. The reasons for changes in the relative abundance of phyla may occur from the different cell wall characteristics of gram-negative and grampositive bacteria. Firmicutes and Actinobacteria are gram-positive organisms probably less affected by DNA degradation due to destruction of the microbial cell wall during storage. In contrast, Bacteroidetes belong to the group of gram-negative bacteria that may be more affected by cell death caused DNA degradation by, e.g., oxygen and enzymes. Using the LefSeq and Friedmann test analyses, the only genus differentially abundant across the storage groups was the gram-negative genus Bacteroides (**Figures 6A–C**). However, our data does not support this hypothesis of alterations due to DNA degradation as the relative abundance of Bacteroides increased by d90 and decreased to reference baseline levels at day 365 (except in forearm samples). This pattern may indicate a technical issue that should be considered in the design of large scale studies although the observed changes between d0 and d365 in forearm samples cannot be fully explained by technical issues. The influence of DNA storage at −20◦C until PCR amplification should also be kept in mind. However, one would expect constant DNA degradation and impaired PCR amplification rather than a change in the relative abundance or ratios of dominating phyla and genera.

To answer the fundamental and pressing questions in microbiome research and their relation to human health, large cohort studies will provide further reliable scientific and statistically valid results. Sample size, storage conditions and quality, as well as data analysis, must be appropriate and standardized. This is particularly important for large cohort studies in which sampling and storing microbiome specimens for several years is necessary. In these studies the effect of storage time and inherent variability in DNA extraction batches, as well as library preparation and sequencing, needs to be reconsidered to ensure reproducibility and standardization across studies and in clinical treatment applications. Standardization of sample storage procedures with instant flash-freezing and continuous evaluation through calibration samples throughout the total project duration should be mandated.

To overcome the difficulty inherent in microbiome specimen storage durations, immediate nucleic acid extraction may seem to be advantageous. Our results may indicate that the risk to create a technical bias through different lots of DNA isolation kits may be higher than the bias found in different sample storage times at recommended conditions (−80◦C or lower) especially for large cohort skin microbiome studies. In addition, it is not recommended to change the nucleic acid isolation technique to keep samples matchable once a method has been established. While the extracted DNA can be stored at −20◦C or even at higher temperatures for long periods of time, it can be exposed to additional sources of variability through degradation induced by enzymatic processes, oxygen degradation or repetitive freeze-thaw cycles, which can bias the results.

Collection of well accepted specimens, such as skin swabs, is critical for future research endeavors. Our study provides a critical reference to elucidate the storage periods for skin microbiome studies. The main biological endpoints and parameters (clustering of samples according to volunteers and body locations but not to storage time points) used for analysis after different sample storage times remained valid throughout our study. Nevertheless, we observed changes in the ratios of the most abundant phyla (Actinobacteria vs. Bacteroidetes), in the relative abundances of the most abundant phyla (minor changes in Actinobacteria, distinct changes in Bacteroidetes), in the relative abundance of the genus Bacteroides as well as in the richness of forearm samples among sample groups of different storage periods. Some of these results should be interpreted with caution as, we cannot rule out a technical issue by the use of another batch of DNA isolation kit at d90. Another limitation of our study is the limited sample size and gender imbalance (with female predominance of participants due to the availability of volunteers). Studies with larger sample sizes are needed to confirm our results with regard on the potential influence of long term storage effects of specimens on specific biological endpoints.

### AUTHOR CONTRIBUTIONS

IK: Study design, analyzed the data, and wrote the manuscript. IB: Study design and performed experiments. ST: Analyzed the data, prepared the figures, and revised and wrote the manuscript. VP: Revised and wrote the manuscript. PW: Study design and revised and wrote the manuscript.

### ACKNOWLEDGMENT

We thank the volunteers for conducting to the study. The work of PW and VP was supported by the Austrian Science Fund FWF (W1241) and the Medical University of Graz through the PhD program DK-MOLIN.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.02012/full#supplementary-material

### REFERENCES

fmicb-07-02012 December 16, 2016 Time: 12:43 # 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 © 2016 Klymiuk, Bambach, Patra, Trajanoski and Wolf. 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) or licensor 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.

# Sampling Modification Effects in the Subgingival Microbiome Profile of Healthy Children

Elisabeth Santigli <sup>1</sup> , Slave Trajanoski <sup>2</sup> , Katharina Eberhard<sup>2</sup> and Barbara Klug<sup>1</sup> \*

<sup>1</sup> Division of Oral Surgery and Orthodontics, Department of Dental Medicine and Oral Health, Medical University of Graz, Graz, Austria, <sup>2</sup> Center for Medical Research, Medical University of Graz, Graz, Austria

Background: Oral microbiota are considered major players in the development of periodontal diseases. Thorough knowledge of intact subgingival microbiomes is required to elucidate microbial shifts from health to disease.

Aims: This comparative study investigated the subgingival microbiome of healthy children, possible inter- and intra-individual effects of modified sampling, and basic comparability of subgingival microprints.

#### Edited by:

Gabriele Berg, Graz University of Technology, Austria

#### Reviewed by:

Yvonne Vallès, New York University Abu Dhabi, UAE Renee Maxine Petri, Veterinärmedizinische Universität, Austria

> \*Correspondence: Barbara Klug barbara.klug@medunigraz.at

#### Specialty section:

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

Received: 30 April 2016 Accepted: 19 December 2016 Published: 18 January 2017

#### Citation:

Santigli E, Trajanoski S, Eberhard K and Klug B (2017) Sampling Modification Effects in the Subgingival Microbiome Profile of Healthy Children. Front. Microbiol. 7:2142. doi: 10.3389/fmicb.2016.02142 Methods: In five 10-year-old children, biofilm was collected from the upper first premolars and first molars using sterilized, UV-treated paper-points inserted into the subgingival sulcus at eight sites. After supragingival cleaning using an electric toothbrush and water, sampling was performed, firstly, excluding (Mode A) and, secondly, including (Mode B) cleansing with sterile cotton pellets. DNA was extracted from the pooled samples, and primers targeting 16S rRNA hypervariable regions V5 and V6 were used for 454-pyrosequencing. Wilcoxon signed rank test and t-test were applied to compare sampling modes. Principal coordinate analysis (PCoA) and average agglomerative hierarchical clustering were calculated with unweighted UniFrac distance matrices. Sample grouping was tested with permutational MANOVA (Adonis).

Results: Data filtering and quality control yielded 67,218 sequences with an average sequence length of 243bp (SD 6.52; range 231–255). Actinobacteria (2.8–24.6%), Bacteroidetes (9.2–25.1%), Proteobacteria (4.9–50.6%), Firmicutes (16.5–57.4%), and Fusobacteria (2.2–17.1%) were the five major phyla found in all samples. Differences in microbial abundances between sampling modes were not evident. High sampling numbers are needed to achieve significance for rare bacterial phyla. Samples taken from one individual using different sampling modes were more similar to each other than to other individuals' samples. PCoA and hierarchical clustering showed a grouping of the paired samples. Permutational MANOVA did not reveal sample grouping by sampling modes (p = 0.914 by R <sup>2</sup> = 0.09).

Conclusion: A slight modification of sampling mode has minor effects corresponding to a natural variability in the microbiome profiles of healthy children. The inter-individual variability in subgingival microprints is greater than intra-individual differences. Statistical analyses of microbial populations should consider this baseline variability and move beyond mere quantification with input from visual analytics. Comparative results are difficult to summarize as methods for studying huge datasets are still evolving. Advanced approaches are needed for sample size calculations in clinical settings.

Keywords: oral microbiome, subgingival biofilm, healthy children, next generation sequencing (NGS), 454-pyrosequencing, paper point, subgingival sampling

#### INTRODUCTION

Oral bacterial biofilm research is an emerging field. During the last decades, the profiling of oral microbial communities has evolved from bacterial culture experiments to biofilm characterization by detailed classification using cultureindependent methods (Jenkinson, 2011; Diaz, 2012; Simón-Soro et al., 2013). High throughput next generation sequencing (NGS) like 454-pyrosequencing and metagenome analysis have replaced fingerprinting methods (Ahn et al., 2011; Griffen et al., 2011; Alcaraz et al., 2012; Li et al., 2012, 2013; Siqueira et al., 2012; Abusleme et al., 2013; Trajanoski et al., 2013; Chen et al., 2015; Park et al., 2015). Instead of identifying single bacteria, operational taxonomic units (OTUs) based on sequence similarities (of mostly 97%) are assigned to identify groups of bacteria. This has led to a new research avenue leaving single germ detection behind and looking ahead to a fingerprinting of the whole bacterial community. With this unique microbial fingerprint, even forensic analyses could be made possible, as the composition of bacterial biofilm differs from person to person, whether sampled from the oral cavity (Aas et al., 2005) or the skin (Fierer et al., 2010). The oral microbiome displays a large variability; various microhabitats like gingival tissue, tongue, saliva, supra- or subgingival locations facilitate biofilm formation and growth already at early ages (Papaioannou et al., 2009). Keijser et al. (2008) showed that the vast majority (namely 99.6%) of sequences in saliva and subgingival plaque samples of adults belong to one of the seven major phyla: Actinobacteria, Bacteroides, Firmicutes, Fusobacteria, Proteobacteria, Spirochetes, or candidate division TM7. Lazarevic et al. (2010) could prove these findings in salivary samples. However, not only bacterial phyla can be tagged; these new methods can show bacterial diversity on all taxonomic levels from the phylum through the genus level. This identification of bacteria takes place over nine hypervariable regions (V1 through V9) of the 16S rRNA gene used to distinguish thousands of species sequences of one sample from another (Chakravorty et al., 2007; Huse et al., 2008). The huge amount of sequence data gained with these methods puts common knowledge of pathogens into a new perspective. Many bacteria previously known to be pathogens were now also found in healthy subjects. Certain bacterial species like Streptococci or Acinetobacter were more related to health while other like Treponema, Fusobacteria, and Prevotella were associated with oral disease states in adults (Ledder et al., 2007; Abusleme et al., 2013; Wade, 2013). At the same time, hundreds of rare bacteria have been neglected in analysis which may be due to their being difficult to cultivate and/or detect, or because their detected numbers do not allow for statistical analysis. Focusing on single species can lead to distortion of the real picture of disease. But, how can we compare patients, possible treatment effects, sampling methods, etc. when the information we get consists not only of 20 bacteria but of thousands of species? In addition, how can clinicians translate this information? In this work, we test and show the exemplary comparison of two subgingival biofilm sampling modes for 454-pyrosequencing. We hypothesize that a modification of the clinical sampling mode can lead to a difference in the microbiome composition. We discuss statistical analyses and bioinformatics to provide information on how to compare on an inter- and intra-individual level the microbiota of the subgingival biofilm of healthy children. Issues related to small sample sizes and sample size calculation are also addressed. The overarching aim of this study is to reach the community of dentists and orthodontists with yet scarce knowledge of the potential of microbiome studies. We wish to raise interdisciplinary awareness for the clinical perspective of oral microbiome research in view of translational medicine from bench-side to patient. According to this announcement we, firstly, address the influence of external factors (i.e., clinical sampling methods) on the stability of a microbiome and, secondly, aim to support methods, possibilities and approaches to change and control the subgingival microbiome in human disease through our clinical work and toward a standardized pipeline. Finally, we look at interdisciplinary collaborations to facilitate the transfer of oral microbiome data to real clinical application.

#### MATERIALS AND METHODS

#### Subjects

For this comparative study, we included five ten-year-old children of both sexes (two male, three female). All recruited children had fulfilled the following criteria for participation in this study: late mixed dentition with first premolars fully erupted in the upper arch, good general and periodontal health, no bleeding on probing, a plaque index below 30%, no antibiotic intake within the previous 3 months, and no use of antiplaque solutions. Prior to enrollment, written informed consent was obtained from each participant and one of his or her parents. The study was approved by the institutional review board at the Medical University of Graz. Written consent was also obtained explicitly for the publication of the intraoral photo in **Figure 1**.

### Sampling of Subgingival Biofilm: Modes A and B

Clinical examination and sampling was performed by a single, experienced investigator. Sampling was always performed after

FIGURE 1 | Sampling in the maxilla after full arch isolation.

standardized oral hygiene instructions over a period of 3 weeks. Prior to subgingival sampling, supragingival plaque as disclosed by an indicator—was removed with a previously unused toothbrush and water performed by the children themselves. No toothpaste was used. Full arch isolation in the maxilla was obtained by NOLA Dry Field © system as shown in **Figure 1**. Subgingival biofilm was then collected from the upper first premolars and first molars. Biofilm sampling had to be performed very carefully, so as not to traumatize the young gingival tissue in the absence of pockets commonly seen in periodontal disease. Healthy subjects and especially children have small compartments that make probing subgingival biofilm very challenging. Sterilized and UV-treated paper points (ISO15, Antaeos <sup>R</sup> ) were inserted into the subgingival sulcus parallel to the gingival margin at eight sites located mesio- and distobuccally of the four index teeth in two run-throughs differing slightly in their sampling mode. Sampling was done, firstly, excluding (Mode A) and, secondly, including (Mode B) supragingival cleansing with a sterile cotton pellet (see **Figure 2**). So the main difference between Mode A and Mode B refers to the supragingival cleaning. Samples were taken in sequence during the same sampling procedure from the same eight sites and then pooled and stored at −20◦C until processed (see **Figure 3**).

#### DNA Extraction

Bacterial DNA was prepared by first placing the paper points in a mixture of 380 µl of MagNA Pure Bacteria Lysis Buffer (Roche Applied Science, Mannheim, Germany) and 20 µl of proteinase K solution (20 g/l). The suspension (including

the paper points) was incubated at 65◦C for 10 min and subsequently at 95◦C for another 10 min. After removal of the paper points, the suspension was transferred into the MagNA Pure Compact Sample Tube (Roche Diagnostics, Mannheim, Germany). Automated DNA extraction was performed on the MagNA Pure Compact instrument (Roche) according to manufacturer instruction using the MagNA Pure Compact Nucleic Acid Isolation Kit I (Roche Diagnostics, Mannheim, Germany). Prior to the start of DNA extraction, the instrument adds the heterologous IC automatically. For extraction of bacterial DNA, the DNA Bacteria Purification protocol was used according to manufacturer instructions. DNA was eluted in 50 µl dH2O and stored at <sup>−</sup>20◦C until use.

#### 454-Pyrosequencing

Pyrosequencing was performed by DNAvision (avenue George Lemaitre 25B, 6041 Charleroi, Belgium, http://www.dnavision. com/). Microbial diversity was analyzed targeting 16S rRNA hypervariable regions V5 and V6. Pyrosequencing primers used are shown in **Supplementary Table S1** containing 16s rRNA target specific primer sequences 784F-5′ -AGAGTTTGA TCCTGGCTC-3′ and 1061R-5′ ATTACCGCGGCTGCTGG-3′ (italic) according to Andersson et al. (2008), MID sequence (underlined), four bases key sequence and the Roche Titanium adaptor sequences (bold). For each sample, a PCR mix of 100 µl was prepared containing 1 × PCR buffer, 2U of KAPA HiFi Hotstart polymerase and dNTPs (Kapa Biosystems), 300 nM primers (Eurogentec, Liege, Belgium), and 60 ng total DNA. Thermal cycling consisted of initial denaturation at 95◦C for 5 min, followed by 25 cycles of denaturation at 98◦C for 20 s, annealing at 56◦C for 40 s, and extension at 72◦C for 20 s, with a final extension of 5 min at 72◦C. Amplicons were visualized on 1% agarose gels using GelGreen Nucleic Acid gel stain in 1xTAE (Biotium) and were cleaned using the Wizard SV Gel and PCR Clean-up System (Promega, Mannheim, Germany) according to manufacturer instructions. Amplicon DNA concentrations were determined using the Quant-iT PicoGreen dsDNA reagent and kit (Life Tech, Carlsbad, USA) following manufacturer instructions. After quantitation, cleaned amplicons were mixed in equimolar ratios into a single tube. The final pool was again purified using Agencourt Ampure XP purification systems according to manufacturer instructions (Agencourt Biosciences Corporation-Beckman Coulter, USA) and then eluted in 100 µl of 1xTE. The concentration of the purified, pooled DNA was determined using the QuantiT PicoGreen dsDNA reagent and kit (Life Tech, Carlsbad, USA) following manufacturer instructions. Pyrosequencing of an equimolar pool of 10 samples on 1/8 PTP was carried out using primer A on a 454 Life Sciences Genome Sequencer FLX instrument (Roche, Mannheim, Germany) and following GS FLX Titanium Sequencing Kit XLR70 chemistry (Roche 454 Life Science, Branford, CT, USA) according to manufacturer instructions which resulted in 4131–19,943 raw reads per sample. Sequences are available at NCBI, accession number: SRP080750.

#### Sequence Data Analysis

In the first step, generated sequence data was assessed for quality. By using our own perl script only sequences with a minimum length of 150 bases, average Phred score of 25 and no ambiguous bases were selected for use in the downstream analysis. The remainder of the analysis was performed with the Quantitative Insights Into Microbial Ecology (QIIME) pipeline version 1.3.0 using standard parameters, including uclust (Edgar, 2010) for building OTUs with a similarity threshold of 0.97, pyNAST (Caporaso et al., 2010) for representative sequence alignment, FastTree (Price et al., 2009) to generate the phylogenetic tree and RDP classifier (Wang et al., 2007) for taxonomic assignment. Chimeric sequences were removed using ChimeraSlayer with default QIIME settings after OTU picking and taxonomic assignment on aligned representative sequences.

Estimation of the within-samples diversity (alpha diversity) was performed with the Simpson (1949), Shannon (1948), and Chao (1984) metrics.

In the final step, we generated PCoA plots and performed hierarchical clustering analysis based on distance matrix from an unweighted UniFrac phylogenetic method (Lozupone et al., 2011) which enabled the between-samples comparison (beta diversity) of the microbial communities.

For the beta diversity analysis and normalization, sample heterogeneity was excluded by rarefication of all samples to the sample with the lowest number of reads.

#### Statistical Analysis

Statistical analyses were performed using SPSS version 22.0 (SPSS Inc., Chicago, IL), R version 3.11 (R Core Team, 2015) and PASS 2012 (NCSS, LLC. Kaysville, Utah). Data are presented as median and as interquartile range (lower quartile 25-percentile and upper quartile 75-percentile). Inter-individual differences of the median relative abundances served for the comparison of the two sampling modes. Wilcoxon tests with Bonferroni correction for multiple testing were used to compare Mode A (excluding supragingival cleansing) and Mode B (including supragingival cleansing) on phylum level (n = 6), on class level (n = 14), on order level (n = 19), on family level (n = 27) and on genus level (n = 29). All names of the specific bacteriae are provided in **Tables 2.1**–**2.5** and in **Figure 6**. Paired t-tests were performed additionally, since the small sample size did not allow to verify the assumption of normality for the data. All reported p-values were two-sided. After Bonferroni correction statistical significance was considered with p < 0.0083 at the phylum level, p < 0.0036 at the class level, p < 0.0026 at the order level, p < 0.0019 at the family level and p < 0.0017 at the genus level.

To test differences in abundance for a total of n taxa between two groups, the rank-sum test including multiple testing with Bonferroni correction was used to estimate the power and the sample size for different effect sizes for alpha level of 0.05/n.

Significance for PCoA (beta-diversity) analyses was checked with multivariate permutation tests using the nonparametric method "Adonis" (999 permutations) included in the package "vegan" of the QIIME-incorporated version of "R."

### RESULTS

### Pyrosequencing and Diversity Indices

A total of 92,680 sequences were derived from pooled DNA of 10 samples in the pyrosequencing assay. Data filtering and quality control resulted in 67,218 sequences with an average sequence length of 243 bp (SD 6.52; range 231–255), read numbers per sample ranging from 2937 to 14,629 sequences.

Rarefaction curve analysis showed that the sequencing effort was not sufficient to cover the whole microbiota in the analyzed samples. It is very likely that rare taxa and taxa with low abundances have been missed (**Supplementary Figure S1**). Nevertheless, this should not significantly influence results, since low abundant taxa do not shift the complete microbiota profiles and the tools used for their comparison are robust enough to compensate for low deviances.

The number of OTUs defined at 97% identity ranged from 532 to 1107 (as shown in **Table 1**). Sample richness, which in this analysis equals to the number of OTUs, as well as sample diversity (Shannon Index range 4.26–5.31) did not demonstrate major differences between the two sampling modes.

#### Taxonomic Summary

Analysis over the whole microbiota showed a predomination of the five main phyla: Actinobacteria (2.8–24.6%), Bacteroidetes (9.2–25.1%), Proteobacteria (4.9–50.6%), Firmicutes (16.5–57.4%) and Fusobacteria (2.2–17.1%) (**Figure 4**).

The median relative abundances for all representatives in the profiled microbiomes on different taxonomic levels (phylum, class, order, family and genus) are given in **Table 2**.

**Figures 4**, **5** show sampling modification effects on relative abundances on phylum level (barchart) and on class level (heat map). A concordant qualitative pattern within individuals and differences between individuals could be shown regardless of the sampling mode.

### Statistical Analyses of Sampling Modes and Sample Size Calculation

#### Differences between Sampling Modes A and B

Effects in the subgingival microbiome profiles possibly due to sampling modification are displayed by area graphs in **Figure 6**. P-values from Wilcoxon signed rank tests and the median


TABLE 1 | Sequencing information and diversity estimates for the subgingival microbiome profiles in five healthy children before (Mode A) and after (Mode B) supragingival cleansing.


Total numbers of OTUs similarity 97% 5601

relative abundance at all five taxonomic levels were used to display differences between Mode A (excluding supragingival cleansing) and Mode B (including supragingival cleansing) for all bacterial species on all levels (**Tables 2.1**–**2.5**). Nearly statistically significant differences (p = 0.063) between sampling Modes A and B could be shown for the phylum of Bacteroidetes based on Wilcoxon signed-rank tests: Bacteroidia (class), Bacteroidales (order), Prevotellaceae (family), and Prevotella (genus). The latter was shown to be statistically significant (p = 0.047) when the paired t-test was applied. Paired t-tests were assessed additionally to Wilcoxon signed-rank tests due to the small sample size in the study so as to prove that nearly statistically significant results with Wilcoxon signed-rank tests become significant. In general, the Wilcoxon signed-rank test cannot be significant for a sample size smaller than 6, for two sided testing. For one sided testing, a sample size of at least 5 is needed for the result to be significant. For the paired t-test there is no such limitation. Notably, after correction for multiple testing, almost all differences were no longer nearly significant (**Table 2.1** through **Table 2.5** and **Figure 6**).

#### Sample Size

Based on the Wilcoxon signed rank test and the assumption of a power of 85% (as required by the local Ethics Committee), a high variety of different sample sizes are thus needed for the bacterial representatives on different taxonomic levels. The bigger the effect size and the smaller the standard deviations, the fewer samples are needed. For example **Table 2.3** shows that the calculated sample sizes needed for the 19 bacterial species on order level ranged between 8 for Bacteroidales (median A = 3.7; median B = 7.9) and 82,194 for Actinomycetales (median A = 7.9; median B = 8.6), despite the huge sample size of 110,445 needed for more or less rare and undefined representatives. At class level, only two more subjects for Bacteroidetes (phylum)-Bacteroidia and, at order level, only three more subjects for Bacteroidetes (phylum)-Bacteroidales would have been needed to reach a power of 85% and to obtain a significant result for the Wilcoxon signed rank test, assuming that effect size and standard deviation remain constant (see **Table 2**).

#### Multivariate Analysis: Principal Coordinate Analysis (PCoA) and Hierarchical Clustering

Principal coordinate analysis (PCoA) on distance matrices calculated with unweighted UniFrac showed a grouping of the paired samples (**Figure 7**). Pairs, shown in the same color, are close together in all three dimensions, except for Sample

supragingival cleansing.

5A that shows a respective deviation and shows similarity with both samples from Subject 1. The results using Adonis (Permutational MANOVA) revealed no grouping of the samples according to Mode A or Mode B (p = 0.914 by R <sup>2</sup> = 0.09) and thus no significant effects between Mode A and Mode B. PCoA results showed greater variability BETWEEN than WITHIN individuals. This observation was also supported by agglomerative hierarchical clustering analysis with average linkage on unweighted UniFrac distance (**Figure 8**).

species present in all samples before (Mode A) and after (Mode B) supragingival cleansing (phylum, class, family and genus taxon).

#### DISCUSSION

Oral microbiota are considered one of the main risk factors for periodontal diseases affecting up to 90% of the world population (Pihlstrom et al., 2005). Oral biofilms have become increasingly important as a source of caries and periodontal disease as well as other bacterial infections in the human organism (Benítez-Páez et al., 2014). Some studies reveal evidence that oral pathogens play a role in various inflammatory diseases (Offenbacher et al., 2008). Few studies have deeply analyzed the composition of subgingival biofilm and elucidated the phylotypes/species associated with health or disease (Paster et al., 2001; Socransky and Haffajee, 2005; Ledder et al., 2007; Diaz, 2012; Abusleme et al., 2013).

The presented study analyzed using 454-pyrosequencing the data of five healthy 10-year-old children whose subgingival biofilm was examined excluding and including supragingival cleansing (Mode A and Mode B, respectively). The study aimed at assessing the effect of a slight modification of the clinical sampling technique for its accuracy in reflecting subgingival microbiome sequence data.

Retrieving adequate and reproducible samples is a challenge but awareness of the natural variability within subgingival microprints would enable us to distinguish pathological patterns at an early stage of disease. Corresponding in vivo conditions can best be studied in healthy children as shown in previous studies.

However, very few oral microbiome studies in healthy children have been performed so far (Papaioannou et al., 2009; Xin et al., 2013), some including pyrosequencing (Crielaard et al., 2011; Stahringer et al., 2012; Ling et al., 2013; Lif Holgerson et al., 2015). The study design of Crielaard et al. differs from ours in that they investigated microbial profiles of saliva collected from caries-diseased Dutch children aged 3–18 years. The biggest difference in the comparable age strata was the relative abundance of Firmicutes at 58% in the saliva group and at 30% in our subgingival samples, while the latter presented a higher proportion of Proteobacteria (22 vs. 12%) and Fusobacteria (6 vs. 2%). Ling et al. used parallel barcoded 454-pyrosequencing to study the diversity and richness of salivary bacteria in 10 healthy children and adults. The bacterial diversity was found to be more complex in children than in adults (Ling et al., 2013) which could be interpreted as evidence for the relationship between biodiversity and health. In their sample comprising 60 children aged 3–6, the eight predominant phyla in supragingival plaque and saliva were present in proportions that were comparable to our study: 23–42% Firmicutes and 16–37% Bacteroides (Ling et al., 2010). In a longitudinal study, Holgerson et al. looked at the oral microbiota of 207 Swedish babies at the age of 3 months and again at 3 years. The pyrosequencing data referred to 11 children with and 11 without caries. A significant increase in species richness and taxa diversity was described. Several taxa within the oral biofilms of the 3-year-olds could be linked to the presence or absence of caries. However, quantitative comparisons of the oral microbiota of children are possible only to a limited extent, since the investigators dedicated work differs in parameters such as study population (age, country, caries


TABLE 2.1 | Comparison of the median relative abundance corresponding to 6 bacterial species on phylum taxon present in all samples before (Mode A) and after (Mode B) supragingival cleansing: median and IQR, p-values for paired t-test and Wilcoxon signed-rank test and sample size calculation.

§Power of 0.85 is assumed.

After Bonferroni correction p < 0.0083 is significant.

TABLE 2.2 | Comparison of the median relative abundance corresponding to 14 bacterial species on class taxon present in all samples before (Mode A) and after (Mode B) supragingival cleansing: median and IQR, p-values for paired t-test and Wilcoxon signed-rank test and sample size calculation.


§Power of 0.85 is assumed.

After Bonferroni correction p < 0.0036 is significant.

status), sampling sites (saliva, mucosal, and supragingival plaque) and molecular methods (DNA-DNA checkerboard, micro arrays, pyrosequencing). Papaioannou et al., for example, looked at five different oral habitats (saliva, tongue, soft tissue, subgingival, and total supragingival plaque) of 93 children from three different age groups using whole genomic probes for 38 species and the checkerboard DNA-DNA hybridization technique. The authors suggest a gradual maturation of the oral microbiota in children displaying patterns of colonization similar to those seen in adults (Papaioannou et al., 2009). However, until now most studies have analyzed salivary biofilm, as it is easier to sample (Luo et al., 2012; Stahringer et al., 2012; Ling et al., 2013; Gomar-Vercher et al., 2014). Interestingly, Luo et al. who studied PCR-amplified bacterial DNA from the saliva of 20 children with caries and of 30 healthy ones found higher microbial diversity in samples from diseased oral cavities. In contrast to these findings, Gomar-Vercher et al. used pyrosequencing to analyze 110 saliva samples from children split into six groups according to caries severity and found the bacterial diversity to decrease with progressing disease. At the same time, intra-group differences were considerable (Gomar-Vercher et al., 2014). Stahringer and colleagues presented a longitudinal survey of salivary microbiota from twins using PCR amplification and 454 pyrosequencing of the 16S rDNA hypervariable regions V1 and V2. Their findings


TABLE 2.3 | Comparison of the median relative abundance corresponding to 19 bacterial species on order taxon present in all samples before (Mode A) and after (Mode B) supragingival cleansing: median and IQR, p-values for paired t-test and Wilcoxon signed-rank test and sample size calculation.

\*Taxa marked with asterisk could not be assigned to any of the ordera and are shown on class level as lowest common taxon. §Power of 0.85 is assumed. After Bonferroni correction p < 0.0026 is significant.

point to the environment as the microbiome-determining factor showing greater differences between non-related subjects than within individuals or between twins (as long as they share a common habitat; Stahringer et al., 2012).

Standardized sampling procedures are a prerequisite for comparing subgingival microbiome data derived from research worldwide. The lack of heterogeneity and standardization for clinical protocols poses a limitation to data quality which should be noted by clinicians and microbiologists. In this context, we need to consider the diverse sampling methods reported for the collection of samples from a healthy oral cavity, not to mention the variability of pocket sampling in periodontally diseased patients. This can be illustrated by the example of just 10 published manuscripts dealing with the collection of samples from an intact oral cavity. They report using saline oral wash rinse (Ahn et al., 2011) or unstimulated whole saliva (Xin et al., 2013) for fluid collection; dental explorers (Xin et al., 2013), metal loops (Ling et al., 2010), metal curettes (Papaioannou et al., 2009) and wooden tooth picks (Keijser et al., 2008) for supragingival sampling; or wet and dry swabs and brushes (Aas et al., 2005; Papaioannou et al., 2009; Cortelli et al., 2012) and spatulas (Gohler et al., 2014) for mucosal sampling. Finally, subgingival sampling is currently being performed using either metal curettes (Papaioannou et al., 2009; Abusleme et al., 2013) or paper points (Cortelli et al., 2012; Griffen et al., 2012; Jünemann et al., 2012). For clinical and research purposes even exotic micropipettes or microelectrodes are used (Geibel, 2006). Potential sampling variability springs not only from the different instruments that can be utilized but also from processes taking place prior to sampling, such as plaque control, tooth cleaning, tooth isolation and drying, as well as from inadequate specifications regarding the sampling technique and time lines. Compared to standards that apply in other medical and laboratory settings, our clinical sampling is much like an elephant in a porcelain shop. Appropriate scientific input facilitates the development of a systematic and precise methodology which in turn can deliver reliable, high-quality clinical samples to the pipeline required in the field of molecular biology and medicine. Some authors have reported on the recovery of putative pathogens from paper point and curette sampling (Jervøe-Storm et al., 2007; Teles et al., 2008; Angelov et al., 2009; Sahl et al., 2014). Hartroth and colleagues have evaluated paper point sampling on bench (Hartroth et al., 1999), but these findings have yet to be tested under clinical conditions to establish the best practice.


TABLE 2.4 | Comparison of the median relative abundance corresponding to 27 bacterial species on family taxon present in all samples before (Mode A) and after (Mode B) supragingival cleansing: median and IQR, p-values for paired t-test and Wilcoxon signed-rank test and sample size calculation.

\*Taxa marked with asterisk could not be assigned to any of the family taxon and are shown on ordera level as lowest common taxon.

\*\*Taxa marked with asterisk could not be assigned to any of the family taxon and are shown on class level as lowest common taxon.

§Power of 0.85 is assumed.

After Bonferroni correction p < 0.0019 is significant.

An aspect on which clinical researchers are in agreement is the removal of supragingival plaque before subgingival sampling. It is as obvious to them as taking off the shoes in the hallway before entering the living room. However, it is still debatable to what extent this cleansing should be performed to be efficient enough.

Generally, clinical sampling within the oral cavity of children can be tricky and calls for an experienced investigator. The clinical method in this study is designed around a younger study population with intact and tight subgingival compartments. The subgingival sulcus itself can best be imagined as an interface (of two millimeters) with a tight epithelium toward the periodontium but with a seamless junction (orifice) toward the supragingival surface. Thus, not only the removal of nonattached bacteria but also the microbial exchange between suband supragingival biofilm has to be taken into account in addition to the difficulty of precise sampling in this extremely limited subgingival space. Limited space makes sampling the subgingival sulcus of children a challenge. The deeper the sulcus, the more likely it is to strip supragingival biofilm before actually reaching the sulcus depth. In our case, sampling was performed by a single, experienced clinician excluding interrater variability. Paper points were gently slid parallel to the gingival margin in order to facilitate a painless and quick examination. This


TABLE 2.5 | Comparison of the median relative abundance corresponding to 29 bacterial species on genus taxon present in all samples before (Mode A) and after (Mode B) supragingival cleansing: median and IQR, p-values for paired t-test and Wilcoxon signed-rank test and sample size calculation.

\*Taxa marked with asterisk could not be assigned to any genera and are shown on family level as lowest common taxon.

\*\*Taxa marked with asterisk could not be assigned to any genera and are shown on order level as lowest common taxon.

\*\*\*Taxa marked with asterisk could not be assigned to any genera and are shown on class level as lowest common taxon.

§Power of 0.85 is assumed.

After Bonferroni correction p < 0.0017 is significant.

contributes to better cooperation on behalf of the child and a short procedure prevents the paper point from becoming saturated with saliva. Paper points were used rather than the more invasive metal curette, as the latter could traumatize the subgingival sulcus and cause bleeding which was to be avoided at all costs. During the sampling procedure, the focus was placed on the drier subgingival areas of the upper arch, so as to optimize sample quality for DNA analysis. To ensure reproducibility, biofilm sampling followed a strict protocol (see also Methods above). Two modes (A and B) of the same sampling method were used for comparison. After supragingival cleaning using an electric toothbrush and water, sampling was performed, firstly, excluding (Mode A) and, secondly, including (Mode B) cleansing with sterile cotton pellets. The samples from a total of eight sites were pooled, so no inter-site comparisons were studied. Based on the paired samples, results of the PCoA intraindividual differences were relatively small despite the modest sample size in the present study. Also, permutational MANOVA showed no grouping of the samples according to Mode A or Mode B. It can be speculated that any existing deviation

between the two sampling modes is very likely to correspond to a natural variation in oral biofilm of the individual subject and supragingival cleansing with a sterile cotton swab does not affect the composition of the subgingival biofilm of an individual. Importantly, it seems that there are no major effects due to the described sampling modification. However these "non-effects" between the two sampling modes refer to interindividual differences and obviously surpass the intra-individual "non-effects" which comes up to an overarching effect with relevance for future clinical studies

Analyses of the pooled DNA data using pyrosequencing is a timely and potentially interesting approach that also has numerous limitations. **Table 1** shows that richness and evenness as well as Shannon diversity index do not indicate any differences in the above mentioned sampling modes (A and B). However, it should be noted that the number of reads can influence the sensitivity of data; this issue is for example evident when comparing the quotient of the number of reads and richness for subjects 4A and 5A. Such differences in the number of reads are practically unavoidable, therefore it is necessary to incorporate a normalization step into the data analysis which we did by rarefication of all samples to the sample with the lowest number of reads. Another option is the use of relative abundances as also applied in this study for statistical comparisons. The field under study here is so complex that it is impossible to ascertain at which exact point in the analysis problems occur, and whether the same amount of DNA was available from the participating children and/or if data loss had occurred even earlier. Laboratory workup is not discussed here in detail but the possibility of passive errors (e.g., during 16S RNA amplification for PCR) does exist despite standardized procedures. It also has to be remembered that this study looked at 16S rRNA hypervariable regions V5 and V6 only and not at the whole metagenome. This limitation also applies to other studies (Wu et al., 2010; Ahn et al., 2011; Griffen et al., 2011; Jünemann et al., 2012; Stahringer et al., 2012).

An asset of our study is the fact that (under the aforementioned conditions and for the afore-mentioned subjects) sample size calculations are presented for the bacterial species on all five levels as shown in **Tables 2.1**–**2.5**. Our small sample size poses a challenge for pyrosequencing and statistical testing, nevertheless different effects can be observed as visualized in **Figures 6**–**8**. In our study, sample sizes are part of the findings. Our work should emphasize that the challenge is the translation of sample size estimations to clinical feasibility. So far, statistically given sample sizes that would explain significantly and clinically relevant differences in the subgingival microbiome of children are neither practical nor ethical. Even a generous increase in samples, i.e. children, in our study would not have solved the problem. However, our data can serve as a pilot for future studies on the topic showing that large sample sizes are needed to elucidate microbial structures at different levels. The demanding task is to reflect the bacterial diversity as well as possible. However, as opposed to more common bacteria, rare species require huge sample sizes in order to unveil any significant differences. This task becomes even more complex with a higher number of rare species in a given sample. In order to study these issues, statistical methodology will have to be developed further. While appropriate technology is becoming increasingly available and affordable, sample sizes remain primarily a matter of practicality and ethics. Including healthy people, in particular children, or patients into clinical studies involves substantial costs for human resources and efforts beyond the daily routine for both sides: the study participants and the clinical staff. One way to practically increase sample sizes are standardized clinical protocols that would allow multi-site sampling in diverse populations.

In our analysis, the bacteria are only analyzed down to the genus level which is limiting. However, from the clinical perspective the data is noteworthy. Interestingly, the smallest calculated sample size roughly corresponds to the 20 bacteria available in commercial bacteria test kits applied in periodontology. However, some abundant bacteria are apparently not included in such test kits. In addition, numerous bacteria have not yet been identified and are assigned as "other" to superordinate taxonomic levels (see **Tables 2.1**–**2.5**). In this context, the limitation is the unattainable sample size for some phyla.

Another general issue that should be mentioned is the need for standardized protocols to facilitate the comparability of data generated in microbiome studies. Considerable interindividual differences in bacterial communities necessitate large samples. At the same time, intra-individual variability should also be considered in comparative studies. For microbiome data, new statistical methods like Adonis are needed and should be combined with methods from bioinformatics. For example, PCoA was used in our study to verify findings based on a small sample size, i.e., grouping of the paired samples for the within-comparison as intra-individual pairs clustered in all three dimensions. Importantly, many decisions regarding study design are made based on investigator experience (e.g., which distances to analyze with UniFrac). Future studies should aim at standardizing methodology to prevent bias and distortion of data.

Our work points at many challenges in the study of oral microbiomes. Our data, though based on a modest sample size, could serve as a reference for healthy children or may serve as a baseline for microbiome function in healthy individuals shedding new light on the frontiers of health and disease. The number of the species known is high (presently amounting to more than 600 taxa) and includes very rare ones whose role is yet unknown as well as other microbial representatives that are not bacteria (Moissl et al., 2002, 2003, 2005). Methods like DNA/RNA/metagenome sequencing need to be employed to begin to uncover the exact role of microbiota. Similarly, visualized analytics can give additional insight into individual species. However, we still need to learn which microbiota are imperative for the functioning of the whole. And we need to ask further questions: How does diversity make healthy? To what extent may individual health be attained by comparison with other individuals? The presented work employs modern approaches from several research areas but the focus remains on the clinical application and a contribution toward the standardization of procedures across all relevant disciplines.

### AUTHOR CONTRIBUTIONS

ES: work conception and design; acquisition, analysis and interpretation of data; drafting and critical review of the manuscript; final approval of the work for publication. ST: analysis and interpretation of data; critical review of the manuscript; final approval of the work for publication. KE: analysis and interpretation of data; critical review of the manuscript; final approval of the work for publication. BK: study design; analysis and interpretation of data; draft and critical review of the manuscript; final approval of the work for publication.

## FUNDING

This study was supported by the Hygiene Fund of the Institute of Hygiene, Microbiology and Environmental Medicine at the Medical University of Graz, Graz, Austria.

#### ACKNOWLEDGMENTS

The authors are grateful to Michael Bozic (Laboratory at the Research Unit Molecular Diagnostics, Medical University of Graz) for his excellent technical assistance in DNA isolation.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.02142/full#supplementary-material

### REFERENCES


Supplementary Figure S1 | Ten rarefaction curves generated from subgingival microbiome profiles of five healthy children based on two performances of paper point sampling before (Mode A) and after (Mode B) supragingival cleansing with a sterile cotton pellet (colors designate sampling modes).

Supplementary Table S1 | Barcoded primer sequences used in this study: PCR ampicons were sequenced from the Titanium A adaptor (CCATCTCATCCCTGCGTGTCTCCGAC), followed by a 4 bases key sequence (TCAG) and the barcode. The reverse primer was used with the Titanium B adaptor (CCTATCCCCTGTGTGCCTTGGCAGTC), the key sequence and the target specific sequence but without barcode sequence.


**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 © 2017 Santigli, Trajanoski, Eberhard and Klug. 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) or licensor 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.

# From Mouth to Model: Combining in vivo and in vitro Oral Biofilm Growth

Barbara Klug1, <sup>2</sup> , Elisabeth Santigli <sup>2</sup> \*, Christian Westendorf <sup>1</sup> , Stefan Tangl 3, 4 , Gernot Wimmer <sup>5</sup> and Martin Grube<sup>1</sup>

*1 Institute of Plant Sciences, University of Graz, Graz, Austria, <sup>2</sup> Department of Dental Medicine and Oral Health, Division of Oral Surgery and Orthodontics, Medical University of Graz, Graz, Austria, <sup>3</sup> Karl Donath Laboratory for Hard Tissue and Biomaterial Research, Department of Oral Surgery, Medical University of Vienna, Vienna, Austria, <sup>4</sup> Austrian Cluster for Tissue Regeneration, Vienna, Austria, <sup>5</sup> Department of Dental Medicine and Oral Health, Division of Preventive and Operative Dentistry, Periodontology, Prosthodontics and Restorative Dentistry, Medical University of Graz, Graz, Austria*

Edited by:

*David Berry, University of Vienna, Austria*

#### Reviewed by:

*David A. C. Beck, University of Washington, USA Anna Edlund, J. Craig Venter Institute, USA Christian T. K.-H. Stadtlander, Independent Researcher, St. Paul, USA*

\*Correspondence: *Elisabeth Santigli elisabeth.santigli@medunigraz.at*

#### Specialty section:

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

Received: *29 April 2016* Accepted: *30 August 2016* Published: *21 September 2016*

#### Citation:

*Klug B, Santigli E, Westendorf C, Tangl S, Wimmer G and Grube M (2016) From Mouth to Model: Combining in vivo and in vitro Oral Biofilm Growth. Front. Microbiol. 7:1448. doi: 10.3389/fmicb.2016.01448* Background: Oral biofilm studies based on simplified experimental setups are difficult to interpret. Models are limited mostly by the number of bacterial species observed and the insufficiency of artificial media. Few studies have attempted to overcome these limitations and to cultivate native oral biofilm.

Aims: This study aimed to grow oral biofilm *in vivo* before transfer to a biofilm reactor for *ex situ* incubation. The *in vitro* survival of this oral biofilm and the changes in bacterial composition over time were observed.

Methods: Six human enamel-dentin slabs embedded buccally in dental splints were used as biofilm carriers. Fitted individually to the upper jaw of 25 non-smoking male volunteers, the splints were worn continuously for 48 h. During this time, tooth-brushing and alcohol-consumption were not permitted. The biofilm was then transferred on slabs into a biofilm reactor and incubated there for 48 h while being nourished in BHI medium. Live/dead staining and confocal laser scanning microscopy were used to observe bacterial survival over four points in time: directly after removal (T0) and after 1 (T1), 24 (T2), and 48 h (T3) of incubation. Bacterial diversity at T0 and T3 was compared with 454-pyrosequencing. Fluorescence *in situ* hybridization (FISH) was performed to show specific taxa. Survival curves were calculated with a specially designed MATLAB script. Acacia and QIIME 1.9.1 were used to process pyrosequencing data. SPSS 21.0 and R 3.3.1 were used for statistical analysis.

Results: After initial fluctuations at T1, survival curves mostly showed approximation of the bacterial numbers to the initial level at T3. Pyrosequencing analysis resulted in 117 OTUs common to all samples. The genera *Streptococcus* and *Veillonella* (both *Firmicutes*) dominated at T0 and T3. They make up two thirds of the biofilm. Genera with lower relative abundance had grown significantly at T3. FISH analysis confirmed the pyrosequencing results, i.e., the predominant staining of *Firmicutes*. Conclusion: We demonstrate the *in vitro* survival of native primary oral biofilm in its natural complexity over 48 h. Our results offer a baseline for cultivation studies of native oral biofilms in (phyto-) pharmacological and dental materials research. Further investigations and validation of culturing conditions could also facilitate the study of biofilm-induced diseases.

Keywords: native oral biofilm, in vitro growth, dental splint, human enamel-dentin slabs, live/dead staining, 454 pyrosequencing

#### INTRODUCTION

At the beginning of the twenty-first century, natural biofilm modeling still poses a great challenge. The biofilm lifestyle of oral bacteria is difficult to simulate, as a normal human oral microbiome comprises more than 700 different bacterial taxa (Aas et al., 2005). The composition of the bacterial community and its spatial distribution have been studied in various ways to reveal a highly structured organization of the biofilm. The matrix surrounding and protecting the biofilm has been compared to a bacterial "house" (Flemming et al., 2007) that provides structures like channels for nutrition supply and communication. Varying surrounding conditions inside a biofilm modify bacterial lifestyles and lead to adaptations, e.g., they render bacteria more resistant to antibiotics (Høiby et al., 2011). Under healthy conditions, this oral ecosystem is in a homeostatic state (Marsh, 2006), but this does not imply uniformity across the microbiome composition. Quite the opposite is the case, as each human being hosts a genuine oral microbiome with an individual bacterial composition on the whole and within each oral compartment (Arweiler et al., 2004; Trajanoski et al., 2013; Langfeldt et al., 2014), e.g., the salivary microbiome composition differs from that of the subgingiva, and the tongue's microbiome is different from that of the cheek (Aas et al., 2005; Simón-Soro et al., 2013). Moreover, our knowledge about natural co-occurrence patterns and patterns of mutual exclusion is still incomplete (Human Microbiome Project Consortium, 2012; Segata et al., 2012). In addition to the bacterial composition, factors outside the biofilm can have an influence on it. They include parameters such as shear stress through salivary flow, natural temperature oscillations, pH value changes, host immunity factors, stress or dietary variation, all of which still need to be fully explored and understood (Rittman, 1982; van Houte et al., 1982; Saunders and Greenman, 2000; Picioreanu et al., 2001; Wimmer et al., 2005; Al-Ahmad et al., 2007; De Filippo et al., 2010; Fierer et al., 2010; Hajishengallis, 2010; Schlafer et al., 2011). It is almost impossible to include all these factors as model parameters, especially considering the fact that they have not yet all been identified. As a result, most of the experimental setups in in vitro models so far have generally focused on a reduced number of these parameters. However, the processes and interactions in complex oral biofilms are difficult to interpret based on simplified experimental setups. Many bottom up assays in microbiology are still limited by two main factors: firstly, the number of bacterial species included and, secondly, the artificial media used to feed the biofilm. They explore individual functional roles and inter-individual interactions (Hansen et al., 2000; Mazumdar et al., 2008; Periasamy and Kolenbrander, 2009; Standar et al., 2010; Agostinho et al., 2011). Alternatively, a top down approach compiles information about structure, spatial distribution and community composition (Filoche et al., 2010; Zijnge et al., 2010; Klug et al., 2011; Edlund et al., 2013; Nyvad et al., 2013; Jorth et al., 2014; Sintim and Gürsoy, 2015; Zheng et al., 2015). Also, advanced experimental setups that include different media and several bacterial species only insufficiently reflect natural conditions, particularly those in the oral habitat. Combining all these setups can lead to better models that realistically mimic natural conditions. Even if the biological parameters cannot be fully reconstructed, tracking the bacterial composition of multispecies biofilms can provide new insights into their interactions. One important step is to enable the transfer of native oral biofilm to experimental setups, and to keep it vital and diverse under laboratory conditions. In the present study, we introduce a standardized workflow to grow native oral biofilm in vivo, to transfer this biofilm into an in vitro environment and to keep it alive in that environment. With this "mouth to model" procedure, we demonstrate the survival of primary oral biofilm grown natively under simplified laboratory conditions and the changes that take place in the bacterial composition.

### MATERIALS AND METHODS In vivo Oral Biofilm Growth—Study Participants and Dental Splints

This study was approved by the institutional review board at the Medical University of Graz. Written informed consent was obtained from all study participants in accordance with the Declaration of Helsinki. The study design and the role of study participants were communicated to them in advance.

Twenty-five dental students aged between 20 and 25 years were recruited via notice-board at the School of Dentistry of the Medical University of Graz. Prior to enrollment a short history was taken to ensure that the following inclusion criteria were met: non-smoker, good general health, no present medication and no antibiotic intake 3 months prior to this study. Due to possible hormonal shifts, only male students were included. For each study participant a dental splint was fitted individually to the upper jaw including six standardized (6 × 4 mm) enameldentin slabs for native oral biofilm collection. The slabs had been prepared from patients' teeth that had been extracted for medical reasons at the outpatient clinic of the University Department of Dentistry in Graz. The dentin-enamel slabs were sterilized and then cut, grinded and polished at the Karl Donath Laboratory for Hard Tissue and Biomaterial Research in Vienna. After that, they were integrated buccally in individually fitted dental splints facing the surface of the teeth (**Figures 1Aa,b**), leaving a small gap between slab and tooth (**Figures 1Ac,d**). Participants were asked to wear the dental splint continuously for 48 h. They were not allowed to drink alcohol or brush their teeth during this time to guarantee undisturbed biofilm growth.

#### In vitro Biofilm Growth–Biofilm Reactor

Dental splints were removed carefully directly in the lab after 48 h and immediately placed into a pre-warmed Brain Heart Infusion (BHI, Roth, Austria) medium. Permanently covered with BHI, the enamel-dentin slabs were consecutively clipped out and transferred into the DFR 110 biofilm reactor (Biosurface Technologies Corporation, Montana, USA). Biofilm was then incubated for another 48 h at 34◦C with a BHI flow rate of 0.2 ml/min (**Figures 1B,C**). Measurements were performed at four points in time: T0—directly after removal from the mouth, T1—after 1 h incubation, T2—after 24 h incubation, and T3 after 48 h incubation (**Figure 1D**).

#### Live/Dead Staining

Biofilm on enamel-dentin slabs was stained with LIVE/DEAD <sup>R</sup> BacLightTM Bacterial Viability Kit (Molecular Probes <sup>R</sup> ) according to protocol. Slabs were fully submerged in the staining solution [Syto 9, green fluorescence, and Propidium iodide (PI), red fluorescence] for 20 min at room temperature. After washing with sterile ddH2O, the biofilm was analyzed directly on the slabs with a Leica TCS-SP confocal laser scanning microscope (CLSM). The slabs were covered with water, and water-immersible objectives (HCX APO L 20x/0.5 W UVI/D 3.5 and HCX APO L 63x/0.90 W) were used to generate stack data. Filters were set at 501–531 nm for Syto 9 and 600–672 nm for PI. At least five stacks at random positions were recorded for each slab.

#### Image Processing

The set of all non-zero pixels of an individual confocal stack was clustered into four different clusters, using the kmeans function in MATLAB <sup>R</sup> (Statistics and Machine Learning Toolbox). The cluster with the lowest mean represented the background and noise, and was subtracted from the image stack, whereas the three remaining clusters represented the foreground. Occasionally present yeast and oral mucosa cells were manually selected and removed from the binary masks of both corresponding confocal stacks. The area covered by stained bacteria was simply calculated as the count of all non-zero pixels in the entire binarized confocal stack. As a mixed environmental biofilm, a fraction of cells was labeled by both dyes contained in the LIVE/DEAD <sup>R</sup> BacLightTM Bacterial Viability Kit. Double-labeled pixels (orange) were always counted as dead and removed from the corresponding binary mask of living bacteria. Finally, for each corresponding pair of confocal stacks, the fraction of living and dead bacteria was computed, with 100% being the sum of both.

### DNA Extraction for Microbial Community Analysis

Enamel-dentin slabs (T0 and T3) were glued into the lids of 1.5 ml Eppendorf tubes with epoxy resin adhesive that covered

#### FIGURE 1 | Experimental outline. (A) Dental splint with six human enamel-dentin slabs (white squares, a, top view; b, side view) fixed in the upper jaw (c, front view; d, side view). (B) Sketch of the biofilm reactor setup (from left to right): Supply bottle filled with BHI medium (orange), peristaltic pump, bubble trap, DFR 110 biofilm reactor on heating plate (34◦C), arrows indicate medium flow direction. (C) Top view sketch of the enamel-dentin slab distribution in biofilm reactor chambers. (D) Timeline and slab use (orange, live/dead staining; blue, pyrosequencing).

all sides except the standardized surface with the biofilm on it. Sterile and DNA-free glass beads and 200µl ultra-pure water were inserted, and the biofilm in the vials was shredded for 2 min. For total DNA isolation, the lysate was mixed with 380µl of MagNA Pure Bacteria Lysis Buffer (Roche Applied Science, Mannheim, Germany) together with 20µl of proteinase K solution (20 mg/ml) and incubated at 65◦C for 10 min. Proteinase K was heat-inactivated at 95◦C for another 10 min. The liquid samples were transferred to MagNA Pure Compact Sample Tubes. DNA isolation was performed on the MagNA Pure Compact instrument according to manufacturer instructions using the MagNA Pure Compact Nucleic Acid Isolation Kit I and following the bacteria purification protocol (Roche Diagnostics, Mannheim, Germany). The DNA was eluted in 50µl elution buffer and stored at −20◦C pending further processing.

#### 454-Pyrosequencing and Data Analysis

A 505 bp fragment targeting the V1-V3 region of the 16S rRNA gene was amplified using FLX 454 one way read fusion primers F27—AGA GTT TGA TCC TGG CTC AG and R534—ATT ACC GCG GCT GCT GGC (Watanabe et al., 2001; Baker et al., 2003). QPCR was used to ensure equal DNA amounts for the FLX 454 run. All samples were run on the same plate to exclude bias. Purified amplicon DNAs were quantified using the QuantiT PicoGreen kit (Invitrogen, Carlsbad, CA) and pooled for pyrosequencing.

Roche GS FLX raw sequences were denoised and qualitychecked using Acacia (Bragg et al., 2012). A minimum length of 150 bases was used with a Phred score of more than 25. No ambiguous bases and two-bases maximum edit distance in the forward primer were allowed. Acacia also assigned sequences to the according tag and trimmed the primer and barcode sequences. The Quantitative Insights Into Microbial Ecology (QIIME) pipeline version 1.9.1 was then used for downstream analysis (Caporaso et al., 2010b). Sequences were clustered into Operational Taxonomic Units (OTUs) with a 97% identity. Alignment of representative sequences was then performed with Greengenes 16S rRNA gene database using pyNAST (Caporaso et al., 2010a). FastTree was used to generate phylogenetic trees (Price et al., 2009). Taxonomies were assigned with the RDP Classifier (Wang et al., 2007). ChimeraSlayer implementation was used to perform chimera check on aligned representative sequences. Alpha-diversity estimates were then calculated using PD whole tree (Chao et al., 2010), Shannon and chao1 (Chao, 1984) metrics. Finally, beta-diversity was evaluated using Principal Coordinate Analysis (PCoA) plots based on unweighted UniFrac distance matrices (Lozupone et al., 2011). Rarefaction to the read size with the lowest number was performed to adjust samples for UniFrac analysis.

All statistical analyses were performed using SPSS version 21.0 (SPSS Inc., Chicago, IL). Shapiro Wilk's Test was used to test for normal distribution of the data. Data were presented as median, and 25th and 75th percentile. Wilcoxon signed-rank tests with Bonferroni correction for multiple comparisons were used for comparing T0 and T3. All reported values of p < 0.05 were considered statistically significant after Bonferroni correction.

Heat maps of the relative abundances were created in R version 3.3.0 using the phyloseq package and the plot\_heatmap function within (Rajaram and Oono, 2010; McMurdie and Holmes, 2013). Correspondence analysis was performed in R 3.3.1 using the Vegan Package 2.4. OTUs with zero counts at one of the points in time were removed to ensure that previously reported bias does not occur in the plots (Zuur et al., 2007). Site-specific scaling was chosen for the biplots. Log<sup>2</sup> fold change was calculated with Matlab R2016 on OTU level plotting the respective genera subsequently. Absolute abundances from the heat map data were used and 7 OTUs excluded as their average abundance in one of the points in time was zero.

### Fluorescence In situ Hybridization

For fixation of the biofilm, enamel-dentin slabs from T0–T3 were inserted into ice-cold 4% PFA solution directly after removal. Slabs were then incubated for 8 h at 4◦C. Subsequently, PFA was removed and slabs were washed two to three times with 1× PBS. Samples were stored in 1× PBS/96% ethanol (v/v), unless they had been used immediately. After that, Fluorescence In Situ Hybridization (FISH) was performed in 1.5 ml vials as described previously (Klug et al., 2011). Probes Bac303 (staining most Bacteroidaceae and Prevotellaceae, and some Porphyromonadaceae), EUB338mix (EUB338, EUB338II, EUB338III staining most bacteria), and LGC354mix (LGC354A, -B and -C staining Firmicutes) were used (Loy et al., 2007). FISH analysis was performed on the TCS-SP CLSM, as was the live/dead analysis. Filters were set at 500–535 nm for FITC, 560–612 nm for Cy3, and 656– 721 nm for Cy5. AMIRA 3D software (FEI, Europe) was used to generate 3D reconstructions of the confocal stack data.

## RESULTS

### Survival of Bacteria

An example of CLSM data from a biofilm containing yeast is given in **Figure 2**. A maximum projection of a CLSM stack is shown in **Figure 2A** with living cells in green, and dead and yeast cells both in red. The 3D reconstruction of the same stack is presented in **Figures 2B–D**. The large orange structures in the 3D reconstruction are presumably yeast cells and thus they were excluded from the analysis. **Supplementary Video 1** shows the performance of our MATLAB script cleaning the data for evaluation of the live/dead ratio.

**Figure 3** shows exemplarily one subject's representative live/dead stained 3D reconstruction of confocal stack data over all four sampling times. The figure exemplifies that, on the whole, the relation of living (green) and dead (red) bacteria remained the same at all points in time. An increase in biofilm mass was found at T3. Long chains of coccoid bacteria, probably the Streptococci found in the pyrosequencing analysis, on top of large staples of cocci dominated the stacks at T3 as shown in **Supplementary Figure 2**.

Survival curves reveal a quite stable growth of microbes inside the biofilm reactor over 48 h (**Supplementary Figure 1**). Half of the curves show a slight increase in the number of living bacteria during the first 24 h (T2), the other half shows a slight decrease. At T2 (24 h), most curves showed values close to those found at T3. After 48 h, the mean number of living bacteria (blue curve)

FIGURE 2 | Example of yeast cells embedded in the bacterial biofilm. Maximum projection of the entire confocal stack of a life/dead stained biofilm (A). Green, living bacteria; red, dead bacteria and yeast. The prominent red cells are probably yeast cells and are visualized in orange in a 3D reconstruction of the biofilm (B–D). (B) Gives the top view while (C,D) show side views of the 3D reconstruction.

eventually approximated the initial level measured at T0. The staining for living and dead cells also revealed that a substantial fraction of the natural biofilm contains dead bacteria at T0. Average values and standard deviations of bacterial survival are given in **Figure 4**.

#### Biofilm Composition

Compositional shifts during in vitro growth at T0 and T3 were revealed with 454 pyrosequencing and are shown in **Table 1**. Median values of the major phyla found at T0 and T3 were 98.67 and 87.71% for Firmicutes, 0.01 and 3.2% for Bacteroidetes, 0 and 2.06% for Proteobacteria, and finally 0.11 and 0.99% for Actinobacteria. Fusobacteria, Cyanobacteria and TM7 represented groups with a relative abundance of around 0.1%. SR1, Spirochaetes, and Thermi were found in very small numbers and in only some samples. The fraction "others" includes all sequences that could not be classified so far.

**Figure 5A** shows a heat map of the 117 OTUs common to all samples assigned to genus level with a relative abundance of more than 2%. The heat map is ordered such that T0 and T3 of each subject are plotted next to each other. Samples stay diverse in T3 including anaerobic and aerobic species.

Comparing in vivo and in vitro growth, the dominating genus found on the enamel-dentin slabs after 48 h of in vivo biofilm growth were Streptococci (with a mean number of 60.83%) and Veillonella (with a 13.38% relative abundance). The dominance of these bacteria remained quite stable even after 48 h of incubation in vitro (**Table 1**). This is also reflected in the log<sup>2</sup> fold change analysis in **Figure 5B**. Lactococci, Lactobacilli, and Staphilocci were found in reduced numbers at T3 while other common oral genera like Phorphyromonas, Actinomyces, Neisseria showed a positive log<sup>2</sup> fold change.

Sample counts analyzed on OTU level with a 97% identity ranged from 1841 to 3863. Assigning this data we found 6 phyla, 9 classes, 12 orders, 15 families, and 17 genera of oral bacteria. A statistical analysis on shifts in bacterial composition over further taxonomic levels is presented in **Supplementary Tables 1**–**4** (excluding values <0.1% relative abundance). For a better understanding, we added the next higher hierarchical level in parenthesis for unassigned "others" at lower levels. Below we will talk about a significant growth of certain bacteria based on an increase in their relative abundance. All bacterial phyla except Firmicutes showed a significant increase over 48 h of in vitro incubation. Firmicutes decreased significantly. On class level significant changes at T3 could not be found in the two dominating groups, Bacilli and Clostridia, although their absolute numbers decreased. All other groups with a relative abundance below 3.15% showed over time a relative increase that was


*Statistical analysis on phylum level. A p* < *0.05 was considered significant after Bonferroni correction.*

#*p-value Wilcoxon signed-rank test.*

##*p-value Wilcoxon signed-rank test adjusted according Bonferroni correction.*

statistically significant. Bacteroidia and Gammaproteobacteria were the two groups with the greatest increase.

No order belonging to the phylum Firmicutes showed significant changes at T3, although their numbers decreased. Orders occurring in lower numbers also showed a statistically significant increase in their relative abundance (**Supplementary Table 2**).

Looking at family levels, Lachnospiraceae and Carnobacteriaceae in the phylum Firmicutes showed a significant increase over time. All the other Firmicutes did not change significantly. Coriobacteriaceae were the only family in the phylum Actinobacteria that increased significantly. Prevotellaceae (Bacteroidales) and Pasteurellaceae (Gammaproteobacteria) also grew statistically significantly (**Supplementary Table 3**).

On genus level only Actinomyces and Rothia (Actinobacteria), Prevotella (Bacteroidetes), Granulicatella (Firmicutes), and Haemophilus (Proteobacteria) showed a significant increase (**Supplementary Table 4**). All other genera did not show a significant change over the 48 h of incubation in BHI medium.

A PCoA showed an incomplete clustering of the samples in two levels at T0 vs. T3 (**Figures 6A–C**). The largest coordinate explains 14.32% of the variation due to time, while the second and the third largest account for 9.29 and 5.77%, respectively. Approximately 70% of the variation is due to other factors. A clustering of T0 and T3 can be seen in **Figures 6A,C**. Correspondence analysis showed an even distribution of subjects at T0 and T3 (**Figures 6D–F**). No clustering of one of the points in time was found. The first three axes of the CA explain 20.4, 12.2, and 10.3% of the total inertia of the respective data. There is an even distribution of T0 and T3 samples in all dimensions shown. The respective OTUs appear near the samples.

The distribution of Eigenvalues explaining the variance and the fraction of total inertia are shown in **Figure 6G** (PCoA) and **Figure 6H** (CA), respectively.

#### Fluorescence In situ Hybridization

FISH analysis showed clear signals of all bacteria stained in green (EUB338mix with Cy3) and Firmicutes (EUB338mix with Cy3 and LGC354mix with FITC) in light blue (**Figure 7**). The majority of bacteria was stained in light blue confirming the pyrosequencing results where Bacilli and Clostridia, both Firmicutes, represented the largest groups. Signals could also be detected from probe Bac303 representing Bacteriodaceae, and some Porphyromonadaceae and Prevotellaceae (red signal).

### DISCUSSION

Modeling native oral biofilm growth is tricky, as many different taxa play a role in co-aggregation and bacterial succession. In order to extend knowledge on initial biofilm colonization, many studies have stained native oral biofilm, e.g., with FISH (Thurnheer et al., 2004; Hannig et al., 2007; Jung et al., 2010; Zijnge et al., 2010). Some of these studies used biofilm sampled directly from the oral cavity, some of them used carrier materials on which biofilm was grown. Hydroxyapatite discs were often chosen to simulate the tooth surface and to study primary colonization. Different attempts were made using in vitro and in vivo assays (Walker and Sedlacek, 2007; Guggenheim et al., 2009; Ledder et al., 2009; Rudney, 2012). Hannig et al., for example, fixed individual splints in the upper jaw with bovine enamel discs as biofilm carriers (Hannig et al., 2007). They found an initial colonization of Streptococci as early as after 3 min. These first studies were based on the analysis of the primary colonization and co-aggregation of bacteria on hydroxyapatite or bovine enamel. They explored biofilm formation and composition directly after removing the sample from the oral cavity. We have gone a step further. Our aim was to improve previous systems by, firstly, using real human enamel-dentin slabs as biofilm carriers, secondly, transferring the biofilm to the laboratory without any disturbance, and, thirdly, keeping the biofilm alive under in vitro conditions. Based on the dental splints used in Jung et al. and Al-Ahmad et al., we designed a dental splint carrying the human enamel-dentin slabs of a standardized size and grid (Al-Ahmad et al., 2007; Jung et al., 2010). Growing the biofilm directly in the human mouth on human enamel-dentin slabs leads to the formation of a native biofilm which is normally found in the supragingival area after pellicle formation (Nobbs et al., 2011; Teles et al., 2012; Jakubovics, 2015). The dental splint developed in our study enables us to insert up to six enamel-dentin slabs measuring 4 × 6 mm. As the slabs are positioned adjacent to the

average abundance in one of the points in time was equal to 0.

supragingival area, the biofilm finds similar conditions to those encountered directly on the individual's tooth. Although the slabs are sheltered from strong shear forces, saliva can bathe them and nourish the biofilm. Waste products can be washed away, as they would be under natural conditions. In our study, dental splints were carried intraorally for continuous 48 h. This way we could enrich a native primary biofilm directly in the oral cavity under native conditions ("from mouth") prior to transfer to the laboratory ("to model"). The easy accessibility of the enameldentin slabs in the dental splint allows for a quick transfer to in vitro systems without any disturbance of the biofilm caused by temperature shifts, excess oxygen or other factors.

For our experiments we used BHI medium as an alternative to real saliva, as it is similar to sulcus fluid (Standar et al.,

2010). The quality of initially used sterilized human saliva treated with 2.5 mM Dithiothreitol (Foster et al., 2004) was uncontrollable regarding workflow standardization and resulted in almost complete loss of viability in most attempts. In contrast, BHI works for both, anaerobes and aerobes, and thus was deemed an appropriate surrogate for our experiments, as we aimed to keep the model standardized.

Treating the biofilm with the LIVE/DEAD <sup>R</sup> BacLightTM Bacterial Viability Kit directly on the enamel-dentin slabs and using water immersible lenses for microscopy avoided damage to the biofilm structure. This was useful for detecting larger structures in the transferred biofilm. We observed (>100 µm) long chains of cocci at T3, similar to those reported in direct observations of oral biofilms. Living and dead cells frequently coexist. Looking at the survival of bacteria, the curves showed variable courses but, all in all, the proportion of living and dead bacteria after 48 h of incubation approximated the level observed at T0. In the biofilm reactor the bacteria found stable temperature conditions and BHI as a very rich food source, leading to perturbation at T1. Already after 24 h, live/dead proportions appeared to be almost at the level seen at T0 again. It seems as if the perturbation in biofilm growth at T1 originated from the transfer to the biofilm reactor. Finally, at 48 h, the initial proportions of living and dead bacteria were reached. We therefore conclude that the biofilm in our system stayed vital for at least 48 h in vitro. Netuschil et al. (2014) reported that several groups had analyzed the staining behavior of the LIVE/DEAD <sup>R</sup> BacLightTM Bacterial Viability Kit and other live/dead staining kits. SYTO 9 and PI work best with a previously tested mixture for each individual bacterium. These tests are not feasible when

working with natural biofilms comprised of several hundred species, because sometimes both stains penetrate the same cell. Therefore, a compromise has to be made for such samples. For proper evaluation of these ambiguously stained cells, however, the obtained images have to be analyzed carefully. In our study, we excluded orange (red + green) signals from further analyses. We are aware that this might have led to an underestimation of viable cells. As there is no software available that consistently excludes the misleading signals, a specific MATLAB script was designed for this purpose.

Prevotellacea). Panel (D) shows a maximum projection of all 3 channels. The presented biofilm was sampled at T3.

In our analysis of the biofilm composition, the phylum Firmicutes—and here Streptococci (facultative anaerobes) and Veillonella (anaerobes)—appear to be the primary colonizers forming a "base" on which other bacteria can dock (Rickard et al., 2003; Zijnge et al., 2010). Streptococci spp. and Veillonella spp. have been reported to show a strong co-occurrence and co-aggregation in native oral biofilm and to interact in in vitro tests (Egland et al., 2004; Palmer et al., 2006; Chalmers et al., 2008; Santigli et al., 2016). They also showed a similar behavior over the 48 h of incubation as demonstrated in the log<sup>2</sup> fold change analysis. This leads us to the assumption that those two genera further interact in our in vivo system. We also found that Streptococci made up the biggest bacterial group with around 60% of the population. These did not change significantly in number after 48 h of in vitro incubation. Interestingly, other genera like Actinomyces (mainly anaerobic growth), Prevotella (obligate anaerobes) and Rothia (facultative anaerobes) increased significantly. Kolenbrander et al. (2006) showed co-aggregation of Actinomyces naeslundii T14V with Streptococcus, Prevotella (obligate anaerobes), and Capnocytophaga strains. These genera play an important role in the "pre-organization" phase of the biofilm which is the period in biofilm development lasting between 18 h and up to 4 days (Jakubovics, 2015). Together with Streptococci and Veillonella they also tended to remain the predominant microorganisms although their relative abundance stagnated (Diaz et al., 2006). This increase after several days has been previously shown in vivo by Takeshita et al. (2015). The growing numbers seen in the other genera, i.e., facultative and obligate anaerobes, prove that our in vitro model using the BHI medium works without an anaerobic chamber. The ability to keep these genera alive over several generations is a good foundation for further assays. This is also supported by a heat map analysis on OTU level. The α-diversity calculated with PD-whole tree is higher across all samples at T3. Sterility was proven for the biofilm reactor system, so we can argue that the reason for higher values at T3 is a relative abundance at T0 which was too low to be detected by 454 pyrosequencing. As the biofilm sampling is discontinuous due to two different slabs used for T0 and T3, bacteria found at T3 can also derive from this.

PCoA explaining around 30% of the variance due to time shows a clustering of the points in time in two dimensions, no clustering can be found in the third dimension. To be able to better interpret this environmental data, correspondence analysis was used to model the change between T0 and T3 and the OTU distributions based on the same data as PCoA. The CA shows a clear proximity of the samples at T0 and T3 with around 40% of total inertia. OTUs appear in high abundance at both points in time reflected through the data points shown in strong vicinity to the sample points. Correspondence analysis thus supports our hypothesis that there is no difference between T0 and T3.

To get more information on cell viability and to confirm the biofilm composition found by pyrosequencing, we also performed the FISH analysis. FISH probes that only bind to viable cells prove that our biofilm is not only vital, but still able to live. Based on the strong signals gained, we conclude that the biofilm is also vital. LGCmix, staining Firmicutes, represented the main group also in FISH analysis. This is consistent with our pyrosequencing data showing Firmicutes as the largest group. Furthermore, signals were recorded from Bac303 staining most Bacteroidaceae and Prevotellaceae, and some Porphyromonadaceae. This result goes along with previous findings that detected these groups in healthy adults (Aas et al., 2005).

Our "mouth to model" system allows for native oral biofilm growth in vivo, a simple transfer of this biofilm to laboratory setups and further growth in vitro in biofilm reactors. Our setup can be easily reconstructed and settings used in miscellaneous studies. With this setup the biofilm stays alive and diverse over 48 h of in vitro incubation. This is an important outcome making our study a sound basis for a new biofilm model to be used in (phyto-) pharmacological assays or dental materials research. Further investigations and validation of the appropriate conditions for in vitro cultivation of native oral biofilms could facilitate the study of all biofilm-induced diseases.

#### AUTHOR CONTRIBUTIONS

BK: work conception and design; acquisition, analysis and interpretation of data; drafting and critical review of the manuscript; final approval of the work for publication. ES: study design; analysis and interpretation of data; draft and critical review of the manuscript; final approval of the work for publication. CW: analysis and interpretation of data; critical review of the manuscript; final approval of the work for publication. ST: analysis and interpretation of data; critical review of the manuscript; final approval of the work for publication. GW: analysis and interpretation of data; critical review of the manuscript; final approval of the work for publication. MG: analysis and interpretation of data; critical review of the manuscript; final approval of the work for publication.

#### FUNDING

We gratefully acknowledge the support by "Land Steiermark" (Human Technology Interface grant: OraSim).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01448

Supplementary Table 1 | Statistical analysis on class level. A p < 0.05 was considered significant after Bonferroni correction.

Supplementary Table 2 | Statistical analysis on order level. A p < 0.05 was considered significant after Bonferroni correction.

Supplementary Table 3 | Statistical analysis on family level. A p < 0.05 was considered significant after Bonferroni correction.

Supplementary Table 4 | Statistical analysis on genus level. A p < 0.05 was considered significant after Bonferroni correction.

Supplementary Video 1 | Analysis of the recorded biofilm and exclusion of suspected yeast cells. The movie shows an entire confocal stack of a life/dead stained biofilm and its corresponding analysis. (A) Living bacteria stained in green. (B) Dead bacteria and suspected yeast cells stained in red. The boundary of the detected foreground of (A,B) is displayed as a green line (C) and an orange line (D), respectively. Additionally, the purple line shows the boundary of the manually selected yeast. These regions were excluded from further analysis.

Supplementary Figure 1 | Evolution of the Life/Dead Ratio over time. Each individual plot displays the evolution of the fractions of living and dead bacteria over time. Each plot belongs to one subject and the running number (SXX) is given on the top left corner. Within an individual plot, each data point represents the mean over all fractions of living (top plot) and dead bacteria (bottom plot) at the particular point in time. The colored region gives the sample standard deviation.

Supplementary Figure 2 | Structure of the biofilm. Exemplary images of life/dead stained biofilms with the different observed structures. Each image is the maximum projection of the respective recorded confocal stack. Often observed structures are cocci filaments (A,C) and cocci staples (A,B).

#### REFERENCES


wound MRSA biofilms. J. Appl. Microbiol. 111, 1275–1282. doi: 10.1111/j.1365- 2672.2011.05138.x

Al-Ahmad, A., Wunder, A., Auschill, T. M., Follo, M., Braun, G., Hellwig, E., et al. (2007). The in vivo dynamics of Streptococcus spp., Actinomyces naeslundii, Fusobacterium nucleatum and Veillonella spp. in dental plaque biofilm as analysed by five-colour multiplex fluorescence in situ hybridization. J. Med. Microbiol. 56, 681–687. doi: 10.1099/jmm.0.47094-0


**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 © 2016 Klug, Santigli, Westendorf, Tangl, Wimmer and Grube. 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) or licensor 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.

# **Microbiome interplay: plants alter microbial abundance and diversity within the built environment**

*Alexander Mahnert 1, Christine Moissl-Eichinger 2, 3 and Gabriele Berg1 \**

*<sup>1</sup> Institute of Environmental Biotechnology, Graz University of Technology, Graz, Austria, <sup>2</sup> Interactive Microbiome Research, Section of Infectious Diseases and Tropical Medicine, Department of Internal Medicine, Medical University Graz, Graz, Austria, <sup>3</sup> BioTechMed Interuniversity Cooperation Centre, Graz, Austria*

The built indoor microbiome has importance for human health. Residents leave their microbial fingerprint but nothing is known about the transfer from plants. Our hypothesis that indoor plants contribute substantially to the microbial abundance and diversity in the built environment was experimentally confirmed as proof of principle by analyzing the microbiome of the spider plant *Chlorophytum comosum* in relation to their surroundings. The abundance of Archaea, Bacteria, and Eukaryota (fungi) increased on surrounding floor and wall surfaces within 6 months of plant isolation in a cleaned indoor environment, whereas the microbial abundance on plant leaves and indoor air remained stable. We observed a microbiome shift: the bacterial diversity on surfaces increased significantly but fungal diversity decreased. The majority of cells were intact at the time of samplings and thus most probably alive including diverse Archaea as yet unknown phyllosphere inhabitants. LEfSe and network analysis showed that most microbes were dispersed from plant leaves to the surrounding surfaces. This led to an increase of specific taxa including spore-forming fungi with potential allergic potential but also beneficial plant-associated bacteria, e.g., *Paenibacillus*. This study demonstrates for the first time that plants can alter the microbiome of a built environment, which supports the significance of plants and provides insights into the complex interplay of plants, microbiomes and human beings.

**Keywords: interplay of microbiomes, indoor plants, built environment, 16S gene and ITS region amplicons,** *Chlorophytum comosum***, qPCR, LEfSe analysis, network analysis**

#### **Introduction**

In recent years, deeper insight into the microbial diversity associated with plants and humans was gained using novel omics approaches; both are now recognized as meta-organisms: a functional unit of eukaryotic cells and microorganisms (Berg et al., 2014a). In contrast, the connection between microbiomes as well as the mutual exchange between them is less understood (Blaser et al., 2013). Although we live in a highly interconnected world, until the present date only a few examples of synergistic microbiomes have been discovered, which have shown that there are important relationships between single microbiomes (Berg, 2015). The rhizosphere is a well-investigated example that presents the root-soil interface influenced by the plant via root exudates as well as by the soil microbiome (Philippot et al., 2013). For instance, the rhizosphere mainly selects bacteria from soil but also contains indigenous plant-associated bacteria, e.g., bacteria derived

#### *Edited by:*

*M. Pilar Francino, FISABIO Public Health, Spain*

#### *Reviewed by:*

*Elisabeth Margaretha Bik, Stanford University School of Medicine, USA David Andrew Mills, University of California, Davis, USA*

#### *\*Correspondence:*

*Gabriele Berg, Institute of Environmental Biotechnology, Graz University of Technology, Petersgasse 12/I, 8010 Graz, Austria gabriele.berg@tugraz.at*

#### *Specialty section:*

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

*Received: 27 May 2015 Accepted: 14 August 2015 Published: 28 August 2015*

#### *Citation:*

*Mahnert A, Moissl-Eichinger C and Berg G (2015) Microbiome interplay: plants alter microbial abundance and diversity within the built environment. Front. Microbiol. 6:887. doi: 10.3389/fmicb.2015.00887* from seeds (Fürnkranz et al., 2012). While the rhizosphere is an example of particular importance for plant health, human health is for instance strongly dependent on the gut microbiome. David et al. (2014) recently provided evidence for the foodgut connection by analyzing the survival and metabolic activity of foodborne microbes from a plant-based diet after transit through the digestive system. Whereas this study highlighted the influence of plant-associated microbiota on the human gut microbiome, nothing is known about the impact of the phyllosphere-associated microbiota (Vorholt, 2012) on microbial abundance and diversity in the built environment. Indoor environments are considered to have big impact on human health (Reponen et al., 2012), since people in developed countries spend most of their lifetime indoors.

Built environments are not only habitats for humans; they also can be considered as biotopes for diverse microbes, whereas their abundance was mainly attributed to the presence of humans and their pets (Hanski et al., 2012; Kelley and Gilbert, 2013; Lax et al., 2014). Until now the significance of plants for humans and the built environment was mainly seen in psychological effects like mood and comfort behavior or VOCs (volatile organic compounds) as well as removal and improvement of indoor air (Sriprapat et al., 2014), but has never been linked to plantassociated microorganisms. However, is it possible that indoor plants function like humans as important or even valuable microbial dispersal sources? Our hypothesis that indoor plants contribute substantially to the microbial abundance and diversity in the built environment was already published as opinion (Berg et al., 2014b). Our hypothesis was based, amongst others, on the observation that hospital rooms that were window ventilated, contain plant-associated bacteria with potential beneficial traits for the eukaryotic hosts (Oberauner et al., 2013).

The objective of this study was to confirm our hypothesis by performing an experiment as a proof of principle, where we tracked the *Chlorophytum comosum* microbiome toward its surroundings inside an enclosed indoor environment. The spider plant *C. comosum* (Thunb.) Jacques is a monocotyledonous plant (Family *Asparagaceae*) and one of the most common indoor plants world-wide. Spider plants have been shown to have a positive impact on indoor air quality by efficiently reducing air pollution such as formaldehyde, toluene, and ethylbenzene (Sriprapat et al., 2014). Our results indicate that the plant associated microbiome spreads into the environment and might thus allow an interaction of human and plant associated microbiomes inside the built environment, which could be much more important than it had ever been assumed before.

### **Materials and Methods**

#### **Experimental Design**

The common indoor plant *C. comosum* was kept isolated in a pre-cleaned chamber (2.27 m3) for almost half a year. During the period of isolation the microclimate was monitored with respect to temperature and relative humidity. Samples for molecular analysis covered the indoor air (1.17 m3), plant leaves (0.16 m2), and surrounding surfaces (glass and press board walls and, floor tiles; 0.811 m2) within the chamber. The plant had been part of an office inventory before it was transferred to the clean chamber. The surfaces of the chamber and all other abiotic surfaces (e.g., plant pot) were cleaned in several steps to remove microbial and DNA remnants, to be able to identify the plant's contribution to the indoor microbiome after the incubation period. First, surfaces were cleaned with water and detergents (all-purpose cleaner, Denkmit, dm-drogerie markt GmbH + Co. KG, Karlsruhe, Germany), followed by cleaning with 70% (w/v) ethanol (Carl Roth GmbH & Co KG, Karlsruhe, Germany) and Bacillol-<sup>R</sup> plus (Bode Chemie GmbH, Hamburg, Germany) to remove most microbes. Chlorine bleach (DNA away, Molecular Bio Products, Inc. San Diego, CA, USA) and UV light (254 and 366 nm, Kurt Migge GmbH, Heidelberg, Germany) was used to fragment and remove remaining DNA in the chamber. The plant was placed on a pedestal in the chamber and watered once a week. Natural tab water was selected to sustain hydration of the plant. This procedure was preferred over a supply with sterilized water and soil to be more comparable with common house plants. Beside the sampling events, watering of the plant was the only period of time where the chamber was opened for some seconds and potentially susceptible to the surrounding laboratory environment. This potential input from the adjacent built environment was covered by a control (see below). Supply with light was guaranteed by natural sun light through glass windows and supported by an artificial light source according to the day/night cycle. Samples were taken in the following order: First samples from the surfaces of the cleaned chamber were received from floor and wall surfaces (surface\_t0). Then samples from plant leaves were sampled before the plant was transferred to the cleaned chamber to avoid any artificial spreading of microbes due to the sampling procedure itself (plant\_t0). Sampling the air (air\_t0) of the chamber with the plant inside finalized all sampling steps for time point and sample group t0. Plant growth could be observed during the time of incubation. The plant was positioned on a pedestal with a reasonable distance (radius of ∼80 cm) to the surrounding wall and floor surfaces (distance of ∼36 cm to wall and floors, ∼100 cm to the ceiling). The plant had an initial volume of about 225 cm3 and doubled its volume during the incubation period. Incubation was stopped, when the first plant leave made direct contact with the surrounding indoor surface (contact to the floor surface due to leave growth of ∼36 cm), to avoid direct transfer of phyllosphere associated microbes onto surfaces. However, throughout the incubation period, seed and flower particles were shed onto the floor surface. After the incubation period samples were taken in the following order to obtain sample group t1: First the air was sampled inside the chamber (air\_t1). Then the plant was carefully removed from the chamber and the plant leaves were sampled (plant\_t1). Finally surfaces of the empty chamber were sampled (surface\_t1).

#### **Sampling Procedure**

Indoor air samples were obtained using the SKC BioSampler-R (SKC Inc., PA, USA). All parts of the air sampler were autoclaved at 121◦C for 30 min to achieve sterility and treated with dryheat at 170◦C for 24 h to degrade DNA (Probst et al., 2013). Four air sampling replications were processed in a serial manner at a flow rate of 13 l/min to allow an entire room volume to pass through the impinger (sampling of particles from the air into PCR-grade water, Sigma-Aldrich Chemie GmbH, Stiegheim, Germany, or Carl Roth GmbH & Co KG, Karlsruhe, Germany) in about 20 min. For one replica the procedure was repeated three times (within an hour) and resulting samples (10 ml each) were pooled (30 ml total volume). For sampling plant leaves and surrounding chamber surfaces in four replications, sterile (autoclaved) and DNA-free (dry heat treatment) Alpha Wipes-<sup>R</sup> (TX1009, VWR International GmbH, Vienna, Austria) were used. Alpha Wipes-<sup>R</sup> were extracted in 100 ml PCRgrade water, vortexed and sonicated at 40 kHz for 2 min. Sample extracts of air, plant leave and surface samples were concentrated 100-fold to 1 ml using Amicon Ultra-15 centrifugal filter tubes (Ultracel-50K, Merck Millipore KGaA, Darmstadt, Germany). Negative controls, field blanks, sequencing controls for prokaryotes and eukaryotes and additional PMA treatment of a sample subset were processed in parallel with all samples. This procedure allowed a quality control for the sample equipment, used reagents, background signals of the indoor environment and to which extent sequences were obtained from actual intact microbial cells. Results presented in this study are based on only those samples, which passed these rigorous quality controls through PCR-testing of respective samples and controls.

#### **PMA Treatment and DNA Extraction**

PMA (propidium monoazide, GenIUL, S.L., Terrassa, Spain) treatment and DNA extraction of samples was applied as optimized and reported before (Moissl-Eichinger et al., 2015). PMA helps to determine the proportion of dead cells and free DNA in a sample, by masking free and non-membrane encased DNA in downstream processes such as PCR. Hence, after observing an over-proportional amount of intact cells compared to other enclosed indoor environments (Moissl-Eichinger et al., 2015) this procedure was not applied to all samples and represented an additional control for possible DNA contaminants and drawn conclusions of this study in general. Afterwards cells were mechanically lyzed in Lyzing Matrix E tubes filled with glass beads (MP Biomedicals, Heidelberg, Germany) on a FastPrep-<sup>R</sup> -24 Instrument (MP Biomedicals, Illkirch, France) at 6.5 m/s for 2x 30 s. DNA was extracted according to the XS buffer method applicable for low biomass environments (Moissl-Eichinger, 2011).

#### **Quantitative PCR (qPCR)**

For determining microbial abundance, qPCRs with bacterial (515f—927r; 10µM each); fungal (ITS1—ITS2; 10µM each); and archaeal (344aF—517uR; 5µM each) directed primers were conducted (see **Supplementary Table S1** for sequence of primers). The qPCR reaction mix for bacteria and fungi (7.04µl) contained 5µl QuantiTect SYBR-<sup>R</sup> Green PCR kit (QIAGEN GmbH, Hilden, Germany), 0.2µl BSA, 0.12µl forward and reverse primers, 0.8µl PCR grade water and 0.8µl of the extracted genomic DNA as a template. For archaea targeted qPCR, the reaction mix (10µl) comprised 3µl PCR grade water, 5µl QuantiTect SYBR-<sup>R</sup> Green PCR kit (QIAGEN GmbH, Hilden, Germany), 0.5µl forward and reverse primers (5µM each), and 1µl template DNA. A modified reaction mix (7µl) was used for e.g., plant samples with observed amplification inhibitions, which might arose from plant associated inhibitory substances. 1.06µl PCR grade water, 3.5µl KAPA Plant PCR buffer (KAPA3G Plant PCR Kit, Peqlab, VWR International GmbH, Erlangen, Germany), 0.42µl forward and reverse primers, 0.056µl of KAPA3G Plant DNApolymerase (2.5 u/µl), 0.78µl of SYBR-<sup>R</sup> Green (4x concentrate, Invitrogen™, Eugene, OR, USA), and 0.8<sup>µ</sup>l extracted DNA template.

Amplification of DNA templates and quantification of fluorescence was achieved on a Rotor-Gene™ 6000 real- time rotary analyzer (Corbett Research, Sydney, Australia) via the following PCR programs. Bacteria: 20 s at 95◦C, 15 s at 54◦C and 30 s at 72◦C for 40 cycles followed by a melt curve from 72 to 95◦C. Fungi: 40 cycles of 30 s at 94◦C, 35 s at 58◦C, 40 s at 72◦C was used, and concluded with a melt curve. For archaea, 40 cycles of 15 s at 94◦C, 30 s at 60◦C, 30 s at 72◦C was used followed by a melt curve. Ten individual qPCR runs with a mean reaction efficiency of 90% and R2 values of standard curves of 0.94 were performed separately and measured in triplicate. Occasional gene copy numbers found in negative controls were subtracted from their respective samples.

#### **Preparation of 16S rRNA Gene and ITS Region Amplicons**

Amplicons were prepared with two different barcoded primer combinations: 520f—802r specific for bacteria and ITS1f— ITS2rP regions specific for fungi (see **Supplementary Table S1** for sequence of primers). Due to scattered PCR inhibitions (e.g., plant samples) for some samples Taq&Go™ Mastermix (MP Biomedicals, Heidelberg, Germany) was substituted with KAPA3G Plant PCR Kit and nested PCR procedures were applied to add barcoded primers. 1µl template DNA was amplified on a Whatman Biometra-<sup>R</sup> Tpersonal and Tgradient thermocycler (Biometra GmbH, Göttingen, Germany) and a TECHNE TC-PLUS gradient thermocycler (Bibby Scientific Ltd, Stone, UK) with the following cycling conditions: initial denaturation 95◦C 5 min, denaturation 95◦C 50 s, annealing 60◦C 30 s (62◦C 35 s for ITS regions), extension 72◦C 60 s (40 s for ITS1-2). Four individual PCR reactions à 30µl (6<sup>µ</sup>l Taq&Go™ polymerase, 18µl PCR grade water, 1.5µl forward and reverse primer (5µM), 1µl template DNA) or 50µl (17.6µl PCR grade water, 25µl KAPA3G Plant PCR buffer, 0.4µl KAPA3G Plant DNA-polymerase (2.5 u/µl), 3µl forward and reverse primer (5µM) and 1µl template DNA) were pooled and transferred on a DNA free 96 well plate. The following pre-sequencing preparations were conducted by Eurofins Genomics GmbH, Ebersberg, Germany. According to HT DNA-QC (Agilent Technologies Sales & Services GmbH & Co.KG, Waldbronn, Germany) samples were pooled in equimolar concentrations in 2 pools (Pool\_Bac520\_Gelex and the Pool\_Fungi\_Gelex with 24 barcoded samples each). Library pools were provided with 2 different adaptor versions to increase complexity of samples. After quality control libraries were purified via gel extraction, quantified, and mixed. Sequencing was achieved on an Illumina MiSeq instrument with chemistry version 3 (2 × 300 bp). Reads were filtered and sorted according to inline barcodes and individual sequencing tags. Raw reads were deposited in the European Nucleotide Archive (www.ebi.ac.uk) under project PRJEB8807 (ERP009846).

#### **Bioinformatics and Statistics**

Filtered and sorted reads were additionally length- (200–400 bp) and quality filtered (phred q20) in QIIME (Caporaso et al., 2010). Chimeric sequences were identified and removed with usearch (Edgar, 2010) using either Greengenes gg\_13\_8 for 16S rRNA gene reads or UNITE ver6\_99\_s\_04.07.2014 for ITS region amplicons as a reference. OTUs (operational taxonomic units) were picked according to the open reference given above and any sequence not present in the respective reference was clustered denovo with usearch (according to 16S analysis tutorial in QIIME) and uclust for ITS reads (according to the Fungal ITS analysis tutorial in QIIME). After OTU picking, representative sequence alignment, taxonomy assignment, and tree construction, an OTU table with all metadata was generated. The rarefied OTU tables (520f—802r 4062 sequences; ITS1f— ITS2rP: 6839 sequences) served as the main input for following alpha and beta-diversity analysis. Core OTUs at 100% were calculated for each category (air\_t0, air\_t1, plant\_t0, plant\_t1, surface\_t0, surface\_t1) and served as input for network analysis (see Moissl-Eichinger et al., 2015 for more details) and LEfSe analysis (Segata et al., 2011) calculated with Galaxy modules provided by the Huttenhower lab. Adonis, ANOSIM, MRPP and mantel tests were calculated in QIIME (using the vegan package in R) with 999 permutations (R Core Team, 2014). One and Two Way ANOVA and *t*-tests were calculated in R (R Core Team, 2014) and MS Excel.

### **Results**

#### **Abiotic Parameters**

Abiotic parameters (temperature, moisture) were constantly monitored to assess their impact on the microbial dispersal. The average temperature was 21.9 ± 3.2◦C and reflects common conditions inside European buildings. A decrease of 13.4◦C from 30.4◦C (maximum temperature) in August to 17◦C (minimum temperature) in December was observed (**Supplementary Figure S1**). Similarly, the average relative humidity showed a decrease from 66.7% (maximum) at the beginning of September to 18.7% (minimum) at the end of November with an average of 49.2 ± 9%. The day/night cycle resulted in a daily in/decrease of the average temperature from 0.3 ± 0.1 – 0.7 ± 0.6 ◦C (minimum 0.1◦C in December to a maximum of 1.4◦C in November) and 1.8 ± 1.1 – 3.4 ± 1.4% (minimum 0.1% in December to a maximum of 7.1% in September).

#### **The Plant Increased the Microbial Abundance in its Environment**

The statistically significant increase (*t*-test *P* = 0.05) of microbial abundance on surfaces (walls and floor) was visible after 6 months of plant isolation in an indoor environment (**Figure 1** and **Table 1**). The extent of increase was variable: the highest

targeting the ITS region of fungi. Samples from surfaces are calculated per 1 m2 and samples from the air are given per 1 m3.

increase was determined for fungi (ITS region copies; up to 5 logs). For 16S rRNA gene copy numbers of Bacteria and Archaea an increase of up to 2 logs was detected. In contrast to the surrounding surfaces, the microbial abundance in the air and on plant leaves remained constant. An analysis of variance (ANOVA) showed significant variation of samples obtained from the indoor air, plant leaves, and surfaces for Archaea (*P* = <sup>7</sup>*.*9∗10−5), Bacteria (*<sup>P</sup>* <sup>=</sup> <sup>1</sup>*.*5∗10−3) and fungi (*<sup>P</sup>* <sup>=</sup> <sup>7</sup>*.*9∗10−4).


**TABLE 1 | Summary of changes in abundance**

 **and diversity of an isolated indoor plant** 

**(***Chlorophytum*

 *comosum***) according to sampled indoor spaces (air, plant leaves, floor, and wall surfaces) and**

#### **The Plant Increased the Microbial Diversity in its Environment**

Microbial diversity was assessed by analyzing amplicon pools, which comprised 1,351,533 (bacteria) and 1,903,469 (fungi) quality sequences with 56,298 (bacteria) and 185,252 (fungi) picked OTUs at a 97% similarity level (**Supplementary Tables S2**–**S4**). The diversity changed during the time of incubation (**Table 1**). Whereas the mean bacterial diversity (calculated with the Shannon-Wiener index: H') remained almost stable on plant leaves and in the air (H' 6.15– 6.94 and H' 5.39–5.31), bacterial diversity increased significantly on surrounding wall and floor surfaces (H' 4.82–6.9, *t*-test *<sup>P</sup>* <sup>=</sup> <sup>7</sup>*.*8∗10−33). On the contrary fungal diversity decreased significantly on surfaces (H' 7.14–4.98, *<sup>t</sup>*-test *<sup>P</sup>* <sup>=</sup> <sup>1</sup>*.*2∗10−17) and in the air (H' 3.87–6.53, *<sup>t</sup>*-test *<sup>P</sup>* <sup>=</sup> <sup>8</sup>*.*62∗10−19), but remained again almost stable on plant leaves (H' 7.2–6.28).

At the beta-diversity level, three distinct clusters appeared in a principal coordinate analysis based on Bray-Curtis distances of bacteria (**Figure 2A**). The first cluster was composed of samples from the air and the surrounding chamber surfaces prior to the plant isolation and the control. This cluster showed reasonable distance along PC1 axis (with a high variation of 32.6% explained) to the second cluster formed by plant leave samples prior to the isolation and the third cluster comprising samples from plant leave samples and surrounding surfaces after the isolation period. The ordination for fungi (**Figure 2B**) showed no distinct clusters of different sample groups, but similar changes in diversity along the PC1 axis (with a high variation of 22% explained). One of the most important findings was that indoor surfaces showed higher similarity to plant leaves after the isolation period. For bacteria, the calculated mean Bray-Curtis

**based on Bray-Curtis distances of rarefied OTU tables (4062 sequences for bacteria and 6839 sequences for fungi). (A)** shows results of the bacterial 16S rRNA gene amplicons. **(B)** shows results of the fungal ITS amplicons. Spheres are colored according to the indoor space and the time points as highlighted in **Figure 1**. The control in gray was a sample from the lab environment outside the chamber after the isolation period.

distances changed significantly (*t*-test *<sup>P</sup>* <sup>=</sup> <sup>1</sup>*.*7∗10−10) from 0.9 (surface\_t0 vs. plant\_t0) to 0.67 (surface\_t1 vs. plant\_t1) with a mean distance of all samples at 0.63. Likewise for fungi the calculated mean Bray-Curtis distances changed significantly (*t*test *<sup>P</sup>* <sup>=</sup> <sup>2</sup>*.*6∗10−10) from 0.75 (surface\_t0 vs. plant\_t0) to 0.37 (surface\_t1 vs. plant\_t1) with a mean distance of all samples at 0.59. However, a similar trend for samples from the indoor air although less significant (*t*-test *P* = 0*.*001, due to a high sample dispersal) could only be perceived for the fungal communities 0.86 (air\_t0 vs. surface\_t0) to 0.73 (air\_t1 vs. surface\_t1). An adonis test (55% variation explained for bacteria and 44% for fungi) and an analysis of similarities (ANOSIM, R-statistic = 0.68 for bacteria and 0.3 for fungi) showed significant (*P* = 0.001) grouping of samples by their categories at an alpha of 0.05 with a stronger grouping per individual for bacteria. A Monte-carlo permutation based analysis (MRPP) between samples obtained from air, plant leaves, and wall and floor surfaces before and after plant isolation, resulted in a delta of 0.001 and a chance corrected within-group agreement of 0.2038 for bacteria and 0.1628 for fungi. Hence, the MRPP revealed significant differences between the overall sampled communities.

#### **LEfSe Analysis Revealed Plants as a New Source for the Microbiome within the Built Environment**

The linear discriminant analysis of the effect size [LEfSe; (Segata et al., 2011)] of bacterial and fungal core OTUs revealed features that most likely explained differences between sampled indoor classes. According to this analysis 47 OTUs could be identified to be responsible for discriminating between the different sampled spaces and microbiomes (**Figure 3**). Hence, amongst other OTUs from lower taxonomic levels, *Acidovorax, Methylobacterium* (for air\_t1 samples); *Caulobacter* (for control samples); *Cellvibrio, Clostridium intestinale, Devosia, Dyadobacter, Luteimonas, Rhizobium, Sphingopyxis* (for plant\_t1 samples); *Bradyrhizobium* (for surface\_t0 samples); and *Heterobasidion* (for surface\_t1 samples) were significantly responsible to explain differences of their respective indoor space. For a deeper insight some of these OTUs are shown as abundance histograms in relation to the sampled indoor environment (**Figure 4**). This analysis showed that mainly OTUs from plant samples and surrounding floor and wall samples were significantly responsible for discriminating the different categories of indoor environments and revealed that the plant serves as a source of microbes within the built environment.

The distribution of core OTUs according to their sampled indoor spaces substantiated results obtained by the LEfSe analysis and was visualized as a core OTU network for bacteria (**Supplementary Figure S2**) and fungi (**Supplementary Figure S3**). A detailed analysis of these distribution patterns showed that most core OTUs were shared between samples from time point t1. The surrounding floor and wall surfaces were the only category where an increase from 14.1% (bacterial OTUs) and 13.5% (fungal OTUs) before plant incubation (surface\_t0) to 19.8% (bacterial OTUs) and 23.1% (fungal OTUs) after plant incubation (surface\_t1) could be determined. On the contrary OTUs detected in control samples

period.

were shared to the lowest proportion (0.8% bacteria and 6.6% fungi).

The air was dominated (*>*10,000 sequences) by sequences assigned to *Deinococcus, Bosea genosp., Delftia, Caulobacter, Methylobacterium, Volutella, Schizophyllum commune, Trametes versicolor,* and *Aspergillus ochraceus*. The same fungal genera (the last three named genera) and species could be found to high proportions on plant leaves together with the bacterial genera *Paenibacillus, Enhydrobacter,* and *Pseudomonas*. The surfaces showed a complex mixture of these genera and species. From these taxa especially *Methylobacterium* is a common resident of the plant phyllosphere, whereas *Caulobacter* for instance is mainly associated to aquatic environments but also with phosphate-solubilizing abilities and *Delftia* is an example of a well-known genus that colonizes abiotic and biotic surfaces such as the phyllosphere. As displayed on a heatmap (**Figure 5**), many taxa were increased on the surfaces after the incubation period with the plant. A *t*-test showed for instance a significant increase for sequences of *Aspergillus ochraceus* (*P* = 0.03), *Agrobacterium* (*P* = 0.03), *Planctomyces* (*P* = 0.01), on surrounding surfaces during plant incubation. *Planctomycetes* were only recently detected since they often belong to the hitherto-uncultured bacteria (Nunes da Rocha et al., 2009). *A. ochraceus* is a soilborne ascomycetous fungus capable of producing a variety of mycotoxins; however its airborne spores are one of the potential causes of asthma in children and lung diseases in humans.

### **Discussion**

In the past, humans and pets were identified as important dispersal sources for microbes into the built environment. Single persons can emit up to 10<sup>6</sup> microbes per person and per hour (Qian et al., 2012; Dunn et al., 2013). We identified an additional effect of house plants, beyond melioration of our mood and indoor air quality, for the quality and quantity of the indoor microbiome. In a proof of principle analysis, we show in this study that plants are an additional important dispersal source in the built environment.

Our study supports our hypothesis that indoor plants contribute substantially to the microbial abundance and diversity in the built environment presented in Berg et al. (2014b) in a pilot experiment. Since plants in general influence abundance and diversity of microbes, they might be important for human wellbeing inside the built environment also from the perspective of plant-associated microbiota. Bacteria and fungi are well-known plant inhabitants, but plant-associated Archaea (*Thaumarchaeota* like *Nitrososphaera* and *Euryarchaeota* like *Halobacteriacae* and *Methanobrevibacter*) have only recently been discovered in olive leaves (Müller et al., 2015). To date, the role of Archaea in the phyllosphere is completely unknown, but their constant occurrence in many common environments might indicate basic functions in many ecosystems (Oxley et al., 2010; Bates et al., 2011; Moissl-Eichinger, 2011; Probst et al., 2013). On average 61% of detected bacterial and fungal sequences were derived from intact cells or spores as revealed by PMA (propidium monoazide) treatment of a subset of samples from all indoor spaces prior to DNA extraction, which masks DNA from compromised cells (**Supplementary Table S5**). This high rate (relative values) of intact cells from all domains of life might be due to the DNA removal and degradation procedures applied to the chamber prior to plant isolation. This uncommon, and very rigorous procedure might explain a higher proportion of intact cells compared to other indoor environments with strict cleaning procedures such as cleanrooms, with only 1% intact microorganisms, compared to 45% in garment areas (Moissl-Eichinger et al., 2015). Nevertheless, a high proportion of intact cells allow active interactions of microbes in the presence of water and nutrients, which could be tackled by metabolome studies.

The general increase of the microbial population on indoor surfaces was not surprising after such a long time of isolation in an enclosed system, but we were especially interested in identifying differences in diversity as well as sources of the microbial dispersal. Microbial diversity shifted with an increase for bacteria but a decrease for fungi on surrounding wall and

**taxa, which relatively increased on different indoor spaces (indoor air, plant leaves or wall and floor surfaces) sorted according to** *P***-value**

**discovery rate (fdr) and Bonferroni corrected** *P***-values (bonf.) are shown as well).**

floor surfaces as well as plant leaves. This transition could be due to unknown plant properties, but more obvious they might be the result of the altered microclimate inside the chamber after half a year of incubation (**Supplementary Figure S1**). Hence, the decrease in relative humidity might explain the lowered diversity for fungi on surfaces over time. An increasing microbial diversity on surfaces as well as the higher similarity to plant leaves could be of importance, since microbial diversity was shown to determine the invasion by a bacterial pathogen (Van Elsas et al., 2012). Due to the fact that several microbial indoor pathogens are known to be able to cause severe health problems (Nunes da Rocha et al., 2009), a higher diversity could help to avoid settling of these pathogens. LEfSe and partly the network analysis (**Figures 3**, **4** and **Supplementary Figures S2**, **S3**) revealed the importance of phyllosphere associated microbiota for the transfer of microbes and the general increase of abundance and diversity on the surrounding wall and floor surfaces. This shows that all microenvironments share a part of the microbiome and that house plants act as a bio-resource.

Altogether, plant incubation led to an increase of beneficial plant-associated bacteria *Paenibacillus* (Rybakova et al., 2015), plant-associated *Plantomycetes* with unknown function (Nunes da Rocha et al., 2009) and the spore-producing fungi *Aspergillus ochraceus*, *Wallemia muriae* and *Penicillium* spp. with allergenic potential (Reponen et al., 2012). The plant microbiome can be altered by the application of biological control agents or stress protection agents (Yang et al., 2009; Berg et al., 2013). This opportunity can be used to develop control agents with beneficial effects to plants as well as to humans. In this context it should also be possible to reduce the proportion of spore-producing fungi, since many of them harbor an allergenic potential (Reponen et al., 2012). Bacterial and fungal biocontrol agents for certain purposes have already been developed (Berg et al., 2013), but the potential of Archaea is completely unknown. Due to the fact that none of the archaeal representatives was judged to be pathogenic so far, they may be a healthy alternative.

Although the plant was identified as major source for microorganisms in a closed cabinet, our experimental design still has several limitations, which will be discussed in detail: Firstly, the study design has some artificial components. The study setup presented here might ignore many other influences between interactions of house plants with their surrounding built environment. However, to limit potential influences and make a compromise of artificial and common environmental parameters, we decided to conduct the experiment in a closed chamber, with ordinary water supply and growing substrate. Secondly, we investigated only one house plant in one incubation system. Due to limitations to reproduce identical indoor environments we focused on one incubation system to limit divergent environmental parameters with unknown effects. As a third point, we only studied two time points. The selection of two sampling points was a compromise to guarantee a low disturbance by the invasive sampling methods. Although more sampling points would help to identify the source of microbial dispersal, we decided against this procedure since regular sampling would disturb microbial abundance and diversity and might increase the level of potential contaminations of the chamber from outside to a critical magnitude.

Additional studies with labeled microorganisms can provide further evidence for microbial dispersal from house plants. House plants are normally grown in soil, which contain a highly diverse microbiome and can influence the environment as well as the phyllosphere as shown by Rastogi et al. (2012).

Indoor plants have the potential to influence the microbiome of the built environment similar to humans and pets. Hence, aside from determining other factors like architecture, ventilation, and room maintenance etc. the microbiome of the built environment is particularly defined by its eukaryotic habitants. The embellishment of built environments with indoor plants does not have an aesthetic relevance alone, indoor plants can act as a simple but efficient way to stabilize and increase diversity of beneficial microbes in the built environment and other enclosed systems for humans in the future such as space stations or manned space missions to successfully colonize other planets.

#### **Author Contributions**

AM: study design, performed experiments, analyzed the data, wrote the manuscript; CM: reviewed the study design and manuscript; GB: study design, wrote manuscript.

#### **Acknowledgments**

We thank Henry Müller and Christian Berg (Graz) for critical discussions throughout data analysis, Timothy Mark (Graz) for

#### **References**


critically reading of the manuscript and Tobija Glawogger for support during sampling and experiments.

#### **Supplementary Material**

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2015.00887

**Supplementary Figure S1 | Microclimate recordings of 1000 measured points of temperature and humidity in the isolation chamber from August to December 2013 (x-axis).** Black line indicates temperature values in ◦C and gray line indicates relative humidity recordings in % (y-axis).

**Supplementary Figure S2 | Core OTU network of 16S rRNA gene amplicons from plant (green triangles), floor and wall surfaces (squares) and the surrounding indoor air (blue hexagons).** OTUs (circles) are spring embedded eweighted due to their abundance and distribution (shared OTUs are colored according to their sample origin). Details of network visualizations are given in Moissl-Eichinger et al. (2015).

**Supplementary Figure S3 | Core OTU network of ITS region amplicons from plant (green triangles), floor and wall surfaces (squares) and the surrounding indoor air (blue hexagons).** OTUs (circles) are spring embedded eweighted due to their abundance and distribution (shared OTUs are colored according to their sample origin). Details of network visualizations are given in Moissl-Eichinger et al. (2015).

**Supplementary Table S1 | Sequence of primers used in this study.**

**Supplementary Table S2 | Summary of obtained raw and quality filtered sequences.**

**Supplementary Table S3 | Bacterial OTUs and taxa detected in this study according to sampled indoor spaces (unassigned reads were summarized due to file size limitations).**

**Supplementary Table S4 | Fungal OTUs and taxa detected in this study according to sampled indoor spaces (unassigned reads were summarized due to file size limitations).**

**Supplementary Table S5 | Summary of controls with counts or hits in this study.**


with specific molds. *J. Allergy Clin. Immunol.* 130, 639–644.e5. doi: 10.1016/j.jaci.2012.05.030


**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 © 2015 Mahnert, Moissl-Eichinger and Berg. 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) or licensor 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.*

# Microorganisms in Confined Habitats: Microbial Monitoring and Control of Intensive Care Units, Operating Rooms, Cleanrooms and the International Space Station

#### Edited by:

Mike Taylor, University of Auckland, New Zealand

#### Reviewed by:

Marius Vital, Helmholtz Centre for Infection Research, Germany Ravindra Pal Singh, John Innes Centre, UK

#### \*Correspondence:

Christine Moissl-Eichinger christine.moissleichinger@medunigraz.at

†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: 10 May 2016 Accepted: 20 September 2016 Published: 13 October 2016

#### Citation:

Mora M, Mahnert A, Koskinen K, Pausan MR, Oberauner -Wappis L, Krause R, Perras AK, Gorkiewicz G, Berg G and Moissl -Eichinger C (2016) Microorganisms in Confined Habitats: Microbial Monitoring and Control of Intensive Care Units, Operating Rooms, Cleanrooms and the International Space Station. Front. Microbiol. 7:1573. doi: 10.3389/fmicb.2016.01573 Maximilian Mora<sup>1</sup>† , Alexander Mahnert<sup>2</sup>† , Kaisa Koskinen1,3, Manuela R. Pausan<sup>1</sup> , Lisa Oberauner-Wappis<sup>4</sup> , Robert Krause<sup>1</sup> , Alexandra K. Perras1,5, Gregor Gorkiewicz3,4 , Gabriele Berg<sup>2</sup> and Christine Moissl-Eichinger1,3 \*

<sup>1</sup> Department for Internal Medicine, Medical University of Graz, Graz, Austria, <sup>2</sup> Institute of Environmental Biotechnology, Graz University of Technology, Graz, Austria, <sup>3</sup> BioTechMed-Graz, Graz, Austria, <sup>4</sup> Department of Pathology, Medical University of Graz, Graz, Austria, <sup>5</sup> Department for Microbiology, University of Regensburg, Regensburg, Germany

Indoor environments, where people spend most of their time, are characterized by a specific microbial community, the indoor microbiome. Most indoor environments are connected to the natural environment by high ventilation, but some habitats are more confined: intensive care units, operating rooms, cleanrooms and the international space station (ISS) are extraordinary living and working areas for humans, with a limited exchange with the environment. The purposes for confinement are different: a patient has to be protected from infections (intensive care unit, operating room), product quality has to be assured (cleanrooms), or confinement is necessary due to extreme, healththreatening outer conditions, as on the ISS. The ISS represents the most secluded man-made habitat, constantly inhabited by humans since November 2000 – and, inevitably, also by microorganisms. All of these man-made confined habitats need to be microbiologically monitored and controlled, by e.g., microbial cleaning and disinfection. However, these measures apply constant selective pressures, which support microbes with resistance capacities against antibiotics or chemical and physical stresses and thus facilitate the rise of survival specialists and multi-resistant strains. In this article, we summarize the available data on the microbiome of aforementioned confined habitats. By comparing the different operating, maintenance and monitoring procedures as well as microbial communities therein, we emphasize the importance to properly understand the effects of confinement on the microbial diversity, the possible risks represented by some of these microorganisms and by the evolution of (antibiotic) resistances in such environments – and the need to reassess the current hygiene standards.

Keywords: microbiome, built environment, indoor, confined habitat, microorganisms

### INTRODUCTION

fmicb-07-01573 October 8, 2016 Time: 16:28 # 2

Nowadays, people spend most of their time indoors (up to 90% in industrialized countries; Hppe and Martinac, 1998). In particular, the process of increasing urbanization has created new types of microbiome settings that surround us in our living and work space, such as air conditioned residences and highly populated offices. The microbiome of a built environment is determined by numerous parameters, such as geographic location, type of usage, architectural design, ventilation and occupancy, but mainly by the living inhabitants (humans, animals, and plants), as the major source of microorganisms (Califf et al., 2014; Mahnert et al., 2015a; Meadow et al., 2015). For example the human body is a holobiont and thus the home of billions of microbes. Every second of our lives, we interact with microorganisms that support our life and health. This cohabitation has evolved over 1000s of years, and is characterized by a balanced interaction of three domains of life, namely the Archaea, Bacteria, and Eukaryota (Parfrey et al., 2011; Human Microbiome Project Consortium, 2012; Probst et al., 2013; Gaci et al., 2014). It was calculated that a human body can emit up to 3.7 × 10<sup>7</sup> bacterial and 7.3 × 10<sup>6</sup> fungal genome copies per hour (Qian et al., 2012).

In the study by Ruiz-Calderon et al. (2016) different housing types were analyzed with respect to the indoor microbial community, starting with jungle villages to highly urbanized living areas in Manaus. Although all of the analyzed living areas were well ventilated, the housings of higher urbanization level were characterized by a reduced influence of the outer, natural environmental microbiome whereas the portion of human-associated microorganisms was substantially increased. As a logical conclusion, more confined environments, with less or no contact to the outdoor environment, should be totally dominated by human associated microorganisms. There are many reasons that necessitate stricter confinement for living and work environments than is typical for most people. For the purposes of this review, we are interested in confined habitats as defined by human-populated environments restricted by a number of parameters. The parameters are a restriction of area and space, and restrictions of physical, chemical and biological exchange with the surrounding, natural environment. Such confined habitats include areas such as intensive care units (ICUs) and operating rooms, where patients need to be protected from infection; cleanrooms, where the quality of products needs to be assured; and the ISS, which is encapsulated due to life-threatening environmental conditions. A summary of the characteristics of the confined habitats addressed in this review is given in **Figure 1**.

All these environments require microbiological monitoring, and control, since they harbor their own, possibly adapted, microbial community, which is greatly influenced by the maintenance regime.

In this review, we detail the setting, architecture, and control measures of such environments, which influence the internal microbiome tremendously. We hypothesize that all these environments have parameters in common, which shape, in a similar way, the inhabiting microbial community – with a potential effect on humans living and/or working in these areas.

### THE MICROBIOLOGY OF INTENSIVE CARE UNITS

#### Intensive Care Units and Hospital Acquired Infections

Intensive care units are special departments in hospitals that provide intensive medical care for patients suffering from severe and life-threatening diseases or injuries. These units can be divided into several categories, including neonatal ICUs, pediatric ICUs, psychiatric ICUs, cardiac ICUs, medical ICUs, neurological ICUs, trauma ICUs, and surgical ICUs. Depending on the underlying disease, duration of stay and treatment in ICUs, patients may show higher susceptibility for hospitalacquired infections (HAIs) than healthy individuals due to an overall weak condition, immunosuppression, or disrupted physiological barriers. ICUs are considered potential reservoirs for (opportunistic) pathogenic microbial strains (Russotto et al., 2015). These microorganisms may thrive on the medical equipment, in other patients, personnel, and the surrounding environment of the hospital (Gastmeier et al., 2007). HAIs are a serious problem worldwide: in the United States, HAIs are the sixth leading cause of death, killing more people than diabetes or influenza combined (Anderson and Smith, 2005; Klevens et al., 2007), and similar results have been reported from Europe as well (Peleg and Hooper, 2010). For instance, Vincent et al. (1995) have estimated the risk for gaining a nosocomial infection in a European ICU to be 45%. In general, the risk of acquiring pathogenic infection, in hospital environments is higher than in other environments, and the course of an infection is more often fatal (Centers for Disease and Prevention, 2002, 2010; Klevens et al., 2007).

Already in the 1980's, specialists in infectious diseases detected that patients in ICUs are infected by nosocomial bacteria, as e.g., Pseudomonas aeruginosa and Acinetobacter baumannii, considerably more often than patients in other wards in the hospital (Donowitz et al., 1982). Many factors contribute to the increased infection rate in ICUs, including the underlying disease of the patient, the length of the hospitalization, frequency of contact with medical personnel, the number of colonized or infected patients in the same ward, ICU structure (single bed vs. double bed rooms), and the lack of compliance with existing infection prevention guidelines (Siegel et al., 2007). Even the season affects the incidence: in wintertime the risk of acquiring a HAI is smaller compared to other seasons (Schröder et al., 2015). Patient groups that are most often affected are the elderly, premature infants and patients suffering from immunodeficiency (Unahalekhaka, 2011); in the latter, even nonvirulent bacteria may cause serious infection and death (Poza et al., 2012).

The risk of infection is increased by invasive, clinically necessary procedures (like insertion of catheters), but also from architectural properties of the hospital environments (such as ventilation systems; Unahalekhaka, 2011) or deficient hygiene procedures. For instance, significantly higher risk for the acquisition of antibiotic resistant microorganisms was observed when newly arrived patients were placed in

rooms that were previously occupied by carriers, despite terminal cleaning of the ICU bed space (Huang et al., 2006; Russotto et al., 2015). This transfer was confirmed by another study, reporting that the infection of the previous room occupant was the most important independent risk factor for infection with Pseudomonas aeruginosa and Acinetobacter baumannii, two bacteria causing nosocomial infections (Nseir et al., 2011). The majority of the HAIs is believed to be transmitted directly from patient to patient, but increasing evidence demonstrates that also the medical personnel as well as the clinical environment (i.e., surfaces and equipment) often are a source of infection (Tringe and Hugenholtz, 2008; Caporaso et al., 2012; Passaretti et al., 2013; Salgado et al., 2013). One major vector for cross-contamination are hands of medical personnel, contributing to approximately 20–40% of nosocomial infections (Agodi et al., 2007; Weber et al., 2010). Since infected patients themselves act as a source of microorganisms, frequently touched surfaces close to the patient were heavily contaminated (Wertheim et al., 2005; Pittet et al., 2006). Specifically, Salgado et al. (2013) observed that the risk of acquiring a nosocomial infection increased significantly when the total microbial burden exceeded 500 CFU/100 cm<sup>2</sup> .

The link of invasive equipment and the emergence of nosocomial infections has clearly been shown. However, there is also evidence of non-invasive devices to cause ICU outbreaks. Especially, electrical equipment and devices that are difficult to clean (irregular shape, no cleaning regime) have been reported as a source for infection (Russotto et al., 2015).

Hospital textiles are another potential source of HAIs. These textiles are usually reusable and include uniforms, bed linen and pajamas, as well as privacy curtains and protective clothing of health care personnel. The liberation and dispersal of bioaerosols and fomites from textiles takes place during handling of soiled textiles that have been used by or have been in close contact with an infected patient. It has been shown that antibiotic resistant Staphylococcus strains can aerosolize from bed linen during routine handling of bedding and be transmitted via air (Handorean et al., 2015). However, microbial transfer from textiles can be easily prevented by proper laundry procedures (Fijan and Turk, 2012).

### The ICU Microbiome

fmicb-07-01573 October 8, 2016 Time: 16:28 # 4

Previous studies have shown that pathogenic bacteria, such as Staphylococcus aureus, various Enterococcus species, Escherichia coli, Pseudomonas aeruginosa, Klebsiella pneumonia, different Enterobacter species, Acinetobacter baumannii and Klebsiella oxytoca are, despite efficient cleaning procedures and disinfectants, commonly found on surfaces such as stethoscopes (Marinella et al., 1997), electronic thermometers (Livornese et al., 1992), and other equipment routinely used in hospitals (Myers, 1978; Schabrun et al., 2006; Safdar et al., 2012).

Bacteria living in diverse communities at ICUs include pathogenic strains, opportunistic pathogens, as well as harmless and beneficial bacteria. Bacteria found in ICU environments are typically human associated and, due to confinement and strict cleaning procedures, less diverse than indoor environments with unlimited and uncontrolled access. In addition to the above mentioned common hospital pathogens, several genera of opportunistic pathogens have been detected in hospital environments by cultivation and using next generation sequencing methods, including Actinomyces, Burkholderia, Clostridium, Flavobacterium, Neisseria, Propionibacterium, Roseomonas, Streptococcus, and Vibrio (e.g., Kim et al., 1981; Heeg et al., 1994; Triassi et al., 2006; Hewitt et al., 2013; Oberauner et al., 2013). Bacterial communities in different locations at an ICU vary in species composition and diversity. In general, objects and surfaces near patients, including textiles such as pajamas, bedlinen, pillows and mattresses, carry more human gut-, hair- and skin-associated bacteria like Staphylococcus, Propionibacteria, Corynebacteria, Lactobacillus, Micrococcus and Streptococcus, whereas floor and other sites with greater distance to the patient carry more environmental strains. In addition, the abundance of bacteria was higher if samples were taken close to the patient (Handorean et al., 2015; Hu et al., 2015). However, according to current knowledge, most of the detected bacteria are harmless or beneficial and include, for example, Bradyrhizobium, Corynebacterium, Delftia, Lactobacillus, Melissococcus, Prevotella, Paracoccus, Sandaracinobacter, and Sphingobium (Hewitt et al., 2013).

(Opportunistic) pathogenic bacteria are typically resistant to various stresses. Due to the extreme selective pressure that confinement and cleaning practices induce, microorganisms living in ICUs develop or acquire resistance mechanisms that allow them to survive in the presence of a vast range of antimicrobial agents used in cleaning and antibiotic treatment, to adapt to extremely low nutrient content, and to persist on dry surfaces for a long time (Poza et al., 2012). In particular biofilms (including multispecies biofilms (Fux et al., 2005)) can resist common cleaning protocols. Their cells, embedded in the matrix of a biofilm, are considerably more tolerant to desiccation, detergents and disinfectants than planktonic bacteria (Burmølle et al., 2006), making them a highly dangerous infection source for susceptible patients and a critical target for bacterial burden control (Kramer et al., 2006; Hu et al., 2015). The presence of these multispecies biofilms on various surfaces may contribute to the stability of harmful bacteria in ICUs. In a recent study, Hu et al. (2015) showed that these diverse biofilms can even tolerate terminal cleaning procedures of ICU facilities and harbor viable bacteria even after 1 year (Vickery et al., 2012). Biofilms have been detected in various locations in ICUs, including a box for sterile supplies, a privacy curtain, a glove box, a noticeboard, and catheters (Perez et al., 2014; Hu et al., 2015). According to Hu et al. (2015) up to 93% of studied surfaces carried bacterial biofilms. In addition, the biofilm lifestyle of microorganisms bears a high risk for horizontal gene transfer, consequent spreading of antibiotic resistance and high possibility for recurrence (Fux et al., 2005). Common examples of multidrug resistance (MDR) are methicillinresistant Staphylococcus aureus (MRSA) and vancomycinresistant enterococci (VRE) that are also typical components of the ICU microbiome. Often similar cellular mechanisms are used in virulence, antibiotic resistance and resistance to toxic compounds, such as cleaning agents (Daniels and Ramos, 2009; Beceiro et al., 2013).

### The Microbiome Of Neonatal ICUs

Neonatal intensive-care units (NICUs) are specialized in the treatment of seriously health-threatened or prematurely born infants. In general, infants acquire their microbiome from their mother's vagina (natural birth), skin (cesarean birth) and environment (including the breast milk) emphasizing the role of the NICU's microbiome for the development of a healthy microbiome (Penders et al., 2006; Dominguez-Bello et al., 2010; Brooks et al., 2014). Babies treated in NICUs are often underweight, from low birth weight (<2500 g) to extremely low birth weight (<1000 g). They have congenital abnormalities, or undergone surgery, and are therefore susceptible to nosocomial infections (Stover et al., 2001; Urrea et al., 2003; Couto et al., 2007). As in other ICUs, also NICU patients often develop life-threatening infections. Potentially pathogenic bacteria are found in various locations, such as diaper scales, drawer handles, keyboards, sink counters, and door buttons (Hewitt et al., 2013). Epidemiological studies have shown that infective bacteria can spread particularly well via air (Adler et al., 2005), infant incubators (Singh et al., 2005; Touati et al., 2009), sink drains (Bonora et al., 2004), thermometers (Van den Berg et al., 2000), as well as soap dispensers (Buffet-Bataillon et al., 2009) and toys (Naesens et al., 2009). Brooks et al. (2014) found that tubing, surfaces, incubators, and hands are the most important reservoirs and sources for colonizing the premature babies. They also detected that bacteria which later colonize infants' guts can initially be discovered in NICU environmental samples (Brooks et al., 2014). At genus level, typical bacteria on NICU surfaces include Staphylococcus, Enterococcus, Acinetobacter, Bacteroides, Burkholderia, Clostridia, Pseudomonas and Streptococcus (Hewitt et al., 2013; Brooks et al., 2014), which are all known to include opportunistic pathogens that potentially are of great risk for immunocompromised patients. However, most of the bacterial genera detected on NICU surfaces are harmless to humans. If and how these interact with patients and other bacteria is still not understood.

### Cleaning Procedures At ICUs

fmicb-07-01573 October 8, 2016 Time: 16:28 # 5

Cleaning practices at ICUs are an important part of preventing the spread of multidrug resistant organisms, such as MRSA and VRE, which are associated with HAIs, prolonged stays in hospitals, increased mortality rates and higher healthcare costs (Daxboeck et al., 2006). The cleaning procedures in ICUs are strict, though the practices may vary between hospitals. Depending on the frequency and type of use, dedicated ICU staff and additionally outsourced cleaning personnel are responsible for cleaning hospital interior fittings thoroughly daily, weekly, monthly, or yearly. As one example, the hygiene and cleaning protocol of the ICU, Department of Internal Medicine Graz, Medical University of Graz, is shortly mentioned (listed frequencies are minimum demand): E.g., floor is cleaned daily, toilets are cleaned daily (staff toilet) or twice (visitor toilet), shower heads are cleaned once a week, waste is evacuated as necessary and garbage bins are cleaned daily; windowsills, racks, sinks and showers are cleaned daily; laundry is washed daily, vacuum cleaning is done weekly, umbrella holders are cleansed monthly, and telephones and shutters yearly. Exposed surfaces with direct human contact, such as door handles and sinks are cleaned at least daily with cleaning detergents and surface disinfectants. In case of contamination of highly infectious material, including certain viruses and bacteria such as Norovirus and Clostridium difficile, a detailed procedure for hand and surface contamination is given: the hands have to be decontaminated with a specific disinfectant detergents under a specific exposure time, depending on which pathogen has caused the epidemic (Cleaning and disinfection protocol, guideline 2000.3116, 7.4.2014. ICU, Department of Internal Medicine, Medical University of Graz).

Despite precise protocols and appropriate disinfectants, statistical analyses of data from hospitals have revealed that fatal infections are increasing with more efficient cleaning practices, suggesting that current procedures are inadequate to protect the susceptible patients from serious, life-threatening infections (Arnold, 2014). Efficient cleaning practices are known to decrease, but not eradicate the multidrug resistant organisms living on hospital surfaces (Dancer, 2008). Consequently, new cleaning technologies are being developed. These new methods include for example technologies that are both microbiologically effective and safe to use, such as hydrogen peroxide vapor, and UV light decontamination for terminal cleaning, as well as ultra-microfibers associated with a copperbased biocide (Blazejewski et al., 2011). Hydrogen peroxide vapor and UV light can reduce the amount of bacterial cells by at least four orders of magnitude, leading to far smaller risks for patients to acquire any multidrug-resistant bacterial infection (Boyce, 2016 and references therein). These cleaning methods are particularly effective with uneven surfaces

and textures that are difficult to access with other methods (Blazejewski et al., 2011). Additionally, bacterial contamination and growth can be reduced by selecting antimicrobial material, such as copper, that can reduce bacterial burden and the possibility for patients to acquire HAI (Schmidt et al., 2015).

Other important factors for preventing infections in ICUs, beside strict cleaning protocols, are monitoring of microbial colonization and educational interventions of the cleaning procedures and results (Goodman et al., 2008; Carling, 2013). The Centers for Disease Control and Prevention (CDC) published guidelines for monitoring programs for health care workers to improve the environmental hygiene in hospitals, and to provide instant feedback and a possibility to improve the current procedures. These monitoring methods include direct observation of staff performance and protocol compliance, quantitative microbial detection by swab and agar slide cultures, fluorescent markers to identify the frequently touched surfaces, as well as adenosine triphosphate (ATP) bioluminescence for detecting both microbial and nonmicrobial ATP present in monitored surfaces (Guh and Carling, 2015).

### Summary And Outlook

Research has already shown that objective monitoring can significantly reduce the contamination of surfaces near patients, and can point out the weaknesses of current protocols (Goodman et al., 2008; Carling, 2013). Monitoring projects have shown that flat surfaces and textiles are easier to keep at the required cleanliness level, whereas more complex surface types, including doorknobs, handles and other irregular surfaces, including electronic equipment are more often cleaned with unsatisfactory quality (Goodman et al., 2008). Time pressure and lack of adequate instructions may also play a role when the set cleaning standards are not met (Goodman et al., 2008). For example, the 2010 CDC tool kit "Options for Evaluating Environmental Cleaning" offers specific instructions on how to implement monitoring and intervention programs (Carling, 2013). When HAIs are reduced in number via these infection controls and prevention programs, also substantial economic benefit can be achieved (Raschka et al., 2013).

Recently, a new and completely different perspective in defeating hospital pathogens has emerged: the interest has shifted from pathogenic bacteria toward the whole microbial communities thriving on different surfaces in hospitals and ICUs, and to a more microbial ecological perspective on how the microbes interact with their environment and other species (Arnold, 2014). It has been shown that a higher microbial diversity can prevent pathogenic infections (van Elsas et al., 2012; Pham and Lawley, 2014), and the idea of supporting the beneficial hospital microbiome by increasing the microbial diversity has raised great interest (Hewitt et al., 2013; Berg et al., 2015). However, the interaction between pathogenic bacteria, opportunistic strains, and harmful and beneficial microbes in ICUs, as well as in hospitals in general, are not yet understood and more research is still needed.

## OPERATING ROOMS

fmicb-07-01573 October 8, 2016 Time: 16:28 # 6

### Modern Operating Rooms Structure and Air Quality Monitoring

Operating rooms (ORs) are important hospital wards where most surgical procedures are performed. These areas are subjected to strict cleaning procedures such as sterilization, disinfection and removal of contaminants (e.g., dust and organic waste). Cleaning and maintenance schedules are implemented for each OR according to the surgical procedures performed. All ORs should be cleaned at the beginning of the day, between each surgical procedure, and at the end of the day, followed by a weekly or each second week total cleanup of the entire OR including walls, floor and ventilation system. In addition, guidelines propose the daily exposure to UV radiation (Rutala et al., 2008; Lives, 2009; Gupta et al., 2015).

Operating rooms are part of operating theater complexes and these complexes are architecturally divided into four different zones based on the level of cleanliness with the bacterial burden decreasing from the outer to the inner zones. These zones are maintained by a differential decreasing positive pressure to prevent unfiltered air flow toward the inside of the ORs (Spagnolo et al., 2013; Külpmann et al., 2016). The four zones can be divided as follows: (a) a protective area that includes the changing rooms for all the medical personnel, administrative staff rooms, pre and post-operative rooms and the sterile and non-sterile stores; (b) a clean area that connects the protective area to the aseptic zone; (c) the aseptic zone which includes the ORs; (d) and the disposal area for each OR (Harsoor and Bhaskar, 2007).

Modern ORs are equipped with HVAC (Heating, Ventilation and Air Conditioning) systems to control environmental factors, namely temperature, relative humidity and air flow. The ventilation systems (e.g., with vertical flow, horizontal flow, or exponential laminar flow) are equipped with different filters according to the surgical procedures performed. Most ORs have a conventional ventilation system with filters that have an efficiency of 80–95% in removing particles ≥5 µm (Dharan and Pittet, 2002). In ORs used for orthopedic and other implant surgeries, the air is filtered through HEPA filters. These filters have an effectiveness of 99.97% in eliminating airborne particles of 0.3 µm size and above (Dharan and Pittet, 2002; Sehulster et al., 2003; Lives, 2009; Spagnolo et al., 2013).

Monitoring the air quality is recommended for each OR and is often checked by particle count, a method derived from industrial cleanroom standards. This method has been proposed to determine both the effectiveness of the filters in the ventilation system as well as to establish the level of biological contamination (Pasquarella et al., 2000; Gupta et al., 2015).

Many studies have argued that the results of the particle count method do not correlate with the bacterial count results (Landrin et al., 2005; Scaltriti et al., 2007; Cristina et al., 2012). Only two studies have shown that there is a correlation between the number of airborne particles and the number of CFUs. The presence of particles >5 µm size indicate microbiological contamination in the aerosol (Seal and Clark, 1990; Stocks et al., 2010).

To date, there is no international standard of allowed airborne microbial contamination in ORs. Most countries have their own standards: for example, in France the microbiological limits are between 5 and 20 CFU/m<sup>3</sup> , which are lower than the limits of the United Kingdom (35 CFU/m<sup>3</sup> ) and Switzerland (25 CFU/m<sup>3</sup> ) (Landrin et al., 2005; Cristina et al., 2012). However, facing the increasing use of particle count over microbiological sampling, many countries have established their standards in accordance with the International Standards Organization (ISO) 14644 – Cleanrooms and associated controlled environments<sup>1</sup> . It is proposed that ORs should meet the requirements of a cleanroom of ISO 6 or 7 (explanations see also section on cleanrooms). In contrast, in the ORs equipped with HEPA filters, the levels of an ISO 5 class should be reached (Scaltriti et al., 2007; Chauveaux, 2015).

Active microbial monitoring has been used in most studies as the main method to determine the air cleanliness (Edmiston et al., 2005; Landrin et al., 2005; Wan et al., 2011; Cristina et al., 2012; Birgand et al., 2015b). This method uses an air sampler to collect a known volume of air which is then blown on agar plates for cultivation-based analyses (Napoli et al., 2012).

Besides this method, Friberg et al. (1999) have shown that in ORs with laminar air flow the CFU counts on sedimentation plates is a more relevant indicator of bacterial contamination, with CFU levels not exceeding 350 CFU/m<sup>2</sup> /h (Friberg et al., 1999). In addition to the particle counter and microbial monitoring, other methods (e.g., ATP test, fluorescent particle counter) have been implemented to determine the microbiological contamination of the air and surfaces in the ORs. Griffith et al. (2000) proposed the use of ATP test together with bacterial culture to identify the contaminated surfaces in ORs, while Dai et al. (2015) suggested the use of fluorescent particle counter for real-time measurements of microbes present on aerosol particles (Griffith et al., 2000; Dai et al., 2015).

### Surgical Site Infections: Factors, Sources And Prevention

In OR environments, the presence of microorganisms is closely linked to increased incidence of acquired surgical site infections (SSIs). About 14–20% of all hospital acquired infections are SSIs, leading to an increase in morbidity and mortality, along with rising costs to the healthcare system due to an extended stay in the hospital (Birgand et al., 2015a). Most of the microbes causing SSIs have an endogenous source, the patient's microflora. Occasionally, microorganisms acquired from an exogenous source, such as the ORs environment or health care personnel, can be the cause of the development of SSIs (Mangram et al., 1999; Spagnolo et al., 2013).

The factors that may lead to SSIs development are multifarious and can be divided into 3 main categories: (i) patient-related characteristics (e.g., age, obesity, diabetes mellitus and other diseases); (ii) characteristics of surgical procedures (e.g., duration of the operation, type of procedure, surgeon skills, hypothermia control, antibiotic therapy, surgical personnel behavior and

<sup>1</sup>https://www.iso.org/obp/ui/#iso:std:iso:14644:-1:ed-2:v1:en

equipment) and (iii) the OR environment (Mangram et al., 1999; Cristina et al., 2012; Spagnolo et al., 2013).

In most studies, the relation between these factors and the development of SSIs has been explored mainly by determining the number of particles in the OR under different conditions. The number of airborne particles varies during a surgical procedure being higher at the beginning due to patient installation and surgical bed preparation, and an increased movement of the medical personnel (Knobben et al., 2006).

Additionally, the surgical personnel and patients release skin particles (especially when the skin is dry), respiratory aerosols, dust particles and textile fibers containing viable microorganisms in the OR environment, therefore increasing the overall count of airborne particles (Dineen and Drusin, 1973; Mangram et al., 1999). Moreover, Cristina et al. (2012) have shown that the use of certain instruments (e.g., ultrasonic scalpel, laser tissue coagulation), which produce surgical smoke, increases the number of particulates in the OR air during surgical procedures, but the increasing number of particulates was not correlated with the microbial load.

Besides the presence of surgical personnel, their behavior can also lead to an increased number of microbiological particles. Several studies have shown that the number of persons present during a surgical procedure influences the number of airborne particles to a big extent, their movement leads to resuspension of any dust particle settled and the door opening rates cause an increase in the number of bacteria that can enter the ORs (Scaltriti et al., 2007; Lynch et al., 2009; Wan et al., 2011). To lower the particles shed by the health care personnel and to decrease the incidence of SSIs, different guidelines suggest the use of alcohol-based hand rubs, double gloves, face masks, hoods for covering the hair as well as the use of disposable impermeable garments made of non-woven particles during surgical procedures (Sehulster et al., 2003; Howard and Hanssen, 2007; Humphreys, 2009; Lives, 2009; Salassa and Swiontkowski, 2014). In some studies, the incidence of SSIs increased when the health care personnel wore the suits and shoes or used mobile devices both in and out of the ORs (Amirfeyz et al., 2007; Hee et al., 2014; Venkatesan et al., 2015).

Up to 30% of all SSIs are known to be caused by Staphylococcus aureus, especially the methicillin-resistant strains (Anderson et al., 2007). S. aureus is one of the most commonly isolated microorganisms from the ORs environment and a typical skinassociated microbe, indicating that ORs are dominated by human associated microbiota (Shin et al., 2015).

In two different studies the number of S. aureus has been investigated in different zones of the ORs. The number was increased in the critical zone (in close proximity of the patient) in comparison with the intermediate and peripheral zone (Edmiston et al., 2005; Genet et al., 2011).

Besides Staphylococcus ssp., other microorganisms have been isolated from ORs such as: Enterobacter spp., Micrococcus spp., Acinetobacter spp., Brevibacterium spp., Pseudomonas spp., Klebsiella spp., Bacillus spp., and Escherichia coli (Edmiston et al., 2005; Wan et al., 2011; Al Laham, 2012; Venkatesan et al., 2015; Verde et al., 2015).

Commonly, the microbiota associated with SSIs are investigated by culture-dependent methods and include well known opportunistic pathogens (e.g., S. aureus, Enterococcuss spp., Pseudomonas spp., and Escherichia coli). However, a study performed by Wolcott et al. (2009) shows that the vast majority of the microorganisms linked to SSIs is unidentifiable using standard culture methods (Wolcott et al., 2009) and consists mostly of anaerobes (the majority belonging to the genus Bacteroides).

#### The Microbiome Of Operating Rooms

Knowing that only a small fraction (around 1%) of the microbial diversity can be cultured and described (Amann et al., 1995), the usage of molecular methods arises as a prerequisite not only for identifying the microorganisms present in the ORs environment, but also for uncovering the mechanisms of their dispersal and exploring the sources of microbiological contamination.

To date, only one study has explored the entire microbiome of an OR by using molecular techniques (Shin et al., 2015). Shin et al. (2015) performed next generation sequencing of the microbial communities present in three OR environments (found in two different hospitals), and proved that the OR dust contained a microbial community similar to the one found on human skin (dominated by Staphylococcus and Corynebacterium). Moreover, Staphylococcus strains have been isolated from the dust present on ORs mobile surgery lamps, pointing out a high infection risk associated with the formation of microbial plumes. Overall, the study showed that the microbial communities present in all three ORs were similar, and that the bacteria present belonged to the phyla Proteobacteria, Firmicutes, Bacteroidetes, Actinobacteria, and Cyanobacteria (Shin et al., 2015).

More studies on the microbiome of the OR environment are needed to identify the main sources of microbial contamination, to understand how these microbes thrive in these controlled environments and how they are transmitted from humans to surfaces and vice versa. This would help to optimize stringent maintenance and cleaning procedures and to lower the microbial burden. Furthermore, health care personnel should be instructed on how to perform safer surgeries and how to minimize the microbial shedding during surgical procedures. The recommendations of WHO and CDC guidelines (Sehulster et al., 2003; Lives, 2009) should be applied in each OR to prevent SSIs and avoid unwanted expanse for both the patient and health care facilities.

### CLEANROOMS

#### Cleanrooms: Definition, Architectures and Classes

Cleanrooms are facilities used for ensuring quality and safety of many production processes. They are either mainly particulatecontrolled (e.g., microelectronics, semi-conductor industry), or additionally biocontamination-controlled in case of food

technology, pharmaceutical industry, medical processes (e.g., biosafety labs), aeronautics and many other application areas (Whyte, 2010).

The idea to use a biocontamination-controlled, clean environment to increase hygiene standards was first implemented by the two physicians Semmelweis and Lister in the 19th century. They realized the presence of an "invisible threat," which we nowadays have identified as the presence of (opportunistic) pathogenic microorganisms or viruses. By their developed countermeasures they were able to significantly decrease mortality rates in hospitals (Semmelweis, 1988). However, it was Willis Whitfield who created the basis of the modern cleanroom in 1960 and solved the problem of contaminating particles and unpredictable airflows by the application of a constant highly filtered air flow to flush out air impurities (Whitfield, 1964). A "clean" production process results in a product, which is free of contaminants of concern. Such contaminants can be microorganisms themselves and their remnants, biomolecules in general, as well as any (inorganic) particulate matter that could affect the production process and the quality of the end product. Nowadays any outdoor air entering the cleanroom is filtered and air inside the facility is constantly recirculated through HEPA (high-efficiency particulate air) and/or ULPA (ultra-low particulate air) filters to prevent contaminants to enter the cleanroom or settle on its surfaces. In addition, most cleanrooms are operated at higher pressures than their outside environment to prevent inadvertent airflows into cleaner areas<sup>2</sup> .

The installation of a clean production line requires proper planning prior to the operation itself, including consideration of specific requirements of the product (Whyte, 2001). Specific decisions have to be taken with respect to operation (i.e., exchange of materials (products) and personnel), maintenance and monitoring (i.e., measurements of air conditions, particles, flow dynamics, acoustics, electrostatics, electromagnetics, contaminating sources, risk and hazard assessments, concepts of air flow facilities, laminar flow cabinets, filter fan units), calculations of energy and media consumptions, as well as hygiene protocols and evaluations (i.e., disinfection, decontamination).

A cleanroom class is defined by its amount of particles of a certain size according to the ISO classification criteria (see also above). Hence, a cleanroom of ISO Class 6 is for instance allowed to contain 10<sup>6</sup> particles equal to and larger than 0.1 µm in size per m<sup>3</sup> of air. This number is then decreasing by 1 log per ISO category resulting in 10<sup>5</sup> for ISO 5, 10<sup>4</sup> (ISO 4), 10<sup>3</sup> (ISO 3), 10<sup>2</sup> (ISO 2), and 10 particles for ISO 1, which represents the cleanest level. In case even higher cleanliness is required, so-called insulators can be installed inside a cleanroom environment. Cleanrooms of ISO classification 7–8 represent the most common and appropriate levels of cleanliness for many different production lines. Here, classification is based on 0.5 µm- sized and larger particles with limits at 3.5 × 10<sup>5</sup> for ISO 7 and 3.5 × 10<sup>6</sup> for ISO 8 per m<sup>3</sup> air, whereas ISO Class 9 (3.5 × 10<sup>7</sup> particles) corresponds already to the particle concentration observed in uncontrolled areas. Besides the presence of particles, cleanrooms are controlled with respect to temperature and humidity (HVAC systems; heating, ventilating and air conditioning), the kind and quality of gaseous substances, the light source, electrostatics and electromagnetics (Whyte, 1999; Hortig, 2002).

#### Technologies for a Clean Production

Cleanrooms are often arranged in a sequential manner to guarantee desired conditions on each level. For this purpose, cleaner areas are only accessible after passing other cleanrooms of higher ISO classes in decreasing manner. Passages between different ISO classes and into cleanrooms are often sealed by airlocks or sluice systems, which sometimes include additional air showers and tacky mats. These systems intend to remove dust, soil, skin flakes and many other contaminating particles associated with a person or item. Work processes, as well as people behavior and interaction with respective products are strictly predefined to avoid needless spreading of particles. Hence, people in general are advised to perform their duties with slow body movements inside a cleanroom environment. In addition, the staff is equipped with special cleanroom garment that has to be donned in a specific area in a pre-defined order and often includes an overall, pants, bonnet, mustache cover, glasses, gloves, shoe covers, boots, and hoods. Previous studies have shown that dispersion rates of microbe carrying particles (MCPs; ≥0.5 µm) were substantially reduced from 2.1 × 10<sup>6</sup> to 1 × 10<sup>6</sup> per minute, when staff wore cleanroom garment compared to normal indoor clothing (Whyte and Hejab, 2007), emphasizing the effectiveness of such control measures.

Since cleanrooms can harbor entire production lines, these rooms are modular and scalable up to enormous sizes. Depending on the mode of use, cleanrooms can be equipped with diverse machines and furniture. Regardless of its special requirements, installed devices have in common that they should generate minimal air contaminations and are easy to clean. Hence, materials from natural fibers are often excluded from devices used in cleanrooms (Whyte, 2010).

Microbial decontamination actions are performed regularly but without leaving any residues behind. Standard cleaning reagents include alcohols (e.g., 70% (v/v) isopropanol), hydrogen peroxide (e.g., Klercide-CR) and alkaline cleaning reagents (e.g., Kleenol 30 or Jaminal Plus), and could be supplemented with, e.g., UV light, γ – irradiation and vaporphase H2O<sup>2</sup> treatments. Cleaning schedules can be rather elaborative including extensive repetitions of vacuuming and mopping as well as other cleaning protocols. As a result, microbial abundance is often intensively reduced compared to uncontrolled adjoining facilities. However, harsh environmental conditions and selective pressures in the cleanrooms also result in a microbial shift toward survival specialists like bacterial spore formers or archaea (Mahnert et al., 2015b).

<sup>2</sup>http://www.thomasnet.com/articles/automation-electronics/Cleanroom-Air-Flow-Principles

### Microbial Monitoring in Cleanrooms

Microbial monitoring in biocontamination-controlled cleanrooms is often executed according to standard, cultivation dependent approaches based on the usage of contact plates (nutrient agar plates), witness plates (if specific surfaces are too sensitive to be sampled) or air sampling directly onto nutrient agar plates. Besides pharmaceutical cleanrooms, also industrial cleanrooms are sometimes required to operate under biocontamination control. Examples are spacecraft assembly cleanrooms that house mission vehicles, intended to land on extraterrestrial areas of elevated risk for contamination with Earth-borne microbes. Such missions are subject to strict planetary protection regulations (Kminek and Rummel, 2015).

First studies that examined the microbial contamination of such industrial cleanrooms were conducted in the 1960s (Nicholson et al., 2009), especially in preparation for the Viking mission to planet Mars (Puleo et al., 1977), starting with the microbial characterization of laminar flow cleanrooms (Powers, 1965). A first report on a comprehensive analysis of a horizontal laminar flow, three conventional industrial cleanrooms, and three open factory areas for the presence of microbial contaminants using witness plates was published by Favero et al. (1966). It was found that the number of CFUs was reduced along with the reduction of particles in samples from the air and surfaces and reached a plateau after several weeks of exposure. Microbial contaminations (mainly vegetative microorganisms of human origin like Staphylococcus, Micrococcus, Corynebacterium, Brevibacterium) could be clearly associated with the density and activity of personnel in the cleanroom (Favero et al., 1966). In the 60's, general microbial levels on flat surfaces were evaluated using Rodac (Replicate Organism Detection and Counting) plates. These plates contained Trypticase Soy Agar (TSA) and were, after sampling, incubated at 32◦C for 43 h (Vesley et al., 1966). Similar procedures are still used today. Later on, industrial cleanrooms were brought into a broader perspective after comparing their microbial contamination type and levels with those found in hospital ORs. The hospital environment harbored at least 1 log higher microbial abundances (based on colony forming units) than the investigated cleanrooms (Favero et al., 1968).

In the case of bioaerosol characteristics, Li and Hou (2003) observed only weak relationships among different cleanroom class levels in hospitals and air particle concentrations. The index of microbial air contamination (IMA) was proposed as a reliable tool for monitoring surface contamination by settling of microbes from the air and was tested in environments like hospitals, food industries, art galleries, aboard the MIR space station and in open air (Pasquarella et al., 2000).

Several authors discussed the effectivity of microbiological methods and analytical tools to assess the risk of typical microbial contaminants, such as Staphylococcus, Microbacterium and Bacillus (Wu and Liu, 2007) during pharmaceutical production (Whyte and Eaton, 2004a,b) or in aseptic processing cleanrooms (Hussong and Madsen, 2004). Thomas et al. concluded that the aseptic techniques applied by the personnel were more critical in avoiding contamination, than the general level of cleanliness of the environment (e.g., a cleanroom) for compounding drugs (Thomas et al., 2005).

Nevertheless, besides modeling the spreading of contaminants, risk assessments, improving sampling strategies from air and surfaces in various cleanroom settings, most studies that tried to expand applied methods beyond routine microbial monitoring were conducted in spacecraft assembly cleanroom settings due to planetary protection requirements (Nicholson et al., 2009; Kminek and Rummel, 2015). For planetary protection purposes, the profound knowledge and understanding of the cleanroom and spacecraft associated microorganisms is an important prerequisite for mission success. Besides standard assays based on cultivation of aerobic mesophilic and heat-shock resistant microorganisms, more sophisticated methods have been established. These included for instance the cultivation of microbial contaminants on anoxic TSA, resulting in a collection of more than 100 strains of facultative (Cellulomonas, Paenibacillus, Staphylococcus, Arsenicicoccus, Dermabacter, Pseudomonas, Stenotrophomonas, Corynebacterium, Enterococcus) and obligate anaerobes like Clostridium and Propionibacterium (Stieglmeier et al., 2009; Probst et al., 2010). Isolated bacteria from several spacecraft assembly cleanrooms were extensively tested for their resistance against numerous environmental stresses like desiccation, UV-C irradiation, γ-radiation, 5% (v/v) hydrogen peroxide, temperature extremes from 4 to 65◦C up to a heat shock of 80◦C, pH 3 and 11, and hypersalinity of 25% NaCl (w/v), in order to understand their potential capacity to survive space flight or under extraterrestrial conditions. Besides extremotolerant Alphaproteobacteria, Betaproteobacteria, Gammaproteobacteria (Acinetobacter radioresistens), Actinobacteria and fungi (Aureobasidium), highly tolerant spore forming isolates were found, including numerous bacilli, Geobacillus (thermophilic), Paenibacillus (obligate anaerobes), and other species that revealed halotolerant and alkalo-tolerant characteristics (La Duc et al., 2003a, 2007).

The application of diverse cultivation strategies and regular monitoring and isolation of microbes from spacecraft assembly cleanrooms resulted in a rich culture collection of extremotolerant microorganisms from confined built environments that is now open to the scientific community at the German Collection of Microorganisms and Cell Cultures DSMZ (Moissl-Eichinger et al., 2012, 2013) or through the U.S. Department of Agriculture's Agricultural Research Service Culture Collection (Venkateswaran et al., 2014b).

#### Targeting Microbial Communities Of Cleanrooms With Molecular Cultivation-Independent Technologies

However, beside cultivation based methods, several studies conducted in spacecraft assembly cleanrooms included also (molecular) cultivation independent assays to target microbial diversity and abundance in NASA (National Aeronautics and Space Administration) and ESA (European Space Agency)

affiliated spacecraft assembly cleanrooms. La Duc and coworkers used molecular methods in 2003 in addition to culture-based methods to characterize microbial diversity of a cleanroom encapsulation facility and the collocated Mars Odyssey spacecraft. Predominant species in clone libraries included Variovorax, Ralstonia and Aquaspirillum. The application of various biomarkers such as ATP, LPS (lipopolysaccharides), and DNA to assess contamination of spacecraft and associated environments were reviewed by La Duc et al. (2004) including even samples from the International Space Station (ISS). In 2009, DNA microarrays (PhyloChip) were added and compared in-depth to standard cloning methods in a study covering cleanrooms before and after spacecraft assembly at Lockheed Martin Aeronautics Multiple Testing Facility (LMA-MTF), Kennedy Space Center Payload Hazard and Servicing Facility (KSC-PHSF), and the Jet Propulsion Laboratory Spacecraft Assembly Facility (JPL-SAF; La Duc et al., 2009). Three geographically distinct spacecraft-associated cleanrooms (Jet Propulsion Laboratory, Kennedy Space Flight Center, Johnson Space Center), including air samples, were analyzed in another study to determine if microbial populations are influenced by the surrounding environment or cleanroom maintenance. Only a small subset of microorganisms (e.g., Acinetobacter, Deinococcus, Methylobacterium, Sphingomonas, Staphylococcus, and Streptococcus) was common to all locations, whereas samples from Johnson Space Center featured the greatest diversity of bacteria, Kennedy Space Flight Center samples were characterized by a high presence of Proteobacteria and areas in the Jet Propulsion Laboratory assembly facility harbored mainly Firmicutes. The air of these spacecraft assembly facilities contained for instance Massilia timonae, Agrobacterium tumefaciens and Agrobacterium sanguineum, Janthinobacterium lividum, Wautersia metallidurans, Acidovorax temperans, Deinococcus geothermalis, Delftia acidovorans, Gemmata obscuriglobus, and Methylobacterium fujisawaense (Moissl et al., 2007). In addition to NASA operated spacecraft assembly cleanrooms, their European counterparts used by ESA were investigated for their microbial abundance and diversity as well (Stieglmeier et al., 2012). However, not only Bacteria could be associated to human-controlled environments but also signatures of Archaea (Thaumarchaeota, closely related to Nitrososphaera gargensis; and Euryarchaeota like halophilc and alkaliphilic Halalkalicoccus, and the methanogen Methanosarcina) were detected by molecular methods and could be visualized by FISH (fluorescence in situ hybridization; Moissl et al., 2008; Moissl-Eichinger, 2011).

Similarly like Bacteria, Archaea seem to be transferred by humans into cleanroom environments (Probst et al., 2013). Although they were found to be less (3 logs) abundant than bacteria (2.2 × 10<sup>4</sup> archaeal cells per m<sup>2</sup> cleanroom surface determined via quantitative PCR), they seem to be a constant microbial contaminant. Recently, an shotgun metagenomic approach using multiple displacement amplification (MDA) completed the picture of microbial life in a cleanroom by the detection of Eukaryotes (Acanthamoeba and fungi, e.g., Leotiomyceta, Exophiala, Mycosphaerella) and diverse viruses (Weinmaier et al., 2015).

New molecular methods like next generation sequencing nowadays allow not only a much better assessment of the total microbiome inside confined habitats like cleanrooms, but can additionally be enriched by different assays to target potential viable microbial communities. For instance the application of propidium monoazide (PMA), a chemical compound that masks DNA of dead cells from further downstream molecular analysis, revealed a remarkable proportion of dead cells (up to 99%) compared to other uncontrolled built environments (Vaishampayan et al., 2013; Mahnert et al., 2015b). The viable portion of the cleanroom environment included bacterial spore formers, such as Ammoniphilus, Bacillus, Brevibacillus, Clostridium, Cohnella, Desulfosporosinus, Geobacillus, Paenibacillus, Planifilum, Sporosarcina, Terribacillus, Thermoactinomyces, Virgibacillus) and Archaea (Haloferax and Candidatus Nitrososphaera; Vaishampayan et al., 2013; Mahnert et al., 2015b).

Moreover, viability assays using PMA were shown to increase the traceability of low abundant taxa of the rare viable biosphere (Mahnert et al., 2015b) and help to assess the entire complexity of microbiomes in confined environments which are dominated by DNA signatures of dead cells (Weinmaier et al., 2015). Hence, the importance to include differentiated methods targeting the total microbiome and that of viable or intact cells is of particular relevance in microbially controlled low biomass environments, to allow a less biased picture of the microbial DNA-based inventory.

The investigation of a whole cleanroom facility including adjoining facilities besides actual controlled cleanrooms highlighted the critical role of the gowning area. These areas are located in front of restricted clean zones, and were identified as the major location and source of microbial contaminant dispersal into cleanrooms (Moissl-Eichinger et al., 2015). Moreover, the authors of this study applied a broad spectrum of methods and compared standard cultivation techniques (TSA, R2A), adapted cultivation protocols for anaerobes (anoxic TSA), alkaliphiles (R2A at pH 10), halophiles, oligotrophes (RAVAN agar), methanogens (Methanosarcina medium) and various (molecular) cultivation-independent methods including 16S rRNA gene cloning, micro-array technology (PhyloChip) and next generation sequencing (454-pyrosequencing). Interestingly, against expectations, high throughput next generation sequencing technologies could not cover all cultivated microbes (Moissl-Eichinger et al., 2015). However, due to targeting 16S rRNA genes, this study missed the entire microbial complexity as accessible through broader or even untargeted approaches (Vaishampayan et al., 2013; Mahnert et al., 2015b). Hence all methods, even state-of-the-art, have their individual advantages, disadvantages and limitations. However, in combination they have the potential to lead to a more complete picture of microbes inside the extreme environment of the cleanroom (Moissl-Eichinger et al., 2015).

#### Conclusion

In conclusion, from a microbial perspective, a cleanroom is an extreme environment, where strict maintenance and overall lack of nutrients complicate microbial growth. The

human body serves as a continuous source of microbial contaminants, although also environmental sources (such as soil, dust particles and aerosol droplets) represent another common source of cleanroom microbes. Once transferred to the cleanroom environment, microbes adapt their metabolism (Weinmaier et al., 2015) to withstand harsh conditions, responding to starvation, by reduction of overall metabolic activity (dormancy) and spore formation. Hence, cleanroom maintenance selects especially for microbial adaptation and survival specialists – and thus enriches microbes posing a higher risk for planetary protection. For those purposes, cleanroom maintenance and the design of its infrastructure should be reconsidered and the necessity as well as impracticality of overall sterility in a cleanroom should be critically discussed in the future.

#### ISS AND HUMAN LONGTERM SPACE TRAVEL (MARS AND BEYOND)

#### The International Space Station as a habitat

Another confined man-made habitat exists about 400 km above ground: The ISS, one of the biggest and most complex international scientific projects in history, is circling our planet in low Earth orbit. As joint venture of the five space agencies of USA (NASA), Europe (ESA), Russia (Roscosmos; Russian Federal Space Agency), Canada (CSA; Canadian Space Agency), and Japan (JAXA; Japanese Aerospace Exploration Agency), the ISS is organized in modules. The first module, namely the Russian Zarya module, was launched in 1998 and since 30th October 2000, the ISS has been constantly inhabited by humans. While the ISS kept growing by the addition of new modules over the years, also the crew size increased from initially three crew members to six international astronauts and cosmonauts wo are now routinely inhabiting the ISS. Naturally, the presence of humans also imposes the presence of their associated microorganisms in this confined habitat. Besides the arrival of new crew members roughly every 6 months and about one cargo transporter per month, delivering supplies and scientific equipment for experiments, the ISS is cut off from any other biological environment. Therefore, the ISS composes the most confined man-made and inhabited environment to date. In addition to its confinement, the ISS represents a very unusual microbial biotope. Higher radiation levels than on Earth, low nutrient levels due to reduced introduction of new material, constant temperature (∼22◦C), stable humidity (∼60%) and microgravity characterize the ISS habitat and make it a unique and extreme-situated indoor environment (Coil et al., 2016).

### Microbial Safety Measures And Risk Factors

The microbiology on the ISS has been under surveillance since its first inhabitation. Standardized monitoring of surface and air samples onboard the ISS as well as more detailed postflight investigations thereof have been conducted (Pierson, 2001; Castro et al., 2004; Alekhova et al., 2005, 2016; Novikova et al., 2006; Vesper et al., 2008; Satoh et al., 2011; Venkateswaran et al., 2014c; Checinska et al., 2015; Ichijo et al., 2016; Yamaguchi et al., 2016). Moreover, cleanliness of the ISS water supplies has been investigated (La Duc et al., 2003b; Bruce et al., 2005). The greater part of the first microbial investigations were mainly based on cultivation of bacteria and fungi on commercial high-nutrient media and under moderate conditions (Castro et al., 2004; Novikova et al., 2006; Van Houdt et al., 2012). Since Roscosmos could observe serious problems due to microbial contaminations during operation of the space station Mir, all involved space agencies agreed on preventive measures to protect spacecraft, cargo, and crew from harmful microorganisms (e.g., Novikova, 2004; Ott et al., 2014).

For example, the air regeneration system is equipped with HEPA or equivalent filters (POTOK 150MK in Russian modules) to remove airborne microorganisms and particles ≥0.3 µm. The acceptability limits for airborne bacteria and fungi were set to 10,000 and 100 CFUs/m<sup>3</sup> of air, respectively. For surfaces the respective limits were defined with 10,000 CFUs/100 cm<sup>2</sup> and 100 CFUs/100 cm<sup>2</sup> . The microbial limits for the ISS water supplies differs between the US and the Russian segments: US water must be free of coliforms, with a total heterotrophic content of less than 100 CFUs/100 mL, while the Russians allow heterotrophic bacteria up to 10,000 CFUs/100 mL (Pierson, 2001; Van Houdt et al., 2012).

In order to avoid higher levels of microbial contamination, a rigorous housekeeping program is in place that includes weekly cleaning, biweekly disinfection and standard monitoring of ISS air and surfaces for viable bacterial and fungal contaminants every 90 days. The used disinfection agents are either based on a quaternary ammonium compound which is supplied by the US or on the combination of a quaternary ammonium compound with hydrogen peroxide, which is supplied by the Russians (Directorate, 2000; Pierson, 2001; Castro et al., 2004; Novikova et al., 2006; Duane et al., 2011; Van Houdt et al., 2012). Monitoring of the microbial community onboard the ISS is highly important to evaluate material integrity of the spacecraft and to assess risk factors to the health of crew members. It is known that the human immune system is compromised under space conditions. For example, there is a significant decrease of lymphocytes and also the activity of innate and adaptive immune response is reduced compared to terrestrial controls (Sonnenfeld and Shearer, 2002; Aponte et al., 2006). Additionally, it has been shown that the virulence of most microorganisms is affected by microgravity. For some species virulence is enhanced in space conditions, such as Salmonella typhimurium (Wilson et al., 2007) and for some other species virulence is reduced, such as Listeria monocytogenes or Enterococcus faecalis (Hammond et al., 2013). It is also debated that the efficacy of antibiotics and other medications decreases under space conditions (Taylor, 2015).

Even the integrity of the spacecraft itself can be compromised by microorganisms. So-called technophilic microorganisms, in particular fungi, are able to corrode alloys and polymers used in spacecraft assembly (Alekhova et al., 2005). These technophilic microorganisms caused major problems on the former Russian space station Mir (Novikova et al., 2001; Novikova, 2004).

### The International Space Station Microbiome And Its Origin

fmicb-07-01573 October 8, 2016 Time: 16:28 # 12

The main fungal genera detected onboard the ISS by cultivation were Aspergillus and Penicillium (Alekhova et al., 2005; Novikova et al., 2006; Venkateswaran et al., 2014c). These fungi were also found in higher abundance using different molecular approaches; however, Satoh et al., 2011 did not find any Penicillium in the Japanese Kibo module 1 year after its installation, but detected a predominance of skin-associated Malassezia (Satoh et al., 2011).

The main bacterial phyla detected onboard the ISS in air and on surfaces, by either cultivation or molecular methods, were Firmicutes and Actinobacteria. In cultivationbased assays, Bacillus and Staphylococcus species were the most detected Firmicutes, whereas signatures of Staphylococcus utterly dominate the Firmicutes-affiliated signatures detected by molecular methods. The most probable reason for this observed discrepancy might be the disability of standard DNA isolation protocols to open spores adequately (Venkateswaran et al., 2014c).

This finding emphasizes that cultivation approaches – although generally not able to record the whole diversity of a given environment (also stated above) – are still necessary for regular monitoring procedures. However, the ability of modern culture-independent molecular methods to assess the total microbial diversity present in a given environment is a powerful tool which enables researchers to elucidate the microbial community structure within the ISS beyond the standard cultivation assays. Next generation sequencing is nowadays also facilitating the microbiome analysis of the ISS. For instance, vacuum cleaner dust and filter debris collected from HEPA filters within the US American part of the ISS were analyzed in detail (Venkateswaran et al., 2014c) and their microbial inventory was also compared to the microbial inventory from spacecraft assembly cleanrooms (Checinska et al., 2015). Overall, there are several current projects which aim to broaden the knowledge about the ISS microbiome, including NASA's "Microbial Observatory" project (Venkateswaran et al., 2014a), JAXA's "Microbe" experiment series (Satoh et al., 2011; Ott et al., 2014; Ichijo et al., 2016; Yamaguchi et al., 2016) and ESA's ARBEX project (Moissl-Eichinger et al., 2016).

Almost all studies which investigated the ISS microbiome agree in one major aspect: the crew members act as the main source for the ISS microbial community, since most of the detected microorganisms are known to be human associated. The only studies which did not report a dominance of microorganisms of a presumable anthropogenic origin were studies conducted on the water supplies of the ISS, which is reasonable since these should normally not come in extensive physical contact with humans. Most of the organisms in the ISS water supplies were gram negative Proteobacteria, such as Methylobacterium, Sphingomonas, Ralstonia and Pseudomonas (La Duc et al., 2003b; Bruce et al., 2005).

Besides the human body, the other possible contamination source in this secluded habitat is the cargo delivered to the ISS including food, general equipment and material for scientific experiments. Cargo is always subjected to adapted cleaning procedures before upload and should be at least "visibly clean" before sent to the ISS (Pierson, 2001; Mord, 2009).

The crew on the ISS wears clothing, which does not impede the dispersal of microorganisms off the respiratory tract or skin and thus is certainly the major reason for the predominance of Staphylococcus (Firmicutes), Corynebacterium and Propionibacterium (Actinobacteria), which were also proven to be present in a viable status (Venkateswaran et al., 2014c).

### Conclusion And Outlook

Many human associated fungal and bacterial species are known to be opportunistic pathogens which are able to infect people with a (severely) compromised immune system. As mentioned above, the human immune system is proven to be compromised in space and the virulence of some (opportunistic) pathogens could even be enhanced under space-flight conditions. Additionally, if left uncontrolled in a confined environment where environmental strains are not present, which would normally outcompete human associated microorganisms under such conditions, human associated microorganisms can easily proliferate quickly and thereby pose a health hazard, as has been shown in artificial closed ecosystems on Earth (e.g., Sun et al., 2016). However, to date, there has been no serious infection reported on board the ISS, and the above mentioned CFU limits were exceeded only in a few cases in which appropriate countermeasures succeeded in a timely manner (Van Houdt et al., 2012).

Taking all the publicly available information into consideration, one can conclude that the preventive measures which are in place on board the ISS are currently sufficient to ensure the safety of crew and spacecraft from the microbiological perspective. Nevertheless, the longitudinal analysis of microbial community behavior under space conditions is necessary to deliver crucial knowledge to enable future long term space missions, as e.g., a flight to Mars and beyond. For such a long-term spaceflight, not only the maintenance of a healthy microbiome in the human body and the surrounding environment has to be considered, but also the safe production of food and recycling of water. Spaceflight simulations, such as MARS 300 and MARS 500, and microbial monitoring thereof (Project: MICHA, DLR Cologne) are extremely helpful in order to elucidate potential pitfalls during a flight to Mars and beyond. However, much more research in this area is needed to ensure the health and well-being of the crew during such missions.

Recent and current studies on the overall microbial communities onboard the ISS help to understand the influence of microorganisms on this special inhabited confined environment, as well as on other man-made environments on Earth (and vice versa). The overwhelming majority of detected microorganisms are, however, no threat toward human health or material but provide tremendous resources for human body function, sustainable waste remediation, recycling and purification of water and/or air supplies as well as nutrients for renewable food sources or even as a renewable food source themselves

(e.g., Nitta, 1999; Pierson, 2001; Czupalla et al., 2005; Bekatorou et al., 2006; Habib et al., 2008). In addition, the presence of beneficial microorganisms within a closed environment can help to suppress the harmful microbes and can thereby promote human health. As discussed in Mahnert et al. (2015a), this could potentially be achieved by installing plants in such confined environments, which could support indoor air quality, mental health, provide a food source and support human's health and well-being by providing a natural microbiome source (Mahnert et al., 2015a). However, more research needs to be done in this regard, also to ensure that no harmful microorganisms are introduced by such plants.

#### ADDENDUM: HIGH-SECURITY LABORATORIES

High-security laboratories are facilities developing customized technological solutions covering functional and security needs in natural scientific sectors. The purpose of such a laboratory is to reduce or eliminate exposure of laboratory staff and the outside environment to potentially hazardous agents. Different biosafety levels (BSL) for bio-containment are defined to work with dangerous biological agents in an enclosed laboratory facility. Biological safety levels are ranked from one (BSL-1) to the highest level four (BLS-4) where high security labs are categorized into BSL-3 and BSL-4 based on the agents or organisms on which the research or work is being conducted. High-security laboratories, in particular of level 4, are thus the most confined environments, where humans work. Although not much is known about the indigenous microbial diversity in such environments, for the sake of completeness, these environments shall be mentioned shortly in this review, pointing to a lack of knowledge in this regard. BSL-3 includes safety equipment and construction which are applicable to clinical, diagnostic, teaching, research, or production facilities in which work is done with dangerous agents causing serious and potentially lethal infections. Mycobacterium tuberculosis, St. Louis encephalitis virus, and Coxiella burnetii are representative of the microorganisms assigned to this level (Wilson and Chosewood, 2007). Primary hazards to personnel working with these agents relate to autoinoculation, ingestion, and exposure to infectious aerosols. At BSL-3, more emphasis is placed on primary and secondary barriers to protect personnel in contiguous areas, the community, and the environment from exposure to potentially infectious aerosols. Thus, more protective barriers are used in BSL-3 laboratories, including tightly closed wraparound protective suits made of special materials like DuPontTMTyvek <sup>R</sup> <sup>3</sup> and respirators if required. A high-security laboratory does also comprise self-closing double-doors access apart from general building passageways and the ventilation must supply ducted systems for directional airflows without recirculation. BSL-4 is defined for working with dangerous and exotic agents that pose a high individual risk of life-threatening disease, which may be transmitted via the aerosol route and for which there is no available vaccine or therapy. Agents with a close or identical antigenic relationship to BSL-4 agents should also be handled at this level. BSL-4 microorganisms are the Ebola virus, the Lassa virus, and any agent with unknown risks of pathogenicity and transmission. Thus, BSL-4 facilities provide the maximum protection and containment. In addition to the BSL-3 level, there are requirements for complete clothing change in a special lock and decontamination of all materials when entering and leaving the laboratory. A BSL-4 facility is generally located in a separate building or a totally isolated zone within a building with proper supply and exhaust ventilation systems, where high-efficiency particulate air (HEPA) filters exhaust the air, depending on the agents used (World Health Organization, 2005). Additionally, the laboratories also have their own air, electricity and water supply and multilevel security systems to prevent that pathogens reach the outside.

Besides biological agents, BSLs comprise safe work practices, specialized safety equipment (primary barriers) and facility design (secondary barriers), which are summarized in different reports (US Department of Health and Human Services, 2009; McLeod, 2010). In the United States, the Centers for Disease Control and Prevention (CDC) have specified all BSL levels (Richmond and McKinney, 1999), whereas in the European Union, the same biosafety levels are defined in directives. The most important EU regulations for biosafety laboratories are directive for biological agents at work (2000/54/EC), workplace safety (89/391/EC), contained use of microorganisms (98/81/EC), deliberate release into the environment (2001/18/EC), hazardous waste (94/31/EC), or directive on harmful organisms, plants, plant products and other objects (95/44/EC). The CDC and the National Institutes of Health (NIH) are our main sources for biological safety information for infectious agents. High security laboratories are characterized by strict hygienic guidelines comprising qualified employees, measures, and disinfectionand cleaning plans. Disinfectants, dosage and applications are defined within SOPs. The common decontamination strategies are hydrogen peroxide (H2O2), formaldehyde (CH2O) or chlorine dioxide (ClO2). A study (Beswick et al., 2011) reported that these methods were tested of their efficacy where only chlorine dioxide and formaldehyde showed a high disinfection efficacy.

In comparison to other indoor environments, biosafety laboratories are even more confined. Wearing of special clothes and protections prevents the humans from microbes as well as particles, which can be a carrier of microorganisms. Microbial communities of mentioned indoor habitats within this review are well analyzed by several next generation sequencing methods, but no research study about microbiome analysis in high security laboratories exists. It can be assumed that microorganisms of this extreme habitat adapt to low nutrient and dry conditions as well as strict hygienic guidelines (e.g., decontamination procedures or lock systems), as it was observed for microorganisms of other confined habitats. Further, monitoring of high-security labs is becoming more and more important to prevent outbreaks of

<sup>3</sup>www.dupont.com

disease and to maintain public faith. Generally, the awareness is low, but in cases of epidemic, e.g., Ebola or SARS, the interest increases. Thus, the World Health Organization (WHO) has recently witnessed a worldwide increase in the demand for biosafety guidance and support that culminated in 2005 with the adoption by the World Health Assembly of resolution WHA 58.29 on enhancement of laboratory biosafety (World Health Organization, 2005).

## CONCLUSION

Due to similar maintenance, architecture and type of confinement, the environments presented here harbor a very specific microbial core-community. ICUs, ORs, cleanrooms and even the ISS share a number of typical microbial inhabitants, as displayed in **Figure 2**. In particular Bacillus and humanassociated microbial species are cultivated from all confined areas, reflecting the typical microbial community being composed of survival specialists (such as spore formers) and mainly representatives of the human microbiome, defining the human body as major source of microbial contamination.

The purposes for confinement are different. In the hospital area, the risk of infection is the major driving factor for confinement. Interestingly, higher efforts in cleaning (i.e., sterilization and bioburden reduction) do not necessarily decrease the risk for infections, in contrary: they were even correlated with a higher incidence of infections and presence of multi-resistant strains. Similar findings exist for cleanroom environments: the microbial inhabitants frequently showed

higher resistances against physical and chemical stresses than their naturally occurring counterparts. All of these discussed habitats are extreme and pose stresses toward the internal microbiome, which entails a positive selection pressure for microbes which are adapted to such stresses and therefore promotes the development and establishment of survival strategies within these habitats.

Interestingly, the ISS seems to be a safe work space: despite allergic reactions (Venkateswaran et al., 2014c), so far no severe incidences of outbreaks have officially been reported. Certainly, although most confined, this is also an area where a higher number and diversity of microorganisms can be accepted, since neither persons, nor products are exposed to instantaneous risk.

Although cleanrooms are not living places for human beings, they have been subjected to comprehensive microbial analyses during the last years, using most sophisticated molecular and cultivation-based methods. While the overwhelming majority of cleanroom microorganisms appears to be dead, the survivors are specifically resistant and are considered possible contaminants, of e.g., spacecraft targeting potential extraterrestrial biotopes.

In all habitats considered in this review, the routes of microbial transmission are not clearly resolved yet, leading to uncertainty with respect to optimal maintenance and risk management. Based on our experience and the information summarized in this review, we argue, that hygiene and maintenance strategies need to be critically reviewed, and the role of beneficial microorganisms,

### REFERENCES


that naturally suppress unwanted microorganisms, need to be reassessed. The most-likely healthy transfer of beneficial microorganisms through, e.g., pets or plants into patient rooms is currently restricted, due to uncontrollable risks. However, a controlled spreading of selected, beneficial microorganisms in certain settings could help tremendously to improve quality of living and human health and to reduce long-term risks emanating from multi-resistant microbial strains.

#### AUTHOR CONTRIBUTIONS

MM wrote chapter about ISS and organized manuscript writing. AM wrote chapter about clean rooms and prepared figures. KK wrote chapter about ICUs. MP wrote chapter about operating rooms. LO-W wrote chapter about biosafety laboratories. RK, AP, GG, and GB critically read the manuscript and provided discussion. CM-E wrote and composed the manuscript.

#### ACKNOWLEDGMENTS

We thank all colleagues that provided information. This work was supported by FFG No. 847977. MM and MP were trained within the frame of the Ph.D. Program Molecular Medicine of the Medical University of Graz.



Methicillin-Resistant Staphylococcus aureus. Astrobiology 13, 1081–1090. doi: 10.1089/ast.2013.0986


tolerating the extreme conditions of clean room environments. Appl. Environ. Microbiol. 73, 2600–2611. doi: 10.1128/AEM.03007-06


the Herschel spacecraft during assembly, integration, and test operations. Astrobiology 13, 1125–1139. doi: 10.1089/ast.2013.1024


**405**


fmicb-07-01573 October 8, 2016 Time: 16:28 # 18


Wertheim, H. F., Melles, D. C., Vos, M. C., van Leeuwen, W., van Belkum, A., Verbrugh, H. A., et al. (2005). The role of nasal carriage in Staphylococcus aureus infections. Lancet Infect. Dis. 5, 751–762. doi: 10.1016/S1473-3099(05)70295-4 Whitfield, W. (1964). Ultra-clean room. US 3158457A

Whyte, W. (2001). Cleanroom Technology-Fundamentals of Design. Testing, and Operation. West Sussex: Johnson Wiley & Sons.

Whyte, W. (1999). Cleanroom design. West Sussex: John Wiley & Sons Ltd.


**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 © 2016 Mora, Mahnert, Koskinen, Pausan, Oberauner-Wappis, Krause, Perras, Gorkiewicz, Berg and Moissl-Eichinger. 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) or licensor 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.

# Functional Metagenomics of Spacecraft Assembly Cleanrooms: Presence of Virulence Factors Associated with Human Pathogens

Mina Bashir 1, <sup>2</sup> , Mahjabeen Ahmed1, 3, Thomas Weinmaier <sup>4</sup> , Doina Ciobanu<sup>5</sup> , Natalia Ivanova<sup>5</sup> , Thomas R. Pieber <sup>2</sup> and Parag A. Vaishampayan<sup>1</sup> \*

*<sup>1</sup> Biotechnology and Planetary Protection Group, Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, USA, <sup>2</sup> Division of Endocrinology and Diabetology, Medical University of Graz, Graz, Austria, <sup>3</sup> Department of Biological Sciences, California State Polytechnic University, Pomona, CA, USA, <sup>4</sup> Division of Computational Systems Biology, Department of Microbiology and Ecosystem Science, University of Vienna, Vienna, Austria, <sup>5</sup> Department of Energy, Joint Genome Institute, Walnut Creek, CA, USA*

#### Edited by:

*Martin Grube, University of Graz, Austria*

#### Reviewed by:

*Tomislav Cernava, Austrian Centre of Industrial Biotechnology, Austria Tim Sandle, University of Manchester, UK*

> \*Correspondence: *Parag A. Vaishampayan vaishamp@jpl.nasa.gov*

#### Specialty section:

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

Received: *09 March 2016* Accepted: *10 August 2016* Published: *09 September 2016*

#### Citation:

*Bashir M, Ahmed M, Weinmaier T, Ciobanu D, Ivanova N, Pieber TR and Vaishampayan PA (2016) Functional Metagenomics of Spacecraft Assembly Cleanrooms: Presence of Virulence Factors Associated with Human Pathogens. Front. Microbiol. 7:1321. doi: 10.3389/fmicb.2016.01321* Strict planetary protection practices are implemented during spacecraft assembly to prevent inadvertent transfer of earth microorganisms to other planetary bodies. Therefore, spacecraft are assembled in cleanrooms, which undergo strict cleaning and decontamination procedures to reduce total microbial bioburden. We wanted to evaluate if these practices selectively favor survival and growth of hardy microorganisms, such as pathogens. Three geographically distinct cleanrooms were sampled during the assembly of three NASA spacecraft: The Lockheed Martin Aeronautics' Multiple Testing Facility during DAWN, the Kennedy Space Center's Payload Hazardous Servicing Facility (KSC-PHSF) during Phoenix, and the Jet Propulsion Laboratory's Spacecraft Assembly Facility during Mars Science Laboratory. Sample sets were collected from the KSC-PHSF cleanroom at three time points: before arrival of the Phoenix spacecraft, during the assembly and testing of the Phoenix spacecraft, and after removal of the spacecraft from the KSC-PHSF facility. All samples were subjected to metagenomic shotgun sequencing on an Illumina HiSeq 2500 platform. Strict decontamination procedures had a greater impact on microbial communities than sampling location Samples collected during spacecraft assembly were dominated by *Acinetobacter* spp. We found pathogens and potential virulence factors, which determine pathogenicity in all the samples tested during this study. Though the relative abundance of pathogens was lowest during the Phoenix assembly, potential virulence factors were higher during assembly compared to before and after assembly, indicating a survival advantage. Decreased phylogenetic and pathogenic diversity indicates that decontamination and preventative measures were effective against the majority of microorganisms and well implemented, however, pathogen abundance still increased over time. Four potential pathogens, *Acinetobacter baumannii, Acinetobacter lwoffii, Escherichia coli* and *Legionella pneumophila,* and their corresponding virulence factors were present in all cleanroom samples. This is the first functional metagenomics study describing presence of pathogens and their corresponding virulence factors in cleanroom environments. The results of this study should be considered for microbial monitoring of enclosed environments such as schools, homes, hospitals and more isolated habitation such the International Space Station and future manned missions to Mars.

Keywords: cleanroom, pathogens, indoor environments, microbiome, spacecraft, virulence factors, Acinetobacter, functional metagenomics

#### INTRODUCTION

Detection of signs of life on other planets is of particular interest for many of NASA's planetary missions. In order not to mistake earthborn microorganisms for unknown potential extraterrestrial life, planetary missions are subject to internationally accepted standards of planetary protection, established by the Committee of Space Research (COSPAR) (National Research Council, 2006). Jet Propulsion Laboratory's Planetary Protection Group has undertaken huge efforts (NASA Policy Directive (NPD) 8020.7G, 1999) to avoid inadvertent contamination of other planets with earthborn organisms, and to minimize the bioburden on spacecraft. Spores are of particular interest, given their high resistance to multiple sterilization techniques, including radiation (Venkateswaran et al., 2003; La Duc et al., 2007; Vaishampayan et al., 2012).

All spacecraft parts undergo extensive cleaning and sterilization steps, such as exposure to dry heat, vaporized hydrogen peroxide, radiation and alcohol on surfaces. Additional protocols to reduce the influx of particulate matter include daily vacuuming and mopping of floors, HEPA air filtration, regular replacement of tacky mats at all entry points, and strict gowning procedures. These precautions are routinely taken but with high frequency and stringency during the spacecraft assembly. All personnel that enter the cleanroom are required to put on cleanroom garments. This includes a full body suit, hair and beard nets, facemasks, additional head covering, gloves, shoe covers, and cleanroom boots. These are necessary measures since humans are the major source of contamination in these environments (La Duc et al., 2004; Probst et al., 2013). To monitor contamination levels, cleanrooms are regularly sampled for biological activity, particularly when spacecraft parts are being assembled and cleaned (La Duc et al., 2007; Vaishampayan et al., 2010a).

Multiple sterilization methods are chosen, because there is no known method that can eradicate all microbes, which is still compatible with spacecraft components. Only very resistant microorganisms, such as spores, pathogens, and extremophiles, can overcome these strict decontamination procedures (Ghosh et al., 2009; Derecho et al., 2014). Some microorganisms are even able to survive the harsh conditions of interstellar travel. Researchers placed spore-forming bacteria, isolated from cleanroom environment, outside the International Space Station for 18 months along with exposure to simulated Marslike conditions, including atmospheric pressure and selective UV-radiation and some of them were still able to survive (Vaishampayan et al., 2012).

Our goal was to elucidate whether decontamination measures lead to selection of hardy microorganisms, including pathogens, in the cleanrooms and therefore posing a potential threat to human health. Pathogens might thrive in these environments perhaps due to their selective phenotypic characteristics, metabolic capabilities and reduced competition for scarce nutrients and niches. We were particularly interested in human pathogens, given that humans are the main source of contamination in cleanrooms (La Duc et al., 2004; Probst et al., 2013), and also because they are exposed to these constantlyevolving microbes. Most studies aiming at determining the microbiome of cleanrooms (La Duc et al., 2009; Sandle, 2011; Vaishampayan et al., 2013; Mahnert et al., 2015; Moissl-Eichinger et al., 2015), other indoor environments (Adams et al., 2015) or even the International Space Station (Checinska et al., 2015) have used 16S rRNA amplicon sequencing. 16S rRNA amplicon sequencing is often used to screen for potential pathogens (Case et al., 2007; Stadlbauer et al., 2015; Bashir et al., 2016). However, the lack of discriminability does not allow differentiating between potential and true pathogens. Previous functional metagenomic studies investigated pathogens in other indoor environments (Tringe et al., 2008; Afshinnekoo et al., 2015), but this is the first study, which focuses on the detection of pathogens as well as virulence factors in cleanrooms.

Three geographically distinct cleanrooms were sampled during the assembly of three NASA spacecraft: Phoenix in Cape Canaveral, Florida, DAWN in Fort Worth, Texas, and Mars Science Laboratory (Curiosity) in Pasadena, California. Sample sets from Phoenix mission were collected from the cleanroom at three time points: before arrival of the spacecraft, during the assembly and testing of the Phoenix spacecraft, and after removal of the spacecraft from the facility. All samples were subjected to whole metagenome shotgun sequencing on an Illumina HiSeq 2500 platform. We screened for pathogens and virulence factors, which determine pathogenicity. Clinically relevant pathogens were identified by searching taxonomic classification and potential virulence factors were identified by comparing reads to the Microbial virulence database.

#### MATERIALS AND METHODS

#### Sample Collection and Processing

Multiple samples were collected from the floor of the Kennedy Space Center's Payload Hazardous Servicing Facility (KSC-PHSF), where the Phoenix spacecraft was assembled. Sample sets were collected from the KSC-PHSF surfaces at three time points: before arrival of the Phoenix spacecraft (10 samples; PHX-B), during the assembly and testing of the Phoenix spacecraft (8 samples; PHX-D), and after removal of the spacecraft from the KSC-PHSF facility (10 samples; PHX-A). 10 samples from the Lockheed Martin Aeronautics' Multiple Testing Facility (LMA-MTF) floor were collected during the DAWN spacecraft assembly. Samples were collected from the Ground Support Equipment (GSE) at Jet Propulsion Laboratory's spacecraft assembly facility (JPL-SAF) during the Mars Science Laboratory (2 samples; MSL) spacecraft assembly. These three cleanroom facilities were certified at ISO 8 (3,520,000 particles >0.5µm m−<sup>3</sup> ) level and maintained according to the standard cleaning practices. Each sample was collected from 1 m<sup>2</sup> of the cleanroom floor or GSE by a wet surface sampling technique using Biological Sampling Kits (BiSKits, QuickSilver Analytics, Abingdon, Md.) and polyester wipes, respectively. Samples from each sampling event were concentrated using Amicon Ultra-15 centrifugal filter tube (Millipore, Jaffrey, NH, Ultracel-50 membrane) as described earlier (La Duc et al., 2009). DNA was extracted from each concentrated sample using bead beating and an automated DNA extraction instrument (Autolyser A-2 DNA, Axcyte Genomics, Menlo Park, CA) and pooled equimolar, as described earlier (Vaishampayan et al., 2010b). DNA samples were archived at −80◦C until further use. Negative controls such as field control (sampling devise control), reagent control (during DNA extraction) at each step were collected. None of the negative controls had a sufficient DNA concentration for library preparation and were thus not included in further downstream analysis.

#### Metagenomic Sequencing

Sample processing was performed in a sodium hypochlorite (bleach) treated laminar flow hood in an ultra-clean environment. Operators were using single-use lab-coats, bleached gloves, hairnets, and booties. Due to low DNA concentrations, samples were subject to multiple displacement amplification (MDA) (Dean et al., 2002). Each sample was divided into 1 ml aliquots, which were amplified via MDA using Repli-g single-cell whole genome amplification kit (Qiagen part #150345). All plastic ware and water were ultraviolet (UV) treated in a Stratalinker 2400 UV Crosslinker (Stratagene, La Jolla, CA) with 254-nm UV for 30–90 min on ice (Woyke et al., 2011). This represents a UV dose range of 5.7–17.1 J/cm2, calculated by measuring the distance from inside the tubes to the light bulb (4 cm). Buffer and enzyme come pre-cleaned and don't require UV-radiation. MDA reaction was prepared following manufacturer protocol for single cells, scaling reaction volume down to 15µl final volume and addition of Syto13 dye for real-time monitoring. MDA reaction was stopped when sample amplification reached saturation.

Amplified fractions of each sample were combined, and this pooled DNA product (100 µl) was sheared using a Covaris E210 instrument (Covaris, Woburn, MA) set to: 10% duty cycle, intensity 5, and 200 cycles per burst for 1 min. The concentration and fragment size of each sheared product was determined using Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA) in accordance with the manufacturer's recommended conditions. The sheared DNA was end-repaired, A-tailed, and ligated to Illumina adaptors according to standard Illumina PE protocols (Illumina, San Diego, CA). The concentration of the resulting Illumina-indexed libraries was again determined using Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). Libraries were pooled and normalized to a final concentration of 400 mM each, and the primary bands corresponding to the sizes were gel-purified and dissolved in 30µl TE. One flow-cell was generated from a pooled library, which was subsequently subjected to sequencing in an Illumina HiSeq2500 instrument (2 × 250 bp), in accordance with manufacturer-provided protocols. The raw sequence data are available within IMG/M (https://img.jgi.doe.gov/cgi-bin/mer/ main.cgi) and NCBI's short read archive under the accession number SRP077843.

#### Sequence Data Analysis

We started with a total of 15,001,132 paired reads for PHX-B, 14,654,014 for PHX-D, and 22,355,430 for PHX-A before quality filtering and pairing. For MSL and DAWN we had 57,892,216 and 2,899,364 reads, respectively (**Table 1**).

FastQC v0.10.1 (Andrews, 2010) was used to determine the base quality throughout the 250 bp HiSeq-generated pairedend reads. PEAR v0.9 (Zhang et al., 2014; default parameters) was used to merge paired reads. Unmerged forward and reverse reads were retained. Merged and unmerged reads were processed using prinseq-lite v0.20.3 (Schmieder and Edwards, 2011) with the following parameters: "-min\_len 100 -trim\_qual\_right 20 -trim\_qual\_left 20 -trim\_left 8." Adapter sequences and overrepresented sequences were identified with FastQC and removed using Cutadapt v1.1 (Martin, 2011). PhiX174 and

TABLE 1 | Data statistics: number of reads per sample starting with raw reads coming from the sequencer until final taxonomic and functional classification.


a JGI-standard collection of potential contaminant genomes (Supplementary Table 1) were removed by mapping trimmed high-quality reads using BBMap short read aligner v31.18 (Bushnell, 2014) to the respective genomes. Any reads matching any of these contaminant genomes were removed from the dataset.

To generate the human DNA sequence free dataset, all remaining high-quality reads were mapped with BBMap short read aligner against the human genome GRCh38 (including mitochondrial DNA). All positive matches were removed from the dataset.

Both, datasets including and excluding human DNA sequences were compared to NCBI non-redundant database using DIAMOND BLASTX v0.7.1 (Buchfink et al., 2014) with default parameters. Results were imported to MEGAN v5.10.5 (Huson et al., 2007; minimal bit score of 80%; "minscore 80") for taxonomic binning, functional assignments to KEGG functions, and generation of rarefaction curves (phylogenetic diversity on genus level). After removal of unassigned and unclassified reads, taxonomy and KEGG pathways (Kanehisa and Goto, 2000; Kanehisa et al., 2014) were visualized using Krona Tools v2.4 (Ondov et al., 2011). Taxonomic and metabolic diversity calculations were done in QIIME 1.9.1 (Caporaso et al., 2011) with all samples subsampled to the smallest sample size observed.

Potential virulence factors were identified by comparing contaminant- and human-DNA-sequence-free reads to the Microbial Virulence Database MvirDB (Zhou et al., 2007) using DIAMOND BLASTX (Buchfink et al., 2014) with a 80% sequence similarity cut-off and maximum of target sequences of one. Sequences which passed these criteria were extracted and compared to NCBI non-redundant database using DIAMOND BLASTX (Buchfink et al., 2014) with a maximum of target sequences of one for virulence factor validation (Data Sheet 1 in Supplementary Material). Classified sequences were searched for clinically relevant pathogens (http://www.bode-science-center. com/center/relevant-pathogens-from-a-z.html accessed on Dec 1 2015; Supplementary Table 2).

## RESULTS

#### Phylogenetic Diversity of Cleanroom Samples

Alpha rarefaction curves indicate, besides sufficient sampling efforts, that diversity is drastically lower during the actual spacecraft assemblies (PHX-D, DAWN, and MSL) compared to before or after. This confirms that the very strict gowning, cleaning and sterilization procedures were well executed, and highly effective as previously described (Ghosh et al., 2010). MSL had the highest sampling depth but lowest bioburden (**Figure 1A**) as GSE undergo stringent cleaning procedures and are exposed to less handling and human contact compared to the cleanroom floors. Interestingly, microbial community profiles during active spacecraft assembly (PHX-D, DAWN, MSL) were more similar to each other than to samples from one location (**Figure 1B**). Moraxellaceae was the dominating family, making 83, 73, and 62% of all classified sequences for PHX-D, DAWN, and MSL, respectively. The majority of all Moraxellaceae, 94% to 100%, were Acinetobacter spp. (**Figure 2** and Presentation S2 in Supplementary Material), making it the most dominating taxa during spacecraft assembly.

In general, bacteria were the most dominant kingdom present in all tested cleanrooms with 63 to >99% of all classified sequences. Archaea and viruses on the other hand accounted for less than 0.1% relative abundance combined (**Figure 2** and Presentation S2 in Supplementary Material). Surprisingly, the amount of potentially human DNA was minimal. Only 0.04–2% of all sequences were classified as primates: PHX-B 2%, PHX-D 0.08%, PHX-A 0.2, DAWN 0.05, and MSL 0.2% (Presentation S2 in Supplementary Material).

In PHX-B eukaryotes made 36% of all classified sequences. Most of these sequences (22% of total; 60% of eukaryotes) belong to the class of arthropods, such as insects and arachnids. In all other samples less than 0.1% of all classified sequences were arthropods. Probably arthropod sequences originated from free DNA associated with dust particles, given that no living spiders or insects are present in any cleanrooms. In MSL all eukaryotic sequences were assigned to craniate. Fungi were also not prominent in our cleanrooms. The fungal abundance ranged from 0.0008% (MSL) to 1% (PHX-B) (Presentation S2 in Supplementary Material).

#### Metabolic Diversity During Spacecraft Assembly

Functional assignment resulted in 13,360 KEGG orthologous (KO) for PHX-B, 24,916 KOs for PHX-D, 298,350 KOs for PHX-A, 557 for DAWN and 664,699 for MSL (**Table 1** and Data Sheet 1 in Supplementary Material). **Figure 3** indicates that the majority of the functional classification was assigned to metabolism (PHX-B 67%, PHX-D 67%, PHX-A 90%, DAWN 75%, MSL 29%; **Figure 3** and Data Sheet 1 in Supplementary Material). Although the percentage of sequences assigned to metabolism did not differ much across samples (**Figure 3** and Data Sheet 1 in Supplementary Material), we saw a higher metabolic diversity during assembly compared to before or after spacecraft assembly samples (**Table 1**). We also found that the metabolism of pantothenate and coenzyme A is higher during assembly (PHX-D 4%, DAWN 2%) compared to PHX-B or PHX-A, 0.2% respectively (Data Sheet 1 in Supplementary Material). Nevertheless, no function associated with pantothenate and coenzyme A was found in MSL during assembly. Fifty-two percent of all functional classification from MSL was assigned to Holliday junction DNA helicase RuvB (Genetic Information Processing; Replication and Repair). In PHX-B, 9% were assigned to Genetic Information Processing, 12%, in PHX-D 3%, in PHX-A 11% in DAWN, and 52% in MSL (**Figure 3**).

### Potential Pathogens in Cleanroom Samples

After taxonomic classification we selectively screened the classified binned sequences of all our samples for clinically relevant pathogens (Supplementary Table 2). In total we found 48 different human pathogens in all cleanrooms, responsible for various diseases, from gastrointestinal, to the nervous system. Twenty nine different pathogens were detected in PHX-B, 18 in PHX-D, 33 in PHX-A, 10 in DAWN and 11 in MSL

(**Table 2**). Strikingly, four pathogens, namely Acinetobacter baumannii, Acinetobacter lwoffii, Escherichia coli and Legionella pneumophila, were detected in all cleanrooms, even though they are geographically separate. Besides these four pathogens, present in all samples, we found pathogens that were exclusive to KSC-PHSF, during all three time points, namely, Bacillus cereus, Burkholderia pseudomallei, Enterobacter cloacae, Enterococcus faecalis, Listeria monocytogenes, Pseudomonas aeruginosa, Staphylococcus aureus and Staphylococcus epidermidis. In case of PHX-D, 83% of all classified reads were identified as Acinetobacter spp. (Presentation S2 in Supplementary Material), which is listed as a clinically relevant pathogen (Supplementary

Table 2). Additionally, we found a decreased pathogen diversity during the actual spacecraft assembly in KSC-PHSF, while pathogen abundance almost triples over time (PHX-B 1.52, PHX-D 2.34, and PHX-A 4.26%; **Table 2**).

#### Pathogens and Corresponding Virulence Factors in Cleanrooms

Virulence factors are features, which distinguish pathogens from commensals or symbionts (Das et al., 2011). We found that the fraction of sequences identified as potential virulence factors increased over time in case of Phoenix (**Table 3**), although overall diversity was lower during assembly (**Figure 1A**). DAWN

had approximately half the virulence factor fraction compared to Phoenix, but MSL, which was sampled from GSEs, had approximately 20 times less potential virulence factors compared to Phoenix.

To evaluate if we could find pathogens and their corresponding virulence factors, we identified potential virulence factors that are specifically associated with the pathogens found in our samples. We found 14 different potential virulence factors, which correspond to the classified pathogens in PHX-B, 48 for PHX-D and 41 for PHX-A. Nine different virulence factors were found to correspond with the classified pathogens in DAWN, and 6 were found for the MSL mission.

We were particularly interested in detecting potential virulence factors of the four pathogens, A. baumannii, A. lwoffii, E. coli, and L. pneumophila, which were found in all geographically separated cleanrooms. We found a Acinetobacter sp. specific aminoglycoside 6′ -N-acetyltransferase lv and tetA, which are kanamycin B and tetracycline resistance genes, respectively, in PHX-D. Moreover, we found adeABC, which is an A. baumannii specific multidrug efflux pump and beta-lactamase TEM-1 (Supplementary Table 3).

We found that the abundance of potential virulence factors with associated pathogens increased over time (**Table 4**). Although, virulence factors diversity did not change over time, we observed a trend toward increased pathogens with associated virulence factors (**Table 4**, pathogenic diversity). Again, MSL had the smallest pathogenic diversity.

#### DISCUSSION

In this study, we demonstrated for the first time the presence of pathogens and their corresponding virulence factors in spacecraft assembly cleanrooms. Our approach allowed us not only to prove the presence of pathogens in the spacecraft assembly cleanrooms, but also their associated potential virulence factors. Most studies investigating the cleanroom microbiome have only used 16S rRNA amplicon sequencing (Vaishampayan et al., 2013; Mahnert et al., 2015). For example, the archived samples from KSC-PHSF during the Phoenix mission used in this study have previously been described using a cultivation based (Ghosh et al., 2010) and cultivation independent technique (Vaishampayan et al., 2010a). On one hand, cultivation based techniques offer a very limited insight into the wide spectrum of microbial diversity, given that most microorganisms are not cultivable, while 16S rRNA amplicon sequencing on the other hand shows a more broad picture, but does not allow a reliable phylogenetic classification below genus level and does not provide any information regarding virulence factors and potential pathogenicity.

We observed that cleanroom samples are dominated by bacteria as reported previously (Weinmaier et al., 2015). Contrary to previous studies, which found substantially more human, archaeal and viral sequences in cleanrooms (Moissl-Eichinger, 2011; Weinmaier et al., 2015), we found significantly less of each taxon in all cleanrooms tested during this study. These previous studies have sampled uncontrolled gowning area and ISO-8 cleanrooms, where no active spacecraft assembly was ongoing. Moreover, each cleanroom is unique, because of factors such as geographical location (Moissl et al., 2007), assembly activities, different decontamination procedures and most importantly, different workers, which are the main source of contamination.

We saw an increased metabolic diversity in samples collected from cleanrooms during spacecraft assembly. Cleanrooms are sometimes referred to as extreme environments; not only due to strict decontamination procedures, but also due to the lack of nutrients, water and cofactors (La Duc et al., 2007; Ghosh et al., 2010). Since there are very few resources to rely on in an area that is maintained to be uninhabitable, any microbes able to survive here would have to synthesize all necessary factors themselves. Sterilization procedures and gowning requirements are even stricter during assembly, making it even harder for microorganisms to survive. Strict gowning protocols and increased frequency of cleaning leads to decrease in introduction of human associated microbes, despite high human activities in the cleanroom during assembly. This might also explain lower phylogenetic diversity during Phoenix spacecraft assembly compared to before or after assembly. In addition to the decreased phylogenetic diversity, pathogenic diversity was also lower during spacecraft assembly, however, we observed an increase in pathogen abundance over time. This suggests that strict decontamination procedure favor the growth of pathogens. Nevertheless, studies with bigger sample sizes need to confirm our descriptive findings. A considerable amount, in case of MSL more than 50%, of all reads was assigned to genetic information processing. This highlights the importance of genetic information processing, including DNA repair in such a harsh environment. Surprisingly, microbial profiles during assembly were very similar. Although, DAWN and MSL samples were collected from geographically distinct locations, they were more similar to PHX-D than PHX-B or PHX-A. This suggests that decontamination procedures have a bigger effect on the cleanroom microbiome than location. Taken together, these results show that decontamination and gowning measures were not only sufficient, but also well executed.

Most virulence factors are organized in so-called pathogenicity-islands (Schmidt and Hensel, 2004). Commensals

#### TABLE 2 | Pathogen diversity is lowest during assembly: pathogens found in the different cleanroom samples.


*(Continued)*

#### TABLE 2 | Continued


−*...Not present.*

\**...Pathogens found in all cleanroom samples.*

can turn into pathogens by acquiring pathogenicity-island through phages, or horizontal gene transfer. For example, wild type Vibrio cholerae are not able to cause deadly diarrhea. Only upon infection by the CTX prophage they acquire a pathogenicity island coding for virulence factors, such as the

cholera toxin and pili, needed for recognition host and disease induction (Das et al., 2011). Therefore, virulence factor detection is the only reliable method to identify pathogens.

Acinetobacter baumannii, Acinetobacter lwoffii, Escherichia coli and Legionella pneumophila were found in all samples,

TABLE 3 | Accumulation of virulence factors over time: total number of virulence factors and hits normalized to hits per million reads found in cleanrooms.


TABLE 4 | Virulence factors with their corresponding pathogens.


*norm: normalized to counts per million reads.*

\**Number of pathogens with* ≥*1 corresponding virulence factors.*

although samples were collected from three geographically distinct sites. These prevalent pathogens have to be very resistant to overcome all the cleaning and decontamination procedures. Acinetobacter spp., such as A. baumannii and A. lwoffii are non-fastidious and can rely on a single energy source from different substrates (Rathinavelu et al., 2003). They are resistant to radiation (Firstenberg-Eden et al., 1980b) and several disinfectants and can survive in a wide range of temperatures (Firstenberg-Eden et al., 1980a) and even in low pH. These features might explain why Acinetobacter spp. were the most dominating species during spacecraft assembly in this study. Acinetobacter spp. have also been reported in high abundance in cleanrooms in previous studies (Vaishampayan et al., 2010a; Mahnert et al., 2015). Acinetobacter baumannii has been isolated from water and soil (Yeom et al., 2013), but it has also been found in other hostile environments such as intensive care units. Although A. baumannii is not pathogenic to healthy individuals, it can be an opportunistic pathogen in immunocompromised patients. Hence, it is one of the ESKAPE pathogens (Boucher et al., 2009), which are multidrug-resistant bacteria, responsible for the majority of nosocomial infections (Rice, 2008).

We found A. baumannii specific beta-Lactamase TEM-1, AdeABC and another cation/multidrug efflux pump, which might be responsible for A. baumannii's resistance against all decontamination measures. AdeABC alone is responsible for resistance to aminoglycosides, tetracyclines, erythromycin, chloramphenicol, trimethoprim, fluoroquinolones, some beta-lactams, and also recently tigecycline since they have been described as substrates for this multidrug efflux pump (Wieczorek et al., 2008). We did not detect A. Iwoffii associated potential virulence factors in our data set. MvirDB has only three A. Iwoffii (formerly known as Acinetobacter calcoaceticus) associated virulence factors (two beta-lactamases and a chloramphenicol acetyl transferase). Nevertheless, the presence of this opportunistic pathogen in all our sample collection from locations separated by hundreds of miles, its resistant features, and our finding that Acinetobacter spp. were dominating in all three locations during assembly, suggests that A. lwoffii and A. baumannii are actually viable in the spacecraft cleanroom environment. L. pneumophila, another pathogen present in all three distinct locations, is the causative agent of the Legionnaires' disease (Shevchuk et al., 2011), with symptoms such as fever, chills, and coughing. We found Legionella secretion pathway protein E (LspE), which is part of a type II secretion system required for its full virulence and environmental persistence (Hales and Shuman, 1999). In addition, other L. pneumophila associated virulence factors, such as the catalase-peroxidase KatB and superoxide dismutase were present, potentially explaining why this species is resistant to hydrogen peroxide treatment, one of the decontamination procedures. The last potential pathogen we found in all cleanrooms was E. coli. Although, we have found several virulence factors such as transposases and antimicrobial resistance genes, we cannot confirm whether or not this specific E. coli is a pathogen, given that more and more antimicrobial resistance genes are being found in commensal E. coli (Kaesbohrer et al., 2012; Tadesse et al., 2012; Wasyl et al., 2013). While we think that the four pathogens found in all geographically separated cleanrooms are alive, given their resistant features, we are not able to tell if the classified taxa and functions derive from intact living cells or if they are originating from dead cells. In an ongoing study we're including propidium monoazide staining, enabling us to differentiate between sequences coming from intact live and dead microorganisms.

Interestingly, potential virulence factor abundance increased over time, despite lower phylogenetic diversity during assembly. We have concluded that virulence factors may provide a survival advantage in this very hostile environment (Rathinavelu et al., 2003). Multidrug efflux pumps might be pumping out harmful compounds before they are able to execute their deadly effect (Yoon et al., 2013). This virulence factor accumulation seems to be species dependent, as we also see an increase in pathogenic abundance over time. We also found pathogens not belonging to the bacterial kingdom; such as Candida parapsilosis, a fungus, which plays an important role in wound and tissue sepsis of immunocompromised patients and makes up to 15% of all Candida infections.

One limitation of this study is the low biomass in cleanroom samples, due to the repeated strict cleaning and decontamination practices. MDA was necessary to acquire DNA concentrations sufficient for library preparation. MDA can introduce bias, by favoring some DNA fragments over others (Direito et al., 2014). Therefore, some microorganisms might not have been detected in our approach, while others might be overrepresented. Although, we performed stringent quality filtering of our reads, it's impossible to get rid of all errors and biases. Homology based approaches such as BLASTx against specialized databases such as MvirDB are biased, because a sequence with an 80% sequence similarity might have a better hit to a reference which is not in the database. However, we searched all positive MvirDB hits against NCBI non-redundant database, and the majority was classified as virulence factors (see Presentation S1 in Supplementary Material). Moreover, the circumstantial evidences of the presence of virulence factors associated with human pathogens in cleanroom samples should be confirmed by implementation of selective cultivation based approaches and viability-based molecular assays in future missions.

Humans spend most of their lives indoors (Höppe and Martinac, 1998). Recent studies have speculated that human microbiome is the major contributor to the overall indoor microbiome (Lax et al., 2014). Stringent cleaning and maintenance practices in highly controlled indoor environments such as cleanrooms, hospitals and intensive care units may lead to a relative increase of human pathogens in these environments. This may have serious impact on health of the inhabitants. Monitoring pathogens and virulence factors in these indoor environments may prevent diseases such as nosocomial infections and sustain human health.

The results of this study could be used to develop fast and cost-efficient tests (Craw et al., 2015) to detect the presence of specific pathogens or their virulence factors in enclosed environments such as public transport, pharmaceutical cleanrooms, hospitals, and intensive care units. This study has broadened our understanding of the role of pathogens in such highly controlled environments and should be considered for microbial monitoring of the ISS during sustained presence of humans in space and future manned missions to Mars.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

Designed project: PV. Performed wet or computational experiments: MB, PV, DC and NI. Analyzed data: MB, NI and TW. Drafting the manuscript: MB and MA. Generated figures and tables: MB and MA. Wrote and critically reviewed the manuscript: all authors.

#### ACKNOWLEDGMENTS

Part of the research described in this study was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. This research was funded by Planetary Protection Research program in NNH11ZDA001N, ROSES 2011 awarded to PV and NI. The authors are grateful to Drs. Catharine Conley and Ying Lin for valuable discussion and input. We would like to thank Dr. Kasthuri Venkateswaran (JPL) for making archived DNA available. MB is thankful for the financial support of the FWF Austrian Science Fund (W1241) and the Medical University of Graz through the PhD Program Molecular Fundamentals of Inflammation (DK-MOLIN), as well as the Bank Austria Visiting Scientist Program of the Medical University of Graz.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb. 2016.01321


**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 © 2016 Bashir, Ahmed, Weinmaier, Ciobanu, Ivanova, Pieber and Vaishampayan. 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) or licensor 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.