# UNDERSTANDING AND EXPLOITING HOST-COMMENSAL INTERACTIONS TO COMBAT PATHOGENS

EDITED BY : Sudhanshu Shekhar, Fernanda Cristina Petersen and Xi Yang PUBLISHED IN : Frontiers in Immunology

#### Frontiers eBook Copyright Statement

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

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

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

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

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

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

ISSN 1664-8714 ISBN 978-2-88963-336-4 DOI 10.3389/978-2-88963-336-4

#### About Frontiers

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

#### Frontiers Journal Series

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

#### Dedication to Quality

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

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

#### What are Frontiers Research Topics?

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

# UNDERSTANDING AND EXPLOITING HOST-COMMENSAL INTERACTIONS TO COMBAT PATHOGENS

Topic Editors: Sudhanshu Shekhar, University of Oslo, Norway Fernanda Cristina Petersen, University of Oslo, Norway Xi Yang, University of Manitoba, Canada

Trillions of commensal microbes reside on and inside the human body, including the intestinal and respiratory tracts, which encompass various microbial taxa, such as bacteria, fungi, archea and viruses. The close proximity of microbes with the host provides an opportunity to continually interact with each other. Advances in research approaches, including high-throughput sequencing, have allowed us to study host and commensal microbes at the genetic and functional levels. Recent studies suggest that commensal microbes play a crucial role in the development of the host's immune system and induce innate and adaptive immune responses against pathogens. It is also becoming apparent that the gut commensals are endowed with a capacity to alter immune responses in organs beyond the intestine, such as the lungs, highlighting the significance of the gut commensals in controlling systemic immunity. On the other hand, the host's immune system possesses the ability to shape the repertoire of commensal microbes and contribute to the establishment of beneficial relationships with them. A better understanding of host-commensal interactions will be important for designing effective vaccines and therapeutics against pathogens.

This Research Topic sheds light on our current understanding of the interplay between the host's immune system and commensal microbes and how this interplay can be exploited for prophylactic and/or therapeutic strategies against pathogens.

Citation: Shekhar, S., Petersen, F. C., Yang, X., eds. (2020). Understanding and Exploiting Host-Commensal Interactions to Combat Pathogens. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-336-4

# Table of Contents

*05 Editorial: Understanding and Exploiting Host-Commensal Interactions to Combat Pathogens*

Sudhanshu Shekhar, Fernanda Cristina Petersen and Xi Yang

*08 Chronic Mucocutaneous Candidiasis in Autoimmune Polyendocrine Syndrome Type 1*

Linda Humbert, Marjorie Cornu, Emmanuelle Proust-Lemoine, Jagadeesh Bayry, Jean-Louis Wemeau, Marie-Christine Vantyghem and Boualem Sendid

*17 Exploiting the Oral Microbiome to Prevent Tooth Decay: Has Evolution Already Provided the Best Tools?*

Jonathon L. Baker and Anna Edlund


Xiao-Li Luo, Jian-Xiong Li, Hua-Rong Huang, Jie-Lin Duan, Ruo-Xuan Dai, Ru-Jia Tao, Ling Yang, Jia-yun Hou, Xin-Ming Jia and Jin-Fu Xu


Lu Xiao, Baolin Chen, Di Feng, Ting Yang, Tingyu Li and Jie Chen


Hong-Yu Cheng, Meng-Xia Ning, De-Kun Chen and Wen-Tao Ma

*118 Microbiota Inhibit Epithelial Pathogen Adherence by Epigenetically Regulating C-Type Lectin Expression* Vivienne Woo, Emily M. Eshleman, Taylor Rice, Jordan Whitt, Bruce A. Vallance and Theresa Alenghat

*128 Commensal Bacteria: An Emerging Player in Defense Against Respiratory Pathogens*

Rabia Khan, Fernanda Cristina Petersen and Sudhanshu Shekhar


Andrew J. Forgie, Janelle M. Fouhse and Benjamin P. Willing

*177 Infectious Threats, the Intestinal Barrier, and its Trojan Horse: Dysbiosis* Simona Iacob and Diana Gabriela Iacob

# Editorial: Understanding and Exploiting Host-Commensal Interactions to Combat Pathogens

Sudhanshu Shekhar <sup>1</sup> \*, Fernanda Cristina Petersen<sup>1</sup> and Xi Yang<sup>2</sup>

*<sup>1</sup> Faculty of Dentistry, Institute of Oral Biology, University of Oslo, Oslo, Norway, <sup>2</sup> Department of Immunology, Max Rady College of Medicine, University of Manitoba, Winnipeg, MB, Canada*

Keywords: commensal, host, immunity, pathogen, vaccine, therapeutics

**Editorial on the Research Topic**

#### **Understanding and Exploiting Host-Commensal Interactions to Combat Pathogens**

The human body harbors an astonishing number of diverse commensal microbes, including bacteria, fungi, and viruses, providing a suitable milieu for microbial growth and multiplication (1, 2). These microbes rarely cause disease and communicate with the host in a mode that is advantageous to both host and microbes (3, 4). In recent years, phylogenetic, metagenomic, and functional studies have been conducted to better understand the complexities of the microbial genome and the effect of the microbiota on the host's immunophysiology. Accumulating evidence has shown that gut commensals regulate the ontogeny and function of the immune system, and contribute to shaping the outcome of immune responses (5). It has also become clear that these commensals have the ability to influence the immune responses at extraintestinal tissues/organs, underscoring their profound impact on local as well as systemic immunity (6). However, alterations in composition, diversity, and metabolic activities of commensal microbes can lead to dysbiosis, which may have detrimental consequences, such as autoimmunity, allergy, asthma, inflammatory bowel disease, cancer, and infection (7). Commensals also hold the potential to cause disease depending on multiple microbial and host factors (8). A much deeper understanding of how commensals communicate with the host is crucial for developing new strategies to prevent and treat diseases.

In this Research Topic, a series of 16 articles, encompassing review, original research, and general commentary articles, provide crucial information on how the interplay between host and commensals takes place and how this could be exploited for designing novel prophylactics/therapeutics against a wide spectrum of disorders, including infectious diseases. In an original research article, Yang et al. report the identification of genetic factors that are involved in the lysis of human neutrophils by Staphylococcus aureus, underscoring the mechanism by which this commensal bacterium evades the neutrophilic immune barrier during infection. Likewise, Dai et al. demonstrate that S. aureus employs the vancomycin resistance-associated sensor/regulator (VraSR) to increase its survival within macrophages, thereby modulating the process of host-cell autophagy. Overall, these data throw light on important virulence factors used by S. aureus to escape innate immunity, highlighting why the innate immune response is incapable of eradicating S. aureus. Gao et al. provide evidence that underscores the impact of hydrogen peroxide (H2O2) produced by Streptococcus pneumoniae on the host's immune responses against pneumococcal

#### Edited and reviewed by:

*Ian Marriott, University of North Carolina at Charlotte, United States*

\*Correspondence:

*Sudhanshu Shekhar sudhanshu.shekhar@odont.uio.no*

#### Specialty section:

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

Received: *14 August 2019* Accepted: *25 October 2019* Published: *12 November 2019*

#### Citation:

*Shekhar S, Petersen FC and Yang X (2019) Editorial: Understanding and Exploiting Host-Commensal Interactions to Combat Pathogens. Front. Immunol. 10:2645. doi: 10.3389/fimmu.2019.02645* lung infection. S. pneumoniae-secreted H2O<sup>2</sup> causes damage and leakage of mitochondrial DNA into the cytoplasm, which not only mediates mitochondrial oxidative stress but also promotes IFN-I cascades in lung cells (Gao et al.). Furthermore, Tian et al. show that gut microbiome-derived propionate levels are inversely proportional to the lung inflammation, but not to bacterial immunity, using mouse models of ischemia reperfusion injury and S. aureus pneumonia. These studies indicate that microbial products, such as H2O<sup>2</sup> and propionates, play a significant role in the outcome of host immunity.

Luo et al. describes the inhibitory effect of human cathelicidin antimicrobial peptide LL37 on Aspergillus fumigatus infection in mice by directly binding to the fungal mycelia, which follows reduced pulmonary inflammation characterized by decreased histopathological changes and proinflammatory cytokine levels. Woo et al. show a novel mechanism by which the commensal microbiota epigenetically regulate intestinal epithelial cells to downregulate expression of the cell surface glycoprotein C-type lectin 2e (Clec2e), which reduces the efficiency of epithelial cell interaction with the murine enteric pathogen Citrobacter rodentium. On the other hand, host immunity can also regulate the composition of the mucosal-associated microbiota. Xiao et al. demonstrate that Toll-like receptor 4 (TLR4), which is a membrane-bound protein expressed on immune cells that identifies microbe-associated molecular patterns, may play a role in the regulation of the distribution and structure of the intestinal mucosal-associated microbiota by vitamin A. Thus, there exists a bidirectional communication between the microbiota and host, which may contribute to maintaining homeostasis in the gut.

The review and commentary articles presented in this Research Topic focus on a variety of interesting areas that include commensal-immune cell interaction and its implications for therapy and prophylaxis of diseases. Humbert et al. review the current literature on the pathophysiology of chronic mucocutaneous candidiasis, which is caused by the fungus Candida albicans, in autoimmune polyendocrine syndrome type 1, whereas Iacob and Iacob focus on the importance of the relationship between the intestinal barrier and microbiota, and how dysbiosis can alter this relationship to make the host prone to pathogens. Additionally, it is becoming clearer that the gut microbiota modulates the innate and adaptive immune responses to impact the disease outcome. Cheng et al. shed light on the major mechanisms by which the commensal microbiota boosts the host's innate immunity against infectious agents. Pandiyan et al. review recent evidence on how the gut microbiota influences the function of adaptive immune cells, such as regulatory T cells (Tregs) and Th17 cells, and how the microbiota can be targeted to promote mucosal immunity and ameliorate pathology. Natural killer T (NKT) cells constitute an innate and unconventional population of T cells that perform protective as well as detrimental roles in diverse disease models. A general commentary by Jia focuses on the crucial role played by NKT cells in liver cancer, which is regulated by the gut microbiota-mediated bile acid metabolism.

An interesting review by Forgie et al. discusses the hostmicrobiota interactions from a dietary point of view. This provides knowledge on how dietary components, such as carbohydrates and proteins, can modulate the hostmicrobiota interactions to promote resistance against pathogens (Forgie et al.) Studies on the therapeutic applications of the hostmicrobiota dialogue are crucial for ensuring further translation of the acquired knowledge into health benefits. Li et al. provides a comprehensive review on the beneficial and harmful role of the commensal microbiota in dealing with viral infections and the effect of these infections on the microbiota homeostasis. Khan et al. evaluate emerging data on the contribution of commensal bacteria to host defense against respiratory pathogens and the mechanisms whereby bacteria induce protective immunity. They also discuss how commensal bacteria can be exploited to treat and prevent respiratory infections (Khan et al.). In line with this, Baker and Edlund discuss the therapeutic potential of the oral microbiome in developing strategies to exploit the protective effects of the oral microflora in order to prevent dental caries.

Cumulatively, this Research Topic provides significant knowledge on the mechanisms underlying host and commensal microbe interactions, and the profound impact that these microbes exert on the host's health and disease. This has clinical implications for the prevention and treatment of diseases. The targeting of commensals is gaining momentum as an effective strategy to combat various diseases, including infectious diseases. Successful treatment of severe intestinal infections, caused by antibiotic-resistant bacterial pathogens, using fecal microbiota transplantation, offers an excellent example of how commensal microbes can be used for disease therapy. It is, however, notable that most data on this topic stem from mouse studies. Although mouse experiments remain critical to understand the contribution of the commensal microbiota to health and disease, they pose a concern for the scientific community because of their poor recapitulation of human conditions. For example, the murine and human intestinal microbiota exhibit significant differences in abundance and gene identity, and the murine microbiota composition depends on multiple factors such as rearing facilities and genetic background (9). It is important to consider these facts while translating the knowledge acquired from mouse to human.

Since several commensals/probiotics have been shown to be safe in animal models and humans, it would be worth examining the long-term consequences of their use for the host's well-being in toto. With technological advances in this field, approaches that harness the beneficial effects of commensals to prevent diseases and promote health will continue to grow in number. In-depth studies are also needed to focus on the pathogenic potential of commensals/probiotics in individuals under immunosuppression due to malnutrition, chemotherapy, or viral infections. This is crucial because immunocompromised individuals possess an altered microbiota, along with an impaired immune system, which make them highly susceptible to opportunistic infections and cancer.

Finally, we are extremely grateful to all the authors for their significant contribution to this Research Topic as well as to the reviewers for taking the time to review the submitted manuscripts.

# AUTHOR CONTRIBUTIONS

All authors wrote the editorial manuscript, and approved it for publication.

#### REFERENCES


#### FUNDING

SS and FP received the funding from the Norwegian Research Council (241011), whereas XY received financial support from the Canadian Institutes of Health Research (MOP-130423).


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

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

# Chronic Mucocutaneous Candidiasis in Autoimmune Polyendocrine Syndrome Type 1

Linda Humbert <sup>1</sup> , Marjorie Cornu2,3, Emmanuelle Proust-Lemoine<sup>4</sup> , Jagadeesh Bayry <sup>5</sup> , Jean-Louis Wemeau<sup>1</sup> , Marie-Christine Vantyghem1,6,7 and Boualem Sendid2,3 \*

<sup>1</sup> Department of Endocrinology and Metabolism, CHU Lille, Lille, France, <sup>2</sup> Department Parasitology-Mycology, CHU, Lille, France, <sup>3</sup> Inserm, U995-LIRIC, Fungal Associated Invasive & Inflammatory Diseases, Lille, France, <sup>4</sup> Department of Endocrinology, Polyclinique Aguilera, Biarritz, France, <sup>5</sup> Inserm, Center de Recherche des Cordeliers, Sorbonne Université, Université Paris Descartes, Sorbonne Paris Cité, Paris, France, <sup>6</sup> UMR 1190, Translational Research in Diabetes Inserm, Lille, France, <sup>7</sup> European Genomic Institute for Diabetes, Univ Lille, Lille, France

#### *Edited by:*

Sudhanshu Shekhar, University of Oslo, Norway

#### *Reviewed by:*

Jeanette Wagener, University of Aberdeen, United Kingdom Marc Swidergall, UCLA Department of Medicine, United States Desa Lilic, Newcastle University, United Kingdom

#### *\*Correspondence:*

Boualem Sendid boualem.sendid@univ-lille.fr

#### *Specialty section:*

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

*Received:* 24 August 2018 *Accepted:* 18 October 2018 *Published:* 19 November 2018

#### *Citation:*

Humbert L, Cornu M, Proust-Lemoine E, Bayry J, Wemeau J-L, Vantyghem M-C and Sendid B (2018) Chronic Mucocutaneous Candidiasis in Autoimmune Polyendocrine Syndrome Type 1. Front. Immunol. 9:2570. doi: 10.3389/fimmu.2018.02570 Autoimmune polyendocrinopathy candidiasis ectodermal dystrophy (APECED) is an autosomal recessive disease caused by mutations in the autoimmune regulator (AIRE) gene, characterized by the clinical triad of chronic mucocutaneous candidiasis (CMC), hypoparathyroidism, and adrenal insufficiency. CMC can be complicated by systemic candidiasis or oral squamous cell carcinoma (SCC), and may lead to death. The role of chronic Candida infection in the etiopathogenesis of oral SCC is unclear. Long-term use of fluconazole has led to the emergence of Candida albicans strains with decreased susceptibility to azoles. CMC is associated with an impaired Th17 cell response; however, it remains unclear whether decreased serum IL-17 and IL-22 levels are related to a defect in cytokine production or to neutralizing autoantibodies resulting from mutations in the AIRE gene.

Keywords: autoimmune polyendocrinopathy candidiasis ectodermal dystrophy (APECED), chronic mucocutaneous candidiasis (CMC), autoimmune regulator (*AIRE*) gene, IL-17, IL-22

# INTRODUCTION

Autoimmune polyendocrine syndrome type 1 (APS-1), also known as autoimmune polyendocrinopathy candidiasis ectodermal dystrophy (APECED) syndrome, is characterized by the clinical triad of chronic mucocutaneous candidiasis (CMC), hypoparathyroidism, and adrenal insufficiency. This syndrome was formerly known as Whitaker syndrome (1). Accurate diagnosis of this syndrome requires the presence of at least two of these three major components, or only one if a sibling has already been diagnosed with the disease (2). Other autoimmune disorders have also been described, such as hypergonadotrophic hypogonadism, thyroid disease, type 1 diabetes, coeliac disease, liver disease, alopecia, vitiligo, chronic atrophic gastritis, and hypophysitis. These autoimmune disorders are associated with ectodermal dystrophy, asplenia, and the presence of several autoantibodies, even in the absence of corresponding organ dysfunction (3).

APS-1 is a monogenic, autosomal, recessive disease caused by a mutation in the autoimmune regulator (AIRE) gene on chromosome 21 (gene map locus 21q22.3) (4). The AIRE gene is composed of 14 exons and codes for a 545 amino acid protein (5, 6). The AIRE gene is mainly expressed in thymic medullary epithelial cells, which play an important role in the presentation of self-antigens (7, 8), but is also expressed at low levels in the spleen, lymph nodes, pancreas, adrenal cortex, and peripheral blood mononuclear cells. The AIRE gene codes for a nuclear

**8**

transcriptional regulator protein involved in the ectopic expression of self-antigens in the thymus, leading to the removal of self-reactive thymocytes and generation of peripheral tolerance. The role of peripheral AIRE expression, which has been confirmed by mRNA analysis, remains unclear. To date, more than 100 different mutations in this gene, both homogeneous and heterogeneous, have been reported worldwide (9–12).

APECED is a rare syndrome, which has been reported worldwide, but is more prevalent in some historically-isolated homogeneous populations in Finland (1/25000) (4, 13), Sardinia (1/14500) (14), and Iranian Jews (1/9000) (15). APECED is also seen at a lower incidence in Norway, Sweden, Slovenia, Great Britain, Italy, Ireland, and North America (16–21).

Patients with APS-1 suffer from CMC without displaying susceptibility to any other pathogen. CMC is associated with the Finnish mutation c.769C>T (p.Arg257stop) (22). CMC usually affects the oral mucosa, but the nails and skin may also be involved. Esophageal candidiasis results in pain and dysphagia. CMC can be complicated by systemic candidiasis or oral squamous cell carcinomas (SCCs), and may lead to death (23, 24).

#### CHRONIC MUCOCUTANEOUS CANDIDIASIS

CMC is characterized by recurrent or persistent symptomatic mucocutaneous infections caused by Candida species, predominantly Candida albicans, affecting the nails, skin, oral cavity, and genital mucosa. The diagnosis of CMC is based on clinical symptoms, associated in most cases with the isolation of Candida from body sites (25). The first case of syndromic CMC was described by Thorp and Handley (26).

C. albicans is a ubiquitous, diploid, dimorphic yeast that resides as a commensal organism on the mucosae and in the gastrointestinal tract of healthy individuals. Mucosal candidiasis results from a change in mucosal homeostasis leading to disequilibrium between the yeast and its host. Opportunistic mucosal infection, deep organ, or systemic infection in immunocompromised patients usually arise from Candida colonizing the digestive tract (27). Systemic candidiasis can be diagnosed using a number of non-culture based assays (28) but no biological markers are currently available for the diagnosis of "culture-negative" CMC.

Most cases of CMC are sporadic and are secondary to other medical conditions such as HIV infection with T-cell deficiency, diabetes, immunosuppressive therapies like anticytokine blockers, antibiotic or steroid therapy (25, 29, 30). CMC is also more rarely favored by genetic disorders (i.e., familial CMC) that can be inherited. These cases of primary CMC are due mainly to innate immunodeficiency disorders. They have been reviewed by Puel et al. (31) and were classified as primary immunodeficiency disease by the International Union of Immunological Societies Expert Committee for Primary Immunodeficiency in 2015 (32).

Three types of immnuodeficiency can be distinguished depending on the genetic abnormalities associated with CMC:


CMC and invasive Candida infection have also been associated with a deficiency in autosomal recessive caspase recruitment domain-containing protein 9 (CARD9) (39–41). CARD9 is an intracellular adaptor involved in Dectin 1 and Dectin 2 signaling, the main pathogen recognition receptors for C. albicans glycans. The number of circulating IL-17-producing cells and IL-17 secretion have been reported to be decreased in CARD9-deficient patients (40, 42, 43). However, these findings remain open to debate (33). With regard to Candida infections, mutations in Dectin-1 and Dectin-2 have also been studied in murine models. In humans, a mutation in the early-stop codon for Dectin-1 (Y238X) has been reported in a family with recurrent vulvovaginal candidiasis. Experiments in vitro demonstrated that monocytes and neutrophils from homozygous patients lacking Dectin-1 expression are defective in cytokine production, including IL-17, when stimulated with C. albicans. However, phagocytosis and yeast cell killing remained normal (44, 45). Another report demonstrated that patients receiving hematopoietic stem cell transplants who were heterozygous for Y238X had an increased incidence of gastrointestinal Candida colonization (46). Although Dectin-2-deficient mice had higher mortality and a higher kidney fungal burden after infection with C. albicans (47), the impact of Dectin-2 mutations on the human host response to C. albicans infection remains unclear.

Other studies have demonstrated that patients with IL-12 or IL-23 signaling defects have an increased risk of developing CMC (31, 33, 48). Patients with CMC, autoimmune manifestations, other mild bacterial or viral infections, intracranial aneurysms, or SCC also had heterozygous missense gain-of-function (GOF) mutations of STAT1. The development of IL-17-producing T-cells is impaired in these patients, as a result of hyperactivity of STAT1 that inhibits STAT3 signaling. This phenomenon results in increased STAT1-dependent cellular responses that repress IL-17 producing T-cell responses, such as IFN-γ, and/or enhanced IL-6, IL-21, and IL-23 STAT1 responses, which normally activate STAT3 and induce IL-17 T-cell production (31).

(iii) Isolated CMC has also been described in subjects with IL-17RA, IL-17RC, ACT1, IL-17F, and RORγt deficiency, where CMC is the only presenting feature of the disease. Some families have also been identified with autosomal-dominant mutations in the gene coding for IL-17F, or autosomalrecessive mutations in the gene coding for IL-17 receptor A or IL-17 receptor C as risk factors for CMC (31–33).

These familial cases of CMC demonstrate that IL-17 plays a pivotal role in human epithelial immunity to C. albicans. Another piece of evidence supporting the central role of Th17 cytokines in CMC and mucosal immunity to yeasts is the observation of cases of CMC in the Phase 2 trial of secukinumab, a human anti-IL17 receptor antibody for the treatment of Crohn's disease (49, 50). This review aims to focus on the characteristics of familial CMC associated with APECED syndrome.

#### CHRONIC MUCOCUTANEOUS CANDIDIASIS AND APECED

#### Clinical Description

CMC is the most common infection occurring in APECED patients (77–100%) (19, 21, 23, 51–53), except in Iranian Jews (17%) (15). CMC is also the most common first clinical manifestation of APECED syndrome (40–93%) (20, 23, 52, 54, 55). Median age at diagnosis is usually <5-years-old (1.0– 6.5years) (20, 21, 23, 54, 55). According to the Finnish series, one-sixth of patients had developed CMC by 1.0 year, half by 5 years, 70% by 10 years, 94% by 20 years, and 97% by 30 years of age (56).

The clinical course of CMC varies from periodic to chronic, and its severity varies between individuals. The oral cavity was involved in 100% of patients in the Finnish cohort (23). In the Norwegian cohort, 40% of patients had angular cheilitis (53). In the mild oral form, CMC causes ulceration, redness, and soreness of the corners of the mouth. In more severe cases, the entire mouth is involved making it impossible to consume acidic or spicy foods. In the hyperplastic form, the tongue and buccal mucous membranes are covered by white or gray plaques and hyperkeratosis. In the atrophic form, the mucosa is erythematous and may be speckled with areas of leukoplakic or nodules (56). Candida onychomycosis is often associated with mucosal Candida lesions in childhood and is very difficult to eliminate (56). CMC affected the nails in 72% of patients in the Irish cohort of Collins et al. (52), and less frequently the skin (10– 17%) (21, 23). Esophageal CMC occurred in 5–22% of patients in a European series (23, 52, 53), and in 51% of patients in a recent American study (21). Esophageal candidiasis often occurs without the typical form or symptoms of oral candidiasis, and can be complicated by substernal pain, dysphagia, and stenosis. Esophageal stenosis requires endoscopic dilation (23, 52–54). In the digestive tract, CMC can cause abdominal pain, flatulence, and diarrhea, which may be severe. Symptomatic intestinal candidiasis may also be present in the absence of oral disease (56).

In some patients, CMC may also be complicated by systemic candidiasis, although evidence is lacking that dissemination occurs from the oral cavity. Systemic candidiasis is very rare, even in APECED patients, and is frequently associated with immunosuppressive therapy. In an Italian cohort of 41 patients, one patient died from systemic candidiasis after the onset of immunosuppressive treatment (54), and in a French cohort, one patient died from systemic candidiasis after the onset of immunosuppressive treatment for large granular lymphocytic leukemia (57). In a Finnish cohort, one patient developed an abscess on the pericardium and small intestine (23). One isolated case of chronic Candida otitis has been described (53).

#### Genotype-Phenotype Correlation

The prevalence of CMC is reportedly higher in patients with the major Finnish AIRE mutation R257X than with other mutations (22). The prevalence is <20% in Iranian Jews affected by the Y85C mutation (15). Kisand et al. studied 160 APECED patients with the most severe mutations, R257X (Finnish) and R139X (Sardinian) and the Norwegian mutation, 967-979113 In contrast to the study of Puel et al. (58), CMC was less prevalent in patients with the homozygous mutation 967-979113 than with the other two mutations (59).

#### Squamous Cell Carcinoma and CMC

CMC has been reported to be involved in carcinogenesis as cancer often develops at the site of fungal lesions. Several cases of oral carcinoma have been described in association with CMC of the oral cavity and esophagus, suggesting that oral candidiasis may be carcinogenic. The most common morphological entity of these cancers is SCC (29, 60, 61). In patients with chronic Candida infection, oral CMC is often associated with esophageal cancer (62), and good clinical practice should include regular monitoring, every 2–3 years, by endoscopy. CMC is induced by immunosuppressive therapy rather than SCC itself.

The first report of oral SCC associated with APECED syndrome was published in 1975 in a patient who died of metastatic disease at 27-years of age (63). A case report published in 2010 recorded the presence of three separate oral SCCs in a 40-year-old woman with APECED (61). In the Finnish case series, 6/91 APECED patients >25-years of age (10.5%) developed SCC, four died from the disease, and two developed oral colonization with Candida with decreased susceptibility to azole antifungals. One patient developed SCC without symptoms of oral candidiasis (23, 60). In the Norwegian cohort, 3/52 patients (6%) developed SCC at an early age (53). SCC was not reported in a recent American case series (21).

In the general population, mean age at diagnosis for oral and esophageal SCC is >62-years, and the disease is uncommon in young adults (64). The patients diagnosed with oral SCC in the Finnish cohort were between 29- and 44-years of age (23). The high rate of oral SCC in young patients with APECED demonstrates the possible carcinogenic potential of C. albicans when associated with the specific immunodeficiencies characteristic of this syndrome. Therefore, appropriate screening and adequate management of the infection and areas of oral dysplasia is necessary to reduce the risk of malignancy. Each erosive, ulcerated lesion should be biopsied, and each dysplastic lesion should be treated surgically (65).

In the general population, there are many risk factors for oral and esophageal SCC. Extrinsic factors include alcohol, tobacco, betel quid, immunosuppression, radiation, oncogenic viruses (human papilloma virus), and Candida infection, while intrinsic factors include immune defects, iron or vitamin A deficiency, malnutrition, and defects in tumor suppressor genes (64, 66, 67). Of the six APECED patients who developed SCC in the Finnish cohort, four were smokers and one had received immunosuppressive therapy (23). Therefore, extrinsic factors for SCC should be controlled in APECED patients as well as in CMC.

The role of chronic Candida infection in the etiopathogenesis of oral SCC is unclear. Possible mechanisms by which oral Candida infection might contribute to cancer development include: (i) metabolism of procarcinogens (such as the conversion of ethanol to acetaldehyde by Candida); (ii) production of carcinogens (such as the production of nitrosamine by Candida species); and (iii) induction of chronic inflammation, with the production of cytokines that enhance cell proliferation and inhibit apoptosis (24, 64, 65, 68, 69).

#### *C. albicans* and Decreased Susceptibility to Antifungal Treatment

Oropharyngeal candidiasis (mainly C. albicans and C. glabrata) is the most common fungal infection in patients with human immunodeficiency virus (HIV), and long-term use of azoles in this population has been reported to cause loss of susceptibility of C. albicans to fluconazole (70, 71). Candidiasis is usually caused by the yeast C. albicans in APECED patients, unlike in HIV patients (23, 52, 72–75). In a group of Finnish patients with APECED, non-C. albicans spp. were reported in only 7/56 patients (12.5%) (72). Because of the high prevalence of CMC in APECED patients and the risk of SCC, lifelong management of candidiasis with antifungal treatment is necessary (76). Topical intermittent treatment is more frequently prescribed than systemic antifungals, which are restricted to periods of severe symptoms and systemic candidiasis. Unlike in HIV patients, APECED patients with CMC treated with fluconazole have a high risk of becoming colonized with C. albicans with decreased susceptibility to fluconazole (76). Emergence of resistance does not appear to be a problem during short-term use, as shown in 43 isolates of C. albicans from 23 Finnish APECED patients.

Resistance seems to be correlated to the number of years of antifungal drug use, and is mainly due to the use of triazoles. Rautemaa et al. identified C. albicans isolates with decreased susceptibility to fluconazole in 11/56 (20%) Finnish patients (72). In the Irish cohort study of 16 patients with APECED, McManus described 11/16 (69%) patients with clinical signs of oral Candida infection and oral Candida isolates were recovered from 12/16 (75%) patients. Surprisingly, clinical signs suggestive of candidiasis did not always correlate with microbiological evidence of infection, and yeasts were isolated from only 32% of patients. The susceptibility of sequentially recovered isolates to fluconazole and itraconazole was examined; 11/16 (69%) patients had received antifungal treatment with fluconazole or itraconazole. Four of these 11 patients (36%) had isolates that were resistant or had reduced susceptibility to azoles (74).

Multiple mechanisms could lead to azole resistance in C. albicans. Some genes of C. albicans involved in ergosterol synthesis can be upregulated and in this way confer resistance to azoles, in particular the ERG11 gene and CDR1, CDR2, and MDR1 encoded efflux pumps. In the study of McManus, acquisition of such mutations was revealed in C. albicans strains isolated during the longitudinal follow-up of patients. C. albicans strains recovered sequentially from nine Finnish patients were studied. The major molecular mechanisms leading to azole resistance were GOF mutations in TAC1, contributing to overexpression of CDR1 and CDR2, point mutations in ERG11, and six new TAC1 mutations were detected (77). To avoid the emergence of azole resistance, McManus et al. proposed only prescribing azole therapy in patients with mycologicallydocumented Candida infection.

#### How Should CMC Be Managed in APECED Patients?

Husebye et al. suggested that oral infection should be strictly controlled to prevent the development of cancer. These authors highlighted the importance of having good oral hygiene with abstention from smoking and excessive alcohol consumption, and to avoid eating acidic or spicy food, and toothpastes containing powerful whitening agents. Meticulous oral hygiene is recommended in CMC patients, using both toothpaste and chlorhexidine solution, at bedtime and long-term, with the continual use of two topical polyene drugs. In line with this recommendation, patients should hold 1–2 mL of nystatin suspension in their mouth for several minutes and then allow an amphotericin B lozenge to dissolve without chewing. Both drugs should be spread to every part of the mouth with the tongue and finally swallowed. This regimen should be continued for 4–6 weeks or for at least 1 week after the resolution of symptoms. The use of these two polyene antifungal drugs is important in clinical practice and both are well-tolerated due to a lack of absorption from the gut and a low rate of yeast resistance to these drugs.

Azole agents should be restricted to 2–3 courses per year to avoid decreased susceptibility. This treatment can be followed by prophylactic treatment consisting of 1 week of a polyene antifungal every 3 weeks and 1 week of chlorhexidine mouth rinse twice a day, if CMC becomes recurrent. Prophylactic treatment should be administered more frequently if symptoms persist, and even become daily treatment. In the case of failure of antifungal therapy, it is necessary to alert microbiologists to the possibility of a mixed infection and to use chromogenic media to detect mixed species, which would not be picked up with classical Sabouraud glucose agar. It is also recommended that all species isolated undergo antifungal susceptibility testing so that treatment can be adapted accordingly. Angular cheilitis should be treated by application of natamycin, amorolphine hydrochloride cream, or chlorhexidine gel several times a day, continuing for 4–5 days after the corners of the mouth have healed. Miconazole gel can also be used, while bearing in mind the risk of selection of azole-resistant strains. In order to avoid CMC relapses, antifungal treatment should be prescribed simultaneously for oral candidiasis and angular cheilitis, and biopsy of any lesion should be considered if mucositis with ulceration fails to respond to treatment within 2 weeks. Candida esophagitis and digestive CMC with diarrhea should also be treated with the same drug regimen for 1 or 2 weeks. If esophagitis persists, esophagoscopy and mycological sampling should be performed. The use of highdose fluconazole (200–300 mg once a day for 1 week) must be restricted to severe cases and failure of topical therapy. Vaginal Candida infection should be treated with a short course of vaginal fluconazole, while fingernail candidiasis is very hard to eliminate and systemic medication is often necessary for 6 weeks. In general terms, prolonged intravenous (IV) antifungal therapy may be required and administered after obtaining expert medical advice (56, 75).

Adjunct immunotherapy with normal pooled immunoglobulin or IV immunoglobulin G (IVIG) in combination with antifungal agents is another treatment option. IVIG is a pooled IgG therapeutic preparation obtained from the plasma of several thousand healthy donors. In addition to its safety profile, IVIG could benefit CMC-associated APECED as a result of multiple mechanisms, including a reduction of inflammation by targeting various arms of the immune system, inhibition of autoantibody production by B-cells, and B-cell anergy (78). Further, antibodies in IVIG to fungal antigens may also help in pathogen neutralization (79, 80). As high-dose IVIG (1–2 g/kg) is known to inhibit Th17 responses in vitro (81, 82) and in vivo, including experimental allergic bronchopulmonary aspergillosis (78, 83), studies are required regarding appropriate dose of IVIG for CMC-associated APECED.

#### Pathophysiology of CMC Associated With APECED

The most prevalent autoantibodies detected in APECED patients are those neutralizing cytokines, especially type I IFNs and

FIGURE 1 | Familial CMC and APECED syndrome. APECED syndrome is characterized by the association of endocrine autoimmune disorders (such as hypoparathyroidism, hypothyroidism, adrenocortical insufficiency, and gonadal failure), non-endocrine autoimmune disorders (such as autoimmune hepatitis, celiac disease, and chronic atrophic gastritis), and chronic mucocutaneous candidiasis (CMC). These manifestations are related to the presence of tissue specific antibodies and cytokine antibodies. IL-17 mediated immunity is represented by cooperation between cells recognizing C. albicans (phagocytes and epithelial cells) and IL-17 cytokine-producing cells (T-cells). On C. albicans recognition by PRRs (pathogen recognition receptors; including Dectin-1, Dectin-2, or Mincle), the adaptor molecule CARD9 mediates the induction of pro-inflammatory cytokines, such as IL 1β, IL-6, and IL-23. On binding to their receptors expressed on T-lymphocytes, pro-inflammatory cytokines, such as IL-6 or IL-23, activate T-lymphocytes via the transcription factor STAT3 resulting in their differentiation into IL-17-producing T-cells. Genes in which mutations are associated with CMC are indicated in pink: dectin 1, CARD 9, STAT 1, STAT3, IL22, and IL17F. IL-17RA and IL-12RB1 are not represented. Y designates cytokine-neutralizing autoantibodies that develop in AIRE-deficient (APECED and rare thymoma cases) patients.

TH17-related cytokines, with a prevalence of 100% for IFNω (84, 85) and >90% for IL-22. IFN-ω autoantibodies are also found in thymomas (86), with a high rate of specificity and sensitivity. These autoantibodies are highly disease specific. Antibodies neutralizing IFN-ω are not correlated with CMC, and high titers of anti-IFN-ω antibodies are found in patients without CMC. Antibody titers in patients with APECED are almost always higher at diagnosis and persist for decades, representing a reliable biomarker for APECED syndrome (84, 85). Neutralizing autoantibodies against IL-17A, IL-17F, and IL-22 are also present at diagnosis. Autoantibodies against IL-22 and IL-17F seem to be more prevalent in APECED patients with CMC than in those without CMC, suggesting that type Th17 cytokines are central in human epithelial immunity against Candida infection (**Figure 1**). In contrast to the study of Kisand et al. (59) who reported

that autoantibodies neutralizing IL-22 and IL-17F (but not those against IL-17A) were correlated with CMC in a study of 162 APECED patients, a study by Sarkadi et al. (87) reported high levels of autoantibodies against IL-17A in APECED patients with severe CMC. However, recent experimental data have shown that autoantibodies against IL-17A that develop in older AIREdeficient mice do not confer susceptibility to oropharyngeal candidiasis, while monoclonal antibodies that cross-react with murine IL-22 derived from patients increase the mucosal fungal burden (88).

Occasional or weakly binding autoantibodies against IL-6, IL-9, IL-12, IL-21, IL-23, IL-26, IL-29, and RANTES have been reported in APECED patients in addition to autoantibodies against Th17 cytokines; however, their role in the development of CMC has not been demonstrated clearly (59)**.**

The production of type I IFNs by dendritic cells is not impaired in APECED patients (89). In vitro, IL-22 and IL-17F production by peripheral blood mononuclear cells of APECED patients, stimulated by Candida antigens or polyclonal stimuli, was decreased in patients with CMC (59, 90, 91). The production of IL-17A was not impaired and even appeared to be increased (90, 91). The pathogenesis of CMC is believed to be associated with impaired Th17 cell responses, similar to several other primary immunodeficiencies associated with CMC. Th17 cytokines (IL-17A, IL-17F, and IL-22) influence epithelial cells by inducing the production of chemokines and antimicrobial peptides that exert direct antifungal activity. Additionally, IL-22

#### REFERENCES


promotes epithelial barrier integrity, especially in synergy with TNF-α co-secreted by Th22 cells (92). Moreover, the production of IL-22 is severely impaired by skin-populating T-cells from APECED patients (93).

#### CONCLUSION

In APECED patients, CMC is associated with an impaired Th17 cell response. However, it remains unclear whether decreased serum IL-17 and IL-22 levels are related to a defect in cytokine production or to neutralizing autoantibodies resulting from mutations in the AIRE gene. Further investigations to develop new host- or pathogen-derived biomarkers are needed to improve the diagnosis of CMC and for a better understanding of human epithelial immunity against C. albicans infection.

#### AUTHOR CONTRIBUTIONS

BS and LH conceived the framework of the review. BS, LH, MC, JB, and M-CV wrote the manuscript. EP-L and J-LW revised the manuscript. LH and MC created the Figure. All authors have read and approved the final version of the manuscript.

#### ACKNOWLEDGMENTS

The authors would like to thank Gwendal Geoffroy who actively participated in the graphical improvement of **Figure 1**.

polyendocrinopathy-candidiasis–ectodermal dystrophy protein. Am J Hum Genet. (2000) 66:378–92. doi: 10.1086/302765


gene. J Clin Endocrinol Metab. (2007) 92:595–603. doi: 10.1210/jc.20 06-1873


chronic mucocutaneous candidiasis with and without autoimmune polyendocrinopathy–candidiasis–ectodermal dystrophy. J Allergy Clin Immunol. (2010) 126:1006–15. doi: 10.1016/j.jaci.2010. 08.027


provides evidence in support of an IL-22 defect. Autoimmunity (2014) 47:556– 62. doi: 10.3109/08916934.2014.929666.

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

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

# Exploiting the Oral Microbiome to Prevent Tooth Decay: Has Evolution Already Provided the Best Tools?

Jonathon L. Baker and Anna Edlund\*

Genomic Medicine Group, J. Craig Venter Institute, La Jolla, CA, United States

To compete in the relatively exposed oral cavity, resident microbes must avoid being replaced by newcomers. This selective constraint, coupled with pressure on the host to cultivate a beneficial microbiome, has rendered a commensal oral microbiota that displays colonization resistance, protecting the human host from invasive species, including pathogens. Rapid increases in carbohydrate consumption have disrupted the evolved homeostasis between the oral microbiota and dental health, reflected by the high prevalence of dental caries. Development of novel modalities to prevent caries has been the subject of a breadth of research. This mini review provides highlights of these endeavors and discusses the rationale and pitfalls behind the major avenues of approach. Despite efficacy, fluoride and other broad-spectrum interventions are unlikely to further reduce the incidence of dental caries. The most promising methodologies in development are those that exploit the exclusive nature of the healthy oral microbiome. Probiotics derived from the dental plaque of healthy individuals sharply antagonize cariogenic species, such as Streptococcus mutans. Meanwhile, targeted antimicrobials allow for the killing of specific pathogens, allowing reestablishment of a healthy microbiome, presumably with its protective effects. The oral microbiota manufactures a massive array of small molecules, some of which are correlated with health and are likely to antagonize pathogens. The prohibitive cost associated with sufficiently rigorous clinical trials, and the status of dental caries as a non-life-threatening condition will likely continue to impede the advancement of new therapeutics to market. Nevertheless, there is room for optimism, as it appears evolution may have already provided the best tools.

Edited by:

Sudhanshu Shekhar, University of Oslo, Norway

#### Reviewed by:

Gena D. Tribble, The University of Texas Health Science Center at Houston, United States Thomas Thurnheer, University of Zurich, Switzerland

> \*Correspondence: Anna Edlund aedlund@jcvi.org

#### Specialty section:

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

Received: 20 November 2018 Accepted: 20 December 2018 Published: 11 January 2019

#### Citation:

Baker JL and Edlund A (2019) Exploiting the Oral Microbiome to Prevent Tooth Decay: Has Evolution Already Provided the Best Tools? Front. Microbiol. 9:3323. doi: 10.3389/fmicb.2018.03323 Keywords: oral probiotics, oral treatment, Streptococcus, caries, antimicrobial small molecules

# INTRODUCTION

Evolution within a microbiota is driven by the requirement of each taxa to compete and persist within the host. Meanwhile, hosts are under strong selective pressure to modulate their microbiota to ensure that it confers a benefit. Unlike individual microbes, the sophisticated immune system of mammals can easily influence an entire resident microbial community, and benefit from doing so. For this reason, the human microbiota has been described as an "ecosystem on a leash" (Foster et al., 2017). With multiple microenvironments allowing for a large diversity of taxa, as well as consistent exposure to the external environment and food, the oral cavity presents a highly unique circumstance for the interaction of the human microbiota and the host. Constant exposure to foreign microbes selects for oral taxa which are particularly skilled at direct competition—they

must avoid being replaced! As a consequence, the oral microbiome displays colonization resistance, which is beneficial to the host (He et al., 2014). Therefore, it is probable that the immunology of the oral cavity has also evolved to tolerate, and even facilitate, maintenance of a commensal, yet fiercely territorial, oral microbiome which prevents the establishment of foreign invaders, including pathogens.

Humans have a long history of co-evolution with our resident bacteria, and evidence suggests that our ancient hominid microbiota was more diverse and stable than that of modern humans (Adler et al., 2013; Moeller et al., 2014, 2016). Two dietary shifts, brought about by the development of agriculture (∼7,500 years ago) and the Industrial Revolution (∼200 years ago) (Adler et al., 2013), significantly and rapidly increased the consumption of carbohydrates (particularly sucrose, in the case of the latter). These changes have perturbed the homeostasis of the oral microbiome and dental health, causing dental caries to become the most common chronic disease worldwide, affecting 60–90% of children and adults in industrialized countries (reviewed in Pitts et al., 2017). This review highlights therapeutic strategies, both contemporary and developing, that exploit the protective effects of the healthy oral flora in an effort to prevent dental caries.

## A BRIEF OVERVIEW OF DENTAL PLAQUE ECOLOGY

Typically, the earliest colonizers of the tooth surface are commensal streptococci, such as Streptococcus mitis, Streptococcus sanguinis, Streptococcus gordonii, and other closely related taxa. These species are the most avid binders of the naked, pellicle-coated tooth surface. Once these species have bound, they provide a more complex substrate to which other species can now bind. To help ensure their continued success, the majority of taxa within the mitis and sanguinis groups stanchly antagonize newcomers using the production of alkali, bacteriocins, and H2O2. In the absence of a carbohydrate-rich diet, these commensal streptococci tend to remain at high abundances in dental plaque. This dominance is strongly associated with good dental health. With frequent consumption of carbohydrates, particularly when concurrent with a lack of oral hygiene, increased bacterial production of a glucan matrix is favored, emeshing cells and preventing diffusion of metabolites. This allows for development of emergent properties of the dental plaque, such as acidic microenvironments resulting from carbohydrate fermentation. Typically, the saliva in the mouth has sufficient buffering capacity to neutralize the organic acids produced by bacterial metabolism, and repair acid-damaged enamel. However, the increased thickness and density of exopolysaccharide-rich plaque prevents both diffusion of saliva into the biofilm and diffusion of acids out of the biofilm. The commensal early colonizers are comparatively not well-adapted to acidic conditions, allowing for a further enrichment of acidtolerant (aciduric) taxa such as Streptococcus mutans, Veillonella spp., and Lactobacillus spp. With progression of this positive feedback loop, the rate of net acid damage (demineralization) of the tooth enamel outpaces repair (remineralization), leading to clinical disease.

With an arsenal of extracellular glucosyltransferases (Gtfs), S. mutans is particularly adept at producing a glucan matrix from sucrose, and therefore is considered a keystone species in caries pathogenesis (Bowen, 2016; Bowen et al., 2018). Competition between the early colonizers of the teeth and cariogenic species, particularly S. mutans, has been well-documented and acknowledged for decades (Marquis, 1995; Huang et al., 2018). Interested readers are directed to four excellent and recent reviews covering the above topics in more depth (Abranches et al., 2018; Bowen et al., 2018; Lamont et al., 2018; Redanz et al., 2018). This battleground over the ecological niche of the tooth surface represents a significant opportunity for intervention and subsequent prevention of caries. If the balance of power can be tipped in favor of health-associated organisms, it is possible that pathogenesis of caries can be halted. **Figure 1** provides an overview of dental caries pathogenesis and the major preventative strategies discussed in this review.

### CURRENT CONTROL MEASURES: DIET AND FLUORIDE

#### Diet

Caries is not a classic infectious disease, but the consequence of an ecological shift. Indeed, pathogenic species are necessary, but not sufficient, to cause disease—a constant supply of carbohydrates is also required. In addition to dental caries, the carbohydrate-laden, highly processed modern Western diet has led to pandemics of obesity, type II diabetes, cardiovascular disease, as well as related metabolic disorders and cancers. As with caries, a large and growing body of evidence is linking these conditions to diet via microbial mediators (reviewed in Gilbert et al., 2018; Zmora et al., 2018). Education of the general public to the importance of diet, healthy dietary habits, and the significant association of diet, the microbiota, and health issues remains paramount. In addition to the dissemination of current dietary recommendations, improving the accessibility of healthy foods is a goal worthy of considerable attention. A return to a more primitive and unprocessed diet is likely to have significant health benefits by supporting the microbial profiles with which we have the proper evolutionary rapport to underpin a mutualistic relationship. This includes, of course, the microbial profiles on the tooth surface.

#### Fluoride

Fluoride treatments, including fluoridated toothpaste and drinking water, have been used to combat dental caries for more than 50 years. Fluoride prevents and treats dental caries by promoting favorable remineralization of the tooth enamel while concomitantly impairing bacterial metabolism (Pitts et al., 2017). Although the efficacy of fluoride treatments is well-documented, clearly the current prevalence of the disease illustrates that fluoride alone is insufficient to prevent dental caries in many situations. Other, more stringent, antimicrobial agents are available for dental use (e.g., chlorhexidine and triclosan),

establishment and outgrowth of pathogenic species, such as S. mutans. Immune priming via vaccination leads to elevated levels of secretory IgA, which binds target epitopes on S. mutans and other cariogenic targets, preventing binding and biofilm formation, and promoting clearance from the oral cavity. Antimicrobial peptides,

but all are similarly broad-spectrum. As such, reengineering of a dysbiotic oral microbiome is likely to generate a more positive outcome than its total destruction. The development of approaches to specifically alter plaque composition and prevent outgrowth of cariogenic species, such as S. mutans, remains a highly attractive objective. These approaches fall into several broad categories, discussed below.

STAMPS, small molecules, and phage promote targeted killing of specific cariogenic taxa, such as S. mutans.

### PREVENTATIVE APPROACHES IN DEVELOPMENT

#### Prebiotics

Prebiotics are food or supplements which are administered to modulate the microbiome for the benefit of the host. Arginine has demonstrated success as a prebiotic to prevent dental caries (reviewed in Nascimento, 2018). Arginine can be broken down by commensal arginolytic species (e.g., S. sanguinis and S. gordonii) to generate ammonia, an alkaline molecule that buffers the organic acids in dental plaque. These reactions are performed by the arginine deiminase system (ADS). In addition to contributing to a more alkaline pH, the breakdown of arginine by the ADS also generates ATP, providing a bioenergetic advantage to the commensal streptococci (Bowen et al., 2018). Furthermore, arginine has been shown to inhibit the growth, pathogenic potential, and stress response mechanisms of S. mutans, thereby preventing caries pathogenesis through multiple mechanisms (Chakraborty and Burne, 2017). Higher cost, as well as controversy regarding the protective effects of arginine and the integrity of several clinical trials

(Astvaldsdottir et al., 2016; Richards, 2017), have impeded dentifrices containing both fluoride and arginine from becoming widely available. As laboratory evidence for the protective effects of arginine continues to accumulate (Agnello et al., 2017; Huang et al., 2017; Zheng et al., 2017), more rigorous clinical trials could perhaps lead to widespread availability of arginine-containing commercial therapeutics. Recent studies have identified several other compounds; Met-Pro, succinic acid, beta-methyl-D-galactoside and N-acetyl-D-mannosamine as prebiotic candidates for caries prevention. These molecules were able to promote the dominance of health-associated organisms in a multispecies in vitro culture (Slomka et al., 2017, 2018). Whether these effects can be translated into in vivo studies remains to be investigated.

#### Exploiting Oral Immunology

Caries is not immediately life-threatening, thus selective pressure on the host to thwart the condition is not terribly strong when compared to a disease like smallpox. On the other hand, teeth are a highly valued organ involved in obtaining and digesting food, self-defense, speech/communication, and even sexual attraction (Koussoulakou et al., 2009). It is likely no accident that moieties in the saliva provide binding sites and nourishment for specific species (i.e., the early colonizers), which are largely benign. Saliva flow, and the components of saliva have great influence over which taxa are able to persist in the mouth, and which are cleared (Marsh et al., 2016). Individuals with reduced salivary flow have a greatly increased prevalence of caries. Approaches to increase salivary flow are likely to assist in buffering acids, supplying antimicrobial peptides and antibodies, and preventing dysbiosis and caries from occurring. Chewing gums containing polyols, such as xylitol, provide salivary stimulus without fermentable carbohydrates. Furthermore, certain polyols, particularly erythritol (de Cock, 2018), also have antimicrobial activities, furthering their utility as a preventative modality (Makinen, 2010). Although auspicious, more rigorous research into the effects of these polyol molecules on overall systemic health is warranted; several other sugar substitutes have been recently shown to wreak havoc on the gut microbiota and promote disease (Suez et al., 2014; Rother et al., 2018).

Considering the adaptive arm of the immune system, levels of salivary IgA against immunogenic S. mutans epitopes, such as GTFs and glucan-binding proteins (GBPs), inversely correlate with colonization of S. mutans and caries prevalence (Taubman and Smith, 1993; Nogueira et al., 2005, 2012). Research investigating the feasibility of active or passive immunization against dental caries has been sporadic. Early investigations on the topic were excellently reviewed (Taubman and Nash, 2006). More recent studies have explored vaccination using a recombinant P1 adhesin antigen (Batista et al., 2017), a DNAbased vaccine against glucosyltransferases and surface proteins (Jiang et al., 2017), and a glycoconjugate vaccine based on rhamnan surface polysaccharides (St Michael et al., 2018). Unfortunately, past, present, and likely future, translational endeavors to move anti-caries vaccine research into clinical trials face significant regulatory and investment headwinds due to the fact that it is a non-life-threatening disease. There are currently no licensed vaccines to prevent dental caries.

#### Probiotics

Aside from providing "food" for the oral microbiome to modulate ecology, ecology can also be directly altered by either selectively adding or removing particular species from the oral community. Attempted probiotic strategies to prevent caries either have sought to add health-associated taxa to bolster the capacity of the microbiome to resist dysbiosis, or to replace cariogenic strains with genetically modified mutants which are competitive, yet less pathogenic. A number of studies have explored the use of Lactobacillus and Bifidobacterium, traditionally the genera used in probiotic formulations for digestive health, in the prevention of dental caries. Despite some encouraging results (Lin et al., 2018), there is widespread skepticism concerning the use of these genera as anti-caries probiotics. Both Lactobacillus and Bifidobacterium, are acidogenic and aciduric, meaning that they may actually contribute directly to caries formation under the proper conditions, a fear supported by several studies (reviewed in Philip et al., 2018). In addition, most lactobacilli and bifidobacteria are residents of the gut, meaning they are not welladapted for long-term persistence in the human mouth. They lack capabilities to bind to the salivary pellicle or even nascent dental plaques.

Species with a higher likelihood of outcompeting S. mutans are found already residing in the healthy oral cavity. Studies in other environments have illustrated that the best probiotics for preventing the growth of pathogens both occupy the same ecological niche as the pathogen, and produce compounds that directly antagonize the pathogen (Schlatter et al., 2017). Streptococcus dentisani and Streptococcus A12 are two recently described species which show particular promise as potential probiotics (Huang et al., 2016; Lopez-Lopez et al., 2017). Both species are active colonizers of the tooth surface, increase the pH of dental plaque through the arginolytic pathway, and inhibit the growth of mutans streptococci. In addition, Streptococcus A12 produces a challisin-like protease that disrupts pheromone signaling by S. mutans, inhibiting production of the bacteriocins which S. mutans utilizes to poison its competitors (Huang et al., 2016). Meanwhile, S. dentisani utilizes its own arsenal of bacteriocins to kill multiple cariogenic species (Lopez-Lopez et al., 2017). Streptococcus salivarius has also been examined in a probiotic context (Kurasz et al., 1986; Di Pierro et al., 2015). However, similar to lactobacilli and bifidobacteria, broad skepticism remains over the use of S. salivarius strains as dental plaque probiotics. S. salivarius is typically an inhabitant of the soft surfaces of the mouth and is thought to have limited ability to colonize the tooth surface and directly compete with S. mutans in situ (Philip et al., 2018).

The other major strategy used in probiotic approaches to prevent caries is displacement of native S. mutans strains with S. mutans strains engineered to have low pathogenicity. Two examples of this technique have been reported, utilizing S. mutans mutants deficient in intracellular polysaccharide metabolism (Tanzer et al., 1982) or lactic acid production

(Hillman et al., 2007). Despite preliminary results supporting the potential of these strains as anti-caries probiotics, no further research or studies in humans have been reported. Overall, although newer candidates, such as S. dentisani and Streptococcus A12, provide encouragement, no formulations of probiotics have been tested in rigorous clinical trials and successfully received endorsement for the prevention of dental caries from a regulatory agency or professional organization (Gruner et al., 2016; Burne, 2018).

Recent studies have explored the capacity of the biosynthetic gene clusters (BGCs) encoded by the oral microbiome to produce compounds which modulate oral ecology (Donia et al., 2014; Aleti et al., 2018). Specific BGCs appear to be associated with health or disease states. Comparative statistical modeling illustrated that certain BGCs were inversely correlated with the abundance of cariogenic species, such as S. mutans and Lactobacillus spp. This indicates that the molecular products of these BGCs may be priority therapeutic leads and that the strains harboring these BGCs are prime probiotic candidates, inviting further investigation.

#### Antimicrobial Peptides/STAMPS

As opposed to adding species to the community to prevent or alleviate dysbiosis, various approaches to remove problematic species, such as S. mutans, have been explored. Such targeted methods would presumably restore a healthy oral microbiome. Recent reports showed that the antimicrobial peptides ZXR-2 and CLP-4 efficiently killed S. mutans biofilms, however, specificity for S. mutans was not shown (Chen et al., 2017; Min et al., 2017). Specifically targeted antimicrobial peptides (STAMPS) are synthetic peptides consisting of a targeting domain to invoke specificity and a killing domain to exert antimicrobial action against the targeted species (Eckert et al., 2006b). C16G2 is a STAMP designed to specifically kill S. mutans, and several studies have shown that C16G2 is capable of targeted killing of S. mutans while leaving commensal streptococci unharmed. Furthermore, C16G2 was able to remodel the composition of a complex bacterial community, eliminating S. mutans and allowing for enrichment of organisms associated with dental health (Eckert et al., 2006a; Guo et al., 2015). C16G2 has proceeded to clinical trials in a number of formulations, the results of which will be of significant interest.

#### Small Molecules

Several small molecules have been proposed as agents to prevent caries through disruption of S. mutans biofilms. The small molecule 3F1 selectively dispersed S. mutans biofilms and served to modestly reduce caries in rodent model, however, no changes in the oral microbiomes were reported aside from a moderate reduction in S. mutans as measured by CFUs (Garcia et al., 2017). A 2016 study identified a quinoxaline derivative capable of inhibiting the GtfC enzyme of S. mutans. The compound indeed successfully reduced the ability of S. mutans to form biofilms and reduced caries in a rat model (Ren et al., 2016). Although both of these approaches dispersed S. mutans biofilms, there was minimal effect on overall dental plaque ecology. It is likely that this will allow rapid reformation of the problematic community and require perpetual application of these therapeutics. A recent study utilized a drug repositioning approach to identify 126 compounds with activity against S. mutans (Saputo et al., 2018). How many of these leads are specific to S. mutans remains yet to be determined.

#### Phage

Conceptually, bacteriophage is a very attractive approach to combat cariogenic pathogens, and one that has received relatively little attention. Although the few phages known to infect S. mutans were lytic, and completely eliminated viable counts from single-species biofilms, the phage demonstrated a highly stringent host specificity, which was considered a significant disadvantage, particularly in light of the high intra-species diversity exhibited by S. mutans (reviewed in Szafranski et al., 2017). No testing in multi-species communities or further studies have been reported to date.

# CONCLUSION AND PERSPECTIVES

Because of its status as a key pathogenic species, much of the development of novel caries therapeutics has focused on S. mutans, specifically. S. mutans is certainly a justifiable target, associated with caries in the majority of cases and unparalleled in its ability to form insoluble glucans from sucrose. However, it is not the singular cause of the disease, and caries does occasionally occur without detectable S. mutans levels. Future research efforts would benefit from embracing this perspective and tempering expectations for approaches that fail to do so. It is clear that a dramatic reduction in the prevalence of dental caries through current modalities (such as fluoride and dietary modification) alone is unlikely to be realized. Fortunately, evolution has shaped territorial commensal taxa which antagonize cariogenic species. Exploitation of this relationship, whether by directly supporting the dominance of commensal taxa, or via targeted killing of their pathogenic competitors, is a promising course of therapeutic development. Although several of these approaches have produced encouraging results, properly controlled rigorous human studies are needed, the cost of which is likely to be a significant deterrent. Nevertheless, there is room for optimism, as it appears evolution may have already provided the best tools in the form of our commensal defenders and their natural products.

# AUTHOR CONTRIBUTIONS

JB and AE reviewed literature and wrote the review.

#### FUNDING

Research in the laboratory of the authors were supported by NIH/NIDCR F32-DE026947 (JB) and by R00-DE024543 (AE).

#### REFERENCES

fmicb-09-03323 January 11, 2019 Time: 12:56 # 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 © 2019 Baker and Edlund. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# A Genome-Wide Screen Identifies Factors Involved in *S. aureus*-Induced Human Neutrophil Cell Death and Pathogenesis

Dingyi Yang1,2, Yin Xin Ho1,2, Laura M. Cowell <sup>1</sup> , Iqra Jilani <sup>1</sup> , Simon J. Foster 2,3 and Lynne R. Prince1,2 \*

*<sup>1</sup> Department of Infection, Immunity and Cardiovascular Disease, University of Sheffield, Sheffield, United Kingdom, <sup>2</sup> Florey Institute, University of Sheffield, Sheffield, United Kingdom, <sup>3</sup> Department of Molecular Biology and Biotechnology, University of Sheffield, Sheffield, United Kingdom*

#### *Edited by:*

*Sudhanshu Shekhar, University of Oslo, Norway*

#### *Reviewed by:*

*Dane Parker, New Jersey Medical School, United States Victor J. Torres, New York University, United States Rabia Khan, University of Oslo, Norway*

> *\*Correspondence: Lynne R. Prince l.r.prince@sheffield.ac.uk*

#### *Specialty section:*

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

*Received: 30 October 2018 Accepted: 09 January 2019 Published: 31 January 2019*

#### *Citation:*

*Yang D, Ho YX, Cowell LM, Jilani I, Foster SJ and Prince LR (2019) A Genome-Wide Screen Identifies Factors Involved in S. aureus-Induced Human Neutrophil Cell Death and Pathogenesis. Front. Immunol. 10:45. doi: 10.3389/fimmu.2019.00045* *Staphylococcus aureus* is a commensal organism in approximately 30% of the human population and colonization is a significant risk factor for invasive infection. As a result of this, there is a great need to better understand how *S. aureus* overcomes human immunity. Neutrophils are essential during the innate immune response to *S. aureus*, yet this microorganism uses multiple evasion strategies to avoid killing by these immune cells, perhaps the most catastrophic of which is the rapid induction of neutrophil cell death. The aim of this study was to better understand the mechanisms underpinning *S. aureus-*induced neutrophil lysis, and how this contributes to pathogenesis in a whole organism model of infection. To do this we screened the genome-wide Nebraska Transposon Mutant Library (NTML) in the community acquired methicillin resistant *S. aureus* strain, USA300, for decreased ability to induce neutrophil cell lysis. Out of 1,920 *S. aureus* mutants, a number of known regulators of cell lysis (including the master regulators accessory gene regulator A, *agrA* and Staphylococcus exoprotein expression protein S, *saeS*) were identified in this blinded screen, providing validity to the experimental system. Three gene mutations not previously associated with cell death: *purB, lspA,* and *clpP* were found to be significantly attenuated in their ability to induce neutrophil lysis. These phenotypes were verified by genetic transductants and complemented strains. *purB* and *clpP* were subsequently found to be necessary for bacterial replication and pathogenesis in a zebrafish embryo infection model. The virulence of the *clpP* mutant was restored in a neutrophil-depleted zebrafish model, suggesting the importance of ClpP in mechanisms underpinning neutrophil immunity to *S. aureus*. In conclusion, our work identifies genetic components underpinning *S. aureus* pathogenesis, and may provide insight into how this commensal organism breaches innate immune barriers during infection.

Keywords: *Staphylococcus aureus*, neutrophils, cell death, methicillin resistant *S. aureus* (MRSA), zebrafish

# INTRODUCTION

Staphylococcus aureus has long been recognized as a highly adaptive and dangerous human pathogen, yet this microorganism colonizes the nose of ∼30% of the human population without any ill effects (1). Highly virulent methicillin resistant S. aureus (MRSA) strains can also be carried asymptomatically by healthy individuals (2). Interventions such as hospitalization or episodes of immunosuppression can result in invasive S. aureus infection, which can manifest in multiple forms from superficial skin abscesses to necrotising pneumonia or life-threatening bacteraemia. Colonization is a significant risk factor for pathogenic infection (3, 4). Considering the speed and efficiency with which S. aureus acquires resistance to antibiotics, the shift from silent passenger to life-threatening pathogen is all the more concerning.

Neutrophils are a critical defense in controlling colonization and active infection with S. aureus (5, 6). Yet this microorganism uses multiple evasion strategies to avoid killing by these innate immune cells [reviewed in (7)]. Perhaps the most catastrophic of these strategies is the induction of neutrophil cell death. This not only eradicates a critical element of the early immune response, but also results in inflammation and tissue damage which intensifies disease. S. aureus has been shown to upregulate cell death pathway genes and promote neutrophil apoptosis as well as programmed necrosis (8–10). S. aureus also produces a number of cytolytic toxins including Panton-Valentine leukocidin (PVL), phenolsoluble modulins (PSMs), α-hemolysin, and the leukotoxin LukAB (11). Cytolytic toxin production is intimately linked with pathogenesis, including in community-acquired MRSA (CA-MRSA) infection (12, 13). Understanding host-pathogen interactions in the context of CA-MRSA is imperative since these strains spread rapidly between individuals and are able to cause disease in healthy people. The increased virulence of CA-MRSA has been attributed in part to its resistance to neutrophilmediated killing, including via the induction of neutrophil lysis (14).

Previous studies have taken candidate approaches to the study of components involved in neutrophil lysis, which may have resulted in an incomplete picture of the genetics underpinning cell death. Here we have performed an unbiased genome-wide study to identify additional routes to neutrophil cell death. Using the NTML we have screened all non-essential S. aureus genes to identify genetic components that are involved in inducing neutrophil cell lysis. The NTML was created in the CA-MRSA strain USA300, and has been successfully applied in diverse phenotypic screens to identify genes involved in polymicrobial interactions, antimicrobial resistance, and pathogenicity (15–18). A high-throughput, flow cytometric human neutrophil cell death assay has revealed three genes (purB, lspA, and clpP) that are required for cell lysis. The role of these factors in disease has been determined using a zebrafish embryo model of infection. Our study has provided further evidence for the complex interaction between pathogen and host that determines disease outcome.

# MATERIALS AND METHODS

#### Bacterial Information and Culture

All S. aureus strains were grown in brain heart infusion (BHI) broth at 37◦C with aeration at 250 rpm unless otherwise stated. USA300 S. aureus strain JE2 was used as a positive control. NTML strains were cultured in liquid BHI in 96-well plates. Transductant and complement strains were grown on BHI agar plates followed by overnight inoculation in 250 ml BHI broth. Where required, selection for antibiotic resistance markers was carried out using the following concentrations: ampicillin (Amp, 100µg/ml); chloramphenicol (Cm, 30µg/ml); erythromycin (Ery, 2.5µg/ml); lincomycin (Lin, 12.5µg/ml); kanamycin (Kan, 50µg/ml); tetracycline (Tet, 5µg/ml). The NTML was acquired from the Network on Antimicrobial Resistance in S. aureus (NARSA) strain repository. The NTML was constructed based on USA300 FPR3757 chromosomal genome sequence mapped transposition of bursa aurealis from the delivery plasmid pBursa into the non-essential protein coding sequences of the wild-type JE2 strain (15). Colony forming unit (CFU) count results were obtained using the Miles and Misra method (19). Bacterial density was quantified by spectrophotometric reading at 600 nm (OD600) using a Jenway 6100 spectrophotometer. NTML strain OD600 was measured using a Perkin VICTOR x3 2030 plate reader, with orbital shaking for 0.2 s before reading.

# Genetic Transduction and Complementation

Staphylococcus aureus transduction was performed with Φ11 as described previously (20). Transformation of S. aureus RN4220 and E. coli was carried out by electroporation based on previous methods (21, 22). For genetic complementation, the lspA operon was amplified from JE2 genomic DNA with Phusion polymerase (NEB), using primers containing appropriate restriction sites (forward, GAATTCGTACGG GCCCGGGCTTACTTAACCTCCTTCTCC; reverse, CCA TGTAGGCCAAGTCAAATGAATAATTAAGTTCATATT TAATGTCAAAA). The pKASBAR-Kan<sup>R</sup> , a plasmid carrying an attP encoding site, was inserted with the PCR product to integrate the insert into the S. aureus genome via the attB site, in the presence of an integrase (23). The plasmid providing integrase, pYL112119, was propagated into S. aureus RN4220 strain. From RN4220, the insert was transduced into lspA and control strains. The clpP mutant and Φ85 complemented clpP<sup>+</sup> strain in 8325 background were kindly provided by Knut Ohlsen (24).

# Chemical Complementation

Chemical complementation of purB was conducted with adenine (20 mg/ml in 0.5 M HCl) and inosine (50 mg/ml in dH2O). A final concentration of 20µg/ml was achieved in BHI agar or RPMI (+10% FCS) media accordingly.

# Neutrophil Isolation and Culture

Human neutrophils were isolated by dextran sedimentation followed by plasma-Percoll gradient centrifugation from whole blood of healthy volunteers as previously described (25, 26). Written informed consent and ethical approval from the South Sheffield Research Ethics Committee (study number STH13927) were obtained. The purity of isolated neutrophils was determined from Diff-Quik (Sigma-Aldrich, St. Louis, MO) stained cytocentrifuge preparations by light microscopy. Neutrophils were suspended at 5 × 10<sup>6</sup> /ml in RPMI 1640 (Thermo Fisher Scientific, Waltham, MA, United States) + 10% fetal calf serum (FCS, PromoCell, Heidelberg, Germany) and cultured in 96-well plates at 37◦C, 5% CO2. Phagocytosis assays were performed by incubating neutrophils (in RPMI + 10% FCS) with S. aureus strains at a multiplicity of infection (MOI) of 10 for 1 h after which they were cytocentrifuged (Cytospin, Shandon) onto microscope slides (27). Cells were stained with Quik-Diff dyes and S. aureus visualized within neutrophils by oil immersion light microscopy. The phagocytic index was calculated using the following formula: (total number of engulfed bacteria/total number of neutrophils) × (number of neutrophils containing bacteria/total number of neutrophils) × 100.

#### Intracellular Killing Assay

Neutrophils (4.5 × 10<sup>5</sup> cells/well) were infected with S. aureus at an MOI of 5 at 37◦C, 5% CO<sup>2</sup> in RMPI <sup>+</sup> 10% FCS. The number of internalized viable S. aureus was determined after 30 min. Cells were centrifuged at 2000 RPM for 2 min, resuspended in 200 µl 1% saponin (in PBS), and incubated at RT for 10 min with constant vortexing to lyse neutrophils. The number of viable S. aureus in cell lysates was determined by the Miles and Misra method (19). In parallel wells, lysostaphin (20µg/ml) was added for 30 min to kill extracellular S. aureus (28). Viable intracellular S. aureus was determined after a further 60 min (120 min in total) as above.

#### NTML Neutrophil Cell Death Assay Screen

NTML strains were grown overnight in 96-well plates containing BHI broth following which 10 µl was sub-cultured into 190 µl BHI for 3 h at 37◦C. OD<sup>600</sup> measurements varied from 0.3 to 0.7. Cultures were centrifuged at 5,000 g for 10 min (RT) and pellets were resuspended in 200 µl of RPMI (+10% FCS) for use in cell death assays. Five microliters of each strain were added to 2.5 × 10<sup>5</sup> neutrophils in individual wells of a 96-well plate to achieve an MOI of 10. The following conditions were also included for each experiment: media control (neutrophils in RPMI + 10% FCS) and JE2 (WT) challenge. Plates were incubated at 37◦C, 5% CO<sup>2</sup> for 3 h. Following this, 100 <sup>µ</sup><sup>l</sup> cold PBS containing ToPro-3 (100 nM) was added to each well and samples were immediately subjected to flow cytometry using an Attune Autosampler (ThermoFisher, Waltham, MA). To avoid lengthy plate reading times a maximum of two 96 well plates plus controls were assessed in any one experiment. Plates were acquired at a speed of 500 µl/min and stopped once 70 µl of each sample has been aspirated. Cell counts were automatically generated. Flow cytometric data was analyzed by FlowJo software (TreeStar, Ashland OR). FSC/SSC dot plot profiles of media control conditions were used to set a gate around viable neutrophils and absolute cell numbers were automatically enumerated in this gated region for all plots. Events in the viable gate were exclusively ToPro-3 negative (data not shown), verifying cell viability in this population. The genetic identity of mutant strains was not identified until after analysis was completed to avoid bias.

## Lactate Dehydrogenase Assay (LDH) Cytotoxicity Assay

Neutrophils (4.5 × 10<sup>5</sup> cells/well) were incubated in media (RPMI + 10% FCS) alone or challenged with S. aureus strains at MOI of 5 for 3 h after which the cultures were centrifuged at 300 g for 5 min. LDH activity was quantified in 50 µl supernatant using PierceTM LDH Cytotoxicity Assay Kit, according to the manufacturer's instructions.

# Zebrafish Models of Infection

Zebrafish embryos <5 days postfertilization (dpf) are not protected under the Animals (Scientific Procedures) Act 1986, but all zebrafish work was carried out according to the details set out in Project License PPL 40/3574. London Wild Type (LWT) strains were maintained in E3 medium at 28◦C by following standard protocols and used for all experiments (29). Embryos were dechorionated 1 day prior to bacterial injection. Zebrafish embryos at 24 hpf were anesthetized in 0.02% (w/v) tricaine for 8 min prior to bacterial injection. The stock solution of 0.4% (w/v) 3-amino benzoic acid ester tricaine (Sigma-Aldrich) was made in 20 mM tris-HCl (pH 7). S. aureus was microinjected into the circulation valley located ventral to the yolk sac as described previously (29). The inoculum was determined retrospectively by Miles and Misra method. Zebrafish viability was determined by visual assessment based on cessation of heartbeat and circulatory flow, and assessed at time points of 20, 26, 44, 50, 68, 74, and 92 h post-injection. Bacterial growth was assessed from homogenized embryos plated onto BHI agar.

# Morpholino Oligonucleotide Depletion of Neutrophils

Morpholino oligonucleotides to pu.1 (sequence 5′ -3′ : GATATACTGATACTCCATTGGTGGT) were microinjected into the yolk sac of embryos within 30 min of fertilization (1–4 cell stage) (29, 30). Infection studies were carried out as described above. All survival studies and morpholino data were collected over 3 independent experiments, each comprising of 30 (morpholino and survival studies) or 120 (bacterial growth studies) embryos per group.

#### Statistical Analysis

The Kaplan-Meier method was used to generate survival curves.

The log-rank (Mantel Cox) test was performed to compare survival curves of different strains. All statistical analysis was completed by GraphPad Prism Version 6.0 by one-way ANOVA or as otherwise stated. Significant differences were indicated as: <sup>∗</sup>p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ns, notsignificant.

# RESULTS

# Establishing a High-Throughput *S. aureus*-Induced Neutrophil Lysis Screen

We took an unbiased approach to identifying novel genes related to the induction of neutrophil cell death by highthroughput and blinded screening of a genome-wide S. aureus mutant library. Neutrophils were infected with individual NTML strains in 96-well plates at MOI 10 for 3 h following which they were subjected to flow cytometry (**Figure 1**). A total of 1,920 strains were assayed across 20 plates (IDs: 1A-5D) over 11 non-consecutive days and 5 independent neutrophil donors (not pooled). Media treated and JE2 (WT) infected controls were included for each experiment. The absolute number of neutrophils was automatically calculated by an Attune flow cytometer for each strain. **Figure 2** shows the gating strategy [based on known neutrophil FSC/SSC profiles (31)] used to determine the viable neutrophil population. Events outside the rectangular gate include cell debris and contaminating mononuclear and red blood cells. Viable neutrophils in media control treated conditions typically comprised between 80 and 90% of the total events (**Figure 2A**), which equated to absolute numbers of 28,944 ± 2,212 (mean ± SEM) across all experiments. A profound loss of viable neutrophils was seen in JE2 infected samples where absolute numbers equated to 14,936 ± 2,171 (mean ± SEM) across all experiments. **Figure 2B** shows dot plots of 4 representative NTML strains, showing 2 strains (L-lactate permease; lctP and ABC transporter ATP-binding/permease protein; SAUSA300\_2375) that induced neutrophil cell death at comparable levels to JE2, and 2 "hit" strains (maltose ABC transporter, permease protein, ganP, and ABC transporter, permease protein, vraG) that were attenuated in their ability to induce neutrophil cell death. The attenuated strains have a greater number of events in the viable gated region compared to JE2. For each 96-well plate, the mean and 2 standard deviations from the mean (+2 STDEV) viable neutrophil number for all NTML strains were calculated (**Figure 2C** presents representative data from a single plate, 1A) and a box is drawn around strains identified as hits. A number of strains resulted in increased neutrophil lysis, where viable neutrophils were almost undetectable. While this is of great interest, and may reflect a loss of a negative regulatory mechanism of cell death, we did not further pursue these mutants, since our objective was to identify genes that played a positive role in the induction of neutrophil lysis. As a measure of screen robustness and "hit" sensitivity we were able to identify genes required for neutrophil lysis including the master regulators saeS and agrA (32, 33), both of which resulted in increased numbers of viable neutrophils (**Figure 2D**).

## The *S. aureus* Genes *purB, lspA,* and *clpP* Regulate Neutrophil Lysis

The objective of this screen was to identify genetic mutations that resulted in a defect in neutrophil cell death. Attenuated strains were identified as those with a viable neutrophil count > +2 STDEV of the plate mean and/or as those with a visibly different FSC/SSC profile (see **Figure 2D**). This latter strategy was adopted in order to maximize the number of hits identified in this initial screen round. Any false positives among these would be weeded out in the subsequent focused screen. 118 NTML strains were taken forward into a second round of focused screening, where each strain (incubated with neutrophils at MOI 10 for 3 h as per the primary screen) was tested in 3 independent experiments and presented as a box and whiskers plot (**Supplemental Figure 1**). As in the primary screen, attenuated hits were identified based

FIGURE 2 | viable neutrophils is calculated from viable gated region for each strain and plotted by strain (solid circles represent individual strains). Panel shows data from a single plate for illustration. Mean of all NTML strains (solid line) and mean +2 STDEV (dotted line) is shown. Open icons denote media treated (red square) and JE2-infected (green circle) controls for comparison. The rectangular box is drawn around strains identified as "hits." (D) FSC/SSC dot plots of "hits" *saeS* and *agrA* mutant strains showing increased viable neutrophils compared to JE2 infected samples. Data is generated from a single representative experiment.

on viable cell number > +2 STDEV of the plate mean. Thirty four strains were found to be attenuated (**Supplemental Table 1**). Strains were ranked based on the level of attenuation, that is to say the more viable neutrophils remaining, the greater the attenuation, and listed in descending order. As expected, some of the most attenuated strains included mutations in genes known to be profound modulators of neutrophil death, including leukocidins, saeS, and agrA. Note inclusion of the genes: ganP and vraG as highlighted in **Figure 2B**. To determine whether the attenuated neutrophil lysis was due to a bacterial growth defect and therefore reduced MOI, the number of viable neutrophils was correlated with OD<sup>600</sup> following 3 h growth in BHI. There was no correlation between number of viable neutrophils and OD<sup>600</sup> (**Supplemental Figure 2**), suggesting the extent of the attenuation was not because of differences in MOI.

Of the attenuated strains identified in the second screen, 17 of the most attenuated mutants were further validated in transduction studies. To do this the transposon insert for each strain identified in the screen was transduced back into the parent strain (S. aureus JE2) and transductants were rescreened to establish that the mutant phenotype was specifically associated with each Tn insertion. Neutrophils were infected with transductants at an MOI 10 for 3 h and viable neutrophils enumerated by flow cytometry as above. Transductants (3 clones of each mutation) of 5 original mutants were attenuated (> +2STDEV of the WT strain, JE2) including the known cytolytics lukAB and saeS, as well as adenylosuccinate lyase (purB) and the ATP-dependent Clp protease proteolytic subunit (ClpP, **Figure 3A**). ClpP in addition to the lipopeptidase lspA, which was moderately attenuated, were taken further into genetic complementation studies. Genetic complements of lspA and clpP (indicated by <sup>+</sup>) were able to induce neutrophil cell lysis to levels comparable to JE2 (**Figure 3B**, <sup>∗</sup>p < 0.05 mutant vs. complement). The integration of empty pKB had no effect on neutrophil lysis. Representative flow cytometry plots showing FSC/SSC profiles and viable neutrophils are shown in **Supplemental Figures 3A–D**. To determine neutrophil lysis by an alternative method, we performed an LDH assay and show that clpP and purB, but not lspA resulted in significantly attenuated cytotoxicity (**Supplemental Figure 3E**). The lack of attenuation by lspA in this assay may reflect membrane damage leading to leakage of LDH, but without complete cell lysis (as indicated by flow cytometry plots that are consistent with intact cells, **Supplemental Figure 3B**). Since PurB is required for purine synthesis, the purB mutant was chemically complemented by the addition of adenine and inosine to BHI agar during overnight growth of S. aureus and/or during neutrophil infection. The presence of adenine and inosine to BHI had no effect on neutrophil viability. Addition of adenine and inosine to RPMI during the infection partially restored the ability of the purB mutant to induce neutrophil cell lysis, although this was not statistically significant (**Figure 3C**). This indicates that S. aureus-induced neutrophil cell death is dependent on purines and highlights the specific requirement of purines during the infection period. Phagocytosis assays revealed the clpP mutant was phagocytosed significantly less avidly than the purB and lspA mutants (**Figure 4A**), but all strains were killed equally well by human neutrophils (**Figure 4B**). This suggests that the attenuation in neutrophil cell death is not a result of altered killing of S. aureus.

### PurB and ClpP Are Necessary for Bacterial Replication and Pathogenesis in a Zebrafish Embryo Infection Model

Since clpP, lspA, and purB mutants were defective in causing neutrophil lysis and therefore may not overcome neutrophil defenses during infection, we hypothesized these strains would have altered pathogenicity in vivo. To test this, mutants were studied in a zebrafish embryo infection model (29). Survival rate in PBS injected embryos was >90% (data not shown). As expected, infection with JE2 resulted in profound embryo death (**Figures 5A,B**, solid line). The lspA mutant (dotted line) caused significant mortality at rates comparable to JE2 (**Figure 5A**). In contrast, purB and clpP mutants (dashed lines) failed to kill embryos, with almost maximum survival at 92 h (**Figures 5A,B**). To determine whether a phagocyte response was critical for host immunity to strains, myeloid cells were depleted in zebrafish embryos with a morpholino to pu.1 (30, 34). Consistent with previous studies (34), depletion of neutrophils increased the speed at which embryos died following infection with JE2 with >95% embryos dead by 24 h (**Figure 5C**). Compared to JE2, all mutants delayed embryo death by 24 h but with the exception of purB, all went on to completely overcome the zebrafish by 48 h. The clpP mutant was able to mount an overwhelming infection in the absence of neutrophils but not in the presence of neutrophils which suggests neutrophils are key in the control of this strain.

To define whether increased zebrafish survival was associated with lack of bacterial replication, CFU counts from viable and dead embryos were determined. CFU counts of up to 10<sup>7</sup> were recovered from dead embryos infected with JE2, even at early timepoints, indicating rapid replication in vivo (**Figure 6A**). For mutants that were less efficient at killing zebrafish (clpP and purB), CFU counts in both dead and viable zebrafish were markedly lower at between 10<sup>2</sup> and 3 × 10<sup>5</sup> (**Figures 6B,C**). The results suggest that clpP and purB mutants had limited capacity to replicate within embryos, and were therefore unable

for 3 h at MOI 10 and viable neutrophils enumerated (*n* = 3). Data expressed at mean ± SEM and analyzed by ANOVA with Bonferoni post-test \**p* < 0.05, \*\**p* < 0.01. Comparisons were between JE2 and transductant (A) or as indicated (B,C).

to overcome the zebrafish. As seen in other zebrafish bloodstream infection models, none of the S. aureus strains were completely cleared from any viable embryo over the timecourse studied (29).

#### DISCUSSION

The importance of neutrophils both in the control of colonization and during active S. aureus infection makes them an ideal target for bacterial immune evasion strategies (5, 6). Preserving neutrophil function during infection by preventing S. aureus-induced cell death is therefore an attractive therapeutic strategy and here we describe three genes (purB, clpP, and lspA) with previously unidentified roles in neutrophil cell death. Mutations in two of these genes, purB and clpP, led to significantly reduced pathogenesis in a zebrafish model of infection. Virulence for clpP but not purB was restored by depleting neutrophils in zebrafish. Our work suggests that ClpP is a key element of S. aureus pathogenicity and therapeutically targeting ClpP may improve outcome during S. aureus infections (35–37).

Staphylococcus aureus is well-known to cause a lytic or necrotic like neutrophil cell death via factors that directly damage the cell (38–41). It is unlikely however that any of the gene

performed by addition of both inosine and adenine (I/A) to solid BHI agar during bacterial growth or RPMI media during neutrophil infection (indicated by single <sup>+</sup>) or both (indicated by double ++) at a final concentration of 0.02 mg/ml. The absence of I/A in RPMI or BHI is indicated by (−). Neutrophils were incubated with *S. aureus*

*ClpP, purB,* or *lspA* strains at MOI 5 and an intracellular killing assay performed. Graph shows % killing of *S. aureus* by calculating the reduction in the number of viable bacteria at 120 min compared to 30 min for each strain. Data expressed as mean ± SEM (*n* = 3). Data were analyzed by ANOVA with Bonferoni post-test \*\*\**p* < 0.001.

products identified in our study are directly lytic to neutrophils. In the case of PurB, a deficiency in purine biosynthesis and therefore a failure to replicate and express virulence factors in a nutrient poor environment is likely to explain the reduction in neutrophil lysis. This was in part confirmed by the addition of adenine and inosine during infection of neutrophils, which partially increased neutrophil lysis. S. aureus purB mutants have been recently shown by our group to have reduced pathogenesis, and purA mutants were found to be attenuated in a S. aureus murine abscess study (16, 42). In support of this we observed an attenuated phenotype of the purA mutant in the primary screen and this was taken forward into the secondary screen. The level attenuation however, was not as robust as purB and therefore was not carried forward into further studies.<underline >We found the attenuated virulence of the purB mutant in

zebrafish was unaffected by the absence of neutrophils. This further suggests that a fundamental deficiency to replicate, rather than a failure to overcome innate immunity, was the cause of the attenuation.

The lipoprotein signal peptidase product of lspA is localized in the bacterial membrane and required for biogenesis of bacterial lipoproteins (43). Lipoproteins themselves can play essential roles in host-pathogen interactions, for example as pathogen associated molecule patterns (PAMPs) acting via TLRs (44–46). Bacterial lipoproteins can induce apoptosis in other

\*\**p* < 0.01, \*\*\**p* < 0.001.

non-viable (closed black circles) homogenized embryos at selected time points were determined by Miles Misra.

cell types (47). Although we measure a non-apoptotic cell death in this study, it is conceivable that activation of an apoptosis programme by lipoproteins in concert with other death inducing toxins may lead to an overall net effect of lysis and therefore by removing lipoproteins a lytic pathway is prevented. Failure to mature lipoproteins due to loss of LspA or other peptidases has previously been shown to impair growth and pathogenicity, but our study is the first to describe a specific immune modulating role for LspA in S. aureus (44, 48). In murine models of S. aureus infection an lsp mutant failed to induce disease (44) but we show no attenuation in zebrafish embryos. Since S. aureus has evolved as a human associated organism it is possible that this tropism may extend to the targets of LspA, explaining the species specificity of the observed effect.

The principal function of Clp proteases is protein degradation, although they are associated with a number of physiological processes (36). ClpP mutants are highly susceptible to stress and the attenuated phenotype caused by the clpP mutant in our study may be due to lack of stress adaptation in the neutrophil phagosomal environment. S. aureus is adept at escape from the richly anti-microbicidal phagosome (49, 50). ClpP deficient Legionella pneumophila fails to escape from the endosome-lysosomal pathway in a macrophage cell line and S. aureus mutants are unable to replicate intracellularly (35, 51). Interestingly this has also been shown for lspA in Listeria monocytogenes (52). It is possible that containment within the neutrophil phagosome may prevent post-phagosomal cell lysis for these mutants, but further studies are required. The clpP mutant was not phagocytosed as readily, which may be as a result of an impairment in lipoprotein-dependent recognition by neutrophils, leading to a reduction in post-phagocytic cell lysis.

ClpP regulates the expression of a number of bacterial virulence factors such as hemolysin, which may in part account for the attenuation in neutrophil cell death seen in our study (24, 53). As a result, clpP mutants are found to be less virulent in animal models (53, 54). We also show attenuation in vivo, where the clpP mutant fails to replicate in zebrafish embryos, and which is supported by others that demonstrate a failure to replicate in the host (35). A growing number of studies highlight the therapeutic potential of targeting ClpP. A recent study describes a selective, small-molecule inhibitor of ClpP, identified via high-throughput screening and which attenuates virulence in mouse models of S. aureus USA300 infection (55). Our work and others suggest therapeutically targeting ClpP may have great promise in treating invasive S. aureus infections. In conclusion, our work identifies genetic components underpinning S. aureus pathogenesis and provides further evidence for the complex interaction between pathogen and host. These findings provide a greater insight into how this commensal organism breaches innate immune barriers during infection.

#### AUTHOR CONTRIBUTIONS

DY, SF, and LP wrote the manuscript. DY, YH, IJ, and LC performed the experiments. All authors contributed to experimental design and data analysis.

### FUNDING

This work was funded by the University of Sheffield 2022 Futures Program.

#### ACKNOWLEDGMENTS

We are grateful to volunteers who donated blood to this study, and the use of the Bateson Centre aquarium at the University of Sheffield. We thank Kenneth Bayles for providing the NTML,

#### REFERENCES


Rebecca Hodges and Bartlomiej Salamaga for technical assistance with the zebrafish embryo work, Mark Crooke and Katarzyna Wacnik for assistance with the complementation works, and Knut Ohlsen for providing the clpP<sup>+</sup> strain.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00045/full#supplementary-material


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

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

# Elevated Gut Microbiome-Derived Propionate Levels Are Associated With Reduced Sterile Lung Inflammation and Bacterial Immunity in Mice

Xiaoli Tian<sup>1</sup> , Judith Hellman<sup>1</sup> , Alexander R. Horswill<sup>2</sup> , Heidi A. Crosby<sup>2</sup> , Kevin P. Francis<sup>3</sup> and Arun Prakash1,4 \*

<sup>1</sup> Department of Anesthesia and Perioperative Care, University of California, San Francisco, San Francisco, CA, United States, <sup>2</sup> Department of Immunology and Microbiology, Anschutz Medical Campus, University of Colorado, Aurora, CO, United States, <sup>3</sup> Preclinical Imaging, PerkinElmer, Hopkinton, MA, United States, <sup>4</sup> San Francisco General Hospital, University of California, San Francisco, San Francisco, CA, United States

Short-chain fatty acids (SCFA) are important dietary and microbiome metabolites that can have roles in gut immunity as well as further afield. We previously observed that gut microbiome alteration via antibiotics led to attenuated lung inflammatory responses. The rationale for this study was to identify gut microbiome factors that regulate lung immune homeostasis. We first investigated key factors within mouse colonic lumen filtrates (CLF) which could elicit direct inflammatory effects in vitro. We identified lipopolysaccharide (LPS) and SCFAs as key CLF ingredients whose levels and inflammatory capacity changed after antibiotic exposure in mice. Specifically, the SCFA propionate appeared to be a key regulator of LPS responses in vitro. Elevated propionate: acetate ratios, as seen in CLF after antibiotic exposure, strongly blunted inflammatory responses in vitro. In vivo, exposure of lungs to high dose propionate, to mimic how prior antibiotic exposure changed SCFA levels, resulted in diminished immune containment of Staphylococcus aureus pneumonia. Finally, we discovered an enrichment of propionate-producing gut bacteria in mice with reduced lung inflammation following lung ischemia reperfusion injury in vivo. Overall, our data show that propionate levels can distinctly modulate lung immune responses in vitro and in vivo and that gut microbiome increased production of propionate is associated with reduced lung inflammation.

#### Keywords: lung injury, short-chain fatty acids, SCFA, acetate, propionate, IR, inflammation

#### INTRODUCTION

The human body coexists with a vast commensal microbiome that is increasingly recognized to play important roles in human health, physiology, and disease (Cho and Blaser, 2012; Shreiner et al., 2015; Sender et al., 2016; Alverdy and Krezalek, 2017; Young, 2017). Interactions between the microbiome and host are best understood at specific interfaces, such as the gut, oral cavity and the skin. Additionally, there appears to be a prominent role for gastrointestinal dysbiosis and pathobionts in the etiology of inflammatory bowel disease, other intestinal disorders, and critical illness in general (Alverdy and Krezalek, 2017). From an evolutionary viewpoint, human physiology has evolved closely with specific gut microbiota which, prior to the advent of antibiotics, were likely not exposed to sudden perturbations in composition

#### Edited by:

Fernanda Cristina Petersen, University of Oslo, Norway

#### Reviewed by:

Jeremy McAleer, Marshall University, United States Koshika Yadava, University of Oxford, United Kingdom

> \*Correspondence: Arun Prakash arun.prakash@ucsf.edu

#### Specialty section:

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

Received: 14 September 2018 Accepted: 22 January 2019 Published: 14 February 2019

#### Citation:

Tian X, Hellman J, Horswill AR, Crosby HA, Francis KP and Prakash A (2019) Elevated Gut Microbiome-Derived Propionate Levels Are Associated With Reduced Sterile Lung Inflammation and Bacterial Immunity in Mice. Front. Microbiol. 10:159. doi: 10.3389/fmicb.2019.00159

over the life cycle of the human host (after early postnatal colonization). These symbiotic organisms may serve purposes besides food digestion and vitamin production, such as during immune education and host defense. While it may be conceptually challenging at first to explain how gut commensal microbial communities affect physiological processes in distant organ systems, a few studies have supported this paradigm (Trompette et al., 2014; Vieira et al., 2015, 2016).

We and others have reported that the gut microbiome influences lung inflammatory responses and these data support the current hypothesis for the existence for a gut-lung axis of communication (Trompette et al., 2014; Marsland et al., 2015; Prakash et al., 2015; Samuelson et al., 2015; Budden et al., 2017). However, the key questions of how communication along this axis occurs still remain unanswered. We speculate that commensal-associated molecular patterns (CAMPs) and metabolites in concert modulate lung inflammatory responses by either directly acting on resident lung immune cells or indirectly through immune reprogramming of circulating immune cells or both. Released CAMPs and metabolites may transit through the intestinal mucosa and portal circulation and travel to the lungs via the liver and heart. Since the pulmonary and gastrointestinal systems are both exposed to environmental and infectious threats, it is plausible that the commensal microbial community influences these two systems. This concept is supported by our earlier published data that lung alveolar macrophages from mice with antibiotic-altered gut microbiota were less responsive to inflammatory ligands than their unaltered counterparts (Prakash et al., 2015). The gut-lung axis may contribute to establishing lung immune homeostasis that could be akin to a baseline state of immune tonicity or injury readiness (reviewed in Lloyd and Marsland, 2017). By understanding the impact of commensal-derived factors on lung immunity, it may be possible to selectively or temporarily regulate lung immunity: for example, by bolstering host defenses during pneumonia or mitigating sterile pneumonitis after gastric acid aspiration.

In this study we focused on identifying factors produced by the gut microbiome that could regulate lung responses to sterile and infectious injuries. We discovered that antibioticmediated dysbiosis led to large alterations in SCFA levels, specifically, propionate and acetate levels. Furthermore, propionate appeared to have a concentration-dependent ability to modify LPS inflammatory responses in vitro, with low and high concentrations augmenting and blocking inflammatory responses, respectively. High propionate:acetate ratios appeared to skew LPS responses toward less inflammation while low ratios appeared to skew LPS responses toward more inflammation in vitro. This was confirmed by the in vivo observation that direct administration of high propionate concentrations to mouse lungs resulted in worsening of an experimental pneumonia, likely through the inhibition of beneficial inflammation needed to fight the infection. Finally, examining the microbiomes of mice with varying degrees of sterile lung injury revealed that high propionate-producing bacteria (Lachnospiraceae) were enriched significantly in mice that exhibited low lung inflammation phenotypes.

Overall, our study strongly suggests that gut commensal bacteria continuously prime resident cells in distant organ systems, such as the lungs, through shed toxins and specific metabolite compositions. Furthermore, alterations in gut microbiome composition can profoundly change inflammatory responses to sterile injury and pathogenic infectious responses. We had previously shown that reducing lung inflammation results in a deficient host immune response to infection (Tian et al., 2017). Being able to control and modulate lung inflammatory responses through gut microbiome or metabolite manipulation could aid in preventing pulmonary complications from prolonged hospitalization and antibiotic exposure in critically ill patients.

#### MATERIALS AND METHODS

#### Animals

Male mice (12–15 weeks old) were either purchased (The Jackson Laboratory, Bar Harbor, ME, United States) or bred at the animal facility at University of California, San Francisco. Purchased mice were allowed to acclimatize to their new housing for at least 1 week before any experiments on them were conducted. Wild-type C57BL/6 and C3H/HeOuJ mice were used in this study.

Only male mice were used in our experiments primarily to reflect the fact that trauma disproportionately affects human males. Based on our previous studies, we used group sizes of 6–10 for all experiments (Prakash et al., 2012, 2015). All mice for a given experiment were either littermates or purchased/bred such that they were age-matched. Mice used in these experiments were randomly chosen either to undergo the various surgeries (sham vs. IR) or treatments (+/− specific treatments); therefore there was no attempt made to blind the individuals conducting the experiments. However, in situations where mice received a treatment or control before IR surgery or infection after surgery, the individual collecting the organs/plasma and generating the ELISA data was unaware of which mice received which specific treatment.

#### Reagents and Cell Lines

Short-chain fatty acids (SCFA) (acetate, butyrate, propionate, formate), Trichostatin A, polymyxin B, lipopolysaccharide (LPS), and lipoteichoic acid (LTA) were purchased from Sigma-Aldrich, St. Louis, MO, United States. Pam3CysSKKK and FSL-1 were both obtained from EMC Microcollections, Tubingen, Germany. The following cell lines were used in this study: HUVEC (primary human umbilical vein endothelial cells, used at passage 6 or less, Promocell, Heidelberg, Germany), EOMA (129 background endothelial cell line, ATCC, Manassas, VA, United States), SVEC-40 (C3H/HeJ TLR4 mutant background endothelial cell line, ATCC), and MH-S (wild-type BALB/c alveolar macrophage cell line, ATCC).

#### Antibiotic Treatment

A group of wild-type C3H/HeOuJ background mice were treated with antibiotic containing water ad libitum or control

group was given standard drinking water. Antibiotic treatment consisted of neomycin and polymyxin B (final concentrations: 0.6 mg/mL for neomycin and 120 units/mL for polymyxin B, both from Sigma-Aldrich, St. Louis, MO, United States) given in drinking water for 8 weeks as described earlier (Prakash et al., 2015). This combination of antibiotics was chosen specifically for the ability of both to remain within the gut and not be absorbed into circulation; by using this combination to target the gut microbiome, we could focus primarily on its effects without disruption of other microbiomal niches. **Supplementary Figure S6** demonstrates the effects of this combination on the richness, evenness, and diversity of the microbiome after 7 weeks of treatment.

#### Colonic Lumen Filtrate (CLF)

Stool from mice was obtained in two ways: by scruffing mice and collecting stool as it was produced by the mice; or by collecting small and large intestines of mice and expressing it out of the intestinal cavity. Homogenized stool was then sterile filtered to create CLF and protein concentration measured by bicinchoninic acid (BCA) method. Protein concentrations of CLF were measured using standard assay (PierceTM BCA Protein Assay Kit, Thermo Fisher Scientific, Waltham, MA, United States). CLF sterility was confirmed by plating on LB agar plates. Stool from cages of mice (3 cages, 15 mice total per group) that received either control or antibiotic water were collected, pooled, and CLF was prepared. CLF was added to tissue culture supernatant for various experiments described in the concentrations noted.

#### SCFA and Amino Acid Measurement

Stool or colonic lumen filtrate (prepared as described above) was sent to PennCHOP metabolomics core facility<sup>1</sup> for SCFA and amino acid analysis. CLF was further filtered using 1.2, 0.65, and 0.22 µm filter plates (Millipore, Billerica, MA, United States). The filtrate was loaded into total recovery vials (Waters, Milford, MA, United States) for analysis. Short chain fatty acids were quantified using a Water Acquity uPLC System with a Photodiode Array Detector and an autosampler (192 sample capacity). Samples were analyzed on a HSS T3 1.8 µm 2.1 × 150 mm column. The flow rate was 0.25 mL/min, the injection volume was 5 µL, the column temperature was 4◦C, the sample temperature was 4◦C, and the run-time was 25 min per sample. Eluent A was 100 mM sodium phosphate monobasic, pH 2.5, eluent B was methanol, the weak needle wash was 0.1% formic acid in water, the strong needle wash was 0.1% formic acid in acetonitrile, and the seal wash was 10% acetonitrile in water. The gradient was 100% eluent A for 5 min, gradient to 70% eluent B from 5 to 22 min, and then 100% eluent A for 3 min. The photodiode array was set to read absorbance at 215 nm with 4.8 nm resolution. Samples were quantified against standard curves of at least five points run in triplicate. Standard curves were run at the beginning and end of each metabolomics run. Quality control checks (blanks and standards) were run every eight samples. Results were rejected if the standards deviate by greater than ± 5%. Concentrations in the samples were calculated as the measured concentration minus the concentration of the solvent; the range of detection was at least 1 – 100 µmol/g stool.

#### Ventilated Lung Ischemia Reperfusion (Unilateral Left Pulmonary Artery Occlusion) Surgery

A mouse model of unilateral left pulmonary artery (PA) occlusion was used, as we have described previously (Prakash et al., 2012). Briefly, anesthetized mice (using IP tribromoethanol (Avertin <sup>R</sup> ); Sigma-Aldrich) were orally intubated, given buprenorphine (IP; Harry Schein, Melville, NY, United States), and placed on a mini-vent mouse ventilator (Harvard Apparatus, Holliston, MA, United States), using tidal volumes of 0.225 mL (7.5 mL/kg), and a respiratory rate of 180 breaths/min (assuming an average mouse weight of 30 g). A left thoracotomy via the interspace between the 2nd and 3rd ribs was performed and the left PA was identified and ligated using a slip knot suture with 7-0 or 8-0 prolene monofilament suture. The end of the suture was externalized through a narrow bore (27 g) needle to the anterior chest wall. Prior to closure of the thorax, the left lung was reinflated with positive end expiratory pressure (PEEP). Local anesthetic (3-4 drops of 0.25% bupivacaine) was applied topically prior to skin closure. The total period of mechanical ventilation and surgery was approximately 20–25 min. After skin closure, mice were extubated and allowed to recover from anesthesia. After 60 min of ischemia, the ligature on the PA was released and left lung reperfusion started. At the experimental end-point times, mice were euthanized and the blood and lungs were collected.

Blood was collected from anesthetized mice via cardiac puncture using a heparinized syringe, centrifuged (14,000 g, 5 min) and the plasma separated, flash frozen in liquid nitrogen and stored at −80◦C. Lower portions of the left lungs were excised and placed in Trizol <sup>R</sup> (Thermo Fisher Scientific, Waltham, MA, United States) at −80◦C for RNA isolation. Levels of cytokines and chemokines (described later) were quantified in plasma.

Mice received equivalent durations of mechanical ventilation (20–25 min), and were left spontaneously breathing during their recovery from anesthesia and the remainder of the ischemia period and subsequent reperfusion or equivalent periods in the sham mice.

While this lung IR procedure has high initial survival rates of 80–90% on average, some mice die from irreparable damage to the PA or left bronchus during the slip-knot placement. Mice that did not survive the surgery or the reperfusion period due to technical complications in the surgical procedure (predominantly, left bronchus or left PA injury) were excluded from the study. The overall attrition rate was 10–20%.

#### Lung Injury Scoring

Lung injury was scored in histology images by one of two methods: semiquantitative visual scoring and by ImageJ analysis of the images for counts of inflammatory cells and %area occupied. The former (semi-quantitative) scoring method (1 = no lung injury and 5 = severe lung injury) was performed

<sup>1</sup>https://metabolomic.research.chop.edu/

as previously described (Prakash et al., 2015). The latter (quantitative ImageJ) scoring method was performed follows: in brief, the ImageJ freehand selection tool was used to trace the perimeter of each region of interest. The area extending beyond the perimeter of the vessel was cleared, and the color threshold of the image was adjusted using the default method with the following parameters: hue = 0– 255, saturation = 0–255, brightness = 130–255, threshold color = white, background = dark, color space = HSB. The image was converted to an 8-bit gray scale, and the threshold was adjusted using the B&W defaults and a range of 0–150. Counts were outlined and summarized using the analyze particles window (size = 0-infinity, circularity = 0.0–1.00). Average percent Area was also calculated. Lung Injury Scores and cutoffs for high vs. low lung injury designation for this study are included in **Supplementary Figure S7**.

## S. aureus Experimental Pneumonia

C57BL/6 wild-type mice were pretreated with high (1 mM) or low (0.1 mM) propionate intratracheally (IT) 2 h prior to IT administration (10<sup>8</sup> CFU) of luminescent strain of S. aureus (Newman-lux strain generously provided by Alex Horswill, University of Colorado, Denver). IT administration was done under isoflurane anesthesia and with direct visualization. Six hours after infection, mice were live imaged using IVIS <sup>R</sup> in vivo imaging system (see below for more details). Lungs were then collected and imaged ex vivo using IVIS <sup>R</sup> and luminescence was also measured using Cytation5 cell imaging multi-mode reader (BioTek, Winooski, VT, United States).

Staphylococcus aureus (Newman-lux strain) was grown as follows: after an overnight inoculation in LB broth, serial dilutions of the overnight stock was grown at 37C in a 24 well tissue culture dish with orbital shaking in the Cytation5 cell imaging multi-mode reader (BioTek, Winooski, VT, United States). Every 5 min, an OD reading and a luminescence reading (to verify healthy growth of the luminescence producing strain) were obtained. When the mice were ready for infection, the wells that were at OD 0.3 (mid-log phase) were removed and used for IT infection. Later CFU measurements of these innocula provided the actual CFU count administered to the mice as described above.

#### In vivo Imaging (IVIS <sup>R</sup> )

C57BL/6 mice after IT propionate and S. aureus administration were imaged at the time points noted. Imaging was conducted on the IVIS <sup>R</sup> Spectrum Instrument (PerkinElmer, Hopkinton, MA, United States) as previously described (Zhang et al., 2013). Luminescence imaging was performed with an open filter for 5 min.

# Sandwich Enzyme-Linked Immunosorbant Assay (ELISA)

Levels of IL-6 produced were determined using the corresponding mouse duoset or Quantikine kits (R&D Systems, Minneapolis, MN, United States). A multiplex ELISA to measure an immune panel of cytokines was used to identify the pattern of expression of other inflammatory and associated cytokines. This measurement of protein levels of cytokines and chemokines were performed once with a 20plex immune array kit (Thermo Fisher Scientific, Waltham, MA, United States). Analytes included in the panel: FGFβ, IL-1β, IL-10, IL-13, IL-6, IL12, IL-17, MIP-1α, GMCSF, MCP-1, IL-5, VEGF, IL-1α, IFNγ, TNFα, IL-2, IP-10 (CXCL10), MIG, KC, IL-4. Those analytes not shown in **Supplementary Figure S2** were detected at low levels or below the level of detection of the assay. All assays were performed according the manufacturer's supplied protocol. All ELISA measurements (except for the multiplex immune panel) were repeated 2–3 times from independently conducted experiments and representative data shown. Standard curves were generated and used to determine the concentrations of individual cytokines or chemokines in the sample.

# Microbiome Analysis

Single stool pellets from 23 wild type C3H/HeOuJ mice: 11 that received (Neo/PMB) or 12 that received control water for 8 weeks were processed by the UCSF Colitis and Crohn's Disease Microbiome Research Core Facility as previously described (Fujimura et al., 2016; Mar et al., 2016). Briefly, 46 mouse fecal samples (23 from the week prior to starting antibiotic/control water and 23 from the week prior to lung IR surgery) were processed for DNA extraction, PCR amplification of the V4 hypervariable region of the 16S rRNA gene, and DNA sequencing on the Illumina NextSeq. DNA was extracted from all samples using a modified CTAB extraction protocol. Each DNA sample was PCR amplified in triplicate using primers that (1) targeted the V4 hypervariable region of the 16S rRNA gene, (2) contained a unique barcode sequence to enable demultiplexing of pooled samples, and (3) contained an adapter sequence that enables the amplicon to bind to the NextSeq flow cell. Successful amplicons were pooled in equimolar concentrations and sequenced on the Illumina NextSeq.

Downstream analysis: Merged sequencing read pairs containing less than two expected errors were binned into OTUs (operational taxonomic units) using a 97% sequence similarity threshold. OTUs determined to be chimeric or not of bacterial origin were discarded. Additionally, OTUs known to be common contaminants observed in greater than 50% of extraction controls were discarded and the maximum read count of each remaining OTU in any single extraction control was subtracted from the reads counts of that OTU for all sample. Read counts for OTUs which summed across all samples that were less than 1/1000th of a percent of the total read count for the entire dataset were discarded to minimize noise. Sample read numbers were representatively rarefied to 17,444 reads resulting in a rarefied OTU table. Forty-five of 46 (98%) samples had quality filtered read numbers above the specified rarefying threshold and were included in the downstream analyses.

#### DNA Extraction

Individual murine fecal samples were placed into lysing matrix E (LME) tubes pre-aliquoted with 500 of hexadecyltrimethylammonium bromide (CTAB) DNA extraction buffer and incubated at 65◦C for 15 min. An

equal volume of phenol:chloroform:isoamyl alcohol (25:24:1) was added to each tube and samples were homogenized in a Fast Prep-24 homogenizer at 5.5 m/s for 30 s. Tubes were centrifuged for 5 min at 16,000 × g and the aqueous phase was transferred to individual wells of a deep-well 96-well plate. An additional 500 µl of CTAB buffer was added to the LME tubes, the previous steps were repeated, and the aqueous phases were combined. An equal volume of chloroform was mixed with each sample, followed by centrifugation at 3000 × g for 10 min to remove excess phenol. The aqueous phase (600 µl) was transferred to a deep-well 96-well plate, combined with 2 volume-equivalents of polyethylene glycol (PEG) and stored overnight at 4◦C to precipitate DNA. Plates were centrifuged for 60 min at 3000 × g to pellet DNA and the PEG solution was removed. DNA pellets were washed twice with 300 µl of 70% ethanol, air-dried for 10 min and suspended in 100 µl of sterile water. DNA samples were quantitated using the Qubit dsDNA HS Assay Kit and diluted to 10 ng/µl.

#### DNA Amplification and Sequencing

The V4 region of the 16S rRNA gene was amplified in triplicate as previously described (see citation below). Triplicate reactions were combined and purified using the SequalPrep Normalization Plate Kit (Invitrogen) according to manufacturer's specifications. Purified amplicons were quantitated using the Qubit dsDNA HS Assay Kit and pooled at equimolar concentrations. The amplicon library was concentrated using the Agencourt AMPure XP system (Beckman-Coulter) quantitated using the KAPA Library Quantification Kit (KAPA Biosystems) and diluted to 2 nM. Equimolar PhiX was added at 40% final volume to the amplicon library and sequenced on the Illumina NextSeq 500 Platform on a 153 bp × 153 bp sequencing run.

#### OTU Table Generation

Raw sequence data was converted from bcl to fastq format using bcl2fastq v2.16.0.10. Paired sequencing reads with a minimum overlap of 25 bp were merged using FLASH v1.2.11. Index sequences were extracted from successfully merged reads and demultiplexed in the absence of quality filtering in QIIME (Quantitative Insights Into Microbial Ecology, v1.9.1), and reads with more than two expected errors were removed using USEARCH's fastq filter (v7.0.1001). Remaining reads were dereplicated at 100% identity, clustered into operational taxonomic units (OTUs) at 97% sequence identity, filtered to remove chimeric sequences, and mapped back to OTUs using USEARCH v8.0.1623. Taxonomy was assigned using the Greengenes database (May 2013). OTUs detected in Negative Extraction Controls (NECs) were considered potential contaminants and filtered as follows: any known common contaminant OTU present in more than half of the NECs for this study was removed from all samples; the maximum read count for any OTU found in fewer than half of the NECs was subtracted from all samples; and any remaining OTU with a total read count less than 0.001% of the total read count across all samples was removed.

#### Alpha-Diversity

Alpha-diversity indices were computed in QIIME. Comparisons between mouse genotypes were assessed using the Kruskal–Wallis one-way analysis of variance test. Results with a p-value of < 0.05 were considered statistically significant.

#### Beta-Diversity

Beta-diversity dissimilarity matrices (Bray-Curtis, Canberra, weighted and unweighted uniFrac distances) were generated in QIIME. Variables were assessed for their relationship to bacterial beta-diversity by permutational analysis of variance (PERMANOVA) using the Adonis function (vegan package) in the R environment; variables of p < 0.05 were considered statistically significant.

#### Taxonomic Differences

Enriched taxa were identified using a "three model" approach where Poisson, negative-binomial, and zero-inflated negativebinomial models were applied to each taxon individually, and the model that minimized the Akaike information criterion value (AIC) was selected for each taxon. Before applying the models, the OTU table was de-noised by removing taxa that were present in fewer than 25% of the samples. To adjust for multiple-testing, the false-discovery rate was calculated for each taxon; a q-value of < 0.20 was considered significant.

#### Statistical Analysis

Data in the figures are expressed as mean ±SD. Data from in vivo studies comparing two conditions were analyzed using two-tailed unpaired non-parametric Mann–Whitney analyses. Data from in vitro studies comparing two conditions were analyzed using two-tailed unpaired parametric t test with Welch's correction. GraphPad Prism was used for statistical analyses (GraphPad Software, La Jolla, CA, United States). For all in vivo and in vitro experiments, p values < 0.05 were considered statistically significant. P-values are represented as follows in the figures: <sup>∗</sup> < 0.05; ∗∗ < 0.01; ∗∗∗ < 0.001; ∗∗∗∗ < 0.0001. For multiple comparisons, one-way ANOVA was used, with p-values represented as follows: α < 0.05; αα < 0.01; ααα < 0.001; αααα < 0.0001. When comparing treatment conditions against an untreated or control condition (indicated in figure legends), two-tailed unpaired parametric t-test with Welch's correction was used, with p-values represented as follows: δ < 0.05; δδ < 0.01; δδδ < 0.001; δδδδ < 0.0001. Experiments were repeated two or more times, as indicated in the figure legends.

#### Study Approval

All mouse studies were approved from an ethical and methodological standpoint by the Institutional Animal Care and Use Committee (IACUC) at the University of California, San Francisco and followed the ARRIVE guidelines.

# RESULTS

fmicb-10-00159 March 25, 2019 Time: 15:50 # 6

# Colonic Lumen Filtrate (CLF) Is Pro-inflammatory and Contains Lipopolysaccharide (LPS)

After previously reporting that antibiotic treatment led to diminished inflammatory changes in the mouse lung following lung ischemia reperfusion (Prakash et al., 2015), we tested whether colon contents of mice contained inflammationinducing factors. CLF from wild-type C57BL/6 mice was prepared (**Figure 1A**) and used to stimulate human endothelial cells (HUVEC). Similar to LPS, CLF challenge resulted in dose-dependent IL-6 production (**Figure 1B**). Interestingly, CLFs pro-inflammatory effects on human cells were poorly blocked by polymyxin B (PMB – binds and inactivates LPS) (**Supplementary Figure S1**). However, in murine alveolar macrophages (AMs), PMB was able to block the majority of CLF inflammatory effects (**Figure 1C**). Additionally, like LPS, CLF effects were also blocked both in HUVEC and AMs by histone deacetylase (HDAC) inhibition both by high dose SCFA (butyrate) as well as trichostatin A (TSA) (**Supplementary Figure S2**). Taken together, CLF likely contains both LPS as well as other inflammatory agents produced or shed by the gut commensal microbiome.

#### CLF From Antibiotic-Treated Mice Is Less Inflammatory and Contains Less LPS

Since antibiotic-treated mice displayed blunted inflammatory responses to sterile lung injury and diminished alveolar macrophage responses to LPS (Prakash et al., 2015), we investigated the effects of CLF from these mice. CLF from antibiotic-treated mice (2-week treatment) generated less inflammation from both HUVEC (**Figure 2A**) and murine AMs (**Figure 2B**). Consistently, antibiotic CLF contained greatly reduced levels of LPS (**Figure 2C**). Furthermore, while control CLF was blocked by PMB, antibiotic CLF was not (**Figure 2B**). Overall, these data suggest that antibiotic treatment altered the composition of CLF and reduced its stimulatory effect partially due to a significant reduction in LPS levels. This is consistent with the fact that the antibiotics chosen have primarily excellent gram negative coverage.

### Short-Chain Fatty Acid (SCFA) and Amino Acid Metabolomic Analysis of CLF Reveals Antibiotic-Induced Changes in Acetate and Propionate Levels

While comparing the effects of LPS and CLF, we noted the similarities in responses to HDAC inhibition by millimolar levels of the SCFA, butyrate, as described earlier. However, we also noted that lower dose (micromolar) butyrate displayed a paradoxical effect in enhancing LPS mediated and CLFmediated inflammation for IL-6 and other specific inflammatory genes (**Supplementary Figure S3**). This paradoxical effect was more pronounced during stimulation with lower doses of LPS (data not shown). Furthermore, when we challenged HUVEC to increasing levels of antibiotic CLF, it appeared that antibiotic CLF contained inhibitory factors in contrast to control CLF (**Figure 3A**). Since SCFAs could alter the inflammatory potential of LPS, we hypothesized that perhaps

endothelial cells (EC) were stimulated with LPS or CLF (increasing doses from 1 to 100 mg/mL) and IL-6 measured. (C) LPS or CLF stimulation of mouse alveolar macrophages (AM) were both inhibited with polymyxin B (PMB) treatment. P-values are represented as follows: <sup>∗</sup> < 0.05; ∗∗ < 0.01; ∗∗∗ < 0.001; ∗∗∗∗ < 0.0001. For multiple comparisons using one-way ANOVA, p-values are represented as follows: α < 0.05; αα < 0.01; ααα < 0.001; αααα < 0.0001. For comparisons against an untreated or control condition, p-values are represented as follows: δ < 0.05; δδ < 0.01; δδδ < 0.001; δδδδ < 0.0001. All experiments were conducted at least twice and representative data are shown.

high levels of an inhibitory SCFA were present in antibiotic CLF that could reconcile with the corresponding diminished lung inflammation we had previously observed in vivo (Prakash et al., 2015). CLF itself contains SCFAs as well as other metabolites, so we measured SCFA and amino acid levels present in CLF samples used thus far (Wild-type C57BL/6 CLF, Control C3H/HeOuJ CLF, Antibiotic C3H/HeOuJ CLF) (**Supplementary Table ST1**). Focusing on three important SCFAs, namely, acetate, propionate and butyrate, we noted that antibiotic treatment did not change butyrate levels, but instead drastically altered the propionate:acetate ratio within CLF (**Figure 3B**). Similarly to butyrate, propionate can also act as an HDAC inhibitor at millimolar concentrations (Chang et al., 2014).

## Propionate Effects on LPS-Exposed Cells in vitro Also Depend on Concentration and Requires LPS Sensing by TLR4

We next tested the effects of low- and high-dose propionate on LPS responses in mouse AMs and ECs. We observed that low and high-dose propionate effects on ECs and AM LPS responses largely mirrored the effects of low and high-dose butyrate (**Figures 4A,B**), namely, low-dose propionate was able to augment LPS inflammatory responses and high-dose blunted LPS responses. This effect was dependent on TLR4 and not caspase-11 sensing of LPS (**Supplementary Figure S4**). TLR2 ligands' inflammatory effects could also be similarly modulated by low dose and high dose SCFAs (**Supplementary Figure S5**). When we used caspase-11 and TLR4 mutant ECs to study CLF inflammatory effects in the presence of low and high-dose SCFAs, we noted that SCFA effects were largely dependent on the ability of cells to recognize extracellular LPS through TLR4 (data not shown). This indicated to us that LPS (and/or other gut microbiome-derived TLR4 ligands) is likely the major bacterial ligand present in our CLFs.

# Propionate:Acetate Ratios Can Result in Augmentation or Reduction of LPS Inflammatory Responses in vitro

SCFAs are not uniformly diluted into the systemic circulation after absorption in the gut. In fact, various groups have estimated the levels absorbed by the transit organs between the colonic

αααα < 0.0001. For comparisons against an untreated or control condition (e.g., 10 mcg/mL CLF treatment), p-values are represented as follows: δ < 0.05;

contents and systemic circulation (Cummings et al., 1979, 1987; Boets et al., 2015). Based on the estimates for intestinal and liver absorption reported by Boets et al. (2015), namely 95% butyrate absorption and 90% propionate, respectively, we estimated the propionate:acetate ratio in the peripheral circulation for control and antibiotic-treated mice (**Supplementary Figure S6**). To examine the inflammation modulating effects of these SCFAs in circulation and in the lung, we exposed mouse lung macrophages in vitro to LPS in the presence of varying propionate:acetate ratios. We hypothesized that low propionate:acetate ratios would augment the LPS responses while high ratios would inhibit inflammation. After subjecting AMs to LPS in conjunction with inverse ratios of propionate and acetate, we observed exactly that (**Figure 4C**). Therefore, these in vitro data strikingly recapitulated our in vivo observations (Prakash et al., 2015) that antibiotictreatment in mice that targeted mostly gram-negative bacteria within the gut microbiome and resulted in a switch from low to high propionate:acetate ratios within CLF, which in turn, similarly altered lung inflammatory responsesin vivo and in vitro.

# In vivo Lung Pretreatment With High Dose Propionate Results in Worsening of S. aureus Pneumonia

Acetate levels within the peripheral circulation likely remain constant given their use and production by most gut microbiome species and the lack of major absorption by the intestine and liver (in contrast to propionate and butyrate). Therefore to establish in vivo significance to our in vitro propionate findings, we chose to address the question of whether altering propionate levels in the lung could affect lung immune responses. We pretreated C57BL/6 wild type mice with high and low dose propionate [1 and 0.1 mM, intratracheally (IT)] and 2 h later subjected them to a S. aureus pneumonia with 10<sup>8</sup> CFU (IT). We had previously examined the kinetics of this infection and found that maximal infection was detected 6 h after inoculation and at 24 h the infection had resolved (data not shown). IVIS <sup>R</sup> in vivo imaging was used to image the luminescent S. aureus within the lungs in vivo (**Figure 5A**) and ex vivo (**Figure 5B**) 6 h after infection. Low dose propionate pre-treatment did not alter the levels of S. aureus present at 6 h as compared to control mice. However, the high dose propionate group displayed two- to three-fold greater levels of bacteria within the lungs (**Figure 5C**), suggesting that SCFA-mediated attenuation of lung inflammation prior to and during infection diminished the control of the bacterial pneumonia.

### Propionate-Producing Bacteria Are Enriched in Gut Microbiomes of Mice With Attenuated Lung Inflammatory Responses

To attempt to correlate physiologic/pathologic responses to lung IR with specific composition of resident gut microbiota within

δδ < 0.01; δδδ < 0.001; δδδδ < 0.0001.

Control CLF ratios of propionate:acetate (1:19) as well as estimated ratios in peripheral circulation (1:99; see Supplementary Figure S7) and antibiotic CLF ratios of propionate:acetate (5:1) as well as estimated ratios in peripheral circulation (5:4; see Supplementary Figure S7). IL-6 levels were measured in the supernatant by ELISA after overnight treatment. For multiple comparisons using one-way ANOVA (A,B), p-values are represented as follows: α < 0.05; αα < 0.01; ααα < 0.001; αααα < 0.0001. For comparisons against an untreated or control condition (e.g., LPS only treatment, C), p-values are represented as follows: δ < 0.05; 2δ < 0.01; 3δ < 0.001; 4δ < 0.0001. All experiments were conducted at least twice and representative data are shown.

a given mouse, we subjected C57BL/6 wild type mice to an 8-week course of either control water or antibiotics (Neo/PMB) and collected stool samples weekly. At the end of 8 weeks, the mice underwent lung IR injury and lung injury was assessed by histology. Surprisingly, we observed greater than expected variation in the mice from each group vis-à-vis their lung injury (**Supplementary Figure S7** and **Supplementary Table ST2**). To understand the source of this variation as well as the effects of the antibiotic treatment on the gut microbiome, we analyzed stool samples after 7 weeks of antibiotic-treatment and prior to lung IR surgery by 16S sequencing. Antibiotic treatment resulted in significantly reduced alpha diversity (increased evenness with reduced richness and alpha diversity by the Faith phylogenetic diversity index but not by Shannon and Simpson phylogenetic diversity indices) (**Supplementary Figure S8**). Firmicutes dominated the microbiome of the control group, while levels of Verrucomicrobia and Bacteriodetes were enriched after the course of antibiotics (**Supplementary Figure S9**). In mice that had a higher lung injury score, alpha diversity was significantly reduced (increased evenness with reduced richness and alpha

diversity by the Faith and Simpson phylogenetic diversity indices) (**Supplementary Figure S10** and data not shown). Beta diversity was also significantly different between the control and antibiotic treatment groups as well as between those mice that had a higher vs. lower lung injury score. Differences in bacterial composition was estimated to account for ∼15–20% of the observed lung injury differences (data not shown). Additionally, differences in bacterial community composition were significantly influenced by phylogenetic relatedness, bacterial/presence absence, relative abundance and presence of more and lesser abundant taxa (data not shown). Interestingly, specific OTUs (operational taxonomic units) were enriched (by a Three Model Approach including Negative Binomial Regression, Zero-inflated Negative Binomial Regression, and Poisson Regression with a false discovery rate corrected p-value < 0.2) in mice with lower lung injury scores. Eighty percent of the significantly enriched OTUs were members of the phylum Firmicutes, and the orders Bacteriodales and Clostridiales, with the family Lachnospiraceae strongly represented (**Supplementary Figure S10**). We observed that low lung injury correlated with higher levels of Lachnospiraceae (**Figure 6A**), and this correlation was significant (r<sup>2</sup> 0.2123, **Figure 6B**). Coincidently, Bacteriodales and Clostridiales (specifically Lachnospiraceae) are the primary producers of propionate within the gut microbiome (Reichardt et al., 2014; Salonen et al., 2014; Louis and Flint, 2017). We also confirmed by metabolomic SCFA analysis that stool of mice with the highest levels of Lachnospiraceae (9-fold higher Lachnospiraceae than the comparison group) contained ∼150-fold greater levels of propionate (45 mM vs. 300 µM, data not shown).

# DISCUSSION

The main findings of these studies indicate that levels of microbiome-derived metabolites, specifically propionate, can influence lung immune and inflammatory responses in vivo

and in vitro. We began by focusing on colonic lumen filtrate (CLF) from control mice or from mice that received antibiotics given that we had previously showed that antibiotic treatment caused significant alterations in in vivo and ex vivo lung inflammation to sterile injury (Prakash et al., 2015). First, we provided evidence that LPS was one key component of CLF that resulted in immune activation of cells, and that antibiotic CLF contained less LPS and caused less immune stimulation. We observed an unexpected proinflammatory effect of micromolar concentrations of propionate and butyrate on LPS responses in vitro. Compared to control mice, we observed a significant enrichment of CLF propionate concentration in antibiotic-treated mice relative to acetate. Speculating that high propionate:acetate ratios in vivo might explain observed attenuated inflammatory responses after lung IR in antibiotictreated mice, we exposed lung cells in vitro to low and high propionate:acetate ratios and reproduced pro- and antiinflammatory modulation of LPS responses, respectively. To confirm the anti-inflammatory effect of high propionate in vivo, we pretreated mouse lungs with high dose propionate, which resulted in worse control of a S. aureus pneumonia (unlike low dose propionate). Finally, we found gut microbiome enrichment of high propionate-producing Lachnospiraceae among mice with reduced lung inflammation phenotypes. Collectively, the experimental data presented here strongly support a gut-lung immune axis model in which gut-derived SCFAs in combination with CAMPs in circulation profoundly influence lung physiology and immunity.

Evidence in currently published studies support the concept that gut microbiota can influence non-gut disease states. Germfree mice often have local and systemic physiologic perturbations (reviewed in Erturk-Hasdemir and Kasper, 2013). However, these differences have largely been attributed to altered development of the immune system in the absence of early establishment and maturation of the gut commensal microbiome. Seemingly in conflict with this microbiome-driven immune development concept, other studies have shown that repopulating or altering the gut microbiome in adult mice can rapidly alter their physiology and disease processes (Shreiner et al., 2015). This suggests that highly accessible and circulating factors from the gut microbiome have the capacity to change immune responses in remote host locations (Vieira et al., 2015). Candidate factors that fit this factor profile well include microbially released or dietary metabolites. In fact, segmented filamentous bacteria which can ferment non-digestible starches to produce butyrate and propionate have been observed to confer protection in a number of disease models in mice (Gaboriau-Routhiau et al., 2009 and reviewed in Meyerholz et al., 2002; Ericsson et al., 2014).

Metabolites have been reported to have strong immunomodulatory effects on the host (Arpaia et al., 2013; Chang et al., 2014). Specific metabolites used and produced by the commensal microbiome include short, medium, and long-chain fatty acids, indoles, carbohydrates, gylcolipids, bile acids, vitamins, and other co-factors (reviewed in Shapiro et al., 2014; Levy et al., 2015). These metabolites can regulate different aspects of cell function based on engaging cognate receptors as well as epigenetically. SCFAs, namely acetate, propionate, and butyrate, have known roles as sources of energy (butyrate), effectors of epigenetic changes (propionate and butyrate), and signal transduction (acetate, propionate, and butyrate). At millimolar concentrations, both butyrate and propionate can act as histone deacetylase (HDAC) inhibitors, which silence the transcription of specific inflammatory genes (Chang et al., 2014). Butyrate, in its role as an energy source for colonocytes, can also form part of the switch that converts cells from a metabolic program based on oxidative phosphorylation to one based on glycolysis (i.e., the Warburg effect) (reviewed in Burgess, 2012; Donohoe et al., 2012). At micromolar concentrations, SCFAs

engage and signal through free fatty acid receptors (FFARs) and this engagement has been shown to be important not only locally but also for immune responses in niches not directly in contact with the gut microbiome, such as synovial joints (Vieira et al., 2015) and distal lung airways (Trompette et al., 2014). Other roles, yet to be discovered, are also likely.

We propose that pulmonary immune responses may be calibrated by the levels of commensal-derived SCFAs and CAMPs that transit through the lung and result in immune priming or dampening. In support of this concept, an in vitro study has suggested a similar priming phenotype to low-dose SCFA that we report here (Mirmonsef et al., 2012). Other examples of low-dose SCFA effects on immune responses exist, including one in which low levels of circulating acetate specifically promote intestinal IgA responses to microbiota through FFAR2 (GPR43) (Wu et al., 2017). We further propose that low propionate:acetate levels in the lungs create immune priming to support proinflammatory responses. This priming may thus contribute to the generation of healthy baseline immune tone in the homeostatic lung. Conversely, switching to high propionate:acetate levels through gut dysbiosis may reprogram the lung and invoke pathologic or abnormal responses to sterile and infectious challenges.

Studies in mice describe how the gut microbiome is protective against pneumococcal and S. aureus pneumonia (Clarke et al., 2010; Gauguet et al., 2015; Brown et al., 2017). Therefore, direct and indirect mechanisms by which the gut commensal microbiome communicates with the lung are of great interest. Some key questions that still need to be answered include the identity of the free-fatty acid receptors (FFARs) that are important for these immune effects of SCFAs and whether or not those FFARs are druggable. Our data that propionateproducing Lachnospiraceae are significantly associated with a low lung inflammation phenotype in vivo confirm our in vitro cell culture data. Since acetate is widely produced by most if not all commensal bacteria and its levels are largely stable in circulation (Reichardt et al., 2014; Louis and Flint, 2017), alterations of propionate production, say by antibiotic exposure, may therefore significantly alter lung immune responses. We have also shown that limiting lung inflammation downstream of lung sterile reperfusion injury can result in a disseminated experimental pneumonia in mice (Tian et al., 2017). Therefore, caution must be exercised when manipulating gut microbiome lest dampening or augmenting lung inflammatory responses cause unintended consequences.

This study has its limitations. SCFAs are just one of many metabolite classes that may have physiologic and pathologic effects, and these include long-chain fatty acids, bile acids, succinate, lactate, and aromatic amino acids (Krishnan et al., 2015) – none of which were studied here. We used the SCFA composition of the stool to estimate SCFA levels in the lung and performed experiments directly administering propionate to the mouse lung. However, direct measurement of SCFA levels in mouse tissue and plasma, while technically challenging, would be more definitive. Our in vitro approach to study the effects of SCFAs on lung immune cell and endothelial cell responses will need in vivo correlation perhaps through the use of lung-specific FFAR2 and FFAR3 conditional knockout mice. We used human cells early in our studies to confirm that the effects we observed were not limited to mouse cell lines and could be translatable, but further correlation with human data is clearly required. The pulmonary microbiome contributes to lung immunity in disease states such as COPD and though unlikely may also do so in healthy lungs (Dickson et al., 2016). Finally, our work does not exclude the possibility of other members of the microbiome (besides Lachnospiraceae) playing important roles as well through their metabolites or released factors that could also modulate lung immunity.

We conclude that specific SCFA metabolites, namely propionate and possibly acetate, are important contributors to the gut-lung immune axis of communication that may augment and suppress lung immune responses. Specifically, high propionate:acetate levels in the lung may be beneficial in situations where lung inflammation suppression is the goal, such as following lung transplant when ischemia reperfusion injury is a threat to graft survival. On the other hand, high propionate:acetate levels that may result from poor antibiotic stewardship and resulting gut microbiome dysbiosis could adversely affect the course of bacterial pneumonia. Propionate may be an unusual non-native compound (believed to be only derivable from the diet or gut microbiome) that has perhaps been adapted for specialized functions making it a fascinating topic for further investigation. Overall, we believe that SCFA levels achieved in the pulmonary parenchyma may be critical for "healthy" or "normal" primed immune responses to lung injury and as such determine the establishment of an overall homeostatic resting lung immune tone. The identification of simple but powerful metabolites and the microbiota from which they originate as important controllers of the gut-lung immune axis may help explain the vast pathophysiologic diversity of human lung injury responses as well as the ever expanding contribution to human heath of diet, lifestyle, environment, immune history, antibiotic and medication use, and hospital-setting exposure. By expanding this knowledge base and understanding, we hope to pave the way toward devising strategies to positively modulate lung immune responses within diverse clinical scenarios.

#### ETHICS STATEMENT

All mouse studies were approved from an ethical and methodological standpoint by the Institutional Animal Care and Use Committee (IACUC) at the University of California, San Francisco and followed the ARRIVE guidelines.

# AUTHOR CONTRIBUTIONS

XT performed all in vitro and some in vivo experiments, analyzed data, and edited manuscript. JH assisted with experimental design, analyzed data, and edited manuscript. AH, HC, and KF engineered and developed the bacterial strains used in the study. AP designed all the experiments, performed mouse surgeries, analyzed data, and wrote and edited manuscript.

#### FUNDING

Research reported in this publication was supported by the National Institutes of General Medical Sciences of the National Institutes of Health under Award Number K08GM110497. Additional support was provided by the Departments of Anesthesiology and Perioperative Care at University of California San Francisco and San Francisco General Hospital.

#### ACKNOWLEDGMENTS

fmicb-10-00159 March 25, 2019 Time: 15:50 # 13

We would like to acknowledge the following individuals for assistance with providing reagents, mice, advice, helpful

#### REFERENCES


discussions, and critical reading and editing of the manuscript: Douglas Fadrosh (UCSF), Susan Lynch (UCSF), Michael Matthay (UCSF), Mervyn Maze (UCSF), Clifford Lowell (UCSF), Kevin Wilhelmsen (MarinBiologics), Milo Vassallo (Methodist Hospital, Brooklyn, NY), Gautam Prakash (Arlington, VA). The authors would also like to acknowledge the role of THFC (COYS) on these studies.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.00159/full#supplementary-material

Erturk-Hasdemir, D., and Kasper, D. L. (2013). Resident commensals shaping immunity. Curr. Opin. Immunol. 25, 450–455. doi: 10.1016/j.coi.2013.06.001


by commensal microbiota in mice. Shock 44, 272–279. doi: 10.1097/SHK. 0000000000000415


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

Copyright © 2019 Tian, Hellman, Horswill, Crosby, Francis and Prakash. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# LL37 Inhibits *Aspergillus fumigatus* Infection via Directly Binding to the Fungus and Preventing Excessive Inflammation

Xiao-Li Luo1†, Jian-Xiong Li 1†, Hua-Rong Huang<sup>1</sup> , Jie-Lin Duan<sup>2</sup> , Ruo-Xuan Dai <sup>1</sup> , Ru-Jia Tao<sup>1</sup> , Ling Yang<sup>1</sup> , Jia-yun Hou<sup>3</sup> , Xin-Ming Jia<sup>2</sup> \* and Jin-Fu Xu<sup>1</sup> \*

*<sup>1</sup> Department of Respiratory and Critical Care Medicine, Shanghai Pulmonary Hospital, Tongji University School of Medicine, Shanghai, China, <sup>2</sup> Clinical Translational Research Center, Shanghai Pulmonary Hospital, Tongji University School of Medicine, Shanghai, China, <sup>3</sup> Zhongshan Hospital Institute of Clinical Science, Shanghai Institute of Clinical Bioinformatics, Zhongshan Hospital, Fudan University, Shanghai, China*

#### *Edited by:*

*Sudhanshu Shekhar, University of Oslo, Norway*

#### *Reviewed by:*

*Agostinho Carvalho, University of Minho, Portugal Huafeng Wang, California Institute for Biomedical Research, United States*

#### *\*Correspondence:*

*Jin-Fu Xu jfxucn@gmail.com Xin-Ming Jia jiaxm@tongji.edu.cn*

*†These authors have contributed equally to this work*

#### *Specialty section:*

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

*Received: 13 October 2018 Accepted: 04 February 2019 Published: 20 February 2019*

#### *Citation:*

*Luo X-L, Li J-X, Huang H-R, Duan J-L, Dai R-X, Tao R-J, Yang L, Hou J-y, Jia X-M and Xu J-F (2019) LL37 Inhibits Aspergillus fumigatus Infection via Directly Binding to the Fungus and Preventing Excessive Inflammation. Front. Immunol. 10:283. doi: 10.3389/fimmu.2019.00283* The incidence of *Aspergillus fumigatus* infection and the rate of resistance to antifungal drugs have sharply increased in recent years. LL37 has been reported as a host defense peptide with broad-spectrum antibacterial activities. However, the role of LL37 during *A. fumigatus* infection remains unclear. Here, we examined the interaction between LL37 and *A. fumigatus* and found that synthetic LL37 could directly bind to the surface of *A. fumigatus*, disrupting the integrity of the cell wall *in vitro*. LL37 inhibited mycelial growth in a concentration-dependent manner, rather than fungicidal effect even at high concentration (e.g., 20µM). Interestingly, low concentrations of LL37 (e.g., 4µM) significantly attenuated mycelial adhesion and prevented the invasion and destruction of epithelial cells. Following LL37 treatment, the levels of proinflammatory cytokines released by *A. fumigatus*-stimulated macrophages decreased significantly, accompanied by downregulation of M1 type markers. In a mouse model of pulmonary *A. fumigatus* infection, LL37-treated mice showed lower amounts of fungi load, moderate pathological damage, and reduced proinflammatory cytokines. Further, LL37 transgenic mice (LL37+/+) were examined to investigate the effects of endogenous LL37 in an *A. fumigatus* infection model and showed lower susceptibility to *A. fumigatus* infection in comparison with wild-type mice. In addition, LL37 also played a protective role in an immunosuppressed mouse model of *A. fumigatus* infection. Thus, LL37 inhibits *A. fumigatus* infection via directly binding to mycelia and reducing excessive inflammation. LL37 or its analogs may therefore constitute potential drug components for *A. fumigatus* infection.

Keywords: LL37, *Aspergillus fumigatus*, mycelium, inflammation, adhesion

# INTRODUCTION

Aspergillus fumigatus, an opportunistic pathogen widely distributed in nature, is the leading cause of pulmonary Aspergillosis (1). Pulmonary Aspergillosis includes three subtypes—specifically, chronic pulmonary aspergillosis (CPA), allergic bronchopulmonary aspergillosis (ABPA) and invasive pulmonary aspergillosis (IPA)—which are associated with different immune statuses of susceptible hosts. The morbidity and mortality of IPA have sharply increased in recent decades

**49**

due to a rise in immunocompromised individuals, such as patients receiving organ transplants or chemotherapy (2). Furthermore, the emergence of antifungal drug resistance limits the effectiveness of clinical treatment (3–5). Even worse, excessive inflammation and severe tissue damage can deteriorate a patient's condition and increase treatment difficulty (6).

LL37, a short fragment composed of 37 amino acids, is the unique member of human cathelicidin antimicrobial peptides (CAMPs). It presents with broad-spectrum antimicrobial activity against various pathogens, including prokaryotic, and fungal organisms (7, 8). Its positive charge allows it to bind to negatively charged phospholipid membranes of prokaryotic cells, prompting membrane penetration, the formation of transmembrane pores, and bacterial lysis (9). On the other hand, LL37 possesses diverse modulating properties on immune system such as the recruitment of inflammatory cells and the release of inflammatory factors, showing both proinflammatory and anti-inflammatory effects (10). LL37 is expressed in a variety of immune cells and epithelial cells. In different types of cells and tissues, LL37 has different physiological concentrations and often increases during infection.

Recently, a few studies have shown that LL37 expression is significantly upregulated in corneal epithelium and nasal tissue in response to A. fumigatus (11, 12), suggesting that LL37 may play an important role in A. fumigatus infection. Therefore, in the present investigation, we explored the possible effects of LL37 against A. fumigatus infection. We tested whether or not LL37 could bind to A. fumigatus, destroy cell wall structures and inhibit mycelium growth and adhesion in vitro. To test the immunomodulatory effects of LL37, the release of tumor necrosis factor alpha (TNF-α) and interleukin (IL-6) from A. fumigatusstimulated bone marrow-derived macrophages (BMDMs) after LL37 treatment was evaluated. Subsequently, exogenous LL37 treated mice and LL37+/+ mice were involved, respectively to determine the in vivo effects of exogenous and endogenous LL37 on fungi clearance, pathological injury, neutrophil infiltration, and cytokine production during A. fumigatus infection. Overall, this study demonstrates that LL37 not only directly inhibits A. fumigatus hyphae growth and adhesion but also prevents A. fumigatus-induced excessive inflammation, thus providing new evidence for the dual therapeutic value of LL37 against A. fumigatus infection.

# MATERIALS AND METHODS

#### Animals

Specific pathogen-free C57BL/6 mice and FVB mice breeding pairs were purchased from the SLAC Laboratory Animal Center (Shanghai, China). LL37+/+ mice were produced via the microinjection of linearized plasmids expressing hCAP18/LL37 into fertilized eggs of mice bred with an FVB genetic background (**Figures S1A,B** in Supplementary Material). All of the mouse strains were housed in specific pathogen-free conditions within an animal care facility (Center of Laboratory Animal, Tongji University, Shanghai, China) until the day of sacrifice. All of the animal experiments were performed under the guidance and with approval from the Institutional Animal Care and Use Committee of Tongji University (Permit Number: TJLAC-015-002).

#### Reagents

Human cathelicidin LL37 (LLGDFFRKSKEKIGKEFKRIVQRIK DFFRNLVPRTES) with a purity of 95% was purchased from Rockland Immunochemicals (Pottstown, PA, USA). Scrambled form sL37 (RSLEGTDRFPFVRLKNSRKLEFKDIKGIKREQFVK IL) with a purity of 95% was synthesized by GL Biochem (Shanghai, China). Anti-cathelicidin antibody was purchased from Abcam (Cambridge, UK).

## *A. fumigatus* Strains and Culture

Conidia (A. fumigatus train, Af293) harvest and growth into swollen conidia and hyphae were performed as described previously (13). Briefly, fungi were inoculated on Sabouraud Dextrose Agar slant and cultured at 37◦C for 7 days. Conidia were collected with phosphate-buffered saline (PBS), filtered through a 40-µm nylon mesh, then stored at 4◦C for use. To obtain swollen conidia (SC) and hyphae, resting conidia (RC) were incubated in Roswell Park Memorial Institute (RPMI)-1640 media at 37◦C for 8 h to achieve swelling and for an additional 2 h to achieve germination.

### Cell Culture

BMDMs from mice were prepared as previously described (14). Bone marrow was extracted from the femur and tibia of 6–8 week-old female C57BL/6 mice. Cells were centrifuged following the removal of erythrocytes and then were differentiated into BMDMs in Dulbecco's Modified Eagle medium supplemented with 10% fetal bovine serum, 30% L929 supernatant, 1% antibiotic-antimycotic, and 0.1% β-Mercaptoethanol.

#### Transmission Electron Microscopy (TEM)

Aspergillus fumigatus conidia were treated with 4µM LL37 or sL37 and incubated in RIPM-1640 medium for 24 h. Subsequently, the mycelia were pelleted by centrifugation and prefixed in a solution of 5% glutaraldehyde in 0.1 M sodium cacodylate buffer for 2 h. Samples were then washed three times with 0.1 M sodium cacodylate buffer and postfixed with 1% osmium tetroxide for 3 h and were dehydrated with increasing concentrations of ethanol or acetone (i.e., 50, 70, 90, 100%), respectively. Following embedding and fixation, samples were cut into ultrathin sections using an ultramicrotome. After staining with uranyl acetate and lead citrate, the ultrathin sections were viewed under a JEOL JEM-1230 (80KV) transmission electron microscope.

#### Hyphae Growth Inhibition Assay

Aspergillus fumigatus conidia were incubated in RPMI-1640 medium with different concentrations of LL37 or sL37 at 37◦C for 12 h. Hyphae length was then measured under microscopy. Each group included at least 10 visual fields containing ≥50 hyphae.

#### Adhesion Assay

Adhesion assay was performed as previous described (15). A. fumigatus conidia were incubated in a 96-well plate with RPMI-1640 media in the presence of different concentrations of LL37 or sL37 at 37◦C for 24 h. The supernatant was removed and then the wells were washed three times with PBS. Adhesive capacity was estimated by staining the biofilms that had not been washed off with 0.5% crystal violet for 15 min. Then excess stain was washed with PBS for three times. Afterwards, the biofilms were decolorized with 95% ethanol. The density of the biofilms was analyzed by determining the absorbance of the decolorized solution at 570 nm. At the same time, the wells both before and after washing were photographed under a microscope.

#### Epithelial Cell Damage Assay

A mouse alveolar epithelial cell line (MLE12) were plated on a 48-well plate in a monolayer formation and infected with A. fumigatus conidia in the presence of LL37 or sL37. Following incubation for 16 h, the supernatant was collected and transferred to a 96-well plate. The LDH released in the supernatant was detected using the CytoTox 96 <sup>R</sup> non-radioactive cytotoxicity assay kit (Promega Corp., Madison, WI, USA) according to the manufactures' instructions. Then the corrected values in the following formula were used to compute percent cytotoxicity:

Percent cytotoxicity = 100 × Experimental LDH Release (OD 490) Maximum LDH Release (OD 490)

The Maximum LDH Release means a positive control (i.e., treatment with detergent).

#### Binding of LL37 and *A. fumigatus* Analysis

Aspergillus fumigatus RC, SC or hyphae were incubated with or without 4µM of LL37 dissolved in PBS for 30 min at room temperature. The samples were then washed three times with PBS and fixed in 4% paraformaldehyde for an hour before being blocked with 5% fetal bovine serum albumin (BSA) for an hour. Following washing with PBS for three times, samples were incubated with anti-LL37 antibody overnight at 4 ◦C. Following washed with PBS for three times, samples were incubated with a FITC-labeled secondary antibody for an hour at room temperature. The RC samples were analyzed using a flow cytometer (Becton Dickinson, Franklin Lakes, NJ, USA) while SC and hyphae samples were visualized and imaged using a confocal microscope (Nikon Inc., Tokyo, Japan).

#### Cytokines Secretion

The 3 × 10<sup>5</sup> BMDMs were plated on a 48-well plate in a monolayer and infected with 1.5×10<sup>6</sup> UV-killed A. fumigatus swell conidia or 3 × 10<sup>5</sup> hyphae in the presence of different concentrations of LL37 or sL37. After stimulation for 16 h, the supernatant was collected. The concentration of TNFα and IL-6 in the supernatant was detected using ELISA kits (eBioscience, San Diego, CA, USA) according to the manufactures' instructions.

#### *In vitro* Killing Assay

Here, the 3 × 10<sup>6</sup> BMDMs were plated on a 12-well plate in a monolayer and infected with 6×10<sup>5</sup> viable A. fumigatus conidia in the presence or absence of 4µM LL37 or sL37. After incubation for 6 h at 37◦C, the fungi were scraped off and diluted with sterile PBS. Afterwards, SDA plates were inoculated with diluted solution and cultured at 37◦C for 48 h. The killing capacity was estimated by counting the number of fungi colonies.

#### Phagocytosis Assay

A phagocytosis assay was based on and developed as described previously(16). Briefly, A. fumigatus conidia were pretreated with a Fluoro TagTM FITC Conjugation Kit (Sigma-Aldrich, St. Louis, MO, USA) and peritoneal macrophages were isolated from C57/BL6 mice as described previously (17). Then peritoneal macrophages were cultured on a 12-well plate in a monolayer and each well was inoculated with FITC-labeled A. fumigatus conidia to achieve a multiplicity of infection (MOI) equal to 10. After 1 h of coincubation at 37◦C with 5% CO2, the wells were washed with PBS before digested by 0.25% trypsin to collect the adherent cells. Then, the collected macrophages were labeled with anti-CD11b PerCP cy5.5 (M1/70) (eBioscience, San Diego, CA, USA) and samples were analyzed using a BD flow cytometer. The percentages of phagocytosis were calculated by the ratio of FITC<sup>+</sup> macrophages to all macrophages.

## Murine Models of Pulmonary Aspergillus Infection

Mice were anesthetized by isoflurane inhalation, and then (2- 10) × 10<sup>6</sup> A. fumigatus conidia in 35 µl of PBS were instilled into the trachea by pressing tongue intratracheal instillation. For the immunosuppressed model, mice were administered with 40 mg/kg of the corticosteroid, triamcinolone acetonide (TargetMol, Boston, MA, USA) injected subcutaneously 1 day prior to infection, as previously described (18). At the indicated times, mice were killed and lung tissues were isolated for detection of CFU, inflammatory cells, and cytokines as well as histopathological analysis.

#### Fungal Burden Analysis

Mice were killed at the indicated times after infection and lungs were dissected carefully, excised, and homogenized in PBS. For CFU, the homogenates were serially diluted and spread onto SDA plates. After incubated at 37◦C overnight, the colonies were counted and normalized to lung weights.

#### Histopathological Analysis

Lung tissues isolated from mice were fixed in 10% buffer formalin, dehydrated, and embedded in paraffin. Then the lungs were cut into sections and stained with hematoxylin & eosin (HE) or Gomori's methenamine silver stain (GMS) according to standard staining procedures at pathology platform of Servicebio Technology, Wuhan, China.

#### Flow Cytometry Analysis

Lung tissues from mice were digested with collagenase to establish a single-cell suspension as described previously (13). Briefly, lungs were dissected and cut into very small pieces before incubated in collagenase digestion solution (10 mg collagenase type II in 10 ml PBS) for 1 h at 37◦C. Next, the digested lungs were sieved through 40 mm sieve and treated with RBC lysis buffer. Then, cells were washed and stained with the following monoclonal antibodies: anti-mouse CD45-FITC (30- F11), anti-mouse Ly6G-BV421 (1A8), anti-mouse CD11b-PerCP cy5.5 (M1/70) (eBioscience, San Diego, CA, USA), and antimouse CD11c-APC (HL3), anti-mouse SiglecF-PE (E50-2440) (BD Biosciences, San Jose, CA, USA), after which the cells were analyzed by a flow cytometer (Becton Dickinson, Franklin Lakes, NJ, USA). Next, the data were analyzed with the Flow-Jo software: First, all of the cells were defined via FCS-A and SSA-A and then single cells were determined via SSC-A and FCS-W. Afterwards, we defined CD45<sup>+</sup> cells as immune cells in lung tissue, from which Ly6G+CD11c<sup>+</sup> cells were gated as neutrophils and SiglecF+CD11c<sup>+</sup> cells were gated as macrophages, respectively (**Figure S2** in Supplementary Material).

# RNA Extraction and Real-Time Quantitative RT-PCR

Total RNA was isolated using TRIzol reagent (Takara, Kusatsu, Japan). Reverse transcription was performed by utilizing a Primescript RT reagent kit (Takara, Kusatsu, Japan). Realtime quantitative PCR measurement was performed using SYBR Green reagent kit (Vazyme, Nanjing, China) as per the manufacturer's instructions, and the ABI 7500 sequence detection system. The amounts of transcript were normalized to GAPDH. Relative mRNA expression was plotted as fold changes calculated using the 11Ct method. The primers used for realtime RT-PCR were as follows: iNOS (forward: 5′ - ACATCGA CCCGTCCACAGTAT−3 ′ ; reverse: 5′ - CAGAGGGGTAGGCT TGTCTC−3 ′ ); CXCL9 (forward: 5′ - ATCTCCGTTCTTCAGT GTAGCAATG-3′ ; reverse: 5′ - ACAAATCCCTCAAAGACCTC AAACAG-3′ ); CXCL10 (forward: 5′ - AGGGGAGTGATGGAG AGAGG-3′ ; reverse: 5′ - TGAAAGCGTTTAGCCAAAAAAGG-3 ′ ); GAPDH (forward: 5′ - AGGTCGGTGTGAACGGATTTG-3′ ; reverse: 5′ - TGTAGACCATGTAGTTGAGGTCA-3′ ).

#### Statistical Analysis

Statistical analysis was performed using GraphPad Prim 5.0 software. The graph represents at least three dependent experiments and data were presented as mean ± Standard Deviation (SD). Log-rank testing was used to evaluate the equality of survival curves. The other statistical differences were determined by the two-tailed unpaired t-test and one-way ANOVA. It was considered statistically significant when p < 0.05.

# RESULTS

### LL37 Binds Directly to *A. fumigatus* and Destroys Cell Wall Integrity

To investigate the direct antimicrobial activity of LL37 against A. fumigatus, we initially performed immunofluorescence staining of LL37. The direct binding of LL37 to the surface of A. fumigatus resting conidia was proved by flow cytometer after 30 min of coincubation, as well as swell conidia and hyphae observed under confocal microscopy (**Figures 1A–C**). Furthermore, to assess the effect of LL37 on the structure of A. fumigatus, mycelia grown from conidia in the presence of 4µM LL37 were visualized by transmission electron microscope. The images revealed obvious alterations of cell wall integrity prompted by LL37—that is, the cell walls were more likely to aggregate into clumps and the apical plasma membrane to separate from the cell wall to form folds (**Figure 1D**). Together, these data indicate that LL37 binds directly to A. fumigatus and destroys cell wall integrity.

# LL37 Inhibits *A. fumigatus* Hyphae Growth and Adhesion *in vitro*

To explore the effect of LL37 on the biological activity of A. fumigatus, we measured the hyphal length of A. fumigatus following 10 h of incubation with LL37 or sL37 at concentration gradients. Our results showed that hyphal growth was significantly inhibited by LL37 in a concentration-dependent manner (**Figure 2A**). Unexpectedly, we noticed that A. fumigatus were still viable even at high concentration of 20µM LL37, which indicates that LL37 plays a role in inhibiting the growth of mycelium rather than sterilization. Furthermore, this inhibiting effect on hyphae growth resulted in impaired biofilm formation (**Figure 2B**). Crystal violet staining was applied to quantify the adherent biofilm density and it was shown that adherent biofilm formation was markedly attenuated even at a low dose of LL37 treatment (≤ 4 µM) (**Figure 2C**). To further exam whether or not LL37 impacts A. fumigatus invasion and destruction of epithelial cells, alveolar epithelial cells were incubated with A. fumigatus in the presence of LL37 or sL37 for 16 h, and then the secretion of LDH in the supernatant was measured to identify the cell death rate. The results indicated that LL37 attenuated A. fumigatus invasion and cytotoxicity on epithelial cells (**Figure 2D**). Overall, these findings suggest that LL37 inhibited A. fumigatus hyphae growth and adhesion in vitro.

#### LL37 Inhibits *A. fumigatus*-Induced Pro-Inflammatory Cytokine Production in Macrophages

To assess the influence of LL37 on inflammatory activity in A. fumigatus-infected macrophage, BMDMs were stimulated with ulraviolet-killed or heat-killed A. fumigatus hyphae or swollen conidia in combination with different concentrations of LL37 or sL37 for 16 h, after which point, TNF-α and IL-6 in the supernatant were detected. In comparison with sL37 treatment, LL37 significantly reduced the A. fumigatus-induced TNF-α and IL-6 production in macrophages and it performed as a dosedependent manner in ulreaviolet-killed groups (**Figures 3A–F**).

Meanwhile, we examined the transcriptional levels of macrophage polarization-related markers. Consistent with the reduced production of proinflammatory cytokines, LL37 also downregulated the mRNA expressions of iNOS, CXCL9 and CXCL10 (**Figures 3G–I**), which related to M1-type macrophages. However, the levels of M2-type macrophage—related markers, such as arginase and Fizz, were very low in all of groups following A. fumigatus stimulation (data not shown).

In addition, we wondered whether LL37 affected macrophage phagocytosis and killing capacity on A. fumigatus as it decreased inflammatory cytokines production, so we incubated A. fumigatus conidia with macrophages in the presence of LL37 or sL37. However, both LL37 and sL37 did not attenuate the phagocytosis or killing capacity of macrophages against A. fumigatus (**Figures 3J,K**). Together, those data suggest that LL37 inhibits A. fumigatus-induced proinflammatory cytokine production from macrophages and downregulates M1 type markers without affecting macrophage phagocytosis or killing capacity.

were incubated on polystyrene plates with different concentration of LL37 or sL37(0, 0.5, 1, 1.5, 2, 4µM) at 37◦C for 24 h. The wells were washed with PBS and stained with crystal violet. Then the wells were photographed under a microscope and the absorbance of decolorized solution at 570 nm after decoloring by ethanol was determined. (D) Alveolar epithelial cells (1 × 10<sup>5</sup> ) were incubated with *A. fumigatus* (1 × 10<sup>6</sup> ) in the presence or absence of 4µM LL37 or sL37 at 37◦C for 16 h. Cell viability was analyzed by the LDH released in the supernatant. The bars represent the mean values and the standard deviation. The data are representative of three independent experiments. \**P* < 0.05; \*\**P* < 0.01; \*\*\**P* < 0.001. AF, *Aspergillus fumigatus*. sL37, scrambled-LL37.

## LL37 Promotes Fungi Clearance and Alleviates Lung Pathological Injury in an *A. fumigatus*-Infected Mouse Model

The findings described above indicate that LL37 inhibits A. fumigatus in vitro, however, the role of LL37 in the A. fumigatus infection in vivo is not clear. To investigate this, we infected mice via intratracheal administration of 2 × 10<sup>7</sup> A. fumigatus conidia followed by LL37 peptide solution or PBS instillation. As compared with mice in the control group, LL37-treated mice infected with A. fumigatus showed significantly lower fungi burdens in the lungs at both 2 and 7 days (**Figures 4A,B**; **Figure S3** in Supplementary Material). Subsequently, we performed histological analysis of lung tissues from mice at 7 days after A. fumigatus infection. PASM staining was used to visualize the fungal distribution and revealed that the majority of fungi were confined around the trachea and

FIGURE 3 | Effects of LL37 on the activation of *A. fumigatus*-infected macrophages. (A–F) Macrophages were incubated with ultraviolet-killed or heat-killed *A. fumigatus* swell conidia (MOI = 5:1) or hyphae (MOI = 1:1) in the presence of LL37 or sL37(0, 1, 2, 4µM). TNF-α and IL-6 levels were determined at 16 h after stimulation. (G–I) Macrophages were incubated with ultraviolet-killed *A. fumigatus* swell conidia (MOI = 5:1) in the presence or absence of 4µM LL37 or sL37. The relative mRNA expression of iNOS, CXCL9, CXCL10 were determined at 12 h after stimulation. (J) Macrophages were cocultured with FITC-labeled *A. fumigatus* conidia (MOI = 10) in the presence or absence of 4µM LL37 or sL37 for 1 h. Then the macrophages were stained with anti-CD11b PerCP cy5.5 before being analyzed using a BD flow cytometer. The percentages of phagocytosis were calculated by the ratio of FITC<sup>+</sup> macrophages to all macrophages. (K) Macrophages were incubated with viable *A. fumigatus* conidia (MOI = 1:5) in the presence or absence of 4µM LL37 or sL37 for 6 h. The fungi were scraped off and diluted with PBS and then cultured on SDA plates to count colonies. The bars represent the mean values and standard errors of the means. The data are presented as the mean ± SD and representative of three independent experiments. \**P* < 0.05; \*\*\**P* < 0.001. TNF-α, tumor necrosis factor-α. IL-6, interleukin-6. SC, swell conidia. UV, ultraviolet. iNOS, inducible nitric oxide synthase. CXCL, Chemokine (C-X-C motif) ligand. sL37, scrambled-LL37.

bronchi in the lungs of LL37-treated mice, whereas A. fumigatus hyphae had spread into the alveolar tissues of the lungs of the control group (**Figure 4C**). Additionally, hematoxylin and eosin (H&E) staining of lung tissues revealed that a greater number of immune cells accumulated and the more severe phenomenon of congestion and structural damage happened in the mice of the control group vs. LL37-treated mice (**Figure 4D**). These results collectively indicated that LL37 prompted A. fumigatus clearance and controlled the hyphae invasion, which alleviated pathological damage.

# LL37 Inhibits *A. fumigatus*-Induced Pro-inflammatory Cytokines *in vivo*

Since LL37 inhibits A. fumigatus-induced pro-inflammatory cytokines in vitro, we sought to investigate whether the antiinflammatory effects of LL37 were also available in vivo. The cytokine levels in lung homogenate were analyzed at 2 days after A. fumigatus infection. In line with our hypothesis, the release of TNF-α and IL-6 from the lungs of LL37-treated mice was significantly lower than that from the control group mice (**Figures 4E,F**).

# Endogenous LL37 Plays a Protective Role in *A. fumigatus* Infection

Considering the complex interactions among molecules in the immune system, we constructed LL37 transgenic mice to explore whether endogenous LL37 played a protective role in A. fumigatus infection. We infected LL37+/+ and wild-type mice via intratracheal administration of 2 × 10<sup>7</sup> A. fumigatus conidia. Both 2 days and 7 days after infection, the fungi load showed that the LL37+/+ mice had significantly smaller number of A. fumigatus in the lungs than did the wild-type mice (**Figures 5A,B**). At 7 days after infection, histopathological analysis was performed. PMSF staining revealed that wild-type mice had significantly more A. fumigatus invading lesions vs. the LL37+/+ mice (**Figure 5C**). H&E staining suggested that the lungs of wild-type mice were heavily infiltrated by inflammatory cells and the congestion and tissue destruction were more obvious in these mice vs. in the LL37+/+ mice (**Figure 5D**).

## Endogenous LL37 Alleviates Inflammation Caused by *A. fumigatus* Infection

To examine the effects of endogenous LL37 on the inflammatory response mediated by A. fumigatus infection, we analyzed inflammatory cells and inflammatory cytokines in mouse lung. The results showed that the lung tissues of LL37+/+ mice released less TNF-α and IL-6 than did those of wildtype mice after 2 days of infection (**Figures 5E,F**). Using flow cytometry, we additionally observed that LL37+/+ mice performed significantly reduced number and percentage of neutrophils in the lung vs. mice of the control group, who showed plenty of neutrophils (i.e., over 40% of total cells) recruited to the lung (**Figures 6A–C**). In contrast, there was little difference in the number or percentage of macrophages between LL37+/+ mice and control group mice, which both showed an obvious descent of macrophages after A. fumigatus infection (**Figures 6D–F**).

#### LL37 Plays a Protective Role in an *A. fumigatus*-Infected Immunosuppressed Mouse Model

Considering the data above, we still wondered whether the protective effects of LL37 in vivo were directly implemented on the fungus or indirectly enacted by reducing the associated inflammation. Further, we examined the role of LL37 during infections in corticosteroid-treated mice, which imitated the effects of immunosuppression. Under these conditions, LL37 treated mice infected with A. fumigatus for 2 days showed obviously lower fungi burdens in the lungs as compared with mice in the control group (**Figure 7A**). Furthermore, the survival rate curve suggested that LL37-treated mice were significantly less susceptible to A. fumigatus infection vs. control group ones (**Figure 7B**). These results indicated that the direct inhibition of LL37 on A. fumigatus contributed to its protective role in vivo.

# DISCUSSION

LL37, as a host defense peptide, is known to have a broad antibacterial spectrum, including against both Gram-negative and Gram-positive bacteria (10). Meanwhile, the fungicidal activity of LL37 against Candida albicans (19–21) and CAMP from other species against Cryptococcus neoformans have also been reported recently (22). Therefore, our interest was piqued regarding exploring the biological effects of human-derived LL37 on A. fumigatus, which is the most important pathogen of aspergillosis. The current study provided evidence that LL37 bind to A. fumigatus and destroyed its architecture, resulting in inhibiting hyphae growth and adhesion in vitro. Meanwhile, LL37 prevented A. fumigatus-induced aggressive macrophages activation. Further, both exogenous and endogenous LL37 prompted the elimination and prevented the invasion of fungi in vivo, and also reduced pathological damage and inflammation.

As mentioned above, the bactericidal activity of LL37 has been described as the model of peptide-membrane interaction. In contrast to the structure and composition of the bacterial cell membrane, fungi show thicker cell walls and different polysaccharides that contain more zwitterionic phospholipids and sterols (23, 24). Thus, LL37 may exert different mechanism against fungi. Regarding another common fungal pathogen, C. albicans, studies have reported that LL37 performs its fungicidal effect through membrane permeability as well as by affecting fungal structural integrity and altering cell wall composition (20, 21, 25). Our results suggested that LL37 could bind to the surface of A. fumigatus, which was confirmed by fluorescence confocal microscopy. Further, we observed that LL37 disrupted the cell wall structure and caused the cytoplasmic membrane at the apical side to separate from the cell wall to form a wavy structure under the electron microscope. However, the mechanism of LL37 acting on A. fumigatus directly via activating glycanase or forming pores on cell wall/membrane, or indirectly via triggering stress response pathways, remains to be further studied.

Our experiments additionally suggested LL37 inhibited mycelial growth in a dose-dependent manner, but we noted that A. fumigatus could still grow into hyphae that were just a little shorter than the normal at high concentration of LL37 (e.g., 20µM). This is consistent with the findings of previous literature reports that indicated the minimum inhibitory concentrations of five CAMP peptides derived from cattle, sheep and pigs against Candida and Cryptococcus are in the range of 0.5–32µM, while filamentous fungi, such as Aspergillus, are less susceptible to these peptides (22).

Interestingly, while the mycelial growth of A. fumigatus was almost unaffected at low concentrations of LL37 (e.g., 4µM), its adhesion ability in comparison was significantly diminished. Therefore, the primary function of LL37 on A. fumigatus may tend to inhibit invasion rather than act in a fungicidal manner, which is not the same effects as on bacteria. This is in line with our in vivo experiments wherein pathological foci showed that A. fumigatus in the lungs of LL37 transgenic mice was mainly confined around trachea and bronchi but disseminated to the alveolar and pulmonary vessels in wild-type mice. As with A. fumigatus, LL37 significantly inhibited the adhesion of C. albicans, which is related to the elevation of β-1,3-exoglucanase activity (26). However, further research is needed to investigate the mechanism(s) how LL37 inhibits adhesion and invasion of A. fumigatus.

Other studies have demonstrated that LL37 inhibits inflammation by neutralizing LPS and inhibiting TLR4 activation

in bacterial infections (27–30). The present investigation suggested that TNF-α and IL-6 levels from A. fumigatusstimulated macrophages incubating with LL37 also decreased. Given that cell wall components often serve as PAMPs to trigger pro-inflammatory responses in innate cells such as macrophages, there is reason to speculate that impaired cell wall integrity by LL37 accounts, at least partly, for reduced production of proinflammatory cytokines. However, there was no change of TNF-α and IL-6 production in LL37 treated-macrophages in the absence of infection vs. in untreated group, neither in macrophages infected by Aspergillus pre-treated with LL37 or sL37 for an hour vs. by untreated A. fumigatus (data not shown). In order to distinguish the direct fungistatic and immunomodulatory effects of LL37, we treated A. fumigatus by ultraviolet inactivation and heat inactivation treatment, followed by stimulation. The results showed that LL37 could still significantly inhibit the production of TNF-α and IL-6 both in ultraviolet- and heat- killed groups, which indicated LL37 mainly acted by immunomodulation rather than via direct fungistatic effects to reduce inflammation in the situation we set.

To explore how LL37 regulates inflammation in A. fumigatus-infected macrophages, we examined the expression of macrophage polarization-related makers and found that LL37 downregulated M1 type markers, such as iNOS, CXCL9 and CXCL10. It has been reported that LL37 can directly induce macrophages to differentiate into M1 type with

mean values and standard deviation. \**P* < 0.05. WT, wild type.

proinflammatory effects (31). Therefore, we can assume that the immunomodulatory effect of LL37 is closely related with the type of pathogens.

Inflammatory response is a reasonable manifestation of a host's defense against invasive pathogens; however, excessive inflammation can cause damage to the host tissues and even adversely affect the pathogen clearance (6, 32, 33). As seen in our animal experiments, the percentage of neutrophils in the lungs of wild-type mice infected with A. fumigatus was as high as 40%, which resulted in severe edema, congestion and destruction of lung tissue. In contrast, the lungs of LL37+/+ mice experienced a relatively lesser degree of inflammatory cells infiltration, more effective pathogen clearance, and reduced lung damage. Although A. fumigatus is clinically susceptible to immunosuppressed patients, it also occurs in patients with high levels of inflammation such as chronic granulomatous disease or cystic fibrosis, who show host damage and impaired fungi eradication (34, 35). Therefore, for the treatment of aspergillosis,

survival), followed by supplement of 35 µL LL37 (0.1 mg/mL) or PBS. (A) The fungal burden in lung tissues was determined after 2 days of *A. fumigatus* infection. (B) Survival was monitored for 2 weeks. \*\**P* < 0.05. \**P* < 0.05.

in addition to considering the killing of fungi, the promotion of protective immune response regulation is also crucial.

Furthermore, our data demonstrated that LL37 still reduced the degree of susceptibility to A. fumigatus in immunosuppressive mouse models and suggested that the direct inhibition of LL37 on A. fumigatus contributed to its protective role in vivo. We suppose that LL37 significantly inhibits mycelium adhesion and invasion in vivo as shown in the in vitro experiments.

However, in vitro experiments cannot explain the role of LL37 in different microenvironments in vivo (36, 37). Even if exogenously synthesized LL37 was used in mouse experiments, it could not fully reflect the complex interactions that may occur in the innate immune system. In this study, we established LL37 transgenic mice by microinjection of linearized plasmids expressing hCAP18/LL37 into fertilized eggs and demonstrated that endogenous LL37 had a protective effect against A. fumigatus infection. As this transgenic mouse overexpressed LL37 systemically, further experiments are needed to identify the alveolar epithelium or neutrophils where the LL37 derived from that play a major role.

It is worth noting that a recent study reported that LL37 promoted the growth of A. fumigatus (38). However, in addition to the differences between fungal strains we used (Af293 vs. ATCC strain), the authors were mainly concerned about the effects of much lower levels of LL37 (i.e., 5µg/ml, approximately 1µM) than us at mostly higher than 1µM. In growth experiments, their lower concentrations of LL37 (7.8–1.9µg/ml) were more conducive to A. fumigatus growth, while we focused on the inhibition of hyphal length and adhesion by larger doses of LL37. As the authors explained in the discussion, a low level of stress may activate compensatory growth pathways in A. fumigatus. Therefore, it is necessary to carry out further rigorous experiments to determine the LL37 concentration threshold that causes completely opposite effects before LL37 can be used to treat A. fumigatus infection in the future.

In conclusion, LL37 directly inhibits A. fumigatus mycelial growth and its adhesion and invasion ability. Meanwhile, there is beneficial immunomodulatory effect—LL37 attenuates the A. fumigatus-induced excessive inflammatory response. Taken together, it can be surmised that LL37 exerts a dual protective effect in the treatment of A. fumigatus infection and it and its analogs may be potential drug components for use against A. fumigatus infection.

### AUTHOR CONTRIBUTIONS

X-LL designed, performed, and analyzed the experiments and wrote the manuscript. J-XL designed, performed and analyzed the experiments. J-FX and X-MJ designed, analyzed and provided overall guidance for the experiments, wrote and revised the manuscript. H-RH helped with the experiments and interpretation of the data. J-LD, R-XD, R-JT, LY, and JH contributed to the experiments.

# FUNDING

This work was supported by the National Science Foundation of China (NSFC81670006; 81630001; 31622023; 81571611), Shanghai Leading Talent Program (No. 2016036), the Key Project of Fundamental Research Funds for the Central Universities (2017ZD002), and the Project of the Shanghai Hospital Development Center (16CR3036A); Outstanding academic leader program of Shanghai health and Family Planning Commission (2017BR024), Shanghai laboratory animal research fund (16140902600), Shuguang Program of Shanghai Education Development Foundation and Shanghai Municipal Education Commission (17SG24).

#### ACKNOWLEDGMENTS

We thank LetPub (www.letpub.com) for its linguistic assistance during the preparation of this manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu. 2019.00283/full#supplementary-material

#### 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 © 2019 Luo, Li, Huang, Duan, Dai, Tao, Yang, Hou, Jia and Xu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Commentary: Gut microbiome–mediated bile acid metabolism regulates liver cancer via NKT cells

Baolei Jia\*

*School of Bioengineering, Qilu University of Technology (Shandong Academy of Sciences), Jinan, China*

Keywords: gut microbiome, bile acid, liver cancer, NKT cells, immune regulation

#### **A Commentary on**

#### **Gut microbiome–mediated bile acid metabolism regulates liver cancer via NKT cells**

by Ma, C., Han, M., Heinrich, B., Fu, Q., Zhang, Q., Sandhu, M., et al. (2018). Science 360:eaan5931. doi: 10.1126/science.aan5931

A study by Ma et al. (1) showed that gut microbiome composition in mice closely associates with liver cancer by influencing the immune system. This group provided evidence showing that changing commensal gut bacteria in mice affected the accumulation of hepatic CXCR6<sup>+</sup> natural killer T (NKT) cells through mediation of CXCL16 expression in liver sinusoidal endothelial cells. CXCL16 is the only ligand for the chemokine receptor CXCR6, which mediates NKT cell survival and accumulation in the liver (2, 3). The accumulation of CXCR6 in hepatic NKT cells enhances the production of interferon-γ upon antigen stimulation, which contributes to the inhibition of tumor growth. The accumulation of NKT cells is known to be mainly regulated by a type of Clostridium species that metabolizes primary bile acids to secondary bile acids because depletion of Clostridium by vancomycin increases hepatic NKT cells and colonization of C. scindens induces a rapid decrease in liver NKT cells (1). This evidence highlighted the significant contribution of the gut microbiome to regulating anti-tumor immunity in liver and hepatic cancers.

Human microbiota plays a critical role in maintaining metabolic and immune homeostasis and protecting the host against pathogens (4, 5). The gut microbiota provides a prominent benefit to the host; however, there is also increasing evidence of the involvement of the gut microbiota in human disease (6). The liver is closely linked to the gut because of its anatomical connection via the portal vein. The liver is the first system to acquire nutrient-rich blood via a portal vein from the gastrointestinal tract. Accordingly, the liver is also the first target of metabolites from the gut microbiota, including bile acids, choline, short-chain fatty acids, indole derivatives, and lipopolysaccharides (7). Bile acids can be classified into primary bile acids and secondary bile acids, which are synthesized by the liver and by bacterial metabolism in the colon, respectively. Recently, emerging evidence has also indicated direct associations between obesity, gut microbiota, secondary bile acids, and hepatocellular carcinoma (HCC) (8, 9). Dietary obesity induces a clear expansion of gram-positive gut microbiota, especially Clostridium clusters XI and XVIa, in mice with a high-fat diet (8, 9). The elevation in the strains increased the levels of deoxycholic acid (DCA), a secondary bile acid, and lipoteichoic acid (LTA), a major cell wall component in gram-positive bacteria. The accumulation of the two molecules in the livers of HFD mice treated by chemical carcinogen cooperatively enhanced the Toll-like receptor 2 (TLR2)-mediated signals by the upregulation of the receptor, which induced overexpression of cyclooxygenase2 (COX2), catalyzing the production of prostaglandin E<sup>2</sup> (PGE2). Accumulation of PGE<sup>2</sup> suppressed

#### Edited by:

*Sudhanshu Shekhar, University of Oslo, Norway*

Reviewed by: *Sunil Joshi, University of Miami, United States*

> \*Correspondence: *Baolei Jia baoleijia@cau.ac.kr*

#### Specialty section:

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

Received: *04 December 2018* Accepted: *04 February 2019* Published: *20 February 2019*

#### Citation:

*Jia B (2019) Commentary: Gut microbiome–mediated bile acid metabolism regulates liver cancer via NKT cells. Front. Immunol. 10:282. doi: 10.3389/fimmu.2019.00282*

**62**

anti-tumor immunity through a PTGER4 receptor on CD8 cells, thereby contributing to HCC progression (9).

Compared with previous studies, this study contributes to advances in the related field in the following ways. First, the results from Ma et al. (1) indicated that altering the gut microbiome caused the accumulation of both CD8 cells and NKT cells; however, depleting CD8 cells alone had minor effects on the tumor inhibition caused by elimination of commensal gut bacteria, and antibiotic treatment of tumor-bearing mice lacking NKT cells did not reduce liver tumor size. These results suggest that NKT cells are critical for effects on hepatic tumor growth induced by alterations in the gut microbiome. Second, Ma et al. (1) also provided evidence that increasing primary bile acids increased hepatic NKT cells and enhanced tumor inhibition but that increasing secondary bile acids had opposing effects. These analyses ascertained the beneficial effect of primary bile acids functioning as a regulator to enhance tumor inhibition. These findings indicate an axis of bile acids and CXCL16, CXCR6, and NKT cells that regulate live cancer. Third, DCA, a secondary bile acid, has always been speculated to be a promoter of liver cancer (10). This study showed that other secondary bile acids also played important roles. For example, ω-muricholic acid (ω-MCA) but not DCA decreased cxcl16 mRNA expression. Increasing ω-MCA expression by feeding was shown to inhibit the activation of liver sinusoidal endothelial cells. These findings present new knowledge of the function of different secondary bile acids.

The findings from Ma et al. shed light on the prevention and treatment of liver cancer by targeting the gut microbiota in clinical application. The data directly indicated that elimination of gram-positive bacteria by vancomycin from the gut prevents tumorigenesis (1). The data from the study also solidified the evidences of influence of liver health by diets, probiotics, and antibiotics, which affect the composition of the human gut microbiota. This research cautioned that Clostridium colonization in gut promotes tumor growth, on the other hand, the commensal Bifidobacterium can enhance antitumor immunity and regulate the therapy efficacy by blocking programmed cell death 1 ligand 1 (PD-L1) (11). Because there is individual variability in response to diets, endobiotics, and xenobiotics (12), the studies of precision editing of the gut microbiota are needed to prevent live cancer. Furthermore, the results from Ma et al. also raised the questions on the influence of gut microbiota on the monoclonal antibodies therapies by PD-L1 or PD-1 (programmed cell death 1) blockade. Nivolumab, an anti-PD-1 monoclonal antibody, has been approved by the FDA for liver cancer in 2017 (13). Recent studies showed that gut microbiomes modulate the efficacy of immunotherapies against melanoma and epithelial tumors (14, 15). On the basis of this study, further studies should be performed to assess the effect of gut microbiomes on the immunotherapies to cure liver cancer in clinical trials.

This study provided a comprehensive analysis of the relationship among the gut microbiome, the immune system, and liver cancer. However, this research invokes three related questions. First, the mechanism of bile acids regulating Cxcl16 expression is still unclear. CXCL16 is a small cytokine with a C-X-C motif with an O-glycosylated mucin-like stalk, a transmembrane helix and a cytoplasmic domain with a potential tyrosine phosphorylation site. These features allow CXCL16 to be expressed as a soluble chemokine as well as a cell surface-bound molecule (16). Further analysis should be performed to elicit if CXCL16 bind bile acids directly or through other molecules. Second, which secondary bile acids did contribute significantly to liver cancer? It has been reported that DCA induced liver cancer and nodules in rats in 1991 (17). Yoshimoto and Loo further showed that DCA was one of the factors facilitating liver cancer development (8, 9). The current research indicated that ω-MCA should be one of the critical players to promote liver cancer. ωMCA is a transformed from primary bile acid βMCA by three strains in a cooperative way, including one Eubacterium lentum strain and two Fusobacterium sp. strains (18). While DCA is transformed from cholic acid by Clostridium clusters XI and XVIa. Then it is critical to elucidate the contributions of different bacteria and secondary bile acids to promote live cancer. Third, can the findings be applied to humans? Approximately 1% of hepatic lymphocytes are NKT cells in humans; however, the cells constitute up to 40% of hepatic lymphocytes in mice. Promisingly, it was shown that primary bile acid CDCA levels in human samples were correlated with CXCL16 expression, whereas secondary bile acid glycolithocholate (GLCA) levels were inversely correlated. Furthermore, mucosal-associated invariant T (MAIT) cells, which are prevalent in human liver, can also express CXCR6 that can bind CXCL16. This evidence suggests that the current study could be translated into clinical practice. cancer. However, comprehensive analysis of human liver tissue is necessary for clinical application considering the differences between humans and mice.

#### AUTHOR CONTRIBUTIONS

BJ wrote the commentary.

#### REFERENCES


immunotherapy in melanoma patients. Science (2018) 359:97–103. doi: 10.1126/science.aan4236


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

# TLR4 May Be Involved in the Regulation of Colonic Mucosal Microbiota by Vitamin A

Lu Xiao, Baolin Chen, Di Feng, Ting Yang, Tingyu Li and Jie Chen\*

Children's Nutrition Research Center, Children's Hospital of Chongqing Medical University, Ministry of Education Key Laboratory of Child Development and Disorders, China International Science and Technology Cooperation Base of Child Development and Critical Disorders, Chongqing, China

Objectives: To investigate the specific role of Toll-like receptor 4 (TLR4) in the regulation of the intestinal mucosa-associated microbiota by vitamin A (VA).

Methods: Both TLR4−/<sup>−</sup> (knockout, KO) and wild-type (WT) female mice were randomly fed a VA normal (VAN) or VA deficient (VAD) diet for 4 weeks to establish the following four mouse model groups: TLR4−/<sup>−</sup> mice fed a VAN diet (KO VAN), TLR4−/<sup>−</sup> mice fed a VAD diet (KO VAD), WT mice fed a VAN diet (WT VAN), and WT mice fed a VAD diet (WT VAD). Then, the mice from each experimental group were mated with male mice with the same genetic background. The pups in the KO VAD and WT VAD groups were subsequently fed the VAD diet after weaning, while the pups in the KO VAN and WT VAN groups were fed the VAN diet continuously after weaning. The serum retinol levels of 7-week-old offspring were determined using high-performance liquid chromatography, and colons were collected from mice in each group and analyzed via 16S rRNA gene sequencing using an Illumina MiSeq platform to characterize the overall microbiota of the samples.

#### Edited by:

Xi Yang, University of Manitoba, Canada

#### Reviewed by:

Erguang Li, Nanjing University, China Paulina Schmitt, Pontificia Universidad Católica de Valparaíso, Chile

\*Correspondence: Jie Chen jchen010@hospital.cqmu.edu.cn

#### Specialty section:

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

Received: 31 July 2018 Accepted: 01 February 2019 Published: 22 February 2019

#### Citation:

Xiao L, Chen B, Feng D, Yang T, Li T and Chen J (2019) TLR4 May Be Involved in the Regulation of Colonic Mucosal Microbiota by Vitamin A. Front. Microbiol. 10:268. doi: 10.3389/fmicb.2019.00268 Results: The abundance and evenness of the colon mucosa-associated microbiota were unaffected by dietary VA and TLR4 KO. VAD decreased the abundance of Anaerotruncus (Firmicutes), Oscillibacter (Firmicutes), Lachnospiraceae \_NK4A136 \_group (Firmicutes) and Mucispirillum (Deferribacteres) and increased the abundance of Parasutterella (Proteobacteria). TLR4 KO decreased the abundance of Bacteroides (Bacteroidetes) and Alloprevotella (Bacteroidetes). However, the abundance of Allobaculum (Firmicutes), Ruminiclostridium\_9 (Firmicutes), Alistipes (Bacteroidetes), and Rikenellaceae\_RC9 (Bacteroidetes) impacted the interaction between VA and TLR4.

Conclusion: TLR4 may play a pivotal role in regulation of the intestinal mucosaassociated microbiota by VA to maintain the intestinal microecology.

Keywords: toll-like receptor 4 (TLR4), vitamin A normal (VAN), vitamin A deficiency (VAD), intestinal mucosaassociated microbiota, intestinal innate immunity

**Abbreviations:** HPLC, high-performance liquid chromatography; IBD, inflammatory bowel disease; Mean ± SEM, mean ± standard error of mean; OTU, operational taxonomic units; PBS, phosphate-buffered saline; PCR, polymerase chain reaction; RARβ, retinoic acid receptor β; TLR4−/−, Toll-like receptor 4 knockout; TLR4, Toll-like receptor 4; VA, vitamin A; VAD, vitamin A deficiency; VAN, vitamin A normal; WT, wild-type; ZO-2, zonula-occludens 2.

## INTRODUCTION

fmicb-10-00268 February 20, 2019 Time: 18:55 # 2

Vitamin A is an essential fat-soluble vitamin that maintains normal growth and development, participates in the immune response, promotes reproduction, and maintains visual function (Stephensen, 2001; Clagett-Dame and DeLuca, 2002). VAD remains a significant public health concern in many regions of the world (World Health Organization [WHO], 1995). Children experiencing VAD are especially prone to gastrointestinal (GI) tract infections (Thornton et al., 2014). Studies by us and others have shown that VAD can decrease gut integrity and impact the immune response of the GI tract in humans and animals (Quadro et al., 2000; Liu et al., 2014). The diversity and balance of the gut microbiota is important for maintaining the normal biological barrier function of the intestine (Topping and Clifton, 2001; Round et al., 2010). Recent studies have shown that the VA nutritional status can affect the total amount of bacteria in the GI tract and alter the intestinal microflora (Amit-Romach et al., 2009). Our preliminary study found that VAD impacts the structural segregation of the gut microbiota in children with persistent diarrhea (Lv et al., 2016).

Toll-like receptors (TLRs) are membrane-anchored proteins that are expressed on immune cells and enterocytes (Takeda et al., 2003). TLRs act as pathogen recognition receptors (PRRs), identifying microbe-associated molecular patterns (MAMPs) to activate specific signaling pathways (Frosali et al., 2015). A total of 10 TLRs are expressed in humans, and TLR4 is the best characterized PRR. Recognition of MAMPs by TLR4 is involved in protective innate immune response mechanisms against bacterial invasion (Furuta et al., 2006). In addition, TLR4−/<sup>−</sup> mice exhibited a striking reduction in acute inflammatory cells, impaired epithelial cell proliferation and marked bacterial translocation during injury compared with WT mice (Fukata et al., 2005). Moreover, mouse epithelial cells overexpressing the TLR4 signaling pathway exhibited increased bacterial density in the colonic mucosa and increased bacterial translocation (Dheer et al., 2016). In our previous study, we confirmed that RARβ enhanced ZO-2 expression by regulating TLR4 to improve intestinal epithelial barrier function both in vivo and in vitro (Li et al., 2017). However, the role of TLR4 in regulation of the gut microbiota by VA is unclear.

Therefore, the purpose of this study was to determine the effect of TLR4 on the intestinal mucosal microbiota associated with VA nutritional levels. In the present study, TLR4−/<sup>−</sup> and WT mice were acquired to establish both VAN and VAD mouse models. 16S rRNA deep sequencing was used to examine the distribution and structural characteristics of the intestinal mucosa-associated microbiota.

#### MATERIALS AND METHODS

#### Animals, Diets and Sample Collection

This study was approved by the Animal Experimentation Ethics Committee of Chongqing Medical University (Chongqing, China) and was conducted in accordance with the guidelines of the Animal Care Committee of Chongqing Medical University. TLR4−/<sup>−</sup> (knockout, KO) and WT mice obtained from Jackson laboratories (Maine, United States) were purchased from the Model Animal Research Center of Nanjing University (MARC). The TLR4−/<sup>−</sup> mouse strain was C57BL/10ScNJNju, which is based on the C57BL/10JNju mouse strain (WT). The mice were housed in the same room with a constant airflow system, controlled temperature (22–24◦C), and a 12-h light/dark cycle. The VAN and VAD animal models were constructed according to methods described previously (Liu et al., 2014). Half of the female KO and WT mice (3 weeks of age), which were randomly selected, were fed a VAD-inducing diet comprising 400 IU/kg VA for 4 weeks to establish a TLR4−/<sup>−</sup> mouse model with VAD (KO VAD) and a WT mouse model with VAD (WT VAD), and the other half received a VAN diet containing 6,500 IU/kg VA for 4 weeks to establish a VAN TLR4−/<sup>−</sup> mouse model (KO VAN) and a VAN WT mouse model (WT VAN). Then, the (female) mice from each experimental group were mated with the corresponding male mice with the same genetic background. Pregnant mice were fed either the VAD or VAN diet during both gestation and lactation to maintain stable serum retinol levels. Once the pups had weaned, their mothers were sacrificed, and blood was collected from the eyeballs. The serum retinol levels of the maternal VAN mice increased to 1.05 µmol/L, and those of the maternal VAD mice decreased to 0.7 µmol/L. The offspring were used for subsequent experiments. The pups in the KO VAD and WT VAD groups were subsequently fed the VAD diet continuously for 4 weeks, while the pups in the KO VAN and WT VAN groups were fed the VAN diet continuously for the same time period. Next, the mice were sacrificed, and blood was immediately harvested from the eyeball. The colons were extracted from the mice in each group, and after cleaning with 0.01 M PBS, the colons were stored at −80◦C until further study.

## Serum Retinol Detection

The serum retinol levels in the collected mouse blood were determined using HPLC. VA standard curve preparation and testing methods were modified slightly following methods described previously (Li et al., 2017), and VA standard compound was purchased from Sigma (R7632, United States). Briefly, 200 µL of serum was deproteinized with the same volume of anhydrous ethanol. Then, 1000 µL of hexane was used to extract the retinol from the serum, and the hexane was evaporated using nitrogen gas. The retinol residue was dissolved in 100 µL of the mobile phase mixture (methanol:water = 97:3). Finally, the prepared sample was measured using an HPLC apparatus (DGU-20As, Shimadzu Corporation, Japan). The retinoids were separated by chromatography on an analytical column (Hypersil phenyl 120 A 5 mm, 250 mm × 4.6 mm, Phenomenex, United States) via gradient elution of the mobile phase in a liquid chromatograph equipped with a 315-nm ultraviolet photodiode array detector.

#### DNA Extraction and PCR Amplification

Microbial DNA was extracted from colon samples using an OMEGA DNA Kit (Omega Bio-Tek, United States) according to the manufacturer's protocol. The final concentration of the

purified DNA was determined with a NanoDrop 2000 UVvis spectrophotometer (Thermo Fisher Scientific, Wilmington, DE, United States), and DNA quality was checked via 1% agarose gel electrophoresis. The V3-V4 hypervariable regions of the bacterial 16S rRNA gene were amplified with the primers 338F (5<sup>0</sup> -ACTCCTACGGGAGGCAGCAG-3<sup>0</sup> ) and 806R (5<sup>0</sup> - GGACTACHVGGGTWTCTAAT-3<sup>0</sup> ) (Xu et al., 2016) using a PCR thermocycler system (GeneAmp 9700, ABI, United States). The PCRs were conducted using the following program: 3 min of denaturation at 95◦C; 27 cycles of 30 s at 95◦C, 30 s for annealing at 55◦C, and 45 s for elongation at 72◦C; and a final extension at 72◦C for 10 min. The PCRs were performed in triplicate in 20-µL mixtures containing 4 µL of 5 × FastPfu buffer, 2 µL of 2.5 mM dNTPs, 0.8 µL of each primer (5 µM), 0.4 µL of FastPfu polymerase and 10 ng of template DNA. The resulting PCR products were extracted from a 2% agarose gel, further purified using an AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, United States) and quantified using QuantiFluorTM-ST (Promega, United States) according to the manufacturer's protocol.

#### Illumina MiSeq Sequencing

Colon samples were collected from 40 mice from 4 groups: KO VAN group, KO VAD group, WT VAN group, and WT VAD group, with 10 samples per group. After DNA extraction and PCR amplification, the target band size and concentration of the samples to be sequenced were correct. Purified amplicons were pooled in equimolar amounts, and paired-end (2 × 300) sequencing was performed on an Illumina MiSeq platform (TruSeqTM DNA Sample Prep Kit, United States) according to the standard protocols recommended by Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China). The raw reads were deposited into the NCBI Sequence Read Archive (SRA) database (Accession Number: SRP: 158355).

#### Processing of Sequencing Data

Raw fastq files were demultiplexed, quality-filtered with Trimmomatic and merged using FLASH with the following criteria: (i) The reads were truncated at any site that received an average quality score <20 over a 50-bp sliding window. (ii) Primers were exactly matched, allowing 2-nucleotide mismatches, and reads containing ambiguous bases were removed. (iii) Sequences with overlaps longer than 10 bp were merged at the overlap sequence.

Operational taxonomic units were clustered with a 97% similarity cutoff using UPARSE (version 7.1<sup>1</sup> ), and chimeric sequences were identified and removed using UCHIME. The taxonomy of each 16S rRNA gene sequence was analyzed using the RDP Classifier algorithm<sup>2</sup> compared against the Silva (SSU128) 16S rRNA database using a confidence threshold of 70%.

Sets of sequences with 97% identity were clustered into OTUs using USEARCH (version 7.0<sup>3</sup> ). OTUs that reached 97% similarity levels were used for community richness (Chao, ACE), community diversity (Shannon, Simpson), and rarefaction curve analyses. The β-diversity was estimated by computing unweighted UniFrac distances and visualized with principal coordinate analysis. To effectively distinguish between the four groups, a partial least squares discriminant analysis (PLS-DA) was conducted. In addition, linear discriminant analysis (LDA) effect size (LEfSe) was determined using LEfSe software to determine the community or species that most influenced the group division. After features that were significantly different at various bacterial taxonomic levels were identified by LEfSe, the nonparametric factorial Kruskal-Wallis (KW) sum-rank test and LDA were performed to determine whether these features were consistent with the expected behaviors of the different bacterial taxonomic levels; genera with LDA scores greater than three were defined as having a significant impact on the group.

# Statistical Analyses

All data were obtained from ten biological replicates and are presented as the mean ± SEM. Significant differences were calculated via two-way analysis of variance (ANOVA) with a Bonferroni post hoc test using the GraphPad Prism version 5.0 software package. The interaction between the effect of the different VA nutrition levels and the effect of TLR4 deletion was investigated with a Bonferroni post hoc test. When there was a statistically significant interaction, all the experimental groups were compared using a Bonferroni post hoc test. However, when no interaction was observed, the effect of the different VA nutrition levels or TLR4 deletion was assessed using Student's t-test. Only the relevant comparisons of the combined groups are presented in the Results section. Significance was accepted at P < 0.05.

# RESULTS

## The VAD Diet Decreased Serum Retinol Levels in Mice

To explore whether TLR4 participates in regulation of intestinal microbial homeostasis by VA, we established VAN and VAD mouse models in both TLR4−/<sup>−</sup> (KO VAN and KO VAD) and WT mice (WT VAN and WT VAD). As shown in **Figure 1A**, the serum retinol levels of offspring in WT mice fed the VAD diet (0.393 ± 0.027 µmol/L) were significantly lower than those in WT mice fed the VAN diet (1.027 ± 0.067 µmol/L) (P < 0.001). Similar results were observed for the TLR4−/<sup>−</sup> mice, and the serum retinol levels were significantly repressed in the KO VAD group (0.328 ± 0.054 µmol/L) compared with those in the KO VAN group (1.171 ± 0.104 µmol/L). However, there was no significant interaction between the effects of VA and TLR4, determined using two-way ANOVA with a Bonferroni

Microbial Analysis

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

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

<sup>3</sup>http://drive5.com/uparse/

post hoc test, but differences in VA nutritional levels had an effect on the serum retinol levels (P < 0.001, **Figure 1A**). After combining the data for the WT and KO groups, we found that the serum retinol levels in the combined VAD group were markedly lower than those in the combined VAN group (P < 0.001) (**Figure 1B**). The above data demonstrate that the VAD diet was an important factor associated with the decreased serum retinol levels in mice, providing us with a solid foundation for subsequent studies.

### Rarefaction Curves and Alpha Diversity Index

After optimization, a total of 1,487,466 high-quality sequences were obtained from 40 samples, and there were 37,187 highquality sequences per sample on average according to MiSeq sequencing. We acquired numerous OTUs from valid sequences that exhibited 97% similarity for further statistical analyses. Along with an increase in the number of reads, the rarefaction curves for all the samples shown in **Figure 2A** exhibited smooth increasing trends and approached saturation plateaus, demonstrating that the sequencing data volume acquired was suitable for the present study.

In general, Chao and ACE diversity indexes reflect the richness of the microbiota, while Shannon and Simpson diversity indexes are considered to be indicators of colony richness and evenness. After two-way ANOVA with a post hoc test, no significant interaction was found between the effects of VAD and TLR4 KO on the Chao, ACE, Shannon and Simpson diversity indexes (**Figures 2B–E**). Furthermore, these four indexes were not affected by either VAD or TLR4 KO, even though the Simpson index of the KO VAD group was slightly higher than that of the WT VAD group (P < 0.05, **Figure 2E**) and the Shannon index of the KO VAD group was significantly decreased when compared with that of the WT VAD group (P < 0.05, **Figure 2D**). The above data demonstrate that the abundance and evenness of the colonic mucosa-associated microflora were unaffected by dietary VA and TLR4 KO.

#### Different VA Levels in the Diet May Affect the Microbial Community Structure of the Colonic Mucosa in Both the TLR4 KO and WT Mice

To further evaluate structural differences in the microbial communities among the four groups, an unweighted UniFrac distance matrix was calculated based on the OTUs of each sample. **Figure 3A** shows that samples from the WT VAN, WT VAD, and KO VAN groups were relatively concentrated compared with the KO VAD group samples. Principle component analysis (PCA) revealed a separation of the TLR4−/<sup>−</sup> and WT mice fed VAN or VAD diets based on the first two principal component (PC) scores, accounting for 25.5 and 16.77% of explained variances (**Figure 3A**). Meanwhile, ANOSIM analysis showed that the difference among the four groups was significantly greater than the difference within the group (R <sup>2</sup> = 0.3478, P = 0.001, **Figure 3A**), indicating that our grouping was meaningful. The results of the subsequent PLS-DA are shown in **Figure 3B**, and the variance among the KO VAD group samples was greater than that among samples from the other three groups. These data suggest that there may be differences in the distribution of the colonic mucosal microbiota due to the different levels of VA in the diet and due to TLR4 deletion.

# Both VA and TLR4 Are Involved in the Community Abundance of the Colonic Mucosal Microbiota at the Phylum Level

**Figure 4A** shows the composition of the dominant microflora with a relative abundance of more than 1% at the phylum level in the TLR4 KO and WT mice fed the VAN or VAD diet. In the WT VAN group, the predominant phyla were Firmicutes

the WT VAD group; blue curves represent the KO VAN group; yellow curves represent the KO VAD group (n = 10). Comparison of (B) Chao index, (C) ACE index, (D) Shannon index and (E) Simpson index among the four groups (n = 10). Mean ± SEM; ns. = not significant. "Interaction" indicates an effect of the different VA nutritional levels in the TLR4 knockout vs. WT mice; <sup>∗</sup>P < 0.05. VAN, vitamin A normal; VAD, vitamin A deficiency. WT VAN refers to WT mice fed a VAN diet; WT VAD refers to WT mice fed a VAD diet; KO VAN refers to TLR4−/<sup>−</sup> mice fed a VAN diet; KO VAD refers to TLR4−/<sup>−</sup> mice fed a VAD diet.

(34.47%), Bacteroidetes (31.34%), Proteobacteria (30.62%), and Deferribacteres (2.372%); however, the predominant phyla in the WT VAD group were Bacteroidetes (37.33%), Firmicutes (30.91%), Proteobacteria (28.38%), and Deferribacteres (1.718%). The most abundant phyla in the KO VAN group, in decreasing order, were Firmicutes (40%), Proteobacteria (24.66%), Bacteroidetes (23.42%), and Deferribacteres (8.928%). Notably, the percentage of Actinobacteria was highest in the KO VAD group compared with that in the other three groups, and the most abundant phyla, in decreasing order, in the KO VAD group were Proteobacteria (46.58%), Bacteroidetes (22.92%), Firmicutes (22.69%), Actinobacteria (3.644%), and Deferribacteres (1.357%).

FIGURE 3 | PCA and PLS-DA of samples from the TLR4−/<sup>−</sup> and WT mice fed VAN or VAD diets. (A) PCA scores were plotted based on the relative abundance of the OTUs (n = 10). (B) PLS-DA was plotted based on the unweighted UniFrac distance metrics (n = 10). The green circles represent the WT VAN group; red triangles represent the WT VAD group; blue diamonds represent the KO VAN group; orange squares represent the KO VAD group (n = 10). VAN, vitamin A normal; VAD, vitamin A deficiency. WT VAN refers to WT mice fed a VAN diet; WT VAD refers to WT mice fed a VAD diet; KO VAN refers to TLR4−/<sup>−</sup> mice fed a VAN diet; KO VAD refers to TLR4−/<sup>−</sup> mice fed a VAD diet.

FIGURE 4 | The relative abundances of bacterial phyla in the colonic mucosa of the TLR4−/<sup>−</sup> and WT mice fed the VAN or VAD diet. (A) The dominant bacterial phyla with relative abundances greater than 1% in the four groups (n = 10). The combined effects of the different VA nutritional levels and the TLR4 deletion on the relative abundance of (B) Proteobacteria and (C) Firmicutes determined by two-way analysis of variance with a Bonferroni post hoc test (n = 10). (D) The main effect of TLR4, independent of VA nutritional level, on the relative abundance of Bacteroidetes in the four groups (n = 10). (E) The main effect of VA, independent of TLR4 deletion, on the relative abundance of Deferribacteres in the four groups (n = 10). Mean ± SEM; ns. = not significant. "Interaction" indicates an effect of the different VA nutritional levels in the TLR4 knockout vs. WT mice; <sup>∗</sup>P < 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001. VAN, vitamin A normal; VAD, vitamin A deficiency. WT VAN refers to WT mice fed a VAN diet; WT VAD refers to WT mice fed a VAD diet; KO VAN refers to TLR4−/<sup>−</sup> mice fed a VAN diet; KO VAD refers to TLR4−/<sup>−</sup> mice fed a VAD diet.

As shown in **Figure 4B,** although VA and TLR4 had no effect on the abundance of Proteobacteria (P > 0.05), the P-value of the interaction between the VA and TLR4−/<sup>−</sup> challenges was 0.0251 after two-way ANOVA with a Bonferroni post hoc test. The Firmicutes abundance in the VAD group was lower than that in the VAN group in the TLR4−/<sup>−</sup> mice (**Figure 4C**). The VA nutritional level had significant effects on the Firmicutes abundance (P = 0.0024), and the P-value of the interaction between the VA and TLR4−/<sup>−</sup> challenges for the Firmicutes abundance was 0.0217 according to the post hoc test (**Figure 4C**). However, for the abundance of Bacteroidetes and Deferribacteres, no significant interaction was observed between the VA and TLR4−/<sup>−</sup> challenges, determined using two-way ANOVA (**Figures 4D,E**). After combining the VAN and VAD

groups, the Bacteroidetes abundance in the combined KO group was significantly lower than that in the combined WT group (P < 0.05, **Figure 4D**). The Deferribacteres abundance in the combined VAD group was significantly lower than that in the combined VAN group (P < 0.05, **Figure 4E**). Based on these data, VA and TLR4 have interactive effects on the abundance of Proteobacteria and Firmicutes; the Bacteroides abundance was affected by TLR4 and that of Deferribacteres was affected by VA.

## VA and TLR4 Altered the Community Structure of the Colonic Microbiota at the Genus Level

To further understand the effect of VA and TLR4 on colonic bacteria, we analyzed the community structure at the genus level in the four groups. **Figure 5** shows the composition of the dominant microflora with relative abundances greater than 2% at the genus level in the TLR4−/<sup>−</sup> and WT mice fed the VAN or VAD diet. We noticed that the percentage of Helicobacter in the KO VAD group was highest compared with the percentage in the other three groups.

Two-way ANOVA was conducted for each strain, as shown in **Figures 6**, **7**. There were five genera of the phylum Firmicutes that were affected by VA or both VA and TLR4. A significant interaction was observed between the effects of VA and TLR4 on the abundance of Allobaculum and Ruminiclostridium\_9 according to a Bonferroni post hoc analysis, and the P-values were 0.0125 and 0.0345, respectively (**Figures 6A,B**). On the other hand, the main effects of VA, independent of TLR4, were on the relative abundance of Anaerotruncus (P = 0.0003), Lachnospiraceae \_NK4A136 \_group (P = 0.0224), and Oscillibacter (P = 0.0029) in the four groups (**Figures 6C– E**). After combining the WT and KO groups, the abundance of Anaerotruncus (P < 0.001), Lachnospiraceae\_NK4A136\_group (P < 0.05), and Oscillibacter (P < 0.01) was markedly decreased in the combined VAD group compared with that in the combined VAN group (**Figures 6C–E**).

Vitamin A and TLR4 also impacted the community structure at the genus level within the phyla Bacteroidetes, Proteobacteria, and Deferribacteres. According to two-way ANOVA, there was a significant interaction between the effects of VA and TLR4 on the abundance of Alistipes and Rikenellaceae\_RC9 in the phylum Bacteroidetes, and the P-values were 0.001 and 0.0026, respectively (**Figures 7A,B**). However, no significant interaction was observed between the effects of VA and TLR4 on the abundance of Bacteroides (**Figure 7D**) and Alloprevotella (**Figure 7E**). After combining the VAN and VAD groups, the abundance of Bacteroides and Alloprevotella in the combined KO group was markedly lower than that in the combined WT group. TLR4 seems to be the main factor affecting the abundance of Bacteroides and Alloprevotella at the Bacteroidetes level. However, the Parasutterella abundance (Proteobacteria) was significantly increased in the combined VAD group compared with that in the combined VAN group, and the abundance of Mucispirillum (Deferribacteres) was significantly reduced in the combined VAD group compared with that in the combined VAN group.

Based on these data, both VA and TLR4 affected the abundance of Allobaculum (Firmicutes), Ruminiclostridium\_9 (Firmicutes), Alistipes (Bacteroidetes), and Rikenellaceae\_RC9 (Bacteroidetes), while the abundance of Anaerotruncus (Firmicutes), Lachnospiraceae\_NK4A136\_group (Firmicutes), Oscillibacter (Firmicutes), Parasutterella (Proteobacteria), and Mucispirillum (Deferribacteres) was mainly affected by VA, and that of Bacteroides (Bacteroidetes) and Alloprevotella (Bacteroidetes) was primarily affected by TLR4.

## Key Phylotypes in the TLR4−/<sup>−</sup> and WT Mice With Different VA Levels

A metagenomic analysis approach (LEfSe) was used to identify the key phylotypes responsible for the differences among the TLR4 KO and WT mice fed the VAN or VAD diet. **Figure 8** shows a comparison of the bacterial populations in the four groups at the genus level. The results indicated that the key genera in the WT VAN group were Alloprevotella (LDA = 4.34, P = 0.003751), Lachnospiraceae\_NK4A136\_group (LDA = 4.14, P = 0.000845), Clostridium\_innocuum\_group (LDA = 3.45, P = 0.000844), and Blautia (LDA = 3.68, P = 0.000227). In the WT VAD group, the key genera were Aeromicrobium (LDA = 3.81, P = 0.004725), Escherichia\_Shigella (LDA = 3.63, P = 0.000193), Lactobacillus (LDA = 3.29, P = 0.007436), Tyzzerella (LDA = 3.58, P = 0.005026), Rikenellaceae\_RC9\_gut\_group (LDA = 4.26, P = 0.000131), and Allobaculum (LDA = 4.13, P = 0.010607). Oleibacter (LDA = 3.19, P < 0.003641), Pseudomonas (LDA = 3.00, P = 0.000502), Mucispirillum (LDA = 4.54, P = 0.012631), Ruegeria (LDA = 3.51, P = 0.000837), Shewanella (LDA = 3.52, P = 0.023659), and Pseudoalteromonas (LDA = 3.07, P < 0.0001) were the key genera in the KO VAN group. However, only two genera played key roles in the KO VAD group: Acetivibrio\_ethanolgignens\_group (LDA = 3.07, P = 0.004929) and Eubacterium\_coprostanoligenes\_group (LDA = 3.15, P = 0.022025).

# DISCUSSION

The GI tract harbors a complex community of bacteria in the mucosa, lumen and feces. Feces may primarily contain bacteria that are not adherent (Zoetendal et al., 2002), and the luminal microbiota is more variable than the mucosal microbiota (Li et al., 2015). Therefore, in our study, the mucosal bacteria may better reflect the actual intestinal microbiota (Amit-Romach et al., 2009). Some ecologists have noted that the Chao and ACE diversity indexes reflect the richness of the microbiota. The Shannon and Simpson diversity indexes are considered indicators of colony richness and evenness. The Simpson index is sensitive to dominant species, and the Shannon index is sensitive to rare species (Magurran, 1988). In the present study, TLR4 KO reduced the Shannon index and upregulated the Simpson index in the VAD group. TLR4 KO may decrease the abundance of rare species and increase the dominant intestinal mucosaassociated microbiota in VAD rats. However, VA had no effect on the Shannon and Simpson indexes in WT mice; therefore, these two indexes were not found to be affected by either VA

or TLR4 after two-way ANOVA with a post hoc test. Although VA and TLR4 had no effect on the abundance and evenness of the colonic mucosa-associated microflora, the results of PCA and PLS-DA showed that VA and TLR4 affected the distribution and composition of the colonic mucosa-associated bacteria.

Amit-Romach showed that VAD led to a reduction in the proportion of Lactobacillus spp. and resulted in the appearance of pathogenic Escherichia coli strains among the mucosa-associated bacteria (Amit-Romach et al., 2009). Matthew C. found that acute VAD had an effect on bacterial community structure, leading to an increase in the abundance of Bacteroides vulgatus (Hibberd et al., 2017). Studies on changes in gut microbial diversity have shown varying results. However, in the present study, at the genus level, VAD decreased the abundance of Anaerotruncus, Oscillibacter, Lachnospiraceae\_NK4A136\_group, and Mucispirillum and increased the abundance of Parasutterella.

Anaerotruncus is a newly described bacterial genus isolated from human stool (Lawson et al., 2004). Currently, Anaerotruncus colihominis and Anaerotruncus massiliensis are the identified species of this genus (Lau et al., 2006). A. H. Togo isolated Anaerotruncus massiliensis from an obese patient after bariatric surgery (Togo et al., 2016). Although Anaerotruncus is not well known, studies have shown that Anaerotruncus species might be optimal probiotic strains because these species express enzymes that favor the production of butyrate (Polansky et al., 2015). Butyrates are important nutrients for cells lining the mammalian colon. As critical short-chain fatty acid derivatives that regulate colon homeostasis, butyrates participate in colon inflammation (Donohoe et al., 2011). In a study of human gut microflora, Oscillibacter valericigenes was identified in a significantly greater number of samples from healthy controls than from patients with Crohn's disease (Man et al., 2011; Mondot et al., 2011). Anaerotruncus and Oscillibacter appear to play a positive role in maintaining intestinal immune homeostasis. Lachnospiraceae species are also associated with maintenance of gut health, and members of this family may protect against colon cancer in humans by producing butyric acid (Tap et al., 2009; Meehan and Beiko, 2014; Liu et al., 2017). Mice administered retinoic acid (RA) orally and then subjected to partial hepatectomy had higher levels of Lachnospiraceae than mice in the control group, which were not treated with RA (Liu et al., 2016). In the present study, VAD reduced the abundance of Anaerotruncus and Oscillibacter in the colonic mucosa, and the abundance of Lachnospiraceae \_NK4A136 \_ group was also lower in the combined VAD group than in the VAN group, while VAD upregulated the abundance of Parasutterella. Previously, the Parasutterella abundance was found to be increased and the Lachnospiraceae abundance was decreased in the submucosal tissues of patients with Crohn's disease (Chiodini et al., 2015). In addition, the Parasutterella abundance was increased significantly in rats with hypertriglyceridemia-related acute necrotizing pancreatitis (Huang et al., 2017). These results suggest that VA may be involved in regulation of the intestinal mucosa-associated microbiota.

Mucispirillum is a core member of the laboratory mouse microbiota and can colonize the intestinal tract from the stomach to the colon; this genus is represented by a single species, namely, Mucispirillum schaedleri (Robertson et al., 2005). As part of the phylum Deferribacteres, Mucispirillum has been shown to be associated with both inflammatory markers (El Aidy et al., 2014) and active colitis in a T-bet−/<sup>−</sup> Rag2−/<sup>−</sup> mouse model (Berry et al., 2012; Rooks et al., 2014)

mice fed the VAN or VAD diet. The combined effects of the different VA nutritional levels and the TLR4 deletion on the relative abundance of (A) Proteobacteria and (B) Ruminiclostridium\_9 (n = 10). The main effect of VA, independent of TLR4 deletion, on the relative abundance of (C) Anaerotruncus, (D) Lachnospiraceae\_NK4A136\_group and (E) Oscillibacter in the four groups (n = 10). Mean ± SEM; ns. = not significant. "Interaction" indicates an effect of the different VA nutritional levels in the TLR4 knockout vs. WT mice; <sup>∗</sup>P < 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001. VAN, vitamin A normal; VAD, vitamin A deficiency. WT VAN refers to WT mice fed a VAN diet; WT VAD refers to WT mice fed a VAD diet; KO VAN refers to TLR4−/<sup>−</sup> mice fed a VAN diet; KO VAD refers to TLR4−/<sup>−</sup> mice fed a VAD diet.

and was even associated with Citrobacter rodentium infection (Hoffmann et al., 2009). A recent study showed that M. schaedleri possesses specialized systems to handle oxidative stress during inflammation (Loy et al., 2017). Interestingly, our data showed that VAD downregulated the abundance of Mucispirillum. We speculate that this downregulation may be a self-regulatory effect of the intestinal mucosa-associated microbiota.

Activation of TLRs by commensal microflora is critical for protection against gut injury (Rakoff-Nahoum et al., 2004; Fukata et al., 2005; Fukata et al., 2006). A large amount of research has indicated that TLR4 signaling affects the intestinal microbiota (Anitha et al., 2012). However, the effects of TLR4 on the intestinal mucosa-associated microbiota are complex and remain unclear. Our results suggest that TLR4 KO decreased the abundance of Bacteroides and Alloprevotella. Bacteroides, a commensal bacterium that colonizes the lower digestive tract, can strongly affect the host immune system (Swidsinski et al., 2009). A study by Erin B showed that Bacteroides species produce a capsular polysaccharide, polysaccharide A (PSA), to repress proinflammatory cytokines (Troy and Kasper, 2010). Many studies have implicated decreased levels of Bacteroides in the development of IBD (Zhou and Zhi, 2016). In our study, TLR4−/<sup>−</sup> mice exhibited a defective immune response, which may be associated with the marked reduction in Bacteroides abundance. Alloprevotella is a genus of Prevotellaceae, and the clinical significance of this genus remains unclear. Decreased inflammatory cytokine expression in the mouse intestine following interferon tau (IFNT) supplementation led to increased Alloprevotella abundance in the colon (Ren et al., 2016), indicating that Alloprevotella has positive effects on the intestinal mucosa. Intriguingly, the abundance of Alloprevotella and Bacteroides was distinctly reduced in the TLR4 KO group, indicating a complex relationship among TLR4, the microbiota, and intestinal immunity.

Notably, at the genus level, VA and TLR4 had a combined effect on the abundance of Allobaculum (Firmicutes), Ruminiclostridium\_9 (Firmicutes), Alistipes (Bacteroidetes), and Rikenellaceae\_RC9 (Bacteroidetes). Allobaculum has been shown to prevent dextran sulfate sodium (DSS)-induced inflammation (Wang et al., 2015), and the abundance of this genus was seen to be positively correlated with markers of ileal immunity (Cox et al., 2014). However, in the present study, the abundance of Allobaculum in the WT VAD group was significantly higher than that in the WT VAN group, but there was no difference in the abundance of this genus between the VAN and VAD groups of TLR4 KO mice, suggesting that VA might regulate Allobaculum via TLR4. Ruminiclostridium\_9 belongs to the Ruminococcaceae family and is associated with the

release of inflammatory and cytotoxic factors from the gut for maintenance of a stable intestinal microecology (Cheng et al., 2017; Ma et al., 2017; Wu et al., 2017). Although there were no differences among the WT VAN, WT VAD, and KO VAD groups in terms of the abundance of Ruminiclostridium\_9, we found that the Ruminiclostridium\_9 abundance in the KO VAN group was significantly higher than that in the other three groups. These results suggest that TLR4 may be involved in the regulation of intestinal microbiota by VA, but the specific mechanism of regulation and the clinical significance of this regulation need to be further explored. A cross-study analysis showed that Alistipes species are associated with healthy subjects rather than with individuals with gut disease (Mancabelli et al., 2017). In a murine model of DSS-induced colitis, Alistipes finegoldii (Alistipes) was seen to be protective against colitis (Dziarski et al., 2016). However, the Alistipes genus was found to be associated with colorectal cancer, and the abundance of this genus exhibited a negative correlation with the consumption of fruits and vegetables (Feng et al., 2015; Dai et al., 2018). Rikenellaceae\_RC9, similar to Alistipes, belongs to the Rikenellaceae family. The abundance of both these genera was affected by VA and TLR4. However, the specific regulatory mechanisms remain unclear.

A large number of studies have shown that the VA nutritional state affects the abundance and composition of the intestinal microbiota (Amit-Romach et al., 2009; Hibberd et al., 2017). There is also considerable variation and discrepancy associated with identification of bacterial markers of VAD among different studies (Chiodini et al., 2015; Liu et al., 2016) and different disease models, such as models of persistent diarrhea and necrotizing enterocolitis (Lv et al., 2016; Xiao et al., 2018). On the other hand, the intestinal microflora was shown to affect the bioavailability of dietary α- and β-carotene and the activity of VA in rats (Grolier et al., 1998). These results indicate that VA levels and the intestinal microbiota are interrelated. Many studies have shown that VAD impairs GI mucosal barrier integrity by altering bacterial populations, the expression of innate immunity-related genes and the number of immune cells (Amit-Romach et al., 2009; Liu et al., 2014; Li et al., 2017). Our study further demonstrated the independent effects of VA and TLR4 on intestinal mucosa-related bacteria. Our study is the first to indicate that TLR4 is involved in regulation of the colonic mucosal microbiota by VA, providing a foundation for elucidating the relationships among VA, the intestinal microecology and intestinal innate immunity. This result further shows that VA regulates TLR4 to improve intestinal barrier function, as described in our previous study (Li et al., 2017). However, elucidation of the specific regulatory mechanism remains challenging and requires further study and the development of novel approaches.

### CONCLUSION

VAD decreased the abundance of Anaerotruncus (Firmicutes), Oscillibacter (Firmicutes), Lachnospiraceae \_NK4A136 \_group (Firmicutes), and Mucispirillum (Deferribacteres) and increased the abundance of Parasutterella (Proteobacteria). TLR4 KO decreased the abundance of Bacteroides (Bacteroidetes) and Alloprevotella (Bacteroidetes). However, the abundance of Allobaculum (Firmicutes), Ruminiclostridium\_9 (Firmicutes), Alistipes (Bacteroidetes), and Rikenellaceae\_RC9 (Bacteroidetes) impacted the interaction between VA and TLR4. Therefore, TLR4 may play a pivotal role in the regulation of the intestinal mucosa-associated microbiota and maintenance of the intestinal microecology mediated by VA.

### AUTHOR CONTRIBUTIONS

fmicb-10-00268 February 20, 2019 Time: 18:55 # 12

LX performed the experiments and analyzed the data. BC and DF assisted in completion of the experiments. TY provided technical guidance regarding the use of HPLC. LX and JC wrote

#### REFERENCES


the manuscript. TL and JC designed the study. JC provided financial support for the study. All authors read and approved the final manuscript.

## FUNDING

This work was supported by grants from the National Nature Science Foundation of China (81770526 and 81470799) and from the Science & Technology Project of Yuzhong District of Chongqing (2012) to JC.


bacteria and microbiota diversity in finishing pigs. Front. Cell. Infect. Microbiol. 7:526. doi: 10.3389/fcimb.2017.00526


**Conflict of Interest Statement:** Our team bears a patent (Patent No. ZL201010233032.8) on the formula of the vitamin A normal and vitamin A deficiency animal feed in China.

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

# Microbiome Dependent Regulation of Tregs and Th17 Cells in Mucosa

Pushpa Pandiyan<sup>1</sup> \*, Natarajan Bhaskaran<sup>1</sup> , Mangge Zou2,3, Elizabeth Schneider <sup>1</sup> , Sangeetha Jayaraman<sup>1</sup> and Jochen Huehn2,3

*<sup>1</sup> Department of Biological Sciences, School of Dental Medicine, Case Western Reserve University, Cleveland, OH, United States, <sup>2</sup> Experimental Immunology, Helmholtz Centre for Infection Research, Hamburg, Germany, <sup>3</sup> Cluster of Excellence RESIST (EXC 2155), Hannover Medical School, Hannover, Germany*

Mammals co-exist with resident microbial ecosystem that is composed of an incredible number and diversity of bacteria, viruses and fungi. Owing to direct contact between resident microbes and mucosal surfaces, both parties are in continuous and complex interactions resulting in important functional consequences. These interactions govern immune homeostasis, host response to infection, vaccination and cancer, as well as predisposition to metabolic, inflammatory and neurological disorders. Here, we discuss recent studies on direct and indirect effects of resident microbiota on regulatory T cells (Tregs) and Th17 cells at the cellular and molecular level. We review mechanisms by which commensal microbes influence mucosa in the context of bioactive molecules derived from resident bacteria, immune senescence, chronic inflammation and cancer. Lastly, we discuss potential therapeutic applications of microbiota alterations and microbial derivatives, for improving resilience of mucosal immunity and combating immunopathology.

#### Edited by:

*Sudhanshu Shekhar, University of Oslo, Norway*

#### Reviewed by:

*Nobuhiko Kamada, University of Michigan Health System, United States Shimpei Kawamoto, Osaka University, Japan*

#### \*Correspondence:

*Pushpa Pandiyan pxp226@case.edu*

#### Specialty section:

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

Received: *02 December 2018* Accepted: *18 February 2019* Published: *08 March 2019*

#### Citation:

*Pandiyan P, Bhaskaran N, Zou M, Schneider E, Jayaraman S and Huehn J (2019) Microbiome Dependent Regulation of Tregs and Th17 Cells in Mucosa. Front. Immunol. 10:426. doi: 10.3389/fimmu.2019.00426* Keywords: microbiome, mucosa, Treg, mucosal immunity, inflammation, Th17, antibiotics, resident microbes

# INTRODUCTION

Mammals harbor a highly diverse microbiome of at least 1000 species, and an astounding number of 10–100 trillion microbial cells, co-existing in a remarkable balance with the host immune system. Healthy human microbiome is mostly bacteria although other microbial domains such as archaea, viruses, and eukaryotes (principally fungi and protists) are also present (1). While these microbes are distributed in skin, and mucosa of ocular, nasal, oral, eye, and reproductive organs, gastrointestinal (GI) tract mucosa is the major reservoir of resident microbes in terms of abundance and species diversity (2, 3). The human colon harbors approximately 3.8 × 10<sup>13</sup> microorganisms, followed by skin in the range of ∼1011(4). Since the resurgence of microbiome research in recent years, there has been a sharp increase in understanding of how resident microbiome shapes immunity, health and disease of humans. Only a perennial holiday on a lonely island could excuse an immunologist's incognizance on intimate interrelationships between intestinal microbiota and immune balance. Direct crosstalk between resident microbes and host immune cells in mucosa emerges as a pivotal determinant of such an immune balance. Dysbiosis of resident microbes has strong association with a number of immunological disorders, including opportunistic and pathogenic infections (5–13).

Mucosal immune system has not only evolved to protect the mucosal barrier surface against external insults, it has also coevolved with resident microbes in an interdependent harmonious relationship with them (14–21). The resulting immune balance is crucial to drive optimal immune responses without causing an over-exuberant inflammation (22–25). Past few decades have seen that an increase in hyper-hygiene mentality, mindless use of antibiotics and diet changes, have led to reduced diversity and impaired resilience in resident microbiota (26). Consequently, a disruption in aforementioned immune balance leads to rise in autoimmune and inflammatory disorders. Therefore, understanding the mechanisms of these mutualistic relationships between resident microbiota and different components of innate and adaptive immunity is vital to our understanding of immune diseases. Although gut microbiota in laboratory mice and humans differ significantly, murine models have provided a powerful tool to explore host-microbiota-pathogen interactions in mucosa (27, 28). Here we review the effects of resident microbiota on Tregs and Th17 cells, important players in determining immune balance, mucosal barrier integrity and host protective functions in mucosa. These cells mucosa can develop in mucosa independent of commensal microbiota. For example, there is evidence in germ free mice that Treg cells can be induced by dietary antigens from solid food (29). These Treg cells are of limited life span, but are distinguishable from microbiotainduced Treg cells and capable of repressing inadvertent immune responses to ingested protein antigens. Similarly, in oral mucosa, mechanical damage from mastication of food induces barrier protective Th17 cells, independent of oral commensal microbiota under homeostatic conditions (30). However, dysbiosis can lead excessive Th17 cells and lead to periodontal inflammation (31). Thus, while it is known that these cells can develop independent of microbiota, resident bacterial dysbiosis is strongly associated with alterations in these cells, causing mucosal inflammation seen in many diseases including HIV immunopathogenesis (32– 41). Although other cells also play important roles in mucosal tolerance and immunity, we will not review them here.

# TREGS AND TH17 CELLS IN MUCOSA UNDER STEADY STATE-CONDITIONS

Majority of the studies on mucosa-microbiota interactions discuss GI tract. Indeed, GI mucosa harbors by far the largest and most diverse microbiota, as well as abundant and dynamic population of Tregs and Th17 cells. Tregs are defined by the expression of CD25 and Foxp3, and are predominantly known for their immunosuppressive properties. These cells also express other molecules such as Cytotoxic T Lymphocyte Antigen-4 (CTLA-4), PD-1, interleukin 10 (IL-10), transforming growth factor beta 1(TGF-β1), and amphiregulin. Each of the aforementioned proteins has been shown to be either important, or dispensable for different mechanisms of Tregmediated immunosuppression. Divergent conclusions derived from various Treg mechanism investigations have been strikingly similar to those in the popular parable of the "Blind men and an elephant." It is now increasingly clear that suppressive and non-suppressive functions of Foxp3<sup>+</sup> cells are largely variable, depending on local tissues, disease phenotypes, responding effector cells, and cytokine milieu (42–49).

While CD4<sup>+</sup> effector T cell responses contribute to overt intestinal inflammation, Tregs are associated with controlling immunopathology (42, 43, 50). It is well known that Tregs are also pivotal for commensal tolerance (51–53). There have been contentions regarding the Tregs found in colon mucosa (colon Tregs; cTregs); whether they develop in thymus (thymic Tregs; tTregs), or periphery (peripheral Tregs; pTregs). The usage of Nrp-1 and Helios as markers of tTregs, and the extent to which the TCR repertoire of cTreg overlaps with that of tTregs have been debated (54, 55). Nevertheless, it is well established that cTregs require the presence of microbiota for their development, sustenance and function (56–58). There is also evidence that mucosal sites are the primary sites of development and maintenance of pTregs (59– 61). First formal proof for the requirement of microbiota for the induction and maintenance of intestinal Tregs was provided by studies using germ-free (GF) animal models. GF mice show a several-fold reduction in the frequency of Helios<sup>−</sup> <sup>T</sup>regs, when compared with conventionally housed specific pathogen free (SPF) mice. Association of individual bacterial isolates or defined consortia in GF mice is sufficient to induce intestinal Tregs (56, 57). Even antibiotic treated mice, which show depletion in resident microbiota correlating with a drastic reduction in the frequency of Tregs, lend further credence to the positive role of microbiota in sustenance of Tregs (53, 55, 62). In addition to commensal tolerance, mucosal Tregs have been shown to regulate excessive immune responses during infections (43, 63– 65). Recently, they are also shown to accumulate in other tissues and provide functions such as non-suppressive tissue repair functions in muscle (66). While Tregs play diverse and often opposite roles in mucosal infections (**Table 1**), effects of microbiome on Tregs during these infections are largely ignored in many studies.

Th17 cells are RORγt <sup>+</sup>, CCR6+, IL-17A+, IL-17F+, with some cells expressing IL-21 and IL-22, and have been implicated both in mucosal barrier functions. Th17 cells are an important subset of effector T cells that are protective during extracellular bacterial and fungal invasion (83, 88–91). However, excessive Th17 responses are also associated with a variety of pathogenic conditions, depending on the pro-inflammatory cytokines they co-produce (30, 91–95). Littman and colleagues showed for the first time that commensal microbiota play important roles in the development of intestinal Th17 cells (22, 53, 96–100). Th17 development and differentiation is controlled by cytokine and epigenetic regulation (91, 92, 101, 102), but the mechanistic details of microbiome dependent control of Th17 development during mucosal infection is largely unclear.

#### IMPACT OF MICROBIOME ON TREGS AND TH17 CELLS DURING GI INFECTION AND INFLAMMATION

"Healthy" GI microbiota is mainly composed of the phyla Actinobacteria, Bacteroidetes, Firmicutes, Fusobacteria, TABLE 1 | Foxp3+Treg functions in mucosal infections.


Proteobacteria, and Verrucomicrobia. Small intestine is dominated by Enterobacteriaceae and Lactobacillaceae, whereas colon contains the members of Bacteroidaceae, Lachnospiraceae, Prevotellaceae, Rikenellaceae, and Ruminococcaceae respectively (3). A number of factors including diverse environmental conditions, intake of diet and medication, as well as host genetic factors determine the dynamic composition of gut microbiota in individuals (103–107). Gut microbiota are capable of restraining the mucosal colonization by enteric pathogens, a process defined as colonization resistance (108). Thus, administration of antibiotics, and altering the resident microbiota during a mucosal infection is known to lead to post-antibiotic expansion of the pathogens. Loss of overall diversity, or even deficit in single group of bacteria can alter the susceptibility to gastrointestinal infections. For example, Clostridium difficile (C. difficile) infection, the most common cause of nosocomial diarrhea is often preceded by antibiotic usage. Colonization of C. difficile in healthy mice in fact requires a pre-exposure to a cocktail of antibiotics to alter the microbiota composition (109). However, mono-colonization of GF mice with a murine isolate from the family Lachnospiraceae could limit the colonization of C. difficile, suggesting that individual bacterial species are sufficient to confer colonization resistance to C. difficile (110). Enhanced susceptibility toward other infections after antibioticmediated disruption of the intestinal microbiota composition has also been reported for vancomycin-resistant Enterococcus Spp and Salmonella enterica serovar typhimurium (S. typhimurium) (108, 111). Mechanistically, mucosal carbohydrates such as fucose and sialic acid liberated by resident microbiota have been shown to control the growth of enteric pathogens. Antibiotics cause spikes in sugars that can worsen S. typhimurium and C. difficile infections (112). Microbiota alterations reduce the numbers of germinal centers in IL21-receptor knockout mice, resulting in diminished IgA<sup>+</sup> B cells and reduced activationinduced cytidine deaminase in Peyer's patches. These events lead to the expansion of Tregs and Th17 cells, and higher bacterial burdens, but dampening of Citrobacter rodentium-induced immunopathology (113). Resident microbiota at mucosal interfaces can govern transmission and progress of parasitic protozoan infections such as Toxoplasmosis and Amoebiasis (114). In the case of Toxoplama gondii infection in mice, reduction of microbiota in the gut by prolonged antibiotic treatment leads to impaired Toll like receptor (TLR)-11 and Myeloid differentiation response 88 (MyD88) signaling and subsequent deficit in Th1 immunity, substantiating that gut commensals serve as natural molecular adjuvants during T. gondii infection (115). In a mouse model of Giardia duodenalis infection, antibiotic induced alteration of the microbiome prevents CD8 T cell activation by G. duodenalis. Conversely, GI infection can also modulate microbiota specific adaptive immunity (116). For example, a pathogenic GI infection, in parallel to specific immune reactions against the pathogen, induces immune responses to commensals and generates long-lived commensal-specific T cells. Thus an adaptive response against commensals is an integral component of mucosal immunity. However, such a commensal specificadaptive response in a dysbiosis setting can also contribute to excessive inadvertent inflammation. In the context of HIV-1 infection, damages in GI tract and gut microbial translocation (Proteobacterial species) are associated with reduction of systemic and gut/rectal mucosal Th17 cells and Tregs (despite increased Treg/Th17 ratio) (36, 71, 72, 117, 118). A large body of evidence suggests that increased Tregs in circulation correlate to reduced immune activation in HIV+ patients, underlining the anti-inflammatory protective roles of Tregs in patients (71– 73, 118–125). While combined anti-retroviral (cART) therapy in HIV<sup>+</sup> patients generally ensures immune reconstitution in the peripheral blood, dysbiosis and Treg/Th17 abnormalities persist in gut and other mucosae (41, 126–132). This can present residual inflammation and heightened morbidities in cART treated HIV<sup>+</sup> patients. However, in cART-treated HIV<sup>+</sup> patients with elevated levels of immune activation, it is not clear whether altered levels and function of mucosal Tregs/Th17 cells are associated with local microbial dysbiosis (131), and if these alterations contribute to residual inflammation in HIV disease. Collectively, these findings highlight the role of microbiota in restraining pathogens and inflammation by having significant impact on Tregs and Th17 cells.

Alterations in resident microbiota and host immune cells, caused by host genetic makeup also play a role in the pathogenesis of inflammatory bowel diseases (IBD). One of the adaptive arms of immunity that is impacted by such changes is Tregs (133). Bacteroides fragilis for example, has been found to invade mucosa and cause excessive activation of the host intestinal immune response in genetically susceptible patients (134), while under steady-state conditions the same bacterium can enhance Treg differentiation and ensure intestinal homeostasis. Loss of autophagy protein ATG16L1 in Tregs results in aberrant type 2 responses and spontaneous intestinal inflammation (135). It is unclear whether microbiota directly induce the expression of ATG16L1 in Tregs, but it is evident that ATG16L1 and autophagic process directly promote Treg survival and metabolic adaptation in the intestine. Similarly, other genetic risk variants associated with IBD such as: NOD2, CARD9, ATG16L1, IRGM and FUT2 significantly influence the gut microbiota changes (136). For example, a decrease in Roseburia spp (known acetate to butyrate converters), Clostridiaceae family, the genera Bifidobacterium, Ruminococcus and Faecalibacterium has been observed in patients with IBD. Although many of these communities are strongly implicated in Treg maintenance, direct mechanisms of Treg regulation in the context of these genetic variants and IBD are unclear. Combined deficiency of MyD88 and JH gene, which disrupts innate interactions of immune cells with intestinal microbiota and IgA responses respectively, causes overt inflammation, highlighting the requirement of Treg-IgA mediated mechanism in tolerance (51, 137). It has also been shown that microbiota-specific Foxp3<sup>+</sup> <sup>T</sup>reg cells can convert to interferon-γ-producing Foxp3<sup>+</sup> T cells that have a potential to establish mucosal tolerance (138). Disruption of TLR/MyD88 signaling in Foxp3-deficient mice protect them from excessive inflammation at the environmental interfaces of skin, lungs, and intestine, showing that Tregs normally also restrain commensal dependent tonic MyD88-dependent pro-inflammatory signals (139). Mice lacking CLEC7A gene (Dectin-1), thus having dysregulated interactions with fungal microbiome (mycobiome) show an increased susceptibility to dextran sulfate sodium (DSS) induced colitis (140). The role of Th17 cells and Tregs in this model is unknown. Certain proportion of intestinal Tregs coexpresses RORγt, the master transcription factor of the Th17 lineage, with up to 35 % in small intestine and 65 % in colon (141– 143). Some of these RORγt <sup>+</sup> <sup>T</sup>reg co-produce IL-17A (Treg17), and are substantially diminished in GF or antibiotics-treated mice. Mono-association of GF mice with a panel of 22 bacterial species from the human gastrointestinal tract shows that a number of microbes, not only Clostridiales, are capable of induce colonic RORγt <sup>+</sup> <sup>T</sup>regs (142). Segmented filamentous bacteria (SFB) were only mediocre inducers of RORγt <sup>+</sup> <sup>T</sup>regs in that study (142). These studies demonstrate that intestinal RORγt + Tregs are highly microbiota-dependent and have functions in promoting host immunity (62). Yet, RORγt is not a perfect marker for pTregs, because recent reports show the existence of RORγt <sup>+</sup> tTregs, particularly developing under inflammatory conditions (143–145).

While most studies have focused on in-depth characterization of mechanisms by which microbiota engage to counter-regulate their immunostimulatory properties, the reciprocal effect of Tregs on the composition and function of the intestinal microbiota was largely ignored (53, 56, 99, 146, 147). Very recently, analysis of mice harboring a reduced number of TGF-βdependent pTregs demonstrated numerous underrepresented metabolic processes and a limited overall diversity of the microbiome, including a significant reduction of Lactobacillus johnsonii and Mucispirillum schaedleri (148). Mechanistically, it was confirmed that the impaired pTreg generation could adversely affect the microbiota niche by elevating type 2 immune responses in the host, thereby declining the microbiota abundance during the process of community assembly. In conclusion, the presence of pTregs in the intestinal immune system has a strong impact on the composition and function of the intestinal microbiota. Similarly, IL-17F deficiency induces Treg cells in the colon and modifies the composition of the intestinal microbiota and mediates protection against colitis (149). Taken together, two-way interactions between resident microbiota and host intestinal immunity confer intestinal tolerance and immunomodulation.

#### IMPACT OF MICROBIOTA ON TREGS AND TH17 CELLS IN ORAL MUCOSA

Oral microbiome is vital to maintaining both oral and systemic immune homeostasis because oral mucosa is the primary gateway for the GI tract, the biggest component of the immune system (150). While a vast majority of microbiota studies has focused on intestinal mucosae and their interactions with gut microbiota, little is known about oral mucosal microenvironment colonized with a large array of resident microbes, which is structurally and functionally distinct from the GI tract (151– 160). Actinobacteria, Bacteroidetes, Firmicutes, Fusobacteria, and Proteobacteria are the major phyla accounting for ∼96– 99% of the oral microbiome, while SR1, TM7, Cyanobacteria, Spirochaetes, Synergistetes, and Tenericutes, are also found (<1% distribution). It is well established that oral-resident microbiota in poly-microbial interactions and soft-tissue biofilms avert oral diseases, but direct effect of such interactions on host oral immune cells is less clear (161–166). Oral mucosa maintains subsets of dendritic cells (DC), which produce immunomodulatory cytokines such as IL-10, TGF-β1 and Prostaglandin E2, and are predominantly tolerogenic (89, 167– 169). These cells may be in intimate cross-talk with oral mucosal Tregs (58, 62, 170, 171), albeit details of such interactions between these cells are unexplored in oral mucosa. However, alterations in Tregs and Th17 functions have been implicated in human oral Candida infections and periodontitis (36, 38, 40, 69, 172– 176). We and others have shown the presence of oral mucosal Foxp3<sup>+</sup> <sup>T</sup>regs with protective functions during local infection (89, 158, 169, 170). The interrelationship between these cells and oral commensals during an oral infection was also explored (58, 170). In the context of oropharyngeal candidiasis (OPC) infection, Treg cells play a critical role in reducing fungal burden and establishing homeostasis during post anti-fungal response (177). Tregs play rather an unconventional role of enhancing the Th17 cell response and neutrophil infiltration during early acute response, but are associated with reduced TNF-α expression in CD4 T cells at resolution phase (83, 91, 178). Candida infection in mice by itself increases the proportion of Foxp3+Tregs, in a TLR2/MyD88 dependent manner in oral mucosal tissues and draining cervical lymph nodes (58, 83, 91). A small proportion of those Foxp3<sup>+</sup> cells co-express RORγt and IL-17A (Treg17). Antibiotic mediated depletion of resident bacteria significantly diminishes the frequency of Foxp3+Treg IL-17A<sup>−</sup> and Treg17 cells, as well as conventional Th17 cells not expressing Foxp3. Reduction of these cells is concomitant with an increase in tissue pathology and fungal burden in oral mucosa, demonstrating that resident bacteria are important for controlling Foxp3<sup>+</sup> cells and Th17 cells, as well as mucosal immunity (**Figure 1**). Interestingly, Candida can also promote Th17 and Treg responses in oral mucosa (83, 179, 180). The impact of oral resident microbiome in periodontal inflammation, which is now considered a "resident microbial perturbation" rather than a disease caused by a single pathogen, is well known (181). Resident bacterium P. gingivalis, the keystone pathogen contributes to altering the abundance and composition of other normal microbiota. Shift and accumulation of gram-positive aerobes to gram-negative anaerobes such as P. gingivalis, T. denticola, F. nucleatum, and Prevotella sp. are strongly associated with damage in gingival barrier, loss of immune balance and destruction of oral tissue in periodontal disease (150). During this process, bacterial antigens from skewed microbiota can access connective tissues causing abnormal activation and expansion of inflammatory CD4+CD69+CD103<sup>−</sup> memory T cells and Th17 cells (182). Another recent study showed that periodontitisassociated expansion of Th17 cells required both IL-6 and IL-23, and was dependent on the local dysbiotic microbiome (31). Shift in resident microbiota can also include increase in C. albicans, a part of resident mycobiome in ∼50–70% of healthy humans, which can rapidly transition to a pathogen and cause infections in immune-compromised and cancer patients. C. albicans is also shown to heighten P. gingivalis accumulation, worsening the series of inflammatory events associated with periodontitis severity (183, 184). It is known that Treg17 cells exist in periodontitis lesions and could be involved in inflammatory responses against periodontopathic bacteria (185). While there may be only small changes in oral microbiome in HIV+ individuals, underlying mechanisms causing dysbiosis and its association with HIV associated periodontitis during SIV/HIV infection are unclear (117, 186, 187). Precise events defining Th17 and Treg dysfunctions in the context of underlying dysbiosis and aggravating oral inflammation in HIV disease and periodontitis remain to be seen.

#### MICROBIOME IN MUCOSAL IMMUNITY AND INFLAMMATION IN OTHER MUCOSAE

Lung, previously thought to be sterile, is now known to harbor a complex and dynamic microbial community of ∼500 species, with a high resemblance to oral microbiome (188, 189). Lung microbiome strongly influences the development and progression of allergic responses and asthma (190). Disrupting the normal microbiome with childhood antibiotic exposure increases the risk of childhood asthma. Proteobacteria abundance in lower airway secretions correlates with proinflammatory Th17 cell proportions in asthmatic individuals (191, 192). Similarly, in cystic fibrosis patients, alterations of some groups in the polymicrobial community significantly affect the disease progression. Also, in chronic obstructive pulmonary disease (COPD) patients, microbial dysbiosis associated with mucus hyper-secretion and reduced airway clearance results in chronic aberrant inflammation and airway damage (193). Lung microbiota alterations are also associated with differences in pneumococcal clearance (194).

Multiple genera of microbiota exist in vaginal mucosa, often dominated by species of Lactobacillus, and a diverse array of anaerobic microorganisms, including Atopobium,

Anaerococcus, Corynebacterium, Eggerthella, Gardnerella, Mobiluncus, Peptoniphilus, Prevotella, Sneathia, and Finegoldia genera (195). Lactobacilli largely impact the susceptibility to T. vaginalis infection in women. Although mechanisms are still under investigation, there is precedence that Th17 cells and Tregs can have protective and anti-inflammatory effects during T. vaginalis infection (196). During a vaginal herpes simplex virus-2 (HSV-2) infection, mice lacking Tregs fail to timely accumulate HSV-2-specific CD4 T cells and control the infection. This finding underscores the protective role of Tregs in facilitating productive mucosal immunity in vaginal mucosa (197, 198). However, mechanisms of direct control of vaginal microbiome on Tregs and Th17 cells and infection responses remain to be seen. In ocular mucosa, Corynebacterium mastiditis induces commensal specific IL-17 response γδ T cells, recruiting neutrophils and protecting the ocular mucosa from pathogenic infections (199). In nasal mucosa, on the one hand there is evidence that butyric acid-producing microorganisms associate with an impaired olfactory function (200–202). On the other, nasal microbiome is structured by IL-17 Signaling that that supports resistance to S. pneumoniae colonization in the nasal mucosa of mice (203). Collectively, while mcrobial dysbiosis and Tregs/Th17 changes are associated with many of these infections, detailed mechanisms remain to be investigated.

# MOLECULAR MECHANISMS OF MICROBIOTA-ASSOCIATED ALTERATIONS OF TREG/TH17 CELLS IN MUCOSAE

Resident microbes have a variety of mechanisms for conferring mucosal colonization resistance (17, 204–207). They include: (1) directly competing for shared metabolites, (2) expression of inhibitory bacteriocins, (3) induction of protective mucus layer, and (4) priming of protective immune responses (208, 209). Some of the examples include commensal dependent metabolism of secondary bile acids to deoxycholate, production of organic acids, induction of antimicrobial peptides in Paneth cells, and promoting elevated antibacterial T cell responses preventing colonization and dissemination of pathogens (210– 213). Although resident bacteria are known to modulate energy metabolism producing pyruvic acid, citric acid, fumaric acid and malic acid (214), how pH changes determine the mucosal immunity and T cells warrants further investigation. Resident microbiota employ multiple mechanisms that contribute to coordination of Treg/Th17 axis and safeguarding of mucosa (**Figure 2**). For example, microbiota dependent TLR signaling in host is one of the important mechanisms by which microbiota control inflammation and tolerance. TLR2/MyD88signaling is required for generation and expansion of Nrp1low Foxp3+ cells and Treg17 cells in oral and gut mucosa (58). In gut mucosa the capsular polysaccharide A of the Bacteroides fragilis stimulates production of IL-10 by Foxp3<sup>+</sup> cells in a TLR2 dependent manner, thus facilitating mucosal tolerance (215). Recently it was found that this commensal also delivers immunomodulatory molecules to immune cells via secretion of outer membrane vesicles through a non-canonical autophagy pathway for inducing IL-10 expressing Foxp3<sup>+</sup> cells. This mechanism requires the expression of host genes ATG16L1 and NOD2, whose polymorphisms are known to be associated with IBD (216). Selective deletion of Atg16l1 in T cells in mice also results in loss of Foxp3<sup>+</sup> <sup>T</sup>reg cells and spontaneous intestinal inflammation characterized by aberrant Th2 responses. These data indicate microbiota-host interactions intimately involve the processes of autophagy and Treg differentiation. Moreover, loss of MyD88-STAT3 signaling in Tregs causes loss of mucosal Tregs and impaired T follicular regulatory cell interactions, resulting in poor IL-21 and anti-microbial IgA responses (217). Failure of this pathway results in overgrowth of pathobionts, overt Th17 cell expansion and intestinal inflammation. However, the requirement of resident microbiome induced MyD88 signaling specifically in Tregs, to promote Treg sustenance and intestinal tolerance is still debated (217–219). Similar to B.fragilis, colonic Clostridium rhamnosus also potently induces IL-10+Tregs in a TGF-β1 dependent manner, which is correlated to increase in systemic IgE and resistance to colonic inflammation (56, 99). Similarly, microbiota and immune cell networks are known to control the production of IgA, which is central for mucosal barrier and intestinal tolerance. For example, Mucispirillum spp. and SFB have been directly implicated in production of intestinal IgA (137, 220, 221). Tregs are also known to promote IgA secretion, and maintenance of diversified and balanced microbiota, which in turn facilitates their expansion through a symbiotic regulatory loop, and prevent overt inflammation (222, 223). Moreover, RORγt <sup>+</sup> Th17 cells, as well as IL-17A from other cells also promote epithelial polymeric Ig receptor and intestinal IgA expression, further contributing to intestinal homeostasis (224, 225). SFB also control commensal tolerance and anti-microbial host responses through intestinal epithelial cell fucosyl tranferase 2 expression and fucosylation, a process that is dependent on RORγt + group 3 innate lymphoid cells (ILC3s) and IL-22 expression (226, 227). Loss of intestinal fucosylation results in increased susceptibility to infection by Salmonella typhimurium. ILC3s can also express major histocompatibility complex class II (MHCII) and mediate intestinal selection of CD4<sup>+</sup> T cells in order to limit commensal bacteria-specific CD4 T-cell responses (228). Although IL-6, induction of Tregs, or Th17 cells were shown to be not required for ILC-mediated tolerance, alterations in Treg17 and Th17 cells in the context of fucosylation remain to be studied. Treg/Th17 cell differentiation and expansion are also independently controlled by specific members of anaerobic bacteria producing short chain fatty acids (SCFAs), such as acetate, propionate and butyrate (229, 230). Some of these bacteria include Bacteroides, Bifidobacterium, Feacalibacterium genera, and Enterobacteriaceae family, Porphyromonas gingivalis, Fusobacterium nucleatum (mouth), Clostridium cochlearium, Eubacterium multiforme (intestine), and Anaerococcus tetradius (vagina). These bacteria ferment indigestible oligosaccharides and cell surface fucosylated proteins by anaerobic glycolysis, resulting in SCFA production. SCFAs are present in the intestinal lumen at a total concentration of ∼100 mM at a ratio of ∼6:3:1, for acetate, propionate and butyrate respectively. Although this ratio hinges on carbohydrate availability, microbiota composition and intestinal transit time, acetate and butyrate appear to be the highest and least in abundance respectively (231). Emerging data show that SCFAs contribute to immune homeostasis in mucosa, although excessive and suboptimal levels of SCFAs are often associated with inflammation and cancer. Intestinal SCFAs have been shown to potentiate Foxp3<sup>+</sup> cell differentiation and immunomodulatory activity in the colon (53, 99, 147, 232). Mechanistically, in addition to direct histone deacetylase (HDAC) inhibition, SCFAs can induce the expression of retinal aldehyde dehydrogenase 1 family member 1a (Aldh1a) and TGF-β1 in intestinal epithelial cells and DCs (100, 221, 233, 234). Aldh1a could further convert vitamin A into its metabolite retinoic acid in G protein–coupled receptor43 (GPCR43) and Gpr109a manner, which is capable of facilitating Treg induction. These tolerogenic DCs express CD103, sample antigens in the intestinal lamina propria, and migrate to the draining mesenteric lymph node (MLN) to induce immunomodulatory T cells (235–237). Whether SCFA mediated induction and or sustenance of mucosal Tregs require these aforementioned processes is unclear and remain to be studied. However, antibiotics precipitously decrease the oral SCFAs in saliva, showing that in the oral resident bacteriaderived-SCFA is functionally involved in controlling oral mucosal immunity and inflammation (62). Lending credence to this tenet, antibiotics treated mice show not only increased oral inflammation, but also intestinal immunopathology, when infected with oral Candida. Mechanistically, antibiotic treatment results in reduced Tregs, Th17 and Treg17 cells in oral mucosa and tissue draining cervical and axillary lymph nodes in infected mice. Intestinal inflammation in oral Candida infected mice is characterized by an increase in IFN-γ producing Th1 cells and co-producers of IFN-γ and IL-17A (Th1<sup>∗</sup> ) cells. Although the exact mechanism of antibiotic mediated reduction of Tregs, Th17 cells and Treg17 cells is unclear, administration of SCFA partially restored these populations and reduces oral immunopathology during the infection. SCFA administration however, only moderately ameliorates the intestinal inflammation. Therefore, the mechanism of Th1-mediated gut inflammation during oral Candida infection in the context of altered microbiota remains to be addressed. Recently, Atarashi et al. showed that oral bacterium Klebsiella spp. isolated from the salivary microbiota elicits a severe Th1 gut inflammation in the context of intestinal dysbiosis, in a genetically susceptible host (238). This finding underscores the role of oral resident microbes such as Klebsiella spp. and C. albicans in modulating T cells, possibly translocating to gut and causing overt inflammation in the gut in the context of resident microbial dysbiosis. Supporting this tenet, post oral gavage of C. albicans-infected mice pre-treated with antibiotics showed significantly altered composition of intestinal microbiota as well as CD4<sup>+</sup> T cell mediated lung inflammation, following aerosol introduction of an allergen. However, mice without any antibiotics pre-treatment did not develop an allergic response in the airways (239, 240). Whether changes in SCFA, or Treg and Th17 cells in the lung contribute to the inflammation is unknown.

Mechanistically, SCFAs also cause acetylation of p70 S6 kinase and phosphorylation rS6, promoting the mTOR activity. mTOR activity was shown to be required for generation of Th17 (T helper type 17), Th1, and IL-10<sup>+</sup> T cells (241). Moreover phosphoinositide 3-kinase and mTOR pathways play pivotal roles in integrating growth signals in CD4<sup>+</sup> T cell differentiation (242–249). Multiple studies support the role of mTORC1 and mTORC2 proteins in regulating Th17 and Treg fate decisions (247, 250, 251). mTORC1 signaling is constitutively active in Treg cells, and disruption of mTOR protein as well as unrestrained mTOR hyper-activation, both have been shown to cause autoimmunity by impairing Foxp3 expression and Treg functions (252–260). Another study has also shown that mTORC1 and its downstream target hypoxia-inducible factor-1α (HIF-1α) are needed for Foxp3 induction, Treg lipid and cholesterol biosynthesis from glucose, and proliferation and suppressive function in vivo (244, 254). Taken together, while direct role of SCFA in mediating mTOR activation and subsequent Treg induction in mucosa is unclear, these studies highlight the importance of how immunologically relevant microbiome can control Tregs and mucosal homeostasis through multiple mechanisms.

## MICROBIOTA AND TREG/TH17 CELL REGULATION OF IMMUNE SENESCENCE AND CHRONIC INFLAMMATION

While resident microbes have aforementioned protective functions in mucosa, they can also trigger and sustain inflammation during aging and other chronic inflammatory conditions. Some studies demonstrate direct relationship between aging and changes in microbiota, albeit the mechanisms remain largely unstudied. Aging causes increased accumulation of gut Enterobacteria, Streptococci, and yeasts but declining levels of Akkermansia muciniphila, Bifidobacteria and Bacteroides (261–266). Reduced Akkermansia muciniphila is associated with reduced butyrate and impaired intestinal barrier. Consequently, aged mice display endotoxin leakage, and triggering of 4-1BB receptor signaling and insulin resistance. In oral mucosa, aging causes higher levels of RANKL<sup>+</sup> cells, and increased inflammatory Th17 cell accumulation, with concomitant loss of alveolar bone, which are dependent on the presence of commensal microbiota (30, 267, 268). In contrast, these events do not occur in in germfree mice periodontium, showing potentially pathogenic roles of commensal microbiota in aging associated dysbiosis setting. Similarly, resident microbiota have been implicated in the onset and progression of experimental autoimmune encephalomyelitis (EAE) (269). GF mice exhibit lower levels of the pro-inflammatory cytokine IFN-γ and IL-17A producing cells, and a reciprocal increase in Tregs in the intestine and spinal cord. These changes in GF mice correlate with a significantly attenuated EAE, compared with conventionally raised mice. Remarkably, intestinal colonization with SFB alone can promote Th17 cells in the gut and in the central nervous system (CNS), enhancing disease progression (270). Furthermore, partial elimination of intestinal microbiota ameliorates established collagen-induced arthritis by dampening Th17 responses in mice (271). Some bacteria also provide inflammatory signals resulting in chronic inflammation and tumorigenesis, likely by inducing genetic and epigenetic changes in host cells. For example, Fusobabacteria spp. has been implicated in increased risk of IBD and colorectal cancer (272– 275). Also, in oral mucosa, the abundance of Fusobacterium increases, while the number of Streptococcus, Haemophilus, Porphyromonas, and Actinomyces decreases with cancer progression in oral squamous cell carcinoma (276). Interestingly, Fusobabacteria, and several other bacteria of oral mucosal origin, including genera of Streptococcus, Staphylococcus, Peptostreptococcus may translocate to intestine in the context of gut inflammation and carcinogenesis (277–279), similar to Klebsiella spp and C. albicans in susceptible host (62, 238). It is tempting to speculate that loss of Treg functions in the context of dybiosis, excessive SCFA and oral microbial translocation may have contributed to exuberant intestinal inflammation and predisposition to carcinogenesis in these studies (**Figure 2**). However, whether the mouth- to -gut translocation is a cause, or consequence of dysbiosis and intestinal inflammation, and the underlying mechanisms still remain to be understood and warrant further investigation.

#### THERAPEUTIC APPLICATIONS OF MICROBIOTA ALTERATIONS AND MICROBIOTA DERIVED METABOLITES

As we discussed above, studies on patient cohorts, mechanistic studies on mice and epidemiological studies have led to a better understanding of how microbiota changes impact mucosal immunity, and vice versa. Mechanistic "proof-of principle" studies using disease models have opened ways to manipulate these processes, providing therapeutic approaches. Some of the widely used approaches include administration of sodium butyrate and pre- and pro-biotics, and transplantation of fecal microbiota (280–283). However, there are hurdles in pro-biotic and microbiota transplantation approaches. Existing microbiota, whether it is healthy or dysbiotic is largely stable over time in an individual. Without profound perturbation of the existing microbiota, it is challenging to introduce microbiota exogenously. The effects of exogenous bacteria introduced by probiotic and transplant approaches are greatly influenced by existing microbiota in a competitive niche, and are inconsistent. Therefore, approaches to target these niches in favor of exogenous bacteria are being studied (283, 284). Direct administration of microbial derivatives appears to be a promising venue. Butyrate has been shown to alleviate high-fat-diet induced non-alcoholic fatty liver disease. It potently down modulates peroxisome proliferator-activated receptor α-mediated activation of β oxidation, causing reduced inflammation (285). For cART

FIGURE 2 | Cross talk between microbiota and immune cells during homeostasis and dysbiosis–Role of Th17 cells and Tregs in oral and intestinal mucosa. During homeostatic conditions, normal microbiota promote the stimulation of epithelial cells, Th17 cells and Tregs, and maintain barrier function and commensal tolerance. In oral mucosa, Th17 cells are induced by mastication induced mechanical damage, independent of commensals. However, in both mucosae SCFA mediated induction of Tregs is key for mucosal barrier function and immunomodulation. During inflammation and cancer, excessive SCFAs can increase inflammatory Th17 cells and Treg population that may be dysfunctional. The nature of their interaction with Th17 cells, tumor associated M2-type macrophages and other cells remain unclear.

treated HIV<sup>+</sup> individuals, aside from cART treatment, probiotics have been studied to combat persistent systemic inflammation. This approach in the context of cART may lead to improved and holistic management of inflammatory events and higher cancer susceptibility in HIV+ patients. Application of probiotics has also shown positive effect on the course of pneumonia, acute exacerbation of bronchial asthma and COPD in mice models, but warrants further studies in humans (286). SCFA has been shown to have therapeutic potential in microbiome-targeted interventions in anti-aging medicine. Butyrate and dietary fibers have been shown to promote anti-inflammatory effects in the context of aging associated neuro-inflammation in mice (287). Adult and aged mice fed with 5% inulin (high fiber) diet for 4 weeks show an altered gut microbiome and increased butyrate, acetate, and total SCFA production, coinciding with a reduction in neuro-inflammation. High fiber supplementation in aging is a non-invasive strategy to increase butyrate levels, and these data suggest that an increase in butyrate through added soluble fiber such as inulin could counterbalance the age-related microbiota dysbiosis, potentially leading to neurological benefits (287, 288). Similarly, dietary fiber also suppresses colon carcinogenesis in polyposis mice (289). Mechanistically it has been shown to inhibit colorectal cancer cell migration through micro-RNA regulation (290). In summary, alterations of mechanisms of microbiota-host interactions are proving to hold promise for treating a variety of disorders in humans.

# CONCLUSION

It is now well established that resident microbes provide enormous advantages to the host, while dysbiosis can trigger acute and chronic inflammatory conditions. One of the mechanisms by which these microbes regulate immunity id through controlling Tregs and Th17 cells. These cells present in various mucosal locations and share various signaling pathways for their development and sustenance, as stated above. However, signals modulating these subsets unique to each mucosal environment in different epithelial cell contexts are unclear. Most mechanistic studies showing Treg/Th17 developmental regulation were performed using the in vitro cultures using cells isolated from blood (human), spleen and lymph nodes (mice). While there is enough evidence to show that these cells could be regulated by overlapping signaling mechanisms, cells from these mucosae were not directly compared for similarities and differences in their development and functions. Such studies are warranted to get further insights in to homeostatic and dysbiotic conditions in different mucosae. Such studies in the context of microbial manipulation approach will offer new avenues to manipulate their interactions with the host for treating immune-mediated and metabolic disorders. While mono-association of certain genera in GF mice have proven to alter mucosal Tregs and Th17 cells and offer some beneficial effects in some experimental settings (98), from a therapeutic perspective, the field is still at its infancy and warrants intense

#### REFERENCES


mechanistic investigations. Taken together, further research in microbiota targeted approaches will enable the field to take the center stage in the management of health and disease in humans.

#### AUTHOR CONTRIBUTIONS

PP and JH wrote the manuscript. NB, ES, MZ, and SJ contributed to the discussion.

#### FUNDING

PP was supported by startup funding by SODM, CWRU and NIH 1RO1DE026923 grant to PP. JH was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany's Excellence Strategy—EXC 2155 RESIST— Project ID 39087428, and the Priority Program Intestinal Microbiota (SPP1656).


T Cells Is Essential for Suppression of T-Cell-Mediated Inflammation. Immunity. (2015) 42:903–15. doi: 10.1016/j.immuni.2015.04.012


defense against prenatal bacterial pathogens. Cell Host Microbe. (2011) 10:54–64. doi: 10.1016/j.chom.2011.06.005


albicans Th17 cell infection model. Immunity. (2011) 34:422–34. doi: 10.1016/j.immuni.2011.03.002


monkeys are associated with protection against AIDS. J Clin Invest. (2005) 115:1082–91. doi: 10.1172/JCI23006


T cells upon antigen stimulation. PLoS ONE. (2012) 7:e51665. doi: 10.1371/journal.pone.0051665


mucosal Candida albicans infections. Mucosal Immunol. (2013) 6:900–10. doi: 10.1038/mi.2012.128


cells and gut flora and prolongs life of immune-deficient mice. Aging Cell. (2015) 14:945–56. doi: 10.1111/acel.12380


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

Copyright © 2019 Pandiyan, Bhaskaran, Zou, Schneider, Jayaraman and Huehn. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Mitochondrial DNA Leakage Caused by Streptococcus pneumoniae Hydrogen Peroxide Promotes Type I IFN Expression in Lung Cells

Yue Gao1,2† , Wenchun Xu1,2† , Xiaoyun Dou<sup>3</sup> , Hong Wang<sup>3</sup> , Xuemei Zhang1,2 , Shenghui Yang1,2, Hongyi Liao1,2, Xuexue Hu1,2 and Hong Wang1,2 \*

<sup>1</sup> Key Laboratory of Diagnostic Medicine Designated by the Ministry of Education, Chongqing Medical University, Chongqing, China, <sup>2</sup> School of Laboratory Medicine, Chongqing Medical University, Chongqing, China, <sup>3</sup> Institute of Life Sciences, Chongqing Medical University, Chongqing, China

#### Edited by:

Fernanda Cristina Petersen, University of Oslo, Norway

#### Reviewed by:

Dane Parker, New Jersey Medical School, United States Jessica Lynn Humann, Florida A&M University, United States

> \*Correspondence: Hong Wang

wanghongljf@cqmu.edu.cn †These authors have contributed

equally to this work as first authors

#### Specialty section:

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

Received: 26 December 2018 Accepted: 13 March 2019 Published: 28 March 2019

#### Citation:

Gao Y, Xu W, Dou X, Wang H, Zhang X, Yang S, Liao H, Hu X and Wang H (2019) Mitochondrial DNA Leakage Caused by Streptococcus pneumoniae Hydrogen Peroxide Promotes Type I IFN Expression in Lung Cells. Front. Microbiol. 10:630. doi: 10.3389/fmicb.2019.00630 Streptococcus pneumoniae (S. pn), the bacterial pathogen responsible for invasive pneumococcal diseases, is capable of producing substantial amounts of hydrogen peroxide. However, the impact of S. pn-secreted hydrogen peroxide (H2O2) on the host immune processes is not completely understood. Here, we demonstrated that S. pnsecreted H2O<sup>2</sup> caused mitochondrial damage and severe histopathological damage in mouse lung tissue. Additionally, S. pn-secreted H2O<sup>2</sup> caused not only oxidative damage to mitochondrial deoxyribonucleic acid (mtDNA), but also a reduction in the mtDNA content in alveolar epithelia cells. This resulted in the release of mtDNA into the cytoplasm, which subsequently induced type I interferons (IFN-I) expression. We also determined that stimulator of interferon genes (STING) signaling was probably involved in S. pn H2O2-inducing IFN-I expression in response to mtDNA damaged by S. pn-secreted H2O2. In conclusion, our study demonstrated that H2O<sup>2</sup> produced by S. pn resulted in mtDNA leakage from damaged mitochondria and IFN-I production in alveolar epithelia cells, and STING may be required in this process, and this is a novel mitochondrial damage mechanism by which S. pn potentiates the IFN-I cascade in S. pn infection.

Keywords: Streptococcus pneumoniae, hydrogen peroxide, mitochondrial damage, mtDNA, IFNβ, STING

# INTRODUCTION

Streptococcus pneumoniae (S. pn), a member of the human nasopharyngeal microbiota, can cause pneumonia and other invasive pneumococcal diseases, such as otitis media, meningitis and bacteremia, especially in children and the elderly (Weiser et al., 2018). The virulence of S. pn is known to be dependent upon several factors, including its polysaccharide capsule, surface proteins, enzymes, and the cholesterol-dependent cytolysin, pneumolysin (Ply) (Mitchell and Mitchell, 2010). These virulence factors play an important role in the invasion of S. pn into the host. Interestingly, S. pn can secrete a substantial amount of hydrogen peroxide (H2O2) via an enzymatic reaction catalyzed by pyruvate oxidase, SpxB, during the aerobic metabolism. The spxB gene is not only a virulence determinant in S. pn, but it is also essential for resistance to the toxic by-product

produced by itself (Li-Korotky et al., 2009). It can influence the synthesis of acetyl-phosphate, a potential source of ATP under the aerobic conditions in S. pn (Pericone et al., 2003). Meanwhile, it can affect the sugar utilization pattern and capsule biosynthesis (Carvalho et al., 2013). Research has shown that strains of S. pn that lack spxB during growth produce significantly reduced levels of H2O<sup>2</sup> (Yesilkaya et al., 2013; Echlin et al., 2016). And compared with the wide type S. pn, the spxB mutant strain showed reduced virulence in animal models for nasopharyngeal colonization and pneumonia (Spellerberg et al., 1996). Through its ability to produce H2O2, S. pn is able to not only induce autolysis (Regev-Yochay et al., 2007), but also inhibit a variety of competing organisms, such as Haemophilus influenzae (Pericone et al., 2000) and Staphylococcus aureus (Regev-Yochay et al., 2006) in the aerobic environment of the respiratory tract.

Previous research has shown that S. pn-secreted H2O<sup>2</sup> influences the host physiology and immune defense. During pneumococcal meningitis, both S. pn-secreted H2O<sup>2</sup> and Ply are sufficient to induce mitochondrial damage, trigger the release of apoptosis-inducing factor (AIF) from mitochondria, and ultimately mediate apoptosis (Braun et al., 2002). S. pn-secreted H2O<sup>2</sup> has been shown to induce endoplasmic reticulum (ER) stress, activate the mitogen-associated protein kinase (MAPK) signaling pathways, and regulate target genes (Loose et al., 2015). Furthermore, H2O<sup>2</sup> secreted by S. pn is required for the induction of cardiomyocyte cell death, which is involved in the pathogenesis of S. pn infection in the heart (Brissac et al., 2017). However, the mechanism of how S. pn-secreted H2O<sup>2</sup> activates the immune system are not fully understood in acute pneumonia.

Mitochondrial deoxyribonucleic acid (mtDNA) is an important damage-associated molecular pattern (DAMP), which contains a large number of unmethylated CpG sequences (Nakayama and Otsu, 2018). It is thought that mtDNA is more susceptible to damage, owing to an inefficient DNA repair mechanism and the lack of protective histone packaging. Damaged mtDNA released to into the cytoplasm or circulation has been shown to induce the transcription of pro-inflammatory cytokines, including MMP-8, TNFα, IL-6, and IL-1β (Fang et al., 2016). Moreover, mtDNA is involved in the induction of endothelial inflammation (Mao et al., 2017) and cardiomyocyte ischemia/reperfusion-injury (Hu et al., 2018). Previous research has confirmed that S. pn-secreted H2O<sup>2</sup> leads to cytotoxic DNA damage in lung cells (Rai et al., 2015). However, the source of this oxidative DNA damage and whether it is involved in the host immune response remains unclear.

The ability to sense aberrant nucleic acids is a cornerstone of the innate immune system against pathogens. Stimulator of interferon genes (STING), a key innate immune signaling adaptor, responds to various forms of DNA species, including self-DNA from the nucleus of damaged cells. Self-DNA may cause various autoimmune diseases such as systemic lupus erythematosus (SLE) (Barber, 2015). Certain bacteria, such as Listeria monocytogenes, secrete cyclic dinucleotides (CDNs) that induce STING signaling within the host (Sauer et al., 2011). Similarly, S. pn DNA stimulates type I interferons (IFN-I) (IFNα, IFNβ) production in a STING-dependent manner (Parker et al., 2011; Koppe et al., 2012). In addition, mtDNA, which exists as a closed circular doubles-stranded DNA species, is able to activate STING, resulting in the upregulation of IFN-I and other interferon-stimulated genes (ISGs), under cellular damage and stress (Fang et al., 2016). However, there is little data to indicate mtDNA damaged by S. pn-secreted H2O<sup>2</sup> induces IFN-I expression.

In this study, we demonstrated that S. pn-secreted H2O<sup>2</sup> is capable of causing mitochondrial damage and mtDNA leakage into the cytosol of human alveolar epithelial cells, which further trigger the expression of IFNβ mediated by STING signaling probably. Thus, this study revealed a new strategy by which S. pn activates the host immune response.

# MATERIALS AND METHODS

## Ethics Statement

All animal experiments in this study were conducted in accordance with the guidelines of the Institutional Animal Care and Use Committee of Chongqing Medical University and were authorized by the Animal Ethics Committee of Chongqing Medical University.

#### Bacterial Strains and Culture Conditions

The S. pn standard strain D39 (NCTC 7466, serotype 2) was purchased from the National Collection of Type Cultures (London, United Kingdom). The S. pn spxB knockout mutant strain (D391spxB) was constructed by long flanking homologypolymerase chain reaction (LFH-PCR), as described before (Wu et al., 2014). Briefly, the spxB gene was substituted with an erythromycin resistant sequence. The positive clones were selected on blood agar plates containing 0.25 µg/ml erythromycin and identified by PCR. All S. pn strains were grown in C plus Y medium at 37◦C in 5% CO<sup>2</sup> until the optical density at 600 nm equaled 0.5 (OD600 = 0.5).

#### Mouse Model of Acute Pneumonia

Female C57BL/6 mice (6–8 weeks old) were purchased from Chongqing Medical University (Chongqing, China) and were maintained under specific-pathogen-free conditions in a temperature-controlled room of the animal facility at Chongqing Medical University. All animal experiments were approved by the respective ethics committees of Chongqing Medical University. Mice were inoculated intranasally with 1 × 10<sup>8</sup> CFU of D39 (NCTC 7466, serotype 2) or D391spxB in 30 µL of sterile PBS (n = 5 mice/group). Catalase was given intravenously (at 6, 12, 18, 22, 23, and 24 h) in the other five mice inoculated with 1 × 10<sup>8</sup> CFU of D39. Mice were sacrificed 24 h post-infection, and blood and lung homogenate supernatants were collected.

#### Cell Culture

The human lung alveolar carcinoma (type II pneumocyte) A549 cell lines were cultured in DMEM (Hyclone, United States) supplemented with 10% fetal bovine serum (FBS) (BI, United States) and 1% penicillin-streptomycin (Hyclone,

United States) at 37◦C in 5% CO2. Wild-type (WT) and STING knockout mouse embryonic fibroblast (MEF sting−/−) cells were also cultured in DMEM supplemented with 10% FBS (Gibco, United States) and 1% penicillin-streptomycin at 37◦C with 5% CO2. MEF sting−/− cells were kindly provided by Professor Chen Wang (School of Life Sciences and Technology, China Pharmaceutical University, Nanjing, China) (Lu et al., 2018). In order to construct a mtDNA-depleted cell line, A549 cells were cultured in DMEM supplemented with 10% FBS, 1% penicillin-streptomycin, and ethidium bromide (EtBr) (300 ng/ml) for 5 days at 37◦C in 5% CO2.

#### Mitochondrial Transmembrane Potential (19m) Assay

Mitochondrial transmembrane potential was assessed using a JC-1 kit (Solarbio, China). JC-1 is a fluorescent probe that indicates mitochondrial membrane potential loss. In normal cells, JC-1 aggregates in intact mitochondria (red fluorescence) but becomes a monomer (green fluorescence) in cells with disrupted mitochondrial membrane. After incubation with specific stimulators, including D39 with or without catalase (Cat) and D391spxB (MOI = 200) at 2 h. A549 cells were incubated in DMEM containing 10 µM JC-1 at 37◦C and protected from light for 20 min. The cells were then washed with ice-cold 1× JC-1 staining buffer twice and then imaged with a fluorescent microscope (Nikon ECLIPSE 80i, Japan).

## Lung Histology and Immunohistochemistry

Lung tissue was removed and fixed in 4% paraformaldehyde. The tissues were then embedded in paraffin, and then 5-µm sections were cut. The sections were stained with hematoxylin and eosin (Sigma-Aldrich, United States) and analyzed using a light microscope (Nikon ECLIPSE 80i, Japan). The degree of peribronchial inflammation was semi-quantitatively graded as described methods (Blanquiceth et al., 2016). The tissues were scored as follows: 0, normal; 1, a few cells; 2, a ring of cells 1 cell layer deep; 3, a ring of cells 2–4 cells deep; 4, a ring of cells 5–6 cells deep; and 5, a ring of cells of > 6 cells deep.

For immunohistochemistry of lung tissue sections, citrate buffer was used for antigen retrieval. Lung sections were then incubated with an anti-PINK1 antibody (Novus Biologicals, United States), according to standard protocols. The mean integrated optical density (IODs) of PINK1 expression was measured using Image-Pro Plus (Media Cybernetics, Silver Spring, MD, United States).

#### Transmission Electron Microscopy

A549 cells were infected with D39 or D391spxB for 2 h. The cells were then harvested, washed with sterile PBS twice, and fixed with ice-cold 4% glutaraldehyde. Fixed cells were sectioned according to the Electron Microscopy Research Service of Chongqing Medical University and observed with a Hitachi H-7500 transmission electron microscope (Hitachi, Japan).

# Extraction of Cytoplasmic DNA and Transfection

Cytoplasmic DNA was extracted as previously described (Holden and Horton, 2009). Briefly, 2 × 10<sup>6</sup> cells were stimulated as indicated. The cells were then harvested, washed with phosphatebuffered saline (PBS), and then mixed on a rotator in 500 µl of digitonin solution (25 µg/ml) containing 150 mM NaCl and 50 mM HEPES (pH 7.4) for 30 min at room temperature. The lysate was then centrifuged at 1000 × g for 5 min thrice in order to remove the nuclei and intact cells. The supernatant was transferred to a new tube and then centrifuged at 17,000 × g for 10 min in order to pellet the remaining cellular debris. Total cytosolic DNA were extracted using a DNA Blood Mini Kit (Qiagen, Germany), precipitated with 100 µl absolute ethanol, and stored at −20◦C. A549 cells were transfected with the cytosolic DNA for 6 h using Lipofectamine 2000 (Lipo 2000) (Invitrogen, United States) according to the manufacturer's instructions.

# mtDNA Copy Number and Transcription Level

In order to quantify mtDNA copy number, total DNA was extracted from the lungs of mice and A549 cells using a DNeasy Blood & Tissue Kit (Qiagen, Germany). For the quantification of mtDNA transcription level, total RNA was extracted from the lungs of mice and A549 cells using RNAiso plus reagent (Takara Bio, China) following the manufacturer's instructions. Cytochrome B (Cytb) and cytochrome c oxidase subunit 3 (CoxIII) were amplified in order to analyze the copy number and transcription level of mtDNA. Gapdh was used as the internal control (Hu et al., 2018). The primers used in this study are listed in **Table 1**.

#### Immunofluorescence and Confocal Imaging

Treated A549 cells were incubated with MitoTracker <sup>R</sup> Red CMXRos (Yeasen, China) (mitochondrial red fluorescent probe) at 37◦C for 45 min. Cells were fixed, permeabilized with 0.1% Triton X-100, blocked with 10% donkey serum, and then incubated with an anti-8-hydroxyguanine (8-OHdG) antibody (Santa Cruz Biotechnology, United States) at 4◦C overnight. Then the cells were incubated with fluorescent-labeled secondary antibodies (Bioss, China). Nuclei were stained with DAPI (Beyotime, China) at room temperature for 15 min. Cells were observed with a Nikon ECLIPSE Ti confocal microscope (Nikon, Japan). The images were captured and analyzed by using NIS-Elements Viewer (Nikon, Japan).

#### Western Blot Analysis

A549 cells were lysed with RIPA buffer (Beyotime, China) supplemented with PMSF and phosphatase inhibitor (100:1:1) (BioTools, United States). Protein samples were separated onto sodium dodecyl sulfate (SDS)-polyacrylamide gels and then transferred onto PVDF membranes (Millipore, United States). The membranes were incubated with a primary antibody

#### TABLE 1 | The sequences of PCR primers.

fmicb-10-00630 March 27, 2019 Time: 10:46 # 4


overnight at 4◦C, including anti-human STING and antihuman GAPDH (Cell Signaling Technology, United States). The membranes were then incubated with an HRP-conjugated secondary antibody for 1 h at 37◦C. Bands were visualized by using Image Lab (Bio-Rad Laboratories, Hercules, CA, United States). GAPDH was used as a loading control.

#### Real-Time PCR

Total RNA was extracted from the lungs of mice and cells using RNAiso plus reagent (Takara Bio, China), according to the manufacturer's protocol. The mRNA was then reverse transcribed into cDNA using the PrimeScriptTM RT reagent kit (Takara Bio, China). All real-time PCR reactions were performed using TB Green Premix Ex TaqTM II on a Bio-Rad CFX-96 cycler (Bio-Rad Laboratories, United States). The expression of mRNA was normalized against GAPDH. The data shown are representative of three separate experiments. The primers used in this study are listed in **Table 1**.

#### ELISA

Cytokine levels in the blood, lung homogenate supernatants, and cell culture supernatants were measured by using ELISA assays, according to the manufacturer's instructions. The following ELISA kits were used in this study: LEGEND MAXTM Mouse IFNβ (Biolegend, United States) and Human IFNβ (Cloudclone, China).

### Statistical Analysis

All experiments were performed at least three times, and the data are presented as the mean ± SD. Student's t-test was used for statistical analysis, and GraphPad Prism 5 software was used to perform statistical analysis for all experiments. P < 0.05 was considered statistically significant.

# RESULTS

#### H2O<sup>2</sup> Secreted by S. pn Induces IFN-I Expression in vivo and in vitro

In order to assess whether H2O<sup>2</sup> secreted by S. pn induces the production of IFN-I in vivo, we used a mouse model of acute pneumonia. C57BL/6 mice were intranasally infected with S. pn D39 or D391spxB for 24 h, catalase was given intravenously in the other five mice inoculated with 1 × 10<sup>8</sup> CFU of D39. ELISA assays demonstrated significant upregulation of IFNβ in the blood and lung homogenate supernatants of D39-infected mice compared to D391spxB-infected mice and catalase treatment of D39-infected mice (**Figure 1A**). Similarly, Ifnβ mRNA levels were increased in the lung tissue of D39-infected mice, but not in the lung tissues of D391spxB-infected mice and catalase treatment of D39-infected mice (**Figure 1B**). We also infected the human alveolar epithelial cell line A549 with D39 or D391spxB and analyzed the expression of IFN-I at 1 and 5 h post-infection. D39 infection was capable of inducing expression of IFNβ and IFNa2 (subtype of IFN-I) at both 1 and 5 h compared to D391spxB infection. Pre-treatment of A549 cells with catalase prior to D39 infection resulted in diminished IFNβ and IFNa2 transcript levels, in which 5-fold reduction was observed in IFNβ and IFNa2 transcript levels at 5 h post-infection (**Figures 1C,D**, left panel). We also found that the production of IFNβ was reduced by 50% in the supernatant of A549 cells infected with D391spxB as compared to D39-infected cells (**Figure 1C**, right panel). We further determined that D39 infection increased the expression of IFNa5 (subtype of IFN-I) at 5 h, but not 1 h post-infection (**Figure 1D**, right panel).

Together, these data suggest that H2O<sup>2</sup> secreted by S. pn is capable of inducing IFN-I expression in lung cells.

### S. pn-Secreted H2O<sup>2</sup> Is Sufficient to Induce Mitochondrial Dysfunction

Given that IFNβ production is triggered by mtDNA following mitochondrial stress (Fang et al., 2016), we theorized that H2O<sup>2</sup> generated by S. pn induces mitochondrial damage in mouse lung tissue. The mRNA level of peroxisome proliferatoractivated receptor γ (PPARγ) coactivator 1α (Pgc1-α), which is the master regulator of mitochondrial biogenesis and functions as a transcriptional co-regulator (Dorn et al., 2015), was markedly decreased by 75% in D39-infected mice, but not

in D391spxB-infected mice and catalase treatment of D39 infected mice (**Figure 2A**). PTEN-induced putative kinase 1 (PINK1) is a regulator of mitophagy, which is rapidly degraded when mitochondria are healthy, but accumulates on the surface of damaged mitochondria (Jin et al., 2010). We observed by immunohistochemistry that infection with D39 significantly induced the accumulation of PINK1 on the mitochondria in the lung tissue of mice, while both infection with D391spxB and catalase treatment of D39 infection did not increase PINK1-mitochondria interactions (**Figure 2B**). In addition, morphological and histopathological analyses of murine lung tissue revealed that severe pulmonary injuries were found in the D39-infected group, with obvious hemorrhage and massive inflammatory cell infiltration in the peribronchial, as compared to the D391spxB-infected group and these pulmonary injuries were not markedly observed in the lung tissue of mice infected with D39 pre-treatment with catalase (**Figure 2C**). These data suggest that S. pn-secreted H2O<sup>2</sup> causes mitochondrial damage in the lungs of mice.

To further understand the extent of mitochondrial damage induced by the H2O<sup>2</sup> generated by S. pn, we assessed mitochondrial function in A549 cells infected with S. pn by measuring the 19m. We used an oxidative phosphorylation uncoupler, carbonyl cyanide-m-chlorophenylhydrazone (CCCP), as a positive control, which causes depolarization of mitochondria and mitochondrial damage (Park et al., 2018). Our results showed an increase in green fluorescence in A549 cells after infection with D39 for 2 h, indicating that the 19m values significantly decreased. Moreover, the addition of catalase significantly attenuated the decrease in the 19m values induced by D39 infection (**Figure 2D**). Similarly, we used transmission electron microscopy to monitor changes in mitochondria morphology. We observed abnormal mitochondrial morphology in A549 cells infected with D39, including mitochondrial swelling, condensation, and abnormal cristae shape. However, these changes were not observed in D391spxB-infected cells (**Figure 2E**). These results suggest that S. pn-secreted H2O<sup>2</sup> induces mitochondrial damage in alveolar epithelial cells.

Taken together, our results demonstrate that H2O<sup>2</sup> produced by S. pn causes mitochondrial dysfunction in lung cells both in vivo and in vitro.

# S. pn-Secreted H2O<sup>2</sup> Mediates Oxidative Damage of Mitochondrial DNA

The expression of 8-hydroxyguanine (8-OHdG) is known to be reflective of oxidative DNA damage. To further clarify if mtDNA is damaged by S. pn-secreted H2O2, we evaluated the level of 8-OHdG in A549 cells infected with D39 by immunofluorescence analysis. The number of 8-OHdG-positive A549 cells significantly increased following D39 infection, but not D391spxB infection (**Figure 3A**). Furthermore, the addition of catalase markedly reduced the number of 8-OHdG-positive A549 cells infected with D39. These results suggest that S. pn-secreted H2O<sup>2</sup> causes oxidative damage to the mtDNA in A549 cells, and catalase pre-treatment may prevent this phenomenon. We also explored the mtDNA copy number in A549 cells infected with D39 at 2 h post-infection. Real-time PCR analysis demonstrated that the mtDNA copy number significantly reduced by 50% over time in A549 cells after D39 infection (**Figure 3B**). Furthermore, the reduction in mtDNA copy number and mtDNA transcript level in D39-infected A549 cells was largely prevented by catalase pre-treatment, which is consistent with our data showing that catalase pre-treatment also decreased 8-OHdG levels in mtDNA after D39 infection. In D391spxB-infected cells, the mtDNA transcript level and copy number were partially restored as compared with D39 infection. As expected, when A549 cells were exposed to 1 mM H2O2, we observed a 40% reduction in mtDNA copy number and a 60% reduction mtDNA transcript level by real-time PCR (**Figure 3C**). Likewise, copy number and transcript level of mtDNA were significantly decreased in D39 infected mouse lung tissue as compared with D391spxB-infected mice and catalase treatment of D39 infected mice (**Figure 3D**).

In short, these results provide evidence that H2O<sup>2</sup> secreted by S. pn induces significant oxidative damage in the mtDNA of lung cells.

## S. pn-Secreted H2O<sup>2</sup> Promotes Mitochondria DNA Leakage and the Induction of IFN-I

To address whether S. pn-secreted H2O<sup>2</sup> could lead to the leakage of mtDNA into the cytoplasm, we assessed the cytosolic mtDNA levels in D39-infected A549 cells. We show that D39 infection significantly elevated the level of mtDNA in the cytoplasm within the first 3 h post-infection. A549 cells infected with D391spxB did not exhibit this increase in mtDNA levels in the cytoplasm (**Figure 4A**). As expected, catalase pre-treatment inhibited D39 induced mtDNA leakage into the cytoplasm (**Figure 4B**). These findings suggest that H2O<sup>2</sup> generated by S. pn triggers mtDNA leakage into the cytoplasm of A549 cells.

We next sought to determine if IFN-I expression is induced by mtDNA damage caused specifically by S. pn-secreted H2O<sup>2</sup> to mtDNA. Thus, we isolated mtDNA from the cytoplasm of A549 cells subjected to various stimulations, including D39 with or without catalase, D391spxB and H2O2. We then exposed untreated A549 cells with the isolated mtDNA. Poly (dA:dT), a synthetic double-stranded DNA sequence, was used as a positive control. The cytosolic mtDNA isolated from D39-infected cells caused an upregulation in the transcription level of IFNβ, and a similar result was obtained with mtDNA from H2O2-stimulated

were transfected with cytosolic DNA isolated from different stimulations, including D39 with or without catalase (Cat) and D391spxB, as well as H2O2, IFNβ mRNA level were determined by real-time PCR. POLY (dA: dT) (2 µg/ml) was applied as positive control. (D) mtDNA in A549 cells were evaluated by real-time PCR after being treated with EtBr (300 ng/ml) for 5 days (left panel). IFNβ mRNA level in A549 cells treated with 1 mM H2O<sup>2</sup> or D39 (MOI = 200) were determined by real-time PCR (right panel). NC, negative control. All data were presented as means ± SD from three independent experiments. <sup>∗</sup>P < 0.05; ∗∗P < 0.01; ∗∗∗P < 0.001 was considered statistically significant and highly statistically significant differences, respectively; ns, not significant.

cells. However, there was no increase in the expression of IFNβ in A549 cells treated with mtDNA from D39-infected A549 cells pretreated with catalase or D391spxB-infected A549 cells (**Figure 4C**).

In order to ascertain the importance of mitochondria in the induction of IFNβ expression in D39-infected A549 cells, we constructed mtDNA-depleted A549 cells. A549 cells were exposed to low concentrations of ethidium bromide over time to effectively reduce the mtDNA content within these cells. Real-time PCR results confirmed the successful construction of mtDNA-deficient cells (**Figure 4D**, left panel). The expression of IFNβ in mtDNA-deficient cells was reduced by about 60% than that in WT cells following treatment with D39 or H2O<sup>2</sup> (**Figure 4D**, right panel), which suggests that mtDNA plays a critical role in S. pn H2O2-induced production of IFNβ.

# STING Signaling Is Probably Involved in the Activation of IFN-I by S. pn-Secreted H2O<sup>2</sup>

IFN-I production has been shown to be triggered by mtDNA through STING signaling (Fang et al., 2016). We examined the protein level of STING in A549 cells following different stimulations, including D39 with or without catalase, D391spxB and H2O2. The results showed that both D39 infection and H2O<sup>2</sup> stimulation up-regulated the expression of STING in A549 cells (**Figure 5A**), and the induction of STING was reduced by 20% by catalase treatment, a 40% reduction was observed after infection with D391spxB. To further confirm that STING is responsible for inducing IFN-I expression in response to detecting mtDNA oxidized by S. pn-secreted H2O2, WT MEFs and STING knockout MEFs (MEFs sting−/−) were stimulated with D39, D391spxB, and H2O2. Real-time PCR results demonstrated upregulation of Ifnβ and Ifna4 in WT MEF cells, but not in MEF sting−/− cells (**Figure 5B**). Curiously, there was no change in the transcriptional level of the IFNresponsive gene, C-X-C motif chemokine 10 (Cxcl10), in either WT MEFs or MEF sting−/− cells upon stimulation with D39, D391spxB or H2O2. These results demonstrated that STING is indispensable in S. pn H2O2-induced production of IFN-I in MEF cells.

Additionally, we also determined the expression of IFN-Istimulated genes ISG15 and OASl-1, as well as RNF185 (which has been reported to positively regulate the STING signaling pathway) in A549 cells infected with D39 (**Figure 5C**) (Wang Q. et al., 2017). We found that ISG15, OASl-1, and RNF185 were induced in A549 cells infected with D39 at 1 h, but not 5 h post-infection. Furthermore, the addition of catalase or infection with D391spxB diminished the expression of these genes at 1 h. Taken together, these findings indicate that STING signaling is probably involved in inducing the expression of IFN-I in response to mtDNA damaged by S. pnsecreted H2O2.

# DISCUSSION

In this study, we ascertained that S. pn-secreted H2O<sup>2</sup> promoted IFNβ production in lung cells, which was mediated by mtDNA leakage from mitochondria damaged by H2O2. Neutralizing

the H2O<sup>2</sup> produced by S. pn H2O<sup>2</sup> with catalase markedly attenuated mitochondrial malfunction and IFNβ expression, suggesting that targeting H2O<sup>2</sup> during S. pn infection may offer therapeutic strategies.

There are several virulence factors of S. pn that are involved in the disease process. Specifically, S. pn is able to secrete substantial amounts of H2O2—up to a concentration of approximately 2 mM under aerobic conditions (Duane et al., 1993; Echlin et al., 2016; Lisher et al., 2017). Other Streptococcus species, such as Streptococcus sanguis (Sumioka et al., 2017), Oral streptococci (Matsushima et al., 2017), which have been reported to secrete H2O2. Several studies indicate that H2O<sup>2</sup> secretion by most Streptococcus species is universal and indispensable. The production of H2O<sup>2</sup> is dependent on the pyruvate oxidase gene spxB, which confers a selective advantage in co-colonization (Pesakhov et al., 2007; Regev-Yochay et al., 2007). Deletion of spxB results in a significant reduction in H2O<sup>2</sup> to approximately 20% of level produced by WT S. pn (Echlin et al., 2016). However, spxB may also play other roles in the virulence of S. pn as there are several different serotypes. It has been shown that the virulence of a spxB knockout mutant of strain D39 (serotype 2) is attenuated in a murine model of nasopharyngeal colonization (Spellerberg et al., 1996), while a spxB mutant of S. pn serotype 1 is hypervirulent (Syk et al., 2014).

Research have reported that H2O<sup>2</sup> inhibited cell migration in a dose-dependent manner, and this would impair airway epithelial cell repair (Hamada et al., 2016). And A549 cells exposed to H2O<sup>2</sup> caused powerful LDH release and a necrotic phenotype rather than programmed cell death (Schmeck et al., 2004). These suggested that H2O<sup>2</sup> is able to damage lung tissue. Some studies have supported that the toxic H2O<sup>2</sup> secreted by S. pn could cause cellular oxidative stress and participate in cellular immune responses through different signaling pathways. A previous study reported that two pneumococcal toxins, Ply and H2O2, led to mitochondrial damage and consequently caused apoptosis of brain cells (Braun et al., 2002). Recent studies have shown that S. pn-secreted H2O<sup>2</sup> induced DNA damage and apoptosis in lung cells, and contributed to the genotoxicity and virulence of S. pn (Rai et al., 2015). Moreover, there are some reports showing that pneumococcal H2O2-induced stress signaling regulated the expression of inflammatory genes (Loose et al., 2015). However, others have reported that the pneumococci-induced oxidative stress was independent of S. pnsecreted H2O<sup>2</sup> and Ply but depended on the pneumococcal autolysin LytA (Zahlten et al., 2015). Here, we showed that S. pn- secreted H2O<sup>2</sup> alone was able to induce mitochondrial oxidative damage, impairing mtDNA replication and decreasing mtDNA content in lung cells. Moreover, H2O<sup>2</sup> is a type of reactive oxygen species (ROS), and as such, is an important

signaling molecule that mediates oxidative stress and cellular damage (Wible and Bratton, 2018). Previous studies on intestinal health have revealed that H2O<sup>2</sup> upregulated intracellular and mitochondrial ROS expression (Jiang et al., 2017). However, whether H2O<sup>2</sup> and the production of ROS induces mtDNA damage still needs to be further clarified.

The unique aspect of mitochondria is that it is the only source of DNA in cells that does not reside in the nucleus. mtDNA-mediated signaling is the basis of the host immune defense in several diseases. In atherosclerosis, oxidative damage and replication errors are the sources of mtDNA defects, which lead to mitochondrial dysfunction and directly promote atherosclerosis (Yu and Bennett, 2014). In intestinal ischemia reperfusion (I/R), mtDNA contributed to the early phase of I/R injury and amplified the inflammatory response (Yue et al., 2015; Hu et al., 2018). During the process of metabolic stress–induced endothelial inflammation and insulin resistance, palmitic acid caused mtDNA leakage into the cytoplasm, and activated STING signaling to mediate the intercellular adhesion molecule (ICAM)-1 expression and endothelial inflammation (Mao et al., 2017). Interestingly, mtDNA could activate several innate immune pathways including TLR9, NLRP3 and STING signaling pathways in the mammalian immune responses (Fang et al., 2016).

IFN-I are pleiotropic cytokines produced in response to viruses, bacteria, and parasites. In bacterial infection, lipopolysaccharide (LPS) or bacterial nucleic acids are recognized by innate immune receptors, triggering IFN-I production (Boxx and Cheng, 2016). IFN-I induce differential effects on the immune response of the host. Listeria monocytogenes was reported to induce IFNβ expression, suppress the production of IFNγ and TNFα, thereby promoting infection (Auerbuch et al., 2004; Rayamajhi et al., 2010). In contrast, S. pn DNA initiated an IFN-I cascade that contributed to pneumococcal clearance, and this process played an important part in the host defense against pneumococci by inhibiting bacterial transmigration (Parker et al., 2011; LeMessurier et al., 2013). Interestingly, our data showed that exposure of alveolar epithelial cells to H2O<sup>2</sup> produced by S. pn was sufficient to induce mtDNA leakage into the cytoplasm and induce IFNβ production.

Growing evidence has indicated that STING signaling can be triggered by DNA from pathogens or damaged self-DNA in the cytoplasm (Barber, 2015). Hartlova et al. (2015) demonstrated that unrepaired DNA lesions promoted the production of IFN-I via the STING signaling pathway, which strengthened antimicrobial immunity. In acute pancreatitis, STING sensed self-DNA from dying acinar cells and promoted inflammation (Zhao et al., 2018). Our results confirmed that S. pn- secreted H2O<sup>2</sup>

#### REFERENCES


resulted in both the production of IFN-I and the activation of IFN-I-stimulated genes, ISG15 and OASl-1, and RNF185 in A549 cells. We verified that deletion of STING notably impaired the expression IFN-I in MEF cells. These data suggested that STING signaling may play an indispensable role in the production of IFN-I induced by S. pn H2O2.

A previous study showed that IFNβ induced ROS production in human myotubes, which contributed to mitochondrial dysfunction and resulted in muscle impairment and continued inflammation in dermatomyositis (Meyer et al., 2017). Likewise, another study reported that caspases controlled antiviral immunity through cGAS cleavage during inflammasome activation, resulting in reduced IFN-I expression, revealing a negative feedback that regulates the output of DNA-sensing pathways (Wang Y. et al., 2017). However, further research is needed in order to elucidate the end result of IFN-I production by S. pn-secreted H2O2. Specifically, studies need to determine if this signaling pathway ultimately favors bacterial clearance or aggravates host cell apoptosis.

Overall, our findings demonstrated that S. pn-secreted H2O<sup>2</sup> induced mtDNA leakage into the cytoplasm, which resulted in the activation of the IFN-I, and this process may be mediated via STING signaling. We also confirmed that S. pn H2O<sup>2</sup> was sufficient to mediate mitochondrial oxidative stress, which underscores the importance of mitochondrial homeostasis during the host immune defense. In summary, we have identified a novel signaling mechanism that may serve as a potential target for controlling S. pn infection.

#### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and/or the supplementary files.

# AUTHOR CONTRIBUTIONS

YG, WX, and XZ conceived and designed the experiments. YG, XD, SY, and HW (fourth author) performed the experiments. YG, XZ, and HL analyzed the data. YG, XH, and HW (corresponding author) wrote the manuscript. WX and HW (corresponding author) reviewed and edited the manuscript.

#### FUNDING

This work was supported by the National Natural Science Foundation of China (No. 81671639).



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

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

# Interactions Between the Gut Microbiota and the Host Innate Immune Response Against Pathogens

#### Hong-Yu Cheng† , Meng-Xia Ning† , De-Kun Chen\* and Wen-Tao Ma\*

*Veterinary Immunology Laboratory, College of Veterinary Medicine, Northwest Agriculture and Forestry University, Yangling, China*

#### Edited by:

*Sudhanshu Shekhar, University of Oslo, Norway*

#### Reviewed by:

*Matthieu Million, IHU Mediterranee Infection, France Fei Teng, Cornell University, United States Esen Sefik, Yale University, United States*

#### \*Correspondence:

*De-Kun Chen cdk@nwafu.edu.cn Wen-Tao Ma mawentao@nwafu.edu.cn*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *30 November 2018* Accepted: *07 March 2019* Published: *29 March 2019*

#### Citation:

*Cheng H-Y, Ning M-X, Chen D-K and Ma W-T (2019) Interactions Between the Gut Microbiota and the Host Innate Immune Response Against Pathogens. Front. Immunol. 10:607. doi: 10.3389/fimmu.2019.00607* The mammalian intestine is colonized by over a trillion microbes that comprise the "gut microbiota," a microbial community which has co-evolved with the host to form a mutually beneficial relationship. Accumulating evidence indicates that the gut microbiota participates in immune system maturation and also plays a central role in host defense against pathogens. Here we review some of the mechanisms employed by the gut microbiota to boost the innate immune response against pathogens present on epithelial mucosal surfaces. Antimicrobial peptide secretion, inflammasome activation and induction of host IL-22, IL-17, and IL-10 production are the most commonly observed strategies employed by the gut microbiota for host anti-pathogen defense. Taken together, the body of evidence suggests that the host gut microbiota can elicit innate immunity against pathogens.

Keywords: gut microbiota, host innate immunity, antimicrobial peptides, inflammasome, IL-22, IL-17, IL-10

## INTRODUCTION

The mammalian intestine is home to a complex and dynamic population of microorganisms, termed the "gut microbiota" (1, 2). These microorganisms, which co-evolved with the host as part of a mutually beneficial relationship (3), include bacteria, fungi and viruses (4, 5). Accumulating evidence indicates that the gut microbiota can participate in the maturation and function of the innate immune system, while also playing many complex roles in the host defense against pathogens (6). On the one hand, the gut microbiota can help repair intestinal mucosal barrier damage (7, 8); on the other hand, gut microbiota mediates host anti-pathogen defenses (9).

In the past decade, studies of germ-free (GF) mice have provided clues to elucidate the complexity of the intestinal microbiota (10, 11) and its importance to host health (12, 13). Mounting research shows that at least a thousand different gut microbiota species, such as Firmicutes, Bacteroidetes, Actinobacteria, Proteobacteria, and others, contribute to host defense against harmful microorganisms (14, 15).

Recently, several studies have begun to elucidate the molecular mechanisms underlying how the gut microbiota regulates host innate immunity against pathogens (16, 17), including a role in helping the host resist pathogen colonization. In this review, we summarize the main mechanisms by which commensal bacteria, including certain probiotic species, actively prevent pathogen colonization of the host.

#### Cheng et al. Control of Pathogens by Gut Microbiota

#### GUT MICROBIOTA AND ANTIMICROBIAL PEPTIDES

#### Defensins

The α-defensins, microbicidal peptides mainly produced by Paneth cells, are key components of innate immunity. They control pathogen growth within the intestine (18–20) and their production can be directly elicited by both Gramnegative and Gram-positive bacteria, as well as by bacterial metabolites (e.g., lipopolysaccharide, lipoteichoic acid, lipid A, and muramyl dipeptide) (21–23). By contrast, live fungi and protozoa do not appear to stimulate Paneth cells and thus fail to elicit Paneth cell degranulation (21). Nevertheless, recent research has found that the gut microbiota plays an important role in induction of α-defensins expression against pathogens (24). In one in vitro study, live E. coli or S. aureus, live or dead S. typhimurium, lipopolysaccharide (LPS), lipid A, lipoteichoic acid (LTA), or liposomes could stimulate isolated intact intestinal crypts, demonstrating that intestinal Paneth cells may contribute to α-defensins secretion by sensing the presence of exogenous bacteria and bacterial antigens (21). To investigate whether gut microbiota possess the same or similar functions, Shipra Vaishnava and colleagues used a CR2- MyD88 Tg mouse model, whereby Paneth cells were the sole cell lineage expressing MyD88, to demonstrate that Paneth cells may directly sense enteric bacteria to trigger the MyD88-dependent antimicrobial program. Furthermore, increased numbers of Salmonella were observed to be internalized by mesenteric lymph node (MLN) cells of MyD88−/<sup>−</sup> and germ-free mice as compared to corresponding numbers observed for wildtype mice (25). Similarly, transcriptional profiles have shown that α-defensin gene (Defa) transcripts were less abundant in intestinal microbiota-free mice and TLRs-deficient or MyD88 deficient mice, but could be recovered after stimulation with toll-like receptor (TLR) agonists, specifically agonists of TLR2 or TLR4 (26). Thus, commensal microbiota appears to protect the host against pathogen invasion by triggering enteric Paneth cell TLR-MyD88 signaling. Notably, this mechanism is distinct from the NOD2-dependent antimicrobial response (25, 27, 28), since the former mechanism entails triggering of expression of multiple antimicrobial factors (25). However, several humanbased studies have demonstrated that mutations in the NOD2 peptidoglycan sensor actually did reduce secretion of α-defensins (29–33). Therefore, these contradictory human and mouse study results warrant further research. Notably, another study has demonstrated that Cd1d−/<sup>−</sup> mice exhibited a defect in Paneth cell granule ultrastructure that specifically resulted in an inability to degranulate after bacterial colonization, with an increased load of segmented filamentous bacteria (SFB) also noted (34). Thus, no clear evidence demonstrates that CD1d mediates regulation of gut microbiota via α-defensins expression.

Meanwhile, more recent research has begun to examine the mechanism of how the gut microbiota influences α-defensins secretion. Studies using the Caco-2 IEC line have demonstrated that lactic acid strongly suppresses transcription of the αdefensin gene, while cecal content may include as yet unidentified factors which enhance concomitant α-defensin 5 expression (35). However, contrary to the aforementioned results, Menendez et al. found that Defa expression was partially restored in vivo by lactobacillus administration to antibiotic-treated mice (26). Notably, an emerging role of vitamin D, a lactobacillus metabolite, has been recently discovered that exerts an effect opposite on α-defensins expression to that exerted by lactate (36, 37). To reconcile these results, Su et al. used a mouse model and certain feed formulations to demonstrate that VDD- and HFD ± VDD-fed mice exhibited reduced levels of expression of αdefensin and MMP7 (a metalloproteinase that can proteolytically convert pro-α-defensins to their mature and active forms) within ileal crypts as compared to results for control and HFD groups. Moreover, their results demonstrated a critical role of vitamin D signaling in maintaining steady-state expression of α-defensins and MMP7 under physiological conditions. Subsequently, Su et al. have demonstrated that dietary vitamin D deficiency resulted in loss of Paneth cell-specific α-defensins, which may lead to intestinal dysbiosis and endotoxemia (38). Of note, oral administration of α-defensin suppressed Helicobacter hepaticus growth in vivo (38). Meanwhile, using complementary mouse models of defensin deficiency (MMP7−/−) and surplus (HD5+/+), Salzman noted defensin-dependent reciprocal shifts in proportions of dominant bacterial species within the small intestine with no changes in total bacterial numbers observed (**Table 1**). Upon further research, this group observed that mice overexpressing HD5 exhibited a significant loss of segmented filamentous bacteria (SFB), resulting in reduced numbers of Th17 cells within the lamina propria (48). However, direct evidence implicating the involvement of SFB in α-defensin production is still lacking and studies on α-defensin regulation by specific commensal microorganisms are still rare, warranting further research. Nevertheless, in view of existing research results, we believe that the discovery of specific microorganisms through research focusing on specific metabolic pathways may be a more fruitful approach.

With regard to β-defensins, which directly kill or inhibit the growth of microorganisms (49), these agents have been shown to exert antimicrobial activity against some species of enteric pathogenic Gram-positive S. aureus and S. pyogenes, as well as against Gram-negative P. aeruginosa, E. coli and the yeast C. albicans (50). In fact, accumulating evidence has shown that, similarly to α-defensins, β-defensins secretion is also regulated by the gut microbiota. For example, using in vitro studies of HT-29 and Caco-2 human colon epithelial cell lines, human fetal intestinal xenografts have been observed to constitutively express hBD-1 but not hBD-2, with upregulation of only the latter in xenografts intraluminally infected with Salmonella (51). Meanwhile, it has been independently shown that preincubation of Caco-2 cells with live E. faecium significantly reduced S. typhimurium internalization by 45.8%, while heat-killed E. faecium pretreatment had no effect on pathogen internalization (49). This result aligns with the latest research, which has shown that only live gut microbiota, as modeled using Lactobacillus acidophilus PZ 1129 and PZ 1130, Lactobacillus paracasei, Lactobacillus plantarum, E. coli K-12, and E. coli Nissle 1917, can strongly induce expression of hBD-2 in Caco-2 intestinal epithelial cells in a time- and dose-dependent manner (39–42)


TABLE 1 | Gut microbiota protects the host against pathogen infections and the relevant mechanisms.

(**Table 1**). Notably, the E. colistrain Nissle 1917, a non-pathogenic Gram-negative strain isolated in 1917 by Alfred Nissle, elicited the most marked expression of induced β-defensin expression in vitro (39–42). Interestingly, Schlee et al. constructed several E. coli Nissle 1917 deletion mutants and pinpointed flagellin as the major stimulatory factor for triggering of β-defensin secretion in the presence of that strain (40). Meanwhile, Wehkamp et al. and others have found that E. coli Nissle 1917-induced βdefensin expression in cell culture was mediated by NF-κBand MAPK/AP-1-dependent pathways (39–42). Nevertheless, in vivo studies are still needed to confirm if gut microbiota can induce β-defensins expression to reduce pathogen colonization and control gut homeostasis (**Table 1**). Recently, to further clarify the relationship between gut microbiota and β-defensin secretion, Miani et al. used a mouse model and antibiotic treatment experiments to study the participation of dysbiotic microbiota and a low-affinity aryl hydrocarbon receptor (AHR) allele in the defective pancreatic expression of mBD14 observed in NOD mice. By utilizing 16S rDNA gene sequencing and AHR ligand activity measurements, they demonstrated that gut microbiota-derived molecules, including AHR ligands and butyrate, promoted IL-22 secretion by pancreatic ILCs that subsequently induced mBD14 expression by endocrine cells. Therefore, dysbiotic microbiota and a low-affinity AHR allele appear to explain defective pancreatic mBD14 expression of mBD14 in NOD mice (24). Because only live gut microbiota can stimulate secretion of β-defensins, we believe that specific gut microbiota that possess special metabolic pathway functionality, including pathways for secretion of AHR ligands, may possess the ability to regulate secretion of β-defensins.

#### C-Type Lectins

The C-type lectins, also key components of innate immunity that control growth of enteric pathogens (52–54), are expressed by multiple small intestinal epithelial lineages (55, 56). REG3γ and REG3β, two C-type lectins, provide protection against infection by specific bacterial pathogens, including Enterococcus faecalis (57–59), Yersinia pseudotuberculosis (60, 61), and Listeria monocytogenes (57). Notably, additional evidence suggests that C-type lectins actually mediate syncytium endosymbiont defenses through prevention of pathogen colonization. To further demonstrate how these lectins control bacterial colonization of the intestinal epithelial surface, Vil-Myd88Tg mice (mice with IEC-restricted Myd88 expression) were used to determine whether surface Myd88 present on epithelial cells was sufficient to restrain bacterial colonization (55). The results showed that secretion of C-type lectins required both activation of the MyD88 pathway (62) and recognition of syncytium endosymbionts by TLRs (63). Furthermore, Earle et al. used a pipeline method to assess intestinal microbiota localization within immunofluorescence images of fixed gut cross-sections. The results indicated that elimination of dietary microbiotaaccessible carbohydrates (MACs) resulted in thinning of mucus within the distal colon that increased microbial proximity to the epithelium and heightened inflammatory marker REG3β expression (64). These results align with those from an earlier study of transcriptional profiles of duodenum, jejunum, ileum and colon samples, which demonstrated that MyD88 was essential for syncytium endosymbiont-induced colonic epithelial expression of antimicrobial genes Reg3β and Reg3γ, with Myd88 deficiency associated with both a shift in bacterial diversity and a greater proportion of SFB in the small intestine (65). In fact, other research found that conventionally raised Myd88−/<sup>−</sup> mice exhibited increased expression of antiviral genes in the colon, which correlated with norovirus infection of the colonic epithelium (65). Therefore, it can be concluded that both the activation of the MyD88 pathway and recognition of syncytium endosymbionts by TLRs are indispensable for triggering C-type lectins secretion (**Figure 1**). Recently, Ju et al. used antibiotictreated mice to study differences between metronidazole-treated and control groups, and observed reduced abundance of Turicibacteraceae, overgrowth of E. coli and higher levels of Reg3β and Reg3γ mRNA for the metronidazole-treated group (66). These results provide a basis for the study of the effects of specific gut syncytium endosymbiont organisms on C-type lectins secretion.

Other accumulating evidence has shown that the mammalian gut contains a rich fungal community that interacts with the immune system through the C-type lectin receptor Dectin-1. To demonstrate whether symbiotic fungi influence C-type lectins secretion that prevents pathogen colonization, Iliev et al. studied mice lacking Dectin-1 and observed increased susceptibility

to chemically induced colitis due to altered responses to indigenous fungi. Moreover, in humans they identified a gene polymorphism for Dectin-1 (CLEC7A) that is strongly linked to a severe form of ulcerative colitis (67). Independently, Eriksson et al. found that CLR-specific intracellular adhesion molecule-3 grabbing non-integrin homolog-related 3 (SIGNR3) is the closest murine homolog to the human dendritic cell-specific intercellular adhesion molecule-3-grabbing non-integrin (DC-SIGN) receptor. Both receptors recognize similar carbohydrate ligands, such as terminal fucose or high-mannose glycans. Notably, using the dextran sulfate sodium-induced colitis model, IGNR3 has been observed to recognize fungal members of the commensal microbiota, with SIGNR3−/<sup>−</sup> mice exhibiting a higher level of TNF-α in colon (68). Therefore, symbiotic fungi appear to communicate with the host via the C-type lectin receptor to maintain intestinal homeostasis. However, as yet no direct evidence has been found to determine whether symbiotic fungi can regulate selectin secretion, warranting further research.

## GUT MICROBIOTA ELICITS INFLAMMASOME ACTIVATION AGAINST PATHOGENS

Inflammasome activation is an important innate immune pathway that prevents pathogen invasion via secretion of proinflammatory cytokines IL-1β and IL-18, as well as through induction of pyroptosis (69–74). It is well-documented that inflammasomes come from two main sources, namely myeloidand epithelial-derived inflammasomes. While they share several common features, it should be noted that inflammasomes of distinct origins may exhibit different features and effector functions. For example, from a mechanistic of view, macrophageand epithelial cell-derived inflammasomes are activated with different intermediate processes. While IL-18 processing is dependent on caspase-11 in IECs, caspase-1 is responsible for the processing of IL-18 in myeloid cells (75). In addition, compared with myeloid cells, IECs constitutively express IL-18, while produce little IL-1β (76–78). Moreover, unlike myeloid inflammasomes, IEC inflammasomes is capable of producing considerable amounts of prostaglandin upon activation (79). Intriguingly, the signaling circuitry between epithelial and myeloid inflammasomes are also different. For example, in homeostasis conditions, both NLRP3 and PYCARD genes have been shown to be highly expressed in murine primary macrophages, while mouse airway epithelial cells can only express a low level of PYCARD and cannot express NLRP3 (80).

Accumulating evidence suggests that gut microbiota can activate NLRC4 and NLRP3 inflammasome pathways against pathogens (81–83). Enterobacteriaceae and the pathobiont Proteus mirabilis, which are members of the normal flora of the human gastrointestinal tract (84, 85), were shown to induce robust IL-1β production through NLRP3 activation mediated by intestinal Ly6Chigh monocytes (86, 87). Indeed, recruited Ly6Chigh monocytes have been shown to express a variety of inflammasome components, such as NAIPs (71, 88, 89), NLRC4 (89), NLRP1 (90, 91), NLRP6 (92, 93), AIM2 (94), caspase-1 (95), caspase-4 (96) (in humans), ASC (93), and IL-18 (87, 97, 98). Meanwhile, Seo et al. have also demonstrated that Proteus mirabilis (a Proteobacteria phylum member) induced NLRP3 activation and IL-1β production (86). Interestingly, bacterial components from other Proteobacteria, such as LPS produced by Pseudomonas spp., have even been shown to induce host mental depression symptoms via NLRP3 inflammasome activation (99). Other interesting lines of research have shown that in addition to gut commensal bacteria, the mammalian gut contains a rich fungal community which also appears to activate the inflammasome pathway. This community includes the human commensal fungus Candida albicans (C. albicans), which colonizes gastrointestinal and vaginal tract mucosal surfaces and appears to promote inflammasome activation during AOM-DSS-induced colitis (100). In further support of this finding, direct peptide administration experiments had previously demonstrated that candidalysin, a peptide derived from the hypha-specific ECE1 gene, acted as a fungal trigger for NLRP3 inflammasome-mediated maturation that was sufficient for inducing IL-1β secretion mature macrophages in an NLRP3 inflammasome-dependent manner (101).

In recent studies, numerous other gut microbiota metabolites have also been demonstrated to elicit inflammasome pathways against pathogens. For example, gut microbiota-derived adenosine triphosphate (ATP) has been shown to co-operate with NLRP3 (also known as CIAS1) (102) via the macrophage P2X7 receptor (103) to induce assembly of a cytosolic protein complex containing ASK and caspase-1 (70, 104–106) that eventually leads to inflammasome activation (106). Another important gut microbiota metabolite, short-chain fatty acids (SCFAs), end products of fermentation of dietary fibers by anaerobic intestinal microbiota, have also been implicated in inflammasome activation (107). SCFAs binding to GPR43 on colonic epithelial cells to stimulate K<sup>+</sup> efflux and hyperpolarization has been shown to lead to NLRP3 inflammasome activation, with subsequent acceleration of cell maturation and secretion of IL-1β (108) and IL-18 (77, 109).

## GUT MICROBIOTA CAN ENHANCE INTERLEUKIN EXPRESSION TO CLEAR INVADING PATHOGENS

#### IL-22

IL-22 is important in maintaining mucosal barrier integrity and is produced by many different types of innate immune cells (110–113). This cytokine has been shown to play a hostprotective role during infection by a wide range of pathogens, including Klebsiella pneumoniae (114), Citrobacter rodentium (115, 116), vancomycin-resistant Enterococcus (117, 118) and Plasmodium chabaudi (119). One IL-22-dependent mechanism involved in pathogen clearance involves the increased presence of antimicrobial proteins within the mucosa (120) that include the following: calprotectin and lipocalin-2, the latter of which binds to the siderophore enterochelin, with both acting to limit iron availability in the gut (120); C-type lectins, which regenerate islet-derivative proteins Reg3β and Reg3γ that control some components of the microbiota (58, 120, 121); and S100A8 and S100A9, two antimicrobial peptides that heterodimerize to form calprotectin, an antimicrobial protein that sequesters zinc and manganese to prevent microbial access to these nutrients (122). Although epithelial antimicrobial defenses also exist, many pathogens can still colonize mucosal surfaces to establish infections (120, 123). Nevertheless, accumulated evidence has shown that IL-22 is rapidly induced in response to pathogen invasion through activation of host AhR via specific gut microbiota-derived molecules (**Figure 2**) (124, 125). For example, Lactobacillus species (specifically, L. reuteri) can activate IL-22 production by gut type 3 innate lymphoid cells (ILC3) (126–128), while other studies have shown that supplementation with three commensal Lactobacillus strains with high tryptophan-metabolizing activities was sufficient to restore intestinal IL-22 production (43, 129). Indeed, additional work has shown that Lactobacillus species could utilize tryptophan as an energy source and produce a metabolite, indole-3-aldehyde (IAld), which could then activate AhRs present on ILCs (126, 130). In addition to Lactobacillus strains, other recent studies have shown that Allobaculum spp. (43), Escherichia coli (44), Clostridium spp. (45), and Bacteroides spp. (46) can also utilize tryptophan to produce IAld and elicit IL-22 production (**Table 1**). Meanwhile, other studies have shown that activated ILCs secrete IL-22 to protect the host against opportunistic pathogens by reducing pathogen colonization (120, 131). In fact, other innate immune cells, such as NKT cells, γδ T cells and macrophages, have very recently been shown to secrete IL-22 under regulation by gut microbiota via the AhR pathway (132). Therefore, gut microbiota may prevent pathogen infection by collectively enhancing IL-22 expression via the AhR pathway.

#### IL-17

IL-17 is a well-established crucial cytokine that is involved in limiting invasion and dissemination of pathogens, including Salmonella typhimurium (133), by both recruitment of neutrophils and by the induction of production of antimicrobial peptides (131, 134). Recent studies have demonstrated that both the abundance and activation status of IL-17-producing intraepithelial lymphocytes (IELs) are modulated by commensal bacteria, with enrichment of the γδT cell population of IELs representing an important source of innate IL-17 production (135, 136). Notably, a comparative study of GF mice and SPF mice has shown that the number of TCRγδ IELs is decreased in GF mice (133). Moreover, in addition to the regulation of IELs numbers, the gut microbiota may also regulate activation of TCRγδ IELs, as reflected by a report showing that production of IL-17 by TCRγδ IELs is decreased in GF mice (137). Meanwhile, antibiotic-treatment and monocolonization of mice have been used to demonstrate that the great majority of γ/δ T cells within peritonea of SPF mice are CD62L<sup>−</sup> γδT cells, which are activated γδT cells, with GF mice possessing far fewer CD62L<sup>−</sup> γ/δ T cells than SPF mice (47). Notably, additional research suggests that specific commensal bacteria, excluding metronidazole-sensitive anaerobes, such as Bacteroides species, are required for maintaining IL-1R1<sup>±</sup> γδT cells (47), a result

that aligns with previous research results by another research group (138) (**Table 1**). In conclusion, gut microbiota influences the abundance and activation status of IL-17-producing TCRγδ IELs to protect the host from pathogen infection and to maintain intestinal homeostasis. In addition, lymphoid tissue inducer (LTi) cells and NCR<sup>−</sup> ILC3 cells also appear to function as important sources of innate IL-17 production (127). However, few studies have investigated how gut microbiota regulate these cell types, warranting further research in this area.

### IL-10

IL-10 is an anti-inflammatory cytokine that plays a central role in regulating the host immune response to pathogens, thereby preventing host damage and maintaining normal tissue homeostasis (139–141). Accumulating evidence suggests that macrophages are an important source of innate IL-10 and that the gut microbiota plays a vital role in mucosal innate IL-10 generation under homeostatic conditions (142– 144). For example, studies in GF mice and SPF mice have shown that colonic lamina propria from germ-free mice exhibited lower IL-10 production (142), a reduction later confirmed to be a 50% reduction in steady-state IL-10 levels (142–144). To elucidate the mechanism by which gut microbiota regulate intestinal macrophage IL-10 production, Hayashi et al. used macrophage-specific IL-10-deficient mice to demonstrate that Clostridium butyricum (CB), a distinct cluster I Clostridium strain, induces IL-10 production to ultimately prevent acute experimental colitis. However, while CB treatment had no effects on IL-10 production by T cells, IL-10 producing F4/80±CD11b±CD11cint macrophages accumulated within inflamed mucosa after CB treatment. Subsequently, more rigorous examination demonstrated that CB directly triggered IL-10 production by intestinal macrophages there via the TLR2/MyD88 pathway (144). Meanwhile, Ochi et al. recently found that dietary amino acids directly regulate Il-10 production by small intestine (SI) macrophages. Using mice fed via total parenteral nutrition, a significant decrease of IL-10-producing macrophages in the SI was observed, while IL-10-producing CD4<sup>±</sup> T cells remained intact. Likewise, enteral nutrient deprivation selectively decreased IL-10 production by the monocyte-derived F4/80<sup>±</sup> macrophage population, but had no effect on non-monocytic precursor-derived CD103<sup>±</sup> dendritic cells. Notably, in contrast to regulation of colonic macrophages, replenishment of SI macrophages and their IL-10 production were not regulated by gut microbiota (145). Contrary to results obtained under steady-state conditions, an injury model used to study participation of microbiota to explain observed IL-10 increases post-injury yielded different results. Specifically, comparison of Il10 mRNA levels in uninjured intact tissue and day-2 post-wound tissue isolated from SPF or GF mice indicated that IL-10 mRNA was induced in post-wound colonic tissue isolated from both SPF and GF mice. Therefore, injury-triggered IL-10 increases appeared to be largely microbiota independent (146), although the reasons remain unclear regarding the differing effects of the gut microbiota observed in different model systems. Nevertheless, we hypothesize that local damage-associated molecular proteins (DAMPs) may regulate immune cells more rapidly and strongly post-intestinal damage, resulting in either a failure of gut microbiota to temporally adjust or a masking of any microbiotabased regulatory effect.

#### CONCLUDING REMARKS

Gut microbiota resists colonization and growth of invading pathogens through the induction of expression of antimicrobial peptides, IL-22, IL-17, and IL-10 while eliciting inflammasome activation. Because the underlying mechanisms of how the gut microbiota resists pathogenic invasion still remain obscure, future studies are clearly needed to identify gut microbiota functions against various pathogens toward the development of promising strategies to treat infectious diseases. For instance, E. coli Nissle 1917 can induce β-defensin expression mediated by NF-κB- and MAPK/AP-1-dependent pathways (39), while Lactobacillus spp. activate IL-22 production against opportunistic pathogens to reduce colonization (147, 148). Therefore, transplanting suitable specific gut microbiota to compete with specific pathogens could be an effective defense strategy. However, since this strategy poses new disease risks, strategies that restore

#### REFERENCES


intestinal homeostasis and promote host immune system may serve to more safely clear pathogens. To this end, identifying specific gut microbiota functions and defining normal gut microbiota populations are necessary first steps toward development of safer strategies for strengthening host defenses against pathogens. Moreover, research on the function and mechanisms of gut microbiota metabolites may facilitate development of novel therapeutic strategies to combat drug-resistant pathogens.

#### AUTHOR CONTRIBUTIONS

D-KC, W-TM, H-YC, and M-XN designed the structure of the mini-review. H-YC and M-XN wrote the manuscript and drafted the first version of the manuscript. M-XN and W-TM helped revise the manuscript. All authors have reviewed the final version of the manuscript.

### FUNDING

This research was supported by the Qinghai province Major R&D and Transformation Project (2018-NK-125), Xianyang Science and Technology Major Project (2017K01-34), Key industrial innovation chains of Shaanxi province (2018ZDCXL-NY-01-06) and PhD research startup fund of Northwest Agriculture and Forestry University (00500/Z109021716).

bifidobacteria in gut homeostasis. World J Gastroenterol. (2014) 20:15163– 76. doi: 10.3748/wjg.v20.i41.15163


and attenuated pathogenic bacteria. Infect Immun. (2005) 73:2312–20. doi: 10.1128/IAI.73.4.2312-2320.2005


by NF-kappaB and MAPK/AP-1 pathways in an in vitro model of Candida esophagitis. BMC Immunol. (2009) 10:36. doi: 10.1186/1471-2172-10-36


Reg3beta, Reg3gamma and angiogenin-4. PLoS ONE. (2012) 7:e38798. doi: 10.1371/journal.pone.0038798


vancomycin-resistant enterococcus. Sci Transl Med. (2016) 8:327ra325. doi: 10.1126/scitranslmed.aad6663


lamina propria macrophages in the small intestine. Sci Rep. (2016) 6:27634. doi: 10.1038/srep27634


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

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

# Microbiota Inhibit Epithelial Pathogen Adherence by Epigenetically Regulating C-Type Lectin Expression

Vivienne Woo1†, Emily M. Eshleman1†, Taylor Rice<sup>1</sup> , Jordan Whitt <sup>1</sup> , Bruce A. Vallance<sup>2</sup> and Theresa Alenghat <sup>1</sup> \*

*<sup>1</sup> Division of Immunobiology and Center for Inflammation and Tolerance, Cincinnati Children's Hospital Medical Center, Department of Pediatrics, University of Cincinnati College of Medicine, Cincinnati, OH, United States, <sup>2</sup> Division of Gastroenterology, Department of Pediatrics, BC Children's Hospital Research Institute, University of British Columbia, Vancouver, BC, Canada*

#### Edited by:

*Sudhanshu Shekhar, University of Oslo, Norway*

#### Reviewed by:

*Elena Verdu, McMaster University, Canada Kyoko Takahashi, Nihon University, Japan Masato Tsuda, Nihon University, Japan*

\*Correspondence:

*Theresa Alenghat theresa.alenghat@cchmc.org*

*†These authors have contributed equally to this work*

#### Specialty section:

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

Received: *02 February 2019* Accepted: *11 April 2019* Published: *07 May 2019*

#### Citation:

*Woo V, Eshleman EM, Rice T, Whitt J, Vallance BA and Alenghat T (2019) Microbiota Inhibit Epithelial Pathogen Adherence by Epigenetically Regulating C-Type Lectin Expression. Front. Immunol. 10:928. doi: 10.3389/fimmu.2019.00928* Numerous bacterial pathogens infect the mammalian host by initially associating with epithelial cells that line the intestinal lumen. Recent work has revealed that commensal bacteria that reside in the intestine promote defense against pathogenic infection, however whether the microbiota direct host pathways that alter pathogen adherence is not well-understood. Here, by comparing germ-free mice, we identify that the microbiota decrease bacterial pathogen adherence and dampen epithelial expression of the cell surface glycoprotein C-type lectin 2e (Clec2e). Functional studies revealed that overexpression of this lectin promotes adherence of intestinal bacterial pathogens to mammalian cells. Interestingly, microbiota-sensitive downregulation of Clec2e corresponds with decreased histone acetylation of the Clec2e gene in intestinal epithelial cells. Histone deacetylation and transcriptional regulation of Clec2e depends on expression and recruitment of the histone deacetylase HDAC3. Thus, commensal bacteria epigenetically instruct epithelial cells to decrease expression of a C-type lectin that promotes pathogen adherence, revealing a novel mechanism for how the microbiota promote innate defense against infection.

Keywords: microbiota, HDAC, CLEC, intestine epithelial cells, citrobacter, epigenetic

## INTRODUCTION

Infections of the gastrointestinal tract are a major cause of morbidity and mortality worldwide. Specifically, enteric infections caused by bacterial pathogens account for well over 200 million individual cases of enteritis resulting in an estimated 5 million deaths annually (1, 2). In addition to local intestinal infections, the gastrointestinal tract is the initial site of adhesion and entry for several pathogens that disseminate to cause systemic disease (3). Thus, adherence and invasion are critical steps in the pathogenesis of both enteric and systemic bacterial infections. In order to establish disease, pathogens can interact with host cells by expressing adhesin molecules which recognize various components such as extracellular matrix proteins, integral membrane adhesion receptors, and cell membrane associated glycoproteins (4). These interactions between bacterial pathogens and host cells are not only critical for initiating infection, but also direct tissue tropism, species specificity, and host susceptibility to infection (4–7). Therefore, understanding how pathogenic adherence is mediated is critical for directing effective approaches that prevent and treat enteric infections.

In addition to pathogenic bacteria, the mammalian gastrointestinal tract harbors trillions of innocuous commensal bacteria. These commensal microbes, collectively termed the microbiota, are required for healthy intestinal development and immune cell activation (8). Importantly, the presence of the microbiota has also been consistently shown to be essential for host defense against pathogenic infections (8, 9). While several mechanisms have been proposed to account for microbiota-dependent protection against infection, many pathways indicate that commensal bacteria can potentiate host-cell intrinsic defenses (10–12). Intestinal epithelial cells (IECs) reside at the direct interface between the microbiota and underlying host immune cells and are in constant contact with both beneficial as well as invading bacteria. Thus, IECs are a key cell type to which enteric pathogens often directly associate with in order to infect and invade the host. In addition to mediating binding and sensing of microbial components, these critically located cells can actively respond to pathogenic challenges by secreting antimicrobial peptides, mucins, and cytokines that prime and regulate innate and adaptive immune cell compartments (13–16). However, it is not clear whether the microbiota restrict enteric infection by regulating pathogen binding to the intestinal epithelium.

In mammalian cells, DNA is packaged around histone proteins that are condensed into a higher order structure called chromatin. In general, chromatin structure itself restricts access of transcriptional machinery to the genome thereby repressing gene expression. However, covalent modifications of the amino-terminal tails of histones, specifically, acetylation, methylation, phosphorylation, SUMOylation, and ubiquitination are associated with conformational changes in the chromatin landscape. For example, histone acetylation is known to generate an open chromatin structure that contributes to active transcription (17–19). These modifications are mediated by epigenetic modifying enzymes such as histone acetyltransferases and histone deacetylases (HDACs). The balance and pattern of these modifications on specific histone tails regulate chromatin reorganization and direct transcriptional machinery. Thus, epigenetic modifications enable environmental signals to trigger transcriptional changes without altering underlying DNA sequence (20–22).

In this study, we aimed to test whether the microbiota affect IEC-intrinsic pathways that alter the ability of pathogens to adhere to the IECs. Citrobacter rodentium, a murine enteric pathogen with a similar pathogenesis to enteropathogenic E. coli in humans, infects the host by initially adhering to IECs. By employing germ-free (GF) mice, we identified that the microbiota reduce pathogen colonization with C. rodentium during infection and instruct decreased IEC interactions with the pathogen. Global gene expression analyses revealed that the microbiota highly suppressed IEC expression of the cellsurface C-type lectin 2e (Clec2e). Interestingly, functional studies showed that overexpression of Clec2e enhanced pathogen bacterial binding to the mammalian cell membrane. Furthermore, microbiota-dependent transcriptional suppression of Clec2e in IECs correlated with decreased histone acetylation and recruitment of the histone deacetylase, HDAC3. Collectively, these data demonstrate a novel mechanism by which commensal bacteria in the intestine epigenetically regulate expression of a pathogen-binding glycoprotein to promote host defense against infection.

# MATERIALS AND METHODS

### Mice and in vivo Infections

Conventionally-housed C57Bl/6J mice were purchased from Jackson Laboratories and maintained in our specific-pathogen free colony at CCHMC. Germ-free (GF) mice were maintained in plastic isolators in the CCHMC Gnotobiotic Mouse Facility, fed autoclaved feed and water, and monitored to ensure absence of microbes. HDAC3FF mice (23) were crossed to C57Bl/6J mice expressing Cre recombinase under control of the IEC-specific villin promoter (24) to generate HDAC31IEC mice (25). Mice were housed up to 4 per cage in a ventilated cage system in a 12 h light/dark cycle, with free access to water and food. For C. rodentium infection, age- and gender- matched mice were orally inoculated with 10<sup>9</sup> colony forming units (CFUs) of C. rodentium (26, 27). To enumerate intestinal bacterial burdens, stool was collected in PBS and homogenized in a TissueLyser II at 30 Hz for 3 min. Homogenates were serially diluted and plated on MacConkey agar. CFUs were counted and normalized to stool weight after 18 h. All experiments were performed according to the animal guidelines upon approval of the Institutional Animal Care and Use Committee at CCHMC.

#### IEC Harvest, RNA Analyses, Western Blotting

IECs were harvested from mouse intestine as described previously (25, 27, 28). IECs from the small intestine were harvested from the most distal 12 cm section. RNA was isolated from cells using the RNeasy Kit (Qiagen) then subjected to reverse transcription with Verso reverse transcriptase (Thermo Fisher). Directional polyA RNA-seq for IECs from the small intestine was performed by the Sequencing Core at the University of Cincinnati (28). Sequence reads were aligned by using Illumina sequence analysis pipeline by the Laboratory for Statistical Genomics and Systems Biology at the University of Cincinnati. Real-time PCR was performed using SYBR (Applied Biosystems) and analyzed with a threshold in the linear range of amplification using primer sequences as follows: Clec2eF: 5′ -AGCAAGG TTCACAGCTCTCC-3′ ; Clec2eR: 5′ -GCTGCTATGGAGTGA TCATGG-3′ ; RegIIIγF: 5′ -TTCCTGTCCTCCATGATCAAA-3′ ; RegIIIγR: 5′ -CATCCACCTCTGTTGGGTTC-3′ ; HPRTF: 5′ - GATTAGCGATGAACCAGGT-3′ ; HPRTR: 5′ -CCTCCCATC TCCTTCATGACA-3′ . Expression analysis in IECs from large intestine of HDAC3FF and HDAC31IEC mice by microarray was described previously (25). For western blot analyses, total cell lysates were probed with anti-histone H3 (Santa Cruz) and anti-DDK (FLAG) (Origene) and imaged using an Odyssey Fc imager (LICOR). Global expression data has been deposited in NCBI's Gene Expression Omnibus (GEO) and is accessible through accession number GSE128362.

#### Woo et al. Microbiota Regulation of C-Type Lectin

#### ChIP-Sequencing

ChIP was performed as described previously with few modifications (29). Briefly, cells were fixed in 1% PFA for 10 min and quenched with glycine. Total cell extracts were sonicated using a Covaris S220 Focused-ultrasonicator and nuclear extracts were immunoprecipitated with rabbit anti-H3K9Ac (Millipore, 06-942) or rabbit anti-HDAC3 (Abcam, ab7030) using a SX-8G IP-STAR robot. Sequencing was performed using Illumina HiSeq 2500, mapped to mus musculus genome mm10 with Bowtie and peaks were identified with MACS (30, 31) and visualized in Biowardrobe (32). ChIP-qPCR was performed using SYBR (Applied Biosystems) and analyzed as fold difference normalized to an unaffected control gene. Reactions were run on a real-time PCR system (QuantStudio3; Applied Biosystems) with custom made primer pairs: Clec2e-ChIPF: 5′ -ACACAAGATGCAGCGGAGAT-3′ ; Clec2e-ChIPR: 5 ′ -GTGAAGGGGTTTTCACTAGGGG-3′ ; Insl-ChIPF: 5′ -CAG AGACCATCAGCAAGCAG-3′ ; Insl-ChIPR: 5′ -TTCTCCCTA AAGTCGCTGGA-3′ ; Albumin-ChIPF: 5′ -AGAGCGATCTTT CTGCACACA-3′ ; Albumin-ChIPR: 5′ -AGGAGAAAGGTT ACCCACTTCATTT-3′ . ChIP-seq data is accessible through GEO series accession numbers GSE50453 and GSE128369.

#### Cell Culture and Immunofluorescence

HEK293T cells were cultured in DMEM containing 10% FBS, 100 U/ml penicillin, and 100 mg/ml streptomycin at 37◦C and 5% CO2. Cells were transiently transfected with pCMV6- Clec2e-myc-DDK (FLAG) vector (Origene, MR202134) using Lipofectamine 3000 (Thermo Fisher). Transfected cells were seeded onto Retronectin (Takara Bio) coated chamber slides (Ibidi) and infected with 10<sup>6</sup> CFUs of GFP-expressing C. rodentium for 6 h in antibiotic-free media (26, 27). Cells were washed in PBS 3 times and fixed in 4% paraformaldehyde for 20 min. Fixed cells were blocked with 2% BSA for 1 h at room temperature and stained in 0.5% BSA with 488-anti-GFP (Thermo Fisher, 1:300), Phalloidin (Invitrogen, 1:200) and DAPI (Invitrogen, 1:1,000) for 1 h at room temperature. Stained cells were visualized using Nikon A1R LUN-V inverted confocal microscope.

#### Intestinal Organoids

Murine organoids were generated from colonic crypts isolated from germ-free and conventionally-housed mice as previously described (33). Briefly, dissected colons were opened longitudinally, scrapped to remove intestinal contents and outer cells, washed repeatedly in ice-cold PBS, and cut into 1-cm pieces. Colonic pieces were incubated in chelation buffer (2 mM EDTA in PBS) for 30 min at 4◦C with rotation. The tissue was transferred into a new tube containing Shaking Buffer (PBS, 43.3 mM sucrose, 54.9 mM sorbitol) and gently shaken by hand for 2–4 min. Colonic crypts were resuspended and plated in Matrigel (Corning) with organoid culture media (60% Advanced DMEM/F12 media supplemented with 10 mM HEPES, 2 mM L-glutamate, 40% L-WRN conditioned media, 1x N2 supplement, 1x B27 supplement, 50 ng/mL murine EGF, and 10µM Y-27632 ROCK inhibitor) overlaid. Culture media was changed every 3–4 days. Organoid cultures were infected with GFP-C. rodentium at a concentration of 10<sup>6</sup> CFUs for 24 h. After incubation, organoids were washed 3 times in ice-cold PBS, dislodged from plate and Matrigel, and fixed in 1% PFA for 1 h at 4 ◦C. GF organoids were stimulated with 10 ng/mL of E. coli LPS for 24 h.

#### Flow Cytometry

Cells were stained for flow cytometry using the following fluorescence-conjugated antibodies diluted in FACS Buffer (2% FBS, 0.01% Sodium Azide, PBS): PE anti-CD326 (EpCAM) (Clone: G8.8, eBioscience), BUV395 anti-CD45.2 (Clone: 104, BD Biosciences), 488 anti-GFP (Clone; FM264G, BioLegend). Dead cells were excluded with the Fixable Violet Dead Cell Stain Kit (Invitrogen). Samples were acquired on the BD LSRFortessa and analyzed with FlowJo Software (Treestar). The geometric mean fluorescence intensity (MFI) for GFP-C. rodentium expression was assessed and the background MFI determined in uninfected controls was subtracted from infected samples.

#### Bacterial Adhesion Assay

Adhesion of pathogenic bacteria to mammalian cells was determined as previously described (34). Briefly, HEK293T cells were seeded at 70–90% confluency and incubated with GFP-C. rodentium or wild-type Salmonella enterica serovar Typhimurium at a multiplicity of infection (MOI) of 5:1 (bacteria:cells) for 6 h in antibiotic-free media. Cells were washed 3 times with PBS and adherent bacteria were detached using a 1% triton-X 100 lysis buffer and plated onto MacConkey agar in 10 fold serial dilutions. Colony forming units (CFUs) were counted after 16 h.

#### Statistical Analyses

Results are expressed as mean ± SEM. Statistical significance was determined with the Student's t-test, with all data meeting the assumptions of the statistical test used. Results were considered significant at <sup>∗</sup>p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. Statistical significance was calculated using Prism version 7.0 (GraphPad Software).

# RESULTS

#### Microbiota Decrease Pathogen Adherence to Intestinal Epithelial Cells

Citrobacter rodentium (C. rodentium) is a murine bacterial pathogen with similar pathogenesis to enteropathogenic E. coli in humans. Germ-free (GF) mice infected with C. rodentium exhibited significantly higher pathogen burdens compared to conventionally-housed (CNV) mice (**Figure 1A**), indicating that the microbiota enhance defense against pathogenic colonization. Interestingly, microbiota-sensitive protection against infection was detected as early as day 4 post-infection, suggesting the presence of the microbiota affect the initial establishment of C. rodentium colonization. Intestinal epithelial cells (IECs) produce antimicrobial peptides and consistent with previous studies (35), the microbiota induced IEC expression of the antimicrobial peptide RegIIIγ that targets bacterial pathogens (**Figure 1B**) (36). In order to investigate how the microbiota induce epithelial-intrinsic defense, intestinal epithelial organoids that are devoid of immune cells were compared. Interestingly, intestinal epithelial organoids generated from CNV-housed mice expressed significantly reduced levels of RegIIIγ (**Figure 1B**). However, despite this impairment in RegIIIγ expression, organoids from CNV mice exhibited lower adherent GFP-expressing C. rodentium compared to GF organoids as measured by flow cytometry (**Figure 1C**). These data suggest that other mechanisms, aside from RegIIIγ, contribute to microbiota-sensitive IEC-intrinsic resistance against pathogenic bacterial adherence.

# Microbiota Exposure Downregulates C-Type Lectin 2e Expression in Intestinal Epithelial Cells

In order to identify potential mediators of pathogen adherence that are regulated by the microbiota, we compared genes expressed in IECs harvested from the intestine of GF and CNV mice by RNA-sequencing. These analyses identified C-type lectin 2 member e (Clec2e; Clr-a) as one of the most significantly downregulated genes in IECs in response to microbial exposure (**Figure 2A**). Clec2e expression was confirmed to be decreased by quantitative PCR (qPCR) in IECs from independent cohorts of GF and CNV mice in both the small intestine (**Figure 2B**) and large intestine (**Figure 2C**). To investigate whether microbiotasuppressed Clec2e expression was maintained in the absence of immune cells or persistent microbial stimulation, intestinal organoid cultures were generated from colonic crypts isolated from GF and CNV mice (**Figure 2D**). Consistent with IECs in vivo, Clec2e expression was repressed in organoids derived from CNV mice (**Figure 2E**), suggesting prior exposure to the microbiota led to sustained downregulation of Clec2e. To determine how the microbiota may suppress Clec2e expression, intestinal organoids derived from GF mice were incubated with LPS (**Figure 2F**). LPS reduced Clec2e expression in IECs, although less than observed in CNV organoids, suggesting that microbial-derived components may collectively regulate epithelial Clec2e expression.

# Expression of Clec2e Increases Cellular Adherence of Enteric Bacterial Pathogens

Clec2e is a homodimeric cell surface glycoprotein expressed in the intestinal epithelium that shares homology with other C-type lectins (37, 38). However, unlike other CLEC2 family members, Clec2e does not interact with Nkrp1 receptors (37, 39). C-type lectin receptor family members contain extracellular carbohydrate binding domains that associate with common pathogen-associated molecular patterns including mannose, fucose, and β-glycan (40, 41), provoking the hypothesis that Clec2e may facilitate bacterial adhesion to host cells. To test whether Clec2e plays a role in bacterial adhesion to mammalian cells, Clec2e-FLAG was overexpressed in HEK293T cells followed by incubation with either Salmonella or C. rodentium, enteric bacterial pathogens that directly bind to IECs. Expression of transfected Clec2e was confirmed by Western analyses (**Figure 3A**). Interestingly, Clec2e-expressing cells exhibited significantly greater adherence of Salmonella (**Figure 3B**) and C. rodentium (**Figure 3C**) compared to negative control cells. Bacterial adherence of GFP-expressing C. rodentium to Clec2e-expressing cells was confirmed at the cellular level by immunofluorescence (**Figure 3D**) and flow cytometry (**Figure 3E**). Together, these data indicate that Clec2e expression promotes adherence of bacterial pathogens to mammalian cells.

# Microbiota Induce Loss of Histone Acetylation and HDAC3 Recruitment Within Regulatory Regions of Clec2e

Environmental factors can regulate mammalian gene expression through epigenetic modifications of the chromatin, such as DNA methylation and histone acetylation. Consistent with this, recent studies have revealed that epigenetic pathways may be essential in mediating host-microbe dynamics (17, 42, 43). Therefore, to determine whether the microbiota epigenetically modify chromatin at the Clec2e gene, histone acetylation was compared in primary IECs harvested from GF and CNV mice. For these analyses, chromatin immunoprecipitation (ChIP)-sequencing (seq) was performed for the histone mark H3K9Ac, which is associated with permissive and actively transcribed chromatin (17). Remarkably, ChIP-seq analyses revealed that H3K9Ac levels were significantly decreased at multiple sites within the Clec2e locus in IECs isolated from CNV mice compared to GF controls (**Figure 4A**). This loss of histone acetylation in regulatory regions of Clec2e due to the microbiota was confirmed by ChIP-qPCR for H3K9Ac (**Figure 4B**). Previous studies have demonstrated that histone acetylation in IECs can be regulated by epigeneticmodifying enzymes called histone deacetylase (HDACs) (44, 45). The class I histone deacetylase 3 (HDAC3) deacetylates histone H3K9Ac and mediates microbiota-dependent regulation of epithelial gene expression (25, 27). Thus, to determine whether HDAC3 epigenetically regulates Clec2e, HDAC3 recruitment was first examined by ChIP. HDAC3 was enriched at the site of differential H3K9Ac in Clec2e (**Figure 4B**) relative to a negative non-HDAC3 target (**Figure 4C**), supporting that Clec2e is a direct target of HDAC3. Interestingly, IECs from CNV mice exhibited significantly higher HDAC3 recruitment to Clec2e compared to IECs from GF mice (**Figure 4D)**. Collectively, these data indicate that the microbiota direct epigenetic regulation of Clec2e through HDAC3.

# HDAC3 Regulates Epithelial Clec2e Expression and Pathogen Adherence

The microbiota-dependent increase in HDAC3 recruitment to Clec2e suggests that loss of H3K9Ac and decreased expression of Clec2e in response to the microbiota could be mediated by HDAC3. Thus, to directly test whether IEC-intrinsic HDAC3 regulates histone acetylation within the Clec2e gene, ChIPseq for H3K9Ac was performed in IECs harvested from mice that lack HDAC3 expression specifically in IECs (HDAC31IEC) compared to floxed HDAC3FF control mice (25). Consistent with histone deacetylation by HDAC3, IECs harvested from the large intestine of mice lacking IEC-HDAC3 (HDAC31IEC) exhibited

significantly higher levels of H3K9Ac in Clec2e compared to IECs from HDAC3FF mice (**Figure 5A**). Increased H3K9Ac enrichment within the microbiota-sensitive regulatory region in the Clec2e gene was also identified in IECs from the small intestine of HDAC31IEC mice (**Figure 5B**), indicating that Clec2e histone acetylation is regulated by epithelial HDAC3. HDAC3 recruitment is associated with transcriptional repression of bound genes. Thus, to test whether elevated H3K9Ac with HDAC3 depletion corresponds with altered expression, Clec2e mRNA expression was measured in IECs from HDAC3FF and HDAC31IEC mice. Consistent with the role of HDAC3

as a transcriptional repressor of direct targets, these analyses revealed significantly increased Clec2e expression in IECs harvested from the small and large intestine of HDAC31IEC mice compared to IECs from HDAC3FF controls (**Figure 5C**), indicating that Clec2e expression in HDAC3-deficient IECs results from impaired HDAC3-mediated deacetylation within the Clec2e gene. Collectively, these studies demonstrate that epigenetic and transcriptional regulation of Clec2e broadly depends on epithelial HDAC3 expression in the intestine.

To next test whether HDAC3-dependent regulation of IECs affects pathogen adhesion, HDAC3FF and HDAC31IEC mice

were infected with C. rodentium. Interestingly, C. rodentiuminfected HDAC31IEC mice exhibited higher pathogen burden (**Figure 5D**) and increased GFP-C. rodentium adherence to IECs (**Figure 5E**) relative to infected HDAC3FF control mice, confirming increased pathogen adhesion in HDAC31IEC mice. Further, to test this in the absence of immune cells, intestinal organoids were generated from the colon of control HDACFF mice and mice lacking HDAC3 in IECs. Consistent with the in vivo findings, HDAC31IEC organoids exhibited elevated GFP-C. rodentium binding compared to cells from floxed controls (**Figure 5F**). Taken together, these data indicate that regulation of Clec2e expression in IECs by HDAC3-mediated histone deacetylation can promote decreased bacterial pathogen adherence to the intestinal epithelium.

# DISCUSSION

The intestinal epithelium not only maintains intestinal homeostasis to innocuous commensals, but it also defends against invading pathogens (13, 46). Our data indicate that the microbiota can promote epithelial defense by epigenetically suppressing Clec2e-mediated pathogen adherence. Consistent with previously published data (10, 27), we show that GF mice are more susceptible to enteric infection relative to microbiota-replete mice. Previous studies have focused on investigating how the microbiota impact immune cell activation and antibacterial immunity (11, 47, 48), however we observed very early susceptibility to C. rodentium infection in GF mice suggesting an important role for innate responses. By employing intestinal organoid cultures, we determined that the microbiota directly impact IEC-intrinsic defense and identified that Clec2e downregulation by the microbiota can reduce pathogen colonization. The microbiota influence several aspects of IEC biology and microbiotasensitive alterations in IEC composition can also impact host resistance to enteric infection. However, Clec2e is expressed throughout the intestinal epithelium including progenitor and differentiated epithelial cells (37), suggesting

SEM. \**p* < 0.05.

FIGURE 5 | HDAC3 regulates epithelial Clec2e expression and pathogen adherence. (A) ChIP-seq for H3K9Ac in primary IECs isolated from the large intestine of HDAC3FF and HDAC31IEC mice. (B) ChIP-qPCR for H3K9Ac in Clec2e in IECs from the small intestine. (C) Clec2e expression in IECs from large and small intestine of HDAC3FF and HDAC31IEC mice. (D) *C. rodentium* CFUs in stool at day 6 post infection. (E) Fluorescence microscopy of colon from HDAC3FF and HDAC31IEC mice infected with GFP*-C. rodentium*. (Green: GFP-*C. rodentium*, Red: Phalloidin, Blue: DAPI). Scale bars, 25µm. (F) MFI of GFP*-C. rodentiu*m infected intestinal organoids derived from HDAC3FF and HDAC31IEC mice. Data are representative of at least 2 independent experiments with 3–4 mice per group. Results are mean <sup>±</sup> SEM. \**p* < 0.05.

that differences in epithelial composition induced by the microbiota are unlikely to be a main mechanism regulating Clec2e expression.

Consistent with our data, studies using GF mice and mouse models that are deficient for pattern recognition receptor mediators have revealed that mucins and antimicrobial peptides require microbial stimulation for expression (35, 49–52). These proteins work in concert to prevent intestinal infection by restricting bacterial adhesion and invasion, limiting microbial growth and colonization, and directly killing bacteria. For example, mucins function by forming a protective barrier that limits access of microbes to underlying IECs and can bind to several enteric pathogens including C. rodentium to prevent adhesion (26, 53). In addition, expression of cathelicidinrelated antimicrobial peptide by IECs plays an important and non-redundant role in preventing C. rodentium adhesion and colonization, especially in early stages of infection (54). Another member of the C-type lectin family with structural similarity to Clec2e, RegIIIγ, binds intestinal bacteria via interactions with peptidoglycan carbohydrates and directly lyses bacteria (35). These studies, combined with our Clec2e data, demonstrate that the microbiota direct multiple IEC-intrinsic host defenses that alter bacterial access to IECs and limit adhesion and colonization.

Similar to enteropathogenic E. coli and enterohemorrhagic E. coli, C. rodentium employs a type 3 secretion system and other virulence strategies to attach to the apical plasma membrane of IECs (55). Salmonella is also equipped with a type 3 secretion system and several fimbriae proteins that enable adherence and invasion to colonic IECs (3, 4, 56). Genetic deletion of type 3 secretion systems or fimbriae molecules drastically reduces bacterial colonization, confirming the necessity of these molecules for pathogenesis (26, 55, 56). Fimbriae and other filopodia-like extensions that enable bacterial adhesion often interact with host plasma membrane associated proteins. Salmonella fimbriae bind to a specific glycosylated moiety that is abundantly expressed in the intestinal epithelium (56). Interestingly, Clec2e is a heavily glycosylated protein whose expression is restricted to the intestinal epithelium and is downregulated with LPS or Poly:(IC) stimulation (37), suggesting it may play a functional role in regulating intestinal host defense. Future investigation will require GF and CNV Clec2e1IEC knockout models in combination with mono-association studies to dissect the contribution of specific commensal microbes or microbial components that influence in vivo regulation of pathogen control by Clec2e.

Through global RNA-sequencing analysis we identified a drastic reduction in the expression of the C-type lectin protein, Clec2e, in IECs isolated from CNV mice compared to GF controls. Clec2e (Clr-a) is an orphan C-type lectin molecule that closely resembles the natural killer (NK) gene complex receptor, Clec2h (Clr-f). However, unlike Clec2h, Clec2e does not bind any known NK cell receptors (37, 39, 57). In addition to being signaling partners for NK cell receptors, C-type lectin molecules play a crucial role in recognition of conserved pathogen-associated molecular patterns. Specifically, C-type lectin receptors bind carbohydrate structures commonly associated with microbial cell wall components including mannose, fucose, and β-glucans (35, 40, 41). Further, expression of Clec2e is restricted to the intestinal epithelium and is downregulated with LPS and Poly:(IC) in a TLR3-dependent manner (37). Here, we demonstrate that overexpression of Clec2e is sufficient to promote bacterial adherence to mammalian cells. While the ligand of Clec2e remains unknown, its structural similarities to Dectin-1 and RegIII microbial binding proteins, along with our bacterial adhesion data, suggests Clec2e may bind conserved microbial cell wall components or bacterial pili and fimbria necessary for cellular adherence. Although expanded studies are needed to fully interrogate how Clec2e interacts with commensal bacterial populations and pathogens beyond Salmonella and C. rodentium, a lack of Clec2e suppression may contribute to heightened susceptibility of GF or antibiotic-treated mice to infection (48).

Epigenetic modifications enable host cells to alter gene expression without modifying the genetic sequence and changes in the host epigenome occur downstream of external environmental signals (20–22). Recent studies, focused predominantly on immune cells types, have supported that the microbiota may imprint or epigenetically prime genes in the host through enzymes such as DNA methyltransferases (58), histone methyltransferases (29, 59), and HDACs (25, 60–63). In addition to HDAC3, other class I HDACs are expressed in IECs (44, 45) and the role of these HDACs as well as other regulatory proteins may further alter epigenetic regulation of the Clec2e gene in IECs. IECs are equipped to sense and respond to common microbial moieties such as LPS, and previous studies showed TLR4 expression in IECs was epigenetically primed by the microbiota (64). Specifically, DNA methylation of TLR4 was decreased in GF mice compared to CNV controls, consistent with reduced TLR4 expression with microbial exposure (64). Histone acetylation is associated with permissive and actively transcribed chromatin. Our data using intestinal organoids revealed that the microbiota mediate durable changes in IECs that remain even after microbial stimulation has been removed. In the presence of the microbiota, H3K9Ac was reduced in the Clec2e gene which directly corresponds with reduced expression of Clec2e in IECs from CNV mice, indicating that the microbiota epigenetically regulate Clec2e expression. This study further demonstrates that H3K9Ac in Clec2e is regulated by HDAC3 as loss of HDAC3 expression leads to increased histone acetylation and loss of transcriptional repression of Clec2e. Taken together, our data demonstrate a novel mechanism by which the microbiota promote host defense through suppression of IEC-intrinsic pathways that are coopted for pathogen adherence and highlights that epigenetic regulation of innate pathways in IECs may represent a potent, long-lasting mechanism by which the microbiota prime host defense.

# DATA AVAILABILITY

The datasets generated for this study can be found in Gene Expression Omnibus (GEO), GSE128362, GSE50453, GSE128369.

## ETHICS STATEMENT

All experiments were performed according to the animal guidelines upon approval of the Institutional Animal Care and Use Committee at CCHMC.

# AUTHOR CONTRIBUTIONS

TA, VW, and EE designed the studies and analyzed the data. VW, EE, TR, and JW carried out experiments. BV provided bacterial strains. TA, VW, and EE wrote the manuscript.

### FUNDING

This research is supported by the National Institutes of Health (DK093784, DK114123, DK116868 to TA), a Crohn's and Colitis

# REFERENCES


Foundation/Janssen/AGA award to TA, and the Arnold W. Strauss Fellowship to EE. TA holds a Career Award for Medical Scientists from the Burroughs Wellcome Fund and is a Pew Scholar in the Biomedical Sciences, supported by the Pew Charitable Trust. BV holds the CH.I.L.D. Foundation Chair in Pediatric Gastroenterology. This project is supported in part by PHS grant P30 DK078392 and the CCHMC Trustee Award and Procter Scholar's Program.

#### ACKNOWLEDGMENTS

We thank the members of the Alenghat lab for critical reading of the manuscript. We also thank CCHMC Veterinary Services, Research Flow Cytometry Core, Confocal Imaging Core and the University of Cincinnati Genomics, Epigenomics and Sequencing Core for services and technical assistance.


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

Copyright © 2019 Woo, Eshleman, Rice, Whitt, Vallance and Alenghat. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Commensal Bacteria: An Emerging Player in Defense Against Respiratory Pathogens

Rabia Khan\*, Fernanda Cristina Petersen and Sudhanshu Shekhar\*

*Faculty of Dentistry, Institute of Oral Biology, University of Oslo, Oslo, Norway*

A diverse community of trillions of commensal bacteria inhabits mucosal and epidermal surfaces in humans and plays an important role in defense against pathogens, including respiratory pathogens. Commensal bacteria act on the host's immune system to induce protective responses that prevent colonization and invasion by pathogens. On the other hand, these bacteria can directly inhibit the growth of respiratory pathogens by producing antimicrobial products/signals and competing for nutrients and adhesion sites. Such mechanisms preserve the niche for commensal bacteria and support the host in containing respiratory infections. Herein, we discuss current evidence on the role of commensal bacteria in conferring protection against respiratory pathogens and the underlying mechanisms by which these bacteria do so. A deeper knowledge of how commensal bacteria interact with the host and pathogens might provide new insights that are poised to aid in the development of vaccines and therapeutics that target infectious diseases.

Keywords: host, commensal, pathogen, lungs, vaccine

# INTRODUCTION

Since the inception of the Human Microbiome Project in 2007, a plethora of knowledge has accumulated that throws light on diverse and crucial roles played by commensal bacteria in homeostasis and disease (1, 2). With the help of advances in omic and systems biology technologies, our knowledge of the composition, genetics, and functional capacity of commensal bacteria is growing at a fast pace. It is becoming clear that commensal bacteria, which reside in various parts of the human body, such as the gut and airways, correspond approximately to the total number of human cells (about 1:1 ratio), and exert a profound impact on regulation of immunophysiological functions, including but not limited to, metabolism, ontogeny, and pathogen defense (3, 4). Several recent studies have shown that commensals promote resistance to gut pathogens that is mutually beneficial to the host and the commensal microbiota (5–7). However, imbalances in the microbial communities can occur, and are linked to many diseases, such as inflammatory bowel disease, allergies, asthma, diabetes, and obesity (8). It remains scantily understood how these bacteria execute their functional activities against respiratory pathogens. Respiratory infections are of utmost importance, as they inflict substantial social and economic burden on people across the world in general and in low and lower-middle income countries in particular (9–11). Additionally, current therapeutic and prophylactic interventions against respiratory diseases have major constraints, such as rapid emergence of anti-microbial resistance and disruption of the normal microbiota by use of antibiotics. Unraveling the interactions between commensals and pathogens may allow the exploitation of inhibitory properties of commensals

#### Edited by:

*Laurel L. Lenz, University of Colorado Denver, United States*

#### Reviewed by:

*Theresa Alenghat, Cincinnati Children's Hospital Medical Center, United States Aude Remot, Institut National de la Recherche Agronomique (INRA), France*

#### \*Correspondence:

*Rabia Khan rabia\_naeemkhan@hotmail.com Sudhanshu Shekhar sudhanshu.shekhar@odont.uio.no*

#### Specialty section:

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

Received: *01 February 2019* Accepted: *13 May 2019* Published: *31 May 2019*

#### Citation:

*Khan R, Petersen FC and Shekhar S (2019) Commensal Bacteria: An Emerging Player in Defense Against Respiratory Pathogens. Front. Immunol. 10:1203. doi: 10.3389/fimmu.2019.01203*

**128**

to combat pathogens causing respiratory diseases. In this review article, we provide an overview of the current state of knowledge about the role of commensal bacteria in protective immunity to respiratory pathogens and the mechanisms involved in commensal bacteria mediated defenses. Understanding the relationship between commensal bacteria, host, and pathogen is a way forward to develop safe and effective prophylactics and therapeutics against pathogens.

## COMMENSAL BACTERIA MEDIATED PROTECTION

#### Protection in Mouse Models

Much of what is known about the direct role of commensal bacteria in protection against respiratory pathogens stems from studies using various mouse models, including germ-free and antibiotic-treated mice (12–23). Following lung infection with Streptococcus pneumoniae, numerous studies have shown that antibiotic-treated mice display significantly higher pathogen loads and increased pathologies in the lungs compared with sham-treated mice (15, 21). Likewise, germ-free mice showed enhanced levels of bacterial burden when subjected to S. pneumoniae and Klebsiella pneumoniae lung infections (15). Similar to the protection conferred by fecal microbiota transplant (FMT) against intestinal pathogens, FMT in gut microbiotadepleted mice restored pulmonary bacterial clearance early after S. pneumoniae infection (21). In case of mouse models of Mycobacterium tuberculosis infection, gut microbiota disruption after pre- and post-antibiotic treatment showed decreased resistance to infection in the lungs, which was associated with severe histopathological changes, such as pulmonary granulomas (24). Furthermore, antibiotic-induced dysbiosis changed the microbiota diversity in the gut and promoted lung colonization by M. tuberculosis (25). Similar protective effect was conferred after mice having antibiotic-induced disrupted microbiota received an intranasal infectious dose of influenza A virus (12). In a different study, FMT into germ-free mice led to reconstitution of the gut microbiota that facilitated increased survival against lethal influenza A virus infection (26). Overall, these studies employing multiple experimental approaches provide in vivo evidence that underscores a profound contribution of commensal bacteria in defense against diverse respiratory pathogens.

Even though antibiotic-treated and germ-free mice have proven to be a crucial tool in understanding the role of the microbiota in pathogen defense, there are potential pitfalls that need consideration while interpreting results from studies involving these animals. Germ-free animals lack all detectable microbes in different organs and have an impaired immune system, whereas antibiotics are used to deplete specific microbiota (27). Although these two approaches provide crucial information on the function of the microbiota in general, the specific contribution of the microbiota found in different body compartments, such as the lung microbiota, in immunity to respiratory infections is unclear. This is important because the lung microbiota, which in healthy adults seems to mainly consist of a small number of bacteria originating from the oral cavity, plays a significant role in respiratory health, and disease (28–30). Additionally, there is a need for models that can answer a more direct question about protection in the presence of a fully developed immune system. To address this issue, researchers have attempted to deplete the lung microbiota in mice by a combination of aerosolized vancomycin and neomycin via nasal route, which resulted in a significant reduction in the lung commensal microbiota, with the advantage of minimally affecting the gut microbiota (31, 32). But the possibility of antibiotic spread to the nearby tissues/organs harboring different microbiota remains, requiring future studies to focus on developing better models to fill in this pitfall.

The microbiota consists of a large number of bacterial species, and therefore, it is of great interest to specifically identify commensal species that protect from respiratory pathogens. Recent studies have evaluated the protective efficacy of commensal bacterial species in respiratory infections. Oral administration of Bifidobacterium longum (BB536), but not saline, in mice significantly reduced viral loads, pulmonary pathology, and body weight loss following intranasal challenge with influenza virus, suggesting a protective role for this commensal bacterium in influenza infection (33). Similarly, oral or nasal inoculation of mice with different strains of Lactobacillus, e.g., L. gasseri (TMC0356), L. rhamnosus (CRL 1505), and L. brevis (KB290), conferred protective immunity to influenza virus infections (34–36). Furthermore, L. rhamnosus (CRL 1505) exerted a protective effect in mice subjected to an intranasal challenge with respiratory syncytial virus infection (37, 38). These data indicate a prophylactic role for commensal bacteria against viral pathogens. In order to assess therapeutic significance, B. longum (MM2) was orally administered in mice infected with influenza virus. Mice that received B. longum (MM2) ameliorated infection, as determined by decreased body weight loss, viral titers, and inflammation, compared with control mice (39). Protective effect induced by these probiotic bacteria is not confined to respiratory infections with viruses, but can be applicable to bacterial pathogens (37–40). Intake of B. longum (51A) in mice not only demonstrated protective effect against infection with K. pneumoniae, but also suppressed inflammatory changes in the lung (40). Very recently, we have demonstrated that intranasal immunization of mice with the commensal Streptococcus mitis conferred protection against lung infection with S. pneumoniae strains D39 (serotype 2) and TIGR4 (serotype 4), which illuminates the unique ability of S. mitis to offer resistance to different pneumococcal serotypes (41). Two recent studies performed by independent groups further show that the gut colonizer bacterium Helicobacter hepaticus influences the composition of the gut microbiota and the outcome of M. tuberculosis infection in mouse models (42, 43). Mice subjected to intestinal colonization with H. hepaticus, when challenged by intranasal route with M. tuberculosis, reflected higher mycobacterial burden in the lungs compared with the controls (42). This increased mycobacterial burden in the H. hepaticuscolonized mice coincided with severe M. tuberculosis-mediated pulmonary pathologies, mainly characterized by granulomas and tissue damage, and accumulation/production of inflammatory leukocytes/cytokines (42). Similar to these results, mice colonized with H. hepaticus eliminated subunit-vaccine-induced protective immunity to lung infection with M. tuberculosis (43). Taken together, these findings indicate that commensal bacteria can be harnessed for prophylactic and therapeutic purposes, provided utmost precaution on the possible negative effects of enriching for specific colonizers of the microbiota.

### Protection in Humans

Relatively little information is available on whether commensal bacteria can prevent respiratory infections in humans. Oral commensals, such as Streptococcus oralis and Streptococcus salivarius, can induce protection against middle ear inflammation, referred to as otitis media, which is primarily caused by respiratory pathogens, such as S. pneumoniae and Haemophilus influenzae (44–46). Upon intranasal administration of S. salivarius and S. oralis, children susceptible to acute otitis media displayed reduced recurrences of disease with no side effects (44). Contrarily, a nasal spray containing oral commensals, e.g., S. mitis and S. oralis, in susceptible children under 4 years of age showed no significant effect regarding the number of episodes of recurrent otitis media compared to the placebo group (46). The discrepancy in these studies might be due to differences in bacterial doses, inoculation regimens, and combinations, which need to be analyzed in light of new technologies (e.g., metagenomics and next generation sequencing) and concepts like dysbiosis. The fact that antibiotics were used together with the streptococcal nasal spray in the first study, but not in the second, is also an important factor to consider. In controlled infection studies in humans, nasopharyngeal colonization by the commensal Neisseria lactamica provided protection against the respiratory pathogen Neisseria meningitidis (47, 48). Furthermore, in a block-randomized challenge trial, 310 healthy individuals (18–25 years) were intranasally inoculated with live N. lactamica or sham and the bacterial carriage was monitored for 26 weeks (48). All those who developed nasopharyngeal colonization by N. lactamica revealed a significant reduction in the N. meningitidis carriage compared with sham-treated ones (48). These studies show that commensal bacteria not only show inhibitory/displacing effects on the carriage of respiratory pathogens but also highlight the ease and safety with which these bacteria can be used to contain infections in humans. It is however notable that most bacteria with high pathogenic potential, such as those in the above examples, are also part of the healthy microbiome (49). The reason as to why these pathogens cause diseases is attributed to various host and microbial factors, including viral infections (49). Dysbiosis in particular, such as a result of antibiotic use has been associated with a reduction in the prevalence of respiratory commensal bacteria like Corynebacterium spp. and Dolosigranulum spp. in the nasopharynx of healthy infants. These are considered to reduce the colonization by S. pneumoniae, H. influenzae, and S. aureus in the respiratory tract (50). It is further shown that respiratory syncytial virus infection in children below 2 years of age was positively correlated with nasopharyngeal H. influenzae and Streptococcus microbiota clusters and inversely correlated with Staphylococcus aureus (51). Transcriptomic analysis of the children infected with H. influenzae and Streptococcus clusters presented greater expression of immune components, suggesting that nasopharyngeal microbiota can influence host immunity (51). In line with this, prolonged antibiotic treatment in early life has also been annexed with an increased risk for respiratory infections in infants (52, 53). Thus, these studies shine light on the effect of the microbiota perturbations caused by antibiotics on host susceptibility to respiratory infections, particularly during the critical life period of immune maturation.

## MECHANISMS OF COMMENSAL BACTERIA MEDIATED PROTECTION

A pertinent question however remains as to what are the underlying mechanisms by which commensal bacteria perform their protective function against respiratory pathogens. Emerging data thus far indicate that commensal bacteria confer protection in two ways: host-mediated immunity (acting on the host's immune system) and direct action (directly inhibiting/killing pathogens and competing for colonization).

# Host-Mediated Immunity

A wealth of emerging evidence indicates that both the lung and gut microbiota are involved in the regulation of immune responses during lung infections (28, 54). However, it is difficult to assess the specific contributions of the lung and gut microbiota to protective immunity to respiratory pathogens, mainly due to three reasons: (1) the gut microbiota is the largest and most diverse community of commensals that significantly influences the outcome of immunity in the lungs as well as gut; (2) the gut is the largest lymphoid organ in the body because of which it occupies a central position in host-microbiota studies; and (3) we do not have optimal models to ascertain their specific roles in immunity. Despite an important role for these commensal bacteria in promoting resistance against respiratory pathogens, the mechanistic basis for this resistance remains unclear. Several studies have shown a potential defect in innate immunity and subsequent adaptive immunity in the lung, when signals from commensal bacteria are abrogated (12, 14, 25, 55). Key innate immune cells that are recruited to the lungs and are involved in protective immunity include macrophages, natural killer (NK) cells, and mucosal-associated invariant T (MAIT) cells (12, 14, 25, 55). The pulmonary macrophages in mice depleted of the microbiota by antibiotics reflected reduced expression of the macrophage-associated antiviral genes, such as Irf 7, Ifnb, Mx1, Tnfa, Il6, and Il1b following influenza virus infection (14). This corresponded with reduced response to IFN-γ, IFN-α, or influenza infection in macrophages from the mice treated with antibiotics. In vivo experiments in mice also indicated that the alveolar macrophage response was impaired during viral infection, which was characterized by down-regulation of most of the antiviral genes activated ex vivo (14). Wang et al. demonstrated a new mechanism in mice colonized with S. aureus where CD11b<sup>+</sup> M2 alveolar macrophages, stimulated with Toll-like receptor (TLR) 2, play a protective role in influenza infection (56). Another innate immune cell type is MAIT cell that is shown to play an important role in microbiota mediated mycobacterial immunity (25). Flow cytometric analysis reflected that mice depleted of the microbiota had reduced number of lung MAIT cells, characterized by MR1-5-OP-RU tetramer+TCRβ + phenotype, which expressed significantly lower IL-17A compared with control mice, suggesting that lung MAIT cells may function to contain early pulmonary M. tuberculosis infection (25). Furthermore, NK cells from germ-free mice did not induce anti-influenza immunity because macrophages and dendritic cells failed to produce type 1 IFN in response to infection (55). Cumulatively, these data suggest that microbiota-derived signals provide a stimulus that maintains the potency of the lung innate immune system needed for invoking effective immunity (**Figure 1**).

Adaptive immunity follows innate immunity and is crucial for specific immunity against respiratory pathogens (57–59). Rabbit antisera raised against S. mitis show cross-reactivity with S. pneumoniae (59). Similar to IgG mediated cross-reactivity, IgA antibodies from the sera, nasal wash, and bronchoalveolar lavage of mice vaccinated with S. mitis cross-reacted with S. pneumoniae serotypes 2 and 4 (41). On the other hand, human CD4<sup>+</sup> T cells expressing IL-17A, which are reactive to S. mitis, show cross-reactivity toward S. pneumoniae in an in vitro culture system (60). Intranasal vaccination of mice with S. mitis led to an increased production of IL-17A by CD4<sup>+</sup> T cells in the lungs compared to PBS-treated control mice (41). These results are in line with the evidence that the gut commensal segmented filamentous bacteria (SFB) regulate pulmonary Th17 immunity to the fungal pathogen Aspergillus fumigatus (19). In a mouse model of influenza viral infection, it is shown that commensal bacteria regulate virus-specific CD4<sup>+</sup> and CD8<sup>+</sup> T cells and antibodies following lung infection with virus (12). Overall, commensal bacteria mediated adaptive immunity to respiratory pathogens include both humoral (IgG and IgA) and T cellmediated responses.

Involvement of the gut microbiota in protective immunity to pulmonary pathogens illustrates a pathogenic nexus between the microbiota and the "gut-lung axis," underscoring a profound protective influence of the gut commensals over several pathogens residing at distant anatomical compartments of the body (61). The gut microbiota mediated control of the lung immunopathology is also evident from studies demonstrating the susceptibility of animals with the altered gut microbiota to allergic lung diseases (20). On the other hand, dysbiosis in the lung microbiota can change the composition of the gut microbiota. For example, mice subjected to lung infection with influenza virus infection or intranasal instillation of lipopolysaccharide disturbed the gut microbiota homeostasis, which supports the fact that the gut and lungs are closely linked in a way that they affect each other's microbiology and physiology (62, 63). Moreover, how the gut microbiota controls the lung immunity has recently been explored by few key studies in mouse models of pulmonary bacterial infections. Brown et al. performed a well-designed and comprehensive study that sheds light on a major innate immune mechanism used by the microbiota to clear lung infections in mouse models (15). In antibiotic-treated mice, there was an increased growth of S. pneumoniae and K. pneumoniae in the lungs after bacterial inoculation compared to sham-treated mice, which was associated with reduced production of innate immune factors, such granulocyte–macrophage colony-stimulating factor (GM-CSF) (15). In vivo neutralization of GM-CSF into antibiotictreated mice, which received the microbiota from the shamtreated mice and had restored pulmonary bacterial clearance, resulted in making these mice prone to infections (15). These findings suggest that GM-CSF is essential for the microbiota to execute their functional activities against both Gram-positive (S. pneumoniae) and Gram-negative (K. pneumoniae) pathogens. It was further demonstrated that GM-CSF programs alveolar macrophage function via an extracellular signal-regulated kinase (ERK)-specific signaling pathway leading to increased pathogen killing via reactive oxygen species (ROS) (15). Several studies have implicated pattern recognition receptor (PRR) ligands produced by the gut microbiota in controlling immune responses outside the intestinal tract (16, 64). Following antibioticmediated depletion of the microbiota in mice, early clearance of K. pneumoniae was impaired and this could be rescued by injection of bacterial Nod-like receptor (NLR) ligands (the NOD1 ligand MurNAcTri(DAP) and NOD2 ligand muramyl dipeptide [MDP]), but not bacterial TLR ligands (16). Defects in early innate immunity were found to be due to reduced ROS-mediated killing of bacteria by alveolar macrophages (16). Interestingly, upon treatment of mice with antibiotics and NLR ligands orally prior to S. pneumoniae lung infection, neutralization of GM-CSF abrogated the rescue of respiratory clearance (15). Taking account of all these data, it is clear that the microbiota and NLR ligands regulate lung innate immunity to respiratory pathogens via GM-CSF, highlighting crucial mechanisms of the gut-lung axis of communication. In addition, the gut commensal SFB has been reported to provide protection in immunocompromised (Rag−/−) mice by partially enhancing neutrophil resolution during pneumococcal lung infection, which corresponded with reduced expression of the anti-phagocytic molecule CD47 (65). Like NK cells, another lymphoid cell population referred to as group 3 lymphoid cells that produce IL-22 (IL-22+ILC3), a cytokine involved in pathogen defense, has been implicated in gut commensal bacteria-induced protection against S. pneumoniae (66). Disruption of commensal bacteria by antibiotics decreased the influx of IL-22+ILC3 cells into the lungs of new born mice, which made them more prone to pneumococcal infection compared with control mice. This effect was reversed when ILC3 cells were adoptively transferred or exogenous IL-22 administered in mice (66). Thus, these immune mechanisms furnish crucial information on how the gut microbiota controls protective immunity to lung infections (**Figure 1**).

#### Direct Action

Commensal bacteria resist colonization of pathogens by using wide range of direct mechanisms for niche competition, such as secretion of inhibitory substances and nutrient competition, enlisting the exploitative, and interference modes of competition (67). Recent studies dissected novel mechanisms used by resident commensals to inhibit and contain respiratory pathogens, such as

ligands, such as nod-like receptor ligands (NOD RL), leading to the activation of Mφ. NK and MAIT cells when activated by commensal bacteria produce large

quantities of IL-17A, whereas ILC3 cells secrete IL-22, aiding in inhibition/killing of various respiratory pathogens.

disruption of biofilms, exploitation of host resources to generate antimicrobial products, and down-regulation of virulence genes. This highlights the complexity and diversity of mechanisms involved in direct inhibition (68–70). The well-documented mechanism by which commensal bacteria can directly inhibit the pathogen growth and compete with them is the production of ribosomally produced antimicrobials called bacteriocins (71, 72). For example, S. salivarius produces a wide range of bacteriocins, which is a major mechanism that antagonizes S. pneumoniae (71, 73–75). More recently, it is also demonstrated that S. salivarius reduces the S. pneumoniae colonization by blocking the adhesion sites, suggesting multiple mechanisms used by this commensal to inhibit pathogens (76). Apart from ribosomally encoded bacteriocins, commensal bacteria encode non-ribosomally produced bioactive antimicrobials to compete with pathogens (77). Zipperer et al. showed that the nasal commensal Staphylococcus lugdunensis directly inhibits the growth of S. aureus through a novel cyclic antimicrobial peptide named "Lugdunin." Lugdunin possessed bactericidal activity against all tested strains of S. aureus in vitro. Moreover, in animal model, the co-colonization of S. lugdunensis and S. aureus resulted in competitive exclusion of S. aureus (77). The use of purified antimicrobials or bacterial strains encoding antimicrobials may serve as a source of new generation of antibiotics to deal with multidrug resistant strains, such as methicillin resistant S. aureus. One mechanism, which contributes to competitive advantage for colonization of commensal bacteria to preserve their niche and to suppress the growth of pathogens, includes the production of hydrogen peroxide. Epidemiological data show a negative correlation between S. pneumoniae and S. aureus and presumably, the reason for increased S. aureus related otitis media after use of pneumococcal vaccine (78, 79). One possible mechanism implicated to define this negative association is hydrogen peroxide mediated inhibition of S. aureus by pneumococcal hydrogen peroxide (80).

Nutrient competition is also a strategy used by commensal bacteria to reduce the fitness of pathogens by competing for the same pool of resources (81). Stubbendieck et al. recently showed that isolates of Corynebacterium spp. inhibited Staphylococcus spp. in vitro. This inhibition was due to reduced iron bioavailability, mediated by siderophore–induced sequestration of iron by Corynebacterium spp. (82). Another novel mechanism of commensal mediated inhibition is through the production of secreted enzymes. Iwase et al. first demonstrated the negative correlation between the commensal Staphylococcus epidermidis and pathogenic S. aureus in human nasal samples. To gain further insight to explain this negative association, they identified the inhibitory factor produced by S. epidermidis as serine protease, which inhibits the biofilm formation and


TABLE 1 | Examples of direct mechanisms of colonization resistance used by commensal bacteria against respiratory pathogens.

human nasal colonization by S. aureus (68). Follow-up study from the same group showed that intranasal colonization of mice with serine protease producing S. epidermidis inhibited colonization with methicillin resistant S. aureus (83). Commensal bacteria also exploit the host resources to generate metabolic compounds with antimicrobial properties to suppress the growth of respiratory pathogens. An elegant study by Bomar et al. investigated the mechanistic explanation for correlation between increased abundance of Corynebacterium species and reduced S. pneumoniae colonization (69). Interestingly, they found that Corynebacterium accolens encodes lipase, which catalyzes the hydrolysis of host triacylglycerolsto to produce free fatty acids with antibacterial properties that suppress the growth of S. pneumoniae (69). Taken together, the above examples evidently suggest that antagonistic interactions exist in the polymicrobial community utilizing wide range of mechanisms by which commensal bacteria inhibit respiratory pathogens. Advanced understanding of existing mechanisms using both in vitro and in vivo models and further elucidation of novel mechanisms may enable us to exploit commensals to inhibit respiratory pathogens. Mechanisms used by commensal bacteria to directly inhibit/kill respiratory pathogens are exemplified in **Table 1**.

#### CONCLUSIONS AND FUTURE INSIGHTS

Advanced research technologies have been applied to evaluate the contribution of commensal bacteria to respiratory infections. Accumulating evidence indicates an important role for commensal bacteria in defense against respiratory pathogens, which paves the way to target these bacteria for the development of vaccines and therapeutics that provide optimal protection with safety and low cost. Moreover, the use of modern experimental tools to decipher the novel mechanisms used by

#### therapeutics with targeted approach focusing exclusively on the pathogen inhibition without disrupting the homeostatic microbial community. Future studies are required to address the following questions: (1) What are the underlying mechanisms by which the trio of commensals, pathogens, and host interact with each other? (2) What could be the long-term consequences of using commensal bacteria-based vaccines/therapeutics on the host, pathogens, and the microbiota? (3) What are the effects of medical manipulations, such as antibiotics and probiotics, on the biology of commensal bacteria? (4) How can we use commensal bacteria-expressed bacteriocins for protection against respiratory pathogens? (5) Which specific commensal bacterial species of the microbiota are directly involved in protection immunity to different pathogens? (6) How can we use commensal microbiota/bacteria to correct dysbiosis? A sincere exploration of these questions may have implications for the clinical use of commensal bacteria with inhibitory properties against pathogens. This may be important to bypass the drawbacks associated with currently available options, such as antimicrobial resistance.

commensals to inhibit pathogens may assist in designing novel

#### AUTHOR CONTRIBUTIONS

RK, FP, and SS wrote and revised the manuscript. All authors assisted in the conception of this review and acquisition of relevant literature. All authors gave approval of the last version to be published.

#### FUNDING

This work was supported by a grant from the Norwegian Research Council (Grant number - 241011).

#### REFERENCES


Biol. (2016) 14:e1002533. doi: 10.1371/journal.pbio.10 02533


Staphylococcus aureus in healthy children. Lancet. (2004) 363:1871–2. doi: 10.1016/S0140-6736(04)16357-5


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

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

# Heterogeneous Vancomycin-Intermediate Staphylococcus aureus Uses the VraSR Regulatory System to Modulate Autophagy for Increased Intracellular Survival in Macrophage-Like Cell Line RAW264.7

#### Edited by:

Sudhanshu Shekhar, University of Oslo, Norway

#### Reviewed by:

Lynne Prince, University of Sheffield, United Kingdom George Sakoulas, University of California, San Diego, United States Roberta Olmo Pinheiro, Oswaldo Cruz Foundation (Fiocruz), Brazil

#### \*Correspondence:

Xiaoling Ma maxiaoling@ustc.edu.cn; xiaolingma@126.com Jiabin Li lijiabin@ahmu.edu.cn

#### Specialty section:

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

Received: 03 February 2019 Accepted: 16 May 2019 Published: 31 May 2019

#### Citation:

Dai Y, Gao C, Chen L, Chang W, Yu W, Ma X and Li J (2019) Heterogeneous Vancomycin-Intermediate Staphylococcus aureus Uses the VraSR Regulatory System to Modulate Autophagy for Increased Intracellular Survival in Macrophage-Like Cell Line RAW264.7. Front. Microbiol. 10:1222. doi: 10.3389/fmicb.2019.01222 Yuanyuan Dai1,2, Caihong Gao1,2, Li Chen<sup>2</sup> , Wenjiao Chang<sup>2</sup> , Wenwei Yu<sup>2</sup> , Xiaoling Ma<sup>2</sup> \* and Jiabin Li1,3,4 \*

<sup>1</sup> Department of Infectious Diseases, First Affiliated Hospital of Anhui Medical University, Hefei, China, <sup>2</sup> Department of Clinical Laboratory, First Affiliated Hospital of University of Science and Technology of China, Hefei, China, <sup>3</sup> Anhui Center for Surveillance of Bacterial Resistance, Hefei, China, <sup>4</sup> Department of Infectious Diseases, Chaohu Hospital of Anhui Medical University, Hefei, China

The VraSR two-component system is a vancomycin resistance-associated sensor/regulator that is upregulated in vancomycin-intermediate Staphylococcus aureus (VISA) and heterogeneous VISA (hVISA) strains. VISA/hVISA show reduced susceptibility to vancomycin and an increased ability to evade host immune responses, resulting in enhanced clinical persistence. However, the underlying mechanism remains unclear. Recent studies have reported that S. aureus strains have developed some strategies to survive within the host cell by using autophagy processes. In this study, we confirmed that clinical isolates with high vraR expression showed increased survival in murine macrophage-like RAW264.7 cells. We constructed isogenic vraSR deletion strain Mu31vraSR and vraSR-complemented strain Mu31vraSR-C to ascertain whether S. aureus uses the VraSR system to modulate autophagy for increasing intracellular survival in RAW264.7. Overall, the survival of Mu31vraSR in RAW264.7 cells was reduced at all infection time points compared with that of the Mu3 wild-type strain. Mu31vraSR-infected RAW264.7 cells also showed decreased transcription of autophagy-related genes Becn1 and Atg5, decreased LC3-II turnover and increased p62 degradation, and fewer visible punctate LC3 structures. In addition, we found that inhibition of autophagic flux significantly increased the survival of Mu31vraSR in RAW264.7 cells. Together, these results demonstrate that S. aureus uses the VraSR system to modulate host-cell autophagy processes for increasing its own survival within macrophages. Our study provides novel insights into the impact of VraSR on bacterial infection and will help to further elucidate the relationship between bacteria and the host immune response. Moreover, understanding the autophagic pathway in vraSR associated immunity has potentially important implications for preventing or treating VISA/hVISA infection.

Keywords: Staphylococcus aureus, autophagy, VraSR, regulation, macrophage

#### INTRODUCTION

fmicb-10-01222 May 31, 2019 Time: 10:3 # 2

Staphylococcus aureus is an important human pathogen responsible for both hospital-associated and communityacquired infections (Monaco et al., 2017). S. aureus causes a wide range of diseases, from minor skin and soft tissue infections to severe systemic illnesses such as pneumonia, arthritis, endocarditis, and bacteremia (Lowy, 1998). In recent years, indiscriminate and extensive use of vancomycin has resulted in the emergence and development of vancomycin-intermediate S. aureus (VISA) and heterogeneous VISA (hVISA) strains, many of which show an increased ability to evade host immune surveillance and enhanced clinical persistence (Gardete et al., 2012; Cameron et al., 2017; Katayama et al., 2017).

Autophagy is a basic physiological process in eukaryotes and plays an important role in cellular repair and homeostasis (Ohsumi, 2014). Degraded intracellular components are removed and recycled into newly emerging double-membrane vacuoles called autophagosomes (Mizushima et al., 2010). These autophagosomes mature to fuse with lysosomes and are digested. This process of autophagosome formation and eventual degradation is termed autophagic flux (Mizushima et al., 2010). Studies have demonstrated that many pathogens have evolved strategies to harness autophagic processes for survival inside the host cell (Campoy and Colombo, 2009; Mostowy, 2013; Gomes and Dikic, 2014; Mostowy, 2014; Soong et al., 2015; Siqueira et al., 2018). In particular, Neumann et al. and Schnaith et al. reported that S. aureus can take advantage of the autophagic mechanism to aid in its own replication or intracellular survival (Schnaith et al., 2007; Neumann et al., 2016).

The two-component regulatory system VraSR is a vancomycin resistance-associated sensor (VraS)/regulator (VraR) that is highly expressed in VISA/hVISA strains. In this study, we found that clinical isolates with high vraR expression showed increased survival in murine macrophage-like RAW264.7 cells and deletion of vraSR in S. aureus resulted in decreased survival in RAW264.7, indicating that VraSR could enhance intracellular S. aureus survival. We hypothesize that the VraSR regulatory system could be involved in regulation of host autophagy pathways to promote the survival of S. aureus. To examine this hypothesis, the autophagic responses of RAW264.7 cells infected with hVISA reference strain Mu3, vraSR deletion mutant Mu31vraSR, or vraSR-complemented strain Mu31vraSR-C were investigated. The results showed that S. aureus uses the VraSR regulatory system to induce autophagy and inhibit autophagic flux, thereby increasing bacterial intracellular survival in RAW264.7. This finding provides novel insights into the impact of VraSR on bacterial infection and may help to further elucidate the relationship between bacteria and the host immune response.

#### MATERIALS AND METHODS

#### Bacterial Strains, Plasmids, and Culture Conditions

The bacterial strains and plasmids used in this study are listed in **Supplementary Table S1**. All S. aureus strains were cultured with shaking (200 rpm) at 37◦C in tryptic soy broth (TSB). Escherichia coli strains were cultured with shaking (200 rpm) in Luria-Bertani medium at 37◦C. The culture media were supplemented with appropriate antibiotics when required (ampicillin, 100 µg/l; chloramphenicol, 10 µg/l; and anhydrotetracycline, 1 µg/ml).

#### Construction of the vraSR Mutant Strain

The vraSR deletion mutant strain was constructed as described previously (Li et al., 2017). Briefly, the upstream and downstream fragments of vraSR were amplified from S. aureus Mu3 genomic DNA using the vraSR-UF/vraSR-UR and vraSR-DF/vraSR-DR primer sets, respectively, and ligated by overlap extension polymerase chain reaction (PCR) to form an updown fragment. The resulting fragment was recombined into the temperature-sensitive shuttle plasmid pKOR1 using Gateway <sup>R</sup> BP ClonaseTM II Enzyme Mix (Thermo Fisher Scientific) to generate recombinant plasmid pKOR1-vraSR. pKOR1-vraSR was then transformed into S. aureusstrain RN4220 by electroporation for modification and then transformed into S. aureus strain Mu3. The mutant strains that had allelic replacement were screened via high temperature and anhydrotetracycline-resistant and chloramphenicol-sensitive colonies and were further confirmed by PCR, quantitative reverse-transcriptase PCR (qRT-PCR) and sequencing. All primers used in this study are listed in **Supplementary Table S2**.

## Complementation of the vraSR Deletion Strain

To generate a complementation strain, vraSR and its promoter region were amplified and cloned into the shuttle plasmid pLI50, producing recombinant plasmid pLI50-vraSR. The recombinant plasmid was then transferred into E. coli DH5α and DC10B successively, and finally electroporated into the S. aureus Mu31vraSR strain. Successful uptake of the complementation plasmid was confirmed by restriction mapping, PCR, and sequencing of PCR fragments. The presence of vraSR transcripts within the transformants was verified by qRT-PCR analysis.

#### Growth Curve Analysis

S. aureus strains were incubated overnight in 5 ml of TSB at 37◦C with shaking at 200 rpm. The overnight cultures were diluted 1/100 in 30 ml of fresh TSB and incubated at 37◦C with shaking at 220-rpm. The optical densities (OD600 nm) of the S. aureus cultures were then monitored at 1 h intervals for a total of 18 h.

## Cell Culture

fmicb-10-01222 May 31, 2019 Time: 10:3 # 3

Murine macrophage-like cell line RAW264.7 was cultured in Dulbecco's modified Eagle's medium (DMEM high glucose; HyClone) supplemented with 10% (v/v) fetal calf serum (FCS; HyClone). Cells were cultured in a humidified incubator containing 5% CO<sup>2</sup> at 37◦C.

## Assessment of Bacterial Intracellular Survival

Intracellular killing assays were performed as described previously (Munzenmayer et al., 2016). For S. aureus infection, early stationary phase bacteria (OD600 nm = 1.0–1.5) were harvested and washed once in cold phosphate-buffered saline (PBS). RAW264.7 cells were then infected with S. aureus at a multiplicity of infection (MOI) of 10. Following incubation for 1 h, infected cells were washed three times with PBS before the addition of 10% (v/v) FCS-DMEM supplemented with 10 µg/ml lysostaphin (Sigma-Aldrich) and 100 µg/ml gentamicin (Sigma-Aldrich) to each well. Plates were then incubated for 1 h to kill extracellular bacteria. Following incubation, the cells were washed with PBS and further incubated in fresh 10% (v/v) FCS-DMEM. At 0, 3, 6, 12, and 24 h post-infection (hpi), infected cells were washed three times with PBS to remove extracellular bacteria and dead cells and lysed by the addition of 0.5% (v/v) Triton X-100 (Sigma-Aldrich). The number of intracellular bacteria (expressed as colony-forming units, CFU) was determined by serial dilution and plating on TSB agar. In addition, replicate plates were incubated with 1.25 mM 3-methyladenine (3-MA, Sigma) or 100 nM bafilomycin A1 (Baf A1, Sigma) for 2 h prior to infection to block autophagy in the RAW264.7 cells.

#### Transmission Electron Microscopy

RAW264.7 cells were incubated with the individual S. aureus strains at a MOI of 50 for 3 h before being collected by centrifugation and fixed with 2.5% glutaraldehyde in 0.1 M phosphate buffer. Cells were then post-fixed in 1% osmium tetroxide and dehydrated through a series of graded acetone washes. The dehydrated cells were embedded in epoxy resin, sectioned, and stained with uranyl acetate and lead citrate in preparation for observation under a transmission electron microscope (HT7700; Hitachi Co.). Autophagosomes were counted as described previously. (Yla-Anttila et al., 2009).

#### RNA Extraction and qRT-PCR Assays

S. aureus RNA was extracted as described previously (Dai et al., 2017) to examine levels of vraR expression. RAW264.7 cells (2.5 × 10<sup>6</sup> cells/well) in 6-well plates were infected with S. aureus at a MOI of 50 and incubated at 37◦C with 5% CO2. At 0, 1.5, 3, 4.5, and 6 hpi, the culture medium was removed and RAW264.7 cells were washed twice with ice-cold PBS. Total RNA was extracted using Trizol reagent (Invitrogen) according to the manufacturer's instructions, and RNA quantity and quality were measured using a NanoDrop ND-1000 spectrophotometer (PeqLab). RNA was reverse-transcribed into cDNA using PrimeScript Reverse Transcriptase as per the manufacturer's instructions (TaKaRa).

qRT-PCR assays were performed using aliquots of cDNA and SYBR Premix EX TaqTM II (Takara) in an ABI 7500 qPCR instrument (Foster). Primers used for expression analysis are described in **Supplementary Table S1**. All gene expression was normalized against that of Actb (β-actin) and the 16S rRNA gene was used as an internal control. All qRT-PCR assays were repeated three times.

### Western Blot Analysis

RAW264.7 (2.5 × 10<sup>6</sup> cells/well) were infected with S. aureus at a MOI of 50 in 6-well plates for 3 h and then collected and washed twice with ice-cold PBS. The total protein from the cells was extracted using a RIPA lysis buffer solution (Wuhan Boster), with the total protein concentration determined using a bicinchoninic acid protein assay kit (Thermo Fisher Scientific). Cell lysis solutions were separated using sodium dodecyl sulfatepolyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membranes (Millipore). The membranes were blocked in 5% skim milk in tris-buffered saline Tween 20 buffer for 2 h at room temperature and then incubated overnight at 4◦C with primary antibodies against LC3-I/II and p62/SQSTM1 (1:1000, Cell Signaling Technology). The membranes were then incubated with a 1:5000 dilution of horseradish peroxidase-conjugated secondary antibody for 1 h at room temperature. Densitometric analysis of the western blot was performed using Image Gauge software (Fujifilm). β-actin was used as an internal control, and the ratio of the intensity of the protein of interest to β-actin was calculated.

## mRFP-GFP-LC3 Puncta Formation Assays

RAW264.7 cells were plated in 6-well plates and allowed to reach 70% confluence by the time of transfection. The cells were then infected with mRFP-GFP-LC3 adenoviral vector (HanBio) at a MOI of 100 for 36 h, cultured with S. aureus (MOI = 50) for 3 h, and then fixed in 4% (w/v) polyoxymethylene for 30 min at 37◦C. Images were captured on a Zeiss LSM 800 laser scanning confocal microscope (Carl Zeiss AG) using a 63× oil objective lens, followed by image analysis using Zeiss ZEN acquisition software. For quantification of LC3 punctate, mRFP-GFP-LC3 and mRFP-LC3 punctate dots were counted in 100 cells. Punctate dots were measured using Image J software (NIH).

# Statistical Analysis

All statistical analyses were performed using SPSS (Version 16.0; SPSS for Windows). When appropriate, one-way analysis of variance (ANOVA) and unpaired t-tests were analyzed to determine statistical significance, with a P-value cutoff of <0.05 to establish significance. Where appropriate, Bonferroni posttests were performed to directly compare experimental means.

# RESULTS

fmicb-10-01222 May 31, 2019 Time: 10:3 # 4

# The VraSR Regulatory System Is Important for Bacterial Intracellular Survival and Proliferation

In our previous study, a total of 24 clinical S. aureus isolates were screened on brain heart infusion agar containing 3 mg/l vancomycin (BHI-V3) (Chang et al., 2015). Based on the resulting data, we selected 12 isolates that grew on BHI-V3 (GV3) were selected as the test group, and other 12 isolates that did not grow on BHI-V3 (NGV3) for examination in the current study as the test and control groups, respectively. The relative ability of the two groups of clinical strains to persist intracellularly was quantified using a lysostaphin/gentamicin protection assay in RAW264.7 cells. We observed that the GV3 group showed increased survival rates in RAW264.7 cells at 24 hpi compared with the NGV3 group (**Figures 1A,B**). Therefore, to further investigate the impact of vraR expression on S. aureus intracellular growth, we constructed isogenic vraSR deletion strain Mu31vraSR and vraSR-complemented strain Mu31vraSR-C. We examined the mRNA expression and protein levels of VraR in the three S. aureus strains by q PCR analysis and western blotting to confirm the successful construction of the mutant strain (**Figures 1C,D**). To rule out the potential influence of the vraSR deletion on S. aureus growth rate, the growth of strains Mu3, Mu31vraSR, and Mu31vraSR-C was monitored hourly for 18 h. Overall, there were no substaintial differences in the growth of the three strains.

For S. aureus infection assay, we selected a low MOI (MOI = 10) instead of a high MOI (MOI = 50), because a high bacterial burden may cause earlier death of RAW264.7 cells, preventing intracellular replication of S. aureus. Our results showed that, in comparison with wild-type strain Mu3, fewer Mu31vraSR cells were present within the infected RAW264.7 cells, and that the survival of Mu31vraSR continued to decrease throughout the infection process. However, complemented mutant Mu31vraSR-C showed similar survival rates to the wild-type strain Mu3 in RAW264.7 cells, indicating that VraSR is important for intracellular S. aureus survival and proliferation (**Figure 1E**).

### The VraSR Regulatory System Contributes to the Formation of Autophagic Vesicles in S. aureus-Infected Cells

Ultrastructural features of infected cells were examined by transmission electron microscopy. Double-membraned structures, characteristic of autophagosomes, containing undigested S. aureus cells were observed within the RAW264.7 cells (**Figure 2A**). Quantification of S. aureus-containing autophagosome-like vacuoles per 50 infected cells showed that Mu31vraSR-infected cells displayed significantly fewer S. aureus-containing autophagosome-like vacuoles compared with cells infected with Mu3 or Mu31vraSR-C (P < 0.01), indicating that the VraSR regulatory system contributes to the formation of S. aureus-containing autophagosome-like vacuoles in S. aureus-infected cells (**Figure 2B**).

# The VraSR Regulatory System Contributes to the Expression of Autophagy-Related Genes

As autophagy is a dynamic process, we performed a series of biochemical assays to examine the activation of autophagy at the molecular level (Levine et al., 2011). qRT-PCR assays were used to determine the transcriptional levels of autophagyrelated genes Ulk1, Becn1, and Atg5. As shown in **Figure 3**, the transcriptional levels of the three genes were gradually upregulated between 0 and 3 hpi but declined after 4.5 hpi. Compared with Mu3-infected cells, the Mu31vraSR-infected cells showed significantly lower transcriptional levels of Becn1 and Atg5 at 3 hpi, while gene expression in the Mu31vraSR-C-infected cells was comparable with that in Mu3-infected cells, indicating that the VraSR regulatory system contributes to increase expression of autophagy-related genes.

# The VraSR Regulatory System Participates in Modulation of Autophagic Protein LC3 Turnover and p62 Degradation

LC3-II is a commonly used marker for autophagosome formation (Barth et al., 2010). Western blot analysis revealed that the levels of LC3-II in the Mu31vraSR-infected cells were significantly lower than those in Mu3- and Mu31vraSR-C-infected cells at 3 hpi (**Figures 4A,B**), indicating that the VraSR regulatory system is important for autophagic activity. Levels of other autophagic substrates that are degraded by autolysosomes, such as p62, can be used to monitor autophagic flux. We therefore monitored levels of p62 degradation by autolysosomes. At 3 hpi, the amount of p62 in Mu31vraSR-infected cells was significantly lower than that in Mu3- and Mu31vraSRC-infected cells (**Figures 4C,D**), indicating that the VraSR regulatory system may inhibit autophagic flux in RAW264.7.

## The VraSR Regulatory System Is Involved in Autophagic Flux Inhibition in S. aureus Infected Cells

Autophagic flux was morphologically traced to distinguish between the two stages. mRFP-GFP-LC3 can be used as an indicator of autophagic flux because it appears yellow (mRFP and GFP signals merged) in autophagosomes and red (only mRFP signals) in autolysosomes, as a result of quenching of the GFP signal by the acidic pH of the lysosomes, while the RFP signal remains stable at an acidic pH. In this study, RAW264.7 cells stably expressing mRFP-GFP-LC3 were infected with different S. aureus strains. All autophagic structures could be measured by confocal microscope. Results showed that Mu31vraSRinfected cells had fewer punctate LC3 structures (yellow dots) at 3 hpi compared with Mu3-infected cells, while the Mu31vraSR-C-infected cells appeared similar to the Mu3-infected cells (**Figures 5A,B**), indicating that the VraSR regulatory system

posttest (P > 0.05). (E) Comparison of the levels of VraR expression in strains Mu3, Mu31vraSR, and Mu31vraSR-C by western blot. Gels are representative of three independent experiments (n = 3). (F) Survival of intracellular S. aureus were determined at each time point. The X-axis represents hours post-infection and the Y-axis represents log<sup>10</sup> CFU/ml S. aureus. The data are presented as the mean ± SD. The SD was calculated from experiments performed in triplicate (n = 3). Statistical significance was determined by one-way ANOVA with Bonferroni posttest (P < 0.01). <sup>∗</sup>P < 0.05; ∗∗∗P < 0.01.

promotes the formation of autophagosomes. Moreover, we observed that Mu31vraSR-infected cells had more single, red LC3 puncta compared with the other two groups, indicating that the VraSR regulatory system could block or delay the fusion of the autophagosome with the lysosome.

To further confirm the effect of the VraSR regulatory system on the autophagic pathway and to investigate whether the autophagic pathway can target intracellular S. aureus, RAW264.7 cells were pretreated with 3-MA (a well-established phosphatidylinositol 3-kinase (PI3K) inhibitor that prevents the induction of autophagy) or Baf A1 (a selective inhibitor of vacuolar H+-ATPase that prevents lysosomal acidification) for 2 h prior to infection with S. aureus. And we examined the effects of 3-MA and Baf A1 on cells viability by Cell Counting Kit-8 (CCK-8) assay. RAW264.7 cells viability was not affected at 12 h treatment with 3-MA or Baf A1 at concentrations up to

examined by transmission electron microscopy (TEM) under 3000× magnification. Uninfected (I) RAW264.7 cells were compared with RAW264.7 cells infected with Mu3 (II), Mu31vraSR (III), or Mu31vraSR-C (IV) for 3 h. Boxed areas from I, II, III, and IV are magnified from the respective main images (white arrow, intracellular S. aureus; black arrow, autophagosome-like vacuole). (B) Number of autophagosome-like vacuoles observed by TEM. The data are presented as the mean ± SD. The SD was calculated from experiments performed in triplicate (n = 3). Statistical significance was determined by a one-way ANOVA with Bonferroni posttest. ∗∗∗P < 0.01.

1.25 mM or 100 nM, respectively (**Supplementary Figure S1**). And the results of growth curve assay showed that 3-MA and Baf A1 had no effect on the growth of S. aureus (**Supplementary Figure S2**). We observed that 3-MA treatment decreased the levels of LC3-II (**Figure 4C**) and significantly decreased the intracellular survival of Mu3 and Mu31vraSR-C in RAW264.7 cells (**Figure 4D**). In comparison, Baf A1 treatment significantly increased the accumulation of autophagic substrate p62 protein in all infected groups (**Figure 4C**), suggesting that autophagic flux was decreased in infected cells. We also determined that the intracellular survival of Mu3 and Mu31vraSR-C was significantly decreased following pretreatment 3-MA while the survival of Mu31vraSR inside cells pretreated with Baf A1 was significantly increased (**Figure 4D**). These results further demonstrated that the autophagy pathway of RAW264.7 cells is inhibited by 3- MA, which decreased the intracellular survival of S. aureus. Baf A1 inhibited autophagic flux and enhanced the survival of Mu31vraSR. Taken together, these results indicate that S. aureus uses the VraSR regulatory system to block autophagic flux for increasing intracellular survival.

# DISCUSSION

While autophagy is a cytosolic catabolic process in eukaryotic cells, it is also an innate defense mechanism against invading pathogenic bacteria (Gong et al., 2012). However, recent reports have shown that autophagy may play different roles during the infection of different bacterial pathogens, in addition to its known involvement in bacterial clearance, coordinating autonomous cell signaling, and, in some cases, promoting bacterial replication (Mostowy, 2013). Although autophagosomes are reported to be intracellular niches for S. aureus, the underlying mechanisms by which S. aureus triggers the autophagy machinery were poorly understood (Schnaith et al., 2007; Neumann et al., 2016). In the current study, we showed that S. aureus uses the VraSR regulatory system to enhance intracellular survival and increase the number of autophagic vesicles in S. aureus-infected cells, indicating that S. aureus uses the VraSR regulatory system to modulate autophagy in RAW264.7 cells.

The formation of the autophagosome involves the assembly of 36 autophagy-related (ATG) proteins into complexes

that are essential for different steps of the process: the ULK1 complex triggers autophagy, the beclin 1 and class III phosphatidylinositol 3-OH kinase (PI3KC3) complex generates an essential lipid component of autophagosomes, and the ATG12-ATG5-ATG16L1 ubiquitin-like conjugation system mediates formation and elongation of the autophagosome. As autophagy is a dynamic process, we selected Ulk1, Becn1, and Atg5 as markers of the three stages of autophagy, qRT-PCR analysis was then performed to examine the mRNA levels of each of the three genes. The transcript levels of Becn1 and Atg5 were significantly decreased in Mu31vraSR-infected cells, indicating that the VraSR regulatory system contributes to increase the expression of autophagy-related genes. At the same time, we observed that the VraSR regulatory system participates in modulation of autophagic protein LC3 turnover and P62 degradation. Confocal microscopy-based morphological analyses revealed that the VraSR regulatory system plays an important role in S. aureus-induced autophagosome maturation and inhibition of autolysosome formation. Therefore, our findings suggest that the VraSR regulatory system positively contributes to bacterial survival in RAW264.7 cells in two ways: first, hVISA strains use the VraSR regulatory system to promote autophagy, thereby recruiting LC3 protein to develop the autophagosomes; and second, hVISA strains use the VraSR regulatory system to interfere with autophagosome and lysosome fusion, thereby enhancing the intracellular survival S. aureus.

According to the literature, various bacterial species exploit autophagy and promote the formation of autophagic vacuoles in which to multiply by regulating their cell wall components or virulence factors (Campoy and Colombo, 2009; Irving et al., 2014; Castrejon-Jimenez et al., 2015). For example, peptidoglycan is one of the most components of the Listeria monocytogenes cell wall. As such, peptidoglycan and its cleavage products are recognized by peptidoglycan-recognition protein, which induces autophagy (Yano et al., 2008). Intracellular L. monocytogenes then hijacks autophagy in macrophages by secreting the virulence factor listeriolysin O to evade killing (Zhang et al., 2019). In S. aureus strains, two-component system VraSR can upregulate the synthesis of peptidoglycan, resulting in an increase in D-alanine-D-alanine residues (Kuroda et al., 2003). However, whether VraSR can induce autophagy via thickening of the peptidoglycan layer and increasing its cleavage products as well as L. monocytogenes needs further verification.

being infected with S. aureus for 3 h. Cells were then lysed to examine LC3 and p62 protein levels by western blot. β-actin are used as the loading control. Gels were representative of three independent experiments (n = 3). (D) Quantification of intracellular bacteria in RAW264.7 cells infected with S. aureus for 6 h in the presence or absence of 3-MA or Baf A1. The X-axis represents different strains and the Y-axis represents log<sup>10</sup> CFU/ml S. aureus. The data are presented as the mean ± SD. The SD was calculated from experiments performed in triplicate (n = 3). Statistical significance was determined by one-way ANOVA with Bonferroni posttest. ∗∗∗P < 0.01.

Our previous study demonstrated the capacity of VraSR to modulate S. aureus virulence by binding the P2–P3 intergenic region of the agr promoter (Dai et al., 2017). The Agr quorum-sensing system is a key regulatory system in Staphylococci, controlling the expression of a number of virulence factors. It is also essential for S. aureus survival within macrophages (Kubica et al., 2008). Some reports have also suggested that factors regulated by Agr are required for an autophagic response to S. aureus infection (Schnaith et al., 2007; O'Keeffe et al., 2015; Soong et al., 2015). Therefore, we speculate that the mechanism via which the VraSR regulatory system regulates autophagy may be related to the expression of Agr, which we plan to investigate in future.

In this work, we determined that S. aureus uses the VraSR regulatory system to block autophagic flux and delay the early stage of autophagosome formation, thereby promoting bacterial survival inside RAW264.7 cells. Although this study is limited by the use of only a single murine cell lineage and one S. aureus strain, to our knowledge, this is first report showing that the VraSR two-component system is responsible for the onset of autophagy in eukaryotic cells. Our findings provide new insights into the impact of VraSR on bacterial infection and will help to further elucidate the relationship between bacteria and host immune response. Importantly, our results suggest that VraSR may be a potential target for preventing or treating VISA/hVISA infection.

#### AUTHOR CONTRIBUTIONS

fmicb-10-01222 May 31, 2019 Time: 10:3 # 9

YD, XM, and JL contributed conception and design of the study. YD organized the database. CG and WC performed the statistical analysis. YD wrote the first draft of the manuscript. LC, WC, WY, and XM wrote sections of the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.

#### FUNDING

This work was supported by grants from the National Natural Science Foundation of China (Grant Numbers 81772248 and 81673242) and the Anhui Provincial Natural Science Foundation (Grant Number 1908085MH264).

#### ACKNOWLEDGMENTS

We thank Prof. Xinhua Liu, Department of Medical Microbiology, Anhui Institute of Innovative Drugs, Anhui Medical University, China, for gifting us the

#### REFERENCES


cell line RAW264.7. We also thank Tamsin Sheen, Ph.D., from Liwen Bianji, Edanz Editing China (www.liwenbianji.cn/ac), for editing the English text of a draft of this manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.01222/full#supplementary-material

FIGURE S1 | CCK8 analysis of cell viability of RAW264.7 cells treated with different concentrations of 3-MA (A) or Baf A1 (B) for 12 h. 3-MA did not show cytotoxicity at concentrations of up to 1.25 mM and Baf A1 did not show cytotoxicity at concentrations of up to 100 nM.

FIGURE S2 | Growth curves of Mu3 (A), Mu31vraSR (B), and Mu31vraSR-C (C) cultured in presence or absence of 3-MA or Baf A1. Bacteria were grown in TSB at 37◦C with shaking at 200 rpm. Values of OD<sup>600</sup> nm were measured every 1 h. Values are from three biological replicates ± SEM. Statistical significance was determined by a one-way ANOVA with Bonferroni posttest (P > 0.05).

TABLE S1 | Strains an plasmids used in this study.

TABLE S2 | Primers used in this study.



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

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

# The Commensal Microbiota and Viral Infection: A Comprehensive Review

Na Li <sup>1</sup> , Wen-Tao Ma<sup>1</sup> , Ming Pang<sup>1</sup> , Qin-Lei Fan<sup>2</sup> and Jin-Lian Hua<sup>1</sup> \*

*<sup>1</sup> College of Veterinary Medicine, Northwest A & F University, Yangling, China, <sup>2</sup> Animal Health and Epidemiology Center, Qingdao, China*

The human body is inhabited by a diverse microbial community that is collectively coined as commensal microbiota. Recent research has greatly advanced our understanding of how the commensal microbiota affects host health. Among the various kinds of pathogenic infections of the host, viral infections constitute one of the most serious public health problems worldwide. During the infection process, viruses may have substantial and intimate interactions with the commensal microbiota. A plethora of evidence suggests that the commensal microbiota regulates and is in turn regulated by invading viruses through diverse mechanisms, thereby having stimulatory or suppressive roles in viral infections. Furthermore, the integrity of the commensal microbiota can be disturbed by invading viruses, causing dysbiosis in the host and further influencing virus infectivity. In the present article, we discuss current insights into the regulation of viral infection by the commensal microbiota. We also draw attention to the disruption of microbiota homeostasis by several viruses.

#### Edited by:

*Fernanda Cristina Petersen, University of Oslo, Norway*

#### Reviewed by:

*Shashank Gupta, Brown University, United States Kyu Han Lee, Johns Hopkins University, United States*

> \*Correspondence: *Jin-Lian Hua jinlianhua@nwafu.edu.cn*

#### Specialty section:

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

Received: *15 December 2018* Accepted: *21 June 2019* Published: *04 July 2019*

#### Citation:

*Li N, Ma W-T, Pang M, Fan Q-L and Hua J-L (2019) The Commensal Microbiota and Viral Infection: A Comprehensive Review. Front. Immunol. 10:1551. doi: 10.3389/fimmu.2019.01551* Keywords: commensal microbiota, germ-free, antibiotics, virus, virus infectivity, antiviral immunity

# INTRODUCTION

Emerging data suggest that the human body is inhabited by a wide range of microorganisms that are collectively referred to as the commensal microbiota. A majority of the microbiota reside in the intestine, while distinct populations can also be found on the surfaces of the mouth, skin, and urinary tract (1–3). A wealth of evidence suggests that this incredibly diverse microbial community is regulated by host genetic factors, and more importantly, environmental and dietary factors (4–6). We now know that the coevolution of the commensal microbiota and their hosts has resulted in a mutually beneficial condition in which the host can benefit from physiological, metabolic, and immunological regulations provided by the microbiota, while the commensal microbiota depends absolutely on the host for nutrient acquisition and propagation sites (7). For example, the gut microbiota has a crucial role in shaping immune development and functionality in the host, as reflected by extensive defects in the development of gut-associated lymphoid tissues, significantly smaller and fewer mesenteric lymph nodes and Peyer's patches, reduced secretory immunoglobulin A (IgA) production, and abnormal intestinal T cell development in germ-free (GF) mice (8–10). In addition, the gut microbiota helps the host break down dietary substances that are too large to be digested, a process that produces critical nutrients and energy for the host and generates active products (i.e., short-chain fatty acids, lactic acid, choline, and bile acids) that are essential for host health (11, 12).

The recent awareness of the essential role of the commensal microbiota in host health has remarkably improved our understanding of the interactions between microbiota and invading pathogens. In fact, a healthy commensal microbiota, as well as its products, is essential for protecting the host against a variety of pathogenic infections, through both direct elimination and indirect suppression, inside or outside of the gastrointestinal tract (13–16). Among the invading pathogens, viruses constitute one of the most common. During their infection processes, various viruses encounter the commensal microbiota of the hosts, making it possible that there are robust interactions between these viruses and the commensal microbiota. Indeed, a plethora of evidence has now shown that the commensal microbiota regulates and is inevitably regulated by invading viruses through a series of mechanisms, thereby yielding harmful or beneficial outcomes for the host (17–19). In the regulation of viral infection, commensal microbiota play varied and critical roles. They can promote viral infectivity through diverse mechanisms and can also exert substantial inhibitory effects on viral infection. On the other hand, a viral infection usually results in substantial perturbations in the commensal microbiota, causing dysbiosis in the host, which may in turn further affect viral infectivity. Although there have been several excellent reviews summarizing the modulation of viral infections by the commensal microbiota (17, 19–21), most of them focused only on infections by enteric viruses. In addition, none of these articles discussed commensal microbiota at sites other than the gut. Moreover, there have not been any literatures describing the effect of viral infections on the compositional and functional alterations of the commensal microbiota. In the present article, we systematically discuss the current progress concerning the modulation of various types of viral infections by the commensal microbiota. We also highlight the relevant mechanisms underlying these observations. In addition, we further describe the disruption of microbiota composition or homeostasis by viral infections and the relevant mechanisms.

# PROMOTION OF VIRAL INFECTION BY THE GUT MICROBIOTA (SUMMARIZED IN TABLE 1)

# Direct Promotion of Viral Infection

#### Facilitating Genetic Recombination

The commensal microbiota can facilitate genetic recombination of viruses to enhance their infectivity. This is true for the poliovirus infection. Several studies have demonstrated that RNA viruses such as poliovirus benefit from the delivery of various viral genomes into a single target cell, thereby allowing the recombination of multiple viral genomes, and this process potentiates the viral progeny with enhanced environmental fitness (36–38). Using polioviruses encoding either DsRed or GFP and HeLa cells as target cells, Erickson et al. found that preincubation of these viruses with certain commensal microbiota significantly increased the percentage of DsRed and GFP dual-positive cells compared with those preincubated with PBS (22). Mechanistically, bacterial adhesion to HeLa cells was the main promoting force for viral coinfection. Importantly, by employing two types of viruses that are either sensitive to the drug guanidine hydrochloride while resistant to high temperature (DrugSTemp<sup>R</sup> ) or resistant to guanidine hydrochloride while sensitive to high temperature (DrugRTemp<sup>S</sup> ), the authors found that preincubation of viruses with bacteria increased recombination yields significantly, as reflected by the generation of polioviruses with a DrugRTemp<sup>R</sup> phenotype, and that the recombination frequencies were positively correlated with coinfection frequencies (22). Collectively, these data suggest that interactions of the commensal microbiota and poliovirus prior to infection increases the possibility that a cell will be infected by two or more viruses, which further facilitates genome recombination of the viruses, thereby generating progenies with more diverse populations and with increased resistance to otherwise restrictive conditions.

#### Enhancing Virion Stability

In addition to facilitating genetic recombination, bacterial surface polysaccharides, i.e., peptidoglycan and lipopolysaccharide (LPS), can enhance virion stability through several mechanisms, which have been demonstrated mostly for poliovirus and reovirus. For example, gut microbiota depletion with antibiotics prior to poliovirus infection results in less susceptibility of mice and minimal viral replication in the intestine (23). Notably, when orally inoculated poliovirus was isolated from the lumen contents of untreated, antibiotic-treated, and germ-free mice, significantly higher infectivity was identified for poliovirus isolated from untreated mice. In addition, at temperatures above 40◦C, markedly increased poliovirus stability was identified when they were preincubated with untreated feces or feces from germ-free mice that had been supplemented with certain bacteria. Importantly, the enhancement of viral stability did not necessarily require live bacteria, as UV-inactivated bacteria, as well as bacterial surface polysaccharides (LPS and peptidoglycan), significantly increased the viral yield over PBS when incubated with poliovirus. Furthermore, using a poliovirus mutant with reduced LPS-binding capacity, which was generated by a single amino acid substitution in the viral capsid protein VP1-T99K, the same group of authors found that while the mutant viruses showed similar replication, attachment, shedding, and pathogenesis with wild-type viruses following peroral inoculation, they displayed poorer environmental stability compared to their wild-type counterparts, as highlighted by the findings that mutant viruses were more unstable in feces and that an additional cycle of infection in mice aggravated this instability (24).

Consistent with the findings shown in poliovirus infection, another enteric virus, reovirus, also uses commensal microbiota or bacterial components to enhance thermostability (25). Similar to poliovirus, the pathogenesis of reovirus is also negatively affected by antibiotic treatment prior to infection (23). Mechanistically, the direct interaction of reovirus virions with Gram-negative and Gram-positive bacteria promotes the attachment to and infection of target cells at a variety of temperatures (23). It should be noted that commensal bacteria do not affect the overall number of viral capsid proteins, indicating that the bacterial effect on the reovirus is not exerted through regulating the overall number of viral capsid proteins (23). Collectively, these findings highlight the notion that interactions with commensal microbiota can increase the infectivity of viruses by enhancing virion stability.


#### Stimulating Lytic Reactivation

The lytic stages during viral infection involve viral gene expression, viral DNA replication and the production of new virions, making this stage indispensable for transmission and persistence of viruses (39). The direct promotion of viral infectivity by the commensal microbiota is also reflected by the stimulation of lytic reactivation by the commensal microbiota. Asai et al. found that short-chain fatty acids (SCFA) present in the culture fluids of oral bacteria induced the synthesis of early antigens in Epstein-Barr viruses (40). In addition, Gorres et al. used several short-chain fatty acids (SCFAs) and their inhibitors to explore the effect of SCFAs on lytic reactivation of Epstein-Barr virus and herpesvirus. Their results showed that all SCFAs that are histone deacetylase inhibitors can reactivate herpesvirus, whereas only several of these SCFAs reactivated the Epstein-Barr virus (26). As is widely reported, the production of SCFAs is the result of a complex interaction between the gut microbiota and diet (41, 42). These results demonstrated that there is likely a link between commensal microbiota and the lytic reactivation of viruses.

#### Driving the Proliferation of Target Cells

Tuft cells are a rare type of intestinal epithelial cells that are the reservoir for fecal shedding and persistence of murine norovirus (43). Similar to certain commensal bacteria that express receptors for human norovirus, tuft cells also express a functional receptor for norovirus, CD300lf, the expression of which dictates norovirus tropism and the efficient establishment of enteric norovirus infections (44). Elegant work from Virgin et al. revealed that both type-2 cytokines and the commensal microbiota are critical in governing the proliferation of tuft cells. In antibiotic-treated mice, a marked decrease in tuft cell-specific gene expression in the colon was observed, accompanied by a reduced number of tuft cells in the colon, a phenomenon that can be rescued by adding the type-2 cytokines interleukin (IL)-4 and IL-25 (27, 28).

#### Stimulating Attachment to Permissive Cells

The elegant work of Kuss et al. revealed that both Gram-negative and Gram-positive bacteria are potent enhancers of poliovirus infectivity (23). The authors used <sup>35</sup>S-labeled poliovirus and HeLa cells and established an in vitro infection model. In this system, when poliovirus was incubated with Bacillus cereus before incubation with HeLa cells, the virus displayed dramatically increased infectivity and enhanced adherence to HeLa cells (23). Further work by the same group revealed that increased viral attachment to target cells was mainly mediated by the direct facilitation of viral binding to the poliovirus receptor (PVR) by bacterial surface polysaccharides (24). Consistent with this observation, pretreatment of HeLa cells with anti-PVR antibody significantly reduced the binding of poliovirus to HeLa cells, regardless of whether the virus was preincubated with LPS. Mechanistically, LPS treatment directly enhanced the PVRbinding ability of poliovirus, thereby stimulating attachment of the virus to target cells.

#### Contributing to Viral Replication

Certain types of viruses have evolved to interact with and use members of the host microbiota or their components to achieve optimal replication. Histo-blood group antigens (HBGAs) have been identified as receptors or coreceptors for human noroviruses. As reported, certain species of enteric bacteria express HBGAs (45). Jones et al. found that the binding of norovirus and HBGA-expressing bacteria determines the transmission and infection process of these viruses in their hosts, as infection of B cells by human norovirus can only be achieved with the presence of HBGA-positive enteric bacteria. Notably, the antibiotic depletion of normal enteric flora resulted in dramatically decreased virus titers, the mechanism of which presumably lay in the control of viral replication by the commensal microbiota (29). However, direct evidence for the control of viral replication by the commensal microbiota is lacking in this study.

#### Indirect Promotion of Viral Infection

Inducing an Immunoregulatory Microenvironment

Emerging evidence suggests that a rich and diverse commensal microbiota plays an essential role in modulating the development of the host immune system, both inside and outside of the gut (46–49). This is true not only for the eliciting of effector immune responses by stimulating the production of various proinflammatory cytokines such as interferon (IFN)-γ during infection, but also for the establishment of an immunotolerant microenvironment by contributing to the generation of immunoregulatory cells such as Treg cells to maintain homeostasis (50–52). In fact, the commensal microbiota profoundly dictates the development, differentiation, and activation of colonic regulatory T (Treg) cells, which contribute to the maintenance of homeostasis against components of the commensal microbiota and innocuous food antigens (51). Therefore, it is possible that commensal microbiota-induced Treg cells and Treg cell-related cytokines limit the degrees of antiviral immune responses.

Several lines of evidence add to this idea. In a model of mouse mammary tumor virus (MMTV) infection, interactions between the intestinal microbiota and the invading MMTV led to an immune evasion pathway for the virus, as intestinal microbiotaderived LPS can be utilized by MMTV to generate an IL-6 dependent induction of the immunoregulatory cytokine IL-10, a key cytokine mediating the immunoregulatory functions of Treg cells (30). However, MMTV was rapidly lost in toll-like receptor 4 (TLR4) mutant mice, which exhibited robust antiviral cytotoxic immune responses (31). The same group further found that the interactions between MMTV and LPS could not be achieved without the expression of LPS-binding protein (LBP), as reflected by the fact that MMTV isolated from mice lacking LBPs cannot capture LPS and stimulate TLR4, thereby showing a remarkable transmission defect (32). Interestingly, binding to MMTV would dramatically potentiate the LPS stimulation of TLR4 expression and induction of IL-6 production compared to those with virus-free LPS, indicating that virus incorporation guarantees a greater immunostimulatory ability of LPS (32). Collectively, these data indicate that interactions between the commensal microbiota and MMTV foster the establishment of an immunotolerant microenvironment in the host and lead to persistent viral infection.

#### Suppressing Local Antiviral Immune Responses

In addition to fostering the generation of immunoregulatory Treg cells, the commensal microbiota also directly skews antiviral immunity by suppressing the activation of effector immune cells and by inhibiting the production of various inflammatory cytokines that are pivotal for virus elimination, thus creating a more favoring environment for viral infection. This is true for norovirus. In a murine norovirus infection model, the authors found that antibiotic treatment prevented persistent viral infection, a phenomenon that was reversed by replenishment of the commensal microbiota (33). Interestingly, antibiotics did not directly affect viral replication or prevent tissue infection but acted specifically to trigger the expression of receptors for antiviral cytokine IFN-λ and to stimulate the expression of Stat1 and Irf3. In another murine model of norovirus infection, while IL-10−/<sup>−</sup> SPF mice showed dramatically aggravated intestinal inflammation and mucosa damage, IL-10−/<sup>−</sup> GF mice were free of epithelial barrier disruption, and transplantation of defined flora to these mice was sufficient to restore inflammatory lesions in the intestine (53).

In addition, mounting research has shown that the commensal microbiota also hinders the activation of antiviral humoral responses, mainly through regulating the production of virusspecific antibodies. In a murine rotavirus infection model, commensal microbiota elimination via antibiotic treatment or germ-free housing reduced the level of rotavirus antigen, delayed infection and decreased infectivity significantly (34). Notably, this phenotype was accompanied by a stronger antiviral humoral response, as more enhanced serum IgA, serum IgG and fecal IgA levels were observed. Consistent with these findings, antibiotic treatment results in greater maintenance of virus-specific antibody-secreting cells in the intestine. In contrast, when mice were treated with a low dose of dextran sodium sulfate to generate enhanced exposure to the microbiota, impaired production of rotavirus-specific antibodies following virus infection was identified (34). This finding is further supported by an independent study, which showed that although wild-type mice efficiently controlled endogenous retrovirus to a baseline level, mice with a defective antibody-secreting ability could not prevent viral activation and propagation (35). Importantly, this conclusion was true only when the intestinal microbiota was intact, as viral replication was clearly prevented in the host mice when these mice were kept in a germ-free condition, regardless of whether they had antibodyproducing abilities or not, further supporting that the commensal microbiota promotes viral infectivity through suppressing the antiviral humoral immune response (35).

## SUPPRESSION OF VIRAL INFECTION BY THE GUT MICROBIOTA (SUMMARIZED IN TABLE 2)

#### Direct Suppression of Viral Infection

Because the commensal microbiota is present at sites that are used by certain viruses to gain entry into the host, it is likely that there are substantial interactions between the invading viruses and commensal microbiota that could have suppressive outcomes for viral infection. Supporting this notion, it was shown that lactic bacteria are able to reduce the infectivity of vesicular stomatitis virus through direct binding to the viruses, thereby blocking the cell internalization process of these viruses (54). In addition, Enterococcus faecium can prevent infection by influenza viruses upon direct adsorptive trapping of these viruses (55). Organisms of the commensal microbiota also produce various metabolites with antimicrobial effects to prevent virus infection. This is true for the inhibition of infections by influenza virus. First, it was found that commensal microbiota-derived LPS can bind to and destabilize the morphology of influenza virions, thereby decreasing the overall stability of the virus (56). Second, an extracellular matrix-binding protein produced by Staphylococcus epidermidis, a Gram-positive bacterium that lives in the human nasal cavity as a commensal, can stably bind to influenza virus and thus block further viral infection (57). In addition to influenza virus, the replication of herpes simplex virus (HSV)-2 can also be suppressed by commensal microbiota metabolites. For example, lactic acid, a major end product of


TABLE 2 | Suppression of viral infections by the commensal microbiota and the relevant mechanisms.

the carbohydrate fermentation of all Lactobacillus species, can strongly inactivate HSV-2 in the vaginal mucosa by maintaining an acidic pH in the local environment (58). Consistent with this finding, in an in vitro study, it was shown that metabolites of vaginal Lactobacillus strains (i.e., lactic acid and hydrogen peroxide) exhibited potent virucidal activity, as highlighted by the dramatic suppression of virus replication by these substances (59). Commensal microbiota also exert their antiviral activity through bacterial components. For example, a vaginal strain of Lactobacillus brevis-extracted cell wall-associated component, which was resistant to high temperatures and protease digestion, potently inhibited the replication of HSV-2 in an in vitro model (60).

#### Indirect Suppression of Viral Infection

The commensal microbiota plays a critical role in shaping the host immune response, which essentially guarantees effective elimination of invading viruses. Supporting this notion, mounting studies have shown that intact healthy commensal microbiota help maintain robust antiviral immunity, while microbiota disruption increases viral infectivity due to the impaired capacity of the immune system to limit viral infection. For example, Clostridium orbiscindens, a specific human-associated gut microbe, produces desaminotyrosine to prime the amplification loop of type I IFN signaling, thereby mediating protection against influenza infection (61). In another influenza virus-infected chicken model, antibiotic treatment resulted in significantly higher oropharyngeal and cloacal virus shedding, which was also presumably mediated by reduced type I IFN responses after microbiota depletion, while the antibody-mediated antiviral immunity remained unaffected (**Figure 1**) (62). In simian immunodeficiency virus (SIV) infected rhesus macaques, fecal microbiota transplantation (FMT) treatment after commensal microbiota depletion induced greater antiviral immunity, as reflected by dramatically increased peripheral Th17 and Th22 cells post-FMT (63). In contrast, when antibiotic-treated mice are infected by systemic lymphocytic choriomeningitis virus or influenza virus, their macrophages show decreased expression of genes associated with viral suppression, impaired responses to type I and type II IFNs and defective ability to control viral replication (**Figure 1**) (64). Consistent with this finding, it has been shown that during respiratory influenza virus infection, antibiotic exposure led to a defective generation of virus-specific CD4 and CD8 T cells and antibodies due to an impaired inflammasome-dependent migration of antigen-presenting cells (APC) from the lung to the draining lymph nodes (65) (**Figure 1**). This finding was further supported by another study, which showed that while oral antibiotic treatment had little effect on innate immune responses after HSV infection of the vaginal mucosa, a dramatic increase in the level of IL-33, an alarmin produced in response to epithelial cell damage, was observed after antibiotic treatment. Mechanistically, IL-33 acted as an immune-regulatory factor that suppressed local antiviral immunity by hindering the recruitment of effector T cells to the infection site and thus blocking the secretion of IFN-γ in vaginal mucosa (66).

It seems that insufficient TLR ligand stimulation after antibiotic exposure was partly responsible for the compromised immune cell function. When TLR agonists were applied during virus challenge in antibiotic-treated mice, both cellular and humoral antiviral responses could be largely restored (**Figures 1**, **2**) (65, 67). Moreover, TLR2 activation by bacterial products produced by the gut microbiota is necessary for the recruitment of mast cells to sites of viral infection and the further release of cathelicidin, a mast cell-derived antiviral protein (**Figure 2**) (68). However, this situation seems different in young mice, whose gut microbiota has not been completely established. In a hepatitis B virus infection model, TLR4-intact young mice failed to resolve viruses and developed chronic infections, while their TLR4 mutant counterparts exhibited

rapid viral clearance, suggesting that an immune-tolerant pathway mediated by TLR4 signaling was predominant in young mice (72). Intriguingly, it seems that antibiotic treatmentinduced gut microbiota alteration is transient and recoverable, as a more exacerbated disease condition only appears when antibiotics are used during influenza A virus infections; when such treatment ceases before the infection, neither an antiviral immunity defect, nor enhanced viral susceptibility are observed (73).

direct virus-killing effects to suppress the infection process of the influenza virus.

It should be noted that maternal antibiotic treatment confers an altered commensal microbiota to their offspring, thereby profoundly impacting their antiviral immunity. This idea is supported by a recent study that investigated the effect of antibiotic treatment of pregnant mice on the antiviral immunity of their neonatal offspring following vaccinia virus infection (69). In this study, maternal antibiotic treatment during pregnancy and lactation resulted in remarkable alterations in the composition of the gut microbiota of the infant mice,

interferons to stimulate the activation of T cells or macrophages. As a result, these effector cells secrete virus-specific antibodies or inflammatory cytokines or exert

with Enterococcus faecalis predominating within the infant enteric flora. Notably, maternal antibiotic treatment resulted in an accelerated and increased mortality following vaccinia virus infection of the offspring, which was partly mediated by a defective IFN-γ-secreting ability of virus-specific CD8<sup>+</sup> T cells (**Figure 2**).

In the case of the infection of several viruses, higher levels of immune activation may persist, associated with inflammationinduced comorbidities of the host (74). In these cases, immune recognition of the gut microbiota is necessary for the generation and activation of immunoregulatory cells to diminish local or systemic immune activation. Indeed, Rosshart et al. found that reconstitution of the gut microbiota from wild mice confers potent protective effects to laboratory GF mice during lethal influenza virus infections, an effect mainly mediated through the prevention of excessive inflammation via IL-10 and IL-13 production in the virus-affected mice by the natural gut microbiota (70). In addition, Grayson et al. found that antibiotic treatment before or during murine Sendai virus infection resulted in greatly increased morbidity and mortality, accompanied by an abnormal immune response characterized by increased proinflammatory cytokines (i.e., IFN-γ, IL-6, and monocyte chemoattractant protein 1) and decreased Tregs in the lung (71). Notably, the neutralization of IFN-γ or the adoptive transfer of Treg cells abrogated tissue inflammation and prevented increased mortality (71). In cases of human immunodeficiency virus (HIV) infection, the gut microbiota is intimately associated with activation of the immune system in HIV-infected individuals (75). Importantly, decreased activation of CD4<sup>+</sup> T cells was observed post-FMT in SIV-infected rhesus macaques (63). This conclusion is reinforced by another independent study, which showed that the capacity of NKT cells to produce IL-4 and IL-10 in gastrointestinal-associated lymphoid tissues was associated with fewer markers of microbial transmission and less immune activation, a process dependent on the recognition of Bacteroides species by these cells (76).

Intriguingly, a recent study by Stewart et al. revealed that the interrelations between nasopharyngeal microbiota and host systemic inflammatory responses (reflected by serum metabolomic signatures) likely contribute to bronchiolitis in infants (77). Of note, the relative abundance of Streptococcus, which is specifically pathological in respiratory health, was positively correlated with metabolites associated with more severe disease (77). In comparison, the abundance of Moraxella, another important component of the nasopharyngeal microbiota, showed the opposite correlation patterns (77). Similar findings were reported in an independent study. In this study, Piters et al. showed that severity of respiratory syncytial virus (RSV) bronchiolitis induced by RSV infection was positively associated with abundance of Streptococcus and Haemophilus influenzae and negatively associated with abundance of Staphylococcus aureus in the nasal mucus (78). Interestingly, transcriptome profiles of whole blood from children with RSV infection and Streptococcus- and Haemophilus influenza-dominated microbiota revealed greater overexpression of several proinflammatory genes linked to macrophage and neutrophil activation (78). Thus, although the underlying mechanism is still unclear, these data clearly suggest that airway microbiota play an important role in regulating the systemic immune responses, thereby controlling the outcome of viral infections in the respiratory tract.

#### Suppression of Viral Infection With Unclear Mechanisms

Several clinical cases suggest a definite suppression of viral infectivity by the commensal gut microbiota, although the detailed mechanism is unclear. For example, hepatitis B virus (HBV) e-antigen (HBeAg), which may persist in patients for years after HBV infection, is commonly used as a sensitive indicator of remission activity and improved long-term outcome (79). In a trial of FMT for the treatment of HBV, the authors found a significant decrease in the HBeAg titer in patients after FMT treatment, and the HBeAg titer decreased gradually following each FMT treatment, suggesting the efficacy of modulating the gut microbiota for chronic hepatitis B treatment (80). In another trial investigating the underlying mechanism of the development of lower respiratory tract infection (LRTI) after viral infection, the authors found that patients with a higher abundance of butyrate-producing bacteria in their fecal samples showed a 5-fold lower possibility of developing viral LRTI (81).

# VIRAL INFECTION NOT AFFECTED BY COMMENSAL MICROBIOTA

Although a wealth of evidence has shown that the commensal microbiota regulates viral infectivity, there is also evidence showing that these microbial communities may have no effects on modulating antiviral responses. For example, Gopinath et al. recently found that vaginal application of the aminoglycoside antibiotic neomycin enhanced the host resistance to a broad range of viral infections, i.e., HSVs, Zika virus and influenza A virus. However, the antiviral activity of antibiotics was independent of the commensal microbiota, as the protection was also applicable to germ-free mice and in vitro cultured primary cells. Instead, neomycin increased the expression of IFN-stimulated genes in the host, a process mediated by TLR3 expressed by a specific subset of dendritic cells (82). Consistent with this finding, Zhu et al. found that B cells mediate the early control of murine norovirus infections and that this effect can also be achieved even in the absence of commensal microbiota via antibiotic treatment (83). In this study, the authors found that B cells mainly function as antigen-presenting cells but not as antibody-secreting cells to exert their virus-elimination effects. In murine leukemia virus (MuLV)-infected mice, Wilks et al. found that during retrovirus infection, both the production of virusspecific antibodies and the antibody-mediated virus-neutralizing responses were independent of the commensal microbiota, as both GF and specific pathogen-free (SPF) mice produced similar levels of virus-specific antibodies, and the antibody-mediated virus-neutralizing effects were similar in both mice, suggesting that antibody-mediated immune control of MuLV does not require commensal microbiota (84). It should be noted that several earlier studies found the overall pathogenicity of murine

leukemia virus were affected by the commensal microbiota. For example, compared to GF mice, conventionally reared mice developed higher levels of virus expression and longer latency period following infection of MuLV-Moloney (85, 86). However, when GF mice were stimulated with sheep erythrocytes, a significant increase in leukemia development was observed (86). The authors hypothesized that GF mice may not possess certain lymphoid cells that were required by MuLV replication stimulated by the commensal microbiota (86). Conflicting results were obtained in other studies showing that GF mice were more sensitive than conventional housed mice to MuLV infection (87). One potential explanation for this discrepancy is that the likely contamination of MuLV isolates by lactate dehydrogenaseelevating virus that can potently induce systemic lymphocyte activation (88), as suggested by Wilks et al. (89).

#### MODULATION OF THE COMMENSAL MICROBIOTA BY VIRUSES

While modulation of the commensal microbiota by viruses is still poorly understood, studies to date do suggest an important role of virus infection in inducing microbiota dysbiosis. This is true for HIV/SIV infection, influenza virus infection, HBV or hepatitis C virus (HCV) infection and norovirus infection, as discussed in detail below. In addition to these four types of viral infections, alteration of the gut microbiota following infection has also been described in cases of rotavirus infection in pigs or calves (90, 91), avian leukosis viruses in chickens (92), canine distemper virus infection in giant pandas (93), and white spot syndrome virus infection in crabs (94), although the relevant reports are sporadic and the corresponding mechanistic evidence is very limited.

#### HIV/SIV

A plethora of studies have emphasized that in SIV-infected non-human primates and HIV-infected patients, the commensal microbiota composition is disrupted with the enrichment of potentially pathogenic bacterial families. For example, microbial diversity in saliva of HIV patients was significantly reduced than healthy controls, accompanied by increased abundance of potentially pathogenic Megasphaera, Campylobacter, Veillonella and Prevotella species, and decreased commensal Veillonella and Streptococcus species (95, 96). In a recent study, Mukherjee et al. found that fungal communities differed significantly between HIV-infected and uninfected individuals, with Epicoccum, Candida and Alternaria being the most abundant fungi in HIVinfected individuals, while Pichia, Candida and Fusarium being the most common genera in healthy controls (97). Intriguingly, Pichia can efficiently inhibited Candida colonization (97). In

bronchoalveolar lavage fluid, although there were no significant differences among the microbial composition in HIV-infected and uninfected subjects, specific metabolic profiles were associated with bacterial organisms that potentially play a role in the pathogenesis of pneumonia (i.e., Bacteria from families Nocardioidaceae, Staphylococcaceae, Caulobacteraceae, and genus Streptococcus) in HIV-infected patients (98). In a long-term monitoring of chimpanzees following SIV infection, Moeller et al. observed a marked increase in the genera Selenomonas, Staphylococcus, and Sarcina, all containing opportunistic pathogens that were never detected at high abundances in SIVnegative chimpanzees (99, 100). However, SIV infection had little effect on the frequencies of Enterobacteriales, Bacteroidales, or Pseudomonas, nor did the authors find any differences in alphadiversity between SIV-positive and SIV-negative chimpanzees (99). In fecal samples, HIV infection was associated with consistently reduced overall microbiota richness but selective enrichment of the phyla Firmicutes and Proteobacteria, with the most prominent increase in Bacteroides and arabacteroides at the genus level (101–103). In addition, the alpha-diversity of species in the fecal microbiota is negatively associated with the severity of immunodeficiency in patients (104). Notably, combined antiretroviral therapy can effectively restore the alpha-diversity of the fecal microbiota. In addition to compositional alteration of the commensal microbiota, HIV infection also robustly alters the metabolic activity of gut microbiota (105). In contrast to healthy controls and patients with systemic lupus erythematosus and bacterium-induced diarrhea, HIV infection results in a defective metabolic capacity of gut bacteria to produce three amino acids, namely, proline, phenylalanine and lysine (105). In comparison, 3-hydroxyanthranilate, one of the major metabolites of the kynurenine pathway during the oxidative catabolism of tryptophan, was found to be significantly accumulated in the gut microbiota of all HIV-infected patients (105), which is in agreement with the finding of a previous study showing that gut microbiota with the ability to catabolize tryptophan through the kynurenine pathway are enriched in these patients (102).

Mechanistically, specific immune suppression by HIV is partly responsible for the enrichment of certain potentially pathogenic bacteria. For example, Salmonella typhimurium, a member of the Proteobacteria phylum, is tightly controlled by Th17 cells (106). In SIV-infected rhesus macaques, Th17 cells are markedly depleted, resulting in blunted Th17 responses to Salmonella typhimurium and finally leading to systemic dissemination of S. typhimurium (107), a phenomenon also observed in clinical cases of HIV-infected patients (108). In contrast, it seems that neither HIV-induced B cell dysfunction nor enteropathy affect overall systemic antibody responses to the commensal microbiota (109).

#### Influenza Viruses

Influenza viruses enter the host through the upper respiratory tract (URT) and can alter the microbial composition of the URT significantly following infection. Several studies have demonstrated that influenza virus infection can result in decreased colonization by healthy bacteria and increased abundance of potentially pathogenic microbiota. For example, a case-control study using next-generation sequencing of the 16S rRNA gene to analyze specific bacteria in patients with influenza infection and healthy controls showed that the healthy core microbiota, specially Prevotella spp. and anaerobes, were significantly decreased in influenza virus-infected patients (110). In comparison, eight potentially pathogenic bacteria were significantly enriched in these patients, including Haemophilus influenzae, Staphylococcus aureus, Moraxella catarrhalis, Streptococcus pneumoniae, Corynebacterium propinquum/pseudodiphtheriticum, and Dol osigranulum pigrum (110). Consistent with this study, Li et al. found that Prevotella was decreased after H1N1 virus infection (111). It should be noted that nasopharyngeal and oropharyngeal microbiota show distinct alteration profiles following influenza infection. For example, while nasopharyngeal Streptococcus showed higher abundance after infection (112, 113), oropharyngeal Streptococcus was significantly decreased following influenza infection (112). By contrast, Ramos-Sevillano et al. found that the throat microbiota was resilient to influenza infection, with remarkably stable bacterial communities following influenza infection, which was consistent with a recent murine modelbased study (114, 115). The discrepancy may result from different infection doses, sample collection method, subtypes of influenza virus and environmental factors (i.e., pH, CO2, and O<sup>2</sup> concentrations).

In contrast to the commensal microbiota of the URT, studies have only recently begun to evaluate how influenza virus infections affect gut microbiota. Consensus has been reached that influenza virus infection alters the commensal microbiota of the host, causing corresponding disruptions of the microbiota-host homeostasis, which largely accounts for the mechanisms by which infections are established. However, this general conclusion is based on several contradictory findings. For example, while infection of influenza virus was shown to lead to an increase in Bacteroidetes phyla abundance in one mouse-based study (116), a remarkable drop of Bacteroidetes (mainly S24-7) was observed in influenza A virus-infected birds and mice (115, 117). Interestingly, a murine study even found unchanged Bacteroidetes levels after respiratory influenza virus infection (118). In addition, in avian influenza A-infected migrating whooper swans, fewer Proteobacteria were detected in the fecal sample (117), while this was not recapitulated in another two studies showing that selective enrichment of Proteobacteria (mostly Bdellovibrionaceae) in the gut was the result of influenza virus infection (119, 120). These contradictory findings may result from the differences in virus subtypes and doses, experimental animal types, and the age, diet, as well as the lifestyle of the same animals. However, it seems that alterations in Firmicutes after influenza virus infection are more uniform, as decreases in the richness of Firmicutes (represented by Lactobacillus) were observed by most of the abovementioned studies (115–117). There is also interest in whether vaccination affects commensal microbiota. However, after live attenuated influenza virus vaccination, no changes in gut microbiota composition were discovered, indicating that only live viruses can drive an altered commensal microbiota diversity (116).

Investigating the underlying mechanism, several lines of data have highlighted that the modulation of immune responses by influenza contributes to the dysbiosis of the gut. Deriu et al. found that pulmonary infection of influenza virus induced the production of type I IFNs in the lung, which acted as a central player in upregulating Proteobacteria and depleting obligate anaerobic bacteria (120). Moreover, IFNmediated dysbiosis inhibited the antimicrobial inflammatory immune responses in the gut during Salmonella infection, further promoting Salmonella colonization and systemic dissemination (120). Consistent with the findings of this study, commensal microbiota dysregulation following influenza virus infection could also be the result of overproduction of a type II IFN, IFNγ, which was secreted by a subset of lung-derived CC chemokine receptor 9 (CCR9)+CD4<sup>+</sup> T cells in the small intestine (118). The disturbed gut microbiota further stimulated IL-15 production from intestinal epithelial cells, which subsequently facilitated the polarization of Th17 cells in situ, finally leading to intestinal injury (118).

# HBV/HCV

Quite a few studies have implicated that dysbiosis of the commensal microbiota occurs following infection with HBV/HCV and is relevant to the progression of liver disease. In patients with HBV, a profound alteration in the composition of gut microbiota is reflected by the significantly enriched Actinomyces, Clostridium sensu stricto, Megamonas and Lachnospiraceae, and a concomitant decrease in Alistipes, Bacteroides, Asaccharobacter, Parabacteroides, Butyricimonas, Clostridium IV, Coriobacteriaceae, Escherichia/Shigella, Ruminococcus, Clostridiales, Enterobacteriaceae, Lachnospiraceae, and Ruminococcaceae (121). In a species-level study, bacterial species with an opportunistically pathogenic nature were significantly elevated, while species with potential beneficial effects were downregulated in the fecal samples of HBV-infected patients (122). In HCV, a reduced bacterial diversity, with an increase in the order Streptococcus and Lactobacillus, and a decrease in Clostridiales was observed (123, 124). In another study, a lower abundance of Firmicutes, Proteobacteria, and Actinobacteria and a higher abundance of Bacteroidetes were detected in stool samples of HCV patients (124).

The course of liver disease progression in HBV/HCV patients can also be reflected by profiles of the commensal microbiota. In a study comparing the diversity of gut fungal microbiota in patients with HBV infection, the authors found that the diversity of intestinal fungi was positively associated with disease progression, as reflected by the higher richness of enteric fungal species in patients with HBV-related cirrhosis than in those with chronic infection. Moreover, patients with chronic hepatitis B exhibited higher richness of fungal species compared with asymptomatic HBV carriers and healthy controls (125). In a cross-sectional investigation, it was shown that alpha-diversity of the fecal microbiota decreased significantly from healthy controls to HCV patients without cirrhosis to those with cirrhosis (126). In addition, the ratio of Bifidobacterium/Enterobacteriaceae was suggested to be a sensitive biomarker for the clinical course of HBV, as a gradual decrease in this ratio was observed in asymptomatic HBV carriers, patients with chronic hepatitis B, and patients with HBV-associated cirrhosis (127).

In addition to the gut microbiota, an alteration of the oral microbiota in HBV-affected individuals has also been reported in a recent study, which showed that the ratio of Firmicutes/Bacteroidetes was increased significantly (128). Interestingly, HBV infection resulted in a marked increase in bacteria capable of producing H2S and CH3SH, implicating the potential contribution of the altered microbiota to the oral malodor in these patients (128). Compositional and metabolic changes in the tongue-coating microbiota have also been documented in HBV-infected individuals. As reported, yellow tongue was associated with higher HBV titers compared with those in patients with a white tongue. Moreover, a significant decrease in Bacteroidetes and an increase in Proteobacteria was found in HBV-associated yellow tongues, which also showed a selective enrichment of the metagenomic pathways involved in amino acid metabolism, consistent with the metabolic disorder of these patients (129).

In summary, HBV/HCV infection indeed caused profound changes in the composition and metabolism of the commensal microbiota. However, it should be noted that most of these studies are based on observational data. Thus, further studies exploring the underlying mechanism of HBV/HCV-induced microbiota alterations are clearly needed.

#### Norovirus

Norovirus represents one of the most important causes of acute viral gastroenteritis worldwide (20), usually causing severe diarrhea and occasionally causing chronic infections in immunocompromised individuals (20). In a human-based study investigating the effect of norovirus infection on the gut microbiota, the authors found that while the fecal microbiota in most infected individuals exhibited a similar composition to that of uninfected controls, a significant loss of diversity and richness of the gut microbiota characterized by a clear increase in the relative numbers of Proteobacteria and a corresponding decrease in Bacteroidetes was observed in a small proportion of norovirus-infected patients (130). Further analysis revealed that a single operational taxonomic unit of Escherichia coli was partially responsible for the increase in Proteobacteria in these patients (130). Consistent with this observation, human norovirus can bind to certain human stool-isolated bacteria, including those in the phylum Proteobacteria (i.e., Hafnia alvei, Citrobacter spp., Klebsiella spp., and Enterobacter cloacae), with high efficiency, implicating a direct modulation of the gut microbiota by norovirus (131). Intriguingly, murine norovirus even has the capacity to maintain gut homeostasis and shape intestinal immunity, similar to the functions of the gut microbiota. Kernbauer et al. found that norovirus infection of antibiotic-treated or germ-free mice restored the aberrant lymphocyte compartment and the abnormal intestinal morphology without inducing overt inflammation and disease (132). Importantly, norovirus infection protected antibiotic-treated mice from dextran sulfate sodium-induced intestinal injury and C. rodentium superinfection, suggesting that norovirus has the potential to replace the beneficial functions of commensal microbiota in the intestine (132).

# Theiler's Murine Encephalomyelitis Virus (TMEV)

Emerging evidence supports the intriguing concept of the braingut microbiome axis and has shown bidirectional interactions within it (133, 134). Several systems including the central, autonomic and enteric nervous systems, the neuroendocrine and the neuroimmune systems are at work to guarantee proper functioning of this axis (135). Current data have suggested that this complex communication axis is essentially linked to the regulation of multiple aspects of host physiology ranging from gastrointestinal homeostasis to psychiatric, motivational and cognitive functions (136–139). As a result, perturbation of the brain-gut microbiome axis is involved in several disorders including metabolic dysregulation and psychiatric and nonpsychiatric diseases (140, 141). Thus, it is not surprising that viral infection-associated brain abnormalities can result in gut dysbiosis, which may in turn affect the development and severity of virus-associated tissue pathology. This was evidenced by a recent study investigating the effect of intracerebral TMEV infection on commensal microbiota (142). In this study, Carrillo-Salinas et al. found that TMEV infection was associated with significantly altered gut microbiota, reflected by a reduction in the relative abundance of Alloprevotela (phylum Bacteroidetes) at 14 days post infection and a decrease in Anaerotruncus (phylum Firmicutes) and Akkermansia (phylum Verrucomicrobia) while an increase in Clostridium XIVa (phylum Firmicutes) at 28 days post infection (142). Intriguingly, the effect of TMEV infection on gut microbiota is profound and lasting, as alterations in Firmicutes and Bacteroidetes of the gut microbiota still exist at 85 days post infection (142). In addition, oral administration of antibiotics dampened TMEV infection by enhancing antiviral immune responses during acute phase of infection (142).

# CONCLUSIONS AND FUTURE PERSPECTIVES

We have discussed the current understanding of the modulation of virus infectivity by the commensal microbiota of the host and the underlying mechanisms in this regulation. We have also described the contribution of viral infection to the disturbances of microbiota homeostasis in the host. We do not yet fully understand the extent to which commensal microbiota may determine the efficiency of viral replication, transmission, and persistence, and in most cases reported, the relevant mechanisms underlying the influence of the host microbiota by invading viruses are unclear. However, the data presented do support an intimate interaction between the commensal microbiota and invading viruses, an interaction that always dictates the outcome of an infection. Thus, it is tempting to speculate that antiviral drugs aimed at modulating virus-microbiota interactions may be particularly effective in controlling the activity of many viral diseases. In fact, the pharmaceutical application of FMT and probiotic supplements have already been proven useful in reducing the severity of several diseases in human- and nonhuman primate-based studies, although these efforts may turn out to be ineffective in certain circumstances and may even result in unwanted complications (63, 143–146). Therefore, there are still major gaps in our understanding of the interactions between the commensal microbiota and viruses, and constant optimization of these potential treatment methods is clearly needed to better control viral infections via the modulation of commensal microbiota.

Recent works have shed light on the role of the commensal microbiota in health and many diseases. However, considering the immense diversity of the commensal microbiota, most studies investigating the functional attributes of these microbial communities are based on population-level analyses, and the majority of species in the commensal microbiota have never been isolated and cultured in the laboratory, thus greatly hindering progress in identifying the unique phenotypes and functions of each species of the commensal microbiota to minimize the risk of complications of FMT brought about by unwanted microbiota. Therefore, further efforts into developing more effective approaches for commensal microbiota culture in vitro are urgently needed. In addition, an important gap in microbiota research is that most studies discuss only bacterial microbiota and often overlook fungi or viruses, which are also important components of the commensal microbiota of the host (19, 147). Thus, future studies on components of the commensal microbiota other than bacteria are also needed.

As discussed above, several studies have suggested that the commensal microbiota may potently promote viral infections, and commensal microbiota depletion with antibiotics could conceivably be used as a strategy to treat viral infections. However, one must recognize that all the antibiotic-induced inhibitions of viral infection have been designed only in mouse studies, and usually only the effect of such treatment on viral infection is evaluated, regardless of the potentially negative consequences on a broader scale of infections. For example, when the majority of commensal microorganisms are depleted by broad-spectrum antibiotic treatments, the beneficial effects of the commensal microbiota on host health, the maintenance of host physiological homeostasis and the promotion of host immune functions, will concomitantly disappear, making these adverse consequences far outweigh the benefits of blocking a particular viral infection. In addition, it is now well-recognized that antibiotic overuse leads to the emergence of antibioticresistant bacteria or even superbacteria that may bring about severe or even life-threatening infections. Thus, we do not advocate the use of antibiotics to treat or prevent viral diseases in humans. However, understanding how the commensal microbiota enhances viral infection, especially the molecular requirements for the microbiota-mediated promotion of viral infections, may lead to the development of novel, feasible antiviral strategies.

Emerging data suggest that the newly discovered cyclic-GMP-AMP (cGAMP) synthase (cGAS)-cGAMP-stimulator of interferon genes (STING) pathway as the major pathway in sensing cytosolic DNA following viral infections (148– 150). In fact, the cGAS-cGAMP-STING axis has been shown to be involved in restricting both DNA and RNA virus infections (150, 151). Notably, a recent work has shown that guanylate cyclase C, which is expressed on intestinal epithelial cells and is crucial for the generation of cGMP, critically regulates microbiome composition of the intestine through maintaining barrier integrity by cGMP production (152). However, during viral infections, the effect of cGAMP on the commensal microbiota is unclear. Considering that the mechanism underlying the modulation of commensal microbiota by viral infections has not been fully clarified, this intriguing question undoubtedly warrants further investigations.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

W-TM and J-LH designed the structure of this review. NL and W-TM wrote the manuscript. NL, MP, Q-LF, and J-LH revised the manuscript. All authors have reviewed the final version of the manuscript.

## FUNDING

This research was supported by China Postdoctoral Science Foundation (K3080217064) and Ph.D. research startup fund of Northwest Agriculture and Forestry University (Z109021715).


liver disease: a pilot project. Sci Rep. (2015) 5:17098. doi: 10.1038/srep 17098


152. Mann EA, Harmel-Laws E, Cohen MB, Steinbrecher KA. Guanylate cyclase C limits systemic dissemination of a murine enteric pathogen. BMC Gastroenterol. (2013) 13:135. doi: 10.1186/1471-230X-13-135

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

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

# Diet-Microbe-Host Interactions That Affect Gut Mucosal Integrity and Infection Resistance

Andrew J. Forgie, Janelle M. Fouhse and Benjamin P. Willing\*

Department of Agricultural, Food and Nutritional Science, University of Alberta, Edmonton, AB, Canada

The gastrointestinal tract microbiome plays a critical role in regulating host innate and adaptive immune responses against pathogenic bacteria. Disease associated dysbiosis and environmental induced insults, such as antibiotic treatments can lead to increased susceptibility to infection, particularly in a hospital setting. Dietary intervention is the greatest tool available to modify the microbiome and support pathogen resistance. Some dietary components can maintain a healthy disease resistant microbiome, whereas others can contribute to an imbalanced microbial population, impairing intestinal barrier function and immunity. Characterizing the effects of dietary components through the host-microbe axis as it relates to gastrointestinal health is vital to provide evidence-based dietary interventions to mitigate infections. This review will cover the effect of dietary components (carbohydrates, fiber, proteins, fats, polyphenolic compounds, vitamins, and minerals) on intestinal integrity and highlight their ability to modulate host-microbe interactions as to improve pathogen resistance.

#### Edited by:

Fernanda Cristina Petersen, University of Oslo, Norway

#### Reviewed by:

Fernando Gómez-Chávez, National Council of Science and Technology (CONACYT), Mexico Håkon Valen, Nordic Institute of Dental Materials, Norway

> \*Correspondence: Benjamin P. Willing willing@ualberta.ca

#### Specialty section:

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

Received: 14 January 2019 Accepted: 17 July 2019 Published: 06 August 2019

#### Citation:

Forgie AJ, Fouhse JM and Willing BP (2019) Diet-Microbe-Host Interactions That Affect Gut Mucosal Integrity and Infection Resistance. Front. Immunol. 10:1802. doi: 10.3389/fimmu.2019.01802 Keywords: microbiota, diet, infection resistance, gastrointestinal integrity, disease susceptibility

# INTRODUCTION

Infectious enteric diseases are a major cause of morbidity and mortality worldwide and are of particular concern in hospital settings and developing countries. According to the World Health Organization, infectious enteric diseases are one of the top 10 causes of death leading to over two billion cases and one million deaths worldwide in 2010 (1). Host resistance toward invading pathogens requires tight regulation of the gastrointestinal environment, maintained through a synergistic relationship between the host immune system and microbiome. Disruption to a host's intestinal homeostasis, including insults from diet, stress, antibiotic and drug treatment, allergies, cancer, and related illnesses can leave the host vulnerable to enteric pathogens (2). It is well-understood that diet can play a major role on health by positively and negatively shaping gastrointestinal ecology (3, 4), and therefore should be a major focus in mitigating the severity of infection.

Although humans have successfully reduced pathogen exposure through effective sanitation practices, the adoption of a "Western diet," over-sanitation and lack of physical exercise are hypothesized to have contributed to the rise in autoimmune disorders (5). The "Western diet" is characterized by the excessive consumption of fats, proteins, refined sugar, and low intake of dietary fiber. Other dietary patterns such as the Mediterranean, Vegetarian-based, Japanese-based, and Ketogenic type diets can positively regulate immune responsiveness to reduce immune activity and support health (6). However, human epidemiology studies on diet tend to exclude important interindividual variations that govern the gastrointestinal microbiota and may explain the diverse claims to which foods are known as "protective" and "harmful" (7). Establishing a mechanistic link between individual diet components using microbe-host interactions will aid to provide evidence driven recommendations to help control an overactive immune response.

An overactive immune system is associated with autoimmune disorders such as irritable bowel disease (IBD) that affects host immune activity and leads to increased incidence of infection (8, 9). Likewise, "westernized diets" have shown to enhance Escherichia coli colonization and associated inflammation in mice by altering the host mucus layer, increasing intestinal permeability, and impairing immune function (10). Dietary fiber and other microbiota-accessible carbohydrates (MACs) are a key component missing from the "westernized diet" that when re-introduced provides a beneficial balance to host health and microbiome (11). Fiber is exhaustively studied as a microbial fermentation substrate that produces short chain fatty acids (SCFAs) with known benefits to host intestinal homeostasis and health (12). However, we fear that this focus on the beneficial effects of fiber-associated SCFA production has led researchers to overlook other common dietary components that may positively or negatively influence the host gastrointestinal environment and health.

Diet intervention should be considered a valuable tool to manipulate the host-microbe axis to help sustain intestinal homeostasis and infection resistance. Dietary components such as carbohydrates, lipids, proteins, phytochemicals, minerals, and vitamins all have unique structural and chemical (physicochemical) properties that influence host pathogen resistance directly and indirectly through the microbiome. Bridging the gap between diet, host, and microbiome as they relate to immunity and disease resistance is a multifaceted field that requires an understanding of their combined effects on intestinal homeostasis (**Figure 1**). This review explores the role of common dietary components on host-microbe interactions that modulate host resistance and tolerance toward common infectious diseases. We highlight the opportunity to improve outcomes, yet recognize the current knowledge limits the ability to provide concrete dietary advice. This is partially limited by the fact that diet focused infection resistance research is scarce and difficult to translate to humans.

#### GALT AND MICROBIOME REGULATE HOST DEFENSES

The gut associated lymphoid tissue (GALT) plays a crucial role in regulating intestinal homeostasis and is composed of lymph nodes, lamina propria, and epithelial cells that together provide the host with a protective barrier and immune defense against invading pathogens (13). On the other hand, the microbiota provides a physical presence that can directly prevent pathogen colonization by competing for attachment sites or nutrient resources. Indirectly, the microbiota helps to improve host resistance by modulating intestinal integrity through the mucus layer, tight junction proteins, and antimicrobial peptides (AMPs: cathelicidins, C-type lectins, and defensins) (14, 15). Mucins

secreted by goblet cells provide the first line of defense by forming a physical barrier composed of highly glycosylated and interlinked proteins between luminal bacteria and host epithelial cells (16). The mucus layer provides lubricant and is metabolized by mucin-degrading (mucolytic) bacteria forming the loosely attached layer (17), whereas the adherent layer, when properly formed, secures a balance of host AMPs and immune factors that maintain intestinal homeostasis (18).

Disruptions to the balanced microbial ecosystem greatly increase a host's vulnerability to infection (19). In particular, antibiotic exposure can cause major shifts in microbial communities leading to mucus layer thinning, predisposing, and exacerbating infections, as shown with antibiotic accompanied Citrobacter rodentium infections in mice (20). Antibioticinduced microbiota imbalances are well-documented to alter the production of AMPs, tight junction proteins, and immune factors that normally contribute to intestinal homeostasis and infection resistance (21, 22). Secretory immunoglobulin A (SIgA) antibodies are abundant immune factors of the intestinal lumen that protect epithelial cells from enteric pathogens and toxins by blocking their access to epithelial receptors and entrapping them in mucus to promote clearance (23). Although SIgA targets and disrupts pathogens and antigens, commensal microbes such as Bacteroides fragilis alter their surface proteins to attract SIgA to enhance mucosal colonization (24). Intestinal epithelial cells (IECs) produce reactive oxygen species (ROS) (25) and Resistinlike molecules (e.g., RELMβ) (26) that hinder commensal and pathogenic bacteria colonization, further maintaining intestinal hemostasis. IECs apical surface fucosylation is another useful host strategy that controls commensal microbes and inhibits pathogens. Secreted fucose is metabolized by bacteria to produce bioactive metabolites, reduce virulence factors, and enrich beneficial gut microbes to strengthen colonization exclusion (27). Alternatively, fucose can be fermented by commensal microbes into 1,2-propanediol and utilized by Salmonella during inflammation to drive their fitness in the colon (28).

The host has significant control over microbial communities of the small and large intestine; however, this relationship is complex and is managed in part through gastric acid secretions, intestinal motility, bile secretions, oxygen gradients, and regulation of pattern recognition receptors (PPRs), such as Toll-like receptors (TLRs) (4). The host recognizes commensal bacteria through activation of TLRs and relays an appropriate response in accordance to the specific microbial derived ligands [e.g., peptidoglycan, lipoprotein, lipopolysaccharide (LPS), and flagellin] (29). Innate lymphoid cells (ILCs) have been identified as key immune regulatory cells of the GALT controlling pathogen resistance, inflammation, and metabolic homeostasis (30). ILCs concentrate within mucosal surfaces and relay signals sent between the microbiota, epithelia, immune cells, and metabolites in the intestine to maintain epithelial barrier function. Transcriptomic analysis of 15 ILC subtypes revealed their regulatory functions depend on the presence of the microbiome, nutrients, and xenobiotics (31). Ultimately, it is the combined relationship between the gut microbiota, host, and diet that help improve or worsen a host's ability to tolerate and resist pathogenic bacteria (**Figure 2**). The remainder of this review will focus on specific dietary components and how they stimulate some of these and other host-microbe interactions resulting in impaired or improved host disease resistance.

# CARBOHYDRATES

Dietary carbohydrates are often classified by their degree of polymerization into mono-, di-, oligo-, or poly-saccharides and composition of their monosaccharides: glucose, fructose, galactose, and xylose. Typically, carbohydrates are categorized as either digestible or indigestible (fiber). Binding and structural properties of carbohydrates dictate the glucosidase enzymes required to break bonds into their basic units for absorption (32). The digestible carbohydrates escaping host small intestinal digestion, along with dietary fiber, become available as microbial energy substrates and are able to substantially alter the intestinal ecosystem and community structure (33).

Increasing intake of digestible carbohydrates has been scrutinized for contributing to the worldwide obesity and diabetes epidemics. However, carbohydrates are essential energy substrates for the central nervous system and red blood cells, are required to maintain cellular energy balance after sustained increases in metabolic activity, and to restore energy levels and glycogen stores (34). Humans and animals are able to regulate blood glucose levels; however, excessive dietary carbohydrate consumption can worsen acute hyperglycemia, particularly during times of an illness (35, 36) and stress (37, 38). A medical illness can enhance the negative effects of acute hyperglycemia, which include inhibition of neutrophil migration, phagocytosis, superoxide production, and microbial killing, compromising host innate immunity against bacterial and fungal infections (39). Diets high in simple and refined carbohydrates are shown to negatively impact gastrointestinal microbial communities leading to intestinal barrier dysfunction and greater risk for enteric infection (36). Whereas, balanced diets containing resistant starch and fiber stimulate microbial fermentation leading to a stable diverse microbiome and production of beneficial SCFAs (40). Understanding both negative and positive effects of carbohydrate consumption on gastrointestinal immunity and microbial populations will provide vital insight toward dietary strategies to help maintain pathogen resistance.

Dietary trehalose, a food component used to improve a product's texture, flavor, glycemic index and shelf life, was introduced in the early 2000's and has since been proposed to have contributed to the global Clostridioides difficile epidemic (41). Trehalose is a disaccharide composed of two glucose molecules linked by a resistant α,α−1,1-glucosidic bond found in plants, algae, fungi, yeast, bacteria, insects, and other invertebrates (42). Mammals and other vertebrates lack the ability to synthesize trehalose, and the dietary fate of trehalose depends on the capacity of the small intestinal trehalase enzyme to hydrolyze it into glucose (43). Trehalase deficiency is rare in humans but excessive consumption of trehalose can lead to negative intestinal imbalances similar to those associated with lactose and fructose intolerances. Researchers believe the increased use of trehalose in food production has naturally selected for C. difficile with the capacity to metabolize trehalose more efficiently, thus increasing pathogen fitness and contributing to their hypervirulent outbreaks in the human population (41). To combat reoccurring C. difficile infections a fecal microbial transplant (FMT) from a healthy donor has become a helpful treatment option, however the mechanism of remission remains unclear (44). The success of FMTs to treat C. difficile infections highlights the importance of a "healthy" gut microbiome to promote infection resistance. Additional research is needed to confirm the impact of specific carbohydrates and their malabsorption on immune and microbial networks in the gut as it relates to pathogen fitness. Interestingly, studies in mice comparing fiber-rich and fiber-deprived diets support the detrimental effect of a simple carbohydrate dominated diet and the importance of fiber on infection resistance (11, 33).

# DIETARY FIBER

Health benefits associated with foods rich in non-digestible dietary fiber depend on their type, source, and proportion of water soluble and insoluble carbohydrate components (45). Fruits, vegetables, and grains are excellent sources of numerous fiber types, however, not all fiber sources and types are created equal. The food source, glycosylated chain structures, and their fermentability, along with other inherent components are key parameters for their functional quality within the gastrointestinal tract (12). Non-digestible carbohydrates are composed of monosaccharide units (glucose, fructose, galactose, xylose, fucose, and sialic acid) found naturally in plants, algae, fungi, bacteria, and mammalian milk, or produced by chemical or enzymatic processes (46, 47). Short chain fructo-oligosaccharides (FOS) have received a great deal of attention due to their prebiotic effects (48) and fact that they occur naturally (mostly as inulin) with different degrees of polymerization in foods (47). The consumption of prebiotic fibers have helped with diarrhea and constipation (49–51), however, not everyone benefits from their consumption, and can even lead to excessive gas production, bloating, and discomfort (50, 52). In cases of gastrointestinal

FIGURE 2 | Diet and immune activity load (allergies, cancer, other illness, etc.) determine host intestinal integrity toward invading pathogens. Diet affects intestinal integrity directly by stimulating IECs, ILCs, and microbial communities, and indirectly through microbial fermentation by-products (SCFAs, H2S, etc.). A healthy individual following a balanced diet to maintain symbiosis between host and microbial populations has enhanced intestinal integrity with a thick inner and outer mucus layer that retains AMPs and other compounds to protect the host against pathogen colonization (A). A diseased host with heightened immune activity maintains symbiosis by consuming dietary components that protect and boost host innate defenses (IgA, AMPs, mucus, fucosylation) and adaptive immune responses to prevent pathogen colonization (B). Whereas, diseased individuals with heightened immune activity consuming a poor diet are more susceptible to enteric infections due to impaired host defenses that cannot control the dysbiotic intestinal environment (C).

discomfort, a diet low in fructans (FODMAP-restricted diet) or reducing dietary fiber is often effective but remains controversial, and individualized (53–55).

The effects of various non-digestible fiber on health and microbiota is thoroughly reviewed (12, 45, 46). In general, dietary fiber can modify gastrointestinal function directly through fecal bulking and indirectly through the modification of microbial community structure, and by increasing microbial biomass and fermentation products (45). Fiber fermentation leads to beneficial SCFAs (mainly acetate, propionate, and butyrate) but also undesired gases such as carbon dioxide, hydrogen, and methane (56). Increased gas production, fecal bulking and delayed gastric emptying can lead to discomfort, bloating, and flatus in many individuals (45). Microbial fermentation products such as SCFAs interact with the intestinal epithelium to promote certain defense mechanisms. In particular, microbial production of butyrate provides an energy substrate to epithelial cells (57), maintains the hypoxic environment (58), and promotes improved barrier function through hypoxia inducible factor (HIF) (59). Induction of HIF transcription factor subsequently stimulates downstream signaling to increase mucus production (60) and expression of AMPs (61) ultimately helping to minimize facultative pathogen growth.

According to the Global Burden of Diseases, Injuries, and Risk Factors Study of 2015 infectious diarrhea is a leading cause of death globally among all ages (1.3 million deaths); with a large proportion of those occurring in infants under 5 years of age (499,000 deaths) (62). Providing children with MACs is an important strategy to mitigate infection burden by stabilizing the microbiota and by bolstering intestinal immunity. Infants that are exclusively breast fed have reduced risk of developing diarrheal disease (63), partially due to the naturally occurring human milk oligosaccharides (HMOs) present in breast milk. HMOs are soluble complex carbohydrates that act as prebiotics, providing a substrate for the intestinal microbiota and can prevent pathogenic bacterial adhesion through a variety of mechanisms (64). In vitro studies determined that HMOs act as pathogen decoy receptors to prevent infections and their activities depend on the location and degree of fucosylation (65). Human breast milk contains a multitude of other bioactive factors, immunoglobulins, cytokines, chemokines, growth factors, hormones, and lactoferrin which all likely contribute to the improved disease resistance of breast fed infants and is reviewed elsewhere (66). Human milk has shown the ability to directly inhibit the adherence of Streptococcus pneumonia and Haemophilus influenza to human mucosal cells ex vivo (67). When HMOs were fractioned, it was found that the acidic fraction had greater anti-adhesive properties toward enteropathogenic E. coli (EPEC), Vibrio cholera, and Salmonella fyris compared to the neutral high and low molecular weight

fractions (68). Similarly, HMOs blocked EPEC adherence to epithelial cells in vitro and reduced EPEC colonization in newborn mice, further implying the essential role HMO play in the prevention of infectious disease in human infants (69). Experimentally, it was shown that supplementing formula with HMO reduced the duration of diarrhea in rotavirus-infected pigs and promoted IFNγ and IL-10 expression in the ileum, suggesting HMOs may also protect infants against rotavirus infection (70). Therefore, research efforts have focused on HMO substitutes that can be added to formula fed to infants that are unable to breastfeed. Human and animal studies suggest supplementing formula with fermentable fiber (e.g., soy polysaccharides, fructo- & galacto- oligosaccharide) reduces infection-associated diarrhea burden by improving intestinal homeostasis (71) and increasing beneficial Bifidobacterium species (72–74).

Minimizing infectious diarrheal disease with dietary tools has become the focus of recent research efforts. The importance of non-digestible fermentable fiber or MACs intake in adults has clearly been shown where a greater intake (comparing top vs. bottom quartiles) reduced risk of death from cardiovascular, infectious, and respiratory disease by 24–56% in men and 34– 59% in women (75). Galacto-oligosaccharides (GOS) have shown to increase bifidobacteria and beneficially modulate immune function when supplemented to elderly volunteers. Along with improving phagocytosis and natural killer cell activity, the GOS supplemented volunteers had an anti-inflammatory cytokine profiles with increased IL-10 and reduced IL-1B, IL-6, and TNFα (76). In a double blind placebo controlled trial, those supplemented with GOS had reduced diarrhea incidence, duration, and severity (77). Clostridioides difficile is the leading cause of health care-associated diarrheal infections, commonly affecting the elderly and antibiotic treated hospitalized patients (78). Significant evidence suggests that the inclusion of soluble fiber to the diet, specifically MACs that increase SCFA production, may be a useful strategy to enhance infection resistance (79). In a mouse model, dietary inclusion of MACs or inulin alone was shown to suppress C. difficile infection; whereas diets devoid of MACs exacerbated the infection (11). The mechanisms by which MACs help to mitigate C. difficile infection is through the expansion of fiber fermenting microbiota (via competitive exclusion) and subsequent increases in their immune-stimulatory metabolites (promote host defenses), which limit a pathogen's fitness (11).

β-glucans are one type of fermentable fiber that is frequently studied due to its common occurrence in the cell walls of yeast, fungi, and cereals such as barley and oats. Aside from acting as a microbial fermentation substrate, β-glucans are also of great interest for their direct effect on host immune activities and functions that alter immunity toward infections. In humans, the immune modulating property is due to the binding of β-glucans with host receptor dectin-1 (80), which contributes to macrophages activation, and induce phagocytosis (81). Studies in mice found that oat derived βglucans supplemented at 3 mg every other day stimulated a systemic immune response that reduced fecal oocyst shedding of Eimeria vermiformis by 39.6% post-challenge by increasing specific antibodies against the parasite (81). Oral administration of β-glucan from a fungal source (Sclerotinia sclerotiorum at 80 mg/kg every 2 or 3 days) was shown to directly stimulate proliferative responses of Peyer's patches to both T and B-cell mitogens, suggesting β-glucans may also stimulate a mucosal immune activation (82). Intraperitoneal injection of β-glucans has also shown to work as a potent adjuvant to enhance host resistance to both bacterial (81) and parasitic (Leishmania) infections (83). The use of immunostimulants derived from naturally occurring polysaccharides (e.g., β-glucan or chitosan) has become somewhat commonplace in the aquaculture industry as an alternative strategy for disease prevention. Inclusion of oligo-β-glucans (100–200 mg/kg) to striped catfish has shown to improve growth performance and reduce mortality post Edwardsiella ictaluri challenge via heightened phagocytic and lysozyme activity (84). The inclusion of dietary β-glucans (200 mg/kg) in poultry has also been used effectively to reduce the severity of necrotic enteritis when challenged with Eimeria and C. perfringens (85) and inhibited growth depression when challenged with Salmonella enteritidis (86) by increasing specific antibody levels. In both cases, inclusion of dietary β-glucans reduced pathogen colonization (C. perfringens and S. enteritidis).

Generally, increasing fiber will change the microbiome and improve gastrointestinal heath. As stated previously, the benefits associated from consuming food sources or supplements high in fiber is individualized and should be carefully monitored for side-effects.

# FATS

Fats are an essential dietary macronutrient that have been criticized and are commonly avoided in developed countries with the objective of reducing weight, cholesterol levels, and cardiovascular disease risk. Fat avoidance and subsequent reliance on simple carbohydrates for caloric intake with reduced energy expenditure is believed to have contributed to the unintended rise of obesity worldwide (87). In healthy individuals most fats are emulsified and absorbed in the small intestine; however, in excess and during intestinal stress fats can travel toward the colon as a substrate for the microbiota (88). Human and animal studies have shown that intestinal microbes have the capacity to alter host homeostasis through a variety of metabolites, including carcinogenic and cytotoxic secondary bile acids (89). The effects of the microbiota on host homeostasis is through alteration to hepatic lipid and bile metabolism, reverse cholesterol transport, energy expenditure, and insulin sensitivity in peripheral tissue (90). In this respect, dietary lipids are capable of directly affecting the host and microbiome, while indirectly altering host homeostasis through the microbiome and their metabolites.

The direct effect of microbial fat metabolism on intestinal health has yet to be established but studies have shown that dietary lipid profiles can alter the outcome of enteric infections. Fat consumption with regards to infection have been thoroughly reviewed elsewhere (91), and provides a bases to establish the connection between microbe and host

on enteric pathogen resistance. A study comparing dietary saturated (SFA, milk), monounsaturated (MUFA, olive oil), and polyunsaturated (PUFA, omega-6 corn oil) fatty acids uncovered distinct lipid mediated immune responses in mice after an acute C. rodentium challenge (92). SFA and MUFA dominated diets induced protective T-regulatory cells, interleukin (IL)-10, IL-33, and SCFAs that helped mitigate inflammation during enteric infection (92). Interestingly, in a dextran sodium sulfate (DSS) model, IL-10 knockout mice fed a diet containing milk SFAs, but not lard fat SFAs, resulted in a pro-inflammatory TH1 immune response associated with a bloom of Bilophila wadsworthia and its metabolites, hydrogen sulfide and secondary bile acids (93). Diets high in medium-chain SFAs like coconut oil have antifungal action toward Candida albicans (94) and antibacterial properties against enteric pathogens (95). Moreover, the addition of fish oil, high in omega-3 (n-3) fatty acids to a SFA dominated diet activated intestinal alkaline phosphatase (IAP), an enzyme that detoxifies proinflammatory lipopolysaccharide (LPS) endotoxins from gram-negative bacteria that accumulates during infection; whereas supplementing n-3 to an n-6 rich diet did not enhance IAP activity (92). Previously it has been observed that high levels of dietary n-6 PUFAs in fact reduce IAP activity leading to LPS endotoxemia in mice (96). Transgenic Fat-1 mice, which genetically retain a higher concentration of n-3 in their tissues, demonstrated elevated serum IL-10 and IAP activity (96). In mice, safflower and canola oil based diets (high in n-6) heighten mucosal TH1/TH17 responses and inflammation, whereas a fish oil based diet has shown to have a protective anti-inflammatory effect following a C. rodentium infection (97). Diets rich in n-3 PUFAs have proven protective against many extracellular pathogens (Mycobacterium tuberculosis, Salmonella typhimurium, S. pneumoniae, Pseudomonas aeruginosa, E. coli, Staphylococcus aureus, C. rodentium, Helicobacter hepaticus, H. pylori, and Listeria monocytogene); however, potentially damaging effects were observed during intracellular viral infections (98, 99). Dose and timing of n-3 PUFAs is critical for intestinal immune homeostasis. Sustained high doses alter microbial communities and host immune system toward an antiinflammatory state that could exacerbate infections, especially when proinflammatory responses are essential for infection clearance (98). Interestingly, lipid composition affects hostmicrobial interactions even when administered via a non-enteral route. The inclusion of mixed lipids containing soybean oil, medium-chain triglycerides, olive oil, and fish oil in parenteral formula was shown to reduce intestinal inflammation and alter microbial composition in a piglet model of infant total parenteral nutrition as compared to soybean oil alone (100).

#### PROTEIN

Protein homeostasis is crucial for host health, physiology, and immune development that together foster a fast-acting immune response toward pathogens. The role of dietary protein and amino acids on host immune function related to diet malnutrition and pathogen interactions has been thoroughly reviewed (101, 102). Amino acids play a major role in regulating immune cell activation, cellular redox homeostasis, lymphocyte proliferation, and production of cytokines, cytotoxins, and antibodies (101). Protein deficiency is well-known to impair immunity and infection resistance, especially during stress and illness due to protein malabsorption and protein consuming processes such as tissue repair (103). Protein deficits have been shown to exacerbate parasitic Cryptosporidium infections in mice through disruption of baseline (primary) Th1-type mucosal immunity (104). Furthermore, protein-deprived diets decreased small intestinal macrophage proliferation and IL-10 production independently of the microbiota (105).

In contrast, researchers propose that protein-rich diets can be just as harmful since they lead to an increase in undigested proteins that encourage protein-fermenting bacteria and disease susceptibility (106). Resistant and undigested proteins can interfere with host functions directly as biologically active proteins (BAP) like trypsin and chymotrypsin inhibitors, and indirectly through microbial proteolytic fermentation byproducts [H2, CO2, CH4, H2S, SCFA, branched chain amino acids (BCAA), nitrogenous compounds, phenols, and indoles] with poorly understood health outcomes (107). It is important to note that dietary crude protein can contain a high concentration of BAPs whose activities can be reduced upon hydrolysis digestion (heating, chemical, or enzymatic). A study replacing crude protein (wheat and casein) with purified amino acids to diets fed to weaned pigs reduced proteolytic fermentation before and after an enterotoxigenic E. coli (ETEC) K88 challenge (108). Three days post-infection, ETEC K88 colonized the small intestine of pigs fed the crude protein diet whereas no colonization was observed in the small intestine of pigs receiving the purified amino acid diet. In this context, undigested protein or other components associated with crude protein diets promoted ETEC growth and colonization in the small intestine.

Furthermore, the source of proteins can impact microbial communities depending on the digestibility and total amino acids in the diet (106). For instance, animal proteins tend to be highly digestible in the proximal intestine compared to plantbased proteins (109). Processing proteins with heat can impact their digestibility, for example, rats fed thermolyzed (heated to 180◦C for 1–2 h) casein, soy, or egg white protein had reduced proximal intestinal digestibility, leading to a greater degree of protein fermentation in the cecum (110). The number of aberrant crypts were measured after azoxymethane challenge to assess the carcinogenic promoting properties of casein, soy, and egg proteins. For the heat-treated proteins, the number of aberrant crypts increased with casein, remained unchanged with soy, and decreased with egg white compared to untreated protein diets. In agreement, a DSS mouse model study using multiple custom diets demonstrated that casein and soy proteins worsened DSS associated weight loss, whereas no effect was seen in mice fed the egg white protein diets (111). In contrast, a human trial compared high- and low-fat diets with non-meat protein (legumes, nuts, grains, soy), red meat protein (beef) or white meat protein (chicken and turkey) on the gut microbiome and found only a modest impact of protein source on the microbiome (112). For cardiovascular health, the plant-based proteins outperformed meat protein diets but white meat was no better than red meat for reducing disease risk (113). However, animal protein dominated diets tend to include higher amounts of fats, which ultimately may be more impactful on health than the proteins themselves. Plant-based protein diets may inherently contain detrimental components. For example, soybean isoflavones are suggested to contribute to greater parasitic oocyst fecal output and reduce immune responsiveness in mice fed a soy-based diet compared to casein and whey protein fed groups (114). For this reason, crude protein diet studies make it difficult to identify the bioactive component responsible for the observed phenotype. A study in rats comparing protein from soy, casein, pork, beef, chicken, and fish indicates that protein source alters microbial composition (115). Specifically, white meat (chicken and fish) increased beneficial Lactobacillus species. Blood levels of lipopolysaccharide-binding protein (LBP), a marker for lipopolysaccharide (LPS) endotoxemia, was found to be significantly higher in the soy protein diet group compared to fish, chicken, pork, beef, and casein protein fed groups. Further research is needed in controlled animal models to investigate isolated protein types and processing techniques on host digestion, microbiome, and fermentation products to mechanistically link the impact of protein on infection resistance.

Dietary glutamine supplementation has proven to be an effective therapy to help restore intestinal integrity in patients with post-infectious associated irritable bowel syndrome (116). Although glutamine significantly improved IBS scores compared to a placebo supplemented group, a larger cohort and mechanistic studies are warranted. The effect of glutamine supplementation may be associated with glutamines ability to enhance intestinal cell proliferation (117), decrease the Firmicute population, and activate innate immunity through NF-κB, MAPK, and PI3K-Akt signaling pathways (118). Similar effects have been observed with arginine supplementation (119). Over a 14-day study, daily supplementation of 30 g of L-glutamine to overweight individuals led to a significant decrease in Firmicute populations, including species from the genus Dialister, Dorea, Pseudobutyrivibrio, and Veillonella (120). Since overweight individuals typically have a higher Firmicute/Bacteroidetes ratio than lean individuals (121), a decrease in Firmicutes with glutamine supplementation suggests that dietary glutamine may play a beneficial role in restoring microbiota balance. In accordance, glutamine and arginine supplementation promoted the activation of innate immunity and lowered intestinal pathogen load in ETEC-infected mice (122). In humans, enteral glutamine administration in critically ill patients with severe trauma, burns, and sepsis significantly reduced the number of isolated enteric bacteria such as Pseudomonas sp., Klebsiella sp., E. coli, and Acinetobacter sp., all of which can contribute to pneumonia if transmitted to the lungs (123, 124). Enteral glutamine administration reduced bacterial overgrowth within the gastrointestinal tract, which may have reduced the chance of bacterial exposure to the lungs and explain the reduced incidence of pneumonia in patients. Moreover, a systematic review and meta-analysis concluded that glutamine-enriched enteral formulae can significantly reduce gut permeability in critically ill patients (125). The requirement and importance of enteric glutamine has been extensively reviewed (126), but requires further research in healthy subjects and animals models to understand the impact on the microbiome and enteric infection resistance.

Further emerging evidence suggests that numerous microbially-derived indoles from tryptophan catabolism can promote intestinal homeostasis by activating regulatory T cells (Tregs) through their interaction with the aryl hydrocarbon receptor (AhR) (127). Roager and Licht summarize known microbes responsible for producing tryptophan-derivatives that positively act on tight junctions, gastrointestinal motility, host metabolism, AhR to activate IL-22, along with their systemic anti-oxidative and anti-inflammatory properties (128). In this respect, dietary tryptophan likely contributes to infection resistance by priming host defense strategies. The importance of tryptophan is further supported by the ability of host dendritic cells to metabolize tryptophan into kynurenine using indoleamine 2,3-dioxygenase-1 (IDO1) in order to control host inflammation during a C. difficile infection (129). Kynurenine production during C. difficile infection is proposed to be beneficial as it reduces excessive interferon-γ (IFNγ) cytokine production by limiting neutrophil populations in the lamina propria (129). Clinically, these findings provide important insight into the use of IDO1 inhibitors for cancer treatment which would prevent kynurenine production, and increase the severity of C. difficile infection (129). Like tryptophan, threonine is another essential amino acid that must be obtained from diet with deficiencies leading to immune and barrier dysfunctions (130). Dietary threonine is essential for the production of mucin with deficient diets leading to altered mucosal integrity and persistent diarrhea in neonatal piglets (131). The importance of dietary threonine for mucus production and structure may not only provide protection for host IECs but also could stimulate mucolytic bacteria with unknown functions (**Figure 3**).

Dietary protein source, amount, and processing can alter their impact and effects within gastrointestinal environment. Clearly host protein digestion shares an intimate relationship with the gut microbiome and their fermentation products (132). A balanced macronutrient or low indigestible protein diet is recommended to discourage proteolytic bacteria from overproducing cytotoxic, genotoxic, and carcinogenic byproducts that disrupt intestinal integrity and increase the risk of infection (106).

# PHYTOCHEMICALS

Plants synthesize a large pool of compounds known as phytochemicals to protect themselves from stress, predation, and infection. Complex mixtures of phytochemicals are found in the roots, seeds, leaves, bark, flowers, and fruit of plants and have been intensively studied for their antimicrobial, anti-inflammatory, and antioxidants activities (133). The physicochemical properties of phytochemicals give plants their unique color, smell, and flavor profiles, and dictates their bioactivities and bioavailability within the gastrointestinal tract (134). Condensed tannins, mainly polymeric flavanols can act as antinutritional factors that reduce host digestion

through enzyme inhibition and protein precipitation (135). However, the consumption of phytochemicals is typically associated with beneficial health outcomes from their activities on the resident microbial population and host metabolism (14, 136, 137). Phytochemicals are treated as xenobiotics by the host and because of this, the liver can reintroduce phytochemical derivatives to microbes through enterohepatic circulation, further complicating their effects on host health. Many studies fail to demonstrate and characterize absorbed phytochemical derivatives to investigate whether their impact on host are direct or indirect through the microbiota.

Research has focused on the use of phytochemicals as an alternative to antibiotics and as a dietary supplement to strengthen host pathogen resistance (138). For instance, chickens fed a mixture of pepper (Capsicum) and turmeric oleoresin had less weight loss and reduced intestinal lesions scores in a necrotic enteritis disease model (139). The phytochemical mixture lowered intestinal but increased splenic proinflammatory cytokines/chemokines (IL-8, lipopolysaccharide-induced TNF-a factor, IL-17) levels altering host immunity through immune cell differentiation, proliferation, apoptosis and NO production (139). Reactive nitrogen and oxygen species produced by peripheral leucocytes is an essential defense strategy against pathogens. In fish, dietary supplementation of a grass extract (Cynodon dactylon) to infected Catla catla carp stimulated reactive oxygen and nitrogen species production and decreased mortality in a dose depend manner (140). Screening multiple phytonutrients revealed that the dietary flavonoid naringenin can act as an agonist on the AhR to induce regulatory T cells (Treg) that suppress allergy and autoimmune disease (141). Interestingly, phytochemicals such as indole-3 carbinol (I3C) present in cruciferous vegetables (e.g., broccoli, cabbage) act as ligands for AhR leading to the expansion of the anti-inflammatory IL-22 producing ILCs (142). Functioning AhR has proven to be crucial for immunity because AhR-deficient mice failed to control C. rodentium infections (143). Moreover, mice fed a phytochemical-free diet had a reduced formation of lymphocyte aggregates and follicles, a similar phenotype as seen in AhR-deficient mice (142). Dietary I3C supplementation protected against C. difficile infection through activation of AhR but also through unknown AhR-independent mechanisms likely caused by changes to microbial populations (144).

Anti-adhesion properties are well sought after when studying the direct effects of phytochemicals on pathogen fitness. Cranberry extracts are documented to inhibit pathogenic E. coli adhesins (e.g., fimbriae) limiting their ability to attach to host cells (145, 146). The anti-adhesion activity of cranberry extract is attributed to the polyphenolic flavan-3-ol compounds known as A-type proanthocyanidins (PACs) (147). Cranberry A-type PACs reduced adherence of multiple strains uropathogenic E. coli and Proteus mirabilis in vitro (145). However, in vivo, intestinal and microbial PACs metabolites are found at higher concentrations in urine than the intact PACs and thus may be the bioactive metabolites responsible for the anti-adhesive properties (148). Interestingly, an analysis of urine phytochemical metabolites indicated that they change over-time due to multiple rounds of enterohepatic circulation modifications (148) with poorly understood activities (149). Moreover, cranberry PACs are thought to inhibit host and microbial enzymes (e.g., lipase, glycosidases) protecting against diet-induced obesity (150). PACs are associated with increased Akkermansia sp. abundance; however, it is unknown whether microbiota changes are a direct action of PACs or an indirect result of their effects on host metabolism (151). B-type PACs are known to be less inhibitory to both bacteria and host metabolism (150). Work from our group demonstrates that pea seed coats rich in B-type PACs lead to a significant decrease in the Firmicutes population, increased fecal mucin content, and caused greater pathogen colonization in mice compared to a PAC-poor diet (152). B-type PACs may have led to improper mucus formation leading to a greater concentration of mucin excreted in feces. Phytonutrient supplementation is associated with increases in beneficial Clostridia species and can strengthen mucosal barrier function by increasing mucus production and thickness (153), protecting epithelial cells from invading pathogens and disease. Interestingly, a positive feedback loop may be established between mucolytic bacteria such as Akkermansia sp. that can degrade mucus O-linked glycans, thereby producing SCFAs that could stimulate goblet cells to secrete more mucus (14). Polyphenolic compounds may stimulate the microbiota directly or indirectly through modulation of mucus production, however further research is needed to establish direct links between diet and infection resistance (**Figure 3**).

#### VITAMINS AND MINERALS

Micronutrients are essential for proper metabolic and immune function. Nutrient and mineral deficiencies, typical in those that are critically-ill and in developing countries, can lead to metabolic changes, oxidative damage, immunological defects, weakness, and death (154). The effects of essential minerals, including iron, zinc, copper, selenium, silver, sulfur, calcium, phosphorus, and magnesium have been shown to affect resident microbial populations and health outcomes in both animal and human studies (155). Phagocytes have been shown to utilize the bactericidal actions of copper and zinc to enhance intracellular killing of pathogens (156). For instance, mice fed a zinc-deficient diet and challenged with Enteroaggregative E. coli (EAEC) had reduced leukocyte infiltration and increased virulence factors in luminal content, indicating an impaired immune response and increased infection severity (157). Regular supplementation of vitamin C (1–2 g/day) and zinc (<100 mg/day) reduced the duration of the common cold by 8–14 and 33%, respectively (158, 159). For vitamin C, prophylactic doses >0.2 g/day alleviated respiratory associated problems, particularly in physically strained and stressed individuals, however, its use as a therapy to treat the common cold remains controversial (160). In contrast, zinc supplementation studies support its use as a treatment option to reduce the duration and severity of the common cold (159). Vitamin D had the best overall protection against the common cold, however baseline vitamin D levels and dose must be considered since lower doses and deficient individuals experienced the most benefit (158). More mechanistic research is required to understand the impact of vitamins on immune responsiveness, especially with respect to the microbe-host gut axis in deficient and in excess conditions. Experiments in germ-free, conventionalized and infectious C. rodentium mice models confirm that the microbiota influences vitamin D metabolism by lowering fibroblast growth factor (FGF) 23 through increased activation of TNF-α in the colon (161). The fact that the presence of the microbial community or mono-colonization with C. rodentium increases serum vitamin D levels highlights their role on host homeostasis, especially since vitamin D levels control calcium homeostasis and bone formation (161). Research suggests that proper regulation of vitamins and minerals is key for establishing a proper immune response and intestinal barrier function. Similar to vitamin and mineral deficiencies, excessive supplementation can impair a host ability to resist enteric infections by altering intestinal integrity or enhancing pathogen fitness.

Recently, oral iron and vitamin B12 supplementation are suggested to impair microbiota dependent infection resistance. A systematic review and meta-analysis comprising 6831 adult participants concluded that oral ferrous sulfate (iron) supplementation is associated with a significant increase in gastrointestinal side-effects compared to placebo and intravenous iron delivery (162). This reveals that the effects of iron supplementation are possibly initiated through the microbe-gut axis with unknown consequences and should be used cautiously. For instance, excessive luminal iron affects intestinal integrity through oxygen radical production, encourages pathogen virulence, and alters microbial populations leading to pathogen overgrowth (163, 164). In a dose dependent manner, iron increased epithelial invasion and translocation of S. typhimurium in Caco-2 cells in vitro and reduced the survival of the nematode Caenorhabditis elegans infected with S. typhimuriumi (163, 165). Regulation of luminal iron is extremely important for maintaining intestinal integrity and controlling pathogen expansion (166). Furthermore, lipocalin-2 is a protein produced by neutrophils and epithelial cells during inflammation that directly limits bacterial iron uptake, reducing pathogen overgrowth and severity (167). Unlike iron, vitamin B12 is directly regulated in the gut by intrinsic factors for absorption and in excess, it can escape host absorption and affect microbial competition. The gut commensal bacteria Bacteroides thetaiotaomicron may compete against enterohemorrhagic E. coli (EHEC) to sequester dietary vitamin B12 (168). In vitro competition assays show that B. thetaiotaomicron reduced EHEC shiga toxins but when co-cultured with a mutant B. thetaiotaomicron lacking a vitamin B12 transporter, EHEC had normal shiga toxin production (168). Microbial vitamin B12 transporters have different affinities toward vitamin B12 allowing them to compete with host cells and other microbes to take up exogenous vitamin B12 (169, 170). More research is needed into micronutrient supplementation on host-microbe interactions toward pathogens, especially in the context of over-supplementation, which may be detrimental depending on the micronutrient balance and host intestinal homeostasis. Limiting the expansion of enteric pathogens can be accomplished by reducing their access to vitamin or minerals either through diet or stimulation of gut commensals to compete with pathogen for vital nutrients.

# CONCLUSION

Pathogen resistance and tolerance requires tight host regulation of dietary components and subsequent microbial actions that together influence each other and host immunity. Undigested and unabsorbed dietary components are able to influence microbial populations and their fermentation by-products can indirectly contribute to infection resistance by modulating host intestinal integrity. Dietary intervention studies are difficult to control and compare due to seasonal variations in diets sources. We suggest that dietary intervention studies should include diet backgrounds designed with macro- and micro- nutrients that stress and protect the gastrointestinal environment, as to give a proper assessment of that dietary component on host. In general, a balanced diet of SFA, MUFA, MACs, protein, phytochemicals, vitamins, and minerals with limited sources of n-6 PUFAs, simple carbohydrates, BAPs, and iron may help restore intestinal homeostasis in compromised individuals. Dietary individuality makes it difficult to make general diet recommendations as each individual may have genetic, microbiota, and unforeseen environmental factors that influence diet digestibility and

#### REFERENCES


utilization. Together, these factors ultimately provide the context to which dietary components may influence intestinal integrity and homeostasis.

#### AUTHOR CONTRIBUTIONS

The concept of this review was developed by BW and AF. This review was written by AF and JF, and was edited by BW, AF, and JF.

#### FUNDING

BW was supported by the Canada Research Chair program. JF was supported by a Natural Sciences and Engineering Research Council Postdoctoral Fellowship.


drives salmonella expansion during colitis. PLoS Pathog. (2017) 13:e1006129. doi: 10.1371/journal.ppat.1006129


the effect of fructooligosaccharides in infants with constipation. Nutrients. (2018) 10:E1602. doi: 10.3390/nu10111602


modulate piglet mucosal immunity and colonic microbiota. ISME J. (2014) 8:1609–20. doi: 10.1038/ismej.2014.10


receptor can generate regulatory T cells. J Immunol. (2010) 185:3190– 8. doi: 10.4049/jimmunol.0903670


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

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

# Infectious Threats, the Intestinal Barrier, and Its Trojan Horse: Dysbiosis

Simona Iacob1,2 and Diana Gabriela Iacob<sup>1</sup> \*

1 Infectious Diseases Department, "Carol Davila" University of Medicine and Pharmacy, Bucharest, Romania, <sup>2</sup> National Institute of Infectious Diseases "Prof. Dr. Matei Bal ¸s", Bucharest, Romania

The ecosystem of the gut microbiota consists of diverse intestinal species with multiple metabolic and immunologic activities and it is closely connected with the intestinal epithelia and mucosal immune response, with which it builds a complex barrier against intestinal pathogenic bacteria. The microbiota ensures the integrity of the gut barrier through multiple mechanisms, either by releasing antibacterial molecules (bacteriocins) and anti-inflammatory short-chain fatty acids or by activating essential cell receptors for the immune response. Experimental studies have confirmed the role of the intestinal microbiota in the epigenetic modulation of the gut barrier through posttranslational histone modifications and regulatory mechanisms induced by epithelial miRNA in the epithelial lumen. Any quantitative or functional changes of the intestinal microbiota, referred to as dysbiosis, alter the immune response, decrease epithelial permeability and destabilize intestinal homeostasis. Consequently, the overgrowth of pathobionts (Staphylococcus, Pseudomonas, and Escherichia coli) favors intestinal translocations with Gram negative bacteria or their endotoxins and could trigger sepsis, septic shock, secondary peritonitis, or various intestinal infections. Intestinal infections also induce epithelial lesions and perpetuate the risk of bacterial translocation and dysbiosis through epithelial ischemia and pro-inflammatory cytokines. Furthermore, the decline of protective anaerobic bacteria (Bifidobacterium and Lactobacillus) and inadequate release of immune modulators (such as butyrate) affects the release of antimicrobial peptides, de-represses microbial virulence factors and alters the innate immune response. As a result, intestinal germs modulate liver pathology and represent a common etiology of infections in HIV immunosuppressed patients. Antibiotic and antiretroviral treatments also promote intestinal dysbiosis, followed by the selection of resistant germs which could later become a source of infections. The current article addresses the strong correlations between the intestinal barrier and the microbiota and discusses the role of dysbiosis in destabilizing the intestinal barrier and promoting infectious diseases.

Keywords: dysbiosis, intestinal microbiota, epigenetic modulation, short chain fatty acids, immunity, sepsis, HIV infection, liver infections

#### Edited by:

Sudhanshu Shekhar, University of Oslo, Norway

#### Reviewed by:

Elisabeth Billard, Université Clermont Auvergne, France Eliana Mariño, Monash University, Australia

> \*Correspondence: Diana Gabriela Iacob dianagiacob@gmail.com

#### Specialty section:

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

Received: 14 January 2019 Accepted: 08 July 2019 Published: 07 August 2019

#### Citation:

Iacob S and Iacob DG (2019) Infectious Threats, the Intestinal Barrier, and Its Trojan Horse: Dysbiosis. Front. Microbiol. 10:1676. doi: 10.3389/fmicb.2019.01676

# INTRODUCTION

fmicb-10-01676 August 5, 2019 Time: 13:8 # 2

The intestinal barrier defines the morpho-functional unit responsible for the defense of the intestinal mucosa and consists of the intestinal microbiota, intestinal epithelial cells (IECs) and mucosal immunity tightly linked through a complex network of cytokines, antimicrobial peptides (AMPs), metabolic products, and numerous regulatory molecules (Meng et al., 2017). Given that the intestinal mucosa is the largest body surface at risk of infectious threats, the anatomic and functional homeostasis of the intestinal barrier is a key step in the anti-infectious defense of the human organism.

The intestinal microbiota represents the first defense line of the intestinal barrier. The microbiota entails millions of microogranisms colonizing the gastrointestinal tract most of which are bacteria. This large number of microorganisms withstand the unfavorable intestinal habitat thanks to their symbiotic relationships with the human organism. These symbiotic host-commensal relationships develop after birth and enable the metabolic, immune and antiinfectious processes through which the microbiota contributes to gut homeostasis (O'Hara and Shanahan, 2006). The structural and functional stability of commensal populations is regulated through numerous signaling molecules (quorum sensing) and cellular regulators (miRNAs) as well as through other physiologic and pathologic factors. Qualitative or quantitative alterations of this microbial community broadly defined as dysbiosis impair the relationships between the host and commensal species, modify the balance between commensals and pathogens, decrease the intestinal barrier protection and favor infectious pathogens (McDonald et al., 2016). Consequently, the microbiota loses its anti-infectious role and becomes the weak link responsible for persistent infections.

The article discusses gut barrier defense mechanisms, the key anti-infectious role of microbiota inside of the gut barrier and the impact of dysbiosis in the life-threatening infections.

#### THE GUT BARRIER AND ANTIBACTERIAL DEFENSE MECHANISMS

#### Commensal Flora – The First Line of Defense

The intestinal microbiota encompasses all microbial species populating the gastrointestinal tract. Molecular techniques in healthy individuals have revealed a diverse ecosystem, containing nine bacterial phyla, out of which four are dominant: Firmicutes, Bacteroidetes, Actinobacteria, and Proteobacteria (Rajilic-Stojanovi ´ c et al., 2007 ´ ). The structure of the microbiota is gradually delineated after the age of three through symbiotic relationships with the organism and ensures the intestinal dominance of specific commensals ("symbionts") belonging to the Firmicutes and Bacteroidetes phylum or Bifidobacteriales order. Symbionts will compete with pathogens or potential pathogenic germs ("pathobionts") for pre-existent intestinal niches ("niche competition") as well as for intestinal nutrients ("nutritive competition"). This competition ensures the structural stability of the microbiota and is referred to as "colonization resistance." Colonization resistance employs a network of specific molecules with a critical anti-infectious roles (Sassone-Corsi and Raffatellu, 2015; Sorbara and Pamer, 2019). Among these are molecules with important metabolic and antimicrobial roles such as short chain fatty acids (SCFAs) and bacteriocins.

Short chain fatty acids are end products of the anaerobic fermentation of intestinal microbiota and the major energy source for colonocytes (Roediger, 1980). SCFAs effects are mediated by G protein-coupled receptors (GPRs) (GPR41, GPR43, and GPR109A) expressed on immune cells and a variety of tissues including gut epithelial cells. The predominant SCFAs are present at high mM levels in the colon (butyrate), entero-hepatic circulation (propionate), and systemic circulation (acetate) and are responsible for epithelial protection and regulation of the inflammatory intestinal response, reviewed in Donohoe et al. (2011) and Fukuda et al. (2011). SCFAs ensure a low antibacterial pH around colonic cells, favor mucus synthesis, and contribute to IECs integrity through the upregulation of tight junction proteins, stabilization of HIF transcription factor, IL-18 release, and NLR pyrin domain 3 (NLRP3) inflammasome modulation (Wang et al., 2012; Kelly et al., 2015; Macia et al., 2015; Feng Y. et al., 2018). At the same time SCFAs contribute to the antibacterial defense against pathogenic species through neutrophil recruitment, the release of cytokines and AMPs while also inducing the intestinal immunotolerant response against commensal species as further discussed.

Gut bacteria also, particularly Firmicutes, Proteobacteria, and Bacteroidetes, synthesize multiple microbicidal molecules (bacteriocins) with a broad spectrum of activity (Todorov et al., 2014; Drissi et al., 2015). Bacteriocin-producing strains such as lactic acid bacteria are commonly used as probiotics in the food industry (Parada et al., 2007). Certain bacteriocins (nissin, pediocin) have been approved for oral or topical use and others are studied as oral alternatives to antibiotics (reviewed in Cavera et al., 2015). Nevertheless, the oral use of bacteriocins is still uncertain due to a lack of data on their action mechanisms as well as on their efficiency, cytotoxicity, stability, and immunogenicity.

Dysbiosis modifies not only the balance between commensals and pathogens but also the release of antimicrobial molecules. In turn this disturbs the process of colonization resistance and allows the invasion of the intestinal epithelium by various pathogens.

### The Intestinal Epithelium – The Second Anti-infectious Defense Line

The intestinal epithelium has its own defense mechanisms, both structural and functional. The intestinal epithelial monolayer includes several subsets of IECs united through tight apical junctions, externally covered with a mucus layer. IECs come into direct contact with the lamina propria and immune cells (Goto and Ivanov, 2013). All intestinal epithelial lineages arise from HOPX-quiescent stem cells with immunoregulatory

and tumor suppressing properties (Takeda et al., 2011). IECs include entero-absorptive enterocytes, mucus secreting goblet cells, antigen-sampling cells (M cells), and Paneth cells, each exhibiting specific surface receptors.

The IECs express specific receptors namely "pattern recognition receptors" (PRRs) such as Nod-like receptors (NLRs), toll-like receptors (TLRs) and other PRR families located on the cell membrane or in the cytoplasm which recognize specific microbial-associated molecular patterns (MAMP). Once the activation of these receptors induces cytoplasmic signal transduction cascades and further promotes the NF-kappaB (NF-kB) pathway along with other cellular transcription factors, inflammasome activation and pro-inflammatory cytokines (IL-17, IL-18, IL-22) (Hirota et al., 2011; Jones and Neish, 2011; Levy et al., 2015). The IECs and particularly Paneth cells residing at the bottom of the intestinal crypts release AMPs (human-cathelicidin, defensins) with microbicidal, immunomodulatory and wound-healing properties. Thus IECs further contribute along with microbial colonization resistance to the stabilization of the gut barrier. Therefore Paneth cell MyD88 expression is an essential mechanism for the restriction of intestinal translocation and penetration by enteric commensals or pathogens (Vaishnava et al., 2008).

Short chain fatty acids and GPR43/109A stimulation also protects epithelial integrity via inflammasome activation and epigenetic immunomodulation of FoxP3+ regulatory T cells (Treg) proliferation as shown below (Macia et al., 2015).

Dysbiosis induced by infectious or non-infectious causes (ischemia, inflammation, tumors, or various treatments) favors the distruption of the intestinal epitelium (Kitajima et al., 1999; Nagpal and Yadav, 2017; Stewart A. S. et al., 2017). The injured epithelium permits the translocation of various intestinal bacteria and of their toxins including endotoxins (lipopolysaccharide, LPS). In turn these translocations lead to bacteraemia, endotoxemia and life-threatening infections (sepsis, primitive peritonitis, portal encephalopathy in cirrhotic patients, or immune activation in human immunodeficiency virus (HIV) infection.

#### The Intestinal Immunity – The Third Anti-infectious Defense Line

The intestinal epithelium represents the largest epithelial surface in the human body. It connects with the intestinal ecosystem containing commensal and pathogenic species. Commensal germs need to be immunologically accepted, whereas intestinal pathogens must be eliminated. This differentiation requires the activation of an extremely well-coordinated and efficient immune system. Considering that intestinal commensals are present in the human body ever since birth they represent, in fact, the first contact between the immune system and the exterior environment. As such, this large and diverse microbiota will serve as one of the body's defense mechanisms against the invasion of pathogenic germs. At the same time, the microbiota develops mutually beneficial relationships with the organism and together with the intestinal epithelium and gut-associated lymphoid tissue (GALT) the microbiota forms a complex intestinal barrier against infectious threats.

These symbiotic relationships are facilitated by the presence of PRRs and by the immunomodulatory capacity of the intestinal microbiota against the antigenic structures exposed by the intestinal microbiota.

#### The Role of Cellular Receptors

The intestinal epithelium expresses numerous types of receptors able to recognize different MAMPs expressed by microbial species and to convert them into gut signals for inflammatory cascades (Kim et al., 2004). Thus TLR 1,2,4,5,6 (extracellular sensors) and NLR1,2 and TLR9 (cytoplasmic sensors) are expressed on epithelial cells and act complementary, promoting both innate and adaptive immunity (Kim et al., 2004; Abreu et al., 2005). MAMPs are expressed by commensal species since birth and examples of MAMPs include the LPS found in the outer membrane of Gram-negative bacteria, lipoteichoic acid present on the gram positive bacteria wall, peptidoglycan, a component of the bacterial wall, flagellin, a component of intestinal flagellated bacteria, or release outer membrane vesicles that function as PRR ligands (Cañas et al., 2018). PRRs activation in the presence of MAMPs stimulates the NF-kB signal transduction pathway, induces pro-inflammatory interleukins and enables an innate immune response lastly maintaining a state of controlled inflammation (Hayashi et al., 2001; Bambou et al., 2004; Vora et al., 2004; Rhayat et al., 2019). In the absence of MAMPs or of specific receptors (germ free animals or genetic mutations of receptors) the organism fails to detect microbial antigens and does not mount adequate defense mechanisms (Abreu et al., 2005). Hence, it is considered that the rich and diverse commensal flora plays an active role in the proper development of the immune system since birth.

Antigen-presenting cells (APC) belonging to GALT such as macrophages and dendritic cells (DCs) also exhibit cell receptors which regulate numerous genes and modulate the release of NFκB transcription factor, immunomodulatory cytokines and AMP. Depending on the type of activated receptor DCs may generate a Th1/Th17 pro-inflammatory response. Thus, to exemplify, the NLR2 induced by commensal bacteria appears to play a central role in the downregulation of the GALT inflammatory activity and the ability of DCs to induce a polarized CD4+Th1 response in mice and human experiments (Butler et al., 2007; Barreau et al., 2010).

By activating the IECs or APCs receptors, the commensal flora induces a tolerogenic response in DCs, characterized by the release of immunosuppressive cytokines (IL10, IL4, TGFβ), promotion of Treg cells and of CD4+Th2 phenotype in the periphery (Iwasaki and Kelsall, 1999; Stagg et al., 2003; Atarashi et al., 2013; Maharshak et al., 2013; Martin-Gallausiaux et al., 2018). The induction of mucosal immune tolerance protects the epithelia from a detrimental inflammatory immune response and contributes to the immune system's maturation (Jung et al., 2019). Consequently any disturbance involving this process plays an essential role in intestinal inflammatory processes.

The activation of cellular receptors is intricately linked to the pro-inflammatory response of mesenteric DCs and is particularly

involved in the maintenance of the Th17/Th1 response directed against pathogens (reviewed in Sorini et al., 2018). What is interesting is that DCs may activate Treg cells to certain antigens such as LPS of the commensal flora but not LPS of the pathogenic flora (Shirai et al., 2004). Thus the stimulation of mice with E. coli-derived LPS activates inflammatory mechanisms (IL-12 production, Th1 response) whereas the stimulation with LPS derived from Poprhyromonas gingivalis induces IL-4 production and a Th2 response (Pulendran et al., 2001). Similarly TLR recognize flagellin antigens of pathogenic Salmonella species and trigger an inflammatory response against them, while commensal species lacking this antigen do not induce an inflammatory response (Gewirtz et al., 2001).

All these aspects evince that cell receptors represent a filter for intestinal signals and play a deciding role in the induction of specific defense mechanisms adapted to the commensal flora. This role of cell receptors is indeed significant considering the density and variety of the intestinal bacteria. Dysbsiosis eliminates commensal species that trigger Treg-cell polarization and therefore leads to an excessive Th1 or Th2 response, further promoting an inflammatory or autoimmune response.

#### The Immunomodulatory Role of GALT

Gut-associated lymphoid tissue is the most important lymphatic network in humans and involves isolated or aggregated lymphoid follicles (Peyer's Patches), intra-epithelial lymphocytes, macrophages, DCs, mesenteric ganglia, secretory IgA (sIgA) cells and lymphatics.

Peyer's patches localized in the mucosa and submucosa of the small intestine are covered by a "follicle-associated epithelium" containing specialized "M cells". These cells engulf and transport antigens from the intestinal lumen to intestinal DCs where T cell lymphocytes (LT) are primed (Alpan et al., 2001). LT subsequently return to the intestinal lymphatic compartment for their effector function. At the same time, M cells initiate mucosal sIgA production and humoral responses (Rios et al., 2016). Thus, Peyer's patches directly mediate the interaction between the intestinal flora and the humoral or cellular immune response. Enteric pathogens are adapted to invade and destroy follicleassociated epithelium, especially M cells, further interfering with T-cell differentiation and the immune response; these germs adhere to the intestinal epithelium through different mechanisms (Phillips et al., 2000), invade the enterocytes (Kühbacher et al., 2018), or invade and destroy M cells (Clark et al., 1994; Autenrieth and Firsching, 1996; Penheiter et al., 1997; Corr et al., 2006).

Gut-associated lymphoid tissue also harbors a tolerogenic DCs population namely CD103+DCs involved in Foxp3 expression and intestinal conversion of naïve CD4+T cells into Treg cells. Commensal flora stimulates CD103<sup>+</sup> DCs through the activation of specific receptors or through its own metabolic products such as SCFAs (Furusawa et al., 2013; Nastasi et al., 2015; Kaisar et al., 2017). Thus, butyrate stimulates intestinal DCs, ensures intestinal immune tolerance through IL-10 release (Liu et al., 2012) and T cell polarization toward Treg cells while maintaining the balance between the immunosuppressive IL-10-secreting CD4+T cells, IL-17-secreting Th17 cells, and the CD4+Th1 effector cells (Arpaia et al., 2013; Furusawa et al., 2013). Additionally, SCFAs attenuate the excessive inflammatory response induced by LPS-producing Gram negative bacteria (Cox et al., 2009).

The dysregulation of the Th1/Th17 immune response against commensal species is associated with intestinal inflammatory diseases. On the other hand the excessive polarization toward a Treg response attenuates the immune response to infections (Brenchley et al., 2004). Another example of GALT dysregulation occurs in HIV due to the destruction of the intestinal gut barrier and the ensuing chronic inflammatory response. GALT depends on the microbiota and it is worth noting that the development of GALT and the activation of T cell lymphocytes and B cell lymphocytes cannot occur in the absence of signals released by the intestinal flora.

Experiments on commensals belonging to Clostridium and Bacteroides species further highlighted the ability of the intestinal microbiota to maintain the intestinal homeostasis and to orchestrate an adequate T-cell response through specific MAMPs during bacterial invasion (Ivanov et al., 2008; Lécuyer et al., 2014). Thus, Bacteroides fragilis strains expressing polysaccharide A binding TLR2 on CD4+T cells and gut-indigenous Clostridium belonging to XIVa and IV clusters favor Tregs, suppress Th-17 cells and facilitate mucosal tolerance toward the colon microbiota (Mazmanian et al., 2008; Round et al., 2011; Atarashi et al., 2013). On the other hand segmented filamentous bacteria, a genetic relative of the genus Clostridium promote both effector Th17 CD4+T cells (Ivanov et al., 2009; Farkas et al., 2015; Schnupf et al., 2017) and sIgA antibodies during intestinal invasion (Lécuyer et al., 2014). Other microbial species also modulate the Treg/Th17 axis, potentially controlling the intestinal inflammation and tolerance (Pandiyan et al., 2019). This aspect also explains the disequilibrium between these species during dysbiosis as well as the expansion of pathogens and of mucous lesions.

#### Epigenetic Regulatory Implications

Modulation of gene expression by epigenetic mechanisms in the intestinal environment was studied especially for neoplasms and inflammatory processes (Vdovikova et al., 2018). Epigenetic mechanisms modulate gene transcription by various processes: DNA methylation, histone modifications and modulation of long non-coding RNA and microRNA expression. However the mechanisms through which bacteria are either affected or induce gut epigenetic changes are not well understood.

Histone modifications are correlated with the activation or repression of genetic transcription through the modulation of two antagonistic enzymes, namely histone acetyl transferases (HATs) inducing histone acetylation, and histone deacetylases (HDACs) inducing histone deacetylation. During acetylation, the chromatin structure loosens and can be accessed by transcription factors. HDAC inhibitors increase histone acetylation and subsequently regulate gene expression in numerous immune cells such as epithelial cells, neutrophils, APCs and T cells. On the contrary, deacetylation of histones by HDACs prevents gene transcription. Thus studies performed in vitro showed that HDAC inhibitors promote the release of several transcription

factors such is NFκB, MyoD, p53, or HSP90 (Vinolo et al., 2011; Kumar S. A. et al., 2018; Banik et al., 2019). The NF-kB imbalance has been correlated with numerous inflammatory and antiapoptotic mechanisms that interfere with viral evasion (Le Negrate, 2012; Carrasco Pro et al., 2018), septic shock (Liu et al., 1999; Han et al., 2002), or inflammatory diseases (Makarov, 2001; Lawrence, 2009). Likewise, p53 regulates the cell cycle and apoptosis, functioning as a tumor suppressor (Aubrey et al., 2018). P53-mediated apoptosis was associated with the spread of viral infections (Lazo and Santos, 2011; Aloni-Grinstein et al., 2018), while HSP90 reverse transcriptase mediated activity was, most likely, associated with extensive hepatitis B virus (HBV) infection (Hu and Seeger, 1996). Several viruses among which HIV, HBV, hepatitis C virus, Epstein Barr virus, cytomegalovirus, herpes simplex virus and human T-lymphotropic virus have NFkB activation strategies and cell apoptosis blockage (Santoro et al., 2003) while others, especially the oncogenic ones have p53 suppression mechanisms (Sato and Tsurumi, 2013). It has been demonstrated that HDAC inhibitors could block stellate cell activation thus hindering liver fibrosis in experimental animal models (Park et al., 2014; Ding et al., 2018).

By inhibiting HDAC, SCFAs could therefore control numerous infectious and immune processes (Vinolo et al., 2011; Zhou et al., 2017b; Sun et al., 2018). Although the role of HDAC inhibitors hasn't been completely elucidated yet, most experimental studies evinced an anti-inflammatory dose dependent effect (Yin et al., 2001; Le Poul et al., 2003; Weber and Kerr, 2006; Asarat et al., 2016; Kaisar et al., 2017; Li et al., 2018) and an immunosuppressive role of SCFAs in tumors (Villagra et al., 2010; Tang et al., 2011). Still, while other HDAC inhibitors are already used in oncologic therapies, the exact role of SCFAs in the immune and tumor processes remains purely theoretical (Meijer et al., 2010; Matthews et al., 2012; Ulven, 2012).

MicroRNAs (miRNAs) are small, evolutionary conserved non-coding RNAs of approximately 19-23 nucleotides involved in the post-transcriptional regulation of cellular mRNAs. The biogenesis of mature miRNA includes a two-step cleavage process from primary miRNAs (pri-miRNA). The mature miRNA is then loaded into the effector complex RNA-induced complex (RISC). RISC interacts with target mRNA and induce mRNA cleavage or translational repression hence controlling diverse metabolic or cellular pathways including cell cycle progression, differentiation, apoptosis, immune regulation or oncogenesis (Dalmasso et al., 2011; Singh et al., 2012; Nakata et al., 2017). miRNAs are released extracellularly by most eukaryotic cells and various types of small non-coding RNAs (sRNAs) around 50– 200 nt in length have been observed within extracellular vesicles released by gram negative bacteria ("outer membrane vesicles") (Gong et al., 2011). The general conception is that sRNA, as well as microRNA-size small RNAs (msRNAs) of 15–25 nucleotides in length such as msRNA observed in the model bacterium E. coli (Kang et al., 2013) act as post-transcriptional regulators and function as signaling molecules for bacterial growth and virulence under experimental conditions (Ortega et al., 2012; Zhao et al., 2017) and virulence mechanisms, at least under experimental conditions (Padalon-Brauch et al., 2008; Choi et al., 2017). The role of these structures in intestinal homeostasis, has been, for now, scarcely covered, most information having been acquired from studies on conventional or germ-free mice.

Intestinal miRNAs are released by IECs (Liu et al., 2016) and are regulated by intestinal microbiota through TLR/MyD88 dependent pathway (Dalmasso et al., 2011; Singh et al., 2012; reviewed in Eulalio et al., 2012; Williams et al., 2017). Liu and Weiner (2016) conducted an extensive study identifying numerous types of extracellular miRNAs circulating in exosomes in the gut lumen and feces of mice and humans. The study proved that IECs and Hopx-expressing cells are the main sources of exosomal intestinal miRNAs and also highlighted their uptake by intestinal bacteria and potential role in post-transcriptional regulation of bacterial genes.

Studies on mice indicate a certain pattern of miRNA gut compartimentalization after bacterial colonization (Dalmasso et al., 2011) and reported the dominance of miR-143,-145 in the jejunum and cecum and of miR-200b in the large intestine and caecum (Singh et al., 2012). On the other hand the human cells or animal experimental models with pathogenic species (Helicobacter pylori, Citrobacter rodentium, Listeria monocytogenes, Francisella tularensis, and Salmonella enterica) induce a different miRNA panel mainly represented by miR-155 and miR-146 (Eulalio et al., 2012; Archambaud et al., 2013; Staedel and Darfeuille, 2013).

On a molecular level, there is a growing interest for the intestinal role of miRNAs and their regulatory implications in the gut-barrier functionality but data on this topic is divergent, scarce and fragmented. Cellular and extracellular miRNAs from the intestinal lumen modulate the epithelial integrity, inflammatory response, and probably bacterial gene mRNAs through insufficiently known mechanisms (reviewed in Belcheva, 2017).

Studies on human intestinal cells have documented the miRNA importance in the protection of epithelial tight junctions (e.g., miRNA-122) (Ye et al., 2011), epithelial regeneration (e.g., miR-143, miR-145) (Chivukula et al., 2014), and proliferation (miR-30 family members) (Peck et al., 2016), the modulation of epithelial integrity (e.g., miR-122) (Ye et al., 2011), mucin gene expression (Mo et al., 2016), and epithelial permeability (e.g., miR-21-5p) (Nakata et al., 2017).

Intestinal miRNAs are also key regulators of the immune response against infections. Host miRNAs orchestrate the immune response through PRR families and TLR signaling pathways (miR-146a) (Xue et al., 2011, 2014), while the microbiota downregulates the expression of miR-10a and miR-107 in host DCs, decreases the release of proinflammatory cytokines in mice and controls the excessive inflammatory response in human and mice (e.g., miRNA-146a and miR-193a-3p) (Taganov et al., 2006; Nahid et al., 2011; Singh et al., 2012; Dai et al., 2015). MiRNAs could also potentially influence the microbiota and alleviate colonic inflammation as was suggested by a negative correlation on ulcerative colitis between the release of miR-193a-3p and colonic inflammation (Dai et al., 2015). This concept was further explored in a study in which the oral ingestion of endotoxin in mice led to the upregulation of epithelial miR-146 and promoted innate immune tolerance and epithelial protection in the postnatal

period (Chassin et al., 2010). On a similar note mi-RNA-155 defective mice failed to develop a protective immunity toward H. pylori or Salmonella Typhimurium (Rodriguez et al., 2007).

Pathogenic bacteria also release membrane or outermembrane vesicles containing sRNAs that modulate the host miRNA profile and gene expression (Gu et al., 2017). Some studies on intestinal inflammation due to enteropathogens revealed the upregulation of miR-16,-21,-223,-594 and miRNA-31 or downregulation of miR-124 within the human intestinal lumen (Wu et al., 2010; Koukos et al., 2013; Lin et al., 2014). Their biological relevance is under study.

Various miRNAs have been correlated with intestinal oncogenesis according to human studies. In this respect miR-182,-503 and miR-17∼92 clusters modulate glycan production and correlate with the growth of certain bacteria which could potentially initiate the microenvironmental changes in colorectal cancer (Yuan et al., 2018). Gut miRNAs could therefore induce and perpetuate dysbiosis favoring various infections or intestinal cancers (Liu et al., 2016; Williams et al., 2017). Furthermore, single nucleotide polymorphisms of miRNA-146a,-27a genes promote various infections associated with inflammatory and neoplasic intestinal changes (Song et al., 2013; Shao et al., 2014). Butyrate was also shown to affect miRNA-106b expression in IECs and to modulate carcinogenesis (Hu et al., 2011). Therefore, it is probable that epigenetic regulators such as miRNAs could play a key role in the interaction between host and microbiota and miRNA dysregulations.

## DYSBIOSIS, THE TROJAN HORSE INSIDE THE GUT BARRIER

#### Dysbiosis and Enteral Infections

Enteric pathogens alter the intestinal barrier, antagonize the intestinal microbiota, and trigger enteral infections through various mechanisms including increased intestinal inflammation, the release of bacteriocins and upregulation of AMPs and toxin delivery secretory systems (T6SS, T3SS) as well as the exploitation of nutrients or intestinal niches (reviewed in Rolhion and Chassaing, 2016; Sorbara and Pamer, 2018).

Dysbiosis occurs as a consequence of different enteroinvasive or entorotoxinogenic species which distrupt colonization resistance, reduce the protective species belonging to the Bacteroidetes and Firmicutes phylum (especially the Clostridiales incertae sedis XI or the IV/XIVa cluster) and allow the overgrowth of species belonging to the Proteobacteria phylum (especially Enterococcaceae and Enterobacteriaceae families) (Vincent et al., 2013; Livanos et al., 2018). Antibiotic treatment of intestinal infections additionally modifies the structure of the microbiota and drastically disturbs the process of "colonization resistance" (Panda et al., 2014).

As a result, the first alterations due to dysbiosis include the loss of commensals synthesizing bacteriocins and SCFAs such as Bacteroides thuringiensis and the Lachnospiraceae and Ruminococcaceae families (Rea et al., 2010) along with the loss of Clostridia commensals competing with pathobionts for the same intestinal niche (Sailhamer et al., 2009; Buffie et al., 2015; Geerlings et al., 2018) and of Bacteroides thetaiotaomicron and Akkermansia muciniphila which protect the gut barrier (Hooper et al., 2001; Donato et al., 2010; Martín et al., 2019). In these cases, the normally scarce pathobionts eventually become the dominant species and display various virulence mechanisms. Consequently, pathobionts or their endotoxins trigger intestinal infections, gut sepsis or postantibiotic colitis (Brown et al., 2013; Pérez-Cobas et al., 2013). Moreover, antibiotic treatment favors the development of multidrug-resistance species (MDR) and in turn these become invasive and pathogenic after extraluminal translocation (Hirakata et al., 2002; Ayres et al., 2012). Hence, Pseudomonas MDR species with efflux pump systems like MexAB-OprM develop quorum sensing machinery to sense host stress and express multiple virulence determinants (Hirakata et al., 2002). The wide distribution of efflux pump systems in MDR intestinal species after antibiotic treatment increases the virulence of these species and their associated risk of extraluminal translocation (Nishino et al., 2006; also reviewed in Nikaido, 1996). Therefore dysbiosis may be both the cause and the consequence of enteral infection.

The restoration of the microbiota is particularly difficult and is best ensured through fecal microbiota transplantation (FMT) from a healthy donor. The effectiveness of FMT in the management of C. difficile enterocolitis or sepsis dysbiosis further underlines the critical role of the microbiota for gut health (Gough et al., 2011; Rao and Young, 2015; Wei et al., 2016).

#### Dysbiosis and Sepsis

Sepsis is a life-threatening organ dysfunction caused by a dysregulated inflammatory response to infectious agents or their proinflamatory products. Sepsis has a progressive and fatal course to generalized microvascular injury, cellular hypoxia, and shock (Singer et al., 2016).Under certain conditions enteric bacteria could elicit a dysregulated systemic inflammatory response causing sepsis (Deitch et al., 1994; Mainous et al., 1995). The germs most commonly encountered in gut-derived sepsis are Gram negative bacilli producing either endotoxins or poreforming exotoxins (Wallace et al., 2000) and probably one of the most common species involved in these cases is E. coli, a common gut resident (MacFie et al., 1999). The presence of poreforming exotoxins is followed by the efflux of cellular potassium and activation of NLRP3 inflammasome. In turn these contribute to cellular apoptosis and loss of epithelial integrity and permit the access of toxins and pathogens to the portal and systemic circulation. This further stimulates the releases of various proinflammatory and blood coagulation mediators and a cytokine storm followed by vasoconstriction, ischemia and peripheral necrosis and finally organ dysfunction and shock (reviewed in Los et al., 2013; Sonnen and Henneke, 2013). The excessive activation of the CD14/TLR4/MD2 complex by bacterial endotoxins also induces the inflammasome signaling pathway, along with caspase activation (Hotchkiss et al., 1999), cells lysis and disproportionate release of pro-inflammatory cytokines, further eliciting and exacerbating sepsis, reviewed in Lamkanfi (2011), Gao et al. (2018), and Skirecki and Cavaillon (2019).

Sepsis alters the intestinal barrier through multiple factors which ultimately promote dysbiosis including intestinal ischemia

and inflammatory lesions, aggressive care, antibiotic treatments, intestinal comorbidities, parenteral nutrition, etc. (reviewed in Hassoun et al., 2001; Alverdy and Luo, 2017; Fay et al., 2017). Gut vascular dysfunctions in sepsis and particularly intestinal ischemia stimulates pro-inflammatory cytokines, activates HIF-1 alpha factor, disrupts the epithelial tight junction and finally induces colonic cell apoptosis (Hassoun et al., 2001; Li et al., 2009; Yoseph et al., 2016). All of these events increase the permeability of intestinal epithelia and promote intestinal translocations (Diebel et al., 2003). The intestinal translocation of bacteria or microbial products such as endotoxin-LPS or MAMPs perpetuates and aggravates the systemic pro-inflammatory response further speeding the progression to organ failure and death (Shimizu et al., 2006, 2011; also reviewed in Deitch and Berg, 1987; Doig et al., 1998; Meng et al., 2017).

Dysbiosis entails the decrease of anaerobes Bifidobacterium and Lactobacillus (the main producers of SCFAs) accompanied by a detrimental increase of intestinal pathobionts such as E. coli or Staphylococcus and Pseudomonas species (Shimizu et al., 2006, 2011; Hayakawa et al., 2011; Ayres et al., 2012) replacing the protective microbiome. Additionally, the reduction of SCFAs favor the cascade release of LPS-induced pro-inflammatory mediators, LPS-triggered macrophage migration and massive release of nitric oxide by neutrophils (PMN), thus promoting intestinal inflammation and destabilizing the gut-barrier (Maa et al., 2010; Wang et al., 2017). Certain strains or specific phylogenetic groups of intestinal Gram negative bacilli (E. coli and Pseudomonas aeruginosa) acquire additional virulence genes during colonization or even change their morphotype after translocation to express virulence genes and to avoid host defense mechanisms giving rise to gut-derived sepsis (Johnson et al., 2001; Zaborina et al., 2007; Hickey et al., 2018). Thus, dysbiosis induces and supports sepsis associated mechanisms. In favor of this hypothesis, Souza et al. (2004) administered LPS to germ free mice and reported the absence of a severe inflammatory response and low mortality. Literature data on sepsis dysbiosis in newborns has confirmed significant changes of the microbiome, a decreased bacterial diversity, pre-sepsis gut colonization with invasive species (Carl et al., 2014), intestinal translocations (Madan et al., 2012; Mai et al., 2013; Taft et al., 2015; Stewart C. J. et al., 2017) or the activation of virulence and antibiotic resistance factors in intestinal germs (Mittal and Coopersmith, 2014). Of note, these alterations continue to occur throughout the progression of sepsis.

The administration of probiotics in sepsis (Lactobacillus rhamnosus and Bifidobacterium longum) decreases epithelial apoptosis as well as the release of cytokines and bacterial translocations in experimental mice models (Khailova et al., 2013; Panpetch et al., 2017). Probiotics can also restore the lactic acid-producing flora and ensure colonization resistance toward pathogenic flora. According to Haak and Wiersinga (2017), probiotics further influence the prognosis in sepsis by downregulating colonic TLR-2/TLR-4 via MyD88 and by mitigating the systemic pro-inflammatory response. However, their benefit to reduce the probability of sepsis in critically ill patients is questionable (Jain et al., 2004; Kotzampassi et al., 2006; Jacobi et al., 2011; Panigrahi et al., 2017). On the other hand, data on prebiotic supplementation in infections is discordant and clear recommendations are missing (Srinivasjois et al., 2013; Chi et al., 2019).

#### Dysbiosis and Liver Infections

The liver is connected to the intestine through the portal vein, biliary tract, and numerous signaling molecules, together building the "gut-liver axis." As a result, the liver is continuously exposed to microbial and metabolic molecules produced by intestinal bacteria (SCFAs, MAMPs) and further releases its own antibacterial products (bile salts) involved in the intestinal homeostasis (reviewed in Macpherson et al., 2016). Thus intestinal MAMPs reach the portal system and hepatic sinusoids and are further processed by Kuppfer cells and liver-resident T cells. SCFAs released by the microbiota are absorbed in the colon and filtered intrahepatically where most modulate the immune response, oncogenesis and are involved in the epigenetic control of liver pathogens through HDAC inhibition and transcriptional changes (Candido et al., 1978). As HDAC inhibitors, SCFAs decrease HBV replication (Pollicino et al., 2006), mediate hepatitis C virus replication (Taguwa et al., 2008, 2009), impede liver necrosis and prevent hepatocellular carcinoma, reviewed in Koumbi and Karayiannis (2015). Recently, butyrate was shown to regulate the expression of tumorsuppresive miRNAs (miRNA-26a,-26a-1,-192, etc.) through the bile acid nuclear receptor, farnesoid X receptor (FXR) while also promoting hepatocyte apoptosis through miR-22 upregulation in hepatic cells (Pant et al., 2017; reviewed in Feng Q. et al., 2018).

Host liver miRNAs are able to regulate the replication of viral hepatitis yet the precise roles of intestinal miRNAs in the regulation of hepatotropic viruses remain unclear (Bandiera et al., 2016; Feng Q. et al., 2018; reviewed in Fan and Tang, 2014; Li et al., 2016). Hepatocytes also release immunoregulatory molecules such as primary bile acids (BA). A small proportion of primary BA are dehydroxylated in the intestine through bile salt hydrolases and converted to secondary BA entering the enterohepatic circulation. The latter exert a bacteriostatic effect on the intestinal barrier and an anti-fibrotic and anti-inflammatory intrahepatic activity (Wu et al., 2012).

Dysbiosis is closely related to liver pathology. The liver provides an immunotolerogenic environment with an attenuated immune responsiveness and an increased risk of persistent infections (Crispe, 2014). Chronic liver diseases and their neurologic complications (hepatic encephalopathy and coma) are induced and aggravated by changes of the gut microbiota such as the overgrowth of pathobionts (Enterobacteriaceae, Bacteroidaceae, and Enterococcus species) and the dramatic decline of commensals such as Roseburia, Bifidobacterium, and Lactobacillales (Wu et al., 2012; Bajaj et al., 2014; Schnabl and Brenner, 2014). On the other hand cirrhosis modifies the composition of the microbiota, namely it decreases commensal Clostridia species (the clusters XI and XIVab), the Bacteroidales-Prevotella group and butyrate-producing Roseburia genus (Lu et al., 2011; Bajaj et al., 2012; also reviewed in Schnabl, 2013; Kang and Cai, 2017). HBV associated cirrhosis significantly reduces the Bifidobacterium/Enterobacteriaceae

ratio and increases the bacterial virulence through adhesins and toxins (Lu et al., 2011). The ensuing dysbiosis compromises the integrity of the intestinal barrier and favors bacterial translocations, bacteraemia, or endotoxemia (Mainous et al., 1991; Lin et al., 1995; Bajaj et al., 2014). Given the reduced phagocytosis and canalization of porto-systemic shunts in this setting, bacterial translocations increase the risk of primitive peritonitis and septic shock.

Translocated bacteria and their derived products enter the portal circulation and liver sinusoids, promote liver inflammation (Seki et al., 2007) and aggravate the fibrogenic and oncogenic potential of liver diseases, reviewed in Yang and Seki (2012).

The mechanisms through which intestinal dysbiosis promotes liver injury employ the activation of TLR signaling pathways, reviewed in Seki and Schnabl (2012). Hence, LPS binding to TLR4 receptors of Kuppfer cells triggers a cascade of inflammatory cytokines (Su et al., 2000) and plays a key role in hepatic injury. Bacterial DNA activates TLR9 signaling pathways in Kuppfer and stellate cells which further induces liver inflammation and fibrogenesis in animal models (Miura et al., 2010). NLRP3 and NLRP6 inflammasome deficiency due to intestinal dysbiosis activate TLR4 and TLR9 pathways and exacerbate hepatic inflammation (Henao-Mejia et al., 2012). Therefore dysbiosis ensures a pro-inflammatory intrahepatic environment (Richer et al., 2013; Miura and Ohnishi, 2014; Carrasco Pro et al., 2018) and supports the progression to liver cirrhosis. The imbalance of BA synthesis in dysbiosis contributes to liver inflammation, fibrosis and carcinogenesis as well as to small intestine bacterial overgrowth, which in turn prompts an intestinal inflammatory response (Gunnarsdottir et al., 2003; Jun et al., 2010; Ridlon et al., 2014).

Considering the pathogenic role of microbiota in liver diseases various authors investigated therapeutic and prophylactic interventions. Studies on the latter explored the therapeutic use of probiotics (mostly Lactobacillus and Bifidobacterium species), antibiotics (Rifaximine), gut-derived hormones and even FMT for the manipulation of the gut-liver axis (Ponziani et al., 2015; Zhou et al., 2017a; Román et al., 2019; also reviewed in Wiest et al., 2017). Concurrently, the administration of conjugated BA in rats could normalize bile secretion and reduce intestinal bacterial overgrowth and translocations (Lorenzo-Zúñiga et al., 2003).

#### Dysbiosis and HIV Infection

Human immunodeficiency virus is a retrovirus with CD4+T cell tropism leading to a severe immunodeficiency and AIDS, followed by secondary infections, reactivations of opportunistic germs and HIV-associated cancers. HIV induces systemic immune activation, alters the intestinal barrier and promotes irreversible metabolic, cardiovascular and neurologic changes (Steele et al., 2014). As most experiments on non-human primate model of acquired immunodeficiency have shown, HIV/simian immunodeficiency virus (SIV) induces an ongoing activation of the mucosal immune response and a rapid and massive destruction of CD4+T lymphocytes belonging to the GALT system. The cells most affected are the Th17/22 CD4+T intestinal subsets involved in the maintenance of the gut barrier (Veazey et al., 1998; Guadalupe et al., 2003; Brenchley et al., 2004, 2006; McGowan et al., 2004; Kim et al., 2013). Thus, HIV reconfigures the intestinal barrier early during the infection through immune mechanisms which promote gut alterations, chronic inflammation, and immunodeficiency.

Human immunodeficiency virus-associated intestinal lesions ("HIV enteropathy") initially arise due to the imbalance of the epithelial barrier turnover (Sankaran et al., 2008). In time, the increasing inflammation and intestinal permeability contribute to HIV enteropathy and persist despite antiretroviral therapy (ART) (Olsson et al., 2000; Guadalupe et al., 2003; Neff et al., 2018). The histopathological lesions described in this setting involve intestinal inflammation and mucosal atrophy that nevertheless differ from changes recorded in other intestinal immune-mediated diseases (Magro et al., 2013). On a molecular level, animal studies showed that during HIV enteropathy, IECs display a downregulation of host genes and of miRNAs associated with the homeostasis of the epithelial barrier (Sankaran et al., 2005), epithelial permeability (miRNA-21,-130a,-212) (Gaulke et al., 2014; Zhang et al., 2015; Kumar V. et al., 2018), inflammatory signaling (miR-150) (Kumar et al., 2016), and immune activation syndrome (miR-34a) (Mohan et al., 2015). Some of these miRNAs also regulate the gut microbiota (Johnston et al., 2018) or inflammatory pathways associated with irritable bowel syndrome (Zhou et al., 2010, 2015; Fourie et al., 2014). A study performed on the intestinal lamina propria and leukocytes of SIVinfected macaques indicated that the host miRNA gut profile displays a bitemporal variation initially dictated by the rapid reduction of SIV replication and T-cell activation and afterwards correlated with the consequent boost of the inflammatory response (Kumar et al., 2016). Under these conditions gut epithelia gradually lose their protective role and permit the translocation of bacterial immunogenic molecules (Chege et al., 2011; Marchetti et al., 2013) or enteropathogens (Zeitz et al., 1998; Nazli et al., 2010). In the presence of an impaired immune response, these translocations recur and re-enforce the systemic inflammatory response activating CD4+T cells and HIV replication in a vicious circle (Kotler et al., 1993; Brenchley et al., 2006; Cassol et al., 2010; Dillon et al., 2014). Thus dysbiosis perpetuates enteric and opportunistic infections and aggravates the course of HIV.

Studies on HIV associated dysbiosis are discordant, hindering comparisons, yet available data shows a major depletion of Bacteroides in favor of Proteobacteria (Dillon et al., 2014), the enrichment of pathobionts (Enterococcus, Streptococcus, Staphylococcus, Salmonella, and Escherichia species) and the significant reduction of symbionts throughout the entire intestinal tract (Nishitsuji et al., 2017).Whereas the higher population of Enterobacteriaceae increases the risk of translocations, epithelial cells exposed to E. coli could increase the susceptibility of CD4+Th17 and Th-1 lymphocytes to HIV and viral replication (Dillon et al., 2012).

Intestinal dysbiosis occurs early during HIV and is aggravated by ART itself (Ling et al., 2016; Pinto-Cardoso et al., 2017; also reviewed in Pinto-Cardoso et al., 2018). Hence, patients

with immunologic failure on ART, predominantly host species of Enterobacteriaceae instead of Lactobacillus known for its immunomodulatory and anti-inflammatory activity (Merlini et al., 2011). Otherwise, butyrate synthesized by dominant species in HIV patients (Fusobacterium nucleatum, Clostridium cochlearium, and Eubacterium multiforme) could reactivate both unintegrated HIV-1 genomes and latent HIV proviruses through HDAC inhibition along with other transcription factors and concurrently explains HIV immunosuppression during ART (Kantor et al., 2009; Imai et al., 2012; Lee et al., 2018).The role of SCFAs in the regulation of mucosal immunity during HIV is controversial due to their implication in the reactivation of various infections as well as in the attenuation of mucosal inflammation and microbial translocation (Das et al., 2015; Ye and Karn, 2015; Dillon et al., 2017). Consequently, certain authors proposed the use of chemical HDAC inhibitors but not of SCFAs along with ART for a successful antiretroviral treatment of HIV reservoirs as well as for the attenuation of the LPS-induced inflammatory response (Archin et al., 2009, 2012; also reviewed in Shirakawa et al., 2013; McManamy et al., 2014).

In conclusion, HIV pathogenesis involves the loss of the immunomodulatory gastrointestinal activity, bacterial translocations and intestinal dysbiosis, independent of ART. Current therapeutic strategies are modest and based on few studies in humans or macaques: the administration of probiotics (Hummelen et al., 2011; Ortiz et al., 2016; also reviewed in Hummelen et al., 2010; D'Angelo et al., 2017; Kazemi et al., 2018), anti-inflammatory drugs (reviewed in Deeks et al., 2013), FMT (Hensley-McBain et al., 2016; Vujkovic-Cvijin et al., 2017) stimulating the immune response with a T lymphocyteadjuvated-DNA vaccine (Fuller et al., 2012), or using an oral

REFERENCES


recombinant vaccine with Cl. perfringens expressing HIV-1 Gag protein (Pegu et al., 2011).

#### CONCLUSION

The microbiota, the intestinal epithelia and mucosal immunity form an anti-infectious barrier rigorously regulated through complex mechanisms. The host-microbiota-pathogen interactions employ numerous cell receptors and molecules with antibacterial and anti-inflammatory roles, which also modulate the epigenetic and immune response. Together with epithelial and immune cells, these signaling molecules form a network that is essential for intestinal homeostasis and antiinfectious defense. Dysbiosis renders these defense mechanisms non-functional. It consequently aggravates gastrointestinal infections, favors bacterial and LPS translocations in sepsis, unbalances the immune defense in hepatitis viruses and HIV replications and controls the progression of infectious diseases to an unknown extent. A better knowledge on the interactions driving the antimicrobial response of the intestinal barrier is therefore crucial to improve the current anti-infectious armamentarium.

#### AUTHOR CONTRIBUTIONS

Both authors contributed equally to the acquisition, analysis, and critical revision of the manuscript, gave their consent for the publication, and agreed to be responsible for the accuracy and integrity of the manuscript.

regulatory T-cell generation. Nature 504, 451–455. doi: 10.1038/nature 12726



patients. Am. J. Respir. Crit. Care Med. 158, 444–451. doi: 10.1164/ajrccm.158. 2.9710092



patients: a systematic review and meta-analysis. J. Diet Suppl. 15, 776–788. doi: 10.1080/19390211.2017.1380103



to antiretroviral therapy. PLoS One 6:e18580. doi: 10.1371/journal.pone.001 8580



for lipopolysaccharide binding protein and toll-like receptor 4. Hepatology 31, 932–936. doi: 10.1053/he.2000.5634


Zhou, D., Pan, Q., Xin, F.-Z., Zhang, R.-N., He, C.-X., Chen, G.-Y., et al. (2017b). Sodium butyrate attenuates high-fat diet-induced steatohepatitis in mice by improving gut microbiota and gastrointestinal barrier. World J. Gastroenterol. 23:60. doi: 10.3748/wjg.v23.i1.60

Zhou, Q., Costinean, S., Croce, C. M., Brasier, A. R., Merwat, S., Larson, S. A., et al. (2015). MicroRNA 29 targets nuclear Factor-κB–repressing factor and claudin 1 to increase intestinal permeability. Gastroenterology 148, 158.e–169.e.

Zhou, Q., Souba, W. W., Croce, C. M., and Verne, G. N. (2010). MicroRNA-29a regulates intestinal membrane permeability in patients with irritable bowel syndrome. Gut 59, 775–784. doi: 10.1136/gut.2009.181834

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

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