Mucosal Immunity

Abstract Background Nonspecific induction of local innate immune responses by mucosally administered immunotherapy is a new approach to protection from upper respiratory tract infections. Therefore, a new liposome‐toll‐like receptor complex (LTC) immune stimulant was developed and investigated for its ability to activate innate immune responses in cats, both in vitro and in vivo, as part of an initial evaluation of LTC for use as an immunotherapeutic agent in cats. Objectives We hypothesized that LTC could activate innate immune responses in cats after topical application to nasal and oropharyngeal mucosal surfaces. Animals Mucosal immune responses to topical administration of LTC were assessed in 7 healthy, purpose‐bred cats, and in vitro responses were assessed using blood samples from healthy cats. Methods Cytokine and cellular immune responses to LTC were evaluated in blood samples, nasal lavage specimens, and pharyngeal swabs from cats, using reverse transcriptase polymerase chain reaction assays, ELISA assays, and flow cytometry. Results Liposome‐TLR complexes rapidly activated leukocytes in vitro, including upregulation of costimulatory molecule expression and cytokine production. Topical administration of LTC in healthy cats triggered rapid recruitment of monocytes to the nasal and oropharyngeal mucosa. Conclusions and Clinical Importance Liposome‐TLR complexes were found to effectively activate innate immune responses in cats after mucosal administration. These findings suggest that LTC have potential for use as a new mucosally administered immunotherapy for nonspecific protection from viral and bacterial respiratory tract infections. Abbreviations FBS fetal bovine serum IL interleukin INF interferon LTC liposome‐TLR complexes NALT nasopharyngeal‐associated lymphoid tissue NK natural killer OX40 tumor necrosis factor receptor superfamily, member 4 PBMC peripheral blood mononuclear cells PBS phosphate‐buffered saline RT‐PCR reverse transcriptase polymerase chain reaction TLR toll‐like receptor TNF tumor necrosis factor 1 INTRODUCTION Upper respiratory infection (URI) is a common and serious problem in cats in boarding facilities, animal hospitals, and shelters. Unfortunately, few quickly effective options are available for preventing these infections in newly exposed cats. Immunoprophylaxis may be most effective if focused at the likely portal of pathogen entry. The upper respiratory tract is the site of entry for several important respiratory pathogens of cats, consisting of a diverse array of viral and bacterial organisms, including feline herpesvirus 1 (FHV‐1), calicivirus, Mycoplasma sp., Bordetella bronchiseptica, and Chlamydia felis.1-6 Activation of host innate immune defenses, including airway epithelial cells and immune cells in oropharyngeal lymphoid tissues, results in production of antiviral and antibacterial cytokines, including type I (INF‐α) and II (INF‐γ) interferons, tumor necrosis factor alpha (TNF‐α), and interleukin 12 (IL‐12) and chemokines such as monocyte chemoattractant protein 1 and IL‐8 that recruit monocytes, T cells, and neutrophils. Local activation of innate immune responses may create an environment that is less susceptible to viral entry and replication or to bacterial invasion and colonization.7 Clinically, such an immune protective effect could be manifested as either partial or complete protection from infection or as a decrease in the severity and duration of clinical signs of illness, along with decreased viral or bacterial shedding. One potential solution to the problem of complex upper respiratory tract infections in cats would be an immunotherapeutic agent capable of inducing nonspecific immune activation at mucosal surfaces in the oropharynx, such as liposome‐toll‐like receptor complexes (LTC). These complexes have been shown previously to induce strong nonspecific protection from lethal bacterial and viral respiratory tract infections after intranasal administration in several different mouse infection models.8-12 In addition, we demonstrated induction of therapeutic immune responses in cats with chronic rhinitis after parenteral administration of LTC.13 Therefore, we reasoned that immune stimulatory complexes could be further optimized for mucosal delivery and generation of local mucosal immune responses in the upper respiratory tract of cats. Therefore, the original LTC formulation was modified to improve its innate immune stimulatory potency and to improve adhesion and contact time with mucosal surfaces. In the present study, the immune activating properties of modified LTC were evaluated in cats.12, 14 The modified LTC consisted of cationic liposomes formulated with 2 TLR agonists: polyinosinic‐polycytidylic acid (poly I:C), as a TLR3, retinoic acid‐inducible gene I, melanoma differentiation‐associated protein 5 agonist, and noncoding plasmid DNA (pDNA), as a TLR9 agonist rich in cytosine‐guanine. In addition, the modified LTC contained a mucosal adhesive agent (carboxymethylcellulose) to increase adhesion to mucosal cell surfaces. We also assessed cytokine and cell activation responses by feline leukocytes in vitro and also assessed the local mucosal immune responses to topical administration of LTC in healthy cats. 2 MATERIALS AND METHODS 2.1 Preparation of liposome‐TLR agonist complexes for mucosal immune stimulation Liposomes were prepared by dissolving the cationic lipid, 1,2‐dioleoyl‐3‐trimentylammonium‐propane (DOTAP) and cholesterol (Avanti Polar Lipids, Alabaster, AL) in chloroform, and the solution was dried overnight in a vacuum desiccator to a thin film. The lipids were rehydrated in 5% sterile dextrose in water at 50°C as described previously.12, 14 Poly I:C (InVivoGen, San Diego, California) and pDNA (noncoding commercial plasmid PCR2.1) Thermo Fisher Scientific, Waltham, MA were added to form liposome complexes, as reported previously.12, 14 The endotoxin content of the plasmid was between 0.04 and 0.25 EU/μg. Carboxymethylcellulose (Sigma‐Aldrich, St. Louis, Missouri) was added to the LTCs to increase mucosal adhesion. In studies in which fluorescent LTCs were delivered, the liposomes were formulated to contain 10% (vol/vol) Topfluor‐labeled cholesterol (Avanti Polar Lipids). 2.2 Animal studies Animal protocols were approved by the Institutional Animal Care and Use Committee of an independent contract research facility in Fort Collins, Colorado. Purpose‐bred cats for the studies were purchased from Liberty Research (Waverly, New York). All animals used in the studies were adopted to owners in the Fort Collins area after completion of the studies, and no animals were euthanized as part of these studies. For collection of nasal lavage samples, cats were restrained manually in a head‐down position, and 0.5 mL of prewarmed sterile phosphate‐buffered saline (PBS) solution was administered into each nostril, and the resulting fluid backflow was collected from the nostrils using a 15 mL conical tube. The procedure was repeated once for each nostril, and all of the fluid collected was pooled for later analysis and stored on ice. For oropharyngeal samples, alginate swabs were gently rubbed and rolled against the mucosa of the caudal oropharyngeal region and placed in tubes with 1 mL PBS and stored on ice before analysis. Nasal and oropharyngeal samples were collected 24 hours before LTC administration, 6 hours after LTC administration, and again 20 and 72 hours after LTC administration. Cats were treated with LTC administered on day 0. Control cats were treated with an equivalent volume of sterile PBS. Each cat was given 1 mL of LTC (0.2 mL intranasally in each nostril and 0.6 mL PO) while being gently restrained. In a separate experiment designed to track LTC uptake by mucosal cells in the nostril and oropharynx of cats, 1 mL of LTC labeled with TopFluor fluorescent dye was administered by instilling 0.2 mL intranasally in each nostril and 0.6 mL (equivalent to 60 μg of Poly I:C and PCR2.1) PO. Average weight of the cats was approximately 3 kg; thus, the intranasal dose of each TLR ligand was approximately 20 μg per kilogram body weight. No adverse effects, either acute or longer term, were observed after intranasal LTC administration (data not shown). 2.3 Preparation of peripheral blood mononuclear cells Whole blood was obtained by jugular venipuncture from a separate group healthy young adult cats and collected into EDTA tubes. For separation of peripheral blood mononuclear cell (PBMC), blood was diluted 1:2 with sterile PBS, then layered over a Ficoll (GE Healthcare, Uppsala, Sweden) gradient and centrifuged. The PBMC were collected and washed twice in PBS and then resuspended in complete tissue culture medium before flow cytometric analysis. 2.4 Cell culture and PBMC activation with LTC Peripheral blood mononuclear cells were cultured in complete medium, which consisted of Dulbecco's modified eagle medium (Thermo Fisher Scientific, Watham, Massachusetts) containing 15% fetal bovine serum (FBS; Avanti Polar Lipids), essential and nonessential amino acids, penicillin, and streptomycin (Gibco‐Thermo Fisher Scientific, Pittsburgh, Pennsylvania). Cells were plated in 96‐well flat bottom plates (Celltreat, Pepperell, Massachusetts) at a density of 1 × 106 cells/well in 200 μL medium. For assays involving LTC activation of PBMC, LTC were added at 3 different dilutions (1 μL per well, 0.3 μL per well, and 0.1 μL per well) in triplicate wells of PBMC in 100 μL complete medium, with careful mixing, and then the cells were incubated for an additional 48 hours. Conditioned medium was collected for IFN‐γ assay, and cells were collected for flow cytometric analysis of activation markers (see below) and for analysis of cytokine gene expression by quantitative reverse transcriptase polymerase chain reaction (qRT‐PCR) (see below). Assays were repeated at least twice, using PBMC from different donor animals. 2.5 Quantitative RT‐PCR analysis of cytokine gene expression in vitro Ribonucleic acid was extracted from PBMC incubated for 48 hours with LTC, as noted above, and expression of IL‐1, IL‐12, IFN‐α, IFN‐γ, and TNF‐α genes was determined using quantitative RT‐PCR. Briefly, RNA was isolated and reverse transcribed using a commercial kit (Qiagen, Germantown, Maryland), and the complementary DNA then was amplified using Cyber green primers (Bio‐Rad, Hercules, California) using a qPCR MX3000p system instrument (Agilent, Santa Clara, California). Cytokine transcript copy numbers were determined by quantitative RT‐PCR, as noted previously.13 2.6 IFN‐γ ELISA Cell culture supernatants from PBMC cultures were analyzed for IFN‐γ concentrations, using a feline IFN‐γ specific ELISA (DuoSet Feline IFN‐γ kit; R&D systems, Minneapolis, Minnesota) according to the manufacturer's instructions. 2.7 Flow cytometry Cells were harvested after 48 hours of in vitro stimulation and immunostained with the following antibodies conjugated with fluorochromes: CD5‐PE (clone f43; Southern Biotech, Birmingham, Alabama), CD21‐APC (clone CA2.1Dg; Bio‐Rad), and OX40‐FITC (clone 7D6; Bio‐Rad), or CD14‐APC (clone TÜK4; Bio‐Rad) and MHCII‐FITC (clone TÜ39; BD Biosciences, San Jose, California). Cells were immunostained with primary antibodies for 20 minutes at 4°C in FACs buffer (PBS with 2% FBS and 0.05% sodium azide) after a 5‐minute incubation with normal cat serum (Jackson ImmunoResearch) to block nonspecific binding. Cells obtained by nasal lavage and oropharyngeal swabs were washed once with PBS, then immunostained in 96‐well plates with the directly conjugated antibodies noted above. Isotype‐matched antibodies were used as control for each primary antibody. Cells from in vitro activated PBMC cultures were collected 48 hours after stimulation and measured using the same panel of antibodies, with appropriate controls. Cells were analyzed using a Beckman Coulter Gallios flow cytometer (Beckman Coulter, Indianapolis, Indiana), and data were analyzed using FlowJo software (Tree Star, Ashland, Oregon). 2.8 Preparation and analysis of nasal and pharyngeal samples Swabs were collected from the distal oropharynx and immediately placed in 2 mL of complete tissue culture medium in 15 mL polypropylene tubes on ice. The samples were lightly vortexed to free cells from the swabs. Fluid was transferred to a new tube, and 2 mL of PBS was added to the empty tube containing the swab, vortexed again, and pooled with the first sample. The sample then was filtered using a 70 μm cell strainer (Celltreat, Pepperell, Massachusetts) to remove debris and mucus. Samples were centrifuged, the cell pellets were resuspended in 0.5 mL of PBS, and the cells were counted. Nasal lavage samples were filtered, washed, centrifuged, resuspended, and counted in the same fashion as for oropharyngeal swab samples. Cells were stained in the same manner as described for PBMC, and equal numbers were analyzed by flow cytometry to determine the cell type and activation. 2.9 Confocal microscopy Monocyte‐derived macrophages were generated from cat PBMC by incubation of plastic adherent monocytes with 10 ng/mL human recombinant macrophage colony stimulating factor for 5 days. The macrophages that were generated were then incubated with 0.1, 1, or 2 μL per well of TopFluor‐labeled LTC in complete medium at 37°C for 4 hours, then washed gently 3 times, fixed with 2% paraformaldehyde for 30 minutes, washed again with distilled water, cover‐slipped, and analyzed by confocal microscopy. Confocal images of feline monocyte‐derived macrophages (untreated control cells or cells incubated with labeled LTC) were obtained using an Olympus IX3 confocal microscope (Waltham, Massachusetts). Images were processed and analyzed using Olympus cellSens software. 2.10 Statistical methods Statistical comparisons between data sets with 2 treatment groups were performed using the nonparametric Mann‐Whitney test. Comparisons among 3 or more groups were performed using analysis of variance, followed by the Tukey multiple means post‐test. Analyses were performed using Prism6 software (GraphPad, La Jolla, California). For all analyses, statistical significance was set at P < .05. 3 RESULTS 3.1 B cells, T cells, and monocytes from feline PBMC activated by LTC in vitro Feline lymphoid tissues express TLRs such as TLR3 and TLR9, and feline immune effector cells (T cells and B cells) express TNF receptor superfamily, member 4 (OX40). When bound to its ligand OX40L, OX40 augments immune responses by prolonging survival of T cells, generation of immune memory, preventing emergence of T‐cell tolerance, and mediating decreased immunosuppressive activity of regulatory T cells. Thus, we chose to examine the effect of in vitro treatment of PBMC with LTC on OX40 expression in these cells.13, 15-17 Unfractionated PBMC were used because LTC may have both a direct effect on leukocytes as well as an indirect effect through crosstalk among various cell types in the mixture. B lymphocytes that were CD21+ were found to exhibit significant upregulation of expression of the costimulatory molecule OX40, in a dose‐dependent fashion with respect to the amount of added LTC (Figure 1, panel A). For T‐cell responses to LTC stimulation, the percentage of CD5+ T cells expressing OX40 also increased significantly as a function of LTC concentration (Figure 1, panel B). Finally, a significant increase in expression of MHCII by CD14+ monocytes was observed in cultures of PBMC treated with LTC (Figure 1, panel C). Similar results were observed in 3 separate studies. 3.2 LTC treatment activates cytokine gene expression by PBMC Quantitative RT‐PCR analysis was used to assess induction of cytokine gene expression by PBMC treated with LTC (Figure 2). Significant upregulation of expression of mRNA for INF‐α as well as upregulation of IL‐1β, IL‐12, and TNF‐α expression after LTC stimulation was identified in most cats. 3.3 LTC treatment triggers IFN‐γ secretion by PBMC Supernatants were collected from PBMC after 48 hours stimulation with LTC, for analysis of IFN‐γ production (Figure 3). Liposome‐TLR complexes treatment stimulated a dose‐dependent, significant secretion of IFN‐γ secretion, as determined by ELISA analysis of supernatants. 3.4 Uptake of labeled LTC by macrophages in vitro Monocyte‐derived macrophages were generated from PBMC of cats by plastic adherence, followed by culture for 4 days in complete medium, after which macrophage uptake of LTC was assessed. Extensive uptake of LTC complexes by macrophages was observed (Figure 4), consistent with the known targeting of liposome complexes to monocytes and macrophages and other phagocytic immune cells. 3.5 Retention of labeled LTC by cells in the oropharynx and nasal cavity Seven healthy adult cats were given 0.1 mL of labeled LTC bilaterally into the nostrils and 0.6 mL PO. Nasal lavage and throat swab samples were obtained before administration, and 6, 20, and 72 hours after LTC administration; cell populations were analyzed by flow cytometry, as described in the Materials and Methods section. Compared to pretreatment samples, samples obtained from the nose and oropharynx at 6 and 20 hours after LTC administration contained a readily identifiable population of cells containing fluorescently‐labeled LTC (Figure 5). The LTC‐containing cells were detected in both nasal lavage cells (Figure 5A) and in throat swab cells (oropharynx, Figure 5B). Peak uptake of LTC‐labeled cells in the nasal lavage samples occurred 6 hours after LTC administration, whereas in the oropharynx peak of LTC uptake occurred 20 hours after administration. 3.6 Intranasal and PO administration of LTC triggers influx of inflammatory leukocytes to mucosal surfaces Cells from the nasal cavity or oropharynx were obtained by nasal lavage or swabbing the back of the throat, respectively, as described above. Cells were immunostained with fluorescently conjugated antibodies against specific cell markers for T cells (CD5+) and monocytes (CD14+) and analyzed by flow cytometry. Neutrophils were identified by their characteristic forward versus side scatter profiles. Pretreatment samples from the nasal cavity of 7 healthy cats contained an average of 7% CD5+ T cells, whereas the average percentage of CD5+ T cells in the oropharynx of cats before treatment was 73%. The percentage of CD14+ monocytes in the nasal cavity before treatment was 2.5%, similar to the 3% of CD14+ monocytes found in the oropharynx (Figure 6). Percentages of neutrophils found in the nasal cavity were 10%, and 2% in the oropharynx of untreated animals. Six hours after topical administration of LTC, a marked increase in the percentage of CD14+ monocytes was observed in both the nasal mucosa (Figure 6A, increasing from an average of 2.5% to an average of 14%) and in the oropharynx (Figure 6B, increasing from an average of 3% to an average of 11%). The percentage of CD14+ monocytes from both nasal and oropharynx samples had decreased by 20 and 72 hours compared to the 6 hours samples but still remained higher than in pretreatment samples. Interestingly, the percentages of CD5+ T cells from the nasal mucosa decreased significantly 6 hours post‐LTC treatment (decreasing from 7% to 3%) and from 73% to 51% in the oropharynx. Nasal CD5+ T cells returned to pretreatment values by 72 hours after treatment. In the oropharynx, the decrease in CD5+ cells observed at 6 hours was also transient, with percentages of CD5+ T cells returning to nearly baseline values (67%) at 20 hours post‐LTC treatment. The neutrophil population was essentially unchanged throughout the study time course in both the nose and the oropharynx after LTC administration (Figure 6). 4 DISCUSSION An effective, broad‐spectrum immunotherapeutic agent capable of activating protective mucosal immune responses would provide a valuable means of inducing rapid protection from a variety of upper respiratory tract pathogens in cats. Therefore, we investigated innate immune responses to LTC, a new immune therapeutic agent designed for mucosal application. We found that LTC induced activation of feline leukocytes, including upregulation of costimulatory molecule (OX40 and MHCII) expression. In addition, LTC induced production of cytokines associated with innate immune responses, including INF‐α, INF‐γ, TNF‐α, and IL‐12. in vivo studies in healthy cats demonstrated uptake of LTC by immune cells and epithelial cells in the upper airways, as well as recruitment of monocytes to the upper airways. Our studies indicated that cats were capable of responding to mucosal immune activation by LTC. Significant upregulation of IFN‐γ along with increases in lymphocyte expression of OX40 and monocyte expression of MHCII all would be associated with innate immune system activation for better viral and bacterial pathogen clearance.7, 18, 19 These findings are consistent with LTC activation of mucosal innate immune responses in cats and induction of partial protection against inhalational challenge with FHV‐1.19 Nasopharyngeal‐associated lymphoid tissue (NALT) is a network of immune cells located in the oropharyngeal region that is primed for immune recognition of viral, bacterial, and fungal pathogens.20-23 When activated by infections, innate immune effector cells in the NALT (including monocytes, neutrophils, innate lymphoid cells, and natural killer [NK] cells) respond by producing cytokines and chemokines, and by activating other immune effector cells including tissue macrophages, epithelial cells, and T and B cells. In our study, several key antiviral and antibacterial cytokines were produced after LTC treatment, including significant increases in secretion of IFN‐γ protein, and increases in INF‐α and IL‐12 transcription, as assessed by qRT‐PCR. These results together suggest broad activation of innate immune responses in nasal and oropharyngeal tissues after administration of LTC in cats. Distinct cellular trafficking responses to LTC treatment were noted when monocyte and T‐cell responses were compared in the nose and oropharynx. For example, LTC administration triggered strong recruitment of CD14+ monocytes to both the nasal cavity and the oropharynx, while at the same time triggering significant, transient disappearance of T cells from these mucosal sites (see Figure 6). Interestingly, minimal neutrophil responses to LTC administration were observed at any time point evaluated, in either the nose or oropharynx. The recruitment of monocytes by LTC could be attributed to production of the chemokine CCL2, which has been noted previously after treatment with LTC.11 In contrast, the transient disappearance of T cells from mucosal sites after LTC administration most likely could be attributed to transient redistribution of T cells after adherence to capillary blood vessels or migration into tissues. A similar effect has been observed previously in dogs treated systemically with LTC.24 In support of transient redistribution, we observed that CD5+ T lymphocytes were rapidly restored in mucosal sites within 72 hours after LTC administration. We also observed distinct differences in resident immune cell populations in the nasal cavity and oropharynx, especially with respect to the density of mucosal T cells. For example, the oropharynx was found to be highly enriched in mucosal CD5+ T cells, whereas T cells were much less numerous in the nasal cavity (see Figure 6). These differences are considered real (and not simply a reflection of differences in sampling techniques), because the percentages of unmarked cells (assumed to represent primarily mucosal epithelial cells, see Figure 6) were similar in samples from the oropharynx, compared to nasal lavage samples. Thus, the oropharynx in cats appears to be enriched in superficially located T cells compared to the nasal cavity and thus well suited to induction of mucosal immune responses after PO administration of nonspecific immune stimuli. The ability of LTC to rapidly induce mucosal immune responses in the upper airway and oropharynx of cats suggests potential use as a nonspecific immune therapeutic agent for prevention of viral and bacterial infections in cats. Administration of LTC significantly decreased signs of FHV‐1 infection and viral shedding in experimentally inoculated cats.19 In addition, we have found that topical application of LTC in dogs can induce mucosal activation and cellular recruitment, similar to the responses seen in cats (Dow S, unpublished data). Thus, topical treatment in the upper airways with LTC may have utility for preventing or mitigating infections with common respiratory tract pathogens in both dogs and cats. The present study had several limitations. For example, the lack of available reagents for full phenotyping of feline leukocytes precluded a more complete assessment of cellular responses to topical LTC administration. In addition, the relatively small numbers of cells collected by nasal lavage or oral swabs precluded cytokine analysis after LTC administration. The in vitro studies using PBMC did not determine the cellular source of specific cytokine production after LTC activation. However, previous studies of LTC in mice indicated that dendritic cells were the primary cellular source of IL‐12 production after LTC administration, and that NK cells were the primary cellular source of IFN‐γ.25, 26 In summary, our studies describe the innate immune activation properties of a novel mucosally active immune stimulant in cats. Innate immune activation by LTC was associated with production of key antiviral and antibacterial cytokines, and with upregulation of important costimulatory molecules on immune effector cells. These findings suggest further evaluation of the LTC mucosal immune stimulant for prevention of URI in cats and for treatment of established URI refractory to treatment with antibiotics or other medications such as lysine. ACKNOWLEDGMENTS The authors acknowledge the assistance of Ms. Jade Kurihara with preparation of LTC used in the study, and Ms. Jennifer Hawley for performing RT‐PCR assays. Portions of this research were presented at the 2017 American College of Veterinary Internal Medicine Forum, National Harbor, Maryland. This study was supported by a grant from the state of Colorado Office of Economic Development and International Trade. CONFLICT OF INTEREST DECLARATION Drs. Lappin and Dow both hold stock options and corporate positions in Poudre Canyon Therapeutics, a Fort Collins company developing the LTC immunotherapy platform technology. OFF‐LABEL ANTIMICROBIAL DECLARATION Authors declare no off‐label use of antimicrobials. INSTITUTIONAL ANIMAL AND CARE AND USE COMMITTEE (IACUC) OR OTHER APPROVAL DECLARATION The cats were managed by a contract research facility in Fort Collins under an approved IACUC protocol. HUMAN ETHICS APPROVAL DECLARATION Authors declare human ethics approval was not needed for this study. REFERENCES

Summary The immune system is essential for maintaining a delicate balance between eliminating pathogens and maintaining tolerance to self‐tissues to avoid autoimmunity. An enormous and complex community of gut microbiota provides essential health benefits to the host, particularly by regulating immune homeostasis. Many of the metabolites derived from commensals can impact host health by directly regulating the immune system. Many autoimmune diseases arise from an imbalance between pathogenic effector T cells and regulatory T (Treg) cells. Recent interest has emerged in understanding how cross‐talk between gut microbiota and the host immune system promotes autoimmune development by controlling the differentiation and plasticity of T helper and Treg cells. At the molecular level, our recent study, along with others, demonstrates that asymptomatic colonization by commensal bacteria in the gut is capable of triggering autoimmune disease by molecular mimicking self‐antigen and skewing the expression of dual T‐cell receptors on T cells. Dysbiosis, an imbalance of the gut microbiota, is involved in autoimmune development in both mice and humans. Although it is well known that dysbiosis can impact diseases occurring within the gut, growing literature suggests that dysbiosis also causes the development of gut‐distal/non‐gut autoimmunity. In this review, we discuss recent advances in understanding the potential molecular mechanisms whereby gut microbiota induces autoimmunity, and the evidence that the gut microbiota triggers gut‐distal autoimmune diseases. Introduction The mammalian gastrointestinal tract is home to an enormous and complex community of commensal bacteria.1-3 This gut microbial community (microbiota) has co‐evolved with its host over millennia, suggested by the fact that out of 52 bacterial phyla, two dominant phyla, Bacteroidetes and Firmicutes, represent 98% of the bacteria in the gut.4, 5 These gut microbes provide benefits to their host in many ways, including, but not limited to, digestion, production of nutrients, detoxification, protection against pathogens and regulation of the immune system.1-3, 6, 7 As a result of its collective metabolic activity, and necessity for human health, the gut microbiota is often referred to as the ‘forgotten’ organ.8 Interestingly, many of the metabolites derived from commensal bacteria have recently been found to directly impact the immune system. Commensal‐mediated immunomodulation includes the promotion of T helper cell subset differentiation and T helper cell plasticity, the ability of a differentiated CD4+ T cell to take on characteristics of other T‐cell subsets simultaneously or at different times.9 Plasticity is an especially important subject for mucosal immunity, as mucosal tissues are major sites where T‐cell plasticity has been observed.10-14 Therefore, the mucosa harbors the frontline immune response to commensal bacteria, and their interactions may subsequently control health versus disease status through T‐cell differentiation and plasticity. An autoimmune condition develops when the immune system mistakenly attacks our own self‐tissues. A long‐standing question in the field of microbe–host communication is how interaction of microbes with T cells promotes autoimmune development. Molecular mimicry is one prominent hypothesis, theorizing that microbes trigger autoimmunity by a shared immunogenic epitope between microbes and a self‐peptide, which leads to the activation of autoreactive T cells.15, 16 Another less well‐known theory is that expression of dual T‐cell receptors (TCRs) on T cells promotes autoimmunity by allowing autoreactive T cells to escape thymic clonal deletion.17-19 Alterations in the composition of the gut microbiome, known as dysbiosis, have been implicated in many diseases and disease models, including those that show clear association with the gut such as inflammatory bowel disease, colorectal cancer, enteric infections, and obesity.20-27 Dysbiosis has also been implicated in immune disorders that occur outside the gut, such as asthma, eczema, allergies, and gut‐distal autoimmune diseases (autoimmune arthritis, type 1 diabetes, experimental autoimmune encephalomyelitis).20, 28-37 Although it is easy to see how the gut microbiota can influence immune responses within the gut, over the last decade an appreciation for the impact of the gut microbiome on gut‐distal autoimmunity disease has developed. For example, a strong interest has emerged to characterize the impact of gut microbiota on health and disease status in the lung and brain, namely ‘gut–lung axis’ and ‘gut–brain axis’, respectively.38, 39 This article aims to review the influence of microbiota on the immune system by producing key commensal‐derived metabolites as well as controlling T‐cell subsets and plasticity. Additionally, we discuss the recent findings on how gut commensals affect autoimmunity by molecular mimicry and skewing the expression of dual TCRs. Finally, we discuss the role of dysbiosis in the pathogenesis of common autoimmune diseases. As dysbiosis in inflammatory bowel disease has been extensively reviewed elsewhere (see ref. 40-42), this review will focus on the impact of gut microbiota dysbiosis in the development of gut‐distal autoimmune diseases. Understanding the mechanisms of how gut commensals affect the immune system will be crucial for us to elucidate autoimmune pathogenesis and to generate novel therapies. This is an urgent subject, as dysbiosis‐related diseases have emerged as new epidemics in the industrialized world.43-45 Commensal bacteria‐mediated metabolites and immunity Gut commensals are well‐known for their function in digestion, a process involving extraction and synthesis of many metabolites, some of which are produced only by commensal bacteria and are crucial for host health.46 Recently, several advancements have been made determining how the metabolites generated by commensal bacteria can directly influence the development and function of the immune system, directly impacting health and diseases (Table 1). Here, we review the most recent reports in this field. Commensal Metabolite Disease phenotype Immunomodulation References Clusters IV and XIVa of Clostridium ↑ IDO, enzyme involved in tryptophan catabolism, on IEC ↓ Colitis ↑ Colon Treg ↓ Th1 63, 66 Lactobacillus reuteri ↑ Indole derivatives by metabolizing tryptophan   ↑ Differentiation of DP IEL 69 Clostridium sporogenes ↑ IPA derivatives by metabolizing tryptophan, phenylalanine, and tyrosine ↓ Permeable intestines ↓ Neutrophils, monocytes, and memory T cells 72 General microbiota signal through MyD88 ↑ RALDH, enzyme able to catalyze the synthesis of RA from Vitamin A ↓ EAE ↓ Autoimmune arthritis ↑ Treg ↓ Th17 ↑ Gut‐homing integrin α4β7 ↑ Maintenance of peripheral lymph nodes 78, 79, 92 Ruminococcaceae eubacterium ↑ SCFAs signal through GPR43 ↓ EAE ↓ Type 1 diabetes ↑ Lamina propria Treg ↑ Antimicrobial peptides 57, 53, 55 Abbreviations: DP IEL, double‐positive intraepithelial T lymphocytes; EAE, experimental autoimmune encephalomyelitis; IDO, indoleamine 2,3‐dioxygenase; IEC, intestinal epithelial cells; IPA, indolepropionic acid; RA, rheumatoid arthritis; RALDH, retinal dehydrogenase; SCFA, short‐chain fatty acids; Th1, T helper type 1; Treg, regulatory T. Short‐chain fatty acids Short‐chain fatty acids (SCFAs) including butyric acid, propionic acid, and acetic acid are the main metabolic products of undigested carbohydrates by gut commensal bacteria and have broad effects on the host immune system.47 Among the fatty acid receptors, two orphan G protein‐coupled receptors, GPR41 and GPR43, are activated by SCFAs.48, 49 SCFAs are significantly reduced in the colon of germ‐free mice indicating that the gut microbiota is essential for their production.50 Recently, one group reported that long‐chain fatty acids enhanced the differentiation and proliferation of T helper type 1 (Th1) and/or Th17 cells, whereas SCFAs expanded gut regulatory T (Treg) cells.51, 52 Using experimental autoimmune encephalomyelitis (EAE) as a model for multiple sclerosis, Haghikia et al. showed that long‐chain fatty acids decreased SCFAs in the gut, leading to exacerbated EAE by expanding pathogenic Th1 and/or Th17 cells in the small intestine.51 Treatment with SCFAs ameliorated EAE by inducing lamina‐propria‐derived Treg cells. Consistent with the anti‐inflammatory properties of SCFAs, treatment of germ‐free mice with acetate, a type of SCFA, is sufficient to reduce intestinal inflammation in the dextran sulfate sodium model of colitis.53 The protective effect of acetate was lost in GPR43−/− mice, indicating that signaling of acetate through GPR43 is necessary for the anti‐inflammatory effect of SCFAs. Remarkably, administration of the SCFA butyrate ameliorates inflammation in patients with ulcerative colitis.54 SCFAs also indirectly control type 1 diabetes through the production of antimicrobial peptides. Mechanistically, SCFAs produced by the gut microbiota have been shown to stimulate the production of antimicrobial peptides by pancreatic beta cells.55 Furthermore, systemic administration of these antimicrobial peptides induce Treg cells in the pancreatic islets of pre‐diabetic mice, reducing the incidence of autoimmune diabetes. ‬Additionally, it has been recently demonstrated that administration of the SCFA propionate significantly attenuated HLA‐B27‐associated intestinal inflammation in Fischer 344 HLA‐B27/β2m transgenic rats.56 Interestingly, this propionate‐mediated inhibition of inflammation is independent of its role in regulating the Treg cell response. Amino acids An early study gives an indication that gut bacteria may play an important role in host amino acid homeostasis and health by showing that germ‐free mice had an altered distribution of free amino acids along the gastrointestinal tract compared with conventionalized mice.58 More recent reports suggest that the most abundant amino‐acid‐fermenting bacteria in the human small intestine are bacteria belonging to the Clostridia, the Bacillus–Lactobacillus–Streptococcus groups, Proteobacteria, and Peptostreptococcus.59-61 These bacteria are therefore likely to be important for amino acid absorption in the gastrointestinal tract. Among the 20 amino acids, tryptophan belongs to one of the nine ‘essential’ amino acids that humans and higher vertebrates cannot synthesize, and that must be supplied in their diet.62 Numerous studies suggest that tryptophan metabolites generated by gut commensal bacteria serve as important signaling molecules in host–microbe cross‐talk. For example, the colonic intestinal epithelial cells from mice colonized with clusters IV, XIVa, and XVIII of Clostridium express high levels of indoleamine 2,3‐dioxygenase,63 an enzyme involved in the initial and rate‐limiting step of tryptophan catabolism, which has been implicated in Treg cell induction.64 Specifically, it has been demonstrated that kynurenine, the first product in the indoleamine 2,3‐dioxygenase‐dependent tryptophan degradation pathway, activates the aryl hydrocarbon receptor, leading to aryl hydrocarbon receptor‐dependent Treg cell generation.65 Indeed, an induction of colonic Treg cells was observed in mice colonized with Clostridium species.63 The small intestinal epithelium contains a unique population of CD4+ CD8αα+ double‐positive intraepithelial T lymphocytes (DP IELs) which exhibit anti‐inflammatory properties.67, 54 It has been shown that upon migration to the epithelium, Treg cells lose Foxp3 and convert to DP IELs in a microbiota‐dependent manner, as microbiota depletion by treating mice with broad‐spectrum antibiotics prevents Foxp3 loss.68 Whereas all Treg cells express the CD4‐lineage transcription factor T helper‐inducing POZ/Krüppel‐like factor (ThPOK), CD8αα+ CD4+ and > 50% of Foxp3– CD8α– CD4+ cells in the small intestinal epithelium lack ThPOK expression. Later, it was found that loss of ThPOK corresponds to an IEL‐like behavior in CD4+ T cells. Interestingly, supplement of TCR stimulation can overcome the microbiota requirement for the DP IEL differentiation, as demonstrated by the relatively normal DP IEL differentiation in antibiotics‐treated Rag1−/− OT‐II (ova‐specific) mice exposed to oral ovalbumin. By comparing the microbiota profile between mice housed at two vivaria, with mice at one vivarium displaying significant numbers and another displaying negligible or absent DP IELs, Cervantes‐Barragan et al. discovered that the commensal Lactobacillus reuteri was responsible for the differentiation of DP IELs.69 Lactobacillus reuteri generated indole derivatives by metabolizing tryptophan, which activated the aryl‐hydrocarbon receptor in CD4+ T cells, leading to their ThPOK down‐regulation and differentiation into DP IELs. Hence, L. reuteri together with a tryptophan‐rich diet can promote gut immune homeostasis by allowing the differentiation of intraepithelial CD4+ T cells into anti‐inflammatory DP IELs. Only a few species of gut commensals, including Clostridium sporogenes, can break down tryptophan and produce the metabolite indolepropionic acid (IPA), a deamination product of tryptophan.70 Bacterial tryptophanase catalyzes the conversion of dietary tryptophan to indole and subsequently to IPA.71 A recent study identified a total of 12 compounds that can be produced in this process, nine of which can accumulate in the blood and three of which are produced exclusively by bacteria.72 Specifically, the C. sporogenes‐expressed gene fldC is necessary for the production of IPA. In fact, fldC is essential for the reductive metabolism of all three aromatic amino acids (tryptophan, phenylalanine, and tyrosine). Germ‐free mice receiving wild‐type C. sporogenes exhibit high levels (~80 μm) of serum IPA whereas germ‐free mice receiving the mutated version of C. sporogenes lacking fldC, had undetectable serum IPA. Importantly, mice with undetectable IPA had higher levels of immune cells, including neutrophils, classical monocytes, and memory T cells. In addition, the mice with the engineered version of C. sporogenes had increased intestinal permeability, a defect that is often seen in inflammatory bowel disease. In addition to having a direct impact on the immune system, undigested amino acids have the potential to become supplemental precursors for SCFA generation by the gut microbiota, in addition to indigestible carbohydrates.73 Therefore, amino acids could have indirect impact on the immune system through SCFAs as we discussed earlier. Numerous amino acids including glycine, threonine, glutamate, and ornithine can be metabolized by anaerobic bacteria to generate acetate, whereas threonine, lysine, and glutamate can be used for butyrate synthesis.74 Moreover, at the molecular level, it has been shown that intracellular leucine concentrations can be sensed by the multiprotein complex leucyl‐tRNA synthetase,75, 76 which activates the mechanistic target of rapamycin kinase, proving to be a vital link between immune function and metabolism.77 Retinoic acid Retinoic acid, a metabolite of vitamin A, is one of the most active physiological retinoid metabolites and has a wide range of biological activity including regulating immune responses.78 A major part of retinoic acid's anti‐inflammatory effects depends on the inhibition of Th17 cells and promotion of Foxp3+ Treg cell responses.78, 79 In addition, retinoic acid has been shown to be important for the expression of the gut homing receptor integrin α4β7 on T cells.80-82 The α4β7 integrin receptors are imprinted on lymphocytes by dendritic cells (DCs) from Peyer's patches (PPs), and mesenteric lymph nodes.82, 83 In the absence of microbial toll‐like receptor signaling in Myd88−/− mice, gut DCs express low levels of retinal dehydrogenase (RALDH) required for retinoic acid biosynthesis, so these Myd88−/− mice are unable to generate gut‐homing lymphocytes.84 AM80 is a synthetic retinoic acid that is characterized by higher stability and fewer potential adverse effects compared with retinoic acid.85, 86 It has been reported that both retinoic acid and AM80 ameliorate many autoimmune responses, including experimental autoimmune myositis, experimental autoimmune encephalitis, and collagen‐induced arthritis.87-90 We recently showed that oral administration of AM80 inhibits autoimmune disease in the joints as well as in the lung.91 We elucidated a novel mechanism whereby AM80 suppresses the autoimmune pathology in both the lung and joints by inhibiting T follicular helper (Tfh) cells in addition to inhibiting Th17 responses. Specifically, AM80 increased the expression of the gut‐homing integrin α4β7 on Tfh cells, which diverted Tfh cells from systemic (non‐gut) inflamed sites such as the lung draining lymph nodes into the gut (the non‐immunopathological site) and so reduced systemic autoantibody production. Moreover, the impact of retinoic acid can go beyond the cellular level and impact the development of whole lymphoid tissues.92 It has been reported that cellular expansion in peripheral lymphoid tissues is controlled by gut microbiota in a retinoic acid‐dependent manner.93 The mucosal addressin MAdCAM‐1 is the receptor for the gut‐homing integrin α4β7, and peripheral node addressin PNAd is the receptor for CD62L.94 In neonatal mice, MAdCAM‐1 expression in lymph nodes is elevated shortly after birth,95 followed by a switch to a PNAd over a course of 2–3 weeks.93 Zhang et al. demonstrated that commensal fungi drive a wave of RALDH+ DCs to migrate to the peripheral lymph nodes after birth.92 The arrival of these cells increases the amounts of retinoic acid in situ, mediates the neonatal MAdCAM‐1 to adult PNAd addressin switch on endothelial cells, and directs the homing of lymphocytes to gut‐associated lymphoid tissues. Finally, the authors found that a diet deficient in vitamin A causes reduced homing of RALDH+ DCs into peripheral lymph nodes and a lack of maintenance of peripheral lymph node structures, suggesting a dependence on retinoic acid signaling for structural and functional maintenance of peripheral immune tissues. Microbiota and T‐cell subset determination Understanding the role of microbiota in T‐cell subset commitment and plasticity holds the key for unveiling the pathogenesis and therapeutic options for autoimmune diseases, as well as diseases associated with unbalanced immune responses such as cancers and infections. Not surprisingly, addressing this question has become a focused area of research in recent years. One study using transgenic mice expressing a limited but diverse TCR repertoire, through fixed TCR‐β expression, showed that the TCR repertoire of colonic Treg cells is unique compared with thymically derived Treg cells.96 Furthermore, these colonic Treg cells were shown to react with bacterial isolates, suggesting that encountering gut microbes in the intestines leads to peripheral Treg cell induction, and hence to tolerance to the gut microbiota.96 A more recent study, also using transgenic mice expressing a constrained TCR repertoire, through fixed TCR‐β expression and limited TCR‐α rearrangement (TCRmini mice97) showed that thymic and intestinal Treg cells expressed overlapping TCR repertoires, many of which recognize microbial antigens.98 Alteration of the gut microbiota through treatment of TCRmini mice with a cocktail of antibiotics concurrently altered the colonic and thymic Treg TCR repertoire.98 This suggests that the repertoire of thymically derived Treg cells is heavily influenced by the microbiota. Although these studies have differing results regarding the initial site of gut microbiota influence, in either case it is clear that direct recognition of microbial antigens by T cells can promote Treg cell lineage commitment. Beyond Treg cells, the gut microbiota can also influence T helper subsets. In the presence of a normal microbiota, CX3CR1+ intestinal antigen‐presenting cells act to limit Th1 cell development and expansion.99 However, antibiotic‐induced dysbiosis leads to a shift in the function of CX3CR1+ antigen‐presenting cells, allowing them to promote pathogenic Th1 cell development.99 Our laboratory and others have also demonstrated that a single type of commensal bacteria, segmented filamentous bacteria (SFB), can promote Th17 cell responses. In C57BL/6 mice, SFB specifically induce intestinal Th17 cells.25 More recent studies, however, have demonstrated that SFB can also influence T helper subsets outside the gut. In 2011, Lee et al. discovered that SFB alone was able to promote Th17 cells specifically in the spinal cord of EAE mice.33 Similarly, we have shown that in the K/BxN mouse model of autoimmune arthritis, SFB specifically promotes Th17 cell responses in the lung.100 Additionally, SFB can also drive the differentiation of Tfh cells and promote their migration to systemic sites, leading to exacerbated autoimmune arthritis in K/BxN mice.101 However, the exact molecule(s) employed by SFB to promote the differentiation of T helper cell subsets remain elusive. Although many microbiome studies focus on the microbes residing in the intestines, the entire digestive tract is colonized with bacteria. Hence, the oral microbiome has recently become an area of interest with regards to mucosal immunity. Data from Dutzan et al. showed that oral homeostatic Th17 responses, unlike gut Th17 responses, were unaffected by shifts in the oral or gut microbiomes.25, 102 This suggests that not all mucosal T cells are influenced by the microbiome, and that other environmental cues such as chewing activity in the oral cavity and chewing‐associated damage can maintain the Th17 cell population. Interestingly, a more recent study from this same group found that inflammatory Th17 cells isolated directly from periodontal lesions were in fact heavily influenced by oral microbiome dysbiosis.103 By focusing on Th17 isolated from periodontal lesions of both mice and humans, they found that increased abundance of Th17 cells and enhanced production of inflammatory cytokine IL‐17 was associated with shifts in the oral microbiome. Specific outgrowth of Enterococcus and Actinobacteria along with loss of Streptococcus was associated with periodontitis and enhanced Th17 cell responses in the ligature‐induced periodontitis mouse model. Together the studies suggest that only pathogenic but not homeostatic Th17 cell responses are influenced by dysbiosis. We recently addressed how both age and gut microbiota affect T‐cell subsets in the context of autoimmunity.104 Our results show an augmented autoimmune disease phenotype in both the joints and the lung of middle‐aged compared with young mice. Mechanistically, we saw a soaring accumulation of Tfh, but surprisingly not Th17 cells with age. Our data suggest exposure to immunomodulatory commensals such as SFB may allow the young host to develop an overactive immune system similar to that found in middle‐aged hosts. Our study also shows that age can independently increase the Tfh cell response without the help of immunomodulatory commensals. CD4 T helper cell plasticity and the microbiota The development of specific transcription factor reporter mice has proven to be an invaluable tool in dissecting the plasticity of T‐cell subsets. An early study using adoptive transfer of cells from Foxp3eGFP reporter mice showed that 80% of Treg cells lost GFP expression in the PPs compared with 50% in spleen.12 Within the GFP‐negative PP population, > 60% of the cells expressed CXCR5, consistent with a Tfh phenotype,12 suggesting that PP, but not spleen, is the preferred site for a Treg to Tfh reprogramming. Two other studies found that retinoic‐acid‐receptor‐related orphan receptor γ t (Rorγt), the hallmark Th17 transcription factor, was preferentially expressed by colonic Treg cells, and furthermore, that in the absence of gut bacteria, the Rorγt+ Treg cell population was substantially diminished.10, 14 Interestingly, several bacterial species were able to induce Rorγt+ Treg cells, which were indispensable for maintaining gut homeostasis as determined by enhanced IL‐17, interferon‐γ (IFN‐γ), and colitis in their absence.10 Another study showed that the pathogenic bacterium Helicobacter hepaticus promoted expansion of Rorγt+ Treg cells in a c‐MAF‐dependent manner, which selectively suppressed pro‐inflammatory Th17 cells in the intestines.105 However, in the absence of Treg cells (in Il10−/− mice), H. hepaticus predominantly induced Th17 cells, suggesting that pathobionts can have an impact on T‐cell plasticity by expanding potent suppressive T cells, which in turn, inhibit the inflammatory T helper cells necessary for pathogen elimination. Plasticity among T helper subsets has also been observed. Using cell fate reporter mice in which Th17 cells are permanently marked by YFP (Il17creR26eYFP), Hirota et al. found that selectively in the PP, but not spleen, Th17 cells were reprogramed to a Tfh (PD‐1hi CXCR5hi) phenotype.13 Furthermore, these ex‐Th17‐Tfh cells localized in germinal centers where they assisted in T‐cell‐dependent IgA production. In a Th17 adoptive transfer model of colitis, it was demonstrated that Th17 cell conversion to a Th1 phenotype was necessary for colitis development as Th17 cells unable to produce IFN‐γ (Ifng−/−) were incapable of inducing colitis.106 Interestingly, not only did Th17 cells convert to Th1 and contribute to colitis development directly, but also Th17 were capable of providing help to naive T cells for Th1 lineage commitment. Together these studies suggest that gut microbiota is able to influence T‐cell plasticity within the gut microenvironment and influence the development of autoimmunity and other inflammatory diseases. Gut microbiota induces autoimmunity by TCR‐mediated mechanisms Central to T‐cell function is the TCR, through which T cells recognize their cognate antigen and become activated. An important feature of the TCR is its ability to recognize multiple antigens, known as TCR cross‐reactivity. Sequencing‐based methods estimate that humans and mice express 107 and 106 productive unique αβ TCRs, respectively, yet this is insufficient to cover the possible foreign antigen repertoire.107, 108 Therefore, TCRs possess the ability to recognize multiple antigens in order to recognize the plethora of possible foreign antigens (see ref. 109). One potential mechanism through which pathogenic autoreactive T cells may become activated is through weak recognition of a microbial antigen for which their TCR is cross‐reactive. This is also known as molecular mimicry, a phenomenon in which a foreign antigen shares significant structural or sequence similarities with a self‐antigen.15, 16 Later, this now activated autoreactive T cell can home to the tissue where its cognate self‐antigen is expressed and elicit autoimmune pathology. In fact, cross‐reactive TCR recognition of microbial peptides by autoreactive T cells has been seen in several autoimmune settings (Fig. 1). In 2015, Horai et al. showed that T cells restricted to expressing only the retina‐specific TCR (R161H), became activated in the intestines in response to an unidentified microbial antigen.110 T‐cell activation was in response to TCR‐signaling to a non‐cognate microbial antigen, as T cells in the small intestines of R161H‐Tcrα−/− mice had clear co‐localization of phosphorylated Zap‐70 with CD4, known to be downstream of TCR signaling.110-112 Later, it was shown that Islet‐specific glucose‐6‐phosphate catalytic subunit‐related protein (IGRP)‐specific CD8+ T cells in non‐obese diabetic (NOD) mice were activated through recognition of a peptide from Fusobacteria and elicited autoimmune diabetes.113 Interestingly, the Fusobacteria peptide shared significant homology with the IGRP peptide targeted by the IGRP‐specific transgenic TCR,113 suggesting that molecular mimicry in microbial peptides may pose a threat because they can activate pathogenic autoreactive T cells. Most recently, it was shown that gut commensal bacteria expressing an ortholog to human Ro60, a nuclear RNA‐binding protein and primary targeted self‐antigen in lupus, were able to activate human Ro60‐specific CD4+ T cells through cross‐reactive TCR recognition.114 Furthermore, germ‐free mice monocolonized with Ro60‐ortholog‐expressing commensal bacteria spontaneously developed lupus‐like disease, as indicated by glomerular immune complex deposits. Together, these studies suggest that molecular mimicry is a mechanism through which the intestinal microbiota is able to propagate autoimmune responses leading to disease. In addition to molecular mimicry, it has long been thought that pathogenic autoreactive T cells may arise from a subset of thymocytes expressing two unique TCRs, one of which allows them to evade negative selection along with one that has a high affinity for a self‐antigen.115-117 Approaching from a different angle, outside the central immune system, we found that in peripheral tissues, SFB selectively expand dual TCR‐expressing T cells, leading to the augmentation of lung autoimmune pathology (Fig. 1).100 This is the first study to show that dual TCR T cells can be activated through the recognition of the cognate commensal‐derived antigen and migrate to elicit autoimmune pathology in the lung, a gut‐distal site. At the molecular level, we demonstrate that SFB selectively expand autoimmune T cells co‐expressing SFB‐specific TCRs in addition to their self‐reactive TCRs. This additional SFB‐specific TCR provides a proliferative advantage for autoreactive Th17 cells in SFB(+) hosts. Our data suggest that the induction of robust SFB‐specific Th17 response by transferring SFB(+) recipients with SFB‐specific 7B8 T cells could not rescue the autoimmune activity of Rag−/−.KRN T cells through bystander activation. Instead, our data suggest that the same T cells need to specifically recognize self‐ as well as SFB‐antigen to induce IL‐17‐expressing pathogenic T cells to cause autoimmunity. This is supported by previous data showing that in healthy mice, cognate TCR recognition of SFB‐antigen is crucial for SFB‐mediated Th17 induction.118, 119 Taken together, our study suggests that the same T cell needs to recognize both self‐antigen and be activated by SFB to undergo preferential Th17 cell expansion and so enhance autoimmunity in an SFB‐dependent manner. Beyond molecular mimicry, our study suggests an alternative, dual TCR‐based mechanism for commensal‐mediated autoimmunity. Role of gut microbiota in human and murine gut‐distal autoimmunity Autoimmune disorders have been sharply increasing worldwide in recent decades.26 The dramatic changes in the disease onset rate cannot be explained by genetic basis as it occurs in such a short period of time. On the contrary, these data suggest that environmental factors play a key role in the recent surge of autoimmune diseases. In recent years, we have begun to appreciate that the gut microbiota provides environmental cues controlling human health and disease, including its potential impact on autoimmune responses. Under healthy conditions, the human intestinal microbiota is composed primarily of bacteria of the phylum Firmicutes, with Bacteroidetes and Actinobacteria also represented.120 On the family and species level, the gut microbiota varies widely among individuals, although it remains largely stable within an individual over the course of several years.121 Studies using mouse models or samples from human subjects transplanted into mice have shown that age, gender, diet, smoking, autoimmunity, and other factors can alter the microbiota, and that the altered microbiota can be transferred and exert influence over the metabolism and immune system of the recipient.122 These studies demonstrate that the host–microbiota relationship is not a one‐way street, but rather a dynamic conversation that can have long‐lasting impacts. Many studies have investigated a correlation between gut microbiome composition and autoimmune disease in humans, but these studies are limited in their ability to determine causality. Exploring this field using animal models allows for perturbation of the gut microbiome to better distinguish cause from effect, permit mechanistic dissection and allow pre‐clinical evaluation of suggested therapeutic strategies. Throughout this section, we will highlight the important observations made in both human patients and mouse models. Multiple sclerosis (MS) is an autoimmune disease in which pathogenic CD4+ T cells penetrate the blood–brain barrier and cause damage to the central nervous system.123 Numerous studies have shown that patients with MS display intestinal microbiome dysbiosis.124-126 Two independent studies comparing patients with relapsing–remitting MS and healthy controls found significant alterations in the composition of the microbiome from patients with MS.124, 127 Perhaps even more intriguing, when compared with treatment‐naive patients, those receiving treatment displayed some restoration in their microbiome composition.127 Untreated MS patients displayed increased relative abundance of Methanobrevibacter and Akkermansia, which were lower in healthy controls and treated MS patients. MS patients receiving treatment had increased Sutterella, which has been previously observed to be increased in healthy controls compared with treatment‐naive patients with inflammatory bowel disease (Fig. 2).128 This suggests that although dysbiosis may contribute to a predisposition for developing MS, perhaps treatment may act to normalize pro‐inflammatory microbiota. A recent study by Cekanaviciute et al. showed that intestinal microbiome samples from MS patients displayed specific outgrowth of Acinetobacter and Akkermansia.129 Importantly, they further demonstrated that in the presence of Acinetobacter calcoaceticus, human peripheral blood mononuclear cells preferentially shifted towards an inflammatory phenotype (IFN‐γ) and away from a regulatory phenotype (Foxp3+ CD25+). This suggests that the specific dysbiosis observed in patients with MS favors an inflammatory T‐cell response. More definitive studies using the EAE mouse model of MS have shown that in germ‐free mice lacking gut microflora, EAE disease severity is reduced to an almost undetectable level, due to a more than twofold reduction in IFN‐γ and IL‐17 production from CD4+ T cells.33 Importantly, if CD4+ T cells from germ‐free mice were transferred to specific pathogen‐free mice harboring a diverse intestinal microflora, these cells gained the ability to mediate EAE disease development, suggesting that the gut microbiota is necessary to elicit the pathogenic Th1 and Th17 responses seen in EAE and perhaps MS.33 Importantly, it was recently demonstrated that colonization of germ‐free Myelin oliodendrocyte glycoprotein (MOG)‐specific TCR transgenic mice (RR mice) with gut microbiota from patients with MS resulted in increased incidence of spontaneous EAE development.130 This suggests that the microbiota of patients is sufficient to precipitate autoimmune disease. Together these studies suggest a gut–brain axis of communication that allows gut microbes to elicit autoimmune responses in the brain. Rheumatoid arthritis (RA) is an autoimmune disease that primarily affects joints but may also affect other parts of the body including the lungs and heart.131-133 Several studies have suggested that many patients with RA display antigen–antibody complexes with complement fixation in joints that can cause tissue damage and tumor necrosis factor‐α production.134-139 Recent studies also suggest that anti‐citrullinated protein antibodies are involved in RA pathogenesis.140, 141 Tfh cells are a T effector cell type specialized in helping B cells.142-144 Over‐reactive Tfh responses pose a threat of triggering excessive autoantibody production and autoimmune disease.145-149 In that regard, the presence of circulating CXCR5+ Tfh (cTfh) cells in the blood has been observed in patients with different autoimmune diseases, including RA.20 In a recent study assessing 77 treatment‐naive patients with RA, 21 treated patients with RA, and 80 healthy controls, the gut microbiome was found to be altered in patients with RA.150 In particular, Haemophilus spp. were depleted in individuals with RA and negatively correlated with serum autoantibodies titers, whereas Lactobacillus salivarius was over‐represented in individuals with RA and abundance correlated with disease severity (Fig. 2). Interestingly, similar to what has been observed in patients with MS, patients with RA undergoing treatment displayed partial normalization of their intestinal microflora compared with treatment‐naive patients. By using a photo‐labeling mouse model to track cell migration, our laboratory has demonstrated that a commensal bacterium, SFB, is able to promote the differentiation and migration Tfh cells from the PPs in the small intestine to systemic sites where they elicit autoimmune arthritis development.101 At the molecular level, SFB induce PP Tfh cell differentiation by limiting the access of IL‐2 to PP CD4+ T cells and DCs are required for SFB‐mediated IL‐2Rα suppression and up‐regulation of Bcl‐6, a master regulator of Tfh cells, in PPs. Furthermore, we have recently shown that SFB can also promote RA‐associated autoimmune lung disease through propagation and mobilization of gut‐derived Th17 cells.100 Importantly, this highlights a gut‐lung axis of communication through which specific gut commensal bacteria are able to promote pathogenic T‐cell responses leading to gut‐distal autoimmune pathology. In the collagen‐induced arthritis model of RA, dysbiosis characterized by outgrowth of Lachnospiraceae and Lactobacillaceae was detected before arthritis development.151 Furthermore, antibiotic‐induced depletion of the microbiota before induction of collagen‐induced arthritis reduced disease severity by 40% and was accompanied by reduction in the inflammatory cytokines IL‐17, IL‐22, and IL‐23 in the intestines along with reduced anti‐collagen antibodies. This suggests that the gut microbiota is necessary for autoimmune disease development and that alterations in the composition of the gut microbiota precipitate autoimmune disease. Systemic lupus erythematosus (SLE) is an autoimmune disease affecting almost all organs of the body, in which both autoreactive T cells and autoantibody‐producing B cells contribute to the pathogenesis of SLE.152 Like many gut‐distal autoimmune diseases, recent studies have linked gut microbiota dysbiosis with SLE.153 In a 2016 study from Lopez et al., it was shown that intestinal microbiota isolates from patients with SLE promoted naive T‐cell differentiation into Th17 cells.154 Furthermore, peripheral blood mononuclear cells from patients with SLE were enriched for both Th17 and Foxp3+ IL‐17+ cells compared with healthy controls. As described earlier, due to T cell plasticity, Treg cells are able to shift between regulatory and potentially pathogenic phenotypes. Hence, the presence of Foxp3+ IL‐17+ cells in the peripheral blood of patients with SLE suggests that the shift observed in the gut microbiome of these patients may promote inflammatory T‐cell responses and may negatively impact Treg cell stability. More recently, in a study by Luo et al. which assessed the intestinal microbiota dynamics of both SLE patients and the NZB/W F1 mouse model of SLE, it was shown that alteration in the gut microbiota resulted in decreased diversity with increased representation of Gram‐negative bacteria (Fig. 2).155 Interestingly, as NZB/W F1 mice progressed to overt disease, the relative abundance of Lactobacillaceae increased from 0·1% to 10%. Furthermore, there was a positive correlation between Lactobacillaceae abundance and disease severity in female NZB/W F1 mice. This suggests that reduction in α diversity, a measure of the number of different species present in a given site, and outgrowth of specific commensal microbes may promote autoreactive immune responses in susceptible individuals. Another report found a high frequency of Th17 cells in the kidneys of patients with glomerulonephritis, a form of SLE.156 By using Kaede mice to track intestinal T‐cell migration during glomerulonephritis induction, they observed that Th17 cells egress from the intestine and subsequently migrate to the kidney through CCR6 recognition of CCL20. Furthermore, depletion of gut Th17 cells in germ‐free and antibiotics‐treated mice ameliorated the autoimmune glomerulonephritis, suggesting that targeting the intestinal Th17 cells may offer a therapeutic strategy for autoimmune diseases. Hidradenitis suppurativa (HS) is a chronic inflammatory skin condition of the hair follicles thought to be mediated in large part by Th17 cells (Fig. 2).157 With respect to perturbations in the microbiota, most studies assess changes in the cutaneous microbiota as this is the site of the disease manifestation. Recently, HS has been identified to have a strong concurrence rate with inflammatory bowel disease, with over 17% of Crohn's patients also having HS.158 Hence, while there are definite changes in the cutaneous microbiome of patients with HS,159 taking a closer look at the gut microbiome may prove fruitful in the quest for identifying a causal mechanism. Type 1 diabetes (T1D) is a T‐cell‐mediated autoimmune disease characterized by the selective destruction of insulin secreting β‐cells in the pancreatic islets.160 Many patients with T1D present with increased intestinal permeability or ‘leaky gut’, which has been shown to precede the onset of clinical disease.161 In addition, patients with T1D also display an altered intestinal microbiome characterized by decreased α diversity and increased abundance of inflammatory species compared with healthy controls (Fig. 2).162 Hence, it is tempting to assume that in combination with increased intestinal permeability, dysbiosis could promote an inflammatory environment in the gut–proximal pancreatic tissue leading to activation of an autoimmune response. In a 2016 study from Costa et al., it was determined that autoimmune diabetes development in the streptozotocin‐induced model of T1D is dependent on translocation of gut microbiota to the pancreatic lymph node.163 Interestingly, islet‐infiltrating T cells seem to be of gut origin as they often express α4β7 integrin.83, 164, 165 We now know that the effect of commensal bacteria on autoimmune disease can be model dependent. For example, SFB is pathogenic in animal models of arthritis and multiple sclerosis32, 33 but protective in the NOD mouse model of type1 diabetes.166, 167 The beneficial and detrimental effects of SFB could depend on whether commensal‐mediated immunomodulation is enhancing or inhibiting the pathogenesis of each disease. For example, autoantibodies are a key pathogenic factor and diseases can be induced by passive transfer of autoantibodies in the K/BxN and EAE models.168-170 In contrast, type 1 diabetes in NOD mice is a T‐cell‐mediated autoimmune disease, and although B cells of NOD mice produce autoantibodies, these are not thought to play a diabetogenic role.171, 172 Hence, SFB, with their Tfh and autoantibody boosting effect, are more likely to play a pathogenic role in the K/BxN model and other antibody‐mediated autoimmune diseases than in T‐cell‐mediated autoimmune diseases such as in NOD mice. One feature that many of the autoimmune diseases have in common is evidence of gut‐derived T cells participating in the autoimmune response.83, 101, 156, 164, 165 This would suggest that perhaps one mechanism through which the gut microbiota modulates autoimmune responses is through mobilization of T cells from the tolerogenic environment of the gut to systemic sites where alleviated tolerogenic pressure allows these cells to become active and initiate autoimmune disease development.173-175 However, more studies are required to determine whether the altered microbiome drives changes in immunity or conversely the onset of immunological disease induces changes in the gut microflora in human patients. It is likely a combination, as changes in the gut microbiota precede onset of clinical autoimmune manifestations,162 yet some evidence suggests that perhaps T cells, Foxp3+ Treg cells in particular, regulate gut microbiota diversity.176 Conclusions and future directions There is a surge of recent studies investigating the molecular mechanisms of dysbiosis‐related autoimmune diseases. In both molecular mimicry and dual TCR theories, until recently, only infectious pathogens including viruses and bacteria have been implicated as the primary culprits,15-17 and little is known about the molecular mechanism by which asymptomatic commensal colonization could trigger autoimmunity. Our recent study along with others demonstrates a previously unknown condition that asymptomatic colonization by commensal bacteria in the gut is capable of triggering systemic autoimmune disease by molecular mimicking self‐antigen and skewing the dual TCR expression in the host. Notably, dual TCR expression is not limited to transgenic mouse models. In wild‐type mice and in humans, up to 15% and 33%, respectively, of peripheral T cells have been reported to express dual TCRs, due to incomplete allelic exclusion at the Tcra locus.177, 178 Future investigation is required to determine whether dual TCR expression is involved in dysbiosis‐related immune disorders in human patients. Another important, but poorly understood, aspect of microbiota–T‐cell interaction is how different microbial signals impact T‐cell stability and plasticity at the mucosal frontline. Understanding the plasticity and stability of T‐cell subsets in mucosal tissues is highly relevant for future therapeutic strategies as we will not be able to manipulate T cells for therapeutic purposes without understanding the mechanisms by which T cells shift between alternative programs because of exposure to gut microbes. As illustrated in this review, the use of animal models provides a powerful tool for mechanistic studies as it allows us to establish the causative effect of microbiota in disease development. By using germ‐free and specific pathogen‐free mouse models, numerous studies unravel the crucial roles of gut commensals in immune regulation in the context of health and disease. Although these studies provide promising clinical implications, various challenges need to be overcome by the field to harness these findings for future diagnostic and therapeutic approaches. One of the challenges is that the reductionist experiments in mice often do not reflect important features of the immune system in the adult human. In this regard, Beura et al. demonstrated that laboratory mice – like neonates, but not adult humans – lack effector‐differentiated and mucosal memory T cells.179 These cell populations are present in feral mice and pet store mice with constant and diverse microbial exposure. Furthermore, laboratory mice co‐housed with pet store mice display profound changes in immune systems, resulting in an immune signature that is more closely reflected by adult humans than by neonates. Hence, the ‘dirty’ mouse model may provide a unique advantage over the current reductionist mouse models and should be considered for studies with translational purposes in human disease. One of the biggest dilemmas facing the field of gut microbiota research is how to apply laboratory findings to clinical therapies. Currently, microbiota‐based therapy relies on three major strategies: complete fecal microbiota transplants (FMTs), administering one or more species of bacteria orally (probiotics), or administering substrates to favor the expansion of certain kinds of bacteria or a shift in metabolite production, a strategy known as prebiotics.180 Each of these approaches has limitations; for example, probiotics often colonize only transiently because of ineffective competition with the existing microbiota.181 Additionally, despite the focus on taxonomic composition analyzed by 16S rRNA sequencing of microbial communities in many studies, the functional relevance of microbial communities is more likely to be revealed by their metagenomic gene expression and metabolomics profiles. This is further proved by the fact that bacterial isolates of the same strain can display very different immunoregulatory capabilities.182 FMTs depend on severe depletion of the constituent microbiota through antibiotics and bowel flushing before administration of the transfer.183, 184 Other researchers are working to find alternative methods to circumvent the difficulties of FMTs, such as filtering donor material to the point of removing intact cells, leaving only bacterial byproducts such as metabolites.185 An additional approach is to identify bacterial metabolites able to modulate human immune responses and artificially administer the identified metabolites as novel therapeutics.186 This approach would be appealing from a pharmaceutical perspective, because it would allow the development of drugs that would need to be taken regularly. It would also minimize the factors that need to be controlled, since administering metabolites would not depend on the engraftment of a foreign bacterial species. However, a careful determination on how to achieve a comparable delivery of the metabolite that mimics the one produced by the intestinal microbiota will be crucial for the effectiveness of the treatment. In conclusion, the gut microbiota is a complex community that engages in significant cross‐talk with the host. A better understanding of host–microbe interaction and the underlying microbiota‐derived molecules that modulate the immune system and disease development may help to pave the way for better patient‐tailored interventions and microbial molecule‐based therapies for immune disorders. Disclosures The authors have no competing interests to declare. References

3 RESULTS A total of 10 638 citations were identified from the electronic and manual search. After removing duplicates (n = 767) and non‐applicable references (n = 9620), 251 citations were screened, of which 92 met the inclusion criteria (Figure 1).14-105 Baseline characteristics of the included manuscripts, study quality and studies reporting negative results are provided in Table S1, S2 and S3 respectively. Data from several included studies are only presented in tables.14-29 Figures 2 and 3 provide an overview of different predictive molecular markers grouped by underlying biological mechanism. 3.1 Anti‐TNF therapy (IFX // ADA // GLM // CZP) 3.1.1 (Pharmaco) genomics An overview of predictive genetic biomarkers for anti‐TNF response is presented in Figure 2 and Tables 1 and 4. Reference (year) Response definition Biomarker Association Predictive value Bank (2014) Clin Resp: based on three‐step scale CD SNP rs5030728 (TLR4 G>A) Clin NR OR (adj): 0.38 [0.15‐0.96] P = 0.04 SNP rs352139 (TLR9 1174G>A) Clin NR OR (adj): 0.38 [0.16‐0.94] P = 0.04 SNP rs4149570 (TNFRSF1A‐609G>T) Clin Resp OR (adj): 5.16 [1.14‐23.39] P = 0.03 UC SNP rs11938228 (TLR2 C > A) Clin NR OR (adj): 0.26 [0.08‐0.92] P = 0.04 SNP rs4696480 (TLR2 A > T) Clin NR OR (adj): 0.29 [0.12‐0.70] P = 0.01 SNP rs3804099 (TLR2 597 T>C) Clin Resp OR (unadj): 2.47 [0.98‐6.23] P = 0.05 SNP rs5030728 (TLR4 G > A) Clin Resp OR (unadj): 2.59 [1.08‐6.18] P = 0.03 Bank (2018) Clin Resp: three‐step scale to reflect max response 22 W after initiation CD heterozygous genotype of TLRS 936 T>C (rs5744174) NR OR: 0.24 [0.11‐0.53], P = 0.008 homozygous variant genotype of IFNGR2 T>C (rs8126756) NR OR: 0.36, 95% CI: 0.16‐0.81, P = 0.01 UC heterozygous genotype of IL128B ‐ 10993 G>C (rs3212217) NR OR: 0.09, 95% CI; 0.01‐0.65, P = 0.02 heterozygous genotype of IFNGRI ‐ 56 T> C (rs2234711) NR OR: 0.29, 95°/o CI: 0.11‐0.78, P = 0.01 homozygous genotype of TBX21 ‐ 1514 T>C (rs17250932) NR OR: 0.06, 95% CI: 0.01‐0.80, P = 0.03 homozygous genotype of JAK2 T>C (rs12343867) NR OR: 0.17, 95% CI: 0.03‐0.85, P = 0.03 hetero+homoz. genotypes of IL128 −10993 G > C (rs3212217) NR OR: 033, 95% CI: 0.15‐0.69, P = 0.004 hetero+homoz. genotypes of IL 18‐607 C > A (rs1946518) Clin Resp OR: 2.14, 95% CI: 1.06‐4.32, P = 0.03 CD +UC: heterozygous genotype of IFNGRI ‐ 56 T>C (rs2234711) NR OR: 0.57, 95% CI: 034‐0.96, P = 0.04 homozygous genotype of IL 126 −10993 G > C (rs3212217) Clin Resp OR: 10.40, 95% CI: 131‐82.41, P = 0.03 hetero+homoz. genotypes NLRP3 29940 C > G (rs10754558) Clin Resp OR: 1.60, 95% CI: 1.02‐2.52, P = 0.04 Barber (2016) Clin PNR: Clin NR by W12 or Th cessation within 4 months of induction. DR: Clin response ≥24 M prediction model: 15 SNPs (GRS 5‐20) Clin PNR AUC 0.929; OR 2.65 [1.94‐3.61], P = 6 x 10‐10 prediction model: 16 SNPs (GRS 6‐23) Clin DR AUC 0.83; OR 1.6 [1.41‐1.83], P = 2 x 10‐12 prediction model: clinical and genome Clin DR AUC 0.85 prediction model: clinical and genome Th cessation HR 0.89 [0.85‐0.93], P = 4 x 10‐8 Billiet (2016) Biochem; Not Specified FCGRT promotor VNTR2/VNTR3 genotype Drug expos Not specified Dezelak (2016) Clin Resp: IBDQ >170 or increase in >22 pts compared to baseline. Biochem Resp: CRP <3 mg/L or decrease >25% from baseline SNP rs9373839 (ATG5) Allele C and haplotypes C/C & C/T Clin Resp 4.2‐fold higher chance for CResp; P = 1.13x10‐0 SNPs rs9373839 and rs510432 (ATG5) haplotype C/G Clin Resp 4.9% higher IBDQ score; P = 1.31x10‐06 SNPs rs9373839 and rs510432 (ATG5) Allele C and haplotypes C/G Clin Resp 36% higher chance for CResp; P = 8.34x10‐7 SNP rs1130864 (CRP) ≥1 allele C Clin&Biochem Resp 10‐fold higher chance for CResp and fourfold higher chance for BResp Gonsky (2014) Clin; Not Specified SNP rs1861494 (IFNG +2109) T allele carriage (T/T and T/C) vs C allele (C/C) ASCA+ P < 0.015, No quant pred data Haritunians (2009) MR‐UC: UC requiring colectomy (refractory to drugs). Non‐MR‐UC: colectomy for cancer or no colectomy. Prediction model: 46 SNPs MR‐UC 48% of the variance for colectomy explained. HR 1.313; AUC 0.91; Sens 0.793; Spec 0.858 Score of 44 & 47 generated test Sens and Spec of over 90% Hlavaty (2005) Clin Resp (partial): LCD: decrease≥70 pts (W4). FCD: decrease ≥50% in fistula drainage (W10) Clin Rem: LCD: CDAI<150 (W4). FCD: cessation of fistula drainage (W10). SNP rs1800682 (FAS) haplotype C/C or C/T vs T/T (minor allele: T) Clin Resp OR 0.11 [0.08‐0.56]; P = 0.002 SNP rs4645983 (CASP9) allotypes T/T vs C/C or C/T (minor allele: T) Clin Resp OR 1.50 [1.34‐1.68]; P = 0.04 SNP rs763110 (FAS)Fas ligand −843 CC/CT genotype Clin Resp OR 1.66 [1.21‐2.29], P = 0.002 Hlavaty (2007) Clin Resp (partial): LCD: decrease by ≥70 points at W4. FCD: decrease ≥50% fistula drainage at W10. Clin Rem: LCD: CDAI <150 at W4. FCD: complete cessation of fistula drainage at W10. Prediction model: Apoptotic pharmacogenetic index (API) (ranging from low to high apoptotic response (0‐3)) consisting of three SNPs: Fas ligand −843 C/T (0 pt: if TT and 1 point if CT/CC) Fas −670 G/A (0 pt if TT and 1 pt if AG/AA) Caspase9 93 C/T (0 pt if TT/CT and 1 pt if CC) Clin Resp API ≤1 (n = 59): R in LCD (50%) and FCD (39.5%). API of 2 (n = 158): CR and CResp in LCD (56.1% & 73.8%) and in FCD (44.9% & 85.7%). API =3 (n = 10), CR and CResp in LCD (85.7% & 100%) and in FCD (0% & 100%) No quant pred data Koder (2015) Biochem Resp: decrease of CRP to <3 mg/L or drop in CRP >25% SNP rs10210302 (ATG16L1) genotypes C/T or T/T vs C/C Biochem Resp OR 9.44 [2.49‐35.83], P(corrected) = 0.027 Lopez‐Hernandez (2014) Clin Resp: improvement in HBI or clinical observation. Clin NR: no change in HBI or exacerbation SNP rs1800629 (TNFα −308G/A) allele A and G/A vs G/G Clin NR Cp<0.05 No quant pred data %Resp vs %NR: Clin NR Clin Resp Clin NR ‐308A allele: 6% vs 27% Cp=0.026 ‐308GG genotype: 88% vs 44% Cp=0.029 ‐308GA genotype: 12% vs 55% Cp=0.017 No quant pred data Louis (2004) Biochem Resp (complete): normal CRP at W4. Biochem Resp (partial): decrease ≥25% at W4 SNP rs396991 (FCGR3A) allotype T/T vs T/G or G/G (WT) Biochem Resp RR 1.43 [1.27‐1.61], P = 0.0002 Louis (2006) Biochem Resp: ≥25% reduction in CRP at both W2 and W6 from BL SNP rs396991 (FCGR3A) allotype V/V vs V/F and F/F Biochem Resp W6 CRP reduction: −79.4%, −76.5% and −64.3%; P = 0,085, No quant pred data Mascheretti (2002) Clin Resp: cohort 1: decrease CDAI >70 points from baseline to W4; cohort 2: decrease CDAI>70 points from baseline to W2. Clin Rem: CDAI<150 SNP rs1800629 (TNF), haplotype G/A or A/A vs G/G (minor allele: A) SNP rs1061622 (TNFRSF1B), haplotype T/G or G/G vs T/T (minor allele: G) Clin Resp Clin Resp Not specified OR 0.88 [0.57‐1.3] Matsukura (2008) Clin Resp: improvement in HBI or clinical observation. Clin NR: no improvement or exacerbation SNP rs767455 (TNFRSF1A), A/G and G/G vs A/A (WT) SNP rs1061624 & rs3397) (TNFRSF1B), rs1061624 A allele and rs3397 T allele Clin NR Clin NR OR 0.22 [0.06‐0.76] & OR 0.26 [0.08‐0.90]; P < 0.05 5% Resp vs 18% NR, χ2 = 6.4, P = 0.01, No quant pred data Medrano (2014) Clin Resp (partial): LCD: decrease HBI >3 and no CS. FCD: reduction ≥50% in the number of draining fistulas. Clin Rem: LCD: HBI<4 and no CS. FCD: complete closure of fistulas. SNP rs1061624 (TNFRSF1B), haplotype A‐rs3397_T Clin NR OR 1.78 [1.09‐2.90], P = 0.015 Moroi (2013) Biochem Resp: ΔCRP & ΔCRP% SNP rs396991 (FCGR3A‐158), allotype V/V vs V/F or F/F Biochem Resp ΔCRP, P < 0.001; ΔCRP%, P = 0.044, No quant pred data Netz (2017) Clin Resp: in accordance with the treating physician decision Fas ligand SNP (rs763110) A (minor) TNF gene −308 allele NR NR OR 4.30 [1.45‐12.8], P = 0.009 OR 2.88 [1.1‐8.22], P = 0.049 Pierik (2004) Biochem Resp: >12 mg/L reduction in CRP. Biochem Rem: CRP <3 mg/L. SNP rs767455 (TNFRSF1A), A36G carriage Biochem Resp OR 0.47 [0.234‐0.946], P = 0.0343 Repnik (2016) Clin Resp: increase in IBDQ >22 or IBDQ >170 SNP rs2071303 (HFE), A/G or G/G vs A/A genotype Clin Resp W12: OR 3.467 [1.149‐10.463] P = 0.037. W20: OR 4.832 [1.519‐15.367] P = 0.009. Allele G vs A carriage: W20: OR 3.542 [1.355‐9.258] P = 0.0082, No quant pred data Spalinger (2016) Clin NR: 1. breakthrough/LOR, 2. PNR 3. Stop due to side effects/intolerance. SNP rs1893217 (PTPN2), variant haplotypes C/T or C/C vs T/T Clin Resp (n)Resp vs (n)NR Genotype: CC: 7 vs 5 ‐ CT: 37 vs 39 ‐ TT: 71 vs 109 Existence of C‐allele: CC/CT: 44 vs 44 ‐ TT: 71 vs 109 P = 0.05, No quant pred data Steenholdt (2012) Clin Resp: PGA +biochemical markers SNP rs1061622 (TNFRSF1B) minor allele carriage SNP rs976881 (TNFRSF1B), minor allele carriage Clin Resp DR LOR OR 4.2 [1.2‐18.2], P = 0.014 OR 5.5 [1.5‐25.5], P = 0.007 OR 3,3 [1.2‐9.1], P = 0.014) (Homozygosity: OR 19, P = 0.006) IBD: inflammatory bowel disease; CD: Crohn's disease; UC: ulcerative colitis; LCD: luminal Crohn's disease; FCD: fistulising Crohn's disease; MR‐UC: medical refractory ulcerative colitis; M: months; W: weeks; D: days; FU: follow‐up; CDAI: Crohn's Disease Activity Index; CAI: colitis activity index; HBI: Harvey Bradshaw Index; PCDAI: Paediatric Crohn's Disease Activity Index; PUCAI: Paediatric Ulcerative Colitis Activity Index; PGA: Physician's Global Assessment score; ULN: upper limit of normal; Hb: haemoglobin; CRP: C‐reactive protein; ESR: erythrocyte sedimentation rate; FC: faecal calprotectin; CK: cytokine; PBMC: peripheral blood mononuclear cell; TNF: tumour necrosis factor; IFX: infliximab; ADA: adalimumab: UST: ustekinumab; PCB: placebo; CS: corticosteroids A‐D‐A: anti‐drug antibodies; Clin: clinical; Biochem: biochemical; Rem: remission; Resp: Response; NR: nonresponse; PNR: primary nonresponse; DR: durable response; Sust: stained; LOR: loss of response; Th cessation: Therapy cessation; Drug expos: Drug exposure; SSFR: sustained steroid free remission; RB: rectal bleeding; pts: points; OR: odds ratio; HR: hazard ratio; Sens: sensitivity; Spec: specificity; PPV: positive predictive value; NPV: negative predictive value; SNP: single nucleotide polymorphism; WT: wild type; GRS: genetic risk score; LH: likelihood; FCGRT: Fc Fragment Of IgG Receptor And Transporter; MPV: mean platelet volume; IL: interleukin; Cp: Corrected p‐value; (n): number; No quant. pred. data: No quantitative predictive data. Fc‐receptors Several studies reported on the association between FCGR3A genotype and response to anti‐TNF treatment.28-31 The V/V allotype was associated with a significantly higher proportion of CD patients (100%) achieving a CRP decrease in response to infliximab compared to the V/F or F/F allotypes (70%).30 This finding was confirmed by Moroi et al and in a sub‐analysis of the ACCENT I trial.29, 31 Due to close location of the FCGR3A and the CRP genes, a potential false‐positive association was investigated and excluded.32 Genetic variations within the neonatal Fc‐receptor (FcRn gene) were associated with pharmacokinetic changes for IFX and ADA. Heterozygosity for the VNTR2/VNTR3 genotype led to a 14% (P = 0.03) and 25% (P = 0.007) reduction in induction drug concentrations compared to homozygosity VNTR3 for IFX and ADA respectively.33 ATG16L1 The autophagy related gene ATG16L1 was linked to anti‐TNF response in two studies.28, 34 Patients with the ATG16L1 T/T and C/T genotypes had a significantly greater CRP response to adalimumab than patients with the C/C genotype (85% vs 37.5%; OR 9.44 [2.49‐35.83], P = 0.027).28 More recently however, the association between autophagy and anti‐TNF response was not confirmed in a large retrospective study.35 Apoptosis A number of studies reported an association between genes related to apoptosis and anti‐TNF response.28, 34, 36, 37 Fas ligand TT genotype and caspase‐9 CC/CT genotypes were associated with low response to IFX in luminal CD. In fistulising CD, Fas ligand‐843 CC/CT genotype was the only predictor of response.36 Conversely, the CC genotype vs a TC or TT genotype was associated with primary nonresponse in CD (OR 4.3 [1.45‐12.8], P = 0.009).38 TNF/TNF‐receptor genes Several studies looked at TNF‐signalling pathways and anti‐TNF response.39-47 A genetic alteration in the TNF region (the lymphotoxin α locus, also known as TNF‐β) was associated with clinical nonresponse to a single infusion of infliximab in CD (P = 0.007).39 Other studies found no significant association between variation in the TNF region (eg, the TNF promoter region) and treatment response.40, 41 Recently however, the presence of the minor allele (A) for the TNF gene (rs1800629) was associated with primary nonresponse (OR 2.88 [1.1‐8.22], P = 0.049).38 Several studies reported on single nucleotide polymorphisms (SNPs) within the TNF‐receptor superfamilies (TNFRSF) 1A and 1B. Regarding TNFRSF1A, higher response rates were observed in CD patients homozygous for the TNFRSF1A‐609G>T genotype (OR 5.16 [1.14‐23.39], P = 0.03).42 Minor allele carriage (G) of SNP rs767455 was associated with low clinical anti‐TNF response (OR 0.47 [0.23‐0.95], P = 0.034), a finding that was replicated by a different study group (OR 0.26 [0.08‐0.90]).43, 44 With respect to TNFRSF1B, an increased frequency in the haplotype conformed by two SNPs (rs1061624_A‐rs3397_T) was found in anti‐TNF nonresponders in both a Japanese and Caucasian CD population (P = 0.01 and OR 1.78 [1.09‐2.90], P = 0.015, respectively).43, 45 Conversely, minor allele (G) carriage of the TNFRSF1B SNP (rs1061622) was associated with improved initial (OR 4.2 [1.2‐18.2], P = 0.014) and sustained clinical response to anti‐TNF (OR 5.5 [1.5‐25.5], P = 0.007).46 However, the latter association (OR 8.5 [0.9‐206.3]) could not be confirmed.47 Recently, a similar association between this specific SNP (TNFRSF1B, rs1061622) and clinical response was reported (OR 1.52 [0.79‐3.04], P = 0.023).45 In addition, minor allele carriage of TNFRSF1B (rs976881) was also discovered to be associated with loss of response to IFX (OR 3.3 [1.2‐9.1], P = 0.014).46 Finally, Pierik et al did not find an association between TNFRSF1B (rs1061622) and anti‐TNF response.44 NOD2/CARD15, CRP, IL23R and IL12 No significant association with treatment response was found either for the IBD susceptibility variants of NOD2/CARD15 or for variations of the CRP genotype.32, 48-51 UC patients carrying the homozygous IBD risk‐increasing IL23R SNPs were more likely to respond to infliximab than patients with the risk‐decreasing variants (74.1% vs 36.6%, P = 0.001).52 A recent study observed that the heterozygous genotype of IL12B—10993 G > C (rs3212217) was associated with nonresponse within 22 weeks after initiation in UC (OR 0.24 [0.11‐0.53], P = 0.008).53 Genetic risk models Several predictive genetic models for treatment response to anti‐TNF therapy have been suggested.34, 37, 54, 55 Hlavaty et al developed an algorithm—comprising three apoptotic SNPs (Fas ligand −843 C/T, Fas −670 G/A and caspase‐9 93 C/T)—to predict clinical response to the initial IFX infusion in CD.37 Another risk model with 46 SNPs, was found to be predictive of colectomy as a measure of refractory UC (Area under the curve value (AUC) 0.91, sensitivity 0.79 and specificity 0.85).55 Furthermore, three genetic IBD susceptibility SNPs, rs2836878 (21q22/BRWD1) combined with the microbial biomarker pANCA predicted primary nonresponse to anti‐TNF in paediatric UC patients (AUC 0.98, sensitivity 0.95, specificity 0.88 and positive likelihood ratio 8).54 Recently, two additional models were developed for primary nonresponse and sustained response to anti‐TNF treatment in CD. The genetic risk score—comprising 15 SNPs—was the only independent predictor of primary nonresponse (OR 2.65 [1.94‐3.61], P < 0.0001). In the risk model for sustained response—comprising 16 SNPs—each one‐point increase was associated with a 60% higher likelihood of sustained response (AUC 0.929; OR 1.6 [1.41‐1.83], P = 2x10–12).34 Recently, the combination of two‐risk genotypes, involving both apoptosis and the TNF region, was associated with primary anti‐TNF nonresponse (OR 4.76 [1.35‐16.77], P = 0.015).38 3.1.2 Gene expression An overview of predictive gene expression biomarkers is presented in Tables 2 and 4. Reference (year) Response definition Biomarker Association Predictive value Peripheral blood Li (2016) Endo Resp: Mayo/SES‐CD/Rutgeerts score; Not Specified lamina propria CD4+ T‐cell subsets. Endo Resp Reduced prevalence of Gata3‐expressing cells in the CD cohort (P = 0.002), No quant pred data Intestinal tissue Arijs (2009) Endo Resp: MH = Mayo ≤1. H Resp: Geboes ≤1 osteoprotegerin, (TNFRSF11B), stanniocalcin‐1 (STC1), prostaglandin‐endoperoxide synthase 2 (PTGS2), IL‐13 receptor α 2 (IL13Rα2) and IL‐11 Endo&Hist Resp Accuracy of 89.1% (41/46). Sens: 95% (19/20); Spec: 85% (22/26) Arijs (2010) Endo Resp: MH, Hist Resp: ≥ 3 pts decrease on the geboes score TNF induced protein 6 (TNFAIP6), S100A8, IL11, G0/G1switch 2 (G0S2) and S100A9 Endo&Hist Resp CDc: top 5 genes: (accuracy 100%). CDi: no association, No quant pred data Atreya (2014) Clin Resp: CDAI decrease ≥100 pts (12 W). Endo Resp: MH = absence of ulcers Number of membrane‐bound TNF (mTNF) expressing cells Clin&Endo Resp High mTNF‐group (≥20 cells/confocal image) vs low‐group (<20): higher probability of R (92% vs 15%). Accuracy 88%; Sens 92%; Spec 85%; PPV 85%; NPV 92% Halloran (2014) Endo Resp: total mayo of 0 or 1 Principal component 1 (PC1) (disturbance in gene expression seen in TCMR (kidney/heart) Endo Resp Higher pre‐treatment PC1 (UC: P < 0.03; CD: P < 0.001), No quant pred data Viazis (2017) Endo Resp: MH=endoscopic mayo 0 lower Th1‐defining T‐bet and higher Th17‐defining Rorc‐p Endo Resp P < 0.05, No quant pred data West (2017) Endo&H Resp: complete disease resolution BL Oncostatin M‐associated inflammatory module expression: Module‐low vs Module‐high Endo&Hist Resp Relative risk = 5, 95% CI = 1.4‐17.9, P = 0.0006 Yamada (2017) CD: Clin Resp: decrease ≥100 from BL, or <150 when BL was≤248 UC: Clin Resp: decrease ≥3 full Mayo or ≥2 points in clinical Mayo with decrease ≥1 in RB Higher expression of Bmi1 PNR Relative Bmi1 expression Resp vs NR: 1.28 ± 0.14 vs 2.10 ± 0.23, P = 0.0039, No quant pred data IBD: inflammatory bowel disease; CD: Crohn's disease; CDc: crohn's colitis; CDi; crohn's ileitis: UC: ulcerative colitis; M: months; W: weeks; FU: follow‐up; BL: baseline; CDAI: Crohn's Disease Activity Index; HBI: Harvey Bradshaw Index; SES‐CD: simple endoscopic score for Crohn's disease; TNF: tumour necrosis factor; IFX: infliximab; ADA: adalimumab: Clin: clinical; Endo: endoscopic; Hist: histological; Resp: response; NR: nonresponse; PNR: primary nonresponse; MH: mucosal healing; pts: points; TCMR: T‐cell‐mediated rejection of organ transplants; Sens: sensitivity; Spec: specificity; PPV: positive predictive value; NPV: negative predictive value; No quant pred data: No quantitative predictive data. Peripheral blood A reduced number of Gata3+‐expressing T helper cells was associated with endoscopic response to anti‐TNF in CD but not UC, which could be associated with a reduced number of eosinophils.56 Recently, a decrease in peripheral blood mononuclear cell (PBMC) expression of CD14 and CD86 at both week 2 and week 14 compared to baseline was shown to be associated with short term clinical and endoscopic response to infliximab.57 Intestinal tissue A mucosal gene signature for response to infliximab in both colonic CD and UC was identified.58, 59 In colonic CD, the top five differentially expressed genes (TNFAIP6, S100A8, S100A9, IL‐11 and G0S2) predicted endoscopic and histological response with 100% accuracy.59 In UC, a combination of five innate immunity gene expression profiles (TNFRSF11B, STC1, PTGS2, IL13Ralpha2 and IL11) predicted response with 89% accuracy (sensitivity 95% and specificity 85%).58 An association between mRNA patterns—reflecting general inflammation, dedifferentiation and tissue damage—and nonresponse to anti‐TNF has been reported. This pattern was similar to that seen in organ transplant rejection.60 Ongoing increased regulatory T‐cell CD39 expression was associated with anti‐TNF response and resulted in decreased suppression of inflammation in a consecutive IBD mouse model.61 Another approach to predict response is by direct visualisation of a target molecule. A high number of membrane bound TNF‐expressing mucosal cells, as identified by in vivo molecular imaging techniques, was associated with increased probability of anti‐TNF response (92% vs 15%; accuracy 88%, positive predictive value (PPV) 85%, negative predictive value (NPV) 92%).62 Furthermore, clinical response to anti‐TNF in UC was associated with low baseline mucosal mRNA expression of IL‐6, IL‐1β, IL‐17A and IFN‐γ.63 Finally, mucosal oncostatin M and its receptor expression were recently shown to be associated with disease activity and pre‐treatment expression was associated with subsequent endoscopic anti‐TNF nonresponse in both CD and UC.64 3.1.3 Proteins An overview of predictive protein markers is presented in Tables 3 and 4. Reference (year) Response definition Biomarker Association Predictive value Anti‐TNF Algaba (2014) Clin Resp: ≥5 pt reduction in modified Truelove‐Witts index or ≥70 pt reduction in CDAI VEGF; Ang1 Clin Resp NR Low VEGF levels at BL (R vs NR: 453.5 ± 250.7 vs 667.5 ± 153.9 pg/mL) and before 2nd and 3rd dose (P < 0.05). High Ang1 at BL (AUC 0.7) High VEGF at BL (AUC = 0.8. Cutoff: 658.36 pg/mL; Sens 85.7%; Spec 78.6% Baird (2016) Clin Resp: CDAI<150; CAI ≤4, Biochem Resp: CRP≤10 mg/L TNF, IL‐1β, IL‐10, IFN‐γ, IL‐6, plasmacytoid DC, IRAK4 NR Clin&Biochem Resp UC: Increased TNF, IL‐1β, IL‐10 and IFN‐γ at BL. Lower TNF & IL‐1β production by TLRs and reduced TLR0‐induced IL‐6 and TLR‐3/4/8/9‐induced IL‐10. Lower pDC frequency's (P < 0.01). Increased IRAK4 upon TLR‐9 stimulation compared to BL (P = 0.03). No IRAK4 reduction upon TLR‐2/4/7 stimulation. No quant pred data Lower total IRAK4 levels upon TLR stimulation, No quant pred data Biancheri (2015) Clin Rem: Not Specified, Endo Resp: complete MH Serum levels of cleaved IgG and antihinge autoantibodies PNR NR have higher serum levels of MMP3‐/MMP12‐cleaved endogenous IgG than both healthy controls and Resp and higher levels of antihinge autoantibodies than Resp (both significant), No quant pred data Billiet (2017) Clin Resp: PGA clear clinical benefit after three infusions. PNR: absence of clinical improvement TNF/CRP ratio at BL higher IL8 at BL IFN‐ɣ & IL‐6 decrease between BL and W2 and W6 resp PNR PNR Clin Resp OR 2.8 [1.4‐5.5], P = 0.003 BL levels IL8 Resp vs PNR (median(IQR)): 11.3 pg/mL (7.3‐18.7) vs 20.3 pg/mL (8.9‐61.1), P = 0.01, No quant pred data P < 0.05, No quant pred data Caneparo (2016) UC: Clin Rem: partial Mayo ≤2 with no subscore >1, Clin Resp: decrease partial Mayo ≥3 and ≥30% from BL with decrease RB ≥1 or absolute RB of 1 or 0. CD: Clin Rem: HBI<4, Clin Resp: decrease HBI≥3 compared to previous IgG anti‐IFI16 antibodies at BL Clin Resp OR =0.143; 95% CI 0.027‐0.761. Feng (2017) Clin Rem: CDAI<150. Clin Resp: CDAI decrease >100. Endo Rem: MH =absence ulcers IL‐9 levels at W14 to predict Resp at W30 Clin Rem Endo Rem NR AUC: 0.803, cut‐off 17,4 (Sens 0,80; spec 0.75; PPV 44.4; NPV 93.7) AUC: 0.752, cut‐off 15.5 (Sens 0.63; spec 0.87; PPV 85.0; NPV 66.7) higher IL‐9 levels at W14 P = 0.003, No quant pred data Ferrante (2007) Clin Resp: absence of diarrhoea and RB. Partial Clin Resp: improvement but with RB. pANCA+/ASCA‐ (n = 87) Clin Resp (suboptimal) OR 0.4 [0.16‐1.00] P = 0.049 Kevans (2015) Clin Resp at 1 year: no symptoms, CS, salicylate therapy and no colectomy. Anti‐OmpC positivity NR OR 0.14 [0.03‐0.6], P = 0,04; in multivariable model: OR 0.14 [0.03‐0.6] P = 0.006 Magnusson (2013) Clin Resp: decrease in Mayo ≥3 compared with BL, prior to fourth treatment dose. CD4+ and CD8+CD25+ T cells (W2) Annexin V + CD4+ and CD8+ T cells (W2) Serum sCD25 (W2) Serum TNFRII (W2) Clin Resp Clin Resp Clin Resp Clin Resp Cut‐off CD4+CD25+: decrease ≥0.65%; Sens 84.6%. Spec 87.5%. Cut‐off CD8+CD25+ T cell: decrease ≥0.14%; Sens 69.2%; Spec 93.8% Cut‐off: decrease ≥98 pg/mL with Sens 66.7%; Spec 77.8% Cut‐off: decrease ≥98 pg/mL with Sens 66.7%; Spec 77.8% Cut‐off: decrease ≥255 pg/ml; Sens 75%; Spec 72.2% Magnusson (2015) Clin Resp: reduction in Mayo score ≥3 compared with BL (evaluation before 4th infusion). Prediction model: T‐cell surface activation receptor CD25 & IL‐5 Clin Resp R2=0.63; prediction accuracy: 85% Nishida (2017) Clin Resp: partial mayo reduction ≥3 with a decrease ≥30% from BL and decrease RB ≥1 or absolute RB 0 or 1. LOR: clinical relapse BL higher Neutrophil‐to‐lymphocyte ratio (NLR) LOR ROC: cut‐off 4.488 Sens 78.6% Spec 78.3% HR: 3.86 P = 0.023 Reich (2016) Clin Rem: modified HBI<5. Clin Resp: reduction HBI ≥3. NR: surgery or escalation before W14 Clin Rem W14: OR 3.5 [1.3‐9.7], W22: OR 4.7 [1.3‐17.2] Santos Antunes (2016) Clin NR based on PGA and/or need for surgery ANA (positive if ≥1/100) Vit D deficiency: 25(OH)D < 20 ng/mL) Clin NR Clin NR OR 20.107 [2101‐192 388] Association between VD deficiency and positive pre‐treatment ANA OR 6.015 [1,054‐51,599]) Sobolewska (2016) DR: Clin Resp for 52 W. (Clin Resp: CDAI decrease≥25% & ≥70 pts.) LOR: 1) CDAI≥175 & increase CDAI≥35% & ≥70 at ≥2 visits (≥3 W), 2) CD surgery,3) switch to episodic treatment. MPV and ΔMPV DR LOR MPV >10.3 fl at W14: Sens 67% and Spec 80%; P = 0.0019. ΔMPV between BL and W14 >0.4 fl: Sens 87%; Spec: 93% Lower ΔMPV between BL & W14 (P = 0.0003), No quant pred data Steenholdt (2016) Clin Resp: CDAI reduction ≥70 pts from BL IL‐6 and sTNF‐R2 NR (PD) Lower levels after 12 W of treatment in primary PK than PD failures (<3.1 vs 4.0, P = 0.02; 3209 vs 4740, P = 0.04), No quant pred data Zator (2014) DR: time to discontinuation of the index anti‐TNF (drug with VD level within 3 M prior to start) Vit D (<30 ng/mL) NR HR 2.13 [1.03‐4.39] P = 0.04 (with supplementation at index measure) HR 3.4 [1.14‐10.19] (no supplementation) Anti‐IL23p19 Sands (2017) Clin Rem: CDAI<150 at W8, Clin Resp: CDAI decrease ≥100 BL IL‐22 Clin Resp ≥15.6 pg/mL, P = 0.04, No quant pred data IBD: inflammatory bowel disease; CD: Crohn's disease; UC: ulcerative colitis; MR‐UC: medical refractory ulcerative colitis; M: months; W: weeks; D: days; FU: follow‐up; CDAI: Crohn's Disease Activity Index; CAI: colitis activity index; HBI: Harvey Bradshaw Index; PCDAI: Paediatric Crohn's Disease Activity Index; PUCAI: Paediatric Ulcerative Colitis Activity Index; PGA: Physician's Global Assessment score; ULN: upper limit of normal; Hb: haemoglobin; CRP: C‐reactive protein; ESR: erythrocyte sedimentation rate; GM‐CSF: granulocyte‐macrophage colony stimulating factor: IFN‐γ: interferon‐gamma; FC: faecal calprotectin; CK: cytokine; PBMC: peripheral blood mononuclear cell; TNF: tumour necrosis factor; IFX: infliximab; ADA: adalimumab: CS: corticosteroids A‐D‐A: anti‐drug antibodies; Clin: clinical; Bio: biochemical: Endo: endoscopic; Rem: remission; Resp: response; NR: nonresponse; PNR: primary nonresponse; DR: durable response; LOR: loss of response; SSFR: sustained steroid free remission; RB: rectal bleeding; BL: Baseline; PK: pharmacokinetic; PD: pharmacodynamics; pts: points; OR: odds ratio; HR: hazard ratio; Sens: sensitivity; Spec: specificity; PPV: positive predictive value; NPV: negative predictive value; SNP: single nucleotide polymorphism; GRS, genetic risk score; LH: likelihood; FCGRT: Fc Fragment Of IgG Receptor And Transporter; MPV: mean platelet volume; IL: interleukin; No quant pred data: No quantitative predictive data. Reference (year) Response definition Biomarker Association Predictive value Peripheral blood Dubinsky (2010) CD: Clin PNR: increase HBI (W10 or 4 W after last dose) or no decrease ≥3 pts. UC: Clin PNR: no decrease ≥2 pts in partial Mayo score and no improvement in the PGA subscore (W10 or 4 W after last dose) Prediction model: UC, pANCA, 3 SNPs and rs2836878 (21q22/BRWD1) Clin PNR Model: R2 = 0.82 ≥ 3/6 risk factors: AUC 0.98; Sens 0.95; Spec 0.88; accuracy 0.92 and PLR 8 Jurgens (2010) Clin Resp: CAI decrease ≥5 pts or CAI<10pts. Clin Rem: CAI ≤4 Seronegativity of pANCA IL23R variants (homozygous carriers) Clin Resp Clin Resp %seropos vs %seroneg: Resp: 15.6% vs 84.4% NR: 37.1% vs 62.9%, P = 0.037, No quant pred data Clin Resp vs NR:74.1% vs 36.6%; P = 0.001, No quant pred data Lacruz (2013) Clin Resp: CDAI >70 decrease, Clin Rem: CDAI<150. Biochem Resp: CRP >25% decrease Serum IL1‐β SNP rs1143634 (IL1‐β) (C allele) Clin Resp NR <0.64 pg/ml: 6/7 Resp vs 3/10 NR; P = 0.036, No quant pred data 50% Resp vs 84% NR, P = 0.027, No quant pred data Taylor (2001) Clin Resp: decrease in CDAI score ≥70 pnts at W4. pANCA & sANCA seropositivity LTA TNFc‐aa13L‐aa26 haplotype 1‐1‐1‐1 NR NR P = 0.003 (W4), No quant pred data P = 0.007, No quant pred data Peripheral blood & Intestinal tissue Dahlen (2015) Clin Resp: Mayo decrease ≥3 pts (W14) High mucosal mRNA expression of IL6, IL‐1β, IL‐17A and IFN‐γ Low mucosal mRNA expression of IL‐6, IL‐1β, IL‐17A, TNF and IFN‐γ NR Clin Resp Not Specified Not Specified Magnusson (2015) Clin Resp: Mayo decrease ≥3 compared with BL. Endo Resp: reduction ≥1 in Mayo (endoscopy) score (W14) Reduced CD14&CD86 expression (tissue) Reduced CCL2 (W14) (tissue) Reduced PBMC expression of CD14 & 86 (blood) Reduced CCL2 (W2 & W14) (blood) Reduced tenascin C (W2 & W14) (tissue) NR Clin&Endo Resp Clin&Endo Resp Clin&Endo Resp Clin&Endo Resp Not Specified Not Specified Not Specified Not Specified Not Specified Peripheral blood & intestinal tissue & microbiota Magnusson (2016) Clin Resp: decrease total Mayo ≥3 (incl. Endo Resp: decrease ≥1 endoscopic mayo) higher DEFA5 and ECP; lower CATH (tissue) higher circulating ECP (blood) higher abundance of F. prausnitzii (microbiota) higher levels of cFOS and EGR‐1 (tissue) higher IL‐6, IL‐12a, IL‐17A, TNF, CD154 (tissue) higher IL‐6, IL‐12a, IL‐17A, TNF expression (tissue) Clin&Endo Resp Clin&Endo Resp Clin&Endo Resp Clin&Endo Resp NR NR P = 0.001, P = 0.02, P = 0.04 resp, No quant pred data Not specified P = 0.01, No quant pred data Scatter plot, Not specified Scatter plot, Not specified median [25%‐75%] Resp vs NR: IL‐6:0.77 [0.24‐1.46] vs 1.67 [1.09‐2.34] P = 0.06 IL12a: 0.53 [0.29‐0.83] vs 0.91 [0.53‐2.81] P = 0.03 IL17a: 0.92 [0.46‐1.57] vs 1.65 [1.42‐2.27] P = 0.03 TNF: 0.26 [0.18‐1.01] vs 0.65 [0.47‐0.81] P = 0.05 No quant pred data CD: Crohn's disease; UC: ulcerative colitis; pts: points; BL: baseline; W: weeks; CDAI: Crohn's Disease Activity Index; CAI: colitis activity index; RB: rectal bleeding; SF: stool frequency; HBI: Harvey Bradshaw Index; PGA: Physician's Global Assessment score; CRP: C‐reactive protein; Clin: clinical; Biochem: biochemical: Endo: endoscopic; Rem: remission; Resp: response; NR: nonresponse; PNR: primary nonresponse; CS: corticosteroids; PBMC: peripheral blood mononuclear cell; Ig: immunoglobulin; TNF: tumour necrosis factor; IFNγ: interferon gamma; mRNA: messenger RNA; pANCA: perinuclear anti‐neutrophil cytoplasmic antibody; IL: interleukin; CCL2: C‐C motf chemokine ligand 2; Sens: sensitivity; Spec: specificity; PLR: positive likelihood ratio; SNP: single nucleotide polymorphism; No quant pred data: No quantitative predictive data. Peripheral Blood Cytokines and inflammatory cells Low circulating IL‐1β concentrations were associated with clinical response to IFX in CD.41 Decreased frequencies of CD25+ T cells, increased annexin V+ CD4+CD8+ T cells and decreased concentrations of serum CD25 and TNFR2 were also associated with initial response.65 In addition, a combination of T‐cell receptor CD25 suppression and a decrease in IL‐5 secretion was predictive of treatment response with 85% accuracy.66 Reduction in tenascin C (an extracellular matrix glycoprotein) predicted downregulation of CCL2 (a proinflammatory chemokine), which was associated with response at week 14.57 Nonresponders to anti‐TNF dose intensification had high IL‐6 and sTNFR2 compared to responders.67 In UC, anti‐TNF nonresponse was associated with high baseline TNF, IL‐1β, IL‐10 and IFN‐γ, and decreased TNF, IL‐1β, IL‐6 and IL‐10 production upon peripheral blood mononuclear cell stimulation in vitro.68 The TNF/CRP ratio was associated with clinical response (OR 2.8 [1.4‐5.5], P = 0.003), which may underscore the contribution of TNF‐independent inflammation. An elevated IL‐8 concentration at baseline was also associated with nonresponse.69 Finally, a decrease in serum IL‐9 between baseline and week 14 could accurately predict both clinical remission and mucosal healing at week 30 (AUC 0.803 and 0.752 respectively).70 Vitamin D Vitamin D deficiency (<30 ng/mL) predicted earlier cessation of anti‐TNF therapy (HR 2.13 [1.03‐4.39], P = 0.04).71 Severe vitamin D deficiency (<20 ng/mL) was linked with ANA seropositivity (OR 6.015 [1.054‐51.599]), which was in turn associated with anti‐TNF failure (adjusted OR 20.107 [2.101‐192.388]).72 Conversely, an association between low baseline vitamin D concentration and improved clinical anti‐TNF response was recently reported (80% vs 23%; OR 3.5 [1.3‐99.7]).73 Microbial biomarkers As previously described, ANA seropositivity has been associated with anti‐TNF failure.72 In addition, a significantly larger proportion of pANCA negative CD patients achieved clinical response as compared to pANCA positive patients.39 In a multivariable analysis, pANCA+/ASCA‐ status was the only independent predictor of failure to achieve early clinical response to infliximab in UC patients (OR 0.4 [0.16‐1.00], P = 0.049).74 Similarly, pANCA seronegativity in UC patients was associated with early clinical anti‐TNF response in UC patients (P = 0.037). However, the likelihood of achieving clinical remission was not significantly increased (P = 0.557).52 Furthermore, anti‐OmpC status was associated with both anti‐TNF response (OR 0.14 [0.03‐0.6], P = 0.04) and therapy discontinuation at 1 year (HR 2.2 [1.1‐4.7], P = 0.03).75 Finally, low baseline IFI16 IgG level was associated with clinical response to infliximab induction treatment (OR 0.143 [0.027‐0.761]).76 Albumin, haemoglobin, CRP and neutrophil‐to‐lymphocyte ratio The value of albumin,61, 77 haemoglobin78, 79 and CRP40, 77-83 for predicting anti‐TNF response has been reported in several studies (Table S7). Evidence from one study suggests that baseline neutrophil‐to‐lymphocyte ratio predicts secondary loss of response to IFX treatment in UC patients (HR 3.86 [1.20‐12.4], P = 0.023; AUC 0.798).84 Faeces An overview of faecal biomarkers as predictors of therapeutic response is presented in Table S3. In acute severe UC, nonresponse to IFX (as defined by the need for colectomy during admission) was associated with higher faecal calprotectin (FC) values at time of hospitalisation (1795 µg/g vs 920 µg/g, AUC 0.77, P = 0.06).85 Conversely, in UC patients treated with ADA, high pre‐treatment FC values were shown to be associated with sustained clinical and endoscopic response at week 52 (585.5 µg/g [300‐689] vs 372 µg/g [232.5‐450.5], P = 0.049).88 A steep decrease in FC between baseline and week 2 could predict endoscopic remission at week 10 (P < 0.01; β = 0.7 [0.27‐1.12]).89 Week two concentrations of FC (AUC 0.82 and 0.61), faecal lactoferrin (AUC 0.84 and 0.80) and the dimeric M2 isoform of pyruvate kinase (AUC 0.88 and 0.78) could accurately predict both clinical and endoscopic response to IFX induction treatment.90 Finally, several authors have reported the association between low post‐induction FC levels and the decrease between baseline and post‐induction FC values and subsequent sustained response.90-93 3.1.4 Microbiota High baseline abundance of Bifidobacterium, Clostridium colinum, Eubacterium rectale, uncultured Clostridiales and Vibrio, together with a lower abundance of Streptococcus mitis, predicted response to anti‐TNF therapy in one study (Table S4).94 Recently, a lower baseline dysbiosis index and higher abundance of Faecalibacterium prausnitzii were associated with subsequent clinical and endoscopic response, however these findings lacked statistical significance.95 3.2 Anti‐integrins (VDZ// etrolizumab (ETZ)) Three studies reported on predictive biomarkers for therapeutic response to anti‐integrins (Tables 5 and S6 and Figure 3).96-101 In ETZ‐treated patients, higher mucosal baseline levels of granzyme A mRNA and integrin αE mRNA were associated with clinical remission and mucosal healing. Patients with high, compared to low, granzyme A mRNA had significantly increased CCL5 and decreased CXCL1—proteins implicated in the recruitment of memory T cells—at baseline.97 Baseline mucosal integrin α4β7 expression has been linked to subsequent VDZ response with contradictory findings. High integrin α4β7 expression on both T‐ and B‐cell subsets was associated with subsequent VDZ response in IBD patients (P = 0.0047).99 Conversely, low numbers of baseline integrin α4β7 expressing T lymphocytes were also associated with clinical response to VDZ in UC (n = 17), however, not in CD.100 At the microbiome level, high baseline α‐diversity, an abundance of butyrate‐producing species (including Roseburia inulinivorans and Burkholderiales) and increased branched chain amino acid synthesis were associated with clinical response to VDZ at week 14. With an AUC value of 0.872, the optimal neural network prediction model (which incorporates both clinical and microbial variables) accurately predicted treatment response.101 Reference (year) Response definition Biomarker Association Predictive value Gene expression and proteins Fuchs (2017) UC: Clin Resp: MCS (mayo clinical subscore) decrease ≥2 at W16 α4β7 expression at baseline α4β1 expression on CD4+ Tcells at W6 predicts W16 Clin Resp Clin Resp lower expression P < 0.01,No quant pred data cut‐off +4,2%, Sens 100%, Spec 100% P < 0,01 Tew (2016) Clin Rem: total Mayo Clinic Score ≤2, no individual subscore >1. Endo Rem: MH (=Endo score≤1) BL higher level of granzyme A mRNA BL higher level of integrin αE mRNA Clin Rem Endo Rem PPV/NPV (all pts) 41%/94%; (aTNF‐naïve) 75%/88% P < 0.05 Not specified Proteins Boden (2018) Endo Resp: improvement from BL moderate/severe to mild inflammation or reduction extent >50%. Clin Resp: decrease HBI or SSCAI ≥3 or ≤3 within 180 days higher pre‐treatment α4β7 expression Endo&Clin Resp multiple subsets of T, B, and NK cells, with terminal effector memory (mean CD4: 41% ± 2.6% vs 25% ± 3.3%, P = .0009, mean CD8: 43% ± 3.5% vs 28% ± 3.2%, P = .0043) and NK cells (mean =34% ± 3.2% vs 17% ± 4.4%, P = .0047) Stallmach (2016) SFCR at W30 and W54. (Clin Rem: CD: HBI ≤4; UC: pMayo≤1 with RB subscore of 0 at W54) Lower FC (W14) Lower CRP (W14) SFCR SFCR median 354vs3000µg/kg P = 0.002,No quant pred data P = 0.003, No quant pred data Intestinal microbiota Ananthakrishnan (2017) CD: Clin Rem: HBI<4, Clin Resp: decrease HBI>3. UC: Clin Rem: SSCAI<2, Clin Resp: decrease SSCAI>3. Manually curated list of 40 microbiome variables Community alpha‐diversity at baseline Community beta‐diversity at baseline Roseburia inulinivorans and Burkholderiales species 13 pathways, including branched chain amino acid biosynthesis pathways Clin Rem Clin Rem Clin Rem Clin Rem Clin Rem AUC 0,872, >80% true pos <25% false neg α diversity sign higher (q < 0.1), β diversity sign lower, higher BL(q = 0,0914, q = 0,0614) significantly enriched, CD: Crohn's disease; UC: ulcerative colitis; W: weeks; BL: baseline; HBI: Harvey Bradshaw Index; Clin: clinical; Endo: endoscopic; Rem: remission; Resp: response; SFCR: steroid free clinical remission; RB: rectal bleeding; MH: mucosal healing; FC: faecal calprotectin; CRP: C‐reactive protein; PPV: positive predictive value; NPV: negative predictive value; sign: significant; No quant pred data: No quantitative predictive data. 3.3 Anti‐IL23 Thus far, no studies have explored molecular markers as predictors of response to ustekinumab. However, a recent phase 2a CD study of a selective anti‐IL23p19 antibody demonstrated that a significantly higher proportion of patients with elevated baseline IL‐22 serum concentration achieved clinical response and remission at week 8 compared to patients with low IL‐22 levels.102 So far, this observation has not been confirmed with other p19 inhibitors. 3.4 Immunomodulators Despite an abundance of observational studies examining the relationship between molecular markers and adverse events related to immunomodulator use in IBD, evidence regarding biomarkers as predictors of treatment efficacy are scarce (Table S5 and Figure 3).15, 19-21, 103-105 Homozygous carriers for IL23R‐G9T were more likely to respond to azathioprine therapy.15 Moreover, the Rac1 SNP (rs34932801) genotype GC was associated with lower rates of clinical response compared to the wild type (46% vs 76%; OR 0.26 [0.08‐0.91], P = 0.036).20 However, these results could not be replicated in a paediatric IBD population.104 A complete deletion of the GSTM1 gene (glutathione S‐transferase) was associated with poor clinical response to thiopurine treatment (OR 9.22 [1.018‐78.62], P = 0.042).105 Furthermore, a BMI <25 kg/m2 (HR 3.95 [1.55‐5.83], P = 0.001) and albumin concentration <4 g/dL (HR 1.98 [0.97‐4], P = 0.06) were associated with thiopurine failure.103 No studies of methotrexate met the pre‐defined inclusion criteria.

Abstract Breast milk is not only a completely adapted nutrition source for the newborn but also an impressive array of immune‐active molecules that afford protection against infections and shape mucosal immune responses. Decisive imprinting events might be modulated during the first months of life with potential health long‐term effects, enhancing the importance of breastfeeding as a major influence on the immune system correct development and modifying disease susceptibility. The aim of this review was to clarify the link between breastfeeding and autoimmune diseases, inquiring the related mechanisms, based on data available in the literature. Being breastfed was associated with a lower incidence of diabetes, celiac disease, multiple sclerosis and asthma, explained by the protection against early infections, anti‐inflammatory properties, antigen‐specific tolerance induction, and regulation of infant's microbiome. The protective role of human milk in idiopathic juvenile arthritis, rheumatoid arthritis, and inflammatory bowel diseases remains controversial. On the other hand, the breastfeeding mother faces a health‐challenging period in life. High levels of prolactin may lead either to the development of autoimmune diseases in susceptible mothers or exacerbations of current immune‐mediated disorders. These features raise the question if mothers with autoimmune diseases, mainly systemic lupus erythematosus, should avoid breastfeeding. 1 INTRODUCTION Breast milk is the result of 200 million years of Darwinian pressure on mammalian lactation as the source of infant's nourishment.1 Bo Vahlquist (1909‐1978), a renowned pediatrician once said “In all mammalian species the reproductive cycle comprises both pregnancy and breast‐feeding: in the absence of latter, none of these species, man included, could have survived”.2 Diet is one of the most important environmental exposures in early life.3 Currently, the World Health Organization recommends that infants should be exclusively breastfed up to 6 months of age and continue with a partial pattern until the infant is at least 2 years old.4 Breastfeeding has multifactorial determinants and needs supportive measures on many levels, which can represent a barrier for women who attempt it5; still, many are unaware of its great importance on growth and development.6 In 1932, Grulee et al7 analyzed 20 000 healthy infants during the first year of life, relating breastfeeding with a lower incidence of morbidity and mortality. Previously, in the nineteenth century, antibodies were the first immunological molecules discovered in breast milk, boosting the possibility that human milk could have an influence on the development of children immune system.8 Indeed, human milk supplies an impressive array of immune‐active molecules, metabolites, oligosaccharides, microbial content, vitamins, and other nutrients that provide protection against infections and modulates mucosal immune responses.4, 9 At this time, more than 80 different autoimmune diseases have been described in which the leading role is characterized by an inappropriate immune reaction.10 The possibility that decisive imprinting events might be modulated during the first months of life, with potential long‐term effects, sparks the importance of breastfeeding in the correct development of the immune system, altering disease susceptibility, mainly in immune‐mediated conditions.11, 12 Breast milk is not only a completely adapted nutrition source but also a personalized medicine for the infants, programming to some degree their future health.13 For breastfeeding mothers, this period may represent a great health challenge. As prolactin plays an important role on immune modulation,10 physiologic hyperprolactinemia related to continuous breast stimulation may be involved either on the onset of autoimmune diseases or on the exacerbation of current ones.14 2 HUMAN MILK AMONG THE STAGES OF LACTATION Milk is a complex fluid, the composition of which gradually changes according to the stage of lactation, infant feeding, health status (mother and newborn), maternal diet, environment, and genetics.15 Colostrum is the first milk produced and is characterized for containing over 250 potentially immunologically active proteins, including enzymes (lactoferrin, lysozyme, etc.), hormones, immunoglobulins, cytokines, signaling molecules, soluble receptors and inflammatory mediators.8, 16 The high content of bioactive proteins evidenced that colostrum's main role is immunological rather than nutritional.17 Antimicrobial factors present in human milk possess resistance to degradation by the newborn's digestive enzymes, protecting the mucosal surfaces and eliminating pathogens without initiating inflammatory reactions, while the immune system matures.18 Furthermore, the abundance and diversity of indigestible oligosaccharides promote the optimal microbiota colonization which will condition future immune homeostasis.19 A few days after delivery, tight junctions of the mammary epithelium start to close, the rate of sodium/potassium declines, and the lactose concentration rises, indicating the secretory activation and production of transitional milk. Two weeks postpartum, the milk is considered fully mature.20 It is characterized by decreasing relative concentrations of the immunologically active molecules, while the volume and nutritional requirements increase to fulfill the growing infant needs.21 3 BREAST MILK AND IMMUNE SYSTEM MODULATION The postnatal period is crucial for the newborn's immune system maturation. This process is characterized by the development of a balance between Th1/Th2 response.22 The thymus is the central immune system organ where bone marrow‐derived precursors undergo differentiation and selection before they are apt for migration to the periphery23; therefore, thymus plays a central role in the establishment of T‐cell‐mediated immunity.23, 24 Hasselbalch et al25 reported that the thymus is considerably larger in breastfed infants when compared to formula‐fed. Subsequently, Jeppese et al26 not only corroborated the previous findings but also discovered a correlation between breastfeeding and CD8+ T cells. The mechanism by which breastfeeding may influence thymus size is unclear, although pathways involving interleukin 7 (IL‐7) are considered to play a critical role in thymopoiesis, and Tgd lymphocytes output from the intestinal mucosa.27, 28 In addition, breast milk is known to modulate infant's intestinal microbiome,29 which is the main source of bacterial metabolites (Figure 1). This has been proved to play a leading role in T‐cell development and differentiation and may be the missing link between the benefits of breastfeeding in the intestinal microbiome and enhanced thymus size in breastfed infants.30, 31 3.1 Immunological components of human milk Human milk supplies an impressive array of immunologically active molecules (Table 1) involved in complex interactions and influencing infant's outcomes.9, 32 This effect is due to the amazing nature of the interaction between mother and baby. Breast milk will not only nourish the newborn, but also provide a complete arsenal of immune components that will contribute as protective factors and as development and maturation factors (Table 1). The protective factors are known to contribute against infections by providing direct immunity (passive), and immune regulatory effects based on anti‐inflammatory capabilities. They ensure the adequate environment in the infant's gastrointestinal tract by preventing microbiome dysbiosis, therefore preserving the immune barrier functions of the epithelium.33, 34 The newborn immune system can be looked at as a very complex machine with amazing capabilities, yet it does not know how to use it. The development and maturation factors will promote, as their name implies, the maturation, differentiation and development of the local immune system, known as the gut‐associated lymphoid tissue (GALT), and the systemic immune system. Furthermore, it will also promote the development of immune tolerance.22, 35 Components Role Function P D/M Proteins Lactoferrin ● Antimicrobial, iron carrier Lactadherin, lysozyme, defensins ● Antimicrobial MicroRNAs ● ● Lymphocyte development, inflammatory mediator Cells Lymphocytes ● ● (T cells > 80%); lymphocyte development, inflammatory mediator Macrophages ● ● Phagocytosis, pathogen defense, T‐cell stimulation Neutrophils ● Unknown, may play a role in maternal protection Breast milk stem cells ● Secretion of growth factors; tissue homeostasis, repair, and/or regeneration Immunoglobulins sIgA/IgA ● ● Antimicrobial, pathogen binding inhibition IgM/IgG ● ● Antimicrobial, antibody‐mediated cytotoxicity Cytokines IL‐1b ● ● Inflammatory mediator, intestinal and immune system trophic factor IL‐2 ● ● Modulate T‐cell development IL‐6 ● ● Intestinal trophic factor, inflammatory mediator, B‐cell activation IL‐8 ● ● Intestinal trophic factor, recruitment of maternal leukocytes IL‐10 ● ● Intestinal trophic factor, anti‐inflammatory cytokine, tolerance promoter IFN‐g ● ● Stimulate Th1 inflammatory response, suppress Th2 allergic response TNF‐a ● ● Proinflammatory Chemokines and other soluble factors CXCR‐1/CXCR‐2 ● ● Cytokine receptor CXCL‐9 (MIP) ● ● Antimicrobial, NK and T‐cell chemoattractant sCD14 ● Inflammatory mediator, promotor of lymphocyte differentiation and activation IP‐10 ● Antimicrobial, NK and T‐cell chemoattractant MCP‐1 ● ● T‐cell chemoattractant TNF‐RI ● Reduce TNF‐a activity, mediate inflammation G‐CSF ● Stimulate neutrophil growth and differentiation, intestinal trophic factor Growth factors TGF‐b, TGF‐a, EGF ● ● Intestinal trophic factor, promote tolerance IGF‐1 and IGF‐2 ● Development and maturation of gastrointestinal cells Oligosaccharides HMO's ● ● Promote colonization of commensal bacteria, antimicrobial effects (pathogen binding, change gut′s pH) Gangliosides ● ● Prebiotic, suppress inflammation Other immunological compounds LCPUFAs ● ● Promote tolerance, Th1 and Th2 response Nucleotides ● Intestinal trophic factor, promote differentiation and activation of lymphocytes and macrophage, NK activity Osteoprotegerin ● ● May regulate Th1/Th2 balance Food antigens ● Promote tolerance EGF, epidermal growth factor; G‐CSF, granulocyte‐macrophage colony‐stimulating factor; HMO, human milk oligosaccharides; IFN‐g, interferon gamma; Ig, immunoglobulin; IGF, insulin‐like growth factor; IL, interleukin; IP, interferon gamma (IFN)‐g‐inducible protein; LCPUFA, long‐chain polyunsaturated fatty acid; MCP, monocyte chemotactic protein 1; MIP, monokine induced by IFN‐g; TGF, transforming growth factor; TNF‐a, tumor necrosis factor‐alpha. Recent studies show that milk composition may contain different compounds according to the newborn's condition, suggesting a permanent dynamic interaction between mother and baby. Riskin et al36 found an increased number of leukocytes in breast milk when the infant has an infection. Additionally, there are significant differences between immunological factors for term and preterm breast milk.17, 37 In the last few years, analytical advancements have allowed a better characterization of the human milk composition, discovering new structures like stem cells, microbiome elements, miRNA, increasing our knowledge every day.38, 39 3.2 Breast milk and microbiome Breastfeeding influences the gut microbiome and metagenome. Recent studies revealed that the main imprinting events on the newborn immune system might be mediated through the microbiome, highlighting the early influence of the infant's diet nature.6, 40 Its capability to regulate hosts responses depends on the bacterial composition of milk. Abnormal colonization has a long‐term effect on the immune system, mainly with respect to metabolic and autoimmune diseases.41 The microbiome of breastfed infants has copious amounts of bifidobacteria and Lactobacillus species, promoted by the rich bioactive factors present in the human milk, like oligosaccharides(HMOs) that function as natural prebiotics.42 Therefore, there is indeed specificity between the infant's microbiome and breastfeeding, resulting in different bacterial‐induced effects on metabolism and immunity. Curiously, this interaction is based on the mother's enteromammary axis, with the release of dendritic cells containing live maternal gut microbiome (which select non‐pathogenic bacteria from the gut lumen and carry them to the lactating mammary gland), T cells expressing gut‐derived β7 integrins, and plasma cells producing specific immunoglobulin into the milk.43, 44 Naturally, the cytokines present in breast milk also depend on the mother's immunological experience and this heritage will be transmitted to the infant's immature immune system, encouraging tolerogenic responses.45 Studies have reported similar microbiome profiles in human milk, independently of newborns time of gestation or delivery mode.46 Bergstom et al47 reported that the major factor affecting infant microbiome up to 12 months of age was the cessation of breastfeeding. 3.3 Breast milk and immune tolerance A healthy mucosal immune system is capable of distinguishing nutritional antigens and non‐harmful microbes from pathogens, promoting adequate immunological reactions by suppression of cellular and humoral immune responses to ingested antigens. This “controlled responses” are called “oral tolerance” and are influenced by the intestinal microbiota.48 Diseases due to a defect in immune tolerance, such as celiac disease or type 1 diabetes are less common in children who were breastfed, suggesting a protective effect.49 Recent data reported that breastfeed‐induced protection may be explained by early antigen‐specific tolerance induction, both by the transfer of antigens through milk and the presence of factors that affect the newborn's immune maturation and responses,50 adjusting for the deficiency of Immunoglobulin (Ig) synthesis during the first year of life. Indeed, studies have exposed the presence of antigens involved in the pathogenesis of multiple immune‐mediated diseases, such as soluble maternal semi‐allogenic HLA molecules,51 maternal cells, gliadin, ovalbumin, intact human insulin, and food allergens, such as peanut and bovine beta‐lactoglobulin.52 Each mother will produce different milk in accordance with her previous antigen exposure,53 as well as immune responses to transferred antigens, permeability of the mammary gland epithelium,54 microbiome‐influencing factors, gut growth factors in milk, and the presence of tolerogenic molecules.11, 55 4 BREASTFEEDING AND AUTOIMMUNE DISEASES 4.1 Breastfeeding and diabetes 4.1.1 Type 1 diabetes Type 1 diabetes (T1D) is known to be an immune‐mediated disease causing destruction of pancreatic B cells eventually leading to complete and lifelong dependence on exogenous insulin.56 Although genetic susceptibility variants play a crucial role in the development of T1D, increased incidence over the past 50 years strongly suggests an important role for non‐genetic factors.57, 58 The hypothesis that breastfeeding could protect against type 1 diabetes was proposed more than 30 years ago by Borch‐Johnsen et al,59 and since then, several mechanisms have been implied. The reason why human milk might protect against T1D remains uncertain, although numerous theories have been considered for the protective effect of breastfeeding, such as lower incidence of potentially diabetogenic infections, later exposure to dietary antigens, healthier gut microbiota, and correct maturation of the newborn's gut.6, 60, 61 Several positive health outcomes have been associated with breastfeeding4, 62 so far. Cardwell et al63 reported that being breastfed for more than 3 months or exclusively breastfed for more than 2 weeks after birth was associated with 15%‐30% lower risk of T1D in childhood. This association was also found in genetically high‐risk birth controls,64 although other studies were not able to find such affiliation.65, 66 Recently, Lund‐Blix et al performed a large cohort study involving 155 392 children to clarify the relation between the duration of full or any breastfeeding and risk of type 1 diabetes. The authors concluded that children who were never breastfed had a twofold increased risk for developing type 1 diabetes. In the same cohort, authors found no association between the age of introduction of specific solid foods (vegetables, fruits, and berries), or breastfeeding while introducing solid foods, and T1D. This result might be due to cultural differences in infant feeding practices worldwide.64, 66, 67 In contrast, the Diabetes and68 Autoimmunity Study in the Young (DAISY) suggested that the introduction of any solid food before 4 months and after 6 months of age predicted the development of T1D. Likewise, Virtanen et al69 concluded from the Type 1 Diabetes Prediction and Prevention Project (DIPP) study a close link between the risk of islet autoimmunity and early introduction of root vegetables. A recent field of investigation relates microbiome and immune‐mediated diseases. Following birth, early gut microbiota undergoes several changes triggered by the environment. As dietary changes strongly shift the microbial community composition early in life, breastfeeding could play a main role preventing dysbiosis. Interestingly, recent data suggest that alterations in the proportions of short‐chain fatty acids (SCFAs), such as butyrate, might be a potential mechanism contributing to the risk of developing T1D.70, 71 Ongoing large prospective birth control studies, such as DAISY, TEDDY (The Environmental Determinants of Diabetes in the Young), DIPP, as well as the first nutritional primary prevention study of type 1 diabetes, Trial to Reduce IDDM in the Genetically at Risk (TRIGR), will help to clarify the influence of nutrition in the pathogenesis of T1D. 4.1.2 Type 2 diabetes Type 2 diabetes (T2D) is an endocrine pathological state affecting young individuals and adults, which has seen a significant increase in the last 30 years, particularly in children worldwide.72 The etiology is highly heterogeneous, including genetic and environmental factors. These environmental factors, such as diet, may be altered to prevent the onset and the development of the disease.72, 73 The absence of breastfeeding or its improper pattern is recognized as a prime contributive factor in the future development of obesity and metabolic syndrome. Indeed, previous studies related proper breastfeeding with a 24% decreased risk of T2D in childhood.74 According to earlier data, a systematic review directly correlated the duration of breastfeeding and the risk of development of childhood T2D and obesity in neonates.75 More recently, Halipchuk et al76 performed a large retrospective case‐control study to identify the factors associated with the development of childhood‐onset T2D and suggested that exclusive breastfeeding decreases the risk of development of childhood‐onset type 2 diabetes. Unless preventive and intervening efforts are successful, the T2D pandemic and its complications will continue to have an intergenerational effect, affecting many more individuals worldwide.77 4.2 Breastfeeding and idiopathic juvenile arthritis Juvenile idiopathic arthritis (JIA) is the most common chronic rheumatic disease of childhood.78 The pathogenesis and etiology of JIA are unclear, although interactions among genetic factors, immune mechanisms, and environmental exposures are thought to contribute in most cases. Studies relating breastfeeding and JIA have been controversial. Mason et al79 suggested that breastfeeding may protect against JIA, although subsequent studies did not support this theory.80, 81 A prospective cohort study reported that a longer duration of breastfeeding (both total and exclusive) may protect against development of JIA,82 finding an increased risk of JIA in children who were breastfed for less than 4 months. Recently, a cohort study associated breastfeeding with an earlier but milder presentation of JIA and with different patterns of arthritis. Authors justified the earlier onset with the differences in ILAR (International League of Associations for Rheumatology) categories observed between the groups, with a higher proportion of PsA/ERA (Psoriatic Arthritis/Enthesitis‐related arthritis) in the never‐breastfed group, which tend to present at a later age.83 The reason why human milk can influence the susceptibility to certain subtypes of JIA is unknown, although studies have related it with differences in microbiome between children with and without enthesitis‐related arthritis.83, 84 Indeed, breastfeeding could be indirectly linked to JIA by the gut microbiome, as altered microbial profiles have been identified among JIA patients.84 However, further studies are needed to clarify the role of breastfeeding in JIA disease risk.85 4.3 Breastfeeding and rheumatoid arthritis Rheumatoid arthritis (RA) is the most common autoimmune inflammatory joint disease worldwide, characterized by a distinctive pattern of bone and joint destruction.86 Several studies have investigated the influence of breastfeeding and the risk of developing RA, but the results were inconsistent.87-89 An investigation performed by Colebatch reported that in HLA‐DR4‐negative children, rheumatoid factor‐positive infants were less likely to have been breastfed for >3 months when compared with rheumatoid factor‐negative children, although this effect was not seen in HLA‐DR4‐positive infants.90 A recent meta‐analysis by Chen et al91 included 1672 patients with RA and aimed to estimate the association between breastfeeding and RA risk, and an inverse association was found in the overall study population, relating a decreased risk of RA no matter if breastfeeding was longer or shorter than 12 months. 4.4 Breastfeeding and multiple sclerosis Multiple sclerosis (MS) is the most common immune‐mediated disease of the central nervous system affecting young adults.92 Evidence suggests that environmental factors act long before MS becomes symptomatic, suggesting the existence of a prodromal phase of this disease. Conradi et al93 recently associated breastfeeding for at least 4 months as a protective factor against the risk of developing MS; however, other studies still debate the impact of breastfeeding duration in determining the risk for MS.94 Applying data from a large multinational case‐control study (EnvIMS), Ragnedda et al95 described an increased risk of MS in males who were breastfed for less than 4 months. The protective role of human milk in the pathogenesis of immune‐mediated MS could be related with its capacity of promoting the immune system development, safeguarding from toxic agents and pathogens, and modulation of immune responses,96, 97 including immunomodulatory effects by interleukin (IL)‐10 production and the anti‐inflammatory properties of transforming growth factor (TGF)‐β.98 4.5 Breastfeeding and celiac disease Celiac disease (CD) is a systemic immune‐mediated disorder caused by the ingestion of gluten‐containing grains in genetically susceptible individuals.99 Genetic background plays a crucial role in the predisposition to celiac disease,100 although better long‐term health was observed in CD patients who were breastfed.101 This fact may be due to the delayed introduction of gluten in daily diet and expanding the latency time between the introduction of gluten and the development of the disease,102 lower incidence of gastrointestinal infections in the newborn who is breastfed,103 and the prevention of gut dysbiosis.104 Meta‐analysis and systemic review, including all studies between 1966 and 2004, found that infants who were breastfed had a 52% risk reduction in developing CD when compared with those who were not breastfed at the time of gluten introduction.105 Hanson et al106 reported that CD was prevented when gluten was introduced in small doses during breastfeeding. More recently, a meta‐analysis proved that the risk of CD was significantly decreased in infants who were breastfed at the time of gluten introduction.105, 107 Actually, being breastfed at the time of gluten introduction was related with a latter onset of CD.100, 105, 107 Interestingly, intestinal dysbiosis may promote an abnormal response to gluten in predisposed individuals,108 a fact that could be due to the role of specific bacteria in the modulation of infant's immune responses.109 However, it is still not clear whether human milk truly provides permanent tolerance acquisition or just a symptomatic reduction and delayed diagnosis of CD.110 Future studies should investigate the effect of different intervention strategies in CD, including nutrition, immunomodulation, and the role of the microbiome.111 4.6 Breastfeeding and inflammatory bowel diseases Inflammatory bowel diseases (IBD) are known as a chronic inflammatory condition of the gastrointestinal tract, manifesting has 2 major disorders: ulcerative colitis and Crohn disease.112 The exact pathogenesis of IBD remains unknown, and part of the underlying mechanism is believed to be a deregulated host immune response to intestinal flora, in genetically susceptible individuals.12, 113 The major risk for developing IBD is having a family history of the disease.113 The greatest incidence of IBD is in early adulthood, bolstering the fact that early exposures might force future susceptibility. Human milk compounds not only protect against infections but also influence the immune tolerance and bacterial colonization of the infant's gut.114 Considerable investigations found that breastfeeding was protective for IBD,115-117 mainly on early onsets.118 However, other studies failed to relate breastfeeding with IBD.119, 120 This result heterogeneity might be explained by differences in definition (exclusive or nonexclusive) and duration of breastfeeding. 4.7 Breastfeeding and asthma The allergic response to common environmental allergens has been regarded as an important mechanism in the development of airway inflammation in patients with asthma.121 Antibodies directed toward β‐adrenergic receptors were found in sera of some patients with asthma supporting an autoimmune role in the impairment of β‐adrenergic responsiveness.122 Until now, many studies suggested that prolonged and exclusive breastfeeding reduces the risk of wheezing and asthma in infancy and early childhood. Human milk contains allergens, allergen‐specific immunoglobulins, immune complexes, and immunosuppressive cytokines that are transferred through the milk and might modify the immune response to allergens and decrease susceptibility to immune‐mediated disorders.123 A questionnaire‐based cross‐sectional study performed in 2016 by Huang et al concluded that exclusive breastfeeding for 3‐6 months and either partial or exclusive breastfeeding for more than 6 months were associated with significantly reduced risk of asthma. The infant gut microbiome is emerging as an important hypothesis; accordingly, Johnson et al124 recently related the abnormal gut microbiota with the risk of developing asthma and allergic diseases, suggesting the protective role of human milk in maintaining a healthy gut microbial environment. 4.8 Breastfeeding and other autoimmune diseases In 1990, Fort attempted to assess the prevalence of breastfeeding in patients with autoimmune thyroid disease (ATD), which may have a protective role in the development of this disease. Results were unable to document any relationship between the history and duration of breastfeeding and development of ATD; however, a higher prevalence of feedings with soy‐containing formulas in early infancy was found.125 Recently, Simard et al126 evaluated the link between breastfeeding and systemic lupus erythematosus (SLE) among 2 prospective cohort studies, the Nurses’ Health Study and the Nurses’ Health Study II, but found no statistically significant association between duration and SLE incidence. Further research in this field may lead to novel links and strategies of breastfeeding and immune disorders. To summarize and for better comprehension, the relation between breastfeeding and autoimmune diseases is illustrated in Figure 2. 5 BREASTFEEDING MOTHER, PROLACTIN, AND AUTOIMMUNE DISEASES A high prevalence of autoimmune diseases among females suggests a crucial role for gender and hormonal involvement in the pathogenesis of these diseases. The association between prolactin and the immune system has been extensively investigated in the last 2 decades. Prolactin is a polypeptide hormone and member of the immunoneuroendocrine network. It is synthesized not only in the hypophysis but also in other several sites such as neurons, mammary epithelium, skin, prostate, immune organs, and cells. Prolactin secreted by the hypophysis acts as a hormone while the one produced in other organs functions as a cytokine.127 Even though the main physiological role of prolactin is to regulate the mammary gland and ovaries, sustain lactation, and control maternal behavior, currently the evidence points to a correlation between prolactin and autoimmune diseases (Table 2) specially, systemic lupus erythematosus (SLE).128 Organ‐specific diseases Systemic diseases Diabetes mellitus type 1 Celiac disease Addison′s disease Grave′s disease Hashimoto′s thyroiditis Multiple sclerosis Uveitis Lymphocytic hypophysitis Rejection of heart transplant Systemic lupus erythematosus Systemic sclerosis Sjögren syndrome Rheumatoid arthritis Reactive arthritis 5.1 Prolactin and immune modulation Genes which encode prolactin are located in the short arm of chromosome 6, near to HLA‐DRB1. Polymorphisms of the human prolactin gene might have implications in the pathogenesis of autoimmune diseases.127, 129 Studies have demonstrated that hyperprolactinemia impacts B‐cell tolerance via several pathways: (1) decreasing apoptosis in transitional B cells, (2) impairing B‐cell clonal deletion, (3) deregulating receptor editing, and finally (4) decreasing activation of anergic B cells.130-132 Besides, prolactin has proved to induce the expression of IL‐2 receptor and production of interferon‐γ and IL‐1.133 It regulates the maturation of dendritic cells and thymocytes, leading to an increased interferon‐α production and an enhancement of pro‐B‐cell generation.134, 135 5.2 Prolactin and systemic lupus erythematosus Systemic lupus erythematosus is more common in women of reproductive age. Hyperprolactinemia has been described in multiple autoimmune diseases and has shown an intimate correlation with SLE.136 Recently, Song et al137 performed a meta‐analysis to evaluate the relationship between circulating prolactin level and SLE. The author succeeded to demonstrate a significantly positive correlation between prolactin levels and disease activity. A large cohort study performed by Orbach et al138 suggests a role for hyperprolactinemia in lupus and its manifestations, including serositis and anemia. In addition, bromocriptine therapy proved to be beneficial in decreasing disease activity and improving activity scores.139, 140 The discontinuation of therapy was associated with lupus flares.141, 142 Taken together, these findings raise the question whether mothers with SLE should breastfeed their infants. Current evidence points specially to the benefits of treatment with bromocriptine in non‐responsive lupus cases or in flare prevention in postpartum.10, 143 6 CONCLUSIONS The benefits of breastfeeding are well recognized worldwide. Current scientific evidence supports the beneficial role of breastfeeding in several immune‐mediated diseases, although a few exact pathways are yet to be elucidated. Recent technological progresses allowed a better characterization of the human milk composition, opening new horizons to our knowledge for a better understanding of its capabilities, and plausible mechanisms which play a protective role in a heterogeneous pool of diseases with health long‐term effects. On the other hand, the benefits of breastfeeding for mothers with autoimmune diseases are debatable. The role of prolactin in the modulation of the immune system is undeniable, and recent data showed a positive correlation between hyperprolactinemia and lupus manifestations, disease activity, and relapses. Current evidence proved that treatment with bromocriptine is beneficial to non‐responsive lupus cases or on flare prevention in the postpartum. 7 KEY NOTES According to early breastfeeding patterns, decisive imprinting events on the immune system might be modulated during the first months of life, with potential long‐term health effects. Human milk benefits may be explained by the protection against early infections, anti‐inflammatory properties, early antigen‐specific tolerance induction, and regulation of infant's microbiome. Breastfeeding was associated with a protective role in T1D and CD. Benefits in T2D were related to the prevention of obesity and metabolic syndrome. In AS, breastfeeding showed benefit by promoting healthy microbiota, which is suspected to be influenced by HLA‐B27. Infants who were breastfed for at least 4 months show a decreased incidence of MS, through the immunomodulatory effects of IL‐10 production and the anti‐inflammatory properties of TGF‐β. Breastfeeding was associated with lower incidences of asthma during childhood, mainly because of a shaped immune response to allergens. Association between breastfeeding and JIA, RA, and IBD was controversial. A positive correlation between hyperprolactinemia and SLE manifestations, disease activity, and relapses was found. Evidence proved the benefits of treatment with bromocriptine on non‐responsive lupus cases and on flare prevention in the postpartum. ACKNOWLEDGMENTS The authors would like to express sincere gratitude to Professor Eitan Israeli for his priceless contribution to this article. CONFLICT OF INTEREST The authors have no potential conflict of interests in authorship or publication. REFERENCES Citing Literature

Abstract Helicobacter pylori (H. pylori) infection is associated with an inflammatory response in the gastric mucosa, leading to chronic gastritis, peptic ulcers, and gastric cancer. Increased T‐cell infiltration is found at sites of H. pylori infection. The CCR6+ subset of CD4+ regulatory T cells (Tregs), a newly characterized subset of Tregs, has been reported to contribute to local immune inhibition. However, whether CCR6+ Tregs are present in H. pylori gastritis, and what their relationship is to disease prognosis, remains to be elucidated. In this study, gastric infiltrating lymphocytes were isolated from endoscopic biopsy specimens of H. pylori gastritis patients and analyzed. We found that in gastric infiltrating lymphocytes, CCR6+ CD4+ CD25high Tregs, which express high levels of CD45RO, are positively associated with more severe inflammation in gastric mucosa during H. pylori infection. Furthermore, the frequency of CCR6+ Tregs in gastric infiltrating lymphocytes, but not CCR6− Tregs, is significantly increased in inflamed gastric tissues, which is inversely correlated with significantly lower expression of IFN‐γ+ CD8+ T cells. We also found that the frequency of CCR6+ Tregs is positively correlated with the frequency of CD4+ IFN‐γ+ T cells. In addition, the frequency of CCR6+ Tregs, but not that of CCR6− Tregs, is significantly correlated with increased inflammation in H. pylori gastritis. This study demonstrates that immunosuppression in H. pylori gastritis might be related to the activity of CCR6+ Tregs, which could influence disease prognosis. 1 INTRODUCTION Approximately 50% of the world's population is infected with Helicobacter pylori (H. pylori), a common bacteria often associated with chronic gastritis, peptic ulcer disease, and gastric cancer.1 The H. pylori infection process is influenced by H. pylori virulence factors, host gastric mucosal factors, and other environmental factors. H. pylori colonization induces systemic and mucosal immune responses. Increased T‐cell infiltration is found at sites of H. pylori infection.2 Recent studies have suggested that the number of T helper type 1 (Th1) cells is selectively increased during infection.3, 4 Th1 cytokines, such as gamma interferon (IFN‐γ) and tumor necrosis factor alpha (TNF‐α), can increase the release of proinflammatory cytokines, augmenting apoptosis induced by H. pylori.5 Many studies have confirmed that the pathogenesis of H. pylori infection is probably largely due to the host immune system, particularly the cellular immune response. However, the exact immunological mechanism of H. pylori infection is still not completely understood. In recent years, regulatory T cells (Tregs) and other related negative regulatory molecules have been a focus of research in to immune mechanisms of disease. Tregs represent approximately 5%‐10% of circulating CD4+ T cells in mice and humans,6 and they play an important role in the maintenance of immunological tolerance to both self and foreign antigens by suppressing aggressive T‐cell responses.7-9 Evidence indicates that removal or reduction of Tregs can enhance the immune response against infectious microbes.10-17 The transcription factor forkhead box protein 3 (Foxp3), expressed in CD4+ CD25high T cells, is the hallmark of regulatory T cells. Recent studies have further proposed that C‐C chemokine receptor type 6 (CCR6) expression defines regulatory effector/memory‐like cells within the CD25+ CD4+ T‐cell subset.18 Moreover, a recent study found that in situ prior proliferation of CD4+ CCR6+ Tregs, which could be facilitated by TGF‐β‐secreting dendritic cells (DCs), is crucial for their enrichment and suppression in tumor immunity.19 Our previous work has shown upregulation of CCL20 and recruitment of CCR6+ gastric infiltrating lymphocytes in H. pylori gastritis.20 We also showed expression of CD25high Tregs and PD‐1 in gastric infiltrating CD4+ T lymphocytes in patients with H. pylori infection.21 A recent study also revealed CCL20/CCR6‐mediated migration of Tregs to the H. pylori‐infected human gastric mucosa.22 With all these clues, we speculated that CCR6+ Tregs are crucial in the pathogenesis of H. pylori infection. In this study, we further investigated the distribution of CCR6+ Tregs and their association with the prognosis of H. pylori gastritis patients, to further understand the underlying disease mechanisms. 2 MATERIALS AND METHODS 2.1 Subjects and patients Biopsy specimens of the gastric antrum were obtained from patients undergoing gastric endoscopy for dyspepsia at theChina Medical University Hospital. Informed consent was obtained from all patients. The study was approved by the Medical Ethics Committee of the China Medical University Medical Center (Taichung, Taiwan; DMR97‐IRB‐263). A total of 3‐4 biopsy specimens were obtained from each patient. Peripheral blood samples from the same patients were also collected and served as the control. To avoid the influence of medical treatment, patients treated with proton pump inhibitors or antibiotics prior to the endoscopic exam were excluded from this study. The presence of H. pylori infection was determined by visualization through histology (including Giemsa staining), and/or by a positive rapid urease test performed on at least one additional biopsy sample. A total of 38 H. pylori gastritis patients were enrolled in this study. The degree of inflammation present in the histological specimens was classified according to the updated Sydney system, as per our previous study, with modifications.20, 21 Absent to mild inflammation in the updated Sydney system was referred to as mild inflammation, whereas moderate to severe inflammation in the updated Sydney system was referred to as marked inflammation in this study. 2.2 Isolation of gastric infiltrating lymphocytes Gastric infiltrating lymphocytes (GILs) were isolated using a modification ofour previously described technique.20, 21 Briefly, biopsy specimens were collected into sterile collection medium (calcium and magnesium‐free Hanks’ balanced salt solution [HBSS] with 5% fetal calf serum and penicillin plus streptomycin). The collected tissues were immediately placed in ice‐cold RPMI 1640 complete medium (GIBCO BRL, Gaithersburg, MD), which contained 10% fetal calf serum supplemented with penicillin (50 IU/mL), streptomycin (50 g/mL), L‐glutamine (2 mmol/L), and sodium pyruvate (1 mmol/L). The tissues were washed, diced into 1 mm3 pieces through a 380 µmol/L tissue sieve, and treated with collagenase (type I; 5 g/mL; Sigma Chemical, St. Louis, MO) and heparin (5 U/mL) in RPMI 1640 complete medium at 37°C for 60 min. The resulting cell suspensions were washed, and the viability of the mononuclear cells was determined by trypan blue exclusion. 2.3 Flow cytometry Flow cytometry was performed using a FACS Calibur (BD Biosciences) with CellQuest Pro software, using directly conjugated monoclonal antibodies against the following markers: anti‐CD4‐phycoerythrin (PE)‐Texas Red (ECD; Beckman Coulter, High Wycombe, UK); anti‐CD25‐PE‐cyanine 7 (Pc7), anti‐CD127‐PE, anti‐CD62L‐PE (eBioscience Systems, San Diego, USA); anti‐CCR6‐Alexa Fluor 647 (A647), CD45RO‐PE, and CTLA‐4‐PE, with corresponding isotype‐matched controls (either BD Biosciences or eBioscience Systems). Conjugation was carried out before cells were fixed in 0.5% formaldehyde. Cells permeabilized with Foxp3 Perm buffer set (BioLegend) were stained with anti‐Foxp3‐Alexa Fluor 488 (A488). To calculate the percentage of Tregs, lymphocytes were gated by plotting forward and side scatter plots and then gating for CD4+ T cells. The gated cells were then analyzed for CD25 and CCR6 expression. CD4+ CD25high CCR6+ and CD4+ CD25high CCR6− Tregs from GILs were purified using FACS sorting. CD8+ T‐cell or CD4+ T‐cell subsets and intracellular IFN‐γ staining assays were analyzed by flow cytometry as previously described.23-25 2.4 Proliferation assays We needed at least 5 × 104 GILs to do the proliferation assay. Because of the scarcity of cells, Tregs isolated from several gastric biopsies from different patients were pooled together to do each proliferation assay. We repeated the proliferation assay in six sets of patients, involving a total of 36 H. pylori‐infected patients. Briefly, 5 × 104 CD4+ CD25− cells were treated with 2 μg/mL anti‐CD3 (eBioscience Systems) and anti‐CD28 (eBioscience Systems), as per our previous method,23 for 12 hours as effector cells, then incubated with or without CCR6+ Tregs at a ratio of 2:1 for 72 hours in complete RPMI 1640 medium (Sigma, St. Louis, MO), supplemented with 5% FCS. [3H] thymidine (0.5 μCi/well) was added for the last 18 hours of culture to test the suppression activity of CD4+ CD25high CCR6+ Tregs as previously described.26 2.5 Statistical analysis All data were analyzed using SPSS 15.0 for Windows27 and PRISM 4.0 (GraphPad Software Inc, San Diego, CA, USA). Data are shown with standard deviation. For comparison among multiple groups, data were analyzed by one‐way analysis of variance (ANOVA). For comparison between paired GIL and PBL data from the same patient, a non‐parametric Mann‐Whitney U test was used. A Spearman correlation test was used for correlation analyses. P < 0.05 was considered statistically significant. 3 RESULTS 3.1 CCR6+ CD4+ CD25high Tregs were upregulated at inflammation sites To determine whether CCR6+ Tregs are present at the inflammation site in patients with H. pylori gastritis, peripheral blood lymphocytes (PBLs) and gastric infiltrating lymphocytes (GILs) were isolated as per our previous studies.20, 21 We first analyzed the frequency of CD4+ CD25high Tregs in PBLs and GILs in H. pylori gastritis patients, as shown in our previous article.21 In addition, the frequency of CCR6+ Treg subpopulations was analyzed. We found that the frequency of CCR6+ Tregs was higher in GILs (6.9%) than in PBLs (1.4%). Compared with CCR6− Tregs, the frequency of CCR6+ Tregs in GILs was also significantly higher (6.9% vs 2.5%; Figure 1). These data indicate that CCR6+ Tregs are not only present, but also upregulated, at the site of inflammation in patients with clinical H. pylori gastritis. 3.2 CCR6+ Tregs exhibit a memory‐like phenotype and effectively inhibit the proliferation of CD4+ CD25− T cells Next, we investigated the surface molecule expression in CCR6+ CD4+ CD25high Tregs, and their suppression ability. As shown in Figure 2A, like their CCR6− equivalents, CCR6+ Tregs express high levels of Foxp3, CTLA‐4 and low levels of CD127, representative of the Treg phenotype. In addition, CCR6+ Tregs express higher levels of CD45RO and lower levels of CD62L than CCR6− Tregs (Figure 2A). CCR6+ Tregs can also inhibit the proliferation of CD4+ CD25− T cells more efficiently than CCR6− Tregs (Figure 2B). In addition, when CCR6+ Tregs were separated from responder T cells by a trans‐well membrane,20 the inhibitory effect of CCR6+ Tregs on CD4+ CD25− T cells was abolished, indicating that the inhibitory effect of CCR6+ Tregs is contact‐dependent (Figure 2C). Our studies demonstrate that CCR6+ Tregs exhibit a memory‐like phenotype and stronger inhibitory capacity than CCR6− Tregs. 3.3 The prevalence of CCR6+ Tregs is related to the degree of inflammation in patients with H. pylori gastritis To further investigate the role of CCR6+ Tregs in H. pylori gastritis, patients were divided into two groups according to the updated Sydney score with modification: mild inflammation and marked inflammation. The frequency of CD4+ CD25high Tregs was not significantly different in the PBLs from the two groups. For GILs, however, the frequency of CD4+ CD25high Tregs was significantly higher in the marked inflammation group than that of the mild group (Figure 3). There was no difference between the Treg frequency in PBLs in patients and healthy donors (P > 0.05; Figure 3). With regard to the CCR6+ subset of GILs Tregs, the frequency of CCR6+ Tregs was significantly higher in the marked inflammation group than in the mild inflammation group (Figure 4A). The frequency of CCR6− subsets of T GILs Tregs was not significantly different between the two groups (Figure 4B,C). Our results imply that the frequency of the CCR6+ subset of GILs Tregs is positively correlated with the degree of inflammation. 3.4 CCR6+ Tregs were inversely correlated with the percentage of IFN‐γ+ CD8+ T cells, and positively correlated with the percentage of CD4+ IFN‐γ+ T cells We then analyzed the frequency ofIFN‐γ‐secreting T cells. The frequency of CD8+ and CD4+ GILs was similar in both marked and mild inflammation groups (Figure 5A,B). But as shown in Figure 5C, the frequency of the IFN‐γ+ CD8+ T‐cell subset of GILs is significantly lower in the marked group than in the mild group, indicating a compromised anti‐inflammatory response. It has been shown that gastric Tregs are able to inhibit H. pylori‐induced T‐cell proliferation and CD8+ T‐cell IFN‐γ production.28 In our study, we found that for GILs, the frequency of both CCR6+ and CCR6− Tregs is not related to the frequency of CD8+ T cells (Figure 5D,E). However, the frequency of both CCR6+ and CCR6− Tregs is inversely correlated with the frequency of CD8+ IFN‐γ+ T cells (Figure 5F,G). The frequency of CCR6+ Tregs is positively correlated with the frequency of CD4+ IFN‐γ+ T cells (Figure 6A), while the frequency of CCR6− Tregs is not (Figure 6B). 4 DISCUSSION Previous studies, including our previous work, have found enrichment of CCR6+ T cells in inflammatory tissue under different clinical conditions.20, 29 This study further demonstrated that the frequency of the CCR6+ subset of GIL Tregs was positively correlated with the degree of inflammation in H. pylori gastritis. This observation was consistent with a previous report that CD4+ CD25high Tregs expressed high levels of CCR6 in breast cancer patients.30 Our current study also demonstrated that CCR6+ Tregs can inhibit the proliferation of CD4+ CD25− T cells more prominently than their CCR6− counterpart in vitro, through a contact‐dependent manner, which might be correlated with the localized immunosuppression associated with the pathogenesis of H. pylori gastritis. Moreover, in GILs, the frequency of CD4+ CD25high CCR6+ Tregs is inversely correlated with the frequency of CD8+ IFN‐γ+ cells, which could be correlated with the suppression of the host defense mechanism in H. pylori infection. Previous studies suggest that the inhibition of IFN‐γ expression in CD8+ T cells by Tregs is moderated by direct T‐cell‐to‐T‐cell contact, which was replicated in our current study. CD13731, 32 and GITR expression might be associated with this process.31-33 However, the exact mechanism by which CCR6+ Tregs inhibit the IFN‐γ‐secreting CD8+ T cells still requires further investigation. We also found that the frequency of CCR6+ Tregs is positively correlated with the frequency of CD4+ IFN‐γ+ T cells. One possible explanation for this observation is that CD4+ IFN‐γ+ T cells might be more associated with tissue destruction, while CD8+ IFN‐γ+ T cells might be associated more with host defense in H. pylori infection. Further studies are needed to confirm our speculation. In conclusion, our research discovered that the expression of CCR6+ CD4+ CD25high Tregs is positively correlated with the degree of inflammation in patients with H. pylori gastritis. These CCR6+ CD4+ CD25high Tregs can actively suppress the proliferation of CD4+ CD25− T cells. Moreover, their frequency is inversely correlated with the frequency of CD8+ IFNγ+ T cells. These findings suggest an immunosuppressive role for CCR6+ Tregs in the pathogenesis of H. pylori gastritis. Our study helps elucidate the disease mechanism of H. pylori gastritis and may have clinical implications. However, further studies are still needed to help better understand the underlying disease mechanisms. ACKNOWLEDGEMENTS This work was financially supported by the Chinese Medicine Research Center, China Medical University, from The Featured Areas Research Center Program, within the framework of the Higher Education Sprout Project, by the Ministry of Education (MOE) in Taiwan (CMU107‐CR‐02; CMRC‐CHM‐5) and the National Science Council, Taiwan (NSC 98‐2320‐B‐039 ‐012 ‐MY3). We would like to thank Uni‐edit (www.uni-edit.net) for editing and proofreading this manuscript. DISCLOSURES OF INTERESTS No potential conflicts of interest are disclosed. REFERENCES Notes : Ying‐Ming Chiu contributed equally to correspondence.