Circulating and Tissue-Resident CD4+ T Cells With Reactivity to Intestinal Microbiota Are Abundant in Healthy Individuals and Function Is Altered During Inflammation

Background & Aims Interactions between commensal microbes and the immune system are tightly regulated and maintain intestinal homeostasis, but little is known about these interactions in humans. We investigated responses of human CD4+ T cells to the intestinal microbiota. We measured the abundance of T cells in circulation and intestinal tissues that respond to intestinal microbes and determined their clonal diversity. We also assessed their functional phenotypes and effects on intestinal resident cell populations, and studied alterations in microbe-reactive T cells in patients with chronic intestinal inflammation. Methods We collected samples of peripheral blood mononuclear cells and intestinal tissues from healthy individuals (controls, n = 13−30) and patients with inflammatory bowel diseases (n = 119; 59 with ulcerative colitis and 60 with Crohn’s disease). We used 2 independent assays (CD154 detection and carboxy-fluorescein succinimidyl ester dilution assays) and 9 intestinal bacterial species (Escherichia coli, Lactobacillus acidophilus, Bifidobacterium animalis subsp lactis, Faecalibacterium prausnitzii, Bacteroides vulgatus, Roseburia intestinalis, Ruminococcus obeum, Salmonella typhimurium, and Clostridium difficile) to quantify, expand, and characterize microbe-reactive CD4+ T cells. We sequenced T-cell receptor Vβ genes in expanded microbe-reactive T-cell lines to determine their clonal diversity. We examined the effects of microbe-reactive CD4+ T cells on intestinal stromal and epithelial cell lines. Cytokines, chemokines, and gene expression patterns were measured by flow cytometry and quantitative polymerase chain reaction. Results Circulating and gut-resident CD4+ T cells from controls responded to bacteria at frequencies of 40−4000 per million for each bacterial species tested. Microbiota-reactive CD4+ T cells were mainly of a memory phenotype, present in peripheral blood mononuclear cells and intestinal tissue, and had a diverse T-cell receptor Vβ repertoire. These cells were functionally heterogeneous, produced barrier-protective cytokines, and stimulated intestinal stromal and epithelial cells via interleukin 17A, interferon gamma, and tumor necrosis factor. In patients with inflammatory bowel diseases, microbiota-reactive CD4+ T cells were reduced in the blood compared with intestine; T-cell responses that we detected had an increased frequency of interleukin 17A production compared with responses of T cells from blood or intestinal tissues of controls. Conclusions In an analysis of peripheral blood mononuclear cells and intestinal tissues from patients with inflammatory bowel diseases vs controls, we found that reactivity to intestinal bacteria is a normal property of the human CD4+ T-cell repertoire, and does not necessarily indicate disrupted interactions between immune cells and the commensal microbiota. T-cell responses to commensals might support intestinal homeostasis, by producing barrier-protective cytokines and providing a large pool of T cells that react to pathogens.

See Covering the Cover synopsis on page 1175.

BACKGROUND & AIMS:
Interactions between commensal microbes and the immune system are tightly regulated and maintain intestinal homeostasis, but little is known about these interactions in humans. We investigated responses of human CD4 þ T cells to the intestinal microbiota. We measured the abundance of T cells in circulation and intestinal tissues that respond to intestinal microbes and determined their clonal diversity. We also assessed their functional phenotypes and effects on intestinal resident cell populations, and studied alterations in microbe-reactive T cells in patients with chronic intestinal inflammation. METHODS: We collected samples of peripheral blood mononuclear cells and intestinal tissues from healthy individuals (controls, n ¼ 13À30) and patients with inflammatory bowel diseases (n ¼ 119; 59 with ulcerative colitis and 60 with Crohn's disease). We used 2 independent assays (CD154 detection and carboxy-fluorescein succinimidyl ester dilution assays) and 9 intestinal bacterial species (Escherichia coli, Lactobacillus acidophilus, Bifidobacterium animalis subsp lactis, Faecalibacterium prausnitzii, Bacteroides vulgatus, Roseburia intestinalis, Ruminococcus obeum, Salmonella typhimurium, and Clostridium difficile) to quantify, expand, and characterize microbe-reactive CD4 þ T cells. We sequenced T-cell receptor Vb genes in expanded microbe-reactive T-cell lines to determine their clonal diversity. We examined the effects of microbe-reactive CD4 þ T cells on intestinal stromal and epithelial cell lines. Cytokines, chemokines, and gene expression patterns were measured by flow cytometry and quantitative polymerase chain reaction. RESULTS: Circulating and gut-resident CD4 þ T cells from controls responded to bacteria at frequencies of 40À4000 per million for each bacterial species tested. Microbiota-reactive CD4 þ T cells were mainly of a memory phenotype, present in peripheral blood mononuclear cells and intestinal tissue, and had a diverse T-cell receptor Vb repertoire. These cells were functionally heterogeneous, produced barrier-protective cytokines, and stimulated intestinal stromal and epithelial cells via interleukin 17A, interferon gamma, and tumor necrosis factor. In patients with inflammatory bowel diseases, microbiotareactive CD4 þ T cells were reduced in the blood compared with intestine; T-cell responses that we detected had an increased frequency of interleukin 17A production compared with responses of T cells from blood or intestinal tissues of controls. CONCLUSIONS: In an analysis of peripheral blood mononuclear cells and intestinal tissues from patients with inflammatory bowel diseases vs controls, we found that reactivity to intestinal bacteria is a normal property of the human CD4 þ T-cell repertoire, and does not necessarily indicate disrupted interactions between immune cells and the commensal microbiota. T-cell responses to commensals might support intestinal homeostasis, by producing barrier-protective cytokines and providing a large pool of T cells that react to pathogens.
V ast numbers of microbes populate the gastrointestinal tract and contribute to digestion, epithelial barrier integrity, and development of appropriately educated mucosal immunity. 1 Intestinal immune responses are tightly regulated to allow protective immunity against pathogens, while limiting responses to dietary antigens and innocuous microbes. The "mucosal firewall" prevents systemic dissemination of microbes by confining microbial antigens to the gut-associated lymphoid tissue. 2 In the gutassociated lymphoid tissue, dendritic cells drive regulatory T-cell differentiation in response to dietary antigens and commensal bacteria. 3 Nevertheless, vast numbers of potentially commensal-reactive effector and memory T cells populate intestinal mucosae. 4 Recent evidence suggests that in mice, tolerance to commensal-derived antigens may be lost during pathogen-induced epithelial damage and subsequent transient exposure to commensals. 1,5 In humans, circulating memory T cells recognize peptides derived from gut bacteria and can cross-react to pathogens, which can confer immunologic advantage during subsequent new infections. 6,7 Although this process can be beneficial during homeostasis, deranged responses to commensals may promote inflammatory conditions, such as inflammatory bowel diseases (IBDs).
IBDs (including Crohn's disease and ulcerative colitis) result from a prolonged disturbance of gut homeostasis, the precise etiology of which is uncertain. One hypothesis is that, in genetically susceptible individuals, IBD may be triggered by intestinal dysbiosis that promotes aberrant immune stimulation. 8 Indeed, in mouse models of colitis, intestinal microbiota promote inflammation in part by stimulating microbiota-reactive CD4 þ T cells. 5,9 Whether this drives IBD in humans, however, remains unknown.
Although CD4 þ T-cell responses to intestinal bacteria are known to occur in humans, [10][11][12] several aspects of this topic are largely uncharacterized, including the frequency of human T cells in the gut and periphery that are reactive to phylogenetically distinct intestinal microbes; the T-cell receptor (TCR) diversity and clonotype sharing of these T cells; the functional phenotype of gut microbe-reactive T cells and their impact on tissue-resident cell populations; and how microbe-reactive T cells change during chronic intestinal inflammation. To address this knowledge gap, we extensively characterized CD4 þ T-cell responses to intestinal microbiota in healthy individuals and IBD patients.
Using 2 independent assays, we observed that for almost all enteric bacteria examined, bacteria-reactive CD4 þ T cells were present at a frequency of 40À500 per million CD4 þ T cells in adult peripheral blood. Bacteria-reactive T cells were also prevalent in the gut mucosa, with prominent enrichment for proteobacteria reactivity. Microbiota-responsive T cells showed a diverse TCR Vb repertoire and potently stimulated inflammatory responses by intestinal epithelial and stromal cells. Intriguingly, T cells from IBD patients displayed a normal spectrum of microbial responses, but expressed high amounts of interleukin (IL) 17A, consistent with increased amounts of T-helper (Th) 17-polarizing cytokines in inflamed intestinal tissue. Collectively, these data demonstrate that microbiota-reactive CD4 þ T cells are prevalent and normal constituents of the human immune system that are functionally altered during IBD pathogenesis.

Human Samples and Cell Isolation
Leukoreduction chambers from healthy individuals were obtained from the National Blood Service (Bristol, UK). Peripheral EDTA blood samples were obtained from IBD patients attending the John Radcliffe Hospital Gastroenterology unit or from healthy in-house volunteers. IBD patients (n ¼ 119; ulcerative colitis, n ¼ 59; Crohn's disease, n ¼ 60) diagnosed by endoscopic, histologic, and radiologic criteria were recruited for the study. Healthy volunteers (n ¼ 30) without any known underlying acute or chronic pathologic condition served as control donors. All donors provided informed written consent.

CD154-Based Detection of Antigen-Specific T Cells
CD154 detection was done as described previously. 13,14 Briefly, cells were plated at 5 Â 10 6 /cm 2 for 7À12 hours with heat-inactivated bacteria. Brefeldin A (5 mg/mL; Sigma Aldrich, St Louis, MO) was added at 2 hours. After 8À12 hours, cells were harvested and treated as described in the section on intracellular cytokine, CD154, and transcription factor staining. For magnetic cell separation enrichment of CD4 þ CD154 þ T cells, see Supplementary Experimental Procedures.

BACKGROUND AND CONTEXT
CD4 þ T cell responses to intestinal bacteria are known to occur, however these responses remain poorly characterized in humans.

NEW FINDINGS
Microbiota-reactive CD4 þ T cells are prevalent and normal constituents of the human immune system that are functionally altered during IBD pathogenesis.

LIMITATIONS
The functional relevance of the detected T-cell responses in humans remains to be elucidated.
IMPACT T-cell responses to commensals might support intestinal homeostasis by producing barrier-protective cytokines and providing a large pool of T cells with potential cross-reactivity to pathogens.

November 2017
T-Cell Response to Intestinal Microbiota 1321

Flow Cytometry and Cell Sorting
PBMCs and lamina propria mononuclear cells were stained according to standard protocols. For details, see Supplementary Experimental Procedures.

Intracellular Cytokine, CD154, and Transcription Factor Staining
For intracellular cytokine staining, cells were stained with fixable viability dye eFluor 780 (eBioscience, San Diego, CA) and surface markers, fixed with 2% formaldehyde (Merck, Kenilworth, NJ), and stained for cytokines in magnetic cell separation buffer containing 0.05% saponin (Sigma-Aldrich). Transcription factor expression was analyzed using the FoxP3 staining buffer set (eBioscience) according to manufacturer's instructions.

Results
Healthy Adults Possess Circulating Memory CD4 þ T Cells That Are Reactive to Intestinal Microbiota CD154 (also known as CD40 ligand) is rapidly up-regulated by CD4 þ T cells after antigen stimulation, irrespective of their differentiation phenotype, major histocompatibility complex alleles, or the precise nature of the antigenic epitope. 13,14 We therefore used CD154 to detect naïve and memory CD4 þ T-cell responses among PBMCs after stimulation with heat-inactivated bacteria ( Figure 1A  Bacteroides vulgatus, Roseburia intestinalis, and Ruminococcus obeum (Supplementary Figure  1A and Supplementary Table 1). These bacteria are common in the healthy intestine, but are altered in relative abundance during inflammation. 15,16 Furthermore, we analyzed responses to Salmonella typhimurium and Clostridium difficile due to their association with IBD. 17,18 T-cell responses to these bacteria were compared with those against wellcharacterized barrier surface-related microbes that drive robust Th17 responses (Staphylococcus aureus and Candida albicans) or strong Th1 responses (Mycobacterium tuberculosis). 19,20 The presence of a large pool of M tuberculosisreactive memory Th1 cells in non-exposed individuals has been documented previously. The responses in healthy controls are directed toward non-tuberculous mycobacteria rather than toward M tuberculosis. 19,21,22 The superantigen Staphylococcus enterotoxin B (SEB) was used as a positive stimulation control.
Stimulation with enteric bacteria reproducibly induced detectable numbers of CD4 þ CD154 þ T cells in the peripheral blood. CD4 þ T cells reactive to S aureus, C albicans, and M tuberculosis were generally more abundant ( Figure 1B and Supplementary Figure 1B and C).
Activation markers CD69 and ICOS were up-regulated on activated antigen-reactive CD4 þ CD154 þ T cells (Supplementary Figure 1D). Responses were major histocompatibility complex IIÀdependent (Supplementary Figure 1E), and lipopolysaccharide failed to induce CD154 expression, confirming that CD154 up-regulation was not a consequence of non-specific microbial responses ( Figure 1B and Supplementary Figure 1C). Based on CD154 þ cell frequencies, we calculated that enteric bacteria-reactive CD4 þ T cells were present at precursor frequencies of 40À500 cells per 10 6 circulating CD4 þ T cells for almost all enteric bacteria surveyed ( Figure 1C and Supplementary Figure 1F).
The newborn gut is primarily colonized with maternal vaginal and fecal bacteria after birth. 23 To understand whether T-cell reactivity to microbes develops after birth, we compared CD154 expression in umbilical cord blood with adult blood after enteric bacteria stimulation. As expected, appreciable responses to microbiota were observed only in adult blood. However, CD4 þ CD154 þ T-cell frequencies after SEB stimulation were similar between adult and cord blood ( Figure 1D and Supplementary Figure 1G).
Human memory T cells down-regulate the naïve marker CD45RA and produce cytokines more efficiently than naïve T cells. 24,25 In healthy individuals, the majority of bacteriareactive CD4 þ T cells had a memory phenotype (>80%, on average), indicating that they had been primed in vivo ( Figure 1E and F and Supplementary Figure 2A and B). Figure  1G and Supplementary Figure 2C). Therefore, the circulating pool of memory CD4 þ T cells contains numerous microbiota-reactive cells that arise after birth and produce cytokines, including TNF-a and IL2.

Circulating Microbiota-Reactive CD4 þ T Cells Express Surface Molecules That Permit Mucosal Trafficking
Memory T cells express numerous adhesion molecules and chemokine receptors to access different tissues under steady-state and inflammatory conditions. 4,26,27 For example, a4b7 integrin and chemokine receptor (CCR) 9 regulate T-cell migration to distinct parts of the gut. Blockade of a4b7 integrin has shown clinical efficacy for treating IBD, whereas CCR9 blockade yielded mixed results. 28,29 To identify homing receptors expressed by bacteria-reactive CD4 þ T cells, we enriched CD4 þ CD154 þ T cells using magnetic beads to visualize rare entericbacteria-reactive T cells, and analyzed them by flow cytometry (Figure 2A). Microbiota-reactive T cells had a central memory phenotype, with >60% expressing high levels of CCR7 ( Figure 2B). Furthermore, 5%À10% of CD4 þ CD154 þ T cells expressed the gut-homing surface markers integrin b7 and CCR9 ( Figure 2C and D). Relative to total memory CD4 þ T cells and CD4 þ CD154 À T cells, enteric bacteria-reactive T cells had high expression of the mucosa-homing receptors CCR4 and CCR6 (>60%), low expression of CCR10, and comparable expression of CXCR3 and CCR2 ( Figure 2D and Supplementary Figure 2D). Microbiota-reactive CD4 þ T cells also  expressed high amounts of CD161, a marker enriched on Th17 cells ( Figure 2D and E). The majority of memory CD4 þ CD154 þ T cells co-expressed CCR7, CCR4, CD161, and CCR6 in various combinations, some of which also expressed integrin b7 ( Figure 2E, pie chart). Therefore, circulating microbiota-reactive CD4 þ T cells are equipped with several homing receptors that promote mucosal access.
When comparing gut microbiota-reactive CD4 þ T cells with those reactive to non-enteric organisms (including S aureus, M tuberculosis, and C albicans), enteric bacteriareactive T cells were partially enriched only in CCR4 expression ( Figure 2F). The homing receptor phenotype of enteric bacteria-reactive T cells is consistent with that of T cells reactive to a broad diversity of microbes located at barrier surfaces.

Microbiota-Reactive CD4 þ T Cells Are Enriched in Gut Tissue
The gut harbors >3 Â 10 10 CD4 þ T cells, but their specificity is unknown. 4,5 We therefore estimated the abundance of human microbiota-reactive CD4 þ T cells in the gut by examining non-inflamed colon specimens using the CD154 assay ( Figure 3A). Lamina propria CD4 þ T cells showed a dominant effector memory and central memory phenotype and expressed both tissue-resident and gutrelated markers, with 80% of cells being CD69 þ (Supplementary Figure 3A).
We next stimulated lamina propria mononuclear cells with microbial lysates or SEB. We combined intracellular CD154 detection with TNF-a staining to increase assay sensitivity, as lamina propria CD4 þ T cells expressed low amounts of CD154 without stimulation (Supplementary Figure 3B and C). Compared with peripheral blood frequencies of unrelated donors, there were similar frequencies of S aureus and SEB reactivity, and reduced M tuberculosis-reactivity in the gut. However, gut CD4 þ T cells were enriched in reactivity toward intestinal bacteria and C albicans ( Figure 3B and Supplementary Figure 3D). Bacteria-reactive cells comprised 150À4000 cells per 10 6 gut-resident memory CD4 þ T cells for all enteric bacteria tested. Given that peripheral blood contained 40À500 bacteria-reactive cells per 10 6 memory CD4 þ T cells (for each bacteria tested), this suggests that bacteria-reactive T cells are 3-to 8-fold more frequent in gut tissue as compared with those in circulation. The strong enrichment of S typhimurium and E coli reactivity in the gut was confirmed by assessing CD154 and TNF-a expression in CD4 þ T cells from donor-matched blood and intestinal tissue ( Figure 3C). Because the gut harbors up to 3 Â 10 10 memory T cells (vs 5À10 Â 10 9 in blood), 4 many of which are bacteria-reactive, the absolute number of gut-resident microbiota-reactive CD4 þ T cells is likely to be at least 10 times greater than that in peripheral blood.
Gut-resident bacteria-reactive (CD154 þ TNF-a þ ) T cells produced high amounts of IFN-gamma, IL17A, and IL2, while production of IL22, granulocyte macrophage colonystimulating factor, and IL4 was generally low ( Figure 3D and Supplementary Figure 3E and F). Interestingly, lamina propria T cells showed increased IL17A expression and reduced IFN-gamma production relative to cells with similar reactivity in peripheral blood ( Figure 3D and Supplementary Figure 3E).

Enteric Bacteria-Reactive CD4 þ T Cells Are Clonally Diverse
To assess the clonal diversity of circulating bacteriareactive memory CD4 þ T cells, we expanded CFSE-labeled CD4 þ T cells using whole bacteria and autologous irradiated monocytes as antigen presenting cells (Supplementary Figure 4A). 20 S aureus-, M tuberculosis-, and SEB-reactive T cells served as controls. Antigenreactive T cells proliferated in a major histocompatibility complex IIÀdependent manner and were readily detectible after 3À6 days ( Figure 4A and Supplementary Figure 4BÀD). Proliferating cells expressed several activation markers including ICOS, CD25, and OX40 ( Figure 4A and Supplementary Figure 4B, E, and F). Consistent with the CD154 assay, S aureus, M tuberculosis, and SEB strongly induced T-cell proliferation ( Figure 4A and B and Supplementary Figure 3B and C).
Flow cytometry analysis revealed a diverse TCR Vb repertoire in bacteria-reactive T cells, similar to polyclonal stimulation with phytohemagglutinin but different to stimulation with SEB, which is known to activate a restricted Vb repertoire ( Figure 4C). 30 To directly assess the clonal diversity of bacteria-reactive CD4 þ T cells, we isolated CFSE low bacteria-reactive memory T cells and assessed TCR Vb clonotypes by multiplex polymerase chain reaction and deep sequencing. One hundred and fifty to eight hundred clonotypes were detected for each reactivity (Supplementary Figure 4G). The largest clonal diversity was detected among E coli-and S typhimur-iumÀreactive cells, consistent with frequencies observed in the CD154 assay (see Figure 1B and C and Supplementary Figure 1B and F). While closely related species (eg, E coli vs S typhimurium) had 3%À8% overlap in T-cell clonotypes, little clonotype sharing was observed between T cells reactive to more distantly related bacteria (Supplementary Figure 3H). Indeed, E coli-and B ani-malisÀreactive CD4 þ T-cell lines were strongly restimulated when cultured with autologous monocytes loaded with E coli or B animalis lysates, respectively. In contrast, E coliÀreactive T cells responded weakly to the closely related S typhimurium, while B animalisÀreactive T cells responded weakly to L acidophilus, F prausnitzii, and C difficile (Supplementary Figure 5A). These data confirm the low degree of cross-reactivity predicted from TCR Vb sequencing.

Microbiota-Reactive Memory CD4 þ T Cells Are Functionally Heterogeneous and Produce Barrier-Promoting Cytokines
To functionally characterize circulating microbiotareactive memory cells, we analyzed CFSE low cells using flow cytometry after stimulation with enteric bacteria for 6 days. Enteric microbiota-reactive cells produced Th1-and Th17-related cytokines, including IFN-gamma, IL17A, and IL22, but only low amounts of the Th2 cytokine IL4, comparable to cells reactive toward S aureus or C albicans ( Figure 4D and Supplementary Figure 5B and C). In contrast, memory T cells reactive toward SEB, M tuberculosis, influenza vaccine components, or tetanus toxoid showed a polarized Th1 profile with low expression of IL17A ( Figure 4D and Supplementary Figure 5B and C). Boolean gating revealed a high degree of functional heterogeneity in expanded microbiota-reactive memory T cells, with frequent co-expression of IL17A, IL22, and IFN-gamma ( Figure 4E). Bacteria-reactive cells co-expressed the transcription factors RAR-related orphan receptor gt and T-box expressed in T cells, which are characteristic of Th17 and Th1 cells, respectively ( Figure 4F and G and Supplementary Figure 5D and E). Intriguingly, a subset of CD4 þ T cells reactive to F prausnitzii, L acidophilus, or B animalis produced the immunoregulatory cytokine IL10 in addition to IFN-gamma and IL17A (Supplementary Figure 5F). Compared to T cells that are reactive toward M tuberculosis or vaccine antigens, enteric microbiota-reactive T cells are functionally distinct and produce barrier-promoting and immunoregulatory cytokines.

Microbiota-Reactive Memory T Cells Promote Intestinal Stromal and Epithelial Cell Activation
During periods of epithelial damage and exposure to commensals, activation of microbiota-reactive memory T cells could promote protective immune responses. To assess their tissue-modulating capabilities, cell-free supernatants of microbiota-reactive memory T cells were used to stimulate CCD18Co intestinal myofibroblasts and LIM1863 colonic epithelial cells. CCD18Co and LIM1863 cells were then assessed for expression of various immune response genes that were selected a priori to represent responses to major T-cellÀderived cytokines. Both cell types responded by expressing several cytokine and chemokine genes known to be induced by IL17A (including IL1B, CSF2, IL6, CXCL1, and CXCL8), as well as IFN-gammaÀinducible genes, including CXCL9, CXCL10, and CXCL11 ( Figure 5A and B). 31 Conversely, supernatants from SEB-stimulated memory T cells (which produce little IL17A) mainly induced IFN-gammaÀdependent genes. Thus, stimulation of nonhematopoietic intestinal cells by microbiota-reactive T cells may promote recruitment and activation of myeloid cell populations to facilitate pathogen control and tissue repair.
We next assessed the effects of individual cytokines in E coliÀreactive T-cell supernatants using combinations of neutralizing antibodies. This experiment revealed distinct IFN-gammaÀ and IL17A-/TNF-aÀdependent groups of response genes in both intestinal epithelial cells and fibroblasts. IFN-gamma blockade strongly reduced expression of several chemokine genes, including CXCL9, CXCL10, CXCL11, CCL2, and CCL7 (IFN-gammaÀdependent module; Figure 5C and D). Intriguingly, single blockade of IL17A, IL22, or TNF-a did not affect stromal or epithelial cell activation ( Figure 5C and D). However, combined blockade of IL17A and TNF-a influenced a large number of genes, including CSF2, IL1B, TNF, CXCL1, CXCL8, CXCL5, CXCL6, and CCL20 (IL17A/TNF-aÀdependent module). Triple blockade of IFN-gamma, IL17A, and TNF-a completely inhibited stromal and epithelial cell activation. IL22 blockade did not affect cytokine or chemokine production, but attenuated induction of the antimicrobial peptide REG3G in LIM1863 cells. Given that the products of T-cellÀstimulated stromal and epithelial cells are highly expressed in the inflamed mucosa of IBD patients ( Figures 5E and 7D), this signature might reflect the activation of microbiota-reactive T cells after epithelial disruption, a key feature of IBD.

Microbiota-Reactive CD4 þ T Cells in Inflamed Intestinal Tissue Show a T-Helper 17ÀSkewed Phenotype in Patients With Inflammatory Bowel Disease
IBD is thought to arise in part from aberrant adaptive immune responses to microbiota. 8 Human CD4 þ T cells in IBD have been functionally characterized mainly by polyclonal stimulation. [32][33][34] Therefore, we evaluated microbiota-reactive CD4 þ T-cell responses in IBD patients using the CD154 detection approach. Circulating microbiota-reactive CD4 þ T-cell frequencies were decreased in IBD patients compared with healthy donors,  which might reflect their selective recruitment to the inflamed gut ( Figure 6A and Supplementary Figure 6A). However, intestinal memory CD4 þ T cells from IBD patients did not display reciprocally higher frequencies of microbial specificity ( Figure 6B and Supplementary Figure 6B). We next calculated the frequency of memory CD4 þ T cells in inflamed mucosae using flow cytometry. Memory CD4 þ T cells were present at higher frequencies in inflamed tissue from IBD patients compared with tissue from matched non-lesional sites of IBD patients and healthy controls ( Figure 6C). These findings were confirmed using a previously published bioinformatics approach known as CIBERSORT in an independent cohort 35 (Supplementary Figure 6C). Based on both approaches, memory CD4 þ T cells are typically 2-to 4-fold more frequent in inflamed tissue from IBD patients compared with tissue from healthy controls. Because inflamed tissue contains a higher abundance of memory CD4 þ T cells than healthy mucosa, it can be inferred that gut-resident microbiota-reactive CD4 þ T cells are similarly enriched in patients with active IBD ( Figure 6C, Supplementary Figure 6C).
To evaluate functional alterations in microbiotareactive CD4 þ T cells in IBD, intracellular CD154 detection was combined with cytokine analysis. Compared with healthy controls, circulating microbiota-reactive CD4 þ T cells from IBD patients displayed increased IL17A and IL2 production, but decreased expression of IFN-gamma ( Figure 7A and Supplementary Figure 6D and E). Interestingly, increased IL17A production was observed in all enteric bacteria-reactive responses, but not in S aureus, M tuberculosis, or SEB responses ( Figure 7A and Supplementary Figure 6E and F). These changes were observed in both Crohn's disease and ulcerative colitis and were independent of disease activity or therapy (Supplementary Figure 6G). However, no difference in IL10 production was observed between healthy donors and IBD patients (Supplementary Figure 6H). IFN-gamma and IL17A co-expression is thought to identify pathogenic CD4 þ T cells in mouse colitis models, 36 so we assessed their co-expression in E coliÀreactive memory CD4 þ T cells. Compared with controls, IBD patients displayed significantly increased frequencies of IL-17A þ IFN-gamma À cells and a marginal increase in IL17A þ IFN-gamma þ cells, while the IL17A À IFN-gamma þ fraction was reduced significantly ( Figure 7B). E coliÀreactive CD4 þ T cells from inflamed intestinal tissue showed an increase in IL17A single producers similar to that seen in peripheral blood ( Figure 7C).
Because the Th17-inducing cytokines IL1B, IL6, and IL23A were highly enriched in the inflamed intestinal tissue of IBD patients ( Figure 7D), we reasoned that they might promote Th17 polarization of bacteria-reactive T cells. Indeed, treatment of microbiota-reactive CD4 þ T cells from healthy donors and IBD patients with IL1b, IL6, or IL23 for 1 week during stimulation with E coli, S typhimurium, L acidophilus, or B animalis (CFSE dilution assay) resulted in a 1.5-to 2-fold increase in IL17A production ( Figure 7E and Supplementary Figure 6I).
Taken together, these experiments demonstrate that circulating and gut-resident microbiota-reactive CD4 þ T cells express increased frequencies of IL17A in IBD. Intestinal tissues from patients with active IBD express gene modules driven by Th1/Th17-derived cytokines, suggesting that bacteria-reactive memory cells could contribute to the tissue response.

Discussion
The gastrointestinal tract harbors a large and diverse population of commensal bacteria, and how the immune system interacts with them is subject to intense investigation. Here we used 2 different methodologies to characterize microbiota-reactive CD4 þ T-cell frequencies and phenotypes in the blood and intestinal tissue of healthy individuals and those with IBD. For each bacterial strain tested, the healthy CD4 þ T-cell repertoire contains reactive cells at a frequency of 40À4000 per million, consistent with other antigen-reactive memory T cells. 37 Microbiotareactive CD4 þ T cells were mainly of a memory phenotype, present in both blood and gut tissue, had a diverse TCR Vb repertoire, and showed little clonotype sharing. Notably, microbiota-reactive CD4 þ T cells were functionally heterogeneous in terms of homing receptor expression and = Figure 5. Microbiota-reactive memory T cells promote intestinal stromal and epithelial cell activation. Healthy donor memory CD4 þ T cells from peripheral blood were labeled with CFSE and stimulated with heat-inactivated bacteria in the presence of autologous monocytes. CD4 þ CFSE low ICOS high cells were fluorescence-activated cellÀsorted on day 7 and expanded for 10À14 days with anti-CD3/CD28 beads. Expanded cells were stimulated at equal numbers with phorbol myristate acetate (PMA)/ionomycin for 24 hours to produce conditioned supernatants. (A, B) Cell-free supernatants from different T-cell specificities were used to stimulate CCD18Co intestinal myofibroblasts and LIM1863 colon epithelial cells. Gene expression in stimulated cells was measured by quantitative polymerase chain reaction (qPCR) and normalized to control treatment (media containing PMA/ionomycin alone). Results of independent stimulations were pooled together into the following categories: Proteobacteria-reactive T cells (S typhimuriumÀ and E coliÀreactive); Actinobacteria-reactive T cells (B animalis-reactive); Firmicutes-reactive T cells (F prausnitziiÀ and L acidophilusÀreactive). Data are from 3 independent T-cell donors. (C, D) Supernatants from E coliÀreactive CD4 þ T cells were used to stimulate CCD18Co (C) or LIM1863 (D) cells. Supernatants were pretreated with 1 or more cytokine-neutralizing antibodies as indicated. Gene expression was median-normalized, log 2 transformed, and plotted as a heat map. Data representative of 2À3 independent experiments. (E) qPCR analysis of mucosal biopsies from the Oxford IBD cohort, categorized by endoscopic assessment of disease activity. Demographic and clinical characteristics of IBD patients are summarized in Supplementary Table 5. Statistics: (A, B, E) 1-way analysis of variance with Sidak's multiple comparison test.

November 2017
T-Cell Response to Intestinal Microbiota 1331 effector functions and could stimulate intestinal cells via production of IL17A, IFN-gamma, IL22, and TNF-a. In addition, microbiota-reactive CD4 þ T cells were recruited to sites of inflammation and showed increased IL17A production in patients with IBD.
Characterizing T-cell responses to bacteria is technically challenging due to their complex antigenic makeup. We therefore used the CD154 and CFSE dilution assays, both of which exploit microbial complexity, to provide large numbers of antigens. The combination of CFSE dilution and TCR Vb sequencing allowed us to quantify clonotype heterogeneity and sharing between different bacteria-reactive T cells. Given the phylogenetic similarity of several bacteria used in this study, the paucity of clonotype sharing was surprising. Nevertheless, enteric bacteria-reactive T cells could be cross-reactive to other antigens not assessed in this study, and may have been primed during immune responses to other targets. 38 High interclonal and intraclonal functional heterogeneity in human CD4 þ T-cell responses to microbes and vaccines was observed recently. 19 However, clonotype sharing between different microbial stimuli has not been studied previously and requires further investigation.
Microbiota-reactive CD4 þ T cells showed substantial phenotypic and functional heterogeneity. The majority of circulating enteric bacteria-reactive CD4 þ T cells coexpressed chemokine receptors, including CCR4, CCR6, and CCR7, while a smaller fraction expressed the gut-related homing receptors a4b7 and CCR9. These receptors promote access to secondary lymphoid organs and various mucosal tissues, including the intestine. 4,27 In addition, circulating and gut-resident microbiota-reactive T cells displayed both Th17 and Th1 characteristics and, in some cases, produced IL10. 39 Gut-resident cells showed a clear Th17 bias when compared to circulating populations, which was more pronounced in IBD.
Based on our observations, we can speculate that continuous sampling of luminal antigens by intestinal dendritic cells causes low-level stimulation of gut-resident CD4 þ T cells to produce cytokines that support epithelial integrity, barrier function, and intestinal homeostasis. 40,41 Indeed, cytokine production by commensal-reactive CD4 þ T cells might play a more significant role in supporting gut homeostasis than previously thought (Supplementary Figure 7). However, this homeostatic circuit might be disrupted in IBD due to dysbiotic changes and/or perturbed myeloid cell activity, causing inappropriate T-cell activation and a pathogenic imbalance of cytokine production. 8 While IL17A is frequently cited as a pathogenic cytokine, it is also critical for promoting mucosal barrier function and protection from pathogens. 31 Absence of IL17A was recently shown to increase epithelial injury and compromise barrier function in mouse models of colitis. 42,43 Indeed, IL17A is a critical driver of neutrophil recruitment, and its absence could therefore exacerbate mucosal inflammation by facilitating bacterial invasion and dispersal. 44 Notably, blockade of IL17A in Crohn's disease caused disease exacerbation, despite being well tolerated and therapeutically effective in psoriasis. 45 Thus, IL17A likely plays a key tissue-protective role in humans, suggesting that the increased Th17 polarization of microbiotareactive T cells in IBD patients could reflect an effort to bolster tissue integrity.
Host-microbial homeostasis depends on minimizing contact between micro-organisms and mucosal surfaces via the combined action of epithelial cells, mucus, IgA, antimicrobial peptides, and immune cells. 1,2 Active immune responses to gut flora have been linked to disease. 8 However, this concept should be revisited in light of our current findings and the observation that healthy individuals generate antibody responses to commensals. 46,47 At least 2 plausible mechanisms could explain the genesis of these microbiota-reactive responses. First, mucosal dendritic cells constantly survey the luminal microenvironment and thereafter migrate to secondary lymphoid tissues to initiate B-and T-cell responses. 48,49 Second, during gastrointestinal infections in mice, immune responses against commensals and pathogens are induced in parallel. 50 Continuous luminal sampling of intestinal microbiota and periodic epithelial breaches during gastrointestinal infections might provide a plethora of memory T cells with potential reactivity toward newly encountered pathogens. 6,7 Therefore, contrary to the notion that they promote inflammatory pathology, acquired commensal-reactive Tcell responses may be essential to promote barrier function and IL10-mediated immune regulation-2 cornerstones of intestinal homeostasis.   Table 6. (B) Frequencies (±SEM) of IL17A and IFN-gamma co-expression in CD154 þ TNF-a þ memory CD4 þ T cells after short-term stimulation with heat-inactivated bacteria (n ¼ 23À34 independent donors). (C) Lamina propria mononuclear cells (LPMCs) from inflamed IBD surgical specimens or non-inflamed and tumor-free surgical specimens from colorectal cancer patients were stimulated with heat-inactivated E coli. Boolean gating shows each possible combination of IL17A, IFN-gamma, and IL22 production by CD154 þ TNF-a þ memory CD4 þ T cells (n ¼ 6 and n ¼ 7 independent donors for IBD and controls, respectively). (D) Quantitative polymerase chain reaction analysis of IL1B, IL6, and IL23A in intestinal mucosal specimens categorized by endoscopic assessment of disease activity. Demographic and clinical characteristics of IBD patients are summarized in Supplementary Table 5. (E) CD4 þ CD45RO þ CD45RA À CD25 À CD8 À memory CD4 þ T cells were isolated from healthy donor blood, labeled with CFSE, and stimulated with autologous monocytes pulsed with B animalis in the presence or absence of the indicated cytokines. Data represent mean (±SEM) fold-changes in IL17A or IFN-gamma expression frequencies relative to cells expanded without cytokines. Statistics: (A, B, C) Mann-Whitney test; (D, E) 1-way analysis of variance with Sidak's multiple comparison test.

Human Samples and Cell Isolation
PBMCs were isolated by Ficoll-Paque (GE Healthcare Life Sciences, Little Chalfont, UK) density gradient centrifugation, resuspended in phosphate-buffered saline with 2 mM EDTA and 0.02% bovine serum albumin, and further processed. Gut specimens were obtained from patients with IBD undergoing surgery for severe, chronically active, or complicated disease. Control gut specimens from macroscopically healthy areas were collected from colorectal cancer patients as noninflammatory controls. One intestinal pinch biopsy was obtained from healthy donors (colorectal cancer screening, or other non-IBD related conditions) or IBD patients during routine endoscopy, from lesional, and non-lesional sites, attending the John Radcliffe Hospital (Oxford, UK). Inflammation status of biopsies was binarized into either inflamed or uninflamed categories based on endoscopic assessment. Additional information about the analyzed IBD patients can be found in Supplementary Tables 5À7.
Lamina propria mononuclear cells were isolated as described previously. 1 In brief, mucosa was dissected and washed in 1 mM dithiothreitol at room temperature for 15 minutes to remove mucus. Specimens were washed 3 times with 0.75 mM EDTA at 37 C for 45 minutes to detach epithelial crypts and digested overnight with 0.1 mg/mL collagenase D (Roche, Indianapolis, IN). Cells were centrifuged for 30 minutes in a Percoll gradient and collected at the 40%À60% interface. All solutions were supplemented with antibiotics (penicillin/streptomycin, 40 mg/mL gentamicin, and 0.025 mg/mL amphotericin B).

Flow Cytometry
Cells were stained with the following monoclonal antibodies as described 2  Samples were acquired on FACS LSRFortessa and FACSLSRII (Becton Dickinson); !2 Â 10 5 memory CD4 þ T cells were acquired. Data were analyzed with FlowJo (Tree Star, Ashland, OR) and SPICE. For analysis of cytokine expression by microbiota-specific T cells, a minimum of 20 CD4 þ CD154 þ TNF-a þ T cells was used; donors with lower events were excluded.

CD154 Enrichment of Antigen-Specific T Cells
For magnetic cell separation enrichment of CD4 þ CD154 þ T cells, CD154 Enrichment and Detection Kit was used (Miltenyi Biotec). Briefly, cells were plated at 5 Â 10 6 /cm 2 for 7À12 hours with heat-inactivated bacteria in the presence of anti-CD40 blocking antibody (HB14) and anti-CD28 stimulation antibody (CD28.6). Anti-CD40 blockade prevents down-regulation of CD154, while CD28 co-stimulation optimizes induction of CD154 expression. For major histocompatibility complex II blockade, 10 mg/mL of a pan-HLA class-II blocking antibody (HLA-DR, DP, DQ; [Tü39]) was added 30 minutes before bacterial stimulation. Cord and adult blood analysis was performed identically. Antigenpresenting cell abundance is similar in cord and adult blood. 3,4 Cross-Reactivity Assay CD154 þ CD4 þ T cells were isolated using magnetic cell separating enrichment and fluorescein activated cell sorting from total PBMCs after stimulation with E coli Nissle and B animalis lysates as described. CD154 þ CD4 þ T cells were expanded for 10À14 days with IL2 and anti-CD3/CD28 beads (beads/T cell ratio, 1:4, Dynals). The expanded Tcell lines were washed in IL2 free medium and incubated for 12 hours without IL2 in RPMI-1640 supplemented with 2 mM glutamine, 1% (v/v) non-essential amino acids, 1% (v/v) sodium pyruvate, penicillin (50 U/mL), streptomycin (50 mg/mL; all from Invitrogen), and 5% (v/v) human serum (National Blood Service, Bristol, UK). The expanded T-cell lines were labeled with CFSE, and then were coincubated with autologous monocytes loaded with various bacterial lysates. T cells were co-cultured with the irradiated autologous monocytes at a ratio of 2:1 for 5À7 days. The CFSE dilution was measured at the end of the culture. Autologous CD14 þ monocytes were isolated from PBMC using anti-CD14 microbeads (Miltenyi Biotec) and frozen until the T-cell co-culture. Monocytes were thawed down, irradiated (45 Gy), and then pre-incubated for 3 hours with bacterial lysates before T-cell co-culture.

Preparation of Bacterial Lysates
Different bacteria were cultured in their respective optimal media for 16 hours at 37 C, washed in sterile phosphate-buffered saline and heat-inactivated at 65 C for 1 hour, followed by 3 freezeÀthaw cycles. Extremely oxygenÀsensitive bacteria were provided by Sylvia Duncan, University of Aberdeen, after heat inactivation. Suspensions were centrifuged at maximum speed for 15 minutes and supernatants collected. Protein concentration was quantified using Nanodrop (Thermo Fisher Scientific, Waltham, MA). The following bacterial strains were used:  Table 1. Bacterial lysates were titrated in CFSE dilution assay and an optimal concentration was used (5À15 mg/mL). Tetanus toxoid (Calbiochem, San Diego, CA) and influenza seasonal vaccine (OPTAFLU; Novartis, Basel, Switzerland) were used at 5 mg/mL. Heat-killed M tuberculosis (H37Ra) and C albicans were from InvivoGen (San Diego, CA). Ultrapure lipopolysaccharide-EB from E coli 0111:B4 (InvivoGen) was used as a stimulation control. SEB was used at 1 mEBmL (Sigma).

RNA Extraction, Complementary DNA Synthesis, and Quantitative Polymerase Chain Reaction
Tissue was homogenized using a motor with sterile RNase/DNase-free disposable pestles (both VWR) in RLT buffer (Qiagen, Valencia, CA). Cells were lysed directly in RLT buffer. RNA was isolated using RNeasy Mini kit (Qiagen) or Quick-RNA MiniPrep kit (Zymo Research, Irvine, CA) followed by complementary DNA preparation using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Carlsbad, CA) with random hexamers. Quantitative polymerase chain reaction (PCR) was performed using a CFX96 (Bio-Rad, Hercules, CA) or ViiA7 384-well real-time PCR system (Applied Biosystems) with TaqMan assays (Life Technologies, Carlsbad, CA), and PrecisionPLUS Mastermix (Primerdesign, Southampton, UK). Expression levels were normalized to a housekeeping (hk) gene (RPLP0) and expressed as 2 ˇ -(CTgene-CThk). Heat maps were made using Cluster 3.0 and Java TreeView (Michael Eisen, Stanford University).

CIBERSORT Analysis
To calculate enrichment of cell populations using CIBERSORT, 7 we analyzed the GSE16879 dataset on default settings. For each sample, relative expression of PTPRC (CD45, representing relative leukocyte content) was calculated separately using median-normalized microarray data. This value was then multiplied by CIBERSORT cell type scores (eg, proportion of memory CD4 þ T cells in the total leukocyte fraction) to estimate cell type enrichment levels. Finally, for cell types of interest, fold differences between IBD and control specimens were calculated to estimate relative cell type abundance in active IBD lesions versus healthy tissue.

T-Cell Receptor Vb Sequencing and Analysis
Memory T cells (3.6 Â 10 6 ) were cultured with autologous monocytes pulsed with heat-inactivated bacteria from 3 healthy donors. Expanded CFSE low ICOS þ CD4 þ T cells were sorted into DNase/RNase free water with bovine serum albumin (10 mg/mL) in 96-well plates (100 cells/ well) and stored at À80 C. TCR Vb sequence analysis was obtained by a series of 3 nested PCR reactions, as described previously. 8 Reverse transcription and preamplification were performed with a One-Step RT-PCR kit (Qiagen) using multiplex PCR with multiple Vb region primers and a Cb region primer. After the first reaction, an aliquot was used for the second PCR using a set of multiple internally nested TCR Vb primers and internally nested Cb region primer with HotStarTaq DNA polymerase kit (Qiagen). In the final PCR reaction, an aliquot of the second PCR was used and amplification was performed using barcoding primers containing the common 23-base sequence (incorporated into the second set of Vb primers) and a third internally nested Cb primer and Illumina Pair-End primers. After the third PCR reaction, each PCR product should have a unique set of barcodes incorporated that specifies plate, row, and column and have Illumina Paired-End sequences that enabled sequencing on the Illumina MiSeq platform. The PCR products were combined at equal proportion by volume, run on a 1.2% agarose gel, and a band around 350 to 380 bp was excised and gel purified using a Qiaquick gel extraction kit (Qiagen). Barcoded products were sequenced on the Illumina MiSeq platform.
Sequence reads were trimmed using 'Trim Galore' software (http://www.bioinformatics.babraham.ac.uk/ projects/trim_galore/) with default settings. Reads were de-multiplexed using a custom Python script according to the presence of index sequences. A further custom script then split reads from each well according to the presence of primer sequences used in the PCR amplification. These reads were then analyzed by MICXR, version 1.6 (http:// www.nature.com/nmeth/journal/v12/n5/full/nmeth. 3364.html) using settings appropriate for the anticipated TCR locus of origin (A or B). MIXCR provided CDR3 sequences along with counts of the number of times each sequence was observed. Detected CDR3 sequences were filtered to exclude those with 10 counts per well. There were 48 wells per donor of SEB-stimulated or phytohemagglutinin-stimulated cells while there were 96 wells of bacterially stimulated cells. Furthermore, some wells did not contain any valid TCR sequences after filtering for low read counts, suggesting that amplification