Microbiota-induced TNF-like ligand 1A drives group 3 innate lymphoid cell-mediated barrier protection and intestinal T cell activation during colitis

SUMMARY

Inflammatory bowel disease (IBD) results from a dysregulated interaction between the microbiota and a genetically susceptible host. Genetic studies have linked TNFSF15 polymorphisms and its protein TNF-like ligand 1A (TL1A) with IBD, but the functional role of TL1A is not known. Here, we found that adherent IBD-associated microbiota induced TL1A release from CX3CR1⁺ mononuclear phagocytes (MNPs). Using cell- specific genetic deletion models, we identified an essential role for CX3CR1⁺MNP- derived TL1A in driving group 3 innate lymphoid cell (ILC3) production of interleukin 22 and mucosal healing during acute colitis. In contrast to this protective role in acute colitis, TL1A-dependent expression of co-stimulatory molecule OX40L in MHCII⁺ ILC3s during colitis led to co-stimulation of antigen-specific T cells that was required for chronic T cell colitis. These results identify a role for ILC3s in activating intestinal T cells and reveal a central role for TL1A in promoting ILC3 barrier immunity during colitis.

One Sentence Summary:

Microbial-induced TL1A regulates the innate and adaptive functions of ILC3 in colitis

eTOC/In Brief

Inflammatory bowel disease (IBD) results from a dysregulated interaction between the microbiota and a genetically susceptible host. Castellanos et al. show a protective role for microbial induction of the IBD-linked protein TL1A in promoting ILC3 barrier immunity and uncover a pathogenic role for TL1A-induced expression of OX40L on ILC3s in driving chronic T cell colitis.

INTRODUCTION

Inflammatory bowel disease (IBD) affects over 4 million people worldwide, causing significant morbidity among patients (Kaplan, 2015). IBD pathogenesis is thought to be multifactorial, with both intestinal dysbiosis and genetic components underlying disease pathogenesis. Characteristic alterations in the IBD-associated microbiome exist, with marked contraction of intestinal microbial diversity a hallmark of IBD (Gevers et al., 2014). Genetic studies have identified over 200 genes linked with IBD susceptibility (McGovern et al., 2015), but our functional understanding of how these genes contribute to intestinal homeostasis and IBD pathophysiology is incomplete, limiting the potential for diagnostic and therapeutic intervention.

Genome-wide association studies (GWAS) in IBD have revealed strong associations of genetic variants in TNFSF15 with Crohn’s disease, highlighting a central role for its protein, TNF-like ligand 1A (TL1A), in mucosal immunity (Siakavellas and Bamias, 2015). Variants in TNFSF15 have been linked to the pathogenesis of several autoimmune diseases—including psoriasis, rheumatoid arthritis, and multiple sclerosis—implicating a broader role for TL1A in human inflammatory diseases. In IBD, TNFSF15 variants confer higher risk for more aggressive, penetrating, fibrostenotic, and perianal disease complications (Tung et al., 2014; Yang et al., 2014), but the mechanistic impact of TL1A remains controversial. While early studies showed a pathogenic role for TL1A in driving inflammatory T cell response (Prehn et al., 2004; Zheng et al., 2013), more recent reports in acute colitis models revealed a protective role for TL1A (Jia et al., 2016). Genetic models are needed to understand the cellular mechanisms and potential therapeutic targetability of this IBD-linked pathway in intestinal inflammation.

Group 3 innate lymphoid cells (ILC3s) are an emerging class of innate lymphocytes that play a critical role in regulating mucosal homeostasis. In response to environmental triggers, in particular the cytokines IL-23 and IL-1β, ILC3s produce robust amounts of IL-22 and play an important role in maintaining intestinal barrier integrity and promoting mucosal healing—a major clinical endpoint in IBD (Longman et al., 2014). Co-localization of ILC3s with CX3CR1⁺ mononuclear phagocytes (MNPs) in the lamina propria enables rapid regulation of ILC3 effector cytokines to reinforce barrier immunity and mucosal healing (Satoh-Takayama et al., 2014). Both mouse and human ILC3s express Death Receptor 3 (DR3), the monogamous receptor for TL1A, on its surface, with TL1A stimulation promoting ILC3 proliferation (Ahn et al., 2015) and cytokine production in vitro (Longman et al., 2014); however, the role for TL1A in regulating ILC3 function in intestinal homeostasis, and its potential contribution to IBD pathophysiology, remains unknown.

Here, we generated mouse models with selective deletion of TL1A and its receptor DR3 to evaluate the role of this critical IBD-linked genetic pathway in regulating mucosal immunity. Our data revealed a central role for CX3CR1⁺ MNP-derived TL1A, which is induced by adherent IBD-associated microbiota, in regulating ILC3 production of IL-22 and mucosal healing in acute colitis. In addition to regulating cytokine production, TL1A induced OX40 ligand (OX40L) expression on colonic ILC3 in mouse and human colitis. OX40L expression on MHCII⁺ ILC3s supported antigen-specific T proliferation in vitro and pathogenic T helper 1 (Th1) cell expansion in chronic colitis models. Our findings reveal a central role for IBD-linked TL1A in regulating ILC3 barrier immunity and uncover a mechanistic role for TL1A-induced expression of the costimulatory molecule OX40L on MHCII⁺ ILC3s in regulating colonic T cells.

RESULTS

Mononuclear phagocyte TL1A promotes barrier protection during acute colitis

CX₃CR1⁺ MNPs play a central role in coordinating the mucosal immune response during colitis (Kim et al., 2018; Longman et al., 2014; Panea et al., 2015; Satoh- Takayama et al., 2014). Recent studies have reported increased TL1A protein in the colonic tissue of both mouse and humans during active colitis (Bamias et al., 2013) (Zheng et al., 2013). To evaluate the cellular source of TL1A in the human intestine, we performed TL1A staining of lamina propria mononuclear cells (LPMCs). TL1A expression was significantly higher on CD14⁺HLA-DR^hi intestinal MNPs compared to lymphocytes (Fig. 1A). To evaluate the impact of colitis on MNP expression of TL1A, we performed flow cytometry on lamina propria mononuclear cells from endoscopic biopsies from healthy controls (N=8) or individuals with ileocolonic Crohn’s disease (CD, inactive N=3, active N=8) (Fig. 1B, S1A). MNPs from CD patients with active inflammation showed significantly higher surface expression of TL1A (Fig. 1B). Consistent with these data from human CD, selective depletion of CX3CR1⁺ MNP using mice harboring a diphtheria toxin receptor (DTR) insertion into the Cx3cr1 locus (Cx3cr1^DTR) (Longman et al., 2014) following diphtheria toxin injection significantly reduced TL1A expression in the intestinal lamina propria (Fig. 1C). These data demonstrate that CX3CR1⁺ MNPs are a primary source of intestinal TL1A, which is significantly upregulated during active Crohn’s disease.

Fig. 1. MNP-derived TL1A promotes ILC3 IL-22 and protects against acute colitis.

(A) TL1A surface expression on CD3⁺ T cells, CD19⁺ B cells, and CD14⁺ HLA-DR⁺ MNPs from human intestinal biopsy (N=3). (B) MFI of TL1A surface expression on CD14⁺ HLA-DR⁺ MNPs (Healthy=8, Crohn’s inactive=3, Crohn’s active=8) measured by flow cytometry. (C) Tnfsf15 expression by qPCR from intestinal cells from Cx3cr1^DTR/+ mice or WT controls after DT toxin injection (N=5 mice/group). (D) Weight loss curves of 2% DSS-treated Tnfsf15^flox/floxItgax-cre (called TL1A^ΔCD11c, N=9) or Tnfsf15^flox/flox (called TL1A^flox/flox N=10) littermate mice. Data are pooled from two independent experiments. (E) Weight loss curves of 2% DSS-treated TL1A^floxIflox (N=15), Tnfsf15^flox/flox Cx3cr1- creER (called TL1A^ΔCX3CR1J (N=12), or Cx3cr1^CreER (N= 5) littermate mice. Data are pooled from three independent experiments. (F) Representative colonic histology from TL1A^ΔCX3CR1 or TL1A^floxIflox mice littermate mice following 2% DSS treatment for 9 days (scale bar: 100μm). (G, H) Flow cytometric analysis of intracellular cytokine staining (IFNγ, IL-17) or transcription factor (T-BET, RORγt) in CD4⁺ T cells (G) or intracellular cytokine staining (IFNγ, IL-17, or IL-22) in RORγt⁺ ILC3 from TL1A^floxIflox or TL1A^ΔCD11c mice on day 3 after starting 2% DSS treatment (N=5 miceIgroup). Data are mean ± SEM and analyzed by ANOVA (A, B), two-tailed Student’s t-test (C, G, and H), two-tailed Mann-Whitney test (D), or AnOvA with multiple comparisons test (E); *P<0.05, **P<0.01, ***P<0.001. See also Figure S1.

To evaluate the potential physiological role for MNP-derived TL1A in colitis, we generated mice with a TL1A conditional deletion allele with loxP-flanked Tnfsf15 (Tnfsf15^flox/flox) and crossed to Itgax-cre mice to specifically delete TL1A in cells expressing CD11c (Tnfsf15^flox/floxItgax-cre, subsequently referred to as TL1A^ΔCD11c). TL1A mRNA expression in lamina propria CD11c⁺ cells was used to confirm deletion (Fig. S1B). TL1A^ΔCD11c mice exposed ad libitum to 2% dextran sodium sulfate (DSS) for seven days lost significantly more weight than their littermate controls (Fig. 1D). To specifically test the role for CX3CR1⁺ MNP-derived TL1A, we crossed Tnfsf15^flox/flox mice with mice expressing a tamoxifen-inducible Cre recombinase under control of the Cxc3cr1 promoter (Tnfsf15^flox/floxCx3cr1-creER, subsequently referred to as TL1A^ΔCX3CR1). Transcriptional analysis validated that TL1A^ΔCX3CR1 mice selectively lacked TL1A expression on CD11C⁺MHCII⁺CD11b⁺ MNPs following induction with tamoxifen (Fig. S1C). To examine a role for CX3CR1⁺ MNP-derived TL1A in vivo, TL1A^ΔCX3CR1 and their littermate controls were exposed to 2% DSS for seven days. TL1A^ΔCX3CR1 mice lost significantly more weight than their littermate controls and had increased histopathological damage (Fig. 1E, F).

Previous studies have demonstrated that TL1A can affect both T cell and innate lymphoid cell function (Richard et al., 2015). To evaluate the impact of MNP-specific TL1A deletion in regulating lymphocyte effector function during colitis, flow cytometry was performed on LPMCs. Although no differences were seen in T cell subsets or cytokine production (Fig. 1G, S1D), there was a statistically significant reduction in IL- 22 production by ILC3s in TL1A deficient mice (Fig. 1H,S1E). There were no differences in total ILC3 numbers in the colon (Fig. S1F) or in ILC3 production of IL-17 or IFNγ. These models illustrate a central role for MNP-derived TL1A in regulating the innate response to acute colitis.

Adherent CD-associated microbiota induce CX3CR1⁺ MNP-derived TL1A

CX3CR1⁺ MNPs are uniquely positioned at the intestinal barrier to sample luminal microbes and respond to microbial products through toll like receptor signaling (Diehl et al., 2013). Given the ability of bacterial signals to induce TL1A mRNA in human antigen-presenting cells (Shih et al., 2009), we evaluated the effect of MyD88- deficiency on TL1A expression in gut MNPs. MNPs from mice with MyD88-deletion showed a significant reduction in MNP TL1A expression (Fig. 2A). Although CX3CR1 promotes direct microbial sampling via transepithelial dendrites (Niess et al., 2005), CX3CR1-deletion did not affect TL1A expression. To test the role for microbial regulation of TL1A in vivo, we compared TL1A expression in germ-free mice and specific pathogen free (SPF) mice. Flow cytometric analysis of TL1A expression in CD11c⁺ MHCII⁺ CD11b⁺ MNPs showed that germ-free mice had a significant reduction in intestinal TL1A protein expression compared to SPF mice (Fig. 2B). Previous studies have reported that microbial adherence to the intestinal epithelial surface may play a critical role in regulating immune responses in the lamina propria. Segmented filamentous bacteria (SFB), for example, adheres tightly to the ileal mucosa and induces Th17 and ILC3 effector function (Ivanov et al., 2009; Sano et al., 2015). Given the importance of CX3CR1⁺ MNP in regulating Th17 and ILC3 effector functions (Longman et al., 2014; Panea et al., 2015), we sought to examine the role for SFB in TL1A regulation. Lamina propria MNPs from germ-free and gnotobiotic mice colonized with SFB were analyzed for TL1A expression (Fig. 2C). SFB mono-colonized mice had significantly higher intestinal TL1A mRNA compared to germ-free mice. To further test if this induction of TL1A was selective for adherent bacteria, we used Adherent-invasive E.coli (AIEC) strain 2A and a non-adherent E. coli control strain T75 derived from patients with Crohn’s disease (Viladomiu et al., 2017). At 5 days following colonization, AIEC strain 2A, but not T75, induced TL1A mRNA in the colonic lamina propria (Fig. 2C). To evaluate the role for microbial regulation of TL1A in SPF mice, we assessed TL1A protein expression using antibiotic-treated Cx₃cr1^GFP/WT mice. Flow cytometric analysis of TL1A protein in CX3CR1⁺ MNPs from antibiotic treated mice showed a significant reduction in intestinal TL1A compared to SPF mice (Fig. 2D). The reduction in MNP expression of TL1A protein induced by antibiotics could be restored by colonization with AIEC 2A (Fig. 2E). Previous studies have reported that CX3CR1⁺ MNP increase cytokine production during colitis (Zigmond et al., 2012). To test whether colitis altered CX3CR1⁺ MNP TL1A, mice were exposed to DSS and TL1A mRNA levels were evaluated in sort-purified CX3CR1⁺ MNP. Barrier disruption with DSS induced increased TL1A mRNA expression in SPF mice, which was further enhanced after AIEC 2A colonization (Fig. 2E). Collectively, these data suggest that mucosal-associated microbiota induce MNP TL1A and enhance its production during colitis.

Fig. 2. IBD-associated adherent microbiota induce MNP-derived TL1A and promote protection in acute colitis.

(A) Tnfsf15 expression by qPCR in MHCII⁺CD11c⁺CD11b⁺ MNPs sorted from WT, MyD88-deficient, or CX3CR1-deficient mice (N=3 mice/group). (B) Flow cytometric analysis of TL1A expression in MHCII⁺CD11c⁺CD11b⁺ MNP from germ-free (GF) or specific-pathogen-free (SPF) mice. (C) WT mice were mono-colonized with media control, SFB, non-adherent E. coli T75, Adherent-invasive E. coli 2A for 5 days. MNPs were sorted from the lamina propria and Tnfsf15 expression determined by quantitative PCR (N=3–5 mice/group). One representative of two experiments is shown. (D) Flow cytometric analysis of TL1A expression in MHCII⁺CD11c⁺CD11b⁺CX3CR1⁺ MNP from antibiotic-treated (ABX) or untreated specific-pathogen-free (SPF) Cx3cr1^GFP/WT mice. One of three representative experiments is shown. (E) Tnfsf15 expression by qPCR in MHCII⁺CD11c⁺CD11b⁺ MNPs sorted from WT SPF mice pre-treated with or without broad-spectrum antibiotics (Abx) for 3 weeks prior to subsequent AIEC 2A colonization (2A) or DSS for 5 days (N=3–5 miceIgroup). Data are pooled from two experiments. (F) IL-22 production by ILC3 from WT (N=6), Tnfsf15^flox/flox (called TL1A^floxIflox, N=6), or Tnfsf15^flox/flox Itgax-cre (TL1A^ΔCD11c, N=8) mice pre-treated with or without broad-spectrum antibiotics (Abx) for 3 weeks prior to subsequent AIEC 2A colonization (2A). Data are pooled from two experiments. (G) TL1A^floxIflox and TL1A^ΔCD11c were pre-treated with broad-spectrum antibiotics for three weeks before colonization with 2A E. coli. Mice were then treated with 2% DSS for 7 days. Percent survival is shown. Data are pooled from two experiments. Data in A, C, E, and F are mean ± SEM and were analyzed by ANOVA (A, E, and F), two-tailed Student’s t-test (C), or Log-rank (Mantel-Cox) test (G); *P<0.05, **P<0.01, ***P<0.001.

We have previously shown that AIEC can also induce IL-22 production by ILC3s in gnotobiotic mice, leading to reduced mortality in acute DSS colitis (Viladomiu et al., 2017). To test the role for TL1A in AIEC induction of ILC3 IL-22, mice pre-treated with antibiotics were colonized with AIEC. Similar to our findings in the DSS-induced acute colitis model, TL1A^ΔCD11c mice colonized with AIEC showed reduced production of IL-22 compared to littermate controls (Fig. 2F). To test if the protective effects of AIEC are dependent on MNP-derived TL1A, TL1A^ΔCD11c and littermate control mice were pre-treated with broad-spectrum antibiotics and then colonized with AIEC strain 2A prior to DSS exposure. Colonization with AIEC was sufficient to rescue the more severe weight loss and reduced survival following treatment with antibiotics, and this rescue required MNP-derived TL1A (Fig. 2G). Collectively, these findings reveal a protective link between IBD-associated adherent microbiota and MNP-derived TL1A during acute colitis.

ILC3-specific DR3 deficiency impairs IL-22 production and exacerbates colitis

To evaluate the impact of endogenous TL1A in vivo using experimental colitis models, mice deficient for DR3 (Tnfrsf25^−/−) were exposed ad libitum to 2% DSS. Similar to recent results (Jia et al., 2016), DSS induced more severe acute colitis in mice deficient for DR3, as evidenced by increased weight loss with a protracted recovery (Fig. 3A) and reduced survival (Fig. 3B). Similar to the TL1A deficient mice, analysis of CD4⁺ T cells in the lamina propria did not reveal significant differences in CD4⁺Foxp3⁺ T regulatory (Treg) or IL-17-producing RORγt⁺ T helper 17 (Th17) cells during acute colitis (Fig. S2A, B). Although overexpression of TL1A can alter ILC2 function (Meylan et al., 2014), we observed no effect of DR3 deletion on IL-13 production by ILC2 cells during acute colitis (Fig. S2C). In contrast, Lin^- RORγt⁺ ILC3s from the lamina propria showed a significant decrease in IL-22 (Fig. 3C). To evaluate the functional significance of reduced IL-22 production by ILC3 during colitis, we injected DR3-deficient mice with recombinant IL-22 following exposure to DSS (Fig. 3D). Treatment with rIL-22 rescued the reduced survival of DR3-deficient mice, suggesting a critical role for ILC3-derived IL- 22 in mediating DR3-dependent mucosal healing.

Fig. 3. ILC3 expression of DR3 plays an essential role in regulating IL-22- dependent protection from experimental colitis.

(A, B) Body weights (A) and percent survival curves (B) of Tnfrsf25^+/− (called DR3^+/−, N=34) and Tnfrsf25^−/− (called DR3^−/−, N=25) littermate mice following 7 day treatment with 2% DSS. Data are compiled from three independent experiments. (C) Percentage of colonic Lin^- CD127⁺CD90⁺R0Rγt⁺ ILC3s producing IL-22 from DSS-treated DR3^+/− and DR3^−/− co-housed littermate mice 9 days after starting DSS treatment (N=4–5 mice/group). (D) Percent survival of DR3^+/− (N=5) and DR3^−/− (N=4) littermate mice following 7 day treatment with 2% DSS. i.v. rIL-22 was given on days 4 and 6 following initiation of DSS (N=4). (E) Body weights of 2% DSS-treated Tnfrsf25^flox/flox RAG2^−/− (called DR3^flox/flox, N=13) and Tnfrsf25^flox/flox Rorc-cre RAG2^−/− (called DR3^ΔILC3, N=13) littermate mice compared with DR3^ΔILC3 mice treated with rIL-22 i.v. on days 4 and 6 of DSS treatment (N=4). Data are compiled from two independent experiments. (F) Representative colonic histology from DR3^flox/flox and DR3^ΔILC3 littermate mice following 2% DSS treatment for 7 days (analyzed in E). (G) Representative flow cytometry and pooled analysis of IL-22 production by ILC3s from DR3^flox/flox (N=10) and DR3^ΔILC3 (N=5) littermate mice on day 3 after starting 2% DSS treatment. Data are compiled from two independent experiments. (H) Epithelial cell Reg3g and SAA1 mRNA expression in 2% DSS-treated dr3^flox/flox and DR3^ΔILC3 littermate mice 9 days after starting 2% DSS treatment (N=4 mice/group). Data are compiled from two independent experiments. (I) Percent survival of C. rodentium-infected DR3^flox/flox and DR3 ³ littermate mice (N=5, 4 mice/group, respectively). Data in A, C, E, G, and H are mean ± SEM and analyzed by Mann-Whitney test (A, E); Log-rank (Mantel-Cox) test (B,D,I), or two-tailed Student’s t-test (C, G, H); *P<0.05, **P<0.01, ***P<0.001. See also Figure S2 and S3.

Other cell types, including CD4⁺ T cells, can significantly contribute to IL-22 production (Basu et al., 2012). To specifically test the importance of ILC3-intrinsic DR3 signaling in IL-22 production and intestinal barrier immunity, we crossed mice with a /oxP-flanked Tnfrsf25 (Tnfrsf25^flox/flox) to Rorc-cre mice on a RAG2-deficient background (Tnfrsf25^flox/flox Rorc-cre RAG2^−/− subsequently referred to as DR3^ΔILC3). Transcriptional and flow cytometric analysis revealed that DR3^ΔILC3 mice lacked DR3 expression on ILC3, but not on ILC2, confirming cell type-specific DR3 deletion (Fig. S3A). Absence of DR3 did not change the total number or frequency of ILC3 subsets in the small or large intestines (Fig. S3B).

To evaluate a role for ILC3-intrinsic DR3 in experimental colitis, we exposed DR3^ΔILC3 mice and their littermate controls to ad libitium 2% DSS. Phenocopying DR3- deficiency, DR3^ΔILC3 mice lost significantly more weight (Fig. 3E) and had more severe colitis histopathology (Fig. 3F) than their littermate controls. Flow cytometry analysis revealed lower production of IL-22 by ILC3s from DR3^ΔILC3 mice compared to littermate controls (Fig. 3G). Similar to DR3-deficient animals, treatment with rIL-22 rescued the reduced survival of DR3^ΔILC3 (Fig. 3E). IL-22 is critical for inducing intestinal epithelial cell production of antimicrobial peptides (Kinnebrew et al., 2012) and acute phase reactants (Sano et al., 2015). Reduced expression of both Reg3g and Saa1 in intestinal epithelial cells from DR3^ΔILC3 mice (Fig. 3H) demonstrates the functional impact of reduced IL-22 in ILC3-specific deletion of DR3 during colitis.

To test the role for DR3 in infectious colitis, we used the Citrobacter rodentium infectious colitis model, which is dependent on ILC3s during the acute response (Sonnenberg et al., 2011). Similar to the response in DSS colitis, DR3^ΔILC3 mice lost significantly more weight and had reduced survival compared to littermate controls (Fig. 3I). Collectively, these experiments demonstrate a key role for cell-intrinsic DR3 in regulating ILC3 production of IL-22 and protection from acute experimental colitis.

TL1A synergizes with IL-23 via p38 MAPK to mediate protection in acute colitis

To investigate the mechanism for TL1A regulation of ILC3-derived IL-22, we sorted Lin^- CD127⁺ IL-23R^GFP ILC3s from the small intestinal lamina propria (Fig. S4A) and stimulated ex vivo with either recombinant TL1A, IL-23, or TL1A+IL-23. Although both IL-23 and TL1A induce IL-22 transcription (Fig. 4A), concomitant stimulation with both IL-23 and TL1A induced a robust synergistic IL-22 response (Fig. 4A). We confirmed TL1A synergy with IL-23 was DR3-dependent by sorting ILC3s from DR3- deficient mice as well as heterozygous and WT littermate controls (Fig. 4B).

Fig. 4. TL1A potently and selectively synergizes with IL-23 via MAPK to induce IL- 22 production in ILC3 in vitro and in vivo.

(A, B) Sorted ILC3s from the intestinal lamina propria from WT, Tnfrsf25^+I- (called DR3^+/−) or Tnfrsf25^−/−(called DR3^−/−) mice were stimulated ex vivo with media (−), rIL-23 and/or rTL1A. Quantitative PCR (A) or intracellular cytokine staining for IL-22 (B) was performed after 18h stimulation. One of three independent experiments is shown. (C) IL-22 production by sorted ILC3 subsets was measured by intracellular flow cytometry after 18h stimulation. Brefeldin was added to the cultures 4h before intracellular cytokine staining. One of two independent experiments is shown. (D) Intracellular phosphoflow analysis was performed on sorted ILC3 for phosphorylation of p38-MAPK, ERK and IкBa 30 minute following stimulation. One of two independent experiments is shown. (E) Sorted ILC3s were stimulated for with rIL-23 and rTL1A in the presence of the indicated soluble inhibitors or control carrier DMSO (NF-кB inhibitor, NBD, white; TAK1 inhibitor, 5Z-7-Oxozeaenol, red; p38 MAPK inhibitor, SB203580, blue). IL-22 production was evaluated by ELISA. One of three independent experiments is shown. (F) IL-22 production by sorted ILC3s from WT or Il1r^−/−mice after 18 hour stimulation with media, rIL-1β, rTL1A, or rIL-23 (or in combinations). (G) Survival curves of WT, DR3^−/−, Il23r^GFP/GFPand Tnfrsf25^−/− Il23r^GFP/GFP (called DR3^−/− I/23r^GFP/GFP) mice following 7 day treatment with 2% DSS. Data in are compiled from two independent experiments (with N=26, 18, 7, and 4 mice/group, respectively). Data in A, E, and F are mean ± SEM and analyzed by ANOVA (A, E) or two-tailed Student’s t-test with Bonferroni correction. Data in G are analyzed by Log-rank (Mantel-Cox) test; **P<0.01, ***P<0.001. See also Figure S4.

Distinct ILC3 subsets expressing either the natural cytotoxicity receptor NKp46 (called NCR⁺ ILC3) or chemokine receptor 6 (CCR6) (called lymphoid tissue-inducer-like ILC3) share significant functional overlap in the intestine (Melo-Gonzalez and Hepworth, 2017). Surface DR3 expression was equivalent across all ILC3 subsets (Fig. S4B). Additionally, ex vivo TL1A stimulation of sorted intestinal ILC3 revealed synergistic induction of IL-22 protein in all subsets (Fig. 4C).

Conventional signaling downstream of DR3 can occur through mitogen-activated protein kinase (MAPK) or NF-кB pathways (Bamias et al., 2013). To evaluate the role for these signaling pathways in mediating TL1A and IL-23 synergistic induction of IL-22, we profiled protein phosphorylation by flow cytometry. Our results revealed significant phosphorylation of IкBα and p38-MAPK but not ERK following TL1A stimulation of sorted intestinal ILC3s (Fig. 4D). To test the dependence on MAPK or NF-кB signaling, we sorted ILC3 and pre-treated with soluble inhibitors prior to ex vivo TL1A stimulation. Although the NF-кB inhibitor NBD did not affect IL-22 production, pre-treatment with TAK1 or p38 MAPK inhibitors (5Z-7-Oxozeaenol and SB203580, respectively) completely blocked TL1A and IL-23 synergistic induction of IL-22 (Fig. 4E). No significant increase in STAT3 phosphorylation was detected in ILC3s stimulated with TL1A and IL-23 compared to IL-23 stimulation alone (Fig. S4C), and no transcriptional increase in Rorc, Ahr, HIFIa, or Stat3 was detected in ILC3s stimulated with TL1A to account for the increase in IL-22 (Fig. S4D). In monocyte-derived macrophages, TL1A synergy with muramyl dipeptide requires autocrine IL-1β to induce inflammatory cytokines (Hedl and Abraham, 2014). However, in contrast to this pathway used in macrophages, IL-1R on ILC3 was dispensable for TL1A and IL-23 synergistic induction of IL-22 (Fig. 4F).

To evaluate a physiologic role for IL-23 and TL1A synergy in vivo, we generated mice with both DR3- and IL23R-deficiency (Tnfrsf25^−/−Il23r^GFP/GFP). Mice with both DR3 and IL23R deficiency showed reduced survival following acute DSS-induced colitis compared to mice with a single deficiency of DR3 or IL23R (Fig. 4G). Collectively, these data reveal the direct synergy of these IBD-linked pathways in regulating ILC3 production of effector cytokines and protection from acute colitis.

TL1A induces OX40L-dependent co-stimulation of CD4⁺ T cells by MHCII⁺ ILC3

In contrast to our findings for an essential role for TL1A and DR3 in protection from acute colitis, previous studies have suggested a pathogenic role for TL1A overexpression in IBD (Zheng et al., 2013). To identify potential genes and pathways regulated by TL1A that may account for the pathogenic effects of DR3 in ILC3s during chronic colitis, we performed RNA-seq of TL1A-stimulated intestinal ILC3s. ILC3s were sort-purified from Il23r^GFP/WT mice and stimulated for 18 hours with or without recombinant TL1A. In addition to induction of Csf2 and Il22, RNA-seq analysis revealed significant induction of the co-stimulatory molecule OX40L following TL1A stimulation (Fig. 5A), which was confirmed by quantitative PCR of its gene Tnfsf4 (Fig. 5B). Flow cytometric analysis showed induction of surface expression of OX40L in all ILC3 subsets (Fig. 5C). Previous studies have shown that MHCII⁺ ILC3s lacking surface expression of classical co-stimulatory molecules negatively regulate T cell responses to commensal bacteria (Hepworth et al., 2013). Therefore, to test if TL1A-induced expression of OX40L could promote antigen-specific T cell responses, MHCII⁺ and MHCII^- ILC3 were sort-purified from Il23r^GFP/WT mice and co-cultured with CFSE-labeled naïve OT-II CD4⁺ T cells, which are specific for chicken ovalbumin (OVA) peptide 323–339, in the presence or absence of OVA peptide or recombinant TL1A stimulation. OVA peptide and TL1A stimulation of MHCII⁺ ILC3 induced multiple rounds of ovalbumin- specific T cell proliferation, similar to the proliferative levels seen in OVA-loaded dendritic cells (Fig. 5D, E). In contrast, no proliferation was seen in co-cultures with MHCII^- ILC3 or DR3-deficient ILC3 (Fig. 5D, E, F, S5A). To evaluate the role of ILC3 0X40L in co-stimulating T cells, ILC3s were cultured with CFSE-labeled naive OT-II CD4⁺ T cells in the presence of OVA peptide, recombinant TL1A, and α−0X40L blocking antibodies. Blockade of 0X40L, but not ICOSL, completely inhibited in vitro proliferation, suggesting a key role for 0X40L co-stimulation in regulating ILC3-driven T cell proliferation (Fig. 5G, S5B). To further test for a role for ILC3 0X40L expression, we generated mice with a conditional deletion of Tnfsf4 (Cortini et al., 2017) driven by Rorc- Cre recombinase (Tnfsf4^flox/flox Rorc-cre subsequently referred to as 0X40L^ΔRorc) (Fig. S5C). Sorted intestinal ILC3s from mice without 0X40L failed to support 0T-II T cell proliferation following TL1A stimulation in vitro (Fig. 5H, S5D). Finally, to determine if 0X40L was functioning as a co-stimulatory signal or acting independently of MHCII, we sorted ILC3s from mice with MHCII-deficient ILC3 (H2-Ab1^flox/floxRorc-cre) (Hepworth et al., 2015) and tested their capacity for supporting 0T-II T cell proliferation in vitro. Despite stimulation with TL1A, MHCII-deficient ILC3s were unable to support 0T-II proliferation (Fig. 5I). These experiments revealed a functional role for TL1A in inducing 0X40L co-stimulation of CD4⁺ T cells by MHCII⁺ ILC3s.

Fig. 5. TL1A induces OX40L-dependent stimulation of CD4⁺ T cells by MHCII⁺ ILC3s.

(A) Scatterplot of global gene expression profiles of media- or rTLIA-stimulated sorted Lin^-CD127⁺IL23R⁺ILC3s from ll23r^GFPlWT mice. Red and blue dots indicate significantly up- or down-regulated genes, respectively, after FDR correction for multiple hypothesis testing. Data are compiled from two independent experiments. (B, C) Tnfsf4 mRNA expression (B) and 0X40L surface protein (C) in sorted Lin^-CD127⁺IL23R^GFP+ ILC3s from ll23r^GFP/WT mice stimulated for 18 hours with rTL1A or media alone. (D, E, F) Sort-purified, CFSE-labeled CD4⁺ T-cells from OT-II mice were cultured for 6 days with sort-purified MHCII⁺ DCs, MHCII^- ILC3s or MHCII⁺ ILC3s, from either WT (D, E) or Tnfrsf25^flox/flox Rorc-cre RAG2^−/− (called DR3^ΔILC3) mice (F) with media alone, OVA peptide, or OVA peptide and rTL1A, as indicated. Proliferation index calculated by FlowJo is shown (E, F). One of four independent experiments is shown. (G) Sort- purified, CFSE-labeled CD4⁺ T-cells from OT-II transgenic mice were cultured for 6 days with sort-purified ILC3s in the presence of media alone, rTL1A, or with rTL1A plus co-stimulatory molecule blocking antibodies. All samples received OVA peptide in culture. One of two independent experiments is shown. (H, I) ILC3s sorted from Tnfsf4^flox/flox or Tnfsf4^flox/flox Rorc-cre (called OX40L^ΔRorc) (H) or H2-Ab1^flox/floxRorc-cre (called MHCII ^ΔILC3) (I) were stimulated with rTL1A and co-cultured with CFSE-labeled OTII CD4⁺ T cells. One of two independent experiments is shown. Data in B, E, F, G, H and I are mean ± SEM and were analyzed by Student’s t-test (B) or one-way ANOVA (E, F, G, H, and I); ***P<0.001. See also Figure S5.

ILC3 expression of OX40L in colitis regulates pathogenic T cells

To evaluate ILC3 expression of 0X40L in vivo during colitis, we performed flow cytometry on lamina propria ILC3s at steady state and following DSS-induced colitis. Although low 0X40L was seen on ILC3s at steady state, colitis induced robust surface expression of 0X40L protein (Fig. 6A). Expression of 0X40L was significantly reduced on the MHCII⁺ ILC3s from DR3-deficient mice compared to littermate controls, revealing a key role for TL1A in ILC3 expression of OX40L in vivo (Fig. 6B). To determine if ILC3s express OX40L during inflammatory bowel disease (IBD), we profiled ILC3 from human intestinal biopsy samples (Fig. 6C, Fig. S6). Lin^- CD127⁺ c-Kit⁺ ILC3 from patients with active CD showed significantly higher levels of surface OX40L expression. Collectively, these results reveal TL1A- and colitis-dependent OX40L expression by intestinal ILC3s in mouse and human.

Fig. 6. TL1A induces ILC3 expression of OX40L during colitis in vivo.

(A) OX40L expression on Lin^-CD127⁺ RORγt⁺ ILC3s at the steady-state or following 2% DSS- induced colitis. (B) OX40L expression on Lin^-CD127⁺ RORγt⁺ ILC3s from Tnfrsf25^flox/flox RAG2^−/− (called DR3^flox/flox) and Tnfrsf25^flox/flox Rorc-cre RAG2^−/− (called DR3^ΔILC3) littermate mice following 2% DSS-induced colitis (N=4 mice/group). One of two independent experiments is shown. (C) Human OX40L on Lin^-CD127⁺c-KIT⁺ ILC3 from colonic biopsies of healthy patients or IBD patients with active inflammation (N=5–6 patients/group). Data in B and C are mean ± SEM and were analyzed by two-tailed Student’s t-test; **P<0.01. See also Figure S6.

We next sought to evaluate the role for TL1A-induced ILC3 expression of OX40L in regulating intestinal T cell in vivo. To test the role for ILC3 expression of OX40L in regulating antigen-specific T cells, we sorted naïve CD45.1⁺ OT-II cells and adoptively transferred them into OX40L^ΔRorc mice and littermate controls. Mice were exposed to DSS for 5 days (to induce OX40L) and provided ovalbumin antigen ad lib (Fig. 7A,S7A). One week following T cell transfer, we found reduced numbers of CD45.1⁺ CD4⁺ T cells in the mesenteric lymph node of OX40L^ΔRorc mice compared to littermate controls, indicating a role for OX40L on intestinal ILC3s in regulating an antigen-specific T cell response.

Fig. 7. ILC3 expression of OX40L regulates antigen-specific and T-BET⁺ Th1 cell activation in T cell colitis.

(A) Transfer of naïve CD45.1⁺ CD3⁺CD4⁺Vα2⁺ OT-II T cells into CD45.2⁺ Tnfsf4^flox/flox or CD45.2⁺ Tnfsf4^flox/flox Rorc-cre (called OX40L^ΔRorc) mice treated with 2% DSS and ovalbumin for 5 days. Flow cytometric analysis of CD45.1 ⁺ CD3⁺CD4⁺Va2⁺ OT-II T cells from compiled over two independent experiments (N=6–8 mice/group). (B) RORγt and T-BET in CD4⁺ T cells from intestinal lamina propria of Tnfsf4^flox/flox or OX40L^ΔRorc littermate mice (N=4 mice/group). (C, D) Percent body weight (C) and representative colonic histology (D) of Tnfsf4^flox/flox RAG2^−/− (called OX40L^flox/flox, N=8) and Tnfsf4^flox/flox Rorc-cre RAG2^−/− (called OX40L^ΔILC3, N=7) littermate mice following adoptive transfer of CD4⁺CD45RB^high T cells to induce transfer T cell colitis. (E, F) Flow cytometric analysis of intracellular IFNγ (E) or intranuclear T-BET (F) in CD4⁺ T cells from OX40L^flox/flox and OX40L^ΔILC3 littermate mice in (C). (G, H) Tnfrsf25^flox/flox RAG2^−/− (called DR3^flox/flox, N=6) and Tnfrsf25^flox/flox Rorc-cre RAG2^−/− (called DR3^ΔILC3, N=7) littermate mice following adoptive transfer of CD4⁺CD45RB^high T cells to induce transfer T cell colitis. Body weights, percent survival (G) and representative colonic histology (H) from DR3^flox/flox and DR3^AILC3 littermate mice following adoptive transfer of CD4⁺CD45RB^high T cells (scale bar: 100μm). (I) Flow cytometric analysis of T-BET induction in transferred T cells from DR3^flox/flox and DR3^ΔILC3 littermate mice in (G) and (H). Data in A, B, E, F, and I are mean ± SEM and were analyzed by two-tailed Mann-Whitney test (A, B) or two-tailed Student’s t-test (E,F,I). Data in C and G were analyzed by Log-rank (Mantel-Cox) test; * P< 0.05, **P<0.01, ***P<0.001. See also Figure S7.

At steady state, endogenous CD4⁺ T cells from the intestinal lamina propria of OX40L^ΔRorc mice showed equivalent RORγt⁺ cells, but a significant reduction in T-bet⁺ cells (Fig. 7B, S7B). To test if ILC3 expression OX40L regulated inflammatory T cell colitis, we used the T cell transfer colitis model. Induction of T cell-dependent disease has previously been shown to require OX40 on transferred T cells (Griseri et al., 2010). We sort-purified CD4⁺CD45RB^high naïve T cells from WT mice into Tnfsf4^flox/flox Rorc-cre RAG2^−/− (called OX40L^ΔILC3) or Tnfsf4^fox/fox RAG2^−/− (called OX40L^ΔILC3) littermate controls. In contrast to littermate controls which showed a failure to gain weight and intestinal inflammatory infiltrates consistent with T cell colitis, OX40L-deficient mice continued to gain weight and showed no significant intestinal inflammation (Fig. 7C, D). Colonic CD4⁺ T cell production of IFNγ and T-BET expression was significantly reduced in OX40L-deficient animals compared to controls (Fig. 7E, F). Given the role for TL1A signaling in regulating ILC3 expression of OX40L, we next tested the impact of DR3- deletion in the T cell transfer colitis model. Weight loss and reduced survival consistent with chronic T cell-mediated colitis developed within 4 weeks in the RAG2^−/− littermate controls, but similar to the OX40L-deficient mice, ILC3-specific deletion of DR3 protected mice from T cell transfer colitis (Fig. 7G). No differences were observed in the microbiome composition of OX40L- or DR3-deficient mice compared to littermate controls (Fig. S7C, D). Analysis of colonic tissue in these mice revealed decreased histopathologic damage (Fig. 7H) and a significant reduction in T-BET⁺ CD4⁺ T cells (Fig. 7I) in the DR3^ΔILC3 mice compared to littermate controls. Thus, in contrast to the protective role for TL1A in acute colitis, these data reveal a pro-inflammatory role for ILC3-intrinsic DR3 in chronic T cell colitis.

DISCUSSION

Our data highlight the role for the IBD-linked gene TNFSF15 and its protein TL1A as a central regulator of ILC3 function linking innate and adaptive immunity. Previous reports on the role for TL1A in inflammatory colitis offered contradicting results with TL1A promoting protection in acute models of colitis (Jia et al., 2016) and exacerbating T cell-dependent inflammation in chronic colitis (Zheng et al., 2013). Using cell-specific genetic deletion models, our results now help to reconcile these findings, revealing a central role for ILC3 in mediating acute protection in acute colitis models (DSS and Citrobacter rodentium colitis), but promoting DR3-dependent pathogenic T cell responses in chronic disease (T cell transfer colitis). These findings highlight the protective role for TL1A in promoting mucosal healing during IBD (Bamias et al., 2003; Prehn et al., 2004). In synergy with IL-23, TL1A can act as a rheostat to regulate ILC3 effector functions. TL1A synergy with IL-23 in ILC3 required p38 MAPK, but no significant change in key transcription factors (R0Rγt, Ahr, Hif1a) or STAT3 phosphorylation was observed. The direct and indirect mechanistic targets of TL1A regulation of IL-22 in ILC3s still need to be defined. 0n a population level, genetic variants promoting this TL1A-dependent mucosal healing response may confer protection against infectious pathogens; however, sustained induction of TL1A, in patients with additional genetic susceptibility or environmental exposure, may support pathogenic T cell expansion (Zheng et al., 2013) or effector ILCs during chronic inflammation.

Our results support a key role for IBD-associated adherent bacteria in regulating MNPs as the main producers of intestinal TL1A. As sentinels of barrier immunity, MNPs integrate microbial signals and activate both innate and adaptive effector immunity. By producing IL-23 and IL-1β as well as CXCL16, MNPs are positioned functionally and spatially to rapidly coordinate ILC3 effector cytokine production (Longman et al., 2014; Satoh-Takayama et al., 2014). Adherent microbiota, including SFB and Adherent- invasive E. coli, induce TL1A consistent with their ability to potently induce ILC3 production of IL-22 (Sano et al., 2015; Viladomiu et al., 2017). These findings offer a mechanism by which Crohn’s disease-associated microbiota can act acutely to promote barrier immunity in wild-type hosts. The adherence and metabolic characteristics of the intestinal microbiota may offer a novel approach for regulating TL1A effects in the mucosa.

The discovery of antigen processing and presentation by intestinal ILC3 highlights a central role for these cells in linking innate and adaptive immunity (Hepworth et al., 2013). Although splenic ILC3 can upregulate conventional co-stimulatory molecules CD80 and CD86 following IL-1β stimulation (von Burg et al., 2014), the absence of these co-stimulatory molecules following stimulation on intestinal ILC3s lead to a model in which they limit bacteria-specific inflammatory T cell responses in the gut (Hepworth et al., 2015). Our data demonstrate expression of OX40L by ILC3 after TL1A stimulation and highlight increased expression of 0X40L by ILC3 in inflammatory colitis in mouse models and human IBD. Expression of 0X40L by ILC3s plays a critical role in regulating gut antigen-specific and effector T cell responses. Although the total number of intestinal ILC3s are limited, functional and spatial organization of these cells facilitate efficient regulation of immunity in the tissue (Satoh- Takayama et al., 2014). These data, therefore, extend the model by which MHCII⁺ ILC3s may regulate the intestinal T cell response. In IBD patients with genetic susceptibility or animal models of T cell transfer colitis, ILC3 co-stimulation can support pathogenic inflammatory T cell activation. Under homeostatic conditions, this model may extend to other 0X40L-dependent regulatory responses (Griseri et al., 2010). Recent studies have demonstrated significant heterogeneity in ILC3 populations, including MHCII⁺ and MHCII^- subsets (Bjorklund et al., 2016; Gury-BenAri et al., 2016; Lim et al., 2017). While our data show an essential role for ILC3-specific OX40L in mediating TLIA-dependent T cell responses, further studies are needed to define intrinsic features of MHCII⁺ or MHCII^- ILC3 that contribute to their role in tissue specific T cell regulation.

These studies reveal a central role for TL1A in directing ILC3-dependent mucosal immunity. We now propose a model in which MNP-derived TL1A stimulation of intestinal ILC3 serves as a critical component linking innate and adaptive barrier immunity in colitis. During acute colitis, TL1A synergy with IL-23 promotes mucosal healing, but sustained intestinal damage leads to the upregulation of OX40L on ILC3 and co-stimulation of inflammatory intestinal T cells. Our data from IBD patients reveal a conserved TL1A - OX40L pathway in ILC3 with the potential to guide diagnostic and therapeutic approaches for IBD.

STAR METHODS

Experimental Models and Subject Details

Mice.

C57BL/6, Itgax-cre, Rorc-cre, OT-II, illr^−/−, Cx3cr1-GFP, Cx3cr1-CreER mice were purchased from The Jackson Laboratory. All mouse models were on C57BL/6 background. CX3CRI-DTR mice (Longman et al., 2014) were previously described. Il23r^GFP mice were obtained from M. Oukka (Awasthi et al., 2009). Myd88^−/− were obtained from J. Blander. MHCII^ΔILC3 mice were obtained from G. Sonnenberg (Hepworth et al., 2015). Tnfsf4^flox were provided by T. Vyse and M. Botto (Cortini et al., 2017). Tnfrsf25^−/− mice were obtained from Cancer Research UK (Wag et al., 2001). Generation of Tnfrsf25^flox/flox mice is previously described (Shih et al., 2014). Tnfsf15^flox/flox mice were generated at Cedars-Sinai by D. Shih and S. Targan. Tnfsf15 endogenous locus containing 4.5 bp upstream and 3.4 bp downstream of exon 1 were generated by PCR amplification using proprietary C57BL/6J library developed at genOway. Subsequently, two loxP sites were inserted flanking exon 1. Positive selection neomycin gene flanked by FRT sites was inserted to the intron after exons 1 to generate the targeting vector. Every step of the cloning process was validated through restriction enzyme analysis and sequencing. The Tl1a gene targeting construct was linearized and electroporated into embryonic stem (ES) cells with C57BL/6J background. Homologous recombinants were selected by G418 and confirmed by PCR and Southern blot analysis. ES clones with correct 5’ and 3’ recombination were microinjected into C57BL/6J blastocysts and introduced into pseudopregnant C57BL/6J mice. Male chimeric offspring were bred to obtain germ line mutant mice, which were then bred to flpe delete mouse strain to remove the neomycin cassette and were confirmed by Southern blot. For CX₃CR1⁺ MNP depletion studies using Cx3cr1^DTR mice, diptheria toxin was i.p. injected (200ng/200uL) for two consecutive days and then every 72 hours to maintain depletion. For recombinant IL-22 rescue experiments of DR3^−/− and DR3^ΔILC3, 5ug of rIL-22 was injected i.v. on days 4 and 6. All experiments were performed with 6–8 week old littermates. Both male and female mice were used with random and equal assignment of same sex to each experimental group. All vertebrate work was approved by the IACUC at Weill Cornell Medicine.

Human IBD subjects.

Endoscopic biopsies were obtained under an Institutional Review Board-approved protocol (1103011578) and informed consent was obtained at Weill Cornell Medicine (WCM) including patients >18 years of age. Active inflammation was defined by an endoscopic score of >2 and inactive disease was defined by an endoscopic score of 0. The age and gender of subjects included are as follows: Healthy controls—22, 38, 53, 55, 71, 72, 79 year old males and 36 year old female; Crohn’s inactive—30, 52, and 57 year old males; Crohn’s active—32, 36 year old males and 27, 44, 55, 55, 62, and 68 year old females.

Method details

Antibodies and flow cytometry.

Staining of mouse cells was performed with CD3- FITC (145–2C11), CD4-eFluor780 (RM4–5), CD5-PE (53–7.3), CD45.1-APC (A20), CD90.2-efluor450 (53–2.1), MHCII-Alexa700 (M5/114.15.2), F4/80-PE (BM8), Ly6C- eFluor450 (HK1.4), CD19-FITC (MB19–1), CD11b-eFluor780 (M1/70), CD11c-PE-Cy7 (N418), CD103-APC (2E7), TCRγδ-FITC (11–5711-82), KLRG1-PE (2F1), NKp46- eFluor710 (29A1.4), OX40L-APC (RM134L), and ICOSL-eFluor660 (HK5.3), all from eBioscience. The following antibodies were purchased from BioLegend: CD127-PE- Cy7 (A7R34), CCR6-BV421 (29–2L17), DR3-PE (4C12). Intracellular mouse cytokine staining was performed with ROR0γt-PE (B2D), T-bet-e660 (eBio4B10), Foxp3-e450 (FJK-16s), IL-22-APC (IL22JOP), GMCSF-FITC (MP1–22E9), IL-13-PE-Cy7 (eBio13A), IL-17-PE (eBio17B7), IFNγ-PE-Cy7 (XMG1.2), TL1A-e710 (Tandys1a), all from eBioscience. Extracellular staining of human cells was performed with CD14-PE (61D3) and CD14-Alexa700 (63D3), HLA-DR-Alexa700 (LN3), TL1A-eFluor710 (Tandys1a), CD45-APC (2D1), CD3-eFluor780 (UCHT1), CD5-eFluor780 (UCHT2), CD19-Alexa700 (HIB19), CD11c-Alexa700 (3.9; Invitrogen), CD127-PE-Cy7 (eBioRDR5), ST2L-FITC (B4E6), CD117-efluor450 (104D2), OX40L-PE (11C3.1;Biolegend), all from eBioscience unless otherwise noted. Dead cells were excluded using the LiveIDead fixable aqua dead cell stain kit (Invitrogen). Cell proliferation was measured using CFSE (eBioscience). Phosphoflow was performed using pERK1I2-Pacific Blue (20A), pIKBa- e660 (RILYB3R), p38-MAPK-Alex647 (pT180IpY182). Intracellular cytokine staining was performed according to the manufacturer’s protocol (CytofixICytoperm buffer set; BD) and transcription factor staining for was performed according to manufacturer’s protocol (Intracellular Fixation and Permeabilization kit; eBioscience). Flow cytometry was performed with a BD LSRFortessa and analyzed with FlowJo software (Tree Star). Proliferation indexes were calculated as per FlowJo as the total number of divisions divided by the pool of dividing cells.

Preparation of LPMC.

Mouse intestines or human endoscopic biopsies were washed in PBS and 1 mM DTT with 30 mM EDTA, and then digested in collagenase 8 (Sigma- Aldrich) and DNase-containing media with 10% fetal bovine serum. Digested material was passed through a cell strainer and separated on a discontinuous 40%/80% Percoll gradient. LPMCs were cultured ex vivo in the presence of GolgiPlug (BD) for 4 h or stimulated with phorbol myristate acetate (PMA; 20 ng/ml) and ionomycin (1 μg/ml) or IL-23 (40 ng/ml; eBioscience) in the presence of GolgiPlug (BD) for 4 h before staining.

Intestinal ILC cultures.

Lineage^- KLRG1^- IL23R-GFP mouse ILC3 were sorted from LPMCs and resuspended in RPMI (10% FBS) tissue culture media for stimulation directly ex vivo. Mouse ILCs were stimulated with mouse recombinant IL-23 (eBioscience; 40 ng/ml), IL-1 (eBioscience; 10 ng/ml), or TL1A (R&D Systems; 200 ng/ml), as indicated. After 18 h, supernatants were harvested for IL-22 or GM-CSF ELISA (eBioscience) and Golgi Plug (BD) was added to cells for 4 h for subsequent intracellular cytokine staining.

Co-culture assays.

Sort-purified ILC3 (2×10⁴) and MNP (1×10⁴) populations were co-cultured in a 96-well round-bottom plate in tissue culture media. Cultures were incubated at 37^oC for 18h. Supernatants were harvested for ELISA and remaining cells were incubated with Golgi Plug (BD) for 4 hours and subsequently analyzed by flow cytometry. For antigen-specific T cell proliferation assays, 2×10⁵ sort-purified, CFSE- labeled naive CD4⁺ T cells from OTII transgenic mice were co-cultured with either 2.5×10⁴ sort-purified ILC3 from IL23R^GFP/wt mice or 2×10⁴ MHCII⁺CD11c⁺CD11b⁺ MNPs, with 20ug/mL OVA peptide (OVA 323–339; InvivoGen) and 200ng/mL rTL1A, as indicated, in a 96-well flat bottom plate. For co-stimulatory blocking assays, ILC3/Tcell co-cultures were blocked with 20ug/mL anti-mouse ICOSL antibody (LEAF purified anti-mouse CD275; BioLegend) or 15ug/mL anti-mouse 0X40L antibody (Biotin anti-mouse CD252; BioLegend).

Colitis models.

To induce chemical colitis in mice, 2% DSS (w/v) (M.W. 40,000–50,000; Affymetrix) was added to sterile drinking water and administered ad libitum for 7 days. After 7 days, DSS was replaced with normal drinking water. To induce infectious colitis, mice were orally gavaged with log phase growth Citrobacter rodentium DBS100 (ATCC 51459; American Type Culture Collection) at 1×10¹⁰ in PBS. For T cell transfer colitis, 5×10⁵ CD4+ CD45RB^high CD25^- CD62L^- T cells were transferred into DR3^ΔILC3 or DR3^flox/flox mice. For all colitis experiments, mice were monitored daily (DSS) or weekly (T cell transfer) for weight loss, rectal bleeding, diarrhea and survival throughout the experiments. For in vivo experiments using AIEC 2A, mice were pre-treated for 21 days with a broad-spectrum antibiotic cocktail of ampicillin (1 g/L), neomycin (0.5g/L), metronidazole (0.5g/L) and vancomycin (0.5g/L) in their drinking water. Mice were then orally gavaged with 2×10⁹ of AIEC 2A before colitis induction with 2% DSS in their drinking water. For antigen-specific T cell transfer, 5×10⁵ naive CD45.1 ⁺ CD3⁺CD4⁺ CD44^- CD25^- CD62L⁺Vα2⁺ OTII T cells were sort purified and transferred i.p.

qPCR.

RNA from primary intestinal cell populations was prepared with TRIzol (Invitrogen) and purified RNA was reverse transcribed into cDNA (Quanta qScript). qPCR was performed with a QuantStudio 6 Flex Real-time PCR (Applied Biosystems) with SYBR Green Supermix (Bio-Rad Laboratories). The following primers were used: for Reg3g, 5’-GGGATCTGCAAGACAGACAAG-3’ and 5’-GGGGCATCTTTCTTGGCAAC’3’; for Saa1, 5’-CATTTGTTCACGAGGCTTTCC-3’ and 5’- GTTTTTCCAGTTAGCTTCCTTCATGT-3’; for Tnfsf15, 5’- ATGCTTCGGGCCATAACAGA-3’ and 5’-TGAAGGCCATCCCTAGGTCA-3’; for Il22, 5’- GCTCAGCTCCTGTCACATCA-3’ and 5’-CAGTCCCCCAATCGCCTTGA-3’; for 0×40l, 5’-GTGGTCTCTGGGATCAAGGG-3’ and 5’-TCACATCTGGTAACTGCTCCTC-3’; for Hprt, 5’-GAGGAGTCCTGTTGATGTTGCCAG-3’ 5’- GGCTGGCCTATAGGCTCATAGTGC-3’ (Stephens et al., 2011). The thermocycler program was as follows: initial cycle of 95^oC for 60s, followed 40 PCR cycles at 95^oC for 5s, 60^oC for 15s, 72^oC for 15s. Relative levels of the target genes were determined by using the ∆∆C_t of the target gene compared to Hprt expression.

16S rRNA Gene Sequencing and Analyses.

16S rRNA gene sequencing methods were adapted from the methods developed for the Earth Microbiome Project (Caporaso et al., 2011). Briefly, bacterial genomic DNA was extracted using the Maxwell RSC (Promega). The 16S rDNA V4 and V5 regions were amplified by PCR and sequenced in the MiSeq platform (Illumina) using the 2×250 bp paired-end protocol (Caporaso et al., 2012). The read pairs were demultiplexed based on the unique molecular barcodes, and reads were merged using USEARCH 10.0.240 (Edgar, 2010). Taxonomic labelling of OUT representative sequences was carried out with a confidence threshold of 0.8, using the Ribosomal Database Project training set (version 16) (Cole et al., 2014). A rarefied OTU table from the output files generated in the previous two steps was used for downstream analyses of α-diversity, β-diversity (Lozupone and Knight, 2005), and phylogenetic trends.

Quantification and Statistical Analysis

RNA-seq analysis.

Mouse RNA-Seq data was aligned to the mm10 assembly of the mouse genome using the STAR aligner¹. Read counts for genes across all BAM files were computed using the HTSeq software². Differential analysis of RNA-Seq samples utilized the DESeq package for gene expression analysis³. False discovery rate correction of P-values used for all bioinformatics analyses of this study utilized the Benjamini-Hochberg procedure⁴. Heatmap clustering and production was done using the heatmap function within the NMF package in R⁵. Principal components analysis plots were visualized using the ggbiplot package in R.

Statistical analysis.

Statistical analysis was performed in GraphPad Prism or R software. Results represent mean ± s.e.m. and were analyzed by unpaired Student’s t- test, Mann-Whitney test, one-way ANOVA, Log-rank (Mantel-Cox) test, as indicated in the figure legends. Given that mouse experiments required littermate controls and complex genotyping, experimental group allocation was not blinded. No relevant exclusion criteria were applied.

Highlights.

• IBD-associated adherent microbiota induce MNP-derived TL1A

• MNP-derived TL1A promotes the ILC3-IL-22 axis and protects against acute colitis

• TL1A induces ILC3 expression of OX40L during colitis in vivo

• OX40L on ILC3s stimulates T cell activation and is required for T cell-driven colitis