Group 3 innate lymphoid cells mediate intestinal selection of commensal bacteria-specific CD4⁺ T cells

Pathologic CD4⁺ T cell responses to self are limited by presentation of self-antigens in the thymus on MHCII⁺ thymic epithelial cells (TECs) and dendritic cells (DCs), resulting in clonal deletion (1-6). In contrast, commensal bacteria-specific CD4⁺ T cells, which have been implicated in the pathogenesis of inflammatory bowel disease (IBD) (7-11), do not encounter cognate antigen in the thymus and therefore are not subject to negative selection prior to entering the periphery (12, 13). While physical and biochemical barriers separate the immune system from intestinal commensal bacteria (7, 13-17), antigens derived from commensal bacteria are continuously sampled from the intestinal lumen and presented by DCs in the draining lymph nodes (7, 10, 15, 18, 19). Regulatory T cells (T_reg) can in part limit dysregulated CD4⁺ T cell responses to commensal bacteria (20, 21). However, whether other mechanisms control commensal bacteria-specific CD4⁺ T cells in lieu of thymic negative selection is poorly defined. In recent studies, ILC3s were found to express MHCII and genetic deletion of ILC3-intrinsic MHCII resulted in spontaneous CD4⁺ T cell-dependent intestinal inflammation (22), suggesting additional antigen-presentation pathways control commensal bacteria-specific CD4⁺ T cell populations.

CCR6-expressing lymphoid tissue inducer (LTi)-like ILC3s (CCR6⁺ ILC3s), are a major ILC subset present in the mesenteric lymph node (mLN) (Fig. 1A) and colon lamina propria (cLPL) (fig. S1A) of healthy mice, and constitutively express retinoic acid-related orphan receptor gamma t (RORγt) and MHCII, relative to ST2⁺ group 2 ILCs (ILC2) (fig. S1B-C). ILC3s are regulated by various cytokine, environmental and microbial factors (23, 24). However, interleukin (IL)-23p19, the aryl hydrocarbon receptor (Ahr) or the intestinal microbiota were not required for CCR6⁺ ILC3 expression of MHCII (fig. S1D-F), although ILC3 frequencies were reduced in the intestine of Ahr^-/- mice, as previously described (fig. S1E) (25). Further, in contrast to a recent report (26), expression of MHCII, CD80 and CD86 on CCR6⁺ ILC3s were unaffected by ex vivo stimulation with microbial or inflammatory stimuli, or by the absence of MyD88 or Caspase 1/11 in vivo (fig. S2).

Next we examined whether expression of ILC3-intrinsic MHCII was dependent on the class II transactivator (CIITA), a master transcriptional regulator of MHCII expression (27). MHCII expression was absent on CCR6⁺ ILC3s from Ciita^-/- mice, relative to Ciita^+/+ control mice (Fig. 1B and fig. S3A).

Transcription of CIITA in mice is driven via distinct promoter elements, termed pI, pIII and pIV, that are indicative of the upstream signaling events that induce MHCII expression (27). In contrast to B cells and DCs, CCR6⁺ ILC3-intrinsic MHCII expression was absent in the mLN of both pIII/pIV^-/- and pIV^-/- mice (Fig 1C, fig. S3B-C), indicating that the pIV promoter of Ciita is required for MHCII expression on CCR6⁺ ILC3s. The pIV promoter of Ciita is utilized by multiple cell types, such as epithelial cells, in response to interferon (IFN)-γ signaling (27). However, expression of MHCII on CCR6⁺ ILC3s was not impaired in the absence of IFN-γ, IFN-γR1 or STAT-1 (Fig. 1D, fig. S3D-E). MHCII expression in TECs is also dependent upon the pIV promoter of Ciita (fig. S3A-B) and pIV-dependent, IFN-γ-independent CIITA expression has previously only been described in TECs (27-29), suggesting a previously unappreciated link between these cell types.

These data provoked the hypothesis that TECs and MHCII⁺ ILC3s share similar functional roles in the selection of CD4⁺ T cells. To test this, we examined CD4⁺ T cells in the intestine of mice with an ILC3-intrinsic deletion in MHCII (MHCII^ΔILC3 mice). As we previously reported (22), frequencies and cell numbers of CD44^hi CD4⁺ T cells (T_eff) in the cLPL of MHCII^ΔILC3 mice were increased relative to H2-Ab1^fl/fl controls (fig. S4A-B). In contrast, the total numbers of CD44^lo naïve T cells and FoxP3⁺ T_reg were unchanged (fig. S4B). T_eff cell populations utilized a broad range of T cell receptor (TCR) Vβ chains and T_eff expansion was detected in the cLPL but not in the thymus (fig. S4C). Further, sort-purified CD4⁺ T cells from MHCII^ΔILC3 mice responded to fecal-derived antigen, but not mammalian tissue-derived antigens (fig. S4D).

These data, along with previous studies (20, 30, 31), suggest the majority of CD4⁺ T cells in the steady-state intestine are specific for commensal bacteria and that ILC3-intrinsic MHCII controls commensal bacteria-specific CD4⁺ T cell responses through direct presentation of microbiota-derived antigens. To test this, we crossed MHCII^ΔILC3 mice with either TCR transgenic mice specific for CBir1, an antigen expressed by Clostridia species constitutively present in the murine and human microbiota (13, 32), or TCR transgenic mice specific for ovalbumin (OT-II). Loss of ILC3-intrinsic MHCII had no effect on the frequencies or cell numbers of OT-II T cells or CBir1 T cells in the thymus (Fig. 2A). In contrast, numbers of CBir1, but not OT-II, T cells were increased in the cLPL and mLN (Fig. 2B and fig. S5A), and CBir1^MHCIIΔILC3 mice exhibited increased frequencies of antigen-specific IFN-γ⁺ and tumor necrosis factor (TNF)-α⁺ colonic CD4⁺ T cells, colonic inflammation and neutrophil infiltration, which could be prevented by administration of antibiotics (ABX), and was not observed in Rag1^-/- MHCII^ΔILC3 mice (Fig. 2B-D and fig. S5A-D). Finally, similar to polyclonal MHCII^ΔILC3 mice, the population expansion and increased cytokine production of CBir1 CD4⁺ T cells was associated with a selective increase in CD44⁺CD62L^lo T_eff, whereas numbers of FoxP3⁺ T_reg remained unchanged (fig. S5E). These data suggest that CCR6⁺ ILC3s selectively limit the expansion of commensal bacteria-specific CD4⁺ effector T cells through presentation of antigen derived from endogenous commensal bacteria. In support of this hypothesis and consistent with recent findings (33), MHCII⁺ ILC3s localized in distinct clusters in the mLN at the interface between the B- and T- cell zones in the marginal and subcapsular sinus (Fig. 2E), a site through which antigen-experienced T cells traffic.

To investigate the in vivo mechanisms through which MHCII⁺ ILC3s control commensal bacteria-specific CD4⁺ T cells, mice were generated in which MHCII expression was restricted to only ILC3s. This was accomplished by utilizing mice with a floxed-STOP sequence cassette inserted between the first and second IAβ^b exons (IAβ^bSTOP^fl/fl mice) (34). IAβ^bSTOP^fl/fl mice lack MHCII on antigen-presenting cells in the absence of Cre recombinase (MHCII^neg) (Fig. 2G). In contrast, IAβ^bSTOP^fl/fl mice crossed with Rorc^Cre mice demonstrated a partial restoration of MHCII on CCR6⁺ ILC3s (MHCII^ILC3+ mice) (Fig. 2F and G), but not B cells, DCs and macrophages (Fig. 2G). As MHCII^neg and MHCII^ILC3+ mice lack MHCII on TECs and have disrupted endogenous T cell selection (34), we first employed an adoptive transfer approach with naïve CFSE-labeled CD45.1⁺ CBir1 T cells. In mice with normal MHCII expression (MHCII^pos) administration of CBir1 peptide resulted in dilution of CFSE in transferred CBir1 T cells, expansion of CD44^hiCD62L^lo CBir1 T_eff cells and an increase in FoxP3⁺ T_reg in the mLN (fig. S6A). In contrast, naïve CBir1 CD4⁺ T cells transferred into control MHCII^neg and MHCII^ILC3+ mice failed to proliferate or differentiate into T_eff or T_reg following peptide administration (fig. S6A), suggesting that ILC3-intrinsic MHCII does not influence naïve CD4⁺ T cells. As MHCII⁺ ILC3s localize at lymphoid sites through which antigen-experienced CD4⁺ T cells traffic (Fig. 2E) (33), CBir1 CD4⁺ T cells were pre-activated overnight prior to transfer into MHCII^neg or MHCII^ILC3+ mice. Following CBir1 peptide administration, MHCII^ILC3+ mice exhibited reduced frequencies and numbers of pre-activated CBir1 CD4⁺ T cells in the mLN (Fig. 2H-I), which was due to a selective decrease in effector, but not regulatory CBir1 T cells (Fig. 2I). MHCII⁺ ILC3-mediated effects were antigen-specific, could be driven by endogenous microbiota-derived antigen alone and were specific to ILC3s (fig. S6B-D and fig. S7). Further, in complementary loss-of-function studies, CBir1 CD4⁺ T_eff expanded at higher frequencies and numbers in the mLN and cLPL of mice lacking ILC3-intrinsic MHCII (MHCII^ΔILC3 mice), despite exhibiting comparable proliferation (fig. S8). Finally, in line with ILC3-mediated control of antigen-experienced T cells, the frequency and number of activated CBir1 CD4⁺ T cells reaching the small intestine lamina propria (siLPL) and cLPL of MHCII^ILC3+ mice were significantly decreased relative to MHCII^neg controls (Fig. 2J-K).

We hypothesized that the loss of CBir1 CD4⁺ T cells from the mLN and intestine of MHCII^ILC3+ mice could be the result of altered migration, proliferation or through induction of cell death. However, pre-activated CBir1 T cells did not accumulate in peripheral organs, such as the spleen (fig. S9A) and CBir1 T cells recovered from MHCII^ILC3+ mice exhibited comparable proliferation, relative to MHCII^neg mice (fig. S9B). To further investigate this question, we developed an in vitro ILC3-CD4⁺ T cell co-culture system. Consistent with our in vivo findings, co-culture of activated CBir1 T cells and sort-purified MHCII⁺ CCR6⁺ ILC3s resulted in a significant reduction in T cell numbers following culture in the presence of cognate antigen, which could be reversed by administration of an MHCII-blocking antibody (Fig. 3A). Reduced cell recovery was associated with an antigen and MHCII-dependent increase in Caspase-3 activation (Fig. 3B) and increased Annexin V staining (Fig. 3C) in the remaining CBir1 T cells, indicative of programmed cell death. Despite selectively regulating commensal bacteria-specific CD4⁺ T cells in the steady state, MHCII⁺ ILC3s were also sufficient to influence T cells with other antigen-specificities only if antigen was systemically provided (fig. S10).

Negative selection in the thymus has been shown to be associated with induction of Nur77 and subsequent upregulation of the pro-apoptotic molecule Bim (35). Antigen-dependent interactions between CBir1 T cells by MHCII⁺ ILC3s also resulted in the upregulation of Nur77 as well as Bim, which was required for ILC3-mediated induction of cell death (Fig. 3D-G). Antigen-presentation by ILC3s in vitro selectively led to T_eff death, but did not affect T_reg numbers (fig. S11A). We next analyzed mLN-derived CCR6⁺ ILC3 for expression of surface molecules that directly influence antigen-specific CD4⁺ T cell responses (Fig. 3H). ILC3s demonstrated high levels of MHCII-associated transcripts, but had negligible expression of transcripts for canonical co-stimulatory molecules and inhibitory or death receptors (Fig. 3H). Indeed, CCR6⁺ ILC3s lacked expression of FasL by flow cytometry and antibody-mediated neutralization of FasL did not influence ILC3-induced CBir1 T cell death (fig. S11B-D). Moreover, in contrast to Bim^-/- mice, FasL^gld/gld mice did not exhibit increased frequencies of endogenous commensal bacteria-specific CD4⁺ T cells in gut-associated lymphoid tissues (fig. S11E).

Bim-dependent apoptotic cell death may also be induced via cytokine or growth factor starvation (36, 37). As CCR6⁺ ILC3s constitutively express high levels of the common gamma chain cytokine receptors CD25 (IL-2R) and CD127 (IL-7Rα) (Fig. 1, fig. S1 and Fig. 3H), we hypothesized that MHCII⁺ ILC3s may induce cell death of commensal bacteria-specific CD4⁺ T cells synergistically through TCR-induction of an apoptotic program in concert with cytokine withdrawal. Consistent with this, cell death could be reduced upon addition of exogenous recombinant (r)IL-2, but not rIL-7, to in vitro co-cultures (Fig. 3I). MHCII⁺ ILC3s exhibited over twofold higher capacity to bind IL-2 as compared to activated CBir1 CD4⁺ T cells (fig. S11F-G), suggesting MHCII⁺ ILC3s out-compete activated T cells for pro-survival cytokines. The requirement for IL-2 was T cell-intrinsic as activated CBir1 CD4⁺ cells expressing a constitutively active STAT-5 molecule (S5CA), were resistant to Bim upregulation and ILC3-induced cell death in vitro and in vivo (Fig. 3J-L and fig. S11H-I). Taken together these data indicate that MHCII⁺ ILC3s mediate a negative selection process through antigen-presentation and withdrawal of IL-2 from the local milieu, resulting in deletion of activated commensal bacteria-specific T cells.

Inflammatory CD4⁺ T cell responses against commensal bacteria are causally associated with the pathogenesis of IBD (7-11). Furthermore, inflammatory T cells derived from Crohn's disease patients also exhibit reduced Bim-mediated cell death and cytokine-withdrawal-mediated apoptosis (38, 39). Therefore, we next examined whether ILC3-intrinsic MHCII may be dysregulated in the context of human IBD. Using a previously defined gating strategy (40), all ILC subsets could be identified in intestinal biopsies of pediatric Crohn's disease patients (Fig. 4A), including CD127⁺ c-kit⁺ ST2L^- ILC3s, that expressed NKp44 and RORγt (fig. S12A-B). Human ILC3s expressed MHCII (HLA-DR) in intestinal biopsies from non-IBD controls, whereas MHCII expression was largely absent on other ILC subsets (Fig. 4B). Although no alterations in the total frequency of ILC3s were observed between patient cohorts (fig. S12C), MHCII expression was significantly reduced on ILC3s (Fig. 4C-E), but not CD4⁺ T cells or professional antigen-presenting cells (fig. S12D-E), from pediatric Crohn's disease patients in comparison to non-IBD controls. Moreover, we observed an inverse correlation between MHCII levels on ILC3s and frequencies of effector T helper 17 (Th17) cells (Fig. 4F), and circulating commensal bacteria-specific immunoglobulin (Ig)G titers (Fig. 4G), in pediatric Crohn's disease patients. Taken together, these data indicate alterations in MHCII on human ILC3s are associated with elevated commensal bacteria-specific inflammatory responses.

The mammalian gastrointestinal tract is colonized with trillions of beneficial commensal bacteria that regulate host nutrient metabolism, immune cell homeostasis and protect from pathogen infection (7, 11, 15). As such, commensal bacteria are an essential component of the mammalian “superorganism” required for the host to thrive (41). It is well characterized that self-specific CD4⁺ T cells with the potential to cause pathologic inflammation in mammalian tissues are controlled through antigen-dependent thymic selection (1-5). Here, MHCII-expressing CCR6⁺ ILC3s were found to control intestinal homeostasis through induction of apoptotic cell death and deletion of activated commensal bacteria-specific T cells, a process with multiple similarities to negative selection in the thymus, which we propose to call “intestinal selection” (fig. S13). Thus, intestinal selection controls the peripheral commensal bacteria-specific CD4⁺ T cell pool in concert with other previously described tolerogenic pathways, including T_reg, production of IgA and active maintenance of intestinal barrier function (7, 15, 17, 23). Dysregulated ILC3-intrinsic MHCII in pediatric Crohn's disease patients suggests a possible role for alterations in this pathway in the onset and/or progression of human IBD. Thus, MHCII⁺ ILC3 may represent a novel therapeutic target to control pathologic CD4⁺ T cell responses in chronic human inflammatory disorders associated with dysregulated host-commensal bacteria relationships (7, 10, 15).