The Alarmin IL-33 Promotes Regulatory T Cell Function in the Intestine

Abstract

Foxp3⁺ regulatory T cells are abundant in the intestine where they prevent dysregulated inflammatory responses to self and environmental stimuli. It is now appreciated that T_reg cells acquire tissue-specific adaptations that facilitate their survival and function¹; however, key host factors controlling the T_reg response in the intestine are poorly understood. IL-1 family member IL-33 is constitutively expressed in epithelial cells at barrier sites² where it functions as an endogenous danger signal or alarmin following tissue damage³. Recent studies in humans have described high levels of IL-33 in inflamed lesions of inflammatory bowel disease (IBD) patients^4-7 suggesting a role for this cytokine in the pathogenesis of IBD. In the intestine, both protective and pathologic roles for IL-33 have been described in murine models of acute colitis^8-11 but its contribution to chronic inflammation remains ill defined. Here we show that the IL-33 receptor ST2 is preferentially expressed on colonic T_reg (cT_reg) cells, where it promotes T_reg function and adaptation to the inflammatory environment. IL-33 signaling into T cells stimulates T_reg responses in several ways. Firstly, it enhances transforming growth factor-β1 (TGF-β₁) mediated differentiation of T_reg cells and secondly, it provides a necessary signal for T_reg accumulation and maintenance in inflamed tissues. Strikingly, IL-23, a key pro-inflammatory cytokine in the pathogenesis of IBD, restrained T_reg responses through inhibition of IL-33 responsiveness. These results demonstrate a hitherto unrecognized link between an endogenous mediator of tissue damage and a major anti-inflammatory pathway, and suggest that the balance between IL-33 and IL-23 may be a key controller of intestinal immune responses.

To identify potential tissue-specific modulators of cT_reg cells, we compared the mRNA expression profiles of mesenteric lymph node (MLN) and cT_reg cells. We identified St2, the transcript coding for the IL-33 receptor¹², as one of the top differentially upregulated genes in cT_reg cells (Fig. 1a,b). Flow-cytometric analysis confirmed selective enrichment of ST2⁺ T_reg cells in the colon (Fig. 1c) and these cells expressed high levels of the activation markers KLRG1, CD103 and OX40 (Fig. 1d). Analysis of Helios expression revealed that ST2⁺ T_reg cells are a heterogeneous population containing thymus-derived T regulatory (tT_reg) cells as well as peripherally generated Helios⁻ T_reg cells (Fig. 1e)¹³. A significant proportion of intestinal Foxp3⁺ T_reg cells co-express the transcription factor GATA3^14-16 and GATA3 is known to regulate ST2 expression in Th2 cells¹⁷. Indeed, ST2 expression was largely restricted to GATA3-expressing cT_reg cells (Fig. 1e) and selective ablation of GATA3 in Foxp3-expressing cells, using Gata3^f/f-Foxp3-Cre mice¹⁵, caused a marked reduction of ST2 protein levels (Fig. 1f).

Figure 1. ST2-expressing T_reg cells are enriched in the colon.

a, Change in gene expression in cT_reg cells vs. MLN T_reg cells (n=3 per group) presented as volcano plot. b, Top differentially upregulated transcripts in cT_reg vs. MLN T_reg cells. c, ST2 protein expression on T_reg cells from indicated organs. d, Phenotypic analysis of ST2⁻ or ST2⁺ cT_reg cells. e, Expression of transcription factors in cT_reg cells. f, Representative histograms gated on cT_reg cells from control or Gata3^f/f-Foxp3-Cre mice.

Given that ST2⁺ T_reg cells are prominent in the colon we postulated that IL-33 may modulate in vitro-induced T_reg-cell (iT_reg) differentiation. To test this, we sort-purified naïve CD4⁺ T cells from Foxp3^gfp reporter mice and activated them in the presence of TGF-β₁. Notably, both Gata3 and St2 expression were induced under iT_reg-differentiation conditions (Extended Data Fig. 1). Addition of IL-33 to iT_reg cultures significantly increased both the percent and total number of Foxp3-expressing cells but had no effect on Foxp3 expression in the absence of TGF-β₁ (Fig. 2a). The presence of IL-33 in iT_reg cultures did not affect induction of Th2 cytokines or expression of Th1 and Th17-associated transcription factors Tbx21 and Rorc (Extended Data Fig. 1), suggesting that IL-33 preferentially regulates Foxp3 expression. Thus, our data indicate that the alarmin IL-33 is a novel co-factor in TGF-β₁-mediated iT_reg generation.

Figure 2. Effects of IL-33 on iT_reg and tT_reg cells.

a, Naïve CD4⁺ T cells were cultured with anti-CD3/CD28 plus indicated cytokines and the frequencies and absolute numbers of Foxp3⁺ T cells determined 3 days later (mean ± s.e.m. of three independent experiments). b, Naïve CD4⁺ T cells were cultured for 48h with anti-CD3/CD28 plus TGF-β₁ followed by stimulation with IL-33 for 45 minutes. Blots are representative of two independent experiments. c-d, Cells were cultured and stimulated as in (b) and recruitment of GATA3 or RNA polymerase II (Pol II) to the indicated regions assessed by ChIP-qPCR. Data are from one experiment representative of two (mean ± s.d.). e, Representative plots of T_reg cells cultured with anti-CD3/CD28 plus indicated cytokines and analyzed after 3 days. Data are representative of three independent experiments. f, T_reg cells were cultured with anti-CD3/CD28 for 24h followed by stimulation with IL-33. Blots are representative of three independent experiments. g, Mixed chimaeras were generated containing WT and St2^−/− bone marrow cells. Reconstituted mice were analyzed at steady state or two weeks after infection with Helicobacter hepaticus and anti-IL-10R treatment (inflamed). Absolute numbers of WT or St2^−/− T_reg cells in steady state (n = 3) and inflamed (n = 6) hosts (mean ± s.e.m). h, Analysis of Foxp3 expression in T_reg cells from inflamed chimaeric hosts presented as geometric mean fluorescence intensity (gMFI). *P<0.05, **P<0.01 *** P<0.001 as calculated by 1way-ANOVA with Bonferroni post-test or paired Students’ t-test.

GATA3 is highly expressed in ST2⁺ T_reg cells (Fig. 1e) and IL-33 has been shown to activate GATA3 in Th2 cells¹⁸ as well as innate lymphoid cells¹⁹. Consistent with this notion, we observed serine phosphorylation of GATA3 upon acute stimulation of iT_reg cells with IL-33 (Fig. 2b). The Foxp3 locus contains putative GATA3-binding sites within its promoter and intragenic conserved noncoding sequences (CNS) 1-3¹⁴. To investigate whether IL-33 influences the binding of GATA3 to any of these elements in iT_reg cells, we performed chromatin immunoprecipitation (ChIP) followed by quantitative PCR. Acute stimulation of iT_reg cells with IL-33 induced GATA3 recruitment to the Foxp3 promoter but not CNS1, 2 or 3 (Fig. 2c). In addition, RNA polymerase II (Pol II) was recruited to the Foxp3 promoter upon IL-33 stimulation (Fig. 2d) suggesting that IL-33 directly regulates Foxp3 expression through activation and recruitment of GATA3 to the Foxp3 promoter. In Th2 cells, GATA3 has been shown to promote St2 gene expression by binding to an enhancer element located 12-kb upstream of the St2 transcription start site¹⁷. Consistent with this we detected recruitment of GATA3 to the St2 enhancer upon acute stimulation of iT_reg cells with IL-33 and this correlated with RNA Pol II enrichment at the St2 promoter (Extended Data Fig. 2). Thus, in addition to its role in Foxp3 induction, IL-33 also promoted its own receptor expression in iT_reg cells through direct transcriptional regulation of the St2 locus providing an amplification loop for further enhancement of iT_reg-cell differentiation.

Next we focused on thymus-derived T_reg cells, which constitute a significant proportion of ST2⁺ cT_reg cells (Fig. 1e). In line with published reports^20,21 administration of recombinant IL-33 led to a significant increase in the frequency and total number of splenic T_reg cells (Extended Data Fig. 3a,b) and these IL-33-elicted T_reg cells expressed higher levels of Foxp3 and ST2 (Extended Data Fig. 3c,d). Further analysis of proliferation marker Ki67 showed IL-33 induced proliferation in splenic T_reg cells but not T effector cells (Extended Data Fig. 3e). To examine if IL-33 acts directly on T_reg cells we injected IL-33 into chimeric mice containing a mixture of WT and St2^−/− hematopoietic cells. In this setting, the proliferative capacity of St2^−/− T_reg cells was significantly impaired (Extended Data Fig. 3f), suggesting that IL-33 acts directly on tT_reg cells to promote their proliferation and accumulation in vivo. This is further supported by the finding that sort-purified splenic T_reg cells cultured in the presence of IL-33 expressed higher levels of ST2, showed a more activated phenotype and expressed increased amounts of Foxp3 protein (Fig. 2e). In addition, acute stimulation of TCR-activated splenic T_reg cells with IL-33 induced serine phosphorylation of GATA3 (Fig. 2f) further demonstrating that IL-33 acts directly on tT_reg cells.

To assess the impact of IL-33 on the T_reg response during intestinal inflammation we induced chronic colitis by infection with Helicobacter hepaticus and administration of an IL-10R blocking antibody²² (Extended Data Fig. 4a). We detected an increase in IL-33 protein levels in colon explant cultures and its expression kinetics mirrored that of IL-23, which is essential for the development of intestinal inflammation in this model (Extended Data Fig. 4b). Consistent with its pattern of expression, IL-33 protein levels were elevated in colonic intestinal epithelial cells isolated from the inflamed gut (Extended Data Fig. 4c-d). Interestingly, the onset of intestinal pathology correlated with a marked increase of soluble ST2 (sST2), which is produced primarily by colonic stromal cells (Extended Data Fig. 4b, e-f). sST2 is thought to limit IL-33 bioavailability by acting as a decoy receptor²³ and is increased in patients with active IBD^6,24 suggesting that the chronic inflammatory tissue environment may antagonize IL-33 activity. Despite high levels of sST2, analysis of chimeric mice showed that accumulation of St2^−/− T_reg cells in the colon but not the spleen was significantly impaired during the peak of intestinal inflammation (Fig. 2g). In addition, colonic St2^−/− T_reg cells expressed lower levels of Foxp3 protein on a per cell basis as compared to their WT counterparts (Fig. 2h). Together, these observations indicate that the alarmin IL-33 acts in a cell-intrinsic manner to promote the tissue-specific accumulation and stability of the T_reg phenotype in the intestine under inflammatory conditions. Furthermore, high levels of sST2 during chronic intestinal inflammation may represent a mechanism to further perpetuate pathogenic responses by limiting IL-33-driven T_reg accumulation.

We next sought to compare the suppressive capacity of WT and St2^−/− T_reg cells. St2^−/− T_reg cells inhibited T cell proliferation to the same extent as WT T_reg cells in vitro (Extended Data Fig. 5) and addition of IL-33 did not enhance WT T_reg suppressor function. We then tested the ability of St2^−/− T_reg to protect from colitis induced by adoptive transfer of naïve CD4⁺ T cells. Interestingly, ST2 was highly expressed on WT T_reg cells upon T cell transfer pointing toward a potential role of ST2 in modulating T_reg function in this model (Extended Data Fig. 6a). Indeed, St2^−/− T_reg cells were significantly impaired in their ability to prevent colonic inflammation and cellular infiltration (Fig. 3a,b), demonstrating that IL-33 signaling into T_reg cells is important for their suppressive function in vivo. Analysis of WT or St2^−/− T_reg cells two weeks post transfer, prior to the onset of intestinal pathology, showed similar proliferative capacity and Foxp3 expression between groups (Fig. 3c). The ratio of T effector cells (CD45.1⁺ RB^hi progeny) to T_reg cells (CD45.1⁻Foxp3⁺ T_reg progeny) was also similar (Fig. 3d) suggesting that IL-33 signaling in T_reg cells is dispensable for their ability to expand and index with effector cells in the lymphopenic host at this time point. By contrast, analysis at 8 weeks post-transfer showed that progeny of St2^−/− T_reg cells contained a significantly lower proportion of Foxp3⁺ cells and expressed significantly less Foxp3 on a per cell basis suggesting they had lost Foxp3 expression (Fig. 3e,h). Under these circumstances the ratio of T effector/T_reg cells and the total number of T effector cells (CD45.1⁺ RB^hi progeny) was markedly increased in recipients of St2^−/− T_reg cells (Fig. 3f,g). Importantly, ST2-deficient T_reg cells did not themselves acquire the capacity to produce inflammatory cytokines (Extended Data Fig. 6c). Perturbations of Foxp3 expression have been shown to affect T_reg cell function^25-27 and our data indicate that IL-33 signaling in T_reg cells contributes to the maintenance of Foxp3 expression under inflammatory stress, enabling T_reg cells to compete in the inflammatory niche and control the intestinal effector T cell response.

Figure 3. IL-33 promotes T_reg cell stability and function in vivo.

a, C57BL/6 Rag1^−/− mice were injected with CD45.1⁺ naïve T cells alone (RB^hi, n = 4) or in combination with WT (n = 4) or St2^−/− (n = 6) CD45.1⁻ T_reg cells. Mice were sacrificed after 6-8 weeks post transfer and colitis scores are shown. b, Absolute numbers of colon lamina propria (LP) cells from mice in (a). c, C57BL/6 Rag1^−/− mice were injected as in (a) and sacrificed at 2 weeks post injection. Representative plots are gated on colonic T_reg cell progeny (CD45.1⁻). d, Ratio of RB^hi T cell progeny (CD45.1⁺) to WT or St2^−/− Foxp3⁺ T_reg cell progeny (CD45.1⁻) in the colon (n = 5 per group) from mice in (c), (mean ± s.e.m). e, C57BL/6 Rag1^−/− mice were injected as in (a) and sacrificed at 8 weeks post injection. Representative plots are gated on colonic T_reg cell progeny (CD45.1⁻). f, Ratio of RB^hi T cell progeny (CD45.1⁺) to WT or St2^−/− T_reg cell progeny (CD45.1⁻) in the colon from mice in (e), (mean ± s.e.m). g, Absolute numbers of RB^hi T cell progeny (CD45.1⁺) in the colon from mice in (e), (mean ± s.e.m). h, Analysis of Foxp3 expression in colonic Foxp3⁺CD45.1⁻ T_reg cells presented as gMFI (mean ± s.e.m).

Results are representative of two independent experiments. *P<0.05 **P<0.01 *** P<0.001 **** P<0.0001 as calculated by 1way-ANOVA with Bonferroni post-test or Students’ t-test.

Next we sought to integrate our observations with existing pro-inflammatory pathways. IL-23 is a pivotal mediator of intestinal inflammation and polymorphisms in the IL23R locus are associated with increased susceptibility to IBD in humans²⁸. We previously showed that IL-23 promotes intestinal inflammation in part through inhibition of iT_reg cell differentiation^29,30. However, the mechanism by which IL-23 blocked iT_reg generation remained undefined. Interestingly, whole transcriptome analysis of IL-23 target genes in colonic effector CD4⁺ T cells revealed that IL-23 inhibits expression of Gata3 and St2 (Extended Data Fig. 7). Based on this observation we hypothesized that IL-23 might limit T cell responsiveness to IL-33. Indeed, the co-factor activity of IL-33 on TGF-β₁-mediated Foxp3 induction in vitro was completely abrogated in the presence of IL-23 (Fig. 4a). Notably, addition of IL-23 prevented induction of Gata3 and St2 mRNA under iT_reg differentiation conditions (Fig. 4b) resulting in reduced ST2 protein expression (Fig. 4c). Consequently, acute stimulation of IL-23-exposed iT_reg cells with IL-33 did not lead to recruitment of GATA3 to the ST2 enhancer (Extended Data Fig. 8). We observed a similar role for IL-23 in limiting T_reg ST2 expression during bacterially driven intestinal inflammation (Extended Data Fig. 9). Collectively our data indicate that IL-23 inhibits iT_reg differentiation by regulating T cell responsiveness to IL-33.

Figure 4. IL-23 inhibits the effects of IL-33 on T_reg cells.

a, Naïve CD4⁺ T cells were cultured with anti-CD3/CD28 plus TGF-β₁ as well as indicated cytokines and the frequencies of Foxp3⁺ T cells determined after 3 days later (mean ± s.e.m. of three independent experiments). b, Naïve CD4⁺ T cells were cultured with anti-CD3/CD28 plus indicated cytokines for 48h. Data are from one experiment representative of two (mean ± s.d.). c, Naïve CD4⁺ T cells were cultured as indicated. Representative blots of two independent experiments are shown. d, C57BL/6 Il23a^−/−Rag1^−/− mice were injected with CD45RB^hi WT (n = 9) or St2^−/− (n = 10) T cells. Mice were sacrificed after 6-8 weeks post transfer and frequencies of Foxp3⁺CD4⁺ T cells in colon are shown. e, Colitis scores for mice in (d). f, Expression of indicated cytokines by colonic CD4⁺ T cells from mice in (d). g, T_reg cells were cultured with anti-CD3/CD28 for 24h followed by stimulation with IL-33 in the presence or absence of IL-23. Representative blots of two independent experiments are shown. h, T_reg cells were cultured in the presence of anti-CD3/CD28 for 24h and mRNA expression of indicated genes measured following stimulation with IL-33 for 45 minutes in the presence or absence of IL-23 (mean ± s.e.m. of three independent experiments). i, T_reg cells were cultured with anti-CD3/CD28 for 24h and representative blots of two independent experiments are shown. *P<0.05, **P<0.01 *** P<0.001 as calculated by 1way-ANOVA with Bonferroni post-test or Students’ t-test.

We previously showed that IL-23 restrains T_reg cells in vivo because naïve T cell transfer into Il23a^−/−Rag1^−/− recipients resulted in increased T_reg differentiation²⁹. Therefore, we hypothesized that enhanced responsiveness to IL-33 may contribute to increased iT_reg differentiation in Il23a^−/−Rag1^−/− hosts. To test this we transferred WT or St2^−/− naïve T cells into Il23a^−/−Rag1^−/− hosts and monitored iT_reg generation. Indeed, ST2-deficient T cells were significantly impaired in their ability to differentiate into T_reg cells (Fig. 4d) and this correlated with a significant increase in intestinal pathology (Fig. 4e). Importantly, ST2-deficiency had minor effects on in vivo differentiation of Th1, Th2 or Th17 cells (Fig. 4f) further supporting the notion that the increased colitogenic potential of St2^−/− CD4⁺ T cells is a consequence of deficient iT_reg differentiation rather than dysregulated effector T cell responses. In summary, our data strongly suggest that IL-33 is a major factor responsible for driving iT_reg differentiation in the absence of IL-23.

Finally, we investigated whether IL-23 can interfere with ST2 signaling in T_reg cells. Interestingly, sort-purified ST2⁺ T_reg cells from the colon expressed detectable levels of Il23r (Extended Data Fig. 10). Indeed, exposure of TCR-activated tT_reg cells to IL-23 completely abolished IL-33-mediated GATA3 phosphorylation (Fig. 4g). Furthermore, IL-33 preferentially induced genes co-regulated by Foxp3 and GATA3¹⁶ and this was completely abrogated in the presence of IL-23 (Fig. 4h). Addition of a specific inhibitor of STAT3, the main transcription factor downstream of IL-23 signaling, reversed this inhibitory effect of IL-23 (Fig. 4i). Together our data suggest that IL-23 inhibits ST2 signal transduction and expression of a distinct set of GATA3 regulated genes in tT_reg cells.

Our results identify a new function for IL-33 as an important link between inflammation-driven tissue damage and the local intestinal T_reg cell response. We show that colonic T_reg cells are poised to respond to the release of IL-33 upon tissue damage through selective expression of ST2 and that signaling through this pathway plays an essential role in their capacity to adapt to the inflammatory tissue environment and restrain intestinal inflammation. The ability of IL-33 to amplify regulatory networks in response to tissue injury may represent a more general mechanism by which alarmins limit immune-mediated damage to self at barrier tissues. Strikingly, IL-23 limits this regulatory mechanism through inhibition of T_reg cell responsiveness to IL-33 suggesting that the balance between IL-23 and IL-33 may be a major determinant of the outcome of intestinal immune responses.

Method Summary

A description of the in vitro assays and details of in vivo models can be found in the full Methods.

Materials and Methods

Mice

Wild-type C57BL/6, congenic B6.SJL-Cd45.1, C57BL/6 Il23r^−/−, C57BL/6 Rag1^−/−, C57BL/6 Il23a^−/−Rag1^−/− and Foxp3^gfp reporter mice^31,32 were bred and maintained under specific pathogen-free conditions in accredited animal facilities at the University of Oxford. Where indicated, mice were i.p. injected with recombinant IL-33 (1 μg / injection; Biolegend) for 5 consecutive days and sacrificed 24 h after the last injection. Gata3^f/f-Foxp3Cre mice were kept at the NIH and experiments were performed at the NIH. Spleen, MLN and bone marrow from C57BL/6 St2^−/− mice were obtained from Padraic Fallon (Trinity College Dublin, Ireland) or Max Loehning (Charité – University Medicine Berlin, Germany). All procedures were conducted in accordance with the UK Scientific Procedures Act of 1986. Mice were negative for Helicobacter spp. and other known intestinal pathogens, were age and sex-matched and more than 6 weeks old when first used. Both female and male mice were used in experiments. Wherever possible preliminary experiments were performed to determine requirements for sample size, taking into account resources available and ethical, reductionist animal use. Generally, each mouse of the different experimental groups is reported. Exclusions criteria such as, inadequate staining or low cell yield due to technical problems, were pre-determined. Animals were assigned randomly to experimental groups. Each cage contained animals of all the different experimental groups.

Generation of mixed bone marrow chimaeras

Bone marrow isolated from WT (CD45.2⁺), Il23r^−/− or St2^−/− mice was mixed at a 1:1 ratio with bone marrow taken from B6.SJL-Cd45.1 mice and injected intravenously into gamma-irradiated (5.5 Gy, 550 rad) C57BL/6 Rag1^−/− recipients and chimaeras were used in experiments >8 weeks after injection.

T cell transfer colitis

For naïve T cell transfer colitis 4×10⁵ CD4⁺CD25⁻CD45RB^hi T cells were injected i.p. into Rag1^−/− or Il23a^−/−Rag1^−/− recipients. In co-transfer experiments 2×10⁵ CD4⁺CD25⁻CD45RB^hi T cells from each source were mixed and injected i.p. into Rag1^−/− hosts. For T_reg mediated protection from colitis 4×10⁵ CD4⁺CD25⁻CD45RB^hi T cells and 2×10⁵ CD4⁺CD25⁺ T_reg cells were mixed and injected i.p. into Rag1^−/− hosts³³. Mice were sacrificed at indicated time points or sacrificed when weight loss approached 20% of the original body weight at the start of the experiment.

Induction of Helicobacter hepaticus/anti-IL-10R colitis

Mice were fed 1×10⁸ cfu Helicobacter hepaticus by oral gavage with a 22G curved needle on day 0, day 1 and day 2 of the experiment. In addition mice received 1mg of an anti-IL-10R blocking antibody by i.p. injection once weekly starting at the day of Helicobacter hepaticus infection.

Histological assessment of intestinal inflammation

Proximal, mid, and distal colon samples were fixed in buffered 10% formalin solution. Paraffin embedded sections were cut (5 mm) and stained with hematoxylin and eosin, and inflammation was scored in a blinded fashion using a previously described scoring system³⁴.

Isolation of leukocyte subpopulations and flow cytometry

Cell suspensions from spleen, MLN, and the lamina propria were prepared as described previously³⁵ and first incubated with anti-CD16/CD32 (eBioscience) to prevent nonspecific binding. Single cell suspensions were stained with antibodies against CD4, CD25, TCR-β, CD45.1, CD45.2, CD103, OX40, IL-17A, IFN-γ, IL-13, Foxp3, GATA3, Helios, (all from eBioscience), RORγt, Ki67, (BD Biosciences) and CD45RB, KLRG1 (all from Biolegend) and anti-ST2 conjugated to biotin (mdBioproducts). Intracellular staining was performed as follows: cells were restimulated for 4 h as previously described³⁶, washed and incubated with anti-CD16/CD32. Cells were washed and stained for surface markers indicated above and fixed in eBioscience Fix/Perm buffer, followed by permeabilization in eBioscience permeabilization buffer for 1 h in the presence of antibodies. Cells were acquired with a BD LSRII and analysis was performed with FlowJo (Tree Star) software.

T_reg cell cultures

CD25⁻CD62L⁺CD44^lo GFP⁻ naïve CD4⁺ T cells were sort-purified from Foxp3^gfp reporter mice. Cells were plated at 2×10⁵ cells/well in flat bottomed 96 well plates coated with anti-CD3 (5 μg/ml) in RPMI (Invitrogen) containing 10% FCS, 2 mM l-glutamine, 0.05 mM 2-mercaptoethanol, and 100 U/ml each of penicillin and streptomycin (complete media). Antibodies and cytokines were added at the following concentrations, anti-CD28 (2 μg/ml), TGF-β1 (R&D, 500pg/ml), IL-2 (Peprotech, 1U/ml), IL-23 (R&D, 20ng/ml), IL-33 (Biolegend, 1 ng/ml). Cells were harvested at indicated times, lysed for use for qPCR, western blot or analyzed by flow cytometry. GFP⁺CD4⁺ T_reg cells were sort-purified from Foxp3^gfp reporter mice and cells were plated at 2×10⁵ cells/well in flat bottomed 96 well plates coated with anti-CD3 (5 μg/ml) in complete media and anti-CD28 (2 μg/ml), IL-2 (Peprotech, 1U/ml) or IL-33 (Biolegend, 1 ng/ml) were added at the start of the culture. Cells were analyzed after 3 days by flow cytometry. For acute stimulation, 2×10⁵ cells/well were plated in 96 well plates coated with anti-CD3 (5 μg/ml) in complete media and soluble anti-CD28 (2 μg/ml) and cultured for 24 h. Cells were then stimulated with IL-33 (10 ng/ml) for the indicated time and used for qPCR or western blot.

T_reg cell suppression assay

CD25⁺CD4⁺ T_reg cells were sorted by flow cytometry from WT and St2^−/− mice. CD25⁻ CD62L⁺CD44⁻CD4⁺ T cells (responder cells) were sorted from St2^−/− mice. Antigen presenting cells (APCs) were isolated by magnetic bead separation from splenocytes and LN cells from St2^−/− mice. T responder cells were labeled with Violet cell trace and plated together with the sorted Tregs at ratios of 1:1 and 1:3 together with irradiated APCs in the presence of anti-CD3 (1μg/ml). IL-33 was added at 30 ng/ml. After 4 days proliferation of T responder cells was measured by flow cytometry.

Colon explant cultures

Organ explants were prepared as previously described³⁶ and cultured overnight in complete media. Cytokine levels in the supernatants were determined by ELISA IL-23 and IL-33 (eBioscience) and soluble ST2 (R&D), and concentrations were normalized to the weight of the explants.

Isolation and culture of colonic stromal cells

Colonic tissue was digested and cells from the 30:40% Percoll gradient interface harvested and washed twice in PBS containing 2% BSA. Cells were plated at 10×10⁶ cells/ml in complete DMEM containing 40μg/ml Gentamycin (Sigma). After 24-48 hours, non-adherent cells were removed by vigorous washing, and media replaced. Adherent intestinal stromal cells were cultured for 10 days until confluent. Cells were then plated at 1×10⁶ cells/ml for 48 hours and the concentration of sST2 in supernatants was determined by ELISA (R&D Systems).

qPCR and microarray preparation

RNA was extracted according to the manufacturers’ protocol (RNeasy, Qiagen) and cDNA synthesis was performed using Superscript III reverse transcription and Oligo dT primers (both from Invitrogen). qPCR reactions were performed using TaqMan Gene Expression Assays and normalized to HPRT (all from Applied Biosystems). Alternatively, qPCR reactions were performed using SYBR green PCR SensiMix (Quantace) with primers from³⁷. For soluble ST2 the following primers were used: sST2 forward 5′-tcgaaatgaaagttccagca -3′ and sST2 reverse 5′-tgtgtgagggacactccttac -3′ Samples were assayed in duplicate on a Bio-Rad CFX96 RT-qPCR machine and differences were calculated using the 2^ΔC(t) method.

For microarray gene expression analysis RNA was extracted using the RNAqueous-Micro total RNA Isolation Kit or RiboPure RNA Purification Kit (both Ambion). For RNA amplification and labelling the TargetAmp2-Round Biotin-aRNA Amplification Kit 3.0 (Epicentre, Illumina) or the Illumina TotalPrep RNA Amplification Kit (Ambion) were used and RNA quality was assessed using the Agilent 2100 Bioanalyzer and only samples with RNA integrity number (RIN) values above 7 were used for further processing. Biotinylated cRNA was hybridised to MouseWG-6 v2.0 Expression BeadChip (Illumina). Hybridisations to BeadChips and data acquisition were performed by the microarray core facility at the Wellcome Trust Center for Human Genetics, Oxford. Microarray analysis was performed using GeneSpring GX12 software (Agilent). Data was normalised using 75% percentile shift normalisation algorithm and baseline transformed to the median of all samples. Statistical significance was determined using an unpaired t-test followed by Benjamini Hochberg false discovery rate multiple testing correction. P value cut-off was set to 0.05.

Total protein extracts and immunoblot analysis

Total protein extracts were prepared as described³⁸. Equal amounts of protein were resolved by SDS-PAGE and analyzed with anti-pGATA3 (ab61052, Abcam), anti-total GATA3 (L50-823, BD Pharmingen), anti-ST2 (101001B, mdBioproducts), anti-actin (A5541; Sigma) or anti-IL-33 (396118, R&D). For inhibition of STAT3 activity in vitro, STAT3 VI (10 mM, Santa Cruz) was added 30min before cytokine stimulation.

Chromatin immunoprecipitation

ChIP assays were performed as described³⁸ with antibody to GATA3 (L50-823, BD Pharmingen) or RNA polymerase II (sc-899; Santa Cruz). The immunoprecipitated DNA fragments were then analyzed by real-time PCR with SYBR Premix Ex Taq II master mix (Takara Bio) and the following primers: locus encoding St2 enhancer, 5′-gccaaccacaacagcagatggggaaa -3′ and 5′-actgagatcctgccctggcttccct -3′; locus encoding St2 promoter, 5′-tggcctccttggaaaggcttggt -3′ and 5′-agtgcaggaggggcatggagatga. Primers for analyzing binding to CNS1, CNS2, CNS3 and promoter locus of Foxp3 were from²⁵. Data were analyzed using Rotogene 6000 software (Corbett Research Ltd). Results are normalized to input DNA and presented as fold enrichment relative to unstimulated cells.

Statistical analysis

Where appropriate Student’s t-test was used. For comparison of more than two groups a one-way ANOVA followed by a Bonferroni multiple comparison test was performed. All statistical analysis was calculated in Prism (GraphPad). Differences were considered to be statistically significant when P ≤ 0.05.

Supplementary Material

1

Extended Data Figure 1: Effects of IL-33 on gene expression during iT_reg differentiation

a, Sort-purified CD25⁻CD62L⁺CD44^lo GFP⁻ naïve CD4⁺ T cells from Foxp3^gfp reporter mice were stimulated with anti-CD3/anti-CD28 in the presence of indicated cytokines. mRNA expression of indicated genes was measured at 24h, 48h and 72h and presented as fold change over time 0. Data are representative of two independent experiments and show the mean ± s.d.

2

Extended Data Figure 2: IL-33 directly regulates ST2 expression

Sort-purified CD25⁻CD62L⁺CD44^lo GFP⁻ naïve CD4⁺ T cells from Foxp3^gfp reporter mice were cultured with anti-CD3/anti-CD28 in the presence of TGF-β₁ followed by acute stimulation with IL-33 for 45 minutes. Shown are the recruitment of GATA3 or RNA polymerase II (Pol II) to the indicated regions of the gene encoding ST2 assessed by ChIP followed by qPCR. Results are normalized to those obtained with genomic DNA (input) and presented as fold enrichment relative to unstimulated cells. Data are from one experiment representative of two (error bars, s.d.).

3

Extended Data Figure 3: IL-33 acts directly on tT_reg cells to promote their proliferation and accumulation in vivo

a-e, Foxp3^gfp reporter mice were injected i.p with recombinant IL-33 (1μg/mouse/day) for 5 days. a, Frequencies of TCRβ⁺CD4⁺Foxp3⁺ splenic tT_reg cells. b, absolute numbers of TCRβ⁺CD4⁺Foxp3⁺ splenic T_reg cells. c, gMFI of Foxp3 among TCRβ⁺CD4⁺Foxp3⁺ splenic T_reg cells (n = 10 for PBS and n = 14 for IL-33). d, Frequencies of ST2⁺ cells among splenic CD44^high T cells or Foxp3⁺ T_reg cells. e, Frequencies of Ki67⁺ cells among splenic CD44^high T cells or Foxp3⁺ T_reg cells (n = 5 for PBS and n = 8 for IL-33). f, Mixed bone marrow chimaeras were generated by lethal irradiation of C57BL/6 Rag1^−/− mice followed by i.v. injection of 2.5×10⁶ WT (CD45.1⁺) and 2.5×10⁶ St2^−/− (CD45.1⁻) bone marrow cells (n = 5). Following reconstitution, mixed chimaeras were injected with recombinant IL-33 (1μg/mouse/day) for 5 days and the absolute number of Ki67⁺ among TCRβ⁺CD4⁺Foxp3⁻ or TCRβ⁺CD4⁺Foxp3⁺ cells are shown. **P<0.01 *** P<0.001 as calculated by unpaired (WT mice) or paired (chimaeric mice) Students’ t-test.

4

Extended Data Figure 4: Kinetics of IL-33 and IL-23 during Helicobacter hepaticus/anti-IL-10R colitis

C57BL/6 mice were infected by oral gavage with ~10⁸ cfu Helicobacter hepaticus on three consecutive days with concomitant i.p. injections of a blocking anti-IL-10R mAb (1mg/week) starting on the day of the first infection. a, Colitis scores for the indicated groups and b, IL-33, IL-23 and soluble ST2 protein production in colon explant cultures from Hh/aIL-10R treated mice were determined at the indicated time points (n = 4, data show the mean ± s.e.m.). c, Colonic intestinal epithelial cells (IECs) were isolated from steady state or colitic (day 10) mice. Total protein extracts from each time point were analyzed by immunoblot and biological replicates are shown. d, Densitometric quantification of immunoblot in (c). e, mRNA expression analysis of intestinal epithelial cells (IEC), lamina propria cells (LP) and stromal cells isolated from the colon at steady state or 10 days after induction of colitis. f, Colonic stromal cells were isolated at the indicated time points, cultured and passaged and spontaneous sST2 protein production determined in supernatants from 48hr cultures. Stromal cells were uniformly CD45⁻ EpCAM⁻. Data represent mean ± s.e.m. *P<0.05, **P<0.01 *** P<0.001 as calculated by 1way-ANOVA with Bonferroni post-test or Students’ t-test.

5

Extended Data Figure 5: The in vitro suppressive function of St2^−/− T_reg cells is not impaired.

CD25⁺CD4⁺ T_reg cells were sorted by flow cytometry from WT and St2^−/− mice. CD25⁻ CD62L⁺CD44⁻CD4⁺ T cells (responder cells) and antigen presenting cells (APCs) were sorted from St2^−/− mice. T responder cells were labeled with Violet cell trace and plated together with the sorted T_reg cells at ratios of 1:1 and 1:3 together with irradiated APCs in the presence of anti-CD3 (1μg/ml). IL-33 was added at 30 ng/ml. After 4 days proliferation of T_resp was measured by flow cytometry. Data are representative of two independent experiments.

6

Extended Data Figure 6: ST2 is preferentially expressed on T_reg cell progeny

a, C57BL/6 Rag1^−/− mice were injected i.p. with 4×10⁵ CD4⁺CD25⁻CD45RB^hi CD45.1⁺ naive T cells in combination with 2×10⁵ WT CD4⁺CD25⁺ CD45.1⁻ T_reg cells. Mice were sacrificed at 6-8 weeks post transfer. Representative plots of the colon gated on progeny of CD4⁺CD45RB^hi cells (CD45.1⁺) or WT T_reg cells (CD45.1⁻) cells are shown. b, Histogram overlay of Foxp3 expression by WT (solid line) or St2^−/− (dotted line) CD4⁺CD25⁺ T_reg cells prior to injection into C57BL/6 Rag1^−/− mice. c, C57BL/6 Rag1^−/− mice were injected as in (a) and frequencies of cytokine producing colonic T_reg progeny (CD45.1⁻) among Foxp3⁻ or Foxp3⁺ cells are shown. WT (n = 4) and St2^−/− (n = 6), bars represent mean ± s.e.m..

7

Extended Data Figure 7: St2 is an IL-23 target gene in intestinal CD4⁺ T cells.

C57BL/6 Rag1^−/− mice were injected i.p. with a 1:1 mixture of 2×10⁵ WT (CD45.1⁺) and 2×10⁵ Il23r^−/− (CD45.2⁺) CD4⁺CD25⁻CD45RB^hi T cells or 2×10⁵ WT (CD45.1⁺) and 2×10⁵ WT (CD45.2⁺) CD4⁺CD25⁻CD45RB^hi T cells. Mice were sacrificed upon development of clinical signs of inflammation (~8 weeks). Viable TCRβ⁺CD4⁺CD45.2⁺ WT or Il23r^−/− T cells were sort-purified from the colon and RNA extracted without further manipulation. Whole transcriptome gene expression analysis was performed using the Illumina Bead Array platform. Two-dimensional cluster analysis (hierarchical, Pearson uncentered, average linkage, unsupervised) of the 168 transcripts whose expression was significantly affected in the comparison of Il23r^−/− versus WT CD4⁺ T cells. Each row corresponds to a gene and each column to a sample (n = 6 for WT and n= 5 for Il23r^−/−). The 168 transcripts represent 147 unique genes including genes with unknown function.

8

Extended Data Figure 8: IL-23 inhibits IL-33-induced GATA3 recruitment to the ST2 locus

Sort-purified CD25⁻CD62L⁺CD44^lo GFP⁻ naïve CD4⁺ T cells from Foxp3^gfp reporter mice were cultured with anti-CD3/anti-CD28/TGF-β₁ followed by acute stimulation with IL-33 for 45 minutes in the presence or absence of IL-23. Shown are the enrichment of GATA3 or RNA polymerase II (Pol II) to the indicated regions of the gene encoding ST2 assessed by ChIP followed by qPCR. Results are normalized to those obtained with genomic DNA (input) and presented as fold recruitment relative to unstimulated cells. Data are from one experiment representative of two (error bars, s.d.)

9

Extended Data Figure 9: IL-23 is a negative regulator of ST2 expression in vivo

Mixed bone marrow chimaeras were generated by lethal irradiation of C57BL/6 Rag1^−/− mice followed by i.v. injection of 2.5×10⁶ WT (CD45.1⁺) and 2.5×10⁶ Il23r^−/− (CD45.1⁻) bone marrow cells (n = 5). Reconstituted mice were infected by oral gavage with ~10⁸ cfu Helicobacter hepaticus on three consecutive days with concomitant i.p. injections of a blocking anti-IL-10R mAb (1mg/week) starting on the day of the first infection. Mice were sacrificed at 2 weeks post infection. Frequencies of ST2⁺ and gMFI of ST2 among TCRβ⁺CD4⁺Foxp3⁺ cT_reg cells are shown. *P<0.05, *** P<0.001 as calculated by paired Students’ t-test.

10

Extended Data Figure 10: Il23r mRNA is expressed by ST2⁺Foxp3⁺ T_reg cells

mRNA expression of indicated genes in Foxp3⁻ST2⁻, Foxp3⁺ST2⁻ or Foxp3⁺ST2⁺ populations sort-purified from the steady state cLP of Foxp3^gfp reporter mice. Bars represent the mean ± s.e.m. from three independent experiments.