G9a regulates group 2 innate lymphoid cell development by repressing the group 3 innate lymphoid cell program

To study the role of G9a in ILC development, we generated mice with a hematopoietic cell–specific deletion of G9a (termed Vav.G9a^−/− mice) by crossing mice with a floxed allele of G9a (G9a^fl/fl mice; Lehnertz et al., 2010) with mice expressing Cre recombinase under the control of the Vav promoter. We analyzed the generation of iILC2s in the BM during steady-state hematopoiesis by flow cytometric analysis of Lin⁻ c-kit^low Sca-1^high CD25⁺ CD127⁺ cells (Halim et al., 2012b; Hoyler et al., 2012). Despite having a significant expansion of Lin⁻ c-kit^low Sca-1^high cells, we found a significant reduction in the frequency of iILC2s in the BM of Vav.G9a^−/− mice compared with control G9a^fl/fl mice (Fig. 1 A). During development from a CLP to iILC2, receptors for the cytokines IL-33 (ST2 and Il1rl1) and IL-25 (IL-17RB) are up-regulated (Halim et al., 2012b; Wong et al., 2012; Constantinides et al., 2014). We found that iILC2s from G9a^fl/fl mice expressed high levels of ST2, whereas only ∼50% of iILC2s from Vav.G9a^−/− mice expressed ST2 (Fig. 1 A), resulting in a 75% reduction in the number of Lin⁻ c-kit^low Sca-1^high CD25⁺ CD127⁺ ST2⁺ iILC2s in the absence of G9a (Fig. 1 B). In contrast to ST2, IL-17RB was expressed at equivalent levels by iILC2s from both G9a^fl/fl and Vav.G9a^−/− mice (Fig. 1 C). In addition, BM iILC2s have been found to express the transcription factor GATA3 (Hoyler et al., 2012). We found that nearly all G9a^fl/fl Lin⁻ c-kit^low Sca-1^high CD25⁺ CD127⁺ cells expressed GATA3, but in the absence of G9a, only the ST2⁺ cells expressed GATA3 (Fig. 1 D). Systemic administration of IL-33 results in an expansion of ILC2s in the BM (Spooner et al., 2013). In response to IL-33, control G9a^fl/fl mice showed a significant increase in the number of BM iILC2s, whereas treatment with IL-33 failed to induce comparable ILC2 development in the BM of Vav.G9a^−/− mice (Fig. 1 E). Thus, G9a plays an important role in ILC2 development, potentially through the regulation of IL-33 responsiveness.

To determine whether G9a functioned in a cell-autonomous manner in regulating the development of ILC2s, we reconstituted lethally irradiated CD45.1⁺ mice with mixed BM harvested from congenic WT (CD45.1/2⁺) mice and CD45.2⁺ G9a^fl/fl or Vav.G9a^−/− mice. Although a 50:50 mix of donor BM (WT/G9a^fl/fl) yielded chimeric recipient mice with an ∼50:50 ratio of donor-derived hematopoietic cells, a 20:80 ratio of donor BM (WT/Vav.G9a^−/−) was required to achieve comparable chimerism. Furthermore, despite transplanting four times as many BM cells from Vav.G9a^−/− CD45.2⁺ mice, the G9a-deficient BM failed to give rise to iILC2s (Fig. 1, F and G). As Vav.Cre deletes early during hematopoiesis, it remains to be determined at what stage during ILC development G9a becomes important. We found that the absence of G9a does not affect the number of BM CLPs, but ChILPs and α₄β₇⁺ lymphoid progenitors (αLPs) were both reduced in the absence of G9a (Fig. 1 H), suggesting that G9a is important for differentiation of ILC progenitors downstream of the CLP.

ILC2s have been shown to be an important innate source of type 2 cytokines in the mesenteric LNs (mLNs; Neill et al., 2010), the lungs (Halim et al., 2012a), and visceral adipose tissue (VAT; Moro et al., 2010). We found that Vav.G9a^−/− mice have significantly fewer ILC2s in the mLNs, lungs, and VAT (Fig. 2, A and B) compared with G9a^fl/fl mice, but unlike in the BM, those that were present in the periphery expressed comparable levels of ST2 (Fig. 2 C). Thus, it is likely that the small number of ST2⁺ iILC2s in the BM of Vav.G9a^−/− mice that were able to express GATA3 (Fig. 1 D) populated the peripheral tissues, albeit at significantly lower levels than G9a^fl/fl mice.

Treatment of mice intranasally with the protease papain has been shown to induce ILC2-dependent IL-5 and IL-13 production, eosinophilia, and lung inflammation (Oboki et al., 2010; Halim et al., 2012a). After a short course of papain treatment, Vav.G9a^−/− mice had reduced expression of Il5 and Il13 in the lungs and significantly reduced influx of eosinophils into the bronchoalveolar lavage (BAL) fluid compared with control G9a^fl/fl mice, despite equivalent eosinophil numbers in the steady state (Fig. 3 A). In addition, after papain treatment of Vav.G9a^−/− mice on a Rag1^−/− background (Rag.Vav.G9a^−/−), we observed significantly reduced eosinophilic influx into the BAL fluid compared with Rag.G9a^fl/fl mice (Fig. 3 B). After a longer course of papain administration (Spooner et al., 2013; Halim et al., 2014), we found that Vav.G9a^−/− mice had significantly fewer ILC2s and eosinophils in the lungs than control G9a^fl/fl mice (Fig. 3 C). Thus, G9a controls ILC2 function during papain-induced lung inflammation.

We have recently shown that ILC2s facilitate the local development of a Th2 cell response in the house dust mite (HDM) model of allergic lung inflammation (Gold et al., 2014). Similar to results with papain, we found that numbers of ILC2s in the lungs as well as numbers of eosinophils in the BAL fluid of Vav.G9a^−/− mice were significantly reduced compared with G9a^fl/fl mice (Fig. 3 D). Expression of Il5 and Il13 was also reduced in the lungs of Vav.G9a^−/− mice (Fig. 3 E). Consistent with these results, we observed significantly reduced goblet cell hyperplasia in the lungs of HDM-treated Vav.G9a^−/− mice (Fig. 3, F and G). As numbers of ILC2s are reduced in the absence of G9a, we also wanted to determine whether the remaining ILC2s are functionally competent. To test this, we sorted equivalent numbers of mature ILC2s from the lungs of naive G9a^fl/fl and Vav.G9a^−/− mice and stimulated them with IL-33 and thymic stromal lymphopoietin (Halim et al., 2012a). We found a significant reduction in IL-5 and IL-13 production (Fig. 3 H), highlighting a role for G9a in promoting ILC2-dependent gene expression in mature ILC2s. Thus, these results demonstrate that G9a is a critical regulator of ILC2-dependent lung inflammation.

We next used genome-wide expression analysis to identify the molecular signature of G9a-deficient ILC2s. We sorted CD45⁺ Lin⁻ c-kit^low Sca-1^high CD25⁺ ST2⁺ iILC2s from the BM of G9a^fl/fl or Vav.G9a^−/− mice and expanded them in vitro for 10 d in the presence of IL-2, IL-7, and IL-25. Consistent with our in vivo findings, in vitro expanded iILC2s from Vav.G9a^−/− mice expressed lower levels of well characterized ILC2-specific genes including Il1rl1, Crlf2, Il5, Il13, Areg (amphiregulin), Gata3, and Icos (inducible T cell co-stimulator; Fig. 4 A). Over 70% of the core ILC2 genes (Robinette et al., 2015) show reduced expression in the absence of G9a (Fig. 4 B and Table S1), suggesting that G9a is a central control point in ILC2 development. Strikingly, we found that expression of ILC3-associated genes such as Il1r1, Il23r, Rorc, Il17f, Il22, and Lta was significantly higher in ILC2s from Vav.G9a^−/− mice (Fig. 4 A). Indeed, >80% of the core ILC3 genes were increased in G9a-deficient ILC2s (Fig. 4 B). These results demonstrate that G9a serves an important function in promoting ILC2 development while inhibiting the ILC3-associated gene expression program.

In the absence of G9a, we observed a reduction in the frequency of ILC2s producing IL-5, and the production of IL-17F was only seen in cultures of G9a-deficient ILC2s (Fig. 4 C). Critically, IL-17F–producing cells were distinct from IL-5–producing cells, suggesting that G9a specifically represses ILC3 differentiation in developing iILC2s. We found that ∼7.5% of the ST2⁻ fraction of the Lin⁻ c-kit^low Sca-1^high CD25⁺ cells from the BM of Vav.G9a^−/− mice were RORγt⁺, compared with <1% in the ST2⁺ fraction of either Vav.G9a^−/− or G9a^fl/fl cells (Fig. 4 D), suggesting that loss of G9a in iILC2s leads to ectopic expression of ILC3-associated genes. Next, we examined whether transfer of iILC2s from Vav.G9a^−/− mice into NOD-Scid Il2rg^−/− (NSG) mice (Halim et al., 2012b) would result in the development of both mature ILC2s and mature ILC3s. We transferred 5,000 iILC2s (CD45⁺ Lin⁻ c-kit^low Sca-1^high CD25⁺CD127⁺) isolated from the BM of G9a^fl/fl or Vav.G9a^−/− mice into NSG mice and analyzed the small intestine and lungs 3–4 wk later. We found that all of the ILCs in the small intestine of NSG mice that received iILC2s from G9a^fl/fl mice were mature ILC2s based on KLRG1 (killer cell lectin-like receptor subfamily G member 1) expression (Hoyler et al., 2012), whereas G9a-deficient iILC2s gave rise primarily to ILC3s based on RORγt expression (Fig. 4 E). We found a similar result in the lungs with G9a^fl/fl iILC2s generating mature ILC2s, and G9a-deficient iILC2s were able to generate both mature ILC2s and ILC3s despite the fact that ILC3s are normally restricted from the lungs in the steady state (Fig. 4 E; Halim et al., 2012b). Therefore, G9a is required to repress the ILC3 potential of iILC2s.

Consistent with our results, we found a threefold increase in the frequency of ILC3s in the Peyer’s patches and small intestinal lamina propria of Vav.G9a^−/− mice compared with G9a^fl/fl controls (Fig. 4 F), demonstrating that G9a is required to limit the differentiation of ILC3s. However, it remained to be determined whether G9a is also required to repress the ILC3 gene program in terminally differentiated ILC2s. To test this directly, we treated G9a^fl/fl and Vav.G9a^−/− mice with a short course of papain and isolated cells from the lungs, stimulated them ex vivo, and looked at cytokine production. We found increased production of IL-17A after stimulation of mature lung ILC2s, suggesting that repression of ILC3 cytokines in mature ILC2s requires G9a. We also saw a significant reduction in IL-13 production (Fig. 4 G), highlighting a role for G9a in promoting ILC2-dependent gene expression in mature ILC2s. It is noteworthy that unlike Gfi1-deficient mice, IL-17A–producing cells were mainly distinct from IL-13–producing cells, with only a few cells coexpressing the two cytokines. Thus, G9a is required to silence the ILC3 program in developing iILC2s as well as mature ILC2s.

Next, we cultured sorted iILC2s from the BM of WT mice in the presence of IL-2, IL-7, IL-25, and the G9a-specific methyltransferase inhibitor UNC0638 (Vedadi et al., 2011). Similar to the phenotype of iILC2s derived from Vav.G9a^−/− mice, inhibition of G9a methyltransferase activity in WT iILC2s resulted in reduced expression of Il5 and Il13, with a concomitant increase in expression of the ILC3-associated genes Rorc and Il17f (Fig. 5 A). Based on these results, we examined the levels of H3K9me2 at various ILC2- and ILC3-related gene loci in cytokine-expanded iILC2s from G9a^fl/fl and Vav.G9a^−/− mice by chromatin immunoprecipitation (ChIP). We detected high levels of G9a-dependent H3K9me2 at the Il17f and Rorc promoters in iILC2s from G9a^fl/fl mice, consistent with its lack of expression in ILC2s (Fig. 5 B). Despite reduced Il13 and Il5 expression in G9a-deficient ILC2s, we found no difference in the levels of acetylated H3K9/14 (H3K9/14Ac), an activating histone modification, at the promoters for Il13 or Il5 (Fig. 5 C); however, we did see reduced levels at the Gata3 promoter (Fig. 5 C), suggesting that ILC2 lineage identity is reduced in the absence of G9a. In addition, consistent with the heightened production of IL-17F by G9a-deficient iILC2s, we found increased levels of H3K9/14Ac at the Il17f and Rorc promoters (Fig. 5 C). Despite the importance of G9a-mediated H3K9me2 in the repression of ILC3 genes in ILC2s, we did not see any differences in G9a expression levels in ILC2s versus ILC3s in WT mice (unpublished data). Thus, G9a-dependent H3K9me2 is associated with repression of the Il17f and Rorc loci in ILC2s, and the absence of G9a leads to increased epigenetic activation and expression of ILC3-associated genes.

These results define the first epigenetic regulatory mechanism that controls ILC development, differentiation, and function. G9a-dependent H3K9me2 is required for silencing of ILC3-associated genes in iILC2s and suppression of ILC3 development. Further, G9a is required to promote expression of ILC2-associated genes in mature ILC2s. Interestingly, a recent study demonstrated that ILCs from mice with a deletion in the transcription factor Gfi1 display decreased frequencies of ILC2s and dysregulated expression of ILC3-associated genes (Spooner et al., 2013), a phenotype similar to G9a-deficient ILC2s. As Gfi1 has been shown to directly interact with G9a (Duan et al., 2005), our results suggest that G9a–Gfi1 interactions may be critically important in the epigenetic regulation of ILC development. Collectively, the results presented here identify an additional level of regulation of ILC2/ILC3 development and suggest that modulation of G9a activity may provide a novel therapeutic strategy to treat diseases associated with dysregulated ILC responses.

Vav.G9a^−/− mice and control G9a^fl/fl on a C57BL/6 background, which have been previously described (Lehnertz et al., 2010), B6.SJL-Ptprc^aPepc^b/BoyJ (CD45.1), and NSG mice were bred and maintained at the University of British Columbia Biomedical Research Centre. Rag.Vav.G9a^−/− mice and Rag.G9a^fl/fl mice were generated by crossing Vav.G9a^−/− mice and control G9a^fl/fl mice with Rag1^−/− mice. All experiments were performed using sex-matched littermates. All mice were maintained in a specific pathogen–free environment. All experiments were performed at the University of British Columbia according to institutional and Canadian Council on Animal Care guidelines approved by the University of British Columbia Committee on Animal Care.

G9a^fl/fl and Vav.G9a^−/− mice were given intraperitoneal injections of 500 ng IL-33 (eBioscience) on days 0, 1, 2, and 3 as previously described (Price et al., 2010). Mice were euthanized 24 h after the final injection, and BM was harvested and analyzed.

CD45.1 recipient mice were lethally irradiated with 10 Gy with an X-Rad (Precision X-Ray) and were reconstituted by intravenous injection of BM from CD45.1/2 mice mixed with either G9a^fl/fl or Vav.G9a^−/− BM cells (2.5 × 10⁶ cells) to achieve an ∼50:50 chimera after stable reconstitution. At least 8 wk after transplant, mice were euthanized, and BM and spleens were collected for analysis.

For the short course of papain, G9a^fl/fl and Vav.G9a^−/− mice were anesthetized, followed by intranasal instillation of 10 µg papain in 40 µl PBS on days 0, 1, and 2. Mice were sacrificed on day 3, and BAL fluid and lungs were collected and analyzed. Intracellular cytokines were detected after 4 h of stimulation with cell stimulation cocktail (plus protein transport inhibitors; eBioscience) using an intracellular fixation and permeabilization buffer set (eBioscience) according to the manufacturer’s protocol with the antibodies IL-13 (eBio13A) and IL-17A (eBio17B7). Live cells were identified using fixable viability dye (eFluor 506; eBioscience). For the long course of papain, G9a^fl/fl and Vav.G9a^−/− mice were anesthetized and then sensitized with papain (10 µg in 40 µl) by intranasal instillation on days 0, 3, and 6. Mice were rechallenged on day 14, and BAL fluid and lungs were collected on day 15. HDM (Dermatophagoides pteronyssinus)-mediated allergic inflammation was induced in G9a^fl/fl and Vav.G9a^−/− mice by intranasal instillation with 100 µg HDM antigen (Greer Laboratories) on days 0, 1, and 2 and 25 µg on days 13, 14, 15, 16, and 17. Mice were sacrificed 24 h after the final challenge. BAL cells were enumerated and differentiated by flow cytometry with antibodies to CD3e (145-2C11), CD11c (N418), CD45 (I3/2), B220 (RA3-6B2), Siglec-F (E50-2440), and Ly6B (7/4).

Lung lobes were fixed in 10% buffered formalin solution and embedded into paraffin blocks. 5-µm sections were stained with periodic acid-Schiff (PAS). Airways were assessed for PAS staining as an indication of mucus hyperproduction. PAS⁺ cells in the airways in each lung section were quantified.

iILC2s (live CD45⁺ Lin⁻ cKit⁻ Sca1⁺ CD25⁺ ST2⁺) were sorted from the BM of naive G9a^fl/fl and Vav.G9a^−/− mice. Approximately 5,000 cells were cultured in 200 µl IMDM containing 10% FCS, penicillin-streptomycin, and 150 µM monothioglycerol with 50 ng/ml rIL-2, 10 ng/ml rIL-7, and 100 ng/ml rIL-25. Cells were split and diluted 1:2 with fresh cytokine-containing media on day 3 and every other day thereafter. When used, the G9a inhibitor UNC0638 was added to cytokine-containing media at 0.5 µM. Live cells were counted by trypan blue exclusion. Intracellular cytokines were detected after 4 h of stimulation with 50 ng/ml PMA and 750 µg/ml ionomycin in the presence of 10 µg/ml brefeldin A using an intracellular fixation and permeabilization buffer set (eBioscience) according to the manufacturer’s protocol with the antibodies IL-5 (TRFK5) and IL-17F (eBio18F10). Live cells were identified using eFluor 506 fixable viability dye.

ILC2s were sorted from BM and expanded in IL-2, IL-7, and IL-25 for 10 d. Cells were harvested and total RNA was extracted using an RNeasy Mini kit according to manufacturer’s instructions (QIAGEN). Approximately 500 ng RNA was prepared with an mRNA kit (Truseq Stranded; Illumina) and sequenced on a MiSeq paired-end run (75 × 75; v3; Illumina). Samples were aligned to the 10-mm transcript reference using TopHat2, and differential expression was assessed using Cufflinks. The heat map of ILC2 and ILC3 core genes expressed in G9a^fl/fl and Vav.G9a^−/− iILC2s were generated using the open source R and VisR platforms (Younesy et al., 2015). Data are available through NCBI BioProject database accession no. PRJNA313104.

RNA was isolated from expanded BM ILC2s using an RNeasy Mini kit according to the manufacturer’s instructions (QIAGEN). Lung tissue was homogenized using a TissueLyser (QIAGEN). RNA was extracted using TRIzol (Invitrogen) and reverse transcribed with a high-capacity cDNA synthesis kit (Thermo Fisher Scientific), and gene expression was analyzed by quantitative RT-PCR with SYBR green chemistry (KAPA Biosystems) on an RT-PCR system (7900 Fast; Applied Biosystems). The primers used were obtained from QIAGEN or synthesized de novo: Il5 forward, 5′-GATGAGGCTTCCTGTCCCTACTC-3′ and reverse, 5′-TCGCCACACTTCTCTTTTTGG-3′; Il13 forward, 5′-CCTGGCTCTTGCTTGCCTT-3′ and reverse, 5′-GGTCTTGTGTGATGTTGCTCA-3′; Il17f forward, 5′-TGCTACTGTTGATGTTGGGAC-3′ and reverse, 5′-AATGCCCTGGTTTTGGTTGAA-3′; Rorc forward, 5′-TCCACTACGGGGTTATCACCT-3′ and reverse, 5′-AGTAGGCCACATTACACTGCT-3′; and Actb forward, 5′-GGCTGTATTCCCCTCCATCG-3′ and reverse, 5′-CCAGTTGGTAACAATGCCATGT-3′.

ILC2s were sorted from BM and expanded in IL-2, IL-7, and IL-25 for 14 d. ChIP for H3K9me2 and H3K9/14Ac was conducted as previously described (Dong et al., 2008) using 1 µg of antibody specific to H3K9me2 (ab1220; Abcam) or H3K9/14Ac (06-599; EMD Millipore). After washing, elution, and reversion of cross-links, the DNA was isolated and used in quantitative PCR reactions on an RT-PCR system (ABI 7900 Fast). Data are presented as the ratio of immunoprecipitated to input cycle threshold (C_T) values. The primers used for PCR analysis were: Il5 promoter forward, 5′-AAGTCTAGCTACCGCCAATA-3′ and reverse, 5′-AGCAAAGGTGAGTTCAATCT-3′; Il13 promoter forward, 5′-GGGCAGGTGAGTATCAGTCTA-3′ and reverse, 5′-GTCTATATCCCTCCCACTCGT-3′; Il17f promoter forward, 5′-GGGAATCAAAGGGGGACCCTAA-3′ and reverse, 5′-AAAGCAGAACCCACACGCAGAG-3′; Gata3 promoter forward, 5′-CGTATATGATGAGTCTTCTCTGGGACTTG-3′ and reverse, 5′-AAATCTCAGAACACACACATTTCCAGG-3′; and Rorc promotor forward, 5′-ATTCCATGAGGGCTTGCCT-3′ and reverse, 5′-ACCTGTCATCAGCCTCCCATA-3′.

BM cells were isolated by flushing femurs and tibias with a 26-G needle though a 70-µm strainer. Lungs were minced and digested in 200 U/ml collagenase IV (Sigma-Aldrich) for 45 min at 37°C and passed through a 70-µm cell strainer. RBCs were lysed, and leukocytes were enriched using a Percoll (Sigma-Aldrich) separation. mLNs and Peyer’s patches were passed through a 70-µm cell strainer to form single-cell suspensions. Small intestine lamina propria cells were isolated by digestion in 0.5 mg/ml collagenase/dispase for 30 min after removal of intestinal epithelial cells. Leukocytes were enriched using a standard Percoll separation method.

Staining and antibody dilutions were prepared in PBS containing 2% FCS, 2 mM EDTA, and 0.05% sodium azide. Samples were first blocked in buffer containing 5 µg/ml anti-CD16/32 (2.4G2) and 2% rat serum to block nonspecific antibody binding. Viable cells were identified using the eFluor 506 fixable viability dye. The following antibodies were used to identify ILCs: pan-CD45 (I3/2), lineage markers (CD3e [145-2C11 and KT3], CD5 [53-7.3], CD4 [GK1.5; not used when staining for ILC3s], CD8 [53.67], CD11c [N418], CD11b [M1/70], CD19 [1D3], NK1.1 [PK136], Gr1 [RB6-8C5], and Ter119 [TER-119]), CD25 (PC61.5), CD127 (SB/199), Sca1 (D7), cKit (ACK2), ST2 (RMST2-2), CD90.2 (53-2.1), KLRG1 (2F1), ICOS (7E.17G9), GATA3 (TWAJ), and RORγt (B2D). Additional antibodies used include IL-17RB (MUNC33), CD45.1 (A20), and CD45.2 (104).

NSG-recipient mice were injected intravenously with 5,000 FACS-purified BM iILC2s (CD45⁺ Lin⁻ c-kit^low Sca-1^high CD25⁺ CD127⁺) from Vav.G9a^−/− mice and control G9a^fl/fl. After 3–4 wk, small intestinal lamina propria cells and lung cells were isolated and evaluated for the presence of ILC2s and ILC3s.

Results represent the mean ± SEM. Statistical significance was determined by Student’s t test using Prism 5.0 (GraphPad Software). A p-value <0.05 was considered significant.

Table S1 is a list of core ILC2 and ILC3 genes (Robinette et al., 2015) compared with control and G9a-deficient iILC2s.

We thank Ryan Vander Werff for his technical assistance performing the RNA-seq experiment; the Biomedical Research Centre animal care facility; Taka Murakami of the Biomedical Research Centre genotyping facility; Andy Johnson and Justin Wong of the University of British Columbia flow cytometry facility; Les Rollins, Rupinder Dhesi, and Michael Williams of the Biomedical Research Centre core staff; Ingrid Barta of the Biomedical Research Centre histology facility; and K. Jacobsen for critically reading the manuscript.

This work was supported by the Canadian Institutes of Health Research's (CIHR) Canadian Epigenetics, Environment and Health Research Consortium (grant 128090 to C. Zaph) and operating grants (MOP-84545 to K.M. McNagny and MSH-95368, MOP-89773, and MOP-106623 to C. Zaph) and an Australian National Health and Medical Research Council project grant (APP1004466 to C. Zaph). The development of the mouse strains used in this study was supported by funds from the Leon Judah Blackmore Foundation to F.M. Rossi. F. Antignano is the recipient of a CIHR/Canadian Association of Gastroenterology/Crohn’s and Colitis Foundation of Canada postdoctoral fellowship. M.J. Oudhoff is a Banting Fellow and a Michael Smith Foundation for Health Research Fellow. K.M. McNagny is a Michael Smith Foundation for Health Research Senior Scholar. F.M. Rossi is a Distinguished Scholar in Residence of the Peter Wall Institute for Advanced Studies. C. Zaph is a Michael Smith Foundation for Health Research Career Investigator and a Veski Innovation Fellow.

The authors declare no competing financial interests.