Runx3 specifies lineage commitment of innate lymphoid cells.

Subsets of innate lymphoid cells (ILCs) reside in the mucosa and regulate immune responses to external pathogens. While ILCs can be phenotypically classified into ILC1, ILC2 and ILC3 subsets, the transcriptional control of commitment to each ILC lineage is incompletely understood. Here we report that the transcription factor Runx3 was essential for the normal development of ILC1 and ILC3 cells but not of ILC2 cells. Runx3 controlled the survival of ILC1 cells but not of ILC3 cells. Runx3 was required for expression of the transcription factor RORγt and its downstream target, the transcription factor AHR, in ILC3 cells. The absence of Runx3 in ILCs exacerbated infection with Citrobacter rodentium. Therefore, our data establish Runx3 as a key transcription factor in the lineage-specific differentiation of ILC1 and ILC3 cells.

Innate lymphoid cells (ILCs) reside in mucosal surface to facilitate immune responses, maintain mucosal integrity, and promote lymphoid organogenesis[1]. They do not express rearranged antigen-specific receptors, are dependent on IL-2Rγc for differentiation, and all ILCs in the intestine express IL-7Rα (CD127), which forms a heterodimer with IL-2Rγc. The ILC populations are classified into three groups, ILC1, ILC2 and ILC3, based on the expression of specific cytokines, similar to T cell subsets[1]. ILC1 cells are characterized by their capacity to produce the type 1 cytokine interferon γ (IFN-γ) in response to interleukin 12 (IL-12), IL-15 and IL-18. ILC2 cells respond to IL-25 and IL-33 and secrete a set of TH2 cytokines IL-5, IL-9, IL-13 and amphiregulin. ILC3 cells share many features with TH17 and TH22 cells and can be stimulated by IL-1 and IL-23 to elicit IL-17 and IL-22 production. ILC3 cells are heterogeneous and can be further subdivided into additional subsets by expression of CD4 and NKp46: CD4+ ILC3, NKp46+ ILC3 (also known as NK22 or ILC22) and double negative (DN) ILC3 cells[1]. Fetal ILC3 cells in intestine are CD4− or CD4+ lymphoid tissue inducer (LTi) cells, which are necessary for the development of lymph nodes and Peyer’s patches (PPs)[2]. NKp46+ ILC3 cells specifically produce only IL-22, but not IL-17[1,3,4] and have the potential to differentiate into IFN-γ-producing ILC1 cells[4,5]. Thus, ILCs can be classified into different subsets which can be distinguished and they play distinct roles in immune responses.

With regard to their differentiation and transcriptional regulation, all ILC lineages are derived from common lymphoid progenitor cells (CLPs), which also give rise to B cells and T cells[1]. The earliest progenitor cells specific to ILCs are CXCR6+ integrin α4β7-expressing CLPs (CXCR6+ αLP), which have the potential to differentiate into ILC1, ILC2, ILC3, and splenic NK cells[6]. The transcription factor NFIL3 (E4BP4) is essential for differentiation of CXCR6+ αLPs and all ILC lineages. The common progenitors to all helper-like innate lymphoid cell lineages (CHILP) are defined by the Lin− CD127+ Id2+ CD25− α4β7+ phenotype and give rise to ILC1, ILC2 and ILC3 cells, but not splenic NK cells[5]. In this context, NK cells could be a different subset, distinct from ILC1 cells. The common precursor to ILC (ILCP) is identified by the expression of the transcription factor PLZF and can generate ILC1, ILC2 and ILC3 cells although they do not differentiate into the CD4+ ILC3 subset and splenic NK cells[7]. PLZF is expressed in a proportion of CHILPs, suggesting that they are precursors of ILCPs[5]. However, the ILC lineage specification process downstream of ILCPs remains to be completely elucidated.

Differentiation of each ILC subset requires specific transcription factors[1]. While ILC1 cells in the intestine are DX5− and do not express the transcription factor Eomes, splenic NK cells are DX5+ Eomes+ and appear to be dependent on Eomes for full maturation[1,5]. Although both ILC1 cells and splenic NK cells express T-bet, a TH1 transcription factor, ILC1 cells in the intestine are highly dependent on T-bet, whereas splenic NK cells are only modestly affected by the absence of T-bet[1,5,8]. ILC2 cells require GATA-3, a TH2 transcription factor, and RORα for their development[9-11]. The transcription factor RORγt is required for ILC3 and deficiency of aryl hydrocarbon receptor (AHR) affects all ILC3 subsets[1,12,13], suggesting a potential link between RORγt and AHR in ILC3 cells that has not been elucidated. Both RORγt and AHR transcription factors are also indispensable to TH17 and TH22 cells[14]. Of the ILC3 subsets, only NKp46+ ILC3 cells express and require T-bet. Although earlier studies suggested that GATA-3 is an ILC2-specific transcription factor[1,10], recent studies argue that an intermediate level of GATA-3 is also expressed in ILC1 and ILC3 cells and regulates these populations through maintaining CD127 expression[5,9,15]. Thus, the requirements of transcription factors studied thus far for specification of ILC subsets are generally similar to those in helper T cells.

The Runx family of transcription factors, especially Runx1 and Runx3, play important roles in the development of various hematopoietic lineages, including T cells[16]. Runx1 is essential for the emergence of hematopoietic stem cells from hemogenic endothelial cells and the development of lymphoid and dendritic cell progenitors, megakaryocytes, Foxp3+ regulatory T cells and TH17 cells[16,17]. Runx3 is important for differentiation of CD8+ T cells, TH1 and splenic NK cells[16,18]. The majority of phenotypes resulting from deficiency of Runx1 or Runx3 are more pronounced with deletion of Cbfb, encoding the common obligatory partner (CBF-β) of all Runx proteins, suggesting that Runx family members play overlapping roles in immune cell development and function[16].

An early study, before the identification of ILC progenitor cells, described a reduction in the numbers of PPs and RORγt+ LTi cells in the fetal intestine in mice lacking the main transcript variant of Runx1 or CBF-β[19]. Deficiency of Runx1 or CBF-β in hematopoietic cells affects Flt3+ progenitor cells including CLPs[17], suggesting that reduced numbers of fetal intestine RORγt+ LTi cells in the absence of Runx1 or CBF-β[19] could be due to effects on CLPs, the precursors to all ILCs. Therefore, the requirements for Runx proteins in the differentiation of ILCs have not been clearly defined.

Here we show that among the three Runx family members, Runx3 derived from its distal promoter was specifically expressed in ILC1 and ILC3 cells, but not ILC2 cells. Specific deletion of Cbfb or Runx3 using NKp46-Cre resulted in marked reduction of ILC1 and NKp46+ ILC3 cells in the intestine, which led to poor control of C. rodentium infection. Runx3 deletion in hematopoietic cells did not affect the numbers of CLP, αLP, CHILP and ILCP, but abrogated RORγt expression and subsequent AHR expression by all ILC3 subsets in intestine. Thus, our data reveal the non-redundant role of Runx3 in differentiation of ILC1 and ILC3 cells.

RESULTS

Distal Runx3 is expressed in ILC1 and ILC3, not ILC2

We first examined the expression of Runx1, Runx2, Runx3 and Cbfb in ILC1 cells from murine small intestinal epithelium, ILC2 cells and all ILC3 subsets from murine small intestinal lamina propria. Among Runx family members, Runx3 transcripts were predominantly expressed in all ILCs (Fig. 1a). All Runx family members including Runx3 can be expressed in two forms which originate from a proximal or distal promoter[20], with distal Runx3 transcripts being dominant in lymphocytes[20-22]. Although proximal Runx3 transcripts are poorly translated into protein in CD4+ T cells which express only proximal Runx3, CD8+ T cells use the distal Runx3 promoter for Runx3 protein expression[21,22]. To discriminate between proximal and distal promoter usage in ILCs, we further examined Runx3 expression in ILCs by RT-PCR and in Runx3d+/YFP mice expressing a membrane bound YFP from the distal Runx3 promoter[22]. Whereas ILC1 and ILC3 cells expressed only distal Runx3, expression of Runx3 in ILC2 cells was driven only by the proximal promoter (Fig. 1b,c). Splenic NK cells expressed both Runx3 transcripts but less distal Runx3, as previously observed[18]. The expression of distal Runx3 was highest in ILC1 cells, moderate in ILC3 cells, low in splenic NK cells and undetectable in ILC2 cells (Fig. 1b,c). NK cells in liver and skin are known to segregate into two subsets, DX5− CD49a+ tissue-resident NK (trNK) cells and DX5+ CD49a− conventional NK cells[23]. The trNK cells in the liver expressed more distal Runx3 than DX5+ CD49a− conventional NK cells (Supplementary Fig. 1a). By contrast, distal Runx3 expression by both NK populations in the skin was low. While CD49a and DX5 expression on salivary gland NK cells was not previously described and their tissue-residency has not been determined, their distal Runx3 expression was very high (Supplementary Fig. 1a). Thus, distal Runx3 is preferentially expressed in the ILC1 and ILC3 subsets, but not ILC2 cells.

Figure 1

ILC1 and all ILC3 subsets predominantly express Runx3 from the distal promoter. (a, b, d) Quantitative RT-PCR (a, d) and RT-PCR (b) analysis showing expression of indicated genes in intestinal ILC1 cells (CD45+ CD3− CD19− NK1.1+ NKp46+ CD49a+), intestinal ILC2 cells (CD45+ CD3− CD19− CD127+ KLRG1+ CD25+) and splenic NK cells (CD3− CD19− NK1.1+ NKp46+) from wild-type (WT) mice; in intestinal ILC3 subsets (CD45+ CD3− CD19− CD127+ GFP+) from Rorc(γt)+/GFP mice; in common lymphoid progenitor cells (CLP: CD45+ Lin− cKitlo CD127+ sca1lo Flt3+ α4β7−), α4β7 integrin-expressing CLP (αLP: CD45+ Lin− cKitlo CD127+ sca1lo Flt3− α4β7+), common progenitor to all helper-like innate lymphoid cell lineages (CHILP: CD45+ Lin− CD127+ α4β7+ Flt3− CD25−) and ILC2 precursor (ILC2P: CD45+ Lin− CD127+ α4β7+ Flt3− CD25+) from bone marrow of WT mice; in common precursor to ILC (ILCP: CD45+ Lin− cKit+ CD127+ α4β7+ PLZF-GFP+) from bone marrow of PLZFGFP-Cre+/–. Absolute copy numbers per 40000 copies of Actb are shown in a and d. ND, =not detected. (c, e). Flow cytometry showing expression of distal Runx3-YFP by indicated cells in Runx3d+/YFP mice. Shaded histograms indicate wild-type mice. (f) Flow cytometry of CHILPs showing PLZF and distal Runx3-YFP expression in Runx3d+/YFP mice. Data are representative of at least two experiments (mean and s.d. of triplicates in a and d).

We next examined ILC progenitors in adult bone marrow for distal Runx3 and other Runx expression by RT-PCR and reporter expression in Runx3d+/YFP mice. While expression of all Runx family members was relatively low in CLPs, αLPs and CHILPs from adult bone marrow, ILCPs strongly expressed Runx1 and Runx3 transcripts (Fig. 1d). Distal Runx3 expression in ILCPs was also confirmed in ILCPs isolated from Runx3d+/YFP mice (Fig. 1e). A small fraction of CHILPs in bone marrow of Runx3d+/YFP mice was PLZF+, and these cells expressed YFP driven from distal Runx3 promoter (Fig. 1e,f), suggesting PLZF+ ILC progenitor cells start expressing distal Runx3 before ILC lineage specification. Because ILC2 cells and ILC2 precursors (ILC2Ps) did not have substantial distal Runx3 expression based on absent YFP expression in these cells in Runx3d+/YFP mice (Fig. 1c,e), the apparent loss of distal Runx3 expression seems to be an ILC2-specific event. However, ILC1 cells maintain high level of distal Runx3 expression, and ILC3 cells down-regulate distal Runx3 but still maintain a moderate level of distal Runx3 expression. These data suggest that Runx3 may contribute to ILC lineage specification.

CBF-β is indispensable to NKp46+ ILC3 and ILC1 in intestine

Because Runx family members can bind to the same Runx-binding motif, other Runx proteins could potentially compensate for the deficiency of another Runx family member[16,21]. To examine the roles of Runx family members in ILC1 and ILC3 cells, we sought to delete the function of Runx3 and other Runx family members by deleting their binding partner CBF-β. We generated mice with an intrinsic Cbfb deletion in NKp46-expressing cells, ILC1 and NKp46+ ILC3 cells, by crossing Cbfbf/f mice with NKp46-Cre mice. Cbfbf/f NKp46-Cre mice lacked NKp46+ ILC3 cells in the small intestine, colon and PPs (Fig. 2a,b and Supplementary Fig. 1b,c). In addition, the numbers of other NKp46-expressing cells including ILC1 cells in intestine and NK cells in spleen, liver, salivary gland and skin were also strongly reduced in Cbfbf/f NKp46-Cre mice (Fig. 2c,d and Supplementary Fig. 1d,e). By contrast, differentiation of other ILC3 subsets were not affected by CBF-β deficiency (Fig. 2a,b). Thus, CBF-β expression in NKp46+ cells appears to be required for the development of ILC1 and NKp46+ ILC3 lineages.

Figure 2

Cbfbf/f NKp46-Cre mice lack NKp46+ ILC3 and ILC1 cells in intestine. (a) Flow cytometry of lamina propria lymphocytes (LPLs) showing ILC3 subsets in small and large intestine of Cbfb+/f NKp46-Cre and Cbfbf/f NKp46-Cre mice. Plots were gated on Live CD45+ CD3− CD19− cells. Numbers in outlined areas or quadrants indicate percent cells in each. Arrows in upper panels indicate gated cells shown in lower panels. (b) Absolute cell numbers of total ILC3 cells or indicated ILC3 subsets in LPLs from small intestine and large intestine of Cbfb+/f NKp46-Cre and Cbfbf/f NKp46-Cre mice. (c) Flow cytometry of intraepithelial lymphocytes (IELs) showing ILC1 cells in small intestine of Cbfb+/f NKp46-Cre and Cbfbf/f NKp46-Cre mice. Plots were gated on Live CD45+ CD3− CD19− cells. Numbers in outlined areas indicate percent cells in each. (d) Absolute cell numbers of ILC1 cells (CD45+ CD3− CD19− NK1.1+ NKp46+ CD49a+) in IELs from small intestine of Cbfb+/f NKp46-Cre and Cbfbf/f NKp46-Cre mice. (e) Quantitative RT-PCR analysis showing Blimp-1 expression in indicated ILC3 cells of small intestine from indicated mice, presented as relative to Actb expression (mean and s.d. of triplicates). Data are representative of more than three experiments (mean and s.d. of three mice in b and d). *p < 0.05 by Student’s t-test. NS, not significant.

To address the possibility that NKp46+ ILC3 cells lacking CBF-β were not able to express NKp46, we examined NKp46 expression in splenic NK cells and found that it was not changed by CBF-β deficiency (Supplementary Fig. 1d). Blimp-1 is preferentially expressed in NKp46+ ILC3 cells but not in other ILC3 subsets, and is not necessary for the differentiation of NKp46+ ILC3 cells[24]. This finding was recapitulated here (Fig. 2e), allowing us to examine the expression of Blimp-1 transcripts in Cbfbf/f and Cbfbf/f NKp46-Cre ILC3 (RORγt-GFP+) cells in Rorc(gt)wt/gfp mice to test the possibility that CBF-β deficiency abrogated NKp46 expression, but NKp46+ ILC3 cells were otherwise unaffected. Blimp-1 expression was comparably low in Cbfbf/f and Cbfbf/f NKp46-Cre NKp46− ILC3 (RORγt-GFP+) cells (Fig. 2e), suggesting CBF-β deficiency does not affect only NKp46 expression in ILC3 cells in Cbfbf/f NKp46-Cre mice. These results suggest that CBF-β is required for the development of NKp46+ ILC3 cells, suggesting that Runx family members may be required for ILC subset differentiation.

Runx3 is required for ILC1 and intestinal ILC3 development

To analyze the effect of Runx3-deficiency in all hematopoietic cells including ILCPs and all ILC3 subsets, we studied Runx3f/f Vav1-Cre mice in which Runx3 is deleted in hematopoietic cells. We found a comparable number of CLPs, αLPs, CHILP, ILC2P and ILCPs in the bone marrow of Runx3f/f Vav1-Cre mice compared to Runx3+/f Vav1-cre control mice (Supplementary Fig. 2). Because ILCPs expressed comparable amounts of Runx1 and Runx3 (Fig. 1d), Runx1 might compensate for the loss of Runx3 in these cells. The number of Peyer’s patches (PPs) was reduced in Runx3f/f Vav1-Cre mice compared to Runx3+/f Vav1-cre control mice (Fig. 3a). No RORγt+ ILC3 subsets were detectable in the remaining PPs and lamina propria lymphocytes (LPLs) in Runx3f/f Vav1-Cre mice (Fig. 3b,c). These data suggest a defect in ILC3 cell development in the setting of Runx3-deficiency.

Figure 3

Defects of ILC3 differentiation in Runx3f/f NKp46-Cre and Runx3f/f Vav1-Cre mice. (a) Number of Peyer’s patches (PPs) in Runx3f/f, Runx3f/f NKp46-Cre and Runx3f/f Vav1-Cre mice. (b, c) Flow cytometry of PP lymphocytes (b) or LPLs in large intestine (c) showing ILC3 subsets in Runx3f/f, Runx3f/f NKp46-Cre and Runx3f/f Vav1-Cre mice. Plots were gated on Live CD45+ CD3− CD19− cells. Numbers in outlined areas indicate percent cells. Arrows in upper panels indicate gated cells shown in lower panels. (d) Flow cytometry of LPLs in large intestine of Runx3f/f and Runx3f/f vav1-Cre mice for GATA-3, RORγt, α4β7 and PLZF. Left plots were gated on Live CD45+ CD3− CD19− CD127+ NK1.1− cells. Arrows in left panels indicate gated cells shown in right panels. (e) Absolute cell numbers of GATA-3hi (ILC2) and GATA-3int cells among CD45+ CD3− CD19− CD127+ NK1.1− LPLs from large intestine of Runx3f/f and Runx3f/f vav1-Cre mice. Numbers in outlined areas or quadrants indicate percent cells in each. Data are representative of more than three independent experiments (mean and s.d of four mice in a, of three mice in e). **p < 0.01 by Student’s t-test. NS, not significant.

We then examined if Runx3 deletion in NKp46-expressing cells recapitulated the phenotype of Cbfbf/f NKp46-Cre mice. The number and frequency of NKp46+ ILC3 cells were selectively decreased among ILC3 subsets in the PPs and LPLs of Runx3f/f NKp46-Cre mice compared to Runx3+/f NKp46-cre control mice (Fig. 3b,c). Intestinal ILC1 cells and NK cells in spleen, liver, salivary gland and skin were substantially reduced in Runx3f/f NKp46-Cre mice compared to Runx3+/f NKp46-cre mice, although the reduction was generally more modest than that observed in Cbfbf/f NKp46-Cre mice (Supplementary Fig. 1d,e; Supplementary Fig. 3a,b,c). The milder phenotype of the Runx3f/f NKp46-Cre mice compared to the Cbfbf/f NKp46-Cre mice was probably due to compensatory effects by other Runx family members, as previously described in T cell differentiation[21]. These data indicate that Runx3 is necessary for the development of ILC1 and all ILC3 subsets in the intestine.

Runx3 regulates RORγt expression in ILC3

Previous reports indicate that ILCPs express very little RORγt but express GATA-3, which is important for all ILC differentiation in the intestine[7,9]. We next investigated the hypothesis that Runx3 regulates RORγt expression in ILC3 differentiation between ILCPs and RORγt+ ILC3 cells. Lin− CD127+ NK1.1− LPLs are comprised of two main populations discriminated by GATA-3 and RORγt expression, GATA-3hi RORγt− ILC2 cells and GATA-3int RORγt+ ILC3 cells (Fig. 3d), as previously described[5,15]. GATA-3high ILC2 cell numbers were unaffected in the Runx3f/f Vav1-Cre mice (Fig. 3d,e), suggesting ILC2 differentiation was not abrogated by lack of Runx3 in hematopoietic cells. The overall frequencies and absolute numbers of GATA-3int LPLs in adult large intestine were comparable in Runx3f/f Vav1-Cre mice and control Runx3f/f mice (Fig. 3d,e). However, in contrast to control Runx3f/f mice, the GATA-3int LPLs in Runx3f/f Vav1-Cre mice did not express RORγt. These Lin− CD127+ NK1.1− GATA-3int RORγt− LPLs (referred to as CD127+ ILCLN cells) accumulated in adult large intestine in Runx3f/f Vav1-Cre mice compared to Runx3f/f control mice Fig. 3d.e), and were not ILCPs, because they did not express α4β7 or PLZF.

Next, we investigated if the CD127+ ILCLN cells in the adult intestine have the potential to become RORγt+ ILC3 cells and if Runx3 is necessary for this process. Because KLRG1 can be used as another marker for ILC2 cells among the Lin− CD127+ GATA-3hi cells in LPLs, we sorted Lin− CD127+ α4β7− NK1.1− KLRG1− RORγt-GFP− LPLs (wild-type CD127+ ILCLN cells) from adult small and large intestine of Rorc(γt)+/GFP mice (H-2b) in which GFP is driven by Rorc(γt) promoter to mark ILC3 cells (Fig. 4a). Runx3-deficient CD127+ ILCLN cells were sorted as Lin− CD127+ α4β7− NK1.1− KLRG1− LPLs from adult small and large intestine of Runx3f/f Vav1-Cre mice without excluding RORγt+ cells because RORγt+ ILC3 cells were not detected in LPLs of Runx3f/f Vav1-Cre mice. The sorted cells (H-2b) were injected into non-lethally irradiated C. Rag2 host mice (H-2d). Three months after injection, donor wild-type CD127+ ILCLN cells, but not Runx3-deficient CD127+ ILCLN cells differentiated into intestinal Lin− CD127+ GATA-3int RORγt+ILC3 cells (Fig. 4b). Wild-type CD127+ ILCLN cells also differentiated into ILC1 cells in LPLs (Lin− CD127+ NKp46+ RORγt−) and in IELs (Lin− CD127+ NK1.1+ NKp46+), while ILC1 differentiation from Runx3-deficient CD127+ ILCLN cells was impaired, though less than ILC3 differentiation (Fig. 4b), possibly due to compensatory effects by other Runx family members as mentioned previously. By contrast, no ILC2 cells (Lin− CD127+ GATA3hi) were derived from CD127+ ILCLN or Runx3-deficient CD127+ ILCLN cells, suggesting that CD127+ ILCLN cells sorted as above are progenitor cells for ILC1 and ILC3 cells, and that Runx3 deficiency in these precursors mainly dampens differentiation to RORγt-expressing ILC3 cells.

Figure 4

Runx3 regulates RORγt and subsequent AHR expression in ILC3 cells. (a, b, c, d) Wild-type CD127+ ILCN cells from intestine of Rorc(γt)+/GFP mice and Runx3-deficient CD127+ ILCN cells from intestine of Runx3f/f Vav1-Cre mice were injected into C. Rag2 mice (a, b) or cultured in vitro (c, d). (a) Post sort analysis for indicated ILC markers. (b) Flow cytometry of donor H-2b+ cells in intestinal LPLs and IELs showing indicated ILC markers three months after injection. (c) Flow cytometry showing RORγt-GFP expression in wild-type CD127+ ILCN cells from Rorc(γt)+/GFP mice before culture and post 14 days culture. (d) Flow cytometry showing IL-22 production by cells derived from indicated input cells after 14 days in culture. (e) Luciferase activity in NK-92 cells transfected with pGL3, pGL3 with RORγt promoter (RORγt-pGL3), or pGL3 with mutations in Runx binding site of RORγt promoter (RORγt (mut)-pGL3) together with Runx3-expressing vector (Runx3) or empty vector (EV). (f) Chromatin immunoprecipitation assay showing Runx3 binding to IL4 and Roc(γ)t transcription start sites (TSS) in ILC3 cells. (g) Quantitative RT-PCR analysis showing expression of indicated genes in CD45+ CD3− CD19− CD127+ NK1.1− LPLs of the indicated mice, relative to Actb expression. Numbers in outlined areas or quadrants indicate percent cells in each. Data are representative of two independent experiments (a, b, c, d; mean and s.d. of triplicates in e, f, g). **p < 0.01 by Student’s t-test. ND, not detected.

To further confirm these findings, we cultured sorted wild-type CD127+ ILCLN cells from Rorc(γ)t+/GFP mice and Runx3-deficient CD127+ ILCLN cells from Runx3f/f Vav1-Cre mice on OP9-DL1 stromal cells in vitro with IL-2, IL-7 and stem cell factor, as previously reported[5]. On day 14, wells seeded with Rorc(γ)t+/GFP wild-type CD127+ ILCLN cells contained cells which expressed RORγt-GFP (Fig. 4c), and RORγt protein expression in these cells could also be detected by anti-RORγt staining (Fig. 4d). These cells produced IL-22 in response to IL-23 and IL-1β (Fig. 4d). However, IL-22-producing, RORγt protein+ cells were not detected in wells seeded with Runx3-deficient CD127+ ILCLN cells (Fig. 4d). Taken together, these results indicate that expression of Runx3 is required for CD127+ ILCLN cells differentiation into RORγt-expressing, IL-22-producing ILCs.

We then sought to investigate how Runx3 regulates RORγt expression in ILC3 cells. A putative Runx binding site in Rorc(γt) promoter was reported[25] and Runx1 regulates RORγt expression in TH17 cells[16]. To examine if Runx3 can directly affect RORγt expression with the putative Runx-binding site in any ILC, we performed a luciferase assay with the transfectable human NK cell line NK-92 that predominately expresses Runx3 among Runx family members[26]. Transfection of the mouse RORγt promoter led to increased luciferase activity, as compared to control vector without RORγt promoter (Fig. 4e). Deletion of a Runx-binding motif in the RORγt promoter abrogated the enhanced luciferase activity (Fig. 4e). Transfection of mouse Runx3 in the NK-92 cells further increased the luciferase activity of the intact RORγt reporter, indicating that endogenous human Runx3 and ectopically expressed mouse Runx3 can promote RORγt expression in a Runx3-binding motif-specific manner. Furthermore, using chromatin immunoprecipitation (ChIP) we detected Runx3 binding to the RORγt promoter in ILC3 cells (Fig. 4f), while Runx3 did not bind in the IL-4 promoter in ILC3 cells, as previously shown in CD8+ T cells[27]. These data suggest that Runx3 directly regulates RORγt expression in ILC3 cells.

To examine if RORγt deficiency causes accumulation of CD127+ ILCLN cells in adult large intestine as observed in Runx3f/f Vav1-Cre mice, we analyzed intestinal LPLs of Rorc(γt)GFP/GFP (RORγt-deficient) mice. However, for reasons as yet unclear, the frequency of CD127+ ILCLN cells in intestinal LPLs was comparable between wild-type and RORγt-deficient mice (Supplementary Fig. 4a). It is possible that complete deletion of RORγt in CD127+ ILCLN cells might not allow the cell to survive and the low residual RORγt might be required for the accumulation of the CD127+ ILCLN cells in Runx3f/f Vav1-Cre mice.

Because RORγt is also required for the emergence and function of LTi cells in the fetal intestine[2,28-30], we tested the effect of Runx3 deficiency on these cells. The frequencies and numbers of LTi0 (CD3− CD11c− CD127+ CD4−) and LTi4 (CD3− CD11c− CD127+ CD4+) cells[30] were comparable in Runx3f/f Vav1-Cre mice and Runx3f/f control mice. However, RORγt expression was lower in LTi cells from Runx3f/f Vav1-Cre mice compared to wild-type LTi cells (Supplementary Fig. 4b,c), which may explain the partial reduction of PPs in the intestine of adult Runx3f/f Vav1-Cre mice. These observations suggest that RORγt expression may be differentially regulated in fetal LTi cells compared to adult ILC3 cells.

T-bet is also involved in differentiation of ILC3 as well as ILC1 cells because DN ILC3 cells give rise to NKp46+ ILC3 cells in a T-bet-dependent manner[13,24]. We investigated if T-bet expression was altered by Runx3 deficiency in NKp46+ ILC3 cells in Cbfbf/f NKp46-Cre mice, which were used instead of Runx3f/f NKp46-Cre mice to exclude possible compensatory effects by other Runx family members. We could not obtain a large enough number of RORγt+ NKp46+ ILC3 cells from Cbfbf/f NKp46-Cre mice for analysis. However, some Lin− CD127+ NK1.1− GATA-3int RORγtlo LPLs were NKp46+ in Cbfbf/f NKp46-Cre mice (Supplementary Fig. 5a), possibly suggesting that CBF-β deficiency might abrogate RORγt expression in NKp46+ ILC3 cells. T-bet expression in these Cbfbf/f NKp46-Cre GATA-3int RORγtlo NKp46+ cells was comparable to that in NKp46+ ILC3 cells in Cbfb+/f NKp46-Cre control mice (Supplementary Fig. 5b). In addition, the few residual ILC1 cells in Cbfbf/f NKp46-Cre mice expressed T-bet normally (Supplementary Fig. 5b), suggesting that Runx3 is necessary for ILC1 and NKp46+ ILC3 cells through a T-bet-independent mechanism. Taken altogether, Runx3 regulates ILC3 differentiation by inducing RORγt.

AHR expression is downstream of Runx3 in ILC3

Because AHR is a transcription factor that critically regulates ILC3 differentiation in the intestine[12,13], we next examined whether Runx3 deficiency affects AHR expression in Lin− CD127+ NK1.1− LPLs, which include ILC2, ILC3 and CD127+ ILCLN cells in Runx3f/f mice, and ILC2 and CD127+ ILCLN cells in Runx3f/f Vav1-Cre mice. Ahr as well as Rorc(γ)t mRNA expression was not detected in Lin− CD127+ NK1.1− LPLs isolated from Runx3f/f Vav1-Cre mice, which include ILC2 and CD127+ ILCLN cells (Fig. 4g). In contrast, Runx3f/f Vav1-Cre Lin− CD127+ NK1.1− cells had higher expression of Gata3 than Runx3f/f control mice, similar to observations made in Runx3-deficient CD8+ T cells[31]. These data suggest that AHR is downstream of Runx3 in ILC3 development.

To test if Runx3 controlled AhR expression directly or indirectly, through controlling the expression of RORγt, we determined Ahr mRNA expression in Lin− CD127+ NK1.1− LPLs from small and adult intestine of Rorc(γt)+/GFP mice, which are heterozygous for RORγt expression. RT-PCR analysis showed a 50% reduction in Ahr mRNA expression in Rorc(γt)+/GFP Lin− CD127+ NK1.1− LPLs compared to wild-type (Supplementary Fig. 6a), suggesting that AHR is downstream of RORγt in ILC3 cells. To determine if RORγt can bind either to the Ahr promoter or enhancers, we used published RORγt ChIP-seq data[32] as well as histone acetyltransferase p300 and H3K4 dimethylation (H3K3me2) profiles from TH17 cells (Supplementary Fig. 6b), which express RORγt and AHR similar to ILC3 cells. The alignment of these data sets indicated that RORγt binds to at least three sites in the vicinity of Ahr transcription start site, which according to their H3K4me2 modification and p300 binding, could be Ahr enhancer regions (Supplementary Fig. 6b). These data are consistent with the possibility that RORγt directly regulates Ahr via enhancer interactions.

To assess the possibility of direct regulation, we investigated the binding of Runx3 to the Ahr promoter in splenic NK cells, which express both Runx3 and AHR[18,33]. Published ChIP-seq data showed that active promoter or enhancer regions of Ahr promoter marked by H3K4 mono-methylation in splenic NK cells were almost identical to those marked by H3K4 di-methylation in TH17 cell[18] (Supplementary Fig. 6b). However, Runx3 binding was not detected in the Ahr promoter or enhancers in splenic NK cells (Supplementary Fig. 6b), indicating that Runx3 does not regulate AHR in NK cells, and could be taken as an indication that Ahr is also not directly regulated by Runx3 in ILC3 cells. Collectively, these data suggest that AHR expression is regulated by RORγt, whose expression is under direct control of Runx3 in ILC3s.

Runx3 is required for survival of ILC1 but not ILC3

In splenic NK cells, Runx3 regulates the expression of genes related to cell survival and proliferation downstream of IL-15 signaling[18]. Because IL-15 acts as a survival factor for splenic NK cells and intestinal ILC1 cells[5,8,23] we examined whether Runx3 controls survival of ILC1 cells in intestine. The residual intestinal ILC1 cells and liver-resident DX5− CD49a+ NK cells in the Cbfbf/f NKp46-Cre and Runx3f/f NKp46-Cre mice were more apoptotic (as assessed by Annexin V staining) and had higher proliferation (as assessed by Ki67 staining) than those in Cbfb+/f NKp46-Cre and Runx3+/f NKp46-Cre control mice (Fig. 5a,b and Supplementary Fig. 7a,b). However, the effect of Runx3 deletion among various ILC1 subsets varied somewhat because of apparent tissue-dependent effects. For example, Runx3f/f NKp46-Cre NK cells in spleen and salivary gland had normal or only marginally increased apoptosis compared to control Runx3+/f NKp46-Cre NK cells, while showing increased proliferation compared to control Runx3+/f NKp46-Cre NK cells (Supplementary Fig. 7a,b). Moreover, DX5+ CD49a− liver NK cells are thought to correspond to splenic NK cells, because they share the same gene expression profiles, trafficking properties and transcription factor dependence[23]. However, Cbfbf/f NKp46-Cre and Runx3f/f NKp46-Cre DX5+ CD49a− liver NK cells show less pronounced annexin V staining compared to their spleen counterparts (Supplementary Fig. 7a). Regardless, ILC1 cells in the intestine show enhanced apoptosis in the absence of CBF-β and Runx3.

Figure 5

Runx3 is essential for survival of ILC1 cells but not ILC3 cells. (a) Flow cytometry of CD3+ CD19+ cells, ILC1 cells (CD45+ CD3− CD19− NK1.1+ NKp46+ CD49a+) in intestine of Cbfb+/f NKp46-Cre, Cbfbf/f NKp46-Cre, Runx3+/f NKp46-Cre and Runx3f/f NKp46-Cre mice for apoptosis markers. (b, d) Flow cytometry showing expression of Ki-67 (b) and active caspases (d) in ILC1 cells of Cbfb+/f NKp46-Cre, Cbfbf/f NKp46-Cre, Runx3+/f NKp46-Cre and Runx3f/f NKp46-Cre mice. (c) Intestinal IELs of Cbfb+/f NKp46-Cre (thick line) and Cbfbf/f NKp46-Cre mice (thin line) were cultured with IL-15 (20ng/mL) for 24h and Bcl-2 expression by ILC1 cells were examined by flow cytometry (left, histograms; right, mean fluorescent intensity). Shaded histogram: isotype control. *p < 0.05 by Student’s t-test. (e, f) Flow cytometry of CD3+ CD19+ cells, CD45+ CD3− CD19− CD127+ NK1.1− GATA-3int Peyer’s patch lymphocytes of Cbfb+/f NKp46-Cre and Cbfbf/f NKp46-Cre mice (e) or Runx3f/f and Runx3f/f Vav1-Cre mice (f) for apoptosis markers. Data are representative of three independent experiments (mean and s.d. of three mice in c).

To explore the mechanism of Runx-dependent apoptosis, we examined the expression of regulators of apoptosis in Cbfbf/f NKp46-Cre ILC1 cells. Bcl-2 is induced by IL-15 stimulation and involved in NK cell survival[34]. Bcl-2 expression was significantly reduced in IL-15-stimulated intestinal Cbfbf/f NKp46-Cre ILC1 cells compared to Cbfb+/f NKp46-Cre control ILC1 cells (Fig. 5c). In addition, CBF-β or Runx3 deficiency in ILC1 cells led to increased total caspase activity compared to Cbfb+/f NKp46-Cre control ILC1 cells (Fig. 5d). Therefore, Runx3 may control the survival of ILC1 cells through regulating the expression of anti-apoptotic genes, including Bcl-2.

To determine whether CBF-β and Runx3 also control the survival of ILC3 cells, we used PP lymphocytes instead of LPLs, because the collagenase treatments used to isolate LPLs caused cell death and interfered with the apoptosis assay. NKp46+ Lin− CD127+ NK1.1− GATA-3int cells from Cbfbf/f NKp46-Cre mice, which had lost RORγt expression, had similar apoptosis rates compared to NKp46+ ILC3 cells from Cbfb+/f NKp46-Cre mice (Fig. 5e). In addition, Runx3f/f Vav1-Cre Lin− CD127+ NK1.1− GATA-3int cells, which cannot express RORγt because of their Runx3 deficiency, did not show increased apoptotic rates compared to Runx3+/f Lin− CD127+ NK1.1− GATA-3int cells (Fig. 5f). These data imply that CBF-β and Runx3 are crucial for the survival of ILC1 but not ILC3 in the intestine.

Cell-intrinsic role of Runx3 for ILC1 and ILC3 in intestine

Because crosstalks between innate lymphocytes and CD4+ T cells have been described[35,36] and CD4+ T lineage commitment was also affected in Runx3f/f Vav1-Cre mice, we tested whether Runx3 deficiency affected ILC numbers via cell-intrinsic or cell-extrinsic effects. We generated competitive bone marrow chimeric mice by transferring Cbfb+/f NKp46-Cre, Cbfbf/f NKp46-Cre, Runx3f/f or Runx3f/f Vav1-Cre bone marrow cells (Ly5.2+) mixed in equal ratios (1:1) with wild-type bone marrow cells (Ly5.1+) into lethally-irradiated Ly5.1+ congenic mice. Donor chimerism was examined 8–10 weeks post injection in spleen and intestinal PPs, IELs and LPLs based on Ly5.1 and Ly5.2 expression. Less ILC1 and NKp46+ ILC3 cells were derived from Cbfbf/f NKp46-Cre donor cells (Fig. 6a) and Runx3f/f Vav1-Cre donor cells (Fig. 6b) compared to the wild-type competitor cells, demonstrating the importance of intrinsic Runx3 in the development or maintenance of ILC1 and ILC3 cells in the intestine. In particular, almost no ILC3 cells were recovered from Runx3f/f Vav1-Cre donor bone marrow in PPs (Fig. 6b). On the other hand, Runx3f/f Vav1-Cre derived ILC2 cells repopulated the intestinal lamina propria and PPs to the same extent as wild-type cells. Thus, Runx3 deficiency affects the development or maintenance of ILC1 and ILC3 cells in a cell-intrinsic manner, but does not affect ILC2 cells.

Figure 6

Cell intrinsic requirement of Runx3 for development of ILC1 and ILC3 cells. (a, b) Fifty percent of Ly5.1 WT bone marrow cells and fifty percent of Ly5.2 bone marrow cells from Cbfb+/f NKp46-Cre, Cbfbf/f NKp46-Cre (a), Runx3f/f or Runx3f/f Vav1-Cre mice (b) were transferred into lethally irradiated Ly5.1 mice. Chimerism was determined by Ly5.1 and Ly5.2 expression of indicated cells in the chimeric mice by flow cytometry. ILC1 cells (CD45+ CD3− CD19− NK1.1+ NKp46+ CD49a+) were from small intestine. ILC2 cells (CD45+ CD3− CD19− CD127+ GATA-3high) and ILC3 cells (CD45+ CD3− CD19− CD127+ RORγt+) were from small and large intestine (a, b) and PP lymphocytes (b). RORγt expression was detected by anti-RORγt staining. Numbers in outlined areas indicate percent cells in each. Data are representative of two independent experiments.

Runx3 in ILCs is essential for protection against C. rodentium

We next determined the physiological role of Runx3 expression in ILCs using a model of C. rodentium infection in Runx3f/f NKp46-Cre mice, considering that CD4+ T cells[37], other hematopoietic cells and enterocytes[38] also could be affected in Runx3f/f Vav1-Cre mice. Runx3f/f NKp46-Cre infected mice showed a similar degree of body weight loss and survival compared to Runx3+/f NKp46-Cre mice (data not shown), similar to observations made in T-bet-deficient mice, which do not have intestinal ILC1 and NKp46+ ILC3 cells[4,24]. However, on day 8 after infection with C. rodentium, Runx3f/f NKp46-Cre mice had shorter colon lengths and higher bacterial titers in the spleen than Runx3+/f NKp46-Cre control mice (Fig. 7a,b). Compared to uninfected Runx3wt/f NKp46-Cre control mice, there was similarly no inflammation in the Runx3f/f NKp46-Cre mice without C. rodentium infection (Fig. 7c). However, following infection, Runx3f/f NKp46-Cre mice had more persistent intestinal damage including features of increased epithelial injury, crypt hyperplasia and increased inflammatory cell infiltration compared to Runx3f/f NKp46-Cre control mice (Fig. 7c, d). These observations suggest a specific role for ILC1 and NKp46+ ILC3 cells in controlling C. rodentium infection.

Figure 7

Runx3 in ILC cells is critical to control acute infection with C. rodentium. (a-i) Runx3+/f NKp46-Cre and Runx3f/f NKp46-Cre mice were orally infected with C. rodentium. On day eight after infection, (a) length of colon, (b) titers of C. rodentium in spleen (dotted line, detection limit), (c) histology of colon (scale bars, 100 µm), (d) percentage of mucosal damage of colon, (e,f) IL-22 production by ILC3 cells in colon (e, frequency; f, absolute cell numbers) and (g–i) IFNγ production by ILC1 cells in colon (g, frequency in IELs; h, frequency in LPLs; i, absolute cell numbers) were examined. Numbers in outlined areas or quadrants indicate percent cells in each (e, g, h). Arrows in left panels in h indicate gated cells shown in right panels. Data are representative of more than three independent experiments (mean and s.d. of four mice in a and b, seven mice in d, three mice in f and i). *p < 0.05 and **p < 0.01 by Student’s t-test. NS, not significant.

Because IL-22, especially from ILC3 cells, and IFN-γ are required to control acute infection with C. rodentium[3,13,24,39], we examined the production of these cytokines in C. rodentium–infected Runx3f/f NKp46-Cre mice. IL-22-producing ILC3 cells were strongly reduced in the intestine of these mice (Fig. 7e,f). In addition, IFN-γ-producing ILC1 cells in IELs and LPLs were only scarcely detected in Runx3f/f NKp46-Cre mice after infection as compared to Runx3f/f NKp46-Cre control mice (Fig. 7g,h,i), indicating that Runx3 expression in ILCs is critical for host immunity and cytokine production against C. rodentium infection.

DISCUSSION

In this study, we show that Runx3 regulates the development and/or maintenance of ILC1 and ILC3 cells in the intestine. Runx3 induced the transcriptional regulator RORγt and its downstream target AHR in ILC3 cells. Runx3 regulated the differentiation of ILC3 from progenitor cells specific to ILC1 and ILC3 cells into RORγt+ ILC3 cells in the adult mouse intestine. As a maintenance factor, Runx3 was necessary for the survival of ILC1 cells, but not ILC3 cells. Runx3 deficiency had a limited effect on ILC1 development, probably due to compensatory effects by other Runx family members at steady-state. Competitive bone marrow chimeras revealed the cell-intrinsic contribution of Runx3 to the development of ILC1 and ILC3 cells, but not ILC2 cells. Finally, a requirement for Runx3 in the maintenance of ILCs was evident after C. rodentium infection. Thus, these results establish Runx3 as a key player in ILC lineage-specific differentiation.

Taken together, our data add substantial evidence supporting the parallel but somewhat distinct differentiation of ILCs and CD4+ T cell subsets. CD4+ T cells and ILCs share the same signature transcription factors, although all ILCs in the intestine differentially express GATA-3. CD127+ ILC1 and ILC3 cells also express GATA-3 at intermediate levels while ILC2 cells are characterized by high GATA-3 expression. TH1 cells and ILC1 cells are characterized by high expression of T-bet and Runx3. TH2 cells and ILC2 cells are identified as GATA-3hi populations. TH17 cells use Runx1 to induce RORγt expression, while in ILC3 cells RORγt is driven by Runx3. Because Runx family members orchestrate CD4+ T cell differentiation together with GATA-3 and T-bet, our findings offer additional insights into the mechanism of ILC lineage differentiation.

Without Runx3, ILC3 development was arrested at the CD127+ ILCLN stage. GATA-3 controls CD127 expression[9] and only the PLZF+ cells among Lin− CD127+ bone marrow cells express distal Runx3. Although these findings suggest that Runx3 could be downstream of GATA-3, the inducible deletion of GATA-3 in ILC3 cells did not alter Runx1 and Runx3 expression[9]. The interpretation that Runx3 is not regulated by GATA-3 in ILCs is also supported by the observations that ILC2 cells do not express distal Runx3 in spite of high GATA-3 expression. GATA-3 expression is rather inversely correlated with distal Runx3 expression, because only GATA-3int ILC1 and ILC3 cells, but not GATA-3hi ILC2 cells, express distal Runx3.

Runx3 promotes TH1 commitment and represses GATA-3 through direct binding to GATA-3[40]. Conversely, GATA-3 blocks Runx3 function by direct interaction with Runx3 in TH2 cells[41], suggesting that the balance between GATA-3 and Runx3 controls the lineage determination toward TH1 or TH2 cells. The same transcriptional network could regulate ILC1 and ILC2 differentiation, considering that ILCPs express both GATA-3 and Runx3. ILC1 cells express high amounts of Runx3, which is apparently necessary to overcome their GATA-3 expression. By contrast, ILC2 cells acquire high GATA-3 expression and lose Runx3 expression, which appears necessary for the GATA-3-dependent machinery to drive ILCPs to ILC2 cells. Additionally, Runx family members interact with T-bet to control CD4+ T cell differentiation. T-bet regulates Runx3 to enhance TH1 commitment and attenuates TH17 skewing by inhibiting Runx1-mediated RORγt expression through antagonistic binding to Runx1[27,42,43]. Possible interactions of T-bet with Runx3 during ILC lineage determination seem to be very complex. During ILC differentiation Runx3 expression is induced earlier than T-bet expression, because ILCPs express Runx3 but not T-bet[7]. It is not still clear how T-bet is induced, associates with Runx3 and differentially regulates two different subsets of T-bet-dependent ILCs, ILC1 and NKp46+ ILC3 cells.

Prior data showed that Runx1 or CBF-β deficiency is associated with reduction of LTi cells in the fetal intestine[19]. However, it was challenging to interpret these data at that time in the absence of knowledge of ILC progenitor cells. Here we investigated the expression of Runx family members in ILC progenitor cells and all ILC subsets. Runx3, not Runx1, was mainly expressed by ILC1 and ILC3 cells and contributed to ILC lineage commitment toward these two subsets. Although Runx1 expression is not very high in CLPs, Runx1 is critical for differentiation of Flt3+ progenitor cells including CLPs[17]. Runx1 may also have a critical role in ILC differentiation along with Runx3, just before ILC lineage commitment, because both Runx1 and Runx3 were highly expressed in ILCPs. Further studies will be required to determine the function of Runx1 and Runx3 in ILCPs.

RORγt expression was less severely affected in Runx3-deficient fetal LTi cells than in adult ILC3 cells. This could be due to different requirements for their development[29,44,45] and/or to different compensatory effects. For example, IL-15 or IL-2 stimulation is required to reveal proliferation defects in splenic NK cells lacking Runx3[18]. Also, the reduction of ILC1 cells in Runx3-deficient mice was more evident in C. rodentium-infected Runx3f/f NKp46-Cre mice. Thus, differences in microbiota or inflammatory circumstances in fetal versus adult intestine may affect the Runx3 requirement for RORγt expression in ILC3 cells.

The CD127+ ILCLN cell described here in Runx3f/f Vav1-Cre mice are α4β7-negative. By contrast, some studies found ILC progenitor cells in the Lin− CD127+ α4β7+ population from BM, fetal liver and intestine[5,7,44,46]. However, it should be noted that adult ILCs in the intestine do not express α4β7, but do express CD127[7,10] (data not shown)[7,10]. Therefore, the CD127+ ILCLN cells described here could be ILC1- and ILC3-specific ILC progenitors that have already migrated to the intestine, but were arrested in their ILC3 development because they require Runx3 and RORγt for further differentiation.

Runx3 dysfunction has been discussed for many years in the pathogenesis of colitis[37,47,48]. Aberrant TH17 cells deficient for Runx3 were reported to be sufficient to induce colitis[37]. Herein, we demonstrated that Runx3 is indispensable to development of ILC3 cells, which are main source of IL-22 in intestine to maintain epithelial integrity. Thus, Runx3 mutations and loss of Runx3 function can dampen the host’s ability to protect the intestinal barrier against microbes, implicating Runx3 defects in the pathogenesis of intestinal inflammation.

METHODS

Mice

All the mice were bred and maintained at a specific pathogen-free facility of Washington University School of Medicine and animal protocols were approved by the Washington University Animal Studies Committee. C57BL/6 mice and congenic Ly5.1+mice were obtained from NCI. Rorc(γt)+/GFP and Vav1-Cre transgenic mice were obtained from Jackson. Cbfbf/f, Runx3f/f, Runx3d+/YFP mice were previously described[22] and provided by Dan R. Littman (New York University, New York, NY). NKp46-Cre mice were kindly provided by Eric Vivier[49] (Centre d'Immunologie de Marseille-Luminy, Marseille, France). All Cre mice were heterozygous. Control mice were all littermates.

Cell preparation

Lymphocytes from the small and large intestine were isolated as previously described[8,13] with some modifications. Briefly, small and large intestines were dissected and fat tissues were removed. Peyer’s patches were removed from the small intestine and smashed through 70µm strainer. Intestines were cut open longitudinally, washed with PBS then cut into 5 mm-length pieces, followed by incubation with HBSS buffer including 10% FBS, 5mM EDTA and 15mM HEPES at 37 °C for 20 min. After being vortexed for 20s, the dissociated cells were collected for intra epithelial lymphocytes. To isolate lamina propria cells, rest of the tissues were washed with PBS twice and 0.5 g of the tissues were digested with 25 ml of RPMI1640 media including 5 mg of Liberase TL (Roche Life Sciences), 5 mg of DNase I (Roche) and 50 mg of Dispase II (Roche) at 37 °C for 30 min. The digested tissues were passed through 100 µm and 70 µm strainer after vigorous vortexing for 20s. Then, lymphocytes were isolated from the interphase of 40 % and 80 % Percoll gradient after centrifugation at 1800 rpm for 20 min.

Antibodies and flow cytometry

Supernatant of hybridoma 2.4g2 (anti-CD16/32) was used to block Fc receptor. Cells were stained with Fixable Viability Dye eFluor 506 (eBioscience) to detect dead cells before surface staining. Annexin V apoptosis detection kit, eFlour 450 (eBioscience) and Ki-67 kit (BD Biosciences) were used according to manufacturers’ protocol. Antibodies used for flow cytometry were listed in Supplementary Table 1. Pan-caspase activity was detected by FAM-FLICA™ Poly Caspase Assay Kit (ImmunoChemistry Technologies) according to manufacturer's protocol. For intracellular cytokine staining and transcription factor staining, Foxp3/Transcription Factor Staining Buffer Set (eBioscience) was used as manufacturer’s protocol. IELs (1 × 106 cells/ml) and LPLs (1 × 106 cells/ml) were stimulated with IL-12 (10 ng/mL) or IL-23 (10ng/mL)/IL-1β (10ng/mL) with GolgiPlug at 37°C for 4 hours for cytokine production, or IL-15 (20 ng/mL) at 37°C for 24 hours for Bcl-2 expression. Data were acquired on a FACSCanto II (BD Biosciences) and analyzed with FlowJo software (TreeStar).

RT-PCR

ILC1 cells from IELs, ILC3 subsets from LPLs and CD45+ CD3− CD19− CD127+ NK1.1− LPLs were isolated from small and large intestines and sorted by Moflo. CLPs, αLPs, CHILP, ILCPs, ILC2P were isolated from bone marrow cells. RNA was purified with Trizol (Life technologies) and cDNA was synthesized with Superscript III First Strand Synthesis Kit (Life technologies) according to manufacturer’s protocol. Then, mRNA transcripts were quantified with a StepOnePlus Real-Time PCR System and iQ SYBR Green Supermix (Biorad). The absolute expression of CBFβ, Runx1, Runx2 and Runx3 (standard curve method) and the relative expression of Blimp1, GATA-3, RORγt and AHR (2−deltaCt method) were calculated and normalized to Actb expression. For RT-PCR to detect distal and proximal Runx3, normalized amount of cDNA was applied to PCR (Advantage GC2 PCR kit; Clontech). Specific primers were listed in Supplementary Table 2.

In vitro ILC differentiation and adoptive transfer

In vitro cell culture to support all ILC differentiation was reported elsewhere[5]. Briefly, CD45+ CD3− CD19− CD127+ α4β7− NK1.1− KLRG1− RORγt− or RORγt+ cells from Rorc(γt)+/GFP mice and CD45+ CD3− CD19− CD127+ α4β7− NK1.1− KLRG1− cells from Runx3f/f Vav1-Cre mice were sorted by Moflo. Then, 2.000 to 10.000 cells were injected into C. Rag2 mice irradiated at 450 rad or 100 cells were cultured in individual wells containing a monolayer of mitomycin C-treated OP9-DL1 cells with IL-7 (25ng/mL), SCF (25ng/mL) and IL-2 (10ng/mL) for 14 days. Cultured cells were stimulated with IL-23 (30 ng/ml) and IL-1β (30 ng/ml) with GolgiPlug at 37°C for 4 hours to induce IL-22 production.

Luciferase assay

The RORγt promoter encompassing a region from −242 to +109 relative to the transcription start site was inserted into the Xho1/HindIII site of pGL3. To mutate the Runx binding motif, TGTGGTT (nt −174 to −168 from the transcription start site), we modified it to TACAATT. NK-92 cells were obtained from ATCC and transfected with Amaxa cell line Nucleofector Kit R (Lonza) by manufacturer’s protocol. Briefly, 5 × 106 cells of NK-92 cells were electroporated with 1 µg of pGL3 vector with 1 µg of empty pEF vector or Runx3-pEF vector. After 18 hour incubation at 37°C, 5 % CO2, the cells were harvested, counted for normalization and resuspended in Glo Lysis Buffer. Luciferase activity was measured by 1450 Micro Beta (Wallac).

Chromatin immunoprecipitation

ILC3 cells (1.5 × 105 cells) were isolated from Rorc(γt)+/GFP mice. Cells were cross-linked with 1% paraformaldehyde, lysed and sonicated to generate 100–300 bp DNA fragments. DNA prepared from 5% of sample prior to immunoprecipitation was used as input DNA. DNA-protein complexes were immunoprecipitated with 4µl of rabbit anti-Runx3 sera, developed in the Groner lab[18]. After reverse-cross-linking and DNA purification, immunoprecipitated DNA fragments were analyzed by qRT-PCR using SYBR green. Primers were listed in Supplementary Table 2.

Generation of bone marrow chimeras

Congenic Ly5.1+ mice were lethally irradiated at 950 rad and reconstituted with 1 × 107 cells of Ly5.1+ WT bone marrow cells and 1 × 107 cells of Ly5.2+ bone marrow cells which were obtained from Cbfb+/f NKp46-Cre, Cbfbf/f NKp46-Cre, Runx3f/f or Runx3f/f Vav1-Cre mice. After 8–10 weeks, spleen, PPs, IELs and LPLs in small and large intestine were collected for analysis.

C. rodentium infection

Runx3+/f NKp46-Cre or Runx3f/f NKp46-Cre mice were orally infected with 5 × 109 C. rodentium strain DBS100 (American Type Culture Collection) as described elsewhere[13]. On day 8 after infection, spleen, small intestine and colon were harvested from the infected mice. Spleens were weighed, transferred into a 2 ml tube with 1ml of PBS and a 5 mm metal bead (Qiagen) and homogenized with Mini-BeadBeater-8 (Biospec) at a medium speed for one minute. One hundred µl of homogenates were plated onto MacConkey agar plates and C. rodentium colonies were counted after overnight incubation at 37°C. Colon was fixed with Methacarn at 4°C for 2 hours, then 70% ethanol at 4°C for 1 hour, incubated with 20% sucrose in PBS overnight and embedded into OCT compound (Tissue-Tek) for HE-staining. Colons were used to isolate IELs and LPLs as describe above. To examine colonic pathology based on a previously published pathology scoring system[50], observers were blinded to the experimental conditions. The pathological parameters include 1) the extent of inflammatory infiltration 2) degree of epithelial hyperplasia 3) goblet cell depletion and 4) crypt abscess. Data are represented as the percentage of damaged area along the colonic tissue.

Statistical analyses

Data were analyzed by two-tailed Student’s t-test with Graphpad Prizm 4. P values of <0.05 considered statistically significant. No randomization was used in animal studies. No deliberate attempt was made to study only selected mice except based on genotype. No blinding was done except for histological scoring.