The ETS1 transcription factor is required for the development and cytokine-induced expansion of ILC2

We previously demonstrated that ETS1 is required for the development of NK cells (Ramirez et al., 2012). However, our studies were severely hampered by the neonatal lethality of ETS1 deficiency (Bories et al., 1995; Barton et al., 1998; Gao et al., 2010). To overcome this limitation, we created mice in which the Ets1 gene could be inactivated by Cre-mediated recombination. We flanked the exons coding for the ETS1 DNA–binding domain by loxp sequences such that Cre-mediated recombination results in a mutation analogous to the germline mutation described by Bories et al. (1995), from which no truncated ETS1 protein is produced (Fig. S1 A). We crossed Ets1^f/f mice to Il7ra^Cre mice, which produce Cre in IL-7 receptor–expressing (CD127⁺) cells, including CLPs (Schlenner et al., 2010), the progenitors of all lymphoid cells. No ETS1 protein was detected in thymocytes isolated from Il7ra^Cre/+Ets1^f/f (Ets1^Δ/Δ) mice, consistent with the deletion of Ets1 exons 8 and 9 in all lymphoid cells (Fig. S1 B).

To confirm that Ets1^Δ/Δ NK cells phenocopy NK cells in Ets1^−/− mice, we performed flow cytometry to examine NK cell numbers and receptor expression. As anticipated, Ets1^Δ/Δ mice had a decreased frequency and number of mature NK (mNK) cells, and these cells expressed less of the activating receptor NKp46 compared with littermate controls (LMCs; Fig. S1, C–E; Ramirez et al., 2012). These data demonstrate the utility of this Ets1^f allele for the analysis of ETS1 function after Cre-mediated recombination.

We noted a dramatic reduction in Lin⁻CD127⁺ cells in ETS1-deficient mice, which led us to question whether the development of ILC2s requires ETS1, because ILC2s are the major BM cells with this phenotype. Quantitative PCR (qPCR) analysis revealed that Ets1 mRNA was expressed in BM ILC2s in amounts higher than the mRNA encoding ID2 or ROR-α, both of which are required for ILC2 development (Fig. 1 A; Diefenbach et al., 2014). Analysis of BM cells from Ets1^Δ/Δ mice revealed a reduced frequency and number of Lin⁻CD127⁺SCA1⁺ ILC2s as compared with LMCs (Fig. 1, B and C). The LMC Lin⁻CD127⁺SCA1⁺ cells also expressed the IL-33 receptor component IL1RL1 (ST2) and ICOS, and a subset of these cells expressed KLRG1, further confirming that these cells are ILC2s (Fig.1 B and not depicted). We crossed the Ets1^Δ/Δ mice to Rag1^−/− mice to determine whether Ets1^Δ/Δ T cells influence ILC2 development because these cells have been reported to aberrantly produce cytokines that could affect ILC2s (Garrett-Sinha, 2013). Rag1^−/−Ets1^Δ/Δ (REts1^Δ/Δ) mice also had reduced numbers of BM ILC2s compared with Rag1^−/− LMC (RLMC) cells (Fig. 1 C). These data demonstrate that ETS1 was required for the development of BM ILC2s.

It is possible that ETS1 was required for the generation of ILC2s because it impacted another lymphoid cell type that was required for ILC2 development, as recently described for NK cell maturation (Kim et al., 2014). To rigorously test whether the requirement for ETS1 was intrinsic to ILC2s or their progenitors, we created chimeric mice in which CD45.2⁺ Ets1^Δ/Δ or LMC BM cells were injected into lethally irradiated CD45.1⁺ hosts along with WT CD45.1⁺ BM cells. 8 wk later, CD45.2⁺ cells were equally represented among the CLPs and ILC2s in the BM of RLMC:WT chimeras (Fig. 1, D and E). In contrast, in REts1^Δ/Δ:WT chimeras, CD45.2⁺ cells contributed to the CLP population, but ILC2s were almost exclusively CD45.1⁺ (Fig. 1, D and E). Therefore, ETS1 was required for the generation of ILC2s even when the hematopoietic environment contained ETS1-sufficient BM cells.

A recent study revealed that ILC2s in peripheral tissues and secondary lymphoid organs are tissue resident and that under homeostatic conditions, ILC2s are not replenished from BM progenitors (Gasteiger et al., 2015). Therefore, we tested whether ILC2s in mesenteric LNs (mLNs) were present in Ets1^Δ/Δ mice. We found that the number of mLN ILC2s was substantially reduced in Ets1^Δ/Δ mice compared with LMCs (Fig. 2, A and B). In contrast, the numbers of ILC2s in the lungs of Ets1^Δ/Δ and LMC mice were not statistically different (Fig. 2, C and D). Interestingly, similar to what we found in Ets1^Δ/Δ mice, ILC2s in secondary lymphoid organs and lungs appear to have different requirements for the transcription factor BCL11b (Califano et al., 2015; Yu et al., 2015). Although lung ILC2s are tissue resident, they can expand in response to chronic inflammation (Gasteiger et al., 2015). Therefore, it is possible that mice with ILC2 developmental defects experience more inflammation in the lungs than LMCs and that a few ILC2s that make it to that tissue expand in response to this inflammation. Alternatively, when ILC2 numbers are low, the lung niche may support their proliferation or immigration. Consistent with the tissue-resident status of ILC2s, we were unable to detect significant numbers of RLMC or REts1^Δ/Δ ILC2s in the mLNs or lungs of recipient mice in mixed BM chimeras (Fig. 2 E). Our data demonstrate a requirement for ETS1 for the development of mLN ILC2s.

Ets1 mRNA initiates in lymphoid primed multipotent progenitors (Ramirez et al., 2012) and, therefore, ETS1 could be required for the transition from lymphoid primed multipotent progenitors/CLPs to ILC2s. ILC2s arise from CHILPs, which are phenotypically identified as Lin⁻Flt3⁻ cells that express CD127, α4β7, and SCA1 but not CD25 (Fig. 3 A; Klose et al., 2014). Notably, the frequency and number of CHILPs was similar in REts1^Δ/Δ and RLMC BM (Fig. 3, A and B). In contrast, in Ets1^Δ/Δ mice, which had an intact Rag1 gene, there was an ∼23-fold decrease in the number of CHILPs compared with LMC mice (Fig. 3 B). These data raise the possibility that although ETS1 was not required for the development of CHILPs, Ets1^Δ/Δ CHILPs failed to compete with RAG1-dependent cells. Alternatively, Ets1^Δ/Δ T cells may have exerted an inhibitory effect on CHILPs, for example, by producing inhibitory cytokines or by occupying a limiting BM niche that supports CHILPs. To address these possibilities, we created mixed BM chimeric mice in which REts1^Δ/Δ and RLMC CD45.2⁺ cells were in competition with Rag1^+/+CD45.1⁺ cells. Under these conditions, RLMC cells were able to generate CLPs and CHILPs efficiently (Fig. 3 C). In contrast, in the REts1^Δ/Δ:WT chimeras, REts1^Δ/Δ cells generated CLPs, but they contributed to <5% of CHILPs (Fig. 3 C). These data support the hypothesis that ETS1 deficiency compromised the fitness of CHILPs and that these cells were unable to compete with WT CHILPs or WT B or T lymphocytes.

Our data indicate that ETS1 is a component of the early transcriptional network for ILC2 development. In NK cells, ETS1 is required for the increase in Id2 expression that accompanies NK cell maturation (Ramirez et al., 2012). Therefore, we tested whether ETS1 was required for Id2 expression in CHILPs or ILC2s using an Id2^GFP reporter. Expression of GFP from the Id2^GFP reporter was highly variable in Ets1^Δ/Δ CHILPs, suggesting that ETS1 may be required for stable Id2 transcription; however, the mean intensity of GFP was similar in both strains (Fig. 3, D and E). In the LMCs, Id2^GFP expression was higher in ILC2s than in CHILP, but this increase was not observed in the few ILC2s present in Ets1^Δ/Δ mice (Fig. 3, D and E). In contrast to Id2 mRNA, GATA3 and TCF1 were expressed at similar or higher levels in Ets1^Δ/Δ CHILPs and ILC2s when compared with LMCs (Fig. 3 D). These data indicate that ETS1 was required for stable and high Id2 mRNA expression in CHILPs as they differentiate into ILC2s.

The paucity of ILC2s in the BM of Ets1^Δ/Δ mice impeded an analysis of protein or gene expression. Therefore, we tested whether BM ILC2s could be expanded in vitro. We isolated Lin⁻CD127⁺SCA1⁺ ILC2s from the BM of RLMC and REts1^Δ/Δ mice and cultured them in vitro under defined ILC2 conditions for 2 wk: OP9–delta-like ligand 1 (DL1) in the presence of IL-7 and IL-33 or stem cell factor (SCF), IL-7, and IL-2 (Neill et al., 2010; Klose et al., 2014). Under either condition, RLMC cells expanded well and generated progeny that were Lin⁻ICOS⁺SCA1⁺CD127⁺ ILC2s. In contrast, cultures initiated with REts1^Δ/Δ cells generated few ICOS⁺ cells (Fig. 4 A). Overall, there was a 20-fold decrease in the number of ICOS⁺ cells generated from REts1^Δ/Δ Lin⁻CD127⁺SCA1⁺ cells compared with RLMC cells regardless of the culture conditions used, indicating that ETS1 was required for their in vitro expansion (Fig. 4 B). Identical results were obtained when we isolated the ILC2s from Rag1-expressing Ets1^Δ/Δ mice (not depicted). To rule out that a decrease in ILC2 purity contributed to the reduced expansion, we performed similar experiments starting with CLPs. REts1^Δ/Δ CLPs also failed to generate substantial numbers of ILC2s in vitro compared with RLMC CLPs (not depicted). Therefore, ETS-deficient ILC2s failed to expand efficiently in vitro.

Despite the reduced expansion of Ets1^Δ/Δ ILC2s in vitro, we were able to generate more ILC2s than were available in vivo. In an attempt to gain insight into the requirements for ETS1 in ILC2s, we performed a microarray analysis on mRNA isolated from the ICOS⁺ cells that expanded from LMC and Ets1^Δ/Δ ILC2s cultured on OP9-DL1 with IL-7 and IL-33. There were 986 probe sets that were differentially expressed by at least 1.5-fold (P < 0.05), of which 418 were decreased (1.9% of LMC expressed probes) and 568 were increased (2.4% of Ets1^Δ/Δ expressed probes) in Ets1^Δ/Δ cells (Fig. 4 C). As expected, the expression of Ets1 mRNA was lower in Ets1^Δ/Δ cells (Fig. 4 E). In addition, Klrg1 and Areg, genes associated with mature ILC2s, were decreased in the Ets1^Δ/Δ cells, and KLRG1 protein was not detected on the Ets1^Δ/Δ cells by flow cytometry (Fig. 4, E and G). An analysis of Gene Ontology terms revealed that 54 of the differentially expressed genes encoded proteins involved in cell proliferation (P < 1.5 × 10⁻⁸), including Ccnd2, 34 were involved in the regulation of phosphorylation (P < 3.4 × 10⁻⁷), and 71 encoded proteins involved in intracellular signaling cascades (P < 1.7 × 10⁻⁶; Table S1; Fig. 4 D). GSEA revealed that LMC cells had higher expression of genes associated with T effector cell differentiation than Ets1^Δ/Δ cells (Fig. 4 D). Therefore, despite selection for cells that can grow in vitro, genes involved in cell proliferation and signaling were dysregulated in the absence of ETS1, and no KLRG1⁺ ILC2s were generated.

In addition to the obvious failure to generate mature ILC2s, a KEGG pathway analysis revealed dysregulation of cytokine–cytokine receptor signaling (P = 4.5 × 10⁻⁶). Interestingly, the signature cytokines for ILC2s, IL-5 and IL-13, were not highly expressed in LMC or Ets1^Δ/Δ in vitro–expanded cells. In contrast, Il24 and Il6 were among the most differentially expressed transcripts, and we confirmed that mRNA encoding these cytokines was increased in Ets1^Δ/Δ ILC2s expanded in vitro by qPCR (Fig. 4, E and G). Cd160, which encodes a cell surface protein involved in mucosal barrier function and cytokine production from NK cells, was also more highly expressed in Ets1^Δ/Δ cells compared with LMC cells (Fig. 4 F; Shui et al., 2012; Tu et al., 2015). Because no mature ILC2s were generated in our in vitro cultures, we considered the possibility that CD160 was a marker of immature ILC2s. However, BM KLRG1⁻ or KLRG1⁺ ILC2s did not express CD160, indicating that CD160 was not a marker of this ILC2 progenitor population. Collectively, our data demonstrate that ETS1 was required for the in vitro generation of mature ILC2s and that the cells generated in vitro from Ets1^Δ/Δ Lin⁻ICOS⁺ cells differ from LMC ILC2s in the expression of genes involved in intracellular signaling as well as cytokine expression.

Given that lung ILC2 numbers were not severely affected by ETS1 deficiency, we tested whether ETS1 was required for ILC2 expansion after systemic administration of the alarmin IL-33. Intravenous injection of IL-33 into LMC mice resulted in an approximately eightfold expansion of ILC2s within 72 h (Fig. 5, A and B). In contrast, in Ets1^Δ/Δ mice, IL-33 injection resulted in only a 3.8-fold expansion of ILC2s (Fig. 5, A and B). Therefore, similar to what was observed in vitro, ETS1 was required for the appropriate expansion of lung ILC2s in response to IL-33.

We next tested whether the few lung ILC2s that expanded in response to IL-33 in vivo could produce the Th2 cytokines IL-5 and IL-13. Upon stimulation, ∼70% of LMC ILC2s were able to produce IL-13 and >50% could produce IL-5 (Fig. 5, C and D). In contrast, only 3% of ILC2s from Ets1^Δ/Δ lungs produced IL-13 and <7% produced IL-5 (Fig. 5, C and D). Ets1^Δ/Δ ILC2s also produced less IL-6 compared with LMCs, in contrast to what was observed in the in vitro–expanded BM ILC2s (not depicted). Collectively, our data indicate that ETS1 was required for cytokine production from lung ILC2s after exposure to IL-33.

Recently, iILC2s were described in the lungs that express KLRG1 but lack the receptor for IL-33, IL1RL1/ST2 (Huang et al., 2014). In contrast to natural ILC2s (nILC2s) that respond to IL-33, iILC2s proliferate and become activated in response to IL-25. After activation, iILC2s were proposed to differentiate into nILC2s. To determine whether IL-25 could elicit the expansion of Ets1^Δ/Δ iILC2s, we injected IL-25 into Ets1^Δ/Δ and LMC mice and quantified the number of iILC2s 3 d later. iILC2s could not be detected in the lungs of untreated LMC or Ets1^Δ/Δ mice (Fig. 5, E and F). 3 d after the injection of IL-25, there was a 115-fold expansion in the number of iILC2s in the lungs of LMC mice. In contrast, at the same time point, iILC2s remained near the limit of detection in the lungs of Ets1^Δ/Δ mice (Fig. 5, E and F). Our data demonstrate that in addition to being required for ILC2 development in the BM and expansion of nILC2s by IL-33, ETS1 was required for the expansion of iILC2s in response to IL-25.

In conclusion, we have created a conditional allele for Ets1 that allowed us to demonstrate a role for ETS1 in the regulation of Id2 in ILC2s, the fitness of CHILPs, and the development of ILC2s in the BM and mLNs. Despite the presence of ILC2s in the lungs, ETS1 was required for their expansion and for the expression of IL-5 and IL-13 in response to IL-33 and for the expansion of iILC2s in response to IL-25. Although we could expand a small number of ILC2s in vitro from Ets1^Δ/Δ BM ILC2s, these cells failed to express critical markers of mature ILC2s, and they expressed genes that were not typical of ILC2 progenitors such as Cd160. These findings raise the possibility that the cells we expanded in vitro are not bone fide ILC2s; they could be rare ILC2 progenitors, ILC2 progenitors with dysregulated gene expressions, or distinct cell types. The cells do not appear to be NK cells, ILC1s, or ILC3s because they did not express critical transcription factors for these lineages such as Tbx21, Eomes, or Rorc, but they did express Rora, Gata3, and Id2, suggesting that they had entered the ILC2 program. Because these cells expressed Id2 and had altered expression of numerous signaling proteins, we propose that they were selected for alterations that allow them to overcome the most severe requirements for ETS1 in ILC2 expansion, thus masking the critical targets of ETS1. One candidate factor that could drive the survival of these cells is IL-9, which is highly expressed in the in vitro–expanded Ets1^Δ/Δ cells (Turner et al., 2013). Future studies in which ETS1 is removed after the expansion of ILC2s may allow for more insight into the direct targets of ETS1 in these cells.

In some important aspects, the requirements for ETS1 in ILC2s parallel those in NK cells. Although ETS1 was required for the up-regulation of Id2 mRNA during NK cell and ILC2 development, the requirements for ETS1 in both lineages is likely to extend beyond Id2. Dysregulation of multiple surface receptors and intracellular signaling pathways also occur in ETS1-deficient NK cells and other lymphoid cells (Ramirez et al., 2012). However, the NK cells and B cells that develop in Ets1^−/− mice show hyperreactivity to cytokines and increased activation through some receptors, whereas Ets1^Δ/Δ ILC2s appear less activated and have reduced cellular expansion. Whether these distinctions reflect unique functions for ETS1 in ILC2s or distinct selective pressures placed on these cells in vivo remains to be determined. The availability of a conditional allele for Ets1 makes it feasible to test whether the hyperresponsiveness of NK cells and B cells in the absence of ETS1 is a consequence of selection or an intrinsic function for ETS1 in these mature cell types.

Ets1^f/f mice were generated in the University of Chicago Transgenic Core Facility using 129/SvJ embryonic stem cells. The offspring were backcrossed onto the C57BL/6 background for >12 generations. In brief, a targeting vector containing the floxed sites along with exons 8 and 9 containing the Ets1 binding domain was generated and introduced to germline DNA through homologous recombination. After backcrossing, the Ets1^f/f mice were crossed to Il7ra^cre mice, which had also been crossed onto the C57BL/6 background for >12 generations (Schlenner et al., 2010). C57BL/6 Rag1^−/− mice and CD45.1⁺ C57BL/6 mice were purchased through The Jackson Laboratory. All mouse lines were housed at the University of Chicago Animal Resource Center in accordance with the guidelines of the Institutional Animal Care and Use Committee.

Cells suspensions were incubated with an unlabeled purified CD16/32 (2.4G2.1) blocking antibody before the addition of any biotinulated or flourochrome-conjugated antibodies (FITC, PE, APC, PECY7, Percp-cy5.5, AF780, and Brilliant violet 421). Antibodies were purchased from eBioscience, BD, or BioLegend unless noted otherwise. Lineage cocktails for ILC2s included CD3 (145-2C11), CD4 (GK1.5), CD8 (2.43.1), TCR-β (H57-597), TCR-γδ (UC7-13D3), CD11c (N418), NK1.1 (PK136), GR1 (RB6-8C5), B220 (RA3-6B2), CD19 (1D3), and Ter119 (TER-119). ILC2s were identified using a combination of antibodies against SCA1 (D7), CD127 (A7R34), ICOS (7E.17G9), ST2 (DJ8; MD Biologicals), KLRG1 (2F1), and GATA3 (TWAJ). ILC2s were also stained for the expression of CD160 (7H1). To stain for CHILPs, BM was depleted using B220, Gr1, and Ter119. Cells were then stained for lineage markers (B220, Gr1, Ter119, CD3, CD19, and NK1.1), Flt3 (A2F1D), CD127, SCA1, α4β7 (LPAM1), and CD25 (PC61.5). NK cells and CLPs were stained as previously described (Ramirez et al., 2012).

BM ILC2s were sorted from WT mice. Total RNA was isolated using the RNeasy Micro kit (QIAGEN). cDNA was synthesized using Superscript III (Invitrogen). qPCR was performed using the following primers: Hprt, forward (5′-ACCTCTCGAAGTGTTGGATA-3′) and reverse (5′-CAACAACAAACTTGTCTGGA-3′); Ets1, forward (5′-CTGACCTCAACAAGGACAAGCC-3′) and reverse (5′-TTCCAGAAGAAACTGCCACAGC-3′); Id2, forward (5′-CACAGAGTACTTTGCTATCATTCG-3′) and reverse (5′-CCTGAACACGGACATCAGC-3′); Rora, forward (5′-TGATCGGACCAGCAGAAA-3′) and reverse (5′-CTTGGACATCCGACCAAAC-3′); Il24, forward (5′-GCCCAGTAAGGACAATTCCA-3′) and reverse (5′-ATTTCTGCATCCAGGTCAGG-3′); and Il6, forward (5′-CCGGAGAGGAGACTTCACAG-3′) and reverse (5′-GGAAATTGGGGTAGGAAGGA-3′). Samples were run in duplicate or triplicate and expression was measured relative to HPRT (hypoxanthine-guanine phosphoribosyltransferase).

CD45.2 Ets1^Δ/Δ, LMC, REts1^Δ/Δ, or RLMC BM cells were mixed with CD45.1 Rag1^+/+ BM cells at a 1:1 ratio. A total of 5 × 10⁶ BM cells were injected retroorbitally into lethally irradiated CD45.1 hosts. At 8 wk after reconstitution, the frequency of CD45.2⁺ cells to the CLP, ILC2, or CHILP populations was determined by flow cytometry.

Mice were injected with 0.4 µg IL-33 (BioLegend), 0.4 µg IL-25 (R&D Systems), or PBS for three consecutive days. On the fourth day, the mice were sacrificed and the lungs were analyzed for ILC2s. The lung tissue was minced into small pieces and placed into 10 ml DMEM supplemented with 10% FBS, 1× pen/strep, and DNase I. 1,500 U collagenase I (Gibco) was added to the media, and tissue was digested by shaking at 37°C for 30 min. Tissue was further digested by mechanical disruption to break up the remaining pieces before staining. Lymphocytes were isolated using a Percoll gradient and analyzed by flow cytometry from which the frequency or relative or total number of lung ILC2s or iILC2s was calculated.

Ets1^Δ/Δ and LMC mice were treated with IL-33 or PBS for three consecutive days before isolating the lungs as described in the previous section. Lymphocytes were stimulated in vitro with 1 µg/ml ionomycin and 20 ng/ml PMA in the presence of brefeldin A for 5 h followed by staining for ILC2 cell surface markers. Cells were then fixed and permeabilized using a cytofix/cytoperm kit (BD) and then stained for IL-13 (ebio13A) and IL-5 (TRFK5) before analysis by flow cytometry.

Sorted ILC2s were cultured for 2 wk on OP9-DL1 with 10 ng/ml each of IL-7 and IL-33 or on OP9-DL1 with 10 ng/ml of SCF and IL-7 and 2,000 U/ml of IL-2. OP9-DL1 was provided by J.C. Zuniga-Pflucker (University of Toronto, Toronto, ON, Canada). RNA from 100,000 cultured ILC2s was isolated using the RNeasy Micro kit, and cDNA was prepared to probe Affymetrix MOE 430_2 arrays (GEO accession number GSE79742) as previously described (Dias et al., 2008). Raw data were normalized using RMAExpress.

Statistical differences between groups were calculated using Student’s t test, with p-values <0.05 considered significant. All statistics were performed using Prism 5 (GraphPad Software). All error bars are SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Fig. S1 shows that ETS1 was required for the development of BM ILC2s. Table S1 is provided as an Excel file and shows the gene expression in LMC and Ets1^Δ/Δ Lin⁻ICOS⁺ cells expanded in vitro.

We would like to thank the members of the Kee, Fu, and Bendelac laboratory for their helpful discussions.

This work was funded by National Institutes of Health grants R01 AI106352 and R21 AI115388 (to B.L. Kee) and RO1 HL071063 (to E.C. Svensson) and by the Rafael Rivera Asthma Research Fund (to K. Ramirez and B.L. Kee), as well as National Institutes of Health training grants T32AI007090 (to E.C. Zook and K. Ramirez) and F32CA177235 (to E.C. Zook). B.L. Kee was supported by the Leukemia and Lymphoma Society.

The authors declare no competing financial interests.

Author contributions: E.C. Zook, K. Ramirez, X. Guo, and G. van der Voort performed experiments. M. Sigvardsson performed the microarray. E.C. Svensson designed and generated the Ets1 conditional mice. E.C. Zook, K. Ramirez, Y.-X. Fu, and B.L. Kee designed, analyzed, and interpreted the experiments. E.C. Zook and B.L. Kee wrote the manuscript.