Runx/Cbfβ complexes protect group 2 innate lymphoid cells from exhausted-like hyporesponsiveness during allergic airway inflammation

Group 2 innate lymphoid cells (ILC2s) have tissue-resident competence and contribute to the pathogenesis of allergic diseases. However, the mechanisms regulating prolonged ILC2-mediated TH2 cytokine production under chronic inflammatory conditions are unclear. Here we show that, at homeostasis, Runx deficiency induces excessive ILC2 activation due to overly active GATA-3 functions. By contrast, during allergic inflammation, the absence of Runx impairs the ability of ILC2s to proliferate and produce effector TH2 cytokines and chemokines. Instead, functional deletion of Runx induces the expression of exhaustion markers, such as IL-10 and TIGIT, on ILC2s. Finally, these ‘exhausted-like’ ILC2s are unable to induce type 2 immune responses to repeated allergen exposures. Thus, Runx confers competence for sustained ILC2 activity at the mucosa, and contributes to allergic pathogenesis.

I nnate lymphoid cells (ILCs) are enriched in mucosal tissues, where they function as sentinel cells at the front line of host defense 1 . Although ILCs do not possess rearranged antigenspecific receptors, they exert a helper function similar to T H cells by producing helper cytokines. ILCs are categorized into three main subsets : T H 1-like ILC1s, T H 2-like ILC2s, and T H 17/T H 22like ILC3s [2][3][4][5][6] . Recently, another subset of ILCs named regulatory ILCs (ILCregs) has been reported to provide an immune suppressive function by producing IL-10 in the intestine 7 .
ILC2s are the main population producing IL-5, which recruits eosinophils into tissues under healthy conditions 8 . Upon allergic stimulation, ILC2s are activated by IL-25, IL-33, and TSLP from damaged epithelial cells, IL-2, IL-4, and IL-9 from other haematopoietic cells or from ILC2s themselves, neuropeptides, and lipid mediators 1,[9][10][11] . Activated ILC2s contribute to deterioration of allergic diseases by producing high levels of IL-5 and IL-13, both of which enhance the T H 2 induction and inflammation mediated by eosinophils. An ILC2 subset producing IL-10 (ILC2 10 s) in regions of chronic or severe allergic inflammation is associated with reduction of eosinophils in the lung by unknown mechanisms 12 .
Recurrent stimulation influences the biological properties of ILC2s, as well as T cells. After the effector phase, T cells can become long-lived memory T cells in the tissues or lymph nodes, where they are reactivated by the same antigen. A similar recall response was also observed in ILC2s pre-activated with IL-33 or allergens 13 . In contrast, T cells at sites of chronic inflammation become exhausted and lose their effector functions, including cytokine production and proliferation, in response to repeated stimulation 14 . PD-1, which is a T cell exhaustion marker, is induced on activated ILC2s and negatively regulates this cell pool 15 . However, PD-1 + ILC2s are not considered exhausted because they continue to produce IL-5 normally. Thus, ILC2s with a hyporesponsive phenotype similar to exhausted T cells have not yet been identified.
The mammalian Runx transcription factor protein family is composed of Runx1, Runx2, and Runx3. Each Runx protein requires heterodimer formation with Cbfβ to bind DNA 16 . Runx3 is the main family member expressed in all ILC subsets and is indispensable for the differentiation and function of the ILC1 and ILC3 subsets 17 . However, depletion of Runx3 alone has little effect on ILC2 differentiation, probably due to the redundant functions of other Runx proteins, such as Runx1, which is expressed in ILC2s. Thus, the function of Runx/Cbfβ complexes in ILC2s has not been clarified.
Here, we show that Runx/Cbfβ complexes are not necessary for ILC2 differentiation but modulate ILC2 function. At steady state, Runx-deficient ILC2s are activated and aberrantly secrete IL-5, resulting in increased eosinophil recruitment to the lung. However, after allergic stimulation, ILC2s lacking Runx fail to proliferate and produce various cytokines and chemokines but have increased expression of IL-10 and TIGIT, which are known markers of exhausted T cells. We explore the existence of IL-10 + TIGIT + ILC2s with low reactivity in the physiological setting and find that severe subacute allergic inflammation induces the emergence of hyporesponsive IL-10 + TIGIT + ILC2s, and that this effect is enhanced by Cbfβ deficiency. Collectively, our data reveal that Runx/Cbfβ complexes are required to prevent ILC2s from entering an exhausted-like functional state under allergic conditions.

Results
Runx is not required for development of ILC2s. Of all of the ILCs and ILC progenitors, the highest Runx1 and Runx3 mRNA expression levels are found in the common precursor to ILCs (ILCPs), which is marked by stage-specific PLZF expression and can differentiate into ILC1s, ILC2s, and NCR + ILC3s (a subpopulation of ILC3s) 17 . Analysis of Runx3 reporter mice suggests that downregulation of Runx3 may be required for PLZF + ILCPs to enter the ILC2 pathway, whereas ILC1s and ILC3s require intermediate to high levels of Runx3 for their differentiation 17 . To precisely examine Runx1 protein expression in ILC subsets and progenitors, we took advantage of Runx1 +/P1-GFP: P2-RFP mice, in which GFP or RFP was driven from the distal (P1) or proximal (P2) Runx1 promoter, respectively 18 . PLZF + ILCPs utilized both the P1 and P2 promoters for high Runx1 expression, although ILC2s in the lung and intestine expressed Runx1 from the P1 promoter to a greater extent than ILC1s and ILC3s in the intestine ( Fig. 1a and Supplementary Fig. 1). Thus, Runx1 is expressed by ILC2s despite low Runx3 expression.
To assess the roles of Runx/Cbfβ complexes in the function and differentiation of PLZF + ILCPs into ILC2s, we deleted Cbfβ in the PLZF + ILCPs using PLZF-Cre (Cbfb f/f PLZF-Cre mice). This strategy should result in a complete loss of any Runx protein function in the descendant ILC subsets, including ILC1s, ILC2s, and NCR + ILC3s. The Cbfb f/f PLZF-Cre mice were born and grew to adults, although they were smaller than their littermate controls ( Fig. 1b) and had bone distortion and difficulty walking. The defects in bone formation may be explained by the loss of Runx2, which is critical for osteoblast differentiation, because PLZF is expressed in osteoblasts prior to Runx2 induction 19 . Then, we examined which lymphocytes suffered from the Cbfβ mutation in the Cbfb f/f PLZF-Cre mice to validate our system. We observed highly efficient deletion of Cbfb in the PLZF + ILCP and ILC2 progenitors (ILC2Ps) from the bone marrow, ILC2s from the lung and small intestine, and ILC1s from the small intestine and partial Cbfb deletion in NCR + ILC3s from the small intestine in the Cbfb f/f PLZF-Cre mice (Fig. 1c). However, the Cbfb genes in the CD4 + T, CD8 + T, NK, and B cells were not greatly affected in the Cbfb f/f PLZF-Cre mice. Wild type Cbfb transcripts were efficiently deleted in the ILC subsets of the Cbfb f/f PLZF-Cre mice, although subtle transcript expression of mutated Cbfb was detected in the non-ILC populations (Fig. 1d, e). Thus, the Cbfβ dysfunction is specifically induced in ILC subsets among haematopoietic cell populations of the Cbfb f/f PLZF-Cre mice.
The specificity of this deletion effect for ILC subsets is surprising, because haematopoietic stem cells are fate-mapped by PLZF expression 20 . To confirm previous data, we crossed Cbfb f/f PLZF-Cre mice with Rosa26-tdTomato mice in which PLZF-Cre expression in the progenitor cells could be followed by tdTomato expression in the Cbfb f/f PLZF-Cre mice. As previously described, most haematopoietic cells were labeled with tdTomato in the mice ( Supplementary Fig. 2). However, the Cbfb gene locus flanked by loxps was quite intact in the tdTomato + cells of the major haematopoietic cell populations. These data indicate that PLZF-Cre can reach and excise the Rosa26 locus but not the Cbfb locus in the progenitors of the major haematopoietic populations, probably due to the tight chromatin structure of the Cbfb locus.

ARTICLE
Cbfb f/f PLZF-Cre ILC2s in the recipient lungs showed increased KLRG1 expression, decreased Thy1 expression, and IL-5 overproduction compared to those of the competitor cells (Fig. 2e).
These data indicate that Cbfβ suppresses the basal activity of ILC2s in a cell intrinsic manner.
To address the question of which Runx proteins contribute to the ILC2 activation phenotype in the absence of Cbfβ, we sought to generate Runx1 f/f PLZF-Cre and Runx3 f/f PLZF-Cre mice. However, most of the Runx1 f/f PLZF-Cre mice died soon after birth for unknown reasons. Therefore, instead we generated Cbfb f/f ERT2-Cre, Runx1 f/f ERT2-Cre, Runx3 f/f ERT2-Cre, and Runx1 f/f Runx3 f/f ERT2-Cre mice. Inducible deletion of Cbfβ or both Runx1 and Runx3 by oral tamoxifen administration led to a robust reduction of Thy1 expression and an increase of IL-5 production in the intestine (Fig. 2f, g). On the other hand, single deletion of either Runx1 or Runx3 had little effect on Thy1 expression and IL-5 production by the ILC2s (Fig. 2f, g). These results suggest that Runx1 and Runx3 must work jointly and together with Cbfβ to repress ILC2 functions under steady-state conditions.
To investigate the physiological impact of enhanced ILC2 activity resulting from Cbfβ deficiency on the steady state of the lung, we adoptively transferred Cbfb f/f ILC2s or Cbfb f/f ERT2-Cre ILC2s into Rag2 -/-Il2rg -/mice, which were treated with tamoxifen after transfer. At 3-4 weeks after the tamoxifen treatment, almost no damaged PAS + epithelial cells were observed in the lungs of the Cbfb f/f ERT2-Cre ILC2 recipient mice, although the Cbfb f/f ERT2-Cre ILC2s increased the eosinophil numbers in the bronchoalveolar lavage (BAL) fluid (Fig. 2h, i). Thus, Cbfβdeficient ILC2s are not sufficiently active to acutely damage lung epithelial cells but instead contribute to subsymptomatic eosinophil infiltration into the bronchoalveolar space.
Runx antagonizes GATA-3 function in steady-state ILC2s. A previous study demonstrated that Runx proteins antagonized GATA-3 function in T cells by directly binding to GATA-3 23 . We hypothesized that the same inhibitory mechanism by Runx proteins might function in ILC2s to suppress IL-5 production by antagonizing GATA-3 activity. To test this hypothesis, we performed RNA sequence analysis of ILC2s from the lungs of Cbfb +/f PLZF-Cre and Cbfb f/f PLZF-Cre mice and analyzed the expression profiles of genes that were positively or negatively Cbfb f/f ERT2-Cre Recipients of Cbfb f/f ILC2s In c, d, g, h, *p < 0.05 and **p < 0.01 by Student's t-test. Data are representative of at least two independent experiments (mean ± s.d. of four mice in c, d, mean ± s.d. of three mice in g, h) regulated by GATA-3 in ILC2s 24 . If our hypothesis was correct, the function of GATA-3 as a transcription factor would be enhanced in the absence of Cbfβ. Deletion of Cbfβ in ILC2s led to upregulation of 18 of 156 genes (11.5%) that were positively regulated by GATA-3 and downregulation of 30 of 151 genes (19.8%) that were negatively regulated by GATA-3 ( Fig. 3a, Supplementary Data 1). GATA-3 positively regulated IL-5 expression, and this effect was augmented by Cbfβ deficiency. In contrast, GATA3 was a negative regulator of Thy1 expression, which was even more inhibited in the absence of Cbfβ. In addition, a set of genes positively or negatively regulated by GATA-3 was significantly enriched in the Cbfβ-deficient ILC2s (Fig. 3b). Given that Runx/Cbfβ complexes antagonize GATA-3 function, over-expression of Runx protein should suppress IL-5 production by ILC2s. To explore this possibility, CD45.2 + C57BL/6 bone marrow cells were transduced with a retroviral vector encoding Runx3, followed by IRES-Thy1.1 and were adoptively transferred into lethally irradiated CD45.1 + congenic mice. At 8 weeks after transfer, the Thy1.1 + ILC2s over-expressing Runx3 produced less IL-5 than the Thy1.1non-transduced cells or the Thy1.1 + cells transduced with the control vector ( Fig. 3c, d). With the same over-expression system, we confirmed that GATA-3 overexpression in ILC2s increased IL-5 production, which was inhibited by Runx3 ( Supplementary Fig. 4). Collectively, the Runx/Cbfβ complexes suppress the constitutive activity of ILC2s at least in part by inhibiting the function of GATA-3.
To determine how Cbfβ regulates ILC2 effector functions after activation, first we cultured ILC2s from the lungs of Cbfb +/f PLZF-Cre and Cbfb f/f PLZF-Cre mice with IL-2 and IL-33, which is a cytokine combination that is the most potent ILC2 stimulator and is critical for establishment of allergic inflammation. We expected that this cytokine stimulation would lead to unleashed production of T H 2 cytokines by activated ILC2s due to Cbfβ deficiency. Surprisingly, ILC2s lacking Cbfβ secreted decreased IL-5 and IL-13 levels and did not grow well (Fig. 4a, b) in response to IL-2 and IL-33 in vitro. This hyporesponsiveness of ILC2s did not occur when the basal ILC2 activity was maintained by IL-2 and IL-7 without IL-33, which is a strong inducer of allergy (Fig. 4c). To understand the mechanism of the low reactivity of Cbfβ-deficient ILC2s to IL-33, we conducted RNA sequencing analysis of in vitro-activated lung ILC2s from Cbfb +/f PLZF-Cre and Cbfb f/f PLZF-Cre mice ( Fig. 4d- Genes Genes Genes negatively regulated by GATA-  receptors, including IL7r, Il9r, IL4ra, Icos, and Nmur1, which all show reduced expression in the absence of Cbfβ (Fig. 4e, g). In contrast, Cbfβ-deficient ILC2s expressed high levels of T cell exhaustion markers, including the Tnfrsf18 encoding GITR, Klrg1, Tigit, Prdm1 encoding Blimp1, Il10, Ctla4, and Lag3 genes (Fig. 4e, h) 14 . Gene set enrichment analysis indicated that Cbfβdeficient ILC2s stimulated with IL-33 had a signature of exhausted CD8 + T cells (Fig. 4i). Since the TIGIT and IL-10 expression levels were low in the control ILC2s, we thought that upregulation of TIGIT and IL-10 might be good marker for this hyporesponsiveness and confirmed the elevated expression of the IL-10 and TIGIT proteins in activated-ILC2s lacking Cbfβ in vitro (Fig. 4j, k). Collectively, ILC2s present in the lungs of the Cbfb f/f PLZF-Cre mice show hyporesponsiveness to cytokine stimulation in vitro and acquire unique gene expression signatures similar to those observed in exhausted T cells.
IL-10 and inhibitory signals through TIGIT could be responsible for the low reactivity of the Cbfb f/f PLZF-Cre ILC2s against IL-33 stimulation. However, the IL-10 concentration in the culture supernatant of Cbfb f/f PLZF-Cre ILC2s was not high enough to detect the slight inhibitory effect on ILC2s ( Supplementary Fig. 5a, b). Neutralization of IL-10 did not cancel the low reactivity of the Cbfb f/f PLZF-Cre ILC2s against IL-33 ( Supplementary Fig. 5c). In addition, our RNA sequence analysis showed that the expression levels of TIGIT ligands, such as CD112 and CD155, were quite low in ILC2s; the RPKM values of these molecules were less than 1. Therefore, the low reactivity of ILC2s without Cbfβ function did not result from increased IL-10 or TIGIT expression.
To determine what Runx proteins were responsible for this hyporesponsiveness to cytokine stimulation, we deleted Cbfβ, Runx1, Runx3, or both Runx1 and Runx3 by oral tamoxifen administration to Cbfb f/f ERT2-Cre, Runx1 f/f ERT2-Cre, Runx3 f/f ERT2-Cre, or Runx1 f/f Runx3 f/f ERT2-Cre mice as described above and stimulated the intestinal ILC2s with IL-2 and IL-33 in vitro. Runx1 or Runx3 single deletion resulted in minor changes in IL-5 and IL-13 production by the ILC2s (Supplementary Fig. 6a, b). However, when Runx1 and Runx3 were both deleted, IL-5 and IL-13 production were both reduced to the level observed in the Cbfβ-deleted ILC2s. In addition, we performed ChIP sequence analysis for Runx1 and Runx3 binding in ILC2s activated with IL-33 to examine the differential functions of

Exhaustion marker
Cbfb +/f PLZF-Cre    Runx1 and Runx3 in the hyporesponsiveness of ILC2s. However, generally the binding patterns were comparable between Runx1 and Runx3 ( Supplementary Fig. 6c, d). Collectively, both Runx1 and Runx3 serve as inhibitors of the exhausted-like phenomenon in cooperation with Cbfβ.
GATA-3-dependent and GATA-3-independent functions of Runx in ILC2s. We sought to investigate how pre-activated Cbfb f/f PLZF-Cre ILC2s at steady state showed hyporesponsiveness to IL-33 stimulation. First, we hypothesized that GATA-3 overactivation due to the absence of antagonizing effects by the Runx protein pushed ILC2s into an overactivated hypofunctional state. However, over-expressed GATA-3 still enhanced IL-5 and IL-13 production by ILC2s in response to IL-33 in vitro (Fig. 5a,  b). To examine whether the Runx proteins inhibited ILC2 activity by antagonizing GATA-3 as observed in the steady-state ILC2s, we evaluated cytokine production by IL-33-stimulated ILC2s over-expressing Runx3. Interestingly, Runx3 over-expression dampened cytokine production by ILC2s in response to IL-33. Thus, the Runx proteins apparently inhibit ILC2 activity in a dose-dependent manner even after IL-33 stimulation.
Since high transcription factor expression is not always required for enhancer or repressor function, we hypothesize that a dose-independent and GATA-3-independent function of Runx proteins should exist as an epigenetic modulator for ILC2s to normally respond to IL-33. To test this possibility, we examined Cbfβ binding in ILC2s cultured with or without IL-33 by ChIP sequence analysis and found that new Cbfβ binding peaks appeared in ILC2s cultured with IL-33 (Fig. 5c). The GATA-3 motif was not ranked in at least the top ten motifs of the Cbfβ binding peaks specific for ILC2s stimulated with IL-33 (Fig. 5d). Furthermore, the IL-33-specific Cbfβ binding peaks were located at genes for ILC2 function, including Il5, Il13, Lta, Nmur1, and Vipr2, and genes for exhaustion markers, including Il10, Tigit, Prdm1, Lag3, and Ctla4 (Fig. 5e). Examples of these GATA-3independent Cbfβ biding peaks are shown in Fig. 6a. Thus, Cbfβ binding specific to IL-33 stimulation is independent of GATA-3 and is associated with a gene expression profile of exhausted-like ILC2s, suggesting a GATA-3-independent function of the Runx proteins in ILC2 reactivity against IL-33.
Enhancer or repressor functions of Runx in activated ILC2s.
We assessed the possibility that Runx/Cbfβ complexes bound to enhancers of gene loci related to ILC2 activity and repressors of exhaustion marker gene loci in ILC2s stimulated with IL-33. For this purpose, we determined which Cbfb binding peaks were marked by H3K27 acetylation for enhancer regions or H3K27 trimethylation for repressed gene regions in ILC2s stimulated with IL-33 (Fig. 6a, b and Supplementary Data 2). Cbfβ binding peaks overlapping with H3K27 acetylation peaks were associated with ILC2 functional genes that were downregulated in the hyporesponsive Cbfβ-deficient ILC2s (Fig. 4e). Gene ontology analysis of the genes with both Cbfβ and H3K27 acetylation peaks indicated that Cbfβ globally regulated genes involved in positive regulation of proliferation, cytokine production, cytokinemediated signaling, leukocyte migration, and leukocyte adhesion (Fig. 6c). In contrast, the H3K27 trimethylation status of the exhaustion marker genes was variable. Among the genes listed as exhaustion markers in Fig. 4h, H3K27 trimethylation was observed in the Il10, Prdm1, and Ctla4 loci. Since IL-10 is a good marker for exhausted-like ILC2s, we confirmed the reduced H3K27 trimethylation level in the Il10 promoter region of Cbfb f/f  Runx deficiency in ILC2s ameliorates allergic inflammation. ILC2 10 s can be found in vivo during chronic or severe inflammation 12 . To identify TIGIT + IL-10 + ILC2s in physiological settings, we took advantage of a severe subacute asthma model with IL-10-Venus reporter mice administered a high dose of papain every three days. On day 7 after administration of three papain doses, ILC2s producing IL-10 were observed in the BAL fluid and lung (Fig. 7a, b). However, TIGIT + IL-10 + ILC2s were found only in the BAL fluid, which was the site of severe inflammation, but not in the lung. Since most TIGIT + ILC2s expressed IL-10-Venus, TIGIT + ILC2s can represent TIGIT + IL-10 + ILC2s. The TIGIT + ILC2s in the BAL fluid are a small population of activated ILC2s that are marked by PD-1, GITR, and KLRG1 expression 15,25 and are not inflammatory ILC2s due to normal ST2 expression (Fig. 7c) 22 . TIGIT + ILC2s did not proliferate well, as determined by diminished Ki67 staining and were negative for a dead cell marker (Fig. 7d, e). The RT-PCR assay indicated lower Il5 and Il13 expression in the TIGIT + ILC2s than in the TIGIT -ILC2s. The hyporesponsive TIGIT + ILC2s emerged in the lung, as well as in the BAL fluid, when the mice were treated with papain every three days for a month (Supplementary Fig. 7a). Furthermore, to examine whether the TIGIT + IL-10 + ILC2s in the BAL fluid were similar to the hyporesponsive Cbfβ-deficient ILC2s, we performed RNA sequence analysis by barcoding individual cDNA molecules from one hundred TIGIT + and TIGIT -ILC2s. The sample number (n = 3) and assay sensitivity were not sufficient to observe a significant reduction of Il5 and Il13 expression in the TIGIT + ILC2s, although the Il5 and Il13 transcript levels were significantly decreased in the TIGIT + ILC2s based on a sensitive RT-PCR analysis (Fig. 7f, Supplementary Fig. 8, and Supplementary Data 3). Despite these limitations, the gene set enrichment analysis showed that the signature genes of TIGIT + ILC2s in the BAL fluid were enriched in the low-reactive Cbfβ-deficient ILC2s. Thus, severe allergy induced hyporesponsive TIGIT + IL-10 + ILC2s similar to Cbfβ-deficient ILC2s.
To investigate the impact of increased exhausted-like ILC2s resulting from Cbfβ deficiency on inflammation, we created a subacute asthma model with a high dose of papain using congenic mice adoptively transferred with bone marrow cells from either Cbfb +/f PLZF-Cre or Cbfb f/f PLZF-Cre mice. As expected, the recipients of the Cbfb f/f PLZF-Cre bone marrow cells had reduced eosinophils and ILC2s in the BAL fluid and lungs (Fig. 8a), which were accompanied by a reduction in IL-5 and IL-13 production in the BAL fluid (Fig. 8b), as well as less immune cell infiltration around the bronchi (Fig. 8c, d) than those of the Cbfb +/f PLZF-Cre cell recipients. Furthermore, ILC2s without Cbfβ function in the lung generated an increased TIGIT + fraction and produced less IL-5 and IL-13 but more IL-10 than the control ILC2s ( Fig. 8e-g). However, the increased IL-10 from the Cbfβ-deficient ILC2s did not seem to be responsible for the reduced allergic inflammation, because the IL-10 concentration of the BAL fluid was not increased in the papain-treated recipients of the Cbfb f/f PLZF-Cre bone marrow cells, probably due to higher IL-10 production by lineage + cells than by ILC2s (Figs. 7a, 8b). Cellintrinsic hyporeactivity of Cbfb f/f PLZF-Cre ILC2s was also confirmed by bone marrow competition (Fig. 8h). Thus, the exhausted-like phenomenon in ILC2s was associated with decreased inflammation mediated by eosinophils.
To determine the function of Cbfβ-deficient ILC2s in the chronic allergy model, mice were continuously administered a high dose of papain every three days for one month as a severe chronic allergy model (Supplementary Fig. 7c) or a high dose of papain every three days three times followed by the same course H3K27me3 binding to Il10 promoter (% input) Cbfb f/f of papain treatment after a two week cessation period as a repeated chronic allergy model ( Supplementary Fig. 7d). Cbfβdeficient ILC2s did not respond well to papain stimulation in the model mice. However, the hyporesponsive Cbfb f/f PLZF-Cre ILC2s were associated with reduced eosinophil recruitment only after a 2nd course of papain challenge. Thus, Runx/Cbfβ complexes in ILC2s play a critical role in inflammatory responses to repeated allergen stimulation.
To further determine the effect of Cbfβ deficiency in ILC2s on allergic inflammation, we adoptively transferred Cbfβ-deficient ILC2s prepared by Cbfβ deletion during in vitro culture of lung ILC2s into Rag2 -/-Il2rg -/mice lacking both acquired immunity and innate lymphoid cells. Prior to the adoptive transfer, we confirmed acquisition of an exhausted-like phenotype including high IL-10 and TIGIT expression with low IL-5 and IL-13 production due to Cre-mediated conversion of the floxed allele (Fig. 9a, b). The Rag2 -/-Il2rg -/recipients were intranasally administered papain for 3 consecutive days. Analysis of these mice on day 3 revealed that the transferred Cbfβ-deficient ILC2s were less capable of recruiting eosinophils to the bronchoalveolar space and the lung (Fig. 9c), with a decreased amount of IL-5 in the BAL fluid and increased epithelial damage (Fig. 9d-f). These data indicate that Cbfβ is required for the ability of ILC2s to trigger and extend allergic inflammation and to prevent them from falling into an exhausted-like state in inflamed tissues.

Discussion
We have shown here that Runx/Cbfβ complexes suppress the basal activity of ILC2s, which can recruit eosinophils through IL-5 production under steady-state conditions. However, under allergic conditions, Runx/Cbfβ complexes support the ability of ILC2s to exert their helper functions for type 2 immunity. Diminished activity of ILC2s lacking Runx under allergic inflammatory conditions was accompanied by increased expression of T cell exhaustion markers, such as IL-10 and TIGIT. Mechanistically, Runx/Cbfβ complexes contribute to epigenetic modification for a gene expression profile of the exhausted-like ILC2s. These TIGIT + IL-10 + ILC2s were identified even in mice with allergen-induced subacute or chronic inflammation. Finally, transferred ILC2s with exhausted-like characteristics lacked an appropriate capacity for allergic immune responses in vivo. Thus, our results revealed an essential regulation of ILC2 function by Runx proteins and an accelerated emergence of exhausted-like ILC2s in the absence of Runx/Cbfb complexes. We used PLZF-Cre mice to induce dysfunction of Cbfβ in ILC2s. We obtained a Cbfβ deletion that was rather specific to ILC subsets in Cbfb f/f PLZF-Cre mice, although PLZF-Cre is expressed in unknown progenitor cells for most haematopoietic cells. We cannot completely deny the possibility that any small population of haematopoietic cells may be affected in the Cbfb f/f PLZF-Cre mice. Therefore, using PLZF-Cre mice for an ILC study is quite risky and requires precise examination of the deletion effect on the haematopoietic populations, as described in our paper.
Since both TIGIT and IL-10 are induced by IL-33 12 , a strong IL-33 signal should be required for the generation of TIGIT + IL-10 + ILC2s. The hyporesponsive TIGIT + IL-10 + ILC2s are apparently part of the activated ILC2s, because they express a series of activation markers, including PD-1 and KLRG1, which are also known inhibitory molecules. T cells become exhausted through interaction of their inhibitory receptors, such as CTLA-4 and PD-1, with cognate ligands on other cells 14 . Therefore, we can reasonably assume that the low reactivity of TIGIT + IL-10 + ILC2s results from accumulated signals through those inhibitory molecules. Although we did not test the function of the individual inhibitory molecules on the TIGIT + IL-10 + ILC2s, we clearly showed that the emergence of hyporeactive TIGIT + IL-10 + ILC2s was negatively regulated by Runx/Cbfβ complexes.
Chronic or severe inflammation is required to induce TIGIT + IL-10 + ILC2s. If ILC2s have functional Runx proteins, then hyporesponsive TIGIT + IL-10 + ILC2s are rare even after continuous inhalation of papain for one month. Furthermore, Cbfb f/f PLZF-Cre ILC2s with low reactivity were not associated with attenuation of chronic allergy. These data indicate that TIGIT + IL-10 + ILC2s and ILC2s themselves play a limited role in the pathology of chronic inflammation. However, hyporesponsive ILC2s lacking Cbfβ protect the host from exaggerated allergic inflammation when repeatedly treated with allergen. Therefore, regulation of ILC2 functions by Runx/Cbfb complexes is critical for the pathogenesis of acute exacerbation of chronic allergy. ILC2 10 s are identified by low Tnf, Lta, Il2, Retnla, and Ccl1 expression and high Il10 and Id3 expression compared to the expression levels of the non-IL-10 producers 12 . Cbfβ-deficient ILC2s are somewhat similar to ILC2 10 s in that Tnf, Lta, Retnla and Ccl1 expression is reduced in both cell types and IL-10 production is increased. However, Cbfβ ablation resulted in more global defects in ILC2 function, because ILC2s lacking Cbfβ poorly expressed the main effector cytokines, such as Il5 and Il13, whereas ILC2 10 s produce large amounts of IL-5 and IL-13 12 . Furthermore, the expression of T cell exhaustion markers is not increased on ILC2 10 s, whereas the high Id3 expression observed in ILC2 10 s is not induced in Cbfβ-deficient ILC2s. Thus, the hyporesponsive ILC2s lacking Cbfβ seem to be different from ILC2 10 s. Immunosenescence is another hyporesponsive state of immune cells that results from the effects of ageing. Cell cycle arrest related to telomere shortening or DNA damage is thought to be a common feature of cellular senescence 26 . Runx1 is involved in age-related changes in haematopoietic stem cells 27 . The regenerative capacity of HSCs declines following conditional ablation of Runx1 after an initial expansion 28 . In our study, we utilized a subacute or repeated allergy model to induce ILC2 hyporesponsiveness. Testing whether aging could be a trigger for senescent ILC2s and whether the absence of Runx function was involved in ILC2 senescence would be fascinating.
The concept of exhausted-like ILC2s provides a better understanding of normal ILC2 physiology. Many mouse models have successfully demonstrated the importance of ILC2s during the acute phase of an allergic response 29 . In addition, an important role of ILC2s in worsening of recurrent allergy has been suggested. ILC2s can also be trained by the initial allergic stimulation, resulting in production of higher T H 2 cytokine levels after a second challenge 13 . The immune system always has delicate balancing mechanisms to maintain homeostasis. Exhausted-like hyporesponsiveness is one mechanism by which immune cells are prevented from overactivation and continuous production of harmful inflammatory cytokines. Therefore, ILC2s may possess a capacity to reduce their activity similar to that of exhausted T cells after chronic exposure to allergic stimuli.
The frequency and activity of ILC2s are both increased in patients with chronic type 2 inflammation, including asthma and atopic dermatitis and are also associated with the disease severity [30][31][32] . Given a possible role of chronic inflammation in tissues in providing environmental cues to induce low reactive ILC2s, an examination of whether hyporesponsive ILC2s are present in and beneficial for patients with chronic allergy will be important, especially in cases of acute disease exacerbation. Our study in mice revealed that attenuated Runx function accelerated the differentiation of ILC2s towards a hyporesponsive state. Further understanding of the molecular switch triggering exhausted-like ILC2s may provide a basis for the development of therapeutic targets to suppress allergic immune responses. Cell preparation. Cells were isolated from the spleen, liver, BAL fluid, lung, and small intestine as previously reported 17 . The spleen and liver were dissected and smashed through a 70 µm strainer. Lymphocytes from the liver were resuspended in 40% Percoll and centrifuged at 2000 rpm at room temperature for 20 min. Cells at the bottom of the tube were collected for flow cytometry after red blood cell (RBC) lysis. BAL fluid was obtained by infusion of PBS through a catheter. The dissected lung was cut into small pieces, chopped with a razor blade, and incubated in 8 mL of digestion buffer containing RPMI medium with 2% FBS, 0.5 mg per mL of Collagenase IV (Sigma, C5138), and 0.05 mg per mL of DNase (Wako, 043-26773) at 200 rpm and 37°C for 45 min. Digested cells were smashed through a 70 µm strainer and used for flow cytometry and cell culture after RBC lysis. Lung lymphocytes were further purified with a 40 and 80% Percoll gradient for sorting. Intraepithelial lymphocytes (IELs) and LPLs were isolated from the small intestine. After removing feces and Peyer's patches, the small intestine was incubated in 20 mL of RPMI medium with 2% FBS and 5 mM EDTA at 200 rpm and 37°C for 20

Methods
Cbfb +/f PLZF-Cre 20 min. After vigorous vortexing, floating cells were collected as the IELs. The remaining tissues were cut into small pieces, chopped with a razor blade, and incubated in 20 mL of the same digestion buffer at 200 rpm and 37°C for 30 min. Digested cells containing LPLs or the IELs collected above were purified with a 40 and 80% Percoll gradient for flow cytometry, culture and sorting. Lymphocytes from the lung and intestine were cultured with GoldiPlug without any stimulation to assess ex vivo production of IL-5 and IL-13 or with PMA (50 ng per mL) and Ionomycin (0.5 µg per mL) to analyze cytokine production by ILC2s from mice treated with papain for 4 h. ILC2s were sorted from the lung with the FACSAria (BD Biosciences). Five hundred ILC2s were cultured with IL-2 (10 ng per mL), IL-33 (10 ng per mL), and with/without IL-7 (10 ng per mL). For inducible deletion of Cbfb, 1 µM of 4-hydroxytamoxifen (4OHT) was added to the cell culture for 4 days.
Antibodies and flow cytometry. Cells were blocked with an anti-CD16/32 antibody (2.4g2) first and then stained with Fixable Viability Dye eFluor 506 (eBioscience) for detection of dead cells before staining the cell surfaces. The antibodies used for flow cytometry are all listed in Supplementary Table 1. The Cytofix/Cytoperm Buffer Set (BD Biosciences) and Foxp3/Transcription Factor Staining Buffer Set (eBioscience) were used for staining of intracellular cytokines and transcription factors, respectively, according to the manufacturers' protocols. Data were acquired on a FACSCanto II (BD Biosciences) and analyzed with the FlowJo software (TreeStar). Gating and sorting strategies were described in Supplementary Fig. 9.
PCR and RT-PCR. Cells were directly sorted into DNA extraction buffer (0.1 M Tris-HCL pH 7.5, 0.05 M EDTA pH 8.0, and 1.25% SDS) for DNA or TRIzol (Thermo Fisher Scientific) for RNA by the FACSAria. DNA was purified from the cell lysates with Phenol-Chloroform Isoamyl Alcohol, followed by ethanol precipitation. RNA was purified, and cDNA was synthesized with the PrimeScript™ 1st Strand cDNA Synthesis Kit (Takara) following the manufacturer's protocol. Then, the mRNA transcripts were quantified with TB Green™ Premix Ex Taq™ II (Takara). The relative expression of the indicated transcripts was calculated by the 2 −ΔCt method and normalized to Actb expression. The copy numbers of Cbfb transcripts with or without a floxed region were calculated by quantitative RT-PCR. The specific primers are listed in Supplementary Table 2.
Papain-induced asthma model. After irradiation at 950 rad, CD45.1 + congenic mice were transferred with 1 × 10 7 bone marrow cells from either Cbfb +/f PLZF-Cre or Cbfb f/f PLZF-Cre mice. At 12 weeks after transfer, 100 µg of papain in 50 µL of sterile PBS was intranasally administered to the recipient mice every 3 days. On day 7 or 14, a 20-G catheter was inserted into the trachea, and 1 mL of PBS was infused into the lung through the catheter and aspirated 3 times. The first BAL fluid sample was used to detect IL-5, IL-10, and IL-13 by ELISA. Cells in the whole BAL fluid were analyzed by flow cytometry. After making a knot at the left main bronchus, the left lung was removed to isolate cells for flow cytometry. The right lung was intratracheally infused with 1 mL of 10% formalin, removed, and incubated in 10% formalin at 4°C overnight to make paraffin blocks. HE and PAS staining were performed as previously reported 17 . To expand ILC2s from the lungs of the Cbfb f/f and Cbfb f/f ERT2-Cre mice, ILC2s were cultured in RPMI medium with 10% FBS, 55 µM 2ME, 1% Pen Strep, 10 ng per mL of IL-2 (Peprotech), and 10 ng per mL of IL-7 (Peprotech) for 2-4 weeks. For adoptive transfer of Cbfb f/f or Cbfb f/f ERT2-Cre ILC2s without in vitro tamoxifen treatment into Rag2 -/-Il2rg -/mice, 1 × 10 6 cells were intravenously injected, and the mice were treated with tamoxifen as described below. To delete Cbfb and stimulate ILC2s in vitro, 1 µM of 4-hydroxytamoxifen (4OHT) was added to the culture with 10 ng per mL of IL-33. At day 4 after 4OHT treatment, 4OHT was removed from the cell culture. At day 7 after 4OHT treatment, the ILC2s were collected and injected into Rag2 -/-Il2rg -/mice (1 × 10 6 cells/mouse). Then, 100 µg of papain in 50 µL of PBS was intranasally administered to the Rag2 -/-Il2rg -/mice on days 0, 1, and 2. On day 3, the BAL fluid and lung were collected as described above.
Histology score. Lung injury was scored as previously described 33 . Briefly, scores were given as follows: grade 1, a few inflammatory cells around the bronchus; grade 2, a layer one cell deep around the bronchus; grade 3, a layer two to four cells deep around the bronchus; and grade 4, a layer of more than four cells deep surrounded the bronchus. Six airways per section were randomly selected for scoring. PASpositive lung epithelial cells were counted at a magnification of ×400 when at least one clear PAS-positive cell was found.
Inducible deletion of Cbfβ, Runx1 and Runx3 in vivo. Cbfb f/f ERT2-Cre, Runx1 f/f ERT2-Cre, Runx3 f/f ERT2-Cre, Runx1 f/f Runx3 f/f ERT2-Cre, and Rag2 -/-Il2rg -/-recipient mice transferred with Cbfb f/f or Cbfb f/f ERT2-Cre ILC2s were orally administered 4 mg of tamoxifen (Sigma) in 200 µL of corn oil (Wako) for 5 consecutive days. At 3-4 weeks after tamoxifen treatment, ILC2s were isolated from the small intestine LPLs for flow cytometry and short-term culture to analyze ex vivo IL-5 production as described above. Eosinophils in the BAL fluid and PAS-positive lung epithelial cells of the Rag2 -/-Il2rg -/recipient mice were analyzed by flow cytometry and histology, respectively.