The transcription factor GATA3 controls cell fate and maintenance of type 2 innate lymphoid cells

Abstract

Innate lymphoid cells (ILCs) reside at mucosal surfaces and control immunity to intestinal infections. Type 2 innate lymphoid cells (ILC2) produce cytokines such as IL-5 and IL-13 and are required for immune defense against helminth infections and are involved in the pathogenesis of airway hyperreactivity. Here, we have investigated the role of the transcription factor GATA3 for ILC2 differentiation and maintenance. We showed that ILC2 and their lineage-specified bone marrow precursor (ILC2P), as identified here, were characterized by continuous high expression of GATA3. Analysis of mice with temporary deletion of GATA3 in all ILCs showed that GATA3 was required for the differentiation and maintenance of ILC2 but not for RORγt⁺ ILCs. Thus, our data demonstrate GATA3 is essential for ILC2 fate decisions, and reveal similarities between the transcriptional programs controlling ILC and T helper cell fates.

Introduction

Immunity to infections requires a highly coordinated response by the innate and adaptive immune systems. In order to deal efficiently with different types of pathogens, distinct effector programs are initiated. For example, intracellular infections lead to the induction of a T helper (Th) cell response characterized by the expression of IFN-γ (i.e., Th1 cells). Immunity against worm infections is, in part, coordinated by Th2 cells that release cytokines such as IL-4, IL-5 and IL-13. Such diverse Th cell responses are instructed by a distinct set of signals from the innate immune system. Fate decisions of Th cells are controlled by the induction of cell fate-determining transcription factors such as T-bet, GATA-binding protein (GATA)3 or retinoic acid receptor-related orphan receptor (ROR)γt for Th1, Th2 or Th17 cells, respectively. Strikingly similar effector programs exist within populations of innate lymphocytes.

A group of cells, widely referred to as innate lymphoid cells (ILCs), has attracted particular attention recently (Spits and Di Santo, 2011). They consist of RORγt-expressing lymphoid tissue inducer (LTi) cells (also referred to as RORγt⁺ ILCs) and type 2 ILCs (ILC2). In contrast to IL-22 and IL-17-producing RORγt⁺ ILCs, ILC2 can produce IL-5 and IL-13 and possibly also IL-4 (Saenz et al., 2010). Such an effector profile is reminiscent of Th2 cells and this ILC subset is variably referred to as natural helper cells (Moro et al., 2010), nuocytes (Neill et al., 2010), ILC2 (Spits and Di Santo, 2011), or type 2 innate helper (Ih2) cells (Price et al., 2010). We will refer to these cells as ILC2. Similar to the biological role of Th2 cells, ILC2 are involved in the defense against worm infections (Liang et al., 2012; Moro et al., 2010; Neill et al., 2010) and in tissue repair following influenza virus infection (Monticelli et al., 2011). They also have a role in autoimmunity, specifically the initiation of airway hyperreactivity (Chang et al., 2011). The transcription factors determining ILC2 fate are currently not known and ILC2 are, therefore, defined by: (1) various cell surface markers (i.e., IL-7Rα, Sca1, Kit, ICOS); (2) the expression of receptors for the cytokines IL-33 (ST1-T2) and IL-25 (IL-17RB); or (3) their production of the “type 2” cytokines IL-5 and IL-13. GATA3 is a double zinc-finger transcription factor that is required for the effector fate decision of Th2 cells (Zheng and Flavell, 1997; Zhu et al., 2004). Various reports have shown that ILC2 contain Gata3 transcripts (Moro et al., 2010; Price et al., 2010; Yang et al., 2011) and GATA3 was expressed by subsets of ILC2 following Nippostrongylus (N.) brasiliensis infection suggesting that GATA3 may be upregulated following infection (Liang et al., 2012). It remains a fundamental and unresolved issue as to whether GATA3 is essential for lineage commitment and/or differentiation of ILC2, comparable to the role of RORγt for the differentiation and function of RORγt⁺ ILCs.

While RORγt⁺ ILCs and ILC2 have distinct functional profiles, they share developmental requirements, indicating developmental relationships. Both ILC subsets develop from the common lymphoid progenitor (CLP) in a process that requires Notch signalling (Possot et al., 2011; Wong et al., 2012). Interestingly, development of the two ILC lineages and of NK cells but not of B or T cells requires the inhibitor of DNA binding (Id)2, suggesting that innate lymphocytes may have a common Id2-dependent progenitor (Moro et al., 2010; Yokota et al., 1999). Id2 is a helix-loop-helix (HLH) protein, which heterodimerizes with E proteins, that then can no longer initiate transcription of target genes because their DNA binding is impaired (Boos et al., 2007; Kee, 2009). While the requirement of Id2 for the development of innate lymphocytes is well recognized, its expression by ILCs remains uncharacterized, as does the precise stage during differentiation at which induction occurs.

To assess the role of GATA3 in the differentiation, maturation and function of ILC2, we utilized Id2 and GATA3 reporter mice and also genetically modified mice, allowing for controlled temporary deletion of Gata3 in all innate lymphocyte subsets and in T cells, Gata3^iΔILC,T mice. While all innate lymphocytes (i.e., RORγt⁺ ILCs, ILC2 and NK cells) expressed high amounts of Id2, high GATA3 expression was an exclusive and continuous attribute of all ILC2. We also identified a GATA3^hi population among lineage marker-negative (Lin⁻) Sca1^hiId2^hi cells in the bone marrow that we refer to as Lin⁻ Sca1^hiId2^hiGATA3^hi (LSIG) cells. Transfer studies and clonal in vitro cultures revealed that LSIG cells constitute a lineage-specified progenitor to mature ILC2 that we have termed the ILC2 progenitor, ILC2P. Genome-wide transcriptome profiling demonstrated that ILC2P are highly related to ILC2 but lacked expression of maturation markers such as the killer lectin-like receptor (KLR)G1 and of type 2 cytokine genes (i.e., IL-5 and IL-13). Induced deletion of GATA3 in ILC led to a selective loss of mature ILC2, as well as of the ILC2P in the bone marrow, whereas the RORγt⁺ ILC lineage was not affected. These data demonstrate that GATA3 is required for the differentiation of ILC2P and the maintenance of mature ILC2. Our data establishes GATA3 both as a transcription factor continuously expressed by all ILC2 and as being essential for ILC2 fate decisions.

Results

ILC2 are GATA3-expressing innate lymphocytes

To date, ILC2 are defined by a combination of cell surface markers (e.g., Sca1, CD25 and CD127) and their propensity to produce IL-5 and IL-13. However, these cell surface markers are non-specific markers also expressed by other innate lymphocyte subsets (e.g., RORγt⁺ ILC) and lymphoid precursors (Buonocore et al., 2010; Sawa et al., 2010; Vonarbourg et al., 2010). We tested our hypothesis that GATA3 is the lineage-defining transcription factor of ILC2 by examining GATA3 expression by CD25⁺Sca1^hi or CD127⁺Sca1^hi cells among CD3⁻CD19⁻ cells, marker combinations widely used to “define” ILC2 (Moro et al., 2010; Neill et al., 2010; Price et al., 2010; Saenz et al., 2010). CD25⁺Sca1^hi cells found in the lamina propria of the small intestine were largely GATA3-positive (79.6% ± 7.1) (Figure 1A). The CD127⁺Sca1^hi subset was less well defined but still showed GATA3 expression in over 60% of the population (64.8% ± 11.4) (Figure 1A).

Figure 1. ILC2 cells are GATA3-expressing innate lymphoid cells.

(A) Expression of GATA3 by CD25⁺Sca1^hi (top) or CD127⁺Sca1^hi (bottom) lamina propria lymphocytes from the small intestine. Dot plots were gated on CD45⁺CD3⁻CD19⁻ lymphocytes. Numbers represent percentages of cells in the indicated gates.

(B) Representative flow cytometry analyses of Sca1 and GATA3 expression (top) by CD45⁺CD3⁻CD19⁻ lymphocytes from the indicated organs. Histograms (bottom) represent staining with CD25 antibody (grey) or control Ig (open) by electronically gated GATA3^hiSca1^hi cells. Numbers represent percentage of cells in quadrant.

(C) Lamina propria lymphocytes from the small intestine were co-stained with CD3, CD19, Sca1 and GATA3 antibodies and with antibodies specific for the indicated markers (grey) or isotype control Ab (open). Histograms are electronically gated on Sca1^hi GATA3^hi cells within the CD45⁺CD3⁻CD19⁻ population.

(D,E) Intracellular cytokine staining of lamina propria lymphocytes from Gata3^Gfp/+ mice after 4h stimulation with PMA and ionomycin. (D) Dot plots are gated on CD3⁻CD19⁻ CD45⁺CD90⁺ lymphocytes. Numbers represent percentage of cells per quadrant. (E) Percentage (± SEM; n=5) of cytokine-producing GATA3^hi ILC with (black bars) and without stimulation (white bars).

(F) Flow cytometry analysis of GATA3 and Sca1 expression by CD45⁺CD3⁻CD19⁻ cells from the intestinal lamina propria of the indicated mouse strains.

(G) Absolute cell numbers of of GATA3^hiSca1^hi ILCs (± SD; n=4) in the lamina propria of the small intestine isolated from the indicated mouse strains.

Data are representative of at least three independent experiments. See also Figure S1.

We next examined whether co-staining with GATA3 and Sca1 was sufficient to define ILC2. Staining of leukocytes with antibodies specific for GATA3 and Sca1 revealed a well defined population of GATA3^hiSca1^hi cells that was prominently represented in the lamina propria of small intestine and colon and, to a lesser extent, the mesenteric lymph nodes (mLN) and lungs (Figure 1B and S1A,B). GATA3^hiSca1^hi cells in the CD3⁻CD19⁻ subset co-expressed all of the markers normally associated with ILCs, such as CD25 (Figure 1B), CD127, CD90 (Thy1), and Kit (Figure 1C). GATA3^hiSca1^hi cells also displayed more specific ILC2 markers, for instance, the receptor for the cytokine IL-33 (T1-ST2) (Figure 1C). Importantly, GATA3^hi ILCs did not express markers associated with either the myeloid lineage or IL-33R-expressing granulocytes (FcεRI) (Figure S1C). In addition, GATA3^hi ILCs were negative for markers of RORγt⁺ ILCs such as CD4, RORγt, or CCR6, and they did not express NK cell receptors (e.g., NKG2D or NK1.1) or CD122 expressed by all NK cells (Figure S1D). We noticed a population of Sca1^int-hi innate (CD3⁻CD19⁻) lymphocytes in the small intestine that expressed intermediate levels of GATA3 (GATA3^int) (Figure S1E). Further characterization revealed that the majority of these were RORγt or NKp46-expressing cells of the NK and RORγt⁺ ILC lineage, respectively (Figure S1E). Thus, the markers typically used to define ILC2 also define GATA3^hi ILCs. The other innate lymphocyte lineages and myeloid cells did not contain a GATA3^hi population.

ILC2 are functionally defined by their expression of IL-5 and IL-13. We investigated whether GATA3^hi ILCs could also produce these cytokines. We used mice expressing green fluorescent protein (GFP) under the control of the Gata3 promoter (Grote et al., 2006). Direct comparison of GATA3 reporter expression and intracellular staining with a GATA3-specific antibody revealed that the fluorescent reporter faithfully marked all GATA3⁺ cells (Figure S1F). GATA3^hi ILCs, but not GATA3⁻ or GATA3^int innate lymphocytes, produced IL-5 and IL-13 after incubation with phorbol 12-myristate 13-acetate (PMA) and ionomycin (Figure 1D,E). Thus, IL-5 and IL-13 expression by ILCs was strictly correlated with high GATA3 expression. In contrast, GATA3^int cells but not GATA3^hi cells released IL-22 and IFN-γ, cytokines produced by RORγt⁺ ILCs and NK cells, respectively (Figure 1D). These data further demonstrate that GATA3^int cells also do not fulfill the functional criteria of ILC2.

Further support for our hypothesis that ILC2 are GATA3^hi ILCs arose from the analysis of their representation in various gene-deficient mouse strains. Consistent with previous findings regarding ILC2 development (Moro et al., 2010), GATA3^hi ILCs were reduced in Il7r^−/− and Il7^−/− mice as well as in alymphoid Rag2^−/−Il2rg^−/− mice (Figure 1F,G). In contrast, mice genetically lacking: (1) recombination activating gene 2 (Rag2^−/− mice lacking B and T cell development); (2) IL-15 (Il15^−/− mice lacking the NK cell lineage); (3) RORγt (Rorc(γt)^−/−) or aryl hydrocarbon receptor (Ahr^−/−) both of which have reduced intestinal RORγt⁺ ILCs and; (4) the thymic stromal lymphopoietin receptor (Tslpr^−/−) showed normal development and differentiation of GATA3^hi ILCs (Figure 1F,G). Collectively, our data demonstrate that GATA3-expressing ILCs are functionally, phenotypically and developmentally identical to the ILC2 subset.

KLRG1 is a marker of mature ILC2

Upon screening a large panel of cell surface markers to identify genes co-expressed by GATA3^hi ILCs, we found that KLRG1 was expressed by virtually all ILC2 in adult mice (Figure 2A,B). This is consistent with previously published data showing elevated KLRG1 mRNA expression in ILC2 (Neill et al., 2010). Indeed, virtually all GATA3^hiSca1^hi cells were KLRG1-positive (Figure 2A, top) and all KLRG1^hiSca1^hi cells were GATA3^hi ILCs (Figure 2A, bottom). In further support of this view, IL-5 and IL-13 production was found only in KLRG1⁺ cells. KLRG1⁻ cells were producers of IL-22 and IFN-γ (Figure 2C,D). NKp46⁺ NK cells of the small intestine were found to be KLRG1-negative (Figure 2B). This is consistent with our previous data (Sanos et al., 2009) and in contrast to splenic NK cells (Hanke et al., 1998). Thus, KLRG1 can be described as a reasonably selective marker of mature intestinal ILC2.

Figure 2. GATA3^high ILCs express KLRG1.

(A) Coexpression of KLRG1 and CD25 or CD25 and GATA3 by electronically gated GATA3^hiSca1^hi (top) or KLRG1^hiSca1^hi (bottom) CD45⁺CD3⁻CD19⁻ lamina propria leukocytes. Numbers indicate percent of cells in the indicated regions.

(B) Flow cytometry analysis of KLRG1 co-expression with the indicated cell surface markers and transcription factors by CD45⁺CD3⁻CD19⁻ lamina propria leukocytes. Numbers represent percentage of cells per quadrant.

(C,D) Intracellular cytokine staining of lamina propria lymphocytes after 4h stimulation with PMA and ionomycin. (C) Dot plots are gated on CD45⁺CD3⁻CD19⁻CD90⁺ lymphocytes. Numbers represent percentage of cells per quadrant. (D) Percentage (± SEM; n=6) of cytokine-producing GATA3^hi ILC with (black bars) and without stimulation (white bars).

(E) Lamina propria cells from mice at the indicated age were analyzed for GATA3 and RORγt expression (top). Dot plots are electronically gated on CD45⁺CD3⁻CD19⁻CD127⁺ cells. Numbers represent percentage of cells per quadrant. Histograms represent expression of KLRG1 and integrin α₄β₇ by electronically gated GATA3^hiRORγt⁻ cells.

All data shown are representative of at least 3 independent experiments.

We observed that ca. 10% of ILC2 were KLRG1-negative (Figure 2A,B). KLRG1 has been recognized as a marker of mature lymphocytes (Huntington et al., 2007; Voehringer et al., 2001), prompting us to analyze KLRG1 expression by ILC2 during ontogeny. Although RORγt⁺ ILCs were substantially represented among CD3⁻CD19⁻ cells in the intestinal lamina propria of newborn mice, only a few GATA3^hi ILCs were present (Figure 2E). However, their fraction steadily increased during the first two months after birth (Figure 2E). Intriguingly, GATA3^hi ILCs of newborn mice expressed substantial levels of integrin α₄β₇, an integrin required for lymphocyte homing to and retention in the intestine (Gorfu et al., 2009). Increasing age brought about a steady decline in integrin α₄β₇ levels (Figure 2E). These findings suggest a maturation program of ILC2 characterized by the acquisition of KLRG1 and a loss of integrin α₄β₇.

Lin⁻Id2^hi bone marrow lymphocytes co-express high amounts of GATA3 and Sca1

It has been reported that ILC2 are the progeny of bone marrow CLPs (Wong et al., 2012; Yang et al., 2011). To obtain further insights into the distinct stages of ILC2 differentiation, we analyzed GATA3 and Id2 expression by hematopoietic precursor cells (lineage marker-negative cells) in the bone marrow. We chose to monitor Id2 in conjunction with GATA3 for two reasons: (1) a precursor to the lineage of ILC2 may not yet express GATA3 and; (2) Id2 is required for the development of all innate lymphocytes (Moro et al., 2010; Yokota et al., 1999). While the role of Id2 in innate lymphocyte development is well established, its expression pattern lacks definition. We have studied Id2 expression in Id2^Gfp/+ mice, a genetically modified mouse strain which expresses a fluorescent reporter (GFP) under the control of the Id2 promoter to allow faithful tracking of Id2-expressing cells (Rawlins et al., 2009). Utilizing these mice, we discovered that all mature innate lymphocyte lineages (NK cells, ILC2 and NKR⁺ and NKR⁻ RORγt⁺ ILCs) were Id2^hi (Figure 3A,B). Thymic CD4⁻CD8⁻ (double negative, DN) T cells and CD4⁺CD8⁺ (i.e., DP) T cells were negative for Id2, whereas CD4⁺CD8⁻ or CD4⁻CD8⁺ (i.e., SP) T cells expressed low amounts of Id2 (Figure S2A) which was maintained in peripheral T cells (Figure 3A). As previously reported (Jackson et al., 2011), B cells (Figure 3A) and most mononuclear phagocytes (Figure S2B) did not express appreciable amounts of Id2. CD11c⁺ DCs were Id2^lo (Figure S2B), a finding consistent with a role of Id2 for the development of DC subsets (Hacker et al., 2003; Jackson et al., 2011). Thus under steady-state conditions, high Id2 expression is a specific characteristic of all innate lymphocytes.

Figure 3. Identification of GATA3^hi ILCs in the bone marrow.

(A,B) Expression of Id2 (GFP) (grey) by the indicated cell subsets obtained from spleen or intestinal lamina propria of Id2^Gfp/+ mice. Open histograms represent analysis of the same population from control mice (A). (B) Mean fluorescence intensity (MFI ± SEM; n=4) of Id2 (GFP) expression in comparison to control mice by the indicated cell subsets isolated from the intestinal lamina propria.

(C) Id2 (GFP) and GATA3 (GFP) expression (grey) by the indicated bone marrow cell populations (HSC/MPP: Lin⁻CD127⁻Sca1⁺Kit⁺; CLP: Lin⁻CD127⁺CD135⁺Sca1^loKit^lo). Open histograms represent analysis of the same population from control mice. Numbers indicate percentage of positive cells.

(D) Analyses of Sca1, Id2 (GFP) and GATA3 expression by Lin⁻ cells. The dot plots to the right show CD127 and CD25 expression by electronically gated Sca1^hiId2^hi (top) or Sca1^hiGATA3^hi cells (bottom). Numbers represent percentage of cells in the indicated gates.

(E) Flow cytometry analysis of CD44 and CD90 expression (grey) by bone marrow LSI cells. Open histograms show staining with an isotype-matched control antibody.

(F) The indicated mouse strains were analyzed for the presence of bone marrow Lin⁻ Sca1^hiGATA3^hi cells. Numbers represent percentage of cells in the indicated gates.

Data are representative of at least three independent experiments. ** p ≤ 0.01, *** p ≤ 0.001. LSI: Lin⁻Sca1^hiId2^hi bone marrow cells. See also Figure S2.

Among Lin-negative hematopoietic precursors in the bone marrow, Id2 is not expressed by hematopoietic stem cells (HSCs) and multipotent progenitors (MPPs) or common lymphoid progenitors (CLPs) (Figure 3C). We detected a substantial population of Lin-negative cells that uniformly co-expressed Id2 and Sca1 (Figure 3D). These Lin⁻ Sca1^hiId2^hi (LSI) cells also homogenously expressed CD127 and CD25 (Figure 3D) together with CD44 and CD90 (Figure 3E), markers that are characteristic of ILCs. Conversely, Lin⁻Sca1^hiCD127⁺CD25⁺ cells were uniformly Id2-positive (Figure 3C). These cells also, unexpectedly, showed a uniformly high expression of GATA3 (Figure 3C,D). The population of LSI cells was virtually identical to Lin⁻Sca1^hiGATA3^hi cells (Figure 3C,D) and we will, therefore, refer to these as Lin⁻Sca1^hiId2^hiGATA3^hi (LSIG) cells. LSIG cells were developmentally dependent on Il2rg, Id2, and Il7r but were independent of Rag2, Rorc(γt), Il15 and Tslpr gene expression (Figure 3F), a developmental program reminiscent of intestinal ILC2 (Figure 1F).

Genome-wide expression profiling of LSIG cells reveals close relationship to ILC2

LSIG cells may be multipotent progenitors to various innate lymphocyte lineages or may constitute lineage-specified progenitors to ILC2. We analyzed expression of cell surface markers and the functional profile of LSIG cells. LSIG cells did not express markers associated with the NK cell lineage (Figure S3A), the RORγt⁺ ILC fate (Figure S3B,C), the myeloid lineage (Figure S3D) or of multipotent hematopoietic progenitors (Figure S3E). In addition to Id2 and GATA3, these cells expressed a number of markers also found on mature intestinal ILC2, including receptors for the cytokines IL-33 (T1-ST2) and IL-25 (IL-17RB) (Figure 4A,B). Other innate lymphocyte populations do not express these receptors, which provides support for the lineage-specified progenitor model. However, there were also conspicuous differences compared to mature ILC2. Bone marrow LSIG cells lacked expression of KLRG1 and Kit (Figure 4A,B). Similar to the immature KLRG1⁻GATA3^hi ILCs found in newborn mice (Figure 2E), LSIG cells expressed integrin α₄β₇ (Figure 4A), which was absent from mature KLRG1⁺ ILC2 (Figure 2E). Direct comparison of the functional profile of LSIG cells to that of mature ILC2 using qRT-PCR revealed that Il4, Il5 and Il13 but not Il22, Il17a or Ifng were highly expressed in mature ILC2 (Figure 4C, black bars), whereas expression was low in LSIG cells (Figure 4C, open bars). To confirm these findings, we directly compared IL-5 production by LSIG cells and mature intestinal ILC2 in vitro. While intestinal ILC2 were potent producers of IL-5, bone marrow LSIG cells produced only very low amounts (Figure 4D).

Figure 4. Transcriptome analysis reveals close relationship between LSIG cells and ILC2.

(A) Flow cytometry analysis of the indicated cell surface markers (grey) by bone marrow LSIG cells. Open histograms depict staining with isotype-matched control antibodies.

(B,C) Quantitative RT-PCR analysis (± SEM; n=3) of expression of the indicated genes by LSIG cells (white bars) or ILC2 (black bars).

(D) Percentage (± SEM; n=4) of IL-5-producing bone marrow LSIG cells (white bars) compared to ILC2 from the small intestine (black bars).

(E) Hierarchical clustering of normalized microarray data of replicate RNA samples from the indicated cell subsets. Dendrogram was obtained by analyzing 911 out of 28,441 probesets.

(F) Heat map representation of genes clustering with at least two fold differences in expression pattern across different cell subsets as assessed by k-means. Columns represent the indicated cell subsets in 4 or 5 biological replicates. Each row represents one examined gene. Hierarchical clustering revealed 10 different clusters as indicated. The color code at the bottom defines the expression intensity of each individual gene in all examined cell subsets.

(G) Functional classification of the gene profiles shared by LSIG and ILC2 (i.e., ILC2/LSIG core cluster) and specific for ILC2 as assessed by k-means. Representative genes within these clusters are indicated. Signature gene expression profiles of the respective clusters used in the functional annotation are indicated above.

Data are representative of three independent experiments. ** p ≤ 0.01. LSIG: Lin⁻ Sca1^hiId2^hiGATA3^hi bone marrow cells. See also Figure S3.

We performed genome-wide transcriptome analysis of highly purified LSIG cells, intestinal ILC2, and NKR⁺ or NKR⁻ RORγt⁺ ILCs to better assess the similarities and discrepancies of gene expression among these different ILC populations. The hierarchical clustering analysis of the four ILC subsets generated two distinct cell clusters, the first cluster grouped together NKR⁻ and NKR⁺ RORγt⁺ ILCs, as already revealed by previous data (Reynders et al., 2011) and the second cluster combined LSIG cells and mature ILC2 (Figure 4E). These data demonstrate a profound relationship between the transcriptional programs of the GATA3^hi ILC subsets in the intestine and the bone marrow, whereas the profiles of the RORγt⁺ ILC subsets were quite distant. We clustered and annotated selected genes that showed the most robust and consistent differences in expression (at least twofold) among the various ILC subsets (Figure 4F). A group of genes was found to be common to LSIG cells and mature ILC2, but not the RORγt⁺ ILC subsets (cluster 8, ILC2-LSIG core cluster). Among them were the genes encoding GATA3, CD27 and Sca1 (Ly6A), and receptors of IL-33, IL-25 and IL-9 (Figure 4G, top). Importantly, we could identify clusters of genes differentially expressed by LSIG cells and mature ILC2 (Figure 4G, bottom). Of note, the ILC2 signature cytokine genes IL-5 and IL-13, as well as the gene for KLRG1 were strongly represented in intestinal ILC2 but not in bone marrow LSIG cells (Figure 4G), a finding which concurred with the results described above (Figure 4A-D).

LSIG cells are lineage-specified progenitors to ILC2

Collectively, our findings from functional and genome-wide transcriptional analyses demonstrate that LSIG cells and intestinal ILC2 are highly related, albeit with differential expression of gene clusters between the two ILC subsets. Thus, we propose that LSIG cells are the lineage-specified progenitors of ILC2. We tested this model by examining the developmental potential of LSIG cells in vitro and in vivo. We cultured highly purified LSIG cells on mitotically inactivated OP9 feeder cells or Notch ligand delta-like 1 (Dll1)-expressing OP9 cells, which is required to generate T cells from uncommitted precursor cells (Schmitt and Zuniga-Pflucker, 2002). When cultured on OP9 or OP9-Dll1 cells in the presence of IL-7 or IL-7 plus IL-33, LSIG cells did not show any appreciable T or B cell potential (Figure S4A) but, instead, remained uniformly Id2 and GATA3-expressing cells (Figure 5A,B). We were also unable to find any evidence of innate multi-lineage potential in LSIG cells. When cultured in the presence of cytokines permissive for NK cell differentiation, LSIG cells showed no signs of NK cell differentiation such as upregulation of NKR (Figure 5A). In addition, LSIG cells cultured in the presence of cytokines promoting differentiation of RORγt⁺ ILCs (IL-1β, IL-23, IL-2 and IL-7) did not upregulate expression of IL-17A, IL-22, IL-23R or RORγt indicative of the RORγt⁺ ILC fate (Figure 5C and S4B). Interestingly, bone marrow LSIG cells, which are KLRG1-negative, upregulated KLRG1 expression when cultured in medium containing IL-7 and IL-2, in particular if IL-33 (Figure 5B,D) or IL-25 (Figure 5D) were present. In contrast, culture of LSIG cells in the presence of thymic stromal lymphopoietin (TSLP) did not lead to the upregulation of KLRG1 (Figure 5D). Clonal differentiation assays confirmed that LSIG cells remained stably GATA3 and Id2-expressing ILCs (Figure S4C). In addition a fraction of clones up-regulated KLRG1 (Figure S4D).

Figure 5. LSIG cells are lineage-specified progenitors of ILC2.

(A) Highly purified LSIG cells from Gata3^Gfp/+ mice or Id2^Gfp/+ mice cells were cultured for 6 days on OP9 feeder cells in medium containing IL-7 and IL-15 and tested for the expression of NKp46 and GATA3 (left panel) or NK1.1 and Id2 (right panel). Numbers represent percentage of cells in quadrant.

(B) Highly purified LSIG cells were cultured on OP9 feeder cells in the presence of the indicated cytokines. KLRG1 and GATA3 expression were determined after 7 days of culture. Numbers represent percentage of cells in quadrant.

(C) Quantitative RT-PCR analysis of expression of the indicated genes by ILC2 (black bars), by LSIG cells cultured for 7 days with cytokines favoring differentiation of RORγt⁺ ILCs (IL-2, IL-7, IL-23, IL-1β) (white bars) or by RORγt⁺ ILCs (grey bars).

(D) Percentage (± SD; n = 3) of KLRG1⁺ cells after 7 day culture of purified LSIG cells on OP9 feeder cells in the presence of the indicated cytokines.

(E) Highly purified LSIG cells were cultured on OP9 feeder cells in the presence of the indicated cytokines. After 11 days of culture, IL-5 and IFN-γ production was determined by intracellular cytokine staining. Numbers represent percentage of cells per quadrant.

(F) Percentage (± SD; n=3) of IL5⁺ cells after 7 day culture of purified LSIG cells on OP9 feeder cells in the presence of the indicated cytokines.

(G) Highly purified bone marrow LSIG cells from Id2^Gfp/+ mice (H-2^b) were adoptively transferred into irradiated Rag2^−/−Il2rg^−/− hosts (H-2^d). Six weeks after transfer, donor-derived cells from the intestinal lamina propria were analyzed for the expression of the indicated markers. Numbers represent percentage of cells in the indicated gates.

(H,I) The indicated numbers of highly purified bone marrow LSIG cells (H-2^b) or intestinal ILC2 (H-2^b) were adoptively transferred into irradiated Rag2^−/−Il2rg^−/− hosts (H-2^d). Four months after transfer, the fraction of donor-derived KLRG1⁺ ILC2 among lamina propria leukocytes was determined. As a measure of repopulation efficiency, the ratio (± SEM, n=3) of LSIG cell-derived to ILC2-derived KLRG1⁺ cells was determined (H).

(J) Quantitative RT-PCR analysis of Rora expression by the indicated cell subsets from bone marrow or the small intestine.

All experiments are representative of at least two independent experiments. LSIG: Lin⁻ Sca1^hiId2^hiGATA3^hi bone marrow cells. See also Figure S4.

To further test our theory that LSIG cells differentiate into ILC2, we assessed cytokine production after in vitro differentiation. Culture of LSIG cells with IL-7 and SCF, conditions in which they do not substantially expand (Figure S4E), led to low level production of IL-5 and IL-13 (Figures 5E,F and data not shown), similar to that observed in in vivo non-differentiated bone marrow LSIG cells (Figure 4C,D). TSLP did not show any significant effect on the maturation (Figure 5D) or cytokine production of LSIG cells in vitro (Figure 5F). In contrast, LSIG cells differentiating in the presence of IL-33 or IL-25 became potent producers of IL-5 and IL-13 indicative of differentiation into ILC2 (Figure 5E,F). However, cytokines such as IFN-γ or IL-22, which are associated with different innate lymphocyte fates could not be detected (Figure 5C,E). Collectively, these data demonstrate that LSIG cells exclusively differentiate into the IL-5 and IL-13-producing KLRG1⁺ ILC2 subset.

To provide additional support for our hypothesis that LSIG cells are the lineage-specified precursors of ILC2, we transferred highly purified LSIG cells (H-2^b) into alymphoid Rag2^−/−Il2rg^−/− mice (H-2^d) and tracked their differentiation potential. After transfer, donor-derived cells were found in organs in which ILC2 reside (e.g., intestine) (Figure 5G and S4F,G) but not in lungs, liver or spleen (Figure S4F,G). We tested whether injection of inflammatory cytokines such as IL-25 and IL-33 led to recruitment of LSIG-derived ILC2 to extraintestinal sites. Interestingly, 4 days after cytokine injection a small but reproducible population of LSIG-derived cells was found in lungs, liver and spleen (Figure S4G). While KLRG1⁻ LSIG cells had no T cell, B cell, NK cell or RORγt⁺ ILC potential, donor-derived cells maintained expression of Id2 and GATA3. Intriguingly, virtually all precursor cells differentiated into KLRG1⁺ ILC2 (Figure 5G), collectively demonstrating that KLRG1⁻ LSIG cells are precursors to mature intestinal KLRG1⁺ ILC2. To obtain further evidence that KLRG1⁻ LSIG cells have precursor potential and do not simply constitute mature ILC2 lacking KLRG1 expression (Brickshawana et al., 2011), we directly compared the reconstitution efficiency of bone marrow-derived KLRG1⁻ LSIG cells to that of mature small intestinal KLRG1⁺ ILC2 upon transfer into alymphoid mice. The population of mature KLRG1⁺ ILC2 recovered from alymphoid mice was approximately the same size as that injected. In contrast, the KLRG1⁻ LSIG population generated ca. 20-30-fold more progeny (Figure 5H,I), indicating that they had proliferated after transfer to produce mature KLRG1⁺ ILC2. From now on, we will refer to the bone marrow KLRG1⁻ LSIG population as lineage-specified precursors to ILC2 (i.e., ILC2P).

It was recently reported that mice genetically lacking Rora (encoding RORα) have reduced numbers of ILC2 following injections of IL-25 and that in vitro generation of ILC2 from Rora-deficient CLP was impaired (Wong et al., 2012). However, it remained unknown at which stage RORα is upregulated and whether RORα expression or the effects of its deletion are specific to ILC2. We analyzed Rora expression by hematopoietic progenitors, ILC2P and mature ILC subsets by using qRT-PCR. While HSCs-MPPs and CLPs did not express detectable levels of Rora, all mature innate lymphocyte subsets showed robust expression of the Rora gene. ILC2P reproducibly contained lower but considerable numbers of Rora transcripts, indicating that up-regulation of RORα occurs in two waves, during lineage specification of ILCs and at the transition from ILC2P to mature ILC2 (Figure 5J). Thus, in contrast to GATA3, RORα expression is not specific to ILC2.

ILC2P develop into functional ILC2 conferring immunity to helminth infection

It has been recently demonstrated that ILC2 play an important role in the clearance and prevention of infections with the helminth parasite N. brasiliensis (Liang et al., 2012; Neill et al., 2010). Alymphoid Rag2^−/−Il2rg^−/− cannot control N. brasiliensis infection (Figure 6A) (Moro et al., 2010; Price et al., 2010). However, adoptive transfer of ILC2P prior to worm challenge led to a control of infection comparable to that in lymphoreplete mice (Figure 6A). Donor-derived ILC2 were found in the intestine and only very few cells homed to lungs and spleen prior to infection (Figure 6B). Interestingly, infection of mice with N. brasiliensis led to a significant increase of ILC2P-derived ILC2 in small intestine but also in lungs and spleen. While eosinophils were not recruited to the lungs of alymphoid mice, likely because of the lack of an innate source of IL-5 and/or IL-13 (Liang et al., 2012), mice receiving ILC2P prior to worm challenge showed a profound accumulation of eosinophils in the lungs at day 5 post infection (Figure 6C,D). In the intestine, IL-13 expression has been shown to lead to increased differentiation of mucin-producing goblet cells and to increased angiogenin expression by Paneth cells supporting a milieu conducive for worm expulsion (Steenwinckel et al., 2009). While N. brasiliensis-infected Rag2^−/−Il2rg^−/− mice showed low level expression of angiogenin 4 or mucin 2, transfer of ILC2P led to a substantial increase in the expression of these genes (Figure 6E,F). Only few mucus-producing (i.e., Alcian blue-positive) cells were found in the colon of N. brasiliensis-infected Rag2^−/−Il2rg^−/− mice. Transfer of ILC2P into alymphoid mice, and consecutive infection, led to goblet cell hyperplasia surpassing that of control mice (Figure 6G). Collectively our data demonstrate that ILC2P can differentiate into functional ILC2, conferring protection against murine helminth infection by inducing epithelial expression of the genes involved in the defense against worms.

Figure 6. Transfer of LSIG cells confers immunity to N. brasiliensis infection.

(A-G) Highly purified LSIG cells were transferred into groups of Rag2^−/−Il2rg^−/− hosts. Four weeks after transfer, the indicated groups of mice (n=4-5) were infected with N. brasiliensis and analyzed on day 5 of infection. (A) Total intestinal helminth counts. (B) Percentage (± SD; n=3) of donor-derived (CD45.1⁺) ILC2 in the indicated organs at the day of infection (white bars) and at day 5 after N. brasiliensis infection. (C) Representative dot plot of eosinophil (CD49b^loSiglec-F⁺) accumulation in the lungs of the indicated mouse strains. Numbers in graph represent percentage of cells in the indicated gates. (D) Percentage (± SEM; n=4) of CD49b^loSiglec-F⁺ cells. (E,F) Quantitative RT-PCR analysis (± SD; n=4) of Angiogenin4 (E) and Mucin2 (F) expression by intestinal epithelial cells isolated from the indicated treatment groups. (G) Alcian blue staining of cryosections of the small intestine. Arrowheads indicate goblet cells with high amounts of mucus production. Scale bar, 100 μm. Original magnification, x20.

(H) CCR9 (GFP) expression by intestinal ILC2 and bone marrow ILC2P. Histograms are electronically gated on LSIG cells or ILC2. Open histograms represent analysis of the same population from control mice.

(I,J) Representation of GATA3^hi ILCs (i.e., ILC2 or ILC2P) in C57BL/6 (I, upper panel) and Ccr9-deficient (I, lower panel) mice. Numbers represent percentage of cells in the indicated gates. (J) Percentage (± SEM; n=3) of GATA3^hi ILCs in the indicated organs of control (white bars) or Ccr9-deficient (black bars) mice.

All data are representative of 3 or more independent experiments. * p ≤ 0.05, ** p ≤ 0.01. LSIG: Lin⁻Sca1^hiId2^hiGATA3^hi bone marrow cells.

CCR9 signals are required for the migration of ILC2P to the intestine

We investigated the molecular signals involved in coordinating the migration of ILC2 to the intestine. Our phenotypical characterization of ILC2P and mature ILC2 revealed that both populations expressed high levels of the chemokine receptor CCR9 (Figure 6H), a chemokine receptor with essential function in the homing of lymphocytes from bone marrow to the lamina propria (Zabel et al., 1999). Interestingly, Ccr9 deficiency led to a significant decrease of ILC2 in the intestinal lamina propria. However, Ccr9^−/− ILC2 were not significantly diverted to extraintestinal organs, although we consistently observed a larger fraction of ILC2 in bone marrow, lungs and mesenteric lymph nodes (Figure 6I,J). Collectively, our findings illustrate that migration of ILC2P to the intestine is in part dependent on CCR9-mediated signals.

GATA3 is required for ILC2 cell fate decisions and maintenance of mature ILC2

Our data demonstrate that GATA3 is a transcription factor stably and continuously expressed by ILC2 and by ILC2P, their lineage-specified bone marrow precursors. However, it remains unknown whether GATA3 is required for their development and/or differentiation. We developed a genetic model in which the Gata3 gene can be inducibly ablated in all ILCs (Figure S5A). We crossed Id2^CreERt2/+ mice (Rawlins et al., 2009) to mice carrying a conditional Gata3 allele (Gata3^f/f) (Grote et al., 2008). Tamoxifen application induces nuclear translocation of Cre allowing for the inducible inactivation of Gata3. To faithfully monitor Cre activity, we further crossed these mice to a strain carrying a fluorescent reporter (yellow fluorescent protein, YFP), which comes under the transcriptional control of the ubiquitous Rosa26 promoter once a transcriptional roadblock is excised by Cre-mediated recombination (Figure S5A) (Srinivas et al., 2001). As Id2 is highly expressed in ILC2P, innate lymphocytes and during T cell differentiation in the thymus (Figure 3A,C and S2A), this approach produced mice in which the Gata3 gene can be inducibly deleted in all innate lymphocyte lineages and T cells, Gata3^iΔILC,T. Cre activity correlated with Id2 expression so that Id2^hi cells (e.g., ILC2, ILC2P, RORγt⁺ ILCs) had higher Cre activity than Id2^lo cells (e.g., T cells) (Figure S5B). Development of GATA3^int RORγt⁺ ILCs (Figure S1E) was unaffected by the deletion of GATA3 (Figure 7A) which allowed us to investigate the efficiency of GATA3 inactivation in the YFP⁺ (Cre-on) subset of RORγt⁺ ILCs. Upon induction of Cre expression by tamoxifen, GATA3 expression was abolished in YFP⁺RORγt⁺ ILCs (Figure 7B). Thus, Gata3^iΔILC,T mice allow for efficient inducible deletion of Gata3 in innate lymphocytes.

Figure 7. GATA3 controls cell fate and maintenance of ILC2.

(A) Analyses of RORγt and KLRG1 expression by YFP⁺ (Cre-on) and YFP⁻ (Cre-off) fractions of CD45⁺CD3⁻CD19⁻ lamina propria cells from the indicated mouse strains. Numbers represent percent cells in quadrant.

(B) GATA3 expression by intestinal YFP⁺ (Cre-on) RORγt⁺ ILCs of Gata3^+/+ (grey) or Gata3^f/f (open) animals.

(C) Expression of GATA3 and KLRG1 by YFP⁺ (Cre-on) and YFP⁻ (Cre-off) fractions of CD45⁺CD3⁻CD19⁻ intestinal lamina propria cells from the indicated mouse strains. Numbers represent percent cells in gate.

(D,E) Percentage (± SD; n=4) of KLRG1⁺YFP⁺ cells (D) and of YFP⁺ cells (E) among CD45⁺CD3⁻CD19⁻ lamina propria lymphocytes from the indicated mice.

(F) Lamina propria lymphocytes from the indicated mouse strains were stimulated for 4 h with PMA and ionomycin. IL-5 and IL-13 production was analyzed by intracellular cytokine staining and flow cytometry analysis of electronically gated CD45⁺CD3⁻CD19⁻ CD90⁺ lamina propria lymphocytes. Numbers indicate percent cells in quadrant.

(G) Highly purified YFP⁻ ILC2 were cultured on mitotically inactivated OP9 feeder cells in the presence of IL-2, IL-7 and 4-OH tamoxifen. After 7 days, cells were analyzed for expression of YFP and KLRG1. Numbers represent percentage of cells in the indicated gates.

(H) Expression of Sca1 and YFP among Lin⁻ cells of the bone marrow from the indicated mouse strains.

(I) Expression of CD25 and Sca1 by Lin⁻ bone marrow cells. Numbers indicate percentage of cells in quadrant. The lower panel shows the expression of YFP (Cre-on) in electronically gated Lin⁻CD25⁺Sca1^hi bone marrow cells of the indicated mouse strains. Numbers indicate percentage of YFP-positive cells.

(J) Highly purified YFP⁺ ILC2P were cultured on mitotically inactivated OP9 feeder cells in the presence of IL-2, IL-7 and 4-OH tamoxifen. After 7 days, the cells were analyzed for expression of YFP and KLRG1. Numbers represent percentage of cells in quadrants.

All data are representative of at least two independent experiments. *** p ≤ 0.001, n.s.: not statistically significant. See also Figure S5.

We next treated Gata3^iΔILC,T mice for three weeks with a tamoxifen-containing diet and analyzed ILC2 in the intestinal lamina propria. While the pool of RORγt⁺ ILCs was unaffected by Gata3 deficiency (Figure 7A,B), KLRG1⁺GATA3^hi cells or RORγt⁻ KLRG1⁺ cells (both representing ILC2) were virtually absent in the GATA3-deficient (i.e., YFP⁺) subset (Figure 7A,C). This was also reflected in our analysis of the proportion of KLRG1⁺ cells within all innate lymphocytes with Gata3 deletion (YFP⁺) (Figure 7D). In contrast, ILC2 were normally represented in the YFP-negative (Cre-off) subset spared from Cre-mediated inactivation of Gata3 (Figure 7A,C). While ILC2 are entirely lost in mice with conditional deletion of Gata3, the total pool of ILCs with Gata3 deletion (YFP⁺) was only mildly reduced (Figure 7E) which reflects the normal maintenance of RORγt⁺ ILCs in the absence of GATA3. It is unlikely that KLRG1, CD25 and Sca1 are under the direct transcriptional control of GATA3, meaning that a deficiency in GATA3 is unlikely to have the knock-on effect of obscuring detection of these markers. In support of this, ILC2P do not express KLRG1 despite high GATA3 expression (Figure 3C,D and 4A). Nevertheless, we interrogated cytokine production by CD3⁻CD19⁻ cells in order to visualize residual innate type 2 cytokine producers in the absence of GATA3. While the GATA3-proficient (YFP⁻) fraction of innate lymphocytes produced IL-5 and IL-13 after stimulation comparable to control mice, no cytokine producers were detected in the GATA3-deficient (YFP⁺) population (Figure 7F).

The entire intestinal pool of ILC2 was lost within few weeks of Gata3 inactivation (Figure 7A,C,D). Maintenance of the pool size of intestinal lymphocytes is the aggregate of two parameters; cell half-life and the rate of replacement by newly arriving progenitors. Therefore, our data has two important implications. First, that mature ILC2 cannot be maintained when Gata3 is ablated and, second, that bone marrow-derived ILC2P do not efficiently replenish the peripheral pool of KLRG1⁺ ILC2 when GATA3 is inactivated. Since we could detect ILC2 more than four months after transfer (Figure 5G-I), our data do not favor the view that the absence of ILC2 3-4 weeks after Gata3 inactivation reflects the natural half-life of this subset. To more directly address this issue, we cultured highly purified YFP⁻ (i.e., Cre-off) KLRG1⁺ ILC2 from the lamina propria of Gata3^f/f mice and of control mice in the presence of tamoxifen for 7 days. Roughly 20% of ILC2 from Gata3^+/+ controls switched on Cre expression (i.e., became YFP-positive) during the 7 days of culture (Figure 7G). Interestingly, when Cre activity was induced in ILC2 from Gata3^f/f mice, only very few if any YFP⁺ cells could be found whereas the YFP⁻ (Gata3-proficient) ILC2 were normally maintained (Figure 7G). These data establish that GATA3 is required for the maintenance of mature ILC2.

Given the failure of ILC2P to repopulate the peripheral pool of KLRG1⁺ ILC2 when Gata3 is inactivated, we investigated the role of GATA3 in the development or maintenance of bone marrow ILC2P. After inducible deletion of Gata3, LSI cells (i.e., ILC2P) were 6-10-fold reduced (Figure 7H). The remaining cells phenotypically resembled ILC2P in that they expressed high levels of Sca1 and CD25 (Figure 7I). It was remarkable to find such a population of Lin⁻CD25⁺Sca1^hi bone marrow cells within the YFP⁺ (Cre-on) population (Figure 7I, bottom panel) and we suspected that they represented either precursor cells that could not differentiate in the absence of GATA3 or remaining ILC2P. Culture of YFP⁺ (i.e., Cre-on) Lin⁻CD25⁺Sca1^hi cells from GATA3-proficient control mice led to the generation of KLRG1⁺ ILC2 (Figure 7J) as observed before (Figure 5B). In contrast, culture of equal numbers of the same subset isolated from Gata3-deficient mice did not generate any progeny (Figure 7J). Collectively, our data provide evidence for a dominant role for GATA3 at two central checkpoints of the ILC2 lineage. Firstly, GATA3 is required for the maintenance of mature ILC2 and of ILC2P. Secondly, GATA3 is an important factor in the maturation of ILC2P. Thus, similar to the role of RORγt in RORγt⁺ ILCs, GATA3 is an essential determinant of ILC2 fate.

Discussion

Using genetically modified mice to allow faithful tracking of GATA3-expressing cells and intracellular staining for this transcription factor, we demonstrated that ILC2 are, by any definition (phenotypic, functional and developmental), GATA3^hi ILCs. Given that other innate lymphocyte lineages (i.e., RORγt⁺ ILCs and NK cells) expressed only low amounts of GATA3, high GATA3 expression can be said to be a specific characteristic of ILC2. The presence of Gata3 transcripts in ILC2 has previously been reported (Moro et al., 2010; Price et al., 2010; Yang et al., 2011). One recent report drew an interesting parallel to Th2 cells by showing that a subset of lung-resident ILC2 (i.e., CD4⁻CD8⁻ CD19⁻ICOS⁺ cells) isolated from N. brasiliensis-infected mice expressed GATA3 (Liang et al., 2012). Our data now demonstrate that all ILC2 constitutively express high amounts of GATA3.

We observed only a very small fraction of ILC2 in the intestine of newborn mice. However, their proportion steadily increased during the first two months after birth, indicating that ILC2 only populate the intestine in substantial numbers postnatally. Phenotypical profiling showed that >90% of ILC2 of adult mice expressed the lectin-like receptor KLRG1, which was absent from GATA3^hi ILC of newborn mice. Thus, we propose that KLRG1 acts a marker of mature ILC2. KLRG1 is expressed by a subset of mature splenic NK cells and by subsets of activated CD8 T cells and its ligation can inhibit effector functions in these lymphocyte subsets (Blaser et al., 1998; Hanke et al., 1998; Huntington et al., 2007; Sarkar et al., 2008). KLRG1 is an inhibitory immunoreceptor that binds to classical cadherins (e.g., E-cadherin) (Grundemann et al., 2006; Ito et al., 2006), highly expressed by intestinal epithelial cells. It is tempting to speculate that the interaction of KLRG1 with E-cadherins may dampen the function of ILC2.

Previous data have indicated that ILC2 develop from CLPs and are part of the lymphoid lineage (Wong et al., 2012; Yang et al., 2011). However, the distinct stages of ILC2 development or differentiation have not yet been recognized. Id2 is a transcription factor required for the development of all innate lymphocyte lineages (Moro et al., 2010; Yokota et al., 1999). We identified a distinct population of Lin-negative Id2^hi cells in the bone marrow that co-expressed Sca1, CD25, CD127 and receptors for IL-33 and IL-25. Phenotypic and genome-wide transcriptome analysis of these Lin⁻Sca1^hiId2^hi (LSI) cells revealed that they were very closely related to ILC2 and only distantly related to other innate lymphocyte lineages such as RORγt⁺ ILCs. LSIG cells are lineage-specified precursors to mature KLRG1⁺Kit⁺ ILC2 that we propose to refer to as ILC2P. Previous reports had already indicated the presence of ILC2-like cells in the bone marrow but failed to probe the developmental potential of these cells (Brickshawana et al., 2011; Price et al., 2010). Importantly, CD25⁺Lin⁻Sca1⁺Kit⁻ (CD25⁺LSK⁻) cells that are virtually identical to the LSIG population, were recognized over a decade ago before and considered perplexing because, after adoptive transfer, and despite expression of “stem cell antigen”, they did not generate any appreciable progeny in secondary lymphoid organs (Kumar et al., 2008; Randall and Weissman, 1998). Indeed, LSIG cells failed to home to secondary lymphoid organs but were precursors to mature intestinal ILC2.

GATA3 was required for the maintenance of mature ILC2 and bone marrow-resident ILC2P. It remains formally possible that ILC2P cannot migrate to the intestinal mucosa in the absence of GATA3. This is another striking parallel to the role of GATA3 in Th2 cells as they also require continuous GATA3 expression for their maintenance (Pai et al., 2003; Zhu et al., 2004). Interestingly, a subset of GATA3-negative LSI cells were found in the bone marrow of both GATA3-proficient and GATA3-deficient mice. These cells underwent in vitro differentiation into ILC2 only in the presence of GATA3, demonstrating that GATA3 controls lineage specification in this innate lymphocyte lineage. A recent report indicated that RORα was required for efficient ILC2 development but the precise stage of differentiation at which it acted was not examined (Wong et al., 2012). We have now shown that, in contrast to GATA3, RORα expression is not specific to the ILC2 lineage but it is expressed at high levels by all innate lymphocytes. RORα expression is likely upregulated in two waves, during ILC lineage specification and at the transition of ILC2P to ILC2. Altogether, our data demonstrate that, similar to the role of RORγt for RORγt⁺ ILCs, GATA3 is required for the ILC2 fate.

It is an emerging theme that the distinct ILC subsets use transcriptional programs previously identified to be in control of various Th cell effector fates. For example, RORγt is required for both the Th17 program and the development and function of RORγt⁺ ILCs. We have provided conclusive evidence that another ILC subset resembling Th2 cells requires GATA3 for lineage specification and maintenance. These findings may indicate that these transcriptional programs are conserved in both the innate and adaptive arms of the immune system. However, there are also obvious dissimilarities. Under steady-state conditions, CD4⁺ T cells are naïve cells that do not express GATA3 or RORγt before encountering their cognate antigen. It is only upon TCR ligation that the small pool of antigen-specific T cells expands and effector-fate determining transcription factors are up-regulated. In contrast, the ILC lineages require RORγt or GATA3 at the lineage specification stage in primary lymphoid organs (fetal liver or bone marrow), unlike T cells, which can, in general, develop in the absence of these transcription factors. Such differences may well reflect the divergent design principles of the innate and adaptive immune systems.

Experimental Procedures

Mice

C57BL/6 mice were purchased from Janvier. Information about the genetically modified mouse strains used in this study can be found in the Supplemental Experimental Procedures. Mice were kept under specific pathogen free conditions and experiments were performed in accordance with the guidelines of the animal care and use committees and the Regierungspräsidium Freiburg and Erlangen.

Microarray analysis

RNA for Microarray was isolated using the RNeasy Micro Kit (Qiagen) according to the manufacturer’s protocol. RNA integrity was determined before amplification using the Ovation Pico WTA V2 kit (Nugen). Amplified cDNA samples were further fragmented and labeled using the Affymetrix WT labeling kit. Labeled fragments were hybridized to GeneChip ST1.0 before analysis using the Affymetrix GeneChip Scanner 3000 7G. The gene array data have been deposited to ArrayExpress (http://www.ebi.ac.uk/arrayexpress/) with accession number E-MEXP-3743.

Cell sorting and adoptive transfer experiments

The indicated cell populations were isolated from the respective organs and were highly purified (> 98%) using a MoFlo Astrios cell sorter. Cells were transferred i.v. into non-irradiated Rag2^−/−Il2rg^−/− mice on a BALB/c background or were used for in vitro experiments.

N. brasiliensis infection

C57BL/6 Rag2^−/−Il2rg^−/− mice were reconstituted with 1 × 10⁴ LSIG cells. Four weeks later, reconstituted mice were injected s.c. with 500 purified N. brasiliensis stage 3 larvae (see also Supplemental Experimental Procedures). Injection of IL-25 and IL-33 (250 ng/ml each; BioLegends) was performed at day 0 and day 2 following infection and mice were sacrificed for analysis at day 5 post infection. During infection, mice were provided drinking water supllemented with antibiotics (2 g/l neomycin sulfate, 100 mg/l polymixin B sulfate; Sigma-Aldrich).

Histology

Ileal tissues were fixed in 4% paraformaldehyde and embedded in OCT compound (Tissue-Tek) before freezing at −80°C. Histological sections with a thickness of 7 μm were stained with Alcian blue working solution for 15 min, followed by extensive washing and were mounted using Aquatex (Merck).

In vitro differentiation

The indicated, highly purified cell populations were seeded onto mitomycin C-treated, mitotically inactivated OP9- or OP9Dll-1 feeder cells. Cytokines were added to the culture at a concentration of 20 ng/ml if not indicated otherwise. Clonal differentiation assays are described in detail in the Supplemental Experimental Procedures.

Statistical analysis

Significance level of the data sets was determined by performing a two-tailed Student’s t-test. If equal variances between the groups could not be assumed, Welch’s correction of the t-test was applied. Tests for statistical significance were performed with Graph Pad Prism v4 software. (*p <0.05; **p <0.01 and ***p <0.001; n.s., not significant)

Highlights.

• ILC2 continuously and stably express high amounts of GATA3

• Bone marrow GATA3^hi cells are lineage-specified progenitors to mature ILC2 (ILC2P)

• High Id2 expression is characteristic of all innate lymphocyte lineages

• GATA3 is required for differentiation and maintenance of ILC2 and ILC2P