Research ArticleImmunology

Transforming growth factor–β and Notch ligands act as opposing environmental cues in regulating the plasticity of type 3 innate lymphoid cells

See allHide authors and affiliations

Science Signaling  03 May 2016:
Vol. 9, Issue 426, pp. ra46
DOI: 10.1126/scisignal.aaf2176

Plasticity in innate lymphoid cell function

Like the T cells and B cells of the adaptive immune system, cells in the innate immune system are key to organismal health. Innate lymphoid cells (ILCs) are a heterogeneous type of innate immune cell that regulates immune responses and tolerance at mucosal surfaces, such as in the gut, by rapidly secreting cytokines. Group 3 ILCs (ILC3s) are characterized by the presence or absence of a cell surface natural cytotoxicity receptor (NCR). Two studies now provide evidence of heterogeneity and plasticity within ILC3s. Chea et al. found that a substantial proportion of mouse NCR ILC3s differentiated into NCR+ ILC3s in response to stimulation of the receptor Notch2. In mice with defective Notch signaling specifically in lymphoid cells, NCR+ ILC3s were reduced in number and showed impaired cytokine secretion. Viant et al. showed that Notch signaling was required for the maintenance of NCR+ ILC3s. Furthermore, signaling by the cytokine transforming growth factor–β (TGF-β) antagonized Notch signaling, resulting in reduced numbers of NCR+ ILC3s. Together, these studies indicate that ILC3 subset composition in vivo depends on the balance between different signals found in tissue microenvironments, which has implications for whether immune responses or tolerance will prevail.

Abstract

Group 3 innate lymphoid cells (ILC3s) are composed of subsets that are either positive or negative for the natural cytotoxicity receptor (NCR) NKp46 (encoded by Ncr1). ILC3s are located at mucosal sites, such as in the intestine and lung, where they are exposed to billions of commensal microbes and potentially harmful pathogens. Together with T cells, the various ILC3 subsets maintain the balance between homeostasis and immune activation. Through genetic mapping, we identified a previously uncharacterized subset of NCR ILC3s in mice that transiently express Ncr1, demonstrating previously undescribed heterogeneity within the ILC3 population. In addition, we showed that sustained Notch signaling was required for the maintenance of the NCR+ phenotype and that the cytokine transforming growth factor–β (TGF-β) impaired the development of NCR+ ILC3s. Thus, the plasticity of ILC3s is regulated by the balance between the opposing effects of Notch and TGF-β signaling, maintaining homeostasis in the face of continual challenges.

INTRODUCTION

Type 3 innate lymphoid cells (ILC3s) form a heterogeneous population of ILCs that are dependent on the transcription factor retinoic acid–related orphan receptor γ (RORγt) and are abundant at mucosal sites (17). ILC3s can be divided into three main subsets on the basis of their role during embryogenesis and their cell surface expression of the natural cytotoxicity receptor (NCR) NKp46 (6, 7). NCR ILC3s include lymphoid tissue inducer (LTi) cells, which were initially found in the fetus and are required for the development of lymph nodes and Peyer’s patches (8). LTi cells do not require the transcription factor promyelocytic leukemia zinc finger (PLZF) (9) and can be classified into CD4 and CD4+ subsets (6, 7). A further subset of NCR ILC3s requires PLZF (9), is considered to be a homogeneous population, and gives rise to NCR+ ILC3s in a T-bet– and Notch-dependent manner (10, 11). It is unknown whether other environmental factors contribute to ILC3 maturation.

Transforming growth factor–β (TGF-β) is a ubiquitous, multifunctional cytokine that plays a key role in modulating immunity (12). TGF-β signals through the common serine and threonine kinase receptor complex, which has two subunits, TGF-β receptor I (TGF-βRI) and TGF-βRII. In the intestine, TGF-β is secreted by regulatory T (Treg) cells and nonimmune cells, and it has potent anti-inflammatory effects (13). The production of functional TGF-β is associated with protection against intestinal inflammation and colitis in models of inflammatory disease (14). TGF-β signaling plays a crucial role in the proliferation and differentiation of T cells, and it prevents the differentiation of naïve CD4+ T cells into type 2 helper (TH2) cells by inhibiting expression of the gene encoding GATA3 (14), a transcription factor also required for ILC differentiation (15). TGF-β also regulates the maturation of natural killer (NK) cells (16, 17) and acts as a chemoattractant for ILC2s during the pulmonary allergic response (18). It remains unclear how TGF-β influences the differentiation of ILC3s.

Here, we analyzed the heterogeneity of ILC3 subsets by Ncr1 fate-mapping (FM) in mice. We found that the NCR ILC3 population was heterogeneous, and we revealed the existence of a previously uncharacterized population of NCR ILC3s that transiently express Ncr1. Differentiation of these cells into NCR+ ILC3s required Notch, which was also essential to maintain the identity of this subset. We also developed a mouse model in which the TGF-β pathway was selectively stimulated in NCR+ ILC3s. Using this approach, we showed that TGF-β impaired the differentiation of NCR cells into NCR+ ILC3s in the intestine. Thus, Notch and TGF-β reciprocally regulate the balance between the NCR and NCR+ subsets of ILC3s to ensure the homeostatic maintenance of these subsets during gut homeostasis and inflammation.

RESULTS

Ncr1 FM analysis reveals the existence of a population of NCR ILC3s (FM+) that lack the cell surface expression of NKp46

The development of NCR+ ILC3s has been described as a unidirectional process in which a subset of NCR ILC3s differentiates into NCR+ ILC3s in response to Notch-dependent signals (11, 19, 20). We used R26eYFP/+ Ncr1-iCre reporter mice, in which Ncr-iCre selectively drives the expression of enhanced yellow fluorescent protein (eYFP) in NKp46+ cells, to determine whether this was indeed the case. As expected, the NCR+ RORγt and NCR+ ILC3 populations were eYFP+; however, about 30% of NCR ILC3s were also eYFP+, indicating that these cells had previously expressed the Cre recombinase driven by the Ncr1 locus (Fig. 1A and fig. S1A). We describe these cells as NCR ILC3 (FM+). Such cells were also found in the small intestine, cecum, and colon of R26RFP/RFP Ncr1-iCre mice in experiments in which another gating strategy for ILCs was used that did not involve cell permeabilization and RORγt staining (fig. S1B). The NCR ILC3 (FM+) subset lacked detectable cell surface NKp46 in R26RFP/RFP Ncr1-iCre mice (fig. S1C). The presence of the NCR ILC3 (FM+) subset was associated predominantly with mucosal tissues, including the colon and cecum, but not with bone marrow (fig. S2).

Fig. 1 Analyses of the fate and function of ILC3s in the small intestine.

(A) Characterization of ILCs in the small intestine of R26eYFP/+Ncr1-iCre mice at steady state. Flow cytometric analysis of the cell surface expression of NKp46 and RORγt (left) in LinCD45.2+ ILC3 populations isolated from the small intestine of naïve R26eYFP/+ Ncr1-iCre mice. The histograms (right) show eYFP amounts in NKp46+ RORγt (NCR+ RORγt), NKp46+ RORγt+ (NCR+ ILC3), and NKp46 RORγt+ (NCR ILC3) subsets. Numbers indicate the percentages of cells within the indicated gates. Data are from a single experiment with two or three mice and are representative of four independent experiments. (B) Characterization of ILCs from the small intestine of Il22eGFP/eGFP Ncr1-iCre mice after stimulation with mouse IL-23 for 4 hours in vitro. The relative amounts of eGFP and IL-22 were assessed in NCR+ RORγt cells, NCR+ ILC3s, and NCR ILC3s. Shown are the relative abundances of NKp46 and RORγt in Lin CD45.2+ cells (left) and the relative amounts of IL-22 and eGFP levels (right) in all three indicated populations. Numbers indicate the percentages of cells within the indicated gates. Data are from a single experiment with two or three mice and are representative of three independent experiments.

Secretion of the cytokine interleukin-22 (IL-22), a member of the IL-10 superfamily, is a key function of ILC3s. We therefore characterized the NCR ILC3 (FM+) subset further by performing lineage and functional analyses that focused on the production of this cytokine. We previously generated a mouse model of conditional Il22 deficiency that also acts as an Il22 reporter by crossing Ncr1-iCre mice with Il22eGFP knock-in mice (21). In these mice, IL-22 production was selectively abolished in NCR+ ILCs and was replaced by the enhanced green fluorescent protein (eGFP) reporter protein. Intestinal RORγt NKp46+ cells, which include NK cells, ILC1s, and “ex-ILC3s” [interferon-γ (IFN-γ)–producing ILC1s that previously expressed RORγt] (22), did not produce IL-22 on stimulation in vitro with IL-23 for 4 hours (Fig. 1B). In contrast, the same stimulus led to a percentage of NCR+ ILC3s becoming IL-22 eGFP+ cells in Il22eGFP/eGFP Ncr1-iCre mice, whereas NCR+ IL-22+ eGFP ILC3s were barely detectable, demonstrating the accurate reporting of IL-22 production by Il22+/+ Ncr1-iCre mice (Fig. 1B). Furthermore, IL-22 eGFP+ cells were also observed among NCR ILC3s in Il22eGFP/eGFP Ncr1-iCre mice, which corresponded to the NCR ILC3 (FM+) subset present in R26eYFP/+ Ncr1-iCre mice. Thus, the NCR ILC3 (FM+) subset was able to produce IL-22. These data suggest that Ncr1 expression had been induced in the NCR ILC3s (FM+) at mucosal sites and that these cells were able to produce IL-22, raising the possibility that Ncr1 expression was induced transiently or that NCR+ ILC3s could revert to NCR ILC3s.

The NCR ILC3 (FM+) subset shares transcriptional profiles with the NCR ILC3 (FM) subset

We then sought to compare the NCR ILC3 (FM+), NCR ILC3 (FM), and NCR+ ILC3 subsets more precisely by sorting these cells and analyzing their transcriptomic profiles. Ncr1 transcripts were barely detectable in both the NCR ILC3 (FM) and NCR ILC3 (FM+) subsets, whereas NCR+ ILC3s contained about 10 times more Ncr1 mRNA, as shown by reverse transcription quantitative polymerase chain reaction (RT-qPCR) analysis (Fig. 2A). We then compared the pan-genomic profiles of the three ILC3 subsets by RNA sequencing (RNA-seq). This analysis confirmed that Ncr1 transcripts were barely detectable in the NCR ILC3 (FM) and NCR ILC3 (FM+) subsets (Fig. 2B). It also revealed that although the NCR ILC3 (FM) and NCR+ ILC3 subsets had different profiles, the NCR ILC3 (FM+) subset was more similar to the NCR ILC3 (FM) subset than to NCR+ ILC3s. This was shown, for example, by analysis of the genes expressed by the NCR ILC3 (FM+) subset that were previously identified as part of the transcriptional signature of NCR ILC3s (Fig. 2C) (23). Nevertheless, cells of the NCR ILC3 (FM+) subset had low amounts of Ccr6 and Nrp1 (which encodes neuropilin 1) transcripts (Fig. 2C) and their corresponding proteins (Fig. 2D). This intermediate profile of cells of the NCR ILC3 (FM+) subset suggested that this population was in transition between the NCR ILC3 (FM) subset and NCR+ ILC3s.

Fig. 2 Transcriptomic analyses of FM small intestinal ILC3.

(A) Strategy for sorting ILC3 populations from R26eYFP/+ Ncr1-iCre mice. Left: Live Lin (CD3CD19) CD45+ cells were gated on NKp46 versus NK1.1 expression, and NKp46+ NK1.1 eYFP+ cells were selected as NCR+ ILC3s. NKp46 NK1.1 cells were further gated on c-kit versus KLRG-1 expression, and c-kithigh KLRG1 cells (NCR ILC3) were sorted into eYFP+ (FM+) and eYFP (FM) cells. Right: Ncr1 mRNA amounts in the indicated sorted ILC populations were determined by RT-qPCR analysis. Data are means ± SEM of three technical replicates normalized against Hprt mRNA abundance from a single experiment. DCM, dead cell marker; FSC-A, forward scatter A. (B) NCR ILC3 (FM), NCR ILC3 (FM+), and NCR+ ILC3 were isolated from the small intestine of naïve R26eYFP/+ Ncr1-iCre mice and analyzed by high-throughput RNA-seq techniques. RNA-seq analyses of Ccr6 and Ncr1 are provided as examples of NCR and NCR+ signature genes, together with Rorc and Idb2, which were uniformly expressed by all ILC3 subsets. Data show one experiment. (C) Comparison of normalized reads per kilobase of transcript per million mapped reads (RPKM) for signature genes defined for NCR ILC3s and NCR+ ILC3s (22, 23) in the NCR ILC3 (FM), NCR ILC3 (FM+), and NCR+ ILC3 populations. Data are from one or two independent biological replicates pooled from three to five mice for each experiment. (D) Left: Contour plots showing the cell surface expression of Nrp1 and CCR6 in NCR ILC3 (FM), NCR ILC3 (FM+), NCR+ ILC3, and NCR+ RORγt populations. Right: Histograms show the mean fluorescence intensity for Nrp1 and CCR6 of the indicated populations. Data are pooled from four independent experiments, with two to three mice per experiment. Unpaired Student’s t tests were used for statistical analyses. ****P < 0.0001.

Sustained Notch signals play a key role in maintaining NCR+ ILC3s

The FM analysis of the ILC3 subsets suggested that the NCR ILC3 (FM+) subset might be the progeny of NCR+ ILC3s. We tested this possibility by culturing highly purified cells of the NCR ILC3 (FM+) and NCR+ ILC3 subsets on OP9 stromal cells. After 9 days of culture, about 40% of the NCR+ ILC3s had lost cell surface NKp46 and had become NCR ILC3 (FM+) (Fig. 3A, bottom). No NCR+ ILC3s developed from these FM+ cells under the same culture conditions (Fig. 3A, top). Notch is the only extrinsic factor known to drive the differentiation of NCR ILC3s into NCR+ ILC3s (11), and OP9 stromal cells lack Notch ligands. Therefore, we also cultured highly purified cells of the NCR ILC3 (FM+) and NCR+ ILC3 subsets on OP9-DL1 cells, which ectopically express the Notch ligand Delta-like 1 (Dll1). After 9 days of culture with these cells, more than 90% of the NCR+ ILC3s displayed an unchanged phenotype (Fig. 3A, bottom), whereas about 25% of cells of the NCR ILC3s (FM+) subset had differentiated into NCR+ ILC3s (Fig. 3A, top). Thus, NCR+ ILC3s gave rise to cells of the NCR ILC3 (FM+) subset in vitro and Notch counteracted this reversion. Consistent with previous reports (10, 11), no substantial conversion of NCR+ ILC3s into NCR ILC3s was observed after their adoptive transfer into alymphoid Rag2−/−Il2Rg−/− mice (11, 24). Because Notch is highly abundant in vivo, these data suggest that the Notch-dependent pathway is involved in the conversion of NCR ILC3s into NCR+ ILC3s and the maintenance of this phenotype. However, the lack of models for precise monitoring and control of the intensity and site of Notch protein production in vivo precluded further investigations of the influence of Notch on ILC3s in vivo.

Fig. 3 In vitro and in vivo eYFP expression in ILC3s in the presence or absence of T-bet.

(A) In vitro differentiation of NCR ILC3 (FM+) and NCR+ ILC3 subsets purified from R26eYFP/+ Ncr1-iCre mice and cultured in the presence or absence of Notch ligands. NCR ILC3 (FM+) (top) and NCR+ ILC3 (bottom) subsets were purified from the small intestine of R26eYFP/+ Ncr1-iCre mice (left) and cultured on OP9 stromal cells or OP9-DL1 cells. After 9 days of culture, live cells were analyzed for the cell surface expression of NKp46 and eYFP (middle). Numbers indicate the percentages of cells within the indicated gates. Right: Histograms show the relative amounts of T-bet in NKp46+ cells (red) and NKp46 cells (orange) from the indicated cultures. Numbers indicate the mean fluorescence intensity. Data are from one experiment with two to four replicate wells and are representative of four independent experiments. (B) T-bet abundances in ILC3 populations of the small intestine of R26eYFP/+ Ncr1-iCre mice. Flow cytometry plots showing NKp46 versus RORγt expression within Lin CD45.2+ ILC3s (top left). Heterogeneity of T-bet and eYFP in NKp46 RORγt+ cells (NCR ILC3) (bottom left). Right: Histograms show T-bet abundance in NKp46+ RORγt+ (red line), RORγt+ NKp46 eYFP+ (orange line), and RORγt+ NKp46 eYFP (purple line) cells. The dotted line shows T-bet abundance in NCR ILC3s isolated from Tbx21−/− mice. Numbers indicate the percentages of cells within the indicated gates. Data are representative of three independent experiments, with two to three mice per experiment. (C) Distribution of NCR+ RORγt cells, NCR+ ILC3s, and NCR ILC3s in the small intestine of Tbx21−/−R26eYFP/+ Ncr1-iCre mice. Left: Flow cytometry plots show NKp46 versus RORγt expression within Lin CD45.2+ ILC3s. Histogram shows eYFP abundance in NKp46 RORγt+ cells (NCR ILC3s). Numbers indicate the percentages of cells within the indicated gates. Right: The histogram shows the numbers of cells in the indicated populations. Data are representative of two independent experiments, with three mice per experiment. Student’s t test was used for statistical analyses. *P < 0.05; n.s., not significant. WT, wild type.

T-bet regulates the plasticity of ILC3s

We then investigated the abundance of T-bet in the three subsets of ILC3s in R26eYFP/+ Ncr1-iCre reporter mice. Unlike NCR+ ILC3s, which produced substantial amounts of T-bet, cells of the NCR ILC3 (FM) subset contained barely detectable amounts of T-bet, whereas cells of the NCR ILC3 (FM+) subset had intermediate amounts of T-bet (Fig. 3B). Thus, there was a correlation between the NCR ILC3 (FM+) phenotype and a reduced abundance of T-bet, a result that was also found in the in vitro differentiation assays (Fig. 3A). We further investigated the role of T-bet in the various subsets of ILC3s by crossing R26eYFP/+ Ncr1-iCre mice with Tbx21−/− mice; Tbx21 encodes T-bet. The number of cells of the NCR ILC3 (FM+) subset in the small intestine was reduced in the absence of T-bet (Fig. 3C); however, the effect of T-bet deficiency was less marked for these cells than for the NCR+ ILC3s, which were almost undetectable in the absence of T-bet, consistent with previous reports (10, 11, 25).

Constitutive TGF-β signaling impairs the generation of NCR+ ILC3s in vivo

We investigated those environmental cues other than Notch that might influence the plasticity between the NCR ILC3 (FM), NCR ILC3 (FM+), and NCR+ ILC3 subsets. TGF-β is a widely expressed cytokine that plays a critical role in regulating the proliferation and differentiation of various immune cell types, including CD4+ T cells in particular (13). The transcripts for both subunits of the TGF-βR, Tgfbr1 and Tgfbr2, are present in ILCs (18), which suggests a role for the TGF-β signaling pathway in the regulation of ILC3s. We investigated this possibility while minimizing confounding effects due to the widespread expression of TGF-βRs on hematopoietic and nonhematopoietic cells by performing experiments with TgfbRICA mice, which express a constitutively active (CA) form of TGF-βRI in a conditional manner (26). We crossed the TgfbRICA mice to the Ncr1-iCre strain to investigate the effect of TGF-β on the differentiation of the NCR ILC3 (FM+) and NCR+ ILC3 subsets because these two populations are the only ILC3s that exhibited a history of NKp46 expression. In the small intestine, NCR ILC3 generation was unaffected in TgfbRICA Ncr1-iCre mice, but the development of the NCR+ ILC3 subset was markedly impaired (Fig. 4). A similar trend was seen in the cecum and the colon, although the small size of the NCR+ ILC3 subset in the colon resulted in a less pronounced effect of TGF-β in these regions of the gut (fig. S3). We completed our analysis of the various ILC3 subsets with the alternative markers Nrp1 and CCR6. RORγt+ cells formed three populations that were defined on the basis of the cell surface expression of NKp46 or CD127: NKp46 CD127 cells were CCR6 and Nrp1; NKp46 CD127+ cells were CCR6+/−, but mostly Nrp1+; and NKp46+ CD127+ cells were Nrp1+/−, but mostly CCR6 (fig. S4A). The constitutive activity of the TGF-β pathway in TgfbRICA Ncr1-iCre mice was associated with a decrease in the size of Nrp1 CCR6 subset of RORγt+ CD127+ cells in the small intestine (fig. S4B), which is consistent with the smaller size of the NCR+ ILC3 subset in these mice. Thus, TGF-β signals impaired the generation of NCR+ ILC3s in vivo.

Fig. 4 Characterization of ILCs in the small intestine of TgfbRICA Ncr1-iCre mice at steady state.

Left: Flow cytometric analysis of NKp46 and RORγt expression in Lin CD45.2+ ILC3 populations isolated from the small intestine of TgfbRICA and TgfbRICA Ncr1-iCre mice. Right: Histograms show the percentages and numbers of cells in the indicated populations. Data are pooled from five independent experiments, with two to three mice per experiment. Statistical analyses were performed by two-way analysis of variance (ANOVA) with Bonferroni correction. ****P < 0.0001.

TGF-β signaling inhibits the generation of NCR+ ILC3s in vitro

We further dissected the relationship between TGF-β signaling and Notch by analyzing the fate of the NCR and NCR+ populations with normal or constitutively active TGF-βRI in vitro. Purified NCR ILC3s from TgfbRICA Ncr1-iCre mice did not give rise to NCR+ ILC3s when cultured on OP9 cells, as expected (Fig. 5A). By contrast, control NCR ILC3s gave rise to substantial numbers of NCR+ ILC3s when exposed to Notch ligands (that is, when cultured on OP9-DL1 cells), but markedly smaller numbers of these cells were generated from NCR ILC3s purified from TgfbRICA Ncr1-iCre mice and cultured on OP9-DL1 cells (Fig. 5A). This difference could not be attributed to increased rates of cell death or reduced rates of proliferation (fig. S5). Thus, activation of the TGF-βRI pathway impaired the differentiation of NCR ILC3s into NCR+ ILC3s. Purified NCR+ ILC3s cultured on OP9 cells displayed substantially reduced cell surface expression of NKp46 than similar cells exposed to Notch signaling. NKp46 was more strongly reduced in NCR+ ILC3s from TgfbRICA Ncr1-iCre mice, and this loss was sixfold lower in the absence of Notch signaling (Fig. 5B). Thus, strong TGF-β signaling may also drive the reversion of NCR+ cells to an NCR fate. We hypothesized that if TGF-β was critical for this reversion in vitro, then the removal of the TGF-β signal would release this block on differentiation. We tested this hypothesis by generating TgfbRIIfl/fl Ncr1-iCre mice lacking the TGF-βRII subunit, which forms a complex with TGF-βRI to enable the induction of signaling by different isoforms of TGF-β. Analysis of ILC3 subsets in vivo showed that the percentages of the NCR and NCR+ subsets had returned to those of wild-type mice (fig. S6A). Furthermore, a loss of TGF-β signaling in vitro resulted in substantial increases in NCR+ ILC3 cell development in the presence of Notch signaling, as well as in the generation of NCR+ cells even in the absence of Notch (fig. S6B). Thus, TGF-β signaling promotes the conversion of NCR+ ILC3s into NCR ILC3s in vitro. TGF-β and Notch therefore act in opposition to regulate the balance between NCR and NCR+ ILC3s.

Fig. 5 In vitro differentiation of NCR+ ILC3 and NCR ILC3 subsets purified from the small intestine of TgfbRICA and TgfbRICA Ncr1-iCre mice and cultured in the presence or absence of Notch ligands.

(A and B) Gating strategies and representative flow cytometric profiles for the sorting purity of NKp46 ILC3s (A, top) and NKp46+ ILC3s (B, top). (A) NCR ILC3s and (B) NCR+ ILC3s were purified from the small intestine and cultured on OP9 or OP9-DL1 stromal cells as indicated. After 6 days of culture, live cells were analyzed by flow cytometry for the cell surface expression of NKp46. Numbers indicate the percentages of cells within the indicated gate. Right: Histograms show the percentages of NKp46+ cells (A) or NKp46 cells (B). Data are pooled from three to four independent experiments, with three mice per experiment, each point representing a replicate of three mice. Statistical analyses were performed by two-way ANOVA with Bonferroni correction. ****P < 0.0001. PI, propidium iodide.

DISCUSSION

The generation of NCR+ ILC3s from NCR ILC3s is dependent on an increase in the abundance of T-bet (10, 11). NCR+ ILC3s have already been reported to be able to adopt alternative fates. In mice in response to IL-12, NCR+ ILC3s reduce their amount of RORγt, which is accompanied by increased T-bet abundance and IFN-γ production, resulting in ILC1-like functions (10, 22). In humans, the conversion of ILC3s into ILC1s is driven by tissue-resident CD103+ dendritic cells, which constitutively produce IL-12, whereas IL-23 drives the reverse transition, which can be accelerated by environmental cues, such as IL-1β and retinoic acid (27, 28). These data highlight the plasticity of ILC3s, which is in contrast to the apparent stability of conventional NKp46+ NK cells (29). Here, we revealed the existence of a previously unknown ILC3 population, the NCR ILC3 (FM+) subset, as well as plasticity between NCR and NCR+ ILC3s. The NCR ILC3 (FM+) subset was functional, as shown by its capacity to produce IL-22.

The identification of the NCR ILC3 (FM+) subset in R26eYFP/+ Ncr1-iCre and Il22eGFP/eGFP Ncr1-iCre reporter mice suggested that these cells could be direct precursors of NCR+ ILC3s or that they corresponded to a population of NCR+ ILC3s that had lost NKp46. Our in vitro data indicated that NCR+ ILC3s may be converted into the NCR ILC3 (FM+) subset; however, T-bet is mandatory for the generation of NCR+ ILC3s (10, 11) and a substantial fraction of the NCR ILC3 (FM+) subset remained in the absence of T-bet. Thus, cells of the NCR ILC3 (FM+) subset might not only correspond to NCR+ ILC3 revertants. This hypothesis is further supported by the absence of reversion of NCR+ ILC3s to NCR ILC3s upon adoptive transfer in vivo, although these experimental settings arguably have limitations, because they do not provide information on various signals throughout the life span of mice.

Together, our findings suggest that there are two nonmutually exclusive pathways leading to the differentiation of NCR ILC3s into NCR+ ILC3s. On the one hand, cells of the NCR ILC3 (FM) subset become cells of the NCR ILC3 (FM+) subset and then NCR+ ILC3 cells on induction of T-bet (Fig. 6). In this case, cells of the NCR ILC3 (FM+) subset arose from cells of the NCR ILC3 (FM) subset in which Ncr1 expression was activated, but was transient and not associated with the cell surface expression of NKp46, which requires more stable Ncr1 expression. Our results suggest that this induction of Ncr1 expression in NCR ILC3s is partly independent of T-bet. On the other hand, cells of the NCR ILC3 (FM) subset become NCR+ ILC3 cells once T-bet abundance is increased, but a fraction of these NCR+ ILC3 cells revert to the NCR ILC3 (FM+) subset (Fig. 6). In both conditions, the maintenance of NCR+ ILC3 is dependent on sustained Notch signals. These data are consistent with previous findings implicating Notch signaling in the cell surface expression of NKp46 (11, 19, 20). Thus, our data reveal a previously unsuspected degree of flexibility in the differentiation program, regulating the balance between NCR and NCR+ ILC3s.

Fig. 6 Schematic model of the reciprocal regulation of ILC3 differentiation by Notch ligands and TGF-β.

The identification of the NCR ILC3 (FM+) subset suggested that these cells could be direct precursors of NCR+ ILC3s (arrow 1) or that they corresponded to a population of NCR+ ILC3s that had lost NKp46 (arrow 2). Increasing expression of T-bet is associated with loss of CCR6 and Nrp1 as NCR ILC3s transition to NCR+ ILC3s. Notch is a key driver of the differentiation of NCR ILC3s and the NCR ILC3 (FM+) subset into the NCR+ ILC3 (FM+) subset, which is opposed by TGF-β. In vitro, TGF-β can drive the reversion of the NCR+ ILC3 (FM+) subset to the NCR ILC3 (FM) subset, whereas the constitutive activation of TGF-βR1 in vivo counterbalances Notch signals, maintaining the homeostatic balance of ILC3 subsets.

In the search for factors regulating ILC3 plasticity, we hypothesized that TGF-β could participate in local changes to gut inflammatory and homeostatic signals. Using a model selectively targeting TGF-β signaling to NCR+ cells, we showed that acute TGF-β signaling regulated the balance between NCR and NCR+ ILC3s. Sustained TGF-βR signaling was thus detrimental to the development of NCR+ ILC3s, revealing TGF-β as an important extrinsic signal regulating the balance between NCR and NCR+ ILC3s. However, the absence of any marked alteration of ILC3 number or function in TgfbRIIfl/fl Ncr1-iCre mice indicated that TGF-β played a redundant role in the regulation of ILC3 differentiation at steady state. In TH2 cells, TGF-β impairs cell differentiation by inhibiting the expression of GATA3 (encoded by Gata3) (30). TGF-β may also act through GATA3 in ILC3s, because NCR+ ILC3s require GATA3 for their development (15). Nevertheless, enhanced TGF-β signals impaired the Notch-dependent differentiation of these cells that was orchestrated by T-bet, highlighting the interplay between transcriptional regulators in the homeostatic balance between NCR and NCR+ ILC3s.

ILC3s participate in immunity through IL-22–dependent mechanisms (31, 32), a feature that is particularly important when the adaptive arm of the immune response is impaired (21). ILC3s also display proinflammatory features, for example, by secreting IL-17 in a model of inflammatory bowel disease (33). In addition, a description of ILC3 infiltrates in lung tumors raises the possibility that these cells also participate in tumor immune surveillance (34). It remains unclear whether these outcomes depend on a particular ILC3 subset. Further investigations are required to determine the relative contributions of TGF-β and Notch under these conditions, but these pathways provide critical checkpoints in the modulation of homeostasis and the development of ILC3 subsets.

MATERIALS AND METHODS

Mice

All of the mice used were bred and maintained in the specific pathogen–free facilities of WEHI (Walter and Eliza Hall Institute), CIRI (Centre International de Recherche en Infectiologie), or CIML (Centre d’Immunologie de Marseille-Luminy). Ncr1-iCre (29), R26eYFP/+ [B6.129X1-Gt(ROSA)26Sortm1(EYFP)Cos/J] (35), R26RFP/RFP (36), Il22eGFP/eGFP (21), Tbx21−/− (37), TgfbRIIfl/fl (38), and TgfbRICA (26) mice were used. All animal experimentation was performed with the approval of either the Animal Ethics Committees of the WEHI in accordance with the National Health and Medical Research Council (NHMRC) of Australia guidelines or the Animal Ethics Committees of the CIRI and CIML in accordance with the European law.

Isolation of lymphoid cells

Intestinal lymphoid cells were isolated by incubating the intestine with Ca2+- and Mg2+-free Hanks’ medium or phosphate-buffered saline (PBS) supplemented with 1 mM EDTA, 15 mM Hepes, and 10% fetal calf serum (FCS) for 30 min at 37°C with gentle shaking to remove intestinal epithelial cells. Supernatants were discarded, and the tissues were then incubated with gentle shaking for 45 min in collagenase type III [1 to 1.5 mg/ml (w/v); Worthington], deoxyribonuclease I (200 μg/ml; Roche), and dispase (0.4 U/ml; Sigma-Aldrich)] in mouse tonicity RPMI 1640 and 2% (v/v) FCS or in RPMI 1640 supplemented with 15 mM Hepes and collagenase type VIII (300 U/ml; Sigma-Aldrich). The preparations were filtered, and the cells were isolated on a 40%/100% Percoll gradient (GE Healthcare). Lymphocytes were recovered from the interface and washed twice. Splenocyte suspensions were obtained by the mechanical disruption of mouse spleens, and blood cells were then lysed in red blood cell (RBC) lysis solution (eBioscience). Bone marrow cell suspensions were obtained by flushing femurs and tibias, and blood cells were then lysed in RBC lysis solution (eBioscience). Mice were anesthetized and immediately perfused with PBS for the collection of the liver and lungs. Livers were mechanically disrupted in PBS in a cell strainer, and the cell suspensions were washed three times with PBS. Lungs were incubated for 45 min at 37°C in collagenase IV (Sigma-Aldrich; 400 U/ml in Hanks’ balanced salt solution with Ca2+ and Mg2+), and cell suspensions were then obtained by mechanical disruption. Liver and lung samples were then enriched in lymphocytes by centrifugation on a 37.5%/67.5% Percoll gradient (GE Healthcare).

Flow cytometry

To analyze cell surface molecules, cells were incubated with antibodies against the following surface markers: CD19 (clone 1D3), CD3ε (clone 145-2C11), NKp46 (clone 29A1.4), NK1.1 (clone PK136), KLRG-1 (clone 2F1), CD45.2 (clone 104), CD117 (clone 2B8), Nrp1 (polyclonal), CCR6 (clone 29-2L17), CD122 (clone TM-β1), CD27 (clone LG.3A10), CD11b (clone M1/70), CD11c (clone N418), and CD127 (clone A7R34). Live cells were identified by staining with DCM e506, DCM blue fluorescent (Invitrogen), or by lack of staining with propidium iodide. For intracellular staining, cells were stained for cell surface markers and with Fixable Viability Dye (eBioscience) and then were fixed and permeabilized with the eBioscience intracellular staining kit or the Cytofix/Cytoperm kit (BD Biosciences). The cells were then stained with antibodies against RORγt (clone Q31-378, BD Biosciences or AFKJS-9, eBioscience), T-bet (clone eBio4B10), Ki-67 (clone B56), or IL-22 (clone IL22JOP, or provided by J. C. Renauld; antibody was coupled to Alexa Fluor 647 with the antibody labeling kit from Life Technologies). IL-22 production was induced by stimulating cells with IL-23 (10 to 40 ng/ml) for 4 to 5 hours in the presence of BD GolgiStop and BD GolgiPlug (1:1000; BD Biosciences) in complete RPMI 1640 medium [containing 10% FCS, 50 mM 2-mercaptoethanol, 1 mM l-glutamine, penicillin (100 U/ml), and streptomycin (100 μg/ml)] at 37°C. Flow cytometric analysis was performed with an LSRFortessa X-20 or LSR II flow cytometer (BD Biosciences) and FlowJo software (FreeStar).

Cell purification

Lamina propria lymphocytes from the small intestine (pooled from three mice) were stained with antibodies against cell surface markers and then were purified on a BD FACSAria cell sorter (BD Biosciences). Cells were sorted into 50% (v/v) FCS/PBS or the appropriate cell culture medium in tubes previously coated with heat-inactivated FCS.

In vitro differentiation of ILC subsets on OP9 or OP9-DL1 stromal cells

ILC3 populations were purified from the small intestine by flow cytometry, and 0.5 × 104 to 2 × 104 cells were cultured on OP9 or OP9-DL1 stromal cells in α–minimum essential medium containing GlutaMAX, 10% FCS, 50 μM 2-mercaptoethanol, 1 mM l-glutamine, penicillin (100 U/ml), and streptomycin (100 μg/ml) supplemented with Flt3L (20 ng/ml), 2% supernatant of an IL-7–producing cell line, and stem cell factor (10 ng/ml). The culture medium was changed on day 4 and the cells are analyzed on day 8 (Fig. 3A), or alternately, cultures were analyzed on day 6 (Fig. 5 and fig. S6).

RT-qPCR analysis

Total RNA was prepared from purified ILC populations with the RNeasy Plus Micro Kit (Qiagen). We generated complementary DNA (cDNA) from total RNA with oligo(dT) primers and the ThermoScript Reverse Transcriptase (Invitrogen). Real-time qPCR assays were performed with the SensiMix SYBR No-ROX Kit (Bioline). All qPCRs were performed on a CFX384 Real-Time System (Bio-Rad). Analyses were performed in duplicate or triplicate, and the abundances of mRNAs of interest were normalized with respect to that of Hprt mRNA. The primers used for the analyses were as follows: Hprt, 5′-GGGGGCTATAAGTTCTTTGC-3′ (forward) and 5′-TCCAACACTTCGAGAGGTCC-3′ (reverse); Ncr1, 5′-ATGCTGCCAACACTCACT-3′(forward) and 5′-GATGTTCACCGAGTTTCCATTTG-3′ (reverse).

mRNA-seq analysis

Total RNA was prepared from purified ILC populations for RT-qPCR with an RNeasy Mini Kit (Qiagen). cDNA was synthesized from total RNA with oligo(dT) primers and ThermoScript Reverse Transcriptase (Invitrogen). Real-time PCR was performed with the SensiMix SYBR No-ROX Kit (Bioline). About 5 × 105 Lin (CD3CD19) CD45.2+ cells from each ILC subset (NCRFM, NCRFM+, and NCR+FM+) were sorted from the intestines of FM mice, and one to two biological replicates were generated and subjected to 100–base pair single-end sequencing on an Illumina HiSeq 2000 sequencer at the Australian Genome Research Facility (Melbourne, Australia). About 30 million reads were generated for each replicate and aligned to the GRCm38/mm10 build of the Mus musculus genome using the Subread aligner (39). Genewise counts were obtained with featureCounts (40). Reads overlapping exons in annotation build 38.1 of the National Center for Biotechnology Information RefSeq database were included. Genes were filtered from downstream analysis if they failed to achieve a CPM (counts per million mapped reads) value of at least 1 in at least one library. Counts were converted to log2 CPM, quantile-normalized, and precision-weighted with the voom function of the limma package (41, 42). The log2 RPKM expression values generated from fitting linear models to genes were used for data analysis (43) (Gene Expression Omnibus accession no. GSE79410).

Statistical analysis

Statistical significance was determined by Student’s t tests, Kruskal-Wallis tests, and two-way ANOVA with Bonferroni correction (Prism 5, GraphPad Software). The degree of statistical significance is indicated as follows: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/9/426/ra46/DC1

Fig. S1. Characterization of ILCs in the small intestine of R26RFP/RFP Ncr1-iCre mice at steady state.

Fig. S2. Ncr1 FM of NKp46+ and NKp46 hematopoietic cells.

Fig. S3. Analyses of ILC3 populations in the colon and cecum in the presence of a constitutively active form of TGF-βRI.

Fig. S4. Analyses of small intestinal ILC3 populations expressing a constitutively active form of TGF-βRI in vivo.

Fig. S5. In vitro analyses of the differentiation of small intestinal ILC3 populations expressing a constitutively active form of TGF-βRI.

Fig. S6. Small intestinal ILCs of TgfbRIIfl/fl Ncr1-iCre mice.

REFERENCES AND NOTES

Acknowledgments: We thank F. Guimaraes and the animal and flow cytometry facilities of WEHI and CIML for technical assistance. Funding: The laboratory of E.V. is supported by the European Research Council (THINK Advanced Grant), the Ligue Nationale contre le Cancer (Equipe Labellisée), and institutional grants from INSERM, CNRS, and Aix-Marseille University to CIML. E.V. is a scholar of the Institut Universitaire de France. This work was supported by grants and fellowships from the NHMRC of Australia (GNT 1027472, 1054925, 1049407, and 1078671 to G.T.B., C.S., N.D.H., and M.J.S.), an NHMRC Dora Lush Postgraduate Research Scholarship (to L.C.R.), and an Australian Research Council Future Fellowship (to G.T.B.). This work was made possible through Victorian State Government Operational Infrastructure Support and Australian Government NHMRC IRIIS. Author contributions: C.V. and L.C.R. designed the research, performed the experiments, and analyzed the data; M.J.H.G.-M., C.S., W.S., L.B., T.W., N.D.H., and M.J.S. performed the experiments, analyzed the data, and provided compound mouse strains; C.V., G.T.B., and E.V. devised the concept, designed the research, supervised the study, and wrote the manuscript, with the help of the other coauthors. Competing interests: E.V. is the cofounder of and a shareholder in Innate Pharma. The other authors declare that they have no competing financial interests.
View Abstract

Navigate This Article