IL-7 and IL-15 independently program the differentiation of intestinal CD3⁻NKp46⁺ cell subsets from Id2-dependent precursors

To understand the developmental relationship between distinct intestinal CD3⁻NKp46⁺ subsets, we first assessed the role for the transcriptional inhibitor Id2 in the homeostasis of these cells. Previous studies have shown that Id2 plays a critical role in the development of NK1.1⁺ NK cells (Yokota et al., 1999), and is essential for the development of LTi cells that express the transcription factor Rorc and coordinate programmed lymphoid tissue genesis during fetal life and inducible lymphoid structures in adult mice (Eberl and Littman, 2004). By comparing lamina propria lymphocytes (LPLs) from WT and Id2-deficient (Id2^−/−) mice, we found that normal development of all intestinal CD3⁻NKp46⁺ cell subsets required Id2. Percentages of intestinal CD3⁻NKp46⁺ cell subsets that differentially express CD127 and/or NK1.1 were severely reduced in the absence of Id2 (Fig. 1 A), resulting in dramatically reduced absolute numbers of these cells (WT mice, 1.5 ± 0.2 × 10⁵ cells; Id2^−/− mice, ∼10³ cells). In contrast, intestinal T cell homeostasis in the lamina propria was minimally perturbed (Fig. S1), whereas Id2 deficiency markedly reduced the frequency of splenic CD3⁻NKp46⁺ cells (essentially NK1.1⁺ NK cells; Fig. 1 A), consistent with previous reports (Yokota et al., 1999; Kim et al., 2004; Boos et al., 2007).

Intestinal CD3⁻NKp46⁺ cells that express Rorc can produce IL-22 after stimulation with IL-23 in vitro (Cella et al., 2009). Id2 deficiency resulted in a strong reduction of RORγt expression (Fig. 1 B) within residual intestinal CD3⁻NKp46⁺ cells and a corresponding abrogation of IL-23–induced IL-22 production (Fig. 1 C), consistent with earlier studies showing an essential role for Rorc in constitutive and inducible IL-22 expression by gut CD3⁻NKp46⁺ cells (Satoh-Takayama et al., 2008; Luci et al., 2009; Sanos et al., 2009). Thus, differentiation of gut CD3⁻NKp46⁺ cells requires careful titration of E-box protein activity via Id2. These results demonstrate that the development and normal homeostasis of all intestinal CD3⁻NKp46⁺ cell subsets requires Id2, and reinforce the notion that intestinal CD3⁻NKp46⁺ cells are developmentally related to classical NK1.1⁺ NK cells and LTi cells.

We previously demonstrated that intestinal CD3⁻NK1.1⁺ cells were strongly reduced in mice lacking IL-15 or the IL-2Rβ chain (Satoh-Takayama et al., 2008), reflecting their similar cytokine dependency with classical NK cells found in other tissues. In contrast, homeostasis of intestinal CD3⁻NKp46⁺ cells with the CD127⁺NK1.1⁻ phenotype was not affected by an absence of IL-15 signaling. Although these observations rule out a role for IL-15 in the generation of the intestinal CD3⁻NKp46⁺NK1.1⁻ cell subset, it nevertheless remained possible that these cells might represent IL-15–responsive precursors that could be recruited to develop into NK1.1⁺ NK cells under conditions where IL-15 availability increases. To assess the local inductive effects of IL-15 on the composition of intestinal CD3⁻NKp46⁺ cell subsets, we analyzed transgenic (Tg) mice in which IL-15 is overexpressed in the intestinal epithelium (Ohta et al., 2002). In these IL-15 Tg mice, local production of IL-15 has been previously shown to augment numbers of CD8⁺ T cells that express NK1.1 (Ohta et al., 2002). Total absolute numbers of LPLs were increased in IL-15 Tg mice (WT mice, 6.5 ± 3.3 × 10⁶ cells; IL-15 Tg mice, 17.9 ± 9.4 × 10⁶ cells), which increased the absolute numbers of intestinal lamina propria CD3⁻NKp46⁺ cells two- to threefold compared with non-Tg littermates (WT mice, 1.5 ± 0.2 × 10⁵ CD3⁻NKp46⁺ cells; IL-15 Tg mice, 4.3 ± 0.4 × 10⁵ CD3⁻NKp46⁺ cells). When comparing CD127 versus NK1.1 expression profiles, increased IL-15 availability strongly perturbed the relative frequencies of CD127/NK1.1 subsets with amplified percentages of NK1.1⁺CD127⁻ cells in both the lamina propria and intraepithelial compartments (Fig. 2 A). This resulted in significantly increased absolute numbers of lamina propria NK1.1⁺CD127⁻ cells compared with WT mice (P = 0.09; Fig. 2 B). Remarkably, IL-15 overexpression had no effect on the absolute numbers of CD127⁺NK1.1⁻ or CD127⁺NK1.1⁺ cells (Fig. 2 B). Moreover, Rorc expression and IL-23–induced IL-22 production from intestinal CD3⁻NKp46⁺ cells were similar between WT and IL-15 Tg mice (Fig. 2, C and D). These results reinforce the concept that NK1.1⁻ and NK1.1⁺ subsets of intestinal CD3⁻NKp46⁺ cells represent two distinct and independent lineages with respect to IL-15 responsiveness. Nevertheless, these results do not formally rule out the possibility that the CD127⁺NK1.1⁻Rorc⁺ subset of intestinal CD3⁻NKp46⁺ cells harbors IL-15–independent NK cell precursors.

Our previous studies demonstrated that development of the CD127⁺NK1.1⁻ subset of intestinal CD3⁻NKp46⁺ cells was dependent on γ_c cytokine signaling (Satoh-Takayama et al., 2008). As these cells express CD127, we assessed whether IL-7 was required for their homeostasis. Although the frequency of intestinal CD3⁻NKp46⁺ cells was unchanged in Il7^−/− mice (Fig. 3, A and B), absolute numbers were slightly but not significantly increased (WT mice, 1.5 ± 0.2 × 10⁵ cells; Il7^−/− mice, 2.1 ± 0.9 × 10⁵ cells; P = 0.24). We found that the CD127⁺NK1.1⁻ subset of intestinal CD3⁻NKp46⁺ cells was strongly reduced in the absence of IL-7 compared with WT mice (Fig. 3, A and B). In contrast, the generation of intestinal CD3⁻NK1.1⁺ cells was IL-7 independent (in fact, intestinal CD127⁻NK1.1⁺ cells were significantly increased in Il7^−/− mice), thereby distinguishing these cells from thymic NK cells (Vosshenrich et al., 2006). Intracellular staining revealed an absence of RORγt expression in residual CD3⁻NKp46⁺CD127⁺NK1.1⁻ cells from IL-7–deficient mice (Fig. 3 C), demonstrating that IL-7 is required for normal development of this subset of Rorc⁺ intestinal cells. As such, NKp46⁺CD127⁺NK1.1⁻ cells appear similar to fetal LTi cells in their strong IL-7 dependency (Meier et al., 2007; Schmutz et al., 2009). Lastly, we demonstrate that IL-23–induced IL-22 production by intestinal CD3⁻NKp46⁺ cells is abrogated in Il7^−/− mice (Fig. 3 D). These observations identify IL-7 as an essential cytokine for IL-22–producing CD3⁻NKp46⁺Rorc⁺ cells in the gut. Moreover, they demonstrate that IL-7 and IL-15 have divergent roles in regulating the homeostasis of intestinal CD3⁻NKp46⁺ cell subsets via distinct precursors.

Although analysis of Rorc^−/− mice showed that this transcription factor was not essential for normal NK cell development (Satoh-Takayama et al., 2008; Luci et al., 2009; Sanos et al., 2009), previous studies in both humans and mice demonstrated that lineage-negative populations of immature NK cells from diverse sources (lymph nodes, bone marrow, spleen, tonsils) did express Rorc and had NK cell precursor activity (Mebius et al., 1997; Chiossone et al., 2009; Cupedo et al., 2009; Hughes et al., 2009). As such, it remained possible that Rorc expression marked immature NK cells in the gut and elsewhere. To directly determine if Rorc⁺ cells in mice include precursors that can give rise to classical NK1.1⁺ cells, we performed a genetic cell fate–mapping experiment. Bacterial artificial chromosome Tg mice expressing the Cre recombinase under control of the Rorc(t) gene (Rorc(t)-Cre^TG mice; Lochner et al., 2008) were bred to mice bearing a loxP-flanked YFP reporter gene (Srinivas et al., 2001). In this context, YFP⁺ cells either express RORγt or have expressed RORγt at some previous stage in their developmental history. We confirmed previous results (Eberl and Littman, 2004) showing that all splenic T cells in these Tg mice express GFP (unpublished data), consistent with their development from Rorc⁺ double-positive thymocytes. We next examined intestinal CD3⁻NKp46⁺ cells in these mice. We found clear populations of YFP⁺NKp46⁺ cells in the intestinal lamina propria (51 ± 18%) and intraepithelial compartments (18 ± 12%) that were characterized by uniform CD127 expression and little or low NK1.1 expression (Fig. 4, A and B). This result is in accordance with our previous studies using Rorc-GFP mice that identified intestinal NKp46⁺ cell subsets (CD127⁺NK1.1^−/lo) that constitutively express Rorc (Lochner et al., 2008; Satoh-Takayama et al., 2008). In contrast, intestinal cells with the mature NK cell phenotype (identified as CD3⁻NK1.1⁺CD127⁻) were not represented in the YFP⁺NKp46⁺ subset (Fig. 4 B). Concerning tissues outside of the intestine, we further characterized YFP expression in classical NK cells (defined as CD3⁻NKp46⁺NK1.1⁺CD127⁻) from the bone marrow, spleen, liver and lymph nodes (Fig. 4 C). Similar to that observed in the gut, frequencies of YFP⁺ cells among NK cells from these tissues were extremely low (<1%). These results are inconsistent with the hypothesis that Rorc⁺ cells are obligatory precursors to the NK1.1⁺ cells within the various tissues analyzed, and argue against a major pathway of classical NK cell development from Rorc⁺ precursors. Although we do not rule out the possibility that a small subset of NK1.1⁺ cells may be generated from Rorc⁺ precursors in vivo under steady-state conditions, any such cells are likely to be short-lived because they do not accumulate to appreciable levels. Collectively, the available data strongly suggest that intestinal CD3⁻NKp46⁺NK1.1⁻ cells are a distinct lineage and not precursors to classical NK1.1⁺ NK cells.

The recent discovery of CD3⁻NKp46⁺ cells in the intestinal mucosa that lack typical NK cell features (i.e., “natural” cytotoxicity) but express Rorc and IL-22 has raised questions about the biological roles and developmental origins of these unusual innate cells and their relationship (if any) to classical NK cells (Colonna, 2009; Vivier et al., 2009). Our results provide evidence for a lineage relationship between NCR-expressing, Rorc⁺, IL-22–producing (NCR22) cells, NK cells, and LTi cells, in that all of these cell subsets require the transcription factor Id2 for normal development. Still, our results clearly show that intestinal NK cells and NCR22 cells arise from independent developmental pathways that are driven by distinct cytokines (IL-7, IL-15). How IL-7 stimulates NCR22 cells (during development and/or as a peripheral survival factor) remains to be determined. Previous studies have considered Rorc expression as a potential marker for cells that harbor NK cell precursor activity (Chiossone et al., 2009; Hughes et al., 2009). In contrast, using an in vivo fate-mapping approach, we found that the vast majority of classical NK cells do not pass through a Rorc⁺ stage. It is likely that the NK cell precursor populations characterized in previous studies (Chiossone et al., 2009; Hughes et al., 2009) were heterogeneous mixtures of Rorc⁺ and Rorc⁻ cells, with only the latter having the potential to differentiate into mature NK cells. This conclusion is consistent with the results presented in the paper by Crellin et al. in this issue. These authors found that lineage-negative RORC⁺ LTi-like cells in the human tonsil are precursors to CD56⁺NKp46⁺RORC⁺ cells that can produce IL-22 but lack conventional NK cell precursor activity. Their observations in the human system mirror our results demonstrating that mouse NCR22 cells expressing Rorc are not canonical NK cell precursors. NCR22 cells are also clearly different from the earliest immature NK cells that express NK1.1 (but not NKp46 or DX5) and are committed to the NK cell lineage (Rosmaraki et al., 2001; Vosshenrich et al., 2005; Walzer et al., 2007). These more precise phenotypic definitions may provide new access points to further dissect the signals that direct the specification of NK versus NCR22 innate cell lineages at the precursor stage.

C57BL/6 mice were purchased from Janvier. Ncr1^GFP/+ mice with the GFP reporter under the Ncr1 promoter elements (provided by O. Mandelboim, Hebrew University, Jerusalem, Israel; Gazit et al., 2006), as well as mice deficient in Id2 (provided by Y. Yokota, University of Fukui, Fukui, Japan; Yokota et al., 1999), IL-7 (von Freeden-Jeffry et al., 1995), or IL-15 (Kennedy et al., 2000), have been previously described. Mice with a human IL-15 transgene driven by the intestinal epithelial cell–specific T3b promoter (Ohta et al., 2002) were the gift of H. Kiyono (Institute of Medical Science, Tokyo, Japan) and were generously provided by N. Cerf-Bensussan (Institut National de la Santé et de la Recherche Médicale U793, Paris, France). For fate-mapping studies, mice harboring a recombinant bacterial artificial chromosome encoding RORγt-Cre (Lochner et al., 2008) were crossed with mice containing a loxP-flanked EYFP reporter in the ROSA locus (B6.129X1-Gt(ROSA)26Sor^tm1(EYFP)Cos/J; Srinivas et al., 2001). All mice were housed under specific pathogen-free conditions at the Laboratory Animal Facilities of the Institut Pasteur. Mice were analyzed at 6–12 wk of age. All animal experiments were approved by the Animal Care and Use Committee of the Institut Pasteur and were performed in accordance with French law.

Total splenocytes and small intestinal LPLs were prepared for flow cytometry as previously described (Satoh-Takayama et al., 2008). The cell-surface phenotype was analyzed using commercially available fluorochrome-conjugated mAbs (BD and eBioscience). Acquisitions were made using a flow cytometer (FACSCanto II; BD), interfaced to FACSDiva software (BD), and analyzed using FlowJo software (Tree Star, Inc.).

Total LPL preparations diluted in RPMI 1640 with 10% FCS (10⁶ cells/well) were left unstimulated or stimulated with 40 ng/ml IL-23 for 4 h in the presence of GolgiStop (BD). After surface staining and incubation with a fixable reagent (LIVE/DEAD; Invitrogen), cells were fixed in 4% paraformaldehyde and processed for intracellular cytokine detection. Cells were stained with allophycocyanin-conjugated anti–IL-22 mAb (MH22B2.2; provided by J.-C. Renauld, Ludwig Institut, Brussels, Belgium) in staining buffer (0.1% saponin in HBSS) before extensive washing in PBS. For RORγt detection, an anti–mouse/human RORγt-PE (clone AFKJS-9; eBioscience) was used. Nonspecific staining of the anti-RORγt was assessed using Rorc-deficient mice (Eberl et al., 2004).

Values are presented as means ± SD. The statistical significance of differences between groups was determined by the unpaired Student’s t test. P < 0.05 was considered significant.

Fig. S1 shows the minor effect of Id2 deficiency on T cell development in the lamina propria.

We would like to thank O. Mandelboim, N. Cerf-Bensussan, H. Kiyono, and Y. Yokota for providing access to Tg mice, and J.-C. Renauld for IL-22 antibodies. N. Satoh-Takayama and J.P. Di Santo designed the research; N. Satoh-Takayama, S. Lesjean-Pottier, and C.A.J. Vosshenrich performed research and analyzed data; P. Vieira, S. Sawa, and G. Eberl provided new reagents and analytical tools; and N. Satoh-Takayama and J.P. Di Santo wrote the paper.

This work was supported by grants to J.P. Di Santo from the Institut Pasteur and the Institut National de la Santé et de la Recherche Médicale, and as an Equipe Labelisé by the Ligue Nationale Contre le Cancer. N. Satoh-Takayama was the recipient of fellowships from the Association Pasteur Japon and the Uehara Memorial Foundation, Japan.

The authors have no conflicting financial interests.