1. The Kimmel Center for Biology and Medicine of the Skirball Institute, and,
2. Howard Hughes Medical Institute, Departments of Microbiology and Pathology, New York University School of Medicine, New York, New York 10016, USA
3. Immunology Program Benaroya Research Institute Seattle, Washington 98101, USA
4. Department of Immunology University of Washington School of Medicine Seattle, and,
5. Howard Hughes Medical Institute, Department of Immunology, University of Washington Seattle, Washington 98195, USA

Correspondence to: Dan R. Littman^1,2 Correspondence and requests for materials should be addressed to D.R.L. (Email: littman@saturn.med.nyu.edu).

When T lymphocytes are exposed to microbial antigens, they acquire diverse effector functions depending on which cytokines are produced by activated cells of the innate immune system¹². Differentiation of pro-inflammatory T_H17 cells requires the presence of IL-23, which is produced by activated dendritic cells^{13, 14, 15}. In vitro, however, T_H17 cell differentiation is independent of IL-23 and is induced by TGF- plus IL-6 or IL-21 (refs 6, 9–11). Both in vitro and in vivo differentiation of the T_H17 cell lineage require the upregulation of the orphan nuclear receptor RORt⁷. TGF- is also required to restrain inflammatory autoimmune responses¹⁶. Among its numerous properties is its ability to induce expression of Foxp3 in naive antigen-stimulated T cells, endowing the cells with regulatory or suppressor function⁸. Thus, TGF- can induce both regulatory and pro-inflammatory T cells, depending on whether pro-inflammatory cytokines such as IL-6 and, potentially, IL-23 are present^{11, 17}. Treatment of antigen-receptor-stimulated T cells with TGF- alone induces expression of both Foxp3 and RORt, but not of IL-17 (refs 7, 11). After such treatment, a significant proportion of cells co-expressed the two transcription factors (Fig. 1a and Supplementary Fig. 1a). To determine whether co-expression also occurs in vivo, we examined CD4⁺ T cells from the small intestinal lamina propria of heterozygous RORt–GFP (green fluorescent protein) knock-in mice, in which IL-17 is produced by TCR⁺GFP^int lymphocytes⁷. Foxp3 was expressed in about 10% of sorted GFP^int (RORt⁺) cells (Fig. 1b and Supplementary Fig. 1b). In addition, Foxp3 was expressed in approximately 17–20% of GFP^- lamina propria CD4⁺ T cells, consistent with the relatively large proportion of T_reg cells in the intestine.

Figure 1: Co-expression of Foxp3 and RORt in vitro and in vivo.

a, Naive CD4⁺ T cells were stimulated with anti-T-cell receptor (TCR) and 5 ng ml^-1 TGF- for 48 h, and were stained with 4,6-diamidino-2-phenylindole (DAPI; blue nuclear stain; UV channel), anti-ROR (red; Cy3 channel) and anti-Foxp3 (green; Cy5 channel) monoclonal antibodies. All panels are of the same section. The bottom-left image shows Foxp3-only, bottom-right shows ROR-only, top-right shows overlay of Foxp3 and ROR channels, and top-left shows overlay of all three channels. Foxp3, RORt and double-expressing (DP) cells are indicated with coloured arrows. b, Analysis of Foxp3⁺RORt⁺ cells from the small intestinal lamina propria. CD4⁺GFP^int and CD4⁺GFP^- cells were sorted from lamina propria of Rorc(t)^+/gfp and Rorc(t)^gfp/gfp mice, and Foxp3 expression was examined by intracellular staining. Results are representative of three experiments. The numbers indicate the percentage of total cells in each gate. c, Expression of IL-17 in Foxp3⁺RORt⁺ and Foxp3^-RORt⁺ T cells from small intestine. Foxp3 and IL-17 expression was examined by intracellular staining of sorted TCR⁺CD4⁺GFP^int cells from the lamina propria of Rorc(t)^+/gfp mice.

High resolution image and legend (322K)

We next performed a fate-mapping analysis to determine the proportion of IL-17⁺ small intestinal T cells that had expressed Foxp3 during their ontogeny. Mice expressing Cre recombinase under the regulation of the Foxp3 locus (Y.P.R. et al., submitted) were crossed with Rosa26-stop–YFP reporter mice¹⁸, and female progeny (Rosa26^stop–YFP/+; Foxp3^cre/+) were analysed for expression of yellow fluorescent protein (YFP). When inactivation of X-linked Foxp3 was taken into account, we found that approximately 15% of IL-17^- cells and 25% of IL-17⁺ cells had expressed Cre at some stage of development (Supplementary Fig. 2). The former represent Foxp3⁺ T_reg cells, whereas the latter are the minimal proportion of T_H17 cells that had expressed Foxp3 at some stage of their differentiation. These data suggest that Foxp3⁺ T cells can differentiate into T_H17 cells in vivo in the presence of pro-inflammatory cytokines.

Examination of IL-17 expression in heterozygous RORt–GFP knock-in mice revealed that RORt⁺Foxp3⁺ lamina propria T cells produced much less IL-17 than RORt⁺Foxp3^- cells, suggesting that Foxp3 may interfere with the ability of RORt to induce IL-17 (Fig. 1c). This is consistent with findings showing a >1,000-fold increase in Il17 messenger RNA, but little change in Rorc(t), in T_reg lineage cells that differentiate in the absence of Foxp3 (ref. 19). To investigate how Foxp3 may influence T_H17 cell differentiation, we asked whether its induction would influence the expression of IL-17 in TGF--stimulated T cells. In naive T cells that had been transduced with a retroviral vector encoding RORt, we found that, whereas IL-6 augmented the proportion of RORt-IRES-GFP⁺ cells that expressed IL-17, TGF- had a profound inhibitory effect even when added one day after transduction (Fig. 2a, b). Addition of TGF- was followed by a sharp increase in expression of Foxp3 in the CD4⁺ T cells, and both the level of Foxp3 mRNA and proportion of Foxp3⁺ cells were not affected by the expression of RORt (Fig. 2c).

Figure 2: TGF- inhibits RORt-directed IL-17 production by upregulating Foxp3.

a, The effect of TGF- on IL-17 expression after transduction of RORt. Naive CD4⁺ T cells incubated with TGF- from day -1 were transduced with control vector MSCV-IRES-GFP (MIG) or RORt-IRES-GFP (RORt) on day 0 (24 h after TCR activation), and IL-17 intracellular staining was performed on day 5. b, The inhibitory effect of TGF- when included at different times relative to transduction of RORt. The percentage of IL-17⁺ cells among GFP (RORt)⁺ cells is shown. c, Knockdown of TGF--induced Foxp3 expression with an shRNA against Foxp3 (LMP1066). Naive CD4⁺ T cells were stimulated as in a and co-transduced on days 0 and 1 with control retroviral construct MSCV-IRES-hCD2 (MCD2) or RORt-IRES-hCD2 (RORt) and the specific shRNA vector (LMP1066) or control vector (LMP). After transduction, the cells were cultured with or without TGF-, and Foxp3 expression was measured by intracellular staining on day 5. d, Restoration of RORt-induced IL-17 expression on knockdown of Foxp3. IL-17 expression was assessed in cells co-transduced as in c and gated for the level of GFP expression. The percentage of IL-17⁺ T cells in GFP^hi cell populations is shown. Results with additional shRNA vectors that failed to downregulate Foxp3 expression were similar to those with the control LMP vector. Representative data from three experiments are shown.

High resolution image and legend (186K)

To determine whether the inhibitory effect of TGF- on RORt is mediated by Foxp3, we knocked down expression of Foxp3 by using a short hairpin RNA (shRNA) vector. TGF--induced Foxp3 expression was reduced by the Foxp3-specific shRNA vector, but not by control hairpin vectors (Fig. 2c). Accordingly, TGF--mediated inhibition of RORt-directed IL-17 expression was partially reversed by Foxp3 knockdown (Fig. 2d). Consistent with the idea that this inhibition was mediated by Foxp3 upregulation, the most pronounced rescue of IL-17 expression occurred in cells that had lost the most Foxp3 expression (Supplementary Fig. 3). Thus, Foxp3 induced by TGF- inhibits the function of RORt.

These results prompted us to ask whether Foxp3 interacts with RORt to inhibit its function. Using a yeast two-hybrid screen, we previously found that human FOXP3 interacts with RAR-related orphan receptor A (RORA), and that an alternatively spliced isoform of FOXP3, lacking exon 2 (ref. 20), was deficient in this interaction²¹. We therefore examined whether mouse and human Foxp3 could similarly bind to RORt, and whether such interaction was necessary for inhibition of the RORt-mediated induction of IL-17. When Flag-epitope-tagged mouse Foxp3 was co-expressed with mouse RORt in 293T cells, the two proteins were co-immunoprecipitated (Fig. 3a), even in the presence of DNase I or ethidium bromide, suggesting that the interaction does not involve DNA. A similar interaction was observed between human RORT and FOXP3 (Fig. 3b). However, both mouse and human Foxp3 lacking the conserved exon 2-encoded sequence (Foxp3Ex2) had a substantially reduced association with RORt (Fig. 3a, b). We examined the localization of the two proteins by confocal microscopy of HeLa cells transfected with Flag-tagged mouse Foxp3 constructs with or without mouse RORt. Both Foxp3 and RORt were localized in the nucleus, but Foxp3 lacking the DNA-binding forkhead domain (Foxp3FKH) remained in the cytoplasm due to deletion of the nuclear localization signal in FKH²² (Supplementary Fig. 4). However, Foxp3FKH translocated to the nucleus when it was co-expressed with RORt, indicating that the Foxp3–RORt interaction is independent of FKH. Accordingly, Foxp3FKH co-immunoprecipitated with RORt in extracts of transfected 293T cells (data not shown). A combined Foxp3Ex2/FKH mutant remained in the cytoplasm even when it was co-expressed with RORt, further indicating that Foxp3 interacts with RORt by way of the exon 2-encoded sequence (Supplementary Fig. 4).

Figure 3: Foxp3 interacts with RORt and inhibits RORt-directed IL-17 expression.

a, Co-immunoprecipitation of Foxp3 and RORt from extracts of co-transfected 293T cells with or without DNaseI or ethidium bromide (EtBr). Cells were transfected with mouse RORt and Flag-tagged wild-type (Flag–Foxp3) or exon 2-deleted (Flag–Foxp3Ex2) Foxp3. Anti-Flag immunoprecipitates (IP) and total lysates were immunoblotted with anti-RORt antibody and anti-Flag antibody. b, Cells were transfected with Flag-tagged human RORT and full-length (FL) or the exon 2-deleted isoform of human FOXP3 (Ex2). Anti-Flag immunoprecipitates and total lysates were probed with anti-FOXP3 and anti-Flag antibodies. c, Naive CD4⁺ T cells were co-transduced with retroviruses encoding RORt (MIT vector, Thy1.1 reporter) and various mouse Foxp3 constructs (MIG vector, GFP reporter). IL-17 expression was assessed on day 4 in cells gated for expression of both Thy1.1 and GFP. Representative data from at least three experiments are shown in each of the panels.

High resolution image and legend (368K)

To investigate the role of the interaction between Foxp3 and RORt in the repression of RORt-induced transcription, we co-expressed these transcription factors in naive CD4⁺ T cells and examined expression of IL-17. Both mouse and human Foxp3 blocked RORt-directed IL-17 expression, but full suppression required the presence of the exon 2-encoded sequence in Foxp3, suggesting that the interaction between Foxp3 and RORt is essential (Fig. 3c and Supplementary Fig. 5). The ability of both mouse and human Foxp3 to repress RORt-induced IL-17 expression was abrogated by deletion of the FKH domain or a point mutation in this domain (R397W) that impairs FOXP3 DNA-binding activity and was identified in X-linked immunodeficiency, polyendocrinopathy, enteropathy (IPEX) syndrome in humans^{23, 24} (Fig. 3c and Supplementary Fig. 5). Therefore, Foxp3 can block the activity of RORt at least in part through an interaction involving a sequence encoded by exon 2, but the requirement for an intact FKH domain suggests that its DNA-binding activity also contributes to inhibition of IL-17 expression. Thus, Foxp3 may inhibit RORt-directed transcription through a mechanism similar to that proposed for its inhibition of IL-2 expression, involving its association with NFAT1 and Runx1 (refs 25 and 26). However, Foxp3Ex2 was as effective as the full-length protein in suppressing expression of IL-2 and interferon- (IFN-) in primary mouse T cells, indicating that, like the naturally occurring human spliced isoform²⁰, it retains regulatory functions and can, presumably, associate with both NFAT1 and Runx1 (Supplementary Fig. 6).

Our results suggest that Foxp3 may inhibit RORt activity on its target genes during T_H17 cell differentiation. To extend our analysis from Il17 to other potential RORt transcriptional targets, we examined the effect of TGF--induced Foxp3 on Il23r expression, which also requires the activity of RORt^{10, 11}. Forced expression of wild-type mouse Foxp3 inhibited IL-6/IL-21-induced Il23r expression, whereas Foxp3Ex2 had less inhibitory activity (Fig. 4a); this is consistent with the notion that Foxp3 inhibits the function of RORt through an interaction involving the sequence encoded by exon 2. Similar results were observed with Il22 expression in response to IL-6 or IL-21 (data not shown). Expression of Il22 and of Il23r in response to either IL-6 or forced expression of RORt was also inhibited by high concentrations of TGF- (refs 11, 27 and data not shown). However, at low concentrations, TGF- synergized with IL-6 and IL-21 to enhance expression of Il23r mRNA (Fig. 4b). As a consequence, addition of IL-23 to cultures containing high concentrations of TGF- had no effect on IL-17 expression, but significantly increased the number of IL-17⁺ cells and the level of IL-17 expression per cell when low concentrations of TGF- were used (Figs 4c, d and Supplementary Fig. 7). In contrast, induction of T_reg (Foxp3⁺) cells was optimal at high concentrations of TGF-, but there was little induction at TGF- concentrations at which IL-23 had a synergistic effect on expression of IL-17 (Fig. 4e).

Figure 4: TGF- concentration influences Il23r expression and levels of IL-17 in response to T_H17-inducing cytokines.

a, Foxp3-mediated inhibition of IL-6/IL-21-induced Il23r expression. This was measured in arbitrary units relative to expression of transcripts encoding actin. Naive CD4⁺ T cells were transduced with MIG, full length Foxp3 or Foxp3Ex2 viruses, and were treated with the indicated cytokines. RNA was isolated from GFP⁺ cells at day 2. Il23r expression was measured by real-time RT–PCR and was normalized to the actin level. Error bars represent standard deviations obtained using the standard curve method. b, Induction of Il23r mRNA in response to cytokines. Naive CD4⁺ T cells were stimulated with anti-CD3 and anti-CD28 throughout the culture period in the presence of the indicated combinations of cytokines. TGF- was titrated into the cultures at the concentrations of 5 ng ml^-1, 2.5 ng ml^-1, 1 ng ml^-1, 200 pg ml^-1, 40 pg ml^-1 or 8 pg ml^-1. Il23r mRNA expression was measured after 48 h by real-time RT–PCR and was normalized to the actin expression level. c, d, IL-23 enhancement of IL-17 expression at low concentrations of TGF-. The percentage of IL-17⁺ cells at 96 h of stimulation with the indicated cytokines is shown. Results in histogram format are shown in Supplementary Fig. 7. e, Induction of Foxp3 at different concentrations of TGF-. Representative data from at least three experiments are shown for each set of panels.

High resolution image and legend (138K)

TGF--induced Foxp3 expression is inhibited by IL-6 (ref. 17), IL-21 (ref. 10) and IL-23 (Supplementary Fig. 8). However, a substantial number of Foxp3⁺ cells differentiated in response to TGF-, even in the presence of IL-6, and many of these cells also expressed IL-17 (Supplementary Fig. 9a). Conversely, many of the IL-17⁺ cells also expressed Foxp3. Thus, RORt-dependent IL-17 expression can occur in the presence of Foxp3, but the level of Foxp3 may be insufficient to block RORt function or, alternatively, IL-6 may overcome the inhibitory function of Foxp3. To examine this possibility, we added IL-6 or IL-21 to cultures of cells transduced with both RORt and Foxp3. Under these conditions, the inhibitory effect of Foxp3 on IL-17 induction was largely circumvented, even though the level of Foxp3 protein was not affected (Supplementary Fig. 9b and data not shown); this suggests that IL-6 and IL-21 may have an additional post-translational effect on either Foxp3 or RORt.

Our data collectively suggest that T cells receiving a TGF- signal can acquire the potential to develop into either the T_reg or the T_H17 lineage. Foxp3 induction restrains the differentiation of inflammatory T_H17 cells in response to TGF- in the absence of other pro-inflammatory cytokines by inhibiting the activity of RORt. In the presence of pro-inflammatory cytokines, the suppression of Foxp3 expression and inhibitory function, together with the concurrent upregulation or stabilization of RORt expression, leads to full progression towards the T_H17 lineage (Supplementary Fig. 10). This process may be especially relevant in the intestinal lamina propria, in which TGF- can promote either T_H17 or T_reg cell lineage differentiation, depending on its local concentration. In this setting, a fine balance between RORt and Foxp3 may be critical for immune homeostasis. In line with the observation that more Foxp3⁺ T_reg cells were present in the gut of RORt-deficient mice (Fig. 1b), these mutant mice were also protected from autoimmune disease (ref. 7 and data not shown). Conversely, a decrease of Foxp3 expression and function and an increase of RORt expression tips the T_reg/T_H17 balance towards the T_H17 cell lineage. This may occur in some autoimmune diseases, as suggested by the finding that an Il23r polymorphism correlates with protection from Crohn's disease²⁸. These results therefore have important implications for how peripheral tolerance is maintained in the presence of potentially pro-inflammatory cytokines.

References

1. Weaver, C. T., Harrington, L. E., Mangan, P. R., Gavrieli, M. & Murphy, K. M. Th17: an effector CD4 T cell lineage with regulatory T cell ties. Immunity 24, 677–688 (2006) | Article | PubMed | ISI | ChemPort |
2. Weaver, C. T., Hatton, R. D., Mangan, P. R. & Harrington, L. E. IL-17 family cytokines and the expanding diversity of effector T cell lineages. Annu. Rev. Immunol. 25, 821–852 (2007) | Article | PubMed | ISI | ChemPort |
3. McKenzie, B. S., Kastelein, R. A. & Cua, D. J. Understanding the IL-23–IL-17 immune pathway. Trends Immunol. 27, 17–23 (2006) | Article | PubMed | ISI | ChemPort |
4. Lee, E. et al. Increased expression of interleukin 23 p19 and p40 in lesional skin of patients with psoriasis vulgaris. J. Exp. Med. 199, 125–130 (2004) | Article | PubMed | ISI | ChemPort |
5. Witowski, J., Ksiazek, K. & Jorres, A. Interleukin-17: a mediator of inflammatory responses. Cell. Mol. Life Sci. 61, 567–579 (2004) | Article | PubMed | ISI | ChemPort |
6. Veldhoen, M., Hocking, R. J., Atkins, C. J., Locksley, R. M. & Stockinger, B. TGF in the context of an inflammatory cytokine milieu supports de novo differentiation of IL-17-producing T cells. Immunity 24, 179–189 (2006) | Article | PubMed | ISI | ChemPort |
7. Ivanov, I. I. et al. The orphan nuclear receptor RORt directs the differentiation program of proinflammatory IL-17⁺ T helper cells. Cell 126, 1121–1133 (2006) | Article | PubMed | ISI | ChemPort |
8. Chen, W. et al. Conversion of peripheral CD4⁺CD25^- naive T cells to CD4⁺CD25⁺ regulatory T cells by TGF- induction of transcription factor Foxp3. J. Exp. Med. 198, 1875–1886 (2003) | Article | PubMed | ISI | ChemPort |
9. Korn, T. et al. IL-21 initiates an alternative pathway to induce proinflammatory T_H17 cells. Nature 448, 484–487 (2007) | Article | PubMed | ISI | ChemPort |
10. Nurieva, R. et al. Essential autocrine regulation by IL-21 in the generation of inflammatory T cells. Nature 448, 480–483 (2007) | Article | PubMed | ISI | ChemPort |
11. Zhou, L. et al. IL-6 programs T_H-17 cell differentiation by promoting sequential engagement of the IL-21 and IL-23 pathways. Nature Immunol. 8, 967–974 (2007) | Article |
12. Murphy, K. M. & Reiner, S. L. The lineage decisions of helper T cells. Nature Rev. Immunol. 2, 933–944 (2002) | Article |
13. Langrish, C. L. et al. IL-23 drives a pathogenic T cell population that induces autoimmune inflammation. J. Exp. Med. 201, 233–240 (2005) | Article | PubMed | ISI | ChemPort |
14. Aggarwal, S., Ghilardi, N., Xie, M. H., de Sauvage, F. J. & Gurney, A. L. Interleukin-23 promotes a distinct CD4 T cell activation state characterized by the production of interleukin-17. J. Biol. Chem. 278, 1910–1914 (2003) | Article | PubMed | ISI | ChemPort |
15. Cua, D. J. et al. Interleukin-23 rather than interleukin-12 is the critical cytokine for autoimmune inflammation of the brain. Nature 421, 744–748 (2003) | Article | PubMed | ISI | ChemPort |
16. Letterio, J. J. & Roberts, A. B. Regulation of immune responses by TGF-. Annu. Rev. Immunol. 16, 137–161 (1998) | Article | PubMed | ISI | ChemPort |
17. Bettelli, E. et al. Reciprocal developmental pathways for the generation of pathogenic effector T_H17 and regulatory T cells. Nature 441, 235–238 (2006) | Article | PubMed | ISI | ChemPort |
18. Srinivas, S. et al. Cre reporter strains produced by targeted insertion of EYFP and ECFP into the ROSA26 locus. BMC Dev. Biol. 1, 4 (2001) | Article | PubMed | ChemPort |
19. Gavin, M. A. et al. Foxp3-dependent programme of regulatory T-cell differentiation. Nature 445, 771–775 (2007) | Article | PubMed | ISI | ChemPort |
20. Allan, S. E. et al. The role of 2 FOXP3 isoforms in the generation of human CD4⁺ Tregs. J. Clin. Invest. 115, 3276–3284 (2005) | Article | PubMed | ISI | ChemPort |
21. Du, J., Huang, C., Zhou, B. & Ziegler, S. F. Isoform-specfic inhibition of ROR-mediated transcriptional activation by human FOXP3. J. Immunol. (in the press)
22. Schubert, L. A., Jeffery, E., Zhang, Y., Ramsdell, F. & Ziegler, S. F. Scurfin (FOXP3) acts as a repressor of transcription and regulates T cell activation. J. Biol. Chem. 276, 37672–37679 (2001) | Article | PubMed | ISI | ChemPort |
23. Wildin, R. S. et al. X-linked neonatal diabetes mellitus, enteropathy and endocrinopathy syndrome is the human equivalent of mouse scurfy. Nature Genet. 27, 18–20 (2001) | Article |
24. Lopes, J. E. et al. Analysis of FOXP3 reveals multiple domains required for its function as a transcriptional repressor. J. Immunol. 177, 3133–3142 (2006) | PubMed | ISI | ChemPort |
25. Ono, M. et al. Foxp3 controls regulatory T-cell function by interacting with AML1/Runx1. Nature 446, 685–689 (2007) | Article | PubMed | ISI | ChemPort |
26. Wu, Y. et al. FOXP3 controls regulatory T cell function through cooperation with NFAT. Cell 126, 375–387 (2006) | Article | PubMed | ISI | ChemPort |
27. Zheng, Y. et al. Interleukin-22, a T_H17 cytokine, mediates IL-23-induced dermal inflammation and acanthosis. Nature 445, 648–651 (2007) | Article | PubMed | ISI | ChemPort |
28. Duerr, R. H. et al. A genome-wide association study identifies IL23R as an inflammatory bowel disease gene. Science 314, 1461–1463 (2006) | Article | PubMed | ISI | ChemPort |
29. Eberl, G. et al. An essential function for the nuclear receptor RORt in the generation of fetal lymphoid tissue inducer cells. Nature Immunol. 5, 64–73 (2004) | Article |
30. Sun, Z. et al. Requirement for ROR in thymocyte survival and lymphoid organ development. Science 288, 2369–2373 (2000) | Article | PubMed | ISI | ChemPort |

Supplementary Information

Supplementary information accompanies this paper.

Acknowledgements

We thank P. Lopez and J. Hirst for assistance with cell sorting. We also thank J. Lafaille and D. Unutmaz for critical reading of the manuscript, and members of the Littman laboratory for their suggestions. L.Z., M.M.W.C. and I.I.I. were supported by fellowships from the Irvington Institute for Immunological Research, the Cancer Research Institute and the Crohn's and Colitis Foundation of America, respectively. This work was supported by the Howard Hughes Medical Institute (D.R.L., A.Y.R.), the Sandler Program for Asthma Research (D.R.L.), the National Multiple Sclerosis Society (D.R.L.), the Helen and Martin Kimmel Center for Biology and Medicine (D.R.L.), NIH grant AI48779 (S.F.Z.), and the JDRF Collaborative Center for Cell Therapy (S.F.Z.).

Author Contributions L.Z., J.E.L., M.M.W.C, I.I.I. and Y.S. performed the experiments with assistance from R.M. and G.D.V. Y.P.R. and A.Y.R. provided mice for fate-mapping experiments. L.Z., S.F.Z. and D.R.L. designed the experiments, and L.Z. and D.R.L. wrote the manuscript with input from the co-authors.

Online Methods