Tissue adaptation of regulatory and intraepithelial CD4+ T cells controls gut inflammation

See allHide authors and affiliations

Science  24 Jun 2016:
Vol. 352, Issue 6293, pp. 1581-1586
DOI: 10.1126/science.aaf3892

Location matters for immunosuppression

In the gut, food antigens and resident microbes can trigger unwanted immune responses. Immunosuppressive cell types in the gut, such as regulatory T cells (Tregs) and intraepithelial T lymphocytes (IELs), help to keep these responses at bay. Sujino et al. report that the specific anatomical location within the gut shapes the properties of the suppressive T cell populations that reside there (see the Perspective by Colonna and Cervantes-Barragan). Using mice, they find that Tregs primarily reside in the lamina propria. Tregs migrate to the intestinal epithelium, where they convert to IELs in a process that depends on the microbiota and the loss of a specific transcription factor. Tregs and IELs also play distinct but complementary roles in suppressing intestinal inflammation.

Science, this issue p. 1581; see also p. 1515


Foxp3+ regulatory T cells in peripheral tissues (pTregs) are instrumental in limiting inflammatory responses to nonself antigens. Within the intestine, pTregs are located primarily in the lamina propria, whereas intraepithelial CD4+ T cells (CD4IELs), which also exhibit anti-inflammatory properties and depend on similar environmental cues, reside in the epithelium. Using intravital microscopy, we show distinct cell dynamics of intestinal Tregs and CD4IELs. Upon migration to the epithelium, Tregs lose Foxp3 and convert to CD4IELs in a microbiota-dependent manner, an effect attributed to the loss of the transcription factor ThPOK. Finally, we demonstrate that pTregs and CD4IELs perform complementary roles in the regulation of intestinal inflammation. These results reveal intratissue specialization of anti-inflammatory T cells shaped by discrete niches of the intestine.

The gut mucosa is exposed daily to large amounts of both harmless and potentially pathogenic stimuli; hence, diverse immune regulatory mechanisms must operate to avoid inflammatory diseases (1). Peripheral Foxp3-expressing regulatory T cells (pTregs) mediate suppression of a variety of immune cells and actively prevent inflammatory bowel diseases and food allergies (27). Similar to pTregs, Foxp3CD8αα+CD4+ intraepithelial lymphocytes (CD4IELs) depend on retinoic acid (RA) and transforming growth factor– β (TGF-β) signaling for their development and also have anti-inflammatory properties (4, 813). However, whereas CD4IELs accumulate in the intestinal epithelium, few total Tregs (including pTregs and thymically derived Tregs) can be found at this site (fig. S1, A and B). We asked whether and how the intestinal environment segregates pTregs and CD4IELs, which transcriptional factors are involved in this regulation, and what the implications are for gut inflammation.

We used intravital multiphoton microscopy (IVM) to investigate whether Tregs are actively excluded from the gut epithelium. For tracking in vivo Treg dynamics, we used tamoxifen-inducible Foxp3CreER-eGFP:Rosa26lsl-tdTomato (iFoxp3Tomato) mice (14), which allow Treg fate mapping and the ability to distinguish between cells that currently express and cells that once expressed Foxp3. We compared Treg movement patterns in these mice shortly (24 hours) after tamoxifen administration, which allows the visualization of bona fide Tregs (Tomato+), to thymically derived γδ IEL cell movement patterns (15). We found that more than 80% of TCRγδGFP cells preferentially remained in the epithelium, whereas 68 and 14% of Tomato+ cells were considered lamina propria and IE residents, respectively, throughout the duration of image recordings (Fig. 1, A and B, and movies S1 and 2). Among the remaining (~18%) migrating Tomato+ cells, cell tracking indicated that cells moved into the epithelial layer from the lamina propria more frequently than vice versa (Fig. 1, B and C, and movie S2). Because ex vivo analysis of the intestinal epithelium revealed a low frequency of Foxp3+ Tregs (fig. S1, A and B), these IVM results suggested either preferential cell death of Tregs or Treg conversion into another T cell subtype.

Fig. 1 ThPOK levels correlate with reciprocal Treg and CD4IEL localization and migration dynamics in the intestine.

(A to F) IVM analysis of ileal villi. Mice were injected with Hoechst before imaging to visualize all nuclei (blue). Scale bars, 10 μm. (A) Time-stacked image of TCRγδGFP (left, green channel) mice and iFoxp3Tomato (right, red channel) mice, 24 hours after tamoxifen administration. Images are representative of 20 to 22 villi from at least three independent experiments. (B and C) Frequency of intraepithelial (IE), lamina propria (LP), or migratory TCRγδGFP and iFoxp3Tomato cells. (C) Percentages within migrating cells. (D to F) Sorted naïve CD4+ T cells from OT-II (RFP ThpokGFP) mice were transferred to Rag1–/– mice, and recipient mice were fed an OVA-containing diet for 7 days before IVM analysis. (D) Time-stacked image of GFP+ (left, green channel and blue channel overlay) and GFP+ (yellow) and GFP(red) cells (right, green, red and blue channel overlay). Time-stacked images are representative of at least 50 villi from four independent experiments. (E) Quantification of tracked GFP+ and GFP cell dynamics from four different movies (total six paired villi) in two independent experiments. (F) Percentages within migrating cells. Cells were tracked with Imaris software (Bitplane UK). Data are expressed as mean ± SD from three to six independent movies. ns: not significant, *P < 0.05, **P < 0.01, ***P < 0.001 (Student’s t test).

Whereas all Tregs express the CD4-lineage transcription factor T helper–inducing POZ/Krüppel-like factor (ThPOK), CD8αα+CD4+ and more than 50% of Foxp3CD8α CD4+ cells in the small intestinal epithelium lack ThPOK expression (fig. S1C). We thus asked whether down-modulation of ThPOK influences CD4+ T cell dynamics. We performed IVM using an ovalbumin (OVA)–specific T cell receptor (TCR) transgenic system (OT-II), in which oral OVA exposure leads to incremental ThPOK loss by intestinal CD4+ T cells (10, 16). To visualize T cells upon ThPOK loss, we crossed OT-II (Rag1–/–) with ThpokGFP knockin reporter and ubiquitous monomeric red fluorescent protein 1 (mRFP1) mice. Sorted naïve (RFP+GFP+) CD4+ T cells from OT-II (RFP ThpokGFP) mice were transferred to Rag1–/– mice kept on an OVA-containing diet for 1 week before IVM analysis. Computational tracking revealed that ~60% of the transferred cells that down-regulated ThPOK (RFP+GFP), and only 20% of ThPOKhigh cells (RFP+GFP+), remained in the epithelial layer (Fig. 1, D and E, and movie S3). Migrating ThPOKhigh cells showed movement patterns similar to those of Treg cells, with preferential displacement from the lamina propria into the epithelial layer, suggesting that part of these cells convert into ThPOKlow cells, or die, in that compartment (Fig. 1F). These observations indicate that loss of ThPOK corresponds to an IEL-like behavior in CD4+ T cells. Additionally, the discrepancy between the capacity of Tregs to visit the intestinal epithelium and their low frequency in this layer suggests that this environment may favor Treg plasticity.

To directly examine Treg plasticity in the gut tissue, we performed Treg fate mapping using naïve adult Foxp3Cre-YFP:Rosa26lsl-DsRed (Foxp3DsRed) mice (17). Analysis of peripheral lymphoid tissues isolated from Foxp3DsRed mice revealed an almost complete concurrence between Foxp3 reporter (YFP, yellow fluorescent protein) and DsRed expression, confirming previously described stability of the Treg lineage (14, 18). However, more than 40% of DsRed+ CD4+ T cells in the small intestinal epithelium did not express Foxp3 (YFP), indicating that many Tregs lost Foxp3 expression (fig. S2A). Because previous studies have demonstrated that most “ex-Foxp3” cells in the steady state were derived from uncommitted precursors that transiently up-regulated Foxp3 (18), we also performed fate mapping after pulse labeling iFoxp3Tomato mice with tamoxifen (14), a strategy more likely to target bona fide Tregs (19). Nevertheless, although stable Foxp3 expression was again observed in several peripheral tissues examined, more than 50% of Tomato+ CD4+ T cells that accumulated in the small intestinal epithelium and almost 10% that accumulated in the large intestinal epithelium isolated from iFoxp3Tomato mice no longer expressed Foxp3 5 weeks after tamoxifen administration (fig. S2, B and C). The contribution of Tomato+ cells to the CD8αα+ and CD8αβ+ CD4IEL pools was roughly 10 and 25%, respectively (fig. S2D). Consistent with a ThPOK-dependent process, ex-Tregs that underwent IEL differentiation showed low amounts of ThPOK (fig. S2E). These results indicate that a substantial proportion of intestinal Tregs physiologically convert to CD4IELs.

Commensal bacteria play a major role in the induction of large intestine lamina propria pTregs (3, 57, 20). In contrast, we observed an increased frequency of pTregs (Neuropilin-1 Foxp3+) in the small intestinal epithelium isolated from germ-free (GF) mice when compared to specific-pathogen–free (SPF) controls (Fig. 2A). The total number of Tregs in the epithelial compartment was comparable between GF and SPF mice, even though GF mice showed an almost 10-fold reduction in the number of intraepithelial CD4+ T cells (Fig. 2A). Consistent with a reciprocal ThPOK or Foxp3 expression and CD4IEL differentiation, we found an increased frequency of ThPOKhigh CD4+ T cells and significantly reduced numbers of CD4IELs in GF mice (Fig. 2B). We therefore reasoned that the instability of Tregs in the gut epithelium was influenced by the microbiota. To address this possibility, we treated iFoxp3Tomato mice with broad-spectrum antibiotics for 5 weeks, immediately after tamoxifen exposure. We observed that microbiota depletion prevented Foxp3 loss within the Tomato+ CD4+ T cell population, resulting in an accumulation of Tregs in the epithelial compartment (Fig. 2, C and D). The direct contribution of microbial metabolites versus microbial or dietary antigens to the differentiation of CD4IELs or pTregs occupying the small intestinal epithelium (21) remains to be fully determined. Nevertheless, provision of a TCR ligand can overcome the strict microbiota requirement for CD4IEL differentiation, as demonstrated by the relatively normal CD4IEL differentiation in antibiotic-treated OT-II ThpokGFP Rag1–/– mice exposed to oral OVA (fig. S2F). Along with the above data, these observations substantiate a microbial-induced plasticity of Treg cells in the epithelium, corroborating intratissue specialization and conversion of gut CD4+ T cells.

Fig. 2 Microbiota-dependent plasticity of Tregs in the intestinal epithelium.

(A and B) Flow cytometry analysis of lymphocytes from spleen (spl), mesenteric lymph nodes (mLN), and small intestine epithelium (IEL) of wild-type C57BL/6 mice maintained under specific pathogen-free (SPF) or germ-free (GF) conditions. (A) Surface neuropilin-1 (Nrp-1) and intracellular Foxp3 expression by TCRβ+CD4+CD8β cells. Bar graphs represent frequency and total number of Foxp3+ (left) or Foxp3+Nrp-1 (pTregs) (right) among TCRβ+CD4+CD8β cells. Total cell number for T cell populations isolated from the sIELs is also shown. (B) Surface CD8α and intracellular ThPOK expression by TCRβ+CD4+CD8β cells. Bar graphs represent frequency and total number of CD8α+ (left) or ThPOKhigh (right) among TCRβ+CD4+CD8β cells. Data are expressed as mean ± SD of individual mice (n = 6), representative of three independent experiments. (C and D) Flow cytometry analysis of lymphocytes from spl, mLN, and small intestine IEL of Foxp3CreER-eGFP:Rosa26lsl-tdTomato (iFoxp3Tomato) mice treated with tamoxifen and maintained with broad-spectrum antibiotics for 5 weeks (ABX) or sucralose (Control). (C) Surface CD8α and Tomato expression or intracellular Foxp3 among TCRβ+CD4+ cells. (D) Frequency of Foxp3+ (black) and Tomato+ (white) among TCRβ+CD4+ cells in each tissue. Data are expressed as mean ± SD of individual mice (n = 3), representative of three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001 (Student’s t test).

Next, we analyzed whether the reciprocal properties of pTregs and CD4IELs were influenced by lineage-defining transcription factors. We first determined the role of ThPOK by crossing Thpokfl/fl (22) with Cd4CreER mice (23) for Thpok depletion upon tamoxifen treatment (16) (fig. S3A). In vivo administration of tamoxifen to iCd4Thpok) mice (16) 1 week before analysis resulted in a 32 to 48% reduction in the frequency of Foxp3+ Tregs in lymphoid and intestinal tissues, respectively, whereas an increase in CD4IELs was observed only in intestinal tissues (Fig. 3, A and B). To examine the Treg cell–intrinsic nature of these findings, we crossed Thpokfl/fl with Foxp3CreER-eGFP (fig. S3B) (16). Ex vivo analysis of T cells from iFoxp3Thpok) mice 1 week after tamoxifen treatment did not show a significant decrease in amounts of Foxp3 in the tissues examined, suggesting that ThPOK was not required for short-term stability of the lineage. However, analysis of T cells from iFoxp3Thpok) mice 5 weeks after tamoxifen pulse labeling revealed a significant reduction in the frequency and amounts of Foxp3 in CD4+ T cells located in most compartments analyzed, with the exception of the large intestine (Fig. 3, C and D). Reciprocally, we found an accumulation of CD4IELs in the small intestine (Fig. 3, C and D). These data further support a ThPOK role in the regulation of the Treg de novo conversion and stability.

Fig. 3 ThPOK expression by intestinal epithelial CD4+ T cells plays a key role in the reciprocal development of Tregs and CD4IELs.

(A to D) Flow cytometry analysis of lymphocytes from spleen (spl), mesenteric lymph nodes (mLN), small and large intestine epithelium (IEL), and lamina propria (LPL) of inducible conditional Thpok-deficient mice. (A and B) Cd4CreER:Thpokfl/fl [iCd4Thpok)] and (Cre) Thpokf/f littermate control mice 7 days after tamoxifen administration. (A) Representative contour plot for surface CD8α and intracellular Foxp3 among TCRβ+CD4+CD8β cells. (B) Frequency of CD8α+ (upper) and Foxp3+ (lower) among TCRβ+CD4+CD8β or among total CD45+ cells in the indicated tissues. Data are expressed as mean ± SD of individual mice (n = 3 to 6), representative of six independent experiments. (C and D) Foxp3CreER:Thpokfl/fl [iFoxp3Thpok)] and (Cre+)Thpok+/+ littermate control mice 7 days [shown in (D)] or 35 days after tamoxifen administration. (C) Representative contour plot for surface CD8α and intracellular Foxp3 among TCRβ+CD4+CD8β cells. (D) Frequency of CD8α+ (upper) and Foxp3+ (lower) among TCRβ+CD4+CD8β or among total CD45+ cells in the indicated tissues. Data are expressed as mean ± SD of individual mice (n = 3 to 8), representative of three independent experiments. (E) Frequency and total number of CD8α+ and Foxp3+ among TCRβ+CD4+CD8β cells from sIEL of Ox40Tbx21), Cd4Runx3), E8ITbx21), and wild-type (WT) mice. Data are expressed as mean ± SD of individual mice (n = 3 to 9), representative of three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001 [Student’s t test (B), one-way analysis of variance (ANOVA) with Tukey post-test (D, E)].

To target the CD4IEL lineage defining transcription factors, we generated mice with conditional deletion of Runx3 or Tbx21 (encoding T-bet), which mediate down-regulation of ThPOK in developing CD4IELs (9, 10). We analyzed the epithelial compartment of Cd4Runx3) mice and found a reduction in the frequency and number of CD4IELs and, conversely, enrichment in Tregs (Fig. 3E). Next, we analyzed mice in which Tbx21 was excised early [driven by Ox40Cre (24)] or late [driven by E8ICre (25)] during the CD4IEL differentiation (16). Ox40Tbx21) mice also showed reduced numbers of CD4IELs and roughly a twofold increase in Tregs in the epithelium, whereas E8ITbx21) mice showed reduced numbers of CD4IELs but normal numbers of Tregs in the epithelium when compared to Cre control mice (Fig. 3E). Collectively, the above data provide a possible mechanism for the reduced number of Tregs in the gut epithelium, where collaboration between Runx and T-bet results in down-modulation of ThPOK, and Foxp3, in CD4+ T cells (9, 10).

Oral exposure to TCR ligands results in both pTreg and CD4IEL differentiation in a TGF-β–dependent manner (4, 8, 10, 11, 13, 26). We asked whether the intratissue adaptation of pTregs and CD4IELs influences the outcome of T cell responses to dietary antigens by using a transcription factor–based targeting of these lineages in OVA-specific TCR transgenic mice on a Rag1–/– background (16). Conditionally targeting Runx3 in the OT-II model [OT-II(ΔRunx3)] prevented ThPOK loss and CD4IEL differentiation and also affected pTreg differentiation in the large intestine, although no differences in cytokine production were found when compared to control OT-II mice (fig. S4, A to D). Whereas OVA-challenged control OT-II mice showed few or no signs of intestinal inflammation, OT-II(ΔRunx3) mice readily developed diarrhea and severe pathology, as confirmed by fecal lipocalin-2 concentrations (Fig. 4, A to C). We concluded that prevention of ThPOK loss and CD4IEL differentiation resulted in a local inflammatory response toward dietary antigens, although the reduction in pTreg numbers could also contribute to the exaggerated inflammatory phenotype observed in the large intestine of OT-II(ΔRunx3) mice. In contrast, conditionally targeting Thpok by administering tamoxifen to OVA-fed iOT-II(ΔThpok) mice (16) severely impaired pTreg development in all tissues examined with a concomitant increase in CD4IELs in the intestine, although no inflammatory phenotype was observed (fig. S4, E to G). We then tested whether the OT-II(ΔRunx3) disease phenotype could be rescued, or prevented, by wild-type or Thpok-deficient OT-II cells (16). We found that transferred wild-type CD45.1 OT-II cells, which could differentiate to both pTreg and CD4IELs, as well as CD4+ T cells from tamoxifen-treated iOT-II(ΔThpok) mice, which could only differentiate to CD4IELs, rescued the diarrhea and inflammatory phenotype observed in OT-II(ΔRunx3) mice (Fig. 4, B and C, and fig. S4, H and I). These data indicate that a balance in ThPOK levels regulates inflammatory, CD4+ T cell–mediated responses to dietary antigens. Additionally, these results suggest that CD4IELs exert a cell-extrinsic control of local intestinal inflammation.

Fig. 4 Complementary roles for pTregs and CD4IELs in regulating local inflammatory response toward dietary antigens.

OVA-specific TCR transgenic mice (Rag1–/– background) were fed an OVA-containing diet for 7 days. (A to C) ThpokGFP OT-II(ΔRunx3) and control ThpokGFP OT-II mice were analyzed. (A) Hematoxylin and eosin staining of the small intestine jejunum (SI) (upper panels) and the large intestine colon (LI) (lower panels). Original magnification, 40×. Graphs represent histological scores of the SI (upper) and the LI (lower) (each symbol represents one mouse). (B and C) Sorted naïve OVA-specific TCR transgenic cells (TCRVα2+CD4+CD62L+CD44low) from wild-type CD45.1 OT-II or tamoxifen-treated iOT-II(ΔThpok) were transferred to host OT-II(ΔRunx3) before treatment, as in (A). (B) Frequency of diarrhea-free mice after oral OVA challenge. (C) Quantification of fecal Lipocalin-2. (D to F) TBmc Foxp3sf (scurfy) and TBmc control were treated as in (A) and injected with isotype control or anti-CD8α depleting antibody. (D) Hematoxylin and eosin staining of the SI (upper) and the LI (lower). Original magnification, 40×. Graphs represent histological scores of the SI (upper) and the LI (lower) (each symbol represents one mouse). (E) Frequency of diarrhea-free mice after oral OVA challenge. (F) Quantification of fecal Lipocalin-2. Data are expressed as mean +SD or median ± interquartile range (A and D), representative of at least two independent experiments (n = 3 to 8 per group). Scale bar, 200 μm. *P < 0.05, **P < 0.01, ***P < 0.001 [Student’s t test or Mann-Whitney test (A), one-way ANOVA with Tukey post-test (C and F), log-rank test (B and E), and Kruskal-Wallis with Dunns post-test (D)].

To investigate whether pTregs and CD4IELs play complementary anti-inflammatory roles in the intestine, we compared T cell responses to dietary OVA using BALB/c background monoclonal OVA-specific TCR strains, carrying either wild-type Foxp3 or a scurfy mutation (Foxp3sf) (16), which results in a Foxp3 loss of function (27). In contrast to OT-II mice (C57BL/6 background), TBmc Foxp3wt mice fed an OVA diet showed a high rate of pTreg induction in all tissues examined, but less ThPOK loss and fewer CD4IELs in the epithelium (fig. S4, A to C and J to N). Conversely, TBmc Foxp3sf showed a high degree of ThPOK loss and increased CD4IEL development in the small intestine, with a frequency that mirrored the relative amounts of pTregs in TBmc Foxp3wt mice (fig. S4, J to N). However, no inflammatory phenotype or diarrhea was observed, even in the absence of functional Foxp3 in this monoclonal model (Fig. 4, D to F). To examine whether exaggerated CD4IEL differentiation could compensate for the absence of pTregs in TBmc Foxp3sf mice, we depleted CD4IELs using antibodies against CD8α (anti-CD8α) during OVA feeding (fig. S4, M and N). We found that TBmc Foxp3sf mice treated with anti-CD8α, but not TBmc Foxp3wt treated with anti-CD8α or TBmc Foxp3sf treated with control antibodies, showed severe intestinal inflammation and diarrhea (Fig. 4, D to F, and fig. S4O). These results support a model in which CD4IELs and pTregs cooperate in the regulation of local intestinal inflammation.

The single-layered intestinal epithelium constitutes a uniquely challenging location for immune regulatory processes, given its proximity to highly stimulatory luminal contents and limited spatial organization. It is currently thought that Tregs use several redundant and complementary mechanisms to suppress inflammatory responses, and their capacity to sense specific environmental cues plays a major role in their function (2830). The physiological instability that we observe in the Treg lineage within the intestinal epithelium may represent an important modulation of regulatory activity that is coordinated by this particular environment (3, 20, 28, 31, 32). Although our targeting strategies do not discriminate between the function of “ex-Tregs” and “directly converted” CD4IELs, natural or forced ThPOK down-modulation was previously associated with an impaired helper function in CD4+ T cells, including reduced production of proinflammatory cytokines and reduced expression of costimulatory molecules (8, 10, 33). The data presented here support a cell-extrinsic suppressive role for CD4IELs, although the likely epithelium-specific mechanism used by these cells to actively regulate or prevent tissue inflammation remains unclear. However, a role for CD4IELs in triggering inflammation via their cytotoxic activity, in specific contexts, is conceivable (8, 34). Nevertheless, the observation that particular environmental cues, such as the microbiota, induce plasticity of seemingly stable lymphocyte lineages may contribute to the understanding of how specialized tissue-adaptation pathways function to balance efficient immune protective responses with tissue tolerance.


Materials and Methods

Figs. S1 to S4

Movies S1 to S3

Reference (35)


  1. Materials and methods are available as supplementary materials on Science Online.
Acknowledgments: We are indebted to K. Velinzon and N. Thomas for sorting cells and to members of the Nussenzweig lab and The Rockefeller University employees for continuous assistance. We thank S. Hemmers (Memorial Sloan Kettering Cancer Center) for generating Cd4CreER mice. We especially thank A. Rogoz for outstanding technical support. We thank members of our laboratory, particularly V. Pedicord, and D. Esterhazy, for discussions and critical reading and editing of the manuscript. The data reported in this manuscript are tabulated in the main paper and in the supplementary materials. This work was supported by Leona M. and Harry B. Helmsley Charitable Trust (T.S., D.P.H.v.K., B.S.R., D.M.), Japan Foundation for Applied Enzymology and Uehara Memorial Foundation (T.S.), Alexandre Suerman Stipend, Royal Netherlands Academy of Sciences, and the Prince Bernhard Cultural Foundation (D.P.H.v.K.), Deutsche Forschungsgemeinschaft 1410/1 grant and Swiss National Science Foundation 0310030-11620 grant (T.B.), National Multiple Sclerosis Society (J.J.L.), the Crohn’s & Colitis Foundation of America (B.S.R., D.M.), the Irma T. Hirschl Award (D.M.), and National Institutes of Health grant NIH R01 DK093674 (D.M.). The Rockefeller University Bio-Imaging Resource Center is supported by the Empire State Stem Cell Fund through New York State Department of Health C023046. D.M. conceived of and D.M and B.S.R. supervised this study; T.S., J.J.L., B.S.R., and D.M. designed experiments; T.S., M.L., D.P.H.v.K., T.R., H.M.S., and B.S.R. performed and analyzed experiments; T.B. provided the Cd4CreER strain (23); T.S., M.L., D.P.H.v.K. and B.S.R. prepared figures and helped with manuscript preparation; and D.M. wrote the paper.

Stay Connected to Science

Navigate This Article