Research ArticleIMMUNOLOGY

Bcl11b prevents fatal autoimmunity by promoting Treg cell program and constraining innate lineages in Treg cells

See allHide authors and affiliations

Science Advances  07 Aug 2019:
Vol. 5, no. 8, eaaw0480
DOI: 10.1126/sciadv.aaw0480

Abstract

Regulatory T (Treg) cells are essential for peripheral tolerance and rely on the transcription factor (TF) Foxp3 for their generation and function. Several other TFs are critical for the Treg cell program. We found that mice deficient in Bcl11b TF solely in Treg cells developed fatal autoimmunity, and Bcl11b-deficient Treg cells had severely altered function. Bcl11b KO Treg cells showed decreased functional marker levels in homeostatic conditions, inflammation, and tumors. Bcl11b controlled expression of essential Treg program genes at steady state and in inflammation. Bcl11b bound to genomic regulatory regions of Treg program genes in both human and mouse Treg cells, overlapping with Foxp3 binding; these genes showed altered chromatin accessibility in the absence of Bcl11b. Additionally, Bcl11b restrained myeloid and NK cell programs in Treg cells. Our study provides new mechanistic insights on the Treg cell program and identity control, with major implications for therapies in autoimmunity and cancer.

INTRODUCTION

Regulatory T (Treg) cells have potent immunosuppressive ability and are vital for maintaining immune homeostasis and peripheral tolerance. Treg cell transcriptional program is dependent on the transcription factor (TF) Foxp3, which is also essential for their development and suppression function (13). Alteration in the human TF FOXP3 leads to fatal autoimmune disease, causing immune dysregulation, polyendocrinopathy, enteropathy, X-linked (IPEX) syndrome in humans and mice (Scurfy) (47). Despite the indispensable role of Foxp3, Foxp3 alone is not sufficient to induce the Treg cell program (8), and other TFs are important for Treg program and function, including T-bet, GATA3, Helios, and FoxA1 (911).

The Foxp3 locus has several conserved noncoding sequences (CNS-0-3), each with specific roles in Foxp3 expression (12). CNS-0, located at a Treg super-enhancer (SE), is needed to induce Foxp3 expression in Treg cell precursors in a manner dependent on the genome organizer Satb1 (13). CNS-1 and CNS-2 are located in the intron following the untranslated exon 1 of the Foxp3 gene, and CNS-3 is located in the intron after the first coding exon. CNS-1 is essential for induction of Foxp3 expression and generation of peripherally induced Treg (pTreg) cells, especially in the intestine and lung (12). CNS-2 is bound by the RUNX1-Cbfβ complex, and its hypomethylation is needed for prolonged and stable expression of Foxp3 and Treg stability (14). Its deletion results in loss of Foxp3 expression and unstable Treg cells, which start secreting proinflammatory cytokines (12).

Treg cells use several mechanisms to suppress aberrant immune responses. They constitutively express cytotoxic T-lymphocyte associated protein 4 (CTLA4), which provides an off signal and competes with CD28 on effector T (Teff) cells for interaction with B7 on dendritic cells and other antigen-presenting cells (15). CD25 is the high-affinity interleukin-2 (IL-2) receptor, which can leach IL-2 from the environment, depleting Teff cells of the necessary IL-2 to survive. The triphosphate diphosphohydrolase 1 (Entpd1/CD39) is up-regulated on Treg cells in inflammatory conditions and converts extracellular ATP (adenosine triphosphate) to ADP (adenosine diphosphate) and AMP (adenosine monophosphate), which are further metabolized into suppressive pericellular adenosine (pAde) by another ectonucleotidase, 5-nucleotidase-ecto (Nt5e/CD73). pAde binds to the A2A receptor on Teff cells, and pAde/A2AAR signaling causes accumulation of cAMP (3′,5′-cyclical adenosine monophosphate) that blocks Teff cell proliferation and production of proinflammatory cytokines. Glucocorticoid-induced tumor necrosis factor–related receptor (Tnfrsf18/GITR) is constitutively expressed on Treg cells and has immunosuppressive effects [reviewed in (4)].

Bcl11b is a C2H2 zinc finger DNA binding protein (16, 17), which functions as a transcriptional repressor when associated with the Nucleosome Remodeling and Deacetylase complex (Mi-2/NuRD) (18) and as a transcriptional activator when associated with histone acetyl transferases (19). Bcl11b is first expressed at the DN2 stage of thymocyte development and plays essential roles in thymocyte development, commitment, selection, and survival (2024). In addition, Bcl11b is required for invariant natural killer T (iNKT) cell development by controlling the glycolipid processing and presentation by double-positive (DP) thymocytes and the functional program of iNKT cells (25, 26). Removal of Bcl11b from mature T cells leads to multiple defects including the derepression of the T helper 2 (TH2) program in TH17 cells (27) and failure of TH2 cells to differentiate during asthma and antihelminth responses (28). Absence of Bcl11b in innate lymphoid cells type 2 (ILC2) results in loss of their identity and derepression of the ILC3 program (29). CD8+ T cells fail to expand in an antigen-specific manner and have reduced effector function (30, 31). We previously reported that removal of Bcl11b with the CD4-Cre deleter or with the Foxp3–Cre–GFP (green fluorescent protein) BAC (bacterial artificial chromosome), which deletes slowly because of its reduced Cre levels, caused wasting disease associated with colitis and reduced formation of inducible Treg cells and altered expression of Foxp3 and IL-10, as well as acquisition of proinflammatory genes (32). To investigate the molecular mechanisms by which Bcl11b controls Treg cell programs, we used the Foxp3YFP-Cre strain with high expression of Cre and efficient removal of floxed alleles in Treg cells (33). Bcl11bF/F Foxp3YFP-Cre+ male mice and Bcl11bF/F Foxp3YFP-Cre+/+ female mice [Bcl11b/Treg knockout (KO) mice] died at an early age, presenting fatal systemic autoimmunity. Bcl11b-deficient Treg cells showed altered suppression function even when isolated from mice with no inflammation due to the copresence of wild-type (WT) Treg cells. In addition, Bcl11b-deficient Treg cells showed reduced functional suppression molecules at steady state in inflammatory conditions and in tumors. Bcl11b bound at the upstream and intronic regions of several essential Treg genes both in mice and humans, including at the Foxp3 locus and at other Treg genes, namely, Ikzf2, Ctla4, Il2ra, and Tnfrsf18. Binding of Bcl11b at many of these sites overlapped with Foxp3 binding in mouse Treg cells, suggesting a synergistic role in the control of these genes. Absence of Bcl11b altered chromatin accessibility at these genes and their expression. Bcl11b also bound at the NK cell TF Id2 locus, at several NK receptor loci, at Spi1, and other myeloid genes both in mouse and human Treg cells, suggesting a direct and conserved control of their expression in Treg cells. Thus, we demonstrate that Bcl11b is an essential TF for supporting Treg cell program and restraining innate lineage programs in Treg cells.

RESULTS

Removal of Bcl11b in Treg cells causes fatal systemic autoimmunity

In the current study, we used the Foxp3YFP-Cre mice, which express yellow fluorescent protein (YFP)–Cre from the Foxp3 locus and have very efficient removal of Plox alleles only in Treg cells (33). Bcl11bF/F Foxp3YFP-Cre+ male mice and Bcl11bF/F Foxp3YFP-Cre+/+ female mice (Bcl11b/Treg KO mice) died between 20 and 40 days of age, with few exceptions living 60 days (Fig. 1A), and showed reduced weight compared with sex-matched heterozygous (HT) littermates (Fig. 1B). Bcl11b/Treg KO mice showed enlarged peripheral lymphoid organs with disrupted architecture (Fig. 1C). In addition, Bcl11b/Treg KO mice developed eczematous dermatitis and blepharitis (Fig. 1D and data not shown), similar to Scurfy mice, which are known to have a loss-of-function mutation in the Foxp3 gene (5). Skin, lung, and other internal organs had massive infiltrations of leukocytes (Fig. 1D and data not shown). Frequencies of CD8+ T cells in peripheral lymphoid organs of Bcl11b/Treg KO mice were increased, while CD4+ T cells were variable (Fig. 1E and data not shown). However, frequencies and absolute numbers of both CD44+CD4+ T and CD44+CD8+ T cells were significantly increased in Bcl11b/Treg KO mice (Fig. 1, F and G, and data not shown), which is indicative of an activated phenotype. The percentages and absolute numbers of Treg cells in the peripheral lymph nodes (peLNs) and spleens of Bcl11b/Treg KO mice showed a significant decrease compared with HT mice (Fig. 1, H and I). Thus, Bcl11b/Treg KO mice develop multiorgan fatal autoimmunity and lymphadenopathy with uncontrolled activation of T cells and a reduction in Treg cell frequency.

Fig. 1 Bcl11b removal in Treg cells results in early death of mice and increased CD4+ and CD8+ T cell activation.

(A) Kaplan-Meier survival curves of mice with the following genotypes: Bcl11bF/F Foxp3YFP-Cre+ (Bcl11b/Treg KO) and Bcl11bF/WT Foxp3YFP-Cre+ (HT) male and Bcl11bF/F Foxp3YFP-Cre+/+ (Bcl11b/Treg KO) and Bcl11bF/WT Foxp3YFP-Cre+/+ (HT) female. (B) Matched weights of Bcl11b/Treg KO and HT male littermates and female littermates at 3 to 4 weeks of age. (C) Peripheral lymph nodes (peLNs) and spleens of 3- to 4-week-old Bcl11b/Treg KO and HT mice. Hematoxylin and eosin (H&E) staining showing the architecture of peLN and spleen. (D) H&E staining of skin and lung samples of 3- to 4-week-old Bcl11b/Treg KO and HT mice. (E) Average frequencies of CD8+ T cells in peLNs and spleens of Bcl11b/Treg KO and HT mice. (F) Representative CD44 histograms in CD4+ and CD8+ T cells from peLNs and spleens of Bcl11b/Treg KO and HT mice. (G) Average frequencies of CD44+ CD4+ and CD44+ CD8+ T cells in peLNs and spleens of Bcl11b/Treg KO and HT mice. (H) Representative flow plots of Treg cells in Bcl11b/Treg KO and HT mice isolated from peLNs and spleens. (I) Average frequencies of Treg cells in Bcl11b/Treg KO and HT mice (n = 7). Analysis was conducted on Bcl11b/Treg KO and HT control mice. P values determined by Student’s t test. **P < 0.01, ***P < 0.001, ****P < 0.0001. Photo credit: Mohammad Uddin, University of Florida.

Bcl11b deletion causes a decrease in key Treg markers

Given that Bcl11b/Treg KO mice showed multiorgan inflammation, we made use of Bcl11bF/F Foxp3YFP-Cre+/− mosaic female mice (Bcl11b/Treg KO mosaic female mice), which, in addition to YFP+Bcl11b KO Treg cells, have YFP WT Treg cells as a result of random inactivation of X chromosome–linked genes. We found that Bcl11b/Treg KO mosaic female mice did not develop overt signs of autoimmunity and weight loss. However, YFP+ Treg cells of Bcl11b/Treg KO mosaic female mice showed decreased frequencies and mean fluorescence intensities (MFIs) of CD25, CTLA4, and GITR when compared with YFP Treg cells from the same mouse (Fig. 2, A to F). In addition, frequency of Helios+ Treg cells and MFI were both significantly decreased in Bcl11b KO Treg cells of Bcl11b/Treg KO mosaic female mice compared with YFP WT Treg cells (Fig. 2, G and H). Furthermore, frequency and absolute numbers of YFP+ Treg cells in the Bcl11b/Treg KO mosaic female mice were decreased compared with corresponding YFP+ Treg cells in the Bcl11bF/WT Foxp3YFP-Cre+/− HT mosaic female mice (Fig. 2, I and J), and the MFIs for YFP and Foxp3 were also diminished in the absence of Bcl11b (Fig. 2K). Thus, while the YFP WT Treg cells can prevent autoimmunity in the Bcl11b/Treg KO mosaic female mice, many key Treg cell markers are dysregulated in the absence of Bcl11b at steady state in YFP+ Bcl11b KO Treg cells.

Fig. 2 Bcl11b-deficient Treg cells of Bcl11b/Treg KO mosaic female mice have decreased suppression markers at steady state.

(A, C, and E) Representative histograms and average frequencies of CD25+ (A), CTLA4+ (C), and GITR+ (E) YFP+Foxp3+ (KO) and YFPFoxp3+ (WT) Treg cells within the same Bcl11bF/F Foxp3YFP-Cre+/− (Bcl11b/Treg KO mosaic) female mice. (B, D, and F) Mean fluorescence intensities (MFIs) of surface CD25 (B), CTLA4 (D), and GITR (F) on YFP+Foxp3+ (KO) and YFPFoxp3+ (WT) Treg cells of Bcl11b/Treg KO mosaic female mice at steady state (n = 6). (G) Representative histogram and average frequencies of Helios in YFP+Foxp3+ (KO) and YFPFoxp3+ (WT) Treg cells from peLNs of Bcl11b/Treg KO mosaic female mice at steady state. (H) MFI of Helios in YFP+Foxp3+ (KO) and YFPFoxp3+ (WT) Treg cells from peLNs of Bcl11b/Treg KO mosaic female mice (n = 4). Analysis was conducted on YFP+Foxp3+ (KO) and YFPFoxp3+ (WT) Treg cells of Bcl11b/Treg KO mosaic female mice (A to H). (I) Representative contour plots depicting the YFP+Foxp3+ (KO) and YFPFoxp3+ (WT) Treg cells in Bcl11bF/FFoxp3YFP-Cre+/− (Bcl11b/Treg KO mosaic) and Bcl11bF/WT Foxp3YFP-Cre+/− (HT mosaic) female mice. (J) Frequencies and absolute numbers of YFP+ Treg cells in Bcl11b/Treg KO mosaic and HT mosaic female mice. (K) Foxp3 and YFP MFIs in YFP+Foxp3+ Treg cells from Bcl11b/Treg KO mosaic and HT mosaic female mice (n = 7). Analysis was conducted on YFP+ Treg cells from Bcl11b/Treg KO mosaic and HT mosaic female mice (J and K). P values determined by Student’s t test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Functional markers remain decreased in Bcl11b-deficient Treg cells in inflammation and tumors

We further evaluated functional markers in Bcl11b KO YFP+ and WT YFP Treg cells of Bcl11b/Treg KO mosaic female mice induced with experimental autoimmune encephalomyelitis (EAE) to determine their status in inflammatory conditions. Frequencies of CD25+, CTLA4+, and GITR+ Treg cells, as well as their MFIs, remained reduced during EAE in Bcl11b KO YFP+ Treg cells compared with WT Treg cells (Fig. 3, A to F, and table S1). In addition, there was a major reduction in the CD39+CD73+ Treg cell population in the absence of Bcl11b in EAE conditions (Fig. 3G and table S1). Of note, CD39+CD73+ Treg cell frequency and their MFIs are low at steady state even in WT Treg cells.

Fig. 3 Treg cell markers and suppression function are altered in the absence of Bcl11b during inflammation.

(A, C, and E) Representative histograms and average frequencies of CD25 (A), CTLA4 (C), and GITR (E) in YFP+Foxp3+ (KO) and YFPFoxp3+ (WT) Treg cells of Bcl11bF/F Foxp3YFP-Cre+/− (Bcl11b/Treg KO mosaic) female mice induced with EAE. dLN, draining lymph node. (B, D, and F) MFI of CD25 (B), CTLA4 (D), and GITR (F) in YFP+Foxp3+ (KO) and YFPFoxp3+ (WT) Treg cells of Bcl11b/Treg KO mosaic female mice induced with EAE. (G) Representative contour plots of CD39, CD73, and average frequencies of CD39+CD73+ YFP+Foxp3+ (KO) and YFPFoxp3+ (WT) Treg cells of Bcl11b/Treg KO mosaic female mice induced with EAE (n = 16). Analysis was conducted on YFP+Foxp3+ (KO) and YFPFoxp3+ (WT) Treg cells of Bcl11b/Treg KO mosaic female (A to G). (H and I) Rag1−/− mice transferred with 4 × 105 CD45RBhi CD4+ T cells from CD45.1 WT mice to induce colitis and either 1 × 105 YFP+ Treg cells from Bcl11b/Treg KO mosaic female mice (CD45.2) (red), or from HT mosaic female mice (black) (CD45.2), or no Treg cells (blue). (H) Average weight loss/gain of Rag1−/− mice transferred as indicated above (n = 10). (I) Histopathological scores of Rag1−/− mice transferred as indicated above. Analysis was conducted on Rag1−/− mice transferred as indicated above. P values determined by Student’s t test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. n.s., not significant.

We further induced flank tumors in Bcl11b/Treg KO mosaic and HT mosaic control female mice with B16-Ova melanoma, and as expected, the tumor sizes were not different between the two groups, likely due to the presence of WT YFP cells even in Bcl11b/Treg KO mosaic female mice. We then evaluated the phenotype of Bcl11b KO YFP+ and WT YFP Treg cells in Bcl11b/Treg KO mosaic female mice transplanted with B16-Ova flank tumors and found that similar to steady state and EAE conditions, Bcl11b KO YFP+ Treg cells had reduced suppressive markers compared with the WT YFP Treg cells in the same mouse. This included decreased levels and frequencies of CD25, CTLA4, GITR, CD39, and CD73 (fig. S1, A to H, and table S1). A comparison of the suppression markers in Treg cells in inflammatory conditions (EAE), tumors, and steady state shows an overall decrease of these markers in the absence of Bcl11b in all three conditions. Among these markers, GITR, CD39, and CD73 showed even more decline in inflammatory conditions [Fig. 3 (A to G), fig. S1 (A to H), and table S1].

These results, thus, demonstrate that Bcl11b KO Treg cells have alterations in essential Treg functional markers at steady state, in inflammatory conditions, and in tumors, and some are further exacerbated in inflammatory conditions.

Bcl11b KO Treg cells failed to control CD45Rbhi CD4+ T cell–induced colitis

Given the alterations in key Treg markers in the absence of Bcl11b, we next evaluated the in vivo suppression ability of YFP+ Treg cells derived from Bcl11b/Treg KO mosaic female mice and control mosaic female mice in the colitis model of CD45Rbhi CD4+ T cell transfer in immunodeficient mice (34). Naïve CD45Rbhi CD4+ T cells were transferred into Rag1−/− mice, provided also with YFP+ Treg cells from Bcl11b/Treg KO mosaic female mice or control HT mosaic female mice or with no other cells. While HT Treg cells suppressed the weight loss of Rag1−/− mice transferred with naïve CD4+ T cells, Bcl11b KO Treg cells failed (Fig. 3H). Histopathological analysis showed that mice transferred with naïve CD4+ T cells plus HT Treg cells had significantly lower scores compared with those transferred with naïve CD4+ T cells only or with naïve CD4+ T cells plus Bcl11b KO Treg cells (Fig. 3I). Transferred Bcl11b KO Treg cells were detectable at frequencies similar to that of HT Treg cells at the completion of the study, with a slight, but not significant, decrease in the Foxp3+ cell frequency (fig. S2, A to C). These results demonstrate that deletion of Bcl11b in Treg cells hampers their ability to suppress CD4+ T cell–mediated inflammation.

Bcl11b controls expression of essential Treg program genes

To establish the mechanistic basis for the functional failure of Bcl11b KO Treg cells, we conducted RNA sequencing (RNA-seq) analysis of Bcl11b KO and WT YFP+ Treg cells from Bcl11b/Treg KO and WT mosaic female mice. Gene Set Enrichment Analysis (GSEA) showed that Treg signature genes (3) were enriched in WT compared with Bcl11b KO Treg cells in the absence of inflammation (Fig. 4A). Within this signature, the mRNA for Foxp3 was significantly reduced (Fig. 4B), as was for Helios (Ikzf2) (Fig. 4B), encoding a TF controlling stability of Treg cells and characteristic of thymically derived Treg cells (11). In addition, Treg genes belonging to the functional program, including Ctla4, CD25 (Il2ra), Gitr, and Nt5e (CD73), were down-regulated (Fig. 4B). Of note, protein levels of these markers were also decreased (Fig. 2).

Fig. 4 Bcl11b regulates expression of key Treg cell genes while suppressing inflammatory and innate genes.

(A) GSEA showing enrichment of Treg signature genes in WT Treg cells (n = 2) compared with Bcl11b KO Treg cells (n = 2). RNA-seq was conducted on Treg cells sorted from Bcl11b/Treg KO mosaic and WT female mice. NES, normalized enrichment score. (B) Graphs showing normalized read counts of key Treg genes from RNA-seq of Bcl11b KO and WT Treg cells (n = 2). (C) GSEA plots showing enrichment of inflammatory and allograft rejection pathways in Bcl11b KO Treg cells versus WT (n = 2). (D) Heatmap of normalized read counts of myeloid and NK cell program genes from RNA-seq of Bcl11b KO and WT Treg cells (n = 2). (E) Graphs showing normalized read counts of selected myeloid and NK cell genes (n = 2). (F) Distribution of fold changes of differentially expressed (DE) genes between Bcl11b KO and WT Treg cells at steady state and in EAE. FC, fold change. (G) Venn diagram of the proportion of unique and shared differentially expressed genes between Bcl11b KO and WT Treg cells at steady state and in EAE. (H) Scatter plot of RNA-seq log2 (fold change) of the Treg signature genes at steady state and in EAE. False discovery rate (FDR)–adjusted P values were calculated using the Benjamini-Hochberg method, with a significance cutoff of adjusted P < 0.05.

Bcl11b represses a proinflammatory gene signature in Treg cells

In addition, Bcl11b KO Treg cells displayed enhanced inflammatory and allograft rejection gene signatures (Fig. 4C), which included up-regulation of genes such as Ifng, Il15, Il18, Il2, and Il1b (fig. S3, A and B). Thus, Bcl11b represses proinflammatory genes in Treg cells.

Bcl11b globally represses expression of myeloid and NK genes in Treg cells

Further analysis of the RNA-seq data shows that the absence of Bcl11b resulted in up-regulation of mRNAs for multiple myeloid and NK genes in Treg cells (Fig. 4, D and E). Specifically, the mRNAs for the critical TFs for NK and myeloid lineages, ID2 and PU.1, respectively, were both up-regulated in the absence of Bcl11b (Fig. 4, D and E). From the NK program, several killer cell lectin-like receptor genes, including Klrb1b, Klrb1c (encoding Nk1.1), Klrk1 (encoding Nkg2d), Klrd1, and Cd160, were all up-regulated (Fig. 4, D and E). From the myeloid program, Fcer1g, Fcgr3, Cd14, Cd74, Itgb3, and Itgad were up-regulated (Fig. 4, D and E). In addition, genes with role in hematopoiesis, such as c-kit and Cd9, were up-regulated (Fig. 4D). These results point to a critical role of Bcl11b in repressing expression of NK and myeloid cell programs in Treg cells.

Bcl11b controls the same genes in inflammation and at steady state

We next conducted RNA-seq in WT and Bcl11b KO Treg cells isolated from EAE mice to identify commonly or uniquely controlled genes by Bcl11b in inflammation versus steady state. More genes were differentially expressed between KO and WT Treg cells during steady state than in EAE (Fig. 4F). However, during both EAE and steady state, more genes were up-regulated than down-regulated, and most up-regulated genes displayed a higher fold change (Fig. 4F). Most genes were commonly differentially expressed at SS and in EAE, suggesting that Bcl11b controls similar programs in the presence or absence of inflammation in Treg cells (Fig. 4G). A significant correlation of differentially expressed genes in the Treg gene signature, allograft rejection, and inflammatory response pathways was observed at steady state and in EAE (Fig. 4H and fig. S3, A and B).

Bcl11b predominantly binds to intergenic and intronic regions in Treg cells

We further performed Bcl11b chromatin immunoprecipitation sequencing (ChIP-seq) analysis on Treg cells to identify genes directly controlled by Bcl11b, which we compared with Bcl11b genomic binding in naïve CD4+ T cells (28). Given that numerous genes regulated by Bcl11b are also known Foxp3 targets, we also compared the genomic binding by Bcl11b with Foxp3 binding in Treg cells (35). Peak annotation showed that Bcl11b predominantly bound to intergenic and intronic regions and less to promoter regions in Treg cells (Fig. 5A), similarly to conventional CD4+ T cells (28). In line with our previous observations in conventional CD4+ T cells (28), motif analysis showed that the top 2 motifs were ETS (E26 transformation-specific) and Runx (Runt-related transcription factor) (fig. S4). Further motif analysis of peaks at up-regulated or down-regulated genes at steady state showed a significant overlap between motifs, as 7 of the top 10 motifs were present in both up-regulated and down-regulated gene sets (table S2). However, when comparing the relative enrichment of the top 10 motifs, the top 4 motifs for the down-regulated gene set were significantly enriched, and all constituted ETS-binding motifs (table S2), suggesting that Bcl11b may preferentially bind to ETS motifs to sustain gene expression. To establish whether Bcl11b controls the Treg program in humans in a similar manner to mice, we performed Bcl11b ChIP-seq on sorted CD127loCD25hi CD4+ Treg cells from human peripheral blood mononuclear cells (PBMCs). Motif analysis indicated that, similar to mouse Treg cells, in human Treg cells, the two most enriched motifs were for ETS and Runx (fig. S4).

Fig. 5 Bcl11b binds at key Treg genes in mice and human Treg cells and controls chromatin accessibility.

(A) Peak annotation of Bcl11b-binding sites in Treg cells in mice. (B) Integrative Genomics Viewer (IGV) visualization at the Foxp3 locus of Foxp3 ChIP-seq in Treg cells (track 1) (35), Bcl11b ChIP-seq in mouse Treg cells (track 2), naïve CD4+ T cells (track 3), and Bcl11b ChIP-seq in human Treg cells (track 4). (C and D) IGV visualization of mouse Foxp3 ChIP-seq in Treg cells (track 1) (35), Bcl11b ChIP-seq in Treg cells (track 2), naïve CD4+ T cells (track 3), ATAC-seq (tracks 4 and 5) in Treg cells from Bcl11bF/FFoxp3YFP-Cre+/− (Bcl11b/Treg KO mosaic) and Bcl11bF/WT Foxp3YFP-Cre+/− HT mosaic female mice, and Bcl11b ChIP-seq in human Treg cells (track 6) at Ikzf2 (Helios) (C), and Il2ra (CD25), Ctla4, and Tnfrsf18 (GITR) (D). Blue highlights indicate statistically significant DA ATAC-seq peaks. Gray highlights indicate significant ChIP-seq peaks. ChIP-seq and ATAC-seq track scale bar normalized to sequences per million reads. (E) Venn diagram of Bcl11b ChIP-seq peaks from Treg and naïve CD4+ T cells (28) at Treg signature genes. (F) Peak annotation of differentially accessible ATAC-seq peaks between Treg cells of Bcl11b/Treg KO mosaic and HT mosaic female mice. (G) Metacoverage plot from HOMER of differentially accessible ATAC-seq peaks annotated to Treg signature genes in Treg cells from Bcl11b/Treg KO mosaic and HT mosaic females (left) and naïve Bcl11b KO and WT CD4+ T cells (right) (28). TTS, transcription termination site.

Bcl11b binds at Foxp3 CNS-0 and CNS-2 in mouse and human Treg cells

We then evaluated binding of Bcl11b at the Foxp3 locus in mouse and human Treg cells. Foxp3 CNSs are conserved between mice and humans (12), and CNS-2 is hypomethylated in human Treg cells, similar to CNS-2 in mouse Treg cells, suggesting a common mechanism of regulation of Foxp3 expression in mice and humans. Bcl11b bound at CNS-0 (Fig. 5B), a site located in a Treg SE and bound by the pioneer factor Satb1, which initiates Foxp3 expression by facilitating chromatin accessibility at the Foxp3 gene locus (13). In addition, Bcl11b bound at CNS-2 both in mouse and human Treg cells (Fig. 5B), confirming our previous observations by ChIP–quantitative polymerase chain reaction (32). Although not needed for Foxp3 induction, CNS-2 is required for continuous expression of Foxp3 in Treg cells (12). In addition, we found that, in mouse CD4+ T cells, Bcl11b bound at CNS-0, but not at CNS-2, supporting the idea that CNS-2 binding by Bcl11b is specific to Treg cells (Fig. 5B). The binding to CNS-2 and CNS-0 coincided with Foxp3 binding (35). Thus, Bcl11b likely promotes the continued maintenance of Foxp3 expression in human and mouse Treg cells through binding at CNS-2. This is in agreement with the reduction in Foxp3 mRNA and protein in murine Bcl11b-deficient Treg cells. Bcl11b may also act early to initiate Foxp3 expression through CNS-0.

Bcl11b binds at Treg program genes at common sites with Foxp3

Analysis of Bcl11b and Foxp3 binding at Treg signature genes showed that Bcl11b had overlapping binding with Foxp3 at approximately 50% of all the genes in the Treg gene signature, and many of those genes showed similar changes at the RNA level in Bcl11b-deficient Treg cells during steady state and in EAE (table S3). Given these results, we performed a proximity ligation assay (PLA) to determine whether Bcl11b and Foxp3 interact. We found that Bcl11b and Foxp3 interact, at least, in a subset of Treg cells (fig. S5), further supporting the previous finding that Foxp3 and Bcl11b were present in the same complex (8). Therefore, it is likely that Bcl11b and Foxp3 act cooperatively to regulate the expression of many Treg signature genes. In this line, Bcl11b bound at the promoter, an intronic region, as well as at an upstream location of the Ikzf2 (Helios) locus, in a recently identified Treg SE (Fig. 5C and fig. S6A) (13), and the peaks corresponded to regions bound by Foxp3 (Fig. 5C and data not shown). Ikzf2 mRNA and protein were reduced in Bcl11b KO Treg cells [Figs. 2 (G and H) and 4B). Genes involved in Treg cell–mediated suppression function bound by Bcl11b included Ctla4, Il2ra, and Tnfrsf18 (GITR) (Fig. 5D), which were all down-regulated at the mRNA and protein levels in Bcl11b KO Treg cells [Figs. 2 (A to F), 3 (A to F), and 4B]. Bcl11b binding occurred at the same location as Foxp3 at these Treg signature genes, some in the promoters, the introns, or the vicinity of the gene (Fig. 5D). The binding sites at Il2ra and Ctla4 were in a recently identified Treg SE (fig. S6, B and C) (13). Unexpectedly, Bcl11b also bound many of the Treg-specific genes in naïve CD4+ T cells (Fig. 5, C to E). These results demonstrate that Bcl11b regulates the Treg gene program by directly binding to the regulatory regions of Treg genes, and many times, the binding sites overlap with Foxp3 binding, suggesting that Bcl11b cooperates with Foxp3 in supporting expression of signature Treg genes.

Bcl11b binds at the same Treg signature genes in humans as in mice

In addition to conserved binding at the Foxp3 locus in humans and mice, Bcl11b bound at the same key Treg signature genes in human Treg cells as in mice (Fig. 5, C and D). Specifically, Bcl11b bound at the IKZF2 promoter and intronic regions (Fig. 5C), the IL2RA promoter and intronic regions, the CTLA4 intronic regions as well as a regulatory region upstream of the promoter, and at potential regulatory regions near the gene body of TNFRSF18 (GITR) (Fig. 5D). These results demonstrate that many of the same key Treg signature genes that were bound and regulated by Bcl11b in mice were also bound in human Treg cells, suggesting a conserved regulatory mechanism exerted by Bcl11b in human and mouse Treg cells.

Bcl11b maintains chromatin accessibility at Treg signature genes only in Treg cells

We performed assay for transposase-accessible chromatin sequencing (ATAC-seq) (36) on YFP+ Bcl11b KO and HT Treg cells to determine the global changes in chromatin accessibility dependent on Bcl11b. We found that most of the changes in chromatin accessibility occurred in the intergenic and intronic regions (Fig. 5F), similar to the distribution of Bcl11b-bound regions. A total of 39,031 differentially accessible peaks were identified, of which approximately half (19,816) were less accessible in the KO Treg cells. We then compared chromatin accessibility at Treg signature genes bound by Bcl11b in Bcl11b KO and HT Treg cells, as well as in Bcl11b KO and WT naïve CD4+ T cells. Chromatin accessibility was reduced in Bcl11b KO Treg cells, but not in Bcl11b KO CD4+ T cells (Fig. 5G), despite the fact that Bcl11b bound in naïve CD4+ T cells at many of these Treg signature genes. From the Treg-specific genes, Ctla4, Il2ra, and Tnfrsf18 all had reduced chromatin accessibility in the absence of Bcl11b in Treg cells (Fig. 5D). These results demonstrate that Bcl11b binds essential Treg signature genes and controls their chromatin accessibility in Treg cells. However, even if Bcl11b binds many of the same Treg signature genes in naïve CD4+ T cells, Bcl11b is unable to regulate the chromatin accessibility at those genes in naïve CD4+ T cells.

Bcl11b binds at myeloid TF genes in mouse and human Treg cells

Since myeloid genes were significantly up-regulated in the absence of Bcl11b, we sought to determine whether Bcl11b directly bound these genes to repress their expression. We found binding by Bcl11b at two regions upstream of the myeloid lineage TF Spi1 (PU.1) (Fig. 6A) (37), as well as at the promoter region of Cebpa, which encodes a critical TF for myeloid lineage priming and granulocyte development (Fig. 6A) (38). This is similar to Bcl11b binding in naïve CD4+ T cells and in TH2 cells during helminth infection (28). In addition, in human Treg cells, we identified two Bcl11b-binding sites upstream of SPI1, as well as binding at the promoter region of CEBPA (Fig. 6A), suggesting that Bcl11b may similarly regulate these critical lineage defining TFs in humans and mice.

Fig. 6 Bcl11b binds at NK and myeloid genes in mouse and human Treg cells.

(A and B) IGV visualization of Bcl11b ChIP-seq in mouse Treg cells (track 1) and human Treg cells (track 2) at Spi1 and Cebpa (A), Id2, Klrb1c, and Klrb1b (B). Gray highlights indicate significant ChIP-seq peaks. ChIP-seq scale bar normalized to sequences per million reads. bp, base pairs.

Bcl11b binds at the NK gene complex and Id2 in mouse and human Treg cells

Given the up-regulation of NK cell mRNAs, we next examined Bcl11b binding at NK genes in Treg cells. Bcl11b was recently shown to block the NK cell program in developing T cells by binding and repressing Id2 expression (39), which is critical for NK cell development. We found that Bcl11b bound both in mouse Treg cells and naïve CD4+ T cells at the Id2 promoter, as well as at a 5′ upstream location (Fig. 6B and data not shown). Similarly in human Treg cells, Bcl11b bound at the ID2 promoter and at multiple regions downstream of the ID2 locus (Fig. 6B). In addition, Bcl11b bound at several regions within the murine NK gene complex in Treg cells, more remarkably at two regions located between Klrb1b and Klrb1c loci and at two other regions upstream of Klrb1b (Fig. 6B). Furthermore, Bcl11b bound at an upstream location of KLRB1 (Fig. 6B), suggesting conserved binding of potential regulatory elements despite the divergence of the NK gene complex between mice and humans. These results suggest that Bcl11b blocks NK lineage programs in Treg cells in mice and humans not only through control of Id2 expression but also through direct control of NK gene complex expression.

DISCUSSION

The TF Foxp3 is a hallmark of Treg cells; however, Foxp3 is not sufficient to control the Treg program alone, and other TFs are important in this regard. This study demonstrates the fundamental role of Bcl11b in regulating the Treg signature program while repressing the innate lineage programs in mouse Treg cells. In the absence of Bcl11b, Treg cells were unable to exert their suppression function, which led to the inability to control multiorgan inflammation. Even when isolated from Bcl11b/Treg KO mosaic female mice, in the absence of inflammation, Bcl11b KO Treg cells failed to control CD45Rbhi CD4+ T cell–induced colitis in Rag1−/− mice. This failure was likely due to reduced levels of critical Treg suppression genes, including Il2ra, Ctla4, Nt5e, and Gitr. Mechanistic studies using genome-wide binding analysis of Bcl11b show that Bcl11b directly regulates the expression of many of these genes by binding to genomic regulatory regions, both in mouse and human Treg cells, thus making Bcl11b an essential TF for the Treg program both in mice and humans. All the key Treg signature genes bound and regulated by Bcl11b in mouse Treg cells, including Foxp3, Ikzf2, Il2ra, Ctla4, and Tnfrsf18, were also bound by Bcl11b in human Treg cells. In addition, the top two Bcl11b DNA binding motifs were shared between mouse and human Treg cells, suggesting that our findings on the function of Bcl11b in mice correlate highly with its function in humans.

The absence of Bcl11b caused alterations in chromatin accessibility at these genes and reduction of their expression both at the mRNA and protein levels, resulting in a failure to suppress. We found that binding of Bcl11b at the Treg signature genes overlapped with Foxp3 binding, suggesting a likely cooperative mechanism of action, supported by the observation that, at least, in a subpopulation of Treg cells, Bcl11b and Foxp3 are in proximity. Intriguingly, Bcl11b also bound the majority of the same Treg signature genes in mouse naïve CD4+ T cells, however without affecting chromatin accessibility at these genes in naive CD4+ Treg cells. Thus, Bcl11b is likely poised at certain genomic sites waiting for Foxp3 and other TFs to cooperate in promoting expression of Treg signature genes. The binding of Bcl11b at the same sites in CD4+ T cells as in Treg cells also supports the idea that Bcl11b may play differential roles in the regulation of these genes in Treg cells, where it may act in concert with Foxp3 to positively regulate their expression, and in contrast to CD4+ T cells, where Foxp3 is absent, and thus, Bcl11b may repress their expression. These results again underscore the complex and differential role of Bcl11b not only at different stages of T cell development but also at different subsets of T cells, depending on the T cell subset–specific TFs. Its role in Treg cells in the regulation of Treg signature program together with Foxp3 is one of these context-specific roles of Bcl11b. This specific role of Bcl11b contrasts with the general role of Bcl11b in both developing and mature T cells in directly repressing NK and myeloid cell TFs, such as Id2 and Spi1 (PU.1), respectively, and additionally directly repressing genes residing within the NK gene complex. Genome-wide analysis of BCL11B binding in human Treg cells suggests conserved modalities of functioning and program regulation between humans and mice.

Unbiased ChIP-seq analysis revealed that Bcl11b bound at Foxp3’s CNS-2 both in mouse and human Treg cells, required for sustained expression of Foxp3 in dividing Treg cells. As shown in this study, even in an environment devoid of inflammation, in the absence of Bcl11b, Treg cells cannot maintain Foxp3 expression in the periphery. Given that other TFs have been shown to bind to CNS-2, such as GATA3 (10) and the RUNX1-Cbfβ complex (14), the precise mechanism of sustaining Foxp3 expression remains to be elucidated. Bcl11b was also found to bind CNS-0, located in a Treg SE and critical for induction of Foxp3 expression in the thymus (13), suggesting the possibility that Bcl11b may play a role in the induction of Foxp3 expression in early Treg precursors by cooperating with Satb1, given that Bcl11b is also expressed in DP thymocytes. The pioneer factor Satb1 plays dual roles in regulating Foxp3 expression through binding at CNS-0. In the thymus, Satb1 is critical for initiating Foxp3 expression, while it represses the induction of Foxp3 expression during the generation of pTreg cells (13). Intriguingly, we found that Bcl11b bound at CNS-0 both in naïve CD4+ T cells and in Treg cells. Given our previous report that Bcl11b-deficient conventional CD4+ T cells less efficiently generated induced Treg cells in vitro, we speculate that in naïve CD4+ T cells, Bcl11b may act in opposing Satb1’s activity, thus promoting Foxp3 induction in conventional CD4+ T cells through its activity at CNS-0. This exemplifies not only the context-dependent activity of Bcl11b but also how Bcl11b can act differently with the same factor, in this case Satb1.

Bcl11b was shown to suppress myeloid and NK lineage programs early during T cell development (21, 24). However, it has become obvious that persistent expression of Bcl11b in the late stages of development and further in mature CD4+ and CD8+ T cells, as well as other T cell lineages, is required for repression of those innate cell programs (20, 28). We demonstrate here that Bcl11b is critical for repressing NK cell genes in Treg cells not only by repressing Id2 but also by directly repressing NK receptor genes, similarly to CD4+ T cells (28). It is likely that repression of the myeloid cell program by Bcl11b is through control of PU.1 and Cebpa. We found that all these genes are bound by Bcl11b in both mice and humans, providing evidence that Bcl11b’s function in repressing the NK cell and myeloid cell lineages is conserved between mice and humans.

Bcl11b KO Treg cells also showed reduced suppressive markers in conditions of inflammation, namely, in EAE, as well as in the context of tumors. Thus, it is plausible that ablation of Bcl11b would likely generate weakened Treg cells, which may acquire NK receptors and potentially gain the ability to target tumor cells instead of sustaining a tumor-suppressing microenvironment. However, in the context of autoimmunity, expression of Bcl11b needs to be maintained in Treg cells to promote their suppression program. Thus, Bcl11b may be an attractive target for cancer immunotherapies, but careful consideration will be required, given its role in maintaining Treg cell suppression program in autoimmunity.

MATERIALS AND METHODS

Mice

Mice were bred under specific pathogen–free conditions and kept in-house for experiments in individually ventilated cages under specific pathogen–free conditions. Foxp3YFP-Cre+ mice were purchased from the Jackson laboratory (33) and bred with Bcl11bF/F mice on a C57Bl/6 background to generate (i) homozygous Bcl11bF/F Foxp3YFP-Cre+ male mice and Bcl11bF/F Foxp3YFP-Cre+/+ female mice (Bcl11b/Treg KO), with Bcl11b deleted in all Treg cells, and (ii) Bcl11bF/F Foxp3YFP-Cre+/− (Bcl11b/Treg KO mosaic), in which in addition to Bcl11b-deficient Treg cells, WT Treg cells were also present. The controls for this group were Bcl11bF/WT Foxp3YFP-Cre+/− HT mosaic female mice. All experiments were conducted according to the animal protocols approved by the University of Florida Institutional Animal Care and Use Committee.

EAE was induced in Bcl11bF/F Foxp3YFP-Cre+/− (Bcl11b/Treg KO mosaic) and age-matched control female mice, as previously described (27). Experiments were conducted at day 10 after induction of EAE.

Flank tumor inoculation

B16-Ova melanoma cells (5 × 105) were injected subcutaneously in the right flank of Bcl11b/Treg KO mosaic female mice, tumors were isolated on day 14, and immune cells were isolated as described below.

In vivo suppression assay

The in vivo suppression assay was conducted as previously described by Mottet et al. (34). Briefly, B6.Rag1−/− mice were injected with 4 × 105 CD4+CD45RBhi from CD45.1 WT mice to induce colitis and with 1 × 105 YFP+ Treg cells from Bcl11bF/F Foxp3YFP-Cre+/− (Bcl11b/Treg KO mosaic) (CD45.2) or from Bcl11bF/WT Foxp3YFP-Cre+/− (HT mosaic) (CD45.2) mice, or no additional cells. Mice were weighed twice a week.

Colitis histopathology scores

Colonic tissue was removed from the mouse, and a 1.0-cm section was cut out from the distal portion of the colon, 1.5 cm away from the anus. The colon section was fixed overnight in 10% formalin and then embedded in paraffin. Sections of 5 μm were stained by H&E or Alcian Blue (pH 2.5). Colitis histopathological scores were established using the following criteria: immune infiltration—none = 0 to marked, dense infiltration = 4; extent of infiltration—none = 0 to mucosal, submucosal, and transmural = 3. Epithelial changes included hyperplasia visible as crypt elongation: none = 0 to marked = 4; goblet cell loss: none = 0 to marked = 4; and erosion as loss of surface epithelium; 0 to 4. Mucosal architecture scoring included ulceration: 3 or 4, irregular crypts: 4, and crypt loss: 0 to 4.

Cell isolation and analysis

Cell isolation was performed as previously described by VanValkenburgh et al. (32). For isolation of immune cells from B16-Ova flank tumors, tumors were excised from mice, and tissue was homogenized using the GentleMacs system on m.Tumor.01 (Miltenyi). Tissue was further incubated with collagenase D (2.0 mg/ml; Worthington) and 7.5 of μl deoxyribonuclease (Sigma-Aldrich) for 30 min at 37°C in a shaking incubator. Samples were passed through a 70-μm filter, and tumor cells were removed by a 44/66% Percoll gradient centrifugation. For intracellular and TF staining, the eBioscience Foxp3/Transcription Factor Staining Buffer Set (cat. no. 00-5523-00) was used. Flow cytometry was performed on a BD LSR II, data acquisition was performed on BD FACSDiva Software, and flow cytometry files were analyzed on FlowJo. Antibodies are listed in table S4.

RNA-seq library preparation and data processing

RNA was extracted from sort-purified WT and Bcl11b KO Treg cells from steady state and on day 10 after induction of EAE using the RNeasy Plus Micro Kit (74034, Qiagen), and ribosomal RNA (rRNA) was depleted using NEBNext rRNA Depletion Kit (E63102, New England Biolabs). Complementary DNA libraries were prepared using NEBNext Ultra II RNA Library Prep Kit (E7770S, New England Biolabs) and sequenced on a NextSeq 500 at approximately 30 million reads per library (28). RNA-seq fastq files were trimmed using seqtk, and the quality was assessed using FastQC (28). The trimmed reads were aligned to the mouse genome (mm10) using Hisat2, and transcript counts were obtained using HTseq (28). DeSeq2 was used for differential expression analysis (40).

Assay for transposase-accessible chromatin sequencing

ATAC-seq was performed as previously described (28, 36) on 50,000 sort-purified YFP+ Treg cells from Bcl11b/Treg KO and HT mosaic mice. Library quality and adapter dimer contamination were assessed on a Bioanalyzer (#5067-4626, Agilent). Libraries were sequenced as PE75 on a NextSeq500.

Bcl11b ChIP-seq

Treg cells were isolated from the spleens and peLNs of naïve mice and enriched using the EasySep Mouse CD4+ CD25+ Regulatory T Cell Isolation Kit II (#18783, STEMCELL). Human Treg cells were isolated from PBMCs by sorting CD4+ CD25+ CD127 cells. Bcl11b ChIP-seq was performed on Treg cells using the SimpleChIP Enzymatic Chromatin IP Kit from Cell Signaling Technology (CST #9003), following their recommended protocol up to the library preparation with the following alterations: (i) 107 cells were used in place of 4 × 107, (ii) formaldehyde fixation was performed in 1 ml, (iii) buffer A and B steps were all performed in 1 ml, (iv) cells were sheared in ChIP buffer to an average of 200 to 1000 base pairs using a Diagenode Bioruptor Pico (B01060002), (v) a cocktail of three anti-Bcl11b antibodies were used (4 μg of ab18465, 4 μg of CST #12120, and 4 μg of Bethyl A300-385 per ChIP), and (vi) final DNA purification was performed by phenol:chloroform:isoamyl alcohol extraction and MaXtract high-density columns (#129046, Qiagen). Libraries were prepared using NEBNext Ultra II DNA Library Prep Kit (#E7645S, New England Biolabs) and sequenced as SE75 on a NextSeq500.

ATAC-seq and ChIP-seq data processing

Data analysis for ATAC-seq was performed using a customized pipeline based on a previously published version (28, 41). Briefly, fastq files were first trimmed using Trimmomatic, and read quality was assessed using FastQC. Reads were aligned to the mouse genome (mm10) using Bowtie2. The resulting SAM file was pruned (Q > 30), converted to BAM format, and sorted using SAMtools. Duplicate reads were removed using Picard tools, and MACS2 was used for peak calling and bedgraph file preparation. Bedgraph files were visualized using IGV. ChIP-seq data were analyzed as ATAC-seq except using single-end rather than paired-end settings. Differential chromatin accessibility was conducted using MAnorm with the cutoff set at a minimum fold change of 2 and significance of P < 0.05. P value was adjusted for the FDR by the Benjamini-Hochberg adjustment.

Proximity ligation assay

PLA was conducted by the Duolink PLA–fluorescence-activated cell sorting (FACS) protocol (Sigma-Aldrich). Briefly, WT Treg cells were isolated using the EasySep Mouse CD4+ CD25+ Regulatory T Cell Kit II (#18783, STEMCELL). CD4+ T cells were sorted and isolated from the negative fraction after CD25+ selection. Bcl11b KO Treg cells were sorted from mosaic female mice. Cells (1 × 105) were fixed in 2% paraformaldehyde for 10 min at room temperature and permeabilized with 0.5% Triton X-100 for 30 min at room temperature. Cells were incubated with rabbit anti-Bcl11b (Bethyl A300-385) and rat anti-Foxp3 antibody (126402, BioLegend) for 30 min, followed by addition of oligonucleotide (probes)–conjugated secondary antibodies against rabbit and rat, respectively. Two circle-forming oligonucleotides and a ligation enzyme were subsequently added and ligated to the probes following the Duolink PLA-FACS protocol, with the difference that the amplification was conducted overnight. Rolling circle amplification occurs only when the two probes are <40 nm apart, and fluorescently labeled complementary oligonucleotides bind to the amplified DNA, allowing detection by FACS.

Motif enrichment between activated and repressed genes

Bcl11b ChIP-seq peaks were annotated to genes using HOMER and matched to significantly up-regulated or down-regulated genes. Motif analysis was conducted on each peak set using HOMER. The top 10 known motifs from activated or repressed gene sets were further analyzed. Fisher’s exact test was conducted to determine whether each motif was significantly enriched in either the up-regulated or down-regulated gene set.

Correlation analysis between steady-state and EAE RNA-seq

All differentially expressed genes log2 (fold change) between WT and KO Treg cells at steady state and EAE were paired. These paired values were subsequently matched to the gene lists for pathways enriched by GSEA and used to calculate Spearman correlations for each gene list.

Statistics

For nonsequencing experiments, three or more mice per group were used, and the experiments were repeated at least three times independently. For YFP+ Bcl11b KO Treg cells and YFP WT Treg cells from the same mouse, paired Student’s t test was used for statistical analysis, and P < 0.05 was considered significant. In experiments where single comparisons were made across different mice, two-tailed Student’s t test was used for statistical analysis, and P < 0.05 was considered significant. In the case of multiple groups of mice of different genotypes, we used an analysis of variance (ANOVA) or a Kruskal-Wallis test applicable for continuous variables. Significance for RNA-seq data was determined using DeSeq2, using P < 0.05 and adjusting for the FDR by Benjamini-Hochberg adjustment.

SUPPLEMENTARY MATERIALS

Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/5/8/eaaw0480/DC1

Fig. S1. Treg surface markers remain decreased in Bcl11b KO Treg cells in tumors.

Fig. S2. Bcl11b KO Treg cells in Rag1−/− mice transferred with CD45Rbhi CD4+ T cells to induce colitis.

Fig. S3. Bcl11b controls similar pathways during steady state and inflammation.

Fig. S4. Top Bcl11b ChIP-seq binding motifs in mice and human Treg cells.

Fig. S5. Bcl11b and Foxp3 interact in a subset of Treg cells.

Fig. S6. Bcl11b binds at key Treg gene superenhancers.

Table S1. WT and KO Treg phenotype in EAE and tumor compared with steady-state Treg cells.

Table S2. Top 10 motifs found in peaks annotating to up-regulated and down-regulated genes.

Table S3. Treg signature genes commonly bound by Bcl11b and Foxp3 in Treg cells.

Table S4. Antibodies used for flow cytometry staining and ChIP-seq.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.

REFERENCES AND NOTES

Acknowledgments: We gratefully acknowledge J. Cooper, M. C. Moore, and K. Luque-Sanchez for technical assistance. Funding: This work was supported by NIH grants R01AI067846 (to D.A.), R01AI078273 (to D.A.), and R21AI13120501 (to D.A.), University of Florida Gatorade Trust and UF Health Cancer Center (to D.A.), R01DK105562 (to L.Z.), 2T32DK074367 (to T.T.D.), and T32AI007110 (to A.Z. and K.J.L.). Author contributions: Conceptualization: T.T.D., E.H., and D.A. Experimental design: T.T.D., E.H., M.U., and D.A. Methodology: T.T.D., E.H., N.C., P.K., X.L., M.U., S.M., U.P., A.Z., J.J.C., Z.X., and D.A. Performed experiments: T.T.D., E.H., M.U., S.M., U.P., A.Z., J.J.C., and Z.X. Software: K.J.L., E.H., and X.H. Formal analysis: T.T.D., M.U., E.H., N.C., P.K., Z.H., and D.A. Resources: L.Z. and D.A. Data curation: T.T.D., E.H., and D.A. Writing: T.T.D., E.H., and D.A. Competing interests: The authors declare that they have no competing interests. Data and materials availability: RNA-seq, ChIP-seq, and ATAC-seq data shown in this study have been deposited in the Gene Expression Omnibus (GEO) database with accession GSE120869. All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.
View Abstract

Navigate This Article