Bcl11b sets pro-T cell fate by site-specific cofactor recruitment and by repressing Id2 and Zbtb16.

Multipotent progenitor cells confirm their T cell-lineage identity in the CD4-CD8- double-negative (DN) pro-T cell DN2 stages, when expression of the essential transcription factor Bcl11b begins. In vivo and in vitro stage-specific deletions globally identified Bcl11b-controlled target genes in pro-T cells. Proteomics analysis revealed that Bcl11b associated with multiple cofactors and that its direct action was needed to recruit those cofactors to selective target sites. Regions near functionally regulated target genes showed enrichment for those sites of Bcl11b-dependent recruitment of cofactors, and deletion of individual cofactors relieved the repression of many genes normally repressed by Bcl11b. Runx1 collaborated with Bcl11b most frequently for both activation and repression. In parallel, Bcl11b indirectly regulated a subset of target genes by a gene network circuit via the transcription inhibitor Id2 (encoded by Id2) and transcription factor PLZF (encoded by Zbtb16); Id2 and Zbtb16 were directly repressed by Bcl11b, and Id2 and PLZF controlled distinct alternative programs. Thus, our study defines the molecular basis of direct and indirect Bcl11b actions that promote T cell identity and block alternative potentials.

INTRODUCTION

The zinc finger transcription factor Bcl11b is required for development of αβ T cells and most γδ T cells[1-3]. Its expression initiates precisely during T cell-lineage commitment[4], i.e. before T cell antigen receptor expression and between the DN2a (CD25+CD44+c-Kithi+) and DN2b (CD25+CD44+c-Kit+) pro-T cell stages, and progression through T cell-lineage commitment is blocked or highly abnormal in cells lacking Bcl11b[5-7]. Bcl11b-deficient pro-T cells are less sensitive to Notch signaling and more prone to differentiate into natural killer (NK) cells than wild-type counterparts[6,8]. They also fail to go through β-selection, due to defects in T cell gene expression as well as abnormal persistence of immature features including c-Kit expression[5-9]. Deletion of the Bcl11b gene after β-selection causes abnormal activation of effector genes[10,11] and multiple functional defects in later thymocytes and mature T cells[12-14]. While the importance of Bcl11b for T cell development is clear, its exact mechanism of action is not. Bcl11b can bind to GC-rich sequences in DNA[15] and recruit chromatin-modifying NuRD and SIRT1 complexes[16,17], but in pro-T cells it primarily binds Ets and Runx motif-enriched sites in open chromatin[7,18]. Previous work has implicated Bcl11b in both activation and repression[5,6,8,10,12,19,20], with the most consistent effects across development on a core of genes that apparently require repression by Bcl11b in T cells[7,11]. Finally, Bcl11b effects have a striking overlap with effects of the basic helix-loop-helix protein E2A in early T cells[7], yet the basis for this convergence is not known.

This report addresses three questions about Bcl11b roles in establishing T cell commitment. First, what are the directly regulated target genes of Bcl11b during T cell commitment? Second, what are the mechanisms that Bcl11b deploys to work as an activator or a repressor at its target sites? We identify direct target loci based on a new criterion for functional sites of Bcl11b action, through its role in recruiting specific cofactors. Finally, how many of the effects of Bcl11b are indirect, and how are they mediated? We show that Bcl11b in pro-T cells blocks expression of E-protein antagonist Id2 and the innate-response regulator PLZF (encoded by Zbtb16). Id2-mediated suppression of E protein activity is important for all innate lymphoid cells including NK cells[21-26], while PLZF is crucial for innate-type T cells and for non-cytolytic ILCs[24,27-31], and both govern subsets of myeloid and dendritic cells[32-34]. We show that a gene network relating Id2 and Zbtb16 to Bcl11b function sheds light on the split between the T and innate immune cell families of developmental programs.

Results

Bcl11b impacts on gene expression in DN2/3 stage thymocytes

We previously showed that Bcl11b regulates a distinctive set of genes during initial T cell-lineage commitment of fetal-liver-derived precursors differentiating in vitro, including many targets that appear to be unique to this developmental period[7]. To examine its commitment role in vivo, we compared DN pro-T cells in mice where Bcl11b was conditionally deleted with Vav1-iCre[35], expressed in all hematopoietic cells, against those in mice where Bcl11b was deleted with Lck-Cre (Lck proximal promoter), from an early-expressed transgene[36] first activated in DN2 pro-T cells (Fig. 1a, Supplementary Fig. 1a). The mice also contained a Cre-dependent ROSA26R-YFP reporter, which distinguished cells with deleted alleles from normal DN2a cells. In animals with wild-type (WT) Bcl11b, Vav1-iCre caused all DN thymocytes to express YFP (Fig. 1a, top), whereas Lck-Cre activated YFP only in DN2b and later cells (Fig. 1a, below; Supplementary Fig. 1a). Thus, Lck-Cre only deleted genes after Bcl11b would normally be turned on[4]. Homozygous Bcl11b mice bred with either of these Cre transgenes showed similar-appearing arrests of T-cell precursors with a c-Kithi+CD25+ phenotype resembling normal DN2a cells (Fig. 1a). In the Lck-Cre+ Bcl11b mice, however, the c-Kithi+ DN2a-like cells comprised two populations, a YFP-negative, CD44+ one enriched for true DN2a cells, and a much larger YFP+CD44lo one generated only upon Bcl11b deletion (Supplementary Fig. 1a,b). Thus, Bcl11b excision could generate the YFP+ c-Kithi+CD25+ phenotype by retrograde-like differentiation from cells that had previously reached DN2b stage after activating Bcl11b initially.

Figure 1:

Cellular and molecular phenotypes of in vivo Bcl11b deletion by Vav1-iCre or Lck-Cre.

(a), Representative flow cytometric analysis of DN thymocytes from Bcl11b+/+, Bcl11bfl/+, and Bcl11bfl/fl rYFP Vav1-iCre or Lck-Cre mice, showing gates used for defining DN subpopulations (top) and sorting strategy used for purifying YFP–CD25+ and YFP+CD25+ thymocytes (bottom). n ≥ 5 mice of each genotype.

(b, c), Heatmaps generated from transcriptomic analysis of sorted DN thymocytes using RNA-seq. Columns within each category represent cells from individual mice. (b, c) Hierarchical clustering analysis of Bcl11b-repressed (b), and Bcl11b-dependent (c) genes differentially expressed in both Vav1-iCre and Lck-Cre deletions of Bcl11b. Vav1-iCre DEG are identified by comparing the average of Vav1-iCre;Bcl11b+/+ (WT), and Vav1-iCre;Bcl11bfl/+ (HE) together to Vav1-iCre;Bcl11bfl/fl (KO). Lck-Cre DEG are identified by comparing the average of DN3 WT to Lck-Cre;Bcl11b KO cells. Color scale shows fold change relative to average of WT DN2 samples. For gene names, see Supplementary Table 1.

(d, e), Identification of subsets of Bcl11b DEGs that are expressed at lower (d) or higher (e) levels when Bcl11b is deleted with Vav1-iCre than when it is deleted with Lck-Cre. Hierarchical clustering analysis shows expression differences of significantly differentially expressed genes (|Log2FC|>1, FDR<0.05) between Vav1-iCre-deleted and Lck-Cre deleted Bcl11b-deficient CD25+ thymocytes, with color scale representing fold change relative to average for Lck-Cre KO samples.

DN thymocyte gene expression patterns in mice with either Vav1-iCre or Lck-Cre showed that about 300 genes were reproducibly upregulated in YFP+ homozygous Bcl11b knockout DN2-like thymocytes as compared to YFP+ control WT or heterozygous Bcl11b DN2 and DN3 thymocytes (Fig. 1b), defining Bcl11b-repressed genes. About 220 genes were significantly downregulated in these Bcl11b knockouts (Fig. 1c), defining Bcl11b-dependent genes. These criteria [false discovery rate (FDR) <0.05, |log2Fold Change (FC)|>1, average reads per kilobase million (RPKM)>1; Supplementary Tables 1,2] defined Bcl11b-regulated “differentially expressed genes” (DEGs) in young adult thymocytes. Although the Bcl11b knockout cells resembled normal DN2a thymocytes, their gene expression patterns sharply distinguished the mutant cells from any normal Bcl11b+/+ subsets. Highly robust effects were seen in multiple samples, e.g. for Bcl11b itself, Zbtb16 (encoding PLZF) and Id2 (both Bcl11b-repressed), and Cd6 (Bcl11b-dependent) genes (Supplementary Fig. 2). Interestingly, certain differentially expressed genes also showed partial de-repression in YFP+ Bcl11b heterozygous cells (Fig. 1b; Supplementary Table 1).

At a subset of these loci, the effects of Bcl11b deletion prior to commitment (Vav1-iCre) were more severe than the effects of deletion after commitment (Lck-Cre), and Vav1-iCre-mediated deletion had more severe effects on total thymus cellularity (Supplementary Fig. 1c). About 40 Bcl11b-dependent genes failed to be turned on in Vav1-iCre;Bcl11b deleted cells, but were expressed somewhat in Lck-Cre;Bcl11b deleted cells (Fig. 1d), while ~85 Bcl11b repression targets were more overexpressed in the Vav1-iCre knockout cells than in the Lck-Cre knockout cells (Fig. 1e). This difference suggests that even transient Bcl11b expression in the Lck-Cre deletion model provides some needed function for T-cell development. The most timing-sensitive genes included Cd3g, Cd3d, Cd3e, and Dntt among Bcl11b-dependent genes, and progenitor-associated, γδ-associated, and alternative-lineage associated genes Pou2af1, Tyrobp, Cd7, Itgae, Itgb7, Klf2, Trpm1, Cd163l1 (Scart1), Cited4, and Tnni1 among Bcl11b-repressed ones. However, most DEGs required Bcl11b both during and continuously after commitment, for activation or especially for repression in pro-T cells.

For mechanistic experiments on Bcl11b function, we also defined the genes affected by acute Cas9-mediated disruption of Bcl11b in pro-T cells differentiating in vitro, as shown in Supplementary Fig. 3a (Supplementary Table 2). Bone marrow precursors from B6.ROSA26-Cas9; Bcl2-tg transgenic mice (Cas9-Bcl2 cells), were cultured with OP9-DL1 stroma for 7 days, then transduced with retroviral vectors encoding sgRNA, and 7 d later analyzed and harvested for RNA-seq analysis. Most of these genes also overlapped with the significant DEGs from both Lck-Cre and Vav1-iCre deletion in vivo (Supplementary Fig. 3b; Supplementary Table 3). DEGs significant in all three are listed in Table 1, and include most of the targets previously reported in fetal liver-derived pro-T cells[7] (Supplementary Fig. 3c; see Methods). Importantly, genes encoding transcription factors Id2 and PLZF (Zbtb16) were highly significant Bcl11b repression targets in DN2/DN3 cells in every case. Thus, during pro-T cell commitment, Bcl11b activates and represses several hundred genes important for T cell identity.

Table 1:

HIGHEST CONFIDENCE BCL11B-REGULATED TARGET GENES

A. Genes upregulated in Bcl11b knockout DN2/3 cells: Bcl11b repression targets
1700024P16Rik	1700112E06Rik	2900026A02Rik	41340	5730559C18Rik	9030617O03Rik
9130019P16Rik	Aatk	AB124611	Acot11	Adamtsl3	Aifm2
Ak4	Akap6	Aoah	Apobr	Appl2	Arap3
Arhgap20	Arl10	Arl6	Arsi	Art2a-ps	Art2b
Arvcf	Asph	Bace1	Batf3	Bcar3	Bcas1os2
Calm1	Camk2b	Camkmt	Card11	Cass4	Ccdc157
Ccr2	Cd160	Cd163l1	Cd244	Cd63	Cd7
Cd72	Cd9	Ceacam1	Cers4	Chad	Chn2
Chpf	Cisd3	Cited4	Cldn2	Clnk	Clvs1
Cnbd2	Cnksr3	Cnnm2	Cnr2	Col9a3	Colq
Coprs	Cpd	Cpeb2	Cpne7	Cpt1a	Ctbp2
Cx3cr1	Cxcr5	Dab2ip	Dapl1	Dennd3	Disc1
Dlx1	Dlx1as	Dok3	Drc7	Dscam	Dsp
Dyrk3	Eci1	Eea1	Egln3	Fam129a	Fam151a
Fam179a	Fam46a	Fam71b	Fbxw8	Fcer1g	Fdx1
Fes	Fgf3	Flna	Flt3	Frmd4b	Gas7
Gimap4	Gimap6	Gimap7	Gimap8	Glis2	Golm1
Gpr141	Gpr183	Gstm1	Gstm3	Gzmc	Hey2
Heyl	Hip1r	Ica1l	Id2	Il2rb	Iqgap2
Irak3	Itgb6	Itgb7	Ivns1abp	Kank2	Kcne3
Kif13a	Kif5a	Kifc3	c-Kit	Klhdc2	Klrd1
Lag3	Layn	Ltk	Lyn	Man1c1	Micall1
Myo1e	Myo1f	Myo7a	Nat8l	Nav2	Nectin1
Neurl3	Nfil3	Npffr1	Nr2f6	Nrgn	Nt5e
Osbpl5	Oscp1	Papss2	Pcyox1l	Pde2a	Pde4a
Pear1	Pik3r2	Pkig	Plekhg5	Polm	Pou2af1
Prtg	Ptpn14	Ptpn21	Rab19	Rapgef2	Rassf4
Rgs3	Rhobtb1	Rnh1	Rora	S100a10	S1pr3
Sccpdh	Scn5a	Sema3c	Sema4c	Serpinb9	Sh2d1b1
Sh3bgrl2	Siae	Slc22a15	Slc22a23	Slc2a6	Slc35f5
Slc45a3	Spa17	Spaca9	St6galnac6	Sult5a1	Tas1r1
Tgfb3	Tiam2	Timp2	Tirap	Tjp2	Tmem126a
Tmem198	Tmem231	Tmem67	Tnfrsf25	Tnni1	Trf
Trim2	Trnp1	Trpm1	Tspan32	Tyrobp	Ublcp1
Vsig2	Wipi1	Wnt5b	Xcl1	Zbtb16	Zbtb7b
Zcchc18	Zfp105	Zfp296	Zfp316	Zfp518b	Zfp568
Zfp768
B. Genes downregulated in Bcl11b knockout DN2/3 cells: Bcl11b-dependent genes
1300002E11Rik	Abtb2	Actn1	Acy3	Bbof1	BC025920
Bst1	Camkv	Ccdc153	Ccdc18	Cd3d	Cd3g
Cd5	Cd6	Cldn4	Comp	Dapk1	Dcp1b
Def8	Dgke	Dgkeos	Dgkg	Ehd3	Emp1
Evpl	F730043M19Rik	Fam117a	Fer1l5	Frmd4a	Frmd6
Gbp4	Gbp8	Glyctk	Gm15708	Gm26839	Gmpr
Grasp	Hid1	Hmgcs2	Id3	Ifngr2	Il10ra
Il21r	Kif7	Llgl1	Map4k2	Matk	Mpzl2
Mzb1	Nfatc2	Nfkb2	Pisd-ps1	Pisd-ps2	Pisd-ps3
Plcg1	Plekha7	Plxnd1	Pqlc3	Rab11fip4	Rasal1
Rgs10	Rhbdf2	Sestd1	Sirt5	Slc37a1	Slc5a9
Smim5	Smox	Smyd2	Sox13	Spib	Steap3
Synj2	Tbxa2r	Tmem131	Tmem221	Tmprss4	Tnfrsf1a
Ttc38	Zeb2os

Table lists genes that show significant differential expression from wildtype controls in all three of the following conditions: in Lck-Cre deletion of Bcl11b in vivo, in Vav1-iCre deletion of Bcl11b in vivo, and in Cas9-mediated deletion Bcl11b in DN2b/3 cells differentiating in vitro from bone-marrow derived precursors. See Supplementary Table 3 for complete lists of genes identified in each of the individual comparisons. Genes are listed in alphabetical order; see Supplementary Tables 1,2 for expression values.

A challenge of functional specificity

To determine which genes were directly regulated by Bcl11b, we performed Bcl11b ChIP-seq. However, despite the prevalence of Bcl11b repressive effects, binding of Bcl11b across the genome appeared preponderantly at open chromatin sites with active marks[7,18], globally associated with active genes, and showed no distinction between non-regulated loci and those with any response to Bcl11b deletion (Supplementary Fig. 4a; Supplementary Table 2). We tested whether distinct motifs were bound at functional sites. Most sites bound by Bcl11b in DN3 pro-T cells are enriched for Ets and Runx motifs[7,18]; we compared these with GC-rich sequences originally reported as a cognate site[15] and with three newly reported motifs for Bcl11b binding to protein-binding microarrays[37]. While matches to the newly defined sites (log odds ≥5) were found at 10–18% of Bcl11b occupancy sites, they were not enriched at DEGs (Supplementary Table 4). Bcl11b binding sites in CpG islands were actually strongly depleted near DEGs (Supplementary Fig. 4b). Bcl11b occupancy and target site motifs were thus unable to distinguish sites of Bcl11b function.

Bcl11b interacts with several “repressor” complexes

We reasoned that the sites where Bcl11b exerts regulatory functions could be identified through its local interactions with specific protein factors. To identify the functional components of Bcl11b complexes that control gene expression in pro-T cells, cells of a DN3-like cell line[38,39], Scid.adh.2c2 (Fig. 2a), were transduced with Myc-Flag-tagged Bcl11b, and the Bcl11b-containing protein complexes were subjected into two-step affinity purification followed by SDS/PAGE and silver staining (Fig. 2b). Liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis identified more than 300 molecules with supra-threshold enrichment (Supplementary Table 5). The bifunctional transcription factor Runx1 and proteins annotated as involved in “negative regulation of gene expression”, “transcriptional regulation”, and “chromatin remodeling” were highly enriched (Fig. 2c). The most enriched Bcl11b interaction partners included multiple members of the Nucleosome Remodeling Deacetylase complex (NuRD), RE-1-silencing transcription factor (Rest) complex (NRSF), and Lysine-specific demethylase 1A (Kdm1a or LSD1) complexes, with lower but still substantial scores for Polycomb repressor complex 1 (PRC1), confirming earlier evidence for Bcl11b-NuRD association[16] (Fig. 2d). Association of major components of these complexes, Chd4, Mta2, Rnf2 (Ring1b), Rest, LSD1 and Hdac2, with Bcl11b was validated by immune co-precipitation (Fig. 2e). Bcl11b has also been found in SWI/SNF complexes[40], and the Smarca4 (Brg1) component was also specifically enriched (Supplementary Table 5), but not most SWI/SNF components. Note that although LSD1, Chd4, Mta2 and Ring1b act as components of repressor complexes, in specific contexts they too may contribute to activation of genes[41-45]. Thus, both “repressor complexes” and Runx1 might play roles in Bcl11b-mediated positive or negative gene regulation.

Figure 2:

Identification of Bcl11b interacting molecules in early T cells

(a), Schematic representation of phenotype of Bcl11b-deficient pro-T cells. Bcl11b is turned on during T-lineage commitment. Bcl11b-deficient cells retain a DN2a-like phenotype, when the Bcl11b gene is deleted either before or after commitment. Scid.adh.2c2 cell is a DN3-like cell line.

(b), Identification of Bcl11b complexes. Total extracts from Myc-Flag-Bcl11b-expressing Scid.adh.2c2 cells and mock-transduced controls were subjected to two-step affinity purification followed by SDS-PAGE and silver staining. All of the visible bands from both samples were subjected to liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis.

(c), Bcl11b associates with molecules involving positive and negative regulation of transcription. Gene ontology (GO) analysis for Bcl11b interacting molecules was performed. Top ten GO terms are shown.

(d), Schematic representation of Bcl11b–repressor complexes detected by LC-MS/MS. Magenta indicates cofactors that will be discussed in detail.

(e), Bcl11b interacts with several repressor complexes. Total extracts from Myc-Flag-Bcl11b-expressing Scid.adh.2c2 cells and mock-transduced controls were subjected to immunoprecipitation (IP) with anti-Flag mAb followed by immunoblotting (IB) (left). Nuclear lysates were also subjected to IB in parallel (right). Gels are cropped to focus on protein species migrating near the mobilities of the indicated size markers.

Data are representative of two independent experiments (b, e).

Identification of Bcl11b-dependent cofactor binding sites

To test whether differential gene expression in primary pro-T cells was linked to genomic regions where specific cofactor assemblies might be nucleated by Bcl11b, we performed ChIP-seq analysis of wildtype and Bcl11b-deleted pro-T cells. To obtain the large cell numbers needed, we used in vitro differentiation cultures with Cre-ERT2 activation by 4-OH tamoxifen (4-OHT) to delete Bcl11b (Supplementary Fig. 4c). On day 7, before 4-OHT treatment, most cells showed a DN2a/b phenotype (Lin-CD45+c-Kit+CD25+, Supplementary Fig. 4d). Five days after 4-OHT removal, control cells had efficiently progressed into a c-Kitlo DN3 stage, while Bcl11b-deficient cells were still characteristically c-Kithi+ (Fig. 3a, Supplementary Fig. 4e). ChIP-seq analysis identified more than 25,000 reproducible Bcl11b peaks in Bcl11b+/+;Cre-ERT2 control DN3 cells (including 82% of peaks previously reported in B6 DN3 cells without 4-OHT[7], Supplementary Fig. 4f) and these peaks almost completely disappeared in Bcl11b-deleted cells (Fig. 3b). Of the factors tested, only Brg1 could not be mapped using commercially available reagents (data not shown). In the control cells, Chd4, Mta2, Rest, Ring1b, LSD1 and Runx1 bound at ~6500 (Rest)–~33,000 (Runx1) sites each. These overlapped with Bcl11b peaks to different extents (Fig. 3c). However, the results clearly showed that distinct subsets of peaks for each cofactor depended on the presence of Bcl11b for their recruitment. Bcl11b deletion caused large fractions of some cofactor peaks to disappear (Mta2, Rest) or relocate (Chd4, Ring1b, LSD1, Runx1)(Fig. 3c). In each case, most of the Bcl11b-dependent peaks coincided with sites bound by Bcl11b in wildtype cells (Fig. 3c, green), implying that Bcl11b binding itself was needed to recruit that cofactor to such sites. The number of Mta2 peaks was especially sharply reduced by deletion of Bcl11b, although Mta2 protein abundance in Bcl11b-deficient cells was comparable to that in wildtype DN3 cells (Supplementary Fig. 4g), suggesting that most of its genome-wide associations in these cells depend on Bcl11b.

Figure 3:

Identification of Bcl11b-dependent cofactor peaks in DN3 cells

(a), Flow cytometric analysis of BM-derived precursors after 14 days of OP9-DL1 co-culture were performed. Representative c-Kit/CD25 profiles in Lin- and CD45+ cells are shown with the percentages of cells in each quadrant.

(b), Number of Bcl11b ChIP peaks in DN3 cells. Bcl11b ChIP-seq analyses were performed using wildtype or Bcl11b-deficient cells depicted in (a). Venn diagrams show the number of ChIP peaks in each sample. In parentheses, total number of peaks in each sample is indicated.

(c), Bcl11b-dependent cofactor peaks. ChIP-seq analyses for cofactors were performed. Venn diagrams, approximately to scale, show the number of ChIP peaks in each sample with Bcl11b peaks in wildtype DN3 cells. Bcl11b-dependent cofactor peaks are indicated in green along with the number of peaks.

Data are representative of four independent experiments (a) or are based on reproducible ChIP-seq peaks in two replicate samples (b, c).

Only a small subset of Bcl11b sites showed Bcl11b-dependent cofactor binding, among the tens of thousands of Bcl11b peaks across the genome. Some of the genes repressed by Bcl11b not only had Bcl11b-dependent cofactor peaks (Fig. 4a–d, magenta rectangles), but also showed newly generated cofactor peaks occupying distinct sites if Bcl11b was deleted (Fig. 4a,b, green rectangles). Motif preference for cofactor binding changed when Bcl11b was absent. Genome-wide, not only ETS and Runx family but also bHLH motifs were highly enriched among Bcl11b-dependent cofactor peaks (Supplementary Fig. 5a,b), and a minority of these sites included the motifs defined recently using protein binding microarrays[37], similar to Bcl11b sites overall (Supplementary Table 4). However, the “new” cofactor peaks that appeared specifically when Bcl11b was deleted had sharply altered motif distributions, with bHLH motifs (“E2A” or “Ptf1a”) much less common and bZIP, HMG family, and other motifs more commonly enriched (Supplementary Fig. 5c). Bcl11b at CpG islands overlapped mostly with sites where Chd4, Ring1b, LSD1 or Runx1 were engaged whether Bcl11b was deleted or not (Supplementary Fig. 5d). Thus, Bcl11b both facilitated cofactor binding to a subset of Bcl11b occupancy sites and antagonized cofactor binding to other sites.

Figure 4:

Bcl11b-dependent cofactor peaks around the major Bcl11b target genes

(a–d), Binding patterns of Bcl11b, Chd4, Hdac2, Mta2, Rest, Ring1b, LSD1 and Runx1 in wildtype DN3 cells and Bcl11b-deficient CD25+ cells, with H3K27Ac ChIP-seq and RNA-seq tracks of control and Bcl11b-deficient cells. Data are representative of two independent ChIP experiments. The Id2 (a), Zbtb16 (b), Tnni1 (c) and Cd163l1 (d) loci are shown. Magenta rectangles: Bcl11b-dependent cofactor peaks. Green rectangles: cofactor peaks specifically detected in Bcl11b-deficient cells.

(e, f), Differentially expressed genes in Bcl11b-deficient cells are bound by Bcl11b-dependent cofactor peaks. DEGs were from 4 independent pairs of control and Bcl11b-deleted samples; |log2FC|>1, FDR < 0.05, RPKM >1 (Supplementary Table 2). Numbers of Bcl11b-repressed, Bcl11b-dependent and non-DEGs bound by Bcl11b-dependent cofactor and H3K27Ac peaks (e), or newly generated cofactor and H3K27Ac peaks in Bcl11b-deficient cells (f) are shown. P values were determined by two-sided Fisher’s exact test. Calculations were based on ChIP-seq peaks scored as reproducible in two independent replicate samples.

Bcl11b alters cofactor binding at functional target loci

Bcl11b-dependent cofactor peaks were found around genes that Bcl11b repressed (Fig. 4a–d, magenta boxes)[7], both in primary DN3 cells and in Scid.adh.2c2 cells (Supplementary Fig. 6a-d). At Id2, Bcl11b-dependent cofactor peaks were seen not on the gene body itself but substantially upstream and downstream of the Id2 locus (Fig. 4a, Supplementary Fig. 6a), consistent with the extended regulatory system for this gene[22]. Similarly, at Zbtb16 and Tnni1 only one of several sites appeared to be Bcl11b-dependent, and at Cd163l1, different cofactors varied in Bcl11b-dependence at different sites (Fig. 4b–d, Supplementary Fig. 6b–d), suggesting that Bcl11b may interact separately with distinct complex subcomponents. Finally, selective Bcl11b-dependent cofactor recruitment was also seen at genes that were positively regulated by Bcl11b in pro-T cells, such as the Cd3gde cluster and Cd6 (Supplementary Fig. 6e,f). The bifunctional transcription factor Runx1 was frequently recruited to Bcl11b sites around all classes of targets. Thus, Bcl11b binding recruited different cofactors to specific subsets of its genomic sites including both positively and negatively regulated loci.

Statistical evidence implied that this corecruitment was functionally relevant. Unlike Bcl11b binding itself, Bcl11b-dependent cofactor recruitment sites were enriched at Bcl11b-regulated loci (defined in Supplementary Fig. 4a, Supplementary Table 2). Among the DEGs linked to direct Bcl11b binding, Bcl11b-dependent cofactor peaks and “new” cofactor peaks that appeared only when Bcl11b was deleted were highly overrepresented (Fig. 4e,f), as compared to genes that also had Bcl11b binding but did not change expression when Bcl11b was deleted (Fig. 4e,f, right; non-DEG). Bcl11b-dependent Runx1 recruitment was particularly enriched at Bcl11b-repressed target genes, (Fig. 4e, left), while Bcl11b-dependent Rest was particularly depleted. Mta2 accompanied Bcl11b binding at DEGs and non-DEGs alike, but Bcl11b-dependent Chd4, Ring1b, and LSD1 recruitments were also significantly enriched at Bcl11b repression targets (Fig. 4e, left). As expected for repression mediators, these sites had minimal association with the active histone mark H3K27Ac. At sites linked to Bcl11b-dependent DEGs (Fig. 4e, middle), there was also highly significant enrichment for Bcl11b-dependent Runx1, Chd4, LSD1 and Ring1b recruitment, but here with highly enriched H3K27Ac as well. Interestingly, Bcl11b-repressed DEGs were most specifically enriched for “new” cofactor peaks that appeared only when Bcl11b was absent (Chd4, Ring1b, LSD1 and Runx1 P<2.2E-16, Ring1b P<2.5E-9), as H3K27Ac marking of these genes also increased (Fig. 4f, left). Although “new” sites for Runx1 and LSD1 themselves rarely overlapped with strong Bcl11b occupancy in wildtype cells (Fig. 3c; Fig. 4a-b, green boxes), it appeared that Bcl11b’s presence inhibited their occupancy normally. Thus, Bcl11b-dependent cofactor localization, both recruitment and inhibition of recruitment, was a much stronger predictor of Bcl11b functionality at genomic sites than Bcl11b binding alone.

Gene repression is complex, and it is not known how many binding sites are usually needed for effective repression of a target by Bcl11b. However, we tested whether Bcl11b-dependent co-recruitment could identify sites which exert particularly strong effects in repression. We disrupted either of two sites flanking Id2 (Supplementary Fig. 7a, magenta) or one downstream of Tnni1 (Fig. 4c) in Scid.adh.2C2 cells and scored them for target gene deregulation. Using transduction of Cas9 plus sgRNA followed by cloning, we identified cells in which the target sites were completely disrupted as shown by genomic qPCR (Supplementary Fig. 7b,c, right). Loss of the Tnni1 candidate silencer site (Sil +14k) elevated Tnni1 expression at least 5-fold (P=9.52E-6) (Supplementary Fig. 7c). Deletion of one Id2 candidate site (Sil +40k), though not another (Sil −600k), consistently raised Id2 expression in the Scid.adh.2c2 cells above background (P=9.49E-4), despite the presence of numerous other Bcl11b sites around the locus (Supplementary Fig. 7a, blue, b). Thus, Bcl11b-dependent cofactor recruitment can help to identify functional repression sites in the genome.

Functional impact of cofactors in Bcl11b activities

Recruitment of the cofactors themselves contributed functionally to Bcl11b’s effects, as shown when we compared effects of acute disruption of Bcl11b with those of disrupting the genes encoding the cofactors. We generated sgRNAs against coding regions for Chd4, Mta1_2 (Mta1 and Mta2), Rest, Ring1a_b (Ring1a and Ring1b), LSD1, or Runx1, each with a Cyan Fluorescent Protein (CFP) reporter, and these were confirmed to eliminate target protein expression in Cas9-transduced Scid.adh.2c2 cells (Supplementary Fig. 8a)[46]. We then transduced them into Cas9-Bcl2 transgenic primary cells in parallel with sgRNA against Bcl11b alone (Supplementary Fig. 3a), to compare their effects directly during in vitro T cell development. At the time of transduction, after 7 d of OP9-DL1 coculture, most of the primary Cas9;Bcl2-transgenic cells were in DN2 stage (c-Kit+CD25+, Supplementary Fig. 8b). After seven more days, transduced (CFP+) control cells (Supplementary Fig. 8c) had progressed into DN3 stage (Fig. 5a), whereas Bcl11b sgRNA-transduced cells showed the typical c-Kithi+ DN2a-like phenotype. The effects on surface phenotype of cofactor deletion were milder than those of deletion of Bcl11b (Fig. 5a), although RNA transcript structures confirmed the biallelic deletions at the targeted sites in these loci (Supplementary Fig. 8d; arrowheads).

Figure 5:

Effect of cofactor deletion on the expression of Bcl11b target genes

(a), CRISPR/Cas9-mediated deletion of Bcl11b, Chd4, Mta1_2, Rest, Ring1a_b, LSD1 or Runx1 gene in primary DN cells. Flow cytometric analysis of sgRNA transduced BM-derived precursors from Cas9 mice after 14 days of OP9-DL1 co-culture in Lin-CD45+CFP+ cells are shown.

(b), Distinct repressor complexes coordinately control Bcl11b-mediated gene regulation. Heatmaps show hierarchical clustering analyses of the expression of the Bcl11b-repressed (left) and Bcl11b-dependent (middle) genes, which are directly bound by Bcl11b (Supplementary Fig. 4a), in Bcl11b- or cofactor-deficient DN cells. Expression in these samples of control genes, which are not bound by Bcl11b and do not change expression upon Bcl11b deletion, is also shown for comparison (right). Names are indicated for Bcl11b-regulated genes that also show significant change (FDR<0.1) in response to disruption of Mta1_2 or Runx1 alone.

(c, d), Deletion of cofactors has strong effect on gene expression for Bcl11b-regulated genes, but minimal effect for genes expressed but not regulated by Bcl11b in DN3 cells. The percentages of cofactor-dependent genes (|Log2FC|>1) among the Bcl11b-repressed (Fig. 5b, left) and control (Fig. 5b, right) genes (c), and Bcl11b-dependent (Fig. 5b, middle) and control (Fig. 5b, right) genes (d) are shown, tallied from Supplementary Table 6. P values are determined by two-sided Fisher’s exact test.

Data are based on reproducible ChIP-seq peaks in two replicate samples and two replicate RNA-seq results (b, c, d), or representative of three independent experiments (a).

Cofactor deletion specifically affected RNA expression, more frequently at Bcl11b-regulated genes that were directly bound by Bcl11b (Fig. 5b–d; Supplementary Table 6; |Log2FC|>1) than at background genes which were expressed independently of Bcl11b (|Log2FC|<0.5, RPKM>3 in sgControl) and lacked Bcl11b binding (Fig. 5b, right). Most Bcl11b-repressed genes that were directly bound by Bcl11b were also de-repressed upon deletion of at least one of the cofactors, showing distinct gene-specific patterns of cofactor response (Fig. 5b, left). Among Bcl11b-dependent targets, many were downregulated upon deletion of Mta1, Mta2 or Runx1 (Fig. 5b, middle). As summarized (Fig. 5c,d), deletion of Mta1, Mta2 or Runx1 caused the highest numbers of significant gene expression changes (Supplementary Table 6, p.adj<0.1), and were concordant with effects of Bcl11b deletion in 80–90% of significant Runx1-regulated genes, ~90% of significant Mta1,2-regulated genes. Thus, cofactor recruitment is functionally significant, but with target gene-specific functional requirements.

Cofactor recruitment by Bcl11b in TCR locus marking

Bcl11b-deficient mice can generate γδ lineage T cells but not αβ lineage T cells, associated with a failure of Vβ-DJβ rearrangement[1,3,6,9,10]. In BM-derived pro-T cells developing in vitro, Bcl11b and Runx1 co-occupied multiple sites across both the Tcrb complex and the Tcrg complex. However, these genomic regions differed markedly in their dependence for Runx1 binding on Bcl11b (Supplementary Fig 9). Whereas Runx1 binding across the Vβ coding segments was highly Bcl11b-dependent (Supplementary Fig. 9a), its binding across the whole Tcrg complex was largely Bcl11b-independent (Supplementary Fig. 9b). Importantly, the difference between these loci was not revealed by differential RNA expression, for the in vitro cultured pro-T cells at this early stage showed minimal Vβ transcription with or without Bcl11b (Supplementary Fig. 9a, RNA tracks). Instead, the difference in Bcl11b-dependent cofactor recruitment could indicate locus-specific roles in establishing permissive chromosome structure[18] which could underlie later Bcl11b impacts on DNA rearrangement.

Id2 and Zbtb16 in the Bcl11b gene regulatory network

The direct effects of Bcl11b binding just defined leave open the possibility that Bcl11b impacts on pro-T cell gene expression could include indirect effects as well. Genes encoding the transcription factors Id2 and PLZF, which promote development of innate lymphoid cells including NK cells[24,28,47,48], were always up-regulated in Bcl11b-deficient pro-T cells[6,7] (Table 1; Supplementary Tables 1–3). The E2A antagonist, Id2, was of particular interest. We previously reported that Bcl11b knockout effects in fetal liver-derived pro-T cells[7] showed a surprisingly large overlap with Tcf3 (encoding E2A) knockout effects[49]. Whereas the enrichment of bHLH motifs (Ptf1a or E2A) at Bcl11b sites for cofactor recruitment (Supplementary Fig. 5b) suggested possible complex formation, this was not supported by ChIP data: only 104 of 1,430 E2A published occupancy peaks in DN3 cells[50] overlapped with any of the ~26,000 Bcl11b peaks. This raised the possibility that repression of an intermediate regulator like Id2 could contribute to gene expression effects of Bcl11b.

To examine roles of Id2 and PLZF in Bcl11b-deficient cells, we carried out single and double deletion experiments. Cas9-Bcl2 transgenic BM precursors were co-transduced with sgRNAs targeting Bcl11b and/or Id2 or Zbtb16, singly or in combination (as in Supplementary Fig. 3a). Id2 and Zbtb16 showed increased expression in Bcl11b-deficient cells but not in Bcl11b-Id2 or Bcl11b-Zbtb16 double KO (DKO) cells, respectively. While the characteristic upregulation of c-Kit was comparable in Bcl11b KO cells and Bcl11b-Zbtb16 DKO cells, Bcl11b-Id2 DKO cells expressed slightly lower c-Kit (Fig. 6a), implying that Id2 may be involved in the abnormal elevation of c-Kit in Bcl11b-deficient cells. Interestingly, up-regulation of Zbtb16 in Bcl11b KO cells was also weakened in Bcl11b-Id2 DKO cells (Supplementary Fig. 10), and this was confirmed at the PLZF protein level (Fig. 6b). Thus, Id2 is involved in up-regulation of Zbtb16 expression in Bcl11b-deficient cells.

Figure 6:

Id2 is involved in establishment of phenotypes of Bcl11b-deficient cells

(a), Up-regulation of c-Kit expression in Bcl11b-deficient cells is partially dependent on Id2. Flow cytometric analysis of sgRNA transduced BM-derived precursors co-cultured with OP9-DL1 (Supplementary Fig. 3a) was performed. Representative c-Kit/CD25 profiles in Lin-CD45+CFP+ cells are shown with the percentages of cells in each quadrant. A representative histogram and a summary plot of the mean fluorescent intensity (MFI) of c-Kit expression of each group is also shown. *P< 0.05 (two-sided Student’s t-test).

(b), De-repression of Zbtb16 expression is partially dependent on Id2. PLZF protein expression was determined by intracellular staining. A representative histogram and a summary plot of the MFIs of PLZF expression is also shown. **P<0.01, *P<0.05 by two-sided Student’s t-test.

(c), Expression levels of a subset of Bcl11b-repressed and -dependent genes regulated by Id2 and Zbtb16. CD25+ sgRNA transduced BM-derived precursors in (a) were purified and subjected to RNA-seq analysis. Hierarchical clustering analyses of the expression of the Bcl11b-repressed (upper) and -dependent (lower) genes in Bcl11b, Id2, Zbtb16, Bcl11b and Id2 or Bcl11b and Zbtb16-deficient DN cells are shown.

(d), Id2 is involved in Bcl11b-mediated repression of ‘Cytokine-cytokine receptor interaction’ pathway. In KEGG pathway analysis, ‘Cytokine-cytokine receptor interaction’ (P value: 1.15E-05, n=18) and ‘Natural killer cell mediated cytotoxicity’ (P value: 3.09E-04, n=10) pathways were enriched among Bcl11b-repressed genes (n=410). Genes in ‘Cytokine-cytokine receptor interaction’ and ‘Natural killer cell mediated cytotoxicity’ pathways are shown. Dots represent the mRNA log2 FC values for sgControl vs. sgBcl11b (x-axis) and sgControl vs. sgBcl11b+sgId2 (left) or sgControl vs. sgBcl11b+sgZbtb16 (right) (y-axis).

(e), Id2 and Zbtb16 are involved in Bcl11b-mediated activation of ‘T cell receptor signaling’ pathway. ‘T cell receptor signaling’ pathway (P value; 5.32E-07, n=12) was most enriched among Bcl11b-dependent genes (n=384). Genes in ‘T cell receptor signaling’ pathway were shown. Dots represent the mRNA log2FC values for sgControl vs. sgBcl11b (x-axis) and sgControl vs. sgBcl11b+sgId2 (left) or Control vs. sgBcl11b+sgZbtb16 (right) (y-axis).

Data are representative of three independent experiments (a, b; three biological replicates with mean +s.d.) or are pooled from two experiments (c, d, e)

Most genes up-regulated by Bcl11b-deletion (Bcl11b-repressed genes) showed overall similar responses in Bcl11b KO, Bcl11b-Id2 DKO or Bcl11b-Zbtb16 DKO samples (Fig. 6c, Supplementary Table 7), indicating that most of these 410 Bcl11b repression target genes do not depend on Id2 or PLZF for their expression (Supplementary Table 7; DEGs with FDR<0.05 vs. control: 410 for Bcl11b KO; 299 for DKO)(Fig. 6c, top). In contrast, among 349 Bcl11b-dependent genes, the down-regulation of many T cell program genes was ameliorated if Bcl11b was deleted together with Zbtb16 or especially Id2 (Fig. 6c, bottom). Double deletion of Id2 with Bcl11b gave protection to well over half of the Bcl11b-dependent genes (Supplementary Table 7; DEGs with FDR<0.05 vs. control: 349 for Bcl11b KO; only 72 for DKO), suggesting that many of these genes are dependent on E proteins.

Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis showed that Bcl11b deletion up-regulated genes in ‘Cytokine-cytokine receptor interaction’ and ‘Natural killer cell mediated cytotoxicity’ pathways preferentially, while reducing expression of ‘T cell receptor signaling’ pathways. Although the ‘Natural killer cell mediated cytotoxicity’ genes were upregulated in Bcl11b KO cells with or without Id2 or Zbtb16, there was markedly less upregulation of the ‘Cytokine-cytokine receptor interaction’ genes in DKO samples, especially the Bcl11b-Id2 DKO, than in single Bcl11b KO samples, and less inhibition of ‘T cell receptor signaling’ genes (Fig. 6d,e). Thus, despite the weak overall effects of Id2 and Zbtb16 on gene expression profiles of Bcl11b-deficient cells, they had impacts on regulation of selective sets of genes, related to cytokine-cytokine receptor and TCR signaling pathways specifically.

Id2-Bcl11b double deletion did not always attenuate the effects of Bcl11b loss; for some Bcl11b repression targets, it amplified the effects of Bcl11b disruption (Supplementary Table 7). Certain genes highly enriched in TCRγδ IEL, including Heyl and Cited4, and the B cell and progenitor-cell-associated regulatory gene Bcl11a, were upregulated substantially more in Bcl11b-Id2 DKO than in Bcl11b single KO samples. These results imply that Bcl11a, Heyl and Cited4 depend on positive regulation by E proteins even though they are also a target of repression by Bcl11b. Thus, the upregulation of Id2 could sometimes mediate but in other cases mask the full spectrum of regulatory effects of Bcl11b.

Id2 and Zbtb16 promote different alternative programs

Bcl11b deletion upregulates genes associated with various alternative fates, and it has not been clear whether this response represents one coordinated program or several. Confirming previous reports[5,6,8], we repeatedly detected more Lin+ cells in Bcl11b-deficient cells than control cells, especially cells expressing NK1.1 and CD11c, even in T cell developmental conditions (Supplementary Fig. 11a,b, OP9-DL1)(cf. ref[6]). When Bcl11b-deficient cells were removed from Notch signals on day 10, to allow non-T cell fates to be expressed, the percentage of Lin+ cells rose over the next 4d with NK1.1 and CD11c expression most prominent (Supplementary Fig. 11a,b, OP9-Mig). Bcl11b-deficient pro-T cells express these markers less strongly than mature NK or dendritic cells, respectively (Supplementary Fig. 11c), but they indicate distinct altered regulatory states, and double deletion experiments showed that Id2 and PLZF play distinct roles in these states. Both the percentage and the cell number of Lin+ cells were significantly decreased in Bcl11b-Id2 DKO cells, indicating that they depend on this E protein antagonist. However, Lin+ cells were unexpectedly increased in Bcl11b-Zbtb16 DKO cells, both on OP9-DL1 and on OP9-Mig (note scale change: Fig. 7a; Supplementary Fig. 11d,e). Generation of NK1.1+ cells was attenuated in Bcl11b-Id2 DKO cells and reduced further in Bcl11b-Zbtb16 DKO cells. In contrast, while generation of CD11c+ cells was also reduced in Bcl11b-Id2 DKO, it was sharply enhanced in Bcl11b-Zbtb16 DKO cells (Fig. 7a, Supplementary Fig. 11d,e).

Figure 7:

Id2 and Zbtb16 play key roles in Bcl11b-mediated exclusion of alternative fates

(a), Id2 and Zbtb16 are involved in alternative fate exclusion by Bcl11b. The percentages of Lin+, NK1.1+ and CD11c+ in CFP+ sgRNA transduced cells (Supplementary Fig. 11d) are indicated. **P<0.01, *P<0.05 by two-sided Student’s t-test.

(b), Generation of alternative lineages in Bcl11b-deficient cells is cell-intrinsic. BM-derived precursors were cultured on OP9-DL1 for 7 days, then they were split to three aliquots. One aliquot was transduced with sgRNA against Bcl11b alone (hNGFR marker); another with the indicated combination of sgRNAs and CFP marker; and the third with pMxs-GFP empty vector, separately. Three days after infection, they were pooled as shown, transferred to OP9-Mig and cultured 4 more days. Representative profiles of Forward Scatter (FSC) vs. Lin, NK1.1 or CD11c in CFP+, hNGFR+ or GFP+ cells are shown with the percentages of cells in rectangles.

(c, d), Summary of important roles of Id2 and PLZF (encoded by Zbtb16) in establishment of phenotypes of Bcl11b-deficient pro-T cells. (c) Effects on alternative lineage markers. (d) Effects on specific genes. Bcl11b directly represses Id2, Zbtb16 and c-Kit expression, and Id2 is partially involved in Bcl11b-mediated down-regulation of Zbtb16 and c-Kit expression. Up-regulated Zbtb16 in Bcl11b-deficient cells supports generation of NK1.1+ cells and represses CD11c+ cells. On the other hand, Id2 positively contributes to generation of NK1.1+ and CD11c+ cells (c), activation of genes in ‘Cytokine-cytokine receptor interaction’ pathway and antagonism of E-protein-mediated activation of T cell genes (d). The rare Bcl11b repression of an E-protein dependent gene, Bcl11a, shows gene specificity of effects.

Data are individual values of three biological replicates with mean +s.d. (a), or representative of three independent experiments (b).

The effects of these single and double deletions were cell-intrinsic (Fig. 7b and Supplementary Fig. 12a,b). When we cocultured Bcl11b KO and Bcl11b-Id2 or Bcl11b-Zbtb16 dKO cells together with distinctly marked reference standard Bcl11b single KO cells and GFP+ empty-vector control cells in the same well (Supplementary Fig. 12a), the patterns of altered developmental phenotypes were determined only by the genetic perturbation history of the cells themselves (Fig. 7b, quantitated in Supplementary Fig. 12b).

Thus, PLZF not only supported generation of NK1.1+ cells but also restrained expression of CD11c in Bcl11b KO cells (Fig. 7a,c), forcing the Bcl11b KO phenotype normally closer to an NK-like profile. In turn, Id2 had stronger roles in other features of Bcl11b-deficient cells including DN2a-like high c-Kit expression, up-regulated Zbtb16 expression, increased expression of cytokine and chemokine receptors, decreased expression of T cell program genes and opening the way for generation of NK1.1+ and CD11c+ cells (Fig. 7c,d). The ability of Bcl11b to repress both Id2 and Zbtb16 is thus crucial for completing conventional T cell commitment.

DISCUSSION

The importance of Bcl11b for T cell lineage commitment has raised the question of how this factor works to promote and consolidate T-cell identity[5,6,8]. Unlike Pax5 and EBF1 in the B cell lineage[51-54], Bcl11b effects on T-cell gene expression appear more limited and more related to immune activation thresholds – “adaptive” vs. “innate-like” – than to T-cell identity per se[6-8,10,12,55]. A recent study[18] showed Bcl11b to bind to numerous sites throughout the active topological domains of the genome in early T cells, suggesting a possible global organizational role, but leaving open its gene-specific regulatory role. Here, we have used proteomics and genome-wide transcriptome and factor binding analysis to demonstrate molecular mechanisms through which Bcl11b binding to specific genomic sites controls target gene expression during T cell commitment. We show that Bcl11b can repress targets directly by nucleating complexes of corepressors on the DNA at specific sites where they do not otherwise assemble. Such sites of Bcl11b-dependent cofactor recruitment and redirection turn out to be a far better statistical discriminator of genes that Bcl11b actually controls than Bcl11b binding alone. At the same time, a substantial minority of Bcl11b effects are apparently indirect, mediated through its repression of the regulatory genes Id2 and Zbtb16. Acute double-deletion analysis using the Cas9 system has revealed that many effects of Bcl11b deletion are responses to the resultant increases in Id2 and/or PLZF. In particular, many T lineage-specific genes that apparently depend on Bcl11b for activation during commitment in fact require Bcl11b largely to suppress Id2. Thus, in addition to its direct genomic targets, Bcl11b is an indirectly acting but critical member of the E protein-Id gene regulatory network in pro-T cells during commitment.

The best available statistical criteria for enriching loci that Bcl11b regulates, positively or negatively, have turned out to be the presence of sites where assembly of chromatin modifying complexes depended on Bcl11b binding. Such complexes often included Chd4, Mta2, Ring1b, LSD1, and Runx1; future work should also test Brg1. Bcl11b-dependent nucleation of cofactor complexes was only seen at a small minority of Bcl11b sites and a minority of the sites for most of these cofactors. Deletion analysis of individual chromatin modification complex components showed that these components, stably expressed throughout commitment[56,57], were often important for Bcl11b effects, especially at loci that Bcl11b represses. Among factors redeployed by Bcl11b, Runx1 was most potent for both activation and repression of target genes. Notably, the components that depended most on Bcl11b for recruitment differed from one genomic site to another. Although much remains to be learned about the mechanistic rules for transcriptional repression, this suggests that Bcl11b can interact with a variety of subunit assemblies individually, not only with pre-formed NuRD or PRC1 complexes as a whole.

New cofactor peaks also appeared in Bcl11b-deleted cells at sites never bound by Bcl11b in pro-T cells, with a distinct motif enrichment signature, and these were strongly associated with abnormal activation of Bcl11b-repressed genes. Thus, in addition to recruiting repression complexes, Bcl11b may also repress target genes by preventing cofactors from assembling at other neighboring sites that could otherwise serve different, activating transcription factors. Gene regulation by redirection of limiting pools of cofactors, with or without direct DNA binding, is consistent with recent findings on PU.1-mediated gene regulation in early pro-T cells[46]. Therefore, gene regulation by partner factor redeployment could be a common mechanism for transcription factors that have crucial roles in cell fate decisions.

In summary, we have shown that the advent of Bcl11b expression during commitment re-centers T cell regulatory circuits in multiple ways. Bcl11b physically redirects the binding of Runx1 and chromatin modulating complexes across the genome in site-specific ways that are highly concentrated around Bcl11b-regulated target genes. These biochemical collaborations, especially with Runx1, are likely to be directly functional, especially for repression. At the same time, Bcl11b’s repression of PLZF and Id2 blocks implementation of at least two alternative programs, distinctly regulated by these factors. Finally, its control of Id2 gives it a position of indirect power in a pro-T cell network where the main direct effectors are E proteins and their own interaction partners. This network is particularly important for TCR gene expression and the stringent cell biology of β-selection, and may well explain the importance of Bcl11b for these crucial milestones.

METHODS

In addition to the complete description and explanation of the methods presented here, reagent lists and some general methods are also repeated, along with statistical checklists, in the Life Sciences Reporting Summary that accompanies this paper.

Mice

C57BL/6 (referred to as B6), B6.Cg-Tg(BCL2)25Wehi/J (Bcl2-tg)[58] and B6.Gt(ROSA)26Sortm1.1(CAG-cas9*,-EGFP)Fezh/J (Cas9)[59] mice were purchased from the Jackson Laboratory. Vav1-iCre mice[35] (B6N.Cg-Commd10Tg(Vav1-icre)A2Kio/J) were purchased from Jackson Laboratories and Lck-Cre (Lck proximal promoter) mice developed by C. Wilson’s group (B6.Cg-Tg(Lck-Cre)1Cwi N9)[36] were purchased from Taconic Laboratories. The Cre activity reporter allele ROSA26R-eYFP[60] was also used in our previous studies of Bcl11b gene function[6,7,61]. Except for Vav1-iCre, which was maintained in heterozygotes, the indicated transgenes were bred to homozygosity alone or in combinations on the B6 background. Bcl11bfl/fl;Rosa26-Cre-ERT2 mice[8] were derived from stock originally kindly provided by P. Liu (Wellcome Sanger Institute, Cambridge, UK) as previously described[7], and maintained as a separate line. All animals were bred and maintained in the California Institute of Technology Laboratory Animal Facility, under specific pathogen free conditions, and the protocol supporting animal breeding for this work was reviewed and approved by the Institute Animal Care and Use Committee of the California Institute of Technology.

Cells and cell culture

Thymuses from 4–6wk old Bcl11b+/+, Bcl11bfl/+, and Bcl11bfl/fl ROSA26R-YFP mice with Vav1-iCre or Lck-Cre were removed, and single-cell suspensions were made. Lineage-positive cells were depleted by staining with biotinylated antibodies to CD8α (53–6.7), TCRγδ (GL3), TCRβ (Η57597), Ter119 (Ter119), NK1.1 (PK136), Dx5, and CD11c (N418), CD11b (M1/70), after which the cells were incubated with streptavidin-coated magnetic beads and then passed through an LS magnetic column in accordance with the manufacturer’s instructions (Miltenyi Biotec). Eluted DN cells were stained with CD117 (c-Kit)-PE (2b8), CD25-APCe780 (PC61.5) and CD44-e450, and YFP+CD25+ and YFP–CD25+ cells were sorted by FACSAria (BD Bioscience). For flow cytometric analysis, eluted DN cells were stained as above for FACS purification and analyzed with FlowJo software (Tree Star).

Scid.adh.2c2 cells [38] were cultured in RPMI1640 with 10% fetal bovine serum (FBS, Sigma-Aldrich), sodium pyruvate (Gibco), non-essential amino acids (Gibco), Pen-Strep-Glutamine (Gibco) and 50 μM β-mercaptoethanol (Sigma-Aldrich).

For in vitro differentiation of pro-T cells, bone marrow hematopoietic progenitors were used for input. Bone marrow (BM) was removed from the femurs and tibiae of 2–3 month-old mice. Suspensions of BM cells were prepared and stained for lineage markers using biotin-conjugated lineage antibodies (CD11b, CD11c, Gr1, TER-119, NK1.1, CD19, CD3ε, B220), then incubated with streptavidin-coated magnetic beads (Miltenyi Biotec), and passed through a magnetic column (Miltenyi Biotec). Then, Lin–Sca1+c-Kit+ (LSK) cells were sorted on a FACSAria (BD Bioscience). LSK cells were cultured on OP9-DL1 monolayers using OP9 medium (α-MEM, 20% FBS, 50 μM β-mercaptoethanol, Pen-Step-Glutamine) supplemented with 10 ng/ml of IL-7 (Pepro Tech Inc.) and 10 ng/ml of Flt3L (Pepro Tech Inc.). On day 7, cultured cells were disaggregated, filtered through 40-μm nylon mesh, and re-cultured on new OP9-DL1 monolayers with medium containing 5 ng/ml of IL-7 and 5 ng/ml of Flt3L. In cultures that were continued for longer times, cells were passaged onto fresh OP9-DL1 monolayers at day 10 and maintained up to day 14 in 1 ng/ml each of IL-7 and Flt3L.

In samples that were tested for developmental lineage alterations, the final passage at day 10 was either onto fresh OP9-DL1 or onto OP9-control monolayers without the Notch ligand DL1 (OP9-Mig, referred to as OP9-control in ref. [39]).

Acute deletion of Bcl11b in T-cell development cultures

We used two methods to delete Bcl11b acutely in vitro, at a known timepoint and developmental stage, for the perturbation experiments in this study.

To generate sufficient numbers of control and Bcl11b-deleted cells for ChIP-seq analysis, we used acute deletion of floxed Bcl11b in cells with Cre-ERT2 activated by 4-OH tamoxifen. BM-derived precursor cells from control Bcl11b+/+Cre-ERT2 mice and from Bcl11bfl/flCre-ERT2 mice were cultured in parallel on OP9-DL1 stroma to day 7 as described above. Upon reculture, both controls and experimental samples were supplemented with 120 nM of 4-OH tamoxifen (4-OHT, Sigma-Aldrich). Thus, both control and experimental samples contained the Cre-ERT2 transgenes, were exposed to 4-OHT, and subject to Cre activity. Two days later 4-OHT was removed, and cells were cultured for an additional 5 days (to day 14 overall) on fresh OP9-DL1 with 1 ng/ml of IL-7, 1 ng/ml of Flt3L, as shown in Supplementary Fig. 4c. They were then harvested for further analysis and preparation of DNA for ChIP.

RNA expression responses to real-time perturbation were measured using Cas9 plus guide RNA for acute mutational disruption of Bcl11b, Runx1, Id2, Zbtb16, and/or genes encoding cofactors. This method was superior to the Cre-ERT2 method for RNA-seq comparisons, because all perturbations could be carried out on the identical genetic background and the potentially toxic effects of 4-OHT with Cre[62] could be eliminated. To generate input cells, Cas9 mice were first bred to Bcl2-tg mice to generate heterozygotes for both transgenes. Bone marrow cells from these Cas9;Bcl2-tg animals (called Cas9-Bcl2 mice) were then used to seed in vitro differentiation cultures as above. At day 7, the cells were transduced with retroviral vectors encoding a fluorescent reporter (CFP or hNGFR) and the indicated guide RNAs (sgRNAs) as detailed below, and then returned to OP9-DL1 culture. Cells were analyzed after another 7 days of culture (to day 14 overall), or as indicated in specific experiments.

Flow cytometry analysis

For staining of sgRNA-introduced BM cells, surface antibodies against CD45, c-Kit, CD25 and a biotin-conjugated lineage cocktail (CD8α, CD11b, CD11c, Gr1, TER-119, NK1.1, CD19, TCRβ, TCRγδ) were used for staining. Prior to cell surface staining cells were treated with 2.4G2 cell supernatant. Where indicated, intracellular staining using the BD cytofix/cytoperm Kit (BD Bioscience) was carried out with anti-PLZF AlexaFluor 647 (BD Bioscience). All of the cells were analyzed using a flow cytometer, MacsQuant 10 (Miltenyi) with FlowJo software (Tree Star).

Cloning

Myc-Flag-tagged cDNA for Bcl11b was inserted into a multi-cloning site of the pMxs-IRES-GFP vector. The Cas9-GFP expression vector (pQCXIN-EF1a-mNeonGreen-P2A-Cas9) and sgRNA-CFP expression vector (E42-dTet-CFP, in which mTurquoise is the Cyan Fluorescent Protein reporter) were described previously[46]. 19-mer sgRNAs were designed using the CHOPCHOP web tool (https://chopchop.rc.fas.harvard.edu/) and inserted into the empty sgRNA-expression vector by PCR-based insertion (sgRNA sequences provided in Supplementary Table 8). Three sgRNA-expression vectors were generated for one gene, and pooled retroviral plasmids were used to make retroviral supernatant.

Retroviral infection

The methods used to generate the virus supernatant and for infection were described previously[63]. Cultured Lin– BM cells for 7 days were disaggregated, filtered through 40-μm nylon mesh, transferred onto RetroNectin-coated virus bound plates, and cultured with OP9 medium supplemented with 5 ng/ml of IL-7, 5 ng/ml of Flt3L and 10 ng/ml of SCF. Infected cells were cultured for an additional 7 days on OP9-DL1 and subjected to further analysis. For sorting, cells were stained with CD45, CD25, and a biotin-conjugated lineage cocktail (CD8α, CD11b, CD11c, Gr1, TER-119, NK1.1, CD19, TCRβ, TCRγδ), and were sorted for CD25+ infected cells (Lin–CD45+CD25+CFP+).

Two-step affinity purification of Bcl11b complexes from the DN3-like cell line Scid.adh.2c2

Scid.adh.2c2 cells[38,39] were infected with either mock control (pMxs-IRES-GFP) or Myc-Flag-Bcl11b-containing retrovirus. Three days after infection, Myc-Flag-tagged Bcl11b-infected Scid.adh.2c2 cells were solubilized with the following protease inhibitor-containing IP buffer: 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 10% glycerol, 0.1% Tween, 1 mM EDTA, 10 mM NaF, 1 mM DTT and a protease inhibitor cocktail (Roche Applied Science), and lysed on ice for 30 min with gentle shaking and sonicated on a Misonix S-4000 sonicator (Qsonica) for 3 cycles, amplitude 20 for 30 sec. followed by 30 sec. rest. We did not add benzonase or ethidium bromide to exclude DNA- or RNA-mediated interactions, because we were interested in the functionally relevant complexes that Bcl11b forms as it is working on the DNA. The insoluble materials were removed by centrifugation and immunoprecipitation with anti-Flag M2 agarose (Sigma-Aldrich) was performed overnight at 4 oC. Immune complexes were eluted from the agarose by 3xFlag peptide (Sigma-Aldrich), and the eluted Bcl11b complexes were subjected 2nd immunoprecipitation with anti-Myc gel (MBL). Immune complexes were eluted from the gel with Myc peptide (MBL) and separated by SDS-PAGE. The bands were excised from the gel and subjected to a mass spectrometric analysis to identify corresponding proteins. The gel pieces were washed twice with 100 mM bicarbonate in acetonitrile and the proteins were digested with trypsin. After adding 0.1% formic acid to the supernatant, the peptides were analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) with an Advance UHPLC (Bruker) and an Orbitrap Velos Pro Mass Spectrometer (Thermo Fisher Scientific). The resulting MS/MS data set was analyzed using the Mascot software program (Matrix Science).

Gene Ontology and KEGG pathway analysis

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway annotation was performed using the DAVID analysis tool (https://david.ncifcrf.gov/).

Immunoprecipitation and immunoblotting

Protein extracts from Myc-Flag-tagged Bcl11b-infected Scid.adh.2c2 cells were subjected to immunoprecipitation as described previously[43]. Nuclear extracts were prepared using NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Scientific). The antibodies used for the immunoblot analyses were anti-Chd4 (A301–081A, Bethyl), anti-Mta2 (sc-9447, Santa Cruz Biotechnology, Inc.), anti-HDAC2 (ab12169, Abcam), anti-Rest (12C11–1B11, Caltech Protein Expression Center), anti-Ring1b (A302–869A, Bethyl), anti-LSD1 (ab17721, Abcam), anti-Runx1 (ab23980, Abcam), anti-Bcl11b (ab18465, Abcam), anti-Lamin B (sc-6217, Santa Cruz Biotechnology, Inc.), and anti-Myc (My3, MBL).

Chromatin Immunoprecipitation (ChIP) and ChIP-sequencing

1 × 107 BM-derived DN3 cells were fixed with 1% formaldehyde in α-MEM for 10 min (H3K27Ac), or with 1 mg/ml DSG (Thermo Scientific) in PBS for 30 min at ~22°C followed by an additional 10 min with addition of formaldehyde up to 1% (Bcl11b, Chd4, Mta2, HDAC2, Rest, Ring1b, LSD1 and Runx1). The reaction was quenched by addition of 1/10 volume of 0.125 M glycine and the cells were washed with HBSS (Gibco). Pelleted nuclei were dissolved in lysis buffer (0.5% SDS, 10 mM EDTA, 0.5 mM EGTA, 50 mM Tris-HCl (pH 8) and PIC) and sonicated on a Bioruptor (Diagenode) for 18 cycles of 30sec sonication followed by 30sec rest, with max power. Six μg per 107 cells of anti-Bcl11b Abs (a mixture of A300–383A (Bethyl), A300–385A (Bethyl), ab18465 (Abcam) and 12120 (CST)), or anti-Chd4 Ab (A301–081A), anti-Mta2 Ab (sc-9447), anti-HDAC2 Ab (ab12169), anti-Rest Ab (12C11–1B11), anti-Ring1b Ab (A302–869A), anti-LSD1 Ab (ab17721), anti-Runx1 Ab (ab23980), or anti-H3K27Ac Ab (ab4729) were each separately pre-bound to Dynabeads anti-Rabbit, Dynabeads anti-Mouse or Dynabeads Protein A/G (Invitrogen) and then added individually to the diluted chromatin complexes in parallel aliquots. The samples were incubated overnight at 4 °C, then washed and eluted for 6 h at 65 °C in ChIP elution buffer (20 mM Tris-HCl, pH 7.5, 5 mM EDTA 50 mM NaCl, 1% SDS, and 50 μg/ml proteinase K). Precipitated chromatin fragments were cleaned up using Zymo ChIP DNA Clean & Concentrator. ChIP-seq libraries were constructed using NEBNext ChIP-Seq Library Preparation Kit (E6240, NEB) and sequenced on Illumina HiSeq2500 in single read mode with the read length of 50 nt. Analysis pipelines used are described below under ChIP-seq Analysis and RNA-seq Analysis.

mRNA-preparation and RNA-sequencing

Total RNA was isolated from 3 × 105 of cultured cells and 2–5 × 104 of ex vivo cells using RNeasy Micro Kit (Qiagen). Libraries were constructed using NEBNext Ultra RNA Library Prep Kit for Illumina (E7530, NEB) from ~1 μg of total RNA following manufacturer’s instructions. Libraries were sequenced on Illumina HiSeq2500 in single read mode with the read length of 50 nt. Base calls were performed with RTA 1.13.48.0 followed by conversion to FASTQ with bcl2fastq 1.8.4 and produced approximately 30 million reads per sample.

ChIP-seq analysis

Base calls were performed with RTA 1.13.48.0 followed by conversion to FASTQ with bcl2fastq 1.8.4 and produced approximately 30 million reads per sample. ChIP-seq data were mapped to the mouse genome build NCBI37/mm9 using Bowtie (v1.1.1; http://bowtie-bio.sourceforge.net/index.shtml) with “-v 3 -k 11 -m 10 -t --best –strata” settings and HOMER tagdirectories were created with makeTagDirectory and visualized in the UCSC-genome browser (http://genome.ucsc.edu) [64]. The NCBI37/mm9 assembly was chosen for ChIP-seq sample mapping in this study to ease comparisons with numerous previous data tracks from our lab and others. ChIP peaks were identified with findPeaks.pl against a matched control sample using the settings “-P.1 -LP.1 -poisson.1 -style factor”. The identified peaks were annotated to genes with the annotatePeaks.pl command against the mm9 genomic build in the HOMER package. Peak calls were always based on data from at least two independent biological replicates. Peak reproducibility was determined by a HOMER adaptation of the IDR (Irreproducibility Discovery Rate) package according to ENCODE guidelines (https://sites.google.com/site/anshulkundaje/projects/idr). Only reproducible high quality peaks, with a normalized peak score ≥ 15, were considered for further analysis. Motif enrichment analysis was performed with the findMotifsGenome.pl command in the HOMER package using a 200-bp window.

RNA-seq analysis

RNA-sequenced reads were mapped onto the mouse genome build NCBI37/mm9 with STAR (v2.4.0) [65] and post-processed with RSEM (v1.2.25; http://deweylab.github.io/RSEM/) [66] according to the settings in the ENCODE long-rna-seq-pipeline (https://github.com/ENCODE-DCC/long-rna-seq-pipeline/blob/master/DAC/STAR_RSEM.sh) with the minor modifications that settings “--output-genome-bam --sampling-for-bam” was added to rsem-calculate-expression. STAR and RSEM reference libraries were created from genome build NCBI37/mm9 together with the Ensembl gene model file Mus_musculus.NCBIM37.66.gtf. The resulting bam-files were used to create HOMER [67] tag directories (makeTagDirectory with –keepAll setting). For analysis of statistical significance among differentially expressed genes the raw gene counts were derived from each tag directory with analyzeRepeats.pl with the –noadj -condenseGenes options followed by the getDiffExpression.pl command using EdgeR (v3.6.8; http://bioconductor.org/packages/release/bioc/html/edgeR.html) [68]. For data visualization, rpkm normalized reads were derived using the analyzeRepeats.pl command with the options –count exons –condenseGenes –rpkm followed by log transformation. The normalized datasets were hierarchically clustered with “average” linkage and visualized in MatLab (clustergram).

UCSC Genome Browser bigwig visualization

BigWigs were generated from the aligned SAM or BED-file formats using Samtools [69], Bedtools [70] and the UCSC genomeCoverageBed and bedGraphToBigWig and normalized to 1 million reads. For visualization of RNA-seq tracks, bamToBed and genomeCoverageBed were used with the “-split” setting enabled. BigWig files were up-loaded to the UCSC-genome browser (http://genome.ucsc.edu) [64] for visualization.

Position-weight matrix (PWM) scanning of Bcl11b associated peak lists

Liu and coworkers[37] recently reported new sequence motifs to be specific Bcl11b recognition targets, based on in vitro protein binding microarray reactivities of purified Bcl11b full length (“XL”), or the middle two zinc fingers of Bcl11b (Bcl11b_23), or the C-terminal three zinc fingers of Bcl11b (Bcl11b_456). To use HOMER to quantify the representation of these Bcl11b motifs in Bcl11b or Bcl11b co-factor ChIP peaks, 23-mer enriched protein binding microarray-defined probability matrices (Bcl11b_23_rep2, Bcl11b _456_rep1, Bcl11b _XL) were downloaded from ref. [37]. From these 23-mer matrices, a Homer compatible 12-mer PWM was then created from the 7th to 18th positions in each of the 23-mer matrices with the log-odds detection threshold manually set to 5, for relatively inclusive scoring. To detect presence of these PWMs in the peak files, PWM searches were carried out with annotatePeaks.pl with the options –m {motif files}.

QUANTIFICATION AND STATISTICAL ANALYSIS

Differentially Expressed Genes (DEGs) were defined using EdgeR, typically with FDR < 0.05, |log2FC|>1, and RPKM>1 except where otherwise indicated, based on measurements from at least two biologically independent replicates for each sample type. The statistical significance of differences between datasets was determined by two-sided Student’s t test, Pearson’s correlation coefficient or two-sided Fisher’s exact test using Excel or the R package. Statistical details of experiments can be found in the figure legends. The statistical methods and methods for ensuring reproducibility are also reported in the Life Sciences Reporting Summary for this paper.