TCF1 dosage determines cell fate during T cell development

Abstract

Loss-of-function studies have shown that transcription factor T cell factor-1 (TCF1), encoded by the Tcf7 gene, is essential for T cell development in the thymus. We discovered that the Tcf7 expression level is regulated by E box DNA binding proteins, independent of Notch, and regulates αβ and γδ T cell development. Systematic interrogation of the five E protein binding elements (EPE1–5) in the Tcf7 enhancer region showed lineage-specific utilization. Specifically, loss-of-function analysis revealed that only EPE3 plays a critical role in supporting αβ T cell development, while EPE1, 3, and 5 regulate the γδ T cell maturation and functional cell fate decision. The importance of EPE3 in supporting both lineages may stem from its unique capacity to interact with the Tcf7 transcriptional start site. Together, these studies demonstrate that the precise dosage of TCF1 expression mediated by distinct EPEs generates a balanced output of T cells from the thymus.

E protein–dependent TCF1 expression controls αβ and γδ lineage commitment and stage-specific developmental programs.

INTRODUCTION

T cell factor-1 (TCF1), encoded by Tcf7 gene, plays an essential role in the development of both conventional and innate T cells as demonstrated by germline deletion of the Tcf7 gene (1–4). TCF1 is sometimes referred to as a T cell lineage defining transcription factor as it has been shown to determine the epigenetic identity of T cell development (5). Notch signals in multipotent precursors initiate transcription of the Tcf7 gene in CD44⁺CD25⁻ DN1 cells (6, 7) and induce expression of Bcl11b leading to T lineage commitment in CD44⁻CD25⁺ DN2 cells (8, 9). In CD4⁺CD8⁺ double-positive (DP) thymocytes, TCF1 functions in cooperation with E proteins to regulate gene expression (10). Subsequently, together with other transcription factors, it controls expression of Zbtb7b to promote the CD4 fate and along with Runx3 to silence CD4 in CD8 SP thymocytes (11). TCF1 also promotes the persistence of DP thymocytes, thereby facilitating rearrangement of the distal Vα14-Jα18 T cell receptor α (TCRα) chain that is required to support commitment to the natural killer (NK)T lineage (12). In γδ T cells, TCF1 antagonizes the function of the fate specifying transcription factor RORγt, which promotes adoption of the innate-like interleukin-17 (IL-17) producing γδ T cell fate (13, 14). In the development of regulatory T cells (T_regs), reduced expression of TCF1 in Tcf7^+/− mice permits recruitment of lower TCR affinity T cells, thereby expanding the T_reg lineage (15). Together, these studies demonstrate that TCF1 plays distinct roles in the development of different T lineages in the thymus (16, 17). However, the transcriptional mechanisms that control expression of TCF1 to perform these distinct roles in different precursor thymocyte populations are poorly understood.

E proteins, E2A and HeLa E box binding factor (HEB), encoded by the Tcf3 and Tcf12 genes, are essential for T cell development as demonstrated by loss-of-function studies (18, 19). E2A deficiency attenuates T cell development at early stages, resulting in generation of few thymocytes that rapidly progress to thymic lymphomas (18). Loss of HEB function also leads to reduced thymic cellularity due to a specific block at the CD4⁻CD8⁻ [double-negative (DN)] to CD4⁺CD8⁺ (DP) transition (19). E proteins and Notch have been proposed to regulate the expression of genes that coordinate commitment to the T lineage (6, 7, 20–22). Notch activates Tcf7 gene expression (6, 7); however, whether E proteins directly control expression of Tcf7 has remained unclear. We previously reported that ablation of an E protein binding element (EPE), EPE1, located in the first intron of the Tcf7 gene, reduces TCF1 expression in γδ-precursor cells, enhances γδ T cell generation and promotes adoption of the IL-17 lineage fate, without affecting αβ T cell development (13). However, the role of E proteins in regulating TCF1 expression during αβ T cell development remains undefined.

Here, we report that a systematic analysis of E protein (E2A and HEB) binding to the Tcf7 locus revealed four additional EPEs in the Tcf7 enhancer region. Systematic analysis revealed that CRISPR-mediated ablation of EPE1, 3, and 5 reduced TCF1 expression to differing degrees but EPE2 and 4 showed no reduction in TCF1 expression and no effect on T cell development. Each EPE contributed distinctly toward Tcf7 gene regulation in γδ and conventional αβ T cell precursors; ablation of EPE1 (13), EPE3, and EPE5 (ΔEPE1, 3, and 5 mice) reduced Tcf7 in γδ T cells, resulting in accentuated development and adoption of the IL-17 producing effector fate. Deletion of 207–base pair (bp) encompassing EPE3 (ΔEPE3 mice) was unique in its impact on αβ T cell development. We provide mechanistic insight into EPE3 function by demonstrating that EPE3 is epigenetically marked with H3K27 acetylation (H3K27ac) and interacts with the Tcf7 promoter. Single-cell RNA sequencing (sc-RNAseq) analysis of DN progenitors from ΔEPE3 mice reveals that transcriptomic disruption begins in DN2 (CD44⁺CD25⁺) thymocytes and persists through developmental impairment at the DN3 (CD44⁻CD25⁺) and DN4 (CD44⁻CD25⁻) stages. ΔEPE3 DN4 cells show severe defects in transitioning to the DP stage, which markedly reduced thymic cellularity. We show that the EPE3-adjacent Notch binding site (NBS) on the ΔEPE allele is preserved and bound by Notch in ΔEPE3 thymocytes; however, ectopic expression of Notch fails to restore reduced TCF1 expression in ΔEPE3 thymocytes to wild type (WT) levels. Together, these data demonstrate that while Notch signaling initiates TCF1 expression, thereafter, the expression level of TCF1 is controlled by E proteins binding to EPEs, independently of Notch. In conclusion, the level of TCF1 expression is set by E protein binding to EPE1, 3, and 5, which leads to distinct levels of TCF1 expression determined by pre-TCR and other signals, ultimately setting the precise threshold of TCF1 expression in each precursor population as required for optimal development of γδ T and conventional αβ T cells.

RESULTS

E proteins regulate the level of Tcf7 gene expression to determine lineage decisions in the thymus

TCF1 expression varies in subsets of developing thymocytes (Fig. 1A). To systematically evaluate the role of E proteins in influencing the levels of TCF1 expression, we conducted chromatin immunoprecipitation sequencing (ChIP-seq) analysis using anti-E2A and anti-HEB antibodies (fig. S1A). We identified five EPEs that define the E protein regulome in the Tcf7 regulatory region (fig. S1, A and B). Each EPE contains two to four putative E protein binding sites (fig. S1C). By deleting each EPE using CRISPR technology, we generated five strains of ΔEPE mice (fig. S1C). TCF1 expression was reduced in total thymocytes from ΔEPE3 and, to a lesser extent, in thymocytes from ΔEPE5 mice, but not in total thymocytes from ΔEPE1 (13), ΔEPE2, or ΔEPE4 mice (Fig. 1B and fig. S2A). We conclude that deletion of EPE3 has the largest effect on TCF1 expression.

Fig. 1. E protein–dependent modest reduction in TCF1 expression markedly attenuates αβ T cell development.

(A) Intracellular flow cytometry analysis of expression of TCF1 in lineage⁻ DN1 (CD44⁺CD25⁻), DN2 (CD44⁺CD25⁺), DN3 (CD44⁻CD25⁺), DN4 (CD44⁻CD25⁻) cells, and DP (CD4⁺CD8⁺) thymocytes from WT and Tcf7^−/− mice. Graph represents TCF1 MFI in the indicated population in WT (n = 4) and Tcf7^−/− (n = 3) mice. (B) Graph represents fold change in MFI of TCF1 in total thymocytes from ΔEPE1, ΔEPE2, ΔEPE3, ΔEPE4, and ΔEPE5 mice. (C) Graph represents fold change in MFI of TCF1 in the indicated ΔEPE3 thymic subset relative to WT (n = 4) and ΔEPE3 (n = 8) mice. (D) Flow cytometry analysis of surface expression of CD4, and CD8 in total thymocytes. Graph represents the absolute number of total thymocytes in WT (n = 6), Tcf7^−/− (n = 3), and ΔEPE3 (n = 7) mice; and absolute number of DN (CD4⁻CD8⁻), DP (CD4⁺CD8⁺), single positive (CD4⁺CD8⁻ and CD4⁻CD8⁺) cells in WT (n = 11) and ΔEPE3 (n = 5) mice. (E) Flow cytometry analysis of surface expression of CD44, and CD25 in lineage negative thymocytes of WT, and ΔEPE3 mice. Graphical representation of the frequencies and number of DN2–4 cells in thymi of WT (n = 6), and ΔEPE3 mice (n = 12). Results are representative of three or more experiments performed and P values relative to WT are indicated.

We next wished to assess the impact of EPE ablation on TCF1 expression in subsets of developing αβ thymocytes. TCF1 expression was variably reduced in all developing αβ T cell subsets in ΔEPE3 mice (Fig. 1C) and selectively in DP cells in ΔEPE5 mice (fig. S2, B and C). The relatively modest reduction in TCF1 expression in αβ progenitors nevertheless had a profound impact on αβ T cell development in ΔEPE3 mice, reducing thymic cellularity fourfold, primarily through a marked reduction in the number of DP thymocytes (Fig. 1D). In ΔEPE1, ΔEPE2, ΔEPE4, or ΔEPE5 mice, thymic cellularity and αβ development were not affected (fig. S2B). ΔEPE5 DP thymocytes with modestly reduced TCF1 expression generated marginally increased numbers of SP thymocytes (fig. S2C) with WT levels of TCF1 expression (fig. S2C). We surmise that EPE3 and EPE5 contribute differently to modulation of TCF1 levels required for optimal thymocyte differentiation. Notably, the level of expression of TCF1 in ΔEPE3 thymocyte subsets was not reduced to the same extent in all the subsets but was distinct, with DN, CD4SP, and CD8SP subsets exhibiting the greatest reduction (fig. S2C). We also note that the number CD4SP and CD8SP thymocytes in ΔEPE3 mice was reduced in proportion to the decrease in DP precursors indicating that DP to SP transition was not impaired (Fig. 1D). Numbers of DN2 and DN3 subsets were unaffected in ΔEPE3 thymus, whereas the number of DN4 cells showed a modest decrease (Fig. 1E), indicating that the profound impairment of αβ T cell development in ΔEPE3 mice occurred during the transition of DN4 cells to the DP stage. Similar observations were made in three additional independently derived ΔEPE3 founder lines (fig. S3). These data demonstrate that modest reduction in TCF1 expression in progenitor cells attenuates effective DN4 to DP transition and results in the generation of mature CD4⁺ and CD8⁺ thymocytes with low levels of TCF1.

We previously reported that deletion of EPE1 in the first intron of the Tcf7 gene reduced TCF1 expression and promoted adoption of the γδ fate, without affecting αβ T cell development (13). Because strong TCR signals from the γδ TCR have been shown to reduce the function of E proteins via induction of Id proteins (13), we hypothesized that the reduction in TCF1 expression by genetic ablation of EPEs would mimic the strong γδTCR signals that promote γδ lineage commitment and maturation (13). Gating on thymic γδ cells showed reduced expression of TCF1 in ΔEPE1, ΔEPE3 and ΔEPE5, but not in ΔEPE2 or ΔEPE4 γδ lineage thymocytes (Fig. 2A and fig. S4A). Moreover, we found that proportions of γδ T cells were increased in ΔEPE3 and ΔEPE5 thymi, with ΔEPE5 thymi exhibiting an increase in the absolute number of γδ T cells (Fig. 2, B and C). To evaluate the intermediates in γδ T cell development, we evaluated changes in CD73 (comm marks γδ lineage commitment) and CD24 (mature marks maturation) expression (Fig. 2D), which mark development (23, 24). We found that the number of CD73⁺ γδ lineage committed cells and CD24^lo mature γδ cells were increased in both the ΔEPE5 and ΔEPE3 thymi (Fig. 2D), suggesting that like EPE1 (13), both EPE3 and EPE5 play a role in regulating γδ T cell maturation. No alterations in γδ T cell maturation were observed in ΔEPE2 or ΔEPE4 mice (fig. S4B). Reduced TCF1 expression in γδ thymocytes and enhanced γδ T cell maturation were also noted in three independently derived lines of ΔEPE3 and ΔEPE5 mice (fig. S4, C-H). Last, γδ T cells in ΔEPE3 and ΔEPE5 mice also preferentially adopted the IL-17 producing cell fate (Fig. 2E). Together, these data indicate that ablation of EPE1, 3, and 5 promotes γδ T cell maturation, suggesting that they are critical regulatory nodes that are responsive to strong γδTCR signals that induce Id3, inhibit E protein binding, and regulate commitment to αβ and γδ lineages in the thymus.

Fig. 2. Reduced TCF1 expression in progenitor thymocytes promotes development of γδ T cells and adoption of IL-17 producing cell fate.

(A) Graph represents the fold change in the TCF1 MFI in Tcf7^−/− (n = 4), ΔEPE1 (n = 8), ΔEPE2 (n = 6), ΔEPE3 (n = 5), ΔEPE4 (n = 5), and ΔEPE5 (n = 10) relative to WT (n = 9) thymocytes. (B) Flow cytometric analysis of γδ T cells for the surface expression of TCRδ in the thymi of WT, ΔEPE3, and ΔEPE5 mice. (C) Graph represents the frequency (left) and absolute number (right) of γδ T cells in the thymi of WT (n = 13), ΔEPE3 (n = 11), and ΔEPE5 (n = 10) mice. (D) Flow cytometric analysis of surface expression of CD24 and CD73 on γδ T cells with developmental intermediates (precommitment, CD73⁻CD24⁺; committed, CD73⁺CD24⁺; and mature, CD73⁺CD24⁻) and graphical representation of frequency (left) and absolute number (right). WT (n = 13), ΔEPE3 mice (n = 10), and ΔEPE5 mice (n = 7). (E) Flow cytometric analysis of IL-17 producing γδ T cells from WT, ΔEPE3, and ΔEPE5 mice, following activation with phorbol 12-myristate 13-acetate and ionomycin for 6 hours. Graphical representation of the frequency and absolute number of IL-17 producing γδ T cells in the thymi of WT (n = 8), ΔEPE3 mice (n = 4), and ΔEPE5 mice (n = 3). Results are representative of three or more experiments performed and P values relative to WT are indicated as noted in methods.

Reduced expression of TCF1 in ΔEPE3 DN thymocytes impedes transition to the DP stage

The reduced expression of TCF1 in DN thymocytes (Fig. 1C) and the reduced number of DN4 thymocytes in ΔEPE3 mice (Fig. 1E), suggested that the pre-TCR–induced transition from DN to DP was impaired. To test this possibility, we cultured purified DN4 thymocytes from WT and ΔEPE3 mice and assayed their transition in vitro to the DP stage (Fig. 3A) (25). Purified ΔEPE3 DN4 thymocytes cultured in vitro for 24 hours resulted in fewer DP thymocytes compared to WT cells, with a greater proportion remaining as DN thymocytes (Fig. 3B). This observation suggests that the capacity of pre-TCR signals to promote differentiation to DP thymocytes is impaired in ΔEPE3 DN4 cells, consistent with their reduced expression of TCF1 (fig. S5A). Maturation of DN cells to the DP stage can also be kinetically studied by culture of DN thymocytes on monolayers of OP9-DL1 cells (26). Coculture of ΔEPE3-DN thymocytes on OP9-DL1 monolayers showed that their development to the DP stage was impaired (Fig. 3, C to E). The impairment of development of ΔEPE3-DN thymocytes to the DP stage was associated with decreased proliferation, increased apoptosis, and lineage diversion to the γδ fate (fig. S5, B to D). To confirm that impaired development of ΔEPE3 DN thymocytes to the DP stage was primarily due to reduced TCF1 expression, we ectopically expressed TCF1 in WT and ΔEPE3 DN thymocytes and assessed their developmental potential on OP9-DL1 monolayers (Fig. 3F). WT DN cells progressed to the DP stage over 2 days of culture and were not affected by ectopic expression of TCF1 (Fig. 3, G and H). By contrast, ectopic expression of TCF1 significantly rescued the developmental defect of ΔEPE3 DN thymocytes (Fig. 3, G and H), demonstrating that the predominant cause of developmental arrest for ΔEPE3 DN thymocytes was reduced expression of TCF1. Collectively, these data show that relatively modest changes in the levels of TCF1 expression in DN thymocytes profoundly affect αβ lineage development to the DP stage.

Fig. 3. Reduced expression of TCF1 in ΔEPE3 DN thymocytes impedes developmental transition to the DP stage.

(A) Schematic representation of the suspension culture experimental design. (B) The frequency of DN (top) and DP thymocytes (bottom) was determined by flow cytometry at the indicated times of culture and mean ± SD of triplicate measurements was depicted graphically. (C) Schematic representation of OP9-DL1 culture experiment. (D and E) Development of DN progenitors from WT and ΔEPE3 to the DP stage on OP9-DL1 monolayers was assessed by flow cytometry at the indicated times of culture. Representative dot pots of thymic subsets defined by CD4 and CD8 are depicted in (D) and graphical representation of the absolute numbers of DP thymocytes (E) on the indicated days of culture. (F) Schematic of the experiment to evaluate the effect of ectopic expression of TCF1 on rescue of the developmental defect in ΔEPE3 progenitor cells. (G) DN thymocytes were transduced with empty vector (pMiG) or TCF1 (pMiG-Tcf7) and cultured in triplicate for 2 days on OP9-DL1 monolayers, following which the fraction of DP thymocytes among transduced cells was determined by electronic gating on GFP. Representative flow profiles of CD4 and CD8 staining are depicted. (H) Percent of DP thymocytes in each sample is depicted after 1 and 2 days of culture. Results above are representative of at least three experiments performed; P values are indicated.

sc-RNAseq analysis of ΔEPE3 DN thymocytes shows a markedly altered transcriptome starting at the DN2 stage

To gain additional insight into mechanisms by which EPE3 deletion impaired the DN to DP transition, we performed sc-RNAseq using DN cells isolated from WT and ΔEPE3 mice (fig. S6). Unsupervised clustering based on differential gene expression resolved DN cells into eight clusters with major DN subsets being identifiable in these clusters (Fig. 4, A and B). Tcf7 mRNA was reduced in all the clusters, including the heterogeneous DN1 subset (fig. S7A). In addition, DN3 cells could be further divided into subsets that were enriched in WT (DN3-WT) or enriched in ΔEPE3 (DN3-ΔEPE3 bias) thymocytes. We observed that proportions of cells in the DN2, DN3-ΔEPE3–biased, and DN4 clusters were increased in ΔEPE3 mice compared to WT (Fig. 4C), with concomitant changes in gene expression (Fig. 4D). These data indicate that reduced TCF1 expression caused by EPE3 deletion alters the transcriptomes of thymic progenitors beginning in DN2 cells and continuing to the DN4 stage.

Fig. 4. sc-RNAseq analysis of ΔEPE3 DN thymocytes shows a markedly altered transcriptome starting at the DN2 stage.

(A) Uniform manifold approximation and projections (UMAP) of clusters identified by sc-RNAseq analysis of DN thymocytes isolated from WT and ΔEPE3 mice. (B) Heatmap representation of genes used to define the clusters. (C) Bar graph illustrating relative proportions of each cluster among DN thymocytes from WT and ΔEPE3 mice. (D) Volcano plots displaying DEGs in DN2, DN3-WT biased, DN3-ΔEPE3 biased, and DN4 clusters. (E) Heatmap representation of expression changes of selected genes, including TCF1 targets, in all clusters.

To investigate the basis for the developmental impairment observed in the ΔEPE3 mice, we identified differentially expressed genes (DEGs) in the DN2, DN3, and DN4 clusters (Fig. 4, D and E). Amongst DEGs, we observed previously identified TCF1 targets including Lef1, Dtx1, Ptcra, and Lat (Fig. 4, D and E, and fig. S7B) (16, 17). In addition, there were DEGs that had not been previously implicated as TCF1 targets such as Arpc1a, Ldhb, Myc, and Tcf3 (E2A) (Fig. 4E and fig. S7B). TCF1 ChIP-seq data from DP thymocytes has previously revealed TCF1 binding near all of these genes, suggesting that these genes are also direct TCF1 targets (10). A number of these putative TCF1 target genes encode regulators of metabolism including Ldha, Ldhb, and Gapdh, which were all down-regulated among DN2–4 cells from ΔEPE3 mice (Fig. 4, D and E, and fig. S7B). Gene expression changes were reflected in the sc-RNAseq as well as single-cell assay for transposase-accessible chromatin with sequencing (sc-ATACseq) clusters (fig. S7C). Together, these data reveal that the DN2 to DN3 developmental transition and the capacity of the pre-TCR to promote the DN to DP transition are compromised in ΔEPE3 mice. These effects are likely to reflect reduced expression of essential molecular effectors of signaling and gene regulation (Ptcra, Lat, Myc, and Lef1), including induction of inhibitors of Notch signaling (Dtx1), and metabolic changes associated with cell proliferation (27–29) (Fig. 4, D and E). We hypothesize that optimal expression of the genes that play roles in effective developmental transitions during αβ T cell development requires precise levels of TCF1 expression, which are governed by E protein activity.

The NBS and EPEs play distinct roles in regulating TCF1 expression

Pseudobulk ATACseq analysis of WT and ΔEPE3 T cell progenitors showed that all five EPEs were accessible in DN cells (Fig. 5A). Previous analysis of regulatory elements controlling the Tcf7 locus implicated a 1-kilobase (kb) region that included the 207 bp deleted in ΔEPE3 mice, and extended further to include a NBS located 177 bp away from the EPE3 deletion (Fig. 5A and boxed region in fig. S8) (30). We carried out additional studies to determine whether ΔEPE3 affected Notch function at the Tcf7 locus. We note that the NBS adjacent to EPE3 remains intact in the ΔEPE3 allele (fig. S8), enabling distinction between the effects of Notch signaling from those of E2A/HEB binding. ChIP analysis with anti-Notch1 antibody showed that the EPE3 deletion did not disrupt Notch binding to the NBS near EPE3 in ΔEPE3 cells (Fig. 5B). Next, we investigated whether ectopic expression of the active intracellular fragment of Notch1 (ICN1) (31, 32) could regulate TCF1 expression levels and whether this was affected by the EPE3 deletion (Fig. 5C). To this end, we retrovirally transduced fetal liver DN precursors from WT and ΔEPE3 mice with ICN1-expressing retrovirus (Fig. 5C). ICN1 transduction initiated TCF1 expression in DN1 progenitors from both WT and ΔEPE3 mice that were previously TCF1^lo (Fig. 5, D and E). This demonstrates that ICN1 can activate TCF1 expression and that the ΔEPE3 allele remains Notch responsive. However, ICN1 transduction did not rescue the TCF1 expression level in ΔEPE3 progenitors that were already expressing TCF1 (compare peaks denoted by red lines in right panels, Fig. 5, D and F) to the level expressed by WT progenitors [Fig. 5, D, F, and G, mean fluorescence intensity (MFI) plots]. We propose that Notch activates the Tcf7 locus, but thereafter, TCF1 expression is controlled by E protein binding to EPE3, which maintains the high levels of TCF1 needed to promote αβ T cell development.

Fig. 5. NBS and EPEs play distinct roles in regulating TCF1 expression.

(A) ATACseq analysis of chromatin accessibility near the Tcf7 locus in DN thymocytes from WT and ΔEPE3 mice. A 90-kb interval of Chr11 is depicted. (B) ChIP qPCR analysis of Notch binding to the Gapdh, Hes1 NBS, and EPE3-NBS elements. Hes1 serves as a positive control for Notch binding and Gapdh serves as a negative control. ChIP signals for Hes1 and EPE3-NBS were normalized to Gapdh. (C) Schematic to evaluate the effect of ectopic expression of Notch-IC on TCF1 expression and the ability to rescue T cell development in ΔEPE3 progenitor cells. (D and E) The ability of activated Notch to modulate TCF1 expression was assessed by retroviral transduction of WT or ΔEPE3 fetal liver precursors with empty vector (EV; pMiG) or the active intracellular fragment of Notch1 (pMiG-ICN1) after 4 days of culture on OP9-DL1 monolayers. After three additional days of culture, the effect of activated Notch on TCF1 expression was assessed by intracellular flow cytometry on electronically gated GFP⁺ CD25⁻ DN1 progenitors (D) or CD25⁺ DN2/3 progenitors (E). The fraction of TCF1low/negative cells in triplicate technical replicate cultures of the CD25⁻ DN1 progenitors above was quantified and depicted graphically as the median ± SD. (F) Representative histograms are depicted. Solid red lines mark the mean peak intensity for the TCF1 in WT control TCF1⁺ cells. ICN transduction reduces the intensity of intracellular TCF1 staining. (G) The level of TCF1 expression was assessed by determining the MFI of TCF1 by flow cytometry of the TCF1hi fraction of triplicate cultures of both CD25⁻ DN1 and CD25⁺ DN2/3 progenitors. Results are representative of three experiments performed. P values are shown.

Mechanism by which distinct EPEs control the level of TCF1 expression

To understand why EPE3 is unique amongst all EPEs in controlling Tcf7 expression in αβ T cell progenitors, we carried out Hi-ChIP in DN3 cells using anti-H3K27ac antibodies. Hi-ChIP constitutes enriching a subset of genomic DNA fragments from a Hi-C library using antibodies directed against H3K27Ac before deep sequencing. Whereas Hi-C scores for “all-to-all” proximity ligations, Hi-ChIP identifies those associated with specific proteins (33). Consequently, anti-H3K27ac Hi-ChIP identifies genomic interactions of regions that contain H3K27ac histone modifications associated with active promoters and enhancers (34, 35). We found that the Tcf7 promoter selectively interacted with EPE3 and, to a lesser extent, with EPE5 in DN3 cells, but not in bone marrow–derived pro-B cells lacking TCF1 expression (Fig. 6A). By contrast, long-distance interactions between the Vdac1 and Skp1 genes were present in both DN3 and pro-B cells (Fig. 6A). Ablation of EPE3 reduced interactions between the chromatin adjacent to EPE3 and the transcription start site (TSS) of the Tcf7 gene (Fig. 6, B and C). Interaction between EPE5 and the TSS also appeared to be modestly reduced in ΔEPE3-DN3 cells (Fig. 6B, quantified in C). We infer that the 207-bp EPE3 sequence containing the three E protein binding sites is required for optimal Tcf7 gene expression via interactions with the TSS of the Tcf7 gene.

Fig. 6. Mechanism by which EPEs control the level of TCF1 expression.

(A) Interaction heat maps illustrating H3K27ac Hi-ChIP signals in Rag2^−/− pro-B (bottom left) and DN3 (upper right) cells, presented at 1 kb resolution. The region spanning the five EPEs adjacent to the Tcf7 locus is demarcated by dashed box, enlarged in the inset for closer examination on the right. ChIP-seq peaks featured in (A) are indicated on the top of the heatmaps. The right-side magnified map also showcases the H3K27ac peak track derived from H3K27ac Hi-ChIP data. The DN3 specific interaction is annotated by black circle. The scale bars indicate the contact frequency. (B) Interaction heat maps illustrating H3K27ac Hi-ChIP signals within the Tcf7 promoter region in ΔEPE3 (bottom left) and WT (upper right) DN3 cells, presented at 1 kb resolution. The scale bars indicate the contact frequency. The difference map (ΔEPE3 minus WT) for the same region is shown on the right; regions with increased and decreased interactions are colored red and blue, respectively. The scale bar indicates the difference in contact frequency. ChIP-seq peaks featured in (A) are indicated on the top of the heatmaps. Interactions between EPE3 and Tcf7 TSS are indicated by black circles. (C) Bar graphs displaying the relative interactions between Tcf7 TSS and the five E2A peaks in WT and ΔEPE3 DN3, as featured in (A). These interactions were quantified within both WT and ΔEPE3 DN3 cells, with error bars representing the SEM for two replicates.

These data suggest a model by which E proteins bound to the regulatory region of the Tcf7 gene control Tcf7 expression (Fig. 7). The diagram illustrates our findings that before Id3 induction, interactions between EPE3 and the TSS, and perhaps somewhat for EPE5, create a chromatin loop that supports high levels of Tcf7 transcription. Signals from the pre-TCR induce low levels of Id3 that spare E protein function, leaving the EPE3-TSS loop intact to sustain the robust transcription of Tcf7 needed to support the high levels of TCF1 expression required for αβ T cell development. In ΔEPE3 cells, this interaction is diminished, reducing the level of Tcf7 expression and attenuating αβ T cell development. Strong γδ TCR signals induce high levels of Id proteins that dislodge E proteins from the Tcf7 locus, disassemble the EPE3-TSS loop, and diminish Tcf7 transcription, thereby facilitating γδ T cell maturation. Thus, EPE3 deletion mimics the effect of strong γδ TCR signals by reducing Tcf7 transcription and promoting γδ T cell maturation.

Fig. 7. Schematic model showing how the EPEs control the level of TCF1 expression depending on the signal received by developing thymocyte precursors.

(A) Diagram shows the potential looping of the E protein regulome in the Tcf7 regulatory region containing five EPEs and TSS. (B) Diagram shows the role of signal induced Id proteins that inactivate E protein binding to the EPEs.

DISCUSSION

Because of the critical role of TCF1 at all stages of T cell development and in peripheral T cells, signal-driven transcriptional mechanisms controlling Tcf7 gene expression are of particular interest. Here, we provide new insights into Notch-independent control of TCF1 expression by E proteins. These studies identify the E protein regulome adjacent to the Tcf7 locus and demonstrate distinct dependence on the five E protein–bound elements (EPE) by different T cell subsets. We show that EPE3 exclusively controls development of αβ T cells, such that the modest reduction in TCF1 expression caused by EPE3 ablation attenuates the DN to DP transition. By contrast, EPE1 and 5 selectively control the development and maturation of γδ T cells. We show that EPE3, but not EPE1 or EPE5, is epigenetically decorated and interacts with Tcf7-TSS to control Tcf7 gene expression. We determine that the influence of EPE3 begins very early in development at the DN2 stage and persists through subsequent developmental transitions.

A recurring theme in hematopoiesis is activation of lineage-specific gene expression and repression of lineage-inappropriate gene expression to commit multipotential progenitor cells to the differentiation program of a specific cell type. TCF1 and its targets Bcl11b and Gata3 serve as critical molecular effectors of commitment of thymic progenitors to the T lineage (16, 17). Tcf7, the gene that encodes TCF1, is activated in response to Notch signals provided by the thymic microenvironment (6, 7). We provide evidence that after TCF1 expression is initiated, E box DNA binding proteins expressed by thymocytes (E2A and HEB) play an essential role in maintaining and determining the optimal level of TCF1 expression starting at the DN2 stage. There are five EPEs in the Tcf7 locus that play distinct roles in supporting TCF1 expression in the γδ and αβ T lineages. EPE1, EPE3, and EPE5 play critical roles in regulating TCF1 expression in γδ T cell progenitors, while EPE3 also controls TCF1 levels in αβ T cells. Because ablation of EPE3 eliminates E protein binding sites while retaining adjacent NBS, the Tcf7 locus remains Notch responsive. Consequently, we contend that Notch signaling is responsible for initiating TCF1 expression in early T cell progenitors and then maintenance of distinct levels of TCF1 expression is transferred to E proteins upon arrival at the DN2 stage and beyond.

A particularly interesting aspect of our analysis is the finding that a reduction in TCF1 expression of less than twofold in ΔEPE3 thymocytes causes severe impairment in the DN to DP transition of αβ T cell development, emphasizing that precise level of TCF1 expression is critically important in controlling this developmental transition. We propose that TCF1 dosage during the DN to DP transition is dynamically regulated by pre-TCR and other signals that also regulate the function of E2A/HEB via signal-induced expression of Id3, which represses E protein binding to DNA. Our observation that ectopic expression of TCF1 fails to fully rescue T cell development in ΔEPE3 precursors emphasizes the requirement for signal-driven modulation of TCF1. We infer that our data distinguish between basal expression of TCF1 and signal-induced modulation of TCF1.

Among several genes that are reduced in expression in ΔEPE3 progenitors, Lat and Ptcra are required for the DN to DP transition (29, 36). Moreover, key DNA binding proteins that support the pre-TCR–induced DN to DP transition are also reduced in ΔEPE3 progenitors, including Myc and Tcf3 (E2A). Last, several metabolic regulators (Gapdh, Ldha, and Ldhb) are down-regulated in ΔEPE3-DN cells. Because maintenance of glycolysis by Notch is important for development to the DP stage (31, 37, 38), and because the accumulation of lactate blocks T cell proliferation (32), these changes in gene expression may contribute to metabolic alterations that attenuate development. It should be noted that despite EPE3 ablation causing major reduction in TCF1 expression in SP thymocytes, they were produced in proportion to their DP precursors, suggesting no impairment of development. TCF1, in collaboration with LEF1, plays an essential role in CD4SP development (11). The resistance of CD4SP and CD8SP cells to reduced TCF1 in ΔEPE3 mice may occur either because the reduced levels of TCF1 suffice, or because LEF1 provides adequate compensation during the DP to SP transition. While the changes in gene expression provide potential explanations for the attenuation of the DN to DP transition of αβ T cell development, the targets through which the combined reduction of TCF1 and E protein activity facilitate γδ lineage commitment remain to be elucidated.

EPE3 can be characterized as a classic enhancer that augments TCF1 expression by associating with the Tcf7 promoter and TSS. Several transcription factors have been shown to bind to this region in addition to E proteins and Notch. The ΔEPE3 deletion described here removes putative E protein binding sites but leaves intact the NBS. The profound effect of ΔEPE3 on αβ T cell development clearly demonstrates the importance of E protein binding to EPE3 and the insufficiency of Notch binding to the adjacent NBS to maintain optimal TCF1 expression. Even ectopic expression of ICN fails to restore TCF1 expression in ΔEPE3 DN cells to levels seen in WT progenitors, suggesting that the critical role of E proteins is manifest on a fully Notch-responsive Tcf7 allele. Unlike the effects of deleting E protein binding sites seen here, previous studies show that deleting NBS adjacent to EPE3 does not significantly affect αβ T cell development (30). We surmise that the absence of that particular NBS can be compensated for by Notch binding to other presently uncharacterized regulatory sequences.

These observations raise the question as to how residual TCF1 expression is maintained from ΔEPE3 alleles. One possibility is that Tcf7 transcription is regulated by sequences flanking the E protein binding sites in the EPE3 region. In addition to an NBS, the flanking region also has predicted binding sites for GATA3, TCF1, and Runx. A related study that ablated a larger genomic interval encompassing the EPE3, as well as the TCF1, Gata3, and Runx sites, observed greater reduction in TCF1 expression than did ablation of EPE3 alone, suggesting that these flanking sequences may be responsible for supporting residual expression of TCF1 in ΔEPE3 cells (30). In support of this hypothesis, we found that chromatin accessibility of this region was reduced, but not eliminated, on ΔEPE3 alleles. Alternatively, it is possible that another EPE, such as EPE5, might sustain the lower levels of Tcf7 expression observed in ΔEPE3 progenitors. Although individual ablation of the other EPEs did not affect αβ T cell development, it remains possible that they may do so when combined with EPE3. EPE5 is of particular interest, because we detected contacts between EPE5 and the Tcf7-TSS in Hi-ChIP analysis. These contacts were weaker than the EPE3-TSS contacts and the EPE5-TSS contacts were further reduced in ΔEPE3 progenitors, raising the possibility of cooperation.

During T cell development, E protein function is dynamically regulated posttranscriptionally through the induction of members of the antagonistic inhibitor of DNA binding (Id) family, which block E protein function and are induced in proportion to TCR signaling intensity and/or duration (33–35, 39–41). In support, we recently demonstrated that more intense TCR signals that promote γδ lineage commitment results in reduced TCF1 expression, which acts to facilitate γδ lineage commitment (13). Nevertheless, because Id3 is regulated by the Egr family of transcription factors, which are induced by a diverse array of stimuli, E protein activity is almost certainly also controlled by extracellular signals in addition to those emanating from the TCR, particularly before the DN3 stage.

TCF1 plays multiple roles in the thymus as first shown by germline deletion of Tcf7 which markedly impairs T cell development (3, 4). However, in Tcf7^+/− mice TCF1 expression is modestly and uniformly reduced in all thymocyte subsets but T cell development is not impaired (15, 42). We infer that TCF1 expression from one intact allele that can respond dynamically to signals that regulate developmental transitions is sufficient to sustain αβ T cell development. This observation contrasts with the phenotype of ΔEPE3 mice, suggesting that all forms of reduction in TCF1 expression are not equivalent during T cell development. Specifically, we propose that E proteins modulate Tcf7 gene expression in response to pre-TCR and other signals via EPE3, thereby selectively affecting transition of DN to the DP stage. This reasoning also provides a plausible explanation for a lack of similar phenotype in ΔEPE5 mice despite reduced expression of TCF1. We propose that E protein binding to EPE5 is not dependent on pre-TCR–dependent or other signals that control the DN to DP transition. The particular importance of EPE3 is highlighted by this region being marked with H3K27ac and its interaction with the Tcf7 promoter in DN cells. These observations suggest that the Tcf7 regulome functions distinctly to respond in different situations and the outcome is expression of appropriate level of TCF1.

Last, our finding that E protein binding to EPE3 plays a key role in regulating TCF1 expression in developing thymocytes has implications beyond T cell development. For example, TCR modulation of E protein binding to EPE3 may regulate the expression of TCF1 in peripheral T cells, where it has been shown to play a critical role in effector function in the context of exhaustion due to prolonged TCR stimulation, such as is observed in chronic viral infection or in cancer (36–39). The expression of TCF1 by “stem-like” precursor exhausted T cells, which can be reinvigorated by checkpoint blockade in cancer (38, 39, 43, 44), may be controlled through the TCR-Id3-E protein axis and regulated by EPE3 and/or other EPEs (45, 46).

MATERIALS AND METHODS

Mice, tissue, and cell extraction

WT male and female C57BL/6J mice were obtained from the Jackson Laboratory (Bar Harbor, ME). ΔEPE1, ΔEPE2, ΔEPE3, ΔEPE4, and ΔEPE5 mice were generated using CRISPR-Cas9 technology as described in (47). gRNAs sequences and primers for genotyping are listed in tables S1 and S2. ΔEPE1 mice were generated as described (13). Generation of Tcf7^−/− mice is previously described (1). All mouse experiments were performed under protocols approved by the National Institute on Aging Institutional Animal Care and Use Committee (343-LCI-2025) and Fox Chase Cancer Center (protocol 96-4). All the mice used in the experiments were age matched (8 to 14 weeks old). The studies were carried out in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals. Thymi were isolated and gently pressed through 70-μm strainer to extract the thymocytes. The thymocytes were resuspended in 10 ml of RPMI medium with 2% fetal bovine serum (FBS; Gibco, MD).

Flow cytometry and cell sorting

Thymocytes were stained and acquired on a Symphony-analyzer (Becton-Dickinson, DC.) and analyzed with FlowJo version 10.0 (FlowJo LLC, USA). eBioscience Fixable Viability Dye eFluor-780 (Thermo Fisher Scientific, MA) was used to exclude dead cells. Fluorescent antibodies were purchased from BD Biosciences, eBioscience, or BioLegend and listed in table S3 and other reagents used for FACS are listed in table S4. In brief, cells were incubated with FC block and stained with antibodies for surface markers and fixed with 2% paraformaldehyde. For intracellular staining, cells were permeabilized using the eBioscience FoxP3/transcription factor staining buffer set (Thermo Fisher Scientific, MA) and stained with antibodies against transcription factors. For sorting, CD8α cells were removed by incubating thymocytes with CD8α microbeads (Miltenyi Biotech, USA) as per the user’s manual. CD8-depleted cells were stained with antibodies for surface markers and resuspended in RPMI-1640 supplemented with 2% FBS. Cells were sorted in RPMI-1640 medium supplemented with 10% FBS.

In vitro culture of DN progenitors

CD8-depleted cells were stained with antibodies for surface markers. Cells were resuspended and sorted in RPMI-1640 medium supplemented with 10% FBS. DN4 thymocytes (lineage-depleted with antibodies reactive to: TCRβ, CD4, and CD8, CD11c, CD11b, Gr1, CD19, B220, and Ter119) were isolated by flow cytometry based on the absence of expression of CD44 and CD25 and cultured in 12-well plates overnight at 10⁶ cells/ml at 37°C with 5% CO₂. DP (CD4⁺CD8⁺) thymocytes were quantified by flow cytometry every 8 up to 24 hours. For culture of adult DN thymocytes on OP9-DL1 cultures, DN thymocytes were isolated by magnetic bead depletion using anti-CD4 and anti-CD8 magnetic beads, following which they were plated on OP9-DL1 monolayers as 100,000 cells per well of a 24-well plate in the presence of IL-7 and Flt3 as described (23). After 1 and 3 days of culture, development to the DP stage was assessed by flow cytometry. For proliferation analysis, the thymocytes were stained with carboxyfluorescein diacetate succinimidyl ester (1 μg/ml) before culture on OP9-DL1 monolayers. For retroviral transduction, the DN thymocytes were spin infected with retrovirus encoding empty vector (pMiG) or Tcf7 (pMiG-Tcf7) in the presence of polybrene (8 mg/ml) for 90 min, following which the behavior of transduced progenitors was monitored by flow cytometry by electronic gating on green fluorescent protein–positive (GFP⁺) progenitors at days 1 and 2 posttransduction (13). For transduction of fetal liver precursors, embryonic day 14 (E14) fetal liver precursors from WT or ΔEPE3 mice were cultured on OP9-DL1 monolayers for 4 days as described (13), transduced with empty vector (pMiG) or the Notch intracellular domain (pMiG-ICN1), cultured for 3 days, and then subjected to intracellular staining with anti-TCF1 antibody as described above.

ChIP and Hi-ChIP

E14 fetal liver precursors from WT and ΔEPE3 were cultured on OP9-DL1 monolayers for 7 days, following which DN3 cells were isolated by flow cytometry, fixed in 1% formaldehyde for 10 min, and sonicated to produce chromatin fragments. The resulting fragments were immunoprecipitated with 5 μg of anti-Notch antibody (Cell Signaling Technology, no. 3608) with Protein G Dynabeads (Invitrogen, 10003D). Eluted chromatin fragments were analyzed in triplicate by quantitative polymerase chain reaction (qPCR) using SYBR Green and primer sets spanning sites in Gapdh, Hes1, and EPE3 (Gapdh: F, TGGCGTAGCAATCTCCTTTT; and R, TGGCGTAGCAATCTCCTTTT; Hes1: F, CGTGTCTCTTCCTCCCATTG; and R, CGTGTCTCTTCCTCCCATTG; and EPE3: F, CACCGAGCATTCTCAGCAGCA; and R, ACAGCTTTTATCGCACGTTTATGAAGG). qPCR results from the ChIP analysis were normalized to Gapdh and represented graphically as the mean ± SD of triplicate measurements. To prepare cells for Hi-ChIP, DN3 precursors were isolated by flow cytometry following culture on OP9-DL1 monolayers for 7 days as described above. Hi-ChIP data for WT pro-B cells were obtained from previously published datasets (GSE214438) (48). Hi-ChIP assays were conducted using duplicate samples of WT and ΔEPE3 DN3 cells. The Arima HiC+ kit (Arima Genomics, catalog no. A101020) was used, strictly following the manufacturer’s protocol as outlined in Arima-HiC documents A160168 v00 (Hi-ChIP) and A160169 v00 (library preparation using the Swift Biosciences Accel-NGS 2S Plus DNA Library Kit). For each Hi-ChIP sample, 5 × 10⁶ cells were collected, and a total of 2.5 μg of H3K27ac antibody (Active Motif, catalog no. 39133) was used. The resulting barcoded Hi-ChIP libraries were combined and subjected to sequencing using an Illumina NovaSeq instrument to produce an average depth of 200 million reads. Hi-ChIP reads were processed using Juicer to generate .hic files (49). Juicer was run with the flags “-g mm10 -s Arima”. To compare multiple experiments, Hi-ChIP data were normalized by downsampling to the same number of total reads. Contacts within 2000 bp and reads with mapping quality scores below 30 were removed before downsampling. To quantify the interaction frequency between specific regions, such as EPE sites and promoters, the number of contacts was counted after the experiments had been downsampled to the same number of total reads. Hi-ChIP contact maps were visualized using cooltools and are shown at 1000 bp resolution. Difference maps were created by subtracting the in WT contacts from ΔEPE3 contacts.

Single-cell RNAseq

Cell isolation

CD4⁻CD8⁻ DN cells were obtained from C57BL/6 (WT-DN) and ΔEPE3 (ΔEPE3-DN) mice. In brief, thymocytes were depleted using CD8α microbeads (Miltenyi Biotech, USA) (fig. S7A). The CD8-depleted fraction was incubated with lineage marking antibodies (TCRβ, CD4, and CD8, CD11c, CD11b, Gr1, CD19, B220, and Ter119) and then lineage cells were sorted in RPMI-1640 medium supplemented with 10% FBS. Postsorting, a small fraction of sorted cells was used to check the purity (>98%). A total of 200,000 cells per sample were provided for the library preparation.

Library generation

Using the Chromium Next GEM Single Cell 3′ _GEM, Library & Gel Bead Kit version 3.1 (10x Genomics, PN-1000121) and the Chromium Chip G Single Cell Kit (10x Genomics, PN-1000120), the thymocytes, at 1000 cells/μl, were loaded onto a chromium single-cell controller (10x Genomics) to generate single-cell gel beads in emulsion (GEMs) according to the manufacturer’s protocol. In brief, approximately 10,000 cells per sample were added to chip G to create GEMs. Cells were lysed, and the bead captured poly(A) RNA was barcoded during reverse transcription in a Thermo Fisher Scientific Veriti 96-well thermal cycler at 53°C for 45 min, followed by 85°C for 5 min. cDNA was generated and amplified. Quality control and quantification of the cDNA were conducted using Agilent’s High Sensitivity DNA Kit (5067-4626) in the 2100 Bioanalyzer. sc-RNAseq libraries were constructed using the Chromium Single Cell 3′ Library Kit version 3.1 and indexed with the Single Index Kit T Set A (10x Genomics, PN-1000213). The libraries were sequenced using an Illumina NovaSeq 6000 sequencer with a paired-end, single-indexing strategy consisting of 28 cycles for read 1 and 91 cycles for read 2.

Data processing

Demultiplexing of raw base call files into FASTQ files was completed with 10x Genomics Cell Ranger (version 3.0.2) mkfastq coupled with mouse reference version mm10. The results from Cell Ranger were processed using DoubletFinder and Seurat, following the standard procedures (50–52). DoubletFinder was used on each sample individually to remove doublets. One WT and one ΔEPE3 sample were paired together and integrated with a second pair of WT and ΔEPE3 samples to remove experimental batch effect. Integrated data were clustered with a resolution of 0.4. Clusters comprising less than 5% were removed from further analysis, as were NKT cells. DEGs between genotypes were identified with the FindMarkers function of Seurat.

sc-ATACseq analysis

Cell isolation

Sorted WT-DN and ΔEPE3-DN cells isolated as above were used for sc-ATACseq analyses. For each sample, 100,000 cells were used for nuclei isolation as described (53). In brief, 100,000 cells were centrifuged at 500g for 5 min at 4°C. Cell pellets were washed with 50 μl ice-cold phosphate-buffered saline. Cell pellets were gently resuspended in 50 μl of cold lysis buffer [10 mM tris-Cl (pH 7.4), 10 mM NaCl, 3 mM MgCl2, and 0.1% IGAPAL CA-630] and centrifuged at 500 g for 10 min at 4°C. Isolated nuclei were passed through Flowmi cell strainer (40 μm) (Scienceware, H13680-0040).

Library generation

Using the Chromium Next GEM Single Cell ATAC Library & Gel Bead Kit version 1.1 (10x Genomics, PN-1000175) and the Chromium Next GEM Chip H Single Cell Kit version 1.1 (10x Genomics, PN-1000161), thymocyte nuclei, at 7000 nuclei/μl, were loaded onto a Chromium controller (10x Genomics) to generate single-nuclei GEMs after which library construction was performed according to the manufacturer’s protocol. In brief, nuclear DNA was fragmented at open regions of the chromatin with transposase while adapter sequences were added at the ends. The fragmented nuclear DNA was then loaded on H chips and partitioned into nanoliter-scale GEMS. Illumina P5 sequence, a barcode and read 1 sequence were added by PCR amplification. Unused primers and reagents were removed from the sample by solid-phase reversible immobilization (SPRI) beads. The library construction was completed by the addition of both P7 and a sample index. The libraries were sequenced using an Illumina NovaSeq 6000 sequencer with a paired-end, dual-index strategy consisting of 50 cycles for read 1, 50 cycles for read 2, 8 cycles for the i7 index, and 16 cycles for the i5.

Data processing

Demultiplexing of raw base call files into FASTQ files was completed with 10x Genomics Cell Ranger-ATAC MKFASTQ coupled with mouse reference version mm10. The results from Cell Ranger were processed using Seurat version 3 and Signac versions 0.2.1 and 1.0.0, following the standard procedures (54). Genomic traces were generated from single-cell ATAC data that have been converted to pseudo-bulk tracks of DN cells, as described (52).

Software and statistics

Statistical significance was determined by the student’s unpaired t test.

Supplementary Materials

This PDF file includes:

Figs. S1 to S8

Tables S1 to S4