Chromatin organizer SATB1 controls the cell identity of CD4 + CD8 + double-positive thymocytes by compacting super-enhancers

CD4 + and CD8 + double-positive (DP) thymocytes are at a crucial stage during the T cell development in the thymus. DP cells rearrange the T cell receptor gene Tcra to generate T cell receptors with TCRβ. Then DP cells differentiate into CD4 or CD8 single-positive (SP) thymocytes, Regulatory T cells, or invariant nature kill T cells (iNKT) according to the TCR signal. Chromatin organizer SATB1 is highly expressed in DP cells and plays an essential role in regulating Tcra rearrangement and differentiation of DP cells. Here we explored the mechanism of SATB1 orchestrating gene expression in DP cells. Single-cell RNA sequencing assay of SATB1-decient thymocytes showed that the cell identity of DP thymocytes was changed, and the genes specically highly expressed in DP cells were down-regulated. The super-enhancers regulate the expressions of the DP-specic genes, and the SATB1 deciency reduced the super-enhancer activity. Hi-C data showed that interactions in super-enhancers and between super-enhancers and promoters decreased in SATB1 decient thymocytes. We further explored the regulation mechanism of two SATB1-regulating genes, Ets2 and Bcl6, in DP cells and found that the knockout of the super-enhancers of these two genes impaired the development of DP cells. Our research reveals that SATB1 globally regulates super-enhancers of DP cells and promotes the establishment of DP cell identity, which helps understand the role of SATB1 in thymocyte development. nuclear 4C-seq, 3C-HTGTs. 5’RACE of bioinformatics ChIP-seq, 3C-HTGTs, FACS 5’RACE of Ets2-SE


Introduction
T lymphocytes, a critical component of adaptive immunity, develop as thymocytes in the thymus [1][2][3] . Most of T cells in human and mouse are αβ T cells expressing the TCR consisting of an α and β chain.
The primary purpose of αβ T cell development in the thymus is to generate T lymphocytes with highly diverse T cell receptors (TCRs) to recognize a wide variety of foreign antigens and avoid responses to self-antigens at the same time 1,4 . During differentiation into mature T cells, thymocytes go through several developmental stages, which can be characterized by the cell surface expression of CD4 and CD8 proteins. CD4 + CD8 + double positive (DP) thymocytes are at the center of αβ T cell development 2 . DP cells generate the αβTCR on the cell surface to recognize cortical epithelial cells expressing Class I or Class II MHC plus self-peptides for positive and negative selection 1,4 . TCR signaling induces DP cells differentiation into CD4 + single positive (SP), CD8 + SP, regulatory T cells (T reg ), or invariant natural killer T cells (iNKT) 4 .
The primary event in DP cells is Tcra rearrangement to generate TCRs with an a nity to self-peptide:MHC complexes, which requires that cells do Tcra rearrangement e ciently and sense the TCR signal correctly 3 . Due to the low frequency of positive selection, the Tcra gene can do multiple rounds of Vα-Jα rearrangement on both alleles to increase the chance of passing positive selection 5 . DP cells also upregulate recombinase Rag1 and Rag2 expression for e cient rearrangement 6, 7 . Cell lifespan is another critical factor for generating diverse Tcra repertoire and the proper development of DP thymocytes 8 . DP thymocytes can survive for an average of three to four days to allow Jα rearrangement sequentially from the 5' end to the 3' end of the Jα array. Depleting the factors that regulate the survival of DP thymocytes (e.g., ROR, an orphan nuclear factor, and YY1, Yin Yang 1) shortens the lifespan and reduces the usage of the 3' Jα segments [9][10][11] . However, it is unknown that DP thymocytes orchestrate expression programs for e cient Tcra rearrangement and proper selection.
Special AT-rich binding protein 1 (SATB1) is a chromatin organizer that plays certain roles in many tissues, including skin 12 , tooth 13,14 , and liver 15 , although it is not ubiquitously expressed. Studies also showed that SATB1 is involved in embryogenesis 16, 17 , neurogenesis 14,18 , hematopoiesis 19,20 , and erythropoiesis 21,22 . Nevertheless, most of the studies on SATB1 focused on its role in T cell development, partially due to the highest SATB1 expression during thymocyte development, especially at the DP stage 6, [23][24][25][26] . Satb1 deleted mice displayed a smaller thymus, increased proportion of DP thymocytes, and fewer CD4/CD8 single positive cells, indicating a blockage at the DP stage during thymocyte development 23 . The development blockage may be due to insu cient Tcra rearrangement and impaired positive selection 6, 27 . In post-selection thymocytes, SATB1 also plays roles in activating lineagespecifying genes, including ThPOK, Runx3, CD4, CD8, and Foxp3 25,28 . SATB1-de cient thymocytes display inappropriate T lineage determination after MHC I-and II-mediated selection, and de cient differentiation of regulatory T cells (T reg ) 28 . These ndings indicate the critical role of SATB1 in DP thymocytes, although the detail molecular mechanisms remain elusive. SATB1 can activate or repress gene transcription by recruiting p300/CBP-associated factor (PCAF) or histone deacetylase, respectively 29 . SATB1 functions as a pioneer factor in establishing T reg cell-speci c super-enhancers, which is crucial for T reg cell lineage speci cation in the thymus 28 . More studies focus on the roles of SATB1 in chromatin organization, which was suggested when it was identi ed as a nuclear matrix protein 6, 21,24,25,30,31 . An early report showed that SATB1 induced a unique transcriptionally active chromatin structure at the T helper 2 (T H 2) cytokine locus during T H 2 cell activation 24 . SATB1 directly controls the regulatory elements of several lineage-specifying genes, including Zbtb7b (encoding ThPOK), Runx3, Cd4, Cd8, and Foxp3 25 . We also reported that SATB1 mediates the DP-speci c interaction between the anti-silencer element and Rag1 gene promoter for the high expression of Rag1 and Rag2 in DP thymocytes 6 . However, there is a lack of genome-wide study on SATB1 mediated chromatin architecture and its characteristics 29 .
In the present study, we analyzed the development of Satb1 de cient thymocytes using single-cell RNA sequencing (scRNA-seq) technique and found that Satb1 deletion changed the cell identity of DP thymocytes. Further analysis showed that Satb1 plays an essential role in promoting the activity of superenhancers by reorganizing chromatin interactions. We also analyzed the regulation and function of two genes encoding transcription factors BCL6 and ETS2, which are regulated by SATB1 and super-enhancers in DP thymocytes. These ndings indicate that SATB1 controls the cell identity of double-positive thymocytes by reorganizing super-enhancers.

Results
SATB1 deletion changed the cell identity of CD4 + CD8 + double-positive thymocytes To investigate the role of SATB1 in thymocyte development, we employed single-cell RNA sequencing on all thymocytes from 6-week-old female mice in which the Satb1 gene was deleted in hematopoietic stem cells using vav-cre transgene 6, 27 . A total of 13948 cells consisting of 6844 Satb1 de cient and 7104 control thymocytes passed the quality control criteria. Cells were separated into 16 different clusters according to gene expression and cell cycling progress ( Fig. 1a and S1a). Our previous study showed that SATB1 regulates recombinase Rag1 expression in DP thymocytes 6 . The scRNA-seq data showed that reduced Rag1 expression occurred on the level of a single cell and the number of cells ( Fig. 1b and S1bc). We observed a decreased DN2/3 thymocyte number, an increased DN4/ISP/DP cell number, and a reduced CD4/CD8 SP cell number in the Satb1 de cient thymus ( Fig. 1a and S1d), which is consistent with ow cytometry analysis in the previous report 23 . We noticed that most of the DP thymocytes from Satb1 de cient thymus were enriched in the cluster three, while the cells in the cluster three were rare in the wild type (Fig. 1a), suggesting that the cluster three was generated by Satb1 deletion. The result indicated that Satb1 de ciency changed the cell identity of DP thymocytes.
To explore how Satb1 deletion changed the cell identity, we analyzed the gene expression pro le of Satb1 de cient DP thymocytes with independent bulk RNA-seq data ( Fig. 1c and S2a). There were 576 downregulated genes and 928 upregulated genes in Satb1 de cient DP thymocytes (Fig. S2b-d and Table  S1). Most of the upregulated genes were expressed in thymocyte development earlier stages like DN1 and DN2a, and the expression levels of these genes are higher in the earlier stages (Fig. 1d, e and S2e). To con rm SATB1 repressing the expression of the genes expressed explicitly in earlier stages, we did the Gene Set Enrichment Analysis (GSEA) with the DN1 and DN3 speci c gene sets. The data showed that the DN1 and DN3 gene sets were signi cantly repressed by SATB1 (Fig. 1f). Most of the downregulated genes in Satb1 de cient DP thymocytes were expressed speci cally in DP and SP thymocytes (Fig. 1d). The expressions of these genes were low in the earlier stages, especially from DN2b to ISP, and they were highly expressed in DP and SP (Fig. 1e). These data suggested that SATB1 controls the cell identity of DP thymocytes by activating the DP-speci c genes and repressing the genes expressed in earlier stages.

SATB1 binds and activates super-enhancers of DP thymocytes
It has been showed that super-enhancers drive expression of genes that de ne cell identity 32 . The previous study showed that SATB1 plays a role in activating super-enhancers in Foxp3 + regulatory T cells 28 . To explore the role of SATB1 in organizing super-enhancers of DP thymocytes, we identi ed 246 superenhancers of DP thymocytes using the algorithm ROSE 32 with chromatin immunoprecipitation sequencing (ChIP-seq) data of histone H3 acetylated at Lys27 (H3K27ac) (Fig. 2a and Table S2). Many super-enhancers are associated with known DP signature genes, such as Tcra, Rag1, Cd8a, Cd4, etc. (Fig. 2a). Most of DP super-enhancer-associated genes are highly expressed in the DP stage during thymocyte development (Fig. 2b), suggesting that super-enhancers control the expression of cell identity genes in DP thymocytes.
We noticed that SATB1 binding sites overlapped with the active histone modi cation markers histone H3K4 monomethylation (H3K4me1), H3K4me3, H3K27ac, and the binding sites of the components of chromatin organization complex cohesion Rad21, Nipbl, and CTCF ( Fig. 2c and S3a-d). It was reported that active promoters can produce false positive peaks in ChIP-seq experiments 33 . The SATB1 ChIP-seq experiment in Satb1 de cient thymocytes from Satb1 f/f ⋅ CD4-cre mice showed that the SATB1 binding in active regions were speci c in SATB1-expressing WT cells (Fig. 2c). The SATB1 signals were enriched in gene promoter regions and super-enhancers ( Fig. 2d and S3e-f). The Satb1 deletion impaired both traditional enhancers and super-enhancers ( Fig. 2e and S3g). The average intensities of H3K27ac and H3K4me1 in super-enhancers were reduced in Satb1 de cient DP cells ( Fig. 2e and S3h). Most of the genes associated with the lost and maintained super-enhancers were downregulated in Satb1 de cient DP thymocytes (Fig. 2f-h and S3i). The genes associated with super-enhancers were more sensitive to Satb1 deletion than typical enhancers (TEs) associating genes (Fig. 2o). These data suggested that SATB1 regulates the DP signature genes by activating super-enhancers. SATB1 clusters are associated with cell identity genes.
The SATB1 binding sites were clustered in the genome and displayed a super-enhancer like distribution (Fig. 2c). We did the ROSE analysis with SATB1 ChIP-seq data and identi ed 743 SATB1 clusters (Fig. 3a). The SATB1 clusters associated genes consisted of many DP signature genes, such as Tcra, Rag2, Ets1, Cd8a, Cd4, etc. The GSEA analysis showed that SATB1 clusters associated genes had high level of active histone modi cation H3K27ac (Fig. 3b). Satb1 deletion downregulated most of the SATB1 clusters associated genes (Fig. 3c), suggesting that SATB1 clusters tended to activate the expression of their associated genes. We analyzed the expression of the SATB1 clusters associated genes during thymocyte development. Most of these genes are speci cally expressed in the DP stage and their expression are signi cantly lower in other stages than in DP thymocytes (Fig. 3d). The Gene Ontology (GO) term enrichment analysis showed that these genes enriched in DP associating functions, such as T cell activation and V(D)J recombination (Fig. 3e). These results indicated that SATB1 formed clusters to regulate the DP signature genes. SATB1 mediated the chromatin interactions of super-enhancers.
The previous report showed that SATB1 regulated V(D)J recombinase Rag1 and Rag2 expression in DP thymocytes by mediating an enhancer-promoter interaction 6 . We performed Hi-C experiments with sorted DP thymocytes from Satb1 f/f and Satb1 f/f ⋅vav-cre mice. Visualization of Hi-C data revealed some alterations of chromatin organization in Satb1 de cient DP thymocytes (Fig. S4). Satb1 deletion induced a modest increase in diagonal interactions of most. The contact density decaying curves showed that the interactions less than 2Mb in length slightly increased in Satb1 de cient thymocytes (Fig. S4b). Satb1 deletion has few effects on compartments (Fig. S5a). Satb1 deletion didn't change the number of TADs but slightly reduced sizes of TADs, especially the TADs containing SEs (Fig. S5b). The TAD strength almost kept the same level in Satb1 de cient cells (Fig. S5c). The data suggest that the role of SATB1 on the organization of chromatin is in some speci c regions rather than the whole genome.
To characterize chromatin interactions affected by Satb1 deletion, we generated a chromatin interaction matrix with a 50 kb resolution and identi ed signi cantly differential interactions using the Bioconductor package multilHiCcompare 34 . We compared 766,111 chromatin interactions and found 500 signi cantly increased interactions and 411 decreased interactions. Most of the distances between two anchors of decreased interactions are around 10kb, less than increased interactions (Fig. 4a). In addition, there are high SATB1, H3K27ac, and CTCF ChIP-seq signals in the anchor regions of decreased interactions (Fig. 4b). These data showed that, unlike CTCF and Cohesin, which mediate long-range chromatin interactions, SATB1 mainly mediates chromatin relatively short-range chromatin interactions (from tens to 100kb).
We then explored the chromatin loop structure using the program Fit-Hi-C, a tool for assigning statistical con dence estimates to intra-chromosomal contact maps 35 . Satb1 deletion increased loop numbers but reduced the loop strength ( Fig. 4c and S5d). We also noti ed the reduced strength of enhancer-promoter loops and intra-SE loops ( Fig. 4c and S5d). The loop strength increased in the promoters of the upregulated genes and decreased in the downregulated genes (Fig. S5e). The loop sizes also reduced in Satb1 de cient thymocytes (Fig. 4d). Reduced loop strength in SE may decrease activity of SE. We then analyzed the loop numbers, H3K27ac levels, associated gene expressions of SEs. 80% of SEs had a reduced loop strength and 66% a reduced loop number (Fig. S5f). The loop numbers and loop strength were signi cantly correlated (Fig. S5f). The loop numbers and strength of SEs were correlated with H3K27ac and gene expression ( Fig. 4e and S5g to h), suggesting that Satb1 deletion impaired the organization and activity of SEs. To further validate the relationship of SATB1 binding and chromatin looping, we analyzed the loop strength of SATB1 clusters. The aggregate pile-up analysis showed the reduced loop strength of SATB1 cluster regions in Satb1 de cient thymocytes (Fig. 4f). These data indicated that SATB1 promotes internal interactions of SEs and interactions between SEs and promoters to affect gene expression by forming compacted chromatin organization. SATB1 regulated transcription factors Bcl6 and Ets2 by mediating chromatin topology. SATB1 regulates the DP signature genes, including Tcra, Rag1/2, Cd4, and Ets1, most of which have been proved to play an essential role in DP thymocytes. We also observed that some genes, like Bcl6 and Ets2, were regulated by SATB1 and super-enhancers in DP thymocytes (Fig. 1c). We further explored the regulation of the Bcl6 and Ets2 genes in DP thymocytes. B-cell lymphoma 6 protein (Bcl6) is a zinc nger transcription repressor and is found to be frequently translocated in diffuse large B cell lymohoma 36-39 .
It is a master transcription factor for the differentiation of Follicular Helper T cells (Tfh) 40 . ETS protooncogene 2 (Ets2) belongs to the ETS family of transcription factors and is involved in stem cell development, cell senescence and death, tumorigenesis, and thymocyte development [41][42][43][44][45] . These two genes were reported involved in thymocyte development, but their regulations remain elusive 43,46 .
We observed that Satb1 occupied the loci of the Bcl6 and Ets2 genes in DP thymocytes ( Fig. 5a and 5b). There are super-enhancers located at the Bcl6 upstream and Ets2 downstream, respectively ( Fig. 5a and 5b). These two super-enhancers span more than 100kb and were characterized of H3K27 acetylation, Satb1, cohesin, and CTCF binding ( Fig. 5a and 5b). More important, the super-enhancers and Bcl6 or Ets2 are in a topology associating domain (TAD) or sub-TAD. The Hi-C data showed that the chromatin interactions in the two loci, including super-enhancers and promoters, decreased dramatically in Satb1 de cient DP thymocytes (Fig. 5a, 5b, S6a, S6b, and S6c). 3C-HTGTS data revealed that Bcl6 or Ets2 promoter had strong interactions with the super-enhancer in WT DP thymocytes and the interactions reduced in Satb1 de cient cells ( Fig. 5c and 5d). The H3K27 acetylation of the super-enhancers and promoters also reduced dramatically ( Fig. 5c and 5d), indicating the reduced chromatin interactions the activities of super-enhancers and promoters. RNA-seq and qPCR data showed the reduced expression of Bcl6 and Ets2 in Satb1 de cient thymocytes (Fig. 1c, 5e, and S1c). Analysis of Bcl6 and Ets2 expression during thymocyte development showed that these two genes expression peaks are in DP stages, a similar pattern as Satb1 (Fig. 5f). These results suggested that the speci cally high expressions of Bcl6 and Ets2 in DP thymocytes were regulated by super-enhancers and SATB1.

Super-enhancer regulates Ets2 expression in the thymus
To explore the role of the super-enhancer in Ets2 (named as Ets2-SE) expression in thymocyte, we generated Ets2-SE knockout mice in which the 166kb region containing the Ets2-SE was deleted. The Ets2-SE deletion dramatically reduced the Ets2 expression in thymocytes (Fig. 6a). The 4C assay showed that the Ets2 promoter had much less interactions with the whole locus including upstream and downstream regions (Fig. 6b). The result indicated that the SE regulates Ets2 expression in thymocytes.
Then we analyzed the thymocyte development of the Ets2-SE deleted mice. The cell numbers of thymi reduced by 69% (n=7) in the Ets2-SE −/− mice (Fig. 6c). Flow cytometric analysis showed that Ets2-SE deletion slightly increased the percentage of CD4 − CD8 − double negative 4 (DN4) thymocytes and reduced DP thymocytes (Fig. 6d), indicating the defective transition from DN to DP. We analyzed the cell viability of thymocytes in culture and the result showed that Ets2-SE deleted thymocytes had a shorter lifespan (Fig. 6e), which might explain the defective development. It was reported that short lifespan of DP thymocytes caused impaired Tcra rearrangement 9, 10 . We analyzed Tcra rearrangement using a single primer pair targeting C region of the Tcra gene during 5'rapid ampli cation of cDNA ends (5' RACE) 47 .
MiXCR immune repertoire analysis program was used for Jα and Vα usage 48 . The Ets2-SE deletion didn't affect the Jα and Vα usage ( Fig. S7a-b). These results indicated that the super-enhancer controls Ets2 expression in thymocytes, which plays a role in the DN-to-DP transition and DP lifespan during thymocyte development.
The Bcl6-SE regulated Bcl6 expression and Tcra rearrangement in thymocytes To con rm the role of the Bcl6-SE in Bcl6 regulation in thymocytes, we deleted a 119.2kb region (chr16:24146914-24266171) containing the Bcl-SE in mice. The deletion reduced the Bcl6 expression around fty folds in thymocytes and dramatically changed the chromatin conformation of the locus (Fig. 7a). The cell numbers of thymocytes decreased in Bcl6-SE homozygous mice (Fig. 7b). The proportion of DN, DP, and SP populations was not affected by the Bcl6-SE deletion (Fig. 7c). Within the DN population, the percentage of DN3 was increased signi cantly (Fig. 7c), which is consistent with the observation in the conditional Bcl6 knockout mice with a lck-cre transgene 46 . We also analyzed T cells in spleen, mesenteric lymph nodes, inguinal lymph nodes, and auxiliary lymph nodes ( Fig. S8a-b). We only observed the increased ratio of CD4 + /CD8 + T lymphocytes in inguinal lymph nodes of Bcl6-SE deleted mice ( Fig. S8a-b). The Bcl6-SE deletion didn't affect cell lifespan (Fig. 7d).
We also detected Tcra rearrangement using 5′ RACE sequencing. We noticed that the usages of the proximal Vα genes increased slightly and the usages of the distal Vα genes reduced in thymocytes of Bcl6-SE mice (Fig. S8c). To show the difference more clearly, we combined Vα genes into ve groups: the proximal, the proximal repeats, the central repeats, the distal repeats, and the distal. The usages of the proximal and the proximal repeat increased and the central repeat, the distal repeat, and the distal decreased (Fig. 7e). Consistent with the abnormal usage of the Vα genes, the Jα usage data showed an abnormal pattern with increased 5' Jα and reduced 3' Jα usage in the Bcl6-SE deleted thymocytes (Fig. 7f). Taken together, the results indicated that the super-enhancer regulates Bcl6 expression and plays a role in normal T cell development and Tcra rearrangement.

Discussion
DP cells are at a critical stage of T cell development, which mainly undergo two biological processes: 1) generate highly diverse T cell receptors through Tcra rearrangement and performing positive and negative selection simultaneously; 2) determine the direction of differentiation according to the TCR signal, producing CD4 + SP, CD8 + SP, T reg , and iNKT, respectively. Some transcription factors such as TCF-1 49, 50 , E proteins 51 , c-Myb 52, 53 , and RORγt 9 , are involved in regulating DP cells. However, most of the factors only participate in one of biological processes, rather than acting as a master regulator. SATB1 regulates Tcra rearrangement 6 , positive and negative selection 27 , and lineage decision in DP cells 25 , which makes it as a good candidate of the master regulator of DP thymocytes. Here we provided evidence that SATB1 controls the DP cell identity in a single cell transcription pro le, although SATB1 de cient DP cells still highly express CD4 and CD8. Furthermore, the regulatory effect of SATB1 on DP cell identity is speci c because SATB1 deletion does not change the transcription programs of thymocytes at other stages such as DN2/3 and CD4 + /CD8 + SP.
Since the concept of super-enhancer was proposed, many studies have supported the role of superenhancers in regulating cell identity genes 32,54 . Our data also showed that super-enhancers control genes involved in Tcra rearrangement and positive/negative selection. It was reported that SATB1 acts as a pioneer molecular in establishing T reg cell-speci c super-enhancers 28 . However, SATB1 has a high occupancy at super-enhancer regions and mediates interactions within super-enhancers and between super-enhancers and promoters in DP cells, suggesting that SATB1 regulates DP cell identity genes by reorganizing super-enhancers. SATB1 regulates many genes related to DP thymocyte function, including Rag1, Tcra, Cd4, Cd8a etc. Here we showed that transcription factors Bcl6 and Ets2 are controlled by SATB1 and super-enhancers in DP thymocytes. Super-enhancer knockout mice con rmed that the high expressions of Bcl6 and Ets2 in DP cells play an essential role in DP cells. The research on the dominant-negative truncated Ets2 transgenic mice and a phosphomutant Ets2 (T72A) transgenic mice showed that Ets2 plays an essential role of Ets2 in thymocyte development 43,44 . In this study, Ets2-SE mice have a similar phenotype, indicating that Ets2 high expression regulated by the super-enhancer is critical for the development and survival of DP cells. Bcl6 is a transcriptional repressor that plays an essential role in the germinal center response and is also involved in leukemogenesis 39,40 . Recent studies have shown that conditional deletion of Bcl6 lead to defetive differentiation of DN to DP and abnormal activation of Notch signaling in DP cells 46 . Consistent with the previous report, Bcl6-SE knockout mice also displayed reduced cell number of thymocytes and defective transition from DN to DP. We also noticed that Tcra rearrangement was impaired, and the mechanism remains elusive. These results support the notion that SATB1 orchestrates DP thymocyte function-related genes through reorganizing super-enhancers.
In summary, this study explored the mechanism by which SATB1 controls the cell identity of DP thymocytes and provides evidence that SATB1 promotes the intra-interactions of the super-enhancers, augments the super-enhancer activity, and then enables the high expression of cell identity genes. Thus, SATB1 maintains the cell identity of DP cells and ensures the normal development of thymocytes.

Mouse
Satb1 / vav-cre + mice were generated as previously described 6 , and used in this study as Satb1cKO mice. The Bcl6 SE −/− and ETS2 SE −/− mice were generated using CRSIPR-Cas9 system by Beijing Vitalstar Biotechnology. The deleted regions were chr16: 24146828-24266085 and chr16: 95745432-95912361, respectively. All experiments involving mice were performed using protocols approved by Southern Medical University Animal Studies Committee. Animals were housed and bred in a speci c pathogen-free animal facility.
scRNA-seq Library Construction and data processing For each sample, the cleaned data was generated by Cell Ranger (v2.2.0) and ltered for the low-quality reads and unrelated sequence. The data was aligned to mouse mm10 reference genome. Data merging, thresholding, normalization, principal component analysis, clustering analysis, visualization, differential gene expression analysis, and cell cycle phases analysis were carried out in Seurat (v2.3.4) according to their recommended steps (https://satijalab.org/seurat). In details, cells were sorted based on the barcodes and the unique molecular identi ers (UMIs) were counted per gene for each cell. In total, 8872-9283 (averagely 9077) cells were captured for individual libraries, and 1161-1621 (averagely 1391) genes were detected with UMIs per cell. Cells having total mitochondria-expressed genes beyond 10% were eliminated, along with cells expressing less than 500 or greater than 3000 total genes. After this, we performed global normalization using the SCTransform function in Seurat Seurat (v2.3.4). These preprocessed data were then analyzed to identify variable genes and principal component analysis.
For further analysis, UMAP were used for dimensionality reduction. Cells were represented in a twodimensional UMAP plane, and clusters were identi ed and annotated according to the previously published canonical immune markers. The cell cycle phase score was calculated for each cell using the Seurat function CellCycleScoring. Signi cance of differential expression was calculated using the Wilcoxon rank-sum test.
RNA isolation and bulk RNA-Seq.
The sorted DP cells were isolated using a TRIzol. The total RNA was quanti ed and quali ed by Agilent 2100 Bioanalyzer and NanoDrop2000. 1µg of total RNA was used for following library preparation. The poly(A) mRNA isolation was performed using Poly(A) mRNA Magnetic Isolation Module. First strand cDNA was synthesized using ProtoScript II Reverse Transcriptase and the second-strand cDNA was synthesized using Second Strand Synthesis Enzyme Mix (New England Biolabs). The puri ed doublestranded cDNA was subjected to end repair, 3'-dA tailing, and adapter ligation. Size selection of adapter ligated DNA were performed prior to PCR ampli cation. The ligated DNA was then ampli ed by 10-15 cycles with Illumina P5/P7 primers. The PCR products were cleaned up using AMPure XP beads (Beckman Coulter), validated using an Qsep100 and quanti ed by Qubit3.0 Fluorometer (Invitrogen). Libraries were a 2×150bp paired end (PE) sequenced on a NextSeq 550 on an Illumina HiSeq instrument. Three biological replicates were performed in Satb1WT and Satb1cKO DP T cells.
Differential expression genes analysis was performed using DESeq2 (v1.30.0) with default setting. The increased genes were de ned as log2FC > 1 and adjusted p < 0.05, and the decreased genes as log2FC < −1 and adjusted p < 0.05. To verify the rationality of the DEGs, the volcano plot and heatmaps were draw using ggplot2 package in R. To cluster the samples and calculate the correlation coe cients between the samples, a Spearman correlation test was applied, and the results was visualized using R package pheatmap.
Using WebGestalt (http://www.webgestalt.org) database on DEG set to obtain all Gene Ontology (GO) terms and KEGG pathways, accompanied by number of genes in that GO-term and pathway, enriched pvalue, and FDR. Only the GO-term and pathway with a FDR value < 0.05 were considered as signi cantly enriched.
RNA-seq data from mouse T cell precursors in different developmental stages including DN1, DN2a, DN2b, DN3, and DP, CD4 + SP, CD8 + SP (Gene Expression Omnibus accession: GSE109125) were used to create DN1 and DN3 gene sets. Of the RNA-seq dataset, genes that were differentially up-regulated (P < 0.05 and Log2 fold change (FC) > 1) between DN1 versus DP, DN3 versus DP were used as gene sets for GSEA. GSEA was run on all expressed WT_DP versus Satb1cKO_DP RNA-seq genes, which were ranked by log2 FC value.

ChIP-seq
ChIP-seq data were either generated in this study or downloaded from public resource. The raw data were processed and analyzed according to the following procedure. First, SRA les were converted to fastq format, then aligned to mouse genome (mm10) using Bowtie2 (v2. 3

Identi cation of super-enhancers
Super-enhancers (SEs) were identi ed using the rank ordering of super-enhancers (ROSE) algorithm (http://younglab.wi.mit.edu/super_enhancer_code.html). H3K27ac peaks were used to de ne enhancers, followed by further ltering based on the criteria: brie y, peaks located within ±3kb region of TSSs were excluded. The remaining H3K27ac peaks were de ned as putative enhancers. Enhancers located within ±12.5 kb regions of each other were stitched together, scored, and ranked based on H3K27ac ChIP-Seq signals. Enhancers were plotted with enhancer rank versus enhancer density, and all enhancer regions above the in ection point of the curve were de ned as SEs. An analogous procedure was used to de ne SE regions by enrichment of Satb1 at enhancers. Super-enhancers and typical enhancers were assigned to the genes using the default parameters of the ROSE algorithm. ChIP-seq signal at SE regions were plotted as averaged pro les in ngsplot (v2.41). GSEA was used to identify how SE-gene sets distribute in gene lists ranked by either gene expression fold change values or H3K27ac ChIP-seq enrichment on enhancers. GO analysis for SE was performed as previous described.

Hi-C
We performed in situ Hi-C with 5-10 million cells. Cells were crosslinked at a nal concentration of 1% formaldehydeand in ice bath for 10min and quenched by 200mM glycine for 5min at room temperature.
A/B compartment analysis was performed at 250-kb resolution using a publicly available script (matrix2compartment.pl). The script can be accessed through GitHub The TAD structure (insulation/boundaries) was de ned by the insulation score. The matrices which were used to calculate the insulation score were normalized by ICE method for discarding the bias of raw matrices. And then, insulation score was computed at 10-kb resolution. A publicly available script (matrix2insulation.pl) was used to detect TAD boundaries, with the following options: '--is 600000 --ids 250000 --im mean --bmoe 3'. TADs were called using normalized Hi-C matrices at 10-kb resolution with insulation2tads.pl. The script can be accessed through GitHub (https://github.com/dekkerlab/cworlddekker).
We generated a chromatin interaction matrix with a 50 kb resolution using the Hi-C data, and the differential chromatin interactions between Satb1WT and Satb1CKO DP cells were calculated using the Bioconductor package multilHiCcompare (v1.8.0), which provides functions for the joint normalization and detection of differential chromatin interactions in our two Satb1WT and two Satb1CKO replicates experiments. To identify the signi cantly differential chromatin interactions, we used an adjusted p-value of 0.05 and a log fold change of 0. The chromosomal sequencing was divided into 50 kb bins for the normalized signal le of the Satb1WT ChIP-Seq. Once the bin attained by ChIP-Seq were located within the differential chromatin interactions, and the value in the corresponding bin of the Hi-C matrix indicated the normalized count.
Chromatin loops were called using Fit-Hi-C (v2.0.7). Input les of Fit-Hi-C were created by using a publicly available script (hicpro2 thic.py) from HiC-Pro. Next, loops were called using xed-size bin resolutions from 10 to 20 kb in both cell types. Brie y, signi cant interaction loops (p <= 0.05 and contact frequencies >= 5) were identi ed through jointly modeling the contact probability using raw contact frequencies and ICE normalization vectors with the Fit-Hi-C algorithm. Enhancer-promoter loops were annotated using pgltools (v2.2.0). The anchors of loops were intersected with promoters and enhancers. Promoters were de ned as ±5 kb windows of the TSS of all expressed genes, enhancers are de ned using enhancer dataset of DP cells (http://www.enhanceratlas.org/downloadv2.php). Super-enhancer loops were de ned by H3K27ac-de nded SE sets obtained through ROSE software.
The format of Hi-C data, .hic, was converted into .cool les using hic2cool (v0.8.3) software (https://github.com/4dn-dcic/hic2cool). Hi-C matrices in cool format were used to generate genome-wide aggregate plots at TADs and loops of Satb1WT and Satb1cKO DP cells detected by Hi-C. We used coolpup.py (v0.9.5) to pile-up normalized Hi-C signals at a 10-kb resolution at loops previously identi ed and plotted 500-kb upstream and downstream of the loop anchor coordinates (https://github.com/open2c/coolpuppy). The local rescaled pileups of TADs annotated using insulation score valleys used above in DP cells were analyzed at a 10-kb resolution. We plotted them using plotpup.py (v0.9.5).

4C
Chromatin was crosslinked for 10 minutes at room temperature in 1X PBS/10% FBS containing 4% formaldehyde. Crosslinking was blocked by glycine addition. Following cell lysis, nuclei pellets were resuspended in 1.2X Buffer2 (NEB), followed by SDS addition. The samples were incubated while shaking for 60 minutes at 37°C. TritonX-100 was added to quench SDS. Crosslinked chromatin was primarily digested using MboI. Ligation was performed in the presence of T4 DNA Ligase in diluted conditions. Chromatin reverse crosslinking was carried out using proteinase K, phenol-chloroform and DNA were puri ed by ethanol precipitation. Secondary restriction was performed using NlaIII. Secondary ligation was carried out with T4 DNA Ligase in diluted conditions. Ligated DNA products were then extracted using phenol-chloroform with ethanol precipitation and puri ed.4C-seq library preparation was achieved by inverse PCR using 100 ng template DNA (for 10 reactions in total). The viewpoint-directed inverse PCR primers carrying Illumina P5 or P7 sequencing adapters as below: Ets2 forward primer: 3C-HTGTS fastq data were ltered by removing adaptor and low-quality reads with using fastp (v0.20.0 ).
Filtered reads were extracted from the sequence le after quality control through Cutadapt (v1.18), Pear (v0.9.6), etc..Pair-end reads containing NestPrimer or AdapterPrimer were obtained, and then the reads were ltered by searching restriction sites sequences. The remaining reads were mapping to mouse genome mm10 with bowtie2(v2.3.5.1). The mapping reads were ltered the duplicated reads, self-ligation reads, relegation reads and dumped reads .For visualization, we convert the nal bam les into bedGraph le. The signal peak bedGraph le was obtained by post-comparison ltering, signal statistics, and standardization. We normalized bedGraph le using CPM(Counts Per Million in cis)normalization method, and visualized on IGV. Finally, we organized the results report and visualized it with R.

FACS analysis
Thymus, spleen, mesenteric lymph nodes, inguinal lymph nodes, and auxiliary lymph nodes from 6-8 weeks mice were ground in MACS buffer (1×PBS, 0.5% BSA, 2mM EDTA) and ltered with 40 um nylon mesh. Red blood cells were lysed in RBC buffer (Biolegend, USA) for 10 minutes at room temperature.
FACS analysis was performed on the BD FACS Aria3 machine. DP cells were de ned as CD4 + CD8 + . Cells were gated on CD4 − CD8 − and DN cells were de ned as followed: DN1 CD44 + CD25 − , DN2 CD44 + CD25 + , DN3 CD44 − CD25 + , and DN4 CD44 − CD25 − . The B cells were de ned by the expression of CD19 and the T lymphocytes were de ned by the CD3 expression. Cells from the spleen, mesenteric lymph nodes, inguinal lymph nodes, and auxiliary lymph nodes were gated with CD3 and the mature T cells were de ned as CD4 + or CD8 + . For Bcl6 staining, cells were then xed and permeabilized using True-Nuclear Transcription Factor Buffer Set (Biolegend, USA) after cell surface staining.

5' RACE sequencing
The 1ug RNA was used for the reverse transcription (RT The 5'RACE raw data were ltered by fastp and the adapter sequences, contamination, and low-quality reads were removed. T cell receptor beta chain V, D, and J gene identi cation, CDR3 sequence extraction in clean reads were performed using MiXCR. The corresponding germline sequences were mapped to the reference sequences derived from international ImMunoGeneTics (IMGT) information.   during the differentiation from DN1 to SP thymocytes. **** p value < 0.0001 by Kruskal-Wallis test, followed by Dunn's multiple comparisons test. e) Gene ontology analysis on the genes associated with super-enhancers by enrichment for H3K27ac or Satb1. The top seven biological processes ranked by signi cance (-log(q-value)) are depicted.  loci. The 3C-HTGTS bait is at the Bcl6 or Ets2 promoters. e) Relative expression of Bcl6 and Est2 in Satb1WT and Satb1cKO DP thymocytes detected by reverse-transcripted quantitative PCR. The expressions are normalized to the Actb gene and then to WT. The data represent mean ± SD of three experiments. * p value < 0.05, ** p value < 0.01 by two side Student's t test. f) The expression pro les of Satb1, Bcl6, and Ets2 during the differentiation of DN1 into DP thymocytes.  and homozygous total thymocytes. Results are the mean ± SD of ve mice for each genotype. ** p < 0.01 by two side Student's t test. c) Flow cytometry of thymocytes from WT and Bcl6 SE-/-mice. Right, average frequency of thymocyte subsets. Results are the mean ± SD of three WT and KO mice. *** p < 0.001 by two side Student's t test. d) Thymocytes were cultured from 0 to 48 hours in RPMI 1640 medium with 10% FBS at 5% CO2 and 37 °C. MTS assay was performed to detect the cell viabilities. Each experiment was repeated three times and the data are the mean ± SD of three WT and KO mice. e) Relative Vα and f) Jα usages were determined by deep-sequencing of Tcra transcripts ampli ed by 5'RACE of WT and Bcl6 SE-/-thymocytes, respectively. The relative Vα or Jα usages were calculated by dividing the number of the clonotypes containing the Vα or Jα genes by the total clonotype number. The Vα usage is the sum of the usage frequency of all Vα genes in the region.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. TableS1DEGinSatb1cKODP.xlsx TableS2H3K27acsuperenhancer.xlsx TableS3Satb1cluster.xlsx 2021FengDLSupFig.docx