GATA3-Controlled Nucleosome Eviction Drives MYC Enhancer Activity in T-cell Development and Leukemia

Enhancers are long-range, orientation-independent, cis-acting DNA-regulatory elements that control gene expression through physical interaction with proximal regulatory elements located at gene promoters (1–3). Temporal and spatial transcriptional regulation of key developmental factors is frequently coordinated by clusters of distal enhancers organized in regulatory domains (4, 5). Active enhancers competent for transcription factor binding and transcriptional regulation show low nucleosome occupancy (6, 7), and enhancers that work simultaneously often display coordinated patterns of DNA accessibility, whereas those that work in mutually exclusive modes show divergent chromatin accessibility profiles (8).

Constitutive activation of NOTCH1 signaling plays a prominent driver role in more than 60% of T-cell acute lymphoblastic leukemias (T-ALL) harboring activating mutations in the NOTCH1 gene (9). Oncogenic NOTCH1 drives T-cell transformation, activating a broad transcriptional program that promotes leukemia cell growth and proliferation. Most prominently, NOTCH1 directly activates MYC expression, and NOTCH1 and MYC share multiple common direct target genes driving leukemia cell growth in T-ALL (10). Consistently, the NOTCH1-MYC enhancer (N-Me), a NOTCH1-controlled T cell–specific MYC long-range enhancer, is strictly required for NOTCH1-induced T-ALL (11). Notably, although activating mutations in NOTCH1 are also found in adenoid cystic carcinoma (12, 13), chronic lymphocytic leukemia (14), and mantle cell lymphomas (15), N-Me seems to be selectively active only during early T-cell development and in T-ALL (11). This observation supports that yet unrecognized T cell–specific signaling, transcriptional or epigenetic factors epistatic with NOTCH1 signaling are dominantly required for N-Me enhancer activity and may contribute to leukemic transformation.

T-cell precursors follow an orchestrated developmental program that begins with double negative 1 (DN1) cells, the earliest cell entrants in the thymus, and progresses to uncommitted DN2a progenitors, which become T-cell committed as they mature into DN2b cells (16). These early precursors subsequently progress through highly proliferative DN3, DN4, and intermediate single positive (ISP) thymocyte stages, which then exit the cell cycle as they mature into double positive (DP) and ultimately mature single positive CD4⁺ (CD4SP) and CD8⁺ (CD8SP) T cells (16). Analysis of chromatin accessibility by Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) in sorted mouse thymocyte precursors identified 69,302 highly accessible regions. Most of these correspond to gene bodies (33,294; 51.8%) and intergenic regions (26,947; 38.8%), and only a fraction reside in gene promoters (9,061; 13%). Interestingly, however, an increased representation of intergenic regions (3,194; 46%; P = 2^–28) and decreased frequency of promoters (144; 2%; P = 4.8^–148) is observed in ATAC-seq regions that display variable accessibility through T-cell development stages, suggesting that dynamic control of accessibility at distal regulatory elements may influence thymocyte development. Hierarchical clustering analysis revealed distinct groups of differentially accessible regions that closely clustered thymocyte DN1 and DN2a populations, separate from DN2b and DN3 cells, and DN4, ISP, and DP thymocytes distinct from CD4SP and CD8SP populations (Fig. 1A). Consensus clustering further highlighted developmental transitions between DN1, DN2a, and DN2b cells; placed DN3 closer to the DN4, ISP, and DP thymocyte cluster; and distinguished CD4SP and CD8SP cells (Fig. 1B). In these analyses, the transition from DN1/DN2a to DN2b, which marks T-cell specification, is associated with marked loss of chromatin accessibility consistent with a restriction of transcriptional potential from uncommitted populations to T-cell progenitors (Fig. 1A). Moreover, among the four major differential chromatin accessibility developmental modules, the cluster characterized by high levels of chromatin accessibility in DN1/DN2a cells accounted for 4,763 (68%) of all differentially accessible segments (Fig. 1A). A second cluster composed of 684 (9.8%) segments showed orchestrated opening during T-cell specification in DN2b and DN3 cells (Fig. 1A). This was followed by the opening of 439 intervals (6.3%) characteristically accessible in DN4-ISP-DP populations and, subsequently, 1,044 intervals (15%) selectively open in mature CD4SP and CD8SP cells (Fig. 1A). These results demonstrate a highly dynamic chromatin remodeling landscape during thymocyte development, particularly at nonpromoter regulatory regions with discrete clusters of differentially accessible regions controlled by distinct regulatory circuitries. Consistently, transcription factor binding site analyses identified significantly enriched regulatory sites in each of these clusters with prominent representation of PU-box, GATA, RUNX, HOX, helix-loop-helix, ETS, FOX, and KRAB transcription factor binding motifs (Fig. 1C; Supplementary Table S1).

MYC, a master regulator of cell growth and proliferation in development and cancer, is transcriptionally controlled by a complex array of long-range regulatory elements with tissue- and cell type–specific enhancer activities (17). Myc expression in developing T cells is controlled by N-Me, a discrete long-range enhancer located 1.4 Mb downstream of MYC (11, 18). Given the importance of MYC expression in lymphocyte biology, we examined the regulatory logic and mechanisms responsible for dynamic N-Me regulation during thymocyte development (11, 18, 19).

Circularized Chromosome Conformation Capture (4C) analyses of NOTCH1-driven human and mouse T-ALL lymphoblasts, where the N-Me enhancer is active (11), not only confirmed the long-range interaction between the MYC proximal promoter and N-Me, but also revealed unanticipated chromatin loops connecting both the MYC promoter and the N-Me enhancer with distal elements located centromeric and telomeric from the MYC transcription start site, implying a more complex mechanism of transcriptional regulation (Fig. 2A; Supplementary Fig. S1). Chromatin immunoprecipitation sequencing (ChIP-seq) data showed binding of CTCF and MED1, two factors involved in chromatin–chromatin interactions, to the N-Me site (Fig. 2B). We also observed high densities of chromatin marks characteristic of an active enhancer configuration in the vicinity of N-Me, such as histone H3-K4 monomethylation (H3K4me1) and H3-K27 acetylation (H3K27ac; ref. 11). In addition, analysis of chromatin-associated factors revealed N-Me occupancy by BRD4, a reader of H3K27ac, and KDM6A, which acts to erase histone H3-K27 trimethylation (H3K27me3; Fig. 2B). Moreover, ChIP-seq analysis of T-ALL lymphoblasts showed that, in addition to the expected occupancy of N-Me by NOTCH1 and the RBPJ NOTCH1 DNA-binding partner (11, 18), this enhancer is also bound by numerous other transcription factors involved in hematopoietic and lymphoid development, including ERG, ETS1, GATA3, RUNX1, TCF3, and TCF12 (20, 21), and by bona fide transcription factor oncoproteins with prominent roles in T-ALL pathogenesis, such as HOXA9, MYB, MYC, LMO1, LMO2, TAL1, and TLX1 (ref. 22; Fig. 2B). In addition, mass spectrometry analysis of N-Me pulldown preparations identified 25 high-confidence N-Me–associated proteins in nuclear extracts from several T-ALL cell lines (HPB-ALL, ALL-SIL, and JURKAT), and 17 additional factors associated with N-Me in at least two of these lines (Fig. 2C). Of note, N-Me DNA pulled down developmentally important and ChIP-validated N-Me–associated transcription factors (RUNX1, GATA3, TCF3, TCF12, and MYC), as well as numerous additional DNA-binding proteins with major roles in thymocyte development (BCL11B, LEF1, RUNX3, CUX1, CBFA2T3, and IKZF1; refs. 20, 21).

The landscape of chromatin interactions and DNA-binding factors associated with the N-Me enhancer suggests that N-Me acts as a regulatory hub that receives multiple developmental and oncogenic cues to control MYC expression in T-ALL. Despite this apparent complexity, we hypothesized that N-Me activity would be governed by highly conserved cis-acting elements. Indeed, multispecies conservation analysis of N-Me sequences across vertebrates revealed a tight clustering of enhancers from placental mammals and a more distant relationship of these with the N-Me sequences of marsupials and monotremes, which clustered closer to those of reptilian and avian species (Fig. 2D). Moreover, phylogenetic footprint analyses of vertebrate N-Me enhancer sequences (Fig. 2E; Supplementary Table S2) revealed multiple highly conserved regulatory elements present in mammals, birds, and reptiles. Given the strict requirement of both Myc expression and N-Me activity for early T-cell differentiation in MYC-driven highly proliferative DN3, DN4, and ISP thymocytes (11), we tested the impact of targeted mutations at these highly conserved motifs on mouse T-cell development (Fig. 2E; Supplementary Fig. S2A; Supplementary Table S3). In these analyses, mutant mice harboring homozygous disruption of two independent highly conserved N-Me HOX–PBX motifs, a RUNX binding site, a LEF/TCF binding site, and a PAX binding site showed no clear alterations in the distributions of thymocyte populations (Supplementary Fig. S2B and S2C). Similarly, mice carrying a genetic disruption of a highly conserved RBPJ binding site showed no clear phenotypic alterations in the thymus (Supplementary Fig. S2B and S2C), despite effective abrogation of NOTCH1 binding to DNA induced by this mutation (Supplementary Fig. S2D). These results support a redundant role for multiple transcription factor binding sites in the control of N-Me activity and Myc expression in the thymus. In contrast, mutation of an ultraconserved GATA transcription factor binding motif [hereafter named GATA site 1 (GS1); Supplementary Fig. S3A] revealed an accumulation of DN2, DN3, DN4, and ISP populations with preserved DP and SP cells (P < 0.05; Fig. 3A–E; Supplementary Fig. S2B and S2C), whereas homozygous disruption of a second highly conserved GATA-binding motif [hereafter named GATA site 2 (GS2)], across all mammalian species, showed a more modest phenotype with accumulation of DN3 and ISP populations with otherwise largely preserved thymocyte development (Fig. 3A–E; Supplementary Fig. S2B and S2C). In line with these findings, N-Me-MYC promoter luciferase reporter assays showed a marked reduction in transcriptional activity in a GS1-mutant N-Me construct compared with the wild-type (WT) N-Me control, and a more moderate decrease in enhancer function in the reporter containing a mutation in the less conserved N-Me GS2 site (Fig. 4A). On the basis of these results, we generated mice homozygous for combined mutations in these two regulatory elements. Notably, N-Me GS1- and GS2-deficient mice (hereafter named GS1+2^mut) showed a marked early T-cell developmental defect with small thymi and a dramatic reduction in thymocyte numbers (Fig. 3A and B) accompanied by accumulation of DN3, DN4, and ISP cells, and decreased numbers of DP and SP CD4⁺ and CD8⁺ thymocytes (Fig. 3C–E). Lack of apoptosis in this model supports a developmental block phenotype (Supplementary Fig. S3B). Moreover, analysis of spleen and lymph nodes from GS1+2^mut mice showed a consistent reduction in the numbers of mature CD4⁺ and CD8⁺ T-cell populations (Supplementary Fig. S3C and S3D). These results support a cooperative and partially redundant role of the GS1 and GS2 N-Me GATA sites in the control of N-Me activity and Myc expression during thymocyte development.

To test this hypothesis, we performed RNA sequencing (RNA-seq) analysis of WT and GS1+2 mutant DN3 thymocytes. These analyses revealed an effective downregulation of Myc messenger RNA and MYC-controlled gene expression programs in GS1+2 mutant DN3 cells compared with WT controls (Fig. 4B and C). Moreover, single-cell RNA-seq analyses showed decreased numbers of Myc-expressing cells in the thymus of homozygous N-Me GS1+2^mut animals (Fig. 4D and E). Similarly, IHC and flow-cytometry analysis revealed markedly reduced numbers of MYC-positive cells in the thymi of N-Me GS1+2 homozygous–mutant mice, and reduced MYC levels in the DN4 and ISP populations compared with WT controls (Fig. 4F and G).

Next, we evaluated the impact of N-Me GS1+2 mutations in T-ALL transformation. Toward this goal, we infected hematopoietic progenitors from N-Me WT (N-Me^+/+), N-Me GS1+2 heterozygous (N-Me^+/GS1+2), and N-Me GS1+2 homozygous (N-Me^GS1+2/GS1+2) animals with retroviruses driving the expression of an oncogenic constitutively active form of NOTCH1 (ΔE-NOTCH1), which specifically induces development of MYC-driven T-ALL in mice, and transplanted them into sublethally irradiated recipients (11, 23). Consistent with previous reports (11), mice transplanted with ΔE-NOTCH1–infected N-Me WT cells showed a transient wave of CD4⁺ CD8⁺ DP cells in peripheral blood at 21 days post-transplant (Fig. 5A) and developed overt T-ALL with a median latency of 8 weeks (Fig. 5B and C). In contrast, mice transplanted with ΔE-NOTCH1–expressing N-Me GS1+2^mut heterozygous progenitors showed a blunted DP wave and impaired tumor development (P < 0.01; Fig. 5A–C), whereas animals transplanted with ΔE-NOTCH1 N-Me GS1+2 homozygous–mutant cells remained leukemia-free 100 days post-transplant (P < 0.001; Fig. 5A–C). To explore the requirement of GATA3 binding to N-Me in tumor progression and maintenance, we analyzed the capacity of the N-Me GS1+2 mutant allele to support cell growth and proliferation in established NOTCH1-induced leukemia. To this end, we generated ΔE-NOTCH1–induced T-ALL tumors from mice expressing tamoxifen-inducible Cre recombinase (Rosa26TMCre; Rosa26^+/CreERT2) that are also compound heterozygous for the conditional N-Me knockout (N-Me^flox; ref. 11) and the N-Me GS1+2 mutant alleles (Rosa26^+/CreERT2 N-Me^{flox/GS1+2mut}). Following leukemia development, we injected T-ALL cells into secondary recipients and then treated these animals with vehicle only (N-Me^{flox/GS1+2mut} group) or tamoxifen, to induce Cre-mediated deletion of the N-Me^flox allele (N-Me^–/GS1+2mut group). Vehicle-treated mice bearing compound heterozygous N-Me GS1+2 leukemia cells (N-Me^{flox/GS1+2mut}) developed overt leukemia and died of disease with a median survival of 33 days. In contrast, tamoxifen-treated animals harboring isogenic hemizygous N-Me GS1+2 mutant cells (N-Me^–/GS1+2mut) showed markedly impaired tumor progression with a median survival of 53 days (P < 0.0001; Fig. 5D). Consistently, analysis of Rosa26^+/CreERT2 N-Me^{flox/GS1+2mut} leukemia lymphoblasts treated with vehicle only (N-Me^{flox/GS1+2mut}) or tamoxifen (N-Me^–/GS1+2mut) showed reduced Myc expression and markedly impaired growth, proliferation, and survival in N-Me GS1+2 hemizygous leukemia cells (tamoxifen treated, N-Me^–/GS1+2mut), compared with isogenic N-Me GS1+2 heterozygous (vehicle treated, N-Me^{flox/GS1+2mut}) controls (Fig. 5E–H). In all, these results demonstrate a strict requirement for N-Me GATA site-mediated enhancer activity in the pathogenesis of T-ALL.

GATA3, a GATA-binding transcription factor upregulated during early stages of thymocyte maturation and critically implicated in T-cell development (20, 21, 24–26), prominently interacts with the N-Me enhancer (Fig. 2B and C). To evaluate the potential role of GATA3 as a driver of N-Me activity and effector factor mediating N-Me GATA site function, we analyzed the representation of WT and GATA site mutant sequences in GATA3 ChIP DNA isolated from heterozygous GS1+2^mut DN3 thymocytes (Fig. 6A). In these experiments, GATA3 ChIP DNA preparations contained only WT N-Me sequences in support of complete disruption of GATA3 binding to GS1+2^mut N-Me (Fig. 6A). In contrast, analysis of GATA3 ChIPs from heterozygous GS1^mut and GS2^mut DN3 cells showed only a partial reduction in GATA3 binding to N-Me single GATA site mutant DNA (Fig. 6A), a result in agreement with the observed partial redundancy of the GS1 and GS2 GATA motifs in Myc regulation and thymocyte development.

Active enhancers characteristically correspond to regions of low nucleosome occupancy (7, 27, 28) flanked by areas of high density of nucleosomes containing H3K4me1 and H3K27ac (29, 30). To evaluate the potential impact of N-Me GATA site mutations and impaired GATA3 occupancy in the epigenetic landscape surrounding the N-Me enhancer, we performed ChIP for the H3K4me1 and H3K27ac chromatin marks in DN3 thymocytes from heterozygous GS1+2 mutant mice and evaluated the recovery of the N-Me mutant over WT enhancer by DNA sequencing as before. These analyses revealed no apparent difference in the recovery of active enhancer-associated chromatin marks in N-Me GS1+2 mutant chromosomes over WT (Fig. 6B), suggesting that GATA3 N-Me binding is not required for the establishment and maintenance of active enhancer histone marks. In addition, and to further explore the mechanisms responsible for the loss of effective N-Me–mediated transcriptional activity in N-Me GATA3-binding–deficient thymocytes, we evaluated the effect of GATA site mutations in the establishment and maintenance of N-Me-Myc promoter chromatin loops by interphase FISH using DNA probes mapping to the Myc promoter region and the N-Me enhancer. Analysis of N-Me WT cells showed probe colocalization indicative of effective interaction between N-Me and the Myc promoter in DN3 thymocytes, but not in B cells, where the N-Me enhancer is not active (Fig. 6C). In addition, and most notably, we observed a similar pattern of probe colocalization in DN3 cells from homozygous N-Me GS1+2 mutant mice, ruling out a role for GATA3 N-Me binding in N-Me enhancer–Myc promoter long-range chromatin looping (Fig. 6C). To further explore a potential role of changes in chromatin looping in the loss of N-Me activity in GS1+2–mutant cells, and to rule out a potential confounding effect of secondary effects derived from low MYC expression in GS1+2 mutant thymocytes, we crossed N-Me GS1+2 mutant mice with tamoxifen-inducible Rosa26-Cre-ERT2 and Rosa26-LSL-Myc knock-in animals, generating in this way Rosa26-Cre-ERT2 Rosa26-LSL-Myc N-Me^GS1+2/GS1+2 mice, which are defective in GATA3 N-Me binding, but have the capacity to express Myc ectopically from the Rosa26 locus after tamoxifen treatment. Hematopoietic progenitors from this model were infected with oncogenic ΔE-NOTCH1–expressing retroviruses and treated with tamoxifen to generate T-ALL N-Me GS1+2 mutant Rosa26-Myc tumors. Chromatin configuration by Chromosome Conformation Capture (3C) analysis in N-Me GS1+2 Rosa26-Myc T-ALL lymphoblasts and N-Me WT T-ALL controls confirmed effective interaction between the N-Me enhancer and Myc promoter sequences in GS1+2 mutant cells (Fig. 6D).

Given the broad and dynamic changes in enhancer accessibility observed during thymocyte development, we evaluated whether regulation of chromatin accessibility at the N-Me enhancer could function in the control MYC expression. Analysis of ATAC-seq data obtained at different stages of thymocyte development revealed a closed N-Me enhancer configuration in early DN1 and DN2a progenitors; acquisition of an open chromatin conformation in DN2b cells; high levels of chromatin accessibility in Myc-expressing DN3, DN4, and ISP populations; and a closed chromatin configuration in DP thymocytes and mature single positive CD4⁺ and CD8⁺ T cells, which express low levels of Myc (ref. 31; Fig. 7A and B). These results place N-Me as part of the cluster of regulatory sites gaining accessibility as cells commit to the T-cell lineage, which is characteristically enriched in GATA binding motifs (239/684; 35%), suggesting that N-Me accessibility in the thymus could be controlled by a GATA transcription factor. To test this hypothesis, we performed ATAC-seq in sorted WT and homozygous N-Me GS1+2 mutant DN3 cells. These analyses revealed a marked decrease in chromatin accessibility in N-Me GS1+2 mutant thymocytes compared with WT controls (Fig. 7C). Moreover, nucleosome position analysis demonstrated a nucleosome exclusion area at the N-Me enhancer flanked by two regions of prominent nucleosome occupancy in WT DN3 thymocytes (Fig. 7D). In contrast, and consistent with decreased chromatin accessibility, N-Me GS1+2 mutant DN3 cells showed prominent nucleosome invasion and a consequent marked reduction in the nucleosome-free region surrounding the GS1 and GS2 GATA sites (Fig. 7D). Notably, ATAC-seq analyses in N-Me GS1+2 mutant Rosa26-Myc cells show similar loss of chromatin accessibility and nucleosome invasion of the N-Me enhancer (Supplementary Fig. S4A and S4B), supporting that loss of enhancer accessibility at the N-Me site is linked to loss of GATA3 binding and not an indirect effect resulting of decreased MYC expression. Moreover, GATA3 protein enhanced DNAse I digestion of N-Me histone H1-compacted nucleosome arrays (32), further supporting a role as pioneer factor promoting N-Me enhancer accessibility (Supplementary Fig. S4C and S4D).

Mechanistically, abrogation of GATA3 binding in N-Me GS1+2 mutant cells resulted in loss of N-Me binding by the SMARCA4 SWI/SNF core factor, a result consistent with a role of GATA3 in promoting enhancer chromatin opening via the recruitment of chromatin remodeling complexes implicated in nucleosome repositioning (Fig. 7E). In this context, we predicted that nucleosome invasion could result in broad abrogation of transcription factor binding to N-Me. ChIP analyses in GS1+2 mutant Rosa26-Myc T-ALL cells revealed that GATA site mutations impaired transcription factor binding for RUNX1, ETS1, NOTCH1, TCF1, TCF3, and TCF12 in support of a defect in multi–transcription factor combinatorial regulation at this enhancer (Fig. 7F). In all, these results demonstrate a driving role for GATA3 binding in promoting N-Me enhancer nucleosome eviction, and a distinct and strict requirement for enhancer accessibility in the control of N-Me enhancer activity and Myc expression in immature T cells.

Distinct clinicobiological groups of T-ALL show parallels in their immunophenotypes and gene-expression signatures with those of immature thymocytes at different stages of development, supporting broadly common developmental and transcriptional regulatory circuitries between tumor cells and their normal cell counterparts (33). However, analysis of mouse enhancer accessibility at the N-Me enhancer in normal DP thymocytes and in DP T-ALL cells showed a closed enhancer configuration in normal DP thymocytes and an open conformation in DP T-ALL lymphoblasts (Fig. 7G). ATAC-seq analysis also revealed markedly divergent profiles in human normal and leukemic DP cells, with normal DP thymocytes showing a closed chromatin N-Me profile and human DP T-ALL lymphoblasts displaying high levels of chromatin enhancer accessibility (Fig. 7H). These results support a potential oncogenic role for deregulated N-Me enhancer accessibility as a driver of MYC expression in T-ALL.

Enhancers represent distal transcription regulatory elements with high density of transcription factor binding sites and the capacity to interact with proximal regulatory elements in the vicinity of transcription start sites via long-range chromatin loops (34, 35). The activity of enhancers is dependent on multiple transcription factors with additive and cooperative roles in transcription regulation, which together establish a combinatorial logic responsible for tuning transcriptional control in response to developmental and signaling cues (36). Transcription factor binding motif redundancy and overlapping control by multiple transcription factors are considered key enhancer features responsible for robust and fine-tuned transcriptional regulation (37). In addition, enhancer activity depends on chromatin looping and long-range chromatin interactions and is closely associated with the presence of a distinct pattern of chromatin marks with low levels of H3-K27 trimethylation and high density of H3-K4 monomethylation and H3-K27 acetylation (38). Moreover, active enhancers are highly accessible chromatin regions with highly mobile H2A.Z-containing nucleosomes (39) and low nucleosome occupancy (6). Pioneer transcription factors capable of binding to their target sequences on nucleosomal DNA and eliciting changes in local chromatin structure are particularly important to license enhancer activity during development (40).

By performing global analysis of ATAC-seq data in early T-cell progenitors, we observed dynamic changes in enhancer accessibility consistent with a prominent role of pioneer factors during thymocyte development. Moreover, the presence of distinct enrichment of transcription factor binding motifs in enhancers with coordinated opening at specific stages of thymocyte development supports the presence of a combinatorial logic in enhancer regulation. Among these highly dynamically regulated regulatory elements, N-Me, a long-range Myc enhancer, is prominently dependent on NOTCH1 signaling (11). Consistent with its role as a long-range Myc enhancer, N-Me directly interacts with Myc promoter sequences (11). However, extended mapping of chromatin interactions by 4C shows a more extensive interaction landscape involving multiple intergenic sites conserved between human and mouse of potential functional importance. Moreover, although NOTCH1 is active in multiple tissues during development, N-Me is selectively licensed and active only in early T-cell progenitors, suggesting a role for additional, T cell–specific, transcriptional regulators.

N-Me is highly conserved with multiple transcription factor binding sites commonly present among reptiles, avian species, and mammals in support of a combinatorial multitranscription factor mode of enhancer regulation. However, analysis of mice harboring deleterious mutations in multiple highly conserved transcription factor binding motifs showed no apparent defects in thymocyte development, a result consistent with transcription factor binding site redundancy and overlapping transcription factor activities in enhancer regulation. In contrast, whereas mice harboring mutations in one of two N-Me GATA3 binding sites showed only mild developmental defects, mutation of these two GATA motifs in cis resulted in abrogation of N-Me enhancer activity and marked defects in thymocyte development. These results indicate cooperation between these two GATA sites and a prominent role for GATA3 in N-Me regulation. It is worth noting that N-Me GATA3 binding motifs appeared asynchronously during evolution, as the GS1 site can be found in N-Me sequences in turtle and alligator species dating back 250 million years; whereas the GS2 site is present only in mammals and seems to have emerged early in monotreme development 170 million years ago. It has not escaped our attention that some of the developmental phenotypes in GS1+2 mutant mice are slightly less prominent than those observed in the thymus of mice harboring a full N-Me enhancer knockout (11), a result that indicates at least partial functional overlap with other N-Me–binding transcriptional regulators. Consistent with this hypothesis, reverse ChIP and ChIP-seq data indicate prominent transcription factor occupancy at the N-Me enhancer and point to this regulatory element as a major node controlling Myc expression with input from multiple developmentally important transcription factors including NOTCH1, and from T-ALL transcription factor oncogenes activated by chromosomal translocations (TAL1, LMO1, LMO2, MYB, and TLX1).

Importantly, functional characterization of mechanisms involved in loss of enhancer activity in N-Me GATA site mutant thymocytes and T-ALL cells showed loss of chromatin accessibility and enhancer nucleosome invasion, but preserved active enhancer chromatin marks and chromatin looping between N-Me and Myc promoter sequences. Loss of GATA3 binding in N-Me is linked to defective occupancy by SMARCA4, a core subunit of SWI/SNF chromatin-remodeling complexes mediating an open chromatin configuration. Moreover, nucleosome invasion secondary to GATA3 binding defects results in broad abrogation of enhancer binding by multiple other transcription factors. The model that emerges from these observations is that GATA3 binding facilitates the recruitment of SWI/SNF chromatin-remodeling complexes to N-Me and promotes nucleosome eviction in early T-cell progenitor cells, which in turn enables multi–transcription factor control of Myc expression. Consistently, GATA3 binding and N-Me enhancer accessibility are epistatic over NOTCH1 signaling in T-cell transformation. Finally, the presence of an aberrant open chromatin configuration at the N-Me enhancer in DP T-ALL cells supports a role for deregulated MYC enhancer accessibility in the pathogenesis of T-ALL.

DNA from leukemic ALL blasts were provided by the Princess Máxima Center for Pediatric Oncology tumor bank (Utrecht, the Netherlands). Normal thymus samples were obtained from Morgan Stanley Children's New York-Presbyterian Hospital (New York, NY). Written informed consent was obtained at study entry and samples were collected under the supervision of local Institutional Review Boards for participating institutions and analyzed under the supervision of the Columbia University Medical Center Institutional Review Board (protocol numbers: IRB-AAAB3250 and IRB-AAAC1660).

We performed cell culture of cell lines in standard conditions in a humidified atmosphere at 37°C under 5% CO₂. We obtained ALL-SIL (RRID: CVCL_1805), HPB-ALL (RRID: CVCL_1820), and JURKAT (RRID: CVCL_0065) T-ALL cell lines from the Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ) cell line repository. We purchased HEK293T (RRID: CVCL_006) cells from ATCC.

We cultured primary mouse lineage-negative bone marrow cells in Opti-MEM (Life Technologies 51985091) supplemented with 10% FBS, 100 U/mL penicillin G, 100 μg/mL streptomycin, 55 μmol/L β-mercaptoethanol, 10 ng/mL IL3 (PeproTech 213-13), 10 ng/mL IL6 (PeproTech 216-16), 25 ng/mL IL7 (PeproTech 217-17), 50 ng/mL SCF (PeproTech 250-03), and 50 ng/mL FLT3L (PeproTech 250-31L). Primary mouse leukemia lymphoblasts were cultured in Opti-MEM media supplemented with 10% FBS, 100 U/mL penicillin G, 100 μg/mL streptomycin, 55 μmol/L β-mercaptoethanol, and 10 ng/mL murine IL7 (PeproTech 217-17).

We generated N-Me mutant mice at the Herbert Irving Comprehensive Cancer Center Transgenic Shared Resource (41) by injecting fertilized eggs from B6CBAF1 (GATA site 1, GATA site 2, and RBPJ site) or B6 (RUNX site, LEF/TCF site, PAX site, HOX site 1, and HOX site 2) females with Cas9 mRNA (TriLink Biotechnologies L-6125) and the corresponding single-guide RNA (sgRNA) in each case (Synthego) and transferring them into the oviducts of Swiss Webster foster females.

We performed mouse genotyping by N-Me PCR amplification (GATA and RBPJ sites: forward primer 5′-GTGAAAAATTACAAGGATGGG-3′ and reverse primer 5′-CATCAGAGTAGAGTACAGTGC-3′; HOX, RUNX, LEF/TCF, and PAX sites: forward primer 5′-GACCTTTGCTGCACTTGCATC-3′ and reverse primer 5′-TGACACAATCACCAGGTTCAG-3′) and Sanger DNA sequencing at Genewiz.

All animals were maintained in specific pathogen-free facilities at the Irving Cancer Research Center at Columbia University Medical Campus. The Columbia University Institutional Animal Care and Use Committee (IACUC) approved all animal procedures (protocols AC-AAAL3600 and AC-AAAR3425). Animal experiments were conducted in compliance with all relevant ethical regulations.

To generate NOTCH1-induced T-ALL in mice, we infected lineage-negative enriched cells from bone marrow of N-Me WT, GS1+2 heterozygous mutant and GS1+2 homozygous mutant donors with retroviral particles expressing oncogenic NOTCH1 (ΔE-NOTCH1; ref. 23) and GFP as described previously (11, 42), and transplanted them in sublethally irradiated (500 cGy) NRG mice (RRID: IMSR_JAX:007799).

We crossed N-Me GS1+2 mutant mice (N-Me^+/GS1+2mut), N-Me conditional knockout (N-Me^+/flox; Rr38^+/tm1.1Aafo; ref. 11), and Rosa26TM-Cre (Cre-ERT2) mice (43) to generate NOTCH1-induced leukemia from resulting Rosa26^+/CreERT2 N-Me^{flox/GS1+2mut} mice as before (11, 42). We transplanted lymphoblasts from spleens of diseased animals into secondary hosts and treated mice with 3 mg of tamoxifen or with corn oil vehicle by intraperitoneal injection two days after transplant and every five days thereafter.

To generate GS1+2 mutant MYC-rescued tumors, we crossed N-Me GS1+2 homozygous mutant mice, Rosa26TM-Cre mice, and Rosa26StopFLMYC (LSL-MYC) mice (44). We infected lineage-negative enriched cells from bone marrow of Rosa26^{CreERT2/LSL-Myc} N-Me^{GS1+2mut/GS1+2mut} donors with oncogenic NOTCH1 and transplanted them in NRG mice as described before. We treated mice with 3 mg of tamoxifen by intraperitoneal injection 1 week after transplant.

We performed 4C analysis in Jurkat cells and mouse primary T-ALL lymphoblasts as described previously (45), using the restriction enzymes HindIII and DpnII. We constructed sequencing libraries from 4C DNA including barcoded Illumina adapters to the 5′ end of each PCR primer (human MYC promoter viewpoint: forward primer–5′-AGACGTGGGGGCTAAAGCTT-3′ and reverse primer 5′-TGGGTATTTGGTTTGGCCTAT-3′; human N-Me viewpoint-forward primer 5′-CCAAAGTACCCTACAAGCTT-3′ and reverse primer 5′-GCTGACAGTTGTTAGCAGGG-3′; mouse Myc promoter viewpoint: forward primer 5′-TAAAGGATGACCGGAAGCTT-3′ and reverse primer 5′-GGGAGTAGATGAACCCATCC-3′; mouse N-Me viewpoint: forward primer 5′-CTAATTTATTTTCTAAGCTT-3′ and reverse primer 5′-CATGAAATTCCATTGCTTCAG-3′). We sequenced pooled libraries using a HiSeq 2500 sequencer (Illumina).

We analyzed 4C sequencing results using 4Cseqpipe (46). Samples containing less than 0.6 million mapped reads (following removal of undigested and self-ligated fragments) were discarded. In addition, samples with fewer than 0.2 cis/trans ratio of mapped reads and with fewer than 40% read coverage for all HindIII sites in the 1 Mb surrounding the bait region were discarded.

We performed reverse ChIP assays as described previously (47). Briefly, we generated N-Me DNA bait sequences by PCR from human genomic DNA using an N-Me biotinylated forward primer (5′-CCCTAATTTCTATCCCCACTGTC-3′) and an unmodified N-Me reverse primer (5′-ATTTTTTTCCTGTTAATATGCTGTAC-3′). Then, we conjugated DNA baits to streptavidin beads and incubated them with nuclear protein extracts from ALL-SIL, HPB-ALL, or JURKAT cells. We used nonconjugated beads as negative control. N-Me pulled-down proteins were analyzed by mass spectrometry at the Proteomics Laboratory at the New York University School of Medicine (New York, NY). The MS–MS spectra were searched against the Uniprot human reference proteome database using Sequest within Proteome Discoverer. A 1% false discovery rate cutoff was applied on the peptide level using a standard target–decoy database strategy. All proteins identified with fewer than two unique peptides were excluded from analysis. Thus, we recovered 362 proteins from HPB-ALL extracts, 321 proteins from ALL-SIL extracts, and 219 from JURKAT extracts. We normalized data to background signal (KRT1) and filtered against signal-to-noise ratio (protein signal/KRT1 signal > 0.30) and the Contaminant Repository for Affinity Purification (CRAPome; ref. 48) contaminant list (≤10% of the CRAPome). We identified 79 N-Me–associated proteins in HPB-ALL extracts, 50 in ALL-SIL extracts, and 38 in JURKAT extracts. Proteins identified in all three cell line extracts were considered high-confidence N-Me–associated proteins.

For site-specific reverse ChIP of RUNX- and RBPJ-binding sites, we annealed biotinylated complementary primers encompassing the binding motifs and their mutated forms (WT RUNX site 5′-TGAGATGATCAGTTTTACCACAGTTCACTACACTC-3′, mutant RUNX site 5′-TGAGATGATCAGTTTTAgactAGTTCACTACACTC-3′, WT RBPJ site 5′-CAGAGATGGGGTTCCCAGGGTGTTTCAAGGG-3′, mutant RBPJ site 5′-CAGAGATGGGGTTgCgtGGGTGTTTCAAGGG-3′). We analyzed RUNX1 and activated NOTCH1 binding to WT and mutated sequences by Western blot analysis using RUNX1 (365644, RRID: AB_10843207, Santa Cruz Biotechnology) and activated NOTCH1 (2421, RRID:AB_2314204, Cell Signaling Technology) antibodies, respectively.

We analyzed evolutionary conservation of N-Me sequences (Supplementary Table S2) using the DiAlign TF tool (RRID: SCR_008036; ref. 49). We aligned N-Me sequences from 28 vertebrate species and identified putative transcription factor binding sites in highly conserved regions. We calculated evolutionary conservation scores (phyloP score) at individual alignment sites using Ion Reporter Software (50). We generated a phylogenetic tree using the iTOL tool (51).

We designed sgRNAs overlapping the N-Me transcription factor binding sites of interest and evaluated for potential off-targets using E-CRISPR (Deutsches Krebsforschungszentrum; ref. 52). sgRNA sequences can be found in Supplementary Fig. S2.

All flow cytometry data were collected on a FACSCanto II flow cytometer (BD Biosciences) using FACSDiva software (BD Biosciences, RRID: SCR_001456) and analyzed with FlowJo software (Tree Star, RRID: SCR_008520).

To analyze thymic populations, we stained single-cell suspensions of thymocytes with a lineage marker biotinylated antibody cocktail against CD11b, Gr1, NK1.1, Ter119, CD19, and B220, and then with fluorochrome-conjugated streptavidin and antibodies against CD3e, CD4, CD8a, CD25, and CD44 (Supplementary Table S4). Lineage-negative cells were represented in a CD4 versus CD8a plot and CD4/CD8 DP (CD4⁺ CD8a⁺) and CD4SP (CD4⁺ CD8a⁻) populations were gated. Then CD4/CD8 DN cells were plotted in a CD44 versus CD25 plot and CD8-positive cells in a CD3 histogram, to characterize DN1 (CD4⁻ CD8a⁻ CD44⁺ CD25⁻), DN2 (CD4⁻ CD8a⁻ CD44⁺ CD25⁺), DN3 (CD4⁻ CD8a⁻ CD44⁻ CD25⁺), and DN4 (CD4⁻ CD8a⁻ CD44⁻ CD25⁻) populations, and ISP (CD4⁻ CD8a⁺ CD3e⁻) and CD8SP (CD4⁻ CD8a⁺ CD3e⁺) populations, respectively.

To analyze mature T-cell populations in peripheral lymphoid tissues, we stained single-cell suspensions of spleen and lymph nodes with antibodies against CD4 and CD8a (Supplementary Table S4).

For MYC intracellular staining, we fixed and permeabilized membrane marker–labeled cells using the Fixation/Permeabilization Solution Kit (BD Biosciences, 554714) and stained them with a fluorochrome-conjugated anti-MYC antibody (Supplementary Table S4).

In tumor-generation experiments, we stained bone marrow–infected cell preparations with anti-Sca1 antibody (eBioscience 17-5981, RRID: AB_469488) and evaluated infection efficiency by assessment of the percentage of GFP⁺ Sca1⁺ cells. To analyze the emergence of the post-transplant CD4⁺ CD8⁺ DP wave, we bled mice 21 days after transplant, lysed erythrocytes, and stained white blood cells with antibodies against CD4 (BD Pharmingen 553051, RRID: AB_398528) and CD8a (eBioscience 25-0081, RRID: AB_469584).

For apoptosis and cell-cycle analysis, we harvested primary mouse tumor cells from the spleen of leukemic mice and cultured them with vehicle only (ethanol) or (Z)-4-hydroxytamoxifen for 3 days. To analyze apoptosis, we stained 10⁵ cells with APC-conjugated Annexin-V (BD Biosciences 550475) in Annexin-V Binding Buffer (BD Biosciences 556454) for 15 minutes at room temperature. Then, we incubated cells with 5 μg/mL of DAPI (Invitrogen D3571) and determined the percentages of live (Annexin-V⁻ DAPI⁻), apoptotic (Annexin-V⁺ DAPI⁻), and dead (DAPI⁺) cells by flow cytometry. To analyze cell cycle, we stained 10⁶ cells with 5 μg/mL of Hoechst 33342 (Sigma-Aldrich B2261) in complete Opti-MEM media for 1 hour at 37°C. We then washed the cells and incubated them with 50 ng/mL of 7-AAD (BD Biosciences, 559925) as a viability dye, and analyzed cell-cycle progression in the GFP⁺ 7-AAD⁻ population by flow cytometry.

For isolation of DN3 cells, we enriched CD4⁻ CD8⁻ DN cells from thymi from 6- to 8-week-old mice by labeling with a lineage marker biotinylated antibody cocktail against CD11b, Gr1, NK1.1, Ter119, CD4, CD8a, CD19, and B220 (Supplementary Table S4) followed by magnetic depletion of antibody-labeled cells using streptavidin microbeads (Miltenyi Biotec 130-048-101). We stained nondepleted cells with fluorochrome-conjugated streptavidin and antibodies against CD4, CD8, CD25, and CD44 (Supplementary Table S4), and sorted DN3 cells. For isolation of B cells, we stained splenocytes with a fluorochrome-conjugated anti-B220 antibody (Miltenyi Biotec, 130-102-187, RRID: AB_2660443) and sorted B cells (B220⁺). All sortings were performed on a SH800 cell sorter (Sony Biotechnologies).

We performed luciferase reporter assays using a pBV-Luc−MYC promoter luciferase construct (53) together with a plasmid driving the expression of the Renilla luciferase gene (pCMV-Renilla) used as an internal control in Jurkat cells as described previously (11).

To perform H3K27 acetylation, H3K4 monomethylation, and GATA3 ChIP in GS1+2 heterozygous thymocytes, we cross-linked DN3 cells isolated from 6-week-old mice with 1% formaldehyde in PBS for 10 minutes at room temperature. We quenched the reaction by adding glycine up to 0.125 mol/L and incubated it for 5 minutes at room temperature. We performed cell lysis, chromatin shearing, ChIP, and purification of precipitated chromatin using the Auto iDeal ChIP-seq Kit for Histones ×100 (Diagenode C01010171), the Auto IPure Kit v2 × 100 (Diagenode C03010010), and the Diagenode Automated Platform SX-8G IP-Star Compact, following the manufacturer's protocol and using antibodies recognizing H3K27 acetylation (Abcam ab4729, RRID: AB_2118291), H3K4 monomethylation (Diagenode C15410194, RRID: AB_2637078), and GATA3 (Cell Signaling Technology 5852S, RRID: AB_10835690). For differential allele-binding analysis, we amplified N-Me DNA by PCR from immunoprecipitated chromatin (forward primer 5′-GTGAAAAATTACAAGGATGGG-3′ and reverse primer 5′-CATCAGAGTAGAGTACAGTGC-3′) and performed Sanger sequencing of the PCR products at Genewiz.

To perform SMARCA4 ChIP from GS1+2 heterozygous thymocytes, we cross-linked DN3 cells with 1 mg/mL disuccinimidyl glutarate in PBS for 30 minutes, followed by a second fixation with 1% formaldehyde in PBS for 10 minutes and 0.125 mol/L glycine quenching for 5 minutes. We lysate pelleted nuclei in lysis buffer containing 10 mmol/L Tris-HCl (pH 7.5), 0.1% SDS, 1 mmol/L EDTA, 0.1% sodium deoxycholate, 1% Triton X-100, 150 mmol/L NaCl, and protease inhibitor cocktail and sonicated them on a Bioruptor (Diagenode). Anti-SMARCA4 antibody (Abcam ab110641, RRID: AB_10861578) was adsorbed to Dynabeads Protein A/G (Invitrogen), added to the diluted chromatin complex, incubated overnight at 4°C, washed, and eluted for 1 hour at 65°C in ChIP elution buffer containing 100 mmol/L NaCO₃ and 1% SDS. We treated eluted samples with RNAse A (Invitrogen) and proteinase K (Ambion) and cleaned up chromatin samples using MicroChIP DiaPure columns (Diagenode). We performed differential allele-binding analysis as described previously.

To analyze the binding of transcription factors to N-Me, we used WT and GS1+2 MYC rescued tumor cells and performed ChIP as described before for SMARCA4 using antibodies against cleaved NOTCH1 (Santa Cruz Biotechnology, 6014-R, RRID: AB_650335), ETS1 (Santa Cruz Biotechnology, 350, RRID: AB_2100688), RUNX1 (Abcam ab23980, RRID: AB_2184205), TCF1 (Santa Cruz Biotechnology, 271453, RRID: AB_10649799), TCF3/E2A (Santa Cruz Biotechnology, 349X), and TCF12/HEB (Cell Signaling Technology, 11825, RRID: AB_2797736). We analyzed N-Me enrichment over the input chromatin by quantitative real-time PCR (qRT-PCR) with a QuantStudio 3 Real-Time PCR System (Applied Biosystems) using FastStart Universal SYBR Green (Roche; forward primer 5′-AACCCTGAACCTGGTGATTG-3′ and reverser primer 5′-GCCAAGAACTCCTCTGTGCT-3′).

We fixed thymi in 3.7% buffered formalin and embedded them in paraffin using standard procedures at the Herbert Irving Cancer Center Molecular Pathology Core. We stained 5-μm tissue sections with hematoxylin and eosin and performed MYC and caspase-3 staining following standard procedures (Histowiz). We prepared cell smears from heparinized blood preparations and stained them with May Grünwald (Sigma-Aldrich MG500, 500 mL) and Giemsa (Sigma-Aldrich GS500, 500 mL) following standard protocols.

We extracted RNA from DN3 thymocytes isolated by FACS using RNeasy Micro Kit (Qiagen 74004). We performed RNA library preparations and next-generation sequencing using the SMART-Seq v4 Ultra Low Input Kit for Sequencing (Clontech 634888) for full-length cDNA synthesis and amplification at Genewiz. Illumina Nextera XT Library Preparation Kit (Illumina FC-131-1024) was used for library preparation. Sequencing was performed on an Illumina HiSeq platform (Illumina).

We aligned RNA-seq raw reads to the mouse genome (mm10) using STAR (54) with 2-pass mapping. We computed gene-level raw counts in R using featureCounts (55) for all genes in the Gencode vM16 reference annotation. We performed differential gene expression analysis in R using the Bioconductor package DESeq2 (56).

We loaded thymocyte suspensions (5 × 10³ cells per sample) isolated from 6-week-old mice on a 10x Chromium instrument (10x Genomics) and prepared single-cell RNA-seq libraries using the Chromium Single Cell 3′ v2 Reagent Kit (10x Genomics CG00052) and sequenced them on an Illumina HiSeq instrument (Illumina) at the JP Sulzberger Columbia Genome Center.

We demultiplexed BCL files with 10x Cell Ranger's mkfastq command and performed analysis and alignment with Cell Ranger's count command with Cell Ranger's reference mm10 version 1.2.0 using Cell Ranger 2.0.0. We loaded single-cell data into count matrices (pandas v0.23.0) and removed ribosomal and mitochondrial genes. We normalized each cell for library size using counts per million (CPM) normalization and log-transformed. We then performed principal component analysis on the resulting matrices (scanpy v1.4.3). To identify significant principal components, we exploited Random Matrix Theory methodologies through the randomly algorithm (57). We used UMAP (58) to visualize the distribution of cells in the projection of the significant principal components. We then extracted markers for the different populations from data produced by the IMMGEN project (GSE15907). Briefly, we selected genes that maximized the difference between the mean expression in the population of interest and the rest of the populations. Each gene score was then divided by the SD of the gene in the samples outside the population of interest to penalize genes with high variability. The top genes were defined as markers for each population. These genes were subsequently used to score cells in the scRNA-seq samples. Cells were assigned to a specific population if scoring in the top percentile (for upregulated genes) or bottom percentile (for downregulated genes) for such particular population. The percentiles for each population were defined using known population percentages determined by flow cytometry analysis in each genotype as ground truth. Cells with multiple or no population calls were not assigned to any subset and were excluded from the final analysis.

We performed ATAC-seq analysis from sorted DN3 thymocytes from 6-week-old mice, from NOTCH1-induced WT and MYC-rescued GS1+2 tumors, and from DP leukemia lymphoblasts obtained from NOTCH1-induced leukemia-bearing mice and human lymphoblasts from 2 patients with DP (CD4⁺ CD8⁺) T-ALL. We generated transposed DNA fragments as described previously (59) and amplified them by PCR using NEBNext High-Fidelity 2X PCR Master Mix (NEB M0541) and custom primer indexes to generate ATAC-seq libraries (59). We purified PCR products using Agencourt AMPure XP beads (Beckman Coulter A63880) and sequenced them on an Illumina NextSeq instrument (Illumina).

BCL files were demultiplexed and FASTQ files generated on the BaseSpace platform (Illumina). Reads were trimmed of contaminating adapter sequences using cutadapt and aligned to the GRCM38 (mm10) build mouse genome and the GRCh37 (hg19) build human genome, respectively, using Bowtie2 (60). Peaks of transposase accessible chromatin were called using MACS2 v2.1.1 (61).

We analyzed chromatin accessibility during T-cell development from mouse T-cell precursor ATAC-seq data from the Immunological Genome Project (GSE100738) as described before. All the peaks from each population were merged together using the merge function of bedtools to generate a consensus peak table for the T-cell development program (69,302 peaks). Genomic analysis and visualization of highly variable ATAC-seq peaks was performed using R packages (R version 3.5.0). Briefly, we studied the 10% of peaks (6,930) with higher coefficient of variation. Consensus clustering of the samples was done using the ConsensusClusterPlus package in R. Unsupervised clustering and heat-map representation was done with gplots package. Analysis of motif enrichment was conducted using the Meme suite tools v 5.0.2 with default parameters, using the Joma 2013 database. For downstream analysis, we considered motifs with E-value < 0.05 and a minimum percentage of 30% of true positive (TP).

We designed oligonucleotide fluorescent probes for DNA FISH (Agilent) covering 200 kb windows encompassing N-Me (mm9 chr15:62977370-63177630) or the Myc promoter (mm9 chr15:61721767-61924188). N-Me probe (64% coverage) was conjugated to FITC. The Myc promoter probe (70% coverage) was conjugated to Cy3.

We performed DNA FISH analyses of splenic B cells and DN3 thymocytes from 6-week-old mice. We incubated cells at 37°C for 30 minutes on poly-l-lysine–coated slides and fixed them with 4% paraformaldehyde in PBS at room temperature for 10 minutes. Then, cells were permeabilized in 0.5% Triton X-100 in PBS. Probes were hybridized following Agilent FISH Protocol on formalin-fixed, paraffin-embedded samples. After hybridization, coverslips were mounted on slides using ProLong Diamond Antifade Mountant with DAPI (Invitrogen P36962). We imaged cells with the CSU-X1 confocal spinning disk system (Yokogawa Life Sciences) on an Eclipse TiE microscope stand (Nikon Instruments) using a 100× Oil Apo TIRF oil-immersion objective and an Andor Zyla 4.2 sCMOS camera.

We analyzed FISH images using ImageJ software (NIH, Bethesda, MD; RRID: SCR_003070; ref. 62) and measured distances between FISH foci. Briefly, nuclei were segmented using a mask of the DAPI staining. Green and red channels in each nuclei (FISH labeling) were threshold to create binary images, and FISH foci were detected using 3D Objects Counter plugin. Volumes and distances between objects were measured using 3D ROI Manager plugin (63). Measurements were filtered to analyze the shortest centroid to centroid distance between N-Me and Myc promoter foci.

We performed 3C analysis as described previously (64) using MboI as restriction enzyme. Bacterial artificial chromosome clones were used as control templates to cover the genomic regions under study. We analyzed 3C libraries by quantitative real-time PCR (qRT-PCR) with a QuantStudio 3 Real-Time PCR System (Applied Biosystems) using FastStart Universal SYBR Green (Roche) and primers for the Myc promoter (5′-TGCCTTCCCCGCGAGATGGAGTGGCTGTTT-3′), N-Me (5′-TGCAAGTGGAGTTGGCCATTGGGTGGCACC-3′), and an N-Me neighboring region (5′-TCACCCCAAGCCCAGTGCCTGTCATATGGGA-3′).

(Z)-4-hydroxytamoxifen (Santa Cruz Biotechnology, SC-3542) was dissolved in ethanol and added to the medium at a final concentration of 1 μmol/L. Tamoxifen (Sigma T5648) was dissolved in corn oil to a final concentration of 30 g/mL.

We analyzed expression of Myc and Actb by quantitative PCR (Myc: forward primer 5′-AGTGCTGCATGAGGAGACAC-3′ and reverse primer 5′-GGTTTGCCTCTTCTCCACAG-3′; Actb: forward primer 5′-AGGTGACAGCATTGCTTCTG-3′ and reverse primer 5′-GCTGCCTCAACACCTCAAC-3′) using FastStart Universal SYBR Green (Roche 4913850001) in a 7500 Real-Time PCR system (Applied Biosystems).

We generated N-Me DNA sequences by PCR from mouse genomic DNA using a N-Me biotinylated forward primer (5′-ACTTCTACTGTATGCAGAATG-3′) and an unmodified N-Me reverse primer (5′- GTAATAAAAGACCTCTCTTCC-3′). We assembled extended nucleosome arrays using Chromatin Assembly Kit (Active Motif), following the manufacturer's protocol. We generated compacted nucleosome arrays by adding histone H1 (Active Motif) to extended nucleosome arrays and incubating for 1 hour at 27°C. We incubated extended and compacted nucleosome arrays with GATA3 (Origene) for 2 hours at room temperature in binding buffer containing 10 mmol/L Tris pH 7.5, 1 mmol/L β-mercaptoethanol, 40 mmol/L KCl, 5 mmol/L DTT, 250 μg/mL BSA, 1% Ficoll, and 5% glycerol. We digested nucleosome arrays with 40 U/mL of DNAse I (NEB) for 1 minute at room temperature. DNA was purified, run in an agarose gel, and transferred into a nylon membrane for chemoluminiscence detection using Streptavidin–AP conjugate (Roche) and CDP-Star (Roche). DNA smear products were quantified by plot profile analysis using Fiji and normalized to the total DNA content.

We analyzed N-Me occupancy of chromatin marks, epigenetic factors, and transcription factors using the following T-ALL publicly available ChIPseq datasets from GEO: GSM1697882, GSM1314139, GSM2218755, GSM1689152, GSM1581344, GSM1442004, GSM722168, GSM3243670, GSM1442005, GSM2218756, GSM722167, GSM722166, GSM1193664, GSM449525, GSM2474553, GSM837992, GSM1410327, GSM732905, GSM2274676, GSM722165, and GSM1524254.

For analysis of mouse thymus development phenotypes, MYC intracellular expression in thymic populations, CD4⁺ CD8⁺ DP wave in tumor-generation experiments, MYC promoter luciferase reporter activity assays, and Myc expression, cell growth, apoptosis, and cell cycle in in vitro experiments, we evaluated statistical significance using two-tailed Student t test assuming normality and equal distribution of variance between the different groups analyzed. Distances between foci in DNA FISH experiments were analyzed using nonparametric Kolmogorov–Smirnov test. Survival in mouse experiments was represented with Kaplan–Meier curves and significance was estimated with the log-rank test. We performed all statistical analyses using Prism GraphPad 6 (RRID: SCR_002798) and considered statistical significance to be P < 0.05.

GEO Series accession numbers: RNA-seq data, GSE117483; single-cell RNA-seq data, GSE117412; ATAC-seq data, GSE117573, GSE124175, and GSE124223.

R. Rabadan is a member of the scientific advisory board at Aimedbio. A.A. Ferrando is a consultant at Ayala Pharmaceuticals and SpringWorks Therapeutics, reports receiving commercial research grants from Pfizer, Bristol-Myers Squibb, Merck, and Eli Lilly, and reports receiving other commercial research support from Novartis, EMD Millipore, and Applied Biological Materials. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Belver, A.Y. Yang, A.A. Ferrando

Development of methodology: L. Belver, A.Y. Yang, R. Albero

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Belver, A.Y. Yang, D. Herranz, P. Pérez-Durán, F. Gianni, D. Gurung, J.R. Cortés, A.J. Cooke, A.A. Wendorff, V. Cordó

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Belver, A.Y. Yang, R. Albero, D. Herranz, F.G. Brundu, S.A. Quinn, P. Pérez-Durán, S. Álvarez, F. Gianni, M. Rashkovan, P.P. Rocha, R. Raviram, C. Reglero, A.A. Wendorff, R. Rabadan, A.A. Ferrando

Writing, review, and/or revision of the manuscript: L. Belver, S.A. Quinn, J.P. Meijerink, A.A. Ferrando

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.Y. Yang, S.A. Quinn, J.P. Meijerink

Study supervision: L. Belver, A.A. Ferrando

We are grateful to T. Ludwig (The Ohio State University Comprehensive Cancer Center) for the Rosa26^+/Cre-ERT2 mouse. We thank Esperanza Agullo-Pascual, Sofie Demeyer, Victor Lin, and Beatrix Ueberheide for outstanding technical assistance. This work was supported by the NIH grants R35 CA210065 (to A.A. Fernando), U54 CA193313 (to R. Rabadán), R01 CA185486 (to R. Rabadán), U54 CA209997 (to R. Rabadán), and P30 CA013696 (Confocal and Specialized Microscopy Shared Resource and Transgenic Animal Shared Resource, Molecular Pathology Shared Resource, Herbert Irving Comprehensive Cancer Center). R. Albero and S. Alvarez are supported by Leukemia and Lymphoma Society postdoctoral fellowships. D. Herranz is supported by the U.S. NIH grant K99/R00 CA197869 and an Alex's Lemonade Stand Foundation Young Investigator grant. F. Gianni is supported by the American-Italian Cancer Foundation postdoctoral fellowship. M. Rashkovan is supported by a Damon-Runyon Sohn Pediatric Cancer fellowship. J.R. Cortes is supported by a Lady Tata Memorial Trust fellowship. V. Cordó is supported by the Dutch Cancer League (KWF 2016-10355).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.