All animals were kept on a C57BL6/J strain background and maintained in pathogen-free conditions according to institutional animal care and use guidelines. Lck-NotchIC9 (Notch1^tg) (53), SIL-SCL (A (5)3SCL; SCL^tg) (54), Lck-LMO1 (LMO1^tg) (55), Cd3e^-/- (56), E2a/Tcf3^+/- (57) and Heb/Tcf12^+/- (52) were described previously. Kaplan-Meier survival and statistical analysis was performed using GraphPad Prism 9.0 software (GraphPad Software, Inc.). T-ALL susceptibility was computed from areas under the curve (AUC) of Kaplan–Meier survival curves using Prism (1-AUC). Generation of human xenograft T-ALL blasts (14H025 and 14H148) were described previously (58). Human blast and primary murine thymocytes were cultured in MEM Alpha culture medium (Thermo Fisher) supplemented with 10% FBS, 10mM HEPES, 1mM sodium pyruvate, 55 μM β-mercaptoethanol, 2 mM glutamax, 5 ng/mL human FLT-3 Ligand, 5 ng/mL murine IL-7 and 20 ng/mL murine SCF. The DN T-cell line AD10.1 and Jurkat were cultured as previously described (50). KOPT-K1 and P12-ICHIKAWA cell lines were obtained from the DSMZ collection, Germany and maintained in RPMI-1640 culture media supplemented with 10% Fetal Bovine Serum (FBS).

Single-cell suspensions were prepared from thymi or thymoma of mice. Flow cytometry analysis and cell sorting were done as described previously (39) using antibodies against Thy1.2, CD4, CD8, CD25 and CD44, using propidium iodide to exclude dead cells. Multiparametric flow cytometry analysis was performed on a FORTESSA flow cytometer, and cell sorting was performed on FACSAria (BD Biosciences, San Jose, CA). Cell cycle analysis using DAPI staining was performed using ModFit (Verity Software Software, USA).

For Cdnk1a gene expression analysis, DN thymocytes from wt and Heb/Tcf12^-/- newborn mice were sorted by flow cytometry and cDNAs were prepared as described previously (39). Southern blots of the amplicons were revealed by hybridization using an internal ³²P-labeled oligonucleotide fragment (primer sequences are listed in Supplementary Table l). The ribosomal Rps16 expression was used as a control for cDNA quality and quantity.

For Notch1 sequencing, cDNA was prepared from total RNAs as described previously (39). Amplification of Notch1 exons 26, 27, and 34 from leukemias cDNA were Sanger sequenced in both directions. Quantitative gene expression analysis was performed on StepOne system (Life Technologies) using specific primers and Advanced qPCR mix (Wisent). Primer sequences used for specific mRNA amplification are listed in Supplemental Table S1.

Cells were lysed in RIPA buffer containing a cocktail of protease inhibitors. Protein extracts were resolved on bis-acrylamide gel, transferred on PVDF membranes and hybridized with anti-HEB and anti-E2A (Santa-Cruz Biotechnology Inc., CA) and anti-tubulin-β (Sigma) and anti-ERK (Cell Signaling) as a loading control.

Chromatin immunoprecipitation were performed on either Cd3e^-/- primary thymocytes or AD10.1 extracts as previously described (50). Quantitative PCR was performed on StepOne system (Life Technologies) using specific primers using Advanced qPCR mix (Wisent). Oligonucleotide sequences used for promoter amplification are shown in Supplementary Table 1.

Total RNAs were prepared from freshly isolated thymocytes from Cd3e^-/- (control DN3 thymocytes), Cd3^-/-SCL^tgLMO1^tg T-ALL and pre-leukemic (3-week-old), and Cd3e^+/+SCL^tgLMO1^tg leukemic mice using the RNeasy extraction kit (Qiagen, Mississauga, ON). cDNA synthesis, labeling and hybridization onto Affymetrix Mouse Genome 430A 2.0 arrays were performed at the Ottawa Health Research Institute (Ottawa, ON) as described (2). Raw data were normalized using the RMA procedure implemented in the Affy package from Bioconductor (59).

We obtained raw microarray data for normal thymocyte populations generated by the Immgen project from GEO (accession number GSE15907). Data were normalized using the RMA procedure implemented in the Affy package from Bioconductor (59). We derived “transition” signatures for each differentiation step (i.e. ETP to DN2, DN2 to DN3a, etc.), which contained genes whose expression levels present with at least a 2-fold change (up or down-regulated) in the transition (gene signatures provided in Supplementary File 1). In Figure 1F, gene set enrichment analysis (60) was applied to detect transition signatures that are enriched in the transcriptome of Cd3e^-/-SCL^tgLMO1^tg and Cd3e^+/+SCL^tgLMO1^tg leukemic cells.

Pre-TCR specific signatures (Figure 1G, Supplementary File 1) included genes that increased or decreased at least 2-fold during the DN3a-DN3b transition, and are not regulated by the αβTCR in peripheral CD8⁺ T cells stimulated by antigen (naive versus activated CD8+ T cells). Conversely, TCR-specific signatures (Figure 1G, Supplementary File 1) included genes exclusively regulated by the αβTCR in activated CD8⁺ T cells.

We collected genome-wide chromatin occupancy data for 7 transcription factors implicated in pre-TCR signalling (11 ChIP-seq experiments in total, Figure 2A) from Wang et al. (61), Miyazaki et al. (37), and the HemoChIP project (62). ChIP-seq data obtained for E2A (DN3 and DN4) (37) and NOTCH1 (G4A2 and T6E murine cell lines) (61) were processed according to the following steps: (i) sequence reads were mapped to the mouse genome mm9 using Bowtie with default parameters (maximum 2 mismatches); and (ii) peak coordinates were determined by the MACS tool, using the cutoff P < 10^–9. Peak coordinates for the HemoChIP dataset mapped to the mouse genome mm9 were downloaded from http://hscl.cimr.cam.ac.uk/ChIP-Seq_Compendium/ChIP- Seq_Compendium2.html. Last, all peaks were associated to their closest transcription start sites in the mouse genome using PeakAnalyzer v.1.4 tool (63). Lists of targets bound by transcription factors included all genes containing at least one binding site for the regulator (Supplementary File 1). We tested enrichment of targets using the Fisher’s exact test (Figure 2A).

DNA was extracted from Cd3e^-/-SCL^tgLMO1^tg leukemias (n=4) and control Cd3e^-/- thymocytes (n=2), followed by targeted exome enrichment was performed using the mouse Nimblegen SeqCap EZ kit from Roche. Sequencing was performed in the Illumina HiSeq2000 at the IRIC Genomics platform. Low quality bases (quality below 20) in paired-end reads were trimmed off using the Trimmomatic tool (64), duplicate reads were removed using Picard (http://broadinstitute.github.io/picard), and alignment to the mouse genome (mm10) was performed with bwa (65). The resulting depth of coverage was 20X for at least 85% of the captured exome across all samples. Exome variants in T-ALL samples were predicted using Strelka (66), using Cd3e^-/- thymocytes as matching controls, and annotation was performed using ANNOVAR (67). SIFT scores to determine deleterious variants were computed using the Variant Effect Predictor tool (68).

RNA extracted from Cd3e^-/-SCL^tgLMO1^tg leukemias (n=4) and control Cd3e^-/- thymocytes (n=2) was prepared using the TruSeq RNA kit (Illumina) and sequenced in the Illumina HiSeq2000 in the IRIC Genomics platform. Low quality bases (quality below 20) in paired-end reads were trimmed off using the Trimmomatic, and processed reads were aligned to the mouse genome mm10 using Tophat2/Bowtie2 v2.0.7 (69). Gene levels were quantified (FPKM values) based on the UCSC reference genes annotation using cuffdiff v.2.1.1.

RNA-seq data from pediatric tumors including a minimum of 313 T-ALL, 720 B-ALL and 350 Brain tumor samples were accessible for data analysis via the St. Jude PeCan data portal (https://pecan.stjude.cloud) (70) using Protein paint to capture RNA expression from the above ALL datasets (71).

The first critical event in leukemogenesis is the reprogramming of DN3 thymocytes into pre-leukemic stem cells (pre-LSCs) by the SCL-LMO1 oncogenes (2). These two oncogenes target the DN3 population (4, 72) but are not sufficient per se for progression to T-ALL. While the initiating reprogramming event is pre-TCR independent (2), progression to T-ALL requires both pre-TCR and NOTCH1 signaling (4), thus emulating the requirement for cooperative signaling between the two important pathways for normal thymocyte differentiation (11). To define the precise contribution of each pathway to disease progression, we took a genetic approach to quantitatively estimate T-ALL progression using disease penetrance and the time to leukemia onset as endpoints to measure leukemogenesis. T-ALL induced by SCL^tgLMO1^tg or Notch1^tg separately is affected by the absence of pre-TCR/CD3 signaling (Figure 1B) as previously reported (4, 72, 73). In contrast, SCL-LMO1 together with the hyperactive Notch1 allele (ICN1, hereafter Notch1^tg), induce T-ALL with full penetrance in the presence or in the absence of Cd3e (Figure 1C). Nonetheless, in Cd3e-proficient mice, pre-TCR/CD3 signaling accelerates the disease to 29 days, compared to 112 days in Cd3e-deficient mice. These results indicate that Notch1^tg drives the penetrance of T-ALL while the pre-TCR determines the time to leukemia. Of note, T-ALL induced by the three oncogenes together in the absence of Cd3e reproduce the disease induced by the two transcription factor oncogenes SCL and LMO1 in a Cd3e-proficient background (Figure S1). Therefore, in the context of T-ALL induced by the SCL and LMO1 oncogenes, the Notch1 transgene controls disease penetrance while pre-TCR signaling accelerates disease onset.

While pre-TCR signaling has been known to be important for leukemogenesis (74, 75) and more specifically for SCL-LMO1-induced T-ALL (4, 76), the contribution of the pre-TCR and downstream molecular effectors remain to be uncovered. T-ALL that still develops in the absence of CD3 or RAG, completely lacked the typical Notch1 gain of function mutations (4), providing us with a unique genetic tool to dissect the contribution of these pathways to T-ALL.

We therefore conducted a transcriptomic analysis that capitalized on our identification of DN3a as the cell of origin of T-ALL, and in Cd3e^-/- mice in which pre-TCR signaling is abrogated, causing thymocyte differentiation blockade at the DN3a stage. This allows for a stringent comparison between the pre-leukemic and the leukemic state to define the molecular signature of progression. During the pre-leukemic stage, Cd3e^-/-SCL^tgLMO1^tg thymocytes are blocked at the DN3a stage, as expected from the absence of CD3.

We next compared the transcriptomes of pre-leukemic (n=3) and of SCL^tgLMO1^tg leukemic cells (n=6) (Figure 1D). We applied principal component analysis (PCA) to compare the transcriptomes of normal thymocyte populations (obtained from the Immgen project) with pre-leukemic and leukemic samples. The first component reflected the distinct Affymetrix chips used for profiling whereas the second PCA component organized the transcriptomes according to their differentiation trajectories, from ETP to DP cells. Overall, PCA showed that pre-leukemic samples were comparable to the Cd3ϵ^-/- DN3a thymocytes. Strikingly, progression to T-ALL correlates with the expression profiles of thymocytes that have undergone β-selection, despite the absence of pre-TCR signaling in Cd3ϵ^-/- mice. Consistent with PCA, we observed that leukemic cells acquired a post-β-selection phenotype to become DN3b-DP cells (98% of the thymic mass), compared to 2% at the pre-leukemic stage, despite the complete lack of normal pre-TCR function in Cd3e^-/- mice (Figure 1E).

We also correlated disease progression with gene signatures of thymocyte differentiation using gene set enrichment analysis (GSEA). We observed that only signatures associated to post-β-selection thymocytes (DN3b to DN4 cells) were positively correlated with T-ALL progression (Figure 1F). The strongest correlation was associated with the DN3a-DN3b transition, exactly at the stage where the pre-TCR/CD3 triggers a burst of cell proliferation following a productive TCRβ rearrangement and the formation of a functional pre-TCR/CD3 complex.

CD3 signaling is important for pre-TCR function and β-selection, but also for TCR signaling. To distinguish the contribution of these two pathways to T-ALL progression, we applied GSEA to analyse gene signatures of antigen-independent (pre-TCR) and antigen-dependent (TCR) T-cell stimulation in T-ALL progression. Pre-TCR-induced genes (DN3b-DN4) correlated positively with leukemia progression (Figure 1G, adj. P < 10^-4). In contrast, the TCR gene signature did not correlate positively with T-ALL progression induced by SCL-LMO1 (Figure 1G).

In summary, our results indicate that leukemic cells display gene signatures of post-β-selection thymocytes, suggesting that a pre-TCR/CD3-like proliferation has occurred even in the absence of a functional pre-TCR (Cd3e^-/- background). Moreover, progression to T-ALL overlaps specifically with pre-TCR-driven gene signature. This CD3-independent activation of the pre-TCR molecular signature indicates a strong selective pressure during leukemic progression for pathways that normally control the β-selection checkpoint.

To directly address the role of Heb or E2a, we analyzed thymocyte numbers in E2a/Tcf3^+/- or Heb/Tcf12^+/- mice in the context of the SCL^tgLMO1^tg mice or their wild type (wt) littermates. In adult wt mice, removal of one Heb allele or one E2a allele did not significantly affect thymocyte numbers at the DN3 to DP stages (Figure 3A), although Heb monoallelic deletion resulted in modest but significantly increased cell numbers within populations undergoing β-selection, i.e. DN3b and DN3-4 (Figure S3), concurring with the view that Heb enforces a proliferation checkpoint at this stage (44, 82). We and others previously showed that the SCL-LMO1 or LMO2 oncogenes expand the DN3 populations due to increased self-renewal capacity (2, 4). Interestingly, Heb haploinsufficiency further increased the expansion of the DN3 and DN4 populations induced by the SCL-LMO1 oncogenes (Figure 3A). To directly address the antiproliferative role of HEB, we compared S/G2/M phase progression in Heb/Tcf12^+/- and Heb/Tcf12^+/+ DN3 thymocytes from Cd3e^-/- mice. In absence of pre-TCR signaling, while loss of one allele of Heb increased the proportion of proliferating cells, we found that the SCL-LMO1 oncogenes decreased the proportion of cycling DN3 thymocytes, consistent with a role for SCL in quiescence control (83). In this context, removing one Heb allele re-established proliferating DN3 thymocytes to normal proportions (Figure 3B). Therefore, in the absence of pre-TCR signaling, the SCL and LMO1 oncogenes revealed Heb haplo-insufficiency in cell cycle control, indicating and that HEB anti-proliferative function is required to control oncogenic stress at the β-selection checkpoint. In addition to the previously reported role for Id3-mediated inhibition of E proteins during β-selection in steady state (84), our data indicate a distinctive requirement for HEB in stress response.

The E proteins and Id axis has a well-established tumor-suppressor function (42, 45). We therefore addressed the question whether both HEB and E2A have a tumor suppressor function in the context of SCL-LMO1-induced T-ALL (Figure 3C). In Cd3e-proficient mice, loss of one E2a allele caused a modest decrease in latency from 122 days in littermate controls to 114 days (Figure 3C, left panel). In contrast, deletion of one Heb allele accelerated the time of onset to 83 days compared to 130 days in littermate controls (Figure 3C, right panel). Together, our results indicate a tumor suppressor function for Heb which acts in a gene-dosage dependent manner in T-ALL induced by SCL and LMO1.

In Cd3e-deficient mice expressing the SCL and LMO1 oncogenes, inactivation of a single Heb allele bypassed pre-TCR signalling to increase the proportion and numbers of DN4 cells during the pre-leukemic stage (Figures 4A, B and Figure S4). In addition, monoallelic Heb deletion allowed a minor population of SCL-LMO1 expressing thymocytes to progress to the DP stage (1.4%, Figures 4A, B). Last, progression to the leukemic stage is associated with a transition to a post-β-selection phenotype in Cd3e-deficient SCL^tgLMO1^tg mice, which consistently increased in Heb/Tcf12^+/- T-ALL (Figures 4C, D).

We next assessed the impact of Heb gene dosage on disease penetrance and time to leukemia. Strikingly, decreased Heb (Heb/Tcf12^+/-) compensates for the absence of pre-TCR signaling in Cd3e^-/- mice and allowed SCL^tgLMO1^tg T-ALL to become fully penetrant, in addition to accelerating disease onset by 107 days (Figures 5A, B). Hence, in Cd3e^-/-SCL^tgLMO1^tg mice in which disease penetrance was 80%, loss of one Heb allele recapitulated the effect of the Notch1 oncogene on restoring full disease penetrance as shown in Figure 1C. Since Notch1^tg-induced T-ALL is also dependent on pre-TCR function (Figures 1C and 5C, D), we next addressed the importance of Heb in T-ALL induced by Notch1^tg in pre-TCR proficient and pre-TCR deficient mice. In contrast to SCL and LMO1, decreased Heb gene dosage did not affect Notch1^tg-induced T-ALL in Cd3e^-/- or in Cd3e^+/+ mice (Figures 5C–F), indicating that Heb and Notch1 operate in the same genetic pathway in T-ALL, as suggested during normal differentiation (36). In summary, Heb and Notch1 inversely control the penetrance of T-ALL induced by the SCL and LMO1 oncogenes whereas the pre-TCR determines the DN-DP transition during the pre-leukemic stage and the time of disease onset.

Since the Notch1 oncogene is sufficient to bypass pre-TCR signaling to confer full leukemic penetrance in Cd3e^-/-SCL^tgLMO1^tg mice (Figure 1B), we addressed the question whether the increased penetrance caused by monoallelic Heb deletion shown in Figure 5A could be due to the acquisition of Notch1 gain of function mutations. We analyzed a cohort of mice with Heb/Tcf12^+/- and Heb/Tcf12^+/+ T-ALL for the presence of Notch1 mutations (Figures 5E and S5). As expected, 19 of 20 SCL^tgLMO1^tg T-ALL in Cd3e^+/+Heb/Tcf12^+/+ mice exhibit Notch1 gain of function mutations (4, 85) which affect the PEST domain whereas Cd3e-deficient SCL^tgLMO1^tgHeb/Tcf12^+/+ T-ALL completely lacked Notch1 mutation as reported (4). In comparison, only 1 of 7 Heb/Tcf12^+/- T-ALLs acquired Notch1 mutations (Figure 5E and Figure S5). Therefore, the increased aggressiveness of Heb/Tcf12^+/- T-ALL is unlikely due to Notch1 gain of function mutations.

These results establish Heb as a tumor suppressor that normally enforces a proliferative checkpoint during β-selection to suppress oncogene-induced T-ALL in a gene dosage-dependent manner. Unlike classical tumor suppressors, the human HEB gene is not affected at the genomic level in T-ALL (10, 25). Here, we show that HEB is regulated at the protein level, pointing to a distinctive loss of function mechanism.

While pre-TCR signaling has been shown to modulate the aggressiveness of T-ALL in several mouse models, the pre-TCR is dispensable for T-ALL induced by E47- or Trp53-deficiency (97, 98). Hence, the importance of the pre-TCR in T cell transformation induced by oncogenic events remained to be clarified. We propose that the importance of pre-TCR signaling depends on the primary oncogenic transcription factor. Previous transcriptome analyses of pediatric T-ALL identified five molecular subgroups and showed that the TCR signaling pathway is significantly enriched in the TAL1 group while other KEGG pathways related genes were more generally distributed within the other molecular subgroups (9). We now provide evidence that β-selection but not antigen-specific TCR signals are collaborating events. Indeed, the pre-TCR drives cell proliferation, which is required not only for clonal expansion but also for DN-DP differentiation (99) and thymocyte survival (100). A pre-T cell receptor lacking the TCR beta variable domain causes an expansion of the DP population that precedes overt T cell leukemia, suggesting that abnormal pre-TCR function can be oncogenic (101). In addition, the pre-TCR signal is important for Notch1/ICN1, Notch3- and TEL-JAK2-induced leukemias (74, 102, 103), indicating that developmental processes required for normal thymocyte development can be implicated in the pathogenesis of T-ALL (75). In the present study, even though T-ALL can develop in the absence of Cd3e and of Notch1 mutations, our transcriptomic comparison of preleukemic cells with fully transformed leukemic cells indicate a reactivation of β-selection during the progression to T-ALL. The signal strength providing progression to CD4+CD8+ DP T-ALL can either originate from a hyperactive Notch1 allele or the deletion of a single Heb allele. Our data provide compelling evidence for the importance of signal strength of NOTCH1 or HEB in driving the DN to DP transition associated with leukemogenesis.

The pre-TCR complex induces Id expression and consequently inhibition of E protein activity suggesting that the pre-TCR functions upstream of E proteins (81). Nonetheless, E2a-deficient mice develop T-cell lymphomas (42, 45), associated with Notch1 mutations within the PEST domain (104), suggesting that E2a is not downstream of Notch1. Rather, the latter observations suggest that Notch1 and E2a operate in parallel pathways. Moreover, Heb may function parallel to or downstream of the pre-TCR (51). Since NOTCH1 and pre-TCR functions are cooperative during β-selection (11), and both are believed to affect E protein function (81, 105, 106), pre-TCR signaling becomes a confounding factor in assessing the contribution of NOTCH1 and of E proteins in leukemogenesis. Using Cd3e^-/- mice in which pre-TCR function is abrogated, we now show that oncogenic Notch1 controls disease penetrance in SCL-LMO1-induced T-ALL while the pre-TCR signal governs the time of leukemia onset. Last, our data indicate that Heb but not E2a operate in the same pathway as Notch1 since monoallelic deletion of Heb does not affect Notch1-induced T-ALL but accelerates the disease induced by SCL-LMO1 in Cd3e-sufficient mice. Finally, loss of one Heb allele in Cd3e-deficient mice compensates in part the Notch1 oncogene to restore full penetrance to T-ALL induced by SCL and LMO1.

We and others previously showed that the SCL transcription factor heterodimerizes with E2A or HEB, thereby inhibiting target gene expression such as Cd4 and Ptcra and thymocyte differentiation (39, 40, 50). Nonetheless, E47 deficiency promotes the aberrant development of Rag1 null thymocytes and appearance of DP cells (84), similar to the appearance of a minor population of DP cells reported here in Cd3e-deficient thymocytes lacking one Heb allele.

E proteins are negatively regulated by ID proteins, another class of bHLH protein. At the β-selection checkpoint, the activation of RAS extracellular-signal-related kinase (ERK)–MAP kinase pathway upregulates ID proteins that bind to E proteins preventing their transcriptional activity (84). E2A inhibits proliferation and differentiation at the β-selection checkpoint in the absence of pre-TCR expression (84), suggesting that inhibition of E2A via Id3 would enforce the two distinctive functional outputs of pre-TCR signaling in favoring both cell proliferation and differentiation. However, Id2 and Id3 suppress lymphomagenesis (107), suggesting a necessary balance of E protein/ID levels during thymocyte differentiation. Partial redundancy may explain that both E2a and Heb are haplosufficient for thymocyte differentiation during steady-state conditions. Nonetheless, the oncogenic stress induced by SCL and LMO1 in thymocytes reveal Heb haploinsufficiency at the DN to DP transition, more specifically at the β-selection checkpoint revealing at the same time that E2a cannot compensate for Heb in stress-response, despite their quasi-redundancy during steady-state conditions.

Our results also bring out the importance of monitoring protein levels in primary tumors, since inactivation of tumor suppressors via protein downregulation could represent a distinct mechanism driving T-ALL.

Master regulators of hematopoietic lineages have been implicated in tumor suppressor function, as reported for C/EBPα (108, 109) and SPI1/PU.1 in acute myeloblastic leukemia (AML) (110) or PAX5 loss of function in B-ALL (90, 111). While E2A/TCF3 is annotated as a tumor suppressor gene (TSGene 2.0) with direct experimental and clinical evidence, HEB/TCF12 tumor suppressor function has been overlooked, despite the critical role of HEB in the T lineage (reviewed in (4, 112)). Due to a non-redundant function to enforce T lineage (36) and at multiple stages of T cell differentiation, we propose that the essentiality of HEB precludes the possibility of identifying loss of function mutations in hematopoietic malignancies. Our study reveals the possibility of temporary loss of function via downregulation of HEB protein levels. Moreover, our analysis of the PCGP cohort identifies decreased gene expression levels in T-ALL compared to other pediatric cancers for at least five TSGenes that are HEB targets, regardless of the T-ALL molecular subgroup (9). In addition to well documented genetic alterations of tumor suppressor genes, our study indicates that reduced expression levels of multiple tumor suppressor genes could also contribute to the process of tumorigenesis, as illustrated here for CDKN1A.

Cdkn1a is a well known E2A and HEB target (44) and tumor suppressor gene (113). More recently, the Cdkn1a^SUPER mouse is shown to be more resistant to transformation and exhibit a strong cancer protection phenotype, establishing a direct gene dosage-dependent tumor suppressor function for CDKN1A (87). Our observations indicate that Cdkn1a is down regulated by the SCL-LMO1 oncogenes during the pre-leukemic stage and further down regulated during progression to T-ALL. In Cd3e-deficient T-ALL, our data point to the down-regulation of Cdkn1a as part of the selective pressure to activate molecular mechanisms underlying the β-selection process and drive progression to T-ALL. Supporting this hypothesis, CDKN1A gene expression levels in the PCGP cohort is on average six- to seven-fold lower in pediatric T-ALL compared to B-ALL, AML or brain tumors. In addition to mRNA downregulation, p21^Cdkn1a can also be phosphorylated and degraded via ubiquitin-dependent and ubiquitin-independent proteolysis (113). Much like HEB, the CDKN1A gene is not subject to copy number variations nor deleterious mutations in hematopoietic malignancies. Rather, CDKN1A levels can be downregulated at the mRNA level, as illustrated here, or at the level of protein stability.

HEB and E2A exert an anti-proliferative function in thymocytes prior to Tcrb rearrangement, required for the formation of a functional pre-TCR (44), consistent with a tumor suppressor function reported here for HEB. In addition to a well-timed restriction in cell proliferation, it remains possible that tumor suppression involves additional molecular functions secured either by HEB or HEB downstream targets, as reported for TP53 (114). In addition to CDKN1A, HEB and E2A co-occupy three TSG loci, APC, a haplo-insufficient tumor suppressor, ZMYND11 a chromatin reader and tumor suppressor in breast cancer (115) and CUX1, a homeodomain transcription regulator and haploinsufficient tumor suppressor (116). HEB is also found to occupy three other TSGenes, MYBPP1A, potentially involved in nucleolar stress, NEDD4L, an E3 ligase with tumor suppressor function (117) and PLCB3, encoding phospholipase C beta 3 (phosphatidylinositol-specific), involved in G-protein-linked receptor-mediated signal transduction (Figure 6H). These additional targets may account for the more prominent role of HEB as a tumor suppressor in response to the oncogenic stress induced by SCL and LMO1, exactly at the β-selection checkpoint controlled by HEB. Beyond the SCL/TAL1 molecular group, five HEB target-TSGenes are down regulated in T-ALL compared to other pediatric tumors. In T-ALL, we propose that HEB controls a network of tumor suppressor genes to mitigate the oncogenic stress occurring at the β-selection checkpoint. Unlike classical tumor suppressors, these genes are not inactivated by genomic deletion or deleterious mutations but are down-regulated at the mRNA or at the protein levels. This mechanism cannot be detected by assessing DNA copy number variation or whole-genome sequencing, as exemplified by the total absence of HEB copy number loss or point mutations in T-ALL. In this context, future large-scale proteomics studies in primary tumors will be able to address whether downregulation of HEB protein is a recurrent event in human T-ALL.