Intracellular Cholesterol Pools Regulate Oncogenic Signaling and Epigenetic Circuitries in Early T-cell Precursor Acute Lymphoblastic Leukemia

Acute lymphoblastic leukemias (ALL) are highly proliferative malignancies that require prominent activation of anabolic pathways to support cell growth as well as metabolic activation of energy production and redox potential pathways (1). In these tumors, oncogenic transformation not only hijacks anabolic mechanisms active in highly proliferating early lymphoid progenitor cells, but also rewires these circuitries downstream of driver oncogenic transcriptional and signaling factors that enhance anabolic output, bypassing cell growth control checkpoints. Metabolic rewiring in B-precursor ALL (B-ALL) has been linked with the loss of PAX5 and IKZF1, two developmentally important B-cell lineage–specific transcription factor tumor suppressors (2), whereas T-ALL tumors are metabolically dependent on oncogenic signals from constitutively active PI3K–AKT and transcriptional programs driven by oncogenic NOTCH1 (3–5). These observations support the notion that lineage-specific transcriptional and signaling context, cell of origin–related epigenetic landscape, and differences in the mutational profile converge to activate metabolic pathways enabling increased nucleic acid, protein, and lipid biosynthesis in support of leukemia cell growth. Importantly, drugs inhibiting nucleotide biosynthesis (antifolates) and protein biosynthesis (L-asparaginase) are central to the treatment of this disease (6).

Recent transcriptomic and genetic studies have shed new light on the biology of ETP T-ALLs, a distinct group of immature T-cell leukemias associated with poor prognosis (7–9). The ETP-ALL group is defined by a characteristic immunophenotype and a distinct gene-expression signature indicative of very early arrest in T-cell development (7–10). In fact, analysis of the gene-expression signatures associated with immunophenotypically defined ETP-ALLs has shown that the gene-expression programs of these leukemias are indeed most closely related to those of human hematopoietic stem cells and myeloid progenitors (7, 9). Convergently, mutational profiling of these tumors has revealed a distinct mutational profile characterized by alterations in oncogenic signaling drivers and epigenetic regulators that are frequently mutated in myeloid malignancies, but rarely involved in T-ALL (7, 9). Given the distinct genetic and transcriptional profiles of ETP-ALL, and the characteristic impaired response to therapy of these tumors, we hypothesized a different architecture for metabolic circuitries supporting leukemia cell growth, proliferation, and survival in this disease. A corollary of this model is that the unique metabolic wiring of ETP-ALL may result in selective metabolic dependencies and distinct targetable therapeutic vulnerabilities. Following this premise, we integrated metabolomic and transcriptomic analyses to identify differential metabolic programs in ETP-ALL and T-ALL. These studies revealed lipid biogenesis as a prominent feature of ETP leukemia and a striking selective dependency of these high-risk tumors on the mevalonate pathway for cholesterol biosynthesis. Mechanistically, high-risk ETP-ALL cells are dependent on cholesterol biosynthesis to support oncogenic signaling and transcriptional circuitries that activate MYC expression via epigenetic regulation of chromatin accessibility at a distinct ETP-ALL–associated long-range MYC enhancer, uncovering an unanticipated link between cholesterol homeostasis and epigenetic regulation of leukemia cell growth.

To explore the metabolic landscape of T-cell lymphoblastic tumors, we analyzed flash-frozen bone marrow samples from a cohort of 22 samples by liquid chromatography/tandem mass spectrometry (LC/MS-MS), including ETP-ALL (n = 9) and T-ALL (n = 13) tumors categorized by immunophenotype and genetic alterations. We evaluated the quantitative profiling of 403 named cellular metabolites using unsupervised consensus clustering, resulting in the identification of two major groups, each encompassing 11 samples (Fig. 1A; Supplementary Table S1). Supervised analysis of these two clusters identified 53 differentially represented metabolites (fold change > 1.3; P < 0.05). Cluster I was characterized by increased levels of metabolites related to phospholipid and fatty acid metabolism, and notably encompassed all nine ETP-ALL samples in our series. Interestingly, we also detected lower levels of 3-hydroxy-3-methylglutarate (meglutol), a metabolite structurally similar but antagonistic to 3-hydroxy-3-methylglutaryl-CoA (HMG-CoA), the substrate of 3-hydroxy-3-methylglutaryl-CoA reductase (HMGCR)—the rate-limiting enzyme in the mevalonate pathway—in ETP-ALL compared with T-ALL patient samples. In addition, cluster II, which contained 11 of 13 T-ALL samples in our cohort, showed higher levels of metabolites in the pyrimidine biosynthesis and taurine and ornithine synthesis pathways (Fig. 1B and C).

These results support a differential metabolic profile differentiating ETP-ALL and T-ALL tumor samples, which could be linked to the distinct transcriptomic profiles characteristic of these leukemia groups. To explore this possibility, we first performed hierarchical clustering analysis of RNA sequencing (RNA-seq) data from 50 primary T-lineage leukemia samples using gene signatures characteristically associated with ETP-ALL and T-ALL (8) and classified these tumors into two groups corresponding to ETP-ALL (n = 32) and T-ALL (n = 18; Supplementary Fig. S1A–S1C). Next, and to specifically evaluate the transcriptional space most relevant in the context of metabolic regulation, we performed an unsupervised consensus clustering analysis in the space of 2,417 expressed metabolic genes. Notably, this analysis segregated our cohort into two distinct categories, generally matching the ETP-ALL (n = 27) and T-ALL (n = 23) groups (Fig. 2A). Consistently, supervised analysis revealed a robust signature of 269 metabolic genes differentially expressed between ETP-ALL and T-ALL (fold change > 1.5; P < 0.05), with 105 genes upregulated in ETP-ALL and 164 genes upregulated in T-ALL (Fig. 2B). Supervised clustering analysis in an independent cohort of pediatric samples using our 269 differentially expressed metabolic genes classified these samples into ETP-ALL and T-ALL, indicating that our metabolic signature can be extrapolated to both adult and childhood leukemias (Supplementary Fig. S2A–S2C). Gene set enrichment analysis (GSEA) showed that ETP-ALLs were characterized by upregulation of genes involved in phospholipid and sphingolipid metabolism and in the regulation of reactive oxygen species, whereas T-ALL samples showed significant enrichment in genes involved in nucleotide metabolism, cholesterol biosynthesis, and glycolysis (Fig. 2C and D).

These results point to prominent differences in metabolic programs between ETP-ALL and T-ALL leukemias that are potentially conducive to selective therapeutic vulnerabilities. However, addressing this question is hampered by prominent genetic and metabolic drift in most leukemia cell lines that have been maintained in vitro culture for prolonged periods of time. To overcome this obstacle, we leveraged early passes of two custom cell lines generated from pleural effusion tumor samples of patients with ETP-ALL (CUTLL3) and a T-cell lymphoblastic leukemia/lymphoma (CUTLL1) at relapse. CUTLL3 cells show prototypical immunophenotypic and transcriptional ETP features and are mutant for TP53 (Supplementary Fig. S3A and S3B). CUTLL1 cells are a model T-ALL cell line with constitutive activation of oncogenic NOTCH1 resulting from a TCRB–NOTCH1 chromosomal translocation and are mutant for TP53 (11). Profiling of 582 named cellular metabolites analyzed by chromatography–tandem mass spectrometry revealed broad overlap between the metabolic signatures of CUTLL3 and ETP-ALL tumors, consistent with their limited adaptation to in vitro culture conditions (Supplementary Fig. S3C). Notable differences between CUTLL1 and CUTLL3 lines included high levels of glycolysis products (lactate and pyruvate) and tricarboxylic acid (TCA) cycle intermediates (citrate and aconitate) in CUTLL1 T-ALL cells and elevated levels of most amino acids and sphingomyelin in the CUTLL3 ETP-ALL cell line (Supplementary Fig. S3D). Moreover, analysis of metabolic gene-expression signatures in RNA-seq data showed significant enrichment of metabolic genes upregulated in T-ALL in the CUTLL1 cell line and enrichment of ETP-ALL–related metabolic signatures in CUTLL3 cells (Supplementary Fig. S3E). In all, these results validate the distinct metabolic wiring of ETP-ALL and T-ALL tumors and support a role for the CUTLL3 cell line for the characterization of metabolic circuitries and tumor vulnerabilities associated with high-risk ETP-ALL.

The markedly differential metabolic signatures of ETP-ALL and T-ALL support that these tumor entities may differ significantly in their use of metabolic precursors for anabolic growth, redox homeostasis, and energy production. To investigate if, because of this differential metabolic wiring, ETP-ALL and T-ALL could show specific and differential metabolic vulnerabilities and response to chemical perturbations, we analyzed the response of CUTLL1 and CUTLL3 cells to a library of 6,374 bioactive compounds broadly targeting metabolic and signaling pathways. These analyses identified 52 compounds active against both CUTLL1 and CUTLL3 (>40% cell death) and selective cytotoxic molecules with preferential activity (fold change > 2; P < 0.01) against CUTLL3 ETP-ALL cells (n = 114) or the CUTLL1 T-ALL cell line (n = 227; Fig. 3A; Supplementary Fig. S4A and S4B; Supplementary Table S2). Among the most prominent metabolic hits, we observed marked antitumor activity of small-molecule inhibitors targeting nucleotide metabolism (6-mercaptopurine, 6-thioguanine, gemcitabine, cytarabine, mycophenolic acid, aminopterin, and cladribine) in CUTLL1 T-ALL cells. A higher dependency of CUTLL1 T-ALL cells on drugs inhibiting nucleotide metabolic circuitries is consistent with the observed increase in nucleotide metabolites and nucleotide biosynthesis genes in T-ALL primary patient samples, and the well-established central role of nucleotide antimetabolites in the treatment of ALL (6). In addition, we observed a specific response to statin (pitavastatin, fluvastatin, and mevastatin) inhibition of HMGCR, a central enzyme in the mevalonate–HMG–CoA reductase pathway, in CUTLL3 ETP-ALL cells (Fig. 3B), a finding congruent with our observation of decreased levels of 3-hydroxy-3-methylglutarate in ETP-ALL versus T-ALL primary samples (Fig. 1B).

The mevalonate–HMG–CoA reductase pathway is an essential metabolic anabolic route responsible for the synthesis of isopentenyl pyrophosphate and dimethylallyl pyrophos-phate (DMAPP), the building blocks for the synthesis of isoprenoids, a diverse class of biomolecules that includes cholesterol, important lipid cofactors, precursors for protein lipidation, and all steroid hormones (Supplementary Fig. S4C). Consistently, and in agreement with our compound screening results, dose–response curve analysis revealed markedly increased sensitivity to statin treatment in CUTLL3 cells compared with the CUTLL1 line [pitavastatin (CUTLL3 IC₅₀ 330 nmol/L; CUTLL1 IC₅₀ >5 μmol/L, P < 0.0001), fluva-statin (CUTLL3 IC₅₀ 340 nmol/L; CUTLL1 IC₅₀ >5 μmol/L, P < 0.0001); Fig. 3C; Supplementary Fig. S4D]. Furthermore, mevalonate, cholesterol, and geranylgeranyl pyrophosphate (GGPP), but not CoQ10 or farnesyl pyrophosphate (FPP) supplementation, rescued the effects of pitavastatin- and fluvastatin-mediated HMGCR inhibition (Fig. 3D–F; Supplementary Fig. S4E–S4K), implicating not only cholesterol biosynthesis, but also and convergently inhibition of geranylgeranylation as mediators of the antitumor effects of statin treatment in the CUTLL3 ETP-ALL cell line. Consistently, treatment of CUTLL3 cells with YM-53601, a squalene synthase inhibitor, induced antitumor effects only at high doses (IC₅₀ 10 μmol/L), yet this phenotype may result from nonspecific off-target effects, as it was not rescued with cholesterol (or GGPP) supplementation (Supplementary Fig. S4L and S4M).

Pitavastatin treatment in CUTLL3 cells induced a block of cell proliferation with a marked reduction in the proliferative S and G₂–M fractions (Fig. 3G), decreased cell viability via induction of cell death (Fig. 3H), impaired cell growth manifested as decreased cell size measured by forward scatter in flow cytometry (Fig. 3I), and resulted in a significant decrease in cellular cholesterol concentration (Supplementary Fig. S4N). Interestingly, although both CUTLL1 T-ALL cells and CUTLL3 ETP-ALL cells effectively upregulate LDLR (low-density lipoprotein receptor) mRNA levels following statin treatment (Supplementary Fig. S4O), only CUTLL1 cells are able to induce surface expression of the LDLR protein, a finding potentially related to the differential dependency of ETP-ALL and T-ALL to inhibition of the mevalonate pathway (Fig. 3J).

Analysis of the antileukemic effects of statin treatment in ETP-ALL and T-ALL patient-derived xenografts showed significant loss of cell viability induced by pitavastatin treatment compared with vehicle controls in all three ETP-ALL xenografts examined, but not in T-ALL xenografts, reinforcing a selective role for de novo cholesterol biosynthesis in supporting cell survival in ETP-ALL (Fig. 3K; Supplementary Fig. S4P). Consistently, statin treatment of mice xenografted with ETP-ALL (ETP5) cells improved survival compared with vehicle-treated controls (P = 0.0199; Fig. 3L) and associated with decreased spleen sizes at endpoint (P = 0.0341; Fig. 3M). This finding highlights cholesterol biosynthesis as a key player in ETP-ALL.

Next, to further explore the mechanisms mediating the antitumor effects of statin treatment in ETP-ALL, we investigated the impact of statin-mediated HMGCR inhibition on the global metabolite profile of CUTLL3 cells. In these experiments, chromatography–tandem mass spectrometry analysis of CUTLL3 lymphoblasts treated with a sensitizing dose of pitavastatin (80% cell viability) resulted in significant changes (fold change > 2; P < 0.01) in 62 metabolites implicated in multiple different metabolic circuitries (Fig. 3N). Metabolite changes following statin treatment included accumulation of 3-hydroxy-3-methylglutarate, indicative of a blockade in the mevalonate pathway, and increased levels of multiple products of RNA degradation including 3′ nucleotides (3′-GMP, 3′-AMP, 3′-UMP, and 3′-CMP) and dinucleotides [(3′-5′)-uridylyluridine, (3′-5′)-cytidylyluridine, (3′-5′)-adenylyluridine, (3′-5′)-uridylyladenosine, and (3′-5′)-uridylylcytidine]. Conversely, statin treatment resulted in decreased levels of TCA cycle metabolites (alpha-ketoglutarate, citrate, and aconitate) and in multiple intermediates in the purine and pyrimidine nucleotide biosynthesis pathways (orotate, AICAR, PRPP, thymidine, 2′-deoxyuridine, 2′-deoxycytidine, adenine, 5-methylcytidine, orotidine, and uracil) in support of unanticipated interactions between mevalonate metabolism and the central carbon metabolism and nucleotide biosynthesis pathways.

Following on these results, and to fully explore the landscape of genetic metabolic vulnerabilities in ETP-ALL, we performed a genome-wide forward genetic cell viability CRISPR screen in CUTLL3 ETP-ALL cells (Fig. 4A and B). Analysis of differential gRNA representation at day 7 following Cas9 induction in cells engineered with a genome-wide CRISPR library identified 1,049 essential genes [fold change > 1.3; false discovery rate (FDR) < 0.5] including 272 metabolic factors. Strikingly, top metabolic vulnerabilities in CUTLL3 cells included HMGCR, in support of an essential role of the mevalonate pathway (Fig. 4B; Supplementary Fig. S4Q). In addition, we noted prominent depletion of gRNAs directed against the BCL2 antiapoptotic factor, a finding congruent with the described dependency of ETP-ALL tumors on BCL2 expression (12, 13). Strikingly, combined treatment of CUTLL3 cells with pitavastatin and the BCL2 inhibitor ABT-199 revealed prominent and markedly synergistic antileukemic effects (Fig. 4C and D).

Following on our metabolite profiling analyses and to better understand the downstream effector pathways mediating the effects of statin treatment in cell growth, proliferation, and survival in ETP-ALL, we explored the global transcriptional effects of HMGCR inhibition. RNA-seq analysis of CUTLL3 cells following treatment with vehicle only or a sensitizing dose of pitavastatin (80% cell viability) identified 98 upregulated and 168 downregulated genes (fold change > 2; P < 0.01; Fig. 5A). This transcriptional response to statin treatment included upregulation of cholesterol biosynthesis genes, a well-established feedback mechanism mediated by SREBP2 activation in response to cholesterol depletion (14). Moreover, in agreement with a decreased cell growth phenotype, GSEA showed downregulation of multiple metabolic pathways including glycolysis, amino acid metabolism, and nucleotide metabolism (Supplementary Fig. S5A). Interestingly, extended analysis of signaling and oncogenic signatures identified significant downregulation of signatures related to AKT signaling and upregulation of gene sets indicative of GSK3 activation following pitavastatin treatment (Fig. 5B and C), suggestive of suppression of the AKT–GSK3B signaling axis downstream of HMGCR inhibition. Consistently, Western blot analysis of CUTLL3 cells treated with pitavastatin at an early time point (80% cell viability) showed markedly decreased phosphorylation of AKT, mTOR, and S6, which was effectively rescued by supplementation with mevalonate, cholesterol, or GGPP (Fig. 5D and E; Supplementary Fig. S5B). In contrast, a control experiment in which cells were treated in the same conditions with ABT-199 showed no effects on PI3K–AKT signaling factors (Supplementary Fig. S5C).

The AKT kinase orchestrates cell growth, metabolism, and survival downstream of multiple cytokine and growth factor receptors in support of a mechanistic role for AKT inhibition as a mediator of the antileukemic effects of statins in CUTLL3 ETP cells. To test this possibility, we evaluated the capacity of a myristoylated constitutively active membrane-bound form of AKT (myr-AKT) to antagonize the effects of statin treatment. In this experiment, expression of myr-AKT in CUTLL3 cells effectively antagonized the activity of statin treatment with prominent rescue of proliferation and cell viability (Fig. 5F and G). Furthermore, isolation of lipid rafts from CUTLL3 cells treated with vehicle or pitavastatin at an early time point (80% cell viability) showed a marked decrease in phosphorylated AKT following cholesterol depletion (Fig. 5H and I), further supporting a role for AKT signaling as a key effector of the antileukemic effects of statins in ETP-ALL. Moreover, GSEA of CUTLL3 transcriptional signatures elicited by pitavastatin treatment also revealed significant downregulation of gene sets controlled by MYC (Fig. 6A), a central oncogene driver of leukemic cell growth and proliferation. Of note, we identified MYC as a prominent vulnerability in CUTLL3 cells in our genome-wide CRISPR screen (Fig. 4B). MYC downregulation following pitavastatin treatment was associated with increased phosphorylation at threonine 58 (Fig. 6B–D), a posttranslational modification controlled by AKT–GSK3B, responsible for MYC proteasomal degradation (15–18), but also associated with transcriptional downregulation of MYC expression (Fig. 6E). Moreover, supplementation with mevalonate, cholesterol, or GGPP rescued MYC expression with reduced pMYC/MYC ratios (Fig. 6B–D; Supplementary Fig. S6A). In addition, a control experiment with ABT-199 in similar conditions (80% cell viability) revealed no changes in MYC expression (Supplementary Fig. S6B). To test whether enforced MYC expression could antagonize the effects of statin treatment, we expressed a doxycycline-inducible MYC in CUTLL3 ETP-ALL cells. In this experiment, expression of MYC in CUTLL3 cells effectively antagonized the activity of statin treatment with prominent rescue of cell viability (Fig. 6F and G).

Expression at the MYC locus is controlled by a constellation of long-range regulatory enhancers located in a gene desert region covering over 200 kb 5′ upstream centromeric and over 1.7 Mb 3′ downstream telomeric of the MYC gene (19–23). In the hematopoietic system, expression of MYC in stem and progenitor cells other than T-cell precursors is regulated by the blood enhancer cluster (BENC; refs. 21, 23), a group of distal regulatory elements located 1.7 Mb 3′ from MYC, whereas N-Me, a more proximal discrete enhancer placed 1.4 Mb 3′ from the MYC promoter, is responsible for MYC expression during thymocyte development (22). Analysis of chromatin accessibility by Assay for Transposase-Accessible Chromatin with high-throughput sequencing (ATAC-seq) revealed an open chromatin configuration at defined positions and most prominently in the MYC promoter and the BENC cluster of enhancers in CUTLL3 cells, in agreement with a stem cell–early myeloid progenitor chromatin configuration in ETP-ALL (Fig. 6H). ATAC-seq analysis of three independent ETP-ALL primary samples showed an open chromatin configuration at the BENC enhancer cluster, further supporting an active role for this regulatory site in the control of MYC expression in ETP-ALL (Fig. 6I). Next, and to explore the impact of HMGCR inhibition on the activity of these MYC regulatory elements, we analyzed the distribution and abundance of the transcriptional activation–associated histone 3 lysine 27 acetylation (H3K27ac) mark in CUTLL3 cells in basal conditions and following pitavastatin treatment. Chromatin immunoprecipitation followed by sequencing (ChIP-seq) analysis of H3K27ac at the MYC locus in these experiments revealed high levels of the activation-associated epigenetic mark at the MYC promoter and proximal 3′ regulatory sequences and in a far distal region encompassing the BENC (Fig. 6J). Notably, and in agreement with decreased MYC transcription, H3K27 acetylation levels were markedly reduced across these regions after pitavastatin treatment (Fig. 6J), in the absence of global changes in the levels of this histone mark (Supplementary Fig. S6C and S6D). Moreover, focused analysis of ATAC-seq changes induced by statin treatment showed markedly reduced chromatin accessibility at the BENC module C (BENC-C), a specific enhancer functionally linked with MYC expression in leukemia stem cells (Fig. 6H; ref. 23). Importantly, supplementation with cholesterol effectively restored the reduced chromatin accessibility induced by statin treatment at this site (Supplementary Fig. S6E). In addition, global analysis of the ATAC-seq changes induced by statin treatment identified 49 regions with increased accessibility and 343 regions with decreased accessibility (Supplementary Fig. S6F). Notably, the BENC-C enhancer ranked prominently among the top regions with decreased accessibility following treatment with pitavastatin (Supplementary Fig. S6F). Phylogenetic footprint analyses of the vertebrate BENC-C enhancer sequences revealed strong conservation and high density of evolutionarily conserved transcription factor binding motifs recognized by developmental (ETS, GATA, RUNX, SOX, and EBF) and signaling effector (STAT) transcription factors across most eutheria (placentals) and metatheria (marsupials) mammals, with some sites conserved in prototheria (monotreme), avian, and reptilian vertebrate species (Fig. 6K). Moreover, reverse ChIP mass spectrometry of enhancer-bound proteins identified major transcriptional regulators of hematopoietic stem cell activity (TCF3 and SPI1), lymphoid (MYB, ELF1, CBFA2T3, and TCF12), and myeloid development (SPI1, CEBPA, and RUNX2; refs. 24, 25) associated with BENC-C in CUTLL3 cells (Fig. 6L), highlighting a major role of this enhancer as a regulatory hub dominantly involved in the control of MYC expression in ETP-ALL. In agreement, chromatin conformation capture (3C) analysis of chromatin looping between BENC-C and the MYC promoter showed effective interaction, further implicating this regulatory segment in the control of MYC expression (Supplementary Fig. S6G). Interestingly, and of mechanistic importance, HMGCR inhibition following pitavastatin treatment did not abrogate the interaction between BENC-C and the MYC promoter, supporting a dominant role of changes in BENC-C chromatin accessibility and BENC H3K27ac over chromatin looping in the control of MYC expression following statin treatment (Supplementary Fig. S6G).

Patients diagnosed with ETP-ALL are classified as high risk, as they show high prevalence of primary drug resistance and early relapse with standard chemotherapeutic protocols, calling into question the efficacy of antifolates and other antimetabolites for these patients (8). Altered lipid biosynthesis is a hallmark of cancer metabolism, as increased de novo fatty acid synthesis is essential for rapidly proliferating cancer cells to continuously maintain membrane production and lipid-based posttranslational modifications of proteins (26). Moreover, the mevalonate pathway has emerged as an important essentiality across different tumor types (27–29). By performing global metabolomic profiling of ETP-ALL and T-ALL clinical samples, we observed that these two leukemic subtypes are distinct in their metabolic signatures, reflective of different metabolic wiring. In agreement with their high sensitivity to nucleotide analogues (6-mercaptopurine and nelarabine) and folate antagonists (methotrexate), T-ALLs showed metabolic profiles consistent with high rates of nucleotide biosynthesis (6). In addition, high levels of amino acids in T-ALL samples are in line with the effectiveness of amino acid depletion therapy with L-asparaginase in these leukemias (6). In contrast, the metabolic profiles of ETP-ALL seem to be differentially indicative of prominent cholesterol biosynthesis and lipid metabolism. Consistently, a high-throughput drug screen broadly targeting metabolic pathways identified inhibition of the mevalonate biosynthesis pathway with statins as a differential vulnerability in ETP-ALL compared with T-ALL. Of note, complete genetic inactivation of HMGCR by CRISPR mutation across over 900 cell lines profiled in DepMap induces a broad and extensive lethality, which extends to all blood cell lines tested, in agreement with the essential role for cholesterol biosynthesis in cell homeostasis in eukaryotic cells. These findings highlight the value of our pharmacologic chemical screen for target identification, as they support that the selective response to statin treatment documented across our experiments is reflective of a dose-dependent differential requirement of the mevalonate pathway for ETP-ALL cells over T-ALL.

Statin-mediated inhibition of HMGCR not only blocks cholesterol biosynthesis, but also suppresses N-glycosylation via depletion of dolichol, and protein farnesylation and geranylgeranylation. We observed rescue of the antileukemic effects of statins in CUTLL3 ETP-ALL cells by both cholesterol and GGPP supplementation, implicating both cholesterol depletion and geranylgeranylation as a compound vulnerability in these cells. The requirement of cholesterol depletion for the antitumor effect of statins in ETP-ALL is in contrast with the proposed mechanism of action of these drugs in other cancers in which statin-induced apoptosis seems to result primarily from suppression of prenylation in Rho and Ras GTPases (30–32).

Cholesterol, an essential component of biological membranes, contributes to proper cell membrane rigidity and accumulates at lipid raft signaling microdomains (33, 34). Moreover, molecular characterization of downstream effector pathways mediating the antitumor effects of HMGCR inhibition in ETP-ALL cells identified inhibition of the AKT1 kinase as a prominent mechanism contributing to the antileukemic effects of statins. Ligand engagement of membrane G protein–coupled receptors and tyrosine kinase receptors triggers activation of phosphoinositide 3-kinase complexes, which in turn recruit AKT to the membrane for activation, triggering the phosphorylation of effectors broadly involved in cellular growth, metabolism, and survival (35, 36). AKT1 is lipid-modified, is membrane-bound, and is found in lipid rafts, making it particularly sensitive to changes in membrane cholesterol (37). Consistently, we observed attenuated AKT signaling in ETP-ALL cells following HMGCR inhibition, and expression of a membrane-bound constitutively active form of AKT effectively restored cell growth and survival in statin-treated cells. Suppression of AKT signaling triggers pleiotropic effects including direct and indirect regulation of cholesterol biosynthesis routes via anterograde trafficking of the SCAP–SREBP2 complex toward the Golgi in COPII vesicles (38) and GSK3-induced phosphorylation-mediated targeted degradation of SREBP2 (39). Among these, gene-expression profiling analyses of statin-treated cells support prominent downregulation of transcriptional programs controlled by MYC, which we linked to both increased GSK3-induced MYC phosphorylation and, most prominently, transcriptional suppression of MYC expression. The relevance of MYC as an effector of leukemia cell growth and master regulator of leukemia stem cell activity cannot be understated (40). Dissection of the MYC regulatory programs in ETP-ALL reveals activation of an enhancer repertoire distinct from that typically active in thymocyte development and T-ALL, arguing for activation of early immature-myeloid programs in shaping the regulatory logic controlling leukemia cell growth and proliferation in this group.

Once considered separate circuits, mechanisms of metabolic–epigenetic coordination are now recognized as crucial for cellular adaptation to varying environmental conditions, particularly in terms of nutrient availability fuel supply (41). Thus, metabolites are substrates of writers and erasers of histone and DNA modifications such as acetylation, methylation, and glycosylation, linking the metabolite pools to epigenetic machinery (41). Importantly, cancer-associated metabolic perturbations can modify the epigenetic landscape, generating epigenetic plasticity and promoting epigenetic states conducive of increased self-renewal and transformation that offer unique opportunities for therapeutic intervention (42–44). Here we document suppression of AKT signaling and abrogation of MYC expression coupled with loss of chromatin accessibility at specific long-range MYC enhancers upon inhibition of the mevalonate pathway, revealing an oncometabolic circuitry that ties together nutrient and growth factor sensing, oncogenic signaling, and epigenetic regulation of central oncogenic drivers in support of leukemia cell growth, proliferation, and survival.

Finally, it has not escaped our attention that although specific data on the activity of statin treatment on the incidence of ETP-ALL are not available, statin treatment has been associated with a 26% risk reduction in overall leukemia incidence, an observation that is consistent with the clinically relevant activity of these drugs even in the context of dosing schedules designed for sustained prolonged therapy for prophylaxis of cardiovascular disease (45).

Leukemia lymphoblasts samples from patients with ALL were provided by the ECOG-ACRIN leukemia tissue bank and the Pediatric Oncology Division at the Columbia University Medical Center. The ETP4 and ETP5 leukemia xenografts were provided by David Teachey at the University of Pennsylvania. The T-ALL3 leukemia xenograft was provided by S. Indraccolo (UOC Immunologia e Diagnostica Molecolare Oncologica, Istituto Oncologico Veneto—Istituto di Ricovero e Cura a Carattere Scientifico, Padova, Italy). All samples were obtained with written informed consent at study entry and under the supervision of local Institutional Review Boards for participating institutions and were analyzed for this study under the supervision of the Columbia University Medical Center Institutional Review Board and in compliance with ethical regulations.

We passaged and harvested primary human xenograft ETP-ALL cells from the spleens of female immunodeficient 8- to 10-week-old NRG mice (NOD.Cg-Rag1<tm1Mom> Il2rg<tm1Wjl>/SzJ, The Jackson Laboratory) and cultured them in RPMI media supplemented with 20% FBS, 100 U/mL penicillin G, 100 μg/mL streptomycin, and 10 ng/mL human IL7. We purchased 293T cells for viral production from ATCC and grew them in DMEM supplemented with 10% FBS, 100 U/mL penicillin G, and 100 μg/mL streptomycin for up to 2 weeks. The CUTLL1 cell line, which was generated by continuous culture of a male T-cell lymphoblastic pleural effusion at relapse, has been characterized and reported before (11). We cultured CUTLL1 and CUTLL3 cells in RPMI 1640 media supplemented with 10% FBS, 100 U/mL penicillin G, and 100 μg/mL streptomycin. Cell lines were regularly authenticated and tested for Mycoplasma contamination.

To establish the CUTLL3 cell line, we obtained pleural effusion cells (10⁶) from a female pediatric patient with ETP-ALL at relapse from Morgan Stanley Children's Hospital of New York–Presbyterian/Columbia University Medical Center and plated them in a 10-cm dish in RPMI 1640 supplemented with 20% FBS, 100 U/mL penicillin G, 100 μg/mL streptomycin, and 10 ng/mL human IL7 at 37°C in a 5% CO₂ humidified environment. Cells were frozen and thawed three times over a period of 6 months and were finally cultured in RPMI 1640 supplemented with 10% FBS, 100 U/mL penicillin G, and 100 μg/mL streptomycin at 37°C in a 5% CO₂ humidified environment. We performed G band karyotyping at the Molecular Cytogenetics Shared Resource in the Herbert Irving Comprehensive Cancer Center using standard procedures.

To generate CUTLL3 cells with expression of myrAKT1, we transfected pCDH-puro-myr-HA-Akt1 (Addgene, 46969; ref. 46) or empty vector lentiviral constructs together with packaging (pCMV ΔR8.91) and envelope protein (pMD.G VSVG)–expressing vectors into HEK293T cells using PEI MAX 40000 (Polysciences, 24765). We collected viral supernatants after 48 hours and used them for infection of the CUTLL3 cell line by spinoculation using standard procedures. We selected infected cells with 2.5 μg/mL puromycin for 4 days.

To generate CUTLL3 cells with expression of MYC, we cloned MYC into the pLVX-TetOne plasmid at the EcoRI and BamHI sites in the multiple cloning site. Lentiviral production and puromycin selection were performed as above. We induced MYC expression with 0.01 μmol/L doxycycline.

We purchased pitavastatin calcium (S1759), fluvastatin sodium (S1909), ABT-199 (S0848), and CHIR-98014 (S2745) from Selleckchem. We purchased cholesterol (S5442), mevalonolactone (M4667), and coenzyme Q₁₀ (C9538) from Sigma-Aldrich. We purchased YM-53601 (18113), FPP ammonium salt (63250), and GGPP ammonium salt (63330) from Cayman Chemical.

For experimental therapeutic treatment studies, we injected 10⁶ ETP5 cells into female 6- to 8-week-old NRG mice. We began treating the mice when the percentage of hCD45 (eBioscience, HI30) cells detectable in the peripheral blood reached 5% and then randomized the mice into treatment and vehicle groups. We treated the mice with daily doses of vehicle (PBS) or fluvastatin (80 mg/kg) by oral gavage for four cycles (5 days on, 2 days off). We maintained all animals in specific pathogen–free facilities at the Irving Cancer Research Center at the Columbia University Medical Campus. The Columbia University Institutional Animal Care and Use Committee approved all animal procedures. We conducted all animal experiments in compliance with all relevant ethical regulations. We euthanized animals upon showing symptoms of clinically overt disease (do not feed, lack of activity, abnormal grooming behavior, hunch back posture) or excessive weight loss (10%–15% body weight loss over a week). Survival was represented with a Kaplan–Meier curve, and significance was estimated with the log-rank test (GraphPad Prism, RRID:SCR_002798).

We analyzed leukemic bone marrow samples from patients with ETP-ALL and T-ALL and CUTLL1 and CUTLL3 cell lines untreated or treated with vehicle (DMSO) or pitavastatin. Briefly, we snap froze pellets and extracted and prepared them for analysis using standard solvent extraction methods. We split the extracted samples into equal parts for analysis on the GC/MS and LC/MS-MS platforms of Metabolon, Inc. Identification of known chemical entities was based on comparison with metabolomic library entries of purified standards. We normalized results to protein concentration, log transformed, and imputed any missing values with the minimum observed value for each compound and used Welch two-sample t test to identify biochemicals that differed significantly between experimental groups. We also calculated an estimate of the FDR (q-value). We used MBROLE 2.0 (RRID:SCR_014684) to perform enrichment analysis of metabolites using the Kyoto Encyclopedia of Genes and Genomes (KEGG) annotations.

We quantified intracellular cholesterol levels using the Amplex Red Cholesterol Assay Kit (Invitrogen). We harvested 10⁵ CUTLL3 cells 72 hours after treatment with vehicle (DMSO) or pitavastatin, washed the cells with ice-cold PBS twice to remove residual culture media, and performed the assay based on the manufacturer's instructions.

High-throughput drug screening was performed by the high-throughput screening facility at the Columbia Genome Center. Briefly, we seeded cells into 384-well microplates at a concentration of 4,000 cells per well in 50 μL of RPMI media supplemented with 10% FBS. The next day, we delivered compounds using an automated liquid handling station (Cell::Explorer, PerkinElmer). We transferred compounds in triplicate (1 μmol/L, DMSO) to assay plates using nano-head (PerkinElmer). After 72 hours, we performed a cell viability assay using CellTiter-Glo reagent (Promega, G7573).

For data normalization, we used Pipeline Pilot Lab Analytics 17.1.0 and Chemistry 17.1.0 Collections. We used the percentage of positive control normalization component where each well is expressed as a percentage of the average of the DMSO-only wells in the same plate.

We measured cell growth and drug responses of the CUTLL3 cell line in vitro by measurement of the metabolic reduction of the tetrazolium salt MTT using the Cell Proliferation Kit I (Roche) following the manufacturer's instructions after 72-hour incubation with increasing concentrations of drugs. For drug combination studies, we added single agents simultaneously at fixed ratios and analyzed the results using Calcusyn software (Biosoft). This software was used for measurement of synergy or antagonism and is based on isobologram generation and the method of Chou and Talalay.

For cell death analysis, we harvested 10⁶ CUTLL3 cells 72 hours after treatment with vehicle (DMSO) or 0.5 μmol/L pitavastatin and stained with 50 ng/mL of 7AAD (BD Biosciences, 559925) for 10 minutes at room temperature.

For cell-cycle analysis, we harvested 10⁶ CUTLL3 cells 72 hours after treatment with vehicle (DMSO) or 0.5 μmol/L pitavastatin and fixed with 70% ethanol and stored overnight at 4°C. Cells were stained with propidium iodide solution (0.2 mg/mL RNase A, 500 μg/mL PI in PBS) for 30 minutes at 37°C.

We quantified cell viability of CUTLL3 cells and patient-derived xenografts by flow cytometry using a cell viability kit with the liquid counting beads (BD Biosciences) following 72-hour incubation with vehicle (DMSO) or 0.5 μmol/L (CUTLL3) or 5 μmol/L (xenografts) pitavastatin.

For LDLR quantification by flow cytometry, we harvested 10⁶ CUTLL3 cells 48 hours after treatment with vehicle (DMSO) or 0.5 μmol/L pitavastatin and stained with LDLR antibody (BD Biosciences, 565653, clone C7; RRID:AB_2739325) for 15 minutes at 4°C.

We collected all flow cytometry data on a FACSCanto II flow cytometer (BD Biosciences) using FACSDiva software (BD Biosciences, RRID:SCR_001456) and analyzed the data with FlowJo software (TreeStar, RRID:SCR_008520).

We extracted RNA from CUTLL1 and CUTLL3 cell lines using the RNeasy Mini Kit (Qiagen, 74104). We performed reverse transcription with 1 μg RNA using the SuperScript IV Reverse Transcriptase Kit (Invitrogen, 18090010). We analyzed LDLR expression 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 LDLR (F-GAATCTACTGGTCTGACCTGTCC; R-GGTCCAGTAGATGTTGCTGTGG).

We performed Western blot analysis using antibodies recognizing vinculin (VIN-11-5; NB120-11193, Novus Biologicals) and the following antibodies from Cell Signaling Technology using standard procedures: Akt (9272, RRID:AB_329827), phospho-Akt (S473) (193H12; 4058, RRID:AB_331168), mTOR (7C10; 2983, RRID:AB_2105622), phospho-mTOR (S2448; 2971, RRID:AB_330970), S6 (54D2; 2317, RRID:AB_2238583), phospho-S6 (S240/244) (D68F8; 2215, RRID: AB_331682), flotillin-1 (3253), MYC (D3N8F; RRID:AB_2631168), pMYC (T58) (E4Z2K), and histone H3 (9715, RRID:AB_331563). We extracted proteins using RIPA buffer for total cell lysates or NE-PER kit (Thermo Fisher) for nuclear lysates.

We performed lipid raft isolation using the Minute Plasma Membrane-Derived Lipid Raft Isolation Kit (LR-042, Invent Biotechnologies). We harvested 40 × 10⁶ CUTLL3 cells 48 hours after treatment with vehicle (DMSO) or pitavastatin, washed the cells with ice-cold PBS, and performed the assay based on the manufacturer's instructions.

To generate CUTLL3 cells with inducible expression of Cas9, we first infected this cell line with pCW-Cas9 (Addgene, 50661) lentiviral particles followed by selection with 2.5 μg/mL puromycin for 4 days. We isolated single-cell clones by limiting dilution and identified doxycycline-inducible Cas9-expressing cells by Western blot following doxycycline treatment. Briefly, we infected Cas9-inducible CUTLL3 cells with a genome-wide gRNA library (SureGuide GeCKO v2 Human Exome CRISPR Library; Agilent) at a multiplicity of infection of <1 and selected GFP-positive infected cells by flow cytometry cell sorting. Library representation was evaluated by PCR amplification and sequencing of gRNAs. We induced Cas9 expression with 1 μmol/L doxycycline in cultures containing 250× representation of the library for 12 days in triplicate. We extracted genomic DNA with phenol-chloroform using standard procedures. We amplified DNA containing gRNAs by PCR and then sequenced these on an Illumina NextSeq 500 instrument loaded with a 20% spike-in of PhiX DNA. We analyzed sequencing data and normalized gRNA abundance using MAGeCK. We used MAGeCKFlute to generate rankview plots in R.

We performed reverse ChIP assays as previously described (47). Briefly, we generated BENC-C DNA bait sequences by PCR from human genomic DNA using a BENC-C biotinylated forward primer (5′-GCCCACAAGCCTACATCCTT-3′) and an unmodified BENC-C reverse primer (5′-TCTTAGGGGTGGGCAGTTCT-3′). Then, we conjugated DNA baits to streptavidin beads and incubated them with nuclear protein extracts from CUTLL3 cells, using unconjugated beads as negative control. We analyzed BENC-C pulled-down proteins by mass spectrometry at the Columbia University Irving Medical Center Mass Spectrometry Proteomics Shared Resource. The MS/MS spectra were searched against the Uniprot human reference proteome database using Sequest within Proteome Discoverer (RRID:SCR_014477). A 1% FDR cutoff was applied on the peptide level using a standard target-decoy database strategy. All proteins identified with less than four unique peptides were excluded from analysis. We filtered peptides against the Contaminant Repository for Affinity Purification (CRAPome) contaminant list (≤10% of the CRAPome).

We performed RNA library preparations using the Illumina TruSeq Stranded mRNA kit with polyA selection and performed sequencing on an Illumina HiSeq platform (Illumina) at Genewiz. We aligned RNA-seq raw reads to the human genome GRCh37 (hg19) using HiSat2 (v.4.8.2, RRID:SCR_015530). For the 50 T-ALL patient samples, we quantified transcript abundance using stringtie, using Gencode assembly (v.19). For all other RNA-seq samples, we generated count matrices using HTSeq (v.0.12.4, RRID:SCR_005514) using Gencode assembly (v.33, RRID:SCR_014966). Samples were segregated in ETP-ALL and T-ALL groups based on differential gene expression of gene sets associated with each of these diagnostic categories as before (9). We performed differential gene-expression analysis in R (v.3.6.3, RRID:SCR_001905) using the Bioconductor package DESeq2 (RRID:SCR_000154).

We performed ChIP-seq analysis from CUTLL3 cells treated with vehicle or pitavastatin for 48 hours. To perform H3K27 acetylation ChIP, we cross-linked CUTLL3 cells with 1% formaldehyde in PBS for 10 minutes at room temperature. We quenched the reaction by adding glycine up to 0.125M and incubated it for 5 minutes at room temperature. We lysed 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 using a Bioruptor (Diagenode). We incubated diluted chromatin overnight at 4°C with antibodies recognizing H3K27 acetylation (Abcam, ab4729, RRID: AB_2118291). We incubated protein A/G Dynabeads (Thermo Fisher Scientific) with chromatin–antibody complexes for 1 hour at 4°C. We washed samples with the following buffers: low salt buffer (0.1% SDS, 1% Triton X-100, 2 mmol/L EDTA, 20 mmol/L Tris pH 8, 150 mmol/L NaCl), high salt buffer (0.1% SDS, 1% Triton X-100, 2 mmol/L EDTA, 20 mmol/L Tris pH 8, 500 mmol/L NaCl), LiCl buffer (2 mmol/L EDTA, 20 mmol/L Tris pH 8, 0.25M LiCl, 1% NP-40, 1% sodium deoxycholate), and TE buffer. We carried out elution 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 (Thermo Fisher Scientific) and proteinase K (Sigma-Aldrich) overnight, and cleaned up chromatin samples using MicroChIP DiaPure columns (Diagenode). We used the TruSeq ChIP library preparation kit (Illumina) for library preparation at Genewiz. We performed sequencing on an Illumina HiSeq platform (Illumina). We trimmed reads of contaminating adapter sequences using Trimmomatic (v.0.29, RRID:SCR_011848) and aligned the reads to the GRCh37 (hg19) build human genome using Bowtie2 (v.2.3.4.1, RRID:SCR_016368). We processed mapped reads with Samtools (v.1.7, RRID:SCR_002105) and generated reads per kilobase million–normalized bigwigs with DeepTools (v.3.3.1, RRID:SCR_016366) with default parameters. We used bigwigCompare for computing the reciprocal ratio of vehicle versus pitavastatin-treated ChIP-seq tracks, with a bin size of 1,000 bp and filtering out signal intensity <5.

We performed ATAC-seq analysis from CUTLL3 cells treated with vehicle or pitavastatin for 48 hours and human lymphoblasts from three patients with ETP-ALL. We generated transposed DNA fragments as described before (47) and amplified them by PCR using NEBNext High-Fidelity 2 PCR Master Mix (NEB M0541) and custom primer indexes to generate ATAC-seq libraries. We purified PCR products using Agencourt AMPure XP beads (Beckman Coulter A63880) and sequenced them on an Illumina HiSeq instrument (Illumina) at Genewiz. We processed fastq files as above (ChIP-seq). We used MACS2 (2.1.2, RRID:SCR_013291) to call peaks with default parameters and generated an intersection peak matrix using bedtools (v2.28.0, RRID:SCR_006646). We used bigwigCompare from DeepTools to compute the reciprocal ratio of vehicle versus pitava-statin-treated ATAC-seq tracks with a bin size of 200 bp and filtering out signal intensity <10. For assessing changes upon pitavastatin treatment, signal was quantified using DeepTools (v 3.3.1) and the pitavastatin-to-vehicle intensity ratio was used to rank the changes.

We performed 3C analysis from CUTLL3 cells treated with vehicle or pitavastatin for 48 hours as previously described (48) using DpnII as the restriction enzyme. Bacterial artificial chromosome clones were used as control template to cover the genomic region under study. We analyzed 3C libraries by 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′-GGGTGTCTTTTCTCCCATTCCTGCGCTAT-3′), BENC-C (5′-TGTCAGGGTGGGAGGAGCACAGTGAGCAA-3′), and a BENC-C neighboring region (5′-AAATTAGCCAGGCTTGGTGGCGGGCGCCTGTA-3′).

All statistical tests and values/meaning of n are listed in the figure legends. The following statistical tests were used as appropriate: Mann–Whitney rank-sum test, two-tailed unpaired Student t test, and permutation test (n = 1,000 trials).

RNA-seq, ChIP-seq, and ATAC-seq data are available from the Gene Expression Omnibus (GEO) with accession number GSE164932.

We performed gene-expression analysis on pediatric ETP-ALL and T-ALL patient samples from Affymetrix array deposited in GEO under accession GSE33315 and 50 ETP-ALL and T-ALL patient samples included in the European Genome-phenome Archive under series EGAS00001004810.

M. Rashkovan reports grants from the Damon Runyon Cancer Research Foundation during the conduct of the study. M.S. Tallman reports grants from AbbVie, Orsenix, Biosight, Glycomimetics, Rafael Pharmaceuticals, and Amgen and other support from AbbVie, Daiichi Sankyo, Orsenix, KAHR, Rigel, Delta Fly Pharma, Tetraphase, Oncolyze, Jazz Pharma, Roche, Biosight, Novartis, Innate Pharmaceuticals, Kura, Syros Pharmaceuticals, and Ipsen Biopharmaceuticals outside the submitted work, and is Past President of the American Society of Hematology (2021). A.A. Ferrando reports grants from the NCI and Alex's Lemonade Stand Foundation during the conduct of the study; personal fees from Bristol Myers Squib, the American Society of Hematology, SpringWorks Therapeutics, and VantAI and grants from Hyundai, the Leukemia & Lymphoma Society, Pershing Square Sohn Cancer Research Alliance, and Alex's Lemonade Stand Foundation outside the submitted work; and a patent for 20130064810 issued, a patent for 20110118192 issued, a patent for 20100093684 issued, a patent for 20100087358 issued, and a patent for 20150299801 issued. No disclosures were reported by the other authors.

M. Rashkovan: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. R. Albero: Formal analysis, investigation, writing–original draft, writing–review and editing. F. Gianni: Investigation. P. Perez-Duran: Investigation. H.I. Miller: Investigation. A.L. Mackey: Investigation. E.M. Paietta: Conceptualization and resources. M.S. Tallman: Resources. J.M. Rowe: Resources. M.R. Litzow: Resources. P.H. Wiernik: Resources. S. Luger: Resources. M.L. Sulis: Resources. R.K. Soni: Formal analysis, investigation, and methodology. A.A. Ferrando: Conceptualization, resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

This work was supported by the St. Baldrick's Foundation (A.A. Ferrando); the Chemotherapy Foundation (A.A. Ferrando); a Crazy 8 Pilot Project Award from Alex's Lemonade Stand Foundation (A.A. Ferrando); and the NIH grants P30 CA013696 (Genomics and High-Throughput Screen Shared Resource, Flow Cytometry Shared Resource, Oncology Precision Therapeutics Shared Resource), R35 CA210065 (A.A. Ferrando), UG1 CA233332 (A.A. Ferrando, ECOG-ACRIN), CA180827 (E.M. Paietta), CA196172 (E.M. Paietta), UG1 CA232760 (M.R. Litzow), UG1 CA233290 (M.S. Tallman), CA180820 (ECOG-ACRIN), CA189859 (ECOG-ACRIN), and CA233332 (ECOG-ACRIN). M. Rashkovan is a Damon Runyon-Sohn Pediatric Cancer Fellow supported by the Damon Runyon Cancer Research Foundation (DRSG-2017). R. Albero is supported by a Leukemia & Lymphoma Society postdoctoral fellowship. F. Gianni is supported by an American–Italian Cancer Foundation postdoctoral fellowship.

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.