A Therapeutically Targetable NOTCH1–SIRT1–KAT7 Axis in T-cell Leukemia

T-cell acute lymphoblastic leukemia (T-ALL) is a rare and aggressive hematologic cancer characterized by its frequent infiltration of the central nervous system and other organs (1). T-ALL affects both children and adults, accounting for 10% to 15% of pediatric and 20% to 25% of adult ALL diagnosed cases (1). Patient outcomes in the last decades have improved thanks to intensive chemotherapy regimens (2, 3). Despite the refinement of these therapies, 20% of pediatric and 50% of adult T-ALL cases still experience drug resistance and eventually relapse, showing dismal clinical outcomes with cure rates of less than 30%. In addition, patients who are cured typically suffer from long-term debilitating comorbidities, overall underscoring the need to identify novel therapeutic approaches for the treatment of this disease (4).

The detection of highly prevalent activating mutations in NOTCH1 in 60% of the patients underscores the role of NOTCH1 as the main oncogenic driver in T-ALL (5). Hence, the development of γ-secretase inhibitors (GSI), which block NOTCH1 signaling via inhibition of a proteolytic cleavage required for NOTCH1 signaling (6), offered a potential targeted strategy for the treatment of T-ALL. However, the clinical development of these drugs has been hindered by limited therapeutic responses observed in clinical trials, further exasperated by on-target gut toxicity (7). In this setting, cancer-specific metabolic rewiring and epigenetic remodeling represent critical hallmarks of cancer (8, 9), and previous studies from our lab and others have specifically demonstrated the importance of both metabolic (10) and epigenetic (11) mechanisms mediating resistance to NOTCH1 inhibition in vivo. Thus, we postulated that central regulators that control both the metabolic and epigenetic status of cells could act as master regulators of NOTCH1-induced transformation and yield novel therapeutic targets in T-ALL.

In this context, we investigated the role of the NAD⁺-dependent SIRT1 histone deacetylase in T-ALL. SIRT1 is a well-known metabolic and epigenetic master regulator (12–16), which deacetylates H4K16ac and H3K9ac (14, 17) and also regulates glycolysis and lipolysis through its complex interactions with PGC1α (18), HIF2α (19), PPARγ (20), FOXO1 (21), and AMPK (22). In addition, SIRT1 has been shown to deacetylate key proteins in T-ALL, such as NOTCH1 (23), MYC (24), AKT (25), or PTEN (26), and it has been linked to the deacetylation of other relevant proteins in cancer such as p53 (27, 28) or CDK2 (29). Specifically related to T cells, previous literature showed that SIRT1 loss in the hematopoietic compartment failed to affect normal T-cell development (30). Regarding its role in hematologic malignancies, SIRT1 has been previously shown to play both oncogenic or tumor suppressor roles in different contexts depending on disease type and oncogenic driver. Indeed, SIRT1 was shown to promote leukemogenesis in chronic myeloid leukemia (CML; refs. 31, 32) and FLT3-ITD acute myeloid leukemia (AML; Ref. 33). On the other hand, SIRT1 activation in vivo showed antiproliferative effects against MLL-rearranged leukemia (34) and myelodysplastic syndrome (MDS; ref. 35). However, the potential role of SIRT1 in T-ALL in vivo is not well understood.

Here, we dissect the role of SIRT1 in T-ALL using a comprehensive approach that combined the use of primary mouse models with different genetic dosages of Sirt1 with transcriptional and epigenetic profiling analyses from both human and mouse leukemias, as well as metabolomic and acetyl-proteomic analyses upon SIRT1 loss in T-ALL in vivo.

We first investigated the expression of SIRT1 in ALL cell lines taking advantage of the Cancer Cell Line Encyclopedia (CCLE; ref. 36). These analyses revealed that SIRT1 levels in ALL are the highest among the different types of leukemias analyzed, including AML, CLL, and CML samples (Supplementary Fig. S1A). Subsequent analyses of gene-expression profiling data from pediatric T-ALL patients (37) revealed a significant upregulation of SIRT1 expression in T-ALL compared with healthy thymocyte subsets (Fig. 1A). Interestingly, SIRT1 levels were similar across different T-ALL clinical subgroups (38), with HOXA-driven T-ALLs showing relatively lower levels (Supplementary Fig. S1B). In line with this, we found SIRT1 protein levels to be significantly upregulated in T-ALL cells compared with healthy human peripheral mononuclear cells, CD4⁺ T cells or human thymus samples (Fig. 1B). SIRT1 is not known to be duplicated or mutated in T-ALL (38). As NOTCH1 activating mutations are observed in ∼60% of T-ALL patients (5), these results led us to hypothesize that NOTCH1 might be regulating SIRT1 expression. Arguably, the most accurate way to uncover NOTCH1 direct targets is to use GSI washout methods (39). These assays allow NOTCH1 reactivation in the presence of cycloheximide (CHX) or dominant-negative MAML1 (DN-MAML1, a specific inhibitor of NOTCH1 signaling) upon removal of GSI (39). High-confidence direct NOTCH1 targets are defined by a ≥2-fold increase within 4 hours of GSI washout that is insensitive to CHX (independent of new protein translation) but sensitive to DN-MAML1 (dependent on NOTCH1 binding), such as canonical NOTCH1 targets like HES1 (Supplementary Fig. S1C). Notably, SIRT1 expression in T-ALL cells increased ∼4-fold after GSI removal, and this effect was insensitive to CHX but was rescued by DN-MAML1 (Fig. 1C), demonstrating that SIRT1 is a direct target of NOTCH1 in T-ALL. Indeed, short-term treatment with GSIs in NOTCH1-driven human or mouse T-ALL cell lines in vitro led to a slight but consistent downregulation of SIRT1 protein levels (Fig. 1D). Similar results were observed in NOTCH1 wild-type T-ALL cells (Supplementary Fig. S1D). Next, epigenetic profiling analyses revealed that NOTCH1 binding to the SIRT1 promoter in T-ALL was modest; however, it uncovered a discrete NOTCH1-binding peak located ∼35 Kb upstream of the SIRT1 promoter (Fig. 1E). Importantly, NOTCH1 binding to this region was shown in both CUTLL1 and HPB-ALL, two independent T-ALL cell lines. This region also exhibited additional features of NOTCH1 canonical signaling, such as the binding of RBPJ or ETS1, as well as bona fide enhancer marks such as enrichment of H3K27ac and binding of BRD4 and ZNF143 (Fig. 1E). In addition, Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) data of two independent human primary T-ALL samples revealed that this region shows accessible chromatin (Fig. 1E). Finally, further epigenetic profiling upon NOTCH1 inhibition in CUTLL1 cells showed a slight decrease in H3K27ac signal in this region (Supplementary Fig. S1E), together with increased RNA transcription upon GSI washout (Supplementary Fig. S1F). Thus, we postulated that this region (which we named N-Se, for NOTCH1-bound SIRT1 enhancer) may act as an enhancer of SIRT1 driven by NOTCH1, in an analogous way to our previously identified N-Me enhancer in the control of MYC (40).

Luciferase reporter assays in T-ALL cells testing this hypothesis showed strong, orientation-independent activation of reporter constructs containing N-Se (Fig. 1F). Moreover, CRISPR/Cas9-induced deletion of N-Se in JURKAT cells resulted in reduced SIRT1 expression (Fig. 1G and H). Finally, repression/activation assays targeting N-Se in DND41 T-ALL cells via CRISPRi/a led to decreased SIRT1 levels after CRISPRi repression (dCas9-KRAB) or increased SIRT1 levels after CRISPRa activation (dCas9-VP64), to a similar extent as targeting the SIRT1 promoter itself (Fig. 1I). These results demonstrate that this distal regulatory region controls SIRT1 expression downstream of NOTCH1, which might underlie the broad upregulation of SIRT1 observed in T-ALL patients.

We next hypothesized that targeting SIRT1 might result in antileukemic effects in human T-ALL. Indeed, knockdown of SIRT1 in DND41 T-ALL cells with two independent doxycycline-inducible shRNAs led to reduced SIRT1 levels (Fig. 2A and B), together with a concomitant reduction in proliferation (Fig. 2C) and cytotoxic effects, as revealed by increased apoptosis (Fig. 2D and E).

Moreover, we tested the effects of pharmacologically inhibiting SIRT1 using EX-527, a well-described SIRT1-specific inhibitor (41), alone or in combination with the GSI dibenzazepine (DBZ). Importantly, EX-527 treatment in DND41 cells impaired cell growth (Fig. 2F) and showed strong and synergistic antileukemic activity in combination with DBZ (combination index = 0.3 using isobologram analyses; Fig. 2G). Similar to the genetic inhibition of SIRT1, this effect was primarily driven by increased cytotoxicity, as revealed by increased Annexin V–positive apoptotic cells (Fig. 2H and I). Overall, our results demonstrate that targeting SIRT1, either genetically or pharmacologically, leads to antileukemic and synergistic effects with NOTCH1 inhibition in vitro.

Following up on these results, we tested the effects of different Sirt1 allelic dosages in NOTCH1-induced T-ALL generation. To this end, we generated NOTCH1-induced leukemias in mice as previously described by using retroviral expression of an oncogenic version of NOTCH1 (ΔE-NOTCH1, lacking NOTCH1 extracellular domain) in bone marrow progenitor cells followed by transplantation into irradiated mice (10). In this context, NOTCH1-induced leukemias generated from Sirt1-overexpressing bone marrow progenitors (42) showed significantly accelerated kinetics of leukemia development as compared with those generated from Sirt1 wild-type mice (Fig. 3A–C). Conversely, tamoxifen-induced deletion of Sirt1 only 2 days after transplantation of Sirt1-conditional knockout bone marrow progenitor cells infected with the same oncogenic form of NOTCH1 resulted in delayed leukemia generation and reduced disease penetrance (Fig. 3D–F). Finally, we postulated that SIRT1 might also contribute to mediate resistance to NOTCH1 inhibition in vivo. To test this hypothesis, we overexpressed Sirt1 in already established mouse primary GSI-sensitive NOTCH1-induced leukemias and analyzed its effects on the response to GSI treatment, as previously described (10). In this context, mice were transplanted with a combination of nontransduced (GFP-only) or transduced (GFP/mCherry double positive) T-ALL cells with either the Sirt1 wild-type deacetylase or a catalytically dead, H355A-mutant version of Sirt1 (ref. 18; Fig. 3G). Mice transplanted with these T-ALL cells were then treated with the DBZ GSI daily and leukemia progression was analyzed in peripheral blood (Fig. 3G). Notably, although nontransduced GFP-only cells showed little or no progression under treatment, T-ALL cells overexpressing Sirt1 did progress under GSI treatment in vivo (Fig. 3H–I). Importantly, this effect was mediated by SIRT1 deacetylase activity because overexpression of the catalytically dead, H355A-mutant version resulted in the abrogation of this competitive advantage (Fig. 3H and I). Consistent with our findings, GSI-persister human T-ALL cells upon long-term exposure to suboptimal doses of GSI also showed upregulation of SIRT1 (ref. 11; Supplementary Fig. S2). Together, our results demonstrate that SIRT1 promotes NOTCH1-induced T-ALL development and mediates resistance to NOTCH1 inhibition in vivo in a deacetylase-dependent manner.

We next investigated the therapeutic effects of targeting Sirt1 in already established NOTCH1-induced leukemias in vivo. To this end, we followed the same retroviral-transduction protocol described before using bone marrow cells from Sirt1-conditional knockout mice, but we first let leukemias fully develop before analyzing the effects of secondary SIRT1 loss (Fig. 4A). Here, we tested the effects of SIRT1 loss not only in leukemias driven by a ΔE-NOTCH1 construct, but also in leukemias driven by a weaker oncogenic version of NOTCH1 that harbors the prototypical mutations targeting the NOTCH1 heterodimerization domain (HD) and the PEST domain (HDΔP-NOTCH1), which are recurrently observed in patient samples (ref. 5; Fig. 4A). In this context, tamoxifen-induced isogenic deletion of Sirt1 in already established primary Sirt1-conditional knockout leukemias led to significant antileukemic effects on its own, which were further enhanced by GSI treatment in vivo, regardless of the oncogenic version of NOTCH1 used (Fig. 4B). Interestingly, leukemias that eventually relapsed after tamoxifen injection did show SIRT1 loss (Fig. 4C), suggesting that T-ALL cells can still adapt to survive Sirt1 deletion. Finally, in an effort to better model what would happen in patients treated with a pharmacologic SIRT1 inhibitor, we generated a novel knockin mouse model with conditional expression of the Sirt1 H355A catalytically dead allele (18) from its own endogenous locus (Sirt1^cKI-H355A/+; Supplementary Fig. S3A–S3E). We next bred these mice with our mice expressing a tamoxifen-inducible Cre from the Rosa26 locus in order to generate NOTCH1-induced Sirt1^H355A-conditional knockin leukemias using the same approach as before (Fig. 4D). In this setting, secondary replacement of wild-type Sirt1 by catalytically dead H355A Sirt1-mutant in already established T-ALL in vivo led to significant survival extension (Fig. 4E–G). Overall, our results demonstrate that the antileukemic effects of SIRT1 loss are enhanced by NOTCH1 inhibition in vivo and further highlight the relevance of SIRT1 enzymatic activity in leukemia progression.

To further dissect the acute effects of SIRT1 loss in T-ALL in vivo, we transplanted the same ΔE-NOTCH1 Sirt1-conditional knockout leukemias in mice and let them become overtly leukemic before inducing SIRT1 loss. In this setting, analyses of leukemic spleens only 48 hours after tamoxifen injection revealed drastically reduced SIRT1 levels (Fig. 5A and B), together with reduced tumor burden (Fig. 5C and D) at the expense of increased apoptosis (Fig. 5E and F). Similar effects were observed upon acute loss of SIRT1 in mice harboring HDΔP-NOTCH1 Sirt1-conditional knockout leukemias (Supplementary Fig. S4A–S4F).

SIRT1 is a well-known metabolic regulator, and SIRT1 loss has been shown to impact metabolism in other hematologic malignancies (32); thus, we next investigated the metabolic effects of its loss. In addition to increased acetylation levels of p53 and p65 NFκB, two well-described SIRT1 targets (Supplementary Fig. S5A and S5B), we observed activation of AMPK and downregulation of mTOR signaling as revealed by decreased p4E-BP1 levels, together with reduced levels of the stress response factor ATF4 (Fig. 5G; Supplementary Fig. S5C). We next performed global untargeted metabolomic analyses (LC-MS) upon acute loss of SIRT1 in T-ALL in vivo. These experiments revealed that SIRT1 loss leads to global metabolic changes (Fig. 5H), including accumulation of glycolytic intermediates (Fig. 5I), together with increased levels of glutamine, and glutamine-derived metabolites such as glutamate, aspartate, and asparagine (Fig. 5J). In addition, seahorse analyses in Sirt1-conditional knockout leukemia-derived cell lines revealed a global reduction in oxygen consumption rate (OCR) upon SIRT1 loss (Fig. 5K; Supplementary Fig. S5D). Using these murine tumor–derived cell lines, we also observed synergistic effects of SIRT1 loss with NOTCH1 inhibition in vitro driven by increased cytotoxicity (Supplementary Fig. S5E–S5H), consistent with the effects previously observed in human cells (Fig. 2).

We next asked whether the strong AMPK activation observed might be mediating the therapeutic effects of SIRT1 loss. To address this question, we generated double Sirt1/Ampk-conditional knockout leukemias following the same approach previously described (Supplementary Fig. S6A). However, tamoxifen-induced secondary loss of SIRT1 in combination with AMPK still resulted in significant antileukemic effects with extended survival (Supplementary Fig. S6B and S6C), suggesting that AMPK activation is dispensable for the therapeutic effects of SIRT1 loss. Similarly, given the global metabolic changes affecting glycolysis and glutaminolysis, we hypothesized that metabolic rescues using membrane-soluble compounds that can bypass glycolysis or glutaminolysis blocks, such as methyl-pyruvate (MP) or dimethyl-2-oxoglutarate (DMKG), might rescue the effects of SIRT1 loss, similar to what we previously described upon NOTCH1 inhibition (10). Yet, the addition of either MP or DMKG did not prevent the antileukemic effects and impaired proliferation upon SIRT1 loss in vitro (Supplementary Fig. S6D and S6E). Overall, these data show that SIRT1 loss leads to AMPK activation and global metabolic changes, although these might be dispensable for (or secondary to) its therapeutic effects.

Given the deacetylase-dependent effects of SIRT1 on the resistance to NOTCH1 inhibition and leukemia progression (Figs. 3H and 4E), we next postulated that unbiased acetyl-proteomic experiments might reveal relevant targets for the mechanistic effects of SIRT1 loss in T-ALL. To this end, we performed mass spectrometry analyses upon acetyl-lysine enrichment in Sirt1-conditional knockout T-ALLs treated acutely with tamoxifen or vehicle in vivo (Fig. 6A). These experiments revealed multiple hyperacetylated peptides upon SIRT1 loss (Supplementary Table S1). Interestingly, integrative results using our two independent Sirt1-conditional knockout leukemias revealed only nine targets that were consistently hyperacetylated (Fig. 6A). These notably included hyperacetylation of histone acetyltransferase 7 (KAT7, also known as HBO1) on lysine 277, without affecting total KAT7 levels (Fig. 6A; Supplementary Table S1); we also detected hyperacetylation of Bromodomain-containing protein 1 (BRD1) on Lys 418. Moreover, SIRT1 inhibition with EX-527 in human JURKAT T-ALL cells also led to strong hyperacetylation of KAT7 (Fig. 6B). Because both KAT7 and BRD1 are known to form part of a histone acetyltransferase complex that targets H4K12 and H3K14 for acetylation (43), our results suggested that SIRT1 loss might be affecting the activity of this complex. Therefore, we performed gene-expression profiling upon acute loss of SIRT1 in T-ALL in vivo. Although these analyses revealed global transcriptional changes upon SIRT1 loss (Fig. 6C; Supplementary Table S2), gene set enrichment analyses (GSEA) against the C2 database showed no evident signature related to glycolysis, glutaminolysis, oxidative phosphorylation, and/or PGC1α as significantly enriched upon Sirt1 deletion (Supplementary Table S3). Recent literature has shown KAT7 loss leads to antileukemic effects in MLL-rearranged leukemias (44). Thus, we tested our transcriptional signature after the loss of SIRT1 against the one observed upon KAT7 loss (44) and uncovered a significant correlation between both (Fig. 6D), suggesting that SIRT1 loss leads to hyperacetylation of KAT7, which might be less active. Consistent with this hypothesis, we detected a global reduction in H4K12ac upon acute loss of SIRT1 in mouse T-ALL in vivo (Fig. 6E). Moreover, doxycycline-induced knockdown of SIRT1 in human DND41 T-ALL cells led to similarly reduced levels of H4K12ac (Fig. 6F). Further, ChIP-seq epigenetic profiling of H14K12ac upon deletion of Sirt1 confirmed a global reduction in H4K12ac peaks in vivo (Fig. 6G; Supplementary Table S4). In addition, integration of these data with our gene-expression profiling data revealed that regions showing reduced H4K12ac levels were significantly enriched in the promoter of downregulated genes upon SIRT1 loss (Fig. 6H; Supplementary Fig. S7). Together, our data demonstrate that SIRT1 loss leads to hyperacetylation of KAT7, which correlates with the transcriptional effects of KAT7 loss and results in globally decreased levels of the KAT7-mark H4K12ac.

Our previous data suggested that KAT7 inactivation might drive, at least in part, the antileukemic effects of SIRT1 loss. Thus, we hypothesized that Sirt1-deleted leukemia cells might be more sensitive to KAT7 inhibition than wild-type counterparts. Indeed, experiments using the KAT7-specific inhibitor WM-3835 (45) in our Sirt1-conditional knockout T-ALL cell line revealed significantly stronger antileukemic effects in Sirt1-deleted leukemia cells at different concentrations, resulting in a lower IC₅₀ than in isogenic Sirt1-positive leukemia cells (Fig. 7A). Finally, to formally test whether KAT7 mediates part of the antileukemic effects of SIRT1 loss, we performed rescue experiments overexpressing either wild-type KAT7, an acetyl-mimic (K277Q) or an acetyl-dead (K277R) version of KAT7 in our Sirt1-conditional knockout T-ALL cell line (Fig. 7B). Notably, although wild-type or K277Q KAT7 did not affect the impaired proliferation upon SIRT1 loss (Fig. 7C), overexpression of the acetyl-dead K277R version of KAT7 was able to partially rescue the antileukemic effects of SIRT1 loss, resulting in increased cell proliferation (Fig. 7C) and reduced apoptosis (Fig. 7D). Overall, these results demonstrate that the antileukemic effects of SIRT1 loss are partly mediated by KAT7.

Here, we demonstrate that SIRT1 is a direct target of NOTCH1 in T-ALL through a distal enhancer. Accordingly, SIRT1 is broadly overexpressed in T-ALL. Genetic experiments in vivo demonstrated an oncogenic role for SIRT1 in both leukemia generation and progression such that targeting SIRT1 pharmacologically in vitro, or genetically in vivo, leads to antileukemic and synergistic effects with GSIs. Even if NOTCH1 and SIRT1 are in the same route, the synergism obtained by concomitantly targeting both might be explained by the fact that NOTCH1 inhibition alone does not reduce SIRT1 levels drastically, thus leaving room for further SIRT1 inhibition to obtain synergistic effects. This is conceptually similar to previous findings of synergism between NOTCH1 inhibition and mTOR inhibition (46) or glutaminase inhibition (10). Mechanistically, SIRT1 loss leads to global transcriptional and metabolic changes together with hyperacetylation and inactivation of the KAT7 histone acetyltransferase, which is critical for the therapeutic effects of SIRT1 loss. Consistently, a nonacetylated mutant form of KAT7 partially rescues the effects of SIRT1 loss on leukemia proliferation. Overall, we uncovered a prominent role for SIRT1 in T-ALL generation and progression, as well as a therapeutic target in T-ALL. In this context, it is relevant to note the multiple efforts and compounds described to date to either pharmacologically activate or inhibit SIRT1 (47). Still, the field of SIRT1 pharmacologic modulation has been mired in controversy, with many studies disputing the specificity of these compounds, most prominently resveratrol and other purported SIRT1 activators (48). Although described SIRT1 inhibitors are reported to be more specific, none of the inhibitors described to date have been clinically approved, probably owing to their poor bioavailability in vivo (49). Moreover, EX-527 has been shown to inhibit other targets including SIRT2 (50) or SIRT6 (51), and our experiments with EX-527 in vitro showed stronger effects than genetic knockdown of SIRT1. These results suggest that part of its antileukemic activity might be due to SIRT1 OFF-target effects and that, more broadly, sirtuins might be attractive antileukemic targets. However, our results using novel mouse models demonstrate that not only SIRT1 loss but also Sirt1 genetic inhibition shows antileukemic effects in vivo. These encouraging findings will hopefully spur renewed efforts to discover improved SIRT1 inhibitor drugs for clinical development.

Our discovery of a NOTCH1-bound SIRT1 enhancer (N-Se) adds to the growing literature of relevant enhancers in T-ALL, including the N-Me MYC enhancer (40), the PE PTEN enhancer (52), or the de novo acquired enhancers for TAL1 (53) and BCL11B (54), among others (55). Our results unequivocally demonstrate that N-Se regulates SIRT1 levels. Still, the binding of NOTCH1 to N-Se is relatively weak compared with N-Me or other enhancer regions, and NOTCH1 inhibition results in a modest reduction of SIRT1 protein levels. By contrast, genetic deletion/inactivation of N-Se shows stronger effects on SIRT1 levels. Thus, even if NOTCH1 modulates N-Se, it probably does not play a dominant role in its regulation, and it is likely that other transcription factors might be equally or even more relevant in regulating the effects of N-Se. Further studies to mechanistically dissect the role and mechanisms controlling N-Se activity will answer these questions.

Even if NOTCH1 inhibition should lead to a short-term reduction of SIRT1 levels via inactivation of N-Se, human T-ALL cells resistant to NOTCH1 inhibitors show increased SIRT1 expression, and our own results also demonstrate that SIRT1 overexpression confers resistance to NOTCH1 inhibition. We speculate that these contradictory findings might be due to epigenetic mechanisms to upregulate SIRT1 expression through some transcription factor other than NOTCH1 (either by binding and regulating the N-Se enhancer or directly at the SIRT1 promoter level) or through switching to an alternative enhancer different from N-Se, similar to how GSI-resistant T-ALL cells can maintain MYC levels by shifting from the NOTCH1-controlled N-Me enhancer to the BRD4-controlled BDME enhancer (56).

Mechanistically, acute loss of SIRT1 resulted in broad metabolic, transcriptional, and epigenetic changes. From a metabolic point of view, we observed an accumulation of glycolytic and glutamine-derived metabolites upon SIRT1 loss, together with strong activation of AMPK. This is consistent with previous literature describing SIRT1 activates AMPK in a feedforward positive loop (22). As expected, AMPK activation resulted in the downregulation of mTOR signaling together with a reduction in ATF4 levels, consistent with previous findings demonstrating mTOR directly activates ATF4 (57). We also observed a concomitant decrease in OCR, similar to previously described results in CML models with loss of SIRT1 (32). Thus, we decided to formally test the relevance of these metabolic effects on the therapeutic effects of SIRT1 loss. However, experiments in vivo using double Sirt1/Ampk-conditional knockout leukemias still showed significantly extended survival even after the concomitant loss of SIRT1 and AMPK. We previously demonstrated that isogenic loss of AMPK alone does not affect leukemia progression by itself (58). Thus, because the extension in survival observed upon SIRT1/AMPK loss was similar to the one observed upon SIRT1 loss alone, our results suggest that AMPK activation is dispensable for the survival extension upon loss of SIRT1. In addition, the decrease in OCR upon SIRT1 loss in CML models has been mechanistically linked to inhibition of PGC1α (32); yet, our gene-expression profiling data did not show any significant correlation with an inactivation of PGC1α transcriptional program, suggesting that other mechanisms might mediate these effects in T-ALL. Finally, experiments using membrane-soluble metabolites that bypass blocks in glycolysis and glutaminolysis failed to rescue the impaired proliferation observed upon SIRT1 loss. Together, these results suggest that, although the metabolic effects of SIRT1 loss may play an important role, they are likely not the main mechanistic mediators of its phenotype in T-ALL and, thus, they might be more a consequence rather than a cause of its antileukemic effects.

To gain further insights into the mechanistic effects of SIRT1 loss, we performed global unbiased acetyl-proteomic analyses, together with gene expression and epigenetic profiling in our Sirt1-conditional leukemias. Even if SIRT1 was previously shown to deacetylate NOTCH1 (23), we failed to observe NOTCH1 hyperacetylation upon SIRT1 loss in our mass spectrometry data. Still, it is tempting to speculate with the possibility of circular feedback loops between SIRT1 and NOTCH1, similar to the ones described before between SIRT1 and MYC (24). Similarly, although SIRT1 has been previously suggested to affect leukemia progression via deacetylation of CDK2 (29), we failed to detect CDK2 hyperacetylation upon SIRT1 loss in two independent Sirt1-conditional leukemias. On the other hand, our analyses revealed KAT7 and BRD1 as consistently hyperacetylated upon SIRT1 loss. This finding caught our attention because KAT7 and BRD1 are known to form part of the same histone acetyltransferase complex targeting H4K12 and H3K14 (43). Moreover, KAT7 has been recently shown to play important roles in AML (44, 45, 59), which suggests that its hyperacetylation might be involved in the mechanistic therapeutic effects of SIRT1 loss. Indeed, gene-expression profiling upon SIRT1 loss in T-ALL in vivo revealed a transcriptional signature similar to the one observed upon KAT7 loss. Consistently, we observed globally reduced levels of H4K12ac upon SIRT1 loss in mouse or human T-ALL cells. Finally, overexpression rescue experiments in our Sirt1-conditional knockout leukemia-derived cell line using either wild-type KAT7 or acetyl-mimic (K277Q) or acetyl-dead (K277R) KAT7-mutant versions showed that overexpression of the nonacetylated K277R KAT7 partially rescued the antiproliferative and cytotoxic effects of SIRT1 loss. Overall, our experiments demonstrate that KAT7 mediates, at least in part, the antileukemic effects of SIRT1 loss.

Previous studies have shown that KAT7 is important for the recruitment and processivity of RNApolII (45). Thus, it is tempting to speculate that this effect might be relevant for the broad transcriptional changes observed upon SIRT1 loss. Similarly, KAT7 loss has been shown to impair leukemic stem cell activity in AML (45), as well as to lead to the exhaustion of hematopoietic stem cells (59). Interestingly, SIRT1 itself has been shown to play a role in hematopoietic stem cells and leukemic stem cell maintenance in CML (31, 32, 60); thus, it is also fair to speculate that these phenotypes might be mediated by KAT7. Still, further studies are warranted to fully understand the specific effects of KAT7 in T-ALL, as well as the relevance of its acetylation.

Our results demonstrate that KAT7 partly mediates the effects of SIRT1 loss in T-ALL; still, the proliferative rescue observed is not total. This might be partly due to the high levels of endogenous KAT7 present in leukemia cells, which might obscure the effects of additional wild-type/mutant KAT7 overexpression. In addition, it may be possible that concomitant overexpression of acetyl-dead mutants for both KAT7 and BRD1 is needed to completely rescue the effects of SIRT1 loss, because they are both involved in the same histone acetyltransferase complex. It is also possible that some of the other seven consistently hyperacetylated proteins upon loss of SIRT1 might play relevant roles and partly influence the antileukemic effects of Sirt1 deletion. Among these, the hyperacetylation in NPM1 was particularly interesting, given the well-described roles of NPM1 mutations in AML (61). Still, nothing is known regarding a putative role for NPM1 in T-ALL; thus, further studies are needed to clarify the potential relevance of these findings.

Moreover, SIRT1 has already been described to play pleiotropic roles in the literature, where it has been shown to affect the acetylation of multiple targets with relevant roles in leukemia. Although we did not detect increased acetylation of NOTCH1, CDK2, or MYC in our mass spectrometry data, we detected an expected increase in the acetylation of p53 upon SIRT1 loss, which is known to activate p53 (27, 28). Thus, part of the antileukemic effects observed upon SIRT1 loss might be mediated by p53. Still, we observed antileukemic effects for genetic or pharmacologic inhibition of SIRT1 in human DND41 T-ALL cells, which are known to be p53 mutant (62), suggesting that p53 might not be the central factor mediating the observed effects. Overall, our data revealed wide-ranging effects upon SIRT1 loss in leukemia and demonstrated that part of its antileukemic effects are mediated by the hyperacetylation and inhibition of KAT7.

Finally, our data uncovered a feedback mechanism balancing the activity of histone deacetylases and acetyl-transferases. Our results demonstrate that reduced activity of the SIRT1 histone deacetylase is paralleled by a concomitant increase in the acetylation of KAT7, resulting in inhibition of its acetyltransferase activity (Fig. 7E). Thus, a decrease in the histone deacetylase activity of SIRT1 would be paralleled by a decrease in KAT7 histone acetyltransferase activity. Conversely, an increase of SIRT1 histone deacetylase activity should be paralleled by reduced KAT7 acetylation together with an increase in its histone acetyltransferase activity (Fig. 7E). In this way, the interplay between SIRT1 and KAT7 constitutes a rheostat mechanism regulating the total epigenetic output in terms of histone acetylation levels, which is reminiscent of other well-described relevant feedback mechanisms, such as the tightly regulated interplay between protein kinases and phosphatases in mitosis and beyond (63, 64). Thus, although our data unveil a SIRT1–KAT7 link with clear relevance in T-cell leukemia, our results point to the existence of deacetylase-histone acetyltransferase balancing mechanisms which might be broadly relevant across different biological processes and cancer types.

Overall, our results reveal an oncogenic role for SIRT1 in T-ALL generation and progression downstream of NOTCH1, identify SIRT1 as a therapeutic target for the treatment of T-ALL and uncover a SIRT1–KAT7 link that might represent a more broadly relevant mechanism in cancer.

Cells were cultured in standard conditions in a humidified atmosphere in 5% CO₂ at 37°C in HyClone RPMI-1640 Media (Fisher Scientific, SH3002701) with 10% to 20% FBS (Gemini Bio-Products, 900-108) and 100 U/mL penicillin and 100 μg/mL streptomycin (VWR, 45000-652). DND41 (ACC 525), HPB-ALL (ACC 483), JURKAT (ACC 282), and TALL-1 (ACC-521) cells were obtained from Deutsche Sammlung von Mikroorganismen und Zellkulturen. Tumor-derived cell lines from mouse primary T-ALLs were generated and cultured in Opti-MEM (Life Technologies, 57985091) with 10% FBS, as previously described (40). Drugs used for different experiments include EX-527 (Sigma, E7034), DBZ (Syncom, 29762), WM-3835 (Tocris, 7366), 4-hydroxytamoxifen (Sigma, H7904), doxycycline (Sigma, D9891), puromycin (Sigma, P8833), methyl-pyruvate (Sigma, 371173), and dimethyl-2-oxoglutarate (Sigma, 349631).

To analyze cell survival, cells were plated in triplicates on 24-well plates. Cell lines were seeded at 300,000 cells per well in a final volume of 1 mL RPMI-1640 media under the experimental conditions described in each experiment. 1 μmol/L 4-hydroxitamoxifen was added to induce Sirt1 deletion when indicated. Doxycycline (1 μg/mL) was added to induce SIRT1 knockdown when indicated. DBZ (250 nmol/L) was used to inhibit NOTCH1, when indicated. EX-527 (90 μmol/L) was used to inhibit SIRT1 when indicated. Every 3 days of treatment, cells were counted using a Countess II FL instrument (Fisher Scientific).

Viability and cell growth ratios were determined by analyzing cell density by MTT using the Cell Proliferation Kit I (Roche, 11465007001) in DND41 cells treated with EX-527 (10⁻⁶ to 10⁻³ mol/L) for 3 and 6 days, obtaining an IC₅₀ = 90 μmol/L. Synergism was evaluated using the Chou–Talalay method in DND41 cells treated with EX-527, DBZ (IC₅₀ = 250 nmol/L, as previously published; refs. 10, 65) or with the combination of both EX-527 and DBZ using doses equivalent to 0.25 × IC₅₀; 0.5 × IC₅₀; 1 × IC₅₀; 2 × IC₅₀ or 4 × IC₅₀ to calculate synergy. Isobolograms were used to graphically represent the interaction between the two drugs and the combination index was determined using the Calcusyn software package (Biosoft; ref. 66).

Doxycycline-inducible GFP-expressing lentiviral vectors either non­targeting (shControl, VSC11521) or targeting SIRT1 (V3SH11252-225149601 or V3SH11252-230472600) were obtained from Dharmacon. DND41 cells were transduced with each construct and selected with 5 μg/mL puromycin. Selected cells were treated with 1 μg/mL doxycycline to induce the different shRNAs together with concomitant GFP expression for downstream analyses.

For KAT7 overexpression experiments, we first obtained the Kat7 cDNA from Genecopoiea (EX-Mm30842-Lv224-GS). K277Q- or K277R-directed mutagenesis was performed on this vector using the QuikChange II XL Site-Directed Mutagenesis Kit (Agilent, 200521), following the manufacturer's instructions, and constructs were sequence verified. Then, we subcloned the different Kat7 constructs in the multiple cloning site of the pMSCV-mCherry FP vector (Addgene, 52114). Finally, empty vector or different Kat7-containing vectors were used to transduce and sort the tumor-derived Sirt1-conditional knockout leukemia cell line using an Influx High Speed Sorter (BD Biosciences). Sorted cells were then used for downstream analyses.

Custom Alt-R CRISPR-Cas9 crRNA targeting sequences on both flanks of the N-Se region (5′-GTTCATCGGTGGAGTTGAGG-3′ and 5′-AGCGTTTTCAAGTTATAATC-3′, respectively) were obtained from Integrated DNA Technologies (IDT) and used to generate JURKAT N-Se^−/− cells following the manufacturer's instructions. Briefly, each crRNA and Alt-R CRISPR-Cas9 tracrRNA-ATTO 550 (IDT, 1075927) oligos were mixed in equimolar concentrations to a final duplex concentration of 44 μmol/L. Annealing was achieved by heating up the mixture to 95°C for 5 minutes and subsequent slow cooling it down to room temperature (20–25°C). 22 pmol/L of each crRNA:tracrRNA duplex was incubated with 18 pmol/L of Alt-R S.p. Cas9 Nuclease V3 (IDT, 1081059) for 20 minutes at room temperature. After the formation of the crRNA:tracrRNA:Cas9 complexes, the complexes targeting the 5′ and 3′ end of N-Se were mixed in equimolar concentrations. 0.5 million JURKAT cells were electroporated with 40 pmol/L of crRNA:tracrRNA:Cas9 complexes using the Neon Transfection System with 10 μL electroporating tips (MPK1096, Thermo Fisher Scientific) in duplicate. Electroporation conditions were 3 pulses at 1,600 V and 10 ms width. ATTO 550-positive single-cell clones were sorted using the Influx High-Speed Sorter (BD Biosciences) 24 hours after transfection. Clones were screened for N-Se loss by PCR using REDTaq ReadyMix (Sigma, R2523-100RXN) and custom primers (5′-AGCGAGACTCCGTCTAGAAA-3′, 5′-CTCGAACTCCTGACCTCAATC-3′ and 5′-CAGCAAAATTGAGGGGAAAG-3′).

pLV-U6-gRNA-UbC-DsRed-P2A-Bsr plasmid (Addgene, 83919) was used to express gRNAs targeting human SIRT1 transcriptional start site (TSS) and N-Se. Assembly of custom lentiviral vectors expressing sgRNAs of choice was accomplished by Golden Gate cloning and type IIS restriction enzyme BsmBI (NEB, R0739), which cleaves outside its recognition sequence to create unique overhangs, as previously reported (67). Primers used to clone SIRT1 TSS gRNA: sense primer 5′-CACCGGGGCAGCCAAATTCGCCCCT-3′ and antisense primer 5′-AAACAGGGGCGAATTTGGCTGCCCC-3′. Primers used to clone N-Se gRNA1: sense primer 5′-CACCGGGCTCCAGAAGCTCCGAGCG-3′ and antisense primer 5′-AAACCGCTCGGAGCTTCTGGAGCCC-3′. Primers used to clone N-Se gRNA2: sense primer 5′-CACCGGGGCCTCCTTTGCTCGTATT-3′ and antisense primer 5′-AAACAATACGAGCAAAGGAGGCCCC-3′.

pLV hUbC-dCas9 VP64-T2A-GFP lentiviral vector expressing dCas9VP64 (Addgene, 53192) or pLV hUbC-dCas9 KRAB-T2A-GFP lentiviral vector expressing dCas9KRAB (Addgene, 67620) were used to transduce DND41 cells, followed by the selection of GFP-positive cells by fluorescence-activated cell sorting (FACS) using an Influx High-Speed Sorter (BD Biosciences), in order to obtain DND41 stably expressing either dCas9-VP64-GFP or dCas9-KRAB-GFP, respectively. Subsequently, one lentiviral vector encoding for each gRNA-DsRed was used to transduce DND41-dCas9VP64-GFP⁺ or DND41-dCas9KRAB-GFP⁺ models. Cells coexpressing both GFP and DsRed markers (>90%) were further isolated by FACS and processed for the evaluation of SIRT1 expression.

We performed reporter assays using a pGL4.10 Vector (Promega, E6651) luciferase construct alone or coupled with the N-Se enhancer sequence (hg19; chr10:69,608,815–69,610,235; DNA sequence was synthesized by Genewiz), cloned in the forward and reverse orientations, and following our previously described protocol (52). Briefly, we electroporated JURKAT cells with a Neon Transfection System Device (MPK5000, Thermo Fisher Scientific) using 100 μL tips (MPK10025, Thermo Fisher Scientific). Electroporation conditions were as follows: pulse voltage 1,350 V, pulse width 10 ms, pulse number 3, and cell density 10⁷ cells/mL. Constructs were transfected together with a plasmid driving the expression of the Renilla luciferase gene (pCMV-Renilla) used as an internal control. We measured luciferase activity 42 hours after electroporation with the Dual-Luciferase Reporter Assay kit (Promega, E1980).

OCR rates were measured using an XF24 Seahorse Biosciences extracellular flux analyzer (Agilent Technologies) according to the manufacturer's instructions. Briefly, T-ALL cells were resuspended in Seahorse XF RPMI Medium (Agilent Technologies, 103576-100) supplemented with 10 mmol/L glucose (Agilent Technologies, 1003577-100), 1 mmol/L pyruvate (Agilent Technologies, 1003578-100), and 2 mmol/L glutamine (Agilent Technologies, 1003579-100). 5 × 10⁵ cells per well were plated in XF24 Seahorse Biosciences plates precoated with Cell-Tak (Corning, 354240) and spun down on the plate to ensure that cells were completely attached. OCR was analyzed by sequential injections of 1 μmol/L oligomycin (Sigma, O4876), 1 μmol/L FCCP (Sigma, C2920), and 0.5 μmol/L rotenone (Sigma, R8875) and antimycin A (Sigma, A8674) in each well.

Sirt1^H355A-conditional knockin mice were generated by Biocytogen using a CRISPR/Cas9 approach. Briefly, two sgRNAs were designed with a CRISPR design tool (http://www.sanger.ac.uk/htgt/wge/) to target either a region upstream or downstream of Sirt1 exon5 and then were subsequently screened for on-target activity using a Universal CRISPR Activity Assay [UCATM, Biocytogen Pharmaceuticals (Beijing) Co., Ltd]. To minimize random integrations, a circular donor vector was used. The gene targeting vector containing a 5′ homologous arm, the target fragment (Sirt1 minigene containing exons 5–9 + 3′-untranslated region) and a 3′ homologous arm was used as a template to repair the DSBs generated by Cas9/sgRNA. The two loxp sites were precisely inserted in both sides of the Sirt1 gene target fragment. T7 promoter sequence was added to the Cas9 or sgRNA template by PCR amplification in vitro. Cas9 mRNA, targeting vector, and sgRNAs were coinjected into the cytoplasm of one-cell stage fertilized C57BL/6N eggs. The injected zygotes were transferred into oviducts of Kunming pseudopregnant females to generate F0 mice. F0 mice with the expected genotype (confirmed by tail genomic DNA PCR and sequencing) were mated with C57BL/6N mice to establish germline-transmitted F1 heterozygous mice. F1 heterozygous mice were genotyped by tail genomic PCR, southern blot, and DNA sequencing. In basal conditions, these mice express wild-type Sirt1 as exon 4 is spliced into the exon 5–9 minigene. However, upon Cre recombinase expression (which removes the floxed minigene), exon 4 is spliced into the H355A-mutant exon 5, resulting in the expression of H355A-mutant Sirt1 mRNA driven by the Sirt1 promoter itself. Gentoyping of floxed versus wild-type versus recombined alleles was done by PCR using custom primers (Fwd: 5′-TGTACCTTGCACAACTAGTTTCCGT-3′; Rv: 5′-TGTGAGGGTGTCAGATCCACATGC-3′).

Animals were maintained in ventilated caging in specific pathogen-free facilities at New Brunswick RBHS Rutgers Campus. All animal housing, handling, and procedures involving mice were approved by the Rutgers Institutional Animal Care and Use Committee, in accordance with all relevant ethical regulations.

To generate NOTCH1-induced Sirt1-overexpresing leukemias, we performed a retroviral transduction of lineage-negative enriched cells from Sirt1^WT or Sirt1^TG (ref. 42; JAX, #024510) donors with retrovirus encoding ΔE-NOTCH1-GFP, an oncogenic activated form of NOTCH1 with concomitant expression of GFP, as previously described (10). Cells were then transplanted via retro-orbital injection into lethally irradiated (7.5 Gy) recipient mice. Investigators were not blinded to group allocation. Animals were monitored for signs of distress or motor function at least twice daily, until they were terminally ill, whereupon they were euthanized.

In order to generate Sirt1-conditional inducible knockout (Sirt1^flox/flox) or Sirt1^H355A-conditional inducible knockin leukemias (Sirt1^{cKI-H355A/cKI-H355A}), we first crossed Sirt1^flox/flox mice (ref. 30; JAX, #029603) or our newly generated Sirt1^cKI-H355A/+ mice with mice harboring a tamoxifen-inducible Cre recombinase from the ubiquitous Rosa26 locus (10). Then, we performed a retroviral transduction of lineage-negative enriched cells from Sirt1^flox/flox-Rosa26^Cre-ERT2/+ or Sirt1^{cKI-H355A/cKI-H355A}-Rosa26^Cre-ERT2/+ donors with retrovirus encoding ΔE-NOTCH1-GFP or HDΔP-NOTCH1-GFP, oncogenic activated forms of NOTCH1 with concomitant expression of GFP, as previously described (10). Cells were then transplanted via retro-orbital injection into lethally irradiated (7.5 Gy) recipient mice.

For leukemia-initiation survival studies, mice transplanted with ΔE-NOTCH1-GFP-infected Sirt1^flox/flox-Rosa26^Cre-ERT2/+ progenitors were treated 48 hours after transplantation with vehicle only (corn oil; Sigma, C8267) or tamoxifen (Sigma, T5648; 3 mg per mouse in corn oil), to induce isogenic loss of Sirt1 before leukemic transformation.

For leukemia-progression studies, already generated ΔE-NOTCH1-GFP–induced or HDΔP-NOTCH1-GFP-induced Sirt1^flox/flox-Rosa26^Cre-ERT2/+ leukemias (1 × 10⁶ leukemia cells) or ΔE-NOTCH1-GFP–induced Sirt1^{cKI-H355A/cKI-H355A}-Rosa26^Cre-ERT2/+ leukemias (1 × 10⁶ leukemia cells) were transplanted from primary recipients into sublethally irradiated (4.5 Gy) 6–8-week-old secondary recipient C57BL/6 mice (Taconic Farms) by retro-orbital injection. Forty-eight hours after leukemic cell transplantation, recipient mice were treated with vehicle only (corn oil; Sigma, C8267) or tamoxifen (Sigma, T5648; 3 mg per mouse in corn oil), to induce isogenic loss of Sirt1 or replacement of wild-type Sirt1 by Sirt1^H355A in established leukemias. In the case of mice transplanted with Sirt1^flox/flox-Rosa26^Cre-ERT2/+ leukemias, 5 days posttransplantation, mice in each arm (vehicle or tamoxifen) were divided randomly into two different groups: control groups were subsequently treated with vehicle only (2.3% DMSO in 0.005% methylcellulose, 0.1% Tween-80) or with DBZ (5 mg per kg in vehicle solution) on a 4-day-ON and 3-day-OFF schedule, as previously described (10). Investigators were not blinded to group allocation. Animals were monitored for signs of distress or motor function at least twice daily until they were terminally ill, whereupon they were euthanized.

To generate Sirt1/Ampk double-conditional inducible knockout (Sirt1^flox/flox; Ampk^flox/flox-Rosa26^Cre-ERT2/+) leukemias, we first crossed Sirt1^flox/flox-Rosa26^Cre-ERT2/+ mice with Ampk^flox/flox mice (ref. 68; JAX, #014141), and performed leukemia-progression studies upon transplantation of already generated ΔE-NOTCH1-GFP–induced Sirt1^flox/flox;Ampk^flox/flox-Rosa26^Cre-ERT2/+ leukemias (1 × 10⁶ leukemia cells) similar to before.

For Sirt1 acute deletion analyses in mouse primary leukemias (metabolomics, acetyl proteomics, RNA-seq, ChIP-seq, western blot, and/or apoptosis studies), we transplanted lymphoblasts from spleens of Sirt1-conditional knockout NOTCH1-induced T-ALL-bearing mice into a secondary cohort of recipient mice, as before. We monitored mice until they presented clear leukemic signs with >60% GFP-positive leukemic cells in peripheral blood; then, mice were treated with vehicle or tamoxifen, and, 48 hours after treatment, mice were euthanized and spleen samples were collected for further analyses.

For analysis of the effects of Sirt1 overexpression on the response to DBZ in vivo, we infected NOTCH1-induced GFP-positive mouse primary leukemic cells with a retrovirus driving bicistronic expression of the cherry fluorescent protein together with either Sirt1 (pMSCV-Sirt1-IRES-mCherry FP) or Sirt1 H355A (pMSCV-Sirt1-H355A-IRES-mCherry FP). Sirt1 cDNA was amplified by PCR, using primers containing unique restriction enzyme sites, cloned into the multiple cloning site of the pMSCV-mCherry FP vector (Addgene, 52114), and sequence verified. Sirt1 H355A mutation was generated by performing directed mutagenesis on this vector using the QuikChange II XL Site-Directed Mutagenesis Kit (Agilent, 200521), following the manufacturer's instructions. A mix of transduced and nontransduced cells (i.e., without sorting) was then injected into sublethally irradiated C57BL/6 mice (4.5 Gy). Mice were treated with DBZ (5 mg/kg) on a daily basis and leukemic blasts in peripheral blood were analyzed as previously described (10).

Human peripheral blood mononuclear cells (PBMC) and peripheral blood CD4-positive cells were purchased from Lonza (product numbers CC-2702 and 2W-200, respectively). Samples from normal human thymus were kindly provided by Dr. Adolfo Ferrando (Columbia University). Whole-cell extracts were prepared using standard procedures. After protein transfer, membranes were incubated with the antibodies anti-SIRT1(1:1,000, sc-74504, Santa Cruz), anti-SIRT1 (1:5,000, ab12193, Abcam), anti-AMPK (1:1,000, 5831S, Cell Signaling Technology), anti-p-AMPK (1:1,000, 2535S, Cell Signaling Technology), anti-4E-BP1 (1:1,000, 9644S, Cell Signaling Technology), anti-p-4EB-P1 (1:1,000, 2855S, Cell Signaling Technology), anti-p53 (1:1,000, 2524S, Cell Signaling Technology), anti-ac-p53 (1:1,000, 2570S, Cell Signaling Technology), anti-ac-NFkB (1:1,000, ab191870, Abcam), anti-c-MYC (1:200, sc-40, Santa Cruz), anti-KAT7 (1:1,000, 13751-1-AP, ProteinTech), anti-H4 (1:1,000, ab10158, Abcam), anti-H4K12ac (1:1,000, ab177793, Abcam), anti-GAPDH (1:10,000, 97166, Cell Signaling Technology), and anti–β-actin-HRP (1:50,000; A3854, Sigma). Antibody binding was detected with a secondary antibody coupled to horseradish peroxidase (Sigma, NA934 and NA931) using enhanced chemiluminescence (Thermo Scientific, 34578) or using secondary fluorescence-coupled antibodies (Thermo Scientific) in an iBright FL1500 instrument (Thermo Scientific). Immunoblot quantification was performed using the Fiji image processing package (69).

10M cells were pelleted down and washed with ice-cold 1× PBS. Cells were resuspended in 500 to 1,000 μL of immunoprecipitation (IP) lysis buffer (Thermo Scientific, 87788) and incubated in a rotor for 20 minutes at 4°C. Protein lysate was extracted by centrifugation at 13,000 rpm for 10 minutes at 4°C. Protein concentration was determined using the BCA kit (Thermo Scientific, 23223). Immunoprecipitations were carried out using Dynabeads Protein G Immunoprecipitation Kit (Thermo Scientific, 10007D). Briefly, beads were resuspended well and 40 μL were aliquoted into tubes for each IP reaction. Supernatants were removed from the beads by placing tubes on a magnet and 10 μL of anti-Kat7 antibody (13751-1-AP, ProteinTech) were added in PBST (1:20 dilution). The antibody was allowed to bind to the beads for 1 hour in a rotor at 4°C. Supernatants were removed by placing the tubes on a magnet. Lysates (1 mg) added and incubated overnight at 4°C in a rotor. Supernatants were removed and washed three times with the wash buffer provided with the immunoprecipitation kit. Wash buffer (100 μL) was added to resuspend the protein Ab–coated beads and transferred to a clean tube. Again, supernatants were removed, and beads were finally resuspended in 20 μL of the elution buffer supplied with the kit. Finally, 10 μL of the 4× LDS sample buffer (Thermo Scientific, NP0007) and 10× of reducing buffer (Thermo Scientific, B0009) were added to the bead samples, followed by boiling at 70°C. Supernatants were loaded into gels followed by transfer to PVDF membrane using the same protocol for western blotting described before. Membranes were developed with the antibodies anti-acetyl-Lysine (1:1,000, MA1-2021, Thermo Scientific). Membranes were then stripped with stripping buffer (Thermo Scientific, 21059), and probed for KAT7 (1:1,000, 13751-1-AP, ProteinTech).

To analyze spleen samples, single-cell suspensions were prepared by disrupting spleens through a 70-μm filter. Red cells were removed by incubation with ammonium–chloride–potassium lysing buffer (155 mmol/L NH4Cl, 12 mmol/L KHCO₃, and 0.1 mmol/L EDTA) for 5 minutes in ice. Apoptotic cells in leukemic spleens or human/mouse cell lines after different treatments were quantified with PE–Annexin V Apoptosis Detection Kit I (BD Pharmingen, 559763). All flow cytometry data were collected on an Attune NxT Flow Cytometer (Thermo Fisher Scientific) and analyzed with FlowJo v10.6.2 software (BD).

For leukemic spleens, leukemic mice were euthanized, and spleens were quickly removed, minced, and frozen. Subsequently, 25 mg per sample of frozen leukemic spleens were homogenized under liquid nitrogen flow using a Retsch CryoMill at 20 Hz for 2 minutes. Pulverized samples were then mixed with methanol:acetonitrile:water (40:40:20) with 0.1 M formic acid solution followed by 10 minutes of incubation on ice and 500 μL of the extract was neutralized with 44 μL of 15% (m/v) ammonium bicarbonate. Finally, after centrifugation (14,000 × g, 10 minutes at 4°C), samples were transferred to clean tubes and sent for LC-MS analysis. For media, we diluted 20 μL of each respective media into 980 μL of ice-cold solvent (40:40:20 methanol:acetonitrile:water + 0.1 M formic acid) followed by neutralization with 80 μL of 15% ammonium bicarbonate. Samples were directly submitted to LC-MS analysis.

LC-MS analysis of the extracted metabolites was performed on a Q Exactive PLUS hybrid quadrupole-orbitrap mass spectrometer (Thermo Fisher Scientific) coupled to hydrophilic interaction chromatography (HILIC). The LC separation was performed on Vanquish Horizon UHPLC system with an XBridge BEH Amide column (150 mm × 2.1 mm, 2.5 μm particle size, Waters) with the corresponding XP VanGuard Cartridge. The liquid chromatography used a gradient of solvent A (95%:5% H₂O:acetonitrile with 20 mmol/L ammonium acetate, 20 mmol/L ammonium hydroxide, pH 9.4), and solvent B (20%:80% H₂O:acetonitrile with 20 mmol/L ammonium acetate, 20 mmol/L ammonium hydroxide, pH 9.4). The gradient was 0 minutes, 100% B; 3 minutes, 100% B; 3.2 minutes, 90% B; 6.2 minutes, 90% B; 6.5 minutes, 80% B; 10.5 minutes, 80% B; 10.7 minutes, 70% B; 13.5 minutes, 70% B; 13.7 minutes, 45% B; 16 minutes, 45% B; 16.5 minutes, 100% B. The flow rate was 300 μL/minute. Injection volume was 5 μL and column temperature 25°C. The MS scans were in negative ion mode with a resolution of 70,000 at m/z 200. The automatic gain control (AGC) target was 3 × 10⁶, and the scan range was 75 to 1,000. Metabolite features were extracted in MAVEN (70) with the labeled isotope specified and a mass accuracy window of 5 ppm.

Leukemic mice (ΔE-NOTCH1-GFP-induced or HDΔP-NOTCH1-GFP-induced Sirt1^flox/flox-Rosa26^Cre-ERT2/+ leukemia–bearing mice) were euthanized, and spleens were quickly removed, minced, and frozen. Subsequently, 25 mg per sample of frozen leukemic spleens were homogenized under liquid nitrogen flow using Retsch CryoMill at 20 Hz for 2 minutes. Pulverized samples were then mixed with 2× laemmli buffer with 50 mmol/L DTT in a ratio of 1 mg of tissue: 10 mL of 2× laemmli buffer. Samples were sonicated for 7 minutes, incubated at 95°C for 10 minutes and centrifuged (25,000 × g, 30 minutes at 4°C), the supernatant was saved in a new tube. An equal volume of 8 M urea was used to extract the remaining precipitate with sonication and centrifugation (25,000 × g, 30 minutes at 4 °C), the supernatant was combined with the previous one, and protein concentration was determined by a 660 kit (Thermo Fisher Scientific). Lysate (500 μg) was run into SDS-PAGE gel and digested using standard gel-plug procedures. 450 μg of each sample was labeled with TMT10 plex reagent (Thermo Fisher Scientific) according to the manufacturer's instructions. A test mix of 1% of individual samples was run on nano-LC-MSMS and the reporter ion intensity was used to normalize the mixing of the samples to achieve a ratio of 1:1 for all samples. After mixing, the samples were desalted with SpeC18 (Waters) and dried. TMT10-labeled peptides with acetylated K were enriched with Acetyl-Lysine Motif kit (Cell Signaling Technology) following the manufacturer's instructions. Eluted peptides were desalted with Stagetips (71).

Nano-LC-MS/MS was performed using a Dionex rapid-separation liquid chromatography system interfaced with a QExactive HF (Thermo Fisher Scientific). Samples were loaded onto an Acclaim PepMap 100 trap column (75 μm × 2 cm, Thermo Fisher Scientific) and washed with buffer A (0.1% trifluoroacetic acid) for 5 minutes with a flow rate of 5 μL/minute. The trap was brought in line with the nano analytical column (nanoEase, MZ peptide BEH C18, 130A, 1.7 μm, 75 μm × 20 cm, Waters) with a flow rate of 300 nL/minute with a multistep gradient (4%–15% buffer B [0.16% formic acid and 80% acetonitrile] for 20 minutes, then 15%–25% B for 40 minutes, followed by 25%–50% B for 30 minutes). Mass spectrometry data were acquired using a data-dependent acquisition procedure with a cyclic series of a full scan acquired with a resolution of 120,000 followed by tandem mass spectrometry scans [33% normalized collision energy (NCE) 33% in the higher-energy collisional dissociation (HCD) cell with a resolution of 45,000 of the 20 most intense ions with a dynamic exclusion duration of 20 seconds].

Some samples were analyzed using a Dionex rapid-separation liquid chromatography system interfaced with a Orbitrap Eclipse tribrid mass spectrometer (Thermo Fisher Scientific). The LC conditions were the same as described for Dionex RSLC connectee to QExactive HF (Thermo Fisher Scientific). The scan sequence began with an MS1 spectrum (Orbitrap analysis, resolution 120,000 with a scan range from 375 to 1,600 Th, AGC target 6E4, maximum injection time 50 ms). The top S (3 seconds) duty cycle scheme was used to determine the number of parent ions investigated for each cycle. For proteomic samples without IAP, the SPS method was used. In detail, peptide ions were first collected for MS/MS by collision-induced dissociation and scanned in ion trap with AGC 2E4, NCE 35%, maximum injection time 50 ms, and isolation window at 0.7. Following the acquisition of each MS2 spectrum, 10 MS2 fragment ions were captured as the MS3 precursor population using isolation waveforms with multiple frequency notches. MS3 precursors were fragmented by HCD and analyzed using the Orbitrap (NCE 65, AGC 1.5E5, maximum injection time 105 ms, resolution 50,000 at 400 Th and scan range from 100–500 amu). For acetyl-peptide, the MS/MS method was used. Parent mass was isolated in the quadrupole with an isolation window of 0.7 m/z, AGC target 1.75E5, and fragmented with HCD NCE38. The fragments were scanned in Orbitrap with a resolution of 50,000. The MS/MS scan ranges were determined by the charge state of the parent ion but the lower limit was set at 110 amu.

LC-MS data were analyzed with Maxquant (version 1.6.17.0) with Andromeda search engine. For samples run on QExactive, type of LC-MS run was set to reporter ion MS2 with 10plex TMT as isobaric labels. For data acquired on Eclipse, type of LC-MS run was set to reporter ion MS3 with TMT10 plex as isobaric labels for proteome and MS2 with TMT10 plex for acetylome. Reporter ion mass tolerance was set at 0.003 Da. LC-MS data were searched against the Uniprot mouse database with the addition of potential contaminants. Protease was set as trypsin/P that allowed 2 miss cuts for total proteomic data and 5 miss cuts for acetyl K sample (PTM sample). Carbaidomethylation of cysteine was set as a fixed modification, and N-terminal acetylation, oxidation at methionine as well as acetylation at lysine were set as variable modifications. For quantification, spectra were filtered by min reporter PIF set at 0.5 (spectra purity) for acetyl peptides and 0.75 for proteome data.

The Maxquant search results were analyzed using Perseus (version 1.6.15.0). Data were first filtered for reverse and contaminant hits, and the reporter ion intensity data were further log₂ transformed. The proteome data were then normalized to the median of each channel. For acetyl peptides, the data were also normalized based on the proteome summed intensity of each channel, and each acetyl site was further normalized to the corresponding protein group quantitation value if the group had been identified in proteome data. For group comparison, statistical significance between groups was analyzed using the Student t test with equal variance on both sides, requiring two valid values in at least one group, and the Q value was calculated with permutation. FDR was analyzed using the Student t test with equal variance on both sides and the Q value was calculated using a permutation test. Significance was assigned with consideration of S0 (S 0 = 0.1).

Total RNA was extracted from cells using RNeasy Plus Mini Kit (Qiagen), and cDNA was generated with the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Quantitative PCR was performed on a QuantStudio 3 Real-Time PCR System (Applied Bio­systems) using SYBR Green PCR Master Mix (Takara), and transcript levels were normalized to actin as an internal control. Primers used: SIRT1 (Forward: 5′-AGGATAGAGCCTCACATGCAA-3′; Reverse: 5′-TCGAGGATCTGTGCCAATCATA-3′), Sirt1 (Forward: 5′-GAGCTGGGGTTTCTGTCTCC-3′; Reverse: 5′-CCGCAAGGCGAGCATAGATA-3′), ACTIN (5′-CAACGCCAACCGCGAGAAGAT-3′; Reverse: 5′-CCAGAGGCGTACAGGGATAGCAC-3′) and Actin (5′-GGCTGTATTCCCCTCCATCG-3′; 5′-CCAGTTGGTAACAATGCCATGT-3′).

Sirt1-conditional knockout ΔE-NOTCH1–induced T-ALL–bearing mice were treated with vehicle only (corn oil; Sigma, C8267) or tamoxifen (Sigma, T5648; 3 mg per mouse in corn oil) to induce isogenic loss of Sirt1 via intraperitoneal injection. Forty-eight hours later, single-cell suspensions of total leukemic splenocytes were prepared by pressing leukemic spleens through a 70-μm filter. We removed red cells in spleen samples by incubation with red blood cell lysis buffer (155 mmol/L NH₄Cl, 12 mmol/L KHCO₃, and 0.1 mmol/L EDTA) for 5 minutes on ice. RNA was extracted using QIAshredder (QIAGEN, 79656) and RNeasy Mini (QIAGEN, 74106) kits. RNA library preparations and next-generation sequencing were performed using the Illumina Next-Seq platform (Illumina). We estimated gene-level raw counts using kallisto 0.44.0 (72), with the Ensembl GRCm38 transcriptome as the reference. We evaluated differential expression between the samples in R using DESeq2 (73). GSEAs were performed to assess the enrichment of the KAT7-loss signature (44) based on 10,000 permutations of the gene list (74).

Analyses of genome-wide H4K12ac mark in leukemic cells isolated from Sirt1-conditional knockout ΔE-NOTCH1–induced T-ALL–bearing mice acutely treated as before with vehicle or tamoxifen to induce isogenic loss of Sirt1 was done by Active Motif, following well-established protocols and using a ChIP-validated H4K12ac antibody (Active Motif, #39165). Peaks were called using either the MACS (75) or SICER algorithms (76). MACS default cutoff is P value 1e-7 for narrow peaks and 1e-1 for broad peaks, and SICER default cutoff is FDR 1e-10 with gap parameter of 600 bp. Peak filtering was performed by removing false ChIP-seq peaks as defined within the ENCODE blacklist (77).

SIRT1 expression was analyzed among T-ALL samples (n = 57) and physiologic thymocyte subsets (n = 21) from published literature (37). Quantile normalization was performed across samples. Differential expression was performed using Mann–Whitney U test (P < 0.001), and genes with FDR < 0.05 using the Benjamini–Hochberg correction were shortlisted as differentially expressed.

Statistical analyses were performed with Prism 8.0 (GraphPad). Unless otherwise indicated in figure legends, statistical significance between groups was calculated using an unpaired two-tailed Student t test. Survival in mouse experiments was represented with Kaplan–Meier curves, and significance was estimated with the log-rank test.

RNA-seq and ChIP-seq data from Sirt1-conditional knockout leukemias were deposited in the NCBI Gene-Expression Omnibus repository under the following accession numbers: GSE203387 and GSE203386. We analyzed NOTCH1 transcriptional targets using GSI washout experiments publicly available from GEO: GSE29544. We analyzed SIRT1 promoter occupancy of chromatin marks and epigenetic and transcription factors using the following T-ALL publicly available ChIP-seq and ATAC-seq data sets from Gene-Expression Omnibus (GEO): GSE58406, GSE124223, GSE29611, GSE54379, GSE29600, and GSE138516.

X. Su reports grants from NIH/NCI during the conduct of the study. D. Herranz reports grants from NIH/NCI, American Cancer Society, Leukemia and Lymphoma Society, Ludwig Cancer Research, Alex's Lemonade Stand Foundation, Children's Leukemia Research Association, Gabrielle's Angel Foundation for Cancer Research, and the American Association for Cancer Research during the conduct of the study. No disclosures were reported by the other authors.

O. Lancho: Data curation, formal analysis, investigation, writing–original draft. A. Singh: Data curation, formal analysis, investigation. V. da Silva-Diz: Formal analysis, investigation, writing–review and editing. M. Aleksandrova: Formal analysis, investigation, writing–review and editing. J. Khatun: Formal analysis, investigation. L. Tottone: Formal analysis, investigation, writing–review and editing. P. Renck Nunes: Formal analysis, investigation. S. Luo: Investigation. C. Zhao: Data curation, investigation. H. Zheng: Data curation, investigation. E. Chiles: Data curation, investigation. Z. Zuo: Investigation. P.P. Rocha: Formal analysis, supervision. X. Su: Formal analysis, supervision. H. Khiabanian: Data curation, formal analysis, supervision, writing–review and editing. D. Herranz: Conceptualization, data curation, formal analysis, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

Work in the laboratory of D. Herranz is supported by the NIH (R01CA236936), the American Cancer Society (RSG-19-161-01-TBE), The Leukemia and Lymphoma Society (Scholar Award 1386-23), the Ludwig Cancer Research, the Alex's Lemonade Stand Foundation, the Children's Leukemia Research Association, the Gabrielle's Angel Foundation for Cancer Research, the 2021 AACR–Bayer Innovation and Discovery Grant (Grant Number 21-80-44-HERR), and the Rutgers Cancer Institute of New Jersey. Work in the laboratory of H. Khiabanian is supported by the NIH (R01CA233662) and the V Foundation (T2019-012), as well as by Rutgers Office of Advanced Research Computing (NIH 1S10OD012346-01A1). In addition, Rutgers Cancer Institute of New Jersey shared resources supported in part by the National Cancer Institute Cancer Center Support Grant P30CA072720 were instrumental for this project, including Biomedical Informatics Shared Resource (P30CA072720-5917), Flow Cytometry and Cell Sorting Shared Resource (P30CA072720-5921), and the Pilot Award/New Investigator Award (P30CA072720-5931). Moreover, the purchase of the Eclipse and QE instruments was supported by NIH grants S10OD025140 and S10OD01640. Work in the lab of P.P. Rocha is supported by the Intramural Research Program of the Eunice Kennedy Shriver National Institute of Child Health and Development (NICHD). We also thank the NICHD molecular genomics core and computational resources of the NIH HPC Biowulf cluster (hpc.nih.gov). Fellowships from the New Jersey Commission on Cancer Research supported the work of V. da Silva-Diz (DCHS19PPC008), L. Tottone (DCHS20PPC010) and P. Renck Nunes (COCR22PDF002). V. da Silva-Diz was also supported by the Pediatric Cancer and Blood Disorders Research Center at the Rutgers Cancer Institute of New Jersey. We are grateful to Adolfo A. Ferrando (Columbia University Medical Center) for sharing with us normal human thymocyte samples, and we thank both Adolfo A. Ferrando and Antonio Maraver (IRCM, Montpellier) for their constant constructive criticism and support. We also thank everyone involved with JuanLord for their support.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.