Genetic or pharmacological inactivation of CREBBP sensitizes B-cell Acute Lymphoblastic Leukemia to Ferroptotic Cell Death upon BCL2 Inhibition

B-cell acute lymphoblastic leukemia (B-ALL) is a leading cause of death in childhood and outcomes in adults remain dismal. There is therefore an urgent clinical need for therapies that target the highest risk cases. Mutations in the histone acetyltransferase CREBBP associate with high-risk features in B-ALL and have been implicated in chemoresistance. We performed a targeted drug screen in isogenic human cell lines, identifying a number of actionable small molecules that specifically target CREBBP-mutated B-ALL. The most potent was the BCL2 inhibitor Venetoclax, which acts through a non-canonical mechanism resulting in ferroptotic cell death. CREBBP-mutated cell lines showed differences in cell-cycle, metabolism and response to oxidative stress. Lastly, we demonstrate that small-molecule inhibition of CREBBP sensitizes B-ALL cells, regardless of genotype, to Venetoclax-induced ferroptosis in-vitro and in-vivo, providing a potential novel drug combination for broader clinical translation in B-ALL.


Introduction
B-cell acute lymphoblastic leukemia (B-ALL) is an aggressive hematological malignancy of B-lineage progenitors and is the commonest cancer in children 1 .Whilst the majority of children can be cured with multi-agent chemotherapy, patients with high-risk genetic subtypes, certain age groups and those who relapse remain a clinical challenge, such that B-ALL remains a leading cause of death in childhood.Furthermore, outcomes of adults with B-ALL remain dismal, even when fit enough to be treated intensively.There is therefore a need to better understand drivers of high-risk B-ALL and to develop novel therapeutic approaches targeting these challenging patient cohorts.
CREBBP mutations are found in multiple hematological and solid malignancies, notably B-cell lymphomas 2,3 .Loss-of-function (LOF) mutations affecting CREBBP are recurrent second-hit mutations across multiple genetic subtypes of B-ALL and are associated with adverse features, including high-risk genetic subtypes and persistent measurable residual disease [4][5][6][7] .In addition, they have been mechanistically associated with chemoresistance and are enriched at relapse 4,[8][9][10][11][12] .CREBBP mutations have also been described as an adverse prognostic factor in ALL, acute myeloid leukaemia (AML) and follicular lymphoma [13][14][15] .CREBBP is a large protein with histone acetyltransferase (HAT) enzymatic activity alongside protein scaffolding function mediated through multiple protein-protein interaction domains, including a bromodomain responsible for binding acetylated lysine residues.Alongside its paralogue EP300, CREBBP is considered to primarily function as a transcriptional co-activator, responsible for acetylating histone residues at gene enhancers and promoters.CREBBP LOF mutations can include complete loss of the protein or recurrent point mutations affecting the HAT domain, which appear to exert a stronger phenotype 4 .During B-ALL evolution, CREBBP mutations frequently become bi-allelic and commonly co-associate with activating RAS pathway mutations, suggesting strong oncogenic co-operativity 4,12,16 .
Targeting cells harbouring LOF mutations in tumour suppressor genes (TSG) principally relies on perturbing "synthetic lethal" dependencies acquired upon loss of TSG activity, commonly via inhibition of redundant pathways or protein paralogues.In the context of CREBBP, this has been demonstrated in models of B-cell lymphoma through inhibition of residual EP300 function using small molecule HAT or bromodomain inhibitors 17 .Global analyses of genetic co-dependencies have also implicated a dependency of CREBBP-mutated tumours on EXOC5 function 18 , whilst a number of mechanistic studies have identified potentially targetable roles for CREBBP in modulating key cellular processes including DNA damage response, signaling, apoptosis and metabolism 3,12,16,19 .
In this study we identify novel treatment options for CREBBP-mutated high-risk B-ALL.We generated isogenic human B-ALL cell lines and undertook a synthetic lethal drug screen focusing on clinicallyactionable agents targeting pathways mechanistically associated with CREBBP function.CREBBP LOF resulted in cell cycle and metabolic dysregulation associated with marked sensitivity to ferroptotic cell death upon small molecule inhibition of the anti-apoptotic regulator BCL2.Inhibition of CREBBP function with small molecule inhibitors could phenocopy this synthetic lethal effect, sensitizing diverse subtypes of B-ALL to BCL2 inhibitors in-vitro and producing a significant survival advantage in-vivo, thus providing a potential novel efficacious drug combination across a wider number of B-ALL genotypes.

CREBBP-mutated B-ALL shows increased sensitivity to Venetoclax
To identify candidate therapeutics that specifically target CREBBP-mutated high-risk B-ALL, we undertook a synthetic-lethal drug screen.The CREBBP wild-type (WT) B-ALL cell line 697 (driven by the recurrent E2A::PBX1 fusion) was genome-engineered by CRISPR-Cas9 homologous recombination to introduce a recurrent "hotspot" mutation at arginine 1446 (CREBBP R1446C ), which is implicated as exerting a dominant-negative effect on CREBBP acetyltransferase activity 4 .Several clones were generated, including a homozygous CREBBP R1446C knock-in mutant clone (hereafter, 697 KI ) and a mutant clone containing two frameshift mutations resulting in a complete knockout of CREBBP protein (hereafter, 697 KO ) (Fig. 1a and Extended Data Fig. 1a).For use in validation studies, the ETV6::RUNX1driven cell line REH (containing three WT copies of CREBBP) was also edited, resulting in two compound-heterozygous mutated clones, each including a single allele of the CREBBP R1446C HAT mutation, alongside presumed deleterious mutations of the other two alleles (Extended Data Fig. 1a).
We subjected the 697 isogenic cell lines to a targeted drug screen, using a wide range of concentrations, focussed on clinically-actionable drugs in classes implicated or hypothesized to show differential sensitivity in published models of B-cell lymphoma and other CREBBP-mutated malignancies (Extended Data Table 1) 3,4,[10][11][12]17,19,20 . CREBBP-mtated 697 cells were not differentially sensitive to traditional cytotoxic chemotherapy, and paradoxically showed a degree of sensitization to the glucocorticoid Dexamethasone, used in current ALL induction regimens (Fig. 1b and Extended Data Table 1 and Fig. 1b) 4,11,12 .As anticipated, and validating our screen design, inhibitors of the CREBBP paralogue EP300 (the CREBBP/EP300-specific bromodomain inhibitor Inobrodib and the CREBBP/EP300 acetylase inhibitor A485) exhibited synthetic lethality, consistent with previous reports in B-cell lymphoma (Fig. 1b,c and Extended Data Table 1 and Fig. 1c) 17 .
Unexpectedly, the most potent hit identified from the screen was the clinical-grade BCL2 inhibitor Venetoclax, which showed a 2-log10-fold reduction in IC50, in both 697 KI and 697 KO clones (Fig. 1b,d and Extended Data Table 1 and Fig. 1d).These findings were validated in isogenic REH lines, with a 1-log10fold reduction in IC50 in both mutant clones (Extended Data Fig. 1e).We confirmed this sensitization to Venetoclax in in-vitro proliferation assays by direct cell counting (Fig. 1e).Upon low-dose Venetoclax exposure, CREBBP-mutated 697 cells showed enhanced evidence of markers of programmed cell death, including mitochondrial depolarization, externalization of Annexin-V and induction of cleaved PARP and caspase-3 (Fig. 1 f-h and Extended Data Fig. 1f), consistent with the known mechanism-of-action of Venetoclax in inducing apoptotic programmed cell death.
Overall, this focussed drug screen demonstrates that CREBBP-mutated B-ALL is: i) not uniformly chemo-resistant, and ii) identifies a number of clinically-actionable agents for use in CREBBP-mutated high-risk B-ALL, including Dexamethasone, EP300 inhibitors and a potent sensitization to the BCL2 inhibitor Venetoclax.

Venetoclax exerts its effect on CREBBP-mutated B-ALL by on-target inhibition of BCL2
We sought to explore the mechanism-of-action of Venetoclax.Venetoclax was developed to induce apoptosis through inhibition of BCL2 binding to the pro-cell death proteins BAK and BAX.However, recently, it has been shown to have alternative mechanisms-of-action, in particular on metabolism and self-renewal, including potential BCL2-independent effects [21][22][23] .To test whether the mechanismof-action of Venetoclax in CREBBP-mutated B-ALL was through on-target BCL2 inhibition, we employed a doxycycline-inducible shRNA knock-down system, where shRNA expression was directly linked to a fluorescent reporter (Extended Data Fig. 2a-d) 24 .
Co-culture of 697 WT or 697 KI cells expressing one of two unique BCL2-targeting shRNAs (reported by mCherry) competed against cells expressing an shRNA targeting the negative-control renilla gene (reported by green fluorescent protein (GFP)) showed that 697 KI cells exhibited a marked competitive disadvantage upon BCL2-knockdown, when compared to 697 WT (Fig. 2a,b).This was associated with significant externalization of Annexin-V, consistent with the induction of programmed cell death (Fig. 2c,d).
Collectively these studies demonstrate that Venetoclax induces programmed cell death in CREBBPmutated B-ALL by on-target BCL2 inhibition and that the sensitivity of 697 KI cells is specific to BCL2 and does not occur through other anti-apoptotic proteins.

CREBBP-mutated B-ALL shows significant cell cycle and metabolic dysregulation
To further explore the mechanism of action of Venetoclax in CREBBP-mutated 697 cells, we undertook bulk RNA sequencing (RNAseq) of 697 WT and 697 KI cells, after 24 hours exposure to either dimethyl sulfoxide (DMSO) vehicle, or low-dose Venetoclax (20nM -the IC50 of 697 KI cells).Consistent with the role of CREBBP as a transcriptional co-activator, the majority of differentially expressed genes (DEGs) between DMSO-vehicle treated 697 WT and 697 KI cells were down-regulated (Extended Data Fig. 3a).Gene Set Enrichment Analysis (GSEA) showed marked down-regulation of published Crebbp target genes from a mouse lymphoma model, supporting the role of the CREBBP R1446C mutation as inhibitory to CREBBP transcriptional co-activator function (Extended Data Fig. 3b) 25 .
KEGG pathway and GSEA analysis showed significant down-regulation of signatures associated with apoptosis, in consonance with our functional experiments above (Fig. 1f-h and Fig. 3a,b).However, gene-specific examination of differential expression of apoptotic regulators by KEGG pathway analysis showed a mixed picture, affecting the expression of both pro-and anti-apoptotic genes, including a small but significant up-regulation of BCL2 itself (Fig. 3c and Extended Data Fig. 3c).
More broadly, KEGG and GSEA pathway analyses showed a differential down-regulation of cell cycle and signaling pathways in 697 KI (Fig. 3a,b and Extended Data Fig. 3d).Downregulation of cell cycleassociated transcriptional signatures was associated with a marked up-regulation of the tumour suppressor CDKN2A (encoding the negative cell-cycle regulator P16 INK4a-ARF ), which is commonly mutated in B-ALL (Fig. 3d).We confirmed a relative reduction in proliferative capacity in both 697 KI and 697 KO cells compared to 697 WT by proliferation assays (Fig. 3e).This was associated with a significantly increased proportion of 697 KI cells in G1 phase alongside reduced Early S/G2-S-M phases (Fig. 3f and Extended Data Fig. 3e) confirming a significant defect in cell cycle progression.The majority of transcriptionally up-regulated KEGG and Gene Ontology (GO) pathways were indicative of metabolic dysfunction, and GSEA showed dysregulation of fatty acid oxidation and hypoxic gene signatures (Fig. 3a,b and Extended Data Fig. 3f).Given the close association of cell cycle and metabolism, and the established role of BCL2 and Venetoclax in disturbing mitochondrial respiration 23 , we analysed baseline metabolic differences between 697 WT and 697 KI cells using in-vitro metabolic flux assays.Unexpectedly, and despite lower cell cycle progression, 697 KI cells consistently showed increased rates of both glycolysis and oxidative phosphorylation (OxPhos), including higher rates of both basal and maximal respiration, and an increase in spare respiratory capacity (SRC) (Fig. 3g,h).
Collectively our model suggests that loss of CREBBP acetyltransferase function results in significant transcriptional dysregulation, affecting multiple cellular processes including apoptosis, cell cycle and metabolism.

Venetoclax induces ferroptotic cell death in CREBBP-mutated B-ALL
To explore the transcriptional impact of low-dose Venetoclax treatment of 697 KI cells, we employed a four-way interaction model to identify genes specifically dysregulated in 697 KI cells upon Venetoclax treatment (Extended Data Fig. 4a).KEGG pathway analyses of these genes showed further downregulation of cell cycle-associated genes and enrichment for metabolic pathways and ferroptosis in Venetoclax-treated 697 KI cells (Fig. 4a).Furthermore, GSEA showed marked up-regulation of genes associated with multiple metabolic processes, ROS scavenging, ferroptosis and the unfolded protein response (Fig. 4b).
Ferroptosis is a distinct form of programmed cell death resulting from iron-catalyzed reactive oxygen species (ROS)-mediated damage to unsaturated fatty acids of membrane phospholipids 26 .It commonly co-associates with apoptosis and is associated with expression cell death markers, including Annexin-V externalization 27 .We demonstrated evidence of ferroptosis upon Venetoclax treatment specifically occurring in 697 KI cells using in-vitro assays.Exposure to the cell-permeable pancaspase inhibitor Z-VAD partially rescued viability to high-dose Venetoclax in both 697 WT and 697 KI cells, indicating a role for intrinsic, caspase-mediated apoptosis induced by high-dose Venetoclax in both lines (Fig. 4c).Conversely, exposure to Liproxstatin 1, which specifically inhibits ferroptosisassociated lipid peroxidation, rescued viability in 697 KI cells exposed to either low or high-dose Venetoclax (Fig. 4d), with no effect seen in 697 WT cells.BCL2 inhibition with either Venetoclax or Navitoclax was also associated with ferroptosis in 697 KI/KO and REH Mut cells, as shown by elevated BODIPYC11 staining, an indicator of lipid peroxidation (Fig. 4e and Extended Data Fig. 4b-d).
We tested the intrinsic susceptibility of 697 WT and 697 KI cells to ferroptosis.The inhibition of glutathione phospholipid peroxidase activity using the GPX4 inhibitor RSL3 induced significantly higher levels of lipid peroxidation in 697 KI cells, further demonstrating their enhanced sensitivity to redox stress (Fig. 4f).Metabolically, low dose (20nM) Venetoclax specifically resulted in a reduction in both basal and maximal OxPhos in 697 KI cells, whereas only high-dose Venetoclax (2000nM) resulted in a marked reduction in OxPhos in both 697 KI and 697 WT cells, consistent with BCL2's role in regulating mitochondrial outer membrane permeabilization (Fig. 4g).
Overall, these findings demonstrate that the major driver underlying the sensitivity of CREBBPmutated cells to BCL2 inhibition is ferroptotic programmed cell death, associated with underlying metabolic dysregulation.

Pharmacological inhibition of CREBBP function can sensitize B-ALL to Venetoclax in-vitro
The majority of B-ALL patients do not have CREBBP-mutations.Given the strength of the association with Venetoclax sensitization we hypothesized that pharmacological inhibition of CREBBP WT function could sensitize CREBBP WT B-ALL to BCL2 inhibitors.
Pre-treatment of 697 WT or REH WT cells with the preclinical CREBBP/EP300 HAT inhibitor A485 almost perfectly phenocopied the Venetoclax sensitization seen in our isogenic lines (Fig. 5a and Extended Data Fig. 5a).Furthermore, co-treatment with A485 and Venetoclax showed strong pharmacological synergy (Fig. 5b).Further corroborating our isogenic findings, A485-sensitized Venetoclax cytotoxicity was associated with enhanced lipid peroxidation consistent with ferroptosis (Fig. 5c and Extended Data Fig. 5b).Similar results were also seen with the early-phase CREBBP/EP300-specific bromodomain inhibitor Inobrodib (Extended Data Fig. 5c,d).In both 697 WT and REH WT cells, A485induced HAT inhibition was associated with increased basal and maximal OxPhos with an increase in SRC, similar to that seen in 697 KI cells (Fig. 5d and Extended Data Fig. 5e-g).Single-agent A485-treated 697 WT cells showed no loss of proliferation or evidence of programmed cell death by annexin-V externalization, strongly suggesting that at these doses A485 works by sensitizing cells to Venetoclaxinduced cytotoxicity (Fig. 5e,f).
We extended these findings to other human B-ALL cell lines.A485 sensitized the highly Venetoclaxresistant, BAX-mutated cell line NALM6 to Venetoclax (Fig. 5g).This was associated with increased lipid peroxidation, indicative of ferroptotic death (Fig. 5h and Extended Data Fig. 5h).In Venetoclaxsensitive B-ALL lines driven by high-risk genetic drivers, Venetoclax and A485 showed evidence of pharmacological synergy, indicating that this interaction is conserved across diverse genetic subtypes of B-ALL (Fig. 5i).
Collectively, these findings pharmacologically validate the findings of BCL2 sensitization identified in our CREBBP mutated cell lines and support a possible novel drug combination for clinical translation in B-ALL more broadly.

Genetic or pharmacological inhibition of CREBBP sensitizes B-ALL to Venetoclax in-vivo
Finally, we sought to test whether Venetoclax could target CREBBP-mutated B-ALL in-vivo, where cell extrinsic factors and pharmacodynamic effects can result in reduced efficacy or highlight dose-limiting toxicities.
Luciferase-expressing 697 WT and 697 KI cell lines were engrafted into NOD-SCID-Gamma (NSG) mice (Extended Data Fig. 6a,b).Upon confirmation of engraftment by bioluminescent imaging (BLI), mice were treated with daily oral Venetoclax for up to 30d (Fig. 6a).As anticipated, Venetoclax exposure was associated with limited disease control in 697 WT cells (Fig. 6b,c) resulting in a small, but significant, prolongation of post-transplant overall survival (OS), with all Venetoclax-treated 697 WT recipient animals succumbing within 15d of treatment (median OS 19 vs. 22d, p=0.0011) (Fig. 6f).697 KI recipient mice treated with vehicle control succumbed to disease slightly later than 697 WT , potentially reflecting the proliferation defect characterized in-vitro (median OS 25 vs. 19d, p=0.0004) (Fig. 6f); nevertheless, all vehicle-treated 697 KI mice succumbed by 19d of treatment.Venetoclax-treated mice engrafted with 697 KI cells demonstrated markedly improved disease control by BLI (Fig. 6d,e and Extended Data Fig. 6c,d).Only one Venetoclax-treated 697 KI recipient succumbed to disease within the treatment window, with the remaining 5 recipients gaining up to 2 weeks of survival after cessation of Venetoclax exposure (median OS 47.5 vs. 25.5d,p=0.0006) (Fig. 6f).
Lastly, we tested whether co-administration of the orally-available, early-phase CREBBP bromodomain inhibitor Inobrodib could sensitize 697 WT cells to Venetoclax in-vivo (Extended Data Fig. 6e).Once-daily oral dosing of single-agent Inobrodib did not confer survival benefit compared to recipients treated with either vehicle or single-agent Venetoclax, whereas recipient mice treated with both agents demonstrated a significant survival advantage (median OS combination 34.5 vs. Inobrodib only 24.5d, p=0.0084) (Fig. 6g).
Overall, these findings demonstrate that: i) oral Venetoclax can be highly efficacious in controlling CREBBP-mutated B-ALL in a preclinical in-vivo model; and ii) pharmacological inhibition of CREBBP is a tolerable and efficacious means of sensitizing B-ALL to Venetoclax in-vivo.

Discussion:
There is a pressing clinical need to develop novel treatments for patients with high-risk B-ALL, such as those harbouring mutations in CREBBP.Using a highly-curated panel of clinically-tractable small molecules in an isogenic human cell line model, we demonstrate that CREBBP mutation sensitizes cells to Dexamethasone, inhibitors of residual CREBBP/EP300 function and most potently BCL2 inhibition.CREBBP LOF increases metabolic rate and upon BCL2 inhibition results in ferroptotic cell death, which can be phenocopied by pharmacological CREBBP inhibition in genetically-diverse B-ALL cell lines, providing a readily translatable novel drug combination for B-ALL more broadly.
Our model shows a number of significant similarities to previously published work.The global transcriptional profile of our CREBBP-mutated cells strongly overlaps with those seen in murine Crebbp -/-germinal-centre B-cells 25 .Other models of CREBBP-mutated B-ALL and B-cell lymphoma have implicated changes in cell cycle, metabolism, DNA damage response and apoptosis 3,10,19 .We also note that hypodiploid B-ALL, a genetic subtype highly associated with CREBBP mutations, has been linked with Venetoclax sensitivity 20 .Moreover, preliminary results from an early phase clinical trial of Venetoclax in relapsed pediatric malignancies (NCT03236857) have specifically highlighted responses in the small number of CREBBP-mutated B-ALL patients enrolled, providing supportive evidence that our results will be clinically translatable 28 .
Our CREBBP-mutated model did not show significant differential sensitivity to cytotoxic chemotherapy in the ex vivo setting.Our cell lines also showed a small but significant sensitization to Dexamethasone, suggesting that CREBBP LOF does not significantly alter responses to glucocorticoids 4,11,12 .Previous reports have shown variable association of CREBBP LOF with chemoresistance in-vitro 10,12,19 and overall, our findings add to a growing body of evidence that CREBBP LOF does not in itself provoke significant cell-intrinsic chemoresistance.
A number of possible routes to chemoresistance remain.Recent reports have shown that LOF mutations affecting multiple epigenetic regulators across multiple malignancies can increase tolerance to environmental stress, a characteristic unlikely to read out in standard in-vitro culture conditions 29 .A prominent phenotype of our model was cell cycle retardation, demonstrable both transcriptionally and in in-vitro assays, and potentially resulted in the relatively delayed engraftment of 697 KI cells in NSG mice.It is therefore possible that slower cell-cycle kinetics could result in relative resistance to predominantly S-phase-targeting cytotoxic chemotherapy, as has been shown in preclinical models of chemoresistance in B-ALL 30 .The mechanism underlying delayed cell cycle progression is likely to be multifactorial, but we note recent work implicating CREBBP/EP300's role in maintaining enhancer acetylation marks during cell division 31 .Thus, a failure to efficiently reconstitute the enhancer landscape during mitosis could plausibly result in delayed cell-cycle progression and may also explain why genetic or pharmacologically-induced loss of CREBBP function exerts a more potent phenotype 4 than complete loss of CREBBP protein, through steric competition with residual compensatory EP300 activity.
We show that BCL2 inhibition results in ferroptotic cell death in the context of CREBBP LOF.Unlike apoptosis, ferroptosis is not mediated by a defined biochemical pathway, rather is the output of a combination of underlying metabolic state, ROS scavenging capacity and lipid composition 32 .B-cell progenitors are exquisitely sensitive to redox balance 33 and recent genetic perturbation data has highlighted an underlying propensity for ferroptotic cell death in B-ALL 34 .Reflecting its pleiotropic role in biology 35 , how CREBBP loss alters this balance is likely to be multifactorial, including by direct transcriptional changes, as well as post translational acetylation of proteins affecting metabolism and redox balance (Fig. 6h).Our functional experiments show significant increases in both glycolytic and mitochondrial metabolism upon CREBBP LOF.Thus, in contrast with studies in AML, we show an association between enhanced metabolic state and BCL2 dependence, likely relating to a B-cell-state specific susceptibility to oxidative stress 23,34 .Concordantly, we hypothesize that the deep repression of AGPS seen in 697 KI cells is an adaptation to limit production of the unsaturated lipid substrates responsible for ferroptosis 36 .
Lastly, we demonstrate that small molecule inhibition of CREBBP can sensitize genetically-diverse B-ALL cell lines to Venetoclax, potentially widening the paradigm of our combination to multiple ALL genotypes, and the clinically-actionable combination of Venetoclax and Inobrodib was tolerable and highly efficacious in in-vivo models.We are aware of preliminary reports of this combination being trialled in related haematological malignancies 37 and propose this as a rational drug combination for B-ALL.

MTS
All drugs were purchased from Selleckchem, except A1155463 (Cayman Chemicals), and reconstituted to 10mM in DMSO.72-hour viability assays were performed using CellTiter 96® AQueous One Solution Cell Proliferation Assay (Promega, G3580) following manufacturer recommendations and 490nM absorbance measured using a CLARIOstar plate reader (BMG Labtech).Drug synergy scores were calculated using SynergyFinder web tool (https://synergyfinder.fimm.fi) 38.We defined a peak ZIP threshold of +10 for synergism.

Annexin-V Staining by Flow Cytometry
Annexin-V externalization was assessed using anti-AnnexinV-APC kit from (eBioscience TM 88-8007-74) following manufacturer protocol.Briefly, after incubation with drug/vehicle, cells were washed in DPBS, resuspended in Annexin binding buffer and incubated with 1/40 Annexin-V-APC dilution for 15min at room temperature (RT) in the dark.Cell pellet was resuspended in 7-AAD (BD, 559925) diluted 1/100 in 1x binding buffer and analyzed by flow cytometry.

Lentiviral Production
Lentiviral vectors were produced in HEK-293T cells by co-transfection of the previously sequenced vector constructs with psPAX and pMDG.2 packaging plasmids using Trans-IT LT-1 (Mirus, MIR2700).HEK293T cells were incubated overnight at 37°C and the next day medium was changed for IMDM 10% HI-FBS.Viral particles were harvested at 48 and 72h by centrifugation and 45μm filtration.250µl of viral particles were used to transfect 1x10 6 B-ALL cells by spinoculation (900g for 2h at 32°C) in the presence of 10μg/ml Polybrene (sc-134220).Subsequently, the cultures were diluted in fresh media and washed three times the following morning.
Transfection efficiencies were assessed by flow cytometric reporters and transfected cells selected using either 0.5μg/ml of Puromycin (Gibco) for three days or FACS cell sorting as appropriate to achieve >90% transfection.
Competitive co-culture assays were performed by mixing an equal number of renilla shRNA-GFP and one of two BCL2 shRNA-mCherry transfected 697 WT or 697 KI cells in triplicate.BCL2 knockdown was induced with doxycycline and aliquots of cell culture analyzed daily by flow cytometry.Results are presented as the ratio of 697 KI Cherry:GFP events normalized to the mean ratio of 697 WT Cherry:GFP events.

Gene expression by Reverse Transcriptase Quantitative PCR (RT-qPCR)
Total RNA was extracted using the RNeasy kit (Qiagen) and 1μg was used to synthesize cDNA using TaqMan Reverse transcription (Applied Biosystems, N8080234) following manufacturer's instructions.RT-qPCR was carried out with SYBR Green PCR master mix (Agilent, 600882) on cDNA (diluted 1:5 in water).The expression level of RNA was calculated using the standard curve method, normalized to the expression of GAPDH.Primer pairs: BCL2F:ATTGGGAAGTTTCAAATCAGC; BCL2R:TGCATTCTTGGACGAGGG; GAPDHF:CCACATCGCTCAGACACCAT; GAPDHR:CCAGGCGCCCAATACG.

Seahorse Metabolic Assays
Oxphos and glycolysis assays were performed using the Seahorse XFe96 analyser (Agilent) according to manufacturer's instructions.Briefly, cells were analysed at 48-72h post passage and in drug concentrations/DMSO vehicle as stated in figure legends.For glycostress experiments, cells were harvested and counted in Seahorse XF base medium, pH7.4 supplemented with 2mM L-Glutamine (Sigma Aldrich, G7513).The final concentration of the injected drugs in the glycostress test is 10mM glucose (Sigma, G7021), 1µM Oligomycin A and 50mM of 2-deoxyglucose (2-DG) (Sigma, D3179).For mitostress tests, Seahorse XF base media was supplemented with 2mM L-Glutamine, 10mM Glucose and 1mM Sodium pyruvate (Sigma, S8636).The final concentration of the injected drugs was 1µM Oligomycin, 1µM FCCP supplemented with 1µM Sodium pyruvate and 1µM Antimycin.For experiments using Venetoclax or A485 pretreatment drug exposure continued in the Seahorse media.Cells were plated at a density of 70000 cells/well in XFp tissue culture plates previously coated with CellTak (Corning, 354240) and incubated for 45mins at 37 o C in a CO2-free incubator prior to analysis.

Lipid Peroxidation Assays
Lipid peroxidation was assessed using the lipid peroxidation sensor BODIPYC11-581/591 (ThermoFisher, D3861) according to manufacturer's instructions.Briefly, cells were washed in DPBS and stained with 4µM BODIPY dye for 30 min at 37 o C with gentle shaking in the dark.After washing with DPBS, cells were analyzed by flow cytometry.Replicate data is presented as the ratio of 488nmexcited 530/30 ("FL1") to 610/20 ("FL3") filtered signal.

RNA Sequencing Library Preparation
697 WT and 697 KI treated with 20nM Venetoclax or DMSO vehicle for 24h (Fig 4 ) were harvested, washed and RNA was extracted from 4x10 6 cells using the RNeasy kit (Qiagen) (performed in triplicate).Libraries were prepared using NEBNext Poly(A) mRNA magnetic Isolation module (NEB, E7490) starting with 1µg DNA-free RNA, as per manufacturer's instructions.RNA fragmentation, double-strand cDNA synthesis, end repair, adapter ligation and PCR amplification (9 cycles) was performed using NEBNext Ultra II directional RNA library prep kit for Illumina protocol (NEB, E7760) according to manufacturer's protocol.Library quality and molarity was measured by Qubit (ThermoFisher) and TapeStation (Agilent).Samples were 50-bp paired-end sequenced on the NovaSeq (Illumina) instrument (CRUK CI Genomics Core Facility).

RNAseq analysis
The quality of the paired-end RNA-seq reads were assessed using FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/).All RNA-seq libraries were found to be above the minimum quality thresholds across quality control metrics.Adapter sequences were trimmed using TrimGalore package (https://github.com/FelixKrueger/TrimGalore),and were next mapped against the human genome version 38 (Hg38) using STAR version=2.7.10a 41 .Only uniquelymapping high confidence reads (flags NH:i:1 and MAPQ=255) were retrieved and used for transcriptome quantification with featureCounts version 2.0.1 42 .Genes with a minimum of 10 reads or more in a minimum of two samples were considered for all downstream analysis.Differential expression analysis was based on either a single factor model (~ genotype/treatment/sensitivity) or an interaction model when two factors were considered (~ genotype * treatment), and the interaction term was used to determine the difference of the effect of treatment across genotypes.Differential expression analysis was conducted using gene level transcriptomic reads using the Bioconductor package DESeq2 43 .KEGG Pathway enrichment, Gene Set Enrichment Analyses (GSEA) and Gene Ontology (GO) Analyses were all performed using the R package 'clusterProfiler' version 4.4.4 44.A minimum p-value (corrected for multiple testing) and FDR of 0.05 were used across all tests to establish statistical significance.
For the in-vivo dosing experiments in Figure 7G, 0.25x10 6 697 WT cells were injected by tail vein injection into sub-lethally (2Gy) irradiated 11-13 week-old old male NSG mice.Mice received a daily dose of Venetoclax (100mg/kg) and/or Inobrodib (20mg/kg) or vehicle control (as above), by means of oral gavage until the experimental endpoints were reached.
Mice were housed in a pathogen-free animal facility and were allowed unrestricted access to food and water.All experiments were conducted under a UK Home Office project (under the Animals (Scientific b, Cell-based drug screening was performed using a panel of 32 small molecule compounds predicted to have differential sensitivity between CREBBP WT and mutant lines.72h viability was measured by MTS assays using a wide concentration range between 3pM to 30µM in triplicate and repeated using narrower concentration ranges to define IC50 where appropriate.Results are presented as an IC50 ratio of 697 KI (blue) and 697 KO

Figures:Figure 1 :
Figures: (yellow) clones compared to WT. c, Dose response curves of two CREBBP/EP300 inhibitors Inobrodib (left) and A485 (right) showing enhanced sensitivity of 697 KI (blue) compared to 697 WT (black) in 72h MTS viability assay.Performed in triplicate, mean ± SD. d, Dose response curve of 697 WT (black) and 697 KI (blue) lines to Venetoclax in 72h MTS viability assay.Performed in triplicate, mean ± SD. e, Growth curve of 697 WT (black) and 697 KI (blue) grown in the presence of either DMSO vehicle (dotted lines) or 20nM Venetoclax (solid lines).f, Mitochondrial depolarization as assessed by staining for JC1 by flow cytometry (488nm 530/30) in response to DMSO vehicle, or Venetoclax at 20 or 2000nM in 697 WT (black) and 697 KI (blue) cell lines.g, Representative flow cytometry plots of externalization of Annexin-V (reported by APC) in response to 24h exposure to DMSO vehicle (left) or Venetoclax 20nM (right) in 697 WT (top) and 697 KI (bottom) cell lines.Viability is assessed by 7AAD exclusion.h, Summary of 3 replicate experiments measuring proportion of viable 7AAD -ve Annexin-V +ve early apoptotic cells.Mean ± SD, 2 way ANOVA **** , P <0.0001.

Figure 2 :
Figure 2: Venetoclax exerts its effect on CREBBP-mutated B-ALL by on-target inhibition of BCL2.a,Schematic of doxycycline-inducible shRNA KD system competitive co-culture assay.697 WT and 697 KI cells were stably transfected with two separate doxycycline-inducible shRNAs targeting BCL2, reported by mCherry, or a control shRNA targeting Renilla, reported by GFP.BCL2 and Renilla shRNA-expressing cells were mixed in equal numbers, doxycycline added to the media (500ng/ml) to induce shRNA

Figure 5 :
Figure 5: Pharmacological inhibition of CREBBP function can sensitize B-ALL to Venetoclax in-vitro.a, Dose response curve of 697 WT (black) and 697 WT pre-treated with 3 days of A485 (pale blue) to Venetoclax in 72h MTS viability assays.Performed in triplicate, mean ± SD.

Figure 6 :
Figure 6: Genetic or pharmacological inhibition of CREBBP sensitizes B-ALL to Venetoclax in-vivo.a,Schema of 697 WT vs. 697 KI in-vivo drug dosing protocol.NSG mice were intravenously injected with 697 WT vs. 697 KI luciferase-expressing cell lines.At 8 days post injection animals were administered Venetoclax (100mg/kg) or vehicle control by oral gavage.Mice were imaged by BLI as indicated.N=6 per group.b, Representative BLI images of mice engrafted with 697 WT cells at days 11 and 17 of treatment.c, Average BLI radiance of mice engrafted with 697 WT cells at days 11 and 17 of treatment (log p/s/cm 2 /sr).Mean ± SEM. d, Representative BLI images of mice engrafted with 697 KI cells at days 11, 17 and 21 of treatment.e, Average BLI radiance of mice engrafted with 697 KI cells at days 11, 17 and 21 of treatment (log p/s/cm 2 /sr).Mean ± SEM. f, Overall survival of NSG mice engrafted with 697 WT vs. 697 KI B-ALL cells treated with oral Venetoclax (100mg/kg) or vehicle control.30 day maximal treatment window is indicated.N=6 per group.Significance calculated by Mantel Cox log-rank test.g, Overall survival of NSG mice engrafted with 697 WT B-ALL cells treated with oral Inobrodib (20mg/kg) or a combination of Inobrodib (20mg/kg) plus Venetoclax (100mg/kg) (n=4 per group).Engraftment control groups were treated with Venetoclax only (n=2) or vehicle control (n=1).Treatment began at day 8 post-transplant.Significance calculated by Mantel Cox log-rank test.

b ,
GSEA analysis of ranked genes of RNAseq analysis comparing DMSO vehicle-treated 697 KI with 697 WT .NES: normalized enrichment score; FDR: false discovery rate .Comparison of BCL2 expression levels in 697 WT (black) and 697 KI (blue) cells by RNAseq.Each dot represents one sample.Bars show mean fragments per kilobase of transcript per million fragments mapped (FPKM) value ± SD, n=3, significance calculated by unpaired t test, **, P=0.0038.Comparison of CDKN2A expression levels in 697 WT (black) and 697 KI (blue) cells by RNAseq.Each dot represents one sample.Bars show mean FPKM value ± SD, n=3, significance calculated by unpaired t test, *** , P=0.0002.Proliferation of untreated 697 WT (black), 697 KI (blue) and 697 KO (yellow) cells measured by direct counting.Performed in triplicate, mean ± SD. f, Analysis of cell cycle stage by FUCCI reporter system.Percentage proportion of cells in G1, Early S and G2-S-M phases in 697 WT (black) and 697 KI (blue) cells.N=3, each dot represents mean average of 3 technical replicates, bar shows mean average; significance calculated by 2 way ANOVA, **** , P <0.0001,Early S P=0.0498; G2-S-M P=0.0342.Glycolytic rate measured by extracellular acidification rate (ECAR) using Seahorse (Agilent) Glycostress test in 697 WT (black) and 697 KI (blue) cells.Left panel: Representative ECAR plot over time.Mean ± SEM.Right panel: Summary of maximal ECAR.Each dot represents a single replicate acquired from two separate experiments.Significance calculated by unpaired t test, **** , P <0.0001.Mitochondrial oxygen consumption rate (OCR) measured using Seahorse (Agilent) Mitostress test.Left panel: Representative OCR plot over time.Mean ± SEM.Right panel: Summary of basal OCR (left), maximal OCR (middle) and spare respiratory capacity (right).Each dot represents a single replicate acquired from two separate experiments.Significance calculated by unpaired t test, **** , P <0.0001; c, d, e, g, h, * , P=0.0315.