Targeting mitochondrial respiration and the BCL2 family in high‐grade MYC‐associated B‐cell lymphoma

Multiple molecular features, such as activation of specific oncogenes (e.g., MYC, BCL2) or a variety of gene expression signatures, have been associated with disease course in diffuse large B‐cell lymphoma (DLBCL), although their relationships and implications for targeted therapy remain to be fully unraveled. We report that MYC activity is closely correlated with—and most likely a driver of—gene signatures related to oxidative phosphorylation (OxPhos) in DLBCL, pointing to OxPhos enzymes, in particular mitochondrial electron transport chain (ETC) complexes, as possible therapeutic targets in high‐grade MYC‐associated lymphomas. In our experiments, indeed, MYC sensitized B cells to the ETC complex I inhibitor IACS‐010759. Mechanistically, IACS‐010759 triggered the integrated stress response (ISR) pathway, driven by the transcription factors ATF4 and CHOP, which engaged the intrinsic apoptosis pathway and lowered the apoptotic threshold in MYC‐overexpressing cells. In line with these findings, the BCL2‐inhibitory compound venetoclax synergized with IACS‐010759 against double‐hit lymphoma (DHL), a high‐grade malignancy with concurrent activation of MYC and BCL2. In BCL2‐negative lymphoma cells, instead, killing by IACS‐010759 was potentiated by the Mcl‐1 inhibitor S63845. Thus, combining an OxPhos inhibitor with select BH3‐mimetic drugs provides a novel therapeutic principle against aggressive, MYC‐associated DLBCL variants.


Introduction
Diffuse large B-cell lymphoma (DLBCL) is a heterogeneous disease with variable clinical course. While 50-60% of all patients achieve cure with standard R-CHOP immune-chemotherapy, others succumb despite aggressive salvage regimens, including high-dose chemotherapy with autologous stem cell support, allogeneic transplantation, or CAR T-cell therapy [1]. Multiple genetic and molecular features have been associated with disease course in DLBCL, including translocation and/or overexpression of specific oncogenes (e.g., MYC, BCL2), mutational profiles, and transcriptome-based classifiers such as the socalled cell-of-origin (COO) and comprehensive consensus clustering (CCC) [1][2][3][4][5]. While MYC, BCL2, and COO-associated gene expression signatures are now commonly monitored in the clinic [1], other features have yet to make their way into routine practice, and full elucidation of their clinical relevance is still lacking. Moreover, complex combinatorial arrangements of the above features have far-reaching implications in disease classification and prognosis, some of which are just beginning to emerge [4][5][6].
A series of coincidental observations pointed to a hypothetical relationship between MYC and one of the signatures defined in the CCC model, termed 'OxPhos' (henceforth CCC-OxPhos) based on its enrichment for genes involved in Oxidative Phosphorylation [2]. First, MYC activation (whether assessed by gene rearrangement, protein expression, or associated gene signatures) and CCC-OxPhos have both been linked to poor prognosis in DLBCL patients treated with R-CHOP [2,3,[7][8][9], both behaving independently from the COO classification system [2][3][4]. Second, MYC upregulates multiple genes involved in mitochondrial transcription and translation [10][11][12] and increases the cells' dependency upon those processes [11,12], both of which are required for assembly of the electron transport chain (ETC) and oxidative phosphorylation. In line with this principle, tigecycline-an antibiotic that inhibits the mitochondrial ribosome and impairs OxPhos activity [13,14]-showed increased toxicity toward either MYC-overexpressing cells [11,12] or DLBCL cell lines of the CCC-OxPhos subtype [15]. Yet, whether MYC-and OxPhos-associated gene expression signatures might be related in DLBCL remains to be addressed: we show here that this is indeed the case, with high levels of correlation across multiple patient cohorts.
Altogether, the above observations led us to assess the potential of ETC inhibitors as therapeutic agents against MYC-associated high-grade DLBCL. We focused on IACS-010759 [16], a small-molecule inhibitor of ETC complex I that showed antitumoral activity in preclinical models of acute myeloid leukemia (AML), chronic lymphocytic leukemia, mantle cell lymphoma and lung cancer [16][17][18][19]. Our data show that IACS-010759 preferentially induces apoptosis in MYC-overexpressing cells and cooperates with the BCL2 inhibitor venetoclax in killing MYC/ BCL2 DHL cells, as previously reported for Tigecycline [20]. Furthermore, in BCL2-negative, MYC-translocated lymphoma cell lines, the cytotoxic activity of IACS was potentiated by the Mcl-1 inhibitor S63845. Hence, our data point to the possible use of ETC inhibitors in combination with distinct BH3-mimetic compounds [21] for the treatment of high-grade DLBCL, in a subtypeand patient-specific manner.

Analysis of DLBCL cohorts
For survival and correlation analyses in R-CHOPtreated DLBCL patients' cohorts, we used gene expression data from six publicly available datasets [3][4][5][22][23][24]: Data collection methodologies and patient written consents were as provided in the original studies. For RNA-seq data [3,5,24], RNA FastQ files were processed with the same pipeline used for our RNA-seq data, except for being aligned to hg19; for microarray data [4,22,23], probes were matched to the corresponding gene and, for those genes with more than 1 probe set, the mean expression was calculated. In all cases, data were first normalized to Transcripts Per kilobase Million (TPM); then, for each gene, the zscore across all samples was calculated. The Log-rank test was applied (SURVMINER R package, https://rpkgs. datanovia.com/survminer/index.html, https://cran.rproject.org/web/packages/survival/index.html) for comparing survival curves of R-CHOP-treated patients stratified according to the mean z-score of the expression of the genes belonging to the Hallmark-MYC-V1 or OxPhos signatures. For the purpose of computing the linear correlation between signatures within a dataset, outliers were excluded from the calculation. Outliers were identified with the interquartile range (IQR) method as cases were the expression of the signature was not comprised within the following lower and upper boundaries: Lower boundary, 25th quantile -(IQR * 1.5); Upper boundary, 75th quantile + (IQR * 1.5), where IQR = 75th quantile -25th quantile. Moreover, in order to perform the analysis under stringent conditions, each pairwise correlation was calculated following exclusion of the genes that were shared between the two gene signatures (Table S2).
For the analysis of drug responses in vivo, 5 9 10 6 SU-DHL-6 or DOHH-2 cells were xenografted subcutaneously in irradiated (3 Gray), 8-week-old, female CD1-nude nu/nu mice (Envigo, Indianapolis, IN, USA), and expanded by serial subcutaneous transplantation of tumor fragments. Tumors were allowed to grow for about 2 weeks, followed by exclusion of outliers, randomization of the experimental groups and start of treatments (days 0-11). Tumor volumes were assessed from the start of the treatment every 2 days with a digital caliper and calculated as 1/2 length 9 width 2 (mm 3 ). The following treatment schemes were used: daily oral gavage with venetoclax and IACS-010759 for 5 days, followed by two days off and a repeat of the same scheme, for a total of 12 days. Venetoclax was dissolved in 60% Phosal 50 PG (Lipoid, Steinhausen, Switzerland), 30% Polyethylene glycol (Merck, Darmstadt, Germany), 10% Ethanol; IACS-010759 was suspended in 0.5% methylcellulose (Merck). During the experiment, two animals treated with the highest daily dose of IACS-010759 (15 mgÁkg À1 ), given alone and in combination venetoclax, respectively, showed signs of toxicity, and thus, the treatment was discontinued and the animals were pulled out from the respective group.
The DHL patient-derived xenograft (PDX) model DFBL-20954-V3-mCLP [30] was obtained from the Dana Farber Cancer Institute Center for Patient Derived Models (CPDM). 1 9 10 6 cells were xenografted via tail vein injection into 8-week-old, male NSG mice (Charles River, Calco, Italy). Tumor engraftment was confirmed 7 days after transplant by whole-body imaging on an IVIS Lumina III platform following intraperitoneal injection of 150 mgÁkg À1 XenoLight D-Luciferin (PerkinElmer, Waltham, MA, USA) and anesthesia with isoflurane. The animals were subsequently randomized in the different experimental groups to start the treatment (day 0-11); the response to treatment was assessed by whole-body imaging at day 4, day 11, and day 14. The data were analyzed with the Living Image Software, version 4.2 (Caliper Life Sciences, Hopkinton, MA, USA). Radiant efficiency was quantified bilaterally on femurs based on the epifluorescence signal as indicated in the user manual.
Experiments involving animals were done in accordance with the Italian Laws (D.lgs. 26

Reagents and kits
L-aspartic acid (Merck) was added to the growth medium and the pH adjusted to 7.5 with 1 M NaOH before FBS supplementation; hypoxanthine, adenine, and uridine (all from Merck Life Science) were solubilized and added directly to the growth medium. IACS-010759 [16], venetoclax/ABT-199 (Carbosynth, Newbury, UK) [34], S63845 (Medchemexpress LCC, Monmouth Junction, NJ, USA) [35], Z-VAD-FMK (Santa Cruz Biotechnologies, Santa Cruz, CA, USA) were dissolved in DMSO and added directly to the culture medium; 4hydroxytamoxifen (OHT; Merck) was dissolved in ethanol and added directly to the culture medium.
An ADP/ATP Ratio Assay kit (Merck) was used to quantify cellular ADP and ATP, while Caspase-Glo 3/ 7 Assay (Promega, Madison, WI, USA) was used to evaluate caspase activity, both according to the manufacturer's instructions.

Immunoblot analysis
After collection, cells were divided in two aliquots: One was extracted as previously described [36] to be used for quantification by Bradford protein assay (Bio-Rad Laboratories, Hercules, CA, USA), while the other was directly lysed in Laemmli buffer and denatured at 95°C for 5 min for subsequent SDS/PAGE. After clearance by centrifugation, 10 µg of proteins from each sample was loaded on an acrylamide gel for SDS/PAGE, followed by transfer on a nitrocellulose membrane and immunoblot with the antibodies listed in Table S3. All data shown are representative of at least n = 2 experiments.

Fluorescence microscopy
Nonadherent cells were subjected to cytospin, fixed in 4% paraformaldehyde (PFA), and permeabilized with 0.1% Triton X-100 for subsequent staining as described previously [37]. Single optical sections were acquired with an SP8 confocal microscope (Leica Microsystems, Wetzlar, Germany) equipped with a 639/1.4 oil immersion objective lens. For the analysis of apoptotic Figures, cells were stained with 4 0 ,6diamidino-2-phenylindole (DAPI) and tetramethylrhodamine (TRITC) -conjugated Agglutinin (both from Thermo Fisher Scientific). For cytochrome c localization, cells were stained overnight at 4°C with an antibody against cytochrome c (clone 7H8.2C12, Thermo Fisher Scientific) and counter-stained with DAPI. Secondary staining was performed with an Alexa Fluor 488-conjugated anti-mouse antibody (Jackson Immu-noResearch, West Grove, PA, USA). The total and cytochrome c-positive cellular areas were calculated from > 20 fields for each sample with an in-house developed macro for IMAGEJ (RRID: SCR_003070). Briefly, the cytochrome c signal from each cell in the field of view was segmented using an automatic threshold (Otsu algorithm) and the area of the signal calculated on the binary images. The cytochrome c area was then normalized by the total cell area, which was identified on the bright-field transmission image.

Flow cytometry
Flow cytometry was conducted on a MACSQuant Analyzer (Miltenyi Biotec, Bergisch Gladbach, Germany) and data analyzed with FLOWJO software (version 10.6.1; BD Biosciences, Franklin Lakes, NK, USA; RRID: SCR_008520). For cell number and viability counts, cells were resuspended in ice-cold PBS in the presence of 1 µgÁmL À1 Propidium Iodide (P.I.). Cell viability is given as percentage of live cells on total cells counted, while the proliferation index is defined as the ratio of viable cells in the sample to viable cells in the corresponding IACS-010759-untreated control.
For evaluation of apoptotic cell death, cells were stained with APC-conjugated Annexin V (Thermo Fisher Scientific) and P.I. as previously described [11]. The cytoplasmic release of cytochrome c was detected as described [38] with few modifications: Briefly, cells were permeabilized with 0.002% digitonin for 20" under vortexing before fixation in 4% PFA and permeabilization with 0.1% Triton X-100, for subsequent immunostaining with anti-cytochrome c (clone 6H2.B4, BD Biosciences) overnight at 4°C. Secondary staining was performed with an Alexa Fluor 647-conjugated antimouse antibody (Jackson ImmunoResearch).

Cell cycle kinetics
For the analysis of cell cycle progression, cells were incubated for 20 min in the presence of 33 µM bromodeoxyuridine (BrdU, Merck), washed, resuspended, and incubated in fresh medium: at the indicated time-points ( Fig. S2A), aliquots were collected for ethanol-fixation followed by FITC-conjugated anti-BrdU (BD Biosciences) and P.I. staining. Stained samples were analyzed by flow cytometry on a MACSQuant Analyzer.

Oxygen consumption rate (OCR) and respiratory parameters
Oxygen consumption rate was measures on the Seahorse XFe96 Analyzer (Agilent Technologies, Santa Clara, CA, USA) using SeaHorse XF Cell Mito Stress Test, following the manufacturer's instructions. Briefly, on the day of the assay, cells were counted and attached to 96well Seahorse cell culture microplates, precoated with Corning TM Cell-Tak (Sacco, Cadorago, Italy) according to the manufacturer's instructions, at a density of 80 000 cells per well. Cells were seeded in at least eight wells per experimental condition, in XF RPMI Medium pH 7.4 with 1 mM HEPES (Agilent), supplemented with 2.75 mM glucose, 1 mM sodium pyruvate, 2 mM Lglutamine and, where specified, 100 nM OHT, 135 nM IACS-010759, 100 nM venetoclax. The plates were incubated at 37°C for 1 h in a non-CO 2 incubator. After OCR baseline measurements, oligomycin A, Carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP), and antimycin A/rotenone were added sequentially to each well, to final concentrations of 1, 1.5, and 0.5 lM, respectively. Results were normalized by cell number, measured at the end of the experiment using the CyQUANT Cell Proliferation Assays (Thermo Fisher Scientific). Data are expressed as pmol of oxygen per minute per arbitrary units (pmolÁmin À1 Áa.u. À1 ). Respiratory parameters were calculated with the Wave Desktop 2.6 software and shown as mean AE SEM of three independent experiments.

RNA extraction and quantitative PCR analysis
Total cellular RNA was extracted using the Quick-RNA Miniprep kit (Zymo Research, Irvine, CA, USA) and reverse transcribed with the iSCRIPT cDNA Synthesis Kit (Bio-Rad Laboratories). 10 ng of cDNA was used as template in each real-time quantitative PCR (qPCR), performed with fast SyberGreen Master Mix (Thermo Fisher Scientific) on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad Laboratories). Primer sequences are provided in Table S4.

RNA library preparation and Next-Generation Sequencing
Libraries for RNA-Seq were prepared for each sample from 0.5 lg of total RNA. rRNA removal, RNA-Seq library preparation, and subsequent analyses were performed as previously described [39] with three biological replicates for each experimental condition. Differentially expressed genes (DEGs) were identified using the BIOCONDUCTOR DESEQ2 package (RRID: SCR_015687) [40] as genes whose q-value is lower than 0.05. Gene set enrichment analysis (GSEA) was performed using the DESKTOP tool of the Broad Institute (https://www.gsea-msigdb.org/gsea/index.jsp; RRID: SCR_003199) [41] by querying DEGs for the enrichment of Hallmark Gene Sets from the Hallmark MSigDB collection [42] and CCC DLBCL signatures [2]. Upstream regulator analysis of DEGs was performed with the INGE-NUITY PATHWAY ANALYSIS software package (QIAGEN, Venlo, Netherlands; RRID: SCR_008653).

CRISPR-Cas9 constructs and knockout generation
Gene knockout was obtained by targeting two sites for each gene, selected using the CRISPR Targets Track function from UCSC Genome Browser (https:// genome-euro.ucsc.edu/; RRID: SCR_005780): One was selected close to, and the other ca. 100 bases downstream of the start codon. The genomic sequences targeted by the corresponding sgRNAs are provided in Table S5. Complementary DNA oligonucleotides encompassing the sequence of each sgRNA were annealed and ligated into the PX458 plasmid (Addgene; plasmid # 48138, a gift from Feng Zhang) digested with BbsI, as described [43]. For each construct, 1 µg of DNA was electroporated in 4 9 10 5 FL5.12 cells using the Neon Transfection System (Thermo Fisher Scientific). After two days, GFP positive cells were sorted with FACSMelody (BD Biosciences) and single clones isolated by limiting dilution. Following in vitro expansion, each clone was tested for recombination by extracting genomic DNA with the QuickExtract DNA Extraction Solution (Illumina) and performing PCR amplification with GoTaq polymerase (Promega) with specific primers for each genomic locus (Table S6). Ablation of the targeted protein in the selected clones was confirmed by immunoblot analysis.

Quantification and statistical analysis
Drug interaction landscapes and delta scores for synergy were based on the ZIP model in SynergyFinder [44]. All other statistical analyses were performed with PRISM 8 (GraphPad Software Inc, La Jolla, CA, USA; RRID: SCR_002798); in cell culture experiments, multiple comparisons of treated groups with the untreated control group were performed by one-way ANOVA with Dunnett's test. The numbers of independent biological replicates are indicated in the corresponding Figures. Multiple comparison between groups from in vivo tumor growth experiments were performed by one-way ANOVA with Tukey's test.

MYC-and OxPhos-associated gene expression signatures are correlated in DLBCL
To address a possible relationship between MYC and OxPhos in DLBCL, we determined the pairwise correlations between four reference signatures, comprising either MYC-regulated genes (Hallmark-MYC-V1 and Hallmark-MYC-V2) or genes related to oxidative phosphorylation (Hallmark-OxPhos and CCC-OxPhos), in six independent DLBCL datasets [3][4][5][22][23][24]. Remarkably, we found a significant positive correlation in all patient cohorts for the MYC-and OxPhos-associated signatures (Table S1), in particular between Hallmark-MYC-V1 and Hallmark-OxPhos (Fig. 1A). Most noteworthy, those two signatures also showed similar relationships with patient survival after R-CHOP treatment, albeit with inconsistent predictive value among cohorts: significant association with worse outcome in two of the cohorts, loose association in one, and none in three others (Fig. S1A,B). Thus, while variable in terms of their predictive value in R-CHOP-treated patients, the MYC-and OxPhosassociated signatures were highly correlated in all DLBCL cohorts.
We then addressed the behavior of the Hallmark and CCC-derived gene signatures during MYCinduced lymphomagenesis, based on previous RNAseq data in Eµ-myc transgenic mice [45]. Relative to control nontransgenic B cells, either pretumoral Eµmyc B-cells or late-stage lymphomas enriched not only for the MYC signatures, as expected, but also for Hallmark-OxPhos (albeit with lower significance in the lymphomas, owing most likely to their clonal heterogeneity) [46], while CCC-OxPhos showed no significant enrichment (Fig. 1B).
Altogether, the above data point to OxPhos as one of the positively regulated gene programs in MYCdriven lymphomagenesis. We previously reported that the same is true for up-regulation of the mitochondrial translation machinery [11], which is itself required for OxPhos activity [13,14,20]. We thus hypothesized that, like mitochondrial ribosomes [11,12,20], OxPhos activity might represent a tractable therapeutic target in MYC-associated lymphoma.

MYC sensitizes B cells to IACS-010759induced killing through the intrinsic apoptotic pathway
In order to address whether enhanced Myc activity may increase the sensitivity to OxPhos disruption, we transduced two murine B-cell progenitor lines (FL5.12 and Ba/F3) with retroviral vectors driving constitutive expression of a 4-hydroxytamoxifen (OHT)-dependent MycER chimera (hereafter FL MycER and BaF MycER ). At the phenotypic level, 48 h of OHT treatment had no noticeable impact on either cycle transit times or cell death (Fig. S2A,B), owing most likely to already maximal division rates in the basal state (ca. 12 h) and to the presence of survival factors in the culture medium [47].
Yet, as expected, OHT treatment led to the induction of known MYC-activated mRNAs, as well as suppression of endogenous Myc mRNA and protein (Fig. S2C,D). RNA-seq profiling in OHT-treated FL MycER cells revealed that MycER activation caused both up-and down-regulation of discrete sets of differentially expressed genes (DEGs: ca. 1200 each; Fig. S2E), as in other cell types [45]. A search for upstream regulators confirmed that MYC was the main driver of the observed transcriptional changes (Fig. S2F).
Remarkably, besides the Hallmark-MYC-V1 and -MYC-V2 gene sets, MycER-induced DEGs also enriched for Hallmark-OxPhos ( Fig. 2A). In line with this finding and with the positive effects of MYC on mitochondrial translation [11], OHT treatment induced statistically significant increases in basal respiration and mitochondrial ATP production in FL MycER cells, without significant impact on maximal respiration rate and spare respiratory capacity (Fig. 2B); these effects were specifically due to MycER activation, as they were not observed in parental FL5.12 cells (Fig. S2G). Finally, irrespective of previous OHT treatment, these mitochondrial activities were all suppressed by 24 h of treatment with the ETC complex I inhibitor IACS-010759 [16] (Fig. 2B).
We thus proceeded to assess the phenotypic consequences of IACS-010759 treatment in FL MycER and BaF MycER cells. In the absence of OHT, increasing concentrations of IACS-010759 showed no cytotoxic activity up until 48 h of treatment (Fig. 2C), but a dose-dependent cytostatic effect (Fig. 2D); instead, pretreatment with OHT markedly sensitized both cell lines to dose-dependent killing by IACS-010759. In parental FL5.12 cells, IACS-010759 was cytostatic regardless of prior OHT treatment (Fig. S3A): hence, sensitization to IACS-induced cell death was attributable to MycER activation. In order to confirm that the effect of IACS-010759 was on-target, we transduced FL MycER cells with a vector expressing the S. cerevisiae Ndi1 protein, a IACS-010759-resistant ortholog of mammalian ETC complex I that can restore electron transport in complex I-deficient cells [16,48]: indeed, as expected, Ndi1 conferred full resistance to IACS-010759 (Fig. S3B). Finally, OHT promoted killing by another complex I inhibitor, rotenone (Fig. S3C). Hence, MycER activation sensitized cells to inhibition of ETC complex I.
As seen in other cell types [16,49], the cytotoxic action of IACS was manifest with reduced amounts of glucose (2.75 mM) but not in standard high-glucose media (11 mM) (Fig. S3D). On this basis, cell killing by IACS-010759 was previously ascribed to a drop in cellular energy that could be compensated by A B Fig. 1. MYC activity positively correlates with the expression of OxPhos-associated genes. (A) Correlation between the Hallmark-OxPhos and MYC-V1 gene sets across gene expression profiles from six independent human DLBCL patient cohorts [3][4][5][22][23][24]. The X and Y axes report the mean expression of all genes in the indicated signature, in each patient sample (dots). Black lines represent linear regression fits to the data points. (B) Biological processes enriched during Eµ-myc driven lymphomagenesis. Our previous RNA-seq data [45] were used to address the enrichment of gene sets from the Hallmark collection and the CCC model [2] in pretumoral Eµ-myc B cells (left) and lymphomas (right), relative to control nontransgenic B cells. The plot shows the enriched genes signatures (FDR q-value < 0.1), ranked according to their normalized enrichment score. Only a subset of the Hallmark-associated biological pathways, but none of the CCC signatures, reached this threshold. The arrows indicate the MYC target and OxPhos gene signatures.
increasing glycolytic rates [16,17,49]. However, energy depletion was unlikely to be the cause of cell death in our system, as neither OHT nor IACS-010759 treatment impacted the ATP to ADP ratio before the onset of cell death (Fig. 2E). This conservation of the energy balance following IACS-010759 treatment might be explained by adaptive cellular responses, such as increased glycolysis or suppression of energy- consuming processes (e. g. cell growth and proliferation). Mitochondrial respiration also sustains aspartate biosynthesis, which in turn is required for anabolic reactions, including protein and nucleotide biosynthesis [50,51]. In line with this concept, supplementation with aspartate or nucleotide precursors partially bypassed the anti-proliferative effects of complex I inhibitors in other cell types [16,[50][51][52]. However, the same supplements were insufficient to prevent IACS-010759-induced arrest and cell death in FL MycER and BaF MycER cells (Fig. S3E), implying that these effects involve additional complex I-dependent processes. Cell death in FL MycER cells treated with OHT and IACS-010759 presented clear features of apoptosis, such as external membrane exposure of phosphatidylserine (Annexin V staining: Fig. 3A), chromatin condensation and nuclear fragmentation (Fig. S4A), PARP cleavage (Fig. S4B), and caspase activation-the latter suppressed by treatment with the caspase inhibitor Z-VAD-FMK (Fig. S4C). Unexpectedly, however, Z-VAD-FMK failed to rescue IACS-010759-induced killing of OHT-primed cells (Fig. 3B), implying that caspase activity in not an absolute requirement for apoptosis in this setting, as also observed in other contexts [53,54].
The apoptotic response can be activated by either the extrinsic or the intrinsic pathway, the latter mediated by permeabilization of the mitochondrial outer membrane, determined by the equilibrium between pro-and anti-apoptotic members of the BCL2 protein family [55,56]. In particular, the effectors Bax and Bak must disengage from anti-apoptotic BCL2-family proteins to form oligomeric pores, through which cytochrome c and other proteins are released from the mitochondrial intermembrane space into the cytosol [57]. Indeed, we detected cytochrome c release from the mitochondria of OHT-primed FL MycER cells undergoing IACS-010759-induced cell death (Fig. 3C,  Fig. S4D). In order to further address the role of intrinsic apoptosis in our model, we derived Bak/Baknull FL MycER cell clones through CRISPR-Cas9 engineering: Remarkably, these knockout cells acquired full resistance to IACS-010759-induced cell death (Fig. 3D), while retaining the cytostatic response (Fig. 3E). Hence, IACS-010759-mediated cytotoxicity was mediated by the intrinsic apoptotic pathway.

BCL2 acts as a suppressor of IACS-010759induced cell death
Myc is known to perturb the balance between proand anti-apoptotic BCL2-family members by promoting transcription of Bim [58,59] and Bax [60], while repressing the expression of BCL2 and Bcl-xL [61,62], altogether favoring cytochrome c release [63]. In order to address whether the increased sensitivity to IACS-010759 following MycER activation could be ascribed to alterations in balance within the BCL2 family, we monitored protein levels in FL MycER cells following OHT priming and subsequent treatment with IACS-010759 for 24 and 44 h (both before the actual onset of cell death): Indeed, MycER activation downregulated BCL2 and Bcl-xL, over-riding a slight increase in these proteins promoted by IACS-010759 alone (Fig. 3F). In contrast, neither Bim, nor Bax were significantly affected by MycER, while moderately induced by IACS-010759 (in particular at 44 h: Fig. 3F). Finally, we tested the expression of Mcl-1, an anti-apoptotic BCL2-family protein previously reported to be induced by MYC [64] and suppressed by the ETC complex I inhibitor metformin [65]; within the resolution of these experiments, we did not detect changes in its levels in response to either MycER activation or IACS-010759 treatment (Fig. 3F bottom).
Based on these results, we hypothesized that reduced expression of BCL2 and Bcl-xL upon MycER activation may cause the increased sensitivity to IACS-010759 in FL MycER cells. In support of this concept, the BCL2-specific inhibitor venetoclax sensitized the cells to IACS-010759-induced cell death irrespective of prior MycER activation (Fig. 3G), confirming the importance of endogenous BCL2 in conferring resistance to IACS-010759. Reciprocally, ectopic expression of BCL2 prevented IACS-010759-induced cell death in OHT-primed cells, a protective effect predictably overridden by venetoclax (Fig. 3H). As reported above for the combination of IACS-010759 and OHT (Fig. 3B), cell killing by IACS-010759 and venetoclax was not rescued by Z-VAD-FMK (Fig. S4E), further indicating that cell death induced by the intrinsic apoptotic pathway is not dependent on caspase activity in our model.
Previous data indicated that BCL2 can sustain OxPhos activity in leukemic stem cells [66,67]. Hence, in the above experiments, BCL2 might have acted by boosting mitochondrial respiration and/or protecting it from IACS-010759-mediated inhibition. In contrast with this scenario, however, neither overexpression, nor inhibition of BCL2 significantly impacted respiratory activity or its suppression by IACS-010759 in OHT-primed FL MycER cells (Fig. S4F,G). Finally, Bax/Bak-null FL MycER cells remained resistant to IACS-010759-induced cytotoxicity even in the presence of venetoclax (Fig. S4H), further emphasizing that BCL2 blocks IACS-010759-induced cell death by preventing activation of the intrinsic pathway.

IACS-010759 treatment activates the Integrated Stress Response
In order to characterize possible signals mediating the cytotoxic action of IACS-010759, we profiled transcriptional changes after 24 h of IACS-010759 treatment in either OHT-primed or nonprimed FL MycER cells. Independently from MycER activation, IACS-010759 elicited extensive transcriptional alterations, with over 1000 DEGs in either OHT-primed or nonprimed cells and a close correlation between the two responses (Fig. S3E). Upstream regulator analysis identified EIF2AK3 (also known as PERK) and ATF4, two key components of the Integrated Stress Response (ISR) [68], as controllers of the transcriptional response to IACS-010759 in both OHT-primed and nonprimed cells (Fig. 4A). Moreover, IACS-010759 treatment suppressed the expression of genes involved in sterol biosynthesis, which are controlled by the Sterol regulatory element-binding proteins (SREBPs) and the upstream SCAP-INSIG1 regulatory circuit [69]. Finally, IACS-010759 led to the activation of genes associated with the expression of UCP1, which mediates dissipation of the proton gradient across the mitochondrial inner membrane, thus decoupling electron transport from ATP synthesis [70]: this most likely indicates a common cellular response to impaired mitochondrial ATP production, regardless of its cause (i.e., suppression of membrane potential by UCP1, or disruption of the electron transport chain by IACS-010759; Fig. 2C). The ISR, the highest ranking IACS-010759regulated program in our analysis (Fig. 4A: EIF2AK3 and ATF4), is an adaptive pathway engaged by diverse stress stimuli that selectively activate one of four different kinases (EIF2AK1-4), which converge to phosphorylate a specific residue (S51) in the translation factor eIF2a [68]. While repressing general translation, phospho-eIF2a promotes the translation of a subset of transcripts, including the ATF4 mRNA. ATF4 and several other ISR-induced transcriptional regulators, in turn, promote a pro-survival, stress-resistance program [68]. Most relevant here, under conditions of severe, unresolved stress, the ISR can also induce proapoptotic factors such as the BH3-only BCL2-family proteins Bcl2l11/Bim, Bbc3/PUMA, Hrk, and NOXA [71][72][73][74][75].
Based on the above, we monitored the status of the ISR pathway in FL mycER cells: Treatment with IACS-010759, but not OHT, promoted eIF2a phosphorylation and accumulation of ATF4 and several of its target-gene products, known to be involved in either cell survival (Gpx1, Pck2) or death (Ddit3/CHOP, PUMA, Bim) (Figs 4B and 3F) [68]. Part of these were also significantly upregulated at the mRNA level in our RNA-seq data (i.e., Pck2 and CHOP), while others were not (Gpx1, PUMA, Bim) (Fig. S3G), possibly reflecting activation of the latter at the translational level, as described in other ISR-related contexts [76,77]. Two other known effectors of the ISR, Hrk and NOXA [74], were not expressed in FL mycER cells as judged by our RNA-seq data, precluding assessment of their possible roles in IACS-010759-induced cell death.
In order to assess the contribution of the ISR to the cytotoxic action of IACS-010759, we used CRISPR-Cas9 engineering to inactivate the gene encoding CHOP, the main mediator of the pro-apoptotic branch of the ISR [68]. Previous work suggested that CHOP acts by inducing the BH3-only proteins PUMA and Bim [71][72][73], while suppressing BCL2 expression [78]. However, similar to parental FL MycER cells, CHOPnull cells showed IACS-010759-dependent activation of Bim and PUMA (Fig. 4C). Instead, while still suppressed by MycER activation, BCL2 levels were altered in CHOP knockout cells, with increased amounts in either basal or IACS-010759-treated conditions (Fig. 4C). Consistent with the role of BCL2 in IACS-010759-induced cell death, CHOP KO cells showed increased resistance to the drug, which was overcome by co-treatment with venetoclax (Fig. 4D).

IACS-010759 and venetoclax synergize against MYC/BCL2 double-hit lymphoma
The preferential killing of Myc-overexpressing cells by IACS-010759 ( Fig. 2A) and its suppression by Bcl2 (Fig. 3G,H) were reminiscent of our previous results with Tigecycline [11,20]. We also reported that Tigecycline and venetoclax cooperated in killing MYC/BCL2 double-hit lymphoma cells, and showed synergy against DHL in a preclinical setting [20]. Hence, these observations prompted us to address the potential of combining IACS-010759 and venetoclax to treat DHL. Indeed, dosing either drug against the other in vitro on the human DHL cell lines SU-DHL-6 and DOHH-2 revealed a strong synergy in cell killing (Fig. 5A,B), further substantiated with fixed concentrations of each drug in two other DHL lines (Fig. S5A). It is noteworthy here that as in FL5.12 cells (Fig. 4A,B), IACS-010759 treatment induced an ISR in DHL cells-albeit to variable levels-as indicated by eIF2a phosphorylation and ATF4 accumulation (Fig. S5B).
In order to address the combinatorial effect of the drugs in vivo, CD1 nude mice bearing subcutaneous SU-DHL-6 tumors were treated by oral gavage with    IACS-010759 (1-15 mgÁkg À1 ) and/or venetoclax (50 mgÁkg À1 ), for a total of 10 daily doses over two weeks. While either drug alone or the combination with the lowest dose of IACS-010759 moderately slowed down tumor growth during the treatment period, combinations with higher doses of IACS-010759 showed increased efficacy and a marked delay in regrowth post-treatment (Fig. 5C), most animals showing either partial or complete tumor regression up to one week after the end of treatment (day 19; Fig. S5C). Similar results were obtained in mice transplanted with DOHH-2 cells (Fig. 5D). We then monitored the response of the DHL patient-derived xenograft (PDX) DFBL-20954-V3-mCLP, isolated from a lymphoma refractory to multiple chemotherapy regimens [30]. PDX cells were inoculated by tail vein injection and their growth monitored by whole-body luminescence in NSG mice: again, IACS-010759 or venetoclax alone showed little effect, while the combination strongly suppressed tumor growth after the end of the treatment cycles (Fig. 5E, Fig. S6). Altogether, IACS-010759 and venetoclax showed strong antitumoral activity against DHL in the preclinical setting. The above results pointed to a wider therapeutic strategy against other high-grade lymphomas. In particular, in cases showing translocation of MYC alone, tumor progression and survival may be under the control of distinct anti-apoptotic BCL2-family proteins, against which specific pharmacological inhibitors are also available [21]: the latter compounds-rather than venetoclax-may thus show synergy with IACS-010759 in these tumors. We tested this hypothesis in OCI-LY7, a MYC-translocated, BCL2-negative DLBCL cell line that is fully resistant to venetoclax [20], but expresses Mcl-1 [29]. Indeed, while IACS-010759 alone showed substantial toxicity in those cells, this was significantly reinforced by co-treatment with the Mcl-1specific inhibitor S63845, but not with venetoclax (Fig. 6A). The Burkitt's lymphoma cell line Ramos, which is also characterized by MYC translocation and expression of Mcl-1 but not BCL2, showed a similar sensitivity profile (Fig. 6B). Thus, depending upon the cellular context, combination of IACS-010759 with distinct BH3-mimetic compounds may allow maximal antitumoral activity (Fig. 7).

Discussion
Our work points to oxidative phosphorylation (OxPhos) as one of the critical MYC-activated processes in DLBCL, and as a tractable therapeutic target in highgrade, MYC-associated forms of the disease. First, MYC-and OxPhos-related gene signatures were highly correlated in six distinct DLBCL patient cohorts, and were enriched in MYC-overexpressing mouse B-cells and lymphomas. Most noteworthy here, MYC may drive OxPhos genes not only at the transcriptional, but also at the translational level in B-cells [79]. Second, ectopic MYC activity sensitized B-cells to IACS-010759, an inhibitor of ETC complex I [16]. Third, IACS-010759 showed synergy with BH3-mimetic compounds that inhibit either BCL2 or Mcl-1, allowing contextdependent killing of lymphoma cell lines (Fig. 7). Finally, IACS-010759 and the BCL2 inhibitor venetoclax effectively cooperated against MYC/BCL2 doublehit lymphoma in xenograft-based preclinical models.
It has long been known that MYC overexpression sensitizes diverse cell types to apoptosis in response to shows the drug interaction landscape and synergy score for IACS-010759 and S63845 (as in Fig. 5A).
environmental or cell-autonomous stress conditions [47,63,80], in a manner that can be counteracted by BCL2 [80,81]. In line with this concept, ectopic activation of a MycER chimera sensitized two mouse B-cell lines to killing by IACS-010759, an effect that was blocked by BCL2 and, reciprocally, exacerbated by venetoclax. Yet, unlike reported in AML cells [66,67], venetoclax did not suppress OxPhos activity in B-cells, indicating that it did not directly enhance the inhibitory action of IACS-010759 on ETC complex I. At the mechanistic level, and as described in other contexts [61,62], MycER activation suppressed expression of BCL2 and Bcl-xL. Further analysis confirmed that IACS-010759 induced cell death through the intrinsic apoptotic pathway, as evidenced by the requirement for the effectors Bax and Bak, and cytoplasmic release of cytochrome c. Altogether, as depicted in Fig. 7, MYC promoted OxPhos activity and concomitantly lowered the anti-apoptotic safeguard provided by BCL2/Bcl-xL in B-cells [82], thus sensitizing the cells to IACS-010759. Our findings are also consistent with a recent study in AML, in which venetoclax and IACS-010759 synergized to induce apoptosis in AML cells [83]. The MYC-dependent suppression of both BCL2 and Bcl-xL also suggests that the combination of IACS-010759 with the BCL2/Bcl-xL double inhibitor navitoclax might as well constitute an effective chemotherapy regimen in MYC-overexpressing tumors, if navitoclax-associated toxicity is properly managed [84,85]. We and others have shown that the cytotoxic effects of IACS-010759 in vitro are enhanced by reduced glucose in the medium (Fig. S3C) [16,49]; therefore, reducing glucose supply to the tumor might improve the efficacy of the drug in vivo. In line with this concept, intermittent fasting reduced glycemia and improved the anti-tumoral activity of a different ETC complex I inhibitor, metformin, in a xenograft model of AML [65]. Whether combining mitochondrial inhibitors, such as IACS-010759 or metformin, with dietary restriction bears clinical potential against cancer remains to be investigated.
To identify possible effectors of IACS-010759, we profiled gene expression in B-cells: this singled out the Integrated Stress Response (ISR) as an IACS-010759induced pathway, independently from MycER activation. The ISR is a composite signaling pathway that mediates protective adaptation to multiple stresses, but concurrently promotes cell death when homeostasis cannot be restored [68]: activation of this pathway in response to OxPhos inhibitors was also reported in AML, multiple myeloma and glioblastoma cells, in which it appeared to relay mainly a pro-death signal [67,86,87]. In line with these observations, elimination of CHOP, which controls the pro-apoptotic branch of the ISR [68], conferred resistance to IACS-010759 in our experiments, thus pointing to the ISR as a common effector of OxPhos inhibitors in diverse tumor types. Finally, we shall note here that MYC-induced oncogenic stress can also activate the ISR [88]: while not observed in our experimental setting, this may further contribute to the sensitivity of MYC-driven tumors to ETC inhibitors (Fig. 7).

Conclusions
While IACS-010759 directly inhibits ETC complex I, thus suppressing OxPhos activity [16], tigecycline achieves the same effect indirectly via inhibition of mitochondrial translation [11,13], providing a common denominator for their pro-apoptotic effects on MYCoverexpressing cells, as well as on DHL cells when combined with venetoclax [20]. Thus, targeting OxPhos (whether with IACS-010759, tigecycline, etc.) along with select BCL2-family members [21] may be an effective means to achieve synergy against high-grade B-cell lymphomas (Fig. 7). In support of this concept, the BCL2 inhibitor venetoclax potentiated the cytotoxic activity of IACS-010759 in MYC/BCL2 DHL cells, while the Mcl-1 inhibitor S63845 did so in MYCtranslocated, BCL2-negative lymphoma cell lines.
Altogether, our work supports a wider therapeutic concept, in which the status of MYC and BCL2family members in individual patients may guide the decision to combine OxPhos inhibitors and select BH3-mimetics against high-grade DLBCL, and possibly other refractory malignancies. targeted high-affinity inhibitor of BCL-2, in lymphoid malignancies: a phase 1 dose-escalation study of safety, pharmacokinetics, pharmacodynamics, and antitumour activity.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article.  Fig. S4. Effects of MycER activation, IACS-010759 and other pharmacogenetic interactions in B cells. Fig. S5. Effects of IACS-010759 and BH3-mimetics on human lymphoma cells. Fig. S6. Whole body imaging of PDX-engrafted mice treated with IACS-010759 and/or venetoclax. Table S1. MYC-and OxPhos-associated gene signatures are correlated in DLBCL. Table S2. Genes shared among the MYC-and OxPhos-related signatures. Table S3. Antibodies used for immunoblot analysis. Table S4. Primers for mRNA quantification by qPCR. Table S5. Genomic targets of the sgRNAs used for CRISPR-Cas9 gene knockout. Table S6. Primers for PCR analysis of CRISPR-Cas9 knockout clones.