Actinomycin D targets NPM1c-primed mitochondria to restore PML-driven senescence in AML therapy

Abstract

Acute myeloid leukemia (AML) pathogenesis often involves a mutation in the NPM1 nucleolar chaperone, but the bases for its transforming properties and overall association with favorable therapeutic responses remain incompletely understood. Here we demonstrate that an oncogenic mutant form of NPM1 (NPM1c) impairs mitochondrial function. NPM1c also hampers formation of PML nuclear bodies (NBs), which are regulators of mitochondrial fitness and key senescence effectors. Actinomycin D (ActD), an antibiotic with unambiguous clinical efficacy in relapsed/refractory NPM1c-AMLs, targets these primed mitochondria, releasing mtDNA, activating cGAS signaling and boosting ROS production. The latter restore PML NB formation to drive TP53 activation and senescence of NPM1c-AML cells. In several models, dual targeting of mitochondria by venetoclax and ActD synergized to clear AML and prolong survival through targeting of PML. Our studies reveal an unexpected role for mitochondria downstream of NPM1c and implicate a mitochondrial/ROS/PML/TP53 senescence pathway as an effector of ActD-based therapies.

Introduction

The nucleolar chaperone nucleophosmin 1 (NPM1) exerts a wide range of activities, from ribosome biogenesis to control of MYC or TP53 signaling. In 50% of acute myeloid leukemia (AML) with normal karyotypes, highly clustered mono-allelic NPM1 mutations cause frameshifts that delete tryptophan residues involved in nucleolar targeting and create de novo nuclear export signals (1). While occurrence of NPM1c (for cytoplasmic NPM1) oncoprotein is preceded by mutations sustaining clonal hematopoiesis (2), its continuous presence is required for maintaining the leukemic state of established AMLs, where NPM1c disables TP53 signaling and sustains high expression of Hox genes (3).

Promyelocytic leukemia (PML) nuclear bodies (NBs) are reactive oxygen species (ROS)-responsive domains (4,5) that exert pro-senescent and tumor suppressive functions by enhancing TP53 and Rb activities. Many protein degradation pathways associated with tumor progression target PML, presumably to impede its pro-senescent function. PML NBs have directly been implicated in the therapeutic response of acute promyelocytic leukemia (APL) where the PML/RARA fusion oncoprotein disrupts PML bodies, but therapy-induced PML/RARA degradation restores them to drive cure (6–8). PML has also been implicated in tumor robustness through the control of mitochondrial fitness and stem cell metabolism (9–11).

Mitochondrial dysfunctions can induce the integrated stress response and ROS production, important features of normal or AML stem cells (12). Conversely, mitochondrial fitness and high OXPHOS metabolism have been implicated in resistance to BCL2 inhibitors or conventional chemotherapy (13–18). Accordingly, inhibitors of mitochondrial respiration were proposed to constitute a novel type of cancer drugs (19). In addition to their well-studied effects on nuclear DNA replication or transcription, anti-cancer antibiotics may exert ill-understood mitochondrial toxicities, implicated in long-term doxorubicin-induced cardiomyocytes loss (20).

Here, we demonstrate that NPM1c blunts PML NB-biogenesis and weakens mitochondrial fitness. Actinomycin D (ActD), a drug with established clinical activity in therapy-resistant NPM1c-AML (21–23), induces acute mitochondrial stress, ROS production, restoration of PML NBs, and senescence, contributing to its therapeutic efficacy.

Results

NPM1c targets PML and promotes cell growth

Primary leukemic cells or cell lines from AML patients expressing NPM1c exhibit some defects in NB formation, resembling APL micro-speckles (24,25) (Supplementary Fig. S1A,B). As AMLs harbour multiple gene mutations that may contribute to NBs alteration, we explored non-leukemic primary HSC from a Flp-inducible humanized Npm1c (variant A) knock-in mouse model (Npm1 ^frt-cA/+, R26 ^FlpoER) (26). Four weeks after tamoxifen exposure, these HSCs exhibited a dramatic reduction of NBs numbers (Fig. 1A). Similarly, hematopoietic progenitors differentiated from a mouse embryonic stem cell (mESC) NPM1c knock-in model (Supplementary Fig. S1C), displayed significantly fewer PML NBs.

Fig. 1. NPM1c alters PML NBs biogenesis.

A, Quantification (left) and representative image (right) of PML NBs in HSC purified from NPM1c-expressing knock-in mice. The results are expressed as the mean value ± SD (error bars) of n = 5 NPM1^+/+ and 12 NPM1^cA/+ mice. Unpaired t test ***p < 0.001. Scale bar, 3 μm. B, Quantification (left) and representative image (right) of PML NBs in MEFs stably expressing GFP-PML-III and transiently transfected with NPM1-derived expression vectors. Results are expressed as the mean value ± SD (error bars) of n = 8 analyzed cells per condition. Unpaired t test ***p < 0.001. Scale bar, 10 μm. C and D, Coimmunoprecipitation analyses of 293T cells transfected with indicated constructs. E, Histidine pull-down of disulfide-linked NPM1c-PML complex from transiently transfected 293T cells. *: non-specific binding to NPM1. F, Quantification (left) and representative image (right) of PML-NBs in isogenic AML2 derivatives. Results are expressed as the mean value ± SD (error bars) of n = 18 analyzed cells per condition. Unpaired t test ***p < 0.001. Scale bar, 10 μm. G, Colony-forming ability of isogenic AML2 derivatives. Unpaired t test. n = 3. H, GSEA analyses of E2F and MYC pathways in transcriptomic exploration of NPM1c vs NPM1 expressing AML2 cells. I, Western blot analyses of TP53, ARF, PML and endogenous NPM1 in isogenic AML2 cells. J, Western blot analyses of PML in primary bone marrow samples from 9 NPM1c and 7 NPM1 AML samples explored.

A previous study suggested that overexpressed PML and NPM1c may interact (27). Transient expression of NPM1c (but not NPM1) disrupted PML NBs in mouse embryonic fibroblasts (MEFs) stably expressing GFP-PML-III (Fig. 1B). Cysteine 288 in the de novo C-terminal sequence of NPM1c was implicated in nucleolar export and oxidative stress control (28). Mutation of this residue in NPM1c abolished its ability to oppose NBs formation (Fig. 1B). Conversely, an exquisitely redox-sensitive cysteine residue of PML (C389) (29) was required for NPM1c to target NBs. Immunoprecipitation experiments demonstrated that ectopically expressed NPM1c interacts with PML, a process that requires NPM1c C288 and PML C389, but not a PML cysteine residue implicated in arsenic binding (4) (Fig. 1C and 1D). We determined whether a disulfide bond might govern interaction of NPM1c with PML. For this, we transiently transfected PML and NPM1 variants and purified His-PML proteins under denaturing conditions, followed by NPM1 Western blot analyses, in the absence or presence of the protein-reducing agent TCEP. High molecular weight complexes reactive with both PML and NPM1c antibodies were detected only when NPM1c C288 and PML C389 were both present and TCEP was omitted (Fig. 1E and Supplementary Fig. S1D). Thus, two hyper-reactive cysteines in NPM1c and PML interact, yielding impairment of PML-NB assembly. However, while NPM1c binding is required for PML NB disruption, cytoplasmic localization of NPM1c did not require PML, as observed in NPM1c knock-in Pml ^-/- mESCs (Supplementary Fig. S1E).

To explore functional consequences of NPM1c/PML interactions, we generated an isogenic system by stably expressing GFP-NPM1c, GFP-NPM1c^C288S or GFP-NPM1 fusions in OCI-AML2 leukemic cells (thereafter named AML2) in which NPM1 gene is wild-type. Expression of NPM1c blunted NB formation, whereas NPM1c^C288S failed to do so (Fig. 1F). Functionally, expression of NPM1c, but not NPM1c^C288S, increased clonogenic activity in methylcellulose cultures (Fig. 1G) and activated E2F and MYC target genes (Fig. 1H, Supplementary Table 1), possibly reflecting decreased basal level of TP53 and ARF (Fig. 1I). NPM1c expression also diminished PML protein levels (but not PML mRNA, Supplementary Fig. 1F). Low PML protein expression was confirmed in primary AML patient samples (Fig. 1J) and could independently amplify NPM1c-driven defects of PML-NBs formation. NPM1c also reduced expression of WT NPM1 (Fig. 1I), which, in NPM1c AML blasts, may further aggravate NPM1 haploinsufficiency (30).

NPM1c impacts mitochondria and drives stress response

Since NPM1c impairs NB-biogenesis and PML NBs control mitochondrial fitness (9,10), we assessed the impact of NPM1c expression on mitochondrial status. In tamoxifen-treated NPM1c knock-in mice (26), we found an increase in mitochondrion number (but not volume) in phenotypic HSC, but not in Lin-negative cells (Fig. 2A and Supplementary Fig. S2A). Similarly, in NPM1c-expressing isogenic AML2 cells, number of mitochondria increased, while branching pattern decreased (Fig. 2B) and cristae were lost, as determined by transmission electron microscopy (Fig. 2C). Transcriptomic and proteomic analyses revealed dysregulation of mitochondria-related pathways (Supplementary Fig. S2B). These mitochondrial defects were associated with enhanced production of ROS, mitochondrial superoxides and membrane potential, with no alteration of mitochondrial mass (Fig. 2D and Supplementary Fig. S2C). Mitochondrial impairment was also substantiated by leakage of mitochondrial DNA into the cytoplasm in NPM1c-expressing AML2 or mESCs (Fig. 2E and Supplementary Fig. S2D) and reduced electron transport chain (ETC) proteins, notably complex II, reflecting transcriptional downregulation of SDH genes (Supplementary Fig. S2E-G).

Fig. 2. NPM1c impacts mitochondria and drives an integrated stress response.

A, Number and volume of mitochondria in hematopoietic stem cell (HSC) purified from NPM1c-knockin mice. The results are expressed as the mean value ± SD (error bars) of n = 8 NPM1^+/+ and 11 NPM1^cA/+ HSC. Unpaired t test **p < 0.005. Scale bar, 3 μm. B, Examination of mitochondrial morphology in NPM1c and NPM1c^C288S expressing AML2 cells. Top: immuno-fluorescence analyses of mitochondrial morphology. Bottom: Mitochondrial fragmentation was quantified by number of mitochondria, of junctions and branches length. The results are expressed as the mean value ± SD (error bars) of n = 25 cells per condition. Unpaired t test ***p < 0.001. Scale bar, 10 μm. C, Transmission electron microscopy image of mitochondrial cristae in AML2-derived isogenic cells. Scale bar, 1 μm. D, FACS analyses of mitochondrial status in AML2-derived isogenic lines. The results from are expressed as the mean value ± SD of 3 independent experiments. Unpaired t test. ***p < 0.001. **p < 0.01. E, Quantification of cytosolic and mitochondrial mtDNA. The results are expressed as the mean value of triplicate samples ± SD. Unpaired t test. ***p < 0.001. n=3. F, Immuno-fluorescence analyses of cGAS localization in NPM1c cells. G, Basal 2’3-cGAMP concentration in AML2 cells expressing NPM1c or not. The results are expressed as the mean value of triplicate samples ± SD. Unpaired t test. ***p < 0.001. n=2. H, Gene set enrichment analysis of differentially expressed genes in AML2 cells expressing NPM1c or not. I, Effect of NPM1c expression on nascent protein synthesis. The results are expressed as the mean value of triplicate samples ± SD. Unpaired t test. ***p < 0.001. n=3. J, ATP/ADP ratio in AML2-derived isogenic lines. The results are expressed as the mean value of triplicate samples ± SD. Unpaired t test. *p < 0.05. K, Western blot analyses of TFAM and PGC1α in AML2-derived isogenic cells. L, Acetylation of PGC1α in isogenic AML2 cells. Immuno-precipitation of lysine-acetylated proteins followed by Western blot with an anti-PGC1α antibody. The estimated fraction of acetylated PGC1α is indicated.

Release of mtDNA into the cytoplasm was associated with the ignition of intracellular stress pathways, particularly cGAS (cyclic GMP-AMP synthase) activation, and cGAMP production, resulting in the downstream transcriptional activation of interferon (IFN) and NFκB pathways (Fig. 2F-H) (31,32). Other stress pathways, such as oxidative stress and unfolded protein response, were also activated (Fig. 2H and Supplementary Fig. S2B, Supplementary Tables 2,3), explaining the decrease in global protein synthesis (Fig. 2I). Western blot analyses validated activation of these stress pathways in AML2-NPM1c cells, including the up-regulation of key TP53-repressed metabolic enzymes (Supplementary Fig. S2H). As expected, these mitochondrial alterations decreased the ATP/ADP ratio, reflecting metabolic stress (Fig. 2J). Functionally, activation of these stress pathways could favor AML cell fitness (12). Indeed, inhibition of cGAS activity by G140 selectively suppressed growth of AML2-NPM1c cells (Supplementary Fig. S2I).

Mechanistically, two key regulators of mitochondrial protein expression, PGC1A (peroxisome proliferator-activated receptor gamma coactivator 1-alpha) and TFAM (transcription factor A, mitochondrial), showed decreased protein abundance in AML2-NPM1c (Fig. 2K) and OCI-AML3 cells (hereafter referred as AML3, which are NPM1c-mutant) when compared to AML2 cells (Supplementary Fig. S2J). TFAM reduction may reflect transcriptional down-regulation of PGC1A by IFN signaling (33). Importantly, stable NPM1c expression in AML2 cells drove PGC1α hyperacetylation resulting in its functional inactivation, a known consequence of PML NB downregulation (Fig. 2L) (9,10). In principle, NPM1c could alter mitochondrial functions independently from its ability to interfere with PML NB formation. Exploring mitochondrial status in NPM1c-knockins Pml ^+/+ or PML ^-/- mESCs, revealed that mitochondrial morphology was so profoundly altered in Pml ^-/- mESCs, that it precluded detection further of NPM1c-driven abnormalities (Supplementary Fig. S2K).

Actinomycin D targets mitochondria and promotes superoxide response

Recent clinical studies have shown that the rDNA transcriptional inhibitor ActD (Dactinomycin®) can induce complete remissions and even cure some refractory/relapsed NPM1c-AML patients as a single agent (21–23). Transcriptional inhibition of rDNA genes by ActD induces nucleolar segregation and activates TP53 through ribosomal checkpoint activation upon MDM2-L5/L11 complex formation (34). When used at low doses (1-10 nM), ActD triggers a transient cell-cycle arrest in the absence of TP53 stabilization, DNA damage or apoptosis (35). We found that 5 nM ActD (a concentration in the range of those found in patients (23)) triggered TP53 activation in AML3, but not in AML2 cells (Supplementary Fig. S3A). Nonetheless, we only observed a small increase in the L11/HDM2 interaction in patient in vivo (see below), suggesting that ribosomal checkpoint activation may not solely account for TP53 activation (Supplementary Fig. S3B). Unexpectedly, we found that ActD treatment rapidly fragmented the mitochondrial network in AML3 or NPM1c-mESC, with no or minor effects in AML2 or mESC (Fig. 3A and Supplementary Fig. S3C,D). Moreover, ActD initiated rapid production of ROS and induction of a superoxide transcriptional response, loss of mitochondrial membrane potential and loss of mtDNA through its leakage to the cytoplasm, all preferentially in AML3 and/or AML2-NPM1c-cells (Fig. 3B-3F), with little immediate impact on mitochondrial mRNA expression (Supplementary Fig. S3E).

Fig. 3. Actinomycin D targets mitochondria.

A, Top: representative immuno-fluorescence analyses of mitochondrial morphology in AML3 cells treated with 5 nM ActD. Bottom: Mitochondrial fragmentation was quantified by number of mitochondria, of junction and branch length. The results are expressed as the mean value ± SD (error bars) of n = 25 cells per condition. Unpaired t test ***p < 0.001. Scale bar, 2 μm. B, Heatmap (induction over mock-treated in the pool of samples) showing the expression of superoxide response genes upon ActD exposure. C and D, FACS analyses of mitochondrial membrane potential (C), ROS and mitochondrial superoxide production (D), after ActD exposure in AML2-derived isogenic cells. The results are expressed as the mean value ± SD. Unpaired t test. **p < 0.005, ***p < 0.001. n=3. E, effect of ActD on mtDNA abundance in the cytoplasm (left) or within mitochondria (right). Representative experiment of n=2. F, Southern blot analysis of the different forms of mitochondrial DNA, in AML2 or AML3 cells treated or not with ActD (5 nM) for 1 h. G, levels of cytoplasmic cGAMP before or after ActD treatment. Representative experiment of n=2. H, Western blot analyses of eIF2α and its phosphorylated form after ActD exposure. I, ATP/ADP ratio from metabolomic analyses of AML2 cell lines after ActD exposure. The results are expressed as the mean value of triplicate samples ± SD. Unpaired t test. ***p < 0.001. J, Western blot analyses of AMPK and its phosphorylated form after ActD exposure.

Actinomycin D-triggered mitochondrial toxicity drove stress responses, as indicated by increased cGAMP production, phosphorylation of eIF2α (Fig. 3G,H), transcriptional stress signatures (UPR, ROS signaling), activation of innate immunity (IFN, TNFA and TGFB) (36) and apoptosis/senescence (TP53) (Supplementary Fig. S3F). While responses were often shared between the two isogenic cell-lines, IFN and TP53 activation were particularly pronounced in AML2-NPM1c cells. We found a remarkable similarity between the basal expression of genes induced by NPM1c and those activated by ActD in AML2 cells (Supplementary Fig. S3G), likely reflecting their shared ability to impair mitochondrial function. Consequently, ActD-triggered induction of stress pathways may be functionally more significant in NPM1c-expressing cells where basal activation levels are already high. Finally, ActD exacerbated pre-existing NPM1c-driven metabolic stress, as revealed by decreased ATP levels and increased AMPK (AMP-activated protein kinase) phosphorylation (Fig. 3I,J). Collectively, these observations suggest that ActD targets mitochondria, particularly those primed by NPM1c-expression.

Actinomycin D activates a ROS/PML/TP53 senescence axis in NPM1c-expressing cells ex vivo

We then explored the cellular impacts of mitochondrial poisoning by ActD. Unexpectedly, ActD rapidly restored PML NBs in NPM1c-transfected MEFs, AML2-NPM1c or AML3 cells (Fig. 4A and Supplementary Fig. S4A). PML NB restoration was also observed in primary NPM1c-AML blasts treated ex vivo with ActD (Fig. 4B). In transiently transfected cells, NB restoration by ActD was associated with (and likely caused by) disruption of NPM1c-PML complexes (Fig. 4C). Similarly, in AML3 cells, ActD (as TCEP) restored the normal size of high molecular weight NPM1 or PML species (Supplementary Fig. S4B). Apart from disrupting NPM1c/PML adducts, ActD-induced ROS may also directly enforce PML biogenesis (4,5). Accordingly, NB-restoration in AML2-NPM1c cells was completely blocked by the ROS scavengers N-acetyl cysteine and glutathione (Supplementary Fig. S4C).

Fig. 4. Actinomycin D activates a ROS/PML/TP53 senescence axis ex vivo.

A, Quantification (left) and representative image (right) of PML-NBs upon ActD exposure in MEF stably expressing GFP-PML-III and transiently expressing NPM1 (WT) or NPM1c (C+). The results are expressed as the mean value ± SD (error bars) of n = 10 cells per condition. Unpaired t test ***p < 0.001. Scale bar, 10 μm. B, Ex vivo ActD treatment of primary AML blasts demonstrates PML NB restoration in NPM1c-AMLs, but not NPM1 WT ones. Representative images (left) and quantification (right). C, Histidine pulldown of disulfide-linked NPM1c-PML complex in transiently transfected 293T cells, before or after ActD exposure. D, Western blot analyses of AML3 and AML3 ^PML-/- cells after ActD treatment. E, SA-β-gal staining of AML3 derivatives 7 days after 2 hours ActD pre-treatment. F, Effect of ActD pre-treatment on colony formation in the indicated AML cell lines. The results are expressed as the mean value of triplicate samples ± SD, Representative experiment of n=2. Unpaired t test. ***p < 0.001. G, Western blot analyses of NAC pre-treated AML cells upon ActD exposure. H, Effect of NAC and ActD pre-treatment on colony formation in the indicated AML cell lines. The results are expressed as the mean value of triplicate samples ± SD. Unpaired t test. ***p < 0.001. Representative experiment of n=2. I, Western blot analyses of ActD response of parental and mitochondria-depleted AML2 and AML3 cells. MT-CO1, Mitochondrial cytochrome c oxidase subunit 1. RXR, Retinoid X receptor alpha.

PML NBs are tightly linked to senescence induction. In AML3 cells, ActD-driven PML NB reformation was associated to multiple signs of senescence, including TP53 activation, increase in p21 or Serpine-1 expression, activation of SA-β-Gal activity and loss of clonogenic potential, (Fig. 4D-F, Supplementary Fig. S4D). Importantly, these features were abolished or attenuated by PML or TP53 deletion. Moreover, blocking PML NB reformation using antioxidants blunted ActD-driven TP53 and P21 activation, as well as loss of clonogenic activity (Fig. 4G,H). Collectively, these experiments establish a central role of a ROS/PML/TP53 cascade in driving senescence and ActD response ex vivo.

To substantiate the mitochondrial origin of ActD-induced ROS in driving TP53 activation, we cultured AML cells with low-dose antimycin A (Fig. 4I) to promote progressive mitochondrial clearance through mitophagy (37). Mitochondrial depletion abolished early ActD-triggered TP53 activation (Fig. 4I), suggestive for a key role of mitochondrial ROS in PML-dependent TP53 stabilization. Conversely, we used Thenoyl-trifluoroacétone (TTFA) a well-characterized inhibitor of mitochondrial complex II and major inducer of ROS (Supplementary Fig. S4E). In AML3 cells, TTFA decreased clonogenic activity in a PML-dependent manner (Supplementary Fig. S4F). In stable GFP-PML-III MEFs transiently expressing NPM1c, TTFA restored NB formation (Supplementary Fig. S4G). Collectively, these experiments establish the key role of mitochondrial ROS in ActD-driven activation of the PML NB-activated senescence checkpoint ex vivo.

ActD exerts leukaemia-specific growth suppression in vivo

We first assessed ActD response in primary cells. We found no significant ability of low-dose ActD to promote TP53 stabilization in normal bone marrow progenitors in vivo or ex vivo, or in human or mouse primary fibroblasts (Supplementary Fig. S5A)(35). In immuno-deficient mice xenografted with primary human NPM1c-AML blasts, a 4-day course of ActD therapy significantly stabilized human TP53, but not mouse Trp53, highlighting tumor-specific targeting and implying that NPM1c sensitizes cells to ActD (Fig. 5A, top). Actinomycin D therapy inhibited leukemic growth and induced blast differentiation in that model (Fig. 5A, bottom). Similarly, in a murine NPM1c-driven AML model (38), ActD treatment again induced leukemia-selective Trp53 activation, AML regression and blast differentiation (Fig. 5B). Finally, in AML3 xenografts, PML was required for TP53 activation and anti-leukemic effects of ActD (Fig. 5C). In contrast, ActD failed to cause regression of MLL/ENL-driven murine AMLs after 7 days (Supplementary Fig. S5B). Collectively, these in vivo observations suggest that ActD favors PML-dependent TP53 activation and regression of NPM1c-AMLs, but (at least initially) spares normal cells.

Fig. 5. ActD exerts AML-specific growth suppression in vivo.

A, Top: TP53 or HDM2 stabilization in bone marrow of immuno-deficient mice xenografted with primary patient NPM1c-AML cells. +: ex vivo MLL/ENL ActD treated cells for 3h, as a positive control. Bottom, AML features in this mice model following treated or not with ActD for 5 days. Unpaired t test. *p < 0.05, **p < 0.005, ***p < 0.001. Representative experiment of n=3. B, Top, Trp53 stabilization in CD45.2⁺ NPM1c-driven murine AML blasts vs CD45.2^- normal mouse cells upon ActD exposure. Bottom, AML features after a week of ActD therapy. The results are expressed as the mean value ± SD. Unpaired t test. *p < 0.05, **p < 0.005, ***p < 0.001. Representative experiment of n=3. C, Western blot analyses of AML3 and AML3 ^PML-/- engrafted immuno-deficient mice treated or untreated with ActD for a week (left) and percentage of CD45+ AML3 blasts in the bone marrow (right). Unpaired t test. ***p < 0.001. n=2. D, Top: Immuno-fluorescence analyses of blast-rich blood samples from an AML patient during ActD therapy. Bottom: quantification of PML-NBs. Scale bar, 10 μm. E, Heat map of upregulated TP53 target genes in AML cells from ActD-treated patient. F, Western blot analyses of AML-rich blood samples during the first cycle of ActD therapy. G, Gene set enrichment analyses of differentially expressed genes in the patient treated with ActD. H, Expression of NFΚB targets (left), acute stress response (middle) and proliferation genes (right) in AML-rich peripheral blood from patient during therapy (data from the micro-array experiments).

We then explored a NPM1c-AML patient who exhibited a 10⁴-fold blast decrease 3 weeks after a 5-day course of single-agent ActD treatment, using the previously reported therapeutic scheme (21–23) (Supplementary Fig. S5C). In leukemic blasts sequentially sampled from the peripheral blood of this patient during ActD treatment, the reformation of PML NBs was complete by 12 hours (Fig. 5D). Such ActD-induced NB reformation was accompanied by TP53 stabilization and target gene activation, including P21 (Fig. 5E,F). However, we failed to detect a significant increase in the L11/HDM2 interaction (Supplementary Fig. S3B) or γH2AX foci before 48 hours (arrow in Supplementary Fig. S5D), arguing against potent early activation of ribosomal and DNA-damage checkpoints in vivo. Thus, PML NB-reformation is an immediate response of AML cells to ActD therapy in vivo.

Pathway analyses of transcriptomes from AML-rich peripheral blood revealed activation of acute stress responses (expression of HSP1A, FOS, EGR1) and activation of PML NBs as early as 6 hours following initiation of therapy (Fig. 5G,H). We also observed immediate shutoff of NFκB targets, including extinction of IL8 expression, a distinct feature of mitochondrial dysfunction-driven senescence (37)(Fig. 5H). We detected cGAS aggregation 6 hours after ActD treatment (Supplementary Fig. S5D). Despite daily ActD injections, these transcriptional changes were transient, pointing to the existence of adaptive control mechanisms (Fig. 5G). Nevertheless, ActD treatment triggered features of senescence (PML NBs, TP53 and P21 up, E2F down)(Fig. 5E-H). TUNEL-positive apoptotic cells were also detected 48 hours post-treatment (data not shown) indicating the co-occurrence of apoptosis.

A consolidation course of ActD treatment given when the patient had no residual leukemic blasts was not accompanied by major TP53 target activation or modulation of stress responses in normal blood cells (Fig. 5E and Supplementary Fig. S5E). Collectively, our findings support the idea that ActD-driven acute mitochondrial stress initiates PML-NB-activated senescence/apoptosis in NPM1c AML cells in vivo.

ActD potentiates Venetoclax anti-leukemic effects

In AML and myeloma cells, low basal complex II activity predicts Venetoclax sensitivity (15,39). Having demonstrated NPM1c-driven complex II impairment (Supplementary Fig. S2E-G) and ActD-mediated mitochondrial toxicity (Fig. 3), we hypothesized that ActD might enhance the established anti-AML clinical activity of Venetoclax (40,41). Venetoclax induced mitochondrial fragmentation, reduction of mitochondrial membrane potential, production of ROS and leakage of mtDNA to the cytoplasm of AML3 cells (Fig. 6A-E). Critically, all of these features were strongly potentiated by co-treatment with ActD. In AML2-NPM1c (but not AML2-NPM1), Venetoclax and ActD strongly synergized to activate cGAS activity (Fig. 6F) and PML NB reformation (Supplementary Fig. S6A). A brief pre-treatment with both drugs abolished clonogenic growth of NPM1c-expressing cells, only in the presence of PML (Fig. 6G). Accordingly, in AML3 xenograft models, Venetoclax and ActD strongly synergized to extend survival, an effect that also requires PML (Fig. 6H).

Fig. 6. ActD enhances Venetoclax anti-leukaemic effects.

A, Mitochondrial morphology in AML3 cells treated with 5 nM ActD and/or 1 μM Venetoclax for 3 hours. Left: mitochondrial fragmentation was quantified by number of mitochondria, of junction and branch length. Right: immuno-fluorescence analysis of mitochondrial morphology. The results are expressed as the mean value ± SD of n = 30 cells per condition. Unpaired t test. *p < 0.05, **p < 0.005, ***p < 0.001. Scale bar, 5 μm. B-D, FACS analyses of general ROS (B), mitochondrial membrane potential (C), and mitochondrial superoxide production (D), after ActD and/or Venetoclax exposure in AML3 cells. Cell were treated with indicated drugs for 3 hours in (B) and (D) and 6 hours in (C). The results are expressed as the mean value ± SD from from n=2 experiment. Unpaired t test. ***p < 0.001. E, Leakage of mitochondrial DNA to the cytoplasm in response to a 3 hours ActD or Venetoclax exposure in AML3 cells. Results are expressed as the mean value of triplicate samples ± SD. Unpaired t test. ***p < 0.001. Representative experiment of n=2. F, 2’3-cGAMP concentration in AML cells after 6 hours exposure to Venetoclax and/or ActD. Representative experiment of n=2. G, Methylcellulose colony formation assays of AML cell lines upon 2 hours pre-seeding exposure to Venetoclax and/or ActD. Results are expressed as the mean value of triplicate samples ± SD. Unpaired t test. ***p < 0.001. Representative experiment of n=3. H, Survival of immuno-deficient mice xenografted with AML3 or AML3 ^PML-/- cells and treated with ActD or Venetoclax, 5 days per week. I, AML features in NPM1c-driven AML mice treated with ActD and or not Venetoclax for 5 days. The results are expressed as the mean value ± SD. Unpaired t test. *p < 0.05, ***p < 0.001. n=2. J, Response to ActD and/or Venetoclax in a transplantable AML model initiated by NPM1c + IDH1^R132H mutation. Left, GFP abundance in bone marrow samples collected at treatment interruption (arrow in the survival curve, right), Representative experiment of n=3. K, Abundance of human AML cells in xenografted mice treated for two weeks. Combined therapy induces AML regression. L, Model of Venetoclax/ ActD/ mitochondria/PML/TP53 interplays in NPM1c-AMLs. Dashed lines point to likely mechanisms not directly explored here.

We then explored the ActD/Venetoclax interaction in three other NPM1c-driven AML models in vivo. First, in murine NPM1c-driven AMLs (38) the combination treatment dramatically decreased leukemic burden and enhanced blast differentiation (Fig. 6I). Second, in a double conditional knock-in of NPM1c plus IDH1^R132H ((42), to be fully described elsewhere), only the combined treatment eliminated AML cells from the bone marrow and yielded a significant survival advantage (Fig. 6J). Finally, in xenografted primary AML blasts, ActD and Venetoclax (but neither agent alone) cleared AML cells from the bone marrow (Fig. 6K). Collectively, the ActD/Venetoclax combination synergizes for mitochondrial targeting, yields PML-dependent growth arrest and clears leukemia in several distinct in vivo AML models.

Discussion

NPM1c expression alters PML NB formation in multiple cell types, including in normal HSCs in vivo. NPM1c is an abundant protein which may impede NB-biogenesis through direct PML binding and sequestration, degradation or indirect control of PML oxidation, which drives NB formation (4,43). PML is critical for mitochondrial fitness (9–11,44). NPM1c-induced mitochondrial defects reflect, at least in part, NB disruption, since NPM1c expression in Pml^-/- mESCs did not further aggravate the prominent mitochondrial morphological alterations resulting from Pml loss (Supplementary Fig. S2K). Functionally, NPM1c-mediated NB impairment and mitochondrial defects, which drives the integrated stress response, may all contribute to the transformation of DNMT3A or IDH1/2 immortalized stem cells into full-blown leukemia (2,12,26), through cell autonomous mechanisms (including ROS signaling and impaired senescence) and/or inflammation-mediated remodeling of the micro-environment (45,46).

Targeting mitochondrial function by antibiotics was proposed as a therapeutic option for cancer (19). Low-dose ActD can induce reversible cell-cycle arrest, without DNA-damage or apoptosis (35) and accordingly was well-tolerated by non-leukemic primary cells ex vivo or in vivo (Fig. 5). In contrast, ActD triggers rapid, but transient, alterations in NPM1c-primed mitochondria (Fig. 3A, Supplementary Fig. S3D). ActD also induces mtDNA leakage to the cytoplasm (Fig. 3E) possibly caused by ActD intercalation within the mitochondrial genome. Such mtDNA release, followed by activation of cGAS, and its downstream IFN pathway could activate senescence, as well as anti-leukemic immune responses (31,36,47,48). Besides cGAS activation, ActD induces ROS which are required for ActD-induced senescence ex vivo through PML NB reformation and TP53 activation (Fig. 4, Fig. 6L) (4,5,43). The actual origin of these ROS remains poorly defined. Upon mitochondrial stress, complex II, at the junction between the TCA cycle and the ETC, which is highly expressed in LSK progenitors (49), regulated by PML (50) and by NPM1c (Supplementary Fig. S2F,G), could play a role in ROS production, as the complex II inhibitor TTFA mimics ActD in NPM1c-expressing cells ex vivo (Supplementary Fig. S4E-G). Our ex vivo observations suggest that a significant part of ActD activity is mediated through mitochondrial ROS, PML NB-reformation and TP53 activation. In the ActD-treated AML patient, we also found some evidence for mitochondria dysfunction-induced senescence (37). Later, ActD intercalation into nuclear DNA, inhibition of RNA Pol I activity and activation of the ribosomal checkpoint (34), could also contribute to ActD therapeutic activity. NPM1c-driven PML NB disruption and their ActD-induced reformation (Fig. 5D, 6K) evoke a striking, but unexpected, similarity with APL treatment by retinoic acid and arsenic (6). PML NBs may thus emerge as key hubs inactivated by oncogenes (NPM1c, PML/RARA), but re-activated by anti-tumor drugs (retinoic acid, arsenic, ActD or interferon alpha (51)). Actually, in an APL mouse model, an ETC inhibitor drove prolonged remissions (52), possibly reflecting mitochondrial ROS-induced PML NB restoration. This ROS/PML/TP53 cascade may also promote ActD responses in non-NPM1c-AMLs harboring dysfunctional mitochondria. Doxorubicin similarly fragments mitochondria of NPM1c-expressing cells to drive ROS-induced TP53 activation (Supplementary Fig. S6B-D), suggesting that dual targeting of nucleus and mitochondria may be shared by other anti-tumor antibiotics.

High mitochondrial ETC or OXPHOS activities were repeatedly linked to Venetoclax- or chemotherapy-resistance in AML (13–15,17) or other cancers (18). Conversely, inhibition of mitochondrial translation sensitizes AML cells to Venetoclax (16). Thus, the NPM1c-associated defects in mitochondrial function unraveled here likely underlie the favorable prognosis of NPM1c-AMLs. Finally, ActD and Venetoclax exert synergistic effects to clear NPM1c-AMLs in multiple in vivo models (Fig. 6). Clinically supporting our observations, ActD was very recently proposed to circumvent Venetoclax resistance in salvage therapy of multi-relapsed AML patients (53). Thus, mechanism-based prospective trials associating ActD and Venetoclax could be envisioned in AML patients.

Methods

Cell Lines

Human AML cell lines were maintained in minimum essential medium-a plus GlutaMAX (Thermo Fisher Scientific; Cat # 32561094) supplemented with 20% FBS in the presence of 100 U/mL of penicillin, 100 μg/mL of streptomycin, and were incubated at 37°C with 5% CO₂ (see key resource table for all details). To establish isogenic NPM1c expressing AML2 cells, cells were transduced with a lentivirus expressing Flag-GFP-tagged NPM1, NPM1c or NPM1c-C288S and selected for GFP expression by FACS. WI38 were maintained in Eagle’s Minimum Essential Medium (ATCC; Cat # 30-2003) supplemented with 10% FBS in the presence of 100 U/mL of penicillin, 100 μg/mL of streptomycin. Primary mouse embryonic fibroblasts were obtained from E18 mouse embryos and, maintained in Dulbecco’s Modified Eagle Medium plus GlutaMAX (Thermo Fisher Scientific; Cat # 41965062) supplemented with 10% FBS, 1 × non-essential amino acids solution, 100 U/mL of penicillin and 100 μg/mL of streptomycin. Normal CD34-positive cells were purified by using Dynabeads CD34 positive isolation kit (Thermo Fisher Scientific; Cat # 11301D). Mouse embryonic stem cell line (mESC) E14 was maintained on gelatin-coated dishes and cultured in Dulbecco’s Modified Eagle’s Medium supplemented with 10% fetal bovine serum, 1000 U/ml LIF, 50 U/ml penicillin, 50 μg/ml streptomycin, 2 mM GlutaMAX, 0.1 mM non-essential amino acids, 1 mM sodium pyruvate, and 0.1 mM 2-mercaptoethanol. The medium was changed every other day. All cell lines were repeatedly shown to be Mycoplasma negative by PCR testing (Eurofins). OCI-AML2 and OCI-AML3 were obtained from DSMZ, regularly authenticated by NPM1c expression and genetically modified in house. Details on PML- and TP53- inactivation by CRISPR in AML3 cells or NPM1c CRISPR-in mESC can be found in the Supplementary Material Methods. mESC were obtained from P. Thérizol (INSERM U944), primary and SV40-immortalized MEF cells were derived in house. AML2-NPM1c and AML2-NPM1 were regularly flow-sorted to ensure continuous GFP expression. Actinomycin D (ActD), N-acetyl cysteine (NAC), Glutathione Monoethyl Ester (GSH-MEE), antimycin A, and doxorubicin were purchased from SIGMA (Cat # 01815, Cat # A7250, Cat # G1404, Cat # A8674, Cat # D1515). Human cGAS inhibitor, G140 was purchased from InvivoGen (Cat # inh-g140).

AML patient samples

The study was conducted according to the Declaration of Helsinki and approved by our institutional ethic committee. Primary AML cells were obtained upon written informed consent from the patient during the off-label study. Human primary AML cells were isolated by density gradient centrifugation of either peripheral blood or bone marrow. AML was defined as NPM1-mutated on the basis of targeted sequence analysis.

Animal Studies

For the NPM1c leukemia mouse model, C57BI6-CD45.1 mice were irradiated with 4,5 Gy and then engrafted with 10⁶ CD45.2-NPM1c leukemic cells, a kind gift from P.G. Pelicci (38). In the experiment shown in Fig. 5B, Mice were intraperitoneally injected with ActD (60 μg/kg/d) 19 days after engraftment and sacrificed 24 hours later after an additional ActD injection 1h before euthanasia. CD45.2+ bone marrow cells were sorted by FACS (ARIA2, Becton Dickinson) and lysed directly by Laemmli buffer containing 20 mM NEM and 10% 2-mercaptoethanol, followed by Western blot analysis. For the spleen weight or cytometric analyses, mice were intraperitoneally injected daily with ActD (60 μg/kg/d) for 7 days or 10 days after engraftment. CD45.2 and differentiation markers Mac-1/Gr-1 triple positive cells were examined by FACS. In the experiment shown in Fig. 6H, 10 days after engraftment, mice were intraperitoneally injected with ActD (60 μg/kg/d) and/or orally fed with Venetoclax (INTERCHIM; Cat # 3X552A)) daily for 5 days. Vehicle (60% Phosal 50 PG, 30% PEG400, 10% ethanol) was administered as control. Mice were sacrificed after an additional day without treatment. Spleen weights were measured and percentages of CD45.2 and differentiation markers Mac-1/Gr-1 triple positive cells were analyzed using FACS.

For the patient-derived xenografts, shown in Fig. 5A (top), 10 to 12 weeks old NOD-SCID gamma SGM3 (NSGS) mice were treated with busulfan (Busilvex, 20 mg/kg) followed 24 hours later by 9 × 10⁴ NPM1c-AML patient mononuclear cells engraftment (NPM1c/FLT3 wt). 72 days later, mice were treated daily with ActD (60 μg/kg/d) for 1 or 4 days and sacrificed for bone marrow cells Western blot analysis. In the experiment shown in Fig. 5A (bottom), NSGS mice were irradiated with 1,5 Gy 24 hours before 2.85 × 10⁵ AML patient mononuclear cells engraftment. 87 days later mice received the first intravenously injection with PBS or ActD (120 μg/kg/d) and then treated daily for 4 more days by IP injection, followed by 2 days without any further treatment prior to sacrifice. Percentage of human CD33+ and/or CD45+ and KIT+ bone marrow cells were analyzed using FACS. In the experiment shown in Fig. 6J, 5-10 weeks old NSGS mice were pre-conditioned by 1,25 Gy irradiation, followed by 1 × 10⁶ NPM1c-AML patient mononuclear cells engraftment. 35 days later, mice were ActD- (60ug/kg/d) and/or Venetoclax- (100mg/kg/d) treated daily 5 days/7 for 2 weeks. Percentage of human CD33+ and/or CD45+ bone marrow cells was analyzed using FACS by intra-bones aspiration 5 days before and 15 days after treatment start.

For the NPM1c/IDH1^R132H GFP+ leukemia mice model, C57BI6 mice were irradiated with 4,5 Gy and then engrafted with 10⁶ AML cells, a kind gift from Tak W Mak. 13 days after engraftment, mice were intraperitoneally injected with ActD (60ug/kg/d) and/or fed by Venetoclax (100mg/kg/d) daily, 4 or 5 days/7. Percentage of GFP+ bone marrow cells was analyzed using FACS by intra-bones aspiration 3 days before and 16 days after treatment start. Mice were then followed until they reach humane endpoints and then euthanized.

For the experiment shown in Fig. 5C and 6H, NSGS mice were treated with Busulfan (Busilvex; 20 mg/kg) for 24h, followed by 10⁶ AML3 derivatives engraftment for 7 days. Mice were intra-peritoneally injected with ActD (60 μg/kg/d) for 4 weeks (5 days per week) (Fig. 5C) or 3 weeks (Fig. 6H). Human CD45+ bone marrow cells were examined by FACS and Western blot analysis.

For the experiments shown in Fig. 1A, 2A and Supplementary Fig. S2A, LSK fraction was sorted and stained as previously described (49) from NPM1c^A/+ and control mice (26) four weeks after tamoxifen injection. Briefly, committed hematopoietic cells were depleted through MACS LS column (Mylteni Biotech) and FLSK cells were sorted by FACS using following antibodies: CD3e-biotin, CD8-biotin, IgM-biotin (Life Technologies; Cat # 553728, Cat # 13-0081-85, Cat # 13-5790-85), CD4-biotin, B220-biotin, Mac1-biotin, Tert119-biotin, CD19-biotin, Nk1.1-biotin (BD Bioscience; Cat # 553728, Cat # 553086, Cat # 553309, Cat # 553672, Cat # 553784, Cat # 553163), Gr1-biotin, CD150-perCp/Cy5.5, c-kit-APC/Cy7 (Biolegend; Cat # 108404, Cat # 115922, Cat # 105826), Il7Ra-biotin, CD135-biotin, Streptavidin-Pacific Blue, Sca1-PE/Cy7, CD48-APC (eBioscience; Cat # 13-1271-85, Cat # 13-1351-82, Cat # 48-4317-82, Cat # 25-5981-81, Cat # 17-0481-82). FLSK cells were resuspended in 20 μl of StemSPAN SFEM (StemCell Technologies; Cat # 9600), then seeded on Lab-Tek™ II Chamber Slide (Thermo Fisher Scientific; Cat # 154453) coated with Retronectin (Clonetech; Cat # T202). Rabbit polyclonal anti-TOM20 (Santa Cruz; Cat # FL-145, dilution 1:100), mouse monoclonal anti-B23 (Sigma-Aldrich; Cat # B0556, dilution 1:100) and mouse monoclonal anti-PML (Millipore Sigma; Cat # MAB3738, clone 36.1-104, dilution 1:100) were used for the detection. Z-stack were acquired on a Leica SP5, then deconvolved using the Richardson-Lucy algorithm using measured PSF. Representative iso-surface images and the analysis of number and volume of objects were obtained by Imaris 7 (Bitplane).

All experiments in mice were performed in accordance with protocols approved by the Comité d’Ethique en Expérimentation Animale Paris-Nord no. 121 (project no. 20710-2019071616338948 v2). For the mouse NPM1c leukemia model, C57Bl6-CD45.1, mice were purchased from Charles River Laboratories (agreement n° C 69 208 1301) at 6-weeks old and kept under SPF conditions. For OCI-AML3 or primary patient blasts xenograft experiment, NOD/Shi‐scid IL2rγ^−/− (NSG) mice were obtained from Jackson Laboratories (United States) and kept in SPF conditions.

Immuno-precipitation analyses

For immunoprecipitation analysis, cell extraction was performed with RIPA lysis buffer containing 50 mM Tris (pH 8.0), 0.15 M NaCl, 1% NP40, 1% sodium deoxycholate, 0.1% SDS, 1 mM PMSF, 1 μg/ml aprotinin, and 1 μg/ml leupeptin, 10 mM NEM. Equal amounts of protein lysates were used for immuno-precipitation with indicated antibodies. For protein expression analysis, cells were collected, counted, and lysed in XT sample buffer (Bio-Rad) or 2x Laemmli buffer both containing 20 mM NEM and 10% 2-mercaptoethanol. Samples were heated at 95°C for 10 min. For disulfide-mediated PML-NPM1c complexes, protein samples were divided into two equal aliquots with XT sample buffer containing 20 mM NEM. One of the aliquots was freshly added disulfide bond breaker tris-(2-carboxyethyl) phosphine (TCEP, Thermo Fisher Scientific) and heated at 95°C for 5 min. Details on immunodetection in precipitation sample and cell lysate can be found in Supplementary Material Methods.

Mitochondrial Explorations

Mitochondrial superoxide, membrane potential and mass were measured by using MitoSOX, MitoTracker CMXRos and MitoTracker red FM probes (Thermo Fisher Scientific; Cat # M36008, Cat # M7512, Cat # M22425) respectively. Live cell labeling was done according to manufacturer’s instructions and analysis made by FACS (FACS Calibur, BD Biosciences).

To measure relative mitochondrial gene expression by reverse transcriptase-PCR, whole cell RNA was isolated using RNeasy mini Kit (Qiagen; Cat # 74104). Reverse transcription was performed using iSCRIPT cDNA synthesis kit following the manufacturer’s instructions (Bio-Rad). Quantitative real-time PCR analysis was performed using Roche LightCycler 480 system with SYBR Green PCR Master kit (Roche). Primers are listed in Supplementary Material Methods.

Cytosolic/mitochondria resident mtDNA extraction was carried out as previously described (36). Briefly, 15 × 10⁶ mESCs or 10 × 10⁶ AML cells were washed with ice cold PBS and permeabilized by digitonin buffer containing 150 mM NaCl, 50 mM HEPES (pH 7.4), 1 M Hexylene glycol and 25 μg/ml digitonin for 10 min at 4°C, then centrifuged at 2,000 g for 15 min. The cytosolic supernatants were transferred to fresh tubes and spun at 12000 rpm for 10 min to remove any remaining cellular debris, yielding cytosolic preparations free of nuclear, mitochondrial and any other organelles. Cytosolic DNA was purified by using QIAquick nucleotide removal kit (QIAGEN; Cat # 28304). The intact cell pellets were resuspended in 50 μM NaOH and boiled for 30 min to solubilize DNA. 50 μl of 1 M Tris-HCl pH 8.0 was added to neutralize the pH. Mitochondria and nuclear DNA were purified by using QIAamp DNA Mini kit (QIAGEN; Cat # 51304). Primers are listed in in Supplementary Material Methods. mtDNA was analyzed by Southern blotting. Briefly, 1 μg of mitochondria DNA was separated over a 0.5 % agarose gel overnight in TAE, blotted by capillary transfer onto Hybond-N+ membranes (GE Healthcare). The mtDNA was visualized by Southern blotting using a probe against nucleotide positions 168-604 of human mtDNA labelled using digoxigenin-labelled nucleotide (PCR DIG Probe Synthesis Kit (Sigma-Aldrich; Cat # 11636090910)) and later revealed using anti-digoxigenin antibodies.

To measure 2’3-cGAMP, 40 × 10⁶ AML cells were washed with ice cold PBS and lysed by M-PER™ extraction reagent (Thermo Fisher Scientific; Cat # 78501). 2’3’-cGAMP were quantified by an ELISA kit (Cayman; Cat # 501700) which is based on the competition between 2’3’-cGAMP and a 2’3’-cGAMP-horseradish peroxidase (HRP) conjugate for a limited amount of 2’3’-cGAMP polyclonal antiserum.

Microarray Analysis

Whole cell RNA from one million cells of AML derivates or peripheral blood mononuclear cells of patient were purified using RNeasy Mini Kit (QIAGEN). Microarray were performed at the core facility of Institut de Recherche Saint Louis of Université de Paris. Details can be found in Supplementary material methods. Primary data is accessible in ArrayExpress, accession E-MTAB-10397. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD025507.

Transmission Electron Microscopy

AML2-derived cell lines were washed and fixed in 1.6% glutaraldehyde in 0.1 M phosphate buffer. Cell pellets were secondary fixed with 2% osmium tetroxide and dehydrated using ethanol. Cells were embedded in Epon™ 812. Ultrathin sections and then stained with standard uranyl acetate and lead citrate solutions. Mitochondrial morphology was analyzed using FEI Tecnai 12 electron microscopy equipped with SIS Mega view III CCD camera.

Gene set Enrichment Analysis (GSEA)

For the experiment shown in Fig. 1H and 5G, GSEA was performed using Hallmarks (h.all.v7.1), Gene ontology (c5.all.v7.1) or Curated (c2.cp.KEGG) gene set. 371 genes were selected base on GSEA Hallmarks gene set (h.all.v7.1). For the experiment shown in Supplementary Figure S2B, GSEA was performed using Gene ontology (c5.all.v7.1) or Curated (c2.cp.KEGG) gene set.

Statistical Analysis

Graphpad Prism software was used to perform multiple unpaired two-tailed t-test to determine the P value and look for significant changes in the data generated from the isogenic cell lines and different drug treatment in vivo or ex vivo. All data are expressed as mean ± s.d of technical replicates from a representative experiment of the n performed ones, as indicated. For all graphs, *P = 0.01-0.05, **P = 0.001-0.01, and ***P < 0.001. For the experiment shown in Supplementary Figure S3F, ROS and inflammation-related gene changes from AML2 expressing NPM1c versus NPM1 or AML2 expressing NPM1 treated with ActD versus DMSO were selected to perform linear regression analysis using GraphPad Prism software.

Significance.

Actinomycin D induces complete remissions in NPM1-mutant AMLs. We found that NPM1c affects mitochondrial biogenesis and PML bodies (NBs). Actinomycin D targets mitochondria, yielding ROS which enforce PML NB-biogenesis and restore senescence. Dual targeting of mitochondria with actinomycin D and venetoclax sharply potentiates their anti-AML activities in vivo.