Oxidized mC modulates synthetic lethality to PARP inhibitors for the treatment of leukemia

SUMMARY

TET2 haploinsufficiency is a driving event in myeloid cancers and is associated with a worse prognosis in patients with acute myeloid leukemia (AML). Enhancing residual TET2 activity using vitamin C increases oxidized 5-methylcytosine (mC) formation and promotes active DNA demethylation via base excision repair (BER), which slows leukemia progression. We utilize genetic and compound library screening approaches to identify rational combination treatment strategies to improve use of vitamin C as an adjuvant therapy for AML. In addition to increasing the efficacy of several US Food and Drug Administration (FDA)-approved drugs, vitamin C treatment with poly-ADP-ribosyl polymerase inhibitors (PARPis) elicits a strong synergistic effect to block AML self-renewal in murine and human AML models. Vitamin-C-mediated TET activation combined with PARPis causes enrichment of chromatin-bound PARP1 at oxidized mCs and γH2AX accumulation during mid-S phase, leading to cell cycle stalling and differentiation. Given that most AML subtypes maintain residual TET2 expression, vitamin C could elicit broad efficacy as a PARPi therapeutic adjuvant.

In brief

Vitamin C treatment can slow the progression of AML by enhancing TET2 activity but is not curative as a single agent therapy. Using genetic and compound screening approaches, Brabson et al. identify a rational combination treatment strategy where vitamin C enhances the therapeutic efficacy of PARPis for AML treatment.

Graphical Abstract

INTRODUCTION

Ten-eleven translocation proteins (TET1–TET3) catalyze oxidation of 5-methylcytosine (5-mC) to form 5-hydroxymethylcytosine (5-hmC), 5-formylcytosine (5-fC), and 5-carboxylcytosine (5-caC), oxidized mCs (oxi-mCs) considered to be key intermediates of DNA demethylation and reversal of gene silencing.^1–3 5-hmC can be maintained as a stable epigenetic mark that drives passive DNA demethylation by blocking the ability of DNA maintenance methyltransferase 1 (DNMT1) to recognize these bases upon DNA replication.⁴ 5-fC and 5-caC, however, trigger active DNA demethylation independent of the cell cycle because of their recognition and replacement with an unmethylated cytosine by the action of thymidine DNA glycosylase (TDG) and additional proteins involved in base excision repair (BER).^5–8

Of the three mammalian TET genes, TET2 is the most frequently mutated in myeloid malignancies, including 10% of de novo acute myeloid leukemia (AML), 30% of myelodysplastic syndrome (MDS), and up to 50% of chronic myelomonocytic leukemia (CMML),^9–11 leading to impaired oxidation of 5-mC¹² and a DNA hypermethylation phenotype.^9,13,14 Recent studies have revealed vitamin C to be a potential non-toxic therapy for treatment of myeloid malignancies by enhancing residual TET2 activity. Vitamin C is a co-factor of the superfamily of Fe²⁺ and α-ketoglutarate-dependent dioxygenases, which include the TET proteins, collagen and hypoxia-inducible factor prolyl hydroxylases, and histone and RNA demethylases (reviewed in Lee Chong et al.¹⁵). By acting as a targeted reductant, vitamin C converts ferric iron (Fe³⁺) to ferrous iron (Fe²⁺), which, during TET-mediated catalysis, leads to increased rates of 5-mC oxidation by 2- to 8-fold.^16,17

Patients with hematological malignancies are often vitamin C deficient compared with normal healthy controls,^18,19 and dietary vitamin C deficiency is associated with reduced 5-hmC levels and accelerated AML progression in mice.²⁰ In a recent study, oral vitamin C supplementation has been shown to restore normal plasma concentrations in patients with myeloid cancer who were moderately deficient at diagnosis, leading to increased 5-hmC levels in circulating white blood cells.¹⁸ While the therapeutic benefit of long-term oral supplementation remains unknown, numerous studies have shown that high-dose vitamin C can enhance therapeutic efficacy in the treatment of solid tumors when given alone or in combination with radiation, standard chemotherapy, or immune checkpoint inhibitors.^21–23 Other studies have reported increased efficacy of arsenic trioxide and 5-azacytidine in the treatment of myeloid cancers in combination with vitamin C.^24,25 Our previous work showed that enhancing Tet2 activity, using genetic models of reversible RNAi or high-dose vitamin C treatment, can block the aberrant self-renewal of pre-malignant hematopoietic stem cells and suppress the expansion of TET2-deficient immature and mature myeloid lineage cells.²⁶ The ability to enhance and restore TET2-dependent mC oxidation and gene expression programs could be exploited in combination therapies to inhibit AML growth and improve treatment outcome, specifically in patients with mutant TET2 patients.

Here we show that vitamin C treatment can slow the progression of Tet2-deficient murine AML but is not curative at high dose as a single-agent therapy. To identify rational vitamin C combination therapies for the treatment of AML, we performed two independent screens of AML cell viability using a whole-genome CRISPR knockout approach and a small compound library targeting epigenetic regulators. Several US Food and Drug Administration (FDA)-approved drugs were found to exhibit increased efficacy at blocking AML survival in the presence of vitamin C, but the greatest synergy occurred upon targeted inactivation of BER via co-treatment with poly-ADP-ribosyl polymerase inhibitors (PARPis).

We verified the efficacy of PARPis in combination treatment with vitamin C using a panel of human AML cell lines harboring diverse fusion oncogenes (MLLr, AML1-ETO, and PML-RAR) and primary murine leukemic cells with Tet2^+/+, Tet2^+/−, or Tet2^−/− alleles. Maximal efficacy of combination treatment was found to occur in the presence of residual TET2 activity regardless of oncogenic context and was associated with elevated levels of γH2AX, cell cycle stalling, and myeloid differentiation. These findings were recapitulated in vivo using murine leukemia and patient-derived xenograft (PDX) models. Finally, we show that the mechanism of vitamin C and PARPi combination efficacy is associated with an accumulation of 5-fC in PARP1-bound chromatin and enrichment of these oxi-mCs at sites of replicative stress marked by γH2AX. Vitamin C in combination with PARPis prevents aberrant leukemic hematopoietic stem and progenitor cell (HSPC) self-renewal in a Tet2-dependent manner and inhibits leukemic cell expansion in vitro and in vivo greater than either agent alone. We conclude that enhancing TET function with vitamin C creates a novel mechanism of synthetic lethality to PARPis driven by recruitment of PARP proteins to sites of endogenous replicative stress induced by oxi-mC formation that could be harnessed for improved treatment outcome in myeloid cancers.

RESULTS

Restoration of TET2 activity in vivo with vitamin C promotes differentiation and slows disease progression in a murine leukemia model

To understand how Tet2 functional restoration influences leukemia progression and maintenance, we generated an AML model of reversible Tet2 knockdown using a ROSA-M2-rtTA (RTA) tetracycline (doxycycline [DOX])-regulated transactivator and miR30-based small hairpin RNA (shRNA) transgenic mice to express an inducible GFP-linked Tet2-specific shRNA (shT2) or Renilla luciferase shRNA (shRN) as a negative control²⁶ (Figure S1A). HSPCs isolated from Tet2^+/+, Tet2^+/−, and Tet2^−/− and RTA shT2 or shRN transgenic mice were transduced with AML-ETO9a (AE9a) and injected into lethally irradiated recipient mice to generate primary leukemia. We observed a Tet2-dependent effect on leukemia progression with 100% loss of survival by 230 days for AE9a Tet2^−/− recipient mice compared with 50% loss of survival at ~300 days for AE9a Tet2^+/− and 100% survival at this time point for AE9a Tet2^+/+ mice (Figure 1A). AE9a⁺ RTA shT2 leukemia progression mirrored the kinetics of AE9a⁺ Tet2^+/− mice (Figures 1A, S1B, and S1C), consistent with an approximate 50%–70% knockdown of Tet2 mRNA expression (Figure 1B). All AE9a⁺ leukemias exhibited cKit⁺ staining and low or negligible CD11b surface expression (Figures S1D and S1E).

Figure 1. Restoration of TET2 activity with vitamin C promotes differentiation and slows disease progression in a murine AML model.

(A) Survival curve of primary murine AML-ETO9a (AE9a) leukemia modeled with Tet2^+/+, Tet2^+/−, or Tet2^−/− (left panel) and transgenic Rosa-rtTA2 (RTA) shRN or shT2 mice (right panel).

(B) Tet1–3 mRNA levels in AE9a leukemia cells with Tet2^+/+, RTA shRN, Tet2^+/, and RTA shT2.

(C) Schematic of doxycycline (DOX)-regulated Tet2 knockdown (KD) and restoration (RS) in secondary transplant recipient mice treated with vehicle (PBS), DOX withdrawal, or ascorbate (ASC).

(D–F) ASC treatment of mice transplanted with primary AE9a leukemia cells (AE9a RTA shT2 #12).

(D) Leukemic cell numbers in the peripheral blood (PB) of KD, RS, and ASC-treated mice.

(E) Percent of PB CD11b⁺ leukemia cells upon 5 weeks of treatment.

(F) Survival curve of KD, RS, or ASC-treated mice.

(G–I) Analysis of (G) spleen (SP) weight, (H) leukemia cell numbers in the SP and bone marrow (BM), and (I) H&E staining of liver and lung leukemic infiltration (200-μm scale bar indicated) in mice treated for 4 weeks.

(J) Representative flow cytometry of AE9a RTA shT2 splenic leukemia cells from KD, RS, and ASC mice treated for 4 weeks.

(K and L) Sorted SP and BM leukemia cells from AE9a RTA shT2 KD, RS, and ASC mice treated for 4 weeks were measured for (K) relative Tet2 mRNA levels and (L) relative 5-hmC and 5-fC levels in genomic DNA (normalized to total 5-mC) by ELISA. Mean ± SEM of 3–7 mice per group.

(M and N) RNA sequencing analysis of leukemia treated in vivo.

(M) Heatmap of all significant differentially expressed genes (padj < 0.05) in AE9a RTA shT2 leukemia cells of ASC versus KD compared with RS and untreated primary leukemia (1°TP).

(N) Gene set enrichment analysis (GSEA) plots of genes up- and down-regulated in Tet2 RS HSPCs. Decreased and increased gene expression is indicated in shades of blue and red, respectively. *p < 0.05, **p < 0.005, and ***p < 0.0005.

To model the effect of endogenous Tet2 restoration compared with vitamin C treatment on leukemia progression in vivo, we generated cohorts of secondary and tertiary transplanted mice with AE9a⁺ RTA shT2 leukemia cells. DOX withdrawal allows Tet2 endogenous mRNA expression levels to be restored in the RTA transgenic system. In our previous work, a single intraperitoneal (i.p.) injection of 4 g/kg vitamin C (ascorbate [ASC]) increases 5-hmC levels upto 2-fold incirculating white blood cells.²⁶Titration of ASC delivered i.p. for the treatment of aggressive tertiary AE9a⁺ RTA shT2 leukemia showed effective suppression of peripheral blood leukemic cell expansion when administered at 1–4 g/kg (Figures S1F and S1G). We divided secondary recipient mice into treatment groups of sustained Tet2 knockdown (KD; PBS, +DOX), Tet2 restoration upon DOX withdrawal (RS; PBS, −DOX), or vitamin C treatment (ASC, +DOX) to compare survival and leukemia burden (Figure 1C). Mice were treated with single daily i.p. injections of 4 g/kg vitamin C for 5 days per week, which mimics pharmacological dosing of ASC in human clinicaltrials.^27–30 We found that genetic restoration of Tet2 expression or ASC treatment reduces disease burden in vivo, drives leukemia cell myeloid differentiation, and confers a modest but significant survival advantage for ASC-treated mice, potentially because of effects on the microenvironment or activation of additional TET proteins such as Tet1/Tet3 (Figures 1D–1F and S1H).²⁶

At 4 weeks of treatment, we observed reduced spleen weight and fewer leukemia cells in the spleen and bone marrow of Tet2 RS or ASC-treated mice (Figures 1G, 1H, and S1I). Leukemia cell infiltration of the liver and lung was also reduced (Figure 1I). Furthermore, leukemia cells in the peripheral blood upregulated CD11b in response to Tet2 RS or ASC treatment, but this was not observed in the spleen and bone marrow (Figure S1J). Purified splenic AE9a⁺ leukemia cells from mice treated for 4 weeks in vivo showed induction of Tet2 expression upon genetic restoration but not ASC treatment (Figure 1K). Increased 5-hmC/5-mC and 5-fC/5-mC ratios were also observed in genomic DNA from Tet2 RS and ASC treated cells compared with PBS-treated KD controls (Figure 1L). Importantly, RNA sequencing analysis of splenic leukemia cells from moribund mice revealed that ASC treatment induces a Tet2 RS gene expression signature²⁶ (Figures 1M, 1N, and S1K–S1M; Table S1), providing in vivo evidence that vitamin C treatment can effectively target leukemia cells residing within peripheral hematopoietic organs to enhance Tet2 activity.

Multiple FDA-approved drugs synergize with vitamin C to reduce AML cell viability

To explore the therapeutic efficacy of vitam in C in combination with additional therapeutic agents, we implemented genetic and compound screening approaches in AML cells to identify rational druggable targets. First, we performed a genome-wide CRISPR-Cas9 knockout screen with the Brunello sgRNA library³¹ in MOLM-13 human AML cells treated with vitamin C for ~2 weeks at a high pharmacological concentration of vitamin C (2 mM L-ascorbic acid [LAA]) (Figure S2A). The CRISPR screen identified several druggable targets whose loss of function (negative selection) increased sensitivity to vitamin C, including FASN, KAT6B, HDAC3, BCL2, NFKB, DOT1L and BER proteins, such as TDG and GADD45G (Figures 2A and 2B; Table S2). Targets whose loss of function conferred positive selection included genes involved in ataxia telangiectasia mutated (ATM) signaling (TP53 and CHK2) (Figures 2A and 2B). Gene Ontology pathway analysis of the top 500 significant CRISPR AML targets revealed that loss of function in genes involved in adipogenesis (KAT6B), thyroid hormone signaling (TR/RXR), receptor activation (FASN), sirtuin signaling (GADD45G, DOT1L, and NFKB2) and biotin-carboxyl carrier protein assembly (ACACA) were associated with negative selection (Figure 2C; Table S2). Target genes in pathways of G1/S checkpoint regulation (TP53, CDKN1A, and CDKN2C), epithelial-to-mesenchymal transition (EMT) (WNT genes and HIF1A), and RIG-1 like antiviral innate immunity (RELA and IFNA genes) were associated with positive selection (Figure S2B; Table S2).

Figure 2. Vitamin C synergizes with multiple drugs to reduce AML cell viability.

(A–C) Brunello sgRNA library CRISPR knockout screen in AML cells treated with vitamin C (LAA).

(A) sgRNA target genes ranked by log₂ fold change in LAA- versus PBS-treated MOLM-13 cells.

(B) Volcano plot of sgRNA target genes, highlighting negatively selected fatty acid (FA) synthesis and base excision repair (BER) genes and positively selected ATM signaling and p21 activation genes.

(C) Pathway analysis of negatively selected genes.

(D–G) Epigenetic target compound screen in AML cells treated with or without LAA.

(D) Venn diagram of drugs that elicit a more than 10% decrease in viability with LAA compared with drug alone.

(E) Class distribution of ALL library compounds and those with LAA-enhanced decrease in AML viability.

(F) Heatmap of the compound library ranked by average percent decrease in viability and LAA for all AML cell lines (left panel), with top-ranked significant compounds (>2 standard deviations) shown in red (right panel).

(G and H) Dose-response curves (G) and IC₅₀ summary (H) of select compounds titrated with LAA in MOLM-13 and THP-1 cells. *p < 0.05, **p < 0.005.

(I) Representative synergy plots of Olap and LAA in MOLM-13 and THP-1 cells.

(J) Heatmap of average Bliss synergy scores for select compounds tested in MOLM-13 and THP-1 cells with LAA. Mean of 2 or more experiments.

Using a secondary screening approach, MOLM-13, THP-1, and KASUMI-1 human AML cell lines, all of which are wild type for TET1–TET3 (Figure S2C) were tested for effects on viability upon 72 h of treatment with vitamin C and an epigenetic targeted library of 166 compounds (Table S3). Upon addition of vitamin C, 36 of 166 compounds showed an additive reduction in AML cell viability by more than 10% compared with drug treatment alone in all three cells lines (Figures 2D and S2D). Based on the representation of drug classes in the epigenetic library, the 36 compounds with increased efficacy upon vitamin C co-treatment were enriched for inhibitors of histone deacetylases (HDACis), protein methyltransferases (PMTis), histone acetyltransferases (HATi), and Sirtuin activators or inhibitors (SIRTa/is) (Figure 2E; Table S3). Across all three AML lines with vitamin C co-treatment, several specific agents, including scriptaid and apicidin (HDACis), caused a reduction in viability, but the most significant effect was observed upon combination with a PARPi, olaparib (Figures 2F, S2D, and S2E). Compounds whose effect on viability was abrogated by vitamin C included JIB 04 (a pan-Jumonji histone demethylase inhibitor) and disulfiram (an alcohol dehydrogenase inhibitor) (Figure S2F).

We validated the results of our screens by performing liquid culture titration assays using vitamin C concentrations ranging from physiological doses (62.5 μM) to low and high pharmacological doses (250 μM and 1 mM) in combination with several targeted drugs in MOLM-13 and THP-1 cells (Figures 2G–2J). Vorinostat (a pan-HDACi), venetoclax (a BCL2 inhibitor [BCL2i]), TPCA-1 (a nuclear factor κB inhibitor [NF-κBi]), pinometostat (a DOT1-L inhibitor [DOT1-Li]), C75 (a FASN inhibitor [FASNi]), and olaparib exhibited decreased IC₅₀ by several orders of magnitude upon combination treatment (Figures 2G and 2H). The lethality of vitamin C and olaparib co-treatment in MOLM-13 and THP-1 cells was shown to be synergistic upon Bliss analysis (Figures 2I and 2J). The results of these screens reveal the potential therapeutic efficacy of vitamin C as an adjuvant to several classes of FDA-approved drugs for the treatment of cancer and AML.

Synthetic lethality to PARPis and AML cell differentiation is induced by vitamin C co-treatment

Our screening results suggested that targeting BER pathway genes, and specifically blocking PARP activity in combination with vitamin C treatment, leads to synthetic lethality in AML. To examine whether this combination was efficacious for the treatment of diverse AML subtypes, we expanded our cell line panel to include additional myelomonocytic cell lines harboring MLL rearrangements (THP-1, MOLM-13, MOLM-14, NOMO-1, and MV4-11); stem- or APL-like AML lines such as KASUMI-1 (AML-ETO), KG-1 (FGFR1OP2-FGFR1), and NB4 (PML-RAR); and TP53-deficient HL-60 and U-937 cells. Treatment of the AML panel with olaparib in combination with vitamin C (250 μM LAA) decreased the IC₅₀ in all cell lines (Figure 3A) independent of their sensitivity to either vitamin C or Olaparib alone (Figures 3B and S3A–S3C). MLL-rearranged AMLs have been characterized previously as insensitive to PARPis because of HOXA9-mediated upregulation of genes associated with homologous recombination (HR),³² such as BRCA1/2 and RAD51, however, we found that the addition of vitamin C in combination with olaparib increased the sensitivity of all MLL-AF9⁺ AML cells by 5- to 30-fold (Figure 3C).

Figure 3. Synthetic lethality to PARPi and AML cell differentiation induced by vitamin C.

(A–C) AML cell panel viability screen with Olap.

(A) Representative dose-response curves of Ola titrated in 10 AML cell lines with and without 250 μM L-ascorbic acid (LAA) for 72 h. Mean ± SD of 2 or more experiments.

(B) IC₅₀ of LAA (top) and Olap (bottom) in 10 AML cell lines.

(C) IC₅₀ of Olap with and without 250 μM LAA in 6 AML cell lines with greatest resistance to either Olap or LAA as a single agent (top) and Olap IC₅₀ fold change (bottom). Mean ± SEM of 2 or more experiments.

(D) Dose-response curves of MOLM-13 cells treated for 72 h with veliparib, rucaparib, Olap, or talazoparib in combination with LAA. Mean ± SD, representative of 2 or more experiments.

(E) Cell cycle analysis of MOLM-13 cells treated with LAA and veliparib, rucaparib, Olap, or talazoparib. Representative experiment, mean ± SD.

(F) Cell cycle profiling of MOLM-13 and THP-1 cells treated with Olap and LAA after 72 h. Representative experiment, mean ± SD.

(G) Relative mean fluorescence intensity (MFI) of the myeloid differentiation markers CD14 and CD11b in MOLM-13 and THP-1 cells treated for 72 h with LAA and Olap. Representative experiment, mean ± SD.

(H) Wright-Giemsa stain of MOLM-13 and THP-1 cells treated with LAA and Olap for 72 h. Scale bar, 50-μm.

(I–K) RNA sequencing of LAA + Olap-treated AML cell lines.

(I) Volcano plots of differentially expressed genes (DEGs) in MOLM-13 (top) and THP-1 (bottom) cells treated with LAA (250 μM), Olap (1 μM), or LAA + Olap versus vehicle for 72 h, showing upregulated (red), downregulated (blue), and not significant (black) genes. Significance cutoff of padj < 0.05 and log₂(|fold change|) > 0.5.

(J) Venn diagrams of genes uniquely upregulated (left) or downregulated (right) in LAA + Olap-treated cells.

(K) Enriched gene sets in uniquely upregulated (top, red) and downregulated (bottom, blue) LAA + Olap DEGs from the Molecular Signature Database (MSigDB) Hallmark collection. *p < 0.05, **p < 0.005, ***p < 0.0005.

To determine whether vitamin C could potentiate the effect of other PARPi compounds, we compared veliparib, rucaparib, olaparib, and talazoparib treatment with vitamin C on MOLM-13 cell viability. All PARPis tested showed improved efficacy, as evidenced by decreased IC₅₀ (Figures 3D and S3D). Co-treatment also induced an expansion of S and G2/M cell cycle phases, greater than either treatment alone (Figure 3E), that correlated with the trapping potency of PARPis, which increases from veliparib < rucaparib < olaparib < talazoparib.³³ The S phase expansion upon combination treatment was shown to be vitamin C dose dependent in MOLM-13 and THP-1 cells (Figure 3F).

Previous studies have reported increased differentiation of AML cells in response to PARPis.³² We observed that, upon vitamin C co-treatment, olaparib induced surface expression of mature myeloid markers (CD14 and CD11b) at a 10-fold lower concentration in MOLM-13 and THP-1 cells (Figure 3G). Co-treated cells also exhibited morphological changes consistent with differentiation compared with vehicle- or single agent-treated cells (Figure 3H). The strong differentiation effect of combined vitamin C and olaparib treatment was also confirmed by transcriptomics analysis (Figures 3I–3K and S3E–S3I; Table S4). Vitamin C treatment alone led to greater gene expression changes (121–224 up versus 177–213 down) in MOLM-13 and THP-1 cells than olaparib (13–41 up versus 12–73 down), while the combination created more than 4,000 gene expression changes (Figure 3I), with MOLM-13 and THP-1 cells exhibiting up-regulation of interferon α/γ (IFN-α/γ) signaling genes, endogenous retrovirus (ERVK3), and myeloid differentiating transcription factors (CEBPB/G). Down-regulated genes included the leukemia stem cell marker CD99 and interleukin-6 (IL-6), IL-2, and tumor necrosis factor alpha (TNF-α) cytokine signaling pathway genes (Figures 3J, 3K, and S3E–S3I; TableS4). G2-M cell cycle checkpoint genes were also upregulated upon combination treatment, including MDC1 and BRCA1/2 DNA repair genes (Figures 3K and S3G), suggesting that the combination treatment does not reduce the expression of canonical HR genes as its mechanism of action.

Enhanced myeloid differentiation and suppression of primary leukemia stem cell self-renewal by combined PARPi treatment with vitamin C

We have shown previously that vitamin C mediates Tet2-dependent suppression of colony-forming unit (CFU) capacity and self-renewal in primary mouse and human HSPCs.²⁶ To determine whether combination treatment efficacy with PARPis and vitamin C is also dependent on Tet2 expression, we tested Tet2^+/+, Tet2^+/−, and Tet2^−/− murine HSPCs with no oncogene or in the context of AE9a or MLL-AF9 overexpression using colony formation and liquid culture assays (Figure 4A). Olaparib and talazoparib were first tested for toxicity to wild-type (WT) HSPCs in colony formation assays, alone and in combination with vitamin C. WT cells treated with 1 μM olaparib or 1 μM talazoparib were found to maintain more than 70% CFU capacity with no added toxicity upon co-culture with LAA (250–1,000 μM) (Figure S4A). Consistent with our previous findings, in the absence of an overexpressed oncogene, Tet2^+/− CFUs lose their aberrant replating capacity upon vitamin C treatment alone, while Tet2^−/− CFU cells become insensitive to vitamin C but display sensitivity to olaparib and talazoparib (Figures 4B and S4B). In the context of AE9a⁺ or MLL-AF9⁺ HSPCs, vitamin C alone at 250 μM is not sufficient to significantly block colony formation and re-plating capacity regardless of Tet2 genotype. However, while olaparib treatment reduces colony numbers, the combination of vitamin C and olaparib does so with greater significance compared with vehicle but not with either treatment alone (Figure 4B). MLL-AF9⁺ Tet2^+/− CFUs cell were also significantly suppressed by talazoparib in combination with vitamin C compared with vehicle or talazoparib treatment alone (Figure S4B). When vehicle-treated cells were re-plated with a higher concentration of vitamin C (1 mM), the combination with PARPi significantly reduced overall colony formation capacity, with the greatest sensitivity observed in Tet2^+/+ or Tet2^+/− leukemic cells compared with Tet2^−/− (Figures 4C and S4D). Differentiation of CFU cells toward a more mature myeloid cell phenotype (CD11b⁺, Gr1⁺), was also observed to be most significant for PARPi and vitamin C combination treatment of Tet2^+/− CFU cells with or without oncogene overexpression (Figures 4D, 4E, and S4C).

Figure 4. Combination of PARPis and vitamin C treatment reduces self-renewal and increases myeloid differentiation of primary murine leukemic cells.

(A) Relative Tet1–3 mRNA levels in Tet2^+/+, Tet2^+/−, and Tet2^−/− murine HSPCs with no oncogene and in the context of AE9a or MLL-AF9 overexpression.

(B–E) Effect of vitamin C treatment in combination with PARPis on self-renewal and differentiation of primary murine HSPCs.

(B) Colony formation of HSPCs with no oncogene or in the context of AE9a or MLL-AF9 overexpression, treated with LAA (250 μM) or Olap (1 μM), or the combination. Cells were re-plated for four passages (P1–P4). Mean +SEM, n ≥ 2 experiments.

(C) Vehicle-treated colonies with AE9a and MLL-AF9 overexpression re-plated with higher concentrations of vitamin C (up to 1 mM) for additional passage P4 and P5 and their relative colony numbers normalized to vehicle treatment (n ≥ 6 experiments).

(D and E) Representative flow cytometry of Tet2^+/− CFU cells (D) and relative MFI of a myeloid differentiation marker (Gr1) (E) across representative genotypes, with no oncogene overexpression at P2 or in the context of oncogenic (AE9a or MLL-AF9) expression at P3 (mean +SEM, n ≥ 2 biological replicates).

(F and G) Representative dose-response curves (F) and average IC₅₀ values (G) of HSPCs overexpressing AE9a or MLL-AF9 grown in liquid culture treated with Olap and LAA (250 μM). Mean ± SEM, n = 4 experiments.

(H) Relative MFI of Gr1 upon Olap and LAA treatment compared with vehicle. Mean +SEM, n = 2–4 experiments; *p < 0.05, **p < 0.005, ***p < 0.0005.

In liquid culture assays, low pharmacological doses of vitamin C (LAA, 250 μM) were shown to be sufficient to reduce the IC₅₀ of olaparib or talazoparib by 10- to 100-fold independent of Tet2 genotype (Figures 4F, 4G, and S4F–S4I). However, Tet2^+/− leukemic cells again showed a greater sensitivity to combination treatment than either Tet2^+/+ or Tet2^−/− cells, which correlated with enhanced upregulation of Gr1 and CD11b surface expression (Figures 4H and S4E). Together, these findings suggest that low doses of vitamin C in combination with PARPis are sufficient to suppress the colony formation capacity of pre-malignant Tet2-deficient HSPCs, slow leukemia growth in liquid culture, and enhance myeloid differentiation, whereas high pharmacological doses of vitamin C are needed as an adjuvant to block AML CFUs in methylcellulose assays.

Combination treatment of PARPis with high-dose vitamin C in murine and PDX AML models promotes differentiation in vivo and increases survival

To determine the in vivo efficacy of vitamin C treatment in combination with PARPis, we generated murine AE9a⁺ and MLL-AF9⁺ Tet2^+/− AML transplant models for combination treatment studies. We observed that mice administered ASC (2 g/kg) with olaparib (25 mg/kg) i.p. exhibited a significant block in frequency and number of AE9a⁺ Tet2^+/− leukemia cells in peripheral blood, with increased differentiation toward the myeloid lineage, and 75% survival after 6 weeks compared with no survival upon either treatment alone (Figures 5A–5C). MLL-AF9⁺ Tet2^+/− and Tet2^−/− leukemias exhibit accelerated disease progression compared with MLL-AF9⁺ Tet2^+/+ with no alteration in myelomonocytic phenotype (Figures S5A–S5C). Despite a high disease burden at onset of treatment, MLL-AF9⁺Tet2^+/− leukemia-bearing mice administered ASC (4 g/kg) and olaparib (50 mg/kg) also showed a suppression of frequency and number of circulating leukemia cells, a bias toward increased neutrophilic differentiation driven by ASC, and 50% overall survival after a 6-week treatment window compared with only 25% survival for either treatment alone (Figures 5D–5F). Efficacy of the ASC + olaparib (Olap) combination therapy for the treatment of MLL-AF9⁺ leukemia was also found to be dependent on residual Tet2 levels, with Tet2^+/+ and Tet2^+/− leukemia-bearing mice showing a significant survival advantage compared with Tet2^−/− mice (Figures 5G–5I).

Figure 5. Increased therapeutic efficacy of PARPis with high-dose vitamin C in murine and human AML models in vivo.

(A–C) Secondary AE9a⁺ Tet2^+/− transplanted mice treated with vehicle, ASC (2 g/kg), Olap (25 mg/kg), and ASC + Olap.

(A) Circulating AE9a⁺ cells by percentage (left) and total number (right).

(B) CD11b⁺Gr1⁻ and CD11b⁺Gr1⁺ flow cytometry staining of AE9a⁺ PB cells after 4 weeks of treatment.

(C) Survival curve (treatment window in gray).

(D–F) Primary MLL-AF9⁺ Tet2^+/ transplanted mice treated with vehicle, ASC (4 g/kg), Olap (50 mg/kg), and ASC + Olap.

(D) Peripheral leukemia cell percentage (left) and number (right).

(E) Frequency of CD11b⁺Gr1⁻ and CD11b⁺Gr1⁺ populations in MLL-AF9⁺ PB cells after 4 weeks of treatment.

(F) Survival curve (treatment window in gray).

(G–I) ASC (4 g/kg) + Olap (50 mg/kg)-treated primary MLL-AF9 Tet2^+/+, Tet2^+/−, and Tet2^−/−-transplanted mice.

(G) Frequency (left) and total number (right) of circulating MLL-AF9 cells.

(H) CD11b⁺Gr1 and CD11b⁺Gr1⁺ frequency of peripheral MLL-AF9⁺ cells after 4 weeks of treatment.

(I) Survival curve (treatment window in gray).

(J) Liquid culture differentiation of normal human (samples M24 and 3080) and primary (1°) leukemic patient CD34⁺ stem cells (samples 4103, 4101, and 4837) treated with LAA and Olap for 7 days and MFI (left) and frequency (right) of CD11b⁺ and CD14⁺ cells in normal (top) and leukemic (bottom) CD34⁺ cultures.

(K and L) Primary patient MLL-AF9⁺ TET2mt leukemic cells transplanted in an NSGS mouse model treated with ASC, Olap, and talazoparib.

(K) PB hCD45 cell numbers.

(L) CD14 relative MFI (left) and terminal SP weight (right). *p < 0.05, **p < 0.01, ***p < 0.001.

We next tested the effect of vitamin C in combination with Olap on CD34⁺ cells isolated from normal donors and primary patients with AML (Table S5). In liquid culture, primary patient AML cells upregulated CD14 surface expression in response to vitamin C or Olap treatment, which was enhanced when tested in combination, while little effect on differentiation was observed for normal CD34⁺ cells, suggesting AML-specific induction of differentiation upon co-treatment (Figures 5J and S5D). We also determined the effect of ASC + PARPi treatment on two primary human AML PDX models with MLLA-F9 or FLT3-ITD mutation (Table S5). Upon detection of more than 1% CD45⁺ cells in PB, mice were treated with ASC (4 g/kg, i.p.) and Olap (25–50 mg/kg) or talazoparib (0.25 g/kg) administered i.p. or by oral gavage. Treatment with PARPis in combination with vitamin C significantly reduced peripheral blood leukemia burden and spleen weight and increased differentiation (Figures 5K, 5L, S5E, and S5F). These findings highlight the improved efficacy of PARPis as a therapy for AML when administered in combination with high-dose vitamin C.

Cell-cycle lengthening, γH2AX formation, and S phase stalling correlates with vitamin-C-mediated PARPi synthetic lethality in AML cells

Hallmark features of PARPi-mediated cytotoxicity are induction of γH2AX foci at DNA double-stranded breaks (DSB), replication stalling, and an expansion of cells blocked at G2/M.³³ In combination with vitamin C, we observed in MOLM-13 and THP-1 AML cells a 4-fold higher increase in γH2AX formation than with Olap treatment alone (Figures 6A–6C). The greatest induction of γH2AX upon combination treatment was found to occur during mid- and late S phase, significantly greater than with Olap treatment alone (Figures 6D–6F). The increased frequency of AML cells in S phase and γH2AX formation was evident within 24 h of co-treatment and dose dependent on Olap in combination with vitamin C (Figures S6A–S6C), and 5-hmC was also increased by 72 h (Figure S6D). Furthermore, we observed a decrease in the replication rate of AML cells upon Olap and vitamin C co-treatment, as evidenced by increased bromodeoxyuridine (BrdU) retention in cells during pulse-chase labeling with EdU (Figures 6G, 6H, and S6E). Given the lack of genome fragmentation observed by comet assay (Figures 6I and S6F) and minor effects on apoptosis after 72 h of treatment (Figure 6J), the reduction in cell density observed upon co-treatment of Olap and vitamin C (Figure 6K) leading to loss in cell viability can most likely be attributed to replicative stress-induced cell cycle stalling rather than cell death. Previous studies have also reported that lengthening the cell cycle, by DNA damage or overexpression of the CDK inhibitor p21 (CDKN1A), is sufficient to drive differentiation of normal HSPCs and MLL-rearranged leukemia cells.^34–36 MOLM-13 and THP-1 AML cells exhibit induction of p21 protein expression after 72 h of combination treatment greater than with either treatment alone (Figure 6L). Taken together, these data suggest that Olap treatment in combination with vitamin C mimics a DNA damage response that induces replicative stress and lengthening of the cell cycle, leading to AML cell differentiation.

Figure 6. Replicative stress, γH2AX formation, and oxi-mC accumulation correlates with vitamin-C-mediated PARPi synthetic lethality in AML cells.

(A and B) Immunofluorescence of AML cell lines.

(A) Representative images of MOLM-13 and THP-1 cells with γH2AX, RAD51, and DAPI after 72-h treatment with vehicle, LAA (250 μM), Olap (1 μM), or LAA + Olap. Scale bar, 25 μm.

(B) γH2AX (top) and RAD51 (bottom) focus/cell quantitation.

(C) Relative γH2AX (top) and RAD51 (bottom) MFI in MOLM-13 and THP-1 cells treated with LAA and Olap for 72 h.

(D–F) Cell cycle dynamics of LAA + Olap-treated cells.

(D) Representative density plots of MOLM-13 and THP-1 cells treated for 72 h with vehicle, LAA, Olap (Olap), or LAA + Olap and stained for EdU incorporation and DAPI.

(E and F) Relative frequency of MOLM-13 (top) and THP-1 (bottom) cells (E) and γH2AX MFI within each cell cycle gate (F), normalized to vehicle. Mean ± SEM of 3 or more experiments. *p < 0.05, **p < 0.005 (normalized to vehicle), #p < 0.05 (normalized to Olap).

(G and H) Cellular proliferation and cycling dynamics.

(G) Schematic of BrdU pulse pre-treatment, followed by 72-h treatment with vehicle, LAA, Olap, or LAA + Olap, and EdU chase.

(H) Relative frequency of BrdU⁺ (left) and percentage of EdU⁺BrdU⁺ (right) of THP-1 cells. Representative experiment, mean ± SEM.

(J) MOLM-13 and THP-1 cells treated for 72 h with vehicle, LAA, Olap, or LAA + Olap, stained for Annexin V and DAPI, live (double negative), early apoptotic (Annexin V⁺DAPI⁻), or late apoptotic (Annexin V⁺DAPI⁺). Representative experiment, mean + SD.

(I) Comet assay performed on MOLM-13 and THP-1 cells treated for 72 h with vehicle, LAA, Olap, or LAA + Olap or for 2 h with 20 μM etoposide. Shown is the tail moment of treated THP-1 (left) and representative fluorescence images of treated MOLM-13 and THP-1 cells (right). Scale bar, 200 μm.

(K and L) Quantification of MOLM-13 and THP-1 cells (seeded at 0.2 × 10⁶ cells/mL) (K) and relative p21 MFI (L) upon treatment with vehicle, LAA, Olap, or LAA + Olap for 72 h. Mean ± SEM, n = 2 experiments. *p < 0.05, **p < 0.005, ***p < 0.0005, ****p < 0.00005.

PARP inhibition in combination with vitamin C treatment leads to enrichment of PARP1 binding and γH2AX formation at 5-fC sites in the genome of AML cells

The presence of TET-mediated oxidation products of 5-mC in the genome, such as 5-fC or 5-caC, can lead to DNA polymerase pausing^37,38 and slowing of RNA transcription and elongation.^39,40 If left unresolved, the increased presence of these oxi-mCs in the genome could induce replicative stress or collapse of replication forks, leading to DNA DSB formation and genomic instability.^41,42 5-fC and 5-caC are excised by the mammalian DNA glycosylase TDG, which, in combination with downstream regulators of BER, are essential for active DNA demethylation.^5–7 In our CRISPR knockout screen of MOLM13 cells cultured in the presence of vitamin C, TDG sgRNAs were negatively selected, indicating that loss of TDG enhances vitamin C efficacy (Figures 2A and 2B).

Using inducible shRNAs targeting TDG in MOLM13 cells, we validated these findings and observed increased 5-fC basal levels, a stronger accumulation of 5-fC formation upon vitamin C treatment, and a reduced IC₅₀ for vitamin C and Olap upon combined treatment in TDG KD cells compared with shRNA control cells (Figures 7A–7D and S7A–S7C). TDG KD was also shown to cause increased S phase cell cycle stalling and p21 induction upon Olap treatment alone and increased γH2AX formation in combination with vitamin C (Figures 7E, 7F, and S7D). TDG KD can therefore mimic and enhance the role of vitamin C to sensitize AML cells to Olap treatment. In addition, using primary murine leukemia cells with Tet2^+/+, Tet2^+/−, and Tet2^−/− genotypes, we show that the increased formation of oxi-mCs generated upon combination treatment with Olap and vitamin C is dependent on Tet2 expression (Figure 7G). We also observed that homozygous deletion of Tet2 or deletion of the Slc23a2 gene, which encodes for the primary vitamin C transporter on hematopoietic cells,²⁰ blocks the effect of Olap and vitamin C combination treatment with respect to cell cycle stalling, γH2AX formation, and p21 induction (Figures S7E–S7G). Slc23a2^−/− primary murine leukemia cells also exhibit increased resistance to combination treatment in CFU and liquid culture viability assays and a reduced capacity for differentiation (Figures S7H–S7K) that correlates with a lack of oxi-mC formation capacity (Figure S7L).

Figure 7. 5-fC accumulation correlates with prolonged S phase and proximity to PARP1-binding and γH2AX sites in PARPi-treated cells.

(A) Relative expression of TDG mRNA in MOLM-13 cells transduced with shRNA targeting TDG (two independent clones, shTDG 820 and shTDG 989) or Renilla luciferase control (shRN). Mean of technical triplicates.

(B) Relative 5-formylcytosine (5-fC) content, normalized to 5-methylcytosine (5-mC), of shTDG or shRN MOLM-13 cells treated for 72 h with LAA or vehicle. n = 3 experiments + SEM.

(C and D) Vitamin C and Olap titration in shTDG MOLM-13 cells.

(C) IC₅₀ of LAA (left) and Olap (right) in shTDG- and shRN-transduced MOLM-13 cells. Mean + SEM, n ≥ 3 experiments.

(D) Dose-response curves of LAA (left) and Olap with or without 250 μM LAA (right) in shTDG and shRN MOLM-13 cells. Mean ± SEM, n ≥ 3 experiments.

(E and F) Flow cytometry of shTDG MOLM-13 cells.

(E) Cell cycle analysis of shRN and shTDG MOLM-13 cells after 72-h treatment with 0.8 μM Olap (Olap) and 250 μM LAA. Mean + SEM, n = 1–4 experiments.

(F) Relative MFI of p21 in LAA- or Olap-treated shRN or shTDG MOLM-13 cells after 72 h. Mean + SEM, n = 4 experiments.

(G) Relative 5-hmC (left) and 5-fC content (right), normalized to 5-mC levels in Tet2^+/+, Tet2^+/−, or Tet2^−/− primary murine MLL-AF9⁺ leukemic cells treated with LAA, Olap, LAA + Olap, or vehicle control for 72 h. Mean + SEM, n ≥ 4 experiments.

(H) Genome-wide 5-mC, 5-hmC, and 5-fC content of MOLM-13 and THP-1 cells treated with or without LAA and with or without Olap for 72 h. Mean +SEM, n ≥ 3 experiments.

(I and J) PARP1 and γH2AX chromatin immunoprecipitation (ChIP)-ELISA of LAA + Olap-treated cells.

(I) Schematic of ChIP-ELISA for PARP1 or γH2AX and 5-mC, 5-hmC, or 5-fC quantitation.

(J) Fold enrichment of 5-mC, 5-hmC, and 5-fC in PARP1 (left) and γH2AX (right) immunoprecipitated DNA normalized to input DNA in LAA + Olap-treated MOLM-13 and THP-1 cells. Mean ± SEM, n ≥ 3 experiments. *p < 0.05, **p < 0.005, ***p < 0.0005, ****p < 0.0005.

In MOLM-13 and THP-1 human AML cells lines, TET2 is the most abundantly expressed TET gene (Figure S2C), and 5-hmC levels increase significantly upon vitamin C treatment alone or in combination with Olap, whereas 5-fC levels rise most abundantly in response to combination treatment (Figures 7H and S7M). Given that PARPi cytotoxicity is attributed to its ability to physically trap PARP proteins onto DNA at sites of BER,³³ we performed chromatin immunoprecipitation (ChIP)-ELISA to measure oxi-mC levels in regions of the genome bound by PARP1 and γH2AX (Figure 7I). No enrichment was observed for 5-mC or 5-hmC; however, significant enrichment of 5-fC was detected in MOLM-13 and THP-1 cells (from 5- to 15-fold) upon combination treatment with vitamin C and Olap compared with vehicle-treated cells (Figure 7J). Overall, these findings suggest that PARPi-mediated DNA trapping at 5-fC sites in the genome leads to increased γH2AX marks at these loci, lengthening of the cell cycle during S phase, and myeloid differentiation that effectively blocks AML cell viability and disease progression.

DISCUSSION

Our studies have uncovered a novel mechanism of synthetic lethality to PARPis, driven by TET2-mediated oxi-mC formation for the treatment of AML. We previously reported transcriptional upregulation of BER genes (including PARP, TDG, and GADD45) upon genetic restoration of Tet2 in mouse HSPCs that could be mimicked pharmacologically by vitamin C treatment of human AML cells.²⁶ These data suggested an increased reliance on the BER pathway to remove TET-mediated oxi-mCs in the genome. Our current genetic and compound library screens validated these findings by showing that the strongest synergistic block on AML cell viability occurs upon vitamin C and PARPi combination treatment, leading to 5-fC accumulation and cell cycle stalling, resulting in greater efficacy at impeding the growth of murine and human AML.

A key mechanism by which vitamin C and PARPis can synergize their therapeutic efficacy for the treatment of AML may be via directed trapping of PARP1 at gene loci enriched for TET-mediated oxi-mCs. All FDA-approved PARPis exhibit strong enzymatic inhibitory activity but vary in their ability to trap PARP onto DNA, which is known to enhance their cytotoxicity.^33,43 In our study, the weaker PARP trappers (veliparib and rucaparib) did not elicit as strong a synthetic lethality in combination with vitamin C compared with Olap and talazoparib, which are known to induce the strongest PARP1 trapping.³³ Targeting PARP proteins to specific genomic loci upon PARPi treatment has not been previously explored and could enhance their anti-cancer activity regardless of HR proficiency. Our study is also the first to show that TDG KD can co-operate with PARPi treatment to reduce AML cell viability. TDG is essential in removal of 5-fC and 5-caC modifications in the genome,³⁷ and TDG KD is known to lead to their accumulation, which was exploited in genome-wide mapping studies in embryonic stem cells to show their enrichment at poised and active enhancers, promoters, and transcriptional start sites of actively transcribed genes.^7,44,45

Formation of 5-fC and 5-caC triggers helical unwinding and DNA-histone cross-links that signal recruitment of DNA repair machinery, which can cause steric interference with DNA replication,^38,44 slow RNA elongation^39,40 and pause DNA polymerase extension.³⁷ TDG is most active prior to S phase entry and at G2/M;⁴⁶ therefore, if TDG does not remove 5-fC bases before initiation of S phase, then their persistence during DNA replication could lead to stalling of replication forks and increased replicative stress marked by γH2AX formation. In a recent study, it was shown, using cell models of post-mitotic neuron and myeloid lineage specification, that knockout of TDG prevents formation of single-strand breaks (SSBs) at sites marked by oxi-mC.⁴⁷ The lack of DNA damage induction seen as SSBs in the context of post-mitotic cells differs from our model, in which AML cells are rapidly cycling. In our study, TDG KD or vitamin C treatment causes increased accumulation of 5-fC bases at sites where repair cannot proceed efficiently upon PARPi treatment. Our data therefore suggest that 5-fC accumulation in the genome, either through ineffective removal caused by TDG deficiency or via vitamin-C-mediated TET activation (enhanced further upon TDG deficiency), creates a synthetic trapping bait for PARP1 upon Olap treatment that could further impede RNA elongation and DNA polymerase extension, leading to stalling of cell cycle progression.

Several recent studies have shown that TET2 deficiency plays a causal role in genomic instability and progression of hematological malignancies, which may enhance their sensitivity to DNA repair inhibitors. Low levels of tyrosyl-DNA phosphodiesterase 1 (TDP1) expression, an enzyme important for removing TOP1 cleavage complexes, has been proposed as the mechanism underlying the selective killing of TET2-mutant HSPCs and AML cells by TOP1i and PARPi.⁴⁸ Another study using AML models of oncogenic tyrosine kinases found that malignant TET2-deficient cells downregulate BRCA1 and LIG4, resulting in reduced activity of HR and DNA-PK-mediated non-homologous end joining (D-NHEJ), respectively, and that TET2-deficient cells rely on PARP1-mediated alternative NHEJ (Alt-NHEJ) for protection from the toxic effects of spontaneous and drug-induced DNA DSBs.⁴⁹ Consistent with these findings, we show that Tet2^−/− pre-malignant HSPCs lose sensitivity to vitamin C in colony formation assays but remain sensitive to PARPi treatment alone. PARP1 recruitment is also involved in stabilization of G quadruplexes and R loops, which are enriched at promoters of actively expressed genes, DNA replication origins, and telomeres.^50–52 TET deficiency has recently been shown to cause elevated levels of G quadruplexes and R loops.^53,54 PARPis may therefore target TET2 mutant AML cells more effectively because of an elevated dependency of PARP1 to stabilize and resolve DNA structures. While many studies have modeled therapeutic vulnerabilities in TET2^−/− cells, most TET2-deficient patients with myeloid malignancies are not devoid of TET2 and instead exhibit haploinsufficiency because of heterozygous loss-of-function mutations.^9,11 Using murine models of AML with Tet2 WT alleles compared with heterozygous or homozygous deletion, we observed sensitivity to PARPi and vitamin C combination therapy even when Tet2 levels were decreased but not completely absent. PARPi treatment in combination with vitamin C could therefore effectively target TET2 WT and mutant cells, provided they maintain a threshold amount of residual TET2 expression.

The differentiation of AML cells observed upon vitamin C and PARPi combination treatment correlated with an increased frequency of cells stalled in S phase rather than the typical G2/M block induced by PARPi alone.³³ Previous studies using MLL-AF9⁺ leukemia cells reported that, above a certain threshold of DNA damage, p21 becomes activated, leading to an extension of S phase and terminal differentiation.³⁶ Furthermore, targeted inactivation of DNA repair using ATMi, or loss of p21, imparts resistance of MLLr⁺ leukemia cells to DNA damage-induced differentiation.³⁶ Increased p21 expression at the RNA and protein level is a feature of normal myeloid maturation,³⁴ and cell cycle lengthening by p21 or p27 upregulation has been proposed as a mechanism that allows myeloid transcriptional regulators the time required to accumulate in high enough levels to drive terminal differentiation.³⁵ Vitamin C and TET2 have been shown independently to activate CDKN1A/p21 expression. In hepatocellular carcinoma and liver cancer stem cells, deletion of the vitamin C transporter SVCT2 blocks p21 induction and cell-cycle arrest in response to vitamin C treatment.⁵⁵Moreover, the upregulation of Cdkn1a/p21 normally induced in WT HSPCs in response to DNA damage caused by ionizing radiation has been reported to be significantly impaired upon tet2 deletion in a zebrafish model.⁵⁶ TET2 also regulates hydroxylation of CDKN1A/p21 and CDKN1B/p27 gene loci in solid cancer cell lines (breast, lung, and melanoma), leading to increased expression and cell cycle arrest of tumor-repopulating cells.⁵⁷ The ability of vitamin C to activate p21 expression either indirectly or via enhancement of TET2 activity in pre-malignant HSPCs and leukemia cells may therefore reduce the aberrant self-renewal that fuels Tet2-deficient myeloid malignancy and promote differentiation, which, combined with DNA damage accumulation, leads to improved efficacy upon PARPi co-treatment.

We have shown that enhancing TET2 enzymatic activity and oxi-mC formation creates an increased reliance on BER and PARP proteins that provides a strong rationale for combining vitamin C with PARPi for the treatment of AML. Future studies should focus on identifying additional targeted therapies that prevent removal of oxi-mCs, bypassing the need to directly enhance TET activity with vitamin C for combination therapies with PARPis. This could be achieved via inactivation of additional BER proteins involved in active DNA demethylation or by trapping readers of oxi-mCs, which, in combination with cell cycle or checkpoint inhibitors, may lead to improved PARPi efficacy for the treatment of AML.

Limitations of the study

The ChIP-ELISA experiments performed in this study suggest that 5fC is enriched in regions of chromatin bound by PARP1 and marked by γH2AX upon vitamin C and PARPi combination treatment. Future studies should include mapping via high-throughput sequencing of 5fC, PARP1, and γH2AX genomic loci at base resolution in response to vitamin C and PARPi treatment. These experiments should be performed with poor and strong PARP trapping inhibitors, such as veliparib and talazoparib, respectively. Furthermore, controls for these genome mapping studies should include TDG KD and TET1–TET3-deficient cells to determine PARPi trapping dependencies at oxi-mC sites in the AML genome.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Resource and reagent requests should be directed to Luisa Cimmino (luisa.cimmino@med.miami.edu).

Materials availability

The study did not generate new unique reagents.

Data and code availability

• RNA-seq data has been uploaded to the Gene Expression Omnibus (GSE217587).

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Mice

All animal protocols were approved by the University of Miami Institutional Animal Care and Use Committee. B6.SJL-Ptprc^a Pepc^b/BoyJ (CD45.1), C57BL/6 J (CD45.2), and NOD.Cg-Prkdc^scid ll2rg^tm1Wjl Tg (CMV-IL3, CSF2, KITLG)1Eav/MloySzJ (NSGS) mice were purchased from Jackson Laboratories. Complete blood counts were acquired on a Heska Element HT5 hematology analyzer. ROSA26-M2rtTA (RTA),⁵⁸ TRE-GFP-Ren (shRN),⁶⁷ TRE-TurboGFP-shTet2 (shT2),^26,68 and Tet2-deficient germ-line mice⁶⁹ were previously described. 1 g/kg doxycycline rodent diet (DOX; Bio-Serv) was used to maintain in vivo shRNA expression.

Retrovirus was generated by polyethylenimine transduction of HEK293-FT cells with MSCV-AML1/ETO-IRES-mCherry (AE9a)²⁶ or pMIG-FLAG-MLL-AF0-IRES-GFP (MA9). pMIG-FLAG-MLL-AF9 was a gift from Daisuke Nakada (Addgene plasmid # 71443; http://n2t.net/addgene:71,443; RRID:Addgene_71443).⁷⁰ Freshly dissected femur and tibia from Tet2^+/+, Tet2^+/−, or Tet2^−/− CD45.2 mice were flushed with PBS using 25G needles. For primary leukemic cell studies, cKit⁺ progenitor cells were isolated from ACK-lysed (Quality Biological, 118–156-101) whole bone marrow by AutoMACS cell selection with mouse CD117 MicroBeads. Isolated cells were transduced with retrovirus using Polybrene and pCL packaging vector.

For in vivo AE9a primary transplants, 6–8wk old female recipient mice were lethally irradiated (8Gy) and transplanted with 50,000 primary AE9a⁺ cells with Tet2^+/+, Tet2^+/−, Tet2^−/−, RTA shRN, or RTA shT2 and 500,000 support CD45.1⁺ wild-type bone marrow. When mice developed leukemia, spleens were harvested, mashed, and filtered through a 70μm filter, and cells were viably frozen.

For Tet2 in vivo restoration experiments, 50,000 splenic tumor cells from terminal primary AE9a RTA shT2-transplanted mice were transplanted with 500,000 support wild-type bone marrow into lethally irradiated recipient mice. Mice were divided into 3 groups: Tet2 knockdown (+DOX, PBS), Tet2 restoration (−DOX, PBS), or ascorbate (+DOX, 4 g/kg ASC). Treatment started either 7 days or 4 weeks post-transplant. Experiments were performed with two biologically independent primary AE9a RTA shT2 tumors (#11 and #12).

For mouse leukemia transplantation studies, 6–8wk old female recipient mice were lethally irradiated, transplanted with 50,000 terminal, splenic AE9a⁺ or 5,000 primary MA9⁺ tumor cells and 500,000 support wild-type bone marrow cells by retroorbital injection. (+)-Sodium L-ascorbate (ASC) was prepared in PBS, dosed 2 g/kg, or 4 g/kg, and administered by intraperitoneal injection (IP). Olaparib was prepared in 10% (2-hydroxypropyl)-β-cyclodextrin in PBS (βcd), dosed 25 mg/kg or 50 mg/kg and co-IP’ed. ASC vehicle control was PBS; Olaparib vehicle control was DMSO. Mice were treated 5 days on, 2 days off for the indicated treatment windows, starting 4 weeks post-transplant. Mice were bled retro-orbitally or submandibular every two weeks to monitor tumor progression.

For patient-derived xenograft (PDX) transplantation studies, 12–16-week-old male NSG-SGM3 (NSGS) mice were injected intravenously with 1×10⁶ primary patient leukemia cells. Treatment began when hCD45 in peripheral blood was ≥1%. Mice were treated daily with 4 g/kg ASC by IP, 25 mg/kg or 50 mg/kg Olaparib by IP or oral gavage (OG), or 0.25 mg/kg Talazoparib (prepared in βcd) by OG. Mice were bled weekly to monitor tumor progression.

Cell lines

Human leukemia cell lines MOLM-13, MOLM-14, NOMO-1, THP-1, U-937, NB4, MV4-11, and KG-1 were grown in RPMI 1640 with L-glutamine and 25mM HEPES supplemented with 10% FBS (Sigma) and 1% Pen-Strep (100U/mL penicillin,100 μg/mL streptomycin), and cultured at 37°C, 5% CO₂. Kasumi-1 and HL-60 were supplemented with 20% FBS, all other conditions being the same. HEK293-FT cells were cultured in DMEM with 10% FBS, all other conditions the same.

Primary cell culture

Total bone marrow cells, cKit⁺ cells, or retrovirus-transduced cells were plated in triplicates with methylcellulose medium (Methocult M3434) supplemented with 1% Pen-Strep. Cells were seeded at 10,000 total bone marrow cells or 2,000 purified cKit⁺ cells per replicate. Colony forming units were enumerated using a microscope and re-plated (2,000 cells/replicate) every 7–10 days. Following the fourth passage (P4) for the colonies with AE9a or MLL-AF9 overexpression, any remaining established colonies from the vehicle groups were given a new passage at 2,000 cells per well with the same treatments plus an additional higher vitamin C concentration to the fifth passage (P4-5) as well as re-plated to the sixth passage (P5-6). Primary mouse hematopoietic progenitor cells from Tet2^+/+, Tet2^+/−, and Tet2^−/− mice in combination with AE9a or MLL-AF9 were cultured in Opti-MEM Reduced Serum Media, GlutaMAX Supplement with 10% FBS, 1% Pen-Strep, 50μM 2-mercaptoethanol, and recombinant murine SCF (50 ng/mL), IL3 (10 ng/mL), and IL6 (10 ng/mL) at 37°C, 5% CO₂. The cell culture was optimized to a final seeding density of 0.5×10⁶ cells/mL. LAA was added directly to methylcellulose or liquid culture medium at a final concentration of 250μM unless otherwise indicated. Final concentrations for PARP inhibitors (PARPi) were 1μM Olaparib or 1nM Talazoparib, unless otherwise stated. For liquid culture growth assays, each primary mouse genotype was treated on a 96-well plate or 24-well plate at a density of 0.2×10⁶ cells/mL for 3 days with a fixed concentration of L-AA (250μM) and with a 5-fold serial dilution of either Olaparib or Talazoparib with a top concentration of 100μM or 1000nM, respectively. Each treatment had 2–3 replicates at each concentration and were compared to a PBS/DMSO control.

For primary human liquid culture differentiation assays, primary patient leukemic cells (UM Biospecimen Resource Center) or wild-type CD34⁺ cells (Fred Hutchinson Cancer Research Center) were thawed in IMDM with 20% FBS (R&D Systems), then washed with FACS buffer and counted. Cells were sorted on live (DAPI−) CD38-CD34+. Sorted cells were cultured on a feeder layer of 80–90% confluent irradiated (30Gy) OP9 stromal cells, in IMDM, 20% FBS, with cytokine supplementation: 100 ng/mL SCF, 10 ng/mL Flt3-Ligand, 20 ng/mL IL-3, 20 ng/mL IL-6, 20 ng/mL GM-CSF, and 20 ng/mL G-CSF. Cells were treated with LAA (250μM or 1mM), Olaparib (1μM), the combinations, or vehicle (PBS and DMSO) controls for 7 days. After 7 days, cells were assayed by flow cytometry for CD14 and CD11b.

METHOD DETAILS

RT-qPCR

Cells were homogenized with QIAshredders, and total RNA extracted using RNeasy Plus Mini Kit. RNA quantities were determined by using NanoDrop8000. For mRNA quantification, total RNA was used for cDNA synthesis using the High-Capacity RNA-to-cDNA Kit. Real-time PCR reactions were carried out using SYBR Select Master Mix on a QuantStudio 5 Real-Time PCR System (Applied Biosystems, A28575). Genes of interest were normalized against Hprt (mouse) or GAPDH (human). See Table S6 for oligonucleotide sequences used.

Flow cytometry

For primary mouse cell flow cytometry analysis, single cell suspensions were prepared from bone marrow, spleen, or peripheral blood, red blood cells were ACK-lysed, and the remaining cells were resuspended in 3% FBS in PBS (FACS buffer). Cells were then incubated with 1/100 mouse Fc block or human Fc block for at least 5 min, then surface marker antibodies were added at 1/250 for at least 15 min in the dark. Cells were stained for cKit, CD11b, and Gr1.

Primary human AMLs and normal CD34⁺ cells were labeled with CD34 and CD38 antibodies and double sorted for CD34⁺CD38⁻ on a BD FACS SORP Aria Fusion to >95% purity. CD14 and CD11b were used as markers for differentiation for human AML lines, primary patient samples, and normal CD34⁺ cells.

For intracellular staining, antibodies against 5-hmC and γH2AX (BioLegend) were used either overnight at 4°C or 1 h at room temperature, after paraformaldehyde fixation and membrane permeabilization. Stained cells were quantified using CytoFLEX LX Flow Cytometer (Beckman Coulter). FlowJo 10 software was used to generate flow cytometry plots.

For Annexin V staining, cells were cultured 72hr, then washed twice with PBS, washed once with diluted Annexin V binding buffer (diluted to 1X with water), and then resuspended in staining buffer (1/25 Annexin V and 1/250 DAPI in diluted binding buffer). Cells were incubated for 15–30 min in the dark at room temperature before flow cytometry.

(Oxidized-) methylcytosine quantification

Commercially available 5mC, 5-hmC, and 5-fC ELISAs (Epigentek) were used to quantify the levels of methylcytosine and oxidized methylcytosine in whole genomic and ChIP’ed DNA samples following the manufacturer’s protocol. For whole genomic DNA, 100mg DNA per replicate was used for 5mC and 5hmC ELISAs, and 300μg DNA per replicate for 5fC ELISAs. For ChIP samples, 50ng DNA per replicate was used. Each sample was assayed in technical duplicate or triplicate.

Histology & immunocytochemistry

Dissected mouse tissues were fixed overnight, at minimum, in 10% formalin. Tumor sections were embedded in paraffin, and thin histologic sections were prepared. Tissue sections were deparaffinized in xylene and rehydrated in ethanol following treatment in pre-heated target retrieval solution. Serum-free blocking solution was then applied for 40 min at RT, the slides were stained with hematoxylin and eosin, dried, and mounted with Permount. Micrographs of morphology were captured using an Olympus BX43 dualhead light microscope equipped with an Olympus Q-Color 5 digital camera (Olympus America) at 400x magnification.

For human AML cell lines, 5×10⁴ cells were washed once with PBS, attached to slides by Cytospin (Thermo Scientific), stained for morphology with the Easy III Stain Kit following manufacturer’s protocol, and visualized with an EVOS XL Imaging System (Life Technologies).

For immunocytochemistry, 5×10⁴ cells were fixed with 2% formaldehyde in PBS for 10 min at room temperature (RT), washed again with PBS and bound to poly-L-lysine coated slides. After a 5 min PBS wash, cells were incubated with Blocking Solution (1 mg/mL BSA, 3% goat serum, 0.1% Triton X-100 in PBS) for 30 min at RT. Cells were then incubated with primary antibodies in Blocking Solution for 1 h at RT, washed three times with PBS, and incubated with Alexa 488 or 594 coupled secondary antibodies for 30 min prior to staining with 0.5 μg/mL DAPI in PBS for 5 min. Slides were washed with PBS for 2 min, air dried, and mounted in SlowFade Diamond antifade mounting reagent. Foci counted manually from at least 300 cells per treatment. Antibodies to the following proteins were used in this study: γ-H2AX (Ser 139) clone JBW301 (Millipore), RAD51 (B-Bridge International).

Transcriptome analysis

Splenic tumor cells from terminal primary AE9a shT2, and terminal secondary AE9a shT2 KD, RS, and ASC-treated mice were sort-purified. MOLM-13 and THP-1 cells were treated with vehicle, 250μM LAA, 1μM Olaparib, or LAA + Olaparib for 72 h. Total RNA was isolated from QIAshredder-homogenized cells using the RNeasy Plus Mini Kit. RNA quantity and quality were determined using a Nanodrop 8000 (Thermo) and a 2100 Bioanalyzer (Agilent). Mouse tumor RNA was sequenced at NYU Genome Technology Center, and cell line samples were sent to Novogene for library preparation and sequencing. mRNA was purified using poly-T oligo-attached beads, fragmented, and cDNA amplified using random hexamer primers. Mouse RNA libraries were prepared using NEXTflex RNA-Seq Kit; human RNA libraries were generated with NEBNext Ultra RNA Library Prep Kit for Illumina. Library was evaluated by Qubit, qPCR, and bioanalyzer. Mouse samples were sequenced on an Illumina HiSeq 2500 using 50bp single-end reads; for human samples, cluster generation and 150bp paired-end sequencing was performed on an Illumina NovaSeq 6000. Reads were filtered for adapter contamination and low quality (Illumina Casava 1.8). Primary mouse sample reads were aligned to the genome (mm10/GRCm38) using STAR aligner⁶² and human cell line sample reads were mapped to the genome (hg38) with HISAT2.⁶³ For both datasets, reads were counted with featureCounts,⁶⁴ while DESeq2⁶⁵ was used to determine differential expression and the resulting p values were adjusted using the Benjamini and Hochberg approach for false-discovery control. Principal component analysis was performed with limma⁶⁶ and heatmaps of Z score row-normalized FPKM were generated with heatmap.2 from gplots.⁷¹

Gene set enrichment analysis

Gene set enrichment analysis (GSEA) was performed using Enrichr.^72–74 GSEA was also performed against the 10-day Tet2-restoration gene signature,²⁶ using GSEA software^75,76 with gene set as permutation type, 1,000 permutations and log2 ratio of classes. Gene pathways and functions in the CRISPR screen were assessed using Ingenuity Pathway Analysis (QIAGEN Bioinformatics).

In vitro culture treatments

L-Ascorbic Acid (LAA) was prepared in PBS. The following compounds were solubilized in DMSO: Olaparib, Rucaparib, Veliparib, Talazoparib, Pinometostat, Venetoclax, Vorinostat, TPCA-1, and C75. PBS and DMSO were used for vehicle control treatments.

CRISPR knockout screen

MOLM-13 cells were transduced with lentiCas9-Blast, and Cas9-competent cells were selected for using 10μg/mL Blasticidin. lentiCas9-Blast was a gift from Feng Zhang (Addgene plasmid # 52962; http://n2t.net/addgene:52,962; RRID: Addgene_52962). 10⁹ Cas9-competent MOLM-13 were then infected with the Brunello sgRNA library at an MOI of 0.1 and selected for using 1μg/mL Puromycin for 5 days, then grown without antibiotic for 2 days. 10⁸ library-infected cells were then treated with 2mM LAA or PBS control for 16 days. Experiment was performed in duplicate using two independent infections. Genomic DNA was then extracted, and guideRNA amplified and sequenced as previously published.⁷⁷ Data analysis and hit calling were performed using MAGeCK.⁵⁹ Significant positively- and negatively-selected genes were analyzed by Ingenuity Pathway Analysis (QIAGEN).

Compound library screen and viability assays

MOLM-13, THP-1, and Kasumi-1 were seeded at 4000 cells per well in 20μL media in 384-well plates and cultured for 24 h. 5μL LAA (or PBS control) was then added to 250μM final concentration, and 5μL library test compounds (or DMSO control) were added to final concentration of 1μM. Cells were returned to culture for 72 h. Each treatment was performed in duplicate. Equal volume CellTiter-Glo (CTG) was added, and cells were incubated for 10 min prior to reading. CTG data was analyzed by normalizing drug effect against a 10μM botezomib control. Effect of drug + LAA was subtracted from drug effect without LAA, then ranked. Cell lines were analyzed separately, and then averaged together and drugs were ranked by differential effect.

All other viability assays were performed in 96-well plates, with cells seeded at 0.2×10⁶ cells/mL. LAA (or PBS control) and drugs (or DMSO control) were added to indicated final concentrations and cells were cultured for 72 h. Cells were then mixed 50μL:50μL with CTG reagent, incubated in the dark for 15 min, and luminescence was read on a Synergy 2 plate reader (Biotek).

Cell cycle analysis

For basic cell cycle profiling, 4% paraformaldehyde (PFA) fixed and permeabilized cells were stained with 1μg/mL DAPI (Sigma, D9542) for at least 15 min before flow cytometry. For more granular cell cycle profiling, Click-iT Plus EdU Kit (Thermo, C10634) was used. Briefly, EdU was added directly to the culture media at a final concentration of 10μM, mixed, and cells were returned to incubator for 1 h. Afterward, cells were PFA fixed, permeabilized, transferred to 96-well plates, and stained with 1/1000 γH2AX antibody (BioLegend) overnight. The following day, the cells were washed then stained with the fluorescent dye picolyl azide per manufacturer’s protocol, except at ¼ volume. Cells were washed one more time then stained with DAPI before flow cytometry. For BrdU-pulse, EdU-chase experiments, cells were cultured for 1hr with 10μM BrdU then washed and resuspended in fresh media with indicated treatments. After 6, 24, 48, and 72hr, cells were harvested, EdU added to 10μM, and returned to incubator for 1hr. Cells were then PFA fixed, permeabilized, and resuspended in 150μg/mL DNase in PBS and incubated at 37°C for 1hr. Cells were next diluted with wash buffer, resuspended in fresh wash buffer with 1/200 anti-BrdU and 1/1000 anti-γH2AX (BioLegend), and incubated overnight at 4°C on a rocker. The next day, cells were washed, Click-EdU stained as described above, and incubated with DAPI for 30min prior to flow cytometry.

Comet assay

Comet Assay Kit (Cell Biolabs) used according to manufacturers’ protocol. Briefly, 1% low-gel temperature agarose melted and kept in a molten state at 37°C. 75μL agarose pipeted onto slide wells and chilled 15 min at 4°C to create a base layer. Cells were harvested, washed with chilled PBS, and resuspended to 10⁵ cells/mL in chilled PBS. Cells were then mixed 1:10 with molten agarose, 75μL pipetted on top of the base layer, and chilled for 15 min at 4°C. Slides were then submerged 60min in chilled lysis buffer, 30min in chilled alkaline solution, then electrophoresed in chilled alkaline electrophoresis solution for 30 min at 1 V/cm at 300mA. Slides were washed three times with chilled d.i. H₂O, fixed with cold 70% ethanol for 5 min, and air dried. Slides were stained with 100μL/well of 0.1mg/mL propidium iodide and incubated at least 15 min then imaged on a Thunder Imaging System (Leica). Comet tail moment was quantified using the ImageJ plugin OpenComet⁶¹; poorly fit comets were manually excluded. 47–126 cells per treatment counted for THP-1, and 9–40 for MOLM-13.

Cell density

Cell density data was acquired from cells treated with LAA + Olaparib for 72hr. Cells were sampled on a Vi-CELL BLU Cell Viability Analyzer (Beckman). 100 images acquired per sample, diameter 6–30μm, cell sharpness 7, minimum circularity, medium decluster degree, 3 aspiration and mixing cycles, 50% viable spot brightness, and 5% viable spot area.

TDG knockdown

Two SMARTvector Inducible Human TDG PGK-TurboGFP shRNA clones targeting the ORF and shRN were transduced with lentivirus into MOLM-13 cells and selected using puromycin to generate cell lines stably carrying the desired shRNA. Transduced, selected cells were cultured in standard media supplemented with 1μg/mL doxycycline to induce expression of the shRNA.

Dot Blot analysis

Genomic DNA from treated cells was isolated by phenol-chloroform extraction, ethanol precipitation, and quantified on a NanoDrop 8000 (Thermo Scientific). 2μg (100μL) DNA denatured with 10μL 10X denaturing solution (4M NaOH, 100mM EDTA) and incubated at 95°C for 10 min. Samples were immediately transferred to ice and neutralized with 110μL 2M ammonium acetate, pH 7.0 neutralizing solution. DNA was transferred to a 96-well plate and serially diluted 1:2 into cold Buffer AE. A nylon membrane was sized to a microplate, pre-wet with 6X saline sodium citrate (SSC) and sandwiched into a vacuum-connected Bio-Dot apparatus (Bio-Rad, 1,706,545). Membrane was rehydrated with diH₂O, DNA was applied, washed with 2X SSC, and cross-linked in a UV crosslinker at 1200×100μJ. The membrane was pre-blocked with 5% milk in TBS-T (blocking buffer) for 1 h at RT, then incubated with 1/1000 primary antibody in blocking buffer for 1–2hours at RT or overnight at 4°C with gentle rocking. The membrane was then washed three times with TBS-T for 10 min each, then incubated with 1/3000 HRP-conjugated antibody (anti-mouse for 5mC, and anti-rabbit for 5hmC) in blocking solution for 1 h at RT. The membrane was washed again in TBS-T thrice for 10 min each, then incubated for 5 min at RT with a chemiluminescence mixture. The membrane was exposed on Blue Lite Autorad Film. As a loading control, the membrane was then stained with a solution of 0.02% methylene blue in 0.3M sodium acetate, pH 5.2 for 1–2 min and washed 3–5 times for 2–5 min each with diH₂O. Densitometry was quantified using the FIJI ImageJ package⁶⁰ and the Protein Array Analyzer macro (G Carpentier, 2008; http://rsb.info.nih.gov/ij/macros/toolsets/ProteinArrayAnalyzer.txt). Antibody densitometry was normalized against methylene blue intensity, then averaged across dilutions within a linear detection range.

Chromatin immunoprecipitation (ChIP)-ELISA

8 million MOLM-13 or THP-1 cells per condition were treated for 72 h in 40mL media. ChIP was performed as previously described,⁷⁸ with some modifications. Briefly, after treatment, cells were directly crosslinked in media for 10 min with 1/10 11% formaldehyde solution, which was then quenched for 5 min with 1/20 2.5M glycine. Cells were washed twice with cold PBS and flash frozen or directly prepared for sonication. Prior to sonication, chromatin was isolated by washing with three separate lysis buffers (LB). Cells were suspended in 2mL LB3 and sonicated in 15mL tubes with sonication beads (Diagenode, C01020031) in a Bioruptor Pico (Diagenode). Chromatin was fragmented to 200–1000bp fragments with 15 cycles of 30sec on/30sec off. Sonicated lysate was spun in a 4°C microcentrifuge at max speed for 15min, and supernatant containing fragmented chromatin was transferred to a DNA LoBind tube. 50μL of sonicated chromatin was purified for DNA quantification and fragmentation visualization. 2μg anti-PARP1 antibody (Abcam, ab227244) or 5mg anti-γH2AX antibody (Millipore, 05–636) were bound to 50μL Dynabeads Protein G (Invitrogen, 10003D), rotating overnight at 4°C. Next day, the beads were mixed with 100mg chromatin (PARP1-ChIP) or 50mg chromatin (γH2AX-ChIP) in a 2mL reaction volume. 1% input was reserved, and the remainder immunoprecipitated overnight at 4°C. The following day, the ChIP samples were washed 6 times with RIPA buffer, and once with a TE-NaCl buffer before decrosslinking both the ChIP and input samples at 65°C for 3 h. DNA was then phenol-chloroform extracted and ethanol precipitated. DNA was quantified by NanoDrop 8000 and (oxidized) methylcytosine content was assayed by ELISA as described above.

QUANTIFICATION AND STATISTICAL ANALYSIS

IC₅₀ values were calculated by nonlinear regression in GraphPad Prism 9 software. Synergy scores were calculated using SynergyFinder 2.0.⁷⁹ All p values were calculated using unpaired two-tailed Student’s t test or ANOVA. Analyses were not performed under specific randomization or blinding protocol. Statistically significant differences are indicated with asterisks in figures with the accompanying p value in the legend. Error bars in figures indicate standard deviation (SD) or standard error of the mean (SEM) for the number of replicates, as indicated in the figure legend.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Purified Rat Anti-Mouse CD16/CD32 (Mouse BD Fc Block™) (2.4G2)	BD Biosciences	Cat# 553142, RRID:AB_394657
Human BD Fc Block™ (Fc1)	BD Biosciences	Cat# 564220, RRID:AB_2869554
anti-mouse CD117 (c-Kit) APC (2B8)	BioLegend	Cat# 105812; RRID:AB_313221
anti-mouse/human CD11 b PE/Cyanine7 (M1/70)	BioLegend	Cat# 101216; RRID:AB_312799
anti-mouse Ly-6G/Ly-6C (Gr-1) APC/Cyanine7 (RB6-8C5)	BioLegend	Cat# 108424; RRID:AB_2137485
anti-human CD34 APC (8G12)	BD Biosciences	Cat# 345804, RRID:AB_2686894
anti-human CD38 PE/Cyanine7 (HIT2)	BioLegend	Cat# 303516; RRID:AB_2072782
anti-human CD14 APC (63D3)	BioLegend	Cat# 367118; RRID:AB_2566792
anti-human/mouse CD11b FITC (M1/70)	BioLegend	Cat# 101206; RRID:AB_312789
anti-5-Hydroxymethylcytosine (5-hmC) antibody (pAb)	Active Motif	Cat# 39769; RRID:AB_10013602
anti-human/mouse H2A.X Phospho (Ser139) PE/Cyanine7 (2F3)	BioLegend	Cat# 613419; RRID:AB_2715784
anti-human/mouse phospho-Histone H2A.X (Ser139) (JBW301)	Millipore Sigma	Cat# 05-636; RRID:AB_309864
anti-human RAD51 antibody	B-Bridge International	Cat# 70-001; RRID:AB_2177110
anti-human Recombinant p21 Alexa Fluor® 488 [EPR362]	abcam	ab282187
anti-human PARP1	abcam	ab227244
CD117 MicroBeads, mouse	Miltenyi Biotech	Cat# 130-091-224; RRID:AB_2753213
Goat anti-Rabbit IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor™ Plus 488	Invitrogen	Cat# A32731; RRID:AB_2633280
Donkey anti-Mouse IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor™ Plus 594	Invitrogen	Cat# A32744; RRID:AB_2762826
anti-5-Methylcytosine (5-mC) Monoclonal Antibody [33D3]	Epigentek	Cat# A-1014; RRID:AB_2819207
anti-mouse IgG, HRP-linked Antibody	Cell Signaling Technology	Cat# 7076, RRID:AB_330924
anti-rabbit IgG, HRP-linked Antibody	Cell Signaling Technology	Cat# 7074; RRID:AB_2099233
anti-mouse Recombinant p21 Alexa Fluor® 647 [EPR18021]	abcam	ab237265
Biological samples
Primary human AML samples	Sylvester Comprehensive Cancer Center’s Biospecimen Shared Resource	https://umiamihealth.org/en/sylvester-comprehensive-cancer-center/research/research-resources/shared-resources/biospecimen-shared-resource
Patient-derived xenografts (PDX)	Sylvester Comprehensive Cancer Center’s Cancer Modeling Shared Resource	https://umiamihealth.org/en/sylvester-comprehensive-cancer-center/research/research-resources/shared-resources/cancer-modeling-shared-resource
Chemicals, peptides, and recombinant proteins
Polyethylenimine	Polysciences	23966
ACK Lysis buffer	Quality Biological	118-156-101
Polybrene	Millipore Sigma	TR-1003-G
(+)-Sodium L-ascorbate	Millipore Sigma	A4034
Olaparib	Selleck Chemicals	S1060
(2-Hydroxypropyl)-β-cyclodextrin	Millipore Sigma	389145
Talazoparib	MedKoo Biosciences	204710
10% formalin	VWR	89370–094
Easy III™ Stain Kit	Azer Scientific	ES902
4′,6-Diamidino-2-phenylindole dihydrochloride (DAPI)	Millipore Sigma	D9542
SlowFade™ Diamond Antifade Mountant	Invitrogen	S36967
L-Ascorbic acid	Millipore Sigma	A4544
Rucaparib	ApexBio	A4156
Veliparib	Selleck Chemicals	S1004
Pinometostat	MedChem Express	HY-15593
Venetoclax	MedChem Express	HY-15531
Vorinostat	MedChem Express	HY-10221
TPCA-1	MedChem Express	HY-10074
C75	MedChem Express	HY-12364
Dimethyl sulfoxide (DMSO)	Millipore Sigma	D2650
2-Mercaptoethanol	Gibco	21985023
Recombinant Murine SCF	PeproTech	250–03
Recombinant Murine IL-3	PeproTech	213–13
Recombinant Murine IL-6	PeproTech	216–16
Recombinant Human SCF	PeproTech	300–07
Recombinant Human Flt3-Ligand	PeproTech	300–19
Recombinant Human IL-3	PeproTech	200–03
Recombinant Human IL-6	PeproTech	200–06
Recombinant Human GM-CSF	PeproTech	300–03
Recombinant Human G-CSF	PeproTech	300–23
Etoposide	Sigma Aldrich	E1383
Propidium Iodide - 1.0 mg/mL Solution in Water	Invitrogen	P3566
Paraformaldehyde	Thermo Scientific	43368-9M
Formaldehyde	Thermo Scientific	28908
Saline sodium citrate	Quality Biological	351-003-131
Critical commercial assays
QIAshredders	QIAGEN	79656
RNeasy Plus Mini Kit	QIAGEN	74136
High-Capacity RNA-to-cDNA Kit	Applied Biosystems	4387406
SYBR™ Select Master Mix	Applied Biosystems	4472908
MethylFlash Global DNA Methylation (5-mC) ELISA Easy Kit (Colorimetric)	Epigentek	P-1030-96
MethylFlash Global DNA Hydroxymethylation (5-hmC) ELISA Easy Kit (Colorimetric)	Epigentek	P-1032-96
MethylFlash 5-Formylcytosine (5-fC) DNA Quantification Kit (Colorimetric)	Epigentek	P-1041-96
NEXTflex RNA-seq Kit	Bioo Scientific	5129–02
NEBNext® Ultra™ RNA Library Prep Kit for Illumina	NEB	E7530L
CellTiter-Glo	Promega	G7572
Click-iT™ Plus EdU Alexa Fluor™ 647 Flow Cytometry Assay Kit	Invitrogen	C10634
FITC BrdU Flow Kit	BD Bioscience	559619
Annexin V APC	BioLegend	640920
Annexin V Binding Buffer, 10X	BD Bioscience	556454
Comet Assay Kit	Cell BioLabs	STA-350
Deposited data
Raw and analyzed data (murine AE9a RNA-seq)	This paper	GSE217586
Raw and analyzed data (human cell line RNA-seq)	This paper	GSE217585
Mouse reference genome	Genome Reference Consortium	https://www.ncbi.nlm.nih.gov/data-hub/genome/GCF_000001635.20/
Human reference genome	Genome Reference Consortium	https://www.ncbi.nlm.nih.gov/data-hub/genome/GCF_000001405.26/
Mouse gene annotation	Ensembl	https://nov2020.archive.ensembl.org/Mus_musculus/Info/Annotation
Human gene annotation	Ensembl	https://useast.ensembl.org/Homo_sapiens/Info/Annotation
Experimental models: Cell lines
MOLM-13	DSMZ	ACC 554
MOLM-14	DSMZ	ACC 777
NOMO-1	DSMZ	ACC 542
THP-1	ATCC	TIB-202
U-937	ATCC	CRL-1593.2
NB4	DSMZ	ACC 207
MV4-11	ATCC	CRL-9591
KG-1	ATCC	CCL-246
KASUMI-1	ATCC	CRL-2724
HL-60	ATCC	CCL-240
OP9	ATCC	CRL-2749
HEK293-FT	Invitrogen	R70007
Experimental models: Organisms/strains
Mouse: B6.SJL-Ptprca Pepcb/BoyJ (CD45.1)	The Jackson Laboratory	Strain #: 002014
Mouse: C57BL/6J (CD45.2)	The Jackson Laboratory	Strain #: 000664
Mouse: NOD.Cg-Prkdcscid Il2rgtm1Wjl Tg(CMV-IL3,CSF2,KITLG)1Eav/MloySzJ (NSGS)	The Jackson Laboratory	Strain #: 013062
Mouse: B6.Cg-Gt(ROSA)26Sortm1(rtTA*M2)Jae/J (Rosa26-M2rtTA)	The Jackson Laboratory	Strain #: 006965
Mouse: TRE-shTet2	Cimmino et al. 201726	N/A
Mouse: TRE-shRen	Dow et al. 201458	N/A
Oligonucleotides
See Table S6 for oligonucleotide sequences used in this paper		N/A
Recombinant DNA
lentiCas9-Blast	Addgene	52962
shRenilla	Dow et al. 201458	N/A
SMARTvector Inducible Human shTDG 820	Horizon Discovery Biosciences	V3IHSPGG_4832820
SMARTvector Inducible Human shTDG 989	Horizon Discovery Biosciences	V3IHSPGG_5815989
AML-ETO9a-IRES-mCherry	Cimmino et al. 201726	N/A
pMIG-FLAG-MLL-AF9	Addgene	71443
Software and algorithms
MAGeCK	Li et al. 201459	https://github.com/liulab-dfci/MAGeCK
FIJI	Schindelin et al. 201260	https://imagej.net/software/fiji/
Protein Array Analyzer	G Carpentier 2010	http://image.bio.methods.free.fr/ImageJ/?Protein-Array-Analyzer-for-ImageJ
Open Comet	Gyori et al. 201461	http://cometbio.org/
GraphPad Prism v9	GraphPad Software, LLC	https://www.graphpad.com/scientific-software/prism/
FlowJo v10	Becton, Dickinson and Company	https://www.flowjo.com/solutions/flowjo
STAR aligner	Dobin et al. 201362	https://github.com/alexdobin/STAR
HISAT2	Kim et al. 201963	http://daehwankimlab.github.io/hisat2/
featureCounts	Liao et al. 201464	https://subread.sourceforge.net/featureCounts.html
DESeq2	Love et al. 201465	https://genepattern.github.io/DESeq2/v1/index.html
limma	Ritchie et al. 201566	https://kasperdanielhansen.github.io/genbioconductor/html/limma.html
Other
Methocult M3434	Stem Cell Technologies	03434
Rodent Diet, Grain-Based, Doxycycline (1.0 gm/kg), Green, 1/2″ Pellets-Sterile	Bio-Serv	Custom
Buffer AE	QIAGEN	19077
Nylon membrane	GE Healthcare	RPN303B
Blue Lite Autorad Film	Genemate	490001-950

Highlights.

• Vitamin C synergizes with PARPis, causing S phase stalling and AML differentiation

• Combination treatment efficacy is dependent on TET2 expression in AML

• PARPi treatment with vitamin C causes 5fC accumulation in the genome of AML cells

• PARPi treatment with vitamin C enriches for PARP1 binding and γH2AX at 5fC sites