Poly(ADP-ribose) Glycohydrolase Inhibition Sequesters NAD⁺ to Potentiate the Metabolic Lethality of Alkylating Chemotherapy in IDH-Mutant Tumor Cells

Gliomas characterized by mutations in the isocitrate dehydrogenase (IDH) genes IDH1 and IDH2 are the most common primary brain cancer of adults in the third, fourth, and fifth decades of life (1, 2). Typically presenting as lower-grade lesions within the World Health Organization diagnostic classification (grades II and III; refs. 3, 4), these infiltrative tumors are slow-growing but cause progressive neurologic morbidity in patients. Alkylating chemotherapy has been proven effective for patients with IDH-mutant gliomas, with extended survival demonstrated in international randomized clinical trials combining radiation with the procarbazine, lomustine, and vincristine (PCV) regimen (5, 6) or the oral chemotherapeutic temozolomide (7). However, though effective, these therapies are not curative, and strategies to improve treatment are needed.

Mutant IDH enzymes catalyze the abnormal overproduction of D-2-hydroxyglutarate (2-HG) from alpha-ketoglutarate, resulting in substantial metabolic derangements in cancer cells, including low basal levels of NAD⁺ (8). Many studies have shown that these metabolic derangements have created selective susceptibilities in IDH-mutant cancers (9–14). In addition, recent work has shown that DNA damage response and Poly(ADP-ribose) polymerase (PARP) signaling is a susceptibility in IDH-mutant cancer cells (15–18), which display sensitivity to PARP inhibitors (PARPi; ref. 19).

As a key component of the cellular metabolic response to alkylating chemotherapy, PARP activity is acutely upregulated upon chemotherapeutic exposure (20), transiently consuming NAD⁺ through enzymatic polymerization of monomeric NAD⁺ into PAR chains (21). This PARylation signal then recruits DNA-repair machinery to the sites of chemotherapeutic-induced DNA damage (22–24). Unique to IDH-mutant gliomas, which have low basal levels of NAD⁺, widespread activation of PARP causes available NAD⁺ to be critically depleted after alkylating chemotherapy, exposing a transient metabolic vulnerability (25).

Importantly, PARylation is regulated not only by PARPs, but also by PAR glycohydrolase (PARG). PARG is the primary enzyme that mediates PAR breakdown, by cleaving the glycoside bonds of PAR to release mono ADP-ribose (26). Therefore, PARG inhibition after alkylating chemotherapy could, in theory, result in hyperaccumulation of PAR by blocking enzymatic degradation, with the simultaneous arrest of PAR recycling to NAD⁺ leading to catastrophic collapse of free NAD⁺ levels (27). We hypothesized that, for IDH-mutant gliomas, a therapeutic effect could be maximized by combining alkylating chemotherapy with PARG inhibition, causing both DNA damage and scarcity of available monomeric NAD⁺.

Here, we report that the combination of temozolomide and PARG inhibitors is highly effective against IDH-mutant gliomas. We show that, as hypothesized, temozolomide treatment promotes PARP activation and outflow consumption of cellular NAD⁺ pools, while PARG inactivation then freezes this NAD⁺ as polymerized PAR by blocking subsequent breakdown. This resulting state is lethal in IDH-mutant tumor cells in both in vitro and in vivo models.

We first treated a panel of patient-derived endogenous IDH1-mutant and wild-type glioma lines, as well as the IDH1-mutant fibrosarcoma line HT1080, with the small-molecule PARPi olaparib. PARP inhibition displayed a substantial effect as monotherapy in the IDH1-mutant lines HT1080 and TS603 (Fig. 1A; Supplementary Fig. S1A), and this cytotoxic effect was further augmented by the addition of temozolomide. Both findings are consistent with prior reports (16, 17). In several patient-derived IDH1-mutant glioma lines, however, including MGG152, MGG119, and BT142, the effect of PARPi monotherapy was weaker, and augmentation of that effect with temozolomide was muted. As a control, we tested a panel of IDH wild-type glioma lines and, consistent with recent reports (28, 29), also observed a range of responses to PARPi monotherapy and combination with temozolomide (Fig. 1A; Supplementary Fig. S1A). Although some wild-type glioma lines like Hs683 and T98G were sensitive to PARPi monotherapy and PARPi + temozolomide (Supplementary Fig. S1A and S1B), normal human astrocyte (NHA) cells used as controls were relatively insensitive to either treatment, at doses across the physiologic range (Supplementary Fig. S1C). These data suggest that the association between IDH mutation and PARP-mediated synthetic lethality in cancer cells may be context-specific, depending on additional genomic or metabolite differences in different cancer lines.

To explore this finding further, we performed genomic profiling of our cell line panel; however, these analyses did not reveal an alteration that consistently predicted PARPi responsiveness (Fig. 1B). MGMT is methylated in the majority of lines, although cases where it is unmethylated (i.e., HT1080, TS603) could potentially account for differences in the effect of combination temozolomide and olaparib treatment compared with relative insensitivity to temozolomide monotherapy. Given the utilization of NAD⁺ as a PARP substrate, we then considered the metabolic effects of NAD⁺ levels on PARPi monotherapy across our panel of cancer cell lines. To assess the relationship between PARP inhibition and NAD⁺ consumption, we noted that, at transient time points (6 hours), temozolomide treatment often caused detectable NAD⁺ depletion in IDH-mutant cell lines (Fig. 1C), an effect not seen in wild-type lines that have higher levels of basal NAD⁺ (8). This depletion was reversed by PARPi (Fig. 1C), recovering at 24 hours in all cases. Because of the lower basal levels of NAD⁺, IDH-mutant cells are vulnerable to treatment with inhibitors of the NAD⁺ biosynthesis enzyme NAMPT (8). Indeed, in prior studies we have observed that NAMPT inhibitor plus temozolomide had an additive growth-inhibitory effect in IDH-mutant cells (25). Interestingly, however, it has been noted that PARPi synthetic lethality is mediated through amplification of DNA damage response, as a selective vulnerability in IDH-mutant cancer cells, and not dependent upon NAD⁺ (16). To assess this molecular mechanism, we therefore tested the effect of PARP1 overexpression. We found that PARP1 overexpression does lead to PAR hyperaccumulation and cytotoxicity in IDH1-mutant cells (Fig. 1D and E). Not surprisingly, measurement of NAD⁺ levels revealed they were markedly depleted. Furthermore, attempted rescue with the NAD⁺ precursor nicotinamide mononucleotide (NMN) was only minimally effective when compared with NMN rescue of NAMPT inhibitor (ref. 8; Fig. 1F), consistent with robust PARylation signaling as a primary mechanism of effect.

We also tested the potential contribution of the lowered availability of substrate NAD⁺ for PARP consumption, which could render a pass-through mechanism of “pseudo-PARP inhibition” by NAMPT inhibition. We therefore considered whether PARPi would additively promote the cytotoxicity of NAMPT inhibitors in IDH-mutant gliomas. Notably, we observed the opposite—PARPi treatment consistently reversed the cytotoxic effect of NAMPT inhibition at 24 and 48 hours (Fig. 1G; Supplementary Fig. S1D), potentially highlighting a distinct metabolic weakness of depleted basal NAD⁺ in these tumors, separable from PARP-mediated DNA damage signaling.

Thus, our working model for understanding PARPi-mediated cytotoxicity in susceptible cancer cells is centered on the amplification of DNA damage response by PARP enzymes through PARylation trapping of available NAD⁺ (Fig. 1H). The complex interplay between PARP inhibition, PAR recycling and NAD⁺ equilibrium led us to consider whether this repair cycle could be intercepted in IDH-mutant gliomas at a different point, maximally exploiting both vulnerabilities.

We hypothesized that a therapeutic effect could be optimized by targeting PAR breakdown via PARG inhibition, thereby combining the effect of PAR-mediated damage signaling with an arrest of NAD⁺ captured within PAR to simultaneously drive metabolic stress. In a test of this hypothesis, across our panel of IDH-mutant cancer cell lines, we observed that cytotoxicity was significantly and consistently induced by a combination of temozolomide with the selective small-molecule PARG inhibitor (PARGi) PDD00017273 (hereafter abbreviated as PDD), at doses of PDD that were only modestly cytotoxic as monotherapy (Fig. 2A; Supplementary Fig. S2A). Conversely, we did not observe augmentation of temozolomide cytotoxicity by PDD in 5 of 7 IDH wild-type lines, nor did we observe this effect in NHA cells (Fig. 2B). We further observed that treatment with PDD augmented the temozolomide-mediated inhibition of clone formation of IDH-mutant HT1080 cells, but not IDH wild-type U251 cells (Fig. 2C). Analogously, the temozolomide and PDD combination suppressed sphere formation in IDH-mutant neurosphere lines MGG152 and MGG119, compared with an IDH wild-type control line (Fig. 2D). We used a live cell protease activity assay as a control in parallel with an ATP-based cell viability assay to demonstrate consistency across assays, confirming cell death from combined temozolomide and PDD treatment (Supplementary Fig. S2B).

To assess whether this PDD cytotoxic effect was “on-target,” we engineered IDH-mutant HT1080 and TS603 and IDH wild-type U251 PARG knockout (KO) cells using two independent CRISPR guide RNAs (Genscript gRNA #2 and #3; Fig. 2E; Supplementary S2C). Baseline PAR levels were increased, and temozolomide treatment strongly induced PAR accumulation in PARG KO HT1080 cells compared with nontargeting control gRNA–transduced cells (Fig. 2E). We also observed lower basal NAD⁺ and NADH levels in HT1080 PARG KO lines (Fig. 2F; Supplementary Fig. S2D). Temozolomide treatment potently inhibited cell viability in both PARG KO IDH-mutant cell lines, but not in the control PARG KO IDH wild-type cell line (Fig. 2G; Supplementary Fig. S2E), an effect analogous to those seen above with combined temozolomide and PDD. Of note, baseline cell growth was somewhat slower in HT1080 PARG KO than control (Supplementary Fig. S2F), suggestive of lowered fitness for PARG KO in this genetic background. We therefore compared clonogenicity after temozolomide treatment, confirming marked sensitization in PARG KO compared with wild-type control (Fig. 2H).

After exposure to temozolomide only, the decrease of NAD⁺ and upregulation of PAR is transient in IDH-mutant gliomas (25), with a measurable PAR increase in the 1 to 6 hours following treatment that resolves within 24 hours. The rapid return to baseline levels suggests that enzymatic PAR breakdown by PARG may be contributing to the restoration of NAD⁺ homeostasis. To test whether PARG inhibition after temozolomide was sequestering NAD⁺ as PAR, we next measured NAD⁺ levels in cells treated with combined temozolomide + PDD, and observed significant and sustained depletion in five endogenous IDH-mutant cancer lines compared with IDH wild-type cancer lines (Fig. 3A; Supplementary Fig. S3A), a finding that supports our proposed mechanism. In addition, in these IDH-mutant cell lines, temozolomide + PDD, but not temozolomide alone, led to marked PAR accumulation and concomitant apoptosis (Fig. 3B; Supplementary Fig. S3B), compared with IDH wild-type glioma U251 control. Olaparib potently suppressed temozolomide-induced PAR expression (Supplementary Fig. S3C). Strikingly, olaparib also reversed the cell viability reduction induced by the temozolomide + PDD combination in IDH-mutant cells (Fig. 3C; Supplementary Fig. S3D). Taken together with the previous results that PARP inhibition alone did not decrease NAD⁺ (Fig. 1C), and that temozolomide alone caused only transient PAR accumulation (25), these data suggest that the temozolomide + PDD combination induces cytotoxicity by a distinct mechanism. We considered that the aberrant metabolism of NAD⁺ depletion could be the primary cause of the observed effect, in place of (or in addition to) intensified alkylating DNA damage. Indeed, during the short-term time frame of our in vitro experiments, an effective dose of the temozolomide + PDD combination elicited no, or only minimal, activation of γH2AX, a DNA damage marker, or LC3, an autophagy marker, in HT1080 and MGG119. (Supplementary Fig. S3E). Interestingly, addition of PDD seems to suppress temozolomide-induced γH2AX activation, suggesting possible secondary effects leading to alterations of DNA damage signaling pathways.

Strongly supporting our hypothesis, we were able to partially or fully reverse the effects of temozolomide + PDD on both NAD⁺ levels and cell viability by rescue supplementation with the NAD⁺ precursors nicotinamide (NAM) or NMN (Fig. 3D and E; Supplementary Fig. S3F). The lesser magnitude of effect on NAD⁺ level, when compared with cell viability, that we observed with NAM rescue suggests that cytotoxicity may be offset with NAD⁺ levels recovering above an appropriate threshold in a critical time frame. Furthermore, PAR levels decreased with NAM supplementation (Supplementary Fig. S3G), possibly due to the established mechanism by which NAM acts as a feedback negative regulator on PARPs, suppressing PARylation (30, 31) and affecting the synthetic lethal DNA damage signaling pathways.

To further evaluate our hypothesized metabolic mechanism, we then tested NAD⁺ alterations in our independent PARG KO lines. Temozolomide treatment induced marked NAD⁺ depletion in both PARG KO HT1080 lines, compared with nontargeting control (Fig. 3F). Temozolomide treatment also induced PAR upregulation and apoptosis, as evidenced by cleaved caspase-3 induction, in both PARG KO HT1080 lines (Fig. 3G; Supplementary Fig. S4A); these findings are fully consistent with those observed after PARG inhibition by PDD. Conversely, PARP inhibition with olaparib abrogated temozolomide-induced PAR accumulation and NAD⁺ depletion in PARG KO cells, rescuing them from early (24 hours) apoptosis (Fig. 3F and G). These results constitute independent evidence that inactivation of PARG activity, and resulting PAR accumulation, is required for cell death after temozolomide treatment in PARG KO cells, in contrast to the control cells. Furthermore, we could reverse the NAD⁺ level and cell viability inhibition by rescue supplementation with NMN or NAM in PARG KO cells (Fig. 3H and I; Supplementary Fig. S4B and S4C), a result consistent with the effect seen with temozolomide + PDD treatment. All told, these results indicate that metabolic depletion of NAD⁺ contributes mechanistically to combined temozolomide and PARGi toxicity.

Next, we assessed the relationship between PARG inhibition and metabolic enhancement with other alkylating regents that are clinically effective in patients with glioma, such as procarbazine, carmustine (BCNU), and lomustine (CCNU; Supplementary Fig. S5A–S5C). Interestingly, only the monofunctional DNA alkylating agents temozolomide and procarbazine activated PAR in PARG KO lines (Supplementary Fig. S5D). These monofunctional alkylators produce DNA adducts primarily at O6- and N7-methyl guanine (O6meG and 7meG) and N3-methyl adenine (3meA). On the other hand, nitrosoureas such as BCNU and CCNU, which react with adjacent cytosine bases to generate guanine–cytosine (G-C) interstrand cross-links, did not activate PAR in PARG KO lines. These results highlight the differential role of PARPs in response to various DNA damage adducts, supporting the specificity of our hypothesized mechanism (32). Thus, we conclude that PARG inhibition after monofunctional alkylators sequesters NAD⁺ as PAR, both prolonging the hyperactive PAR-mediated signaling state and critically depleting NAD⁺ levels, resulting in increased cytotoxicity.

We then considered the genomic determinant of sensitivity to alkylating chemotherapy and PARGi in IDH-mutant cancer cells, given the possibility that the combination could be toxic in normal cells. Given our prior evidence of the selectivity of NAD⁺ depletion–mediated toxicity for IDH-mutant cells (8), we used an isogenic model to directly test the causal effect of IDH mutation on the combination-responsive phenotype.

First, we tested temozolomide + PARGi in IDH wild-type MGG18 cells, which were engineered to express a doxycycline-inducible IDH1-mutant protein (Fig. 4A). Altered NAD⁺ equilibrium in IDH-mutant cells is related to downregulation of the biosynthesis enzyme NAPRT1 (8, 33). After 3 months of exposure to doxycycline, a time course consistent with previous findings (8, 34), we observed an induction of mutant IDH1 expression, a high concentration of 2-HG, NAPRT1 inactivation, and decreased NAD⁺ and NADH (Fig. 4B–D). In this model system, we observed that temozolomide + PARGi had a greater and more rapid cell-inhibitory effect after mutant IDH1 induction compared with the noninduced state (Fig. 4E). These results were also recapitulated in an additional IDH wild-type line, MGG123, engineered to express IDH1-R132H (Supplementary Fig. S6A–S6D).

In the IDH1-mutant induced state, as compared with noninduced MGG18, NAD⁺ was depleted to a greater degree and for a longer time after treatment with the combination, inducing apoptotic cell death (Fig. 4F and G). In the MGG18 IDH1-mutant line, transient high expression of γH2AX was observed at 24 hours, but it was repaired at 48 hours (Supplementary Fig. S6E) and no significant change occurred in the autophagy marker LC3. On the other hand, short-term incubation of doxycycline (6 days) caused sufficient expression of IDH1-R132H to increase 2-HG level, but did not result in changes in NAPRT1 expression or NAD⁺/NADH levels (Supplementary Fig. S6F–S6I). Consistent with the lack of difference in NAD⁺ level, no significant change occurred in the sensitivity to temozolomide + PARGi in this cellular context (Supplementary Fig. S6J). The addition of octyl-2HG, a cell-permeable form of 2-HG, similarly increased the intracellular 2-HG concentration, but did not change the sensitivity to temozolomide + PARGi (Supplementary Fig. S6K and S6L). This time course suggests that long-term 2-HG exposure is required for NAPRT1 suppression, which leads to altered NAD⁺ metabolism and temozolomide + PARGi sensitivity.

Also supportive of this mechanism, the patient-derived endogenous IDH-mutant astrocytoma line MGG152, tested with and without prolonged exposure (>3 months) to the IDH inhibitor AGI5198, displayed a reversal of the combination temozolomide + PDD cytotoxic effect after inhibition of mutant IDH1 (Fig. 4H). Long-term IDH inhibitor treatment suppressed 2-HG and was associated with greater steady-state NAD⁺ and NADH accumulation (Fig. 4I and J).

Genetic inactivation of the mismatch repair system (MMR) leads to the emergence of clones resistant to alkylating chemotherapy in cancer cell populations (35), thus demonstrating the key role MMR signaling plays in mediating the cytotoxicity of alkylating DNA damage. In recurrent gliomas, somatic acquisition of MMR pathway deficiency confers resistance to temozolomide, albeit at the cost of a hypermutator phenotype in the post-temozolomide setting (36, 37). MMR deficiency is particularly enriched in post-treatment recurrent IDH-mutant gliomas (38, 39). To experimentally isolate the impact of MMR on the combined effect of TMZ and PARGi in the context of endogenous IDH mutation, we established stable MSH6 knockdown lines of IDH1-mutant HT1080 (Fig. 5A), which is MGMT-unmethylated (Fig. 1B; ref. 25). In this model system, we confirmed the expected acquired resistance to the cytotoxic effects of temozolomide monotherapy (Fig. 5B). Notably, however, MSH6 knockdown lines retain sensitivity to the temozolomide + PDD combination in cell viability and clonogenicity assays (Fig. 5C and D). In addition, in experiments mirroring those above, we could rescue the cytotoxic and NAD⁺ depletion effect by supplementation with NMN or NAM (Fig. 5E and F; Supplementary Fig. S7A).

Furthermore, in MGG152 glioma cells, which are MGMT-methylated (Fig. 1B), we established two independent stable MSH6 knockdown cells and confirmed MSH6 knockdown increased resistance to temozolomide (Fig. 5G and H). Here again, the temozolomide plus PDD combination displayed equal efficacy even in MSH6 knockdown cells (Fig. 5I). The cytotoxic and NAD⁺ depletion effect was again rescued by supplementation with NMN or NAM (Fig. 5J and K; Supplementary Fig. S7B). As a control to test whether this finding was specific for IDH-mutant cells, we tested the MGMT-methylated IDH wild-type glioblastoma, LN229, utilizing engineered MSH6 knockdown lines (Supplementary Fig. S7C). In this IDH wild-type glioma, temozolomide + PDD did not show any benefit, irrespective of the MSH6 status (Supplementary Fig. S7D). Therefore, combining temozolomide and PARGi is effective in IDH1-mutant cells notwithstanding the MMR-mediated resistance mechanism.

Given our observations pointing toward the key role of NAD⁺ in mediating temozolomide + PARGi cytotoxicity, and the potential for variation in metabolism when comparing in vitro and in vivo environments, we tested whether PARG inactivation in vivo would produce the same results. We established a paired test, injecting the left flank of an individual mouse with nontargeting control sgRNA HT1080 and the right flank of the same mouse with PARG-targeting sgRNA #3 HT1080 (Fig. 6A). Pairing tumors in the same animal provides the experimental advantage of an internal control for temozolomide dosing, as each animal harbors an isogenic comparison for the same systemic temozolomide administration. We noted a slowing of tumor growth in PARG KO lines even without chemotherapeutic treatment, a result likely indicative of more stringent growth conditionsin vivo (40, 41), compared with in vitro. Importantly, the tumor-specific effect of temozolomide was augmented in the PARG KO line, which displayed markedly smaller tumor size and weight across treated cohorts (Fig. 6B and C; Supplementary Fig. S8A). Confirming the metabolic mechanism in vivo, we observed significantly lower basal NAD⁺ levels in the PARG KO tumors, in which NAD⁺ and NADH levels were further decreased after 4 days (one cycle) of temozolomide treatment (Fig. 6D), with a concomitant increase in PAR (Fig. 6E). In IHC analysis of the treated tumors, we observed evidence of increased apoptosis in the PARG KO tumors (Fig. 6F and G). PARG KO tumors had a decreased Ki-67 index, indicating suppression of proliferation, consistent with a recent report (41). The animals displayed no significant systemic toxicity during treatment (Fig. 6H). We also tested the orthotopic flank model using a second independent clone of HT1080 with PARG-targeting sgRNA #2 (Supplementary Fig. S8B–S8E) and observed findings mirroring those seen with PARG sgRNA #3. Thus, PARG inactivation potentiates the efficacy of temozolomide in an IDH-mutant xenograft model in vivo.

Here, we report a unique metabolic consequence of PARG inactivation after temozolomide treatment, namely PAR hyperaccumulation and concurrent sequestration of NAD⁺ (Fig. 7). This combination of alkylating chemotherapy and PARG inhibition is highly effective against IDH-mutant cancer cells, as confirmed by isogenic experiments. PARGi prolongs and deepens the temozolomide-induced depletion of free NAD⁺ levels, a metabolic state which is ultimately lethal in these vulnerable cells (8). The cytotoxic effect of the combined regimen is reversed by NAD⁺ supplementation, experimentally demonstrating the key contribution of metabolic stress, and mechanistically consistent with prior studies of NAD⁺ trafficking in IDH-mutant and wild-type cells (25, 27).

A similar mechanism was observed with the monofunctional alkylator procarbazine when combined with PARG inhibition, highlighting the potential translational relevance of this finding to clinically utilized glioma treatment regimens. The most widely used alkylating chemotherapeutic regimens, which include either temozolomide or procarbazine, lack durability in treating patients with IDH-mutant glioma. Recurrences emerge in a meaningful subset of patients within a decade after diagnosis (42), in many cases recurring as malignant tumor subclones that drive lethal disease progression (43, 44). Post-alkylator recurrences of IDH-mutant glioma often display the mutational signature of an on-target selective pressure of chemotherapy, with escape from DNA damage surveillance mediated by MMR-deficient hypermutant genomic evolution (38, 45). This partial effectiveness of chemotherapy has motivated research efforts to identify additional adjuvant agents that are specific for IDH-mutant gliomas and can be combined with standard-of-care alkylators. Encouragingly, we demonstrate that a combination of a monofunctional alkylator plus PARG inhibition can be effective in MMR-deficient IDH-mutant glioma, potentially allowing for more effective treatment outcomes.

From a therapeutic translational standpoint, this combination may prove effective in other cancer cells with metabolic alterations in NAD⁺ trafficking (33, 46, 47). Nevertheless, one hurdle still to overcome is the lack of a suitable PARGi forin vivo combination with temozolomide, and studies on PARGi to date have primarily explored only in vitro models. The cell-permeable PARGi PDD is not amenable for in vivo use, despite its potency, due to its short half-life in circulation (48–50). A more recently available PARGi, COH34, has improved stability in vitro and in vivo, but its potency is limited in otherwise PARGi-sensitive cell lines (51). This situation, however, is rapidly changing, as increasingly stable new agents enter active preclinical development. An important consideration when testing potential combination regimens is that the toxicity of these agents in normal tissues is currently unknown. Reassuringly, however, we did not detect a measurably significant effect of combination treatment in either normal control NHA cells or most IDH wild-type tumor lines. Although further translational studies are needed, a key strength of our work is the support provided to the proposed combination of alkylating chemotherapy and PARG inhibition in IDH-mutant cancers by independent and convergent lines of genetic and pharmacologic evidence in multiple patient-derived models, both in vitro and in vivo.

We also observed a sensitivity to PARP inhibition as monotherapy in our endogenous IDH-mutant glioma lines, consistent with prior reports (16, 17). However, we noted variability in the responsiveness of IDH-mutant lines, suggesting that important modifiers of this effect are yet to be discovered. Notably, we found that PARP inhibition counteracted NAD⁺ biosynthesis inhibition, a metabolic perturbation that could potentially inform the combination therapy of temozolomide and PARPi in IDH-mutant tumor cells. The complexity of cancer PARylation metabolism is further highlighted by the disparate effects observed in other studies of PARG (50), inhibitors of which can, in varying contexts, alternatively antagonize (49) or phenocopy (52) the homologous recombination deficiency (HRD)–associated synthetic lethal effects of PARPi, or can mediate differential killing via HRD-independent mechanisms altogether (40, 53). In addition, the hyper-PARylation we observed after alkylating chemotherapy and PARGi likely magnifies PAR-mediated inhibition of hexokinase activity and glycolysis (54, 55), independently of metabolic effects on NAD⁺. Our findings of a PARGi-mediated metabolic effect therefore suggest that a temozolomide + PARGi combination could represent another therapeutic option to counter clinical PARPi resistance across multiple cancer types.

In conclusion, the combination treatment we propose uses a two-hit disruption of NAD⁺ homeostasis to achieve targeting of an intrinsic IDH-mutant metabolic weakness, while limiting the escape avenues available for the emergence of subclonal resistance mutations. Our proposed mechanism extends and unifies several prior observations of selective vulnerability in IDH-mutant gliomas to chemotherapy, NAD⁺ biosynthesis inhibition, and isolated PARPi monotherapy sensitivity. Importantly, the observed nonoverlapping mechanisms of DNA damage and metabolic cytotoxicity may allow for a treatment strategy that could improve IDH-mutant glioma treatment.

The patient-derived glioma lines used in this study (MGG18, MGG23, MGG119, MGG123, MGG152, and MGG173) were obtained from 2008 to 2018 under institutional review board–approved protocols. The TS603 (IDH1-R132H, WHO grade III anaplastic oligodendroglioma) gliomasphere was derived from a patient who underwent tumor resection at Memorial Sloan-Kettering Cancer Center (New York, NY). The BT142 (IDH1R132H), HT1080 (IDH1R132C), T98G (IDH wild-type), U251 (IDH wild-type), LN229 (IDH wild-type), and Hs683 (IDH1 wild-type) lines were authenticated in 2019 by comparing their short tandem repeat profiles to those of the ATCC public dataset. Normal human astrocytes (NHA), purchased from ScienCell, were cryopreserved at passage number 3 or lower before their in vitro use. Patient-derived glioma neurosphere lines (BT142, TS603, MGG18, MGG23, MGG119, MGG123, MGG152, and MGG173) were cultured in serum-free Neurobasal medium (Gibco) as described previously (43, 56, 57). NHA, T98G, LN229, Hs683, and U251 cells were cultured in DMEM. HT1080 cells were cultured in Eagle Minimum Essential Medium. All standard growth media were supplemented with 10% FBS and penicillin, streptomycin, and amphotericin B. All glioma cell lines were determined to be Mycoplasma free using a LookOut Mycoplasma PCR Detection Kit (Sigma) in 2018 and 2019. All cells were maintained at 37°C in a humidified atmosphere of 5% CO₂/95% air.

The following chemical compounds were used in culture: AGI-5198 (Selleckchem), FK866 (Cayman Chemical), dimethyl sulfoxide (Sigma-Aldrich), doxycycline hyclate (Sigma-Aldrich), nicotinamide (Sigma-Aldrich), nicotinamide mononucleotide (Sigma-Aldrich), Octyl-(R)-2HG (Sigma-Aldrich), olaparib (Selleckchem), PDD00017273 (Tocris), and temozolomide (Cayman Chemical).

MGG18 cells expressing tetracycline-inducible IDH1-R132H, MGG18-IDH1-R132H, were generated with lentiviral transduction and described previously (8). To induce expression of mutant IDH1, MGG18-IDH1-R132H cells were cultured with doxycycline (1 μg/mL; Sigma-Aldrich) for 3 days to 3 months. GBM cell line MGG123 overexpressing IDH1-R132H was generated by lentivirus infection using pLenti6.3/TO/V5 containing IDH1R132H, pCMV-psPAX2 (Addgene), and pCMV-VSVG (Addgene), followed by selection with blasticidin (0.5 μg/mL).

Cells were seeded in a 6-well plate in antibiotic-free medium. Seventy percent to 80% confluent cells were subjected to transfection. PARP1 cDNA ORF clone (catalog no. OHu2551) and control vector (pcDNA3.1+/C-(K)DYK, catalog no. SC1849) were purchased from GenScript. Transfection was carried out using Opti-MEM medium (Thermo Fisher Scientific) and Lipofectamine 3000 reagent (Invitrogen). After 72 hours, PARP overexpression was confirmed by Western blot analysis.

MSH6 knockdown HT1080 and MGG152 cells were generated with lentiviral transduction as described previously (25). Detailed protocols are found in the Supplementary Methods.

Lentivirus vector plasmids containing gRNA sequences for PARG were pLentiCrispr v2-PARG gRNA from Genscript (gRNA2 and 3). Nontargeting gRNA lentivirus vector was from Genscript. To generate lentiviral particles, 293T cells were transfected with lentiviral plasmid, packaging plasmid (pCMV-psPAX2), and envelope plasmid (pCMV-VSV-G) with Fugene HD (Promega). Cells were infected with lentivirus in the presence of polybrene (8 μg/mL) for 8 hours. After 3 days, the cells were selected with puromycin (0.6 μg/mL for HT1080) for another 3 days before use. Single cells were isolated in a clear 96-well plate, and once single colonies were identified, they were scaled up to 12-well plates and plated in 10-cm dishes. Knockout was confirmed by Western blot analysis.

gRNA sequences were as follows:

• Plasmid: pLentiCrispr v2 nontargeting gRNA: ATCTACGGGTTTATGCCAAT

• Plasmid: pLentiCrispr v2-PARG gRNA2: TGCTATTCTGAAATACAATG

• Plasmid: pLentiCrispr v2-PARG gRNA3: CAACACATTATAAAGATTTG

Genomic and bisulfite-modified DNA was extracted using AllPrep DNA/RNA/miRNA Universal Kit and EpiTect Bisulfite Kits (Qiagen) per the manufacturer's protocol. Detailed protocols are described in the Supplementary Methods.

To assess MGMT promoter methylation status, methylation-specific PCR was performed in a two-step approach as described previously (8). MGMT protein expression was performed by Western blot analysis. Tumors with mixed results (i.e., evidence of partially MGMT-methylated MSP, but detectable expression of MGMT protein), were designated “unmethylated” to signify the presence of temozolomide-resistant subclones.

Cells were seeded in 96-well plates at 1,000 to 3,000 cells per well. After a 4-hour incubation, compounds were serially diluted and added to wells, as indicated in the figure legends. Cell viability was evaluated by Cell Titer-Glo (Promega) and CellTiter-Fluor Cell Viability Assay (Promega) according to the manufacturer's protocol at the indicated time points. Cell viability was evaluated on a daily basis after drug exposure to determine the time course of treatment effects and was plotted as % cell viability relative to DMSO control.

To assess the clonogenicity of attached cells, 100 HT1080 and U251 cells were seeded on 6-well plates. Cells were exposed to DMSO, PDD00017273 (5 μmol/L), temozolomide (50–200 μmol/L), or PDD00017273 plus temozolomide for 8 to 14 days. After 1× PBS washes, cells were fixed in 6% glutaraldehyde and stained with crystal violet for 30 minutes. Images were captured with the EVOS FL Auto 2 Imaging System (Thermo Fisher Scientific).

To assess clonogenicity in gliosphere cells (MGG123, MGG152 and MGG119), 100 cells were seeded per well in clear 96-well plates. Cells were exposed to DMSO, PDD00017273 (5 μmol/L), temozolomide (100 μmol/L), or PDD00017273 plus temozolomide. Fresh medium was added every week. The sphere number was counted directly under a microscope at 14 days. To exclude bias, spheres were counted in a blinded manner by S. Rafferty.

To test cell growth, 4,000 cells of HT1080 nontargeting (NT) gRNA, PARG gRNA2, and PARG gRNA3, were seeded on 6-well plates. Cell numbers were counted 0, 2, and 5 days afterward using a Luna-FL Automated Cell Counter (Logos Biosystems).

Caspase 3/7 activities were evaluated with a Caspase-Glo 3/7 assay (Promega) according to the manufacturer's recommendations. Glioblastoma tumor spheres and HT1080 cells were dissociated into single cells and seeded in 96-well plates at 1,000 to 3,000 cells/well. After 4 hours, serially diluted DMSO, temozolomide, and temozolomide with or without PDD00017273 were added to the wells at the indicated concentrations. After 24 to 72 hours, an equal volume of caspase 3/7 reagent was added to the samples and mixed. After a 30-minute incubation, luminescence was recorded with a Synergy HT Multi-Mode Microplate Reader (BioTek).

To obtain qualitative values of NAD⁺ and NADH, the NAD/NADH-Glo Assay (Promega) was used according to the manufacturer's recommendations. Briefly, 1 × 10⁵ cells were lysed with 200 μL of PBS and then 200 μL of 0.2 N NaOH with 1% dodecyltrimethylammonium bromide (DTAB; Sigma-Aldrich) was added. To measure NAD⁺, 50 μL of 0.4 N HCl was added to lysed cell samples (100 μL), which were then heated at 60°C for 15 minutes. After incubation at room temperature for 10 minutes, 0.5 mol/L Trizma base buffer (50 μL; Sigma-Aldrich) was added. To measure NADH, lysed cell samples (100 μL) were heated at 60°C for 15 minutes and then incubated at room temperature for 10 minutes, followed by the addition of an equal volume of HCl/Trizma solution. Finally, samples were seeded in 96-well plates and incubated with NAD/NADH-Glo detection reagent for 60 minutes. Data were compared with DMSO-treated cells and expressed as % control.

To measure NAD⁺ quantitatively, NAD⁺/NADH Quantification Colorimetric Kit (BioVision Incorporated) was used according to the manufacturer's recommendations. For each experiment, 2 × 10⁵ cells, as indicated, were washed with cold PBS and extracted in NADH/NAD extraction buffer via two freeze/thaw cycles.

To measure NAD⁺ in tumor tissue, 4 × 10⁶ HT1080/NT and PARG gRNA3 cells were subcutaneously transplanted into the left and right flank of athymic nude mice, respectively. Four days after transplantation, the mice were randomized to treatment with vehicle (n = 3) or temozolomide (50 mg/kg, n = 3) given daily for 5 days. Six hours after the last treatment, the mice were euthanized and the tumor tissues were collected, washed with PBS, and homogenized in NADH/NAD extraction buffer. Protein concentrations were measured using a BCA assay and used for normalization. After vortex extraction for 10 seconds, samples were centrifuged at 14,000 rpm for 5 minutes and the supernatant was transferred to a new tube. To detect total NADt (NADH and NAD), 50 μL of the extracted samples were transferred into a 96-well plate. To detect NADH, NAD⁺ was decomposed by heating at 60°C for 30 minutes. Then, 50 μL of NAD-decomposed samples were transferred into a 96-well plate. The NADH/NAD supernatant was transferred into a tube and NADt (NADH and NAD) and NADH signals were measured at OD 450 nm (BioTek). NAD⁺ concentrations were calculated by subtracting NADH from NADt. Data are expressed as pmol/1 × 10⁶ cells or pmol/mg protein.

To obtain qualitative values of 2-HG, a D-2-Hydroxyglutarate (D2HG) Assay Kit (Sigma-Aldrich) was used according to the manufacturer's recommendations. Detailed protocols are found in the Supplementary Methods.

Primary antibodies for blotting were as follows: IDH1 R132H (Dianova, #DIA-H09), NAPRT1 (Sigma-Aldrich, #HPA023739), PAR (Trevigen, #4336-BPC), β-actin (Cell Signaling Technology, #3700), cleaved caspase-3 (Cell Signaling Technology, #9661), cleaved PARP (Cell Signaling Technology, #5625), LC3B antibody (Cell Signaling Technology, #2775), MGMT (Cell Signaling Technology, #2739), MSH6 (Cell Signaling Technology, #5424), PARG (Cell Signaling Technology, #66564), PARP (Cell Signaling Technology, #9542), Phospho-Histone H2A.X (Cell Signaling Technology, #2577), and Vinculin (Thermo Fisher Scientific, # 700062). Detailed protocols are described in the Supplementary Methods.

Tumor tissue sections were incubated with anti–Ki-67 (1:200; Dako, # M7240) overnight at 4°C and then with secondary antibody (ImmPRESS HRP Horse Anti-Rabbit IgG [Peroxidase] Polymer Detection Kit; Vector Laboratories) for 30 minutes at room temperature. Three slides were stained per tumor. Six pictures were captured with 20× magnification per section and used for quantitative analysis of immunopositivity. Detailed protocols are described in the Supplementary Methods.

In situ detection of apoptotic cells was carried out using the TUNEL assay kit according to the manufacturer's protocol (Millipore).

Three slides were stained per tumor. Six pictures were captured with 20X magnification per section and used for quantitative analysis of immunopositivity. Detailed protocols are described in the Supplementary Methods.

All animal experiments were approved by the Institutional Animal Care and Use Committee of Massachusetts General Hospital. NT gRNA HT1080 cells and PARG gRNA3 HT1080 cells (4 × 10⁶ each) were subcutaneously transplanted into the left and right flanks of 7- to 10-week-old female athymic nude mice (weight, 21–26 g; Charles River), respectively. Mice were housed under a 12-hour light:dark cycle in single cages (21–24°C with 45% humidity) with standard bedding and enrichment and ad libitum access to food and water. When the maximum tumor diameter reached 5 mm, the mice were randomized to a vehicle group (PBS, i.p., n = 7) and a temozolomide group (50 mg/kg, i.p., n = 7). PBS with 10% DMSO was used as vehicle control. Treatments were administered 5 times a week for 2 weeks with a 1-week-off period. A digital caliper was used to measure tumor diameters 3 times a week. The volume (mm³) was calculated as length (mm) × width (mm)² × 0.5. Tumor volume was normalized to that of day 0.

To evaluate changes in NAD⁺ in tumor tissue, 4 × 10⁶ HT1080/NT cells and HT1080/PARG gRNA3 cells were subcutaneously transplanted into the left and right flank of athymic nude mice, respectively. When the maximum tumor diameter reached 5 mm, the mice were randomized to treatment with vehicle (n = 3) or temozolomide (50 mg/kg, n = 3). Treatment was administered 5 times daily. Six hours after the last treatment, the mice were sacrificed and tumor tissues collected. Tumor tissue was washed with PBS and homogenized in NADH/NAD extraction buffer, followed by NAD⁺ quantitation. For Western blot analysis, a separate set of mice with bilateral flank tumors and the same vehicle or temozolomide (50 mg/kg) treatment were used. Tumor tissues were collected 6 hours after the last treatment, washed with PBS and homogenized in RIPA buffer for tissue lysates. To examine apoptosis and cell proliferating in vivo, the same bilateral tumor model was set up with vehicle or temozolomide (50 mg/kg) treatment in 5 consecutive days. Six hours after the last treatment, the mice were sacrificed and the tumor tissues collected, fixed in neutral-buffered formalin and embedded in paraffin, followed by IHC.

Prism 8 software (GraphPad) was used for all statistical analyses. Data are presented as mean ± SEM and were analyzed using a Student t test (unpaired) for two group comparisons. Details of statistical analyses are provided in the figure legends. Key in vitro experiments were performed at least three times with consonant results. Keyin vivo experiments were performed at least two times with consonant results. P values less than 0.05 were considered statistically significant.

D.P. Cahill reports grants from the National Institutes of Health National Cancer Institute (P50CA165962, R01CA227821 HW/DPC Co-PIs), Tawingo Fund, and Loglio Foundation during the conduct of the study; and personal fees from Lilly, Merck, and Boston Pharmaceuticals outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

H. Nagashima: Conceptualization, data curation, investigation, methodology, writing-original draft, writing-review and editing. C.K. Lee: Data curation, investigation, methodology, writing-review and editing. K. Tateishi: Resources, methodology. F. Higuchi: Resources, methodology. M. Subramanian: Investigation. S. Rafferty: Investigation. L. Melamed: Investigation. J.J. Miller: Conceptualization, supervision, funding acquisition, validation. H. Wakimoto: Conceptualization, supervision, funding acquisition, investigation, writing-original draft, writing-review and editing. D.P. Cahill: Conceptualization, supervision, funding acquisition, investigation, writing-original draft, writing-review and editing.

This work was supported by the NIH R01CA227821 (to H. Wakimoto/D.P. Cahill), P50CA165962 (to D.P. Cahill), and NCI Paul Calabresi Career Development Award in Clinical Oncology for Nervous System Tumors K12CA090354 (to J.J. Miller). We also acknowledge the Tawingo Fund (to D.P. Cahill), Loglio Foundation (to D.P. Cahill), Richard B. Simches Scholars Award (to J.J. Miller), a Seeman Family MGH Scholar in Neuro-Oncology Award (to J.J. Miller), and the Overseas Research Fellowship of the Uehara Memorial Foundation (to H. Nagashima). We thank the patients for their donation of critical research materials, and Donald Glazer, Fred Barker, and members of the Translational Neuro-Oncology Laboratory for helpful discussions and input.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.