BCLX_L blockade rewires cell fate to overcome PARP inhibitor resistance in ovarian cancer

Abstract

Poly(ADP-ribose) polymerase (PARP) inhibitors (PARPis) induce regressions and extend progression-free survival (PFS) in ovarian cancer, especially in tumors with BRCA1 or BRCA2 mutations that impair homologous recombination repair. While recent studies have clarified the key roles of BRCA1, BRCA2, and PARP1 in replication fork stability, the downstream mechanisms that mediate PARPi-induced cytotoxicity and resistance remain incompletely understood. Here we delineate cell fate outcomes following PARPi treatment in homologous recombination-deficient high-grade serous ovarian cancer and identify actionable pathways to overcome acquired resistance. Our findings reveal that PARPi-induced DNA damage simultaneously triggers apoptosis, which primarily occurs through the BAX/BAK-dependent intrinsic apoptotic pathway, while also driving cellular senescence, as manifested by the expression of senescence-associated β-galactosidase, CDKN1A upregulation and a senescence-associated secretory phenotype. Notably, the PARPi-induced senescent cells persist as resistance develops and exhibit multinucleation, a hallmark of nuclear atypia, both in vitro and in patient-derived xenografts (PDXs). Building on the observation that the anti-apoptotic protein BCLX_L restrains pro-apoptotic BCL2 family members after PARPi treatment, we show that addition of the BCLX_L inhibitor A-1155463 to PARPi therapy diminishes resistance in multiple high-grade serous ovarian cancer cell lines in vitro and significantly enhances PARPi-induced tumor response in a PDX model with acquired PARPi resistance in vivo. Overall, these preclinical findings strongly support the potential of combining BH3 mimetics with PARPis to treat resistant ovarian cancer.

Introduction

PARP inhibitors (PARPis) have transformed the therapy of BRCA1- and BRCA2-mutated ovarian, breast, prostate, and pancreatic cancers. Four PARPis, olaparib, niraparib, talazoparib, and rucaparib, are currently FDA approved for these neoplasms. For example, olaparib, niraparib and rucaparib significantly extend progression-free survival when administered as maintenance therapy following surgery and platinum-based adjuvant or neoadjuvant chemotherapy for BRCA1- or BRCA2-mutated high grade serous ovarian cancer (HGSOC).^1–4 Moreover, rucaparib also exhibits single-agent activity against platinum-sensitive relapsed ovarian cancers.⁵ This clinical activity reflects the “synthetic lethal” interaction of BRCA1/BRCA2 loss and PARP inhibition whereby tumor cells harboring defects in homologous recombination become highly dependent on PARP1/2-mediated DNA repair pathways for survival. Although this concept of synthetic lethality has become central to the clinical application of PARPis, the precise molecular mechanisms that determine tumor sensitivity and eventual resistance remain incompletely understood.^6,7 At the cellular level, PARPis trigger multiple genotoxic events, including trapping of PARP at sites of DNA single-strand breaks,^8,9 diminished resolution of stalled replication forks,¹⁰ altered processing of replication intermediates,¹¹ and impaired resolution of transcription-replication conflicts,¹² leading to increased DNA double-strand breaks. The resulting DNA damage has been reported to induce apoptosis, a tightly regulated cell death process that can be triggered either through engagement of death receptors or changes in BCL2 family members.^13–16 The same types of DNA damage can also induce cellular senescence,¹⁷ a durable state of proliferative arrest characterized by extensive metabolic and transcriptional reprogramming together with secretion of inflammatory cytokines, chemokines, and extracellular matrix-remodeling factors collectively referred to as the senescence-associated secretory phenotype (SASP).^18,19 Previous studies raise the possibility that these senescent cells can contribute to therapy resistance and relapse of various cancers.^20–22 While cellular mechanisms driving senescence are regulated at multiple levels, emerging evidence highlights critical roles for p16^INK4A (CDKN2A) and NF-κB in determining ovarian cancer outcomes.^21,23 Nevertheless, the relative contributions of apoptotic mechanisms versus senescence in PARPi-induced tumor cell death as well as their broader impact on long-term therapeutic response to PARPis remain incompletely understood.

Of the various BCL2 family members, the anti-apoptotic protein BCLX_L (also known as BCL2L1) has been implicated in both resistance to apoptosis and prolonged survival of senescent cells, two effects that result from the ability of BCLX_L to inhibit BAX- and BAK-mediated mitochondrial outer membrane permeabilization.^13–15,24 Numerous genomic and proteomic studies have also demonstrated that BCLX_L is frequently overexpressed in multiple cancers due to gene amplification, copy number increases and/or altered transcription.^25–28 Moreover, BCLX_L has been identified as the most active anti-apoptotic BCL2 family member in ovarian cancer cell lines and patient-derived xenografts (PDXs)^29,30; and selective BCLX_L inhibitors have shown monotherapy activity in vitro.²⁹ In addition to its role in inhibiting apoptosis, BCLX_L is often upregulated in senescent cells, where it functions as a critical survival factor that enables persistence of these damaged but viable cell populations.³¹ Early studies demonstrated that ABT-737, a prototypical BH3 mimetic targeting BCL2, BCLX_L, and BCLW, effectively sensitized ovarian cancer cells to carboplatin-induced apoptosis.³² Subsequent studies further demonstrated that ABT-737 and its orally bioavailable analogue navitoclax exhibit potent senolytic activity, selectively eliminating senescent cancer cells across multiple tumor models,^31,33,34 particularly when applied at high concentrations.

Our previous studies examined genomic and transcriptional changes that accompany the development of acquired PARPi resistance in HGSOC PDXs.³⁵ In the PH039 PDX model, which initially lacked RAD51C expression due to promoter hypermethylation, acquisition of PARPi resistance was associated with loss of promoter methylation and re-expression of RAD51C, thereby restoring homologous recombination repair. Notably, however, resistant PH039 xenografts also retained a prominent proinflammatory transcriptional signature suggestive of a persistent stress response.³⁵ In contrast, PH077, a HGSOC PDX bearing a BRCA2 indel, was rendered resistant by a frameshift mutation that restored the open reading frame. The distinct resistance mechanisms observed suggest that restoration of DNA repair alone may not fully explain the resistant phenotype and raised the possibility that chronic inflammatory or senescence-associated programs contribute to tumor persistence after PARPi exposure.

In the present study, PARPi sensitive BRCA1- and BRCA2-mutant ovarian cancer cell lines and their resistant derivatives as well as the two pairs of HGSOC PDXs were utilized to assess the relative contributions of apoptosis and senescence after exposure to clinically achievable PARPi concentrations. Through detailed characterization of these cellular processes after both acute treatment and repeated cycles of PARPi therapy, we identify the pro-apoptotic BCL2 family members activated by PARPi treatment, identify BCLX_L as an important neutralizer of these PARPi-induced BH3-only family members, and show that addition of the selective BCLX_L inhibitor A-1155463 leads to partial PARPi resensitization in PARPi-selected cell lines in vitro and in an aggressive PDX model of PARPi resistant HGSOC in vivo. Collectively, our work provides further insight into the mechanisms governing tumor cell survival after PARPi treatment and supports therapeutic strategies aimed at simultaneously targeting apoptosis resistance and senescence-associated survival pathways in PARPi-resistant ovarian cancer.

Results

Cell fate after PARP inhibition

To mirror the clinical situation where PARP inhibitors are predominantly used to treat BRCA1- and BRCA2-mutated ovarian cancer, we began by examining the behavior of HR deficient ovarian cancer cell lines during continuous PARPi exposure, focusing on multi-nucleation and apoptosis, which have been previously observed in PARPi-treated cells.³⁶ In contrast to prior studies, which utilized clinically unsustainable PARPi concentrations such as 100 µM olaparib,¹⁷ we focused on clinically sustainable low concentrations.³⁷ During the first three days of treatment with 2 µM olaparib, very few cells underwent apoptosis or multinucleation. By treatment day 6, however, a subset of cells began binding Annexin V, indicating phosphatidylserine externalization that is consistent with the induction of apoptosis (Suppl. Fig. 1a, 1b),^38,39 in agreement with previously reported effects in HGSOC cells.^6,40,41 At this same time, ~40% of BRCA2-mutant PEO1 cells or BRCA1-mutant COV362 cells were also multinucleated (Fig. 1a, b, Suppl. Fig. 1c and 1d). Staining for the mitotic marker phospho-Ser¹⁰-Histone H3⁴² indicated that, unlike paclitaxel-treated cells, PARPi-treated cells became multinucleated with little or no prolonged mitotic arrest (Fig. 1c and Suppl. Fig. 1e). When these cells were stained for senescence-associated β-galactosidase, positive cells were observed beginning at day 6, with marked staining in cells containing multiple nuclei (Fig. 1d and Suppl. Fig. 1f). Furthermore, these multinucleated cells were characterized by increased mRNA expression of CDKN1A, IL-6, and IL-8, markers consistent with a senescence-associated secretory phenotype (SASP) (Fig. 1e, Suppl. Fig. 1g, and data not shown). Collectively, these results suggest that, in addition to the widely accepted PARPi-induced outcome of apoptosis, a large fraction of cells become multinucleated and secrete pro-survival proteins characteristic of the SASP.

Fig. 1.

Characterization of PARPi-induced cell fates in HGSOC cell lines. a Following treatment of PEO1 cells with 2 µM olaparib, multinucleated cells were visualized by phalloidin and Hoechst 33258 co-staining. b Quantification of multinucleated cells (from experiments performed as illustrated in Fig.1a), with ≥100 cells analyzed per treatment condition every three days during continuous olaparib exposure. * indicates p < 0.05 by unpaired, t-test. c Flow cytometric quantification of cells staining positive for phosphorylated histone H3 Ser¹⁰ after treatment with 2 µM olaparib for up to six days. DNA content was assayed using propidium iodide staining. Cells treated with paclitaxel (25 nM, 24 h) were included as a positive control for mitotic arrest. d Cytochemical staining of senescence-associated β-galactosidase (β-gal) activity in PEO1 cells treated with 2 µM olabarib over 9 days. Black arrows indicate β-gal positive cells. Bar graph represents quantification of β-gal positive cells, with ≥ 100 cells every three days during continuous olaparib exposure. e Expression of senescence-associated secretory phenotype (SASP) markers was assessed by qRT-PCR in PEO1 cells treated with 2 µM olaparib for up to 12 days. Error bars in (b–e) represent ± SEM from three independent experiments. * in (c–e) indicates p < 0.05 by one-way ANOVA with Dunnett’s multiple comparisons test

Involvement of BCL2 family proteins in PARPi-induced cell killing

To further elucidate the mechanistic basis of PARPi-induced apoptosis, we interrupted BAX and BAK genes, which encode key effectors of the mitochondrial apoptotic pathway.^13–16 Simultaneous deletion of BAX and BAK markedly reduced apoptosis induction by three different PARPis (Fig. 2a and Suppl Fig. 2a). In contrast, interrupting the FADD gene, which encodes a critical adaptor protein in the death receptor pathway,¹³ did not affect apoptosis induction to the same degree (Fig. S2b), underscoring the predominant role of the mitochondrial (intrinsic) apoptotic pathway in PARPi killing.

Fig. 2.

BCL2 family proteins modulate apoptotic response to PARP inhibition. a PEO1 cells transduced with either empty vector (E.V.) or two separate BAX/BAK knockout clones were treated with increasing concentrations of olaparib (left), veliparib (middle), or niraparib (right) for 6 days and assayed for Annexin V binding. Error bars represent SEM. **** indicates p < 0.0001 (olaparib and niraparib) using two-way ANOVA with Dunnett’s test for multiple comparisons. b, c PEO1 (b) or COV362 (c) cells were treated with olaparib + the broad-spectrum caspase inhibitor Q-VD-OPh (10 µM) for 3 days (left) or 6 days (right) and immunoblotted for the indicated proteins. Lanes 6–8 contain two-fold serial dilutions of lane 5. d Densitometric analysis of BIM, PUMA and NOXA relative to housekeeping proteins from 6-day samples in b and c. Error bars, ± sem for three independent experiments. In this and all subsequent bar graphs *, **, *** and **** indicate p < 0.05, p < 0.01, p < 0.001 and p < 0.0001, respectively, using two-way ANOVA with Dunnett’s test for multiple comparisons. e, f After PEO1 (e) or COV362 (f) cells were treated with 2 µM olaparib for 6 days in the presence of 10 µM Q-VD-OPh, cell lysates were subjected to immunoprecipitation with antibodies to BCL2, BCLX_L, MCL1 or, as a control, no antibody. The immunoprecipitates and an aliquot equivalent to 1/5 of total cell lysate used for immunoprecipitation were subjected to immunoblotting with the indicated antibodies

Given that BCL2 family proteins are essential participants in the intrinsic pathway,^13–15 we examined changes in these proteins following PARPi treatment using immunoblotting. After 3 days of treatment with olaparib, minimal changes were observed. However, after 6 days we detected a pronounced upregulation of several BH3-only pro-apoptotic members, including BIM and PMAIP1/NOXA in PEO1 cells and COV362 cells (Fig. 2b–d and Supplementary Fig. 2c). Notably these results were also observed with the PARPis veliparib and niraparib (data in full scans on western blots).

To assess whether the increased expression of BH3-only proteins resulted in their activation, as evidenced by their binding to and neutralization of anti-apoptotic BCL2 family members, we immunoprecipitated BCL2, BCLX_L, and MCL1 from PARPi-treated cells and probed for bound pro-apoptotic family members. This analysis revealed increased binding of BIM and PUMA to BCLX_L in PEO1 cells, as well as enhanced formation of PUMA•BCLX_L and NOXA•MCL1 complexes in COV362 cells (Fig. 2e, f). These studies not only indicate that BH3 proteins are activated by PARPi treatment in HGSOC cells, but also identify BCLX_L as an important neutralizer of pro-apoptotic BH3-only proteins in these cells.

Cellular adaptations driving resistance to PARP inhibitors in BRCA2-reactivated models

To better understand and potentially overcome PARPi resistance, we next examined the impact of PARPis on PEO1-related cells with two distinct mechanisms of BRCA2 reactivation: PEO4 cells, which acquired a secondary mutation restoring BRCA2 expression after the source patient for PEO1 cells was treated clinically,⁴³ and ABT^R2 and ABT^R3 clones, which re-express BRCA2 as a consequence of translational readthrough⁴⁴. As expected, PEO4 cells and both of the veliparib-selected clones were resistant to apoptosis induced by olaparib as well as the selecting agent veliparib (Fig. 3a) but were less resistant to niraparib and rucaparib (Suppl. Fig. 3a, b), consistent with off-target effects reported with the latter PARPis.^45,46

Fig. 3.

Apoptosis and BCL2 family member expression in HGSOC cell lines with acquired PARPi-resistance. a PEO1, PEO4, ABT^R2, and ABT^R3 cells were treated with increasing concentrations of olaparib (top) or veliparib (bottom) for 6 days and assayed for Annexin V binding. Error bars represent ± SEM for 3 independent experiments. * and ** indicate p < 0.05 and 0.01, respectively, for ABT^R2, ABT^R3, PEO4 vs PEO1 using two-way ANOVA with Dunnett’s test for multiple comparisons. b Whole cell lysates from indicated untreated HGSOC cell lines were immunoblotted for the indicated BCL2 family members. Densitometric analysis relative to housekeeping proteins. Error bars, ± sem for two independent experiments

While part of the explanation for resistance to PARPi-induced apoptosis involves HR restoration, we also examined key changes in survival and stress pathways, including cell cycle proteins and BCL2 family members, across this isogenic group of lines. Consistent with a previous report implicating BCLX_L in ovarian cancer drug resistance,⁴⁷ PEO4 cells expressed elevated levels of BCLX_L and reduced levels of the pro-apoptotic NOXA (Fig. 3b). We further investigated these mechanisms by analyzing the same global proteomics dataset generated in our recent study of BRCA2 stop codon readthrough.⁴⁴ While ABT^R2 and ABT^R3 cells did not show similar shifts in BCL2 protein expression, they exhibited heightened replication stress⁴⁸ and a more pronounced SASP (Suppl. Fig. 3c–e) in comparison to the parent PEO1 cells. These observations suggest that, even in the context of restored HR, persistent stress response signaling and SASP activation support tumor cell survival under continued PARPi pressure.⁴⁹

Combination of BCLX_L antagonist and PARPi kills PARPi resistant HGSOC cells in vitro

Building on the preceding characterization of the PARPi-resistant cell line models, we next evaluated the therapeutic potential of combining PARPis with agents targeting pathways dysregulated during resistance acquisition. In ABT^R3 cells, which exhibit elevated replication stress, co-treatment with veliparib and replication checkpoint inhibitors—specifically the ATR inhibitor ceralasertib and the WEE1 inhibitor adavosertib—markedly enhanced PARPi sensitivity (Fig. 4a, b). In contrast, despite the robust single-agent activity of the checkpoint kinase inhibitor prexasertib in both parental PEO1 and ABT^R3 cells (Fig. 4c, 0 nM veliparib), this agent did not further sensitize ABT^R3 cells to veliparib.

Fig. 4.

BH3 mimetics sensitize PARPi-resistant HGSOC to PARP inhibitor therapy. a–f PEO1 (left) or ABT^R3 cells (right) were treated with increasing concentrations of veliparib and ceralasertib (a), adavosertib (b), prexasertib (c), venetoclax (ABT-199) (d), S63845 (e), or A-1155463 (f) in a clonogenic assay. Error bars represent ± SEM for three or more independent experiments

Because PARPi cytotoxicity is largely mediated through BCL2 family-regulated apoptosis, we next investigated whether combining PARPi with BH3 mimetics could enhance efficacy in resistant cells with increased reliance on anti-apoptotic pathways. While minimal sensitization was observed with the BCL2 inhibitor venetoclax and no sensitization was observed with the MCL1 antagonist S63845 (Fig. 4d, e, Suppl. Fig. 4a, b), co-treatment with the selective BCLX_L antagonist A-1155463⁵⁰ demonstrated strong synergistic effects with veliparib (Fig. 4f, Suppl. Fig. 4c). Additional studies demonstrated synergistic effects when additional PARPi-resistant lines as well as the parental cells were treated with veliparib and A1155463 (Suppl. Fig. 4d), providing assurance that the sensitizing effects of A-1155463 were not unique to the ABT^R3 cell line initially tested. When two other FDA-approved PARPis, niraparib and olaparib, were substituted for veliparib, A-1155463 again enhanced PARPi-induced cell killing in multiple PARPi-resistant clones as measured by clonogenic assays and annexin V binding (Fig. 5a–c and Supplementary Fig. 5a, b). Importantly, this enhancement of sensitivity was not limited to the PEO1 cell lineage, but was also observed in COV362 cells with acquired PARPi resistance (Fig. 5d, e and Suppl. Fig. 5c, d). These findings suggest that targeting BCLX_L may be an effective strategy to enhance sensitivity of ovarian cancers to PARPi, particularly those with persistent replication stress or apoptotic priming.

Fig. 5.

BCLX_L inhibitor/PARPi combination inhibits growth of multiple PARPi-resistant HGSOC cell lines. a–c PEO1 (left) or ABT^R#3 cells (right) were treated with increasing concentrations of niraparib (a) or olaparib (b) and A-1155463 in a clonogenic assay (a, b) or assayed for Annexin V binding by flow cytometry (c). Error bars represent ± SEM for three or more independent experiments. d, e Parental COV362p cells (left) or olaparib-selected COV362 olap^R cells (right) were treated with increasing concentrations of niraparib (d) or olaparib (e) and A-1155463 in a clonogenic assay. Error bars represent ± SEM for three or more independent experiments. Combination index plots for (a, b, d, e) are shown in Supplementary Fig. 5

BCLX_L inhibitor improves PARPi therapy outcomes in PARPi resistant ovarian cancer in vivo

To determine whether the cellular fates induced by PARPi in vitro would also be evident in vivo, we turned to the PH039S PDX, a HGSOC model known to be exquisitely sensitive to rucaparib and niraparib due to RAD51C promoter methylation (Fig. 6a).^35,51 During the first 12 days of niraparib treatment, the tumors showed evidence of both senescence and apoptosis, consistent with our in vitro findings. Notably, by day 6 large, multinucleated cells became evident (Fig. 6b, c), indicating disruption of mitotic control in some cells. Immunohistochemical analysis revealed upregulation of CDKN1A and IL-8 as well as degradation of LMNB (Fig. 6d and Supplementary Fig. 6a), which are molecular hallmarks of cellular senescence. In the same PH039 PDX specimens, there was also a significant increase in apoptotic cells as marked by increased cleaved caspase 3 by day 9. However, by day 12 these apoptotic cells had largely been cleared, while giant, multinucleated cells persisted (Fig. 6e, f). PH039R, a niraparib-resistant model (Fig. 6g) derived from PH039S by repeated PARPi exposure,³⁵ also showed elevated expression of various components of SASP (Fig. 6h). Furthermore, PH039R exhibited elevated BCLX_L compared to PH039S, with an even greater increase observed in PH077R (Fig. 6i, and Suppl. Fig. 6b), a model derived from BRCA2-mutated PH077S by repeated niraparib treatment (Supplementary Figs. 6c and 6d).³⁵

Fig. 6.

BCLX_L inhibition potentiates PARP inhibitor response in a PARPi-resistant HGSOC PDX a Niraparib (100 mg/kg) or rucaparib (150 mg/kg) was administered to mice bearing the niraparib-sensitive ovarian cancer PDX PH039S and maximal tumor cross-sectional areas were measured by transabdominal ultrasound.^35,69 Lines show mean and 95% confidence intervals for growth trajectories by groups.⁷⁰ Pairwise comparisons for niraparib vs control, and rucaparib vs control, ***p < 0.0001. b Immunohistochemistry for Lamin B1 in PDX bearing mouse over 12 days of niraparib treatment. Red arrows indicate tumor cells with disrupted Lamin B. c For each group, 10 random fields were imaged from IHC represented in Fig. 6b and cells with disrupted Lamin B1 were counted as multinucleated. Error bars, ± SEM from three separate mice. * and **p < 0.05 and p < 0.01 for pairwise comparisons using one-way ANOVA with Dunnett’s multiple comparisons test. d Immunohistochemical analysis of SASP-related proteins in PH039S tumors treated with niraparib up to for 12 days. For each group 10 random fields were imaged and analyzed. Error bars, ± SEM from three separate mice. ****p < 0.0001 for indicated pairwise comparisons using one-way ANOVA with Dunnett’s multiple comparisons test. e, f cleaved caspase 3 immunohistochemistry in PH039S tumors treated with 100 mg/kg niraparib over 12 days. Red arrows indicate positive cells with cytoplasmic staining (e). 10 random fields were quantified for each PDX-bearing mouse (f). Error bars, ± SEM from three separate mice. Day 9 vs. S0, ****p < 0.0001 using one-way ANOVA with Dunnett’s multiple comparisons test. g Niraparib (100 mg/kg) or rucaparib (150 mg/kg) was administered to mice bearing the niraparib-resistant PH039 derivative PH039R and maximal tumor cross sectional area was measured by ultrasound. Pairwise comparisons for niraparib vs control, ns p = 0.091, and rucaparib vs control, ns p = 0.4433. h Immunohistochemical analysis of SASP-related genes in PH039R tumors harvested throughout resistance development. Insets show representative staining patterns in PDX tumors. i Immunoblotting for indicated BCL2 family proteins in PARP-sensitive (S) and resistant PDXs (R). Each lane represents a tumor from a single mouse bearing the PH039 models (20 µg protein/lane) or PH077 models (50 µg protein/lane). j, k A-1155463 was administered (5 mg/kg) in combination with niraparib (100 mg/kg) in PDX models PH039S with tumor maximal cross-sectional area (j) and overall survival (k) observed for up to 28 days. Pairwise comparisons for niraparib vs control, ***p < 0.0001; A-1155463 vs control, ns p = 0.7788; A-1155463 + niraparib vs control **p < 0.001 for plots in (j). l, m A-1155463 was administered (5 mg/kg) in combination with niraparib (100 mg/kg), in PDX models PH039R with tumor maximal cross-sectional area (l) and overall survival (m) observed for up to 28 days. Pairwise comparisons for niraparib vs control, ns p = 0.607; A-1155463 vs control, ns p = 0.6399; A-1155463 + niraparib vs. control, p = 0.0801 for plots in (l). The number of mice under observation is indicated below x- axis, where text color indicates treatment arm. Global log- rank p values (Mantel-cox) test are shown for plots in (k) and (m)

Given our results showing that A-1155643 enhances PARPi sensitivity in vitro (Figs. 4, 5), as well as previous reports that BCLX_L inhibition can promote lysis of senescent cells,³¹ we evaluated the combination of A-1155643 and niraparib in vivo. As indicated in Fig. 6j, k and Supplementary Fig. 6e, A-11555463 did not exhibit monotherapy activity in either PH039S or PH077S and did not enhance niraparib-induced tumor shrinkage in these sensitive models. Similarly, the combination was ineffective in the indolent PH077R model (Supplementary Fig. 6f, g). In contrast, adding A-1155463 to niraparib slowed tumor progression (p = 0.07) and enhanced overall survival [HR 0.17 (0.03, 0.90) log rank p-value = 0.022] relative to niraparib alone in the aggressive-PARPi resistant PDX model PH039R (Fig. 6l, m and Supplementary Fig. 6h).

Discussion

Despite the clinical success of PARPis in BRCA1- and BRCA2-mutated ovarian cancer, resistance remains a near-universal outcome.^52,53 Once ovarian cancer progresses on PARPi-based therapy, response rates to subsequent therapies are low (20-22%), and median overall survival is less than one year.⁵⁴ These sobering outcomes highlight an urgent and ongoing need to define targetable vulnerabilities that emerge during PARPi response and resistance. In the present study, we provide evidence for two quantitatively similar cell fates – apoptosis versus multinucleation – in BRCA1- and BRCA2-mutated ovarian cancer cell lines treated with therapeutically sustainable PARPi concentrations and demonstrate for the first time that (i) the multinucleated cells accumulating after PARPi treatment exhibit a senescent phenotype in vitro and in vivo, (ii) cell lines and PDXs selected for PARPi resistance have elevated levels and activity of the anti-apoptotic protein BCLX_L, (iii) selective BCLX_L inhibition sensitizes both treatment naïve and PARPi-selected cell lines to PARPi inhibitors ex vivo, with greater effects in PARPi-selected cells, and (iv) selective BCLX_L inhibition enhances PARPi efficacy in some PDX models with acquired PARPi resistance. This duality extends emerging concepts of therapy-induced senescence (TIS) as an adaptive survival state that contributes to residual disease and therapeutic failure.

Experiments presented above were focused on cell lines and PDXs that, like the clinical situation, have acquired PARPi resistance after initially being PARPi sensitive.^44,55 While reactivation of homologous recombination (HR) through BRCA1/2 reversion mutations or epigenetic reprogramming accounts for a proportion of resistance in the clinical setting,⁵⁶ a significant subset of tumors acquires PARPi resistance through alternative mechanisms that remain incompletely understood.^57–59 Our findings point to one such mechanism that can occur in BRCA1- and BRCA2-mutant ovarian cancer cells: A dynamic cell fate bifurcation in response to PARP inhibition wherein some PARPi-sensitive cells initiate apoptosis through a BCLX_L-modulated pathway, but a substantial proportion also become multinucleated, survive for prolonged periods, and exhibit a pro-inflammatory senescence-associated phenotype. Using two separate PARPi-sensitive ovarian cancer cell lines, BRCA2-mutated PEO1 cells and BRCA1-mutated COV362 cells, we observed both multinucleation and apoptosis within six days of PARPi exposure in vitro, underscoring the relevance of both fates across distinct genetic backgrounds. Importantly, in PARPi sensitive lines, these diverging fates were observed at therapeutically sustainable 1–2 µM olaparib concentrations (Fig. 1). Moreover, once formed, the multinucleated cells persisted for at least 12 days without an apparent increase in cell number and exhibited several features of senescent cells, including increased expression of senescence-associated markers such as IL-6, IL-8, CDKN1A and CDKN2A as well as positive staining for senescence-associated β-galactosidase, linking PARPi-induced multinucleation and senescence for the first time.

Mechanistically, we found that PARPi-induced apoptosis in sensitive ovarian cancer cell lines depends on BAX/BAK activation and is driven by transcriptional upregulation of BH3-only proteins, including NOXA, PUMA, and the more potent splice variants of BIM.⁶⁰ NOXA, PUMA, and BIM_EL were sequestered by BCLX_L, identifying this anti-apoptotic protein as a key resistance node even in sensitive ovarian cancer cells (Fig. 2). Our observation that pro-apoptotic proteins accumulate in PARPi-treated cells but are held in check by anti-apoptotic counterparts (Fig. 2e, f) is consistent with the concept of “apoptotic priming”.⁶¹ Based on this conceptual framework, as well as reports that BCLX_L inhibition can induce the demise of senescent cells,^31,33,34, we examined the effects of the selective BCLX_L inhibitor A-1155463 on PARPi-induced cytotoxicity. These studies demonstrated that A-1155463 enhances PARPi-induced apoptosis in PARPi-sensitive cell lines as well as multiple ovarian cancer cell lines selected for PARPi resistance in vitro (Figs. 4, 5, and Supplementary Figs. S4, S5), suggesting that this approach counteracts pro-survival cell states that are favored during resistance development. Thus, targeting the apoptotic threshold in both the non-senescent and senescent cells may offer a therapeutic opportunity to eradicate residual HGSOC and delay or prevent resistance.

This bifurcation of cell fates was recapitulated during PARPi therapy in vivo. On the one hand, several days of niraparib treatment induced an increase in the apoptotic marker cleaved caspase-3 (Fig. 6e, f) in the highly sensitive PDX PH039.^35,51 On the other hand, a substantial population of tumor cells also entered a pro-survival state, as evidenced by the induction of multinucleated cells with hallmarks of senescence (Fig. 6b–d). What is unclear at the present time is whether pre-existing expression of pro- and anti-apoptotic BCL2 family members, extent of DNA damage, cell cycle stage at the time of initial PARPi exposure, or other factors determine whether an individual PARPi treated cell will undergo apoptosis vs. enter a multi-nucleated state.

Our further experiments showed that the effect of BCLX_L inhibition was context-specific. While it had some impact in PARPi-sensitive tumors already undergoing effective apoptosis, its benefit was more pronounced in resistant cell lines or PDXs dominated by pro-survival, multinucleated cells. This underscores the potential of using senescence or apoptotic biomarkers to stratify future patients who may benefit from BCLX_L targeting strategies once improved methods for targeting BCLX_L, e.g., through selective delivery of BCLX_L antagonists or BCLX_L degraders as the payloads of antibody-drug conjugates, become available.

In summary, our study highlights a previously underappreciated duality in cell fate following PARPi exposure, with apoptosis and senescence co-existing within tumors. The pro-survival state of multinucleated, senescent-like cells represents a therapeutic vulnerability as acquired resistance develops, which can be exploited by combining PARPi with BCLX_L inhibition. Together, these findings support further investigation into BCLX_L-targeting strategies—particularly in combination with PARPis—as a rational and potentially impactful approach to treat resistant ovarian cancers.

Methods

Materials

Annexin V conjugated to allophycocyanin (APC) was purchased from BD Biosciences (Franklin Lakes, NJ). N-(2-quinolyl)valyl-aspartyl-(s2,6 difluorophenoxy)methyl ketone (Q-VD-OPh) was from SM Biochemicals (Pasadena, CA). Venetoclax, S63845, and A-1155463 were from Chemietek (Indianapolis, IN). Reagents for the detection of senescence-associated β-galactosidase were purchased from Cell Signaling Technology (Danvers, MA).

Antibodies were from the following suppliers: Murine monoclonal anti-BCL2 (CAT #M0887) from Dako (Carpenteria, CA); goat anti-β-actin (sc-1615) and murine monoclonal anti-PUMA antibodies (sc-3742230 from Santa Cruz Biotechnology (Dallas, TX); murine monoclonal anti-Noxa antibody (cat # ALX-804-408-C100) from Enzo (Farmingdale, NY); and rabbit antibodies to BAK (#6947), BAX (#2772), BCLX_L (#2764), BIM (cat #2819), and MCL1 (cat #4572) from Cell Signaling Technology.

Cell Lines

PEO1 and PEO4 cells,⁴³ which were derived from the same patient when her ovarian cancer was platinum sensitive and subsequently platinum resistant, were cultured in DMEM medium with 4.5 gm/L glucose, 10 μg/ml insulin, 0.04 mM nonessential amino acids, and 10% (vol/vol) FCS (medium A). ABT^R2 and ABT^R3, two clones selected from PEO1 cells by continuous PARPi treatment⁴⁴, were passaged in medium A containing 40 µM veliparib until 3–5 days before clonogenic assays and then transferred to drug-free medium A. Parental COV362 cells as well as COV362 olap^R cells, which were derived by continuous exposure to increasing concentrations of olaparib from 0.5 to 10 µM over a 6-month period in vitro,⁶² were grown in RPMI medium 1640 supplemented with 10% (vol/vol) FCS (medium B). All media contained 40 units/ml penicillin G, 40 μg/ml streptomycin, and 1 mM glutamine. All cell lines were authenticated by short tandem repeat analysis in the Mayo Clinic Cytogenetics Core Facility and assayed for mycoplasma regularly.

CRISPR/Cas9-mediated interruption of the FADD or BAX and BAK genes was accomplished as recently reported.⁴⁸

Immunofluorescence

Cells grown on coverslips were treated with 2 µM olaparib or 25 nM paclitaxel before fixation in 4% (wt/vol) paraformaldehyde. Actin filaments were visualized with Alexa Fluor™ 488-conjugated Phalloidin (Invitrogen, Waltham, MA). Nuclei were stained with 1 µg/ml Hoechst 33258 in PBS. Cells were examined on a Zeiss Axiovert microscope with 63X NA 1.4 lens and photographed using a Zeiss Axiocam MRm CCD camera using Zeiss Zen software. 100 cells from each of ≥ 3 independent experiments were imaged and counted.

Immunoblotting

Whole cell lysates⁶³ or immunoprecipitates were subjected to SDS-PAGE and immunoblotting with enhanced chemiluminescence detection as previously described.⁶⁴ Densitometric analysis was performed using ImageJ software, To ensure quantification within the linear detection range, internal protein dilution controls (25, 12.5, and 5 µg) were included and used to validate signal proportionality.

Immunoprecipitation

After cells were treated for 6 days with 2 µM olaparib in the presence of 10 µM Q-VD-OPh to inhibit apoptosis, anti-apoptotic BCL2 family members were immunoprecipitated to identify activated BH3-only proteins.⁶⁵ All immunoprecipitation steps were at 4 °C. Cells were lysed in CHAPS lysis buffer [20 mM HEPES, 150 mM NaCl, 1% (w/v) CHAPS, 1% (v/v) glycerol, 1 mM PMSF, 10 μg/mL leupeptin, 10 μg/mL pepstatin, 100 mM NaF, 10 mM sodium pyrophosphate, 1 mM sodium vanadate, and 20 nM microcystin, pH 7.4] for 30 min. After removal of insoluble material at 14,000 g for 15 min, lysates were precleared by incubation for 1 h with protein G-agarose beads. Precleared lysates (600 µg protein) were incubated overnight with antibodies to BCL2, MCL1, or BCLX_L that were precoupled to protein G-agarose using dimethyl pimelimidate.⁶⁶ Following 4 washes with isotonic wash buffer containing 1% CHAPS, bound polypeptides were solubilized at 65 °C in SDS sample buffer [2% (w/v) SDS, 4 M urea, 62.5 mM Tris-HCl (pH 6.8) and 1 mM EDTA] for immunoblotting.

Annexin V assays

After treatment for the indicated times, cells were washed and stained with APC-conjugated Annexin V and 0.1 μg/ml propidium iodide in 140 mM NaCl, 2.5 mM CaCl₂, and 10 mM HEPES (pH 7.4). Cells (20,000 events) were analyzed using the FL2 (excitation: 488 nm; emission: 585/21 nm) and FL4 channels of a BD Biosciences FACSCanto II flow cytometer and CellQuest software.

Senescence-associated β-gal staining

After treatment with olaparib (2 µM), PEO1 and COV362 cells were harvested at the indicated times, fixed, and stained for β-galactosidase activity (Cell Signaling #9680) per the supplier’s protocol except that staining volumes were adjusted for cells plated on chamber slides. Cells were examined on an Olympus CKX41 microscope, and seven random fields were imaged using a 40X UplanFL N Ph2 lens (N.A. 0.75) and a Nikon D90 camera.

Clonogenic assays and assessment of synergy

To perform colony-forming assays, 750 cells were plated in 3 ml medium A per 60 mm well or dish, incubated for 4–16 h to allow cells to adhere, and treated continuously with the indicated agent(s) added from 1000X stocks in DMSO or with diluent [0.2% (v/v) DMSO]. When colonies became visible on the control plates (typically 10–14 days), colonies were stained with Coomassie brilliant blue and counted manually under low magnification. Data were analyzed by the median effect method⁶⁷ under the assumption that effects of the drugs are mutually exclusive, which renders this analysis equivalent to isobologram analysis.⁶⁸ According to this method, a combination index (CI) < 1 indicates synergy, CI = 1 indicates additivity and CI > 1 indicates antagonism.

PDX studies

Studies in mice were performed according to institutional guidelines after approval by the Mayo Clinic Institutional Animal Care and Use Committee (protocol A-23114). For all PDX studies, cryogenically preserved human ovarian cancer tumors were rapidly thawed and reestablished in female SCID Beige mice (C.B-17/IcrHsd-Prkdc^scidLyst^bg-J; Envigo, Indianapolis, IN) as previously described.⁶⁹ Briefly, 0.1–0.2 cc of minced tumor was prepared in 1:1 ratio with McCoy’s 5 A Modified Medium (Cat # MT-10-050-CV, Corning Life Science) before intraperitoneal injection. Sensitive tumors (PH039S and PH077S) and their niraparib-selected counterparts (PH039R and PH077R)³⁵ were matched for passage number. When tumor cross-sectional areas reached 0.3–0.5 cm² by transabdominal ultrasound, mice were randomized (6–7 mice/group) to the following treatments administered by daily gavage for 28 days: (i) Niraparib 100 mg/kg/day; (ii) A-1155463 5 mg/kg/day; or iii) Niraparib + A-1155463 at the same doses. Mice in the main study were euthanized when the tumor burden exceeded 10% of total body weight, ascites developed, body conditioning score dropped to 1 or as a cohort at day 28. Ultrasound measurements were taken weekly and normalized to the starting tumor area for each mouse.

Immunohistochemical (IHC) Staining

Mouse tissues were harvested and fixed overnight in buffered formalin (Fisher Scientific; #23-011-120), then processed and stained with hematoxylin and eosin by the Mayo Clinic Histology Core. Immunohistochemical staining was performed by the Mayo Clinic Pathology Research Core using a Leica Bond RX stainer (Leica, Allendale, NJ). Tissues were sectioned at 5 microns, subjected to antigen retrieval for 30 min using Epitope Retrieval 2 (EDTA; Leica), and incubated for 5 min in Protein Block (Dako). The IL8 antibody (LS-B6427, LifeSpan), p16 antibody (ab10849, Abcam), p21 antibody (2947S, Cell Marque), or cleaved caspase 3 antibody (9661, Cell Signaling) were incubated with slides for 30 min at concentrations previously determined to provide optimal staining. Bound primary antibody was detected using a Leica Polymer Refine Detection Kit according to the supplier instructions. Slides were counterstained on-line with hematoxylin for 5 min and rinsed several times in 1X Bond wash buffer followed by distilled water. After removal from the stainer, slides were rinsed for 5 min in tap water, dehydrated in increasing ethanol concentrations, cleared in xylene, and coverslipped.

Seven random fields from each slide were independently scored for staining by two scorers on a Zeiss Axiophot microscope equipped with a Zeiss Plan-Apochromat 63X N.A. 1.4 lens and photographed. The two scores were then averaged.

Quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR)

RNA was extracted from either frozen cell pellets or snap-frozen tissue using a RNeasy Plus Minikit (Qiagen). 100 ng RNA and a TaqMan RNA-to-CT 1-Step Kit (Applied Biosystems, Carlsbad, CA) were used to perform qRT-PCR in triplicate. PCR was completed on a CFX384 Real Time System (C10000 Touch Thermal Cycler, BioRad, Hercules, CA) using probes for GAPDH (4352665; Life Technologies), IL6 (Hs00174131_m1), IL8 (Hs00174103_m1), CDKN2A (Hs00923894_m1), and CDKN1A (Hs00355782_m1). Data were analyzed using the following equations: ΔΔC_t = ΔC_t(sample)- ΔC_t (endogenous control); and Fold Change = 2^−ΔΔCt.

Statistical analysis

Dose-response curves were performed in cell lines at least three times independently unless otherwise indicated. Error bars in all experiments represent mean ± sd of 3 independent experiments. *, ** and *** indicate p < 0.05, p < 0.01 and p < 0.001, respectively, using unpaired t tests. Survival curves were plotted by the Kaplan-Meier method and analyzed using Cox proportional hazard models.

Additional methods

Methods for proteomics and quantitative analysis of PDX growth in vivo are described in Supplementary Methods.

Author contributions

A.V. contributed to the conception of the project, performed and analyzed experiments, prepared figures, and wrote the manuscript; A.S. performed and analyzed experiments, prepared figures, and contributed to manuscript preparation; C.C. performed the bioinformatic analysis, prepared figures, contributed to results interpretation, and manuscript preparation; X.H., C.D.M., E.A.B., and H.D. performed experiments, analyzed data, and helped prepared figures for the manuscript; R.M.H. contributed to the conception of the project, performed experiments, and contributed to funding. O.K.R., C.W., J.M.W., S.S., K.S.F., K.L.P., P.A.S., L.N.D., S.D., X.W., and A.P. performed experiments, analyzed data, and helped prepare figures for the manuscript; A.E.H. contributed to funding and supervision; M.C.L performed the statistical analysis and prepared figures for the manuscript; X.W.M. generated knockout lines; A.L.O provided statistical expertise, analyzed data and contributed to manuscript preparation; H.L. provided computational expertise and funding; S.J.W. contributed to securing funding, planning of animal studies, interpretation of data and preparation of manuscript; S.H.K. conceived the project, secured funding, performed experiments, interpreted data, and wrote the manuscript. All authors have read and approved the article.

Data availability

The mass spectrometry global proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD037989.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41392-026-02870-7.