Mitochondrial superoxide sustains a senescence-like phenotype in PARP- inhibited ovarian cancer cells by stabilizing HIF1α

Abstract

Ovarian cancer is the deadliest gynecological malignant tumor and is known as the “silent killer”. PARP inhibitors are being increasingly used for their excellent efficacy in the treatment of ovarian cancer. While PARP inhibitors are known to interfere with DNA repair and cause DNA damage, the fates of cancer cells and associated metabolic features in response to PARP inhibition are not well characterized. We herein show that ovarian cancer cells treated with PARP inhibitors exhibit a senescence-like phenotype that is characterized by cell cycle arrest, positive staining of senescence-associated β-gal, and increased accumulation of dysfunctional mitochondria. The survival of senescence-like cells is sustained by glycolysis that is driven by an augmented axis of mitochondrial reactive oxygen species (mtROS) and HIF1α. Mitochondrial antioxidant, inhibition of HIF1α activation and restriction of glycolysis can each block the entry into and the sustenance of the senescence-like state in PARP-inhibited ovarian cancer cells. The senescence-like phenotype, HIF1α activation and lactate production were attenuated in tumor xenografts co-treated with PARP inhibitor Rucaparib and mitochondrial antioxidant. The metabolic reliance on mtROS-driven glycolysis in ovarian cancer cells treated with PARP inhibitors has implications in cancer treatment.

Graphical abstract

Highlights

• •PARP inhibition induces a senescence-like phenotype in ovarian cancer cells.
• •PARP inhibition causes mitochondrial dysfunction and mtROS buildup.
• •Survival of senescence-like cells relies on glycolysis driven by mtROS-HIF1α axis.
• •Mitochondrial antioxidant destabilizes HIF1α and abrogates senescence-like phenotype in vivo.

1. Introduction

As one of the most malignant tumors of the female reproductive system, ovarian cancer has the characteristics of low incidence and high mortality rate. The high mortality is because its symptoms in early stage are more insidious, and effective early detection indicators are relatively lacking, resulting in more than 70% of patients being diagnosed in an advanced stage [1,2]. Most patients with advanced disease will relapse within 3 years after primary surgery and chemotherapy [3,4]. A breakthrough was made in 2005 when two groups reported the sensitivity of BRCA1/2 mutant cancers to PARP inhibitors [5,6]. In recent years, many clinical studies have shown that PARP inhibitors have excellent efficacy in the treatment of ovarian cancer with and without homologous recombination deficiency (HRD) [[7], [8], [9], [10]].

PARP is an enzyme that catalyzes the synthesis of poly (ADP-ribose) (PAR), a protein post-translational modification on many target proteins and plays a critical role in the repair of single-stranded breaks (SSBs). By inhibiting the activity of PARP, PARP inhibitors can interfere with SSBs repair. SSBs, if not repaired, are converted to DSBs during S phase and are repaired by homologous recombination. Thus, HRD cells, like BRCA1 or BRCA2 mutant cancer cells, are highly sensitive to PARP inhibitors [11]. Chemicals that downregulate homologous recombination repair can also render cancer cells more sensitive to PARP inhibition [12,13]. PARP inhibitors were also shown to remarkably benefit cancer patients that are not HRD [9,14,15]. The HRD-independent therapeutic effects of PARP inhibitors could be due to increase in oxidative stress [16,17], induction of cGAS-STING mediated immune response [18], among others. It should be noted that although PARP inhibitors have a good therapeutic effect in the clinical application, almost all patients with initial sensitivity to PARP inhibition eventually develop resistance. Therefore, understanding how cancer cells become resistant to PARP inhibitors can help develop more effective treatment options.

Cancer cells usually harbor a distinct metabolic state called Warburg Effect, a metabolic preference for aerobic glycolysis in tumor cells even in the presence of functioning mitochondria [19]. Cancer cells generally exhibit enormous metabolic diversity and plasticity to meet the increased demand for bioenergy and biosynthesis or to adapt to various types of stress, including those induced by radiotherapy or chemotherapy [20]. The metabolic choices in cancer cells can determine cell fate decisions and drug sensitivity [[21], [22], [23]]. In a murine lymphoma model of chemotherapy-induced senescence, for example, blockade of glucose utilization resulted in regression of the tumor mass and improved treatment outcomes [24]. Therefore, the metabolic vulnerability of cancer cells can be exploited for more effective cancer management.

The fates of PARP-inhibited cancer cells range from cell cycle arrest, cellular senescence to cell death [16,[25], [26], [27], [28]]. While cellular senescence can prevent the propagation of damaged or stressed cells, the senescent cells may secrete factors, referred to as senescence-associated secretory phenotype or SASP, that remodel tissue microenvironment and promote regeneration [29]. SASP may also spread adverse effects of chemotherapy and promote cancer relapse. A recent study show that PARP inhibitors can induce reversible cellular senescence in ovarian cancer cells and application of senolytics can have synthetic lethal effects [26]. Thus, clearance of senescent cells should reduce cancer risk. While many senolytic agents have been developed to target the pro-survival or anti-apoptotic proteins to promote cell death, whether senescent or senescence-like cells possess other vulnerable features that can be exploited for their clearance or containment has not been well explored.

Herein, we explored the fate of ovarian cancer cells treated with PARP inhibitors and the associated bioenergetic and metabolic features. We observed that PARP-inhibited (PARPi) ovarian cancer cells exhibited senescence-like phenotype that is sustained by an axis of mtROS-HIF1α-glycolysis. Quenching of mtROS, inhibition of HIF1α and blockade of glycolysis can each diminish the senescence-like cancer cells caused by PARP inhibition.

2. Materials and methods

2.1. Cell lines

HEY, A2780, SKOV3 and HO8910 cells were as previously described [16]. OVCAR3 cells were purchased from Shanghai Cell Bank, Chinese Academy of Sciences (Shanghai, China). The genetic and epigenetic alterations of these ovarian cancer cell lines (except HO8910) have been well characterized [30]. OV6314622 cells were derived from tumor tissues of a patient with high grade serous ovarian cancer who received surgery at Department of Obstetrics and Gynecology of the First Affiliated Hospital of Soochow University. HEY, A2780 and OV6314622 cells were cultured in DMEM medium (Gibco, Invitrogen). SKOV3 cells were cultured in McCoy's 5A medium. HO8910 and OVCAR3 cells were cultured in RPMI 1640 medium (Gibco, Invitrogen). All media contained 10% FBS (Gibco, Invitrogen), 100 units/mL penicillin, and 100 μg/mL streptomycin. All cells were cultured in a humidified atmosphere of 5% CO₂ at 37 °C.

2.2. Reagents and antibodies

Rucaparib phosphate (S1098), Olaparib (S1060), KC7F2 (HIF1α inhibitor, S7946), DMOG (HIF1α agonist, S7483), Doxorubicin (E2516) and oligomycin A (S1478) were purchased from Selleck. Mitoquinone mesylate (MitoQ, HY-100116A) and 2-Deoxy-D-glucose (2-DG, HY-13966) were purchased from MedChemExpress. CCCP (Y109220) was purchased from Beyotime. Anti-Lamin B1 (ab16048) antibody was purchased from Abcam. Anti-Hexokinase II (2867S), anti-HIF1α (36169s) and anti-β-Actin (4970S) were purchased from Cell Signaling Technology. Anti-HIF1α (PA1-16601) was purchased from Thermo Fisher Scientific. Anti-p-H3 (06-570) was purchased from Millipore. Anti-Tomm20 (11802-1-AP) was purchased from Proteintech. Antibody 8-OHdG (sc-393871) was purchased from Santa Cruz Biotechnology.

2.3. SA β-galactosidase detection

SA β-galactosidase activity was measured by senescence-associated β-galactosidase staining kit (Beyotime, C0602) or SPiDER-β-gal reagent (Dojindo, SG02). Cells grown in 6-well plates or frozen sections of tumor grafts were washed once with 1 × PBS and fixed with fixative solution for 15 min, then washed for 3 min with PBS and finally incubated overnight at 37 °C in a staining solution. Afterward, cells or sections were washed twice with PBS and pictures were taken for quantification. SA-β-gal positive cells were quantified using ImageJ. For fluorescence staining of senescent cells by flow cytometry, tumor tissues were digested and single-cell suspensions were prepared, then SPiDER-β-gal staining solution (1 μM) was added for 15 min and washed three times with PBS. The stained cells were then subjected flow cytometry analysis.

2.4. Western blot analysis

Protein samples (30-50 μg) were separated by SDS-PAGE (6-12%) and electro-transferred onto PVDF membrane. The membrane was blocked with 5% no fat milk for 60 min and incubated with specific primary antibodies at 4 °C for overnight. Proteins of interest were measured with appropriate horseradish peroxidase-conjugated secondary antibodies for 60 min at room temperature and developed using ECL-enhanced luminol reagent. The protein levels were normalized by β-Actin. Band intensities were quantified using ImageJ.

2.5. RNA isolation and real-time quantitative PCR

Total RNA was extracted from cultured cells using Trizol reagent (Invitrogen, 15596026), and was reverse transcribed using reverse transcriptase (Vazyme, R323-01). Relative mRNA levels were quantified by qPCR using SYBR GREEN mix (Vazyme, Q712-02). Human β-actin gene was used as the endogenous control. The comparative cycle threshold method (ΔΔCt) was used to quantify the relative expression of each gene. The samples were loaded in triplicate, and the results of each sample were normalized to β-actin. The primers used in the present study were listed in Supplementary Table 1.

2.6. RNA interference

Transient silencing of HIF1α was performed with small interfering RNAs (siRNAs) purchased from Sigma-Aldrich. For control experiments, cells were transfected with a similar amount of non-targeting control siRNA (Sigma-Aldrich). Cells were transfected with siRNAs using lipofectamine 2000 (Sigma-Aldrich) according to the manufacturer's protocol. And siRNA sequences were listed in Supplementary Table 2.

2.7. Bulk RNA sequencing

Total RNA of treated samples was extracted using Trizol reagent (Thermo Fisher, 15596018) following the manufacturer's procedure. The total RNA quantity and purity were analyzed using Bioanalyzer 2100 and RNA 6000 Nano LabChip Kit (Agilent, CA, USA, 5067-1511), high-quality RNA samples were used to construct sequencing library. The RNA libraries were prepared and sequenced on an Illumina Novaseq™ 6000 platform by LC Bio Technology Co., Ltd. (Hangzhou, China). Differentially expressed genes, KEGG, GO and GSEA analyses were performed using the LC Bio Technology online analysis platform.

2.8. Immunofluorescence staining

Immunofluorescence staining was carried out as described previously [31]. Briefly, cells grown on coverslips or frozen sections of tumors were fixed in 4% paraformaldehyde for 10 min. The cells were then permeabilized in 0.2% Triton X-100 for 10 min, and then blocked in 10% normal goat serum overnight at 4 °C. The coverslips were incubated with anti-Lamin B1, p-H3, HIF1α, 8-OHdG and Tomm20 antibodies overnight at 4 °C, washed in PBS, and incubated with secondary antibody for 60 min at room temperature. Cells were washed in PBS three times and counterstained with DAPI (Beyotime, C1002). Fluorescence images were captured under a fluorescence microscope. Immunofluorescence quantification analysis was performed using ImageJ.

2.9. 2-NBDG uptake assay

The glucose uptake was measured using the fluorescent glucose derivative 2- (N- (7-nitrobenz-2-oxa-1,3-diazol-4-yl) amino) -2-deoxyglucose (2-NBDG) (Invitrogen, N13195). Cells treated with or without PARP inhibitors were subjected to 2-NBDG at the final concentrations of 5 μg/mL. Cells were incubated in a humidified atmosphere of 5% CO₂ at 37 °C for 60 min. Then cells were collected and washed once with PBS. The fluorescence signals were measured by flow cytometry.

2.10. Hexokinase activities and lactate production assays

Cellular hexokinase (HK) activities and lactate production were measured by HK activity assay kit (Solarbio, BC0745) and Lactate colorimetric assay kit II (BioVision, K627-100) according to the manufacturers’ instructions. All experiments were normalized by the cell numbers.

2.11. Measurement of oxygen consumption rate and glycolytic capacity

The cellular oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were analyzed by the Seahorse XFe24 Extracellular Flux Analyzer (Seahorse Bioscience) according to the manufacturer's instructions. OCR was measured using the Cell Mito Stress Test Kit (Agilent, 103015-100) in Seahorse XF Base Medium containing 1 mM sodium pyruvate, 2 mM glutamine and 10 mM glucose in basal conditions. ECAR was measured using the Glycolysis Stress Test Kit (Agilent, 103020-100) in Seahorse XF Base Medium containing 2 mM glutamine in basal conditions. The results were analyzed using Wave software (Seahorse/Agilent).

2.12. Determination of glycolytic metabolites

Cells were counted and harvested in liquid nitrogen after treatment. The cell pellets were thawed at 4 °C and 80 mg of each sample was mixed with 25 μL isotope internal standards and 400 μL of cold methanol/acetonitrile solution (1:1, v/v). The lysate was homogenized by MP homogenizer (20 s, thrice), adequately vortexed, then sonicated for 5 min at low temperature, followed by incubation at −20 °C for 60 min. The mixture was centrifuged for 20 min (14000 rcf, 4 °C). The supernatant was transferred to Ostro plate, and collected solution was dried in a vacuum centrifuge, the samples were re-dissolved in 150 μL acetonitrile/water (1:1, v/v) and adequately vortexed, and then centrifuged (14000 rcf, 4 °C, 15 min). The supernatants were collected for LC-MS/MS analysis. Targeted metabolic analysis was performed using an LC-MS/MS system by Applied Protein Technology Co., Ltd. (Shanghai, China).

2.13. Measurement of ROS and mitochondrial mass

Total ROS were measured using oxidation sensitive fluorescent probe (DCFH-DA) according to the manufacturer's protocols (Beyotime, S0033). Mitochondrial ROS were measured using MitoSOX™ Red mitochondrial superoxide indicator (Thermo Fisher Scientific, M36008) or MitoPY1 mitochondrial hydrogen peroxide indicator (Macklin, M995074) following manufacturer's instructions. MitoTracker Green (Beyotime, C1048) was used for the measurement of mitochondrial mass. Briefly, single cells were collected from cultured cells or digested tumor tissues, and resuspended in preheated base medium with 10 μM DCFH-DA, 5 μM MitoSOX Red, 5 μM MitoPY1 or 100 nM MitoTracker Green. Then the cells were incubated in a humidified 5% CO₂ atmosphere at 37 °C for 20 min in the dark. Finally, the cells were washed three times with PBS and analyzed by flow cytometry.

2.14. Determination of Mitochondrial Membrane Potential (MMP)

The mitochondrial membrane potential (MMP) was determined using a JC-1 assay kit (Beyotime, C2006) according to the manufacturer's instructions. Briefly, cells were harvested and incubated with JC-1 staining solution in a humidified 5% CO₂ atmosphere at 37 °C for 20 min in the dark. Then cells were washed three times with PBS and analyzed by flow cytometry. Mitochondrial depolarization is indicated by a decrease in the red/green fluorescence intensity ratio.

2.15. Transmission electron microscopy

Cells (>1 × 10⁶) were collected in a 15 mL centrifuge tube, fixed with 2.5% glutaraldehyde for 4 h. Then cells were rinsed with PBS and fixed with 1% osmium tetroxide (OsO4) in 0.1 M PBS for 60 min at room temperature. Following dehydration with a graded alcohol series, cells were embedded in Epon resin and ultrathin sections (60 nm) of the selected areas were cut using the Leica EM UC7 microtome using a diamond knife. Cells were then stained with a saturated solution of uranyl acetate in methanol (50:50) for 12 min at 45 °C, followed by incubation in an aqueous solution of concentrated bismuth subnitrate for 10 min at 25 °C. Subsequently, ultrathin sections were observed under a transmission electron microscope (HT7700, Hitachi, Japan).

2.16. Flow cytometry measurement of HIF1α

Cells were collected and made to single cell suspensions. Cells were fixed by a fixation/permeabilization kit (Thermo Fisher Scientific, 00-5521-00) at room temperature for 30 min and were then stained with anti-HIF1α antibody (Cell Signaling Technology, 36169s) in the presence of blocking solution (1% FBS in permeabilization buffer) for 30 min. Cells were next incubated with fluorochrome-coupled secondary antibody (Abcam, ab150075) at room temperature for 30 min. After PBS washing, fluorescence signals were recorded using flow cytometry.

2.17. Cell cycle analysis

Cells were harvested at various time points after PARP inhibitor treatment, washed once with cold PBS, and then fixed in 70% cold ethanol at −20 °C overnight. The fixed cells were washed with cold PBS once. The cell pellets were resuspended and incubated with 0.5 mL of PBS containing 50 μg/mL Propidium Iodide (Beyotime, ST511) and 100 μg/mL RNase A (Sangon Biotech, B500474) at room temperature for 30 min. Cell cycle was measured by flow cytometry. The fractions of cells in G1 phase, S phase and G2/M phase were determined using Flowjo software.

2.18. Cell proliferation assay

Cell death was evaluated by flow cytometry. In brief, the same number of cells were seeded in 6-well culture plates. Then cells were cultured in a humidified atmosphere of 5% CO₂ at 37 °C. After treatment, cells were harvested and stained with 50 μg/mL Propidium Iodide (Beyotime, ST511) at room temperature for 20 min. Cells that are stained positive for Propidium Iodide under flow cytometry were regarded dead cells. MTT assay was used to examine cell viability. In brief, cells were seeded in 96-well culture plates and kept in a humidified atmosphere of 5% CO₂ at 37 °C. After cells were subjected to the indicated treatment, 10 μL of MTT reagent was added to each well and incubated in a humidified atmosphere of 5% CO₂ at 37 °C for 4 h. The absorbance of each well was measured at 490 nm. All experiments were repeated at least three times.

2.19. Apoptosis analysis

Apoptotic cells were identified using APC Annexin V (BioLegend, 640941) and Propidium Iodide (Beyotime, ST511). Briefly, cells were harvested and washed twice with PBS. Then cells were resuspended in 100 μL of 1 × annexin-binding buffer and incubated with Annexin V and Propidium Iodide at room temperature for 20 min in the dark. The results were analyzed by flow cytometry.

2.20. Tumor xenografts

Female NCG (NOD/ShiLtJGpt-Prkdc^em26Cd52IL2rg^em26Cd22/Gpt) mice (6- to 8-weeks old) were purchased from GemPharmatech Corporation and kept in pathogen-free conditions and handled in accordance with the requirements of the Guideline for Animal Experiments. Animals were subcutaneously inoculated with 5 × 10⁶ OVCAR3 cells (suspended in 100 μL PBS). Animals were randomly divided into two groups and treated with either control (DMSO) or Rucaparib (20 mg/kg given daily by intraperitoneal injection). Animals were randomly divided into four groups and treated with either control (DMSO), MitoQ alone (2 mg/kg given every other day by oral gavage), Rucaparib alone (20 mg/kg given daily by intraperitoneal injection) or 20 mg/kg Rucaparib plus 2 mg/kg MitoQ. Tumor growth was measured daily with a caliper. Tumor volumes were calculated with the use of the following formula: tumor volume (mm³) = a × b²/2, where ‘a’ is the maximal diameter, ‘b’ is the perpendicular diameter. After the last treatment on day 18, tumors were excised and weighed. All tumors were fixed in 4% paraformaldehyde or digested, and were subjected to subsequent experiments. Cell suspensions were made from digested tumor tissues. Mouse CD45 antibody staining was used to select for CD45-negative cells for analysis. All animal experiments were approved by the Institutional Animal Care and Use Committee of Soochow University.

2.21. Immunohistochemistry of tumor xenografts

Paraformaldehyde-fixed and paraffin-embedded tumor xenografts were sectioned at 4 μm. After deparaffinization and rehydration, the sections were boiled in citrate sodium buffer (pH 6.0) for 15 min for antigen recovery, and immersed in 3% H₂O₂ for 10 min to quench endogenous peroxidase. Nonspecific binding was blocked by 10% donkey serum at room temperature for 60 min. Then sections were incubated with anti-HIF1α antibody (Thermo Fisher Scientific, PA1-16601) overnight at 4 °C. The signal was developed by DAB detection system and counterstained with hematoxylin. The positive signals were quantified using ImageJ.

2.22. Statistical analysis

All statistical data are presented as mean ± standard deviation (S.D.). ANOVA was used to compare significant differences among multiple experimental groups. Two-sided Student's t-test was used for comparisons between two groups of experiments. Differences were considered significant when p values were below 0.05. Statistical analyses were carried out using Graphpad Prism software (version 8). ∗, ∗∗, ∗∗∗ and ∗∗∗∗ indicate p < 0.05, p < 0.01, p < 0.001 and p < 0.0001, respectively. ns stands for no significant difference.

3. Results

3.1. PARP-inhibited ovarian cancer cells exhibit a senescence-like phenotype

We previously reported that PARP inhibitor-treated (PARPi) ovarian cancer cells were arrested at the G2/M phase [16]. We further confirmed this by flow cytometry analysis of cell cycle distribution and by immunostaining of p-H3, a marker of the G2/M phase, in cancer cells treated with PARP inhibitors (Fig. S1A and B). We next profiled gene expression in ovarian cancer cells treated with PARP inhibitor Rucaparib by bulk RNA-seq. The top 20 Biological Process Enrichment confirmed the cell cycle abnormalities in PARPi cells (Fig. S1C). We observed that PARPi ovarian cancer cells tend to undergo senescence, as shown by the positive SA-β-gal staining (Fig. 1A and S1D), in contrast to insignificant or very mild induction of apoptosis (Fig. S1E). Characteristic of senescent cells, they were enlarged in sizes, as revealed by flow cytometry (Fig. S2A). Those cells also exhibited senescence-associated secretory phenotype (Fig. 1B). Moreover, the expression of Lamin B1, which is usually downregulated in senescent cells, was reduced when measured using Western blot and immunofluorescence staining (Fig. 1C and D). We further examined the expression of two classical senescence markers p21 and p16. While the mRNA levels of CDKN1A (encoding p21) increased significantly in all three cell lines treated with either Rucaparib or Olaparib, the mRNA levels of CDKN2A (encoding p16) did not show a uniformed response to the PARP inhibitors in the three cell lines, which is consistent with the reported hypermethylation or deletion of CDKN2A (Fig. 1E and F) [30]. Importantly, three days after PARP inhibitor withdrawal, the positive SA-β-gal staining largely disappeared (Fig. 1G and S2B). Moreover, the accumulation of G2/M phase induced by PARP inhibitors was significantly reduced (Fig. S2C). Thus, when treated with PARP inhibitors, ovarian cancer cells can develop a senescence-like phenotype, which is consistent with previous report of reversible cellular senescence induced by PARP inhibitors in ovarian cancer [26].

Fig. 1.

PARP-inhibited ovarian cancer cells exhibit senescence-like phenotype. (A) SA-β-gal staining of HEY, OVCAR3, A2780 and OV6314622 cells after 10 μM Rucaparib or Olaparib treatment for 48 h. Scale bar: 100 μm. (B) Relative mRNA expression of CCL2, CCL5, CXCL1, CXCL10, IL-1α, IL-1β, IL-6, IL-8, IL-10, TNFα and IFNγ in HEY and OVCAR3 cells after 10 μM Rucaparib or Olaparib treatment for 48 h. (C) Western blot analysis of Lamin B1 protein levels in HEY and OVCAR3 cells treated with 10 μM Rucaparib or Olaparib for 48 h. (D) Immunofluorescence staining of Lamin B1 and quantification in HEY, OVCAR3 and A2780 cells treated with 10 μM Rucaparib or Olaparib for 48 h. Scale bar: 50 μm. Relative mRNA expression of p21 (E) and p16 (F) in HEY, OVCAR3 and A2780 cells after 10 μM Rucaparib or Olaparib treatment for 48 h. (G) SA-β-gal staining in HEY cells treated with 10 μM Rucaparib or Olaparib for 3 days (3 d), 6 days (6 d) or treated for 3 days then without for 3 days (3 d + Release 3 d). Scale bar: 100 μm. Three independent experiments were performed for each experiment and data presented as mean ± S.D. The statistical differences between the two groups were analyzed by two-sided unpaired Student's t-test. (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns stands for no significant difference).

3.2. Increased reliance on glycolysis in senescence-like ovarian cancer cells

Senescent cells tend to utilize glucose for survival due to dysfunctional mitochondria. We therefore explored whether glycolysis is altered in senescence-like cells. We first measured the uptake of 2-NBDG, a fluorescently labeled 2-deoxyglucose analog, in PARPi ovarian cancer cells. The results showed that the glucose uptake in PARPi ovarian cancer cells was significantly increased (Fig. 2A). Furthermore, lactate colorimetric assay indicated that the lactate secretion by ovarian cancer cells increased significantly after treatment with PARP inhibitors (Fig. 2B), suggesting that the glycolytic ability of PARPi ovarian cancer cells was enhanced. In addition, the results of glycolytic stress test using Agilent Seahorse energy metabolism analyzer also showed that the glycolytic ability, as reflected by the extracellular acidification rate (ECAR), was significantly enhanced in PARPi cancer cells (Fig. 2C). GSEA of RNA-seq data also revealed markedly enriched glycolytic process in PARPi ovarian cancer cells (Fig. 2D). We further measured glycolytic metabolites and found that D-glucose, Dihydroxyacetone phosphate, Lactate, Pyruvate, Oxaloacetate and Thiamine pyrophosphate were remarkably elevated in PARPi cells (Fig. 2E and S3A).

Fig. 2.

Glycolysis is upregulated and relied upon in PARPi ovarian cancer cells. (A) Flow cytometric analysis of 2-NBDG uptake in HEY, OVCAR3, A2780, OV6314622, SKOV3 and HO8910 cells treated with 10 μM Rucaparib or Olaparib for 48 h. (B) Lactate levels in HEY, OVCAR3, A2780 and OV6314622 cells treated with 10 μM Rucaparib or Olaparib for 48 h. (C) Extracellular acidification rate (ECAR) of HEY and OVCAR3 cells after 10 μM Rucaparib treatment for 48 h. Glc: glucose, Oligo: oligomycin A. (D) GSEA of glycolytic process (NES = 2.202719, NOM.pval = 0.00, FDR.qval = 0.007) in OVCAR3 cells treated with DMSO or 10 μM Rucaparib for 48 h. (E) Heatmap of significantly changed metabolites associated with glycolysis/gluconeogenesis. The energy metabolism-associated metabolites were measured in OVCAR3 cells treated with DMSO or 10 μM Rucaparib for 48 h. (F) SA-β-gal staining and (G) quantification in HEY and OVCAR3 cells after 10 μM PARP inhibitors (Rucaparib or Olaparib) alone or in combination with 1 mM 2-DG treatment for 48 h. Scale bar: 100 μm. (H) The proliferation of HEY and OVCAR3 cells after 10 μM Rucaparib alone or in combination with 1 mM 2-DG treatment for 48 h. The viable cell numbers were measured using flow cytometry by propidium iodide staining. Three independent experiments were performed for each experiment and data presented as mean ± S.D. The statistical differences between the two groups were analyzed by two-sided unpaired Student's t-test. ANOVA was used to compare significant differences among multiple experimental groups (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns stands for no significant difference).

Hexokinase (HK) is a key rate-limiting enzyme for glycolysis. Although there was no significant change at protein level, the HK activity was enhanced in PARPi ovarian cancer cells (Fig. S3B and C), further indicating increased glycolytic capacity. We next tested whether PARPi ovarian cancer cells are more sensitive to glucose deprivation. The results showed that the combination of PARP inhibitors and 2-DG could indeed further reduce the population of SA-β-gal positive cells (Fig. 2F and G) and the proliferation of ovarian cancer cells (Fig. 2H and S3D). Moreover, these results were also confirmed by glucose-restricted medium (Fig. S3E and F). These results suggest that PARPi ovarian cancer cells become more dependent on glycolysis.

3.3. The increasingly accumulated mitochondria in PARP-inhibited cancer cells are dysfunctional

Senescent cells typically harbor increased mitochondrial mass that determines some of the senescent features. Therefore, we measured the changes of mitochondrial content in PARPi ovarian cancer cells. Flow cytometry analysis of cells stained with MitoTracker showed that the mitochondrial content was significantly increase in PARPi cells (Fig. 3A). Morphological examination using transmission electron microscopy revealed significant loss of mitochondrial cristae and increased mitochondrial vacuolization in PARPi cells (Fig. 3B). Moreover, the levels of mitochondrial component Tomm20 also elevated in PARPi cells (Fig. 3C). GSEA of RNA-Seq data revealed significant enrichment of mitochondrial inner membrane components in PARPi ovarian cancer cells (Fig. 3D). Reduction of mitochondrial membrane potential (MMP) is often associated with mitochondrial dysfunction. Indeed, we found that the MMP was significantly decreased in PARPi cells (Fig. 3E and S4A). As a control, cancer cells treated with Carbonyl cyanide 3-chlorophenylhydrazone (CCCP) also exhibited decreased MMP and reduced mitochondrial mass (Fig. S4B and C). We further evaluated the oxygen consumption rate (OCR) in PARPi ovarian cancer cells by Agilent Seahorse energy metabolism analyzer and found that both the basic and maximum oxygen consumption capacities were generally increased (Fig. 3F). Additionally, the combination of PARP inhibitor with oligomycin A, an inhibitor of mitochondrial oxidative phosphorylation, had no or very modest additive inhibitory effect on cancer cells (Fig. S4D and E), which is in contrast to the more pronounced inhibition caused by PARP inhibitor and 2-DG in combination (Fig. 2H and S3D). These results indicated that PARP inhibitors led to the accumulation of dysfunctional mitochondria and a metabolic shift to glycolysis in ovarian cancer cells.

Fig. 3.

The increasingly accumulated mitochondria in PARPi cancer cells are dysfunctional. (A) Flow cytometric analysis of mitochondrial content in HEY, OVCAR3, A2780, OV6314622, SKOV3 and HO8910 cells treated with 10 μM Rucaparib or Olaparib for 48 h. (B) Transmission electron micrographs of HEY, OVCAR3 and A2780 cells treated with 10 μM Rucaparib or Olaparib for 48 h. Scale bar: 5 μm, 1 μm, 500 nm. (C) Immunofluorescence staining of Tomm20 and quantification in HEY and OVCAR3 cells treated with 10 μM Rucaparib or Olaparib for 48 h. Scale bar: 50 μm. (D) GSEA of mitochondrial inner membrane (NES = 1.8202196, NOM.pval = 0.00, FDR.qval = 0.05) in OVCAR3 cells treated with DMSO or 10 μM Rucaparib for 48 h. (E) Measurements of MMP (by JC-1) in HEY, OVCAR3, A2780 and OV6314622 cells treated with 10 μM Rucaparib or Olaparib for 48 h. (F) Oxygen consumption rate (OCR) of HEY, OVCAR3 and A2780 cells after 10 μM Rucaparib treatment for 48 h. Oligo: oligomycin A, Rot: rotenone, AA: antimycin A. Three independent experiments were performed for each experiment and data presented as mean ± S.D. The statistical differences between the two groups were analyzed by two-sided unpaired Student's t-test (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns stands for no significant difference).

3.4. HIF1α promotes survival in PARP-inhibited ovarian cancer cells

Glycolysis is commonly driven by HIF1α. GSEA also points to HIF1α signaling pathway in PARPi cells (Fig. S5A). We next determined whether the enhanced glycolysis in PARPi ovarian cancer cells is mediated by HIF1α. HIF1α levels were indeed increased in PARPi ovarian cancer cells when examined using fluorescence microscopy (Fig. 4A) and flow cytometry (Fig. 4B). When KC7F2, an HIF1α inhibitor, was used to block the function of HIF1α, the ECAR induced by PARP inhibitor was significantly attenuated (Fig. 4C). Moreover, the HIF1α inhibitor reduced the population of SA-β-gal positive cells (Fig. 4D and E). Meanwhile, the HIF1α inhibitor further reduced the proliferation of ovarian cancer cells when used in combination with PARP inhibitor (Fig. 4F and S5B). The above results indicate that the combined use of PARP and HIF1α inhibitors could hinder cells from entering a senescence-like state or compromise the survival of the senescence-like cells. However, the induction of apoptosis by the drug combination varied among the cell lines and between the two PARP inhibitors used (Fig. S5C).

Fig. 4.

HIF1α inhibition compromises the survival of PARPi ovarian cancer cells. (A) Immunofluorescence staining of HIF1α and quantification in HEY and OVCAR3 cells treated with 10 μM Rucaparib or Olaparib for 48 h. Scale bar: 50 μm. (B) Flow cytometric analysis of HIF1α expression quantity in HEY, OVCAR3, A2780, OV6314622, SKOV3 and HO8910 cells treated with 10 μM Rucaparib or Olaparib for 48 h. (C) ECAR of HEY and OVCAR3 cells after 10 μM Rucaparib alone or in combination with 20 μM KC7F2 treatment for 48 h. Glc: glucose, Oligo: oligomycin A. (D) SA-β-gal staining and (E) quantification in HEY and OVCAR3 cells after 10 μM PARP inhibitors (Rucaparib or Olaparib) alone or in combination with 20 μM KC7F2 treatment for 48 h. Scale bar: 100 μm. (F) The proliferation of HEY and OVCAR3 cells after 10 μM Rucaparib alone or in combination with 20 μM KC7F2 treatment for 48 h. The viable cell numbers were measured using flow cytometry by propidium iodide staining. Three independent experiments were performed for each experiment and data presented as mean ± S.D. The statistical differences between the two groups were analyzed by two-sided unpaired Student's t-test. ANOVA was used to compare significant differences among multiple experimental groups (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns stands for no significant difference).

To further confirm that HIF1α contributes to the survival of PARPi ovarian cancer cells, we knocked down HIF1α by transfecting ovarian cancer cells with HIF1α specific siRNA. RNA interference (RNAi) efficiency was measured by qPCR analysis (Fig. 5A). HIF1α stabilizer DMOG also validated HIF1α knockdown efficiency (Fig. 5B). Importantly, PARP inhibitor-induced HIF1α upregulation was effectively blocked by the knockdown of HIF1α (Fig. 5C–E). The knockdown of HIF1α significantly attenuated ECAR in PARPi ovarian cancer cells (Fig. 5F). As expected, HIF1α knockdown diminished the population of SA-β-gal positive cells induced by PARP inhibitors (Fig. 5G and H). HIF1α knockdown further decreased cell proliferation in PARPi cells (Fig. 5I and S5D). These results indicate that HIF1α function is critical for the survival of PARPi senescence-like cells.

Fig. 5.

HIF1α knockdown reduces senescence-like ovarian cancer cells. (A) RNAi efficiency was measured by qPCR analysis in HEY cells after transfected scramble siRNA (siScr) or siRNA targeting HIF1α (siHIF1α-1 or siHIF1α-2) for 48 h. (B) Flow cytometric analysis of HIF1α protein in HEY cells transfected siScr or siHIF1α-1/siHIF1α-2 treated with or without 1 mM DMOG for 24 h. (C) Immunofluorescence staining of HIF1α and (D) quantification in HEY cells transfected siScr or siHIF1α-1/siHIF1α-2 treated with or without 10 μM Rucaparib for 48 h. Scale bar: 50 μm. (E) Flow cytometric analysis of HIF1α protein in HEY cells transfected siScr or siHIF1α-1/siHIF1α-2 treated with or without 10 μM Rucaparib for 48 h. (F) ECAR of HEY cells transfected siScr or siHIF1α-1/siHIF1α-2 treated with or without 10 μM Rucaparib for 48 h. Glc: glucose, Oligo: oligomycin A. (G) SA-β-gal staining and (H) quantification in HEY cells transfected siScr or siHIF1α-1/siHIF1α-2 treated with or without 10 μM Rucaparib for 48 h. Scale bar: 100 μm. (I) The proliferation of HEY cells transfected siScr or siHIF1α-1/siHIF1α-2 treated with or without 10 μM Rucaparib for 48 h. The viable cell numbers were measured using flow cytometry by propidium iodide staining. Three independent experiments were performed for each experiment and data presented as mean ± S.D. The statistical differences between the two groups were analyzed by two-sided unpaired Student's t-test. ANOVA was used to compare significant differences among multiple experimental groups (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns stands for no significant difference).

3.5. Mitochondrial ROS stabilizes HIF1α in PARP-inhibited ovarian cancer cells

Mitochondrial reactive oxygen species (mtROS) could stabilize HIF1α [[32], [33], [34]]. We speculated that mtROS might drive HIF1α stabilization to upregulate glycolysis in PARPi ovarian cancer cells. We therefore measured the intensity of MitoSOX, a fluorescent probe specific for mitochondrial superoxide, in PARPi cancer cells. We detected a significant increase in mtROS in ovarian cancer cells treated with either Rucaparib or Olaparib (Fig. 6A). Consistent with our previous report of increased oxidative stress in PARPi cancer cells [16], GSEA also showed significant enrichments in positive regulation of reactive oxygen species metabolic process and glutathione transferase activity (Fig. S6A and B). We also included paraquat as a positive control for the measurement of MitoSOX (Fig. S6C). The level of 8-Hydroxydeoxyguanosine (8-OHdG), a marker of oxidative DNA damage, was significantly increased in PARPi cells (Fig. S6F). Mitoquinone (MitoQ, a mitochondria-targeted antioxidant) treatment attenuated the elevation of oxidative base damage and reactive oxygen species induced by PARP inhibitors (Fig. S6D–F). Importantly, scavenging of mtROS by MitoQ effectively blocked PARP inhibition-induced upregulation of HIF1α (Fig. 6B and C). In addition, the results with glycolytic stress test showed that enhancement of glycolysis induced by PARP inhibitors was greatly reduced by MitoQ (Fig. 6D). Moreover, MitoQ led to significant reduction of the senescence-like cells (Fig. 6E and F). While the proliferation of the cancer cells was inhibited by MitoQ treatment alone, PARPi cells were further impaired (Fig. 6G and S6G). However, MitoQ had only a modest effect on inhibition of cell proliferation and apoptosis induced by doxorubicin in MDA-MB-231 cells (Fig. S6H and I). Together, the results indicate that the mtROS-HIF1α axis is critical for the survival of PARPi ovarian cancer cells.

Fig. 6.

Survival of PARPi ovarian cancer cells requires mtROS-induced HIF1α stabilization. (A) Flow cytometric analysis of mitochondrial ROS production by MitoSOX red in HEY, OVCAR3, A2780, OV6314622, SKOV3 and HO8910 cells treated with 10 μM Rucaparib or Olaparib for 48 h. (B) Immunofluorescence staining of HIF1α and quantification in HEY cells after 10 μM PARP inhibitors (Rucaparib or Olaparib) alone or in combination with 500 nM MitoQ treatment for 48 h. Scale bar: 50 μm. (C) Flow cytometric analysis of HIF1α expression quantity in HEY, OVCAR3, A2780, OV6314622 and SKOV3 cells after 10 μM PARP inhibitors (Rucaparib or Olaparib) alone or in combination with 500 nM MitoQ treatment for 48 h. (D) ECAR of HEY and OVCAR3 cells after 10 μM PARP inhibitors (Rucaparib or Olaparib) alone or in combination with 500 nM MitoQ treatment for 48 h. Glc: glucose, Oligo: oligomycin A. (E) SA-β-gal staining and (F) quantification in HEY and OVCAR3 cells after 10 μM PARP inhibitors (Rucaparib or Olaparib) alone or in combination with 500 nM MitoQ treatment for 48 h. Scale bar: 100 μm. (G) The proliferation of HEY and OVCAR3 cells after 10 μM Rucaparib alone or in combination with 500 nM MitoQ treatment for 48 h. The viable cell numbers were measured using flow cytometry by propidium iodide staining. Three independent experiments were performed for each experiment and data presented as mean ± S.D. The statistical differences between the two groups were analyzed by two-sided unpaired Student's t-test. ANOVA was used to compare significant differences among multiple experimental groups (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns stands for no significant difference).

While senescence-like cells were greatly reduced when PARPi cancer cells were concomitantly subjected to glucose restriction, HIF1α inhibition or MitoQ treatment, it is unclear whether the combinations blocked the cancer cell entry into senescence-like state or compromised the survival of senescence-like cells. To address this, we first induced cancer cells to enter senescence-like state by PARP inhibitor, which is confirmed by SA-β-gal staining, and then applied 2-DG, HIF1α inhibitor or MitoQ to disrupt the mtROS-HIF1α-glycolysis axis. The results showed that the 2-DG, HIF1α inhibitor or MitoQ each reduced the population of SA-β-gal positive cells (Fig. S7A–C), indicating that the axis was critical for the survival of PARPi senescence-like cells. Alternatively, when cancer cells were pretreated with 2-DG, HIF1α inhibitor or MitoQ to undermine the axis, and then treated with PARP inhibitor, the senescence-like cells were also greatly reduced (Fig. S8A–C), suggesting that the axis was required for the emergence of senescence-like cells. The above results indicate that the transition to and the sustenance of the senescence-like state require mtROS-HIF1α-glycolysis axis.

3.6. PARP inhibitor induces senescence-like phenotype in vivo

Next, we tested whether PARP inhibition could induce cellular senescence in vivo. We subcutaneously inoculated OVCAR3 cells into NCG mice and then subjected them to Rucaparib (20 mg/kg) by intraperitoneal injection (Fig. 7A). We found that Rucaparib suppressed tumor growth in vivo (Fig. 7B) and the average tumor size was much smaller in the Rucaparib treatment group (Fig. 7C). Increased cellular senescence was also detected in tumor tissue section derived from PARP inhibitor-treated mice (Fig. 7D), which was confirmed by flow cytometry analysis of single cells prepared from tumor tissues (Fig. 7E and Fig. S9A). Consistent with the increased mitochondrial accumulation in PARPi ovarian cancer cells in vitro, the levels of mitochondrial component Tomm20 and mitochondrial content were elevated in the tumor tissues (Fig. 7F and G). The level of HIF1α also increased in tumor tissues from mice treated with PARP inhibitor (Fig. 7H and I). Consistently, mtROS were upregulated in tumor tissues (Fig. S9B).

Fig. 7.

PARP inhibitor induces senescence-like phenotype in vivo. (A) Schematic diagram for the treatment paradigm. OVCAR3 cells were subcutaneously transplanted in NCG mice. Mice were randomized into two treatment groups: vehicle (DMSO, n = 3), 20 mg/kg Rucaparib (PARPi, n = 4) (given daily by intraperitoneal injection). (B) Growth curves of tumors in mice. Tumor volumes were measured every day. (C) Image of tumors and tumor weights in mice. (D) SA-β-gal staining and quantification in tumor tissues. Scale bar: 50 μm. (E) Flow cytometric analysis of SPiDER-β-gal in single cell suspension of tumor tissues. (F) Immunofluorescence staining of Tomm20 and quantification in tumor tissues. Scale bar: 50 μm. (G) Flow cytometric analysis of mitochondrial content in single cell suspension of tumor tissues. (H) Representative IHC images of HIF1α expression and quantification in tumor tissues. Scale bar: 50 μm. (I) Flow cytometric analysis of HIF1α in single cell suspension of tumor tissues. Data presented as mean ± S.D. The statistical differences between the two groups were analyzed by two-sided unpaired Student's t-test (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001).

3.7. MitoQ attenuates PARP inhibitor-induced senescence-like phenotype in vivo

We next evaluated whether MitoQ could sensitize cancer cells to PARP inhibitors in vivo. Rucaparib and MitoQ alone or in combination were used to treat tumor xenografts in NCG mice (Fig. 8A). We found that the combination further inhibited tumor growth (Fig. 8B and C). At the same time, staining for cellular senescence in tumor tissues indicated that the prevalence of senescence-like cells caused by Rucaparib was reduced (Fig. 8D–F). Moreover, HIF1α expression also restored to control levels (Fig. 8G–I). In tumor xenograft tissues, increased production of lactate, an indicator of glycolysis, was also abolished by MitoQ treatment (Fig. 8J). Thus, the survival of the senescence-like cancer cells was likely compromised by MitoQ. However, although tumor growth was further inhibited by the combination of Rucaparib and MitoQ, the effect was still limited, suggesting that there are other unknown factors contributing to the survival of PARPi cancer cells.

Fig. 8.

MitoQ attenuates senescence-like phenotype in vivo. (A) Schematic treatment diagram. OVCAR3 cells were subcutaneously transplanted in NCG mice. Mice were randomized into four treatment groups: vehicle (DMSO, n = 5), 2 mg/kg MitoQ alone (n = 5) (given every other day by oral gavage), 20 mg/kg Rucaparib alone (n = 5) (given daily by intraperitoneal injection) or 20 mg/kg Rucaparib plus 2 mg/kg MitoQ (n = 5). (B) Growth curves of tumors in mice. Tumor volumes were measured every day. (C) Image of tumors and tumor weights in mice. (D) SA-β-gal staining and (E) quantification in tumor tissues. Scale bar: 50 μm. (F) Flow cytometric analysis of SPiDER-β-gal in single cell suspension of tumor tissues. (G) Representative IHC images showing the HIF1α. Scale bar: 50 μm. (H) Quantification of HIF1α expression in tumor tissues. (I) Flow cytometric analysis of HIF1α in single cell suspension of tumor tissues. (J) Lactate levels in tumor tissues. (K) A schematic model. In ovarian cancer cells, PARP inhibitors induce senescence-like phenotype. Cells undergo metabolic reprogramming through the mtROS-HIF1α axis to sustain survival. Data presented as mean ± S.D. ANOVA was used to compare significant differences among multiple experimental groups (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001).

4. Discussion

In this study, we showed that PARP-inhibited ovarian cancer cells exhibit a senescence-like phenotype that is characterized by accumulation of dysfunctional mitochondrial and a metabolic shift to glycolysis. Mechanistically, mtROS stabilizes HIF1α and thus drives glycolysis. Quenching of mtROS and inhibition of HIF1α could each block the cancer cell entry into senescence-like state or compromise the survival and proliferation of the senescence-like cancer cells. Our findings suggest that targeting the mtROS-HIF1α-glycolysis axis in ovarian cancer may circumvent the pro-survival senescence-like phenotype caused by PARP inhibitors in some circumstances (Fig. 8K).

Previous reports have demonstrated that reverse BRCA1/2 mutations and thus HR functional restoration may mediate the acquired resistance to PARP inhibitors in ovarian or breast cancer [35,36]. Herein, we showed that PARPi ovarian cancer cells can adopt a senescence-like state and shift to HIF1α mediated glycolysis for survival due to an increase in mitochondria superoxide. It was reported that cancer cells induced to enter a state of senescence upon exposure to cancer chemotherapeutic drugs can recover self-renewal capacity [37]. Consistent with our results, Zhang et al. showed that PARP inhibitors induced the formation of polyploid giant cancer cells which exhibit multiple characteristics of senescent cells [28]. Therefore, senescence-like state of tumor cells may represent one form of tumor dormancy and contribute to disease recurrence [38,39]. In two studies of leukemia models, senescence-like cells exhibited a higher tumor initiation potential and entry into this senescence-like phenotype was dependent on ATR [40,41]. Therefore, clearance of senescence-like cells may enhance drug sensitivity and reduce tumor recurrence. A study indeed showed that PARP inhibitors in combination with senolytics could exert a synthetic-lethal effect [26].

Among other metabolic features, senescent cells typically exhibit increased accumulation of dysfunctional mitochondria, a manifestation of mitochondrial homeostatic disruption. Glycolysis tends to be enhanced to compensate for the decline in mitochondrial bioenergetic function. We observed enhanced glucose uptake and lactate secretion in PARPi ovarian cancer cells, and concomitant inhibition of glycolysis conferred cancer cells increased sensitivity to PARP inhibitors. While we here demonstrated PARP inhibition leads to the stabilization of HIF1α and thus upregulates glycolysis pathway, it was recently shown that PARP can inhibit hexokinase activity through PAR and impair glycolysis [42]. We indeed detected increased hexokinase activity in PARPi cancer cells. It is possible that multiple mechanisms may contribute to the enhancement of hexokinase activity in PARPi ovarian cancer cells.

It should be noted that while the combinations of PARP inhibitor with 2-DG, HIF1α inhibition/depletion or MitoQ each remarkably reduced the senescence-like cancer cells in vitro, the combination of Rucaparib and MitoQ only achieved a modest additive tumor-inhibitory effect on OVCAR3 tumor sizes in vivo. There are several possible explanations for this less pronounced effect. First, MitoQ application alone was cytostatic (Fig. 6E and G), which makes further proliferation impediment by PARP inhibitor less striking. Second, in addition to the cell autonomous effect associated with cellular senescence, the senescence-like cancer cells may cause remodeling of tumor microenvironment via their secretome, which is enriched in chemokines, cytokines and growth factors that can alter immune cell landscape in tumor microenvironment and promote angiogenesis. Senescence-associated secretory phenotype (SASP) was reported to drive antitumor immunity [[43], [44], [45]]. However, it should be noted that the NCG mice used in this study are severely immune deficient due to the absence of T cells, B cells and NK cells, which drastically limits the evaluation of immune response associated with the emergence, or lack thereof, senescence-like cancer cells. Whether SASP shapes the myeloid compartments in tumor microenvironment to influence tumor growth also needs to be elucidated. Third, PARP inhibitor may directly act on myeloid cells and stromal fibroblasts to exert its anti-tumor or pro-tumor effects. Wang et al. demonstrate that PARP-inhibition reprograms macrophages toward an anti-tumor phenotype through increased mtROS production and phagocytic capacity [46]. MitoQ can potentially neutralize the mtROS and thus compromise macrophage-mediated anti-tumor function. Furthermore, PARP inhibitors were also reported to trigger an autocrine phenotype in ovarian cancer stromal fibroblasts that compromises the tumor-inhibitory effect [47]. However, Olaparib was shown to reduce the recruitment of myeloid-derived suppressor cells by downregulating the expression of SDFα by cancer-associated fibroblasts in mice with breast cancer [48]. PARP inhibition-induced redox dysregulations may thus have different consequences depending on cell types and other contexts, which remains to be further elucidated.

In summary, we showed that PARP inhibitors can induce a senescence-like phenotype in ovarian cancer cells that enables cells to survive in a glycolysis-dependent manner. Mechanistically, ROS derived from the excessively accumulated mitochondria in the senescence-like cells stabilize HIF1α that sustains glycolysis. The senescence-like phenotype caused by PARP inhibition has implications in understanding both cancer chemoresistance and therapy-induced remodeling of tumor microenvironment.

CRediT authorship contribution statement

Tingting Yang: Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing – original draft, Writing – review & editing. Wenqing Bu: Methodology, Validation. Xiaotong Xue: Methodology. Xuemin Wu: Methodology. Yipeng Zhou: Methodology. Yue Chen: Methodology. Yanan Li: Methodology. Jia Zhang: Methodology. Peishan Li: Resources. Fangrong Shen: Resources. Yufang Shi: Resources. Changshun Shao: Conceptualization, Formal analysis, Funding acquisition, Project administration, Resources, Supervision, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

figs1.

figs2.

figs3.

figs4.

figs5.

figs6.

figs7.

figs8.

figs9.

Data availability

Data will be made available on request.