Targeted blockade of HSP90 impairs DNA-damage response proteins and increases the sensitivity of ovarian carcinoma cells to PARP inhibition

ABSTRACT

Pharmacological inhibition of PARP is a promising approach in treating high grade serous ovarian carcinoma (HGSOC). PARP inhibitors (PARPi) are most active in patients with defects in DNA damage repair (DDR) mechanisms, such as alterations in expression/function of DNA repair and homologous recombination (HR) genes/proteins, including BRCA1 and BRCA2. Benefit of PARPi could be extended towards HR-proficient patients by combining PARPi with agents that functionally abrogate HR. An attractive molecular target for this purpose is heat shock protein 90 (HSP90), which mediates the maturation and stability of several key proteins required for DDR. Here, we tested the hypothesis that targeted inhibition of HSP90 with a small-molecule inhibitor ganetespib would sensitize non-BRCA mutant ovarian carcinoma (OC) cells to PARP inhibition by talazoparib. We used commercially available cell lines, along with several novel HGSOC OC cell lines established in our laboratory. Ganetespib treatment destabilized HSP90 client proteins involved in DNA damage response and cell cycle checkpoint, and disrupted γ-irradiation-induced DNA repair. The effects of the combination of ganetespib and talazoparib on OC cell viability and survival were also analyzed, and among the non-BRCA mutant cell lines analyzed, the combination was synergistic in several cell lines (OVCAR-3, OC-1, OC-16). Together, our data suggest that ganetespib-mediated inhibition of HSP90 effectively disrupts critical DDR pathway proteins and may sensitize OC cells without ‘BRCAness’ to PARPi. From a clinical perspective, this suggests that HSP90 inhibition has the potential to sensitize some HGSOC patients without HR pathway alterations to PARPi, and potentially other DNA-damage inducing agents.

Introduction

Ovarian carcinoma remains the most deadly malignancy of the female reproductive tract, in large part due to late detection and its propensity for recurrence.¹ The management of the disease most commonly involves cytoreductive surgery and adjuvant chemotherapy.¹ There is no cure for a recurrent disease, and recent advances in novel therapeutics have focused on the development of targeted molecular therapies, often in conjunction with conventional cytotoxic chemotherapy.

Poly (ADP-ribose) Polymerase (PARP) plays key roles in DNA single-strand repair.² PARP inhibitors (PARPi) are a promising class of inhibitors for the treatment of high grade serous ovarian carcinoma (HGSOC), the most common histologic subtype of ovarian carcinoma.¹ The greatest activity of PARPi is seen in patients with defects in DNA damage repair (DDR) mechanisms, including mutation or epigenetic inactivation of BRCA1 and BRCA2 genes and alterations in expression and/or function of other DNA repair genes/proteins.³ PARPi are approved as second-line and maintenance therapies in recurrent HGSOCs.⁴ Notably, clinical trials have demonstrated that single agent PARPi show activity in a significant number of HGSOC patients in the absence of alterations in BRCA genes, in patients with platinum-sensitive disease, and those with tumors exhibiting defects in homologous recombination (HR), or ‘BRCAness’.^5,6

Combining PARPi with agents that functionally abrogate HR, thus mimicking ‘BRCAness’, could potentially augment the benefit of pharmacologic PARP inhibition in patients without inherent HR deficiency. An attractive molecular target for this purpose is heat shock protein 90 (HSP90). HSP90 is an ATP-dependent molecular chaperone mediating the maturation, stability, and activation of several hundred diverse proteins, including cell cycle regulators CDK1 and CHK1, and key proteins required for DNA repair, such as BRCA1, BRCA2, RAD51, and MRE11.^7-9 Moreover, prior work directly demonstrated that targeted inhibition of HSP90 impairs HR and non-homologous end joining (NHEJ) repair pathways in response to double-strand breaks (DSBs) or interstrand cross-links induced by platinum-based agents.⁹

We and others have shown that the second-generation small-molecule HSP90 inhibitor, ganetespib (STA-9090), has pre-clinical chemo- and/or radio-sensitization activity in different types of solid tumors, including breast, lung, colon, prostate, and ovarian cancers.^10-14 The goal of the current study was to test the hypothesis that targeted inhibition of HSP90 with ganetespib would sensitize HR proficient OC cells to the clinically relevant PARPi talazoparib (BMN 673).¹⁵ In this study, we used previously established OC cell lines (OVCAR-3, UWB 1.289), and novel OC cell lines (OC-1, OC-16, OC-38) established in our laboratory from de-identified tumors isolated from patients with HGSOC. Together our results show that inhibition of HSP90 by ganetespib effectively disrupts expression of DNA repair and cell cycle checkpoint proteins, ionizing radiation-induced DNA-repair, and sensitizes HGSOC cells to PARPi talazoparib. Taken together, our data suggests that pharmacological inhibition of HSP90 remains a promising approach in sensitizing HR-proficient ovarian cancers to inhibitors of PARP.

Materials and methods

Cell lines, culture conditions and antibodies

Identity verified OVCAR-3 and UWB 1.289 cells were obtained from the Fox Chase Cancer Center (FCCC) Cell Culture Facility and cultured as described by the American Tissue Culture Collection. Several novel cell lines, including OC-1, OC-16, and OC-38, were derived in our laboratory from de-identified tumors isolated from patients with HGSOC. Fresh de-identified tumor tissue was obtained from the FCCC Biosample Repository Facility (BRF). The FCCC BRF has an Institutional Review Board (IRB)-approved protocol for collection and banking of blood, tissue and associated clinical data from patients undergoing surgery at FCCC under informed consent. The biospecimens and associated clinical data obtained from the FCCC BRF are de-identified and distributed to investigators with a unique participant and specimen identification barcode numbers. Fresh ovarian tumor tissue specimens were cut into fragments 2–3 mm and enzymatically and physically dissociated using a gentleMACS Dissociator with a human Tumor Dissociation Kit (Miltenyi Biotec, Germany) according to manufacturer’s instructions. The resulting cell suspension was filtered and seeded onto irradiated J2 fibroblast feeder cells in Rho kinase-inhibitor containing a medium, as described.¹⁶ The patient-derived cells were routinely cultured on the irradiated J2 feeder cells, and differential trypsinization was used to separate OC cells from J2 feeder fibroblasts.¹⁶ All cell lines were maintained in a 5% CO₂ atmosphere at 37°C and were periodically checked for mycoplasma contamination. For short-term analyses of drug treatments (up to 120 h), patient-derived OC cells were plated in the absence of feeder cells in the presence of conditioned culture medium.¹⁷ Antibodies used and commercial source are as follows: BRCA2 (Bethyl, A303-434A), CDK1 (Santa Cruz, sc-54); HSP90 (Enzo Life Sciences, ADI-SPA-835-D); cleaved PARP (Millipore, AB3565); PAR (GeneTex, GTX75054); GAPDH (Advanced ImmunoChemical, 6C5); BRCA1 (Cell Signaling, 9010S); MRE11 (Cell Signaling, 4847S); c-MYC (Cell Signaling, 5605S); CHK1 (Cell signaling, 2360S); ATM (Cell Signaling, 2873); pATM^S1981 (Millipore, MAB3806); p-γH2AX^S139 (Millipore, 05-636); RAD51 (Santa Cruz sc-8349 for IF, and Cell Signaling 8875 for WB); FITC and AlexaFluor 594 (Jackson ImmunoResearch, 711-095-152 and 715-585-150, respectively).

Immunoblot analysis and immunofluorescence microscopy

Immunoblot analyses were performed as described.¹⁴ Briefly, cells were seeded on 6-well plates, and treated in the absence or presence of the indicated concentrations of ganetespib or talazoparib for the indicated times. Resulting protein bands were visualized using a KwikQuant imager (Kindle Biosciences, Greenwich, CT). For immunofluorescent staining, irradiated cells were grown on coverslips, fixed with 4% paraformaldehyde, permeabilized by 1% Triton X-100 PBS solution, and incubated with antibodies recognizing phospho-H2AX^S139 (γ-H2AX), phospho-ATM^S1981 and RAD51 diluted in 5% goat serum, followed by incubation in secondary antibodies conjugated to FITC or AlexaFluor 594. Immunofluorescence images were acquired using a Nikon Eclipse E800 Microscope and Nikon NIS Elements BR software (Nikon, Japan). For quantification of foci-positive cells, at least 200 cells were counted in a minimum of 3 fields of view for each condition. Cells having ≥5 foci were considered positive. Numbers of foci per nucleus were counted in at least 5 cells in 3 fields of view per condition. Experiments were performed at least 3 times.

Analysis of ionizing radiation-induced DNA damage

Cells were seeded on sterile coverslips and grown to ~70% confluence. Culture medium was replaced 24 h prior to irradiation with medium containing 25 nM ganetespib and 5 mM HEPES and cells were exposed to 7Gy γ-irradiation by 81-14R Cesium-137 irradiator (JL Shepherd and Associates, San Fernando, CA) or mock-irradiated. Cells were fixed at 2, 8, and 24 h post-irradiation and incubated in the presence of antibodies as indicated.

Drug reagents and cell viability assays

Ganetespib was provided by Synta Pharmaceuticals Corp. (Lexington, MA). Talazoparib was provided by Synta or purchased from Selleckchem (Houston, TX). Drugs were dissolved in DMSO and stored at −20°C as individual aliquots to avoid freezing/thawing. For viability assays, 1000–3000 cells were seeded in the wells of 96-well plates to achieve 80–90% confluence of untreated cells at the end of the experiment. Drug treatment was initiated 24 h after plating and cells were grown for an additional 72–120 h. The effective dose 50 (ED₅₀) values of single drugs on each cell line were determined before initiation of combination treatments. The concentration of DMSO in the culture medium did not exceed 0.02%. Cell viability was assessed by Cell-Titer Glo assay (Promega, Madison, WI) according to the manufacturer’s protocol. All experiments were performed at least 3 times with 3 technical replicates/treatment.

HSP90 gene knockdown

The HSP90AA1 and control non-targeting siRNA pools were purchased from Dharmacon (Lafayette, CO; M-005186-02 and D-001206-13-05). OVCAR-3 cells were reverse-transfected by plating cells and siRNA pools simultaneously in 96-well plates using Lipofectamine RNAiMAX® (Thermo Fisher Scientific, Waltham, MA). Talazoparib treatment was initiated the following day and cell viability was assessed 120 h later. Protein knockdown was confirmed 48 h post-transfection.

Analysis of drug synergy and statistical analysis

Single drugs were first tested at 10 dose levels to estimate the ED₅₀ in each cell line – d_50s. The suffix ‘s’ is used to denote a single drug. Each dose was repeated at least 9 times within 3 experiments. Least-squares estimates of the d_50s were obtained using a four parameter function of the form: S = A + B/[1 +(d/M)^p], where S is the surviving cell fraction at dose d, and A, B, M, and p are parameters estimated from the data. Bootstrap samples (samples of the same size as the original, taken from the original set, but with replacement) were obtained from each dose level. Using the new (bootstrap) samples for each dose level, the d_50s was re-estimated. This process was repeated 100 times yielding a list of 100 bootstrap estimates of the drug’s d_50s against the particular cell line.

When two or more drugs were simultaneously used to treat a cell line, again several dose levels with multiple observations at each level were used to estimate the ED₅₀ of the combinations. If two drugs were used, then the d₅₀ is described by doses of each drug, i.e. d₅₀(1) and d₅₀(2). In all cases, either one drug was held constant (at ED₄₀) and the other varied to obtain the dose response, or the two drugs were varied together in a constant ratio (e.g. 1:10). Thus, the d₅₀ determination was always one-dimensional. The estimated combination index, CI, is d₅₀(1)/d_50s(1) + d₅₀(2)/d_50s(2),¹⁸ where d_50s(1) is the dose of the first drug used singly to be lethal to 50% of these cells; same with d_50s(2).

If the above CI was less than 1, the possibility of synergy was subjected to a statistical test. To determine if this expressed synergy is statistically significant, the set of 100 or 1000 d_50s(1) and 100 or 1000 d_50s(2) described above were used in combination with bootstrap estimates of d₅₀(1) and d₅₀(2). Each pair of d_50s(1) and d_50s(2) was combined with a single bootstrap estimate of d₅₀(1) and d₅₀(2) to form 100 or 1000 bootstrap estimates of the CI. Using these estimates of CI we tested the null hypothesis that the true CI is at least 1.0. If at most 5/100 or 50/1000 of the estimated CI equaled or exceeded 1.0, the null hypothesis was rejected.

Data obtained from γ-H2AX, pATM^S1981 and RAD51 foci counts, and annexin V/7-AAD staining were analyzed using GraphPad Prism by One-way Analysis of Variance (ANOVA) and the non-parametric Kruskal–Wallis test. For all analyses, P values <0.05 were considered significant.

Apoptosis assay

Apoptosis was analyzed using a BD Pharmingen PE annexin V apoptosis detection kit I (BD Biosciences, San Jose, CA) according to the manufacturer’s protocol. OVCAR-3 cells were treated with 6 nM talazoparib, 50 nM ganetespib or both; OC-1 cells were treated with 25 nM talazoparib, 25 nM ganetespib, or both. The cells were incubated in drug containing a medium for 48 h. Experiments were performed 3 times with technical replicates. Quantification was performed on LSRII flow cytometer (BD Biosciences) and analyzed on FlowJo software.

Results

Targeted inhibition of HSP90 destabilizes expression of DNA-damage response and cell cycle checkpoint proteins in OC cells

We first sought to determine whether inhibition of HSP90 affects the stability of DNA repair and cell cycle regulator protein expression in OC cells. To assess this, we used an established HGSOC cell line (OVCAR-3) in conjunction with novel HGSOC cells lines developed in our laboratory, OC-1 and OC-38 cell lines. Ganetespib-treated cells were lysed and expression levels of established HSP90 client proteins, including BRCA1, BRCA2, MRE11, C-MYC, CHK1, CDK1, ATM, and RAD51, were evaluated by immunoblot analysis. Cells were either treated with increasing concentrations of ganetespib (0–500 nM) to assess dose-response (Figure 1A) or with a fixed concentration of ganetespib for increasing times (0–48 h) to assess duration of response (Figure 1B). All three cell lines exhibited prominent dose-dependent decreases in BRCA1 and BRCA2, cell-cycle checkpoint proteins CDK1 and CHK1 and ATM serine/threonine kinase, and homologous recombination protein RAD51 (Figure 1A). In addition, a dose-dependent decrease in MRE11, a member of the MRN complex, was observed in OVCAR-3 and OC-1 cells, and a dose-dependent reduction of c-MYC was detected in OC-1 cells (Figure 1A). The duration of ganetespib-mediated HSP90 client protein depletion was somewhat variable, with maximal inhibition of protein expression observed 6–24 h post-treatment and gradual recovery of some proteins observed after 24 h (Figure 1B). Ganetespib treatment led to a dose- and time-dependent increase of HSP90 protein expression in the cell lines analyzed, suggesting inhibition and recovery of the drug target. These data demonstrate that targeted inhibition of HSP90 with ganetespib results in destabilization of key proteins involved in DNA repair and cell cycle checkpoint control.

Figure 1.

Ganetespib destabilizes DNA repair and cell cycle checkpoint proteins in ovarian carcinoma cells in a dose- and time-dependent manner. OVCAR-3, OC-1 and OC-38 cells were treated with 0 (vehicle), 12.5, 25, 50, 100, or 500 nM/L of ganetespib for 24 h (A), or with 50 nM/L ganetespib for 0, 6, 12, 24, 36, and 48 h (B). Cells were then lysed and subjected to immunoblot analysis with antibodies against BRCA2, BRCA1, HSP90, MRE11, c-MYC, CDK1, CHK1, ATM, and RAD51. Detection of GAPDH levels was used as a loading control.

Targeted inhibition of HSP90 reduces ionizing radiation (IR)–induced DNA damage response in OC cells

As ganetespib resulted in depleted levels of DNA repair proteins, we next asked whether ganetespib-mediated inhibition of HSP90 functionally affects DNA-repair by evaluating the presence and extent of γ-H2AX, RAD51, and pATM^S1981 foci in cells subjected to DNA damaging ionizing radiation (IR). For this, vehicle- and ganetespib-treated OC cells (OVCAR-3, OC-1, and OC-38 cells) were exposed to 7 Gy of γ-irradiation followed by analysis of these biomarkers of DNA damage and repair. Accumulation of nuclear foci expressing histone H2AX phosphorylated on serine residue S129, γ-H2AX, indicates the presence of DNA DSB in cells.¹⁹ At 2 h post-irradiation, the numbers of γ-H2AX foci per nucleus were similarly increased in OVCAR-3, OC-1, and OC-38 cells regardless of the presence or absence of ganetespib pretreatment (Figure 2B). The numbers of γ-H2AX foci were diminished over time (8 and 24 h post-IR) in the vehicle-treated cells, but remained elevated in ganetespib-treated cells in all three cell lines (Figure 2A & 2B). Phosphorylation of H2AX^S129 following DNA double-strand break formation is carried out by ATM (ataxia telangiectasia mutated), a member of the phosphoinositol-3 kinase family.²⁰ Autophosphorylation at ATM^S1981 dissociates the ATM dimer exposing the catalytic site to the substrate.²¹ Consistent with the elevated levels of γ-H2AX foci, the number of pATM^S1981 foci were also increased in ganetespib-treated OVCAR-3 cells at 2, 8, and 24 h post-irradiation (Figure 2C). RAD51 facilitates the homology search and recombination steps of homologous recombination (HR) and the appearance of RAD51 positive foci is a biomarker of HR repair activity.¹⁰ Thus, induction of RAD51⁺ foci was assessed to determine if ganetespib-mediated inhibition of HSP90 affected HR repair of DNA damage. In OVCAR-3 and OC-38 cells, ganetespib treatment resulted in a lower percentage of cells with RAD51 foci at 2 h post IR exposure and remained low at 8 and 24 h post-IR (Figure 2A & 2B). In OC-1 cells, the percent of RAD51 foci in ganetespib-treated cells was elevated similar to vehicle-treated cells but was significantly lower at 8 h post-IR (Figure 2B). Together, these data suggest that ganetespib-mediated inhibition of HSP90 results in the persistence of unrepaired DSBs in OC cells at least in part due to the disruption of HR.

Figure 2.

Ganetespib treatment disrupted ionizing radiation-induced DNA-repair and homologous recombination. OVCAR-3, OC-1, and OC-38 cells were treated with vehicle or 25 nM/L ganetespib for 24 h before IR exposure and analyzed by immunofluorescent detection of γ-H2AX⁺ or RAD51⁺ foci. (A) Representative images of control and irradiated OVCAR-3, OC-1, and OC38 cells 8 h post-IR (scale bar = 20 µm). (B) Quantification of the number of γ-H2AX foci per nucleus and percentage of cells with RAD51⁺ foci. (C) Representative images 8 h post-IR (scale bar = 20 µm) and quantification of pATM^S1981 positive foci in OVCAR-3 cells treated with 25 nM/L ganetespib for 24 h before IR exposure. Data were analyzed by One-way Analysis of Variance (ANOVA) and the non-parametric Kruskal–Wallis test followed by Dunns post-test. For all analyses, P values <0.05 were considered significant (*P < 0.05, **P < 0.01 and ***P < 0.001).

Ganetespib sensitizes OC cells to the PARP inhibitor talazoparib

After demonstrating that inhibition of HSP90 led to the destabilization of proteins involved in DNA repair and cell cycle regulation, accumulation of DSBs and abrogation of HR, we next tested whether ganetespib treatment could sensitize OC cells to the cytotoxic effects of targeted PARP inhibition. To test this, the individual sensitivities of OC cells to ganetespib and talazoparib were determined. Talazoparib is a potent PARPi currently under evaluation in ovarian cancer patients in clinical trials.^15,22-24 The capacity of talazoparib to specifically inhibit PARP activity in OC cells was confirmed by assessing poly ADP ribosylation (PAR) in OC cells, showing a rapid dose-dependent reduction in PAR levels in treated cells (Fig. S1). The effects of ganetespib and talazoparib on the viability of HGSOC cell lines (OVCAR-3, OC-1, OC-16, OC-38, and BRCA1-null UWB 1.289 cells²⁵) were tested by exposure of cells to increasing concentrations of each drug (0–50 nM/L ganetespib and 0–2000 nM/L talazoparib) to establish effective concentration (EC) that reduces cell viability in each cell line (Table 1). Once the ECs of each drug were established for each cell line, the effects of combinations of varying amounts of ganetespib and talazoparib were evaluated in combination. Ganetespib and talazoparib were combined in two ways: (1) a single sub-lethal (EC₄₀) dose of ganetespib added to serially diluted talazoparib (non-fixed molar ratio combination); or (2) both drugs were given in a fixed molar ratio combination (Table 2). The effects of combined drug interactions were assessed as described in detail in the Methods section, and the combination indexes (CI) were calculated to assess additive, synergistic or antagonistic effects. Additional statistical analysis was performed using the bootstrap method to determine the significance of the CI values. Addition of ganetespib further sensitized BRCA1 null UWB 1.289 cells to talazoparib at a non-fixed molar ratio (Table 2, CI = 0.73). While no synergy was observed when talazoparib and ganetespib were used in non-fixed molar ratios in other OC cells, testing at fixed molar ratios revealed a significant synergy between ganetespib and talazoparib was observed in OVCAR-3 cells, OC-1 cells, and OC-16 cells (Table 2 and Figure 3). Notably, the molar ratios found to be synergistic were somewhat cell line dependent: 1:1 (CI = 0.64) and 1:10 (CI = 0.7) molar ratios in OVCAR-3 cells; 10:1 molar ratio in OC-1 cells (CI = 0.62); 1:1 (CI = 0.66), 10:1 (CI = 0.55) and 40:1 (CI = 0.23) molar ratios in OC-16 cells (Table 2 and Figure 3). For OC-38 cells, each of the fixed ratio drug combinations was antagonistic. The calculated EC₅₀ of ganetespib was 14.2 nM (Table 1) and the EC₅₀ of talazoparib was estimated to be as high as 10⁶ nM by linear extrapolation of the experimental data. In the presence of 8nM of ganetespib (EC₄₀) the EC₅₀ of talazoparib dropped to ~3 × 10³ nM. Calculation of CI by d₅₀(1)/d_50s(1) + d₅₀(2)/d_50s(2), shows OC-38 cell values to be approximately equal to 8/14.19 + 3 × 10³/10⁶, which is <1, suggesting sensitization; however, the EC₅₀ values of talazoparib in single drug and combination treatments were quantified by simple linear extrapolation and are thus estimations and bootstrap simulations to calculate statistical significance were not possible.

Table 1.

Effective concentrations of ganetespib, talazoparib in OC cells.

Cell line	Ganetespib EC50, nM	Talazoparib EC50,nM
OVCAR-3	5.35	1.23
UWB 1.289	1.61	1967.52
OC-1	5.76	3584.81
OC-16	5.35	N/Aa
OC-38	14.19	N/Aa

^a Not available: Value is out of tested concentration range and too high to extrapolate

Table 2.

Combination index (CI) of talazoparib and ganetespib in OC cells.

Cell line	CI* talazoparib + ganetespib	Molar ratio talazoparib:ganetespib	CI (p value)
OVCAR-3	1.25	1:10	0.71 (p= 0.007)
		1:1	0.76 (p = 0.002)
		10:1	2.57
		40:1	1.93
UWB 1.289	0.73 (p= 0.01)	1:10	NTb
		1:1	NT
		10:1	NT
		40:1	NT
OC-1	N/Aa	1:10	NT
		1:1	0.95 (p = 0.081)
		10:1	0.65 (p = 0.003)
		40:1	0.83 (p = 0.066)
OC-16	N/A	1:10	1.01
		1:1	0.66 (p = 0.002)
		10:1	0.55 (p = 0.001)
		40:1	0.23 (p = 0.001)
		40:1	0.93 (p = 0.38)
OC-38	0.62	1:10	NT
		1:1	1.69
		10:1	2.85
		40:1	2.35

^a Not available: Value is out of tested concentration range and too high to extrapolate

^b Not tested

* Values indicate: CI > 1 antagonism, CI = 1 additive, CI < 0.9 synergy and CI < 0.5 strong synergy. P-values based on 1000 bootstrap estimates are reported to 3 decimal places. Those based on 100 bootstraps are Reported to 2 places.

Figure 3.

Ganetespib sensitizes ovarian carcinoma cells to talazoparib. Viability curves of OVCAR-3 (A), UWB 1.289 (B), OC-1 (C), OC-16 (D) cells treated with ganetespib and talazoparib alone or in combination at the indicated non-fixed concentrations or fixed molar ratios. Viability curves of OVCAR-3 cells treated with ganetespib in combination with rucaparib (E) and niraparib (F) in 1:1 molar ratio. Only statistically significant (P values <0.05) synergistic PARPi and ganetespib combinations are shown.

Talazoparib is currently under investigation in clinical trials for OC patients and recently gained FDA approval for treatment of BRCA mutant HER2 negative locally advanced or metastatic breast cancer.²⁶ This agent was selected as the PARPi for these analyses based on its high potency and selectivity.²² To confirm that the sensitization observed with ganetespib treatment extends to other drugs in this class, we assessed the effects of addition of ganetespib to two additional PARPi’s, rucaparib and niraparib, that are FDA-approved for treatment of patients with OC.^27,28 Similar to the results observed with talazoparib, combination treatment of OVCAR-3 cells with ganetespib at a 1:1 constant molar ratio with either rucaparib or niraparib significantly sensitized cells (Table 3 and Figure 3E and F). To address the specificity of ganetespib-mediated targeting of HSP90 in this combination, we also evaluated the effects of siRNA-mediated depletion of the HSP90AA1 gene. Consistent with the effects of ganetespib, depletion of HSP90AA1 in OVCAR-3 cells resulted in increased sensitivity to talazoparib-mediated PARP inhibition (Supplemental Fig. S3). Overall, these results suggest that the addition of an HSP90i has the capacity to sensitize many unique OC cell lines to the inhibitory effects of a PARP inhibitor, though the specific concentrations of each drug to achieve maximum efficacy may be tumor cell-dependent.

Table 3.

Combination index (CI) of rucaparib or niraparib and ganetespib in OVCAR-3 cells.

Molar ratio PARPi:ganetespib	CI* rucaparib + ganetespib (p value)	CI* niraparib + ganetespib (p value)
1:1	0.51 (p = 0.004)	0.58 (p = 0.002)
10:1	1.02	0.82 (p = 0.559)
20:1	1.8	1.25

* Values indicate: CI > 1 antagonism, CI = 1 additive, CI < 0.9 synergy and CI < 0.5 strong synergy. P-values based on 1000 bootstrap estimates are reported to 3 decimal places. Those based on 100 bootstraps are reported to 2 places.

Combined treatment with ganetespib and talazoparib increases cell death in OC cells

To determine if the reduction in cell viability observed in cells following combined treatment with ganetespib and talazoparib was linked to increased cell death, ganetespib and talazoparib treated OVCAR-3 and OC-1 cells were analyzed by Annexin V and 7-aminoactinomycin D (7-AAD) staining and expression of cleaved PARP. The results of this analysis show that treatment of cells with ganetespib or talazoparib alone did not result in significant increases in the percent of Annexin V⁺ cells compared to vehicle treated controls, but when cells were treated with a combination of ganetespib and talazoparib, there was a significant increase in cell death evidenced by significantly increased Annexin V⁺ cells compared to vehicle- or talazoparib-treated OVCAR-3 cells or vehicle-treated OC-1 cells (Figure 4A & 4B). The increased activation of apoptosis was independently confirmed by detection of elevated levels of cleaved PARP (Figure 4, A and B).

Figure 4.

Combined treatment with ganetespib and talazoparib significantly increases cell death in HGSOC cells. OVCAR-3 (A) and OC-1 (B) cells treated with vehicle, talazoparib, ganetespib, or talazoparib + ganetespib and analyzed for induction of cell death by Annexin V and 7-AAD staining and increased levels of cleaved PARP. Representative flow cytometry dot plots of Annexin V and 7AAD stained cells are shown with quantification of the Annexin V⁺ cells and immunoblot detection of cleaved PARP and GAPDH (loading control).

Discussion

The utility of PARP inhibitors has been demonstrated in multiple clinical trials.^3,29-33 This class of agents is a particularly promising treatment approach for patients with ovarian carcinoma, with multiple agents receiving FDA-approval as maintenance therapy following response to platinum-based agents.³⁴ The effectiveness of PARP inhibition is increased by deficiency in DNA damage repair by homologous recombination, thus patients with demonstrated vulnerabilities in HR, such as mutations or methylation in BRCA1 or BRCA2, are predicted to be most sensitive to this class of drugs.^35,36 Interestingly, in the recent ARIEL3 trial it was noted that maintenance treatment for recurrent ovarian cancer with the PARPi rucaparib resulted in increased progression-free survival as compared to placebo regardless of BRCA-mutation status or homologous recombination deficiency.³⁷ However, it is important to note that there was a trend towards greater median progression-free survival in patients with BRCA-mutant or other HR-deficient carcinoma (16.6 and 13.6 months) as opposed to the overall rucaparib-treated population (10.8 months), reinforcing observations that PARPi’s exert their greatest effect when the DNA damage response mechanism is already compromised.³⁷ To extend the benefit of PARPi more broadly in OC patients, there has been considerable interest in identifying pathways and targeted agents that could be employed to extend the utility of PARPi in patients without known defects in DDR.³⁸ Among the combinations considered, agents of particular interest are those that functionally abrogate HR.

One approach to achieve the goal of sensitization of HR proficient OCs to PARPi is targeted inhibition of the ATP-dependent molecular chaperone HSP90. The rationale for this is strong, as many of the proteins taking part in DNA repair and HR are clients of HSP90, and HSP90 is an ideal target for pharmacologic inhibition due to its abundant expression, dependence on ATP, and massive interactome.³⁹ The opportunity to affect numerous cellular processes by hitting one target is especially important for HGSOCs, which tend to be highly heterogeneous. Previous work by our group identified HSP90 as a central interactor for numerous proteins affecting cell growth and survival in OC, and targeted blockade of HSP90 inhibited the growth and survival of OC cells in culture and exhibited single-agent activity in orthotopic xenograft and transgenic mouse models of OC.¹⁴ This work further showed that targeted inhibition of HSP90 was synergized by the addition of either cisplatin or paclitaxel, or by addition of several targeted therapeutic agents. Recent work demonstrated that targeted inhibition of HSP90 impairs HR and non-homologous end joining (NHEJ) repair pathways in response to DSBs or interstrand cross-links (ICL) induced by platinum-based agents.^9,40 Abrogation of DNA repair pathways in response to HSP90 inhibition is a direct consequence of destabilization of essential HSP90 client proteins including BRCA1, BRCA2, and RAD51.^8,40,41 Prior work evaluating 17-AAG, a first-generation HSP90i, combined with olaparib support this, showing suppression of HR, decreased levels of BRCA1 and RAD51 proteins and diminished viability of SKOV-3 and 36M2 cells.⁴² In addition to these essential mediators of HR, several additional effectors of HR and cell cycle control are established clients of HSP90 including MRN complex proteins, CHK1, CDK1, FOXM1, and c-MYC.^{9,14,40,43-46} Thus, HSP90 inhibition is predicted to abrogate cell cycle control and HR, resulting in substantial sensitization to PARP inhibition, making it an attractive target to induce synthetic lethality in otherwise HR proficient cancers. From a clinical development perspective, cancer cells have long been presumed to have increased sensitivity to HSP90 inhibitors compared to normal cells, and this presumption has now been confirmed experimentally in isogenic normal and transformed cell lines with multiple small molecule HSP90i.⁴⁷ The findings of this study suggest that the increased sensitivity and selectivity of cancer cells to HSP90i is due to their increased cellular and metabolic activity compared to normal cells.⁴⁷

Herein, we showed that ganetespib effectively sensitized HGSOC cell lines to PARP inhibition, with synergistic effects on cell kill shown in most of the cell lines tested. Our analysis focused on established and novel patient-derived cell lines that did not have alterations of BRCA1 or BRCA2, underscoring the potential to achieve synergistic cell killing in cells from sporadic HGSCs in the absence of BRCA alterations or known defects in HR. The employment of novel patient-derived cell lines cultured at low passage under conditions that preserve their initial characteristics without significant cell culture-induced changes may more closely model response in tumors, thus supporting the potential clinical relevance. Interestingly, several of the cells lines tested exhibited moderate to strong resistance to talazoparib alone (e.g., UWB1.289, OC-1, OC-16, OC-38), yet were significantly sensitized to cell death by the addition of ganetespib. Coupled with the findings that ganetespib-mediated HSP90 inhibition effectively reduced the levels of critical DNA-repair and cell cycle checkpoint proteins and increased the presence of unresolved DNA damage, these synergistic drug effects are consistent with a mechanism for increasing the efficacy of PARPi by disrupting the DNA damage response in cells that have intact or functional DNA repair mechanisms. These data are in alignment with recent reports in breast and colon cancer cells that showed ganetespib-mediated inhibition of HSP90 abrogated expression of DNA-damage and/or cell cycle checkpoint proteins and disrupted gamma-irradiation induced DNA-repair in MCF-7 and HCT-116 cells.^10,11 These findings are also in line with abstract describing synergy effects in HGSOC patient-derived xenograft models treated with the HSP90i AT13387 and olaparib.⁴⁸ Taken together, the findings herein support the hypothesis that via HSP90 blockade, synthetic lethality can be achieved in HGSOC cells that would otherwise not respond well to PARP inhibition. Moreover, our findings may suggest a potential method for overcoming PARP inhibitor resistance, an important emerging clinical issue as the indications for PARP inhibitor therapy continue to expand in both the treatment and maintenance therapy settings. Further, in vitro analyses with additional patient-derived cell lines and in vivo analyses of patient-derived xenograft models are warranted to support the clinical development of this combination.

Our results show that treatment of cells with a combination of ganetespib and talazoparib results in a significant increase in cell death. Pinpointing the precise mechanism(s) of cell death is extremely challenging in the context of HSP90 inhibition as several cell death pathway targets are among the current list of known HSP90 interacting proteins.⁴⁶ Among the known HSP90 clients there are several regulators of cellular death including anti-apoptotic proteins BCL-2, BCL-XL (intrinsic pathway) and c-IAP and c-FLIP (extrinsic pathway), and the key regulator of autophagy, Beclin-1.⁴⁶ Probing levels of MCL-1 and c-FLIP proteins showed ganetespib-mediated depletion of both of these proteins in OC cells suggesting disruption of both intrinsic and extrinsic cell death pathways (data not shown). Disruption of BECN1 (Beclin-1) expression by RNA interference led to increased sensitivity of the cells to the combination of ganetespib and talazoparib, suggesting autophagy may be engaged as a mechanism of OC cell survival in response to the combination of ganetespib and talazoparib (data not shown). These observations may not be surprising given published work showing that autophagy may have a pro-survival role for OC cells and contribute to chemotherapy resistance.^49-51

HSP90 has been an attractive therapeutic target for over two decades; however, until recently, the capacity to target HSP90 clinically has proven intractable owing to the toxicity of first- and some second-generation inhibitors.³⁹ More recently, second- and third-generation HSP90i’s have been developed that mitigate the toxicity issues of earlier agents. Ganetespib is among these drugs with improved pharmacologic properties and based on good performance in early clinical trials was targeted for FDA approval; however, critical endpoints were not reached in late phase trials and clinical development of this drug was halted.^52,53 Notably, although 18 HSP90is have reached clinical development based on very promising pre-clinical data, none have yet received FDA approval. There are several reasons for this, including challenges identifying a reliable and readily accessible tumor-specific biomarker of drug targeting and activity, potential induction of tumor-promoting pathways as a consequence of high dose HSP90i and incompletely understood effects of HSP90 nuclear functions and its role in host immunity (reviewed in^52,53). The lack of clear clinical benefit as a monotherapy has led to the suggestion that the greatest potential for these agents in solid tumors lies in their use in combination with cytotoxic, targeted or immune therapeutics and/or radiation.⁵⁴ These challenges aside, there are currently five HSP90 targeting agents in clinical trials, thus the concept of the combination of an HSP90i with a PARPi to induce synthetic lethality remains clinically viable and merits consideration for further pre-clinical and clinical development for ovarian HGSC.

Funding Statement

This work was supported by a Pilot Study Award from the Rivkin Center for Ovarian Cancer (to DCC) and the FCCC Core Grant NCI P30 CA006927. Development of patient-derived ovarian carcinoma cell lines was supported in part by a grant from the Pennsylvania Department of Health (SAP#4100068716) and generous donations from the Dubrow Fund, the Bucks County Board of Associates and the Mainline Board of Associates (to DCC).

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website.