GAS6/AXL Inhibition Enhances Ovarian Cancer Sensitivity to Chemotherapy and PARP Inhibition through Increased DNA Damage and Enhanced Replication Stress

High-grade serous ovarian cancer (HGSOC) affects over 22,000 women in the United States annually and is the most deadly gynecologic malignancy. Patients with HGSOC are traditionally treated with either neoadjuvant taxane- and platinum-based chemotherapy followed by interval cytoreductive surgery or primary cytoreductive surgery followed by chemotherapy (1). Although the majority of patients respond to initial chemotherapy, over 80% of women develop resistance and die within 5 years of diagnosis (2, 3). Carboplatin, which causes DNA damage and cell death by introducing platinum-DNA adducts, is most effective in HGSOCs with increased genetic instability caused by deficiencies in homologous recombination (HR; ref. 4). Additionally, patients with HR-deficient tumors can be treated with a combination of chemotherapy plus PARP inhibitors such as olaparib. This combination is effective because PARP inhibitors in the setting of HR defects result in synthetic lethality (5). However, 50% of ovarian cancers are HR proficient (6, 7) and therefore are not susceptible to this approach. Thus, to improve care for patients with HGSOC, we need to identify therapeutic agents that can improve response to chemotherapy or PARP inhibitors in HR-proficient tumors.

One candidate target in ovarian and other cancers is growth arrest–specific 6 (GAS6), which binds to and activates the receptor tyrosine kinases Tyro3, AXL, and Mer (8, 9). AXL mRNA expression is upregulated in several ovarian cancer cell lines, and data from The Cancer Genome Atlas as well as other sources indicate that AXL and GAS6 overexpression was associated with poor prognosis in ovarian cancer (10–13). Furthermore, Buehler and colleagues found that epithelial ovarian cancers commonly had high GAS6 mRNA and protein expression, and ovarian cancers from patients with residual disease after primary cytoreductive surgery had higher GAS6 expression than those without residual disease (14, 15).

One promising GAS6/AXL inhibitor is AVB-S6–500 (AVB-500), a highly selective AXL decoy receptor that binds to GAS6 with high affinity. Given the safety profile in healthy volunteers, AVB-500 was recently granted FDA fast-track designation for potential treatment of platinum-resistant ovarian cancer. Here, our objectives were to determine the utility of AVB-500 in combination with standard-of-care treatments in ovarian cancer and to determine whether GAS6 could be associated with treatment response. We show that patients with ovarian cancer with high tumor expression of GAS6 had poor response to neoadjuvant chemotherapy and shorter progression-free as well as overall survival (OS) than those with low tumor expression of GAS6. In cell culture and in vivo mouse models, treatment with carboplatin and paclitaxel plus AVB-500 was more effective than chemotherapy alone. Mechanistically, this response was accompanied by stalled replication fork progression, increased DNA damage, and decreased HR repair (HRR) in both HR-deficient and HR-proficient chemoresistant cell lines. Finally, AVB-500 synergized with olaparib and increased replication fork progression in HR-proficient ovarian cancer cells. Together, our data support combining AVB-500 with carboplatin or olaparib to improve sensitivity to these treatments for HGSOC patients, including those with HR-proficient tumors.

This study was approved by the Washington University in St. Louis Institutional Review Board (#201407156). Patients were included if they had stage III to IV ovarian cancer of serous, endometrioid, or clear-cell histology and were planning to undergo neoadjuvant chemotherapy and interval cytoreductive surgery. Patients were excluded if they had previously undergone chemotherapy. Blood and tissue specimens were collected before and after 3 cycles of neoadjuvant chemotherapy.

Cases were evaluated for response to neoadjuvant chemotherapy according to a modified form of a published pathologic scoring system (16). All cases were scored for histologic response to neoadjuvant chemotherapy and given a chemotherapy response score (CRS) of 1 to 3 as previously published (16). Additionally, they were given a score of 0 to 3 (0, no response; 3, complete response) for radiologic and surgical response(Supplementary Table S1). Radiologic scores were calculated based on RECIST criteria. Specifically, a score of 0 was associated with progressive disease, 1 with no progressive disease but less than a partial response, 2 with a partial response (≥30% decrease in disease, no new lesions), and 3 with a complete response or disappearance of all lesions. Surgical response was extrapolated from the operative report by two separate researchers (one physician, one research assistant). A score of 0 was assigned if there was persistent large volume disease and ascites, 1 for mild treatment effects but large volume disease, 3 if there was no or minimal residual disease, and 2 for a response not otherwise specified. The sum of each response was subtracted from 10 for a final chemoresponse score between 1 and 10. Cases were considered to have a good response if the score was 5 or lower and a poor response if the score was 6 or higher. This cutoff was representative of a moderate response or higher in two of the three categories evaluated. This novel score was utilized due to its comprehensive nature and clinical applicability. Progression-free and OS were calculated from the time of interval cytoreductive surgery.

Serum was collected from patients before 3 cycles of neoadjuvant chemotherapy. The concentration of free, unbound GAS6 was measured by an ELISA in which immobilized AVB-500 was used as a capture reagent. Briefly, the plate was coated with 1 μg/mL AVB-500 in PBS and GAS6 was detected on a Bio Tek 2 plate reader with a biotinylated polyclonal anti-human GAS6 antibody (BAF885, R&D Systems) and streptavidin-conjugated to horseradish peroxidase (DY998, R&D Systems). The lower and upper limits of detection were 2 ng/mL and 128 ng/mL, respectively.

Formalin-fixed, paraffin-embedded human and xenograft tumor slides were deparaffinized and IHC performed as previously published with antibodies listed in Supplementary Table S2 (17). For patients, both pretreatment and posttreatment slides obtained at the time of cytoreductive surgery were stained. Each slide was blindly scored for quantity and intensity of positive staining by two trained members of the research team, and the scores were averaged.

OVCAR5 and OVCAR8 cells were obtained from the NCI. OVCAR3-TPMES cells were isolated from mouse mesentery tumor nodules in an OVCAR3TP intraperitoneal model (18). The OVCAR3TP cells were a generous gift from Branimir I. Sikic (19). These four cell lines were cultured in RPMI plus 10% FBS and 1% penicillin and streptomycin. To obtain POV71-hTERT, ascites from a patient with chemoresistant ovarian cancer was cultured, and tumor cells were identified morphologically. The primary cells were then immortalized via transfection with the lentiviral plasmid human telomerase reverse transcriptase (hTERT)–hygro-pBABE and cultured in DMEM (Sigma-Aldrich) supplemented with 20% FBS and 1% penicillin and streptomycin. COV362 (Sigma-Aldrich) and CAOV3 (ATCC) cells lines were cultured in DMEM with 10% FBS and 1% penicillin and streptomycin. All cells were maintained at 37 °C in a 5% CO₂ incubator. Cell line identities were confirmed by short tandem repeat profiling using the CellCheck9- Human test through IDEXX BioAnalaytics. Cell lines were confirmed Mycoplasma negative with MycoAlert Mycoplasma Detection Kit (Lonza) before experiments.

Cultured cells were lysed in 9M urea, 0.075 M Tris, pH 7.6, and proteins were quantified by the Bradford assay. Protein lysates (60–100 μg) were subjected to reducing SDS-PAGE by standard methods and transferred to a nitrocellulose membrane. Each membrane was probed with primary antibody at 4°C for 1 to 3 nights, washed, and probed with corresponding horseradish peroxidase-conjugated secondary antibodies (all antibodies are listed in Supplementary Table S2). Signal was detected with the Pierce ECL Western Blotting Substrate, and chemiluminescence was measured on a ChemiDoc (Biorad.).

Cells (5,000 per well) were plated in each well of a 96-well plate and incubated overnight. Cells were then serum starved and treated with 0.7 to 1,000 nmol/L paclitaxel (Sigma-Aldrich), 0.01–108 μmol/L carboplatin (Teva Pharmaceuticals), 1 to 5 μmol/L AVB-500 (Aravive), or a combination of these drugs. Carboplatin and AVB-500 were dissolved in saline. Paclitaxel was dissolved in DMSO for all experiments. After 72 hours, MTS (3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium) solution (Promega) was added to the cells (20 μL/well), incubated for 2 hours, and then the absorbance was measured at 490 nm with a 96-well plate reader (Tecan infinite M200 Pro). The drug combination index (CI; ref. 20) was calculated using CompuSyn software to assess for synergy between drugs.

Cells (5,000 per well) were seeded in 6-well plates, serum starved, and treated with vehicle, 0.5 μmol/L carboplatin, 50 nmol/L paclitaxel, 2 μmol/L olaparib, 1 μmol/L AVB-500, or combinations of these drugs for 72 hours. As described previously (21), this assay required a lower carboplatin concentration than that required for the DNA damage assays. Cells were then incubated in media with 10% FBS until vehicle-treated cells formed colonies optimal for visualization (14–21 days). Cells were then washed with PBS, fixed with methanol, and stained with 25% volume for volume (v/v) methanol containing 0.05% w/v crystal violet for 10 minutes. Residual staining solution was removed, and plates were washed with water and air-dried. Images were captured with a Biorad Imager System. For quantification, retained crystal violet was dissolved in 10% acetic acid, the solution was diluted eight-fold with water, and absorbance was measured at 590 nmol/L. Combenefit software was used to calculate synergism from clonogenic assays according to the Loewe model (22).

All animal experiments were done in accordance with the guidelines of the American Association for Accreditation for laboratory Animal Care and the U.S. Public Health Service Policy on Humane Care and Use of Laboratory Animals and were approved by the Washington University Institutional Animal Care and Use Committee in accordance with the Animal Welfare Act, the Guide for the Care and Use of Laboratory Animals, and NIH guidelines. For established cancer cell line xenograft models, 6-week-old female nude (nu/nu) mice (The Jackson Laboratory) were used. For patient-derived xenograft (PDX) models, 7-week-old female NOD/SCID gamma (NSG; NOD.Bg-Prkdcscid Il2rgtm1Wjl/SzJ) mice (The Jackson Laboratory) were used.

To establish subcutaneous cell line tumors, 1.5 × 10⁶ OVCAR8 cells were implanted subcutaneously into the flank of the mouse, and the mouse was allowed to rest until tumor engraftment and growth were observed. For the subcutaneous PDX models, flash frozen PB1 or ICS051 tumors were thawed and implanted subcutaneously (18). Treatment was started when the PB1 tumor volume reached 300 mm³ and when the ICS051 tumor volume reached 2 cm³. We used different tumor sizes before beginning treatment to determine whether established, larger tumors could also respond to treatment compared with smaller tumors. Intraperitoneal tumor models were established by i.p. injection of 2.5 × 10⁶ OVCAR5 or OVCAR8 cells. Treatment was initiated on day 14. After initiating treatment, tumors were measured every three days with calipers, and tumor volume was calculated with the equation V = l*w*π/6 (l, longest diameter; w, shortest perpendicular diameter).

All mice (subcutaneous or intraperitoneal tumor models) were treated as follows: 20 mg/kg paclitaxel i.p. weekly, 80 mg/kg carboplatin i.p. weekly, 30 mg/kg AVB-500 i.p. every 3 days, or 200 μL saline weekly. After 14 days of treatment mice were sacrificed 2 or 4 hours after receiving a single i.p. injection of 30 mg/kg AVB-500. Tumors were weighed, formalin fixed, and embedded in paraffin.

For the PARP inhibitor plus AVB-500 experiments, treatment was as follows: 30 mg/kg AVB-500 i.p. every 3 days, 50 mg/kg olaparib by oral gavage every 3 days, or 200 μL saline i.p. weekly. Mice were sacrificed on day 21, and tumor burden was evaluated via dissection.

OVCAR8, COV362, CAOV3, and OVCAR3-TPMES cells were plated (10,000 cells per well) in an 8-well chamber slide. Cells were then treated with 500 μmol/L carboplatin plus 1 μmol/L AVB-500 for 4 hours, 2 μmol/L Olaparib plus 1 μmol/L AVB-500 for 96 hours refreshing drug treatment daily, or treated with 10 Gy ionizing radiation. The carboplatin dose was based on the literature (23) and optimized in multiple experiments evaluating DNA damage in OVCAR8 cells. Notably, a significantly higher dose of carboplatin was necessary in order to exaggerate the DNA damage response (DDR) that is caused as a result of acute treatment. After treatment, the cells were washed with cold PBS, fixed with 4% paraformaldehyde for 10 minutes, washed with immunofluorescence (IF) Wash Buffer (25 mmol/L Tris-HCl, pH 7.5, 250 mmol/L NaCl, 0.1% Tween 20), and blocked with IF Blocking Buffer (PBS, 5% goat serum, 0.5% Triton-X-100) for 60 minutes. Cells were incubated overnight at 4 °C with primary antibodies (listed in Supplementary Table S2) in 1% goat serum, 0.1% Tween 20. Cells were then stained with secondary antibodies (1:500, Alexa Fluor 647- or FITC-conjugated; Invitrogen) and DAPI (Sigma) and mounted with Vectashield (Vector Laboratories). Cells were imaged on a Leica TCS SPE inverted confocal microscope. Raw images were exported into ImageJ and adjusted for contrast and brightness (24). Foci were counted with FociCounter (25). At least 100 cells were analyzed for each treatment group in duplicate. Student two-tailed t test was used to determine statistical significance.

To classify cell lines as HR-proficient or -deficient, the method described by Graeser and colleagues (26) and the HR RePair CAPacity (RECAP) test (27) were used. Briefly, all cells with ≥ 5 RAD51 foci were considered RAD51 positive. Cells lines were considered HR-proficient if more than 50% of geminin-positive cells were RAD51 positive, HR-deficient if less than 20% were RAD51 positive, and HR intermediate if more than 20% but less than 50% were RAD51 positive (28). This assay was used to confirm that OVCAR8 cells were HR-deficient and that COV362, CAOV3, OVCAR3-TPMES cells were HR-proficient (Supplementary Fig. S1) as previously described (29–31). Although COV362 cells were derived from a BRCA1 mutated tumor, they were confirmed to be HR-proficient. This is likely due to residual BRCA1 function as published by Pettitt and colleagues (31) Additionally, OVCAR3-TPMES and CAOV3 were confirmed to be HR-proficient (28).

A BD FACS LSRII flow cytometer was used for all flow cytometry experiments, and FlowJo software was used for all data analysis. To measure chromatin-bound replication protein A (RPA), cells were treated with 0.5 mmol/L carboplatin plus 1 μmol/L AVB-500 for 4 hours at 37°C, then washed with PBS plus 2% FBS and fixed in cytofix/cytoperm buffer (BD Biosciences) at room temperature. They were then washed with Perm/Wash buffer (BD Biosciences), stained with rat anti-RPA (Supplementary Table S2) overnight, at 4°C, washed, and stained with secondary antibody (1:1000; Alexa Fluor488; Invitrogen). Cells were washed again with Perm/Wash buffer and resuspended in PBS plus 2% FBS containing 7-AAD for analysis (32).

Cell cycle was analyzed by propidium iodide staining (33). Cells were treated with carboplatin (0.5 mmol/L), AVB-500 (1 μmol/L) alone or in combination for 24 hours. Cells were harvested by trypsinization, washed with PBS, and fixed with 70% ethanol for 2 hours on ice. After washing with PBS, cells were resuspended in PBS containing 100 μg/mL RNase A and 50 μg/mL propidium iodide, and incubated overnight at 4 °C before analysis. Fifty thousand events were measured per sample. Each assay was performed in triplicate in two independent experiments.

Exponentially growing cells were pulse-labeled with 20 μmol/L 5-Iodo-2′-deoxyuridine (IdU, Millipore Sigma-Aldrich) for 20 minutes, washed twice with PBS, pulse-labeled with 200 μmol/L 5-Chloro-2′-deoxyuridine (CldU, Millipore Sigma-Aldrich) for 1 hour, then washed twice with PBS (34, 35). In the case of cisplatin and carboplatin treatment, CldU was added concomitantly with 150 μmol/L cisplatin or 500 μmol/L carboplatin. Cisplatin was evaluated initially, as its effects have been well established in these assays (36, 37). In the case of AVB-500 treatment, cells were treated with 1 μmol/L AVB-500 24 hours before the start of the fiber assay, and 1 μmol/L AVB-500 was maintained in the cell media during the entire labeling period. In the case of olaparib treatment, cells were treated with 10 μmol/L olaparib for 24 hours before the start of the fiber assay, and 10 μmol/L olaparib was maintained in the cell media during the entire labeling period. This concentration of olaparib, used previously for DNA fiber assays (38, 39) was selected in order to maximize any potential fork perturbations caused by this drug. In all cases, cells were harvested, pelleted at 300 × g for 5 minutes at 4°C, and resuspended in PBS at 1,500 cells/μL. Then, 2 μL of cells were mixed with 6 μL of lysis buffer (200 mmol/L Tris-HCl, pH 7.5, 50 mmol/L EDTA, 0.5% SDS in water) on top of a positively charged glass slide. After 5 minutes at room temperature, slides were tilted at a 20- to 60-degree angle to spread the fibers at a constant, low speed. After air drying for 15 minutes at room temperature, DNA was fixed onto the slides with a freshly prepared solution of methanol: glacial acetic acid (3:1) for 5 minutes, dried, then stored at 4°C for at least overnight. For immuno-staining of DNA fibers, DNA was rehydrated in PBS twice for 5 minutes, then denatured with 2.5 M HCl for 1 hour at room temperature. Slides were then washed with PBS three times and blocked with 5% BSA in PBS at 37°C for 1 hour. DNA fibers were immuno-stained with rat anti-BrdU (1:100, ab6326, abcam) and mouse anti-BrdU (1:20, 347580, BD Biosciences; Supplementary Table S2) for 1.5 hours at room temperature, rinsed in PBS, washed three times with PBS-0.1% Tween-20 for 5 minutes, rinsed with PBS, then incubated with anti-rat Alexa Fluor 488 and anti-mouse Alexa Fluor 546 (1:100, A21470 and A21123, respectively; Thermo Fisher Scientific) for 1 hour at room temperature. After three washes with PBS-0.1% Tween-20 of 5 minutes each, slides were rinsed in PBS, then mounted with Prolong Gold Antifade Reagent (P36930, Thermo Fisher Scientific). Images were acquired with red and green channels on a fluorescent microscope (LEICA DMU 4000B, 63X/1.4 oil immersion objective) with a LEICA DFC345FX camera or with LAS AF software using a TCS SP5 confocal microscope (LEICA, 63×/1.4 oil immersion objective). The fiber experiments were repeated for two or three independent biological replicates in CAOV3 and OVCAR3-TPMES cell lines, respectively. At least 15 images were taken across the whole slide using only one channel to select the regions for the images. At least 150 to 200 individual tracts were scored for each dataset. DNA fiber tracts were analyzed only if the IdU and CldU signals were contiguous and if the beginning and end of each color were unambiguously defined. The length of each tract was measured manually with the segmented line tool on ImageJ software (NIH). The pixel values were converted into μm by using the scale bar generated by the microscope software. Ratios (CldU/IdU) or total tract lengths (IdU + CldU) from individual DNA fibers were plotted as scatter dot plot with the horizontal line representing the median.

GraphPad Prism 7 software was used for statistical analysis and IC₅₀ calculations. Statistical significance was determined by the Student t test, χ² test, or one-way ANOVA as appropriate. Significance was defined as P < 0.05. For the DNA fiber assays, data were pooled from three independent experiments. Statistical differences in DNA fiber tract lengths were determined by a one-way ANOVA (Kruskal–Wallis test) followed by a Dunn multiple comparison test.

Given the associations between tumor AXL expression and outcomes in cancer patients (14, 15, 18), we asked whether AXL expression correlated with resistance to neoadjuvant chemotherapy in patients with ovarian cancer. To answer this question, we performed IHC on matched serum and tumor samples from 38 patients before neoadjuvant paclitaxel and carboplatin. Seventy-nine percent of patients had stage III disease and 21% had stage IV disease. All patients had at least mixed high-grade serous histology. BRCA1/2 mutations were identified in 15.8% of patients. Median follow up was 33 months. Supplementary Table S3 shows the characteristics of our study population stratified by response to neoadjuvant chemotherapy on a 1 to 10 scale in which a score of 5 or lower was considered a good response, and a score of 6 or higher was considered a poor response (see Methods for details). By this stratification, 24 (63.2%) of the patients had a good chemoresponse and 14 (36.8%) had a poor response. There were no significant differences in the demographics or tumor characteristics between these two groups. We assigned each tumor an AXL expression score between 0% and 100%. Consistent with previous findings, AXL expression was higher in tumors with poor chemoresponse than in those with a good chemoresponse (72% vs. 60%, P = 0.04; Fig. 1A).

We next measured AXL ligand, GAS6, in serum and tumors collected before neoadjuvant chemotherapy from the same patients. In an attempt to stratify patients into high- and low-risk groups based on GAS6 levels, we explored all observed GAS6 values, identified cut-off values that minimized the P value, and made statistical adjustments to resolve the increased false positive rate (40). By this method, patients were grouped into high versus low serum GAS6 concentrations (>25 ng/mL vs. <15ng/mL) and high versus low tumor GAS6 expression (>80% vs. <35%). Supplementary Table S4 shows the characteristics of patients with high and low serum GAS6 concentrations. There were no differences in the demographics or tumor characteristics between these two groups. Patients with high serum GAS6 concentrations had worse chemoresponse scores than those with low serum GAS6 concentrations (mean score of 5.5 vs. 3.4, P = 0.023; Fig. 1B). Similarly, patients with high tumor GAS6 expression had significantly worse chemoresponse scores than those with low tumor GAS6 expression (mean score of 6.3 vs. 2.8, P = 0.002; Fig. 1B). Moreover, patients with high serum GAS6 concentrations had significantly shorter progression-free survival (PFS; P = 0.002) and shorter OS (P = 0.002) than those with low serum GAS6 concentrations. Likewise, patients with high tumor GAS6 expression had shorter PFS (P = 0.015) than those with low tumor GAS6 expression (Fig. 1C). Serum GAS6 concentrations obtained prior to treatment were not significantly different between patients with and without BRCA mutations (23.6 ± 8.7 vs. 21.7 ± 7.14, P = 0.615). However, tumor GAS6 levels were significantly lower in patients with BRCA mutations than without (59.6 ± 20.4 vs. 41.3 ± 17.9, P = 0.46). Thus, high GAS6 concentrations in serum and tumor before neoadjuvant chemotherapy correlated with poor chemoresponse, shorter PFS, and shorter OS.

To better understand how GAS6 expression correlated with tumor burden we evaluated associations between both serum and tumor GAS6 levels and the administration of chemotherapy, the presence of ascites, and levels of CA 125, a prominent biomarker in the diagnosis and management of ovarian cancer. Serum GAS6 levels decreased significantly after 3 cycles of chemotherapy (23.5 ± 8.4 vs. 18.5 ± 9.3, P < 0.001), but tumor levels did not (55.3 ± 19.6 vs. 52.5 ± 20.1, P = 0.413). There was no correlation between serum or tumor GAS6 and CA 125 or ascites.

Given our finding that tumors with poor response to neoadjuvant chemotherapy had high GAS6 expression, we wondered whether a combination of the GAS6 inhibitor AVB-500 plus chemotherapy would be more effective than chemotherapy alone in reducing ovarian tumor cell growth. We evaluated AXL/GAS6 expression in multiple chemoresistant ovarian cancer cell lines (Supplementary Fig. S2; ref. 18). We then performed clonogenic growth assays in the chemoresistant cell lines OVCAR5, POV71-hTERT, OVCAR8, and COV362. In all four cell lines, cells treated with carboplatin or paclitaxel plus AVB-500 formed significantly fewer colonies than cells treated with chemotherapy alone (P < 0.05; Fig. 2A and B). To evaluate the effects of drug combinations, we calculated CIs in which synergism was defined as CI < 1, additive effect as CI = 1, and antagonism as CI > 1. In POV71-hTERT, OVCAR5, and OVCAR3-TPMES cells, AVB-500 synergized with paclitaxel (Supplementary Fig. S3). In OVCAR5 cells, AVB-500 synergized with carboplatin (CI 0.31). We also evaluated synergism using Loewe analysis which revealed that AVB-500 synergized with carboplatin to inhibit colony formation in OVCAR8 and OVCAR3-TPMES cells (Fig. 2C).

Given our in vitro findings, we next wanted to know whether AVB-500 plus chemotherapy was more effective than chemotherapy alone in in vivo ovarian cancer models. We first created three models in which we subcutaneously injected the chemoresistant, HR-proficient PDX model PB1 or ICS051, or the chemoresistant, HR-deficient cell line OVCAR8 into mice. After the PB1 and OVCAR8 tumors reached approximately 300 mm³ and the ICS051 tumors reached 2 cm³, mice were treated with vehicle, carboplatin plus paclitaxel (chemotherapy), AVB-500 alone, or chemotherapy plus AVB-500 for 12 days. In mice that received chemotherapy plus AVB-500, tumors grew more slowly and reached lower volume and weight than those that received chemotherapy alone (Fig. 3A; Supplementary Fig. S4).

To further assess the effectiveness of AVB-500 plus chemotherapy, we created in vivo models in which we intraperitoneally implanted the chemoresistant cell lines OVCAR8 or OVCAR5 into mice. After 14 days, we treated the mice with either vehicle, carboplatin, and paclitaxel (chemotherapy), AVB-500, or chemotherapy plus AVB-500 for 14 days and then assessed tumor burden. Mice that received chemotherapy plus AVB-500 had significantly fewer tumor nodules and smaller tumor mass than mice that received chemotherapy alone (Fig. 3B). IHC analysis revealed that tumors from the mice treated with AVB-500 or chemotherapy plus AVB-500 had significantly less staining for GAS6 and the downstream protein phospho-AKT (pAKT) than those treated with vehicle or chemotherapy alone (Fig. 3C; Supplementary Fig. S5).

Collectively, both in vivo and in vitro results suggest that chemotherapy plus AVB-500 was more effective at reducing ovarian tumor cell growth than chemotherapy alone in both HR-proficient and HR-deficient cells.

We next sought to determine the mechanism by which AVB-500 plus chemotherapy was more effective than chemotherapy alone. Paclitaxel causes cell death by inhibiting microtubule assembly, whereas carboplatin causes cell death by causing platinum-DNA adducts that lead to DNA double-strand breaks. Because increased DNA damage has been associated with increased platinum sensitivity and improved efficacy of DNA damaging agents, we hypothesized that AVB-500 would increase the DNA damage caused by carboplatin in both HR-proficient and HR-deficient cell lines. To test this hypothesis, we treated OVCAR8, COV362, CAOV3, and OVCAR3-TPMES cell lines with vehicle, AVB-500, carboplatin, or carboplatin plus AVB-500 for 4 hours and then performed IF assays. Cleaved caspase 3, a marker of apoptosis, was not detectable at this time point (Supplementary Fig. S6) excluding apoptosis as the cause of DNA damage. To assess the extent of DNA damage, we performed IF for the DNA damage marker γH2AX, as formation of γH2AX foci is a validated approach to evaluate DNA damage that occurs as a result of platinum therapy (23). As expected, cells treated with carboplatin had more γH2AX foci than cells treated with AVB-500 or vehicle. Moreover, consistent with previous results (41), cells from all four lines treated with AVB-500 alone had more γH2AX foci than cells treated with vehicle. Additionally, cells treated with carboplatin plus AVB-500 had significantly more γH2AX foci than cells treated with carboplatin alone (Fig. 4A). These results suggest that AVB-500 increased DNA damage independently and enhanced carboplatin-induced DNA damage. To confirm this finding, we performed comet assays in which cells with damaged DNA have comet "tails". The intensity of the tails reflect the extent of DNA damage. We confirmed that OVCAR8 cells treated with AVB-500 plus carboplatin had longer tails than cells treated with carboplatin alone (Supplementary Fig. S7).

Because increased DNA damage can impair replication, we next asked whether replication fork progression was impaired in ovarian cancer cells treated with carboplatin plus AVB-500. We used the single-molecule DNA fiber approach to monitor fork progression in OVCAR3-TPMES cells treated with AVB-500 alone, AVB-500 plus carboplatin, or AVB-500 plus cisplatin. Cells were pretreated with 1 μmol/L AVB-500 for 24 hours because downregulation of pAKT, a downstream target of GAS6, was seen after 24 hours (Supplementary Fig. S8). Cells were then incubated with the thymidine analog IdU for 20 minutes followed by the thymidine analog CldU and concomitantly treated with carboplatin or cisplatin for 60 minutes (Fig. 4B). In AVB-500–treated samples, AVB-500 was maintained in the cell culture media for the entire duration of the thymidine analog incorporation. The CldU/IdU ratios were similar in cells treated with vehicle, carboplatin, cisplatin, or AVB-500 alone, indicating that these single agents did not affect fork progression in OVCAR3-TPMES cells, at least at the concentrations used in these experiments. However, combined treatment of the OVCAR3-TPMES cells with carboplatin plus AVB-500 or cisplatin plus AVB-500 led to a significant reduction of CldU incorporation and overall lower CldU/IdU ratios (Fig. 4B). We saw a similar effect in the HR-proficient cell line CAOV3 (Fig. 4B). These findings suggest that treatment with carboplatin plus AVB-500 causes DNA damage that significantly hinders replication fork progression, which may contribute to the increased chemoresponse observed in cells cotreated with AVB-500 and carboplatin.

DNA damage caused by platinum chemotherapy results in activation of a signaling cascade referred to as the DDR. The goal of the DDR is to maintain genomic stability by activating DNA repair, cell-cycle arrest, or apoptosis (42). The type of cell response depends on the amount of DNA damage. Enhanced DDR is a well established mechanism for chemotherapy resistance. Therefore, DDR inhibition is a valuable target to sensitize cancer cells to platinum chemotherapy. Given that cells treated with AVB-500 plus chemotherapy had more DNA damage than cells treated with chemotherapy alone, we sought to define the effects of AVB-500 on key modulators of DNA damage. We first evaluated expression of key proteins involved in DNA damage repair and found that treatment of OVCAR8 cells with carboplatin plus AVB-500 resulted in decreased expression of phospho-ATM and increased expression of phospho-Chk1 (Supplementary Fig. S9). We then treated OVCAR8, COV362, OVCAR3-TPMES, and CAOV3 cells with vehicle, AVB-500, carboplatin, or carboplatin plus AVB-500 and performed IF and quantified the number of RAD51 foci per cell. In all 4 cell types, cells treated with carboplatin plus AVB-500 had significantly fewer RAD51 foci than cells treated with carboplatin alone, suggesting that AVB-500 treatment reduced HRR (Fig. 5A). To confirm this finding, we used flow cytometry to measure the number of RPA-positive cells in S/G2 phase in each treatment condition. In all four cell lines, a smaller percentage of S/G2 phase cells treated with carboplatin plus AVB-500 were RPA-positive than cells treated with carboplatin alone (Fig. 5B; Supplementary Fig. S10A). Importantly, cell-cycle analysis confirmed that the treatment regimens did not alter the proportion of cells in different phases of the cell cycle (Supplementary Fig. S10B), suggesting that the effects on DNA damage repair were not due to cell-cycle disruption.

It was possible that this functional decrease of HRR efficiency was a result of increased 53BP1 binding, which inhibits DNA end resection and thus HRR. To test this idea, we treated cells with vehicle, AVB-500 alone, carboplatin alone, or AVB-500 plus carboplatin, performed IF, and counted γH2AX and 53BP1 foci. We found that HR-deficient and HR-proficient cells treated with carboplatin plus AVB-500 had significantly more 53BP1 foci than cells treated with carboplatin alone (Fig. 5C), and the 53BP1 foci appeared to colocalize with γH2AX foci (Fig. 5C). Western blot analysis demonstrated that total amounts of RAD51 and 53BP1 were similar in all treatment groups, indicating that the observed differences in RAD51 and 53BP1 foci were not due to variable protein expression (Supplementary Fig. S8). Taken together, these findings suggested that AVB-500 impaired HRR in ovarian cancer cells treated with chemotherapy.

Given that AVB-500 appeared to impair HRR in response to DNA damage, we asked whether treatment with AVB-500 would increase sensitivity to the PARP inhibitor olaparib. First, we performed in vitro colony formation assays and found that both HR-deficient (OVCAR8) and HR-proficient (COV362, OVCAR3-TPMES, and CAOV3) ovarian cancer cells were more sensitive to AVB-500 plus olaparib than to olaparib alone (Fig. 6A). Moreover, Loewe analysis revealed that AVB-500 synergized with olaparib to inhibit colony formation in OVCAR3-TPMES cells (Fig. 6B). These in vitro findings led us to perform an in vivo experiment in which we intraperitoneally injected OVCAR8 cells into mice, allowed the tumors to grow for 7 days, and then treated the mice with vehicle, AVB-500, olaparib, or AVB-500 plus olaparib until day 21. Mice treated with olaparib plus AVB-500 had fewer tumor nodules and lower total tumor weight than mice treated with olaparib alone (Fig. 6C).

IF analysis demonstrated that OVCAR8 and OVCAR3-TPMES cells treated with AVB-500 with olaparib had more γH2AX foci and fewer RAD51 foci than cells treated with olaparib alone (Supplementary Fig. S11). We then asked whether replication fork progression was impaired in ovarian cancer cells treated with olaparib plus AVB-500. We used the single-molecule DNA fiber approach to monitor fork progression in OVCAR3-TPMES cells treated with AVB-500 alone, olaparib alone, or AVB-500 plus olaparib. Cells were pretreated with 1 μmol/L AVB-500 and 10 μmol/L olaparib for 24 hours. Cells were then incubated with the thymidine analog IdU for 30 minutes followed by the thymidine analog CldU for 30 minutes (Fig. 6D). AVB-500 and olaparib was maintained in the cell culture media for the entire duration of the thymidine analog incorporation. The IdU plus CldU total tract lengths (μm) were similar in cells treated with vehicle or AVB-500 alone indicating that these single agents did not affect fork progression in OVCAR3-TPMES cells. However, single-agent olaparib as well as the combined treatment of the OVCAR3-TPMES cells with olaparib plus AVB-500 led to a significant increase in total tract length (Fig. 6D). Our findings support those previously published by Maya–Mendoza and colleagues which suggest that PARP inhibition increases the speed of replication forks (39) and suggest that the combined treatment of olaparib plus AVB-500 augments this increase in replication fork speed resulting in increased DNA damage and genomic instability. Together, these findings indicated that AVB-500 increased sensitivity to a PARP inhibitor in both HR-deficient and HR-proficient ovarian cancer cells.

In this paper, we present several lines of evidence that the GAS6 inhibitor AVB-500 improved ovarian cancer cell sensitivity to both traditional chemotherapy and PARP inhibitors. First, we demonstrated that AVB-500 improved ovarian tumor cell response to chemotherapy both in vitro and in vivo. This is due to increased DNA damage, replication fork perturbation, and decreased HRR. Accordingly, AVB-500 improved in vitro and in vivo response to the PARP inhibitor olaparib in both HR-deficient and HR-proficient ovarian cancer cells. Lastly, our data suggests that GAS6 in both patient serum and tumor is associated with outcomes in women with ovarian cancer.

AVB-500 is a recombinant protein containing a truncated, modified portion of the extracellular region of human AXL fused to the human immunoglobulin G1 heavy chain to induce dimerization. Because AVB-500 binds GAS6 with 200-fold higher affinity than the soluble receptor AXL, GAS6′s highest affinity receptor (43), AVB-500 is less likely to have off-target effects than traditional small-molecule tyrosine kinase inhibitors. Consistent with this idea, AVB-500 caused minimal side effects at therapeutic doses in a healthy volunteer trial (44). Thus, AVB-500 is currently being evaluated in a phase 1b/2 trial in platinum-resistant, recurrent ovarian cancer (NCT03639246).

In our work, AVB-500 improved response to traditional chemotherapy, leading to decreased cell viability and tumor burden. These findings align with previous reports that AXL inhibition improves sensitivity to chemotherapy (11, 45). Our data are consistent with data indicating that AXL inhibition prevented the normal DDR (41) and gives insight into the mechanism. Specifically, we show that cells treated with chemotherapy plus AVB-500 had more γH2AX foci, more 53BP1 foci, fewer RAD51 foci, less RPA binding, and slower replication fork progression than cells treated with chemotherapy alone, indicating that AVB-500 impaired HRR. Our findings suggest that the etiology of the HRR deficiency lies in increased 53BP1 binding and thus prevention of DNA end resection and subsequent HRR. Our proposed model is presented in Supplementary Fig. S12 (46, 47). Increases in DNA damage upon AVB-500 and platinum chemotherapy treatment likely also underlie the replication fork perturbations we observed by the DNA fiber assay. Given our findings, AVB-500 may improve ovarian cancer response to traditional taxane- and carboplatin-based chemotherapy.

Our finding that cells treated with AVB-500 plus chemotherapy had fewer RAD51 foci than cells treated with chemotherapy alone has important clinical implications, as HRR deficiencies arecorrelated with response to chemotherapy and PARP inhibitors in ovarian cancer (48). Decreased formation of RAD51 foci is strongly predictive of a complete pathologic response to chemotherapy (26). Additionally, RAD51 foci formation defects have been correlated with platinum sensitivity, platinum-free interval, time to progression, and OS in ovarian cancer (26, 28, 49). Importantly, we found that AVB-500 caused decreased RAD51 foci formation in both HR-deficient (OVCAR8) and HR-proficient (CAOV, COV362, and OVCAR3-TPMES) ovarian cancer cell lines. Thus, AVB-500 may be broadly applicable in treating ovarian cancer and other solid tumors.

Our data suggest that AVB-500 can improve ovarian cancer response to PARP inhibitors. In the upfront clinical treatment of patients with advanced HGSOC with HR deficiency caused by BRCA mutations, PARP inhibitor maintenance after traditional chemotherapy offered a strong survival benefit (hazard ratio of 0.3; ref. 6). However, this upfront clinical benefit has not been demonstrated in patients with HR-proficient tumors. Hill and colleagues and Castroviejo-Bermejo and colleagues showed that RAD51 foci formation is likely more predictive of response to PARP inhibitors and DNA-damaging agents than mutations in HR genes such as BRCA given the risk of gene reversions and the fact that only half of HR-deficient tumors harbor mutations in HR genes (50–54). Specifically, in breast and ovarian cancers, decreased RAD51 foci formation is strongly correlated with improved PARP inhibitor response (50–54). Thus, our findings that treatment with AVB-500 resulted in decreased RAD51 foci formation and that AVB-500 plus olaparib decreased in vitro cell viability in both HR-proficient and HR-deficient tumor cells and in vivo tumor burden suggest that AVB-500 could be effectively combined with PARP inhibitors in patients with ovarian cancer with and without BRCA mutations. Balaji and colleagues demonstrated similar findings in small cell lung, breast, and head and neck cancers, further validating our results (41). Additionally, our findings that the addition of AVB-500 to olaparib increases replication fork speed are novel. It has been suggested that PARP inhibitor works primarily by preventing single-stranded breaks (SSB) and therefore the cell must rely on HR. Further research demonstrates that PARP inhibitor is also involved in replication fork progression (55). Specifically, PARP inhibition has been associated with increased speed of replication forks and as a result increased SSB and heavier reliance of HR. Our findings not only confirm these results, but also suggest that AVB-500 in addition to PARP inhibition further increases replication fork speed and thus genome instability. If our findings are validated in clinical trials, combined treatment with AVB-500 and a PARP inhibitor could significantly improve outcomes for patients with ovarian cancer, particularly those with HR-proficient tumors.

Lastly, previous reports have shown that GAS6 expression correlates with survival in ovarian cancer (14, 15, 18), but whether GAS6 concentration could be used to predict response to carboplatin and paclitaxel has not been fully addressed. Our data demonstrates that high GAS6 levels in both serum and tumor tissue is associated with poor chemoresponse. Future studies evaluating GAS6 as a biomarker to predict which patients will have tumors that are resistant to platinum-chemotherapy and thus might benefit from inhibition of GAS6/AXL will be investigated.

L.M. Kuroki reports grants from P50 CA244431, Doris Duke Fund to Retain Clinical Scientists, KL2 TR000450; and nonfinancial support from GOG Foundation New Investigator Award outside the submitted work. I.S. Hagemann reports personal fees from Change Healthcare; and personal fees from serving as an expert witness outside the submitted work. C.K. McCourt reports other support from UpToDate outside the submitted work. P.H. Thaker reports personal fees from Aravive, Seagen, Agenus, Novocure, AstraZeneca, Celsion, Iovance, and Stryker; grants and personal fees from Merck and GlaxoSmithKline; and personal fees from Eisai outside the submitted work. M.A. Powell reports personal fees from Roche/Genetech, Tesaro/GSK, and Clovis Oncology; and personal fees from AstraZeneca outside the submitted work. K.C. Fuh reports personal fees from Aravive outside the submitted work; in addition, K.C. Fuh and E.B. Rankin have patents for US Patent PCT/US2011/022125 with royalties paid. No disclosures were reported by the other authors.

M.M. Mullen: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. E. Lomonosova: Conceptualization, data curation, software, formal analysis, supervision, validation, investigation, methodology, writing–original draft, writing–review and editing. M.D. Toboni: Conceptualization, data curation, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. A. Oplt: Conceptualization, resources, data curation, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. E. Cybulla: Conceptualization, resources, data curation, software, formal analysis, supervision, validation, investigation, methodology, writing–original draft, writing–review and editing. B. Blachut: Data curation, software, formal analysis, investigation, methodology, writing–review and editing. P. Zhao: Conceptualization, formal analysis, investigation, methodology, writing–review and editing. H. Noia: Conceptualization, resources, data curation, software, formal analysis, supervision, validation, investigation, methodology, writing–review and editing. D. Wilke: Conceptualization, resources, data curation, software, formal analysis, supervision, visualization, methodology, writing–original draft, writing–review and editing. E.B. Rankin: Resources, supervision, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. L.M. Kuroki: Resources, data curation, investigation, writing–review and editing. A.R. Hagemann: Resources, data curation, investigation, writing–review and editing. I.S. Hagemann: Resources, data curation, investigation, writing–review and editing. C.K. McCourt: Resources, data curation, investigation, writing–review and editing. P.H. Thaker: Resources, data curation, investigation, writing–review and editing. D.G. Mutch: Resources, data curation, investigation, writing–review and editing. M.A. Powell: Resources, data curation, investigation, writing–review and editing. N. Mosammaparast: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, methodology, writing–review and editing. A. Vindigni: Conceptualization, resources, data curation, software, formal analysis, supervision, validation, methodology, writing–original draft, writing–review and editing. K.C. Fuh: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

Support for this work was provided by a Reproductive Scientist Development Program award (NICHD 2K12HD000849-28, to K.C. Fuh), a GOG Foundation grant (to M.M. Mullen), and a NCI F30 fellowship (F30C254215, to E. Cybulla). M.M. Mullen holds a Dean's Scholars Award from the Washington University Division of Physician-Scientists, which is funded by a Burroughs Wellcome Fund Physician-Scientist Institutional Award (1020047). We thank Deborah Frank, PhD and Anthony Bartley for scientific editing and Andrea Byrum, PhD, for technical support.

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