BLM overexpression as a predictive biomarker for CHK1 inhibitor response in PARP inhibitor-resistant BRCA-mutant ovarian cancer

Abstract

Poly(ADP-ribose) polymerase inhibitors (PARPis) have changed the treatment paradigm in breast cancer gene (BRCA)-mutant high-grade serous ovarian carcinoma (HGSC). However, most patients eventually develop resistance to PARPis, highlighting an unmet need for improved therapeutic strategies. Using high-throughput drug screens, we identified ataxia telangiectasia and rad3-related protein/checkpoint kinase 1 (CHK1) pathway inhibitors as cytotoxic and further validated the activity of the CHK1 inhibitor (CHK1i) prexasertib in PARPi-sensitive and -resistant BRCA-mutant HGSC cells and xenograft mouse models. CHK1i monotherapy induced DNA damage, apoptosis, and tumor size reduction. We then conducted a phase 2 study (NCT02203513) of prexasertib in patients with BRCA-mutant HGSC. The treatment was well tolerated but yielded an objective response rate of 6% (1 of 17; one partial response) in patients with previous PARPi treatment. Exploratory biomarker analyses revealed that replication stress and fork stabilization were associated with clinical benefit to CHK1i. In particular, overexpression of Bloom syndrome RecQ helicase (BLM) and cyclin E1 (CCNE1) overexpression or copy number gain/amplification were seen in patients who derived durable benefit from CHK1i. BRCA reversion mutation in previously PARPi-treated BRCA-mutant patients was not associated with resistance to CHK1i. Our findings suggest that replication fork-related genes should be further evaluated as biomarkers for CHK1i sensitivity in patients with BRCA-mutant HGSC.

Fig. 1.. In vitro drug screens identify ATR/CHK1 inhibitors as active drugs in both PARPi-sensitive and -resistant BRCA-mutant HGSC cell lines.

(A) Heatmap of ranked drug activities for PARPi-sensitive and -resistant BRCA2-mutant HGSC cell lines, including PARPi-sensitive PEO1, PARPi-resistant PEO1/OlaR, and de novo PARPi–resistant PEO4 cells. Single-agent drug activities were screened using the MIPE 5.0 library of approved and investigational drugs (oncology drugs, n = 1082). Activity scores are based on Z-AUC. The average Z-AUC value >0 indicates inactive drugs, whereas <0 represents active drugs (15). The most active drugs based on their primary mechanisms of action, chemotherapy drugs for ovarian cancer, and PARPis are shown on the right. (B) Mechanistic classes enriched among highly active drugs are shown by the Reactome database (https://reactome.org/) analysis. (C and D) Cell growth was validated using XTT (C) and colony-forming assays (D) in BRCA1-null UWB, BRCA2-mutant PEO1, BRCA2 reversion mutation PEO4, and acquired PARPi–resistant UWB (UWB/OlaR) and PEO1 (PEO1/OlaR and PEO1/OlaJR) cell lines. (C) Short-term cell growth was evaluated by the XTT assay. Cells were treated with the PARPi olaparib (top) or the CHK1i prexasertib (bottom) at the indicated doses for 72 hours. IC₅₀ values were calculated using GraphPad Prism v7.1. (D) Long-term cell proliferation was examined using colony-forming assays. Cells were seeded at low density and treated with prexasertib (0.5 nM for UWB and 5 nM for all other cell lines) or olaparib (10 μM) and grown for 12 to 15 days. Colonies were visualized by 0.01% (w/v) crystal violet staining. Quantification was performed by ImageJ. (E) On-target effect of prexasertib was assessed by immunoblotting of p-CHK1 (S296) and total CHK1. Densitometric values of p-CHK1 (S296) relative to total CHK1 are shown. All experiments were repeated at least in triplicate, and data are shown as means ± SEM. ***P < 0.001; ns, not significant. CHK1i, CHK1 inhibitor; HGSC, high-grade serous ovarian cancer; PARPi, PARP inhibitor; UWB, UWB1.289; Z-AUC, Z-transformed area under the curve.

Fig. 2.. CHK1i monotherapy disrupts fork stabilization and HR restoration in PARPi-resistant BRCA-deficient HGSC in vitro and in vivo models.

(A) Replication fork stability in PARPi-sensitive (PEO1 and UWB), acquired PARPi–resistant (PEO1/OlaR, PEO1/OlaJR, and UWB/OlaR), and de novo PARPi–resistant (PEO4) cells was measured by DNA fiber assays. Cells were incubated with CldU and then IdU and coincubated with prexasertib (20 nM) or olaparib (20 μM). Representative images are shown (left). Dot plots of IdU (red) to CldU (green) tract length ratios in treated cells are plotted (right). A lower ratio of IdU/CldU indicates fork destabilization and hindered replication, suggesting higher replication stress. (B to D) Cells were treated with prexasertib (5 nM for all cell lines except 0.5 nM for UWB) or olaparib (10 μM) for 48 hours. (B) HR status was measured by immunofluorescence staining of RAD51 foci (green). Representative images are shown (left), and the percentage of cells with >5 RAD51 foci is plotted (right). Scale bar, 40 μm. DNA damage was examined by (C) the alkaline comet assay and (D) immunofluorescence staining of γH2AX foci (a marker of DNA double-strand breaks). (C) Representative images of DNA fragments are shown (left), and the percentage of DNA in comet tails is plotted (right). Scale bar, 100 μm. (D) The γH2AX foci (pink) were assessed by immunofluorescence staining. Cell nuclei were stained with 4′,6-diamidino-2-phenylindole (blue). Representative images are shown (left). The percentage of cells with 5 to 15 γH2AX foci, representing cells with DNA damage, and cells with pan-γH2AX staining, indicating cell apoptosis, is plotted (right). Scale bar, 50 μm. The above experiments were repeated, at least in triplicate, and data are shown as means ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001. (E) In vivo assessment of prexasertib (8 mg/kg) and olaparib (100 mg/kg) in PARPi-sensitive and PARPi-resistant BRCA2-mutant HGSC xenograft tumors (n = 5 per group). Tumor volumes are plotted (top). The body weight of mice was measured once per week to monitor drug tolerance (bottom). Data are shown as means ± SEM. ***P < 0.001, prexasertib versus vehicle; ###P < 0.001, olaparib versus vehicle. CldU, 5-chloro-2′-deoxyuridine; HR, homologous recombination; IdU, 5-iodo-2′-deoxyuridine.

Fig. 3.. Clinical trial design and activity of prexasertib in BRCA-mutant HGSC.

(A) The CONSORT flow diagram. In total, there were 22 patients enrolled in the study, with 18 evaluable for RECIST response. (B) Waterfall plot showing the best response to treatment in 18 evaluable patients. Best RECIST responses (percentage change from baseline in tumor size) are shown according to the number of previous treatments, PARPi-free interval, presence of BLM and CCNE1 gene amplification/gain/deletion, and the presence of BLM and CCNE1 mRNA up-regulation in pretreatment biopsy samples. The horizontal dotted line indicates the threshold for PR (30% reduction in tumor size from baseline). (C) Swimmer plot showing duration of treatment for 18 evaluable patients. The vertical dotted line indicates the threshold for clinical benefit (CR + PR + SD ≥ 6 months). The cross symbol (+) in (B) and (C) indicates the PARPi-naïve, CR patient. The hash symbol (#) represents cancers with BRCA2 mutation; all others have BRCA1 mutations. Asterisks (*) indicate cancers with BRCA reversion mutations detected (see table S4 for more details). The caret symbol (^) indicates patients with CA-125 response (see fig. S3C for more details). RECIST, Response Evaluation Criteria in Solid Tumors; CR, complete response; PR, partial response; SD, stable disease; PD, progression of disease; N/A, not applicable; no., number.

Fig. 4.. Exploratory biomarker analyses reveal differentially up-regulated pathways associated with clinical benefit.

(A) Schema for exploratory biomarker analyses. (B) Gene set enrichment analysis by Hallmark, KEGG, and Reactome databases of RNA-seq data shows up-regulated (red) and down-regulated (blue) gene set pathways in those with clinical benefit (left) but not in those with no clinical benefit (right). cfDNA, cell-free DNA; IHC, immunohistochemistry; KEGG, Kyoto Encyclopedia of Genes and Genomes; NES, normalized enrichment score; Rep., replication.

Fig. 5.. High mRNA expression of BLM and CCNE1 is associated with clinical benefit.

(A) The mRNA expression of genes was analyzed from RNA-seq data from patients with clinical benefit (n = 4) versus those with no clinical benefit (n = 11). The x axis shows log₂ fold change (>0 indicates genes enriched in patients with clinical benefit, whereas <0 indicates genes enriched in patients with no clinical benefit). The y axis shows the Wilcoxon −log₁₀ P values, and the horizontal dotted line represents a significance threshold of P = 0.05, with significantly enriched genes falling above the line. (B and C) The mRNA expression of RecQ helicase BLM (B) and CCNE1 (C) is shown in patients with clinical benefit and those with no clinical benefit. Lines represent median with 95% confidence interval. CPM, counts per million.

Fig. 6.. BLM overexpression increases the sensitivity of CHK1i in HGSC cell lines.

(A) Basal abundances of BLM in parental (BRCA-mutant PEO1 and UWB), acquired PARPi–resistant (PEO1/OlaR, PEO1/OlaJR, and UWB/OlaR), and de novo PARPi–resistant (PEO4) cell lines were analyzed by immunoblotting. Densitometric values of BLM relative to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (top) and representative colony formation images of cells with or without prexasertib treatment are shown (bottom). (B and C) Cells transfected with BLM overexpression plasmids for 48 hours were treated with or without prexasertib for another 48 hours. Cell viability was measured by the XTT assay (B). IC₅₀ values (C) from (B) were calculated using GraphPad Prism v7.1. Experiments were repeated, at least in triplicate, and data are shown as means ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001.