Therapeutic targeting of CPSF3-dependent transcriptional termination in ovarian cancer

Abstract

Transcriptional dysregulation is a recurring pathogenic hallmark and an emerging therapeutic vulnerability in ovarian cancer. Here, we demonstrated that ovarian cancer exhibited a unique dependency on the regulatory machinery of transcriptional termination, particularly, cleavage and polyadenylation specificity factor (CPSF) complex. Genetic abrogation of multiple CPSF subunits substantially hampered neoplastic cell viability, and we presented evidence that their indispensable roles converged on the endonuclease CPSF3. Mechanistically, CPSF perturbation resulted in lengthened 3'-untranslated regions, diminished intronic polyadenylation and widespread transcriptional readthrough, and consequently suppressed oncogenic pathways. Furthermore, we reported the development of specific CPSF3 inhibitors building upon the benzoxaborole scaffold, which exerted potent antitumor activity. Notably, CPSF3 blockade effectively exacerbated genomic instability by down-regulating DNA damage repair genes and thus acted in synergy with poly(adenosine 5'-diphosphate-ribose) polymerase inhibition. These findings establish CPSF3-dependent transcriptional termination as an exploitable driving mechanism of ovarian cancer and provide a promising class of boron-containing compounds for targeting transcription-addicted human malignancies.

Fig. 1.. Expression and dependence of CPSF subunits in ovarian cancer.

(A) Heatmap of dependency scores for putative regulators of transcriptional termination in ovarian cancer cell lines (n = 58) from the DepMap project. (B) A schematic model of core factors involved in pre-mRNA cleavage and polyadenylation. The CPSF, CstF, CFIm, and CFIIm functional modules are shown in red, purple, aqua, and green, respectively. The subunits essential for cell viability based on the results of gene knockout are highlighted. (C) CPSF1, CPSF2, CPSF3, WDR33, CPSF4, or FIP1L1 was knocked out in SKOV3, COV362, PEO1, HEY, OVCA433, and SNU-251 cells using CRISPR-Cas9 with two independent sgRNAs, and cell viability was assayed by crystal violet staining. (D) SKOV3 cells with or without CPSF subunits depletion were labeled with firefly luciferase and implanted intraperitoneally. Tumor growth in nude mice was monitored by bioluminescence imaging. (E) Ovarian cancer specimens from two patients were analyzed using single-cell RNA sequencing (RNA-seq). The t-distributed stochastic neighbor embedding plots show tumor and stroma cells and relative expression levels of CPSF1, CPSF2, CPSF3, WDR33, CPSF4, and FIP1L1. Each dot represents a single cell. (F) Representative immunohistochemical images of CPSF1, CPSF2, CPSF3, WDR33, CPSF4, and FIP1L1 staining in a tissue microarray containing 135 ovarian cancer samples. Scale bar, 1 mm. (G) Violin plots showing staining intensity of CPSF1, CPSF2, CPSF3, WDR33, CPSF4, and FIP1L1 in the ovarian cancer tissue microarray. The H score system was used for immunohistochemical quantification.

Fig. 2.. The central role of CPSF3 endonuclease in ovarian cancer.

(A) Cell cycle analysis by flow cytometry on SKOV3 cells with CPSF1, CPSF2, CPSF3, WDR33, CPSF4, or FIP1L1 knockout. (B) EdU incorporation assay on SKOV3 cells with CPSF1, CPSF2, CPSF3, WDR33, CPSF4, or FIP1L1 knockout. Cell nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; blue). Quantification on the percentage of EdU-positive (green) cells is plotted as means ± SD (n = 4). *P < 0.05, analysis of variance (ANOVA) followed by Tukey’s posttest. Scale bar, 200 μm. (C) Flow cytometric analysis of cell death using annexin V/PI double labeling in SKOV3 cells with CPSF1, CPSF2, CPSF3, WDR33, CPSF4, or FIP1L1 knockout. (D) Representative transmission electron microscopy (TEM) images of SKOV3 cells with CPSF1, CPSF2, CPSF3, WDR33, CPSF4, or FIP1L1 knockout. Scale bar, 2 μm. (E) A schematic model of core CPSF complex, which consists of mPSF and mCF subcomplex. The mPSF subcomplex, constituted by CPSF1, WDR33, CPSF4, and FIP1L1, is necessary and sufficient for PAS recognition and polyadenylation. CPSF2 and CPSF3 form the mCF subcomplex, which is responsible for catalyzing the cleavage reaction. CPSF1 and WDR33 interact with each other and act as a structural scaffold of the CPSF complex. (F) Myc-tagged CPSF1, green fluorescent protein (GFP)–tagged CPSF2, 3× FLAG-tagged CPSF3, or His-tagged WDR33 was transfected into human embryonic kidney (HEK) 293T cells. CPSF3 was immunoprecipitated, and CPSF1, CPSF2, CPSF3, or WDR33 was analyzed by immunoblotting. IP, immunoprecipitation; GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (G) CPSF1, CPSF2, CPSF3, WDR33, CPSF4, or FIP1L1 was knocked out in SKOV3, COV362, and OVCA420 cells using CRISPR-Cas9 with two independent sgRNAs, and the indicated proteins were analyzed by immunoblotting. (H) CPSF3 was overexpressed in SKOV3, COV362, and OVCA420 cells with CPSF1, CPSF2, CPSF3, or WDR33 knockout, and the indicated proteins were analyzed by immunoblotting. Cell viability was assayed by crystal violet staining.

Fig. 3.. Altered pre-mRNA 3′-end processing and gene expression upon CPSF depletion.

(A) Scatter plots of percentage of distal poly(A) site usage index (PDUI) values reflecting the relative distal-to-proximal PAS usage in SKOV3 cells with CPSF1, CPSF2, CPSF3, or WDR33 knockout. Transcripts with lengthened or shortened 3′UTR upon gene knockout in comparison to control cells are shown in red or blue, respectively. (B) Scatter plots of IPUI intronic poly(A) site usage index (IPUI) values reflecting the relative IpA usage in SKOV3 cells with CPSF1, CPSF2, CPSF3, or WDR33 knockout. Transcripts with increased or decreased IpA usage upon gene knockout in comparison to control cells are shown in red or blue, respectively. (C) Scatter plots of readthrough counts in SKOV3 cells with CPSF1, CPSF2, CPSF3, or WDR33 knockout. Transcripts with significantly increased or decreased readthrough counts upon gene knockout in comparison to control cells are shown in red or blue, respectively. (D) Volcano plots showing differentially expressed genes in SKOV3 cells with CPSF1, CPSF2, CPSF3, or WDR33 knockout. Transcripts with significant up-regulation or down-regulation upon gene knockout in comparison to control cells are shown in red or blue, respectively. (E) GSEA plots showing disproportionate down-regulation of genes involved in cell metabolism, organelle biogenesis, and DNA repair in SKOV3 cells with CPSF1, CPSF2, CPSF3, or WDR33 knockout. (F) GSEA plots showing disproportionate down-regulation of PAX8 gene signature in SKOV3 cells with CPSF1, CPSF2, CPSF3, or WDR33 knockout.

Fig. 4.. Attenuated DDR upon CPSF depletion.

(A) GSEA plots showing disproportional down-regulation of a 276-member gene set related to DDR defined by TCGA in SKOV3 cells with CPSF1, CPSF2, CPSF3, or WDR33 knockout. (B to E) Boxplots showing comparison of gene sizes, coding sequence (CDS) sizes, exon numbers, and PAS numbers between significantly down-regulated and nondifferentially expressed genes in SKOV3 cells with CPSF1, CPSF2, CPSF3, or WDR33 knockout. The gene sizes, coding sequence sizes, exon numbers, and PAS numbers were also compared between 276 DDR genes defined by TCGA and other genes. The difference between each two groups is statistically significant (P < 0.001, two-tailed Wilcoxon test). (F) Immunoblotting analysis of the indicated proteins in SKOV3 and COV362 cells with CPSF1, CPSF2, CPSF3, or WDR33 knockout. Glyceraldehyde-3-phosphate dehydrogenase was used as the loading control. (G) Immunofluorescence staining of γH2AX (green) in SKOV3 and COV362 cells with CPSF1, CPSF2, CPSF3, or WDR33 knockout. Cell nuclei were counterstained with DAPI (blue). Quantification on the percentage of cells containing more than five γH2AX foci is plotted as means ± SD (n = 3). *P < 0.05, ANOVA followed by Tukey’s posttest. Scale bars, 20 μm. (H) Representative images and quantitative analysis of comet assay in SKOV3 and COV362 cells with CPSF1, CPSF2, CPSF3, or WDR33 knockout. Tail moment = tail DNA % × tail moment length. *P < 0.05, ANOVA followed by Tukey’s posttest. Scale bars, 20 μm.

Fig. 5.. Discovery of benzoxaborole-based CPSF3 inhibitors.

(A) Molecular docking analysis of 115 (green), HQY426 (yellow), HQY436 (orange), or XHJ1049 (purple) against the active site of CPSF3 (Protein Data Bank ID: 6M8Q). In the predicted binding mode, 115, HQY426, HQY436, or XHJ1049 adopted a conformation very close to that of JTE-607 (blue) in complex with CPSF3 from their cocrystal structure. (B) Chemical structures of 115, HQY426, HQY436, XHJ1049, and HQY429. Cell viability was assayed by crystal violet staining in SKOV3 and COV362 cells treated with indicated compounds. (C) SKOV3 and COV362 cells were treated with dimethyl sulfoxide (DMSO), 115, HQY426, HQY436, XHJ1049, or HQY429 and subjected to cellular thermal shift assay. CPSF3 was analyzed by immunoblotting. (D) SKOV3 cells were treated with HQY436, HQY54, or XHJ1167 (0.3 μM), followed by ultraviolet irradiation and click chemistry conjugation with biotin-azide. Proteins captured with streptavidin beads were analyzed by immunoblotting. Competitive photocrosslinking assay was performed by pretreating SKOV3 cells with excessive unlabeled compounds (10 μM) before XHJ1167 (0.3 μM) was added. (E) CPSF3 protein was purified and incubated with indicated compounds. A fluorescent probe was added, and the fluorescence polarization was recorded. Dose-response curves were constructed to determine the median inhibitory concentration values as measures of binding affinity. (F) Wild-type or mutant CPSF3 was overexpressed in SKOV3 cells with or without endogenous CPSF3 knockout. Cells were treated with various concentrations of 115, HQY426, HQY436, or XHJ1049 for 72 hours, and cell viability was measured by the Cell Count Kit-8 (CCK-8) assay. (G) Volcano plots showing differentially expressed genes in SKOV3 cells treated with 115, HQY426, HQY436, or XHJ1049 for 6 hours. (H) GSEA plots showing disproportionate down-regulation of CPSF3 gene signature in SKOV3 cells treated with 115, HQY426, HQY436, or XHJ1049 for 6 hours.

Fig. 6.. Potent antitumor efficacy of CPSF3 inhibitors.

(A) Cell viability was measured by the CCK-8 assay in a panel of ovarian cancer cell lines treated with various concentrations of HQY426 or HQY436 for 72 hours. (B) SKOV3 cells were treated with HQY426 or HQY436 at various concentrations, and cell viability was determined by crystal violet staining and phase-contrast microscopy. (C) Cell cycle analysis by flow cytometry on SKOV3 cells treated with HQY426 or HQY436 for 24 and 48 hours. Values are expressed as percentage of the cell population in the G₁, S, or G₂-M phase of cell cycle. (D) Cell apoptosis induced by HQY426 or HQY436 treatment in a time course was documented by immunoblotting analysis of cleaved PARP and cleaved caspase 3/7. (E) Representative images of SKOV3 xenografts from BALB/c nude mice treated with HQY426 (10 mg/kg per day) or vehicle control [5% ethanol (EtOH), 5% DMSO, and 40% polyethylene glycol, poly(ethylene glycol) 400 (PEG-400)] for 21 days. The three largest tumors shown for each mouse were dissected from distinct abdominal sites. Scale bar, 5 mm. (F) Quantification of SKOV3 tumor weight in the vehicle control and HQY426-treated groups. The tumor weight for each mouse was calculated by adding the weights of all resectable implants. *P < 0.05, unpaired Student’s t test. (G) Body weight measurements of BALB/c nude mice during HQY426 (10 mg/kg per day) or vehicle (5% EtOH, 5% DMSO, and 40% PEG-400) treatment. (H) Representative images of hematoxylin and eosin (H&E) and immunohistochemistry staining for Ki67, cleaved caspase 3, or cleaved caspase 7 in SKOV3 tumor slices. Scale bar, 50 μm.

Fig. 7.. Synergistic effects of CPSF3 and PARP inhibition in ovarian cancer.

(A) GSEA plots showing disproportional down-regulation of a 276-member gene set related to DDR defined by TCGA in SKOV3 cells treated with HQY426 or HQY436. (B) Radar plots showing relative enrichment of different DDR pathways in SKOV3 cells treated with HQY426 or HQY436. (C) Immunofluorescence staining of γH2AX (green) in SKOV3 cells treated with HQY426 or HQY436 for 24 or 48 hours. Quantification on the percentage of cells containing more than five γH2AX foci is plotted as means ± SD (n = 3). *P < 0.05, ANOVA followed by Tukey’s posttest. Scale bars, 20 μm. (D) Immunoblotting analysis of γH2AX in SKOV3 cells treated with HQY426 or HQY436 in a time course manner. (E) Representative images and quantitative analysis of comet assay in SKOV3 cells treated with HQY426 or HQY436 for indicated time. Tail moment = tail DNA % × tail moment length. *P < 0.05, ANOVA followed by Tukey’s posttest. Scale bar, 20 μm. (F) Immunofluorescence staining of γH2AX (green) in SKOV3 cells treated with indicated inhibitors for 24 or 48 hours. Scale bars, 20 μm. (G) Crystal violet staining of dose response assays and heatmaps of bliss synergy scores demonstrated synergistic activities of HQY426 or HQY436 and olaparib in SKOV3 cells. (H) Representative images of SKOV3 xenografts from BALB/c nude mice treated with vehicle control, HQY426 (10 mg/kg per day) or olaparib (50 mg/kg per day), and quantification of SKOV3 tumor weight in indicated groups. Scale bar, 5 mm. The tumor weight for each mouse was calculated by adding the weights of all resectable implants. *P < 0.05, ANOVA followed by Tukey’s posttest. (I) Body weight measurements of BALB/c nude mice during indicated treatment.