An in-tumor genetic screen reveals that the BET bromodomain protein, BRD4, is a potential therapeutic target in ovarian carcinoma

Abstract

High-grade serous ovarian carcinoma (HGSOC) is the most common and aggressive form of epithelial ovarian cancer, for which few targeted therapies exist. To search for new therapeutic target proteins, we performed an in vivo shRNA screen using an established human HGSOC cell line growing either subcutaneously or intraperitoneally in immunocompromised mice. We identified genes previously implicated in ovarian cancer such as AURKA1, ERBB3, CDK2, and mTOR, as well as several novel candidates including BRD4, VRK1, and GALK2. We confirmed, using both genetic and pharmacologic approaches, that the activity of BRD4, an epigenetic transcription modulator, is necessary for proliferation/survival of both an established human ovarian cancer cell line (OVCAR8) and a subset of primary serous ovarian cancer cell strains (DFs). Among the DFs tested, the strains sensitive to BRD4 inhibition revealed elevated expression of either MYCN or c-MYC, with MYCN expression correlating closely with JQ1 sensitivity. Accordingly, primary human xenografts derived from high-MYCN or c-MYC strains exhibited sensitivity to BRD4 inhibition. These data suggest that BRD4 inhibition represents a new therapeutic approach for MYC-overexpressing HGSOCs.

Keywords: BRD4; MYCN; in vivo screen; ovarian cancer; targeted therapy.

Fig. 1.

shRNA-based in vivo screen. (A) Experimental scheme of an in vivo shRNA screen. (B and C) Comparison of shRNA fold change values (FC) obtained from different segments of the pilot screen (∼2k shRNA). IP, solid tumors obtained by injecting tumor cells intraperitoneally; ascites, ascitic cells obtained by injecting the cells intraperitoneally; SQ, tumors obtained by injecting the cells subcutaneously. IP and SQ tumors developed in separate animals of the same NSG strain (D) Distribution of shRNA z-scores obtained from the full subcutaneous screen (4xshRNA lentiviral library pools of ∼2k each). The blue dotted line represents the distribution of the CONTROL (CTRL) shRNAs scores. Scores were calculated as described in SI Materials and Methods. (E) Scatter plot representing the 40 identified candidate target genes. Each gene is represented by a dot in the plot that reflects the first and second lowest shRNA-associated z-scores (best and second best shRNA, respectively). Highlighted in red are the genes PLK1 and ERB3, which were routinely used as positive controls in OV8. In yellow are the previously reported candidate therapeutic targets AURKA, CDK2, and FRAP1(mTOR). In green is the BET bromodomain factor BRD4. (F) Table showing the 40 candidate target genes and for each, the second lowest z-score obtained in the screen (second best shRNA z-score). (G) Matrix showing the representative categories of Gene Ontology biological process (GOBP) that were significantly enriched in the pool of candidate target genes with respect to the pool containing all tested genes. Also shown are the genes associated with each category. (P values and GO subcategories are shown in Table S2).

Fig. 2.

Candidate target gene evaluation in OV8. (A) qRT-PCR analysis of OV8 cells after infection (MOI > 1) with each of the second best shRNAs and 48-h selection with Puromycin. For each gene, expression values are normalized by reference to that of the housekeeping gene, Tubulin, and are reported as a percentage of the levels obtained when infecting the cells with the control lentivirus, shLuc. An additional control lentivirus shCTRL is also reported. (B) Cell proliferation assay. OV8 cells were infected at MOI > 1 and selected for 48 h with Puromycin. OV8 cells used in this assay express luciferase. Cell number was titered by Luciferase expression immediately after selection (day 0) and 4 d later (day 4). For each shRNA is reported the fold increase in Luciferase signal over 4 d of culture. Where reported, error bars represent standard deviations of triplicate measurements. (C) GFP competition assay outline. (D) GFP competition assay results.

Fig. 3.

BRD4 depletion or inhibition impairs OV8 proliferation. (A) List of BRD4 targeting shRNAs analyzed in the screen and selected for further testing. For each shRNA is reported the fold change value (FC) and the z-score obtained in the screen. (B) GFP competition assays with multiple shRNAs directed against BRD4 (shBRD4). (C and D) OV8 proliferation curves in the presence of increasing concentrations of the active BRD4 inhibitor JQ1-S (C) or the inactive enantiomer JQ1-R (D). Cell viability was measured by Cell Titer Glow assay (i.e., for ATP concentration) and for each time point is reported as the percent of the value obtained on day 0. (E–G) OV8 cells were infected with each of the indicated shRNAs at MOI > 1, selected for 48 h with Puromycin and plated for further experiments. shCTRL and shLuc are control lentiviruses. shA and shB are two additional control shRNAs that affect proliferation but not BRD4 levels. (E) Western blot analysis of BRD4 expression in OV8 (−) and OV8 infected with each of the indicated shRNAs. (F) JQ1 dose–response curves in OV8 (−) and OV8 infected with each of the indicated shRNAs. Cell viability was measured by the Cell Titer Glow assay, and each value was reported as a percentage of the effect obtained by using the vehicle alone. (G) Comparison of the EC₅₀ values (micromolar) obtained from the JQ1 dose–response curves performed in the various conditions. 95% confidence intervals (95% CI) are also shown. Where reported, error bars represent standard deviations of triplicate measurements.

Fig. 4.

Sensitivity of various primary ovarian cancer cell strains (DF) to JQ1 and its relationship to c-MYC or MYCN overexpression. (A) JQ1 dose–response curves performed on 20 different primary serous ovarian cancer cell strains (DFs). Dose–response curve data were integrated by nonlinear regression. DFs were clustered in five groups on the basis of their JQ1 maximum effect (E_max), which, in turn, corresponds to the minimum measured viability value. (B–D) qRT-PCR analysis of BRD4 (B), MYCN (C), and c-MYC (D) in DF strains. Expression values are normalized by reference to that of the housekeeping gene, 36B4, and are reported as a percentage of the levels obtained in the JQ1-refractory strain, DF59. (E) Western blot analysis of a subset of DF strains. Red, JQ1 highly sensitive (group 5); orange, group 4; black: JQ1-refractory (group1). (F) Anticorrelation of c-MYC and MYCN expression in DF strains and definition of ”c-MYC“ and ”MYCN“ strains on the basis of their relative c-MYC and MYCN mRNA levels expressed as z-scores. (G) Pearson correlation of MYCN expression and JQ1 E_max in the MYCN DF population. (H) Pearson correlation of c-MYC expression and JQ1 maximal effect (E_max) in the c-MYC DF population. (I and J) qRT-PCR analysis of c-MYC and MYCN expression after 24-h exposure to 1 μM JQ1 (+) or vehicle a (−) in the MYCN-high strain DF37 (I) and the c-MYC-high strain DF149 (J). Expression values are normalized by comparison with that of 36B4 and reported for each gene as a percentage of the levels obtained with vehicle alone. (K) Western blot analysis in selected DF strains of c-MYC and MYCN expression after 48-h exposure to 1 μM JQ1 (+), the inactive JQ1 enantiomer (R), or vehicle alone (−). DF14 and DF37 represent MYCN-high strains, DF86 and DF149 represent c-MYC-high strains, and DF181 is a JQ1-refractory MYCN-low/ c-MYC-low strain. Where reported, error bars represent standard deviations of triplicate measurements.

Fig. 5.

BRD4 inhibition has antitumor effects in PDX xenografts derived from MYCN-high and c-MYC-high primary ovarian cancer strains. (A–C) JQ1 antitumor activity was evaluated in three luciferase-producing ovarian PDX models. Mice bearing xenografts derived from primary DF14 (MYCN-high) (A), DF86 (c-MYC-high) (B), or the low N and c-MYC DF181 (MYC-low) (C) were treated with vehicle (−) or JQ1 (+JQ1) (50 mg/kg once a day, intraperitoneally) for the indicated times starting on day 7 after implantation. Tumor growth was measured by weekly bioluminescence (BLI). Statistical significance of the results was evaluated using a two-way ANOVA test. Error bars represent standard deviations of measurements obtained for each group (n = 10).