Computational drug discovery for castration-resistant prostate cancers through in vitro drug response modeling

Abstract

Prostate cancer (PC) is the most frequently diagnosed malignancy and a leading cause of cancer deaths in US men. Many PC cases metastasize and develop resistance to systemic hormonal therapy, a stage known as castration-resistant prostate cancer (CRPC). Therefore, there is an urgent need to develop effective therapeutic strategies for CRPC. Traditional drug discovery pipelines require significant time and capital input, which highlights a need for novel methods to evaluate the repositioning potential of existing drugs. Here, we present a computational framework to predict drug sensitivities of clinical CRPC tumors to various existing compounds and identify treatment options with high potential for clinical impact. We applied this method to a CRPC patient cohort and nominated drugs to combat resistance to hormonal therapies including abiraterone and enzalutamide. The utility of this method was demonstrated by nomination of multiple drugs that are currently undergoing clinical trials for CRPC. Additionally, this method identified the tetracycline derivative COL-3, for which we validated higher efficacy in an isogenic cell line model of enzalutamide-resistant vs. enzalutamide-sensitive CRPC. In enzalutamide-resistant CRPC cells, COL-3 displayed higher activity for inhibiting cell growth and migration, and for inducing G1-phase cell cycle arrest and apoptosis. Collectively, these findings demonstrate the utility of a computational framework for independent validation of drugs being tested in CRPC clinical trials, and for nominating drugs with enhanced biological activity in models of enzalutamide-resistant CRPC. The efficiency of this method relative to traditional drug development approaches indicates a high potential for accelerating drug development for CRPC.

Keywords: castration-resistant prostate cancer; drug discovery; drug repurpose; drug response prediction; enzalutamide.

Fig. 1.

Overview of our computational framework for drug discovery in CRPC. The overall design for drug discovery consists of two steps: (1) prediction of patient drug sensitivities and (2) identification of efficacious drugs. In (1), ridge regression models are constructed based on CCL transcriptomics and measured drug sensitivities in CCL. Trained parameters are integrated with patient transcriptomics to predict patient drug response. In (2), predicted drug sensitivities are examined with patient clinical features to identify clinically meaningful drugs.

Fig. 2.

Second-generation ADT-resistant CRPC patients showed higher sensitivities to COL-3. Based on our method, COL-3 was predicted to be more efficacious among patients who were more resistant to ADTs including enzalutamide and abiraterone. The sensitivity of COL-3 was summarized in AUC, of which a higher value indicates higher resistance. (A) The coefficient and hazard ratio of COL-3 from the cox proportional model. The negative coefficient indicates that patients who had worse survival showed higher predicted sensitivity to COL-3. (B) The relationship between patient response to COL-3 and survival time. A clear trend supported our finding that COL-3 was more efficacious among patients with worse OS.

Fig. 3.

COL-3 inhibits cell growth, migration, invasion; induced cell apoptosis and G1 cycle arrest in CRPC cell line models. (A) Half maximal inhibitory concentrations (IC₅₀s) of growth inhibition of the two cell lines induced by COL-3. Cell growth inhibition was measured by WST-1 assay after 0.5 to 2 μM COL-3 treatment for 72 h. IC₅₀ values were derived from three independent replicates for each cell line. (B) Half maximal effective concentrations (EC₅₀s) of COL-3 induced apoptosis between the two cell lines. Cell apoptosis was quantified by CellEvent™ Caspase-3/7 Green Detection Reagent after 0.5 to 2.8 μM COL-3 treatment for 72 h. EC₅₀ values were derived from three independent replicates for each cell line (^*P < 0.05; ^**P < 0.01). (C) G1 cell cycle arrest assay of the two cell lines treated with COL-3. Cell cycle arrest quantified by FUCCI assay after 1 μM COL-3 exposure. Proportions of cells arrested at G1 were measured longitudinally every 4 h. Each point represents the average rate of G1-phase arrest of each measurement, and error bars indicate standard deviations. (D) Cell migration assay following 1 μM COL-3 treatment. Both cell lines showed inhibited migration activities, whereas R1-D567 cells were more profoundly suppressed by COL-3 (adjusted Tukey’s HSD P-value < 0.001 when compared to R1-AD1 cells). (E) Invasion assay following 1 μM COL-3 treatment. Cells were exposed to 1 μM COL-3 and monitored longitudinally (adjusted Tukey’s HSD family-wise P-value > 0.05 when comparing invasion inhibition in R1-D567 and R1-AD1 cells). (A–D) COL-3 exhibited higher efficacy in ADT-resistant R1-D567 CRPC cells compared to R1-AD1.

Fig. 4.

Differential expression analysis of MMP pathway genes between R1-AD1 and R1-D567 cell lines. We show a heatmap depicting log-FC normalized expression profiles of AR and MMP genes between R1-AD1 and R1-D567, each with three replicating microarray assays. Each row represents a gene in either the AR or MMP pathway; each column represents a microarray of the corresponding cell line (^* indicates an adjusted P-value or FDR < 0.05).

Fig. 5.

COL-3 treatment’s effect on androgen receptor (AR) expression in R1-AD1 and R1-D567 cells. (A) mRNA expression of AR in R1-AD1 and R1-D567 cells following 12-h COL-3 treatments. Cells were treated with increasing concentrations of COL-3 (0, 1, 2, 4, and 8 μM). RNAs were then extracted, reverse transcribed, and analyzed by qPCR for AR mRNA expression. Data were normalized using GAPDH as the housekeeping gene (^*P < 0.05; ^***P < 0.001). (B) Immunoblotting of AR expression following COL-3 treatment. R1-AD1 and R1-D567 cells were treated with COL-3 with incrementing doses (0, 1, 2, 4, and 8 μM) for 12 h. Protein lysates were harvested and immunoblotted with an anti-N terminal AR antibody using tubulin as the loading control. (C) Quantified immunoblotting results showing relative normalized AR protein intensity compared to no COL-3 treatment. For each technical replicate (REP), measured AR intensity was normalized against that of tubulin. Normalized intensity was then scaled relative to no COL-3 treatment. (D) Nascent mRNA expression of AR in R1-D567 cells following a 12-h, 8 μM COL-3 treatment. (E) Schematic overview of actinomycin D (ActD) treatment. R1-D567 cells were pretreated with COL-3 or DMSO for 12 h prior to the addition of ActD or DMSO. Total mRNA was extracted at 0, 3, 6, 9, and 12 h posttreatment for quantification. (F) R1-D567 cells were treated with 8 μM COL-3 and 10 µg/mL ActD and relative AR mRNA levels were quantified by qPCR at the indicated time points. Treatment with COL-3 did not affect the rate of AR mRNA decay when compared to without COL-3.

Fig. 6.

COL-3 is efficacious against docetaxel-resistant CRPC models in vitro. We established docetaxel-resistant R1-AD1 (orange) and R1-D567 (blue) cell lines and exposed them to docetaxel (A–D) and COL-3 (E–H) and monitored cellular growth rates and apoptosis. We summarized longitudinal dose–response curves from three replicates into IC50s and EC50s. (A and B) IC50 values of growth inhibition of R1-AD1 and R1-D567 induced by docetaxel (nM); C and D. EC50 values of R1-AD1 and R1-D567 apoptosis induced by docetaxel (nM). (E and F) IC50 values of growth inhibition of R1-D567 and R1-D567 post COL-3 (µM); (G and H) EC50 values of R1-AD1 and R1-D567 apoptosis induced by COL-3 (µM) (^*P < 0.05; ns: P > 0.05).