Pharmacogenomic analysis of patient-derived tumor cells in gynecologic cancers

Abstract

Background: Gynecologic malignancy is one of the leading causes of mortality in female adults worldwide. Comprehensive genomic analysis has revealed a list of molecular aberrations that are essential to tumorigenesis, progression, and metastasis of gynecologic tumors. However, targeting such alterations has frequently led to treatment failures due to underlying genomic complexity and simultaneous activation of various tumor cell survival pathway molecules. A compilation of molecular characterization of tumors with pharmacological drug response is the next step toward clinical application of patient-tailored treatment regimens.

Results: Toward this goal, we establish a library of 139 gynecologic tumors including epithelial ovarian cancers (EOCs), cervical, endometrial tumors, and uterine sarcomas that are genomically and/or pharmacologically annotated and explore dynamic pharmacogenomic associations against 37 molecularly targeted drugs. We discover lineage-specific drug sensitivities based on subcategorization of gynecologic tumors and identify TP53 mutation as a molecular determinant that elicits therapeutic response to poly (ADP-Ribose) polymerase (PARP) inhibitor. We further identify transcriptome expression of inhibitor of DNA biding 2 (ID2) as a potential predictive biomarker for treatment response to olaparib.

Conclusions: Together, our results demonstrate the potential utility of rapid drug screening combined with genomic profiling for precision treatment of gynecologic cancers.

Keywords: Gynecologic malignancy; ID2; PARP inhibitor; Pharmacogenomic analysis; TP53 mutations.

Fig. 1

Pharmacogenomic analyses of gynecologic malignancies. a Schematic representation of pharmacogenomic analyses in gynecologic tumor-derived PDCs. Genomic and transcriptomics data were analyzed to identify single nucleotide variations and small indels and gene expression profiles. Short-term cultured PDCs were subjected to drug sensitivity screening against 37 molecular targeted compounds. b Mutational landscape of gynecologic tumors including ovarian cancer, endometrial cancer, cervical cancer, and uterine sarcoma. All mutations with an allele frequency of > 5% and depth of > 20 reads are shown. c Three-dimensional bubble plot demonstrating the frequency of non-synonymous cancer-driver mutations exclusively in tissue (black, left axis), PDC (blue, right axis), or shared between the two (gray, upper axis) (upper panel). The position of each dot or mutation is located on the quadrant based on its shared or private rate between primary tumor tissues and matched PDCs, and the distance reflects the number of cases that harbor respective mutation. Comparison of mRNA expression profiles between tumor tissue specimens and matched PDCs (bottom panel). Pearson’s correlation between tissue and PDCs is demonstrated as a heatmap

Fig. 2

Lineage-specific drug sensitivity among gynecologic tumors. a Volcano plot representation of gynecologic tumor type-specific drug response, with fold-change drug comparison (x-axis) and its significance (y-axis). Each circle represents a single tumor type-drug interaction, and the size is proportional to the cohort size of the respective tumor. b Heatmap representation of plot A. Only significant associations have been marked based on sensitivity (red) or resistance (blue). Drugs have been clustered based on their known target classes. c Violin plots demonstrating the pathway enrichment scores of each corresponding pathway. The activity scores were measured using single sample Gene Set Enrichment Analysis (ssGSEA). Horizontal lines within the violin plots represent 0.25, 0.50, and 0.75 quantiles. P values in a–c: two-sided Wilcoxon’s rank-sum test

Fig. 3

Pharmacogenomic landscape of epithelial ovarian cancer. a Mutational landscape of epithelial ovarian cancers. All mutations with an allele frequency of > 5% and depth of > 20 reads are shown. Genomic variations, including single nucleotide variants (SNVs), frameshift insertions/deletions, in-frame insertions/deletions, and non-sense mutations, are shown. Frequency of each genomic alteration within the whole cohort is shown on the left column. b Network-based enrichment map analysis of gene set enrichment results. Gene sets are organized as a network, where each gene set is a node and edges represent genes overlapping between the sets. Related gene sets are laid out as network clusters. c Volcano plot representation of ovarian cancer type-specific drug response, with fold-change drug comparison (x-axis) and its significance (y-axis). Each circle represents a single tumor type-drug interaction, and the size is proportional to the cohort size of respective tumor. d Drug response assessments of VEGFR (left panel) and PI3K-AKT-mTOR (left panel) inhibitors in serous and clear cell carcinomas. Box plots span from the first to third quartiles, and the whiskers represent the 1.5 interquartile range. e, f In vivo drug response assessments of cediranib (e) and PI3K inhibitor (f) in serous and clear cell carcinomas, respectively. Violin plots represent the overall tumor weights of the PDX models from respective groups. Horizontal lines within the violin plots represent 0.25, 0.50, and 0.75 quantiles. P values in c–f: two-sided Wilcoxon’s rank-sum test

Fig. 4

Predictive biomarkers for response to PARP inhibitors. a Volcano plot representation of gene-drug interactions in gynecologic tumors. b Waterfall plot enumerating individual tumor response to olaparib with BRCA1/2 and TP53 mutation status. c Receiver operating characteristic curve plotted for the sensitivity versus 100 - specificity values for predicting olaparib response rates using BRCA1/2 and TP53 mutation status. d Drug response assessment of olaparib on OVISE cell-line that has been stably transduced with/without TP53-R249S, T273H, or R175H mutation. Dose response curves were generated using percent survival of cells under olaparib treatment for 4 days on 9 different doses from 200 to 0.78 μM. P values in a—two-sided Wilcoxon’s rank-sum test, and in c—two-sided binomial exact test

Fig. 5

Transcriptomic correlates of olaparib sensitivity. a Gene Set Enrichment Analysis (GSEA) between olaparib-sensitive and olaparib-resistant PDCs. b Drug response assessment of olaparib and/or saracatinib. c Heatmap representation of SRC pathway encoding gene expressions in olaparib-sensitive and olaparib-resistant PDCs. d A scatter plot demonstrating linear correlation between olaparib AUC and ID2 expression. The correlation coefficient and the P value were obtained using Pearson’s correlation test. e Representative immunohistochemical images of ID2 staining in patient tumor specimens. Scale bars, 50 μm. f The Kaplan-Meier treatment-free survival analysis of patients with high vs. low ID2 protein expression levels. P values in b—two-tailed t test, in d—Pearson’s correlation test, and in f—log-rank test