Multi-omic Characterization of Pre- and Post-neoadjuvant Chemotherapy Treated Ovarian Cancer Reveals Mediators of Tumorigenesis and Chemotherapy Response

Abstract

High-grade serous ovarian cancer (HGSC) accounts for more than 200,000 deaths each year. Despite recent advances in treating HGSC with neoadjuvant chemotherapy, the majority of patients ultimately develop chemotherapy resistance. HGSC is characterized by TP53 mutations and widespread copy number alterations and occurs frequently in the setting of deleterious germline BRCA1/2 variations, but many cases lack putative driver mutations. Here, we performed whole-exome, whole-genome, and whole transcriptome sequencing along with mass spectrometry to characterize the molecular landscape of HGSC in 22 paired samples obtained before and after neoadjuvant chemotherapy. Responsiveness to chemotherapy was determined for each patient. Evidence at the DNA, RNA, and protein level revealed numerous defects in cell-cell and cell-matrix interactions as well as disruption of cell polarity and cytoskeletal regulation in HGSC, indicating that defects in epithelial integrity were present in the majority of HGSC patients. Non-responsive HGSC harbored subclones with putative survival mutations. Additionally, ineffective nonsense mediated decay resulted in the persistence of transcripts with frameshift mutations that were translated into aberrant proteins detectable in HGSC samples. Together, these findings suggest that HGSC may arise through defects in maintenance of epithelial integrity that lead to shedding of malignant cells throughout the peritoneum, and the presence of resistant subclones prior to chemotherapy may decrease the chemosensitivity of patients.

Figure 1:. Patients with significant enrichment in CompBio generated themes

CompBio generated themes are depicted as rows with each column being an individual patient. The top (green) triangle or bottom (purple) triangles are filled if there is a significant enrichment of mutation in that row’s pathway before or after NACT respectively.

Figure 2:. Changes in different IDR ensemble parameters due to mutations in HGSC cohort

Five measurements of IDR ensemble are predicted from their sequence. Any mutation results in a change to that parameter and is plotted below. Most mutations have a modest effect on a given parameter. Mutations that caused a change greater than 90% of all mutations in one or more parameter were included in downstream analysis.

Figure 3:

A) MAF before and after NACT in NACT-NR (left) and NACT-R (right) patients. All nonsynonymous, frameshift, or nonsense mutations are included for all samples. Mutations that underwent a greater than twofold change in MAF between timepoints are depicted in red with all others in grey. B) Boxplots of MAF in pre-NACT samples. There were significantly more mutations with low MAF present in pre-NACT NR samples compared to pre-NACT R samples, consistent with increased clonal heterogeneity in these samples. C) Cumulative distribution plots of permutation analysis Mutations that increased in frequency in NACT-NR patients after chemotherapy are significantly enriched in COSMIC CGC genes. Curves are empirically generated cumulative distribution functions from 100,000 random samples. The observed number of CGC genes in each sample is shown in red with its percentile relative to the random samples.

Figure 4:

A) Volcano plots of pairwise comparisons between NACT-NR, NACT-R and pre-NACT post-NACT. Genes that were significantly differentially expressed as described in the text in two or more of the pairwise comparisons are labelled. B) Venn Diagram of DEGs present in 2 or more of the pairwise comparisons

Figure 5:

A) Percentage of mutant reads in the exome (x-axis) and transcriptome (y-axis) for nonsense and frameshift mutants. The x- and y- axes represent the number of reads with the alternate allele divided by the total number of reads at that position in the exome and transcriptome respectively. Filled circles are mutations that meet all NMD criteria while empty circles do not meet at least one. Many nonsense mutation-containing transcripts are depleted regardless of whether they meet NMD criteria. Conversely, transcripts with frameshift mutations are often expressed in both NMD-triggering and NMD-escape contexts.

Figure 6:

Heatmap of protein abundance for proteins with significantly different abundance across timepoints and response phenotype