RNA splicing-driven transcriptomic changes reveal PARP inhibitor resistance mechanism in ovarian cancer patients

Abstract

PARP inhibitors (PARPi), crucial for the treatment of ovarian and breast cancers, target deficient DNA repair efficiency in cancer cells by hindering the PARP protein responsible for single-strand break repair. This leads to an accumulation of such breaks, evolving into double-strand breaks and ultimately causing cell death in those with defective DNA repair function. The effectiveness of PARPi has prompted research into predictive markers for treatment response, primarily focusing on genomic data like mutations and genomic scars. However, these genetic markers, being essentially 'scars', fail to capture functional restoration, limiting their ability to predict long-term outcomes, including recurrences. In this study, a discovery cohort with pre-post treatment matched ovarian cancer patients (n=40) was analyzed for transcriptomics changes associated with PARPi treatment. Significant differentially used transcripts (DUTs), which are made by alternative splicing, were discovered both in patient groups with acquired and innate resistance. By observing the usage patterns of these DUTs, this study suggests a resistance mechanism through DNA repair functionality restoration. Building upon these DUTs, a machine learning model was constructed and validated for predicting PARPi response in an independent validation cohort (n=130), demonstrating the potential for dynamic biomarkers in assessing treatment efficacy.