Abstract SY05-03: Dissecting genomic correlates of response and resistance to chemotherapy in bladder cancer through clinical computational oncology

Approximately 20,000 new cases of muscle-invasive bladder cancer (MIBC), a localized but potentially lethal stage of disease, are diagnosed annually in the US alone. Standard-of-care therapy for MIBC includes neoadjuvant cisplatin-based chemotherapy followed by definitive bladder resection. In prior work, we identified and validated genomic alterations in DNA repair genes such as ERCC2, which predict response to cisplatin-based chemotherapy (Van Allen et al. Cancer Discovery 2014; Liu et al. JAMA Oncology 2016). However, the majority of patients have disease resistant to chemotherapy with a poor prognosis of 50x, and a tumor purity minimum threshold of 10%, we had data from 30 trios available for analysis. We hypothesized that DNA-damaging chemotherapy may lead to increased mutational load in the post-treatment tumor. However, we observed that while some tumors gained mutations, others lost mutations, with no overall change (mean change = -17.3 mutations, paired t-test p = 0.20) in total mutational load. We found that clonal mutations (found in all tumor cells) were virtually unchanged from matched pre- to post-treatment tumors. In contrast, subclonal mutations (found only in a subset of tumor cells) were private to pre- and post-treatment tumors. These pre- and post-treatment mutation differences may reflect tumor sampling heterogeneity (i.e. taking from different parts of the tumor), but may be also due to selection pressure from therapy (e.g. loss of subclones) and cisplatin-induced mutations.To investigate the latter possibility, we adapted a non-negative matrix factorization (NMF) approach (Lee and Seung Nature 1999) to discover mutational signatures (Alexandrov et al Nature 2013) in the mutations unique to post-treatment tumors. Along with signatures known to be operant in bladder cancers, we discovered a mutational signature dissimilar to any other previously described mutational signature which accounted for ~15% of post-treatment mutations. This signature exhibited a transcriptional strand bias consistent with known mechanisms of cisplatin-induced DNA damage and repair, and was enriched in subclonal mutations consistent with the relatively short time frame between cisplatin exposure and cystectomy. This signature also exhibited similar activity to a cisplatin-induced mutational signature derived in a preclinical model (DT40) exposed to cisplatin therapy (Pearson rho = 0.95, empiric p = 0.004). Finally, we were able to validate this signature in a separate cohort of pre- and post-cisplatin treated bladder cancers (Faltas et al Nature Genetics 2016). We further hypothesized that the degree of tumor heterogeneity itself may be a prognostic factor. We calculated two different measures of intratumor heterogeneity: (1) the proportion of mutations in each tumor that was subclonal; and (2) the number of unique subclones in each tumor, and examined the association of survival with these measures of intratumor heterogeneity using Cox survival analyses. We found that overall survival was associated with heterogeneity, with a 6.6% increase in mortality rate for each 10% increase in post-treatment proportion of subclonal mutations (p=0.013), and 64% increase in mortality rate for each additional inferred subclone (p=0.02). Tumor heterogeneity continued to be associated with survival after adjusting for clinical covariates (p=0.03, p=0.014, respectively).Finally, we analyzed our tumors for genomic alterations associated with resistance. While we did not discover highly recurrent post-treatment mutations in specific genes, we found drivers of cell cycle progression (E2F3 amplification, JUN amplification), biallelic loss of FBXW7 (regulator of protein degradation of multiple onco-proteins including c-MYC, Notch, Cyclin E, and c-JUN), and focal amplification of PD-L1/2 in individual post-treatment resistant tumors.In this study, we found that cisplatin-based chemotherapy did not induce a large increase in the number of mutations. Thus, while there is good empiric data for the efficacy of combination of chemotherapy and immune checkpoint inhibition in specific tumor type and clinical settings (e.g. platinum-doublet therapy + ICB in first-line therapy of non-small cell lung cancer (NSCLC)), our data suggests that alternative mechanisms other than increased neoantigen burden are responsible. We discovered a cisplatin-induced mutational signature in post-treatment tumors which has subsequently been found in other cisplatin-treated tumors (e.g. NSCLC and ovarian cancer). Interestingly, the proportion of mutations inferred to be cisplatin-induced was quite different between resistant tumors, and an area for further inquiry is whether these differences could be associated with different mechanisms of resistance (e.g. upregulation of efflux pumps vs. anti-apoptotic adaptations). Tumor heterogeneity, which has been associated with worse outcomes and resistance in multiple contexts, was prognostic for survival in our cohort, suggesting that this may be clinically useful as part of a prognostic biomarker. We discovered additional association of drivers of cell-cycle progression with resistance, and further identified acquisition of a focal amplification in a region containing PD-L1/PD-L2, suggesting a potential biomarker for a subset of bladder cancers for response to immune checkpoint blockade. Broadly, this study represents the development of algorithms to dissect genomic features associated with survival and resistance in a carefully curated cohort of matched patient tumors within a specific clinical context. These types of approaches can be applied across tumor types, therapies, and clinical contexts to shed light onto biological mechanisms underpinning response and resistance and inform the development of biomarkers to guide clinical management. Citation Format: David Liu, Philip Abbosh, Daniel Keliher, Brendan Reardon, Diana Miao, Kent Mouw, Amaro Weiner-Taylor, Stephanie Wankowicz, Garam Han, Min-Yuen Teo, Catharine Cipolla, Jaegil Kim, Gopa Iyer, Hikmat Al-Ahmadie, Essel Dulaimi, David Y.T. Chen, R. Katherine Alpaugh, Jean Hoffman-Censits, Levi A. Garraway, Gad Getz, Scott L. Carter, Joaquim Bellmunt, Elizabeth Plimack, Jonathan E. Rosenberg, Eliezer M. Van Allen. Dissecting genomic correlates of response and resistance to chemotherapy in bladder cancer through clinical computational oncology [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr SY05-03.