Abstract 2834: Integration of genomic and clinical data from 25,000 patients identifies molecular determinants of metastatic potential and organotropisms

Abstract

Abstract Although most cancer deaths are due to metastatic spread, we are not able to predict if, when and where a tumor will metastasize. So far, large-scale cancer sequencing efforts have either focused on primary tumors (TCGA), or have characterized the overall genomic landscape of metastases without investigating the interplay between genomic profiles and specific routes of metastatic dissemination. Here, we have investigated associations between genomic alterations and metastatic patterns using a cohort of 25,763 patients treated at Memorial Sloan Kettering with available targeted DNA sequencing data (MSK-IMPACT, 468 genes). Our cohort consists of 15,662 (60%) sequenced primaries and 10,101 (40%) sequenced metastases from 44 different tumor types. A total of 99,220 distant metastatic events were automatically extracted from ICD billing codes from electronic medical records and mapped to a reference set of 21 anatomic locations. Metastases had significantly more copy-number alterations (CNA) (inferred by fraction genome altered, FGA) than primaries in 15/44 tumor types. Metastases also exhibited significantly higher fractions of clonal mutations in 11/44 types and higher rates of whole-genome duplication in 7/44 types. Driver genes that were most often altered at higher frequencies in metastases included TP53, CDKN2A, PTEN and MYC, while most frequently altered oncogenic pathways in metastases included TP53, Cell Cycle and DNA damage repair. Importantly, we found that FGA was positively correlated with metastatic burden (defined as the number of distinct metastatic organs throughout the entire clinical history) in 19/44 tumor types. This trend, which was observed even when restricting the analysis to primary tumors, was most pronounced in prostate, breast and endometrioid adenocarcinoma. In other tumor types, including MSS colorectal and serous ovarian tumors, there was no such trend. These findings suggest that CNA burden can be predictive of metastatic potential in some tumor types. The relationship between TMB and metastatic burden was not straightforward: we observed a positive correlation in six tumor types, and a negative correlation in four tumor types. Finally, we explored associations between somatic alterations and metastatic organotropisms. For example, BRAF mutation was enriched in MSS colorectal cancer patients with brain metastasis (7% vs 12%, qval < .05), AR amplification and PTEN deletion were enriched in prostate cancer patients with bone metastasis (5% vs 21%, qval < .05 and 9% vs 19%, qval < .05) whereas ERG fusion was less frequent in patients with bone metastasis (30% vs 24%, qval < .05). In summary, we present a comprehensive, spatiotemporal map of metastasis that integrates clinical and molecular information and serves as an initial proof of concept that genomic profiling of tumors could be used to predict patterns of metastatic spread. Citation Format: Bastien Nguyen, Christopher Fong, Renzo DiNatale, Anisha Luthra, Subhiksha Nandakumar, Henry Walch, Walid K. Chatila, Arjun Rajanna, David B. Solit Solit, Michael F. Berger, Ahmet Zehir, Jianjiong Gao, Francisco Sanchez-Vega, Nikolaus Schultz. Integration of genomic and clinical data from 25,000 patients identifies molecular determinants of metastatic potential and organotropisms [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 2834.