Abstract PR03: Pan-cancer analysis of the etiology and consequences of intra-tumor heterogeneity

Abstract

Abstract Tumors are typically mosaics of mutant clones that have evolved from a common ancestral cell1–3. This intra-tumor heterogeneity (ITH) is thought to drive both neoplastic progression and acquired therapeutic resistance4–6. Availability of just one sample per tumor and moderate sequencing depth have limited systematic analysis of ITH during previous TCGA pan-cancer analyses7–9, confining the study of ITH to a small numbers of tumor samples and cancer types10–13. The molecular and histopathological causes of ITH and its prognostic significance have thus far remained uncertain14–17. To overcome these limitations, we used an analysis method called EXPANDS18 that estimates the proportion of cells harboring specific mutations from exome sequencing data, as well as other methods that quantify ITH15,19. We extrapolate the number of genetically diverse clonal subpopulations in 1,165 primary tumors among 12 different cancer types from TCGA and investigate mechanisms underpinning ITH as well as the correlation of ITH and genomic instability with prognosis. Lastly we validate the prognostic significance of genomic instability in an independent, high-density SNP-array dataset consisting of 2010 tumor samples, across seven additional cancer types. We found evidence of ITH in the vast majority of tumors. Driver gene mutations, prevalence of copy number gains and tumor microenvironment composition were significantly associated with increased ITH. Mutations in driver genes tended to have a characteristic clone size, suggesting differential fitness effects of those mutations. The detection of a mutation in a driver gene that typically appears in a small clone was a predictor of poor survival. In general, ITH was a universal biomarker of prognosis: across cancer types poor prognosis was associated with intermediate, rather than very low or high, levels of ITH. This was also true for the fraction of tumor genome affected by copy number alterations16,20: tumors with intermediate copy number burden (50 to 75% of the genome affected by copy number alterations) progressed faster than tumors with higher copy number burden, independent of cancer type. Chemo-radiation therapy administration was more efficient in decelerating tumor progression among patients with intermediate copy number burden than among patients with low or high copy number burden. These results were validated and confirmed in the independent SNP-array dataset. This study suggests a tradeoff exists between the costs and benefits of genomic instability21,22 that impacts both the evolvability and fitness of the tumor cell population. In the future, this tradeoff might be exploited to improve survival. In summary, we have shown that ITH is a universal feature of human cancers that predicts survival.