Evolutionary dynamics and molecular features of intra-tumor heterogeneity

Abstract

The systematic assessment of intra-tumor heterogeneity is still limited and often unfeasible. In silico investigations of large tumor cohorts can be used to decipher how multiple clones emerge and organize into complex architectures. Here, we addressed this challenge by integrating mathematical modeling of cancer evolution with algorithmic inference of clonal phylogenies in 2,600 human tumors from 15 tumor types. Through numerical simulations, we could discriminate between observable and hidden intra-tumor heterogeneity, the latter characterized by clones that are missed by DNA sequencing of human samples. To overcome this limitation in human tumors, we show that population frequencies of detectable clones can be used to estimate the extent of hidden heterogeneity. Overall, simulated and human clonal architectures were highly concordant and showed that high numbers of clones invariably emerge through branching lineages. Interestingly, high numbers of alterations were not necessarily associated with high intra-tumor heterogeneity. Indeed, tumors with alterations in proliferation-associated genes exhibited high numbers of clonal mutations, but few clones. Instead, mutations of chromatin remodeling genes characterized tumors with high numbers of subclonal alterations and multiple clones. Our results identify evolutionary and genetic determinants of tumor clonal architectures to guide functional investigations of intra-tumor heterogeneity.