We use single-cell derived xenografts to study the expansion of tumours in vivo and performed whole-exome sequencing to determine the underlying evolutionary dynamics.

Next-generation sequencing (NGS) has been used to infer tumor ancestries and mutational timing. However, the inference of subclonal architecture from NGS data without reference to the dynamics of tumor growth may lead to an erroneous interpretation of clonal relationships. Here we analyze multi-region sequencing data from primary colorectal tumors and an in vivo colorectal tumor growth model and demonstrate that clusters of mutations with the same mutant allele frequency (MAF) do not necessarily belong to the same sub-clone. Rather, the characteristic bimodal and symmetrical MAF distribution reflects tumor growth dynamics. Through simulation studies we demonstrate that the observed MAF distribution is compatible with both the null model of neutral evolution and weak selection, but not strong selection. These data indicate that mutational clusters do not capture clonal identity and challenge a general phenomenon of sequential clonal expansions within an established tumor, with significant implications for earlier detection, prognostication, and therapeutic intervention.