Affiliations: 1) University of Washington, Seattle, WA; 2) Fred Hutchinson Cancer Research Center, Seattle, WA; 3) Howard Hughes Medical Institute; 4) Equal contribution
Keywords: Phylogenetics, Macroevolution, and Biogeography
Spatial properties of tumor growth have profound implications for cancer progression, therapeutic resistance and metastasis, yet how space governs tumor cell division remains an open question. Xenograft and organoid studies suggest that tumors grow preferentially on the periphery ( i.e.,“boundary-driven growth”), while sequencing efforts have suggested faster progression in the tumor interior. Boundary-driven growth affects the shape of tumor phylogenies and is therefore theoretically observable from multi-region sequencing data. However, phylodynamic methods have been largely under-utilized to infer growth dynamics in clinical tumors. Here, we show that boundary-driven growth can be well-approximated by a two-state model permitting different growth rates in the tumor edge and center. To quantify these differential growth rates from sequencing data, we develop a State-Dependent Evolutionary phylodynamic model (SDevo), which links growth rate to tree branching and clock rates as a function of state (here, the edge or center position of the sample). We validate this approach on simulated tumors sampled across multiple spatial regions and demonstrate its ability to quantify spatially-varying growth rates under a range of growth conditions and sampling strategies. We then apply SDevo to multi-region sequencing data from hepatocellular carcinomas and find evidence that these tumors divide more rapidly near the tumor edge than in the center. As multi-region and single-cell sequencing increases in resolution and availability, this approach could interrogate spatial growth dynamics in diverse clinically-resected specimens and be extended to test other two-state growth models, e.g. metastasis or driver gene effects. More generally, this approach demonstrates the potential power of phylodynamic models to quantify tumor evolutionary dynamics.