Spatial structure governs the mode of tumour evolution

Abstract

Characterizing the mode -the way, manner, or pattern -of evolution in tumours is important for clinical forecasting and optimizing cancer treatment. DNA sequencing studies have inferred various modes, including branching, punctuated and neutral evolution, but it is unclear why a particular pattern predominates in any given tumour [1]. Here we propose that differences in tumour architecture alone can explain the variety of observed patterns. We examine this hypothesis using spatially explicit population genetic models and demonstrate that, within biologically relevant parameter ranges, tumours are expected to exhibit diverse evolutionary modes including four archetypal "oncoevotypes": rapid clonal expansion (predicted in leukaemia); progressive diversification (in colorectal adenomas and early-stage colorectal carcinomas); branching evolution (in invasive glandular tumours); and almost neutral evolution (in certain non-glandular and poorly differentiated solid tumours). We thus provide a simple, mechanistic explanation for a wide range of empirical observations. Oncoevotypes are driven by differences in cell dispersal and cell-cell interactions, which we show are essential for accurately characterizing, forecasting and controlling tumour evolution.A tumour is a product of somatic evolution in which interacting processes of mutation, selection, genetic drift, and cell dispersal generate a patchwork of cell subpopulations (clones) with varying degrees of aggressiveness and treatment sensitivity [2]. A primary goal of modern cancer research is to characterize this evolutionary process, so as to enable precise, patient-specific prognostic forecasts and to optimize targeted therapy regimens.Much progress has been made in revealing the evolutionary features of particular cancers, yet these findings raise as many questions as they answer. Why do different tumour types exhibit different modes of evolution [1,3,4,5,6]? What conditions sustain the frequently observed pattern of branching evolution, in which several clones diverge and evolve in parallel [1,7]? And why do some pan-cancer analyses indicate that many tumours evolve neutrally [8], whereas others support extensive selection [9]?Factors proposed as contributing to intra-tumour evolution include microenvironmental heterogeneity, niche construction, and positive ecological interactions between clones [10, 2]. However, because such factors have not been well characterized across human cancer types, it remains unclear how they might relate to evolutionary modes.On the other hand, it is well established that tumours exhibit a wide range of architectures. In leukaemia, for example, mutated stem cells in semi-solid bone marrow produce cancer cells that mix and proliferate in the bloodstream (Figure 1a). Colorectal adenomas and early-stage colorectal carcinomas, by contrast, inherit 1

A unified simulation model for understanding the diversity of cancer evolution

“…Our current version of the model completely ignores spatial information, which potentially influences evolutionary dynamics. Recently reported studies have shown that spatial structures regulate evolutionary dynamics in tumors (Noble et al, 2019;West et al, 2019). We also determined that resource bias prompts the driver-branching process, by simulating tumor growth on a one-dimensional lattice (Niida, Hasegawa & Miyano, 2019).…”

Peer Review #1 of "A unified simulation model for understanding the diversity of cancer evolution (v0.1)"

“…Our current version of the model completely ignores spatial information, which potentially influences evolutionary dynamics. Recently reported studies have shown that spatial structures regulate evolutionary dynamics in tumors (Noble et al, 2019;West et al, 2019). We also determined that resource bias prompts the driver-branching process, by simulating tumor growth on a one-dimensional lattice (Niida et al, 2019).…”

ClonArch: Visualizing the Spatial Clonal Architecture of Tumors

“…Although C A successfully visualizes the spatial clonal architecture assuming a large number of biopsies, we asked the question: can this tool be applicable to smaller datasets? Applying a simulation that mimics the invasive glandular model of tumor growth (Noble et al, 2019), we generated trees and frequency matrices at grid sizes of 3 × 3 and 5 × 5 ( Fig. 6a-b).…”