Genomic evolution of cancer models: Perils and opportunities

Precision cancer medicine is based on the idea that therapeutic response of cancer patients could be predicted by the molecular features of their tumors. Cancer models are commonly used to identify such genotype-phenotype associations. Recently, cancer models have also been used as tumor “avatars” in an attempt to predict response of specific patients to anticancer drugs. However, to optimize the use of cancer models in research and in the clinic, we must understand how faithful their genomic features are to those of their tumors-of-origin, how heterogeneous they are, and how they evolve over time.

We combined experimental and computational approaches to address these questions in the most common cancer models. First, we studied tumor evolution in breast cancer mouse models, identified driver-specific trajectories of evolution, and uncovered specific genes that promote recurrent trajectories (Ben-David et al. Nature Communications 2016). Next, we analyzed tumors from patient-derived xenografts (PDXs) and revealed divergent tumor evolution when human tumors are transplanted into mice (Ben-David et al. Nature Genetics 2017). Finally, we analyzed cancer cell lines and characterized how their genomic evolution altered their transcriptional programs and drug response. This work exposed the extent, the cellular origins and the functional consequences of cell line heterogeneity (Ben-David et al. Nature 2018). Together, these studies shed light on the perils and opportunities of the naturally-occurring variation within cancer models (Ben-David et al. Nature Reviews Cancer 2019).

Here, I will discuss emerging themes from these studies, describe ongoing follow-up work, and highlight the implications for precision cancer medicine.