Uncovering the molecular networks underlying non-genetic heterogeneity in cancer cell populations - PROJECT SUMMARY
Tumor heterogeneity is a major contributor to variable response and treatment failure in cancer patients.
Usually, heterogeneity in cancer is thought of in terms of resistance-conferring genetic mutations that pre-
exist or emerge during treatment. However, recent studies, including our own, increasingly point to non-
genetic sources of heterogeneity as critical factors in the early stages of tumor response. Non-genetic
mechanisms are known to underlie cellular processes such as stem cell differentiation and epithelial-to-
mesenchymal transitions. In bacteria, isogenic cell populations have been shown to diversify in the
absence of perturbations (e.g., drugs) into a variety of cellular phenotypes, each with differential fitness to
potential stressors. This “bet hedging” strategy increases the odds that a portion of the population will
survive a future, unknown challenge. We, and others, have recently hypothesized that cancer cells
employ a similar survival strategy to withstand the initial onslaught of anticancer drugs. So-called “drug
tolerant” cells may persist within a patient for extended periods of time before acquiring genetic resistance
mutations that lead to tumor recurrence. The objective of this proposal is to uncover the molecular factors
that control non-genetic heterogeneity in cancer cell populations using a combined computational and
experimental approach. In Aim 1, I propose to construct a detailed kinetic model of the biochemical
signaling networks that control division and death decisions in individual cancer cells. It is well established
that complex biochemical networks can give rise to multiple stable equilibrium states, known as
“attractors.” Each attractor corresponds to a cellular phenotype and can be conceptualized as a basin
within an “epigenetic landscape.” Cells can transition between phenotypes with rates dependent upon the
depths of the basins and the heights of the barriers separating them. Using a dynamical systems analysis
approach, I will mathematically solve for the epigenetic landscape of the biochemical division/death model
and quantify molecule signatures for all attractors. In Aim 2, using BRAF-mutant melanoma and EGFR-
mutant lung cancer as in vitro model systems, I will use clonal and single-cell RNA sequencing and
chromatin accessibility sequencing (ATAC-seq) to enumerate the number and molecular signatures of
non-genetic phenotypic states. I will also utilize whole-exome sequencing to establish the non-genetic
nature of the phenotypes and immunocompromised mouse models to validate model predictions.
Differences between the experimental and in silico molecular signatures will lead to model refinement and
further experimentation. Quantifying the epigenetic landscapes of cancer cells will lay the groundwork for
novel therapies based on rationally modifying the landscape to favor phenotypes with increased drug
sensitivity, an approach termed “targeted landscaping.” This would reduce the size of the drug-tolerant
pool and delay, perhaps indefinitely, the acquisition of genetic resistance mutations and tumor recurrence.
