Modeling the consequences of heterogeneity in microbial population dynamics

ABSTRACT Chapter 1 The rate at which unicellular micro-organisms progress through the cell cycle is a major component of their evolutionary fitness. Measuring fitness phenotypes in a given environment or genetic background forms the basis of most quantitative assays of drug sensitivity or genetic interaction, including genome-wide assays. Growth rate is typically measured in bulk cell populations, inoculated with anything from hundreds to millions of cells sampled from purified, isogenic colonies. High-throughput microscopy reveals that striking levels of growth rate heterogeneity arise between isogenic cell lineages (Levy et al ., 2012). Using published Saccharomyces cerevisiae data, I examine the implications for interpreting bulk, population scale growth rate observations, given observed levels of growth rate heterogeneity at the lineage level. I demonstrate that selection between cell lineages with a range of growth rates can give rise to an apparent lag phase at the population level, even in the absence of evidence for a lag phase at the lineage level. My simulations further predict that, given observed levels of heterogeneity, final populations should be dominated by one or a few lineages. Chapter 2 In order to validate and further explore the conclusions from Chapter 1, I re-analyzed high-throughput microscopy experiments carried out on Quantitative Fitness Analysis (QFA) S. cerevisiae cultures (Addinall et al ., 2011), an approach referred to as μ QFA. To allow for precise observation of purely clonal lineages including very fast-growing lineages and non-dividing cells, I re-designed an existing image analysis tool for μ QFA, now available as an open source Python package. Fast-growing outliers in particular influence the extent of the lag phase apparent at the population level, making the precision of growth rate estimation a key ingredient for successfully simulating population observations. μ QFA data include population observations which I used to validate the population simulations generated from individual lineage data. I explored various options for modeling lineage growth curves and for carrying out growth rate parameter inference, and included the full workflow in an open source R package. Contact [email protected]