We study how cell populations cope with and adapt to environmental change.
We study how cell populations cope with and adapt to environmental change.
To investigate this question, we take several parallel approaches that rely on the continuing development of new high-throughput technologies. One approach assumes little phenotypic variation within an isogenic population, attempting to understand how cellular networks in a typical cell change across environments. To do this, we are developing an interaction sequencing (iSeq) platform that will allow us to assay millions of genetic or physical interactions in parallel using next-generation sequencing. Future studies will attempt to understand how newly arising or segregating genetic variation impacts the dynamics of cellular networks across environments. A second approach is to study how cell-to-cell variation within a population, either genetic or epigenetic, impacts the fitness of the population as a whole. We use newly developed lineage tracking technologies to measure fitness distributions of populations, and to determine how these fitnesses change across environments, at the levels of both a single lineage and the population. These data will be used to investigate the genetic architecture underlying cellular robustness and bet hedging. In a third approach, we collect millions of trajectories of small lineages during long-term evolutions to quantitatively study the dynamics of adaptation. In collaboration with theorists, we use these data to investigate long-standing evolutionary questions such as the adaptive mutation rate and the distribution of fitness effects. Future studies using this powerful approach will focus on: 1) deleveloping the technology in other organisms such as pathogenic yeast and cancer cell models adapting to chemotherapies, 2) understanding how sex impacts evolutionary dynamics, and 3) understanding how environmental dynamics shape evolutionary dynamics.
Ensemble measurements of cell populations describe the average cell phenotype at a particular time but miss the full extent of phenotypes that exist within a population. This is problematic because the behavior of a population is often dependent on the character of the extreme phenotypes rather than the average. For example, the long-term fitness of an evolving population depends most heavily on rare cells that acquire a beneficial mutation. One major technological advance has been the development of single-cell assays that measure phenotypic distributions within a population. Yet, even these do not tell the whole story because a particular cell phenotype must be tied to a fitness in order to understand how it impacts the population. We attempt to address this gap by developing high-throughput technologies that simultaneously observe many cell lineages over time. By studying cell lineages, we can tie a cell phenotype to cell fitness and a distribution of cell phenotypes to population fitness.
Using a high-throughput microscopy platform and homemade computer vision software, we have developed an assay capable of observing the growth of about 100,000 yeast microcolonies simultaneously. The time-lapse movie above shows a single field from one of these experiments. On the left is a bright-field image and on the right is the output from our algorithm. We can measure the area of each colony over time to estimate lag and growth rates. In combination with fluorescent reporters (not shown), we can measure the abundance of proteins in each microcolony. We are using this assay to study how growth variation in yeast populations could be acting as a bet-hedging mechanism across environmental fluctuations. We would eventually like to adapt this technology to mammalian cell lines to investigate if cellular heterogeneity drives malignancy.
To track lineages over longer timescales, we have developed a set of genetic tools by which to stably insert millions of unique random barcodes at a specific genomic location in yeast. By sequencing only the barcodes in pools of competing cells, we can measure the relative frequency of each barcode lineage in the population and follow lineage trajectories over time.
A movie some of lineage trajectories from a batch evolution experiment of 109 cells. Red, orange and blue lineages have acquired an established beneficial mutation of different fitness effects, while grey lineages have not. In collaboration with theorists, we analyze lineage trajectories to quantitatively measure population parameters. For example, the slope of each adaptive lineage provides its fitness advantage and all lineages trajectories together provide the distribution of beneficial fitness effects.
A schematic of the double barcode system. Upon mating, two barcodes are brought together to the same chromosome location via loxP recombination. The recombination event completes a URA3 selectable marker that is interupted by an artificial intron containing the double barcode. Mating and recombination can be done in cell pools and double barcodes are sequenced from these pools en masse to assay the relative frequency of each genotype in the pool.
We are also developing tools for a double barcode system by which barcodes from two haploid cells are brought to the same genomic location upon mating and can be sequenced together in one next-generation sequencing read. This allows us to track lineages through mating events and to study, at high throughput and resolution, how combinations of genotypes or alleles impact fitness.
One application of the double barcode system that we are pursuing is iSeq or interaction sequencing. Here, double barcodes are used in combination with selectable elements that mark either gene deletions or protein constructs from a protein fragment complementation assay. By mating pools of cells and selecting haploids that contain both double barcodes and these selectable elements, we are able to perform massively parallel genetic or protein-protein interaction assays. Pilot experiments are currently underway. We plan to use this system to study, on a global scale, how genetic and protein-protein interaction networks change across environments or genotypes.
The pairwise yeast genetic interaction network (Costanzo et al., 2010, Science, 327: 425) has taken years to generate with liquid handling robots. This massive undertaking presents a static view of the cell: only a single genotype in only a single environment. Yet, cellular networks are in constant motion as environments and genotypes change. We strive to study these global network dynamics using our iSeq double barcode system. Double barcodes will allow us to capture interactions faster and cheaper than previous methods, making these studies technically feasible for the first time.
Quantitative Evolutionary Dynamics
We are using our random barcode system to study the quantitative properties of evolutionary dynamics such as the beneficial mutation rate, the distribution of beneficial fitness effects, the fitness trade-offs of beneficial mutations, and the extent of epistasis between beneficial mutations. In collaboration with theorists, our long-term goal is to use these measures to derive a predictive theory of adaptation. One area we are currently pursuing is an attempt to understand how environmental complexity impacts the speed of evolution.
Two quantitative measures: the establishment times and fitness effects of beneficial mutations that naturally enter a population of 109 cells. Many small-effect mutations establish early in the evolution. But once large-effect mutations become abundant later in the evolution, small-effect mutations are unable to establish.
Quantitative Trait Locus (QTL) studies are frequently used in an attempt to describe the genetic elements underlying quantitative traits, such as fitness. One major caveat of this method is that the contribution of any genetic locus to fitness is averaged across all other segregating variation. That is, we only observe the average fitness effect of a locus, when it might be the extreme fitness of that locus (its fitness contribution in one particular genotype) that is most important to evolution and disease. Furthermore, we lack any understanding of how much any given locus contributes to fitness of either the average or extreme genotypes. We will be using our random barcode and double barcode systems to attempt to fill these gaps in our understanding. Our aim is to measure fitness distributions across genotype space when one or more loci are held constant. This will tell us both the average and extreme fitness contributions of trait loci.
Robustness and Bet Hedging
Genetically identical cells grown in the same culture display striking cell-to-cell heterogeneity in gene expression and other traits. A crucial challenge is to understand how much of this heterogeneity reflects the noise tolerance of a robust system and how much serves a biological function. We have used our high-throughput microscopy assay to show that clonal populations grown in a benign environment display broad distributions of growth rates and that slow growth predicts resistance to heat killing in a probabilistic manner (link). These results suggest that yeast populations are hedging their bets to maximize long-term fitness in an unpredictably changing environment. In a benign environment, most cells adopt the sensible strategy of fast growth, whereas a small proportion of cells adopt the high-risk strategy of slow growth, which could reap large benefits should the environment change to one of high heat. We are currently interested in exploring how environmental dynamics, for example the frequency of fluctuations from a benign environment to high heat, impacts growth rate and stress resistance in populations. Our aim is to explore the tension between the benefits of having a robust phenotype in the current environment and the benefits of preparing for an environmental shift via bet hedging.