José I. Rojas Echenique

My graduate work consists of two main projects, both concerned with the effects of epistatic interactions on the course of evolution. The first is a systematic study of the epistatic relationships between gene deletions and their compensatory mutations. The second is a new technology that makes it possible to observe the effects of epistasis on fine-scale evolutionary dynamics.

Modular epistasis and compensatory evolution

Epistatic interactions between mutations lead to historically contingent adaptive trajectories. If most epistatic interactions are fundamentally idiosyncratic, then our ability to predict the course of evolution will be limited to short time scales. On the other hand, if epistatic interactions are structured in some way, then an understanding of this structure could allow us to predict adaptation at longer time scales.

Recent research suggests that during laboratory evolution, the historical effects of epistasis are mediated through fitness effects. The emerging consensus is that this fitness mediated epistasis causes beneficial mutations to have smaller effects in more fit backgrounds. As a result, given a common pool of adaptive mutations, we can predict that populations with lower initial fitness will adapt more rapidly and that, as any population adapts, the rate of adaptation will decrease.

The idea of fitness mediated epistasis looks strange juxtaposed with the classical view of epistasis as an indicator of functional relationships. In this view, often called modular epistasis, epistasis between gene deletion mutations, for example, is caused by a functional interaction between those genes: positive epistasis can arise when the genes are both essential in maintaining the same functional module, and negative epistasis can arise when the genes are each essential for maintaining two redundant functional modules. This organization of epistasis into a hierarchy of functional modules would, in principle, allow one to predict that the adaptive trajectories of two populations with different perturbations to the same functional module will be more similar to each other than those of two populations with perturbations to different functional modules.

Of course, our ability to predict historical contingency from functional information is entirely dependent on how we assign mutations to functional modules. In this study, we used the yeast genetic interaction map to assign genes to functional modules. The yeast genetic interaction map consists of measurements of epistasis between more than a million pairs of gene deletion mutations. This data set can be used to cluster genes into functionally related modules, which are consistent with Gene Ontology annotations but contain more fine-scaled structure.

figures/knockout.svg

(A) Schematic describing the genotypes of our experimental populations and our subsequent manipulations. (B) The fitness effect of evolved mutations plotted against initial fitness of the deletion. Each point is the mean of 20 replicate population, vertical error bars are standard deviations. Colors denote functional module. (C) The fitness effect of evolved mutations in the ancestral background plotted against their effect in the background they evolved in. Each point is a single clone from one of 100 sample populations.

We evolved groups of gene deletion mutants from several tightly clustered functional modules and found that functional relatedness as revealed by genetic interaction data predicts the rate of compensatory adaptation. For example, strains with deletions of genes in the elongator complex consistently adapt more slowly than would be predicted by their initial fitness alone. This suggests that the dependence of adaptation on initial genotype is, at least partially, mediated by the functional state of those initial genotypes and that systematic genetic interaction screens capture something real about the effects of deletions on the functional state of the cell. Furthermore, we found that this was true in spite of the fact that most evolved mutations were also beneficial in the wild type background, suggesting that the effect of functionally mediated epistasis is not due entirely to the availability of functionally specific compensatory mutations.

Long-term lineage tracking reveals dynamics of second order lineage competition

In an evolving asexual population, the fate of a new mutant lineage is determined both by its intrinsic capacity to leave more offspring than other lineages and by its offspring’s potential to acquire further beneficial mutations. Thus, it’s possible that mutations could be selected for their effects on the evolvability rather than their direct effects on fitness.

In collaboration with Alex N. Nguyen Ba, I’ve developed a lineage tracking technique that allows us to directly observe the dynamics of both of these facets of lineage competition: first order selection and the second order effects of additional beneficial mutations. Our technique records the lineage history of every cell in an evolving population genetically, by the sequential addition of random DNA barcodes into a time sorted barcode array; earlier barcodes identifying more distant ancestors. Deep sequencing of this growing array allows us to infer the frequency and evolutionary history of each lineage in the population.

figures/ilt.svg

Stacked frequency trajectories for a long term lineage tracking population. Gray bars indicated the addition of new barcodes every 100 generations. Colors indicate nested lineage ancestry, where more similar colors have more recent common ancestors. (A) The entire population. (B) and (C) Sublineages highlighted in (A) scaled to total lineage frequency.

We’ve applied our technique to several controlled laboratory populations of budding yeast, We can trace lineage dynamics over a thousand generations of evolution, and directly observe second order competition between the sub-lineages of competing lineages. We’re in the process of analyzing these data to infer the fitness history of each lineage and to determine if there’s a signal of second order selection for evolvability.