Epistasis Analysis Goes Genome-Wide

Epistasis, a term coined by William Bateson in 1909 [1], refers to the interdependence of mutations in their phenotypic effects. Let the phenotypic value of a trait relative to that of the wild type be fA and fB for mutants A and B, respectively, and let the phenotypic value of the corresponding double mutant be fAB. Although variation exists, epistasis is usually defined by ε = fAB − fAfB and is said to be positive when ε> 0 and negative when ε< 0. Life would have been much simpler, and perhaps even boring, if epistasis were completely absent. In reality, however, epistasis abounds, rendering biology full of surprises and complexity. For instance, a commonly encountered type of epistasis is synthetic lethality, where simultaneously deleting two genes from the genome of a normal organism is lethal despite the fact that deleting each of them separately is viable. Using the notation introduced above, we can describe synthetic lethality by fA > 0, fB > 0, and fAB = 0; consequently ε< 0. A simple mechanistic explanation of synthetic lethal epistasis is that the two genes investigated are functionally similar and hence redundant. Clearly, studying epistasis helps us to understand the functional relationship between genes, which is critical to uncovering the inner workings of biological systems. Epistasis can explain why hybrids between species are typically inviable or infertile [2] and is believed to underlie the intriguing phenomenon that some human disease-causing mutations are fixed in other species with no apparent detriment [3]. Furthermore, epistasis is assumed in many evolutionary theories. For example, the mutational deterministic hypothesis of the evolution of sexual reproduction [4] and the hypothesis of reduction in mutational load by truncation selection against deleterious mutations [5] both depend on overall negative epistasis. Thus, verifying these hypotheses requires confirming the prevalence of negative epistasis. Epistasis is typically detected by demonstrating the inequality between fAB and fAfB or some consequences of this inequality. The advent of next-generation sequencing and other genomic technologies is quickly enlarging the scale of epistasis studies. Of special significance is the recent completion of the yeast genetic interaction map, which includes nearly all 36 million epistasis values for pairs of ~6,000 yeast genes estimated from the growth rates of singleand double-gene–deletion mutants [6]. Although this map provides unprecedented data of epistasis between null mutations of different genes, it offers no information on the epistasis between mutations at different sites within a gene or that between non-null mutations in different genes. Complementing the above coarse-grained epistasis map are nucleotide-resolution epistasis maps of individual genes or segments of genes [7–10]. For instance, Li et al. synthesized a yeast tRNA gene with error, creating all possible single-point mutation variants of the gene as well as tens of thousands of variants with multiple mutations [7]. They then used a high-throughput method to measure the fitness of yeast strains, each carrying a variant tRNA gene at the place of the endogenous gene, and estimated epistasis between mutations. Interestingly, negative epistasis was found to be more prevalent than positive epistasis [7]. In principle, such a map can be constructed for every gene in the yeast genome to acquire a general picture of epistasis.