A key challenge in biology is to understand how mutations combine to alter phenotypes. Each genetic variant in a genome can have diverse effects, for example decreasing, increasing, inactivating, or changing the function of a protein or RNA. In contrast, systematic analyses of how mutations interact have typically used a single variant of each gene, most often a null allele. We therefore lack an understanding of how the full range of genetic variants that occur in individuals can interact. To address this shortcoming, we developed an approach to combine >5000 pairs of diverse mutations in a model regulatory network. The outcome of most mutation combinations could be accurately predicted by simple rules that capture the 'stereotypical' genetic interactions (epistasis) in the network. However, for individual genotypes, additional, unexpected pairwise and higher order genetic interactions can be important. These include 'harmonious' combinations of individually detrimental alleles that reconstitute alternative functional switches. Our results provide an overview of how the full spectra of possible mutations in genes interact and how these interactions can be predicted. Moreover, they illustrate the importance of rare genetic interactions for individuals, including the impact of higher order epistatic interactions that dramatically alter the consequences of inactivating genes.
Main textHuman genomes contain millions of genetic variants. Each of these variants can have diverse effects, for example quantitatively increasing, decreasing or changing the activity of individual genes 1 .Understanding and predicting how the particular combination of variants present in each individual affects molecular processes and phenotypic traits is a fundamental challenge for human genetics and evolutionary biology 2 . To date, however, systematic analyses of how mutations in different genes combine to influence phenotypes have used only one or a few mutations in each gene, most often an inactivating null allele 3,4 . A more complete understanding of how mutations combine in individuals will require functionally diverse mutations in individual genes to be combined in large numbers 5 .The GAL regulatory (GALR) system from yeast is a promising model to begin such studies because it is mechanistically well-understood, has relatively few molecular players, and is an important model of gene network function and evolution 6,7 ( Fig 1A). This network is required for sensing the sugar galactose and then inducing transport (via Gal2) and the Leloir pathway proteins Gal1p, Gal7p, and Gal10p necessary to metabolize this sugar as a carbon source for growth. The core of this network consists of three regulatory genes and their protein products: GAL4, a transcriptional activator; GAL80, its repressor; and GAL3, which acts as a GAL sensor by inhibiting Gal80p as an activated Gal3p-Galactose-ATP complex 8 ( Fig 1A). The GAL1 locus, encoding the first Leloir enzyme galactokinase (GALK) Gal1p, is a paralog of GAL3 that can also effect galactose sensing....