The effects of epistasis and linkage on the invasion of locally beneficial mutations and the evolution of genomic islands.

We study local adaptation of a peripheral population by investigating the fate of new mutations using a haploid two-locus two-allele continent-island migration model. We explore how linkage, epistasis, and maladaptive gene flow affect the invasion probability of weakly beneficial de-novo mutations that arise on the island at an arbitrary physical distance to a locus that already maintains a stable migration-selection polymorphism. By assuming a slightly supercritical branching process, we deduce explicit conditions on the parameters that permit a positive invasion probability and we derive approximations for it. They show how the invasion probability depends on the additive and epistatic effects of the mutant, on its linkage to the polymorphism, and on the migration rate. We use these approximations together with empirically motivated distributions of epistatic effects to analyze the influence of epistasis on the expected invasion probability if mutants are drawn randomly from such a distribution and occur at a random physical distance to the existing polymorphism. We find that the invasion probability generally increases as the epistasis parameter increases or the migration rate decreases, but not necessarily as the recombination rate decreases. Finally, we shed light on the size of emerging genomic islands of divergence by exploring the size of the chromosomal neighborhood of the already established polymorphism in which 50% or 90% of the successfully invading mutations become established. These 'window sizes' always decrease in a reverse sigmoidal way with stronger migration and typically increase with increasing epistatic effect.