Compensatory mutation occurs when a loss of fitness caused by a deleterious mutation is restored by its epistatic interaction with a second mutation at a different site in the genome. How many different compensatory mutations can act on a given deleterious mutation? Although this quantity is fundamentally important to understanding the evolutionary consequence of mutation and the genetic complexity of adaptation, it remains poorly understood. To determine the shape of the statistical distribution for the number of compensatory mutations per deleterious mutation, we have performed a maximum-likelihood analysis of experimental data collected from the suppressor mutation literature. Suppressor mutations are used widely to assess protein interactions and are under certain conditions equivalent to compensatory mutations. By comparing the maximum likelihood of a variety of candidate distribution functions, we established that an L-shaped gamma distribution (␣ ϭ 0.564, ϭ 21.01) is the most successful at explaining the collected data. This distribution predicts an average of 11.8 compensatory mutations per deleterious mutation. Furthermore, the success of the L-shaped gamma distribution is robust to variation in mutation rates among sites. We have detected significant differences among viral, prokaryotic, and eukaryotic data subsets in the number of compensatory mutations and also in the proportion of compensatory mutations that are intragenic. This is the first attempt to characterize the overall diversity of compensatory mutations, identifying a consistent and accurate prior distribution of compensatory mutation diversity for theoretical evolutionary models. T HE number of compensatory mutations that will afpesticide, can often be reduced by compensatory mutafect a specific deleterious mutation is a fundamentions (Schrag and Perrot 1996; Maisnier-Patin and tal evolutionary quantity of which little is known. A Andersson 2004). Finally, the number of compensatory mutation is compensatory if it has a beneficial effect on mutations informs us about the ruggedness of the genofitness that is conditional on the presence of a deleteritype-to-fitness landscape (Wilke et al. 2003). ous mutation at a different site in the genome (Kimura Because we expect the number of different compensa-1985). Hence, a compensatory effect is the outcome of tory mutations (C) to vary among deleterious mutations a strong epistatic interaction between two mutations. in different genes and organisms, our aim is to characEach compensatory mutation represents an alternate geterize the shape of a statistical distribution of C that netic solution to adaptation; thus, fitness recovery from represents the overall diversity of compensatory mutathe accumulation of deleterious mutations becomes less tion abundance. Presently, almost no direct empirical likely to occur by back mutation when compensatory information on compensatory mutations can be applied mutations are increasingly common (Whitlock and to this problem. Fortunately, experimental screen...

Genetic load is the reduction in the mean fitness of a population relative to a population composed entirely of individuals having optimal genotypes. Load can be caused by recurrent deleterious mutations, genetic drift, recombination affecting epistatically favourable gene combinations, or other genetic processes. Genetic load potentially can cause the mean fitness of a population to be greatly reduced relative to populations without sources of less fit genotypes. Mutation load can be difficult or impossible to measure. Many species have mutation rates low enough that substantial genetic load is not expected, but for others, such as humans, the mutation rate may be great enough that load can be substantial. In extremely small populations, drift load, caused by the fixation by drift of weakly deleterious mutations, can threaten the probability of persistence of the population. Migration from other populations adapted to different local conditions can bring in locally maladapted alleles, resulting in migration load. Key Concepts: Genetic load is the reduction in mean fitness of a population caused by some population genetic process. Mutation load is the reduction in fitness caused by recurrent deleterious mutations. Mutation load may be as great as 95% for the human population. Drift load is the reduction in mean fitness caused by genetic drift. In extreme cases, deleterious alleles can reach a frequency of one in a population because of genetic drift. Genetic load can also be caused by recombination breaking up beneficial combinations of alleles, segregation reducing the frequency of fit heterozygotes, or migration bringing less fit alleles into a local population.

Genetic load is the reduction in the mean fitness of a population relative to a population composed entirely of individuals having optimal genotypes.