Evolution. In evolution, the sum is less than its parts.

1160 PERSPECTIVES P ropagating bacteria in a lab for thousands of generations may seem tedious, or even irrelevant, to most evolutionary biologists. Nonetheless, such experiments provide an opportunity to deduce quantitative principles of evolution and directly test them in controlled environments. Combined with modern sequencing technologies , as well as theory, recent microbial experiments have suggested a critical role for genetic interactions among mutations, called epistasis, in determining the pace of evolution. Two papers in this issue, by Khan et al. on page 1193 (1) and Chou et al. (2) on page 1190, present precise experimental measurements of these epistatic interactions. Microbial evolution experiments in a simple, constant environment reveal a characteristic pattern: At fi rst, a population rapidly acquires beneficial mutations, but then adaptation progressively slows so that thousands of generations pass between subsequent benefi cial substitutions (3). Unexpected outcomes, however, can and do occur even in these simple experimental conditions. Populations evolve a dramatically elevated mutation rate (4), discover rare phe-notypic innovations (5), or diverge into distinct lineages that either coexist (6) or compete vigorously as each strain races to acquire more adaptive mutations (7). Recent theory suggests that a common cause underlies all these phenomena: the structure of epistatic interactions among mutations. Epistasis describes how the fi tness consequence of a mutation depends on the status of the rest of the genome. In one extreme example , called sign epistasis, a mutation may be benefi cial if it arises on one genetic background , but detrimental on another. Although interactions among genes may seem an obvious fact of biology, the myriad possible forms of epistasis have made it diffi cult to formulate predictive evolutionary models or to infer such interactions from empirical data. Nevertheless , epistasis is at the heart of classical theories, such as the evolution of sex (8), and also of modern concepts such as robust-ness and evolvability (a population's ability to evolve) (9). Moreover, recent theoretical work (10) suggests that the overall dynami-cal pattern of adaptation observed in long-term microbial experiments can be explained by a prevalence of what is called antagonistic epistasis, in which benefi cial mutations confer less benefi t in combination than they do individually. To quantify epistasis among beneficial mutations and to test these theoretical predictions , both Khan et al. and Chou et al. examined the initial substitutions that occurred in populations of bacteria adapting in the laboratory. The …