compensatory genes will remain in linkage disequilibrium for extensive periods, no matter how far apart they lie on the chromosome. As the frequency of the compensatory mutants increases in the population, the intensity of selection for better alleles at the deleterious locus will drop off. Moreover, if the compensatory mutations become fixed, adaptive valleys may be established and effectively preclude the ascent of even more fit mutations at the deleterious locus itself (Schrag et al., 1997).
Although the total sizes of bacterial populations may be enormous, their genetically effective sizes can be quite low as a consequence of periodic selection, i.e., selective sweeps of better-adapted mutants (Atwood et al., 1951; Koch, 1974; Levin, 1981) or as a result of bottlenecks, contractions in population size associated with transmission to new hosts or microhabitats. These low effective population sizes and selective sweeps have at least two major ramifications for adaptive evolution. One is to purge variation accumulating in the population by selection as well as by genetic drift. Another is to make a given population less fit than it would be if it did not have to deal with the stochastic trials and tribulations of periodic selection and bottlenecks. Although, on Equilibrium Day, the most fit genotypes will ascend to their rightful places, on the days before, those best types may well be lost as the population passes through bottlenecks (Gerrish, 1998; Levin et al., 2000).
Because low rates of recombination allow modifier loci and the genes whose effects they modify to remain together for extensive periods, natural selection can favor modifying traits that are advantageous in the long run despite their short-term disadvantages (Eshel, 1973a, b). One example of such a situation is the evolution of mutators, genes that increase the rate at which variation is generated by mutation. Because they generate deleterious mutations and possibly for other reasons as well, mutator genes, which are commonly defective mismatch repair loci, are anticipated to be at a selective disadvantage. However, continuous changes or heterogeneities in the selective environment can cause bacteria bearing these mutators to ascend by hitchhiking with the beneficial mutants they generate (Taddei et al., 1997; Tenaillon et al., 1999). As anticipated by this theory, in long-term experimental (de Visser et al., 1999; Sniegowski et al., 1997) and natural populations of bacteria (Oliver, 1999) adapting to new and/or changing environments, one can find relatively high frequencies of mutator genes.