| description |
Background Evolutionary divergence is common within bacterial species and populations, even during a single bacterial infection. We use large-scale genomic and phenotypic analysis to identify the extent of diversification in controlled experimental populations and apply this data to decide between several theoretical models of evolutionary divergence. Results We defined testable differences between five proposed models and used experimental evolution studies to follow eight glucose-limited E. coli chemostat populations at two growth rates. Simple phenotypic tests identified 11 phenotype combinations evolving under glucose limitation. Each evolved population exhibited 3 to 5 different combinations of the 11 phenotypic clusters. Genome sequencing of a representative of each phenotypic cluster from each population identified 193 mutations in 48 isolates. Only two of the 48 strains had evolved identically. Convergent paths to the same phenotype occurred, but two pleiotropic mutations were unique to slow-growing bacteria, permitting them greater phenotypic variance. Indeed, greater diversity arose in slower-growing, more stressed cultures. Mutation accumulation, hypermutator presence and fitness mechanisms varied between and within populations, with the evolved fitness considerably more uniform with fast growth cultures. Negative frequency-dependent fitness was shown by a subset |