Synonymous Genetic Variation in Natural Isolates of Escherichia coli Does Not Predict Where Synonymous Substitutions Occur in a Long-Term Experiment

Abstract

Synonymous genetic differences vary by more than 20-fold among genes in natural isolates of Escherichia coli. One hypothesis to explain this heterogeneity is that genes with high levels of synonymous variation mutate at higher rates than genes with low synonymous variation. If so, then one would expect to observe similar mutational patterns in evolution experiments. In fact, however, the pattern of synonymous substitutions in a long-term evolution experiment with E. coli does not support this hypothesis. In particular, the extent of synonymous variation across genes in that experiment does not reflect the variation observed in natural isolates of E. coli. Instead, gene length alone predicts with high accuracy the prevalence of synonymous changes in the experimental populations. We hypothesize that patterns of synonymous variation in natural E. coli populations are instead caused by differences across genomic regions in their effective population size that, in turn, reflect different histories of recombination, horizontal gene transfer, selection, and population structure.

Keywords: effective population size; experimental evolution; genetic variation; mutation rate.

Fig. 1.

The expected time to coalescence for individuals from an evolving haploid population is N_e generations. Tick marks show neutral mutation events along two lineages, which occur at some rate µ per generation. The expected number of mutations separating Individuals 1 and 2 is 2N_eµ. If all genes in the genome have experienced the same N_e, then significant variation among genes in the per-site rate of accumulation of neutral mutations would imply gene-specific heterogeneity in the underlying mutation rate.

Fig. 2.

Synonymous substitutions observed in experimental populations of Escherichia coli closely match the null hypothesis of a uniform point-mutation rate across genes, but not the distribution expected if the variability in θ_s across genes in natural isolates is explained by gene-specific differences in the point-mutation rate. Each observed or hypothetical series shows the cumulative proportion of 1,069 synonymous substitutions in 2,834 genes that have been sorted and ranked by their θ_s values (i.e., the synonymous nucleotide diversity seen in natural isolates for each gene). The red line shows the observed distribution of synonymous mutations in 12 independently evolved genomes after 40,000 generations. The dashed curve shows the null hypothesis of a uniform point-mutation rate, where gene length alone predicts the occurrence of synonymous changes. The dotted curve shows the alternative hypothesis where each gene’s point-mutation rate is proportional to θ_s.

Fig. 3.

Hypermutator clones have distinctive spectra of synonymous mutations in addition to elevated mutation rates. Clones with defective mutS or mutL genes (Ara−2, Ara−3, Ara−4, Ara+3) have large numbers of C:G to T:A and A:T to G:C transitions, whereas clones with defects in mutT (Ara−1, Ara+6) have large numbers of A:T to C:G transversions.

Fig. 4.

Synonymous substitutions tend to be found in longer genes. Genes with at least one synonymous substitution after 40,000 generations (green) are on average 1,296 bp long, whereas those without any synonymous substitutions (purple) are on average only 850 bp long. The bin width is 50 bp.