The impact of recombination on dN/dS within recently emerged bacterial clones

Abstract

The development of next-generation sequencing platforms is set to reveal an unprecedented level of detail on short-term molecular evolutionary processes in bacteria. Here we re-analyse genome-wide single nucleotide polymorphism (SNP) datasets for recently emerged clones of methicillin resistant Staphylococcus aureus (MRSA) and Clostridium difficile. We note a highly significant enrichment of synonymous SNPs in those genes which have been affected by recombination, i.e. those genes on mobile elements designated "non-core" (in the case of S. aureus), or those core genes which have been affected by homologous replacements (S. aureus and C. difficile). This observation suggests that the previously documented decrease in dN/dS over time in bacteria applies not only to genomes of differing levels of divergence overall, but also to horizontally acquired genes of differing levels of divergence within a single genome. We also consider the role of increased drift acting on recently emerged, highly specialised clones, and the impact of recombination on selection at linked sites. This work has implications for a wide range of genomic analyses.

Figure 1. Functional classification of genes assigned to core and non-core regions based on an adapted version of Riley's classification.

The non-core is dominated by “extrachromosomal elements”. The positions of non-core along the genome are shown in black, with well characterised mobile elements highlighted (SCCmec = resistance island (plus SCCmer); ϕSa1(TW20), ϕSa3(TW20), ϕSPβ-like(TW20) = prophage; Tn5801-like = transposon; SaPI1 = pathogenicity island; νSaβ = genomic island).

Figure 2. Neighbour-net networks constructed using Splits Tree 4.0.

Four subsets of the SNP data were used: A – synonymous and non-synonymous SNPs in the core, B – synonymous and non-synonymous SNPs present in the non-core, C – all non-synonymous SNPs, and D – all synonymous SNPs. Taxon labels were removed for the sake of clarity. Uncorrected p distances were used and the networks drawn using the equal angle method. The networks show extensive reticulation in B and D, but less in A and C. This is consistent with high rates of recombination in the non-core, relative to the core, and in synonymous SNPs as these are enriched in the non-core. The Phi test for detecting recombination, as implemented in Splits Tree, was also used on these subsets. This was highly significant for B and D (both p values<0.001) but not significant for A and C (both p values>0.1).

Figure 3. Posterior probability densities for the relative substitution rates.

Synonymous and non-synonymous SNPs were considered for the core (A) and non-core (B). One partition contained only non-synonymous SNPs (blue), whereas the other contained synonymous SNPs (grey). In the core data set (A) the synonymous partition (mean = 1.0041; 95% CI 0.9797–1.0281) is not significantly different from the non-synonymous partition (mean = 0.9918; 95% CI 0.9438–1.0407), although the former is slightly higher. In contrast, for the non-core data set (B) the relative rate for synonymous partition (mean = 1.0401; 95% CI 1.0159–1.0635) is significantly higher than the non-synonymous partition (mean = 0.9199; 95% CI 0.8731–0.9681).

Figure 4. The time dependence of dN/dS in the core and non-core genome.

We calculated this ratio over all 1953 pairwise strain comparisons separately for the two sets of genes. We also calculated, for each pairwise comparison, the number of synonymous SNPs per synonymous site as a measure of divergence. We divided the results into bins of 50 pairwise comparisons according to increasing divergence (i.e. the first bin corresponding to the 50 least diverged comparisons). For each bin we plotted the average dN/dS against the average divergence, with standard errors calculated form all 50 data points (main panel). The data for the core is shown in red, the non-core in blue. To clarify the plot for the core data, which is far less diverged than the non-core, we re-scaled the figure (bottom left) to include only the five most conserved non-core bins. The trajectory for the non-core data fits closely to a power law (R² = 0.96) as shown by the linear relationship when both axes are log-transformed (bottom right).

Figure 5. The mean dN/dS of core orthologues plotted against dS.

Three genome sequences were compared in a pairwise fashion to the TW20 (ST239) reference. These genomes represented different levels of divergence, and the analysis confirms that the dN/dS within core genes decreases over time. Error bars represent the standard error from the re-sampled data (see Methods).

Figure 6. Mean dN/dS values are shown (with standard errors in parentheses) between S. aureus ST239 (TW20) and S. aureus USA300 (USA300), and between TW20 and S. aureus MRSA252 (MRSA252).

The brown section in TW20 depicts the large (635 kb) recombination event which originated from a close relative of MRSA252, the full genome is ∼3 Mb (not drawn to scale). On the left, the dendrogram illustrates the relationships between these strains.