To tree or not to tree? Genome-wide quantification of recombination and reticulate evolution during the diversification of strict intracellular bacteria

Abstract

It is well known that horizontal gene transfer (HGT) is a major force in the evolution of prokaryotes. During the adaptation of a bacterial population to a new ecological niche, and particularly for intracellular bacteria, selective pressures are shifted and ecological niches reduced, resulting in a lower rate of genetic connectivity. HGT and positive selection are therefore two important evolutionary forces in microbial pathogens that drive adaptation to new hosts. In this study, we use genomic distance analyses, phylogenomic networks, tree topology comparisons, and Bayesian inference methods to investigate to what extent HGT has occurred during the evolution of the genus Rickettsia, the effect of the use of different genomic regions in estimating reticulate evolution and HGT events, and the link of these to host range. We show that ecological specialization restricts recombination occurrence in Rickettsia, but other evolutionary processes and genome architecture are also important for the occurrence of HGT. We found that recombination, genomic rearrangements, and genome conservation all show evidence of network-like evolution at whole-genome scale. We show that reticulation occurred mainly, but not only, during the early Rickettsia radiation, and that core proteome genes of every major functional category have experienced reticulated evolution and possibly HGT. Overall, the evolution of Rickettsia bacteria has been tree-like, with evidence of HGT and reticulated evolution for around 10-25% of the core Rickettsia genome. We present evidence of extensive recombination/incomplete lineage sorting (ILS) during the radiation of the genus, probably linked with the emergence of intracellularity in a wide range of hosts.

Keywords: Rickettsia; bacterial speciation; horizontal gene transfer; phylogenomic networks; reticulate evolution.

Fig. 1.—

Relationship between δ values for whole-genome genetic distances (GD), genomic conservation (synteny, GC), and genomic rearrangements (GR). Rickettsia species were assigned to two categories depending on their host range: wide (with different insect vectors and vertebrate hosts; dark green circles) and narrow (with a single genus insect vector and few vertebrate host; dark red squares). The 24 taxa are projected on the first two PCA axes, which represent 68% and 16% of the total inertia.

Fig. 2.—

Split phylogenomic network obtained for Rickettsia genome content data. Main groups are color-coded in pink (spotted fever group), yellow (typhus group), and blue (Felis group). The main recombination events are mapped as intercrossing symbols on the phylogeny. The SFG detail shows clades corresponding to the cohesion plot for core proteome. Details on recombination events are provided in supplementary material, Supplementary Material online.

Fig. 3.—

Cohesion plot for every gene tree, showing the consensus typology estimated by Phylo-MCOA. For each species, lines represent the distance between its reference position (squares) and its position in each individual tree. Long lines represent species which position in a given tree is not concordant with its reference position.

Fig. 4.—

Bar plots for the Rickettsia tree comparison, showing the score calculated for each gene (top) and each species (bottom). Red bars represent significant outliers according to the threshold chosen (0.2). Nodal distances were used for the Phylo-MCOA analysis.

Fig. 5.—

Galled rooted hybridization networks displaying putative recombination events that reconcile the differences between the 606 individual core gene trees. Decreasing thresholds for network construction reveal increasing numbers of recombination events. Networks display clusters present in at least (A) 20%, (B) 15%, and (C) 10% of the gene trees.

Fig. 6.—

Primary concordance tree for the Rickettsia core genome data set. Numbers are posterior mean CFs and their 95% credibility intervals, obtained with α = 1. Numbers refer to sample-wide CFs and their associated 95% confidence interval (in parentheses). Short branches (branch length given in coalescent units), most affected by ILS, have lowest CFs.