Phylogeographic reconstruction of a bacterial species with high levels of lateral gene transfer

Abstract

Background: Phylogeographic reconstruction of some bacterial populations is hindered by low diversity coupled with high levels of lateral gene transfer. A comparison of recombination levels and diversity at seven housekeeping genes for eleven bacterial species, most of which are commonly cited as having high levels of lateral gene transfer shows that the relative contributions of homologous recombination versus mutation for Burkholderia pseudomallei is over two times higher than for Streptococcus pneumoniae and is thus the highest value yet reported in bacteria. Despite the potential for homologous recombination to increase diversity, B. pseudomallei exhibits a relative lack of diversity at these loci. In these situations, whole genome genotyping of orthologous shared single nucleotide polymorphism loci, discovered using next generation sequencing technologies, can provide very large data sets capable of estimating core phylogenetic relationships. We compared and searched 43 whole genome sequences of B. pseudomallei and its closest relatives for single nucleotide polymorphisms in orthologous shared regions to use in phylogenetic reconstruction.

Results: Bayesian phylogenetic analyses of >14,000 single nucleotide polymorphisms yielded completely resolved trees for these 43 strains with high levels of statistical support. These results enable a better understanding of a separate analysis of population differentiation among >1,700 B. pseudomallei isolates as defined by sequence data from seven housekeeping genes. We analyzed this larger data set for population structure and allele sharing that can be attributed to lateral gene transfer. Our results suggest that despite an almost panmictic population, we can detect two distinct populations of B. pseudomallei that conform to biogeographic patterns found in many plant and animal species. That is, separation along Wallace's Line, a biogeographic boundary between Southeast Asia and Australia.

Conclusion: We describe an Australian origin for B. pseudomallei, characterized by a single introduction event into Southeast Asia during a recent glacial period, and variable levels of lateral gene transfer within populations. These patterns provide insights into mechanisms of genetic diversification in B. pseudomallei and its closest relatives, and provide a framework for integrating the traditionally separate fields of population genetics and phylogenetics for other bacterial species with high levels of lateral gene transfer.

Figure 1

Bayesian phylogenetic analysis of 11,208 single nucleotide polymorphisms shared among 43 whole genome sequences from six species. Credibility values for major clades are included. B. dolosa and B. ubonensis were used to root this tree.

Figure 2

Phylogenetic analysis of 14,544 single nucleotide polymorphisms shared among 33 whole genome sequences from two species. Splits decomposition analysis provides a visual account of character state conflict within the dataset (A). Bayesian phylogenetic analysis of these SNPs result in clade credibility values of 1.00 for all bifurcations (except 23344/JHU = 0.96) (B). B. thailandensis was used to root this tree (see Additional file 3).

Figure 3

Estimated population structure of B. pseudomallei/mallei using multilocus sequence typing data. Each thin vertical line represents an ST and is divided into K portions that represent the estimated membership in K populations. Geographic affiliations of STs are labeled below the figure.

Figure 4

eBURST population snapshot of B. pseudomallei and closely related species. Dots represent individual MLST sequence types from Australia (red) and other countries (black). Single locus variants are connected by black (to predicted clonal complex founder) and pink lines (alternative single locus variants), whereas double locus variants are connected by blue lines. Isolates with WGSs are indicated, except BCC215, 91, 4, and 700388 as their sequence types are novel and are not yet included in the database.

Figure 5

B. pseudomallei multilocus sequence typing diversity of isolates from Australia, Papua New Guinea, and New Caledonia (red), and the rest of the world (black). Allelic diversity at each locus is greater for the 811 isolates from Australia, Papua New Guinea, and New Caledonia than the 801 isolates from the rest of the world, suggesting an ancestral Australasian population. Conversely, the number of sequence types (STs) found in the rest of the world is greater than the number found in Australasia, suggesting lower levels of recombination in Australasia.

Figure 6

Depiction of higher interconnectivity of Southeast Asian sequence types compared to Australian sequence types. Red bars represent Australian STs and gray bars represent Southeast Asian STs. Height of each bar is the proportion of STs that possess at least one single, double, or triple locus variant. Horizontal lines represent unit increments of the numbers of variants for each category (e.g., 0.23 of the Australian STs have only one single locus variant, whereas 0.06 have two single locus variants and 0.04 have three single locus variants). Sum of columns for each region is greater than one as each ST can have variants in each column.

Figure 7

Comparison of population metrics across 11 bacterial species. The per-allele recombination to mutation parameter (r/m allele) suggests that B. pseudomallei alleles are between 18 and 30 times more likely to change by recombination rather than mutation. This value is higher than for any other bacterial species yet reported. The low standardized index of association suggests that these populations approach panmixia, and Nei's diversity index suggest that these species are less diverse than many other species.