Pan-genome and phylogeny of Bacillus cereus sensu lato.

backgroundBacillus cereus sensu lato (s. l.) is an ecologically diverse bacterial group of medical and agricultural significance. In this study, I use publicly available genomes and novel bioinformatic workflows to characterize the B. cereus s. l. pan-genome and perform the largest phylogenetic and population genetic analyses of this group to date in terms of the number of genes and taxa included. With these fundamental data in hand, I identify genes associated with particular phenotypic traits (i.e., "pan-GWAS" analysis), and quantify the degree to which taxa sharing common attributes are phylogenetically clustered.

methodsA rapid k-mer based approach (Mash) was used to create reduced representations of selected Bacillus genomes, and a fast distance-based phylogenetic analysis of this data (FastME) was performed to determine which species should be included in B. cereus s. l. The complete genomes of eight B. cereus s. l. species were annotated de novo with Prokka, and these annotations were used by Roary to produce the B. cereus s. l. pan-genome. Scoary was used to associate gene presence and absence patterns with various phenotypes. The orthologous protein sequence clusters produced by Roary were filtered and used to build HaMStR databases of gene models that were used in turn to construct phylogenetic data matrices. Phylogenetic analyses used RAxML, DendroPy, ClonalFrameML, PAUP*, and SplitsTree. Bayesian model-based population genetic analysis assigned taxa to clusters using hierBAPS. The genealogical sorting index was used to quantify the phylogenetic clustering of taxa sharing common attributes.

resultsThe B. cereus s. l. pan-genome currently consists of ≈60,000 genes, ≈600 of which are "core" (common to at least 99% of taxa sampled). Pan-GWAS analysis revealed genes associated with phenotypes such as isolation source, oxygen requirement, and ability to cause diseases such as anthrax or food poisoning. Extensive phylogenetic analyses using an unprecedented amount of data produced phylogenies that were largely concordant with each other and with previous studies. Phylogenetic support as measured by bootstrap probabilities increased markedly when all suitable pan-genome data was included in phylogenetic analyses, as opposed to when only core genes were used. Bayesian population genetic analysis recommended subdividing the three major clades of B. cereus s. l. into nine clusters. Taxa sharing common traits and species designations exhibited varying degrees of phylogenetic clustering.

conclusionsAll phylogenetic analyses recapitulated two previously used classification systems, and taxa were consistently assigned to the same major clade and group. By including accessory genes from the pan-genome in the phylogenetic analyses, I produced an exceptionally well-supported phylogeny of 114 complete B. cereus s. l. genomes. The best-performing methods were used to produce a phylogeny of all 498 publicly available B. cereus s. l. genomes, which was in turn used to compare three different classification systems and to test the monophyly status of various B. cereus s. l. species. The majority of the methodology used in this study is generic and could be leveraged to produce pan-genome estimates and similarly robust phylogenetic hypotheses for other bacterial groups.