Phylogenetic analysis
Of the 5014 protein families, 497 were shared by all the 29 strains and contained only a single copy from each strain (did not contain paralogs). This set of 497 single-copy core genes were identified as putative orthologous genes. The sequences were first aligned at the amino acid level using Probalign [49], then backtranslated to DNA. Alignment columns with a posterior probability < 0.6 were removed, and alignments with > 50% of the sites removed were discarded from the analysis. Multiple alignments with Probalign retained 491 reliably aligned genes from a set of the 497 orthologous genes. Gene trees were reconstructed using PhyML (Phylogenetic estimation using Maximum Likelihood) [50,51] with the General Time Reversible plus Gamma (GTR + G) substitution model of DNA evolution, and the Subtree Pruning-Regrafting (SPR) branch-swapping method. Each gene tree search was bootstrapped (500 pseudoreplicates) using PhyML with the Nearest-Neighbor Interchange (NNI) branch-swapping method to detect genes that support or conflict with various bipartitions. A majority rule consensus of the gene trees was constructed using the consense program of PHYLIP 3.69 [52]. All the alignments were also concatenated, and a tree search was performed using PhyML with the same settings as for the gene trees. Macrococcus caseolyticus JCSCS5402 was used as an outgroup to root the trees. We used DendroPy [53] to annotate the nodes of the estimated consensus and concatenated gene trees with the percentage of gene trees in which the node was found. Resulting phylogenetic trees were drawn using the R package APE (Analysis of Phylogenetics and Evolution) [54].