Branch support was assessed using bootstrap sampling as previously reported [11]. Analyses were performed with each gene in a separate partition to which an independent model of evolution was

applied. The resulting ML phylogeny was compared with the consensus topology https://www.selleckchem.com/products/Y-27632.html obtained from Bayesian Inference (BI) [79, 80], with exploration of parameters using the Metropolis-Coupled Monte Carlo Markov Chain (MC3) algorithm with one million generations, as implemented in MrBayes v3.1.2, sampling a tree every 1,000 generations. The log-likelihood scores of sampled points were plotted against generation time to determine when the chain became stationary. All sample points prior to this (300,000 trees) were discarded as burn-in samples. Data remaining after discarding burn-in samples were used to generate a majority rule consensus tree, where percentage of samples recovering any particular clade represented the posterior probability of that clade. Probabilities ≥ 95% were considered indicative of significant support. Branch lengths of the consensus tree were estimated by maximum likelihood [81]. We performed additional phylogenetic reconstructions using Maximum Parsimony (MP) using the PAUP* package v4.0b10 [82]. MP trees were obtained in an equal weighted heuristic search with tree-bisection-reconnection (TBR)

branch swapping. The consensus tree was calculated using majority rule. Bootstrap (1,000 replicates, heuristic search TBR branch GSK2126458 mw swapping) was used to assess support for each node. A similarity matrix of all the concatenated sequences was prepared using the DNADIST program of the PHYLIP package [77] using Kimura distance [83], in order to compare the distances within the “”X. axonopodis”" clade with previous MLSA. Detection of genomic gains and losses The genomic gains and losses were identified and quantified using GenoPlast [57] with 10,000 burn-in iterations followed by 100,000 additional iterations, 10 iterations between sampling and two independent runs with identical parameters. Analyses were performed assuming a single phylogenetic

tree obtained by ML stiripentol inference. The input multiple alignment was conducted with progressive Mauve [84], and post-processed with the tools for developers of Mauve [85] to first obtain a binary matrix of presence/absence by region, and afterwards a matrix of presence/absence patterns counts. GenoPlast processes this matrix for the calculation of probabilities of ancestral events of genomic gains and losses and implements a model-based method to infer the patterns of genome content evolution by Bayesian inference, assuming a Poisson distribution of genomic gains and losses. The phylogeny inferred here was used as scaffold. Assignation of COG functional categories Homology with entries in the Cluster of Orthologous Groups of proteins (COG) database [86] was determined by BLAST searches [72] against the COG sequences database.