Patterns and processes of Mycobacterium bovis evolution revealed by phylogenomic analyses

Abstract

Mycobacterium bovis is an important animal pathogen worldwide that parasitizes wild and domesticated vertebrate livestock as well as humans. A comparison of the five M. bovis complete genomes from the United Kingdom, South Korea, Brazil, and the United States revealed four novel large-scale structural variations of at least 2,000 bp. A comparative phylogenomic study including 2,483 core genes of 38 taxa from eight countries showed conflicting phylogenetic signal among sites. By minimizing this effect, we obtained a tree that better agrees with sampling locality. Results supported a relatively basal position of African strains (all isolated from Homo sapiens), confirming that Africa was an important region for early diversification and that humans were one of the earliest hosts. Selection analyses revealed that functional categories such as “Lipid transport and metabolism,” “Cell cycle control, cell division, chromosome partitioning” and “Cell motility” were significant for the evolution of the group, besides other categories previously described, showing importance of genes associated with virulence and cholesterol metabolism in the evolution of M. bovis. PE/PPE genes, many of which are known to be associated with virulence, were major targets for large-scale polymorphisms, homologous recombination, and positive selection, evincing for the first time a plethora of evolutionary forces possibly contributing to differential adaptability in M. bovis. By assuming different priors, US strains originated and started to diversify around 150–5,210 ya. By further analyzing the largest set of US genomes to date (76 in total), obtained from 14 host species, we detected that hosts were not clustered in clades (except for a few cases), with some faster-evolving strains being detected, suggesting fast and ongoing reinfections across host species, and therefore, the possibility of new bovine tuberculosis outbreaks.

Fig. 1.—

Core- (left) and Pan-genome (right) curves including only genomes of the 38-data set having <100 contigs (33 genomes). The core-genome is close to a plateau, indicating the complete core set is included in the 38-data set analysis (below 3,000 genes). The pan-genome curve indicates the pool of genes is quite high even within relatively closely related M. bovis genomes. Equations refer to the exponential regression for each curve.

Fig. 2.—

ML tree of the 38-data set matrix obtained using the site-congruence method. Left: branch support values (UFBoot); Right: same tree with actual branch lengths according to ML. RI for this tree (retention index for the character “sampling locality”) = 0.81.

Fig. 3.—

Variability of genomic characteristics across the 38 core-coding genes obtained in DAMBE. Each index was normalized to a relative average of 1.0 for better visualization. Numbers in x axis refer to individual genomes, with the 11 US genomes starting at position #28. Nc, effective number of codons; RSCU, relative synonymous codon usage; CAI2, corrected version of the codon adaptation index; GC-content, average genomic GC content; Sum(ACGT), size of the core-coding data set in base pairs.

Fig. 4.—

ML tree of the 76- data set composed of US strains, plus outgroup, based on their core-coding set of genes. Left: chronogram with respective ML support. Right: same tree with branches in substitutions/site. Red: clades of geographically associated genomes (a paraphyletic clade in the case of MN due to inclusion of a sample from TX).