eBURST: Inferring Patterns of Evolutionary Descent among Clusters of Related Bacterial Genotypes from Multilocus Sequence Typing Data

The introduction of multilocus sequence typing (MLST) for the precise characterization of isolates of bacterial pathogens has had a marked impact on both routine epidemiological surveillance and microbial population biology. In both fields, a key prerequisite for exploiting this resource is the ability to discern the relatedness and patterns of evolutionary descent among isolates with similar genotypes. Traditional clustering techniques, such as dendrograms, provide a very poor representation of recent evolutionary events, as they attempt to reconstruct relationships in the absence of a realistic model of the way in which bacterial clones emerge and diversify to form clonal complexes. An increasingly popular approach, called BURST, has been used as an alternative, but present implementations are unable to cope with very large data sets and offer crude graphical outputs. Here we present a new implementation of this algorithm, eBURST, which divides an MLST data set of any size into groups of related isolates and clonal complexes, predicts the founding (ancestral) genotype of each clonal complex, and computes the bootstrap support for the assignment. The most parsimonious patterns of descent of all isolates in each clonal complex from the predicted founder(s) are then displayed. The advantages of eBURST for exploring patterns of evolutionary descent are demonstrated with a number of examples, including the simple Spain(23F)-1 clonal complex of Streptococcus pneumoniae, "population snapshots" of the entire S. pneumoniae and Staphylococcus aureus MLST databases, and the more complicated clonal complexes observed for Campylobacter jejuni and Neisseria meningitidis.