Local mobile gene pools rapidly cross species boundaries to create endemicity within global Vibrio cholerae populations

Abstract

Vibrio cholerae represents both an environmental pathogen and a widely distributed microbial species comprised of closely related strains occurring in the tropical to temperate coastal ocean across the globe (Colwell RR, Science 274:2025-2031, 1996; Griffith DC, Kelly-Hope LA, Miller MA, Am. J. Trop. Med. Hyg. 75:973-977, 2006; Reidl J, Klose KE, FEMS Microbiol. Rev. 26:125-139, 2002). However, although this implies dispersal and growth across diverse environmental conditions, how locally successful populations assemble from a possibly global gene pool, relatively unhindered by geographic boundaries, remains poorly understood. Here, we show that environmental Vibrio cholerae possesses two, largely distinct gene pools: one is vertically inherited and globally well mixed, and the other is local and rapidly transferred across species boundaries to generate an endemic population structure. While phylogeographic analysis of isolates collected from Bangladesh and the U.S. east coast suggested strong panmixis for protein-coding genes, there was geographic structure in integrons, which are the only genomic islands present in all strains of V. cholerae (Chun J, et al., Proc. Natl. Acad. Sci. U. S. A. 106:15442-15447, 2009) and are capable of acquiring and expressing mobile gene cassettes. Geographic differentiation in integrons arises from high gene turnover, with acquisition from a locally co-occurring sister species being up to twice as likely as exchange with conspecific but geographically distant V. cholerae populations. IMPORTANCE Functional predictions of integron genes show the predominance of secondary metabolism and cell surface modification, which is consistent with a role in competition and predation defense. We suggest that the integron gene pool's distinctness and tempo of sharing are adaptive in allowing rapid conversion of genomes to reflect local ecological constraints. Because the integron is frequently the main element differentiating clinical strains (Chun J, et al., Proc. Natl. Acad. Sci. U. S. A. 106:15442-15447, 2009) and its recombinogenic activity is directly stimulated by environmental stresses (Guerin E, et al., Science 324:1034, 2009), these observations are relevant for local emergence and subsequent dispersal.

FIG 1

Phylogenetic relationships among clinical and environmental V. cholerae, V. metecus, and other Vibrionaceae strains. Maximum likelihood phylogeny based on the concatenated DNA sequences of six protein-coding genes (mdh, adk, pgi, gyrB, recA, and rpoB). Bootstrap support values are displayed on the nodes. Isolates are color coded according to species and origins. Unless otherwise noted, all V. cholerae strains are non-O1/non-O139.

FIG 2

Phylogenetic relationship of representative “core” genes from Vibrio cholerae and V. metecus strains from the Dhaka Delta in Bangladesh (BGD) and from the U.S. east coast (USA). Phylogenetic trees derived from individual protein-coding housekeeping genes (A) and tree quartets (B) were built in TREE-PUZZLE (49) by joining pairs of V. cholerae isolates from BGD and the USA. The bars indicate the proportion of all quartets having topologies consistent (structured) and inconsistent (mixed) with geography, as well as the proportion of unstructured tree quartets, which have a multifurcating maximum likelihood topology. Isolates are color coded according to species and origins.

FIG 3

Similarity in integron cassette families among three biogeographic groups. Phylogenetic comparison (A) and Chao-Sørensen index (B) showing that geographically cooccurring V. cholerae and V. metecus are more similar to each other than to geographically distinct V. cholerae. For the phylogenetic comparison, phylogenetic trees for all 54 gene cassette families occurring in all three biogeographic groups were created. The proportions of clades grouping any two of these three biogeographic groups were counted if their bootstrap support was >80%. When multiple nested clades could have been counted, only the most inclusive clade was. The Chao-Sørensen index calculates the similarity of two samples corrected for bias due to incomplete sampling (15). When calculating the index, cassette families were defined as having >95% sequence similarity to account for recent gene transfers (see Table S1 in the supplemental material for additional sequence similarity cutoffs). All index values in this analysis had a 3% confidence interval.

FIG 4

Rapid decay of integron similarity as V. cholerae strains diverge in “core” genes. Integron and core gene similarities are measured as the proportion of all gene cassette families (defined by >90% nucleotide identity) shared by pairs of isolates and as nucleotide sequence identity at six protein-coding housekeeping loci among the same pairs, respectively. Because integron data combine whole-genome sequences and PCR-based gene cassette sampling, a control for the expected shared portion of gene cassettes between identical integrons was included. This is based on random subsamplings of 65 gene cassettes, corresponding to the average number of cassettes available for isolates in this study.