Vibrio cholerae is a pathogen of considerable public health concern because of its potential to cause large epidemics and pandemics and its high case fatality rate when the disease is left untreated. The disease cholera is caused by V. cholerae strains of serogroups O1 and O139 that can produce a potent enterotoxin, cholera toxin, which is encoded by the ctxAB genes on the bacteriophage CTXϕ (1). Seven pandemics of cholera have been recorded since 1817, when the disease first emerged from the Bay of Bengal and spread around the globe (2). The current seventh pandemic of V. cholerae originated in Southeast Asia and has spread across the globe in several waves of transmission (3). In October of 2010, cholera made its appearance in Haiti. Prior to 2010, there were no documented cases of cholera in that country, despite the devastating outbreaks occurring in the Caribbean in the 19th century (4). Its introduction to the island of Hispaniola following the earthquake that occurred there in January of 2010 has resulted in the largest epidemic of cholera in recent times: 604,634 cases and 7,436 deaths were documented in the first two years of the epidemic (5).

Initial epidemiological and genetic studies focused on the origin of the Haiti epidemic and quickly attributed the outbreak to human introduction of a V. cholerae O1 strain from outside the region, most likely South Asia (6, 7). Epidemiological investigations pointed to Nepalese troops serving as United Nations (UN) peacekeepers as the source of cholera, based on reports of unsanitary conditions at the UN camp, the spatial-temporal pattern of disease clusters, and the coincidence of the outbreak with the arrival of the UN troops from Nepal (8). Phylogenetic analysis of time-relevant isolates from Haiti and Nepal provided additional support for the hypothesis that the epidemic strain was imported from Nepal (9).

The single-source introduction and geographic isolation of the Haiti epidemic, along with the extended duration of the outbreak, provide an unprecedented natural experiment for characterizing in detail the intrinsic tempo and mode of genome evolution in this deadly pathogen. We performed whole-genome sequencing on a set of well-characterized isolates collected near or after the 1-year anniversary date of the Haitian outbreak, and compared them with isolates collected early in the epidemic to gain insight into the dynamics of genome evolution in V. cholerae O1. The sample set includes isolates collected at different time points and in different localities (9, 10) as well as phenotypically and genotypically distinct isolates discovered during routine laboratory surveillance by the Centers for Disease Control and Prevention. The variants that have arisen in the course of the outbreak include various pulsed-field gel electrophoresis (PFGE) pattern combinations, serotype Inaba (11), an altered antibiotic susceptibility pattern (ASP), and a nonagglutinating (NAG) V. cholerae strain. We first conducted phylogenetic analysis to determine whether the diverse set of isolates were all part of the same outbreak and then used the genome sequences to compare gene content and structural arrangement of chromosomes.

We sequenced 23 genomes on the Illumina platform (see Table S1 in the supplemental material). The sample set represents geographically dispersed isolates collected over an array of time points and representing multiple PFGE pattern combinations (see Fig. S1 to S3 in the supplemental material). Eighty-seven genomes were downloaded from the Sequence Read Archive (SRA) (see Table S2 in the supplemental material); two were found to have >20% non-Vibrio genetic material and were excluded from the study: hc-17a1 and hc-77a1. Comparing the 108 genomes yielded 566 core genome single-nucleotide polymorphisms (SNPs). Of the 23 isolates, we sequenced 9 on the PacBio platform and resequenced the reference strain 2010EL-1786.

A phylogenetic tree constructed from the 566 core SNPs grouped all Haiti isolates and three Nepal isolates (14, 25, 26) in a single monophyletic group within the context of a global collection of 108 Vibrio cholerae O1 strains (see Fig. S4 in the supplemental material). Next, we uncovered 45 high-quality SNPs (hqSNPs) in the Haiti-Nepal group. The minimum spanning tree (MST) constructed from the hqSNPs was concordant with the clustering of isolates by maximum likelihood analysis (Fig. 1; also, see Fig. S4 in the supplemental material). The MST illustrated the radiation of numerous lineages from a single sequence type that predominated in the early part of the epidemic.

We then examined the 45 hqSNPs for potential effects on function (see Table S3 in the supplemental material). Most notable was a GAA-to-TAA substitution in the wbeT gene of 2012EL-1410, a representative of five serotype Inaba isolates from our Haiti collection. The substitution introduces a premature stop codon into the gene, which predicts a truncated protein, a result that is consistent with other studies showing that serotype conversion results from mutations in the wbeT gene (12). Comparison of three different molecular clock models showed that overall the changes at nucleotide sites were consistent with the epidemic behavior, as the highest likelihood was obtained under the exponential growth model. Using a strict molecular clock, analysis of 10⁸ states from eight independent runs yielded a median estimate of the date of the most recent common ancestor to be 28 September 2010 (95% credible interval, 23 July 2010 to 17 October 2010).

We used genomic approaches to characterize evolutionary changes in the V. cholerae O1 population following the introduction to Haiti. Our results were consistent with previous findings that show that the Haiti cholera outbreak is clonal and that Nepalese isolates are the closest relatives to the Haiti strain identified to date (9), even when placed in a phylogeny with a larger collection of isolates representing recent cholera epidemics. A previous study based on WGS (9) showed that Nepalese isolates were almost indistinguishable from Haiti isolates; however, that phylogeny did not include isolates recovered from recent cholera outbreaks. The phylogeny presented here provides evidence that the observed variants that were detected by our surveillance are part of the same outbreak and not representatives of secondary introductions. The synthesis of PFGE and sequence data demonstrated the utility of WGS in establishing the clonality of isolates that exhibit greater PFGE dissimilarity than would normally be attributed to a single-source outbreak. The use of high-resolution sequence data that are amenable to evolutionary analysis will greatly enhance our ability to discern transmission pathways of virulent clones, such as the one implicated in this epidemic.

The nucleotide polymorphism that we detected was consistent with the observed epidemic behavior, as the most supported model of population growth was exponential. The molecular clock calculated with this model estimated a most recent common ancestor date of 28 September 2010 (95% credibility interval [CI], 23 July 2010 to 17 October 2010) (see Fig. S5 in the supplemental material). The credibility interval encompasses the date that the Nepalese soldiers arrived in Haiti (9 October 2010) (8), as well as the first reported hospitalization of a cholera case (17 October 2010) (although an earlier fatal case with an onset date of 12 October may have been the index case) (15). The consistency between the molecular data and the epidemiological information demonstrates the utility of advanced statistical tools in outbreak investigations where epidemiological information may be lacking. Our results suggest that a population genomic approach can be very powerful in delimiting the time frame of an outbreak.

We observed remarkably few differences in the genetic repertoire of the V. cholerae O1 population (Fig. 2). All genes from Haiti isolates were found in the genome of the reference strain (2010EL-1786), and the differences in gene content could be attributed to loss of genetic material. The NAG isolate (2012V-1060) had a unique ~10-kb deletion, and closer examination revealed that several key components were missing from the rfb region (see Fig. S6 in the supplemental material). The isolate therefore appeared to be a serogroup O1 strain that was unable to synthesize or transport O antigen to the cell surface. Five representative strains were observed to have large deletions in the SXT, an ~100-kb integrative conjugative element that exhibits considerable diversity in gene content (16). One ~10-kb SXT deletion was found in three isolates (Fig. 1; also, see Fig. S7 in the supplemental material) that also shared an altered ASP. The typical ASP for Haiti V. cholerae includes intermediate resistance to chloramphenicol and nonsusceptibility to streptomycin, sulfisoxazole, trimethoprim/sulfamethoxazole, and nalidixic acid, resistance traits that are encoded by floR, strA, strB, sul2, dfrA1, and a point mutation in gyrA (17). The isolates with the altered ASP displayed nonsusceptibility only to nalidixic acid, and both PCR and genome analysis confirmed the loss of floR, strA, strB, and sul2 resistance genes in the SXT region of these isolates. The deletion could not be placed parsimoniously on the MST (Fig. 1); however, it falls within variable region III (16) (see Fig. S7) and is flanked by transposase genes, so it is reasonable to suppose that the same deletion occurred independently in two different lineages. The SXT deletion could be made parsimonious by inferring that the nonsynonymous transversion (Fig. 1, asterisk) reverted to its original state, an event that we concluded is less likely to occur. We note that a parsimonious reconstruction indicates loss of SXT genes in the three closest Nepal isolates after divergence from the common ancestor of the Nepal-Haiti cluster (Fig. 1), as the other two most closely related Nepal lineages from Hendriksen et al. (9) (Nepal-2 and Nepal-3) possess an intact SXT (see Fig. S7). The large deletions in the SXT and rfb regions (Fig. 3) were confirmed by the continuous-long-read (CLR) assemblies. The nearly complete assemblies from the CLR and the BLAST atlas were both consistent with the notion that the Haiti V. cholerae strain has not acquired genes or genomic islands. Thus, we found no evidence that unrelated bacterial strains in the environment have contributed to the diversification of the Haiti outbreak strain.

It is well accepted that HGT is a major force driving evolution in bacteria, including Vibrio (18, 19); thus, the lack of HGT observed in our study might be surprising. Limited data suggest that changes can accumulate over a relatively short time frame (20), and a previous study of V. cholerae from Haiti (10) reported accumulation of diversity early in the epidemic although gene acquisition was not specifically demonstrated in V. cholerae serogroup O1. In Vibrio cholerae, natural transformation, an important mechanism of HGT, occurs on chitinous surfaces and requires quorum sensing (QS) (i.e., the HapR QS transcription factor) (13, 14). Thus, we first considered whether transmission dynamics precluded the establishment of epidemic V. cholerae in the environment so that they lacked exposure to other Vibrio and to chitin, which are critical for HGT by transformation (12). Although it is possible that environmental factors could limit HGT between environmental and epidemic V. cholerae, they are likely in contact, as coisolation has been observed on several occasions from environmental and clinical samples (10) (Vince Hill [CDC], personal communication). We therefore tested the alternative hypothesis that the Haiti V. cholerae strain could be deficient in natural transformation and therefore unable to take up and incorporate foreign DNA. The transformation experiments on chitin showed that unlike naturally competent clinical strain C6706, each Haiti isolate was severely impaired in its ability to take up extracellular DNA derived from C6706 (Table 1) or from itself (Table 2). Further experiments confirmed that the isolates, like C6706, were quorum-sensing proficient and expressed hapR (data not shown). Thus, the Haiti strain appears to be limited in its ability to acquire new genetic material through transformation, but this is not due to QS deficiencies, as have been identified in other nontransformable V. cholerae isolates (13) (B. K. Hammer and E. E. Bernardy, unpublished results). It remains possible that V. cholerae isolates defective for transformation in lab settings may, nonetheless, be naturally competent in nature. However, to our knowledge no such strains have yet been described. Also, a longer time frame may be required for recombination to occur with more distantly related lineages and create mosaics that are successful enough to be observed. We note that the low rates of transformation would presumably not affect HGT via other mechanisms such as phage transduction or conjugation.

In summary, the Haiti cholera epidemic provided a unique opportunity to study the evolutionary dynamics of an isolated population of V. cholerae O1. Although PFGE was initially critical for determining that a single strain caused the outbreak, subsequent changes in PFGE patterns precluded our ability to determine that observed variants traced back to a single epidemic founder, an issue that was addressed by using WGS in our comprehensive surveillance strategy. The sample set was virtually homogeneous in gene content, an observation that led to the discovery that the Haiti strain is poorly transformable. Thus, our study indicates that transformation by unrelated environmental strains of V. cholerae has played no detectable role in the evolution of the outbreak strain. Further studies will define the genetic mutation(s) that rendered the Haiti strain defective in HGT via natural transformation. Once the mutation(s) is identified, its temporal origin and global prevalence can be determined to understand more about the stability and success of this particular genotype and the role that transformation plays in its genome evolution and success in establishing itself in a new environment. Atypical O1 El Tor V. cholerae strains such as the Haiti strain have already displaced prototypical El Tor strains and emerged as the predominant clone circulating in Asia and Africa (21–24). These strains have acquired multidrug resistance and enhanced virulence traits such as classical or hybrid CTX prophage and SXT-ICE, resulting in higher infection rates and harsher symptoms (25). With the tools such as WGS now being available for epidemiological surveillance and case tracking, we argue for renewed efforts aimed at cholera prevention to avert more widespread and difficult-to-treat cholera outbreaks.