V. cholerae O1 serogroup Ogawa biotype El Tor was isolated from 6 (3.4%) of the 179 water samples and 1 (0.7%) of the 144 aquatic animal and plant samples by using modified APW enrichment techniques. Of those 7 environmental isolates, 3 (43%) were confirmed as ctx-positve toxigenic V. cholerae O1 strains, and 4 (57%) were confirmed as ctx-negative V. cholerae O1 strains by using genetic analysis as described below. As shown in Table 1, APW enrichment at 37°C overnight or incubation at 40°C for 6–8 hours, or both, enhanced the rate of isolation of V. cholerae O1 from samples. PCR analysis of the key virulence genes showed that 3 (43%) of the 7 isolates, all from water, were positive for key virulence genes, including cholera toxin genes and tcpA genes, and that 4 (57%) isolates exhibited no cholera toxin bacteriophage (CTXΦ)–related genes (23; Table 2). To further assess the PCR results, we sequenced DNA flanking the CTXΦ from 1 strain, Env-9 (Table 2). Sequence data corroborated PCR results that indicated that Env-9 lacked CTXΦ.

Physical parameters for the environmental water samples are summarized in Table 3. Because the sites varied from mountains to floodplain to estuaries, there was relatively wide variability in salinity (0–21.6 g/L), pH (6.4–8.6), and temperature (24.3–33.7°C). Temperatures tended to increase as rivers approached the sea. As shown in Figure 2, mean water temperature from all sites showed evidence of seasonal variation. Measurement of rainfall was available for the region as a whole (Figure 3). However, site-specific rainfall data were not available; consequently, rainfall was not included in the regression models.

Isolation of V. cholerae O1 strains was most common from the sites at the mouths of the Momance and Gressier Rivers (Figure 1, panel A). In a conditional logistic regression analysis with water quality factors (Table 4), the only variable that emerged as statistically significant was water temperature (odds ratio 2.14, 95% CI 1.06–4.31); all isolations of V. cholerae O1 (toxigenic and nontoxigenic) occurred at water temperatures of >31°C. As shown in Figure 3, there was evidence that V. cholerae O1 isolation was more common in the environment preceding epidemic peaks of disease among humans; however, numbers of isolations were too small to permit statistical analysis. Of 179 samples, the only V. cholerae O1 isolate from aquatic animals or plants was from a shrimp sample and was nontoxigenic; it was collected simultaneously with a water sample that was also positive for nontoxigenic V. cholerae O1.

Non-O1 V. cholerae was much more common in the environment than V. cholerae O1 strains and was isolated from 56 (31%) of 179 water samples. As observed with O1 strains, isolations were more common at the mouths of the rivers and in estuarine areas (Figure 1, panel B); however, the non-O1 strain was found farther upriver than were O1 strains and was isolated from several sites in the mountains. Non-O1 strains were isolated from all sites that were also positive for O1 strains. Non-O1 strains were isolated in all months, without an obvious association with regional rainfall totals or cholera incidence. In a conditional logistic regression analysis, isolation of non-O1 strains was significantly associated (p<0.05) with higher water temperature and moderate levels of dissolved oxygen (Table 4).

Before this study, isolation of 2 toxigenic V. cholerae O1 strains from large-volume water samples (30 L) was reported in the Artibonite region (24); other studies at that time suggested that V. cholerae O1 strains were not present, or present at only minimal levels (2,25) in the environment in Haiti. In contrast, we isolated ctx-positive and ctx-negative V. cholerae O1 serogroup Ogawa biotype El Tor strains (Table 1) in the environment at a frequency comparable to that reported from cholera-endemic areas such as Bangladesh (17). Our successful isolation of the microorganism from the environment may reflect localization of environmental isolates near Gressier and Leogane, where our study was conducted; however, we believe that our findings are more likely to be a reflection of the method used. Data presented here suggest that, in addition to conventional APW enrichment, longer APW enrichment time and enrichment at higher temperatures contributed to an increased rate of isolation of V. cholerae O1 strains from aquatic environmental reservoirs (Table 1), resulting in successful isolation from 1.5-mL water samples. We also note some issues relating to sample transport: Baron et al. (25) transported their water samples on ice in coolers; our samples were transported at room temperature. As has been reported, Vibrio spp. are extremely sensitive to low temperatures (26), and in our experience, transport of samples on ice resulted in a marked reduction in isolation rates.

Water from which we isolated V. cholerae spp. tended to have been sampled at the point where rivers meet the sea, and in adjacent estuarine areas, again following the patterns reported from Bangladesh (17). Water temperature was found to be the single physical parameter that was substantially associated with isolation of these organisms; higher temperatures were concentrated downriver and in estuarine areas. For our analysis, we used a conditional logistic regression model to permit stratification by site. Although we found very low numbers for V. cholerae O1 isolates (6 positive water samples), results coincided with the non-O1 results and the exploratory data analysis. In our studies of aquatic animals likely to be eaten by humans, we did isolate V. cholerae O1 from shrimp in 1 instance. The isolate was nontoxigenic; consequently, its association with disease is unclear.

After analyzing the results of this study, we asked the following question: has V. cholerae O1 become established in environmental reservoirs in Haiti? Toxigenic V. cholerae O1 strains are clearly present in the environment, and it may be that the isolates that we identified are the result of fecal contamination of the environment by persons infected with V. cholera strains. Although data are limited, there was at least a suggestion that isolation of V. cholera strains from environmental reservoirs was more common at the beginning of epidemic spikes of human disease (as has been described in association with environmental reservoirs) (16,17,27) rather than at the height of epidemics among humans, as might have been expected related to fecal contamination. We also found non-O1 strains widely distributed throughout the environment, including mountain river sites, consistent with widespread dissemination in environmental reservoirs. Although we cannot be certain that O1 and non-O1 strains grow under comparable conditions, the clear establishment of non-O1 V. cholerae strains in environmental reservoirs suggests that conditions are appropriate for growth of V. cholerae O1 strains. Of potentially greater interest is the observation that only 3 of the 7 (47%) V. cholerae O1 biotype El Tor strains isolated carried the ctx genes (Table 2). Data from 1 ctx-negative strain (Env-9) was consistent with absence of the entire CTXΦ. We propose that the 3 isolates that are positive for ctx genes be classified as circulating V. cholerae altered biotype El Tor strains in Haiti. To better understand the evolutionary mechanisms involved, we are performing further sequence analysis of clinical and environmental strains.

The apparent introduction of toxigenic V. cholerae O1 in Haiti in 2010, after decades during which no cholera cases were reported, was unquestionably a public health disaster. If these O1 strains establish stable environmental reservoirs in Haiti, in the setting of ongoing problems with water and sanitation, there is a high likelihood that we will see recurrent epidemics within the country. These circumstances clearly have implications for current plans by the Haitian Ministry of Public Health to eradicate cholera in Haiti within a decade (28). The proposed implementation of vaccination programs and efforts to improve water supplies and sanitation will undoubtedly reduce case numbers, but as long as the causative microorganism is present in the environment, eradication of the disease will not be possible. Establishment of environmental reservoirs and recurrent epidemics may also serve as a potential source for transmission of the disease to the Dominican Republic and other parts of the Caribbean (1). Ongoing monitoring of potential environmental reservoirs in the areas near Gressier and Leogane as well as in sentinel sites throughout the country will be necessary to assess this risk and to permit development of rational public health interventions for cholera control.