Figure 1

Figure 1. Location of Wisconsin residents who submitted diarrheic stool specimens to Marshfield Laboratories. The symbol indicates the location of Marshfield, WI. Symbol...

Cultures for Aeromonas were performed on 2,565 diarrheic stool specimens from 2,310 patients. The median age of the patient population was 37 years (range 4 days to 97 years), and 55% were female. Most specimens (97.6%) were from patients residing in Wisconsin, primarily the central portion of the state, where groundwater, either from a municipal system or private well, is the source of drinking water (Figure 1). Some specimens came from communities along Lake Michigan where lake water is the source of drinking water.

Seventeen specimens (0.66%) from 17 patients (0.74%) tested positive for Aeromonas. Three stool isolates were identified as A. hydrophila and 14 isolates were A. caviae. All positive specimens were from Wisconsin residents. The median age of Aeromonas-positive patients was 27 years (range 1 month to 87 years) and 59% were male. Five Aeromonas-positive patients were coinfected with one other enteric pathogen (two patients with Campylobacter, one with Salmonella, one with Cryptosporidium, and one with Clostridium difficile toxin A), suggesting that in these patients Aeromonas may have been a transient colonizer.

Fourteen of the Aeromonas-positive patients agreed to have their drinking water sampled. Five patients resided in a household with a private well, eight were served by municipal wells, and one lived in a municipality that used Lake Michigan for its drinking water. Except for one system, all municipal water was chlorinated. One drinking water source, a private well, tested positive for A. hydrophila. Designing this study, we assumed that the fecal carriage rate of Aeromonas would be similar to its carriage rate in another study conducted in the Midwest (18) and, based on the ubiquitous occurrence of Aeromonas in water, we anticipated that a number of isolates would be collected from patients’ drinking water. Since only one was collected, additional drinking water isolates were obtained by combining samples from 1,500 private wells throughout Wisconsin that had been submitted to the Wisconsin State Laboratory of Hygiene between September and November 1998 for routine coliform testing. The composite samples (composed of 4–10 well samples) were membrane filtered and cultured for Aeromonas as described above. This process yielded an additional 37 A. hydrophila and 17 A. caviae isolates.

Figure 2

Figure 2. Pulsed-field gel electrophoresis patterns and similarity dendrogram of genomic DNA from Aeromonas hydrophila and A. caviae isolates from diarrheic stool (S) and groundwater (W). The number refers to the isolate number....

PFGE of the stool and groundwater Aeromonas isolates yielded 10–20 well-resolved genomic DNA bands, ranging in size from approximately 10–400 kb (Figure 2). Six isolates (one stool, five groundwater) were not amenable to XbaI digestion, resulting in poorly resolved DNA fragments. PFGE patterns indicated extensive genetic diversity. The 65 isolates analyzed by PFGE had 58 distinct patterns. Five patterns grouped two or more identical isolates. Three of those groups included only isolates derived from the same composite water sample, suggesting multiple isolations of the same strain. Analyzing all pairwise comparisons among the 65 isolates, the median similarity was 59% (range 16% to 100%, n=2,080). Isolates from the same ecologic source also exhibited high genetic diversity. The median similarities of stool and groundwater isolates were 58% (range 30% to 76%, n=120) and 60% (range 25% to 100%, n=1,176), respectively. Among the 12 isolates from diarrheic stool specimens that were negative for other enteric pathogens, the median similarity was 58% (range 30%–76%, n=66).

None of the stool isolates was genetically indistinguishable from the groundwater isolates (Figure 2). The two isolates that appeared epidemiologically related, the patient stool isolate (isolate 0209) and the isolate from his private well (isolate 1320), were 72% similar by the Dice coefficient and differed by nine bands, which, following the criteria of Tenover et al. (19), would be generally interpreted as being genetically unrelated. The highest similarity between a stool and water pair of isolates was 88%, between isolates 1251 and 2094, which differed by four bands (Figure 2). However, the stool isolate was from a patient simultaneously positive for Cryptosporidium oocysts, suggesting that the Aeromonas was transient. The second highest similarity was 86%, a four-band difference between isolates 0949 and 1294 (Figure 2). Of the 784 paired comparisons between stool and water isolates, 776 (99%) had similarities <80%, and the median similarity was 58% (range 16% to 88%).

The prevalence rate in this study was lower than rates from other large surveillance studies in the United States for Aeromonas in stool. Aeromonas was cultured from 2,848 diarrheic stool specimens submitted to a Los Angeles, California, hospital in the early 1980s and 80 (2.8%) were positive (20). Among 1,821 patients with diarrhea visiting a clinic in La Crosse, Wisconsin, during an 18-month period, Agger et al. (21) identified 20 (1.1%) that were positive for A. hydrophila. Moyer (18) examined 3,334 diarrheic stool specimens submitted by physicians over a 2-year period to an Iowa public health laboratory and found 238 (7.1%) positive for either A. caviae, A. hydrophila, or A. sobria. Isolation in the latter study included an alkaline peptone water enrichment step, which may explain the higher prevalence rate. In our study, the specimen prevalence rate was 0.66%, and this rate was likely biased upwards because stool specimens were collected only in the summer and fall months when the incidence of Aeromonas gastrointestinal infections is reportedly highest (18,21). The data, albeit limited, do not suggest that the prevalence of Aeromonas enteric infections is increasing in the United States. Worldwide, the isolation rate of Aeromonas from diarrheic stool has been reported as high as 10.8% (22) and as low as 0% (23). In the latter study, recently conducted in Melbourne, Australia, during a 68-week observation period, 795 fecal specimens were collected from city residents with highly credible gastroenteritis. Aeromonas was not detected in any of the fecal specimens, even though 50% of water samples drawn weekly from the drinking water distribution mains serving the study participants were Aeromonas-positive (23).

The A. hydrophila and A. caviae species designations were equivocal in this study. Except for stool isolate 0858, all isolates from stool identified by the API-20E system as A. caviae were β-hemolytic, suggesting that they were A. hydrophila instead. When the Vitek automated microbe identification system (bioMérieux) was used, all stool and water isolates were identified as the A. hydrophila/caviae group. If the most recent edition of the API analytical profile index were used with the A. caviae profiles derived during the study, the new species designation would be A. hydrophila group 1. Some profiles determined in this study are not listed in the most recent index and, given the ever-changing taxonomy of aeromonads (4), we opted for a consistent one-index approach, reporting the species designations for all stool and water isolates on the basis of the index available at the time of the study.

The high level of genetic diversity observed in our study among clinical and environmental strains of Aeromonas has been corroborated by other nucleic acid–based subtyping methods, such as amplified fragment length polymorphisms (11) and ribotyping (24). Talon et al. (17) subtyped 10 epidemiologically unrelated strains of A. hydrophila by PFGE and reported that the median similarity (calculated by using the Pearson correlation coefficient) was 28.4% (range 9.3% to 44.3%). The variation is not likely due to genetic lability, because the PFGE patterns of Aeromonas reportedly do not become unstable during frozen storage and long-term laboratory culture (17,25). As a control in this study, isolate 1320 was digested and underwent electrophoresis five independent times, and each time yielded the same PFGE pattern. The PFGE patterns also did not correspond to phenospecies. Millership and Want (26) reported a similar finding based on whole-cell protein fingerprinting.

The capacity for human enteropathogenicity among clinical isolates may be derived from a unique set of genes that were acquired or evolved in a common ancestor. Alternatively, enteropathogenicity may have arisen independently among several genotypes. In either scenario, one might expect the subset of Aeromonas strains that are pathogenic to have less genetic variation than the environmental strains. In this study, the level of genetic diversity was similar between environmental and clinical strains, even when the clinical strains were restricted to the subset from stool specimens that were negative for other enteric pathogens. One explanation is that some clinical strains were transient gastrointestinal colonizers, and only a few strains that are truly diarrheagenic. This issue will likely remain unresolved until the pathogenicity mechanisms of Aeromonas are better understood.

PFGE is a reproducible, highly discriminatory subtyping method capable of identifying the transmission source of bacterial infections (19). If the Aeromonas-positive patient with the positive well acquired the infection from his drinking water, the molecular fingerprints of the stool and water isolates should at least have been closely related (i.e., <3-band difference). Likewise, if drinking water is a frequent source of Aeromonas infections in this study population, one would expect at least a few of the stool and water isolates to be more closely related than a four-band difference. The analysis for this study was weighted towards identifying similar stool and water isolates. The 1.5% position tolerance that was selected for band calling resulted in matching bands that differed by as much as 17 kb, and the Dice coefficient gives greater weight to matching bands compared to other similarity coefficients.

The study had several limitations, however, that need to be considered when interpreting the data. Only one isolate per stool specimen or water sample was analyzed by PFGE. Multiple strains may have been present in water, and by chance the enteropathogenic strains were not selected for PFGE analysis. Kühn et al. found multiple Aeromonas strains in a single drinking water source, although only one or a few strains were numerically dominant, and these could persist for years (6,11). When duplicate water samples (incubated at 30°C and 35°C) both yielded Aeromonas, then PFGE was performed on a colony from each plate. Some of these isolate pairs were indistinguishable by PFGE, and some were unrelated, showing that some composite water samples had at least two different strains. Both strains were included in the genetic similarity analysis. Another study limitation was that if diarrheagenic Aeromonas strains are very rare in groundwater, the sample size might have been insufficient to find those strains, even though the composite water samples tested represented 1,500 wells. The stool and water isolates were collected from the same geographic area and during the same period, which should have increased the odds of finding genetically similar isolates if the two ecologic sources are linked. Finally, the drinking water sample volume for Aeromonas isolation was 100 mL. Since the study was conducted, the EPA has developed and validated Method 1605 for detecting Aeromonas in drinking water, which specifies a minimum sample volume of 1 L (27). Possibly, if the sample volume had been 1 L, more patient wells would have been Aeromonas positive.

To our knowledge, this is the first study to use PFGE to compare Aeromonas strains from human stool with strains found in groundwater. Other studies have compared the relatedness of strains from drinking water and stool by fatty acid methyl ester profiles (28), ribotyping (29), and randomly amplified polymorphic DNA (30), all highly discriminatory subtyping methods, and found little similarity between clinical and environmental isolates. Aeromonas isolates from stool and drinking water have been linked by biotyping (31) and whole-cell protein fingerprinting (25), but the discriminatory power of these methods with Aeromonas is questionable (28,30). Thus, the evidence to date from using highly discriminatory subtyping methods suggests that human enteropathogenic strains are rare in drinking water. In the group of primarily Wisconsin residents in this study, Aeromonas was identified infrequently in diarrheic stool specimens and drinking water from a groundwater source did not appear to be an Aeromonas transmission route.