This study was carried out as part of a larger research project on the ecology of M. ulcerans in West Africa. Samples were collected in July 2008 from 25 sites in Ghana. Sites were located along a gradient from the Volta Region, where no M. ulcerans or mycolactone-producing mycobacteria have been detected, to the Greater Accra area, where M. ulcerans and other mycolactone-producing mycobacteria have been detected. Buruli ulcer disease is largely absent from the Volta Region, and it is endemic to the Greater Accra area (Fig. 1). A second sampling season occurred in August 2009, in which six sites were sampled in the Ga West District of Accra, Ghana, where M. ulcerans is considered to be endemic, and one site was sampled in the Tema district of Accra, Ghana, where M. ulcerans is not endemic (31, 34) (Fig. 2). It is unclear whether there is a relationship between seasonality and Buruli ulcer disease occurrence in West Africa (5); however, both sampling efforts in the current study were conducted in late July/early August to minimize any potential seasonal impacts.

Sample collection procedures varied between 2008 and 2009 due to differing objectives for each season. In 2008, the primary objective was to conduct an aquatic community analysis to determine whether certain fish or amphibian species were associated with ER-positive water bodies. For the 2008 season, water samples were collected for DNA processing and fish and amphibians were collected for identification but not for DNA analysis. In 2009, the primary objective was to determine whether fish and amphibians were potential reservoirs of M. ulcerans or other mycolactone-producing mycobacteria. For the 2009 season, fish and amphibians were collected for DNA processing. External swabs were taken from a subsample of fish and amphibians in 2009 to clarify whether positive results were due to external contamination. Water samples were not processed for the 2009 season because of the high expense of processing and because most samples were collected from sites that could reasonably be expected to be ER-positive due to either human endemicity data or previously collected ER-positive samples. DNA testing rather than bacterial cultures was used because M. ulcerans has a slow growth rate, making it extremely difficult to culture from environmental samples where other faster-growing contaminating microorganisms are present (5).

Three fish- and amphibian-sampling techniques were employed at each site to maximize the variety of taxa captured. Three collapsible live bait traps (Promar model TR-501, Gardena, CA) were deployed at each site and baited with commercial catfish dough bait (Berkley, Spirit Lake, IA), and the traps were set as far apart as possible around the site. Traps were submerged for approximately one hour at each site. In addition, a seine (Cabela's Inc., Sidney, NE) was used when the sites were wadeable and without abundant aquatic vegetation. A D-net (Bioquip Products, Rancho Dominguez, CA) was used to sample marginal areas and within aquatic plant beds. Approximately 1 hour of combined seining and D-net sampling was performed at each site. When adult frogs were observed at a site, they were captured with D-nets. Up to 30 individuals of each fish and amphibian taxon present at the site were captured, and excess individuals were released. Two exceptions to these methods occurred in 2009. A higher number of specimens were collected from Kwashikuma because a pond as well as an adjacent marsh area was sampled, and at Lake Weija, live fish were purchased from local fishermen and no amphibian collections were attempted. Fish and tadpoles were euthanized with CO₂, and adult frogs were euthanized with benzocaine hydrochloride (IACUC AUF # 06/08-090-00). Buckets used to hold captured fish and amphibians were cleaned and treated with RNase AWAY (Molecular Bioproducts, Inc., San Diego, CA) between sites.

In 2009, external swab samples were taken from 25 randomly chosen fish and amphibian specimens collected from Kwashikuma. This was done to determine the likelihood of positive DNA results being caused by bacteria adhering to the external surface of a specimen. Immediately following euthanasia of each specimen, the entire external surface of each specimen was rubbed with a sterile swab, which was then preserved in ethanol. An additional two swab samples were taken from tadpoles captured at Otuaplem that showed outward signs of potential infection.

Water was collected and filtered at all sites in 2008 to test for the presence of M. ulcerans and other mycolactone-producing mycobacteria. 20 water filters were analyzed per site, 10 from water samples collected in open water and 10 from water collected near aquatic vegetation. Samples were initially filtered through a 1.6 µm fiberglass filter to remove larger debris, and then filtered through a 0.2 µm nitrocellulose filter, as described in previously published methods (31).

In 2008, all tadpoles were preserved in 10% formalin and all fish were preserved in 95% ethanol, and all organisms from a single site were preserved together. For 2009, half of all specimens (fish and tadpoles) were preserved in 10% formalin and half were preserved in 95% ethanol. Formalin was used to preserve voucher specimens for taxonomic identification. Ethanol was used for specimens intended for DNA processing. Specimens in ethanol were either preserved individually or pooled into groups of three depending on size.

Fish and amphibians were identified to the highest taxonomic resolution possible using appropriate keys (35–37). For DNA analysis, specimens were dissected and the intestines and kidneys were removed and stored in 95% ethanol. The kidney was chosen because though mycobacteriosis in fish can potentially involve any of the organs, it most commonly involves the kidney and liver (30). A thin membrane separates the kidney from the other internal organs, minimizing chance of cross-contamination during dissection. While gills were tested in previous fish studies, they were not included in this study because it would be difficult to rule out incidental contamination of gills. With fish artificially infected with M. marinum and M. peregrinum, mycobacteria were subsequently recovered from internal organs of infected fish but never from the gills (38). The intestine was chosen because experimental exposure of Danio rerio to M. marinum and M. peregrinum showed that these mycobacteria are primarily acquired through the intestine before disseminating to other organs (30). Intestine results were used to show whether fish and amphibians ingested mycolactone-producing mycobacteria, while kidney results were used to ascertain if ingested mycobacteria produced an internal infection.

Instruments were cleaned between specimens using 95% ethanol followed by application of RNase AWAY.

Samples from 2009 preserved in 95% ethanol were screened for ER-positive DNA. All DNA extractions and PCR procedures were performed in a hood, under sterile conditions. Negative controls were used throughout the process to ensure sterility and assess possible contamination. Cultured M. ulcerans cells preserved in 95% ethanol were used as a positive control to ensure that procedures were successful. For small samples (less than or equal to 20 mg wet mass), the entire sample was used for DNA extraction. For large samples, homogenate containing approximately 20 mg of tissue was used for DNA extraction. Samples were centrifuged to remove ethanol before DNA extraction. The remaining pellet was rinsed with TE, centrifuged, and the supernatant was removed.

Extraction was first attempted using the one tube method designed for extraction of M. ulcerans in aquatic insects, mollusks, and fish (12). However, satisfactory detection was not achieved with this method. DNA was then extracted using the Käser method, which is optimized for environmental mycobacteria (39). However, in the final step, DNA was re-suspended in 25 µl TE buffer instead of the 100 µl water used in the Kaser method. To test this method, serial dilutions were made from a stock sample of M. ulcerans cells in 95% ethanol that contained 10⁷ CFUs/ml. Aggregates of bacteria were broken apart by passing the bacterial suspension through a 25-gauge needle 10 times, and eight 1:10 serial dilutions were made (31). To determine whether sensitivity was changed by the addition of fish tissue, the internal organs of a commercially purchased goldfish were dissected and processed using the previously mentioned techniques. Also, 180 µl of 0.1 mg/µl (wet mass) tissue homogenate was spiked with 20 µl of each of the previously made dilutions before being subjected to the Käser extraction procedure (39) and ER-PCR.

Extractions were performed in groups of 18 samples, and each group included a negative and a positive control. Extracted DNA from a subset of 27 intestine samples and 23 kidney samples was spiked with M. ulcerans DNA prior to PCR analysis to determine the level of PCR inhibition. Due to the presence of inhibitors, it was necessary to further purify the extracted intestinal DNA according to manufacturer's instructions using PowerClean DNA Clean-Up Kits (MO BIO Laboratories, Inc., Carlsbad, CA), eluting with 50 µl of the provided elution solution. Negative and positive controls were also cleaned using this procedure. Because the clean-up of the intestine samples resulted in more dilute DNA than that of the kidney samples, 8 µl of intestine sample DNA versus 4 µl of kidney sample DNA was used per 25 µl PCR mixture.

PCR was performed targeting the ER domain of the plasmid responsible for mycolactone production. The 25 µl PCR cocktails contained 2 µl of each primer (10 µM), 12.5 µl FailSafe 2×PCR buffer (EPICENTRE Biotechnologies, Madison, WI), 0.5 µl of Go Taq polymerase enzyme (Promega, Madison, WI), and either 4 µl of template DNA and 4 µl of PCR water (for kidney samples) or 8 µl of template DNA (for intestine samples). PCRs were run using previously described primers and cycling conditions (31). Samples were loaded along with positive and negative controls onto 0.8% TBE agarose gels stained with ethidium bromide, and fragment sizes were compared to a 1 Kb DNA ladder (Invitrogen, Carlsbad, CA). All ER-positive samples were then sent to the University of Tennessee for VNTR analysis as previously described (31).

We sought to determine if there was a difference in fish or amphibian communities at sites that tested positive or negative for the presence of mycolactone-producing mycobacteria. We also sought to determine which taxa contributed to these differences. A Jaccard similarity matrix was constructed separately for both fish and amphibian communities using data collected in 2008, and combined data from 2008 and 2009. A Jaccard similarity matrix allows the comparison of species diversity between each pair of sites by analyzing the presence and absence of taxa (40). Separate non-metric multidimensional scaling (NMS) ordinations were performed on fish and amphibian community ranks obtained from Jaccard matrices using ER positivity as a grouping variable. NMS ordinations allow for visualization of relative similarity between sites; sites that appear closer together in the ordination are more similar (41). Analyses of similarity (ANOSIM) were also performed. ANOSIM is an iterative approach that compares within group similarity ranks to between group similarity ranks to determine if there is a statistical difference between groups (42). ANOSIMs were performed using 10,000 permutations to determine if there were differences between ER-positive and ER-negative fish or amphibian communities. A similarity percentage analysis (SIMPER) was also conducted to determine which taxa were most responsible for the differences between ER-positive and ER-negative sites (42). The above statistical tests were performed using PAST 2.16. Site ER classification for data analyses was based upon water filter results for sites visited in 2008, and on fish and amphibian ER results for sites visited in 2009. Separate Fisher's exact tests were performed to test for differences in positivity between sites, fish and amphibians, and fish-feeding guilds (R Development Core Team 2012). All results were considered to be significant at α=0.05, and all P-values were corrected for multiple tests.

A total of 587 fish, representing 13 genera and at least 17 species, and a total of 351 amphibians, representing 10 genera, were collected (Tables 1–4). In 2008, ER-positive water filters were recovered from Otinibi, Danfa, Teiman, Afiaman, and Nsakina. ER-positive specimens were found at all sites visited in 2009 (Table 5). ER positivity was not found to be associated with fish or amphibian community structure in 2008 (Figs. 3 and 4), and this was verified by ANOSIM (fish: P-value=0.672; amphibian: P-value=0.294). Combined amphibian data from 2008 and 2009 also failed to show a significant association with ER positivity (P-value=0.631; Fig. 5). However, when fish community data from 2008 and 2009 were combined, there was a significant difference in fish communities between ER-positive and ER-negative sites (P-value=0.031; Fig. 6). Weija was eliminated from these analyses because the different sampling protocol for that site did not allow for collection of community data. A SIMPER analysis on combined fish data showed that Hemichromis bimaculatus, Sarotherodon sp., and Barbus sublineatus were responsible for 47% of the dissimilarity between fish communities in ER-positive and ER-negative sites. H. bimaculatus (Fig. 7) was the taxon responsible for the most dissimilarity (22%) and was present in 80% of ER-positive sites versus 11% of ER-negative sites. However, H. bimaculatus specimens did not show a significant difference in ER positivity when compared to other fish species (P-value=0.771).

A subsample of extracted DNA from kidney and intestine samples was spiked with M. ulcerans DNA to test for PCR inhibition. High levels of inhibition were observed in intestine samples, but no evidence of inhibition was seen in kidney samples. Use of a MO BIO clean-up kit successfully removed PCR inhibition.

Of the two methods used for DNA extraction, the Käser method was 5,000 times more sensitive than the one-tube lysis method. With the one-tube procedure, the lowest detection limit using ER PCR was approximately 1,600 CFUs. With the Käser extraction procedure, the lowest detection limit was 0.32 CFUs. Addition of goldfish tissue homogenate did not decrease the detection limit, but additional non-specific binding was observed.

Negative controls consistently showed lack of contamination, and after the removal of inhibitors, positive controls consistently showed that the Käser extraction and PCR were successful. Overall, the mean ER positivity for all specimens was 54% (128/238). ER positivity was 39% (25/65) for fish, 59% (96/162) for tadpoles, and 64% (7/11) for adult frogs. Positivity rates for all specimens from a single site ranged from 13% at Lake Weija to 76% at Sarpeiman. However, Weija was removed from Fisher's exact tests due to the differing sampling method. Combined fish and amphibian counts showed a significant difference in the number of ER-positive specimens based on site (P-value=0.0164). As Sarpeiman showed the greatest positivity, this site was examined in a further analysis. Removing Sarpeiman (P-value=0.0523) indicated that this site was responsible for the significant result. A separate analysis indicated that overall, there was no difference in positivity rates between fish and amphibians (P-value=0.106). There was not a significant (P-value=1) difference in ER positivity among fish-feeding guilds (predator, insectivore, omnivore, and planktivore).

Of the 128 positive fish and amphibian specimens, 118 were identified as positive based on DNA from the intestine. Five specimens had positive results from both the intestine and the kidney, and five had positive results only from the kidney.

Twenty-seven external surface swabs were analyzed, and four of these were ER positive. Of the 27 total swabs, 14 came from fish or amphibians in which either the kidney or intestine was ER positive. Of these 14, only 1 external swab was also positive. The other three ER-positive external swabs came from fish and amphibian specimens in which the kidney and intestine were ER negative.

VNTR typing was attempted for all ER-positive samples, but only three of the tissue samples and no swab samples were successfully VNTR typed. Two were determined to be M. liflandii, a mycolactone-producing mycobacterium pathogenic to frogs. These were recovered from the intestine of a Hyperolius tadpole from Kwashikuma and the intestine of a predatory cichlid, H. bimaculatus from Djorse. One specimen, an adult frog from the genus Leptopelis, collected at Otuaplem, was confirmed to be positive for M. ulcerans. This is the first known finding of M. ulcerans in an adult amphibian. VNTR typing of water filters collected in 2008 confirmed M. ulcerans presence on filters from Otinibi and Nsakina and the presence of other mycolactone-producing mycobacteria on filters from Teiman. Filters from all other sites visited in 2008 were ER-negative, and therefore were not VNTR typed.