In June 2004 and August 2005, we sampled 27 water bodies associated with human communities in southern Ghana (Figure 1). The water bodies were located within or very near (<100–200 m) each community of housing structures and were routinely used for daily domestic purposes and reflect habitats of routine human exposure. These water bodies were chosen after discussions with community members who directed us to the main water source for drinking water, recreation, domestic washing, irrigation, or bathing for that community. Six of these sites were sampled in both years, providing information on annual variation: Afuaman, Amasaman, Abbeypanya, Afienya, Odumse, and Weija. Human BU case data for the years 2003–2005 were provided by the Ghana Ministry of Health and used to classify communities into 2 site types: 15 BU–endemic (BU+) and 12 BU–nonendemic (BU–). A site was classified as a BU+ type if at least 1 case of BU had been reported during the 3-year period.

Within each water body, two 10–20-m transects were measured parallel to the shoreline and positioned through the dominant macrophyte community. Along each transect, we randomly placed two 1-m² polyvinyl chloride quadrats and collected invertebrates by sweeping within the quadrat with a 500-µm mesh dip net. The quadrats floated on top of the water and delineated 1 m² of area to be sampled by using an aquatic dip net designed to capture the aquatic life stages of invertebrates. Three sweeps of the dip net were performed from the water surface to the bottom substrate for comprehensive sampling of specimens in the water column. All contents were washed through a 500-µm sieve and preserved in 100% ethanol for laboratory identification and PCR. The 2 quadrats were combined into 1 composite sample.

Samples were analyzed in a 2-step procedure so that an initial screening reduced sample numbers. Small invertebrates were analyzed in pools of 3–15, whereas larger specimens were tested individually. DNA was extracted by using a protocol adapted from Lamour and Finley (27). Samples were ground and vortexed in 400 μL of lysis solution (100 mmol/L Tris, pH 8.0), 50 mmol/L EDTA, 500 mmol/L NaCl, 1.33% sodium dodecyl sulfate, and 0.2 mg/mL RNase A) and 1 g of 1.0-mm glass beads (Sigma-Aldrich, St. Louis, MO, USA), then centrifuged. After 150 μL of 5 mol/L potassium acetate was added, each sample was incubated overnight at –20° C. After a 30-min centrifugation, supernatants were transferred to new tubes containing 0.66 mol/L guanidine hydrochloride in a 63.3% ethanol solution. The samples were added to a spin filter (MO BIO Laboratories Inc., Carlsbad, CA, USA) in a 2-mL microcentrifuge tube (MO BIO Laboratories Inc.). The flow-through was discarded and the filter was rinsed first with 500 μL of wash solution (10 mmol/L Tris, pH 8, 1 mmol/L EDTA, 50 mmol/L NaCl, 67% ethanol) and then with 500 μL of 95% ethanol. The spin filters were dried by centrifugation and transferred to new 2-mL microcentrifuge tubes, immersed in 200 μL elution solution (10 mmol/L Tris, pH 8), and incubated at room temperature for 15 min. The DNA was eluted and stored at –20°C.

Presumptive identification of M. ulcerans in invertebrates was based on detection of the enoyl reduction domain (ER) in mlsA that encodes the lactone core of the mycolactone toxin, the major virulence determinant of M. ulcerans. All samples were screened for the presence of the ER gene, which has been evaluated for M. ulcerans specificity in a companion study that used a multitiered PCR approach (18). Amplification of the ER gene was achieved using a 50-μL reaction mixture containing 1 μL each of forward and reverse primer (15,18), 10 μL 5× Go Taq reaction buffer (Promega, Madison, WI, USA), 1 μL 10 mmol/L PCR nucleotide mix (Promega), 31.7 μL double-distilled water, 1.6 units Go Taq polymerase enzyme (Promega), and 5 μL DNA template. Cycling conditions began with an initial denaturation at 94°C for 5 min, 35 cycles of 94°C for 1 min, 58°C for 45 seconds, 72°C for 1 min, and a final 10-min extension at 72°C. The amplified DNA was subjected to gel electrophoresis by using a 1.5% agarose gel, and band sizes were compared by using a 1-kb DNA ladder (Invitrogen, Carlsbad, CA, USA). PCR products of appropriate size were cloned into the pCR2.1 Topo vector (Invitrogen) and sequenced by using an ABI 3100 automated genetic analyzer (Applied Biosystems, Foster City, CA, USA).

Using all invertebrate data, we initially evaluated differences between site types (i.e., BU+ vs BU–) by comparing total abundance and percentage composition. Only those taxa that represented >3% of total invertebrates collected from all sites were used for subsequent statistical analyses because some taxa were so rare that any comparisons would limit meaningful conclusions. However, because we were interested in evaluating Hemiptera known to bite humans, the families Belostomatidae, Naucoridae, and Nepidae also were included, although each represented <2% of total collections.

To compare abundance differences between site types, t tests were used after data were log + 1 transformed to meet the assumptions of normality and equal variances. For percentage composition differences, data were arc-sine square root transformed, but they still did not demonstrate a normal distribution, so the nonparametric Wilcoxon/Kruskal-Wallis rank sum test was used. Because multiple tests were performed, it was necessary to calculate a Bonferroni adjusted α (and corresponding p value) of 0.006 to assist in interpreting statistically significant differences. However, to evaluate the biological meaning of these multiple tests, Cohen d effect size (and 95% confidence intervals) was calculated with Hedges adjustment (28). To compare overall ER positivity proportions between BU+ and BU– sites, a t test was used after data were arc-sine square root transformed. Lastly, we evaluated correlations between total biting hemipterans (and each individual family) and ER positivity using Spearman rank correlations with a Bonferroni adjusted α = 0.008. This nonparametric test was used after attempts to transform the data for normality and homogeneity of variances failed.

Of 22,832 invertebrates collected, ≈50% came from each group of BU+ and BU– site types (Technical Appendix). A total of 85 taxa were represented among all sites: 80 taxa were collected from BU+ sites compared with 71 from BU– sites. The abundance of specific taxa was not consistent between site types, indicating that the invertebrate communities were highly variable. This variability was confirmed in statistical analyses comparing the most abundant taxa (>3%) with substantial effect size variation within and among taxa (Technical Appendix). The invertebrates found in greatest abundance were 2 families of Diptera (i.e., Chironomidae and Culicidae), 1 family of Ephemeroptera (Baetidae), and several Crustacea. More than 300 individuals of some families of Hemiptera, Coleoptera, and Odonata were encountered (Technical Appendix). The biting Hemiptera were usually rare. For instance, 55 Naucoridae in total were collected, which was about 0.2% of all invertebrates sampled (Technical Appendix).

Insects made up the greatest percentage of the invertebrates collected from BU+ sites but were nearly equivalent to the Crustacea in BU– sites. In BU– sites, Anura made up a relatively higher percentage, but most (1,231 of 1,303 individuals) were from a single site (Figure 2; Technical Appendix). The Crustacea were most often represented by copepods, ostracods, and shrimp (Atyidae); fewer shrimp were collected from BU+ sites. Most shrimp were from BU– sites Adumanya (197) and Keedmos (120). Further, in BU– sites the large copepod abundance occurred primarily at Odumse, where 1,723 were collected from a total 1,884 (Technical Appendix). Insects were reduced by 40% in BU– sites compared with BU+ sites (Figures 2, 3). When individual insect orders were compared, the Ephemeroptera (mayflies) and Diptera (true flies) made up the greatest percentages of insects in both BU+ and BU– site types (Figure 3; Technical Appendix).

When the abundance and percent composition of dominant taxa were statistically compared between BU+ and BU– site types, there were no significant differences for any taxa (Technical Appendix). However, the effect size varied greatly, reflecting a need to collect from more sites in future studies. On average, the Chironomidae (midges) made up the greatest percentage of the invertebrate communities, representing 9%–20% of the total, while the Baetidae (mayflies) ranged from 6% to 15% and the Culicidae (mosquitoes) from 2% to 5%. The biting Hemiptera made up a very small percentage of the dominant invertebrate communities, with Naucoridae <0.5%, Belostomatidae <2%, and Nepidae <0.3% (Technical Appendix).

Presumptive identification of M. ulcerans from a total of 1,032 invertebrate sample pools tested found no significant difference between BU+ and BU– site types (Technical Appendix). Furthermore, there was no detectable pattern of invertebrate taxa ER positivity among sites, indicating that no single taxon was more often likely to be positive at a particular site. The number of ER positive taxa that were detected at any site ranged from 0 to 15 and 0 to 6 in BU+ and BU– sites, respectively (Figure 4). Clearly, not all BU+ or BU– sites had ER positive invertebrates. There were 6/15 BU+ sites without a single taxon positive compared with only 3/12 BU– sites (Figure 4).

Taxon-specific ER positivity was highly variable, and percentage positivity ranged from 0% to 100% among taxa (Technical Appendix). There were 26 taxa positive from BU+ site types compared with only 18 from BU– sites. Only 2 taxa were positive in BU– and not in BU+ sites, and for those taxa, <5 samples were tested from the BU+ type. When only those taxa with >5 samples tested were compared, no observable pattern in ER positivity was apparent among sites or taxa. The most abundant taxa did not always have the greatest ER positivity. For instance, positivity of Chironomidae (19.5% of all invertebrates) was only about 7%, even though positivity of Caenidae (<2% of all invertebrates) ranged from 6% to 17% (Technical Appendix). For taxa with >5samples tested from either BU+ or BU– sites, the ER positivity was >20% for 5 taxa and from 10% to 20% for 12 taxa (Technical Appendix). The biting Hemiptera had neither the highest nor consistently higher ER positivity compared with more abundant taxa (Technical Appendix). Fifteen taxa with >5 samples tested had 0 positivity. These taxa represented all invertebrate functional feeding groups (e.g., predators, shredders, scrapers, collector-gatherers, and filterers).

No significant correlation was found between mean ER positivity and total biting Hemiptera (r = 0.25; p = 0.218) or any individual family: Belostomatidae (r = 0.31; p = 0.118), Naucoridae (r = -0.03; p = 0.850), and Nepidae (r = 0.37; p = 0.060). These results confirmed that biting Hemiptera were not significantly associated with the pathogen in the environment.