Global and local environmental changes as drivers of Buruli ulcer emergence

Introduction Buruli ulcer (BU) is a serious necrotizing cutaneous infection caused by Mycobacterium ulcerans [1]–[7]. Before the causative agent was specifically identified, it was clinically given geographic designations such as Bairnsdale, Searles, and Kumasi ulcer, depending on the country [8]–[11]. BU is a neglected emerging disease that has recently been reported in some countries as the second most frequent mycobacterial disease in humans after tuberculosis (TB) [12]–[14]. Large lesions often result in scarring, contractual deformities, amputations, and disabilities [2]–[4], [7], [14]–[22] (Fig. 1). Approximately 80% of the ulcers are located on the limbs, most commonly on the lower extremities yet some variation exists [3], [13], [23], [24]. In Africa, all ages and sexes are affected, but most cases of the disease occur in children between the ages of 4–15 years [5], [13], [17], [25]–[28]. 10.1371/journal.pntd.0000911.g001 Figure 1 Buruli ulcer on leg and contractual deformity on wrist and hand. (Photo by R. Kimbirauskas). BU is a poorly understood disease that has emerged dramatically since the 1980's, reportedly coupled with rapid environmental change to the landscape including deforestation, eutrophication, dam construction, irrigation, farming (agricultural and aquaculture), mining, and habitat fragmentation [3]–[7], [29], [30]. BU is a disease found in rural areas located near wetlands (ponds, swamps, marshes, impoundments, backwaters) and slow-moving rivers, especially in areas prone to flooding [3], [4], [23], [27], [29], [31]–[36] (Fig. 2). Cases have been reported from at least 32 countries in Africa (mainly west), Australia, Southeast Asia, China, Central and South America, and the Western Pacific [3], [6], [20], [28], [37], [38] (Fig. 3). A number of cases have been reported in non-endemic areas of North America and Europe as a sequel to international travel [20], [39]–[42]. 10.1371/journal.pntd.0000911.g002 Figure 2 Typical Buruli ulcer riverine endemic sites in Ghana and Benin, respectively. (Photos by M. E. Benbow and M. McIntosh, respectively). 10.1371/journal.pntd.0000911.g003 Figure 3 A global map representing countries that have reported cases of Buruli ulcer disease as of 2009 (WHO). Buruli ulcer disease is often referred to as the “mysterious disease” because the mode of transmission remains unclear, although several hypotheses have been proposed. The objectives of this article are to: 1) review the current state of knowledge on the ecology and transmission of M. ulcerans, 2) discuss traditional and non-traditional methods for investigating transmission, and 3) suggest an intellectual framework for establishing criteria for transmission. Methods Data Sources and Search Strategy Selection of the publications cited was based on the following approaches: 1) Direct knowledge of the authors of this manuscript regarding their background in the field of Buruli Ulcer research and knowledge of key papers and unpublished data; 2) Online search engines for Buruli Ulcer and Mycobacterium ulcerans (predominantly PubMed, ISI Web of Knowledge, Web of Science, Centers for Disease Control (CDC); 3) Knowledge in the field of Buruli Ulcer research in that three of the authors (Merritt, Small, Johnson) are on the WHO Technical Advisory Committee for Buruli Ulcer in Geneva, Switzerland; 4) Review of the following websites: Buruli ulcer disease maintained by WHO in Geneva, Switzerland (http://www.who.int/buruli/en), The Buruli Ulcer Disease Ecology Research Consortium (BUDERC) (https://www.msu.edu/~budiseco/index.html); and UBS Optimus Foundation (http://www.stopburuli.org). Results and Discussion The Pathogen M. ulcerans is a slow-growing environmental mycobacterium that can be isolated from primary lesions after a 5–8 week incubation period, although up to 6 months may be required [43], [44]. M. ulcerans falls into a group of closely related mycobacterial pathogens which comprise the M. marinum complex. The M. marinum complex contains mycobacterial species pathogenic for aquatic vertebrates and includes M. marinum (fish), M. pseudoschottsii (fish) and M. liflandii (frogs) [45]–[48]. All of these species are characterized by slow growth rates and low optimal growth temperatures [49]. From a genomic standpoint, the species in the M. marinum complex can be considered a single species based on the fact that they share over 97% identity in the 16sRNA gene sequence [50]. However, practical considerations have led to the establishment of separate names based on differences in host tropism and pathogenesis analogous to other mycobacterial groupings, such as the M. avium and M. tuberculosis complexes. Genomic analysis suggests that M. ulcerans evolved from an M. marinum-like ancestor [21], [51] through the acquisition of a large virulence plasmid and accumulation of multiple copies of insertion sequences, IS2404 and IS2606. The genome has undergone considerable reductive evolution through a number of mutational events including transposon insertion. As a result, the genome has accumulated over 700 pseudogenes [21], [52]. Although it has been reported that micro-aerophilic conditions enhance the growth of M. ulcerans in the BACTEC system [53], the M. ulcerans genome strain lacks both nitrate and fumarate reductase systems, suggesting that M. ulcerans is considerably handicapped in the ability to grow under low oxygen conditions compared with M. marinum. The reported discrepancy in the oxygen requirements of M. ulcerans may be due to strain differences and requires closer investigation. A mutation in crtI, a key gene in the pathway for carotinoid biosynthesis, is suggested to compromise the ability of M. ulcerans to survive in direct sunlight [52]. A number of genes in ion transport and lipid biosynthesis have been lost and the repertoire of PE, PPE genes are considerably reduced compared with M. tuberculosis or M. marinum. Taken together, these results suggest that M. ulcerans is undergoing adaptation to a different and narrower niche than M. marinum. This idea has recently gained support from experimental work in which Medaka fish were infected with M. marinum and M. ulcerans. In these studies, M. marinum produced a lethal infection in Medaka, whereas M. ulcerans was not pathogenic and declined over a 23-week infection period (L. Mosi, unpubl. data). The most important phenotypic characteristic of M. ulcerans is the low optimal growth temperature and the extremely restricted growth temperature range. M. marinum exhibits growth between 25–35°C, although the optimal growth temperature is 30–35°C [54], [55] and many M. marinum isolates are capable of growth at 37°C. In contrast, growth of M. ulcerans strains under laboratory conditions is characterized by a remarkably narrow temperature range between 28–34°C and optimal growth of most strains is found between 30–33°C [56]. The restricted growth temperature of M. ulcerans is thought to play a substantial role in the pathogenesis of BU by limiting infection to the skin. The organism has never been isolated from internal organs of human patients or from bone in cases of osteomylelitis, or from the internal organs or blood of experimentally infected animals [51], [57]–[59]. It has been recently reported that many isolates of M. ulcerans survive at 37°C for 13 days, although numbers decline after the first few days. No one has isolated or derived a strain capable of growth at 37°C [60]. The characteristic pathology of BU is mediated by a polyketide-derived macrolide exotoxin called mycolactone, which is cytotoxic and immunosuppressive [51], [61], [62]. Because of the large metabolic cost of producing mycolactone, it is likely that mycolactone plays an important role in the survival and growth of M. ulcerans in its environmental niche. Ecology and Distribution of the Pathogen and Disease Detecting M. ulcerans in the environment The slow growth rate of M. ulcerans and the complex mix of many faster growing bacteria and fungi in environmental samples have prevented direct culture on artificial media of M. ulcerans from the environment. A major breakthrough in environmental studies occurred with the development of the first PCR probes for M. ulcerans based on detection of IS2404 by Ross et al. [63]. This technique was rapidly adopted by a number of laboratories leading to identification of M. ulcerans DNA in environmental samples including detritus, soil, biofilms, water filtrates, fish, frogs, snails, insects and other invertebrates [18], [35], [64]–[75]. Although IS2404 PCR has become the gold standard for clinical diagnosis of Buruli ulcer, there are several caveats in applying these methods to environmental samples. First, PCR detects DNA, not intact organisms. The death of infected organisms will lead to the release of M. ulcerans DNA into the environment where it may stick to a number of substrates. Although in two different countries in Africa, Williamson et al. [67] found M. ulcerans DNA in 9.7% (8/82) of water filtrant samples and Vandelannoote et al. [59] found 7.7% (1/13) water samples positive for M. ulcerans, the significance of these small quantities of M. ulcerans in an environmental sample is difficult to evaluate. In southeastern Australia, M. ulcerans also has been detected in a range of environmental samples. Recently, Fyfe et al. [76], reported that 30% of selected samples including detritus, plant material, suspended solids, and soil collected from one highly-endemic area were weakly positive by quantitative PCR. However, in a low endemicity area, only 4/156 (3%) of samples (2 soil, 2 terrestrial plant) were positive. Interpretation of results from environmental PCR is complex. PCR methodology detects DNA, but it does not provide definitive proof for the presence of intact bacteria in a matrix. DNA bound to the surface of potential vectors in the water column also will be detected. However, the successful culture of M. ulcerans from an aquatic water bug collected in Benin [71] provides definitive evidence for the presence of M. ulcerans in an aquatic invertebrate. This considerable achievement was based on earlier observations using IS2404 PCR that implicated aquatic water bugs as possible reservoirs or vectors of M. ulcerans [70]. Ecological associations with disturbed water bodies Until recently, a systematic and/or quantitative approach to the ecology of M. ulcerans in the environment has received little attention, despite the fact that nearly all epidemiological studies have associated disease outbreaks with villages in close proximity to human-disturbed aquatic habitats, including both standing and moving water bodies [7], [9]–[11], [19], [20], [25], [33], [77]–[80]. Increased BU incidence has been reported in association with: 1) unprecedented flooding of lakes and rivers during heavy rainfall [9], [16], [30], [37], [81]; 2) the damming of streams and rivers to create impoundments and wetlands [4], [9], [30], [37]; 3) resorts that modify wetlands [16], [30]; 4) deforestation practices and increased agriculture leading to increased flooding [4], [9], [18], [30], [37]; 5) construction of agricultural irrigation systems [4], [30], [81]; 6) rice cultivation [4], [9]; 7); alluvial, pit and sand mining operations [30], [37], [82]; and 8) population expansion, resettlement and migration closer to water bodies [9], [16], [18], [27], [30], [37]. Indeed, many water bodies associated with increased sedimentation and eutrophication have low dissolved oxygen concentrations that may enhance the growth of M. ulcerans [53]. Hayman [9] speculated that in Australia M. ulcerans enters surface waters through deforestation, erosion and run-off contamination. He suggested that populations of M. ulcerans were washed into aquatic habitats where environmental conditions facilitated growth and proliferation, much like an algal bloom. Because most infectious diseases have a strong correlation between infective dose and incubation period for disease, Hayman [9] speculated that slow growth of M. ulcerans might be required for the bacteria to achieve population numbers sufficient to produce infection and the appearance of disease. The way in which M. ulcerans could be washed down into these habitats has never been explained, but is consistent with other reports of increased BU outbreaks associated with deforested and heavily flooded African lands [20], [33]. Further, deforestation leads to lost riparian cover, resulting in increased water temperatures that may facilitate M. ulcerans growth at optimal temperatures of 30–33°C [11], [18], [20]. Associated sedimentation (e.g., turbidity) also would provide ultraviolet light (UV) attenuation and protection for M. ulcerans biofilm near the bottom substrates and on submerged plant surfaces as proposed by Merritt et al. [30]. It has been documented that UV lowers M. ulcerans cell viability [52], and thus deforestation and high-impact agriculture may promote increased nutrients, higher temperatures, UV attenuation and lower dissolved oxygen – environmental conditions that facilitate M. ulcerans growth. Because of the association with freshwater habitats, Eddyani et al. [83] hypothesized that freshwater plankton, specifically protozoans, may act as reservoirs for M. ulcerans, or may even facilitate the multiplication of the bacteria [18]. Although the former authors did not detect M. ulcerans DNA in free-living amoebae collected BU endemic areas in Benin, this area of research definitely warrants further investigation. Landscape ecology of the disease Buruli ulcer has been widely associated with proximity to aquatic habitats. The disease is rare in the savanna regions of West Africa and drier areas of Australia. Its presence in Australia is notably costal however, where water is often saline. This association between ecosystem ecology and disease has not been quantified. Rather, the association is most often anecdotal or related to specific human risk factors (e.g., wading, swimming, fishing, bathing, washing, farming, mining, etc.) in different countries and/or regional districts (see review below). To date, there have been few ecological studies focused on statistically determining why residence near certain water bodies is associated with BU, whereas the disease is absent along others [30], [67], [68]. For example, BU is highly associated with residence along several major river systems in both Benin and Ghana [12], [14], [20], [84], [85], whereas disease is essentially non-existent in communities within a few kilometers of Lake Volta, the largest water system in Ghana, as well as along the Mono River in Benin. Williamson et al. [67] recently found that in Ghana, PCR results suggesting that M. ulcerans and/or other mycolactone producing mycobacteria are widely distributed in water bodies in endemic and non-endemic villages. In these studies, however, the identification of endemic versus non-endemic sites was based on passive surveillance. A community was considered endemic if a case had been identified in the public health center in the past three years. A community that is not listed in the health center records, in association with a case of Buruli ulcer, was considered non-endemic. A preliminary survey to validate the non-endemic status of several communities in the GA district of Ghana through active surveillance showed that Buruli ulcer cases could be indentified in nearly all of the villages visited along the Densu River in the GA district (P. C. Small, unpubl. data). In areas where much of the disease is not reported, this can lead to significant error in the designation of “non-endemic.” There have been case control studies and observational reports of disturbed landscape associations with BU disease [29], [30], [86]; however, there have only been a few recent studies to statistically quantify landscape characteristics and relationships with disease [36], [79], [81], [87]. Duker et al. [79] found that arsenic levels in soil and gold mining were significant covariates related to increased disease risk in the Amansie West district of Ghana, while Wagner et al. [36], [81] addressed larger scale land use/land cover relationships using satellite imagery, GIS, and country wide BU data from Benin. In the latter studies, Wagner et al. [36], [81] reported highest disease in communities surrounded by an agriculture matrix, and thus deforestation, with abundant wetlands and other habitats that experience frequent flooding. These were low-lying areas with complex topography far removed from urban settings [36], [81]. In another country-wide study using GIS, Brou et al. [88] found that in Côte d'Ivoire, communities near landscapes of irrigated rice and other agriculture near dams used for irrigation were related to increased risk of BU. These studies confirm previous epidemiological studies and indicate that there are quantifiable relationships between landscape features and land use that are related to BU disease. It is also clear that communities involved with these activities are at high risk for disease, yet how specific activities are associated with transmission remains unresolved. Risk factors associated with Buruli ulcer disease Recently, Jacobson and Padgett [89] systematically reviewed the risk factors associated with M. ulcerans infection throughout the world and concluded that poor wound care, failure to wear protective clothing, and living or working near water bodies were commonly identified risk factors in most studies. However, a number of epidemiological studies have identified other potential risk factors associated with M. ulcerans infection and these are summarized in Table 1. For each specific risk factor investigated, it is stated as to whether or not there was an increased or decreased risk of infection reported, or if the factor was not considered a risk factor in the analysis. Several of the commonly reported risk factors showed few consistent associations depending on the country, type of analysis conducted, use of different case definitions, and based on the control populations used [89]. For instance, in a case-control study from Ghana, Aiga et al. [25] found that swimming in rivers on a habitual basis was a significant risk factor, whereas drinking, cooking, washing clothing and bathing were not. However, in another Ghanaian study, wading, bathing, and swimming were all confirmed to be significant risk factors for BU [77]. Two studies found a decreased risk of infection with mosquito net use, while another study found no association between bed net use and infection (Table 1). However, in a case control study performed in southeastern Australia, use of insect repellent was associated with reduced risk and the reporting of mosquito bites on the forearms and lower legs was associated with increased risk [90]. Despite the association with water contact, fishermen were not found to be at high risk for the disease (Table 1). Although a review of these potential risk factors suggests that transmission of M. ulcerans might occur through direct inoculation of bacteria into the skin via contact with environmental sources, insect bites or trauma, it was clear that additional comparative studies are required to clarify the potential modes of transmission of M. ulcerans [89]. 10.1371/journal.pntd.0000911.t001 Table 1 A summary of reported risk factors associated with infection Mycobacterium ulcerans. Country Risk Factor(s) Increased Risk of Infection Decreased Risk of Infection Not Considered a Risk Factor Citation Ghana 1) Arsenic-enriched drinking water (from mining) X Duker et al. (2004) Ghana 1) Exposed skin2) Bednet and mosquito coils use3) Insect bites, cuts, scratches, and other wounds4) Exposure to riverine areas (wading and swimming)5) Association between BCG and vaccination or HIV infection6) Not wearing protective clothing7) Fishing XXX XXXX Raghunathan et al. 2005 Ghana 1) Age 2–14 years of age2) Use of water for drinking, cooking, bathing, washing3) Association with agricultural activities4) Swimming in rivers XX XX Aiga et al. 2004 Benin 1) 5–14 years of age2) Unprotected water from swamps3) BCG-vacinated patients >5 years old4) Participated in agricultural activities5) Sex XXXX X Debacker et al. 2004, 2006 Benin 1) Mosquito bed net use2) Association with agricultural activities3) Improper wound care X X X Nackers et al. 2007 Cameroon 1) Living near cocoa plantation or woods2) Wading in swamps3) Wearing protective clothing while farming4) Association with agricultural activities5) Improper wound care6) Bed nets7) Mosquito coils8) Unprotected water sources9) Fishing XXX XX XXXX Pouillot et al. 2007 Cote d′ Ivoire 1) Age group2) Wearing protective clothing during farming activities3) Washing clothes4) Swimming5) Fishing XX XXX Marston et al. 1995 Australia 1) Wearing protective clothing2) Use of insect repellent3) Most patients > 60 years old4) Washing wounds after sustaining minor skin trauma5) Exposure to mosquitoes XX XXX Quek et al. 2007 Although there have been reports of a seasonal distribution in BU cases related to rainfall-influenced patterns of village waterbody usage [32], and by season in southeastern Australia [91], other studies have not shown this relationship [12]. Recording monthly trends for BU cases over a 3-year period in Benin, Sopoh et al. [12] found consistent average monthly BU case occurrence, without an apparent seasonal trend. However, country-wide data can obscure local variation in climate and the issue of seasonal trends needs to be more closely investigated at the local level. The unknown incubation period for Buruli ulcer, which may vary from 2 weeks to 7 months [92], [93], also makes it difficult to analyze seasonal factors with Buruli ulcer occurrence. Duker et al. [4], and more recently Marion et al. [94], discussed seasonal variations and M. ulcerans infections reported from different countries and concluded that there may be a temporal relationship between BU incidences and relatively dry periods; however, it also has been reported that M. ulcerans infections occurred mainly after flooding events [9], [16], [33], [34], [95]. Environmental Reservoirs and Transmission Africa Unlike leprosy and tuberculosis, which are characterized by person-to-person transmission, it is hypothesized that M. ulcerans is acquired through environmental contact. Direct human to human transmission of M. ulcerans is extremely rare. The one reported case occurred following a human bite [96]. In this instance it was hypothesized that the patient's skin surface was contaminated with M. ulcerans from an environmental source (e.g. swamps) and driven into the skin by the playmate's bite. Non-human mammals and reptiles have been tested in the environment without positive findings [95], and several arthropods (i.e., bedbugs, black flies, mosquitoes) in Africa associated with vectoring other disease agents tested negative in early studies [18], [32]. However, few organisms of each taxonomic group were tested in these studies, and insect sampling methods were neither systematically employed nor standardized. Buruli ulcer cases in wild and domesticated animals in Africa have not been reported [97]. Portaels and colleagues [70] were first to suggest that aquatic bugs (Hemiptera) might be reservoirs of M. ulcerans in nature, and recently they described the first isolation in pure culture of M. ulcerans from a water strider (Hemiptera: Gerridae, Gerris sp.) from Benin [71]. A survey study [18] based on detection of M. ulcerans DNA in aquatic insects (Hemiptera, water bugs; Odonata, dragonfly larvae; Coleoptera, beetle larvae) collected from African BU endemic swamps confirmed their earlier findings, and suggested that small fish might also contain M. ulcerans [66], [98]–[100]. Marsollier et al. [64], [66], [98]–[100] conducted a series of laboratory studies and demonstrated that M. ulcerans could survive and show limited replication within the salivary glands of biting aquatic bugs (Naucoridae: Naucoris cimicoides). In their experimental model they demonstrated that M. ulcerans could be acquired from feeding on inoculated insect prey (a blow fly maggot), transmitted to mice via biting; and that the infected mice subsequently developed clinical BU [66]. Although there has been some controversy regarding the interpretation of this work [68], [101], [102] and subsequent follow-up studies on tracing the pathogen through the bug [103], [104], Marsollier and colleagues concluded that biting water bugs belonging to the families Naucoridae (creeping water bugs) and Belostomatidae (giant water bugs) could be considered reservoirs, and most importantly could serve as vectors in the transmission of M. ulcerans to humans in nature. More recently, Mosi et al. [101] investigated the ability of M. ulcerans to colonize aquatic bugs (Belostomatidae) collected from Africa. Using a natural infection model in which M. ulcerans-infected mosquito larvae served as prey that were then fed to the predacious bugs, Mosi and colleagues confirmed Marsollier's finding that infected belostomatid bugs could become infected with M. ulcerans via feeding. However, they concluded that transfer of bacteria through feeding was most likely to have occurred through contact with the heavily colonized raptorial arms and other external parts of the belostomatid, rather than through saliva or contact with other internal organs as originally reported [66]. Together, these experiments indeed support the hypothesis that predaceous aquatic insects may play an important role in maintaining M. ulcerans within food webs in the aquatic environment [1], [30], [68], [70] but, as detailed below, their role in actual transmission to humans remains unclear. The role of other non-insect aquatic invertebrates as intermediate hosts or environmental reservoirs for M. ulcerans has been suggested by several authors [30], [66], [70], [73], [99], and recently confirmed in more field research [67], [68]. It was experimentally confirmed that aquatic snails could be transiently colonized by M. ulcerans after feeding on M. ulcerans-containing aquatic plant biofilms [64]. Aquatic plant extracts stimulated biofilm formation, and increased the uptake of labeled metabolites by M. ulcerans in laboratory experiments [65]. In the field, Kotlowski et al. [73] recorded M. ulcerans DNA in aquatic snails from endemic regions of Ghana and Benin, and other studies have found that average estimates of M. ulcerans increased by two orders of magnitude in detritus compared to water [72]. More recently, Marsollier et al. [104] described an extracellular matrix associated with the biofilm of M. ulcerans that may confer selective advantages to the mycobacteria in colonizing various microhabitats in the environment. Based on these studies and extensive environmental studies by Williamson et al. [67], it is evident that M. ulcerans DNA can be detected within biofilm on the plant surface, and as part of decaying organic matter (detritus) both of which serve as food for certain aquatic invertebrates and fish, suggesting reservoirs and movement throughout the aquatic food web. A conceptual model, expanded and modified from Portaels et al. [70], illustrating the potential reservoirs and movement of M. ulcerans within and among aquatic environments was detailed by Merritt et al. [30] and more recently by Marion et al. [94]. Basically, M. ulcerans has been reported from mud, detritus, water filtrants, and plant biofilms, thereby allowing grazing or filtering aquatic insects (e.g., midges and mosquito larvae) or other invertebrates (snails, crustaceans, plankton) to concentrate mycobacteria through their feeding activities. Then, predatory aquatic vertebrates (i.e., some fish) and invertebrates (e.g., true bugs, beetles and dragonfly larvae) feed on other invertebrate prey or small fish, serving to move M. ulcerans from prey to biting insects. Lastly, aquatic insects capable of flight, and birds that prey on fish and/or aquatic invertebrates may potentially disseminate M. ulcerans to other aquatic environments [30]. Although the potential for different aquatic invertebrates in Africa to serve as environmental reservoirs for M. ulcerans has been clearly demonstrated, direct transmission by biting water bugs, other than by purely accidental means appears very unlikely for the following reasons. First, in Africa M. ulcerans DNA has only been detected in invertebrates that are not hematophagous. Predatory semi-aquatic Hemiptera (i.e., Naucoridae, Belostomatidae, Notonectidae) mainly feed on invertebrates (aquatic insects, Crustacea, snails) by inserting their piercing mouth parts into their prey, injecting saliva containing proteolytic enzymes, and then imbibing the liquefied prey tissues [105], [106]. Most employ an ambush strategy, waiting motionless clinging to vegetation for unsuspecting prey (Belostomatidae), while others may actively swim and pursue their prey (Naucoridae, Notonectidae) [107], [108]. Adults of most species of semi-aquatic Hemiptera possess the ability to disperse by flight, but mainly at night, and end up being attracted to electric lights during the breeding season, often correlated with the lunar cycle. Because of this, they often find their way into houses by accident [107], [108]. However, the very low disease prevalence among children less than three years of age suggests that infection does not occur in the house. When humans accidently come into contact with the bugs in the water, on aquatic vegetation, or away from water, they can be bitten [109]. However, these bugs do not actively search for humans, they do not require a blood meal or protein source to mature their eggs, nor is there any evolutionary history suggesting or supporting a vectorborne/pathogen transmission or co-evolving host/parasite relationship in the semi-aquatic Hemiptera [107], [110]. Therefore, based on the biology and behavior of predaceous aquatic insects, biting humans appears to be a rare event associated with a purely defensive reaction of these bugs [109], [111]. It should be noted, however, that the causative agent of Chagas disease (Trypanosoma cruzi) in humans is transmitted by a terrestrial hemipteran (Reduviidae), but it is through fecal contamination and not by the bite of the bug. Also, in this case the habitat of the vector (bug) is closely tied to that of its host [112]. In general, field studies on the prevalence of biting aquatic invertebrates do not support the hypothesis that biting aquatic bugs are vectors of M. ulcerans in nature; however, a recent study by Marion et al. [94] in Cameroon identified several water bug families as hosts of M. ulcerans in a Buruli ulcer endemic area. However, in Marion et al. [93], only one endemic area and one non-endemic area were evaluated, suggesting no replication, and thus, a limitation to testing how variable M. ulcerans is among endemic versus non-endemic areas/villages. This makes it difficult to compare to studies by Williamson et al. [67] and Benbow et al. [68] where multiple replicate sites were evaluated to test for M. ulcerans variability in standardized ecological samples. Benbow et al. [68] conducted the largest field study to date that examined biting water bugs in 15 disease-endemic and 12 non-disease-endemic areas of Ghana, Africa. From collections of over 22,000 invertebrates, they compared composition, abundance and invertebrate-associated M. ulcerans positivity among sites, and concluded that biting hemipterans were rare and represented a very small percentage of invertebrate communities. When endemic and non-endemic areas were compared, there were no significant differences in hemipteran abundance or invertebrate-M.ulcerans positivity rates (by PCR) between the areas, and there were no significant associations between hemipteran abundance and overall invertebrate-M.ulcerans positivity. Thus, there is little field evidence to support the assertion that biting bugs are major vectors of M. ulcerans in nature. However, as concluded by Marion et al. [94], the detection of M. ulcerans in water bugs in a specific area could possibly be used as an environmental indicator of the risk of M. ulcerans transmission to humans. Australia In Australia, infection with M. ulcerans occurs at low-levels in the wet tropical north where the climate is similar to sub-Saharan Africa [113]–[115]. However, more than 80% of Australia's cases of Buruli ulcer in the past 15 years have been in the temperate southeastern state of Victoria [93]. In comparison to Africa, people in Victoria have less direct contact with the environment, yet in two well-described outbreaks, 1.2–6.0% of the entire resident population in the outbreak areas developed Buruli ulcer [35], [116]. Visitors may also be at risk, and in one case, contact with an endemic town for just one day appeared to be sufficient to develop Buruli ulcer up to 7 months later [35]. In attempting to understand possible modes of transmission, two competing models have been proposed to explain this pattern of limited environmental contact, brief exposure, and high attack rates. Hayman [9] proposed that transmission by aerosol could partially explain outbreaks of M. ulcerans disease and an opportunity arose to test this hypothesis during a three year period when a large cluster of Buruli ulcer cases occurred in East Cowes, Phillip Island. This outbreak was significant in that only part of the town was affected, and there was a newly created wetland and a golf course at the center of the affected area. The golf course used a mixture of ground water and recycled water for irrigation and run-off from the golf course was likely to have drained towards the new wetland, connecting the two systems. Many of the case-patients lived close to the wetland or the golf course, supporting the concept of transmission by drifting aerosols from contaminated irrigation water [116]–[119]. Initially, no method existed for detection of M. ulcerans in environmental samples. However, as part of the outbreak investigation, Ross et al. [63] discovered IS2404, a high copy number insertion sequence in M. ulcerans. A PCR method using IS2404 as a target sequence has rapidly become the diagnostic method of choice for Buruli ulcer due to its high sensitivity, specificity, and its speed compared with traditional culture methods. IS2404 PCR was then adapted for application to environmental samples, and positive results were obtained from the wetland and golf course irrigation system-the first direct evidence that M. ulcerans DNA is present in environmental samples. IS2404 PCR also can be used as a preliminary test for the presence of M. ulcerans in Africa, but aquatic mycobacteria associated with disease in fish and West African clawed frogs (Xenopus tropicalis) also contain IS2404. For this reason, IS2404 lacks sufficient specificity for use as sole criteria for M. ulcerans in Africa. To date, there is no evidence from Australia of the presence of IS2404 in any other environmental mycobacterium. The above findings supported the hypothesis that the golf course irrigation system and nearby wetland at Phillip Island had become contaminated with M. ulcerans, although transmission by aerosol itself was not directly assessed [72], [120]. Drainage of the wetland, reduction in recycled water use, cleaning of the irrigation equipment at the golf course, and subsequent separation of ground water from recycled water were collectively associated with fewer cases in the following years. Buruli ulcer linked to Phillip Island is now rare; however, disease activity in at least one other Victorian endemic area also declined over a similar time frame without a specific intervention, making it difficult to conclude that the environmental alterations made at Phillip Island were directly responsible for the decline in cases. During the same period several possums (Australian native tree-dwelling marsupials) with Buruli ulcer were identified at Phillip Island [18], the significance of which will be discussed further below. In 2002, a new outbreak commenced in a small town on the Bellarine Peninsula about 60 km to the west of Phillip Island, also in coastal Victoria, southeastern Australia. More than 100 people who either live in or have visited Point Lonsdale have now been diagnosed with Buruli ulcer [35]. Several other towns on the Bellarine Peninsula have been linked to cases, but in lower numbers thus far. Although Point Lonsdale also has a golf course, it is not centrally located, and does not use recycled water. In 2004, intense local mosquito activity seemed to be associated in time with new cases of BU and Buruli lesions were observed on ankles and elbows, and on the back where gaps in clothing could allow access for mosquitoes. In one case, Buruli ulcer developed on the ear of a child who was only briefly present in the outbreak area. The child's mother suspected a mosquito bite as the initiating event [35]. These observations led to a series of studies aimed at assessing a possible role for mosquitoes in the transmission of M. ulcerans. Using an improved real-time quantitative IS2404 PCR environmental screening method [74], more than 11,000 adult mosquitoes captured at Point Lonsdale were tested, and M. ulcerans DNA was identified in or on an estimated 4.3/1,000 mosquitoes. Most PCR positive mosquito pools were Aedes camptorhynchus (Thomson), the most common species on the Bellarine peninsula; however, M. ulcerans DNA also was detected in one or more pools of four other species [35]. PCR amplification and sequence analysis of one variable number tandem repeat (VNTR) locus confirmed that mosquitoes were carrying M. ulcerans DNA, indistinguishable from that of the human outbreak strain [74], [121]. A review of notifiable diseases in Victoria in the period 2002-8, demonstrated a statistically significant correlation between notifications of Buruli ulcer and Ross River Virus/Barmah Forest Virus infections (RRV/BFV) – both of which are transmitted by mosquitoes – but there was no correlation with any other non-mosquito borne notifiable disease [122]. A case-control study, conducted on the Bellarine Peninsula including Point Lonsdale, showed that the odds of being diagnosed with Buruli ulcer were at least halved in respondents who frequently used insect repellent, wore long trousers outdoors, and immediately washed minor skin wounds, and were at least doubled for those who received mosquito bites on the lower legs or lower arms. In a multivariate model, after adjusting for age and location, use of insect repellent and being bitten by mosquitoes on the lower legs were found to be independently associated with Buruli ulcer risk [90]. In laboratory experiments using a green fluorescent protein (GFP) labeled M. ulcerans mutant, in which GFP was linked to the mycolactone toxin polyketide synthase promoter, it was shown that when fed as a single pulse to live mosquito larvae, M. ulcerans-GFP was able to persist through 4 larval instars in the mouth parts and midgut of the insect. This was not observed with a closely related M. marinum-GFP mutant that did not produce mycolactone [123]. This permissive effect of mycolactone on allowing M. ulcerans to selectively colonize aquatic insects also was observed in experiments using aquatic water bugs [66], [100], [104]. However, other investigators found equal colonization with mycolactone negative and wild type strains [101], and this earlier selective effect was not observed in a study on M. ulcerans colonization of mosquitoes conducted by Wallace et al. [124].The latter study found a nearly 100% infection rate was obtained when wild type M. ulcerans, an isogenic mycolactone-negative M. ulcerans, and M. marinum (a non-toxin producing potential progenitor of M. ulcerans) were used to infect mosquito larva. These findings are in line with the fact that mosquito larvae do not discriminately feed on specific bacteria or other foods unless ingestion is mediated by particle size [125], [126]. Differences in experimental conditions and bacterial strains used may help to explain these conflicting findings. Collectively, the above transmission research conducted in southeastern Australia lends support to mosquitoes as being a possible vector of the pathogen for Buruli Ulcer disease in this region of the country (see Bradford Hill guidelines for a critical assessment, below). More recently, it also has been discovered that that 38% of ringtail possums (Pseudocheirus peregrinus (Boddaert)) and 24% of brushtail possums (Trichosurus vulpecula Flannery) captured at Point Lonsdale had laboratory-confirmed M. ulcerans skin lesions and/or M. ulcerans PCR positive feces (Fyfe et al. [76]). The exact sequence of events linking mosquitoes, humans, contaminated possum excreta and infected possums has yet to be determined, but direct or indirect mosquito transmission from a possum reservoir presents a parallel model with aerosol transmission from contaminated environmental water sources. Neither the aerosol nor mosquito transmission hypothesis in temperate Australia is incompatible with transmission by direct contact with the environment or by other vectors not yet examined. Future research on the biological relationships within each model will help to resolve the relative probability and plausibility of either mode. Criteria for Establishing the Role of Insect Vectors of M. ulcerans Stringent criteria exist in biomedical research for indicting the roles of living agents as biologically significant reservoirs and/or vectors of pathogens. The application of these criteria to the transmission of M. ulcerans presents a significant challenge. The above review reveals that various routes of transmission may occur, varying amongst epidemiological setting and geographic region, and that there may be some role for living agents as reservoirs and as vectors of M. ulcerans, in particular aquatic insects, adult mosquitoes or other biting arthropods. It is also clear that the exact mode of transmission, if indeed there is a single mode, remains unknown. We briefly discuss the process by which a vector is incriminated to the point of as much certainty as is possible, and then discuss the application of this process to indictment of insect vectors for transmission of M. ulcerans. If Buruli ulcer is a vectored disease, intervention might be designed to reduce the possibility of transmission since there are possibilities other than suppressing vector populations. Vector incrimination traditionally involves satisfying a set of criteria analogous to Koch's postulates, summarized by Barnett [127] as follows: (1) the vector must be shown to acquire the pathogen from an identified source such as an infected vertebrate host or other reservoir, and thereafter become infected with the pathogen; (2) the vector must be shown convincingly to have close associations with infected hosts, including humans, in time and space; (3) individual vectors collected in endemic settings must repeatedly be found infected with the pathogen; and (4) efficient transmission to competent vertebrate hosts must be demonstrated experimentally, under well controlled conditions, by individual vectors, such as by bite or other means of direct contact. These criteria accommodate mechanical transmission if infection includes recovery of the pathogen from the vector's body, without making any assumptions about replication of the pathogen on or in the vector. Further, they do not preclude the possibility of parallel modes of transmission other than vectors. For example, the causative agent of plague, Yersinia pestis, has a flea vector and during sporadic outbreaks is transmitted by flea bites; but these bacteria also are transmitted during epidemics in aerosols generated by sneezing of pneumonically-infected humans or animals such as cats, which is probably the predominant mode of transmission in epidemics [128]. Similarly, human infection with the causative agent of tularemia, Franciscella tularensis, may occur through direct contact with contaminated water, by aerosols, by contact with blood or infected tissues of animals, or by bites of infected ticks, deer flies, or mosquitoes [129], [130]. The causative agent of Rift Valley fever, a Phlebovirus in the family Bunyaviridae, is transmitted amongst infected vertebrate reservoirs (mainly ungulates) by mosquitoes; however, many human infections occur upon exposure to infected animal blood at the time of slaughter, by aerosolization, as well as by mosquito bites [131]. Another useful illustration is that of Chlamydia trachomatis, the causative agent of trachoma, where the transmission to human eyes has been definitively associated with contact by Musca sorbens flies (Diptera: Muscidae) that breed in human feces in various parts of Africa [132]. Despite this observation, other mechanisms of transmission for this disease are known, such as person-to-person contact with contaminated fingers and wash towels [133], [134]. In two of the above examples (plague and Rift Valley fever), the pathogen has a close biological relationship with, and dependency upon, insect vectors; neither pathogen could persist in nature without infecting their respective vectors. For tularemia and trachoma, vectors are not essential to pathogen persistence in nature, even though fly control in the latter case was shown to reduce incidence of disease in humans [135]. However, it is unlikely in the case of tularemia and trachoma that even highly effective fly control could eliminate human infection in endemic areas owing to other modes of transmission [133]. Therefore, using a critical approach to address the issue of insect vector incrimination for M. ulcerans, one must be cognizant of the relative biological dependency of this bacterium on an insect vector, and the potential for facultative and facilitative relationships between these bacteria and various insect “hosts” to exist which may be ancillary or even spurious to the essential and normal transmission modes. The most thorough examination of the role of an insect vector for transmission of M. ulcerans stems from investigations of aquatic, predaceous Hemiptera (true bugs) as reviewed above, which go far in addressing and meeting Barnett's criteria. It is important to recognize that the vast number of studies of M. ulcerans in environmental samples provide qualitative, indirect evidence of M. ulcerans based on very sensitive methods for detecting M. ulcerans DNA. Such studies revealed repeatedly that natural infection by M. ulcerans in field-collected bugs occurred, but it was tempered by detection of M. ulcerans in many other aquatic insects [18], [67]. Thus, definitive incrimination of a single species or group of closely-related aquatic and semi-aquatic Hemiptera to the exclusion of other insects was not initially established. Other studies suggested natural contamination of the surfaces of these insects with M. ulcerans and suggested that M. ulcerans growth could occur as biofilms on the external appendages of such ‘bugs’ [101]. Thus, although aquatic and semi-aquatic Hemiptera and other insects found to harbor M. ulcerans in nature might provide habitat for the bacteria, along with numerous other living and non-living surfaces where biofilms could form [104], this is insufficient evidence for indicating an obligatory or even facultative vectorial role to these insects. Although the experiments reported by Marsollier et al. [64], [66], [98]–[100] suggested modest bacterial replication in internal tissues of bugs, acquisition of bacterial infection from a live source (infected fly maggots meant to simulate an infected prey item), and transmission to mice, this evidence does not establish natural infection coupled with transmission to humans. Finally, there has been no epidemiological association established between spatial and temporal distribution of contacts with aquatic Hemiptera, or bites by them, and development of Buruli ulcer in humans [68]. As reviewed above, the common understanding of the feeding habitats of aquatic and semi-aquatic Hemiptera does not include feeding on humans. More likely, infection in aquatic insects is associated with exposure to M. ulcerans in detritus and on biofilms formed on submerged materials, leading to a generalized distribution of M. ulcerans and M. ulcerans DNA in aquatic environments. In this particular scenario, despite the body of research on the topic, Barnett's criteria have not yet been fulfilled satisfactorily. The recent research by Wallace et al. [124], whilst firmly documenting growth of M. ulcerans in mosquito larvae and transtadial infection after the molt, showed that infection did not persist upon metamorphosis to the adult stage. Thus, the link between presence of M. ulcerans in aquatic environments in which larval mosquitoes are found and adult mosquito infection with M. ulcerans, was not confirmed experimentally. However, these studies did show that M. ulcerans DNA could be detected on surface components of some adult mosquitoes. This brings up an important issue regarding experimental design and suggests that interpretation of PCR results obtained from whole insect lysates must be cautiously interpreted. These findings suggest that further research is required to confirm the association between mosquito bites, adult mosquito infection, and incidence of Buruli ulcer in humans in Australia (reviewed above), where a link between mosquito feeding on infected possums and transmission of the agent via the same species of mosquitoes was proposed (Fyfe et al.[76]). An analysis of blood host choice by mosquitoes, documenting blood feeding on both possums and humans in the area where human cases of Buruli ulcer are occurring, would be required as one element of satisfying Barnett's criterion #2. At best, Barnett's criteria for vector incrimination have not been completely satisfied for a mosquito vector role, but more compelling data may be forthcoming on this matter in the future. A second approach to vector incrimination involves application of the Bradford Hill guidelines for establishing causation of infection and disease in epidemiological/ecological contexts [136]. Rather than rely upon experimental evidence, the Bradford Hill guidelines emphasize epidemiological/ecological association and use of logical inference to build up support and evidence for a strong conclusion of cause and effect, where A represents the “cause” and B the “effect” in the relationships under study [137]. The result is an “evidence hierarchy” that can be used in formal deduction [138], and represents an interdisciplinary approach to causal investigation in disease ecology. Here, “A” would be contact between an insect vector infected with M. ulcerans, and “B” would be human infection with M. ulcerans. The guidelines are qualitative in nature and do not require the clear endpoints of Barnett's criteria, yet represent a logical approach to the problem of cause and effect under epidemiological circumstances [139]. They are as follows (Table 2): 10.1371/journal.pntd.0000911.t002 Table 2 Listing of Hill's guidelines (Bradford Hill guidelines, Hill 1965) for associating a role of insect vectors of pathogens causing human disease. Term Descriptor/Qualifier 1. Plausibility Plausible, rational given knowledge of the biology of the putative vector, biology of the pathogen, and epidemiology of the disease. Specious associations would contraindicate a positive association. 2. Temporality The insect vector must show a temporal association with infection in humans; in particular, infected vectors should be found in endemic areas immediately before human cases occur. 3. Strength The association of the putative insect vector with human cases must be strong in time and space and in an epidemiological context. Correlation analysis supports the conclusion of strength if the correlation is positive. 4. Biological Gradient Prevalence of human cases should co-vary with prevalence of infection in the insect population. 5. Consistency Confirmed human cases should consistently be associated with infected insect vectors in time and space. 6. Alternate Explanations Explanations other than those related to a role of an insect vector should be considered and ruled out, or validated. 7. Experimentation Role of an insect species as a vector should be validated through experimental analysis with adequate controls and with realism in experimental design. 8. Specificity Infection with M. ulcerans in humans occurs when, and only when, a bite by an infected insect occurs first. 9. Coherence The association of human infection with insect transmission must cohere to knowledge of similar relationships in other similar associations. (1) Plausibility. The cause and effect association of A and B must be plausible, that is, rational and lacking in speciousness. By this is meant that the association reflects the common understanding of the normal behavior and other attributes of both A and B, bringing the appropriate factors together in such a way that abnormally implausible (i.e., irrational) explanations must be discounted. In formal philosophy, plausibility must be demonstrated by sets of binary outcomes whose relationships are clearly defined propositions which can be resolved by the application of logical discourse [140]. Although plausibility can be formulated axiomatically, it cannot be analyzed statistically. It is important, therefore, not to confuse “plausible” with “probable” as the latter allows for rare and unusual circumstances and events to be explanatory under the right circumstances, whereas the former involves a rigorous, but non-probabilistic analytical process. Put more simply, plausibility addresses qualitatively how likely or unlikely it is that A results in B. A common problem in epidemiological scenarios that confronts plausibility is the issue of clusters of cases of infection (e.g., [134]), which may or may not have spatial associations with other nearby cases or with the landscape qualities near those cases [136]. In the case of Buruli ulcer and vector transmission of M. ulcerans, it is not implausible that Hemiptera and human cases are associated in time and space, but it is not plausible that there is a direct, causal relationship between the pair except in rare, accidental circumstances. Hence, there is insufficient evidence to conclude that biting hemipterans are a significant vector of M. ulcerans, although they may act as environmental reservoirs. (2) Temporality. If A results in B, then A must consistently precede B in temporal sequence. For Buruli ulcer, there is no evidence that bites of particular insects consistently precede development of patent M. ulcerans infection in humans, although there is evidence that mosquito bites are associated with increased risk [90]. The problem with this guideline is the prolonged period of time between exposure and development of symptoms in Buruli ulcer disease. However, if bites from true bugs always preceded disease, patients are likely to remember these due to the painful nature of a naucorid or belostomatid bite, in contrast to bites by mosquitoes that often go unnoticed. (3) Strength. Is the “strength” of the association great? For example, is there a statistically significant correlation between A and B in space and or time? The association between contact with water sources and M. ulcerans infection in humans is reasonably strong, but between insect bites and infection it is not for hemipterans, nor yet firmly established for mosquitoes in Australia and virtually non-existent for mosquitoes in Africa. (4) Biological gradient or dose-response relationship. Infection in B should increase proportionately as A increases. This principle can operate at the dose-response level, as in a toxicological series; or at the population level, as when, e.g., more dengue virus infected mosquitoes results in more human cases of infection with that virus in space and time. The relationship may not be linear, thus confounding the interpretation of the relationship. There is no evidence that higher infection rate of M. ulcerans in aquatic insects results in higher incidence of infection in humans, although there is evidence that adult mosquitoes caught in highly endemic area in southeastern Australia are more likely to be PCR positive than those caught in areas with lower endemicity [35]. (5) Consistency. Episodes and research data where A and B show spatial and temporal associations commensurate with the other Bradford Hill guidelines must consistently reveal the association to be a positive one. Consistency could be revealed by meta-analysis of many data sets or through replicated, longitudinal studies across time and space. If scenarios emerge in which B occurs, but A does not in space and time, then doubt emerges regarding the veracity of the association. Although there are vignettes, correlations, and observations regarding insect vectors of M. ulcerans, there is no clear consistency among epidemiological scenarios to currently support the notion that insects are the predominant vector in most geographic regions. Consistent data are lacking for the ubiquitous role of vectors in the M. ulcerans transmission system. (6) Consideration of alternate explanations and analogous situations. Explanations other than causation due to A must be carefully weighed as alternatives. Causation may be inferred by analogous correspondence with other scenarios. For Buruli ulcer, a wide range of alternate explanations for transmission exists, such as human behavior linkages involving activities that increase direct skin contacts with contaminated water and inoculation with infective doses of M. ulcerans through lesions. However, as we have seen, several diseases with insect vector associations have alternative transmission modes, such as tularemia, plague, Rift Valley fever, and trachoma. Thus, it is plausible that there are multiple modes of transmission in Buruli ulcer, with certain modes more likely given specific environmental and socio-cultural contexts. (7) Experimentation. If experimental manipulations are feasible and can be structured realistically, then outcomes of the treatment regime conferred upon B (such as exposure to the effects of A) must reflect the association in a positive way. Often, however, Bradford Hill guidelines are utilized because experiments are either not possible, or not sufficiently rigorous or realistic. Experimental data on insect-M. ulcerans relationships have been reviewed above. There seems to be a sufficient body of work with sufficient variation in outcomes that the treatment manipulations do not lead to easily generalized conclusions on the association. Furthermore, it is often difficult to find true replication for large-scale experiments (e.g., treating replicate ponds with a specific chemical agent to test of changes in M. ulcerans), making it difficult to rigorously evaluate and experimentally test complex dynamics related to multiple modes of transmission of M. ulcerans within the environment. (8) Specificity. In this guideline, B follows A, but B does not follow when other plausible explanatory factors and events occur in temporal or spatial association. It is one of the most difficult of the guidelines to satisfy and comes closest to a strict criterion, usually because of incomplete information, multiple causes of B, random effects, and systematic errors of measurement. The review of the literature on cause and effect between insects and Buruli ulcer cases indicates a paucity of data to prove specificity. Furthermore, there are few studies relating disease incidence and insect abundance in time and space especially in Africa, and none of the alternate explanations for transmission reviewed above, such as through aerosols (9), have been discounted. The current available data points to a multiple transmission model for Buruli ulcer, indicating that the Buruli ulcer disease system lacks specificity with regard to vector insects, with the possible exception of southeastern Australia. Therefore, more complete and rigorous qualitative assessments of data are critical to provide evidence for consistency and specificity with regard to the role of vectors and reservoirs in transmission of M. ulcerans. (9) Coherence. The association of B with A must cohere to knowledge of similar relationships in other similar associations. For M. ulcerans, insect transmission is quite unusual, as the remainder of the M. marinum group does not depend upon invertebrate vectors for transmission and infection in fish hosts. Furthermore, there is no scientific precedent for transmission of any disease agent from the direct bites of hemipteran bugs, nor is there precedent for biological transmission of any bacterial pathogen by mosquitoes known. Thus, coherence is overall not strong. However, although closely related to M. marinum, M. ulcerans is a distinct species with a genomic signature indicating it has diverged from its free-living ancestor and now occupies a specialized niche environment. Either a vertebrate gastrointestinal tract (e.g. possums) or insects may provide this unknown microenvironment. In summary, neither the application of Barnett's strict criteria nor the Bradford Hill guidelines support conclusively that bites by M. ulcerans-infected insects' result in human infection with M. ulcerans. However, further research will reveal if any associations might result in higher risk of infection under certain circumstances. Infection with anthrax bacteria, Bacillus anthracis, provides a useful comparison, not as a directly transferable model, but rather as a model for conceptualization of how insects, like mosquitoes, may have ancillary roles in bacterial transmission when other transmission modes also exist [141]. In that system, infection occurs in animals endemically and sporadically. When they are stressed (as in a drought), they become susceptible to low dosages of bacterial spores in soil. As animals die, colonization of necrophilic flies during decomposition results in infection locally and increased bacterial sporulation and more animal cases occur as a result (the so-called “case multipliers” effect of insects). As more animals become infected, an insect-mediated dispersal of bacteria occurs by biting flies such as deer flies and horse flies, whose mouthparts can become contaminated with bacteria during blood feeding (the so-called “space multiplier” effect of insects). The role of flies in both modes furthers epizootics of anthrax. Although these two processes are unlikely to occur for Buruli ulcer, which appears to be mainly an endemic disease, the scenario for anthrax establishes a model by which insects might be envisioned to have ancillary roles in transmission for M. ulcerans as well. Conclusions Recommended research directions on Buruli ulcer disease As stated in the beginning of this review, Buruli ulcer disease has been referred to as the “mysterious disease” because the exact mode(s) of transmission, in the strictest sense, remain unclear, although several hypotheses have been proposed. We have reviewed the hypotheses and reported on studies that provide good evidence of probable reservoirs for the disease, particularly in Australia. An intellectual framework for establishing criteria for transmission followed this. Finally, we recommend that the following research studies be conducted to help better understand transmission of M. ulcerans in nature: 1) in depth studies of human behavior patterns in African endemic villages to better understand exposure to the pathogen in the environment; 2) a search for mammalian and/or other animal reservoirs and potential arthropod vectors in Africa; 3) understanding the relationship between mosquitoes, humans and infected possums who frequently share the same habitats in Australia; 4) laboratory competency studies with Australian mosquitoes using local strains of MU to determine whether transmission could occur vertically (larvae to adult) or horizontally (adult feeds on possum and then on humans); 5) further field and laboratory experiments on vector transmission and vector competence to confirm current hypotheses and experimental evidence on arthropod transmission; and 6) the development of new and innovative studies aimed at satisfying Hill's Criteria to provide strong and logically defendable evidence about the true mode, or modes, of Buruli ulcer transmission in nature. Supporting Information Checklist S1 PRISMA checklist. (0.07 MB DOC) Click here for additional data file.

Introduction Buruli ulcer (BU) is caused by the environmental mycobacterium, Mycobacterium ulcerans. Infection with M. ulcerans often leads to extensive necrosis of the skin and soft tissue with the formation of large ulcers, usually on the leg or arm, due to the production of the destructive polyketide toxin, mycolactone [1]. Although rarely fatal, BU causes serious morbidity and frequently results in permanent disability [2]. The disease has been reported in more than 30 countries worldwide; however, cases mainly occur in regions with tropical and subtropical climates. The majority of cases are found in West and sub-Saharan Africa. Cases of BU often cluster around particular water bodies and are highly focally distributed, with endemic and non-endemic communities often separated by only a few kilometres [2]. Australia is the only developed country reporting significant local transmission of M. ulcerans. In 1948, a cluster of cases linked to the Bairnsdale region in Gippsland was described by McCallum et al. [3]. Since then, foci of infection have been reported in tropical far north Queensland [4] and temperate coastal Victoria, where there have been several outbreaks over the past two decades: Phillip Island (1992–1995), the Frankston/Langwarrin region (1990–1997), St Leonards (2001–2002) and Point Lonsdale (2002-present) (Fig. 1) [5], [6]. The present outbreak in Point Lonsdale, a small coastal town approximately 60 km south-west of the Victorian capital Melbourne, is the largest on record in Australia, with over 100 laboratory-confirmed cases diagnosed since 2002. Geographically, the town is close to sea level, and there are several natural and man-made swamps and water features in the area [6]. Cases of BU have also been described in both native wildlife and domestic mammal species in Victoria, including koalas (Phascolarctos cinereus) [7], common ringtail possums (Pseudocheirus peregrinus) [8], a mountain brushtail possum (Trichosurus cunninghami), a long-footed potoroo (Potorous longipes) (J. Fyfe, unpublished), two horses [9], two dogs (O'Brien et al., manuscript in preparation), an alpaca [8] and a cat [10]. All animal cases were identified in locations where human cases of BU have been reported. 10.1371/journal.pntd.0000791.g001 Figure 1 Map of central coastal Victoria, showing places referred to in the text or associated references. The precise mode(s) of transmission and environmental reservoir(s) of BU are unresolved and continue to be the subject of intense research. Proximity to marshes and wetlands is a recognised risk factor for infection and several studies have explored the role of aquatic invertebrate species as potential vectors and/or reservoirs [6], [11]–[13]. Detection of M. ulcerans in environmental samples is mainly achieved using PCR, as culturing M. ulcerans directly from the environment is extremely difficult [14]. In Australia, M. ulcerans DNA was detected in water and detritus from swamps during the outbreak of BU on Phillip Island in the mid-1990s [15], [16] and more recently in five species of mosquitoes (Aedes sp., Coquillettidia sp. and Culex sp.) captured from Point Lonsdale (infection rate, 4.3/1,000 mosquitoes) [6]. In West Africa, M. ulcerans DNA has been detected in water and aquatic plants [17], insects (Belastomatidae, Naucoridae, Hydrophilidae), crustaceans and molluscs (Bulinus sp. and Planorbis sp.) and small fish (including Tilapia sp.) [11], [13], [18]–[21]. Recent studies of the distribution of M. ulcerans in aquatic sites in Ghana found evidence of M. ulcerans DNA in insects, water filtrate, biofilm and soil [12], [13]. In 2008, Portaels et al. described, for the first time, the cultivation and characterisation of an M. ulcerans strain obtained from an aquatic Hemiptera (common name Water Strider, Gerris sp.) from Benin [14]. Analysis of the whole genome sequence of M. ulcerans has provided further insights into the elusive environmental reservoir and mode of transmission [22]. Complete sequencing of an M. ulcerans strain isolated from a patient in Ghana revealed a 5,631,606 bp circular chromosome with 4160 genes, 771 pseudogenes and a 174,155 bp virulence plasmid pMUM001 that is required for the production of mycolactone [23], [24]. Comparison of the M. ulcerans genome with the genome of M. marinum confirmed the very close relationship between these species; however, it also revealed that there are some striking differences, mostly due to the presence of the plasmid pMUM001 and the many chromosomal deletions and rearrangements that have occurred in M. ulcerans [23]. It is therefore likely that M. ulcerans has evolved from an M. marinum-like ancestor by lateral gene transfer and reductive evolution, through the acquisition of a pMUM001-like plasmid, expansion of the two high copy number insertion sequence elements IS2404 and IS2606, extensive gene disintegration (formation of pseudogenes), genome rearrangements and DNA deletion. These characteristics suggest that M. ulcerans has recently passed through a so-called “evolutionary bottleneck” and is adapting to a new, niche environment. In this study, we investigated potential environmental reservoirs of M. ulcerans in south-eastern Australia with the aim of developing a more comprehensive model of its life cycle and mode of transmission. Specifically, using semi-quantitative real-time PCR and culture to test for the presence of M. ulcerans, we investigated a range of potential abiotic and biotic reservoirs (selected using emerging information in the literature and our own ongoing field based research) in areas of varying BU endemicity. Our findings have led us to propose that M. ulcerans is able to infect small mammals, survive and potentially replicate within their gastrointestinal tracts and raises the possibility that mammals play a major role in the ecology of M. ulcerans. Materials and Methods Environmental samples a. Study sites and sample collection This study was conducted in Victoria, Australia, primarily at Point Lonsdale on the Bellarine Peninsula (a current human BU outbreak zone, and therefore classified as endemic). A number of other sites, classified as areas of low endemicity (where BU infection has occurred in the past or fewer cases have been recorded recently), or non-endemic (no recorded human or animal BU cases), were also sampled (Fig. 1). The number and types of samples collected and tested are shown in Tables 1 and 2. Following collection, all samples were stored in sterile plastic containers or zip-lock bags, transported cool to the laboratory and stored at 4°C prior to DNA extraction, usually within a week of collection. 10.1371/journal.pntd.0000791.t001 Table 1 Detection of M. ulcerans DNA (IS2404, IS2606 and KR) in environmental samples collected from Point Lonsdale (endemic) and sites of low endemicity in Victoria, Australia. Sample type No. samples positive/no. samples tested Point Lonsdale a Bellarine Peninsula b Phillip Island c Gippsland d Suspended solids/water residue 4/4 (100%)e 0/10 0/9 0/10 Aquatic plant biofilm 2/10 (20%) 0/5 0/2 0/2 Aquatic plants 1/9 (11%) 0/5 0/5 0/2 Aquatic macroinvertebrates 0/12 0/15 0/4 0/7 Detritus 3/14 (22%) - - 0/33 Sediment 9/27 (33%) 0/1 - - Soil 22/36 (61%) 2/7 (29%) 0/3 0/3 Terrestrial Plants 9/51 (18%) 0/3 0/4 2/21 (10%) Brushtail possum faecesf 2/5 (40%) - 0/5 - Total 52/168 (32%) 2/51 (4%) 0/32 2/78 (3%) a High endemicity area. b Ocean Grove, Queenscliff, St Leonards (low endemicity areas). c Low endemicity area. d Bellbird Creek, Sale (low endemicity areas). e All four samples collected from the same site in Point Lonsdale on the same day. f Preliminary testing only (see Table 2 for results of large scale testing). 10.1371/journal.pntd.0000791.t002 Table 2 Detection of M. ulcerans DNA in possum faeces collected from BU high-, low- and non-endemic locations, in Victoria, Australia. Location Total human BU cases, past 5 years c Average annual incidence per 1000 population, past 5 years d (range) Detection of M. ulcerans DNA in faeces by PCR f Ringtail possum Brushtail possum No. positive/No. tested (%) Median est. bacterial load e No. positive/No. tested (%) Median est. bacterial load e High endemicity Point Lonsdale 81 4.04 (0.81–8.07) 70/164 (43%) 104 8/28 (29%) 102–103 Low endemicity Barwon Headsa 15 0.87 (0.00–2.00) 44/171 (26%) 104 15/78 (19%) 102–103 Ocean Grove 11 0.18 (0.00–0.44) 0/29 (0%) 0/9 (0%) Queenscliff 6 0.85 (0.00–2.12) 3/43 (7%) 102–103 0/0 Phillip Island 3 0.00 10/90 (11%) 102–103 1/76 (1%) 102–103 Non-endemic Boho South 0 0.00 0/29 (0%) 0/1 (0%) Breamlea 0 0.00 0/16 (0%) 0/0 Greater Melbourneb 0 0.00 0/15 (0%) 0/43 (0%) Torquay 0 0.00 1/24 (4%) 102–103 0/7 (0%) a Appears to be an area of increasing BU endemicity, with seven of the 15 cases diagnosed in 2009. b Comprises metropolitan suburbs of Clifton Hill, Clayton and Parkville. c Laboratory-confirmed human cases in residents and visitors, 2005–09. d Laboratory-confirmed human cases in residents only, 2005–09. e Expressed as organisms/gram of faeces. f All samples positive for IS2404. Subsets from each location were confirmed by IS2606 and KR PCR. b. Sampling methods Aquatic environments were sampled for suspended solids/water residue collected from natural and man-made water bodies in Point Lonsdale and low endemicity sites. Two hundred millilitres (ml) of water was passed through a 1.6 micron fibreglass filter (Whatman Inc.) using a hand pump and/or 60–120 ml water through a 1.6 micron fibreglass filter (Whatman Inc.) using a syringe (volume was dependent on turbidity). Aquatic plant biofilms were collected from the dominant macrophytes (plant species) in natural and man-made water bodies, in Point Lonsdale and low endemicity areas, by placing the macrophyte samples in sterile bags, mixing with 200 ml clean water and scrubbing by hand to remove the biofilm. A 50 ml subsample was retained for each. A section of the stem from each macrophyte was also sampled. Aquatic macroinvertebrates were collected by sweeping a handheld D-frame aquatic net through a section of the water body for 45 seconds. Detritus, sediment and soil samples were collected from terrestrial and riparian sites using a hand held plastic sieve or by placing samples directly into a sterile container. Samples from terrestrial vegetation (leaves, bark, flowers, seeds etc) were collected and identified by botanist Neville Walsh (Senior Conservation Botanist, Royal Botanic Gardens, Melbourne). Faecal samples from common ringtail possums and common brushtail possums (henceforth referred to as ringtail and brushtail possums) were collected directly from the ground, from the branches of trees or from fences, at 100- or 500-metre intervals along transects across areas of varying BU endemicity: Point Lonsdale (high endemicity area); Barwon Heads, Phillip Island, Ocean Grove and Queenscliff (low endemicity areas); Breamlea, metropolitan Melbourne, Boho South and Torquay (non-endemic areas) (Fig. 1). These sampling intervals were chosen to avoid any chance of repeated sampling from the same individual and were based on an estimated home range diameter for ringtail possums of no more than 100 metres (A. Legione, unpublished data). The identity of the animal host was determined by visual identification of the faecal sample (Fig. 2D), by an experienced zoologist (one of the authors) or with the aid of a scat and tracking manual [25]. 10.1371/journal.pntd.0000791.g002 Figure 2 Photographs of Point Lonsdale, common brushtail possums and common ringtail possums. A. Point Lonsdale streetscape showing typical possum habitat. B. Common brushtail possum. C. Common ringtail possum. D. Brushtail possum faeces (left) and ringtail possum faeces (right). E. Ringtail possum tail lesion. F. Ringtail possum nose lesion. Live animal studies a. Capture and sampling of live possums The capture of possums, which are nocturnal, was based on standard operating procedures for the handling of wildlife developed by Dr Kath Handasyde, approved by The University of Melbourne Faculty of Veterinary Science Animal Experimentation Ethics Committee (project no. 0706769) and under permit from the Victorian Department of Sustainability and Environment (DSE permit no. 10004406). Cage traps, designed for live capture of brushtail and ringtail possums, baited with an apple smeared with peanut butter or a bait ball of peanut butter and rolled oats, were set 1–2 hours before dark in public and private properties throughout Point Lonsdale and then checked, commencing at dawn, the following morning. Ringtail possums were also caught at night, directly by hand, using a specifically designed noosing pole and a hand-held net. After capture, animals were transferred into material bags, and transported by car to a quiet, enclosed area, awaiting collection of samples and data. b. Collection of samples and data from live possums To minimise distress during handling and sampling, possums were heavily sedated with an I.M. injection of Zoletil (Virbac Australia Pty Ltd, 5–8 mg/kg) using a 29 gauge needle. Possums were examined for external lesions resembling BU and, if present, a specimen was obtained by swabbing the affected area. Faecal specimens, along with a number of other clinical samples that are described in a separate report (manuscript in preparation), were also collected. Animals were individually marked, via a number tattooed onto the ear, and a small PIT (passive induction transponder) tag, inserted subcutaneously between the shoulder blades, so that they could be identified in the event of recapture. Individual animals were subjected to full handling only once during any particular field trip. After handling and sampling, animals were placed individually into material bags, and held in an animal box in a quiet enclosed area. Animals were then released at dusk, on the same day, at the site of capture. However, in the circumstance that a captured animal was deemed, by a veterinarian, to be too unwell to be released, there was a provision for the animal to be euthanased using an overdose of pentobarbitone (150mg/kg). DNA-based analyses a. DNA preparation DNA was extracted from samples using the FastDNA® SPIN Kit for Soil with the FastPrep® Instrument (Qbiogene, Inc., Carlsbad, CA), after the following sample-dependent pre-extraction procedures: For soil, sediment, vegetation and possum faeces, ∼50–100 mg of wet or dry sample was directly added to the FastPrep Lysing Matrix E tube. Biofilm samples were prepared by centrifuging the Falcon tubes containing the 50 ml subsample at maximum speed for 10 mins. After removing the supernatant, the pellet was resuspended in kit-supplied Sodium Phosphate Buffer and transferred to the Lysing Matrix E tubes. Water residue was prepared by cutting the fibreglass filters into small pieces using a sterile scalpel and adding directly to the Lysing Matrix E tubes. Swabs were placed in sterile bead bottles with 2 ml phosphate buffered saline (PBS), vortexed, and 1 ml added to the Lysing Matrix E tubes. The Lysing Matrix E tubes were then centrifuged at maximum speed for 10 mins and the supernatant removed. After the sample-dependent pre-extraction procedures, DNA extraction was then performed according to the manufacturer's recommendations. DNA preparations were stored at −20°C. b. Detection of M. ulcerans DNA DNA extracts were tested for the presence of M. ulcerans DNA using two semi-quantitative real-time PCR assays targeting the insertion sequences IS2404 and IS2606 and a sequence encoding the ketoreductase B domain, KR, within the mlsA1, mlsA2 and mlsB genes. These assays were developed and validated for use on environmental samples by Fyfe et al. [26] and are able to distinguish between M. ulcerans and other mycolactone-producing mycobacteria (MPM) that contain IS2404, but fewer copy numbers of IS2606, based on the difference in cycle threshold values between IS2606 and IS2404 (ΔCT [IS2606-IS2404]) [26]. All extracts were initially screened singly for the high copy number insertion sequence IS2404. This assay was multiplexed with an internal positive control to monitor PCR inhibition. Inhibited extracts were diluted 1/5 or 1/10 and repeat PCR performed. Extracts that were still inhibited at 1/10 dilution were omitted from analyses. With the exception of the possum faecal samples, all of the IS2404-positive DNA extracts from each sample type were tested in duplicate for IS2606 and KR. In view of the large number of IS2404-positive DNA extracts from possum faecal samples obtained, a subset of these, taken from each of the different locations, was similarly confirmed. The ΔCT (IS2606-IS2404) were calculated to confirm that the sequences detected were attributable to M. ulcerans and not another MPM. To exclude the possibility of contamination, at least one negative control was included in every DNA extraction run, and four negative controls included in every real-time PCR assay. c. Estimation of M. ulcerans bacterial loads in different samples To estimate the M. ulcerans bacterial loads (expressed as M. ulcerans/gram or M. ulcerans/ml) in various sample types, the CT values obtained for IS2404 were compared with a standard curve generated using a series of DNA extracts prepared from environmental samples that had been spiked with known numbers of M. ulcerans organisms [26]. These estimates were determined to provide an indication of the relative numbers of M. ulcerans between samples, rather than a strict quantitation of the number of organisms present in a sample, and hence are generally expressed as a 10-fold range. d. Variable Number Tandem Repeat (VNTR)/Mycobacterial Interspersed Repeat Unit (MIRU) typing VNTR/MIRU typing was performed using the conditions described previously [27]–[29] in 25 µl reactions using 1 µl of DNA template. PCR products were visualised on a 2% agarose gel and PCR product sizes estimated by comparing fragment sizes with a 100 bp DNA ladder (Promega, Wisconsin, USA). Products of the expected size were purified using a Roche High Pure PCR Purification Kit (Roche Diagnostics, Australia) and sequenced. e. DNA sequence analysis Sequence analysis of purified PCR products was performed using the BigDye (R) Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA) according to the manufacturer's instructions. Reactions were analysed on an Applied Biosystems 3730S Genetic Analyzer (Applied Biosystems). Sequence data were edited using Bionumerics v4.0 (Applied Maths BVBA, Ghent, Belgium) and then compared with those derived from an M. ulcerans isolate, cultured from a human patient from Point Lonsdale. f. Whole genome sequencing and assembly Genomic DNA was prepared from a possum M. ulcerans isolate (JKD8170) and a human M. ulcerans isolate (JKD8049), both from Point Lonsdale. Whole genome sequencing was performed using an Illumina Genome Analyzer II with 36 cycle paired-end chemistry. Reads were mapped to the reference strain M. ulcerans Agy99 (GenBank accession CP000325) using SHRiMP [30]. Single nucleotide polymorphisms (SNPs) and micro-indels (DIPs) were detected using Nesoni, a software tool for analysing high-throughput DNA sequence data (used in [31]). Nesoni tallied the raw base counts at each mapped position in each of the reference strains, and then compared them using Fisher's Exact Test to find variable nucleotide positions in JKD8170 relative to JKD8049. To exclude the possibility that additional mutations in JKD8049 may have occurred in regions not present in the reference M. ulcerans Agy99, de novo assembly of JKD8170 and JKD8049 was performed using Velvet [32] and the above SNP/DIP detection procedure was repeated using the resulting contigs as reciprocal reference sequences. The read data for JKD8170 and JKD8049 have been deposited in the NCBI Sequence Read Archive (SRA) as part of Study accession number SRP001289. Culture of M. ulcerans a. Culture of M. ulcerans from environmental samples Culture of M. ulcerans from possum faeces was attempted by homogenising samples in bead bottles with Ringer's solution, decontaminating with an equal volume of 4% sodium hydroxide, incubating at room temperature for 15 mins and neutralising with 10% orthophosphoric acid (modified Petroff method). Samples were centrifuged at 4000 rpm for 20 mins and pellets resuspended in 2 ml Ringer's solution. 400 µl of the decontaminated suspension was used to inoculate Mycobacteria Growth Indicator Tube (MGIT) broths with PANTA added according to the manufacturer's recommendations (BD, Franklin Lakes, N.J.), Brown and Buckle slopes and 7H10 slopes with antibiotics (25 µg/ml piperacillin, 50 µg/ml amphotericin, 25 µg/ml vancomycin, 800 µg/ml actidione, 4 µg/ml aztreonam). MGIT broths and solid media were incubated at 31°C and monitored weekly for up to 16 weeks. b. Culture of M. ulcerans from possum lesions Swabs were placed in bead bottles with 2 ml phosphate buffered saline (PBS), vortexed, decontaminated with 2% sodium hydroxide, incubated at room temperature for 15 mins and neutralised with 10% orthophosphoric acid. Samples were then centrifuged at 4000 rpm for 20 mins and pellets resuspended in 2 ml Ringer's solution. 400 µl was used to inoculate Brown and Buckle slopes and MGIT broths with PANTA added according to the manufacturer's recommendations (BD, Franklin Lakes, N.J.) and were incubated at 31°C with weekly monitoring for up to 12 weeks. Human case definition and BU incidence A case of BU was defined as a human patient with a suggestive clinical lesion from which M. ulcerans was identified by PCR [26] or culture from January 2005 to December 2009 inclusive. The likely geographic origin of infection was determined on the basis of the patient's residential address and/or travel history. A patient was considered as having acquired BU from a particular geographic area if he/she was a resident of, or a visitor to, that area and had not reported recent contact with any other known BU endemic area. Due to the large seasonal fluctuations in the population of endemic areas (most of which are summer holiday destinations), and the difficulty in estimating the number of visitors to a particular area, the average annual incidence of BU in each geographic area over the five-year study period was calculated by dividing the average annual number of cases in residents only (that is, cases in visitors were excluded) by the resident population of the specified geographic area. Resident population numbers were obtained using Australian Bureau of Statistics data derived from the 2006 Census of Population and Housing [33]. Statistical analyses Statistical analyses were performed using STATA version 10.0 (STATA Corporation, College Station, TX). Proportions were compared using the two-sample test of proportion. Results Environmental testing in Point Lonsdale and areas of low BU endemicity Testing of environmental samples commenced in mid-2004, just prior to the peak of the Point Lonsdale outbreak. The initial focus was low-lying, wet areas in which mosquitoes were likely to breed, such as drains, soak pits (covered concrete pits into which storm water and street runoff flows and sits until it gradually seeps into the ground), man-made lakes and natural water bodies. In Point Lonsdale, low levels of M. ulcerans DNA (that is, weak positive real-time PCR signals for IS2404, IS2606 and KR) were detected in sediment from a man-made lake; soil, sediment and detritus from a number of different soak pits and drains; biofilm; aquatic plants; and residue from filtered water (Table 1). The estimated bacterial loads for these samples ranged from 10–100 organisms/ml for residue from filtered water and 103–104 organisms/gram for biofilm. In contrast, only four samples (two soil and two vegetation) from low endemicity areas were positive for M. ulcerans DNA (Table 1). In late 2006, the scope of our environmental testing expanded to samples in dryer areas at higher elevations, including leaf litter, leaves, tree bark, flowers, seeds, stems and faeces from brushtail possums (Table 1). The rationale for this was: (i) soil collected outside drains had previously tested positive for M. ulcerans DNA, (ii) BU patients have reported an association between small penetrating injuries, sustained from vegetation, and subsequent ulcers [34], and (iii) cases of BU are known to occur in arboreal marsupial mammals, including koalas [7] and ringtail possums [8]. Testing revealed that while M. ulcerans DNA could be detected at low levels in some samples of leaf litter and bark from trees (estimated bacterial load 102–103 organisms/gram), much higher levels of M. ulcerans DNA were detected in brushtail possum faeces (estimated bacterial load ≥106 organisms/gram). This important discovery led to the large scale, systematic testing of possum faeces in Point Lonsdale, as well as low and non-endemic sites. Possum faecal testing in BU high-, low- and non-endemic sites Over a two-year period (2007–09), systematic collection of faeces from brushtail and ringtail possums was carried out across Point Lonsdale, nearby low endemicity areas and non-endemic areas (Table 2). A total of 589 faecal samples from ringtail possums and 250 samples from brushtail possums were tested. The difference in the number of samples collected from each geographic location and from each species reflected the relative population densities, with ringtail possums being much more abundant than brushtail possums in many areas sampled (K. Handasyde and A. Legione, unpublished data). In Point Lonsdale, M. ulcerans DNA (IS2404) was detected in 43% of ringtail possum and 29% of brushtail possum faecal samples (Table 2). All samples tested for the presence of IS2606 and KR were PCR-positive for these additional targets. Furthermore, the ΔCt (IS2404-IS2606) was always in the range expected for M. ulcerans (2.17–2.79), rather than another MPM (6.94–8.07) [26]. The median estimated bacterial load was 104 organisms/gram (range: 102–108 organisms/gram) for ringtail possums (Fig. 3), with 17% of positive samples having an estimated bacterial load >106 organisms/gram. The median estimated bacterial load for brushtail possum faeces was 102–103 organisms/gram (range: 102–106 organisms/gram). 10.1371/journal.pntd.0000791.g003 Figure 3 Distribution and estimated bacterial load of M. ulcerans-positive ringtail faecal samples in two towns. Map shows results of faecal surveys conducted in Point Lonsdale (approx. 81 human cases 2005–09) in August 2008 and Queenscliff (approx. 6 human cases 2005–09) in November 2008. In low endemicity areas, the proportion of PCR-positive faecal samples varied by location. For example, in Barwon Heads, where 15 human cases of BU have been reported since 2005, the proportion of positive ringtail and brushtail faecal samples was relatively high (26% and 19% respectively) compared with the other locations where fewer cases of BU have been reported (Table 2). The median estimated bacterial load of positive faecal samples from low endemicity areas also varied. In Barwon Heads the median estimated bacterial load for ringtail possum faeces was 104 organisms/gram (with 16% of the positive samples having an estimated bacterial load >106 organisms/gram). As in Point Lonsdale, the estimated bacterial load of the positive brushtail possum faeces in Barwon Heads was generally lower than for the ringtail possum faeces, with a median estimate of 102–103 organisms/gram. Similarly low M. ulcerans bacterial loads of 102–103 organisms/gram were estimated for faeces (ringtail possum only) collected in Queenscliff [Fig. 3] and Phillip Island. Only one sample collected from a non-endemic area (Torquay) was positive for M. ulcerans DNA and the estimated bacterial load of this sample was low (102–103 organisms/gram). Mapping of the samples collected in Point Lonsdale revealed that M. ulcerans DNA could be detected throughout Point Lonsdale and did not appear to be concentrated in one particular area or limited to one particular point source (Fig. 3). However, in Barwon Heads, positive faecal samples were only detected in the southern part of the town (data not shown). No seasonal trends were observed, with the number of positive samples, and the estimated bacterial loads of those samples, consistent between summer, autumn, winter and spring (data not shown). All attempts at culturing M. ulcerans from possum faeces were unsuccessful. PCR-positive and PCR-negative possum faeces were inoculated into MGIT and onto Brown and Buckle and 7H10 slopes with antibiotics. The MGIT broths and Brown and Buckle slopes exhibited extensive fungal contamination after two weeks and were discarded. Despite the absence of fungal contamination on the 7H10 slopes, no growth of M. ulcerans was detected after 16 weeks incubation. Capture and examination of possums from Point Lonsdale Over a 20-month period from February 2008 to November 2009, 42 ringtail possums and 21 brushtail possums were captured in Point Lonsdale and examined for BU disease. Among the ringtail possum cohort, 16 (38%) animals had laboratory-confirmed (PCR ± culture) M. ulcerans lesions and/or M. ulcerans PCR-positive faeces. Of the 11 animals with BU disease, nine had M. ulcerans PCR-positive faeces, one had M. ulcerans PCR-negative faeces and we were unsuccessful in collecting a faecal sample from the remaining animal (Table 3). Notably, five of the ringtail possums that did not have BU skin lesions had M. ulcerans PCR-positive faeces. Interestingly, as shown in Table 3, there was little difference in the median estimated bacterial loads of faeces from animals with BU skin lesions and animals without BU lesions. The incidence of M. ulcerans infection among the 21 brushtail possums was lower. One animal had a BU skin lesion and M. ulcerans PCR-positive faeces (estimated bacterial load, 103–104 organisms/gram) and four animals without BU lesions were found to be shedding low levels of M. ulcerans DNA in their faeces (102 organisms/gram) (Table 3). 10.1371/journal.pntd.0000791.t003 Table 3 Mycobacterium ulcerans status of ringtail and brushtail possums captured in Point Lonsdale, Victoria, and examined for BU lesions and the presence of M. ulcerans DNA in faeces. M. ulcerans status of possumsa No. possums (median estimated bacterial load/gram faeces) Total possums Ringtail Brushtail BU lesions present; positive faeces 9 (105–106) 1 (104–105) 10 BU lesions present; negative faeces 1 0 1 BU lesions present; no faeces collected 1 0 1 BU lesions absent; positive faeces 5 (105–106) 4 (102–103) 9 BU lesions absent; negative faeces 26 16 42 Total 42 21 63 a M. ulcerans status refers to the presence or absence of external BU lesions (confirmed by PCR ± culture) and M. ulcerans DNA in faeces (detected by PCR). The most common site for BU lesions was the tail (Fig. 2E). Amongst the 12 possums with BU disease, nine had lesions on the tail and four had lesions on the toe/foot (Table 4). Five of the ringtail possums had multiple lesions, with one animal having severe ulcerative and oedematous lesions on her nose (Fig. 2F), left upper lip, both fore paws, right hock, left hind leg and tail. Three of these animals were euthanased and full necropsies performed to determine the extent of the M. ulcerans infection. The results of these necropsies, along with the results of the other clinical samples taken from all 63 possums captured (including blood, buccal swabs and nasal swabs and urine), are described in a separate report (manuscript in preparation). 10.1371/journal.pntd.0000791.t004 Table 4 Characteristics of possums with laboratory-confirmed BU lesions captured in Point Lonsdale, Victoria, 2008–09. ID Species Sex Age Site of BU lesion(s) a 2 Ringtail possum Female Adult Tailb and toeb 9 Ringtail possum Male Adult Tail 20 Ringtail possum Male Adult Tailb 23 Ringtail possum Male Adult Tail 30 Ringtail possum Male Juvenile Hind foot 32 Ringtail possum Female Adult Multiple ulcerative and oedematous lesionsb , c 46 Ringtail possum Male Adult Tail 47 Ringtail possum Male Adult Tailb 49 Brushtail possum Female Adult Toeb 57 Ringtail possum Female Adult Tailb and ear 61 Ringtail possum Male Adult Tailb, nose, arm and face/cheek 62 Ringtail possum Female Adult Tail, nose and eye a All lesions confirmed by PCR ± culture. b Culture confirmed. c Nose, tail, (R) hock, (L) hind leg, (L) front hand, (L) upper lip, (L) hind leg muscle. VNTR/MIRU typing of possum faecal samples demonstrates identity with human outbreak strain The two multiplex real-time PCR assays used in this study to detect M. ulcerans in environmental samples distinguish between M. ulcerans and other MPM that also harbour IS2404 and IS2606 [26]. However, we also sought to determine whether the DNA detected in environmental samples was from the same strain of M. ulcerans that causes disease in humans in Victoria. PCR reactions for 10 VNTR loci and three MIRU loci were performed on a subset of DNA extracts from possum faeces (estimated bacterial load 105–106 organisms/gram), aquatic plant biofilm (estimated bacterial load 103–104 organisms/gram) and water filters (estimated bacterial load 103–104 organisms/filter). The concentration of M. ulcerans DNA in the other sample types (for example, soil) has previously been shown to be insufficient for PCR amplification of these single copy loci [35]. DNA extracted from the possum faeces generated PCR amplicons of the same size (Fig. 4) and sequence as the Victorian human outbreak strain at all loci. As predicted by the lower concentration of M. ulcerans DNA in the samples, DNA extracts from the aquatic plant biofilm and water filter generated PCR amplicons at one locus only (VNTR locus 6 and 19, respectively). In each case the sequence was identical to the Victorian outbreak strain. These data provide evidence that the strain of M. ulcerans detected in these samples is the same as the strain which causes disease in humans in this region. The results also confirm that this method of analysis can only be applied successfully to samples (clinical or environmental) with an estimated M. ulcerans load of ≥105 organisms/gram and should only be used as a confirmatory/epidemiological tool and not as the primary method by which all environmental samples are screened for the presence of M. ulcerans DNA [35]. 10.1371/journal.pntd.0000791.g004 Figure 4 Variable number tandem repeat (VNTR) typing of M. ulcerans DNA in possum faeces demonstrates identity with human outbreak strain. Numbers represent VNTR loci [27]. At each locus: left PCR product, Victorian human patient isolate; right PCR product, DNA extracted from brushtail possum faeces collected in Point Lonsdale. Whole genome sequencing of an M. ulcerans isolate from a ringtail possum Illumina high-throughput short-read sequencing was used to compare the genome of an M. ulcerans isolate from a ringtail possum captured in Point Lonsdale (M. ulcerans JKD8170) and a human clinical isolate from Point Lonsdale (M. ulcerans JKD8049) obtained during the period of the M. ulcerans outbreak. This process generated 31,028,581 reads for JKD8170 and 10,921,914 reads for JKD8049. Bioinformatic analysis involved read mapping to the reference genome M. ulcerans Agy99 and reciprocal comparisons to consensus sequences derived from de novo sequence assemblies of each data set. These analyses revealed that both the possum and human isolates shared 5455 SNP differences compared to the reference genome (an African strain) but were differentiated from each other by only two SNPs (confirmed by PCR and Sanger DNA sequencing) across 5.6 Mb of chromosomal DNA sequence. These data confirm the extremely close genetic relationship between the human and possum isolates. Discussion Elucidation of the mode of transmission and environmental reservoir(s) of M. ulcerans is essential for the development of strategies to control and prevent BU outbreaks. Early epidemiological studies from Uganda in the 1970s suggested that M. ulcerans may be associated with certain grasses growing at the edges of permanent swamps [36], [37], and that transmission to humans was via contact with this environmental source. However, attempts to culture M. ulcerans from a range of plants were unsuccessful [38]. The possible role of rodents in the ecology of M. ulcerans was also considered over 30 years ago [39], however the presence of the organism in the organs of 700 animals from a BU endemic area in Uganda could not be confirmed by culture. The development of IS2404 PCR in the 1990s [16], [40] enabled researchers to detect the DNA of M. ulcerans and other MPM in a range of different samples, leading to a renewed search for the environmental reservoir(s). The PCR detection of M. ulcerans DNA in waterbugs from Benin and Ghana [18] and subsequent culture of M. ulcerans from a waterbug [14], focussed the search to aquatic habitats. Currently, the prevailing dogma is that the environmental reservoir of M. ulcerans is an abiotic or biotic component of aquatic, rather than terrestrial, ecosystems. Indeed, numerous epidemiological and environmental studies support this view [5], [11], [12], [14], [15], [17]–[21], [26], [41]–[43], including some of the data from our current study. We found that M. ulcerans could be detected in various aquatic samples including aquatic plants, biofilm and residue from filtered water (Table 1). The major strength of our study, however, was the use of a suite of real-time PCR assays targeting multiple regions in the M. ulcerans genome which, in addition to being highly sensitive, specific and less prone to contamination than conventional gel-based PCR [12], [13], enabled us to estimate the relative numbers of M. ulcerans in the various samples tested by determining the relative concentrations of M. ulcerans DNA among the different sample types. By following this gradient of M. ulcerans DNA, we discovered that the faeces of two marsupial mammals (ringtail and brushtail possums), contained higher concentrations of M. ulcerans DNA than the other samples tested. The large-scale testing of possum faeces in BU high-, low- and non-endemic sites, and the subsequent capture and examination of possums in Point Lonsdale, generated a number of important findings. Firstly, we discovered that there is a high density of ringtail possums throughout Point Lonsdale that are excreting copious amounts of faeces, almost half of which are estimated to contain M. ulcerans, into the environment (Table 2, Fig. 3). Secondly, we observed a strong positive correlation between the BU endemicity of an area and the proportion and DNA concentration of M. ulcerans-positive possum faeces, with 41% of faecal samples collected in Point Lonsdale testing positive for M. ulcerans compared with less than 1% of faecal samples collected from non-endemic areas (p<0.0001). Similar results were obtained in Benin with a correlation between BU endemicity in patients and environmental results. Environmental studies detected variations in M. ulcerans DNA positivity rates of aquatic insects over time, and these changes were reflected in corresponding alterations of frequency of BU patients in the same foci [44]. Thirdly, 38% of captured ringtail possums and 24% of captured brushtail possums were found to have laboratory-confirmed M. ulcerans skin lesions, mostly on the tail or feet, and/or M. ulcerans PCR positive faeces (Table 3). One explanation for the observation that most lesions occurred on the extremities is that these sites have lower temperatures favouring the growth of M. ulcerans. Another possibility is that, because these sites have less fur, they are more susceptible to insect bites or skin trauma via contact with vegetation or fighting with other possums, which may lead to inoculation of M. ulcerans. Fourthly, we observed that five of the 14 ringtail possums, and four of the five brushtail possums, that were shedding M. ulcerans DNA in their faeces did not have BU skin lesions, indicating that the presence of M. ulcerans DNA in faeces is not limited to clinically diseased animals (Table 3). However, we noted that animals with multiple lesions tended to have higher estimated faecal loads of M. ulcerans than animals with single lesions (data not shown). Finally, whole genome sequencing confirmed the extremely close genetic relationship between the human and possum isolates. Taken together, these findings suggest that possums may be an environmental reservoir for M. ulcerans in south-eastern Australia. If so, the biology of possums prompts a new interpretation/understanding of the life cycle of M. ulcerans. In particular, ringtail possums are exclusively arboreal, feeding on a variety of leaves of both native and introduced plants, as well as flowers and fruits [45], hence are unlikely to be exposed to M. ulcerans in soil or water. They are also caecotrophic. Caecotrophy is the ingestion of soft faeces of high nutritive value derived from caecal contents and is a critical factor in the ringtail possum's ability to utilise eucalypt foliage as a whole or major food source [46]. This behaviour may also favour gastrointestinal persistence of M. ulcerans. Brushtail possums are semi-arboreal, spending a considerable portion of their foraging time on the ground and, although mainly folivorous, have a more varied diet than ringtail possums [45]. The ecology of these species, which occur in strictly terrestrial habitats, contradicts the idea that the environmental host(s) of M. ulcerans are likely to reside primarily in aquatic environments, although the presence of M. ulcerans in aquatic habitats within the same location is also likely, based on data presented in this study. Thus, in light of our data, we suggest that reservoir species could include terrestrial mammals, and that the association of the disease with low-lying, wetter areas might be driven by the dependence of a vector species (such as mosquitoes [47]) on moist habitats. A disease reservoir may be defined as: “one or more epidemiologically connected populations or environments in which a pathogen can be permanently maintained and from which infection can be transmitted to the target population. Populations in a reservoir may be the same or a different species as the target and may include vector species” [48]. Our findings from Point Lonsdale suggest that at least one free-ranging mammal species (the ringtail possum), which can be very abundant in urban environments, forms part of a transmission cycle (Fig. 5) for M. ulcerans that could explain human outbreaks of BU in south-eastern Australia, although they may not necessarily be true maintenance hosts (that is, be able to maintain the organism in the absence of other environmental sources). However, bovine tuberculosis, caused by Mycobacterium bovis, and Johne's disease, caused by Mycobacterium avium subsp. paratuberculosis, are both maintained in wildlife reservoir species. In the United Kingdom, badgers (Meles meles) contribute to the spread of M. bovis between herds of cattle [49]. In New Zealand, where bovine tuberculosis is a major problem, the principle wildlife host for M. bovis is the common brushtail possum, which was originally imported from Australia and now occurs at such a high population density that it is a major agricultural and conservation pest [49]. 10.1371/journal.pntd.0000791.g005 Figure 5 Proposed transmission pathways of M. ulcerans between the environment, mosquitoes, possums and humans. 1. Possums ingest M. ulcerans from the environment and/or infected by an insect vector. 2. Possums amplify and shed M. ulcerans into the environment. 3. Insect vectors become contaminated with M. ulcerans from the environment and/or from contact with infected possums. 4. M. ulcerans transmitted to humans via insect vector and/or direct contact with contaminated environment. The way in which M. ulcerans might be transmitted from an animal to humans is not clear. A similar epidemiology to leptospirosis, the most common zoonosis worldwide [50], in which rodents are reservoirs but the disease is acquired by contact with contaminated water, should be considered. We envisage that the transmission pathway for M. ulcerans may involve vegetation, vertebrate hosts and invertebrate vectors in both terrestrial and aquatic ecosystems (Fig. 5). Such a model represents a fundamental change to the existing views on the ecology of M. ulcerans, although the idea that M. ulcerans is not confined to low-lying swampy areas is not new [36]–[39], [51], [52]. While we lack important information about whether mosquitoes are productive or simply mechanical vectors, and have only limited information on the site of carriage/colonisation, either on or within mosquitoes, a number of lines of evidence implicate mosquitoes as vectors of M. ulcerans in Victoria [6], [53]–[55]. Given that we found active M. ulcerans lesions in 26% of captured ringtail possums, transmission to humans might occur when an adult mosquito that has fed on a diseased possum, or rested on vegetation contaminated by a possum lesion, subsequently bites a human. Another possibility is that heavy environmental contamination with possum faeces containing M. ulcerans would enable mosquitoes (either as larvae or adults) to come into contact with M. ulcerans, in contaminated soil/water in roof gutters or drains (Fig. 5). This is supported by a study by Tobias et al. which showed that, in a feeding experiment where mosquito larvae were fed possum faecal material spiked with M. ulcerans or M. marinum, M. ulcerans accumulated within the mouth and midgut whereas M. marinum did not [55]. Key to determining which of these potential routes of transmission is most likely (or possible) is the demonstration of viable M. ulcerans organisms in possum faeces. We acknowledge that the detection of M. ulcerans DNA in possum faeces does not necessarily indicate the presence of viable organisms. However we, like many others who have attempted to culture M. ulcerans from environmental samples [14], have currently been unable to culture M. ulcerans from possum faeces. This was despite the fact that some of the samples had real-time PCR signals equivalent to those obtained for the lesion swabs from which culture of M. ulcerans was successful (data not shown). We believe that this has been largely due to the presence of fungi or fungal spores in the faecal samples which, despite decontamination methods, rapidly grew in broth cultures and on Brown and Buckle slopes and inhibited the growth of slower growing organisms such as M. ulcerans. Furthermore, on the basis of subsequent real-time PCR studies, it has become evident that the organisms are tightly associated with the particulate matter and that homogenising faeces in bead bottles results in very few bacteria in the suspension that would normally be used to inoculate the culture media (C. O'Brien, unpublished). We have also found that intact DNA can be recovered from possum faeces many months after sampling and that DNase treatment of the faecal homogenate does not lead to a reduction in the PCR signal (data not shown). This suggests that intact M. ulcerans organisms are present (though not necessarily viable), rather than just free M. ulcerans DNA. There is also the question of whether mammals could act as reservoirs in sub-Saharan Africa, where the majority of BU cases occur. Recent studies in Ghana failed to detect M. ulcerans in the organs or faeces of rodents and shrews [17], [56]. However these authors did not reject the hypothesis that these, or other species of small terrestrial mammals, may be part of the reservoir of M. ulcerans in this setting. Recent work conducted by our group, including the post-mortem examination of ringtail possums and rats (Rattus rattus) with and without clinical BU disease, has shown that M. ulcerans can be present in the gastrointestinal tracts of animals but not in the organs of the same individual (manuscript in preparation). We are currently investigating the potential role of other mammal species as hosts for M. ulcerans in the Australian setting. This study has led to a major a shift in our understanding of the environmental distribution of M. ulcerans in south-eastern Australia. It is hoped that the results presented here, along with our continuing laboratory and field research, will take us closer to elucidating the mode of transmission and environmental reservoir(s) of M. ulcerans and in turn the development of strategies to control and prevent this important yet often neglected human disease.

Mycobacterium ulcerans is an emerging environmental pathogen which causes chronic skin ulcers (i.e., Buruli ulcer) in otherwise healthy humans living in tropical countries, particularly those in Africa. In spite of epidemiological and PCR data linking M. ulcerans to water, the mode of transmission of this organism remains elusive. To determine the role of aquatic insects in the transmission of M. ulcerans, we have set up an experimental model with aquariums that mimic aquatic microenvironments. We report that M. ulcerans may be transmitted to laboratory mice by the bite of aquatic bugs (Naucoridae) that are infected with this organism. In addition, M. ulcerans appears to be localized exclusively within salivary glands of these insects, where it can both survive and multiply without causing any observable damage in the insect tissues. Subsequently, we isolated M. ulcerans from wild aquatic insects collected from a zone in the Daloa region of Ivory Coast where Buruli ulcer is endemic. Taken together, these results point to aquatic insects as a possible vector of M. ulcerans.