Advertisement
  • Loading metrics

Ecological Drivers of Mansonella perstans Infection in Uganda and Patterns of Co-endemicity with Lymphatic Filariasis and Malaria

  • Anna-Sofie Stensgaard ,

    asstensgaard@snm.ku.dk

    Affiliation: Center for Macroecology, Evolution and Climate, The Natural History Museum of Denmark, University of Copenhagen, Copenhagen, Denmark

  • Penelope Vounatsou,

    Affiliations: Department of Epidemiology and Public Health, Swiss Tropical and Public Health Institute, Basel, Switzerland, University of Basel, Basel, Switzerland

  • Ambrose W. Onapa,

    Affiliation: Envision/NTD Program, RTI, Kampala, Uganda

  • Jürg Utzinger,

    Affiliations: Department of Epidemiology and Public Health, Swiss Tropical and Public Health Institute, Basel, Switzerland, University of Basel, Basel, Switzerland

  • Erling M. Pedersen,

    Affiliation: Department of Veterinary Disease Biology, University of Copenhagen, Copenhagen, Denmark

  • Thomas K. Kristensen,

    Affiliation: Department of Veterinary Disease Biology, University of Copenhagen, Copenhagen, Denmark

  • Paul E. Simonsen

    Affiliation: Department of Veterinary Disease Biology, University of Copenhagen, Copenhagen, Denmark

Ecological Drivers of Mansonella perstans Infection in Uganda and Patterns of Co-endemicity with Lymphatic Filariasis and Malaria

  • Anna-Sofie Stensgaard, 
  • Penelope Vounatsou, 
  • Ambrose W. Onapa, 
  • Jürg Utzinger, 
  • Erling M. Pedersen, 
  • Thomas K. Kristensen, 
  • Paul E. Simonsen
PLOS
x

Abstract

Background

Mansonella perstans is a widespread, but relatively unknown human filarial parasite transmitted by Culicoides biting midges. Although it is found in many parts of sub-Saharan Africa, only few studies have been carried out to deepen the understanding of its ecology, epidemiology, and health consequences. Hence, knowledge about ecological drivers of the vector and parasite distribution, integral to develop spatially explicit models for disease prevention, control, and elimination strategies, is limited.

Methodology

We analyzed data from a comprehensive nationwide survey of M. perstans infection conducted in 76 schools across Uganda in 2000–2003, to identify environmental drivers. A suite of Bayesian geostatistical regression models was fitted, and the best fitting model based on the deviance information criterion was utilized to predict M. perstans infection risk for all of Uganda. Additionally, we investigated co-infection rates and co-distribution with Wuchereria bancrofti and Plasmodium spp. infections observed at the same survey by mapping geographically overlapping areas.

Principal Findings

Several bioclimatic factors were significantly associated with M. perstans infection levels. A spatial Bayesian regression model showed the best fit, with diurnal temperature range, normalized difference vegetation index, and cattle densities identified as significant covariates. This model was employed to predict M. perstans infection risk at non-sampled locations. The level of co-infection with W. bancrofti was low (0.3%), due to limited geographic overlap. However, where the two infections did overlap geographically, a positive association was found.

Conclusions/Significance

This study presents the first geostatistical risk map for M. perstans in Uganda. We confirmed a widespread distribution of M. perstans, and identified important potential drivers of risk. The results provide new insight about the ecologic preferences of this otherwise poorly known filarial parasite and its Culicoides vector species in Uganda, which might be relevant for other settings in sub-Saharan Africa.

Author Summary

Mansonella perstans is a widespread, but relatively unknown human filarial parasite that occurs in many parts of Africa. In a nationwide survey carried out in Uganda in 2000–2003, the distribution of M. perstans was assessed by screening school children. Here, we studied the underlying environmental drivers and ecologic correlates of the observed M. perstans prevalence patterns, produced a predictive risk map, and investigated associations with Wuchereria bancrofti (causative agent of lymphatic filariasis) and Plasmodium (causative agent of malaria). Several Bayesian geostatistical logistic regression models with and without spatially structured random effects were fitted for comparison. The model that fitted the data best was used to predict M. perstans infection risk for all of Uganda. Positive associations with M. perstans infection status were observed with cattle densities, forested areas, and vegetation greenness, whereas negative associations were observed with land surface temperature. Only a small geographic overlap was observed with W. bancrofti, and the overall level of co-infection was low (0.3%). However, where the two infections overlapped, a positive association was found. Our study presents the first nationwide geostatistical risk map for M. perstans, and gives important clues about the ecologic preferences of the still unknown main Culicoides vector species in Uganda.

Introduction

The human filarial parasite Mansonella perstans has been considered as one of the most prevalent human parasites in Africa [1]. Despite the wide distribution, only very few studies have addressed its epidemiology and associated health consequences, and currently no effective drug therapy for treatment, control, and local elimination is available [2]. Indeed, M. perstans is viewed as one of the most neglected of the neglected tropical diseases (NTDs) [2].

On-going large-scale surveys and control programs for other filarial infections (e.g., lymphatic filariasis and onchocerciasis), considered to be of greater health importance, have largely ignored M. perstans infections, even though these filarial infections frequently co-occur. This lack of attention mainly stems from its predominance in poor rural communities, and from a paucity of a distinct and clearly recognizable clinical picture [2]. However, widespread co-occurrence with lymphatic filariasis and onchocerciasis could cause complications with regards to control program diagnosis and compliance assessment [2], and could potentially trigger adverse events during mass anti-filaricides administration [3]. It has also been suggested that there could be more subtle effects, as M. perstans might interfere with the host’s immune regulation and influence the susceptibility and effect of other, co-occurring pathogens such as Plasmodium spp. and HIV [2].

The geographic distribution and transmission of M. perstans is closely linked to its vectors, biting midges of the genus Culicoides, and their environmental requirements for breeding and feeding. Culicoides species are widespread throughout the world, and known to transmit a variety of pathogenic viruses, bacteria, protozoa and helminths to humans, and to domestic and wild animals [46]. Yet, they remain among the least studied of the Dipteran vectors [7]. As such, only a few studies have tried to incriminate the exact Culicoides species responsible for transmission of M. perstans in endemic areas in Africa [2]. An accurate understanding of the environmental drivers of both vector and parasite distribution is paramount for the development of spatially explicit risk models based on sound ecological principles, which can help optimize disease prevention planning, and control and elimination programs.

In 2000–2003 a national survey was conducted to map the distribution of M. perstans, concurrently with that of Wuchereria bancrofti [8] and Plasmodium parasites, in school-aged children in Uganda. While geostatistical risk and co-endemicity maps have been constructed for the two latter infections [9], M. perstans infections in Uganda have only crudely been mapped [10]. Furthermore, no risk factor analysis has been performed to identify the underlying environmental drivers of M. perstans infection, and the co-infection rates and geographic overlaps (co-distribution) between the three parasites have yet to be investigated. Delineating areas of geographic overlap, where co-infections might occur, is an important operational issue for integrated disease control planning and implementation [11].

The aim of the present study was to determine the underlying environmental drivers and ecological correlates of the observed prevalence patterns of M. perstans infection and to produce statistically robust prevalence estimates at non-sampled locations (smooth prevalence maps) across Uganda. We furthermore investigated the levels of co-infection and co-distribution with bancroftian filariasis and malaria.

Methods

Ethics Statement

The studies which contributed data used in this paper, received ethical clearance from the Uganda National Council for Science and Technology and were approved by the Central Scientific Ethical Committee of Denmark. Prior to each survey, meetings were held with school staff and village leaders, to explain the objectives and implications of the study. Written informed consent to participate was obtained from those examined (or from the parents/legal guardians of participants aged <15 years). At each study site, a clinical officer from a nearby health unit accompanied the team, examined all the children who were not feeling well and, if need be, either treated the children or referred them to a nearby clinic. For a full description, we refer the reader to prior publications [8, 10].

Study Design and Parasitologic Survey Data

The surveys were carried out between October 2000 and April 2003 and included pupils aged 5–19 years from 76 Ugandan primary schools (12,207 pupils in total) covering the major topographical and ecological zones of the country (see S1 Appendix for a list of schools, with names, geographical coordinates and prevalence). Full details of the study design, data, and the procedures for selection of study sites and participants have been described elsewhere [8,10]. In brief, 100-μl blood sample was collected from each consenting child during the school day and used to prepare a thick film to examine for microfilaremia and Plasmodium parasites. After drying, the thick films were dehemoglobinized, fixed in methanol, stained with Giemsa, and examined under a microscope. All microfilariae observed were identified to species, using morphological criteria [12] and counted. Finger-prick samples of blood were also collected, and assayed for W. bancrofti specific circulating filarial antigens (CFA) by use of ICT cards [1]. Boys and girls were examined in approximately equal numbers.

Environmental and Other Predictor Variables

We investigated a series of climatic and other environmental variables (Table 1) known to be of importance for the distribution of arthropod transmitted parasitic infections in the tropics, but also known ecological drivers of Culicoides species transmitting other parasites and viruses [7,13]. These included measures of temperature, known to influence parasite developmental rate and vectorial development rates, as well as habitat-related factors (i.e., vegetation and land use) and livestock densities that possibly influence the breeding and survival of the (unknown) Culicoides species believed to transmit M. perstans in Uganda.

thumbnail
Table 1. Overview, sources, and resolution of remotely sensed and other geographic information system data used for modeling.

http://dx.doi.org/10.1371/journal.pntd.0004319.t001

The central longitude and latitude of each school obtained using a hand-held global positioning system (GPS; Garmin eTrex, Garmin, Olathe, KS, United States of America) was utilized to define an area of 1 km radius (representing the coarsest resolution of the environmental data) encompassing the community. Average values of each environmental layer were then extracted using ArcGIS 10.1 spatial analyst extension (ESRI; Redlands, CA, United States of America). Land cover variables, calculated as the number of pixels of each category of land use, was counted within the 1 km buffer zone (using the ‘Geospatial Modeling’ environment extension [14]), and the percentage of each category calculated. For a full description of these environmental variables, see Stensgaard and colleagues [15,16].

Statistical Analysis

Initially, a non-spatial, frequentist bivariate logistic regression analysis was conducted in Stata version 13 (Stata Corporation; College Station, TX, United States of America) to assess the relation between various environmental and habitat-related predictors of M. perstans infection status. Significant candidate factors based on likelihood ratio test (LRT) with significance levels of 5% were selected as covariates in further multivariate analyses. To avoid over-parametrization and confounding arising from correlated environmental variables within the same “environmental theme”, these were ranked by the Akaike information criterion (AIC) [17], and strongly correlated variables (Spearman rank correlation r >0.75) excluded.

Next, Bayesian multivariate non-spatial and geostatistical logistic regression models were fitted in OpenBUGS version 3.1.1. (Imperial College and Medical Research Council; London, United Kingdom) via Markov chain Monte Carlo (MCMC) methods which provide higher flexibility in fitting complex models and avoid asymptotic inference than frequentist approaches, and overcome the computational challenges encountered in likelihood-based fitting [18]. Bayesian geostatistical modeling represents the current leading edge in spatial statistics, and makes it possible to incorporate both spatial dependence and covariates, but also enables full representation of uncertainty in model outputs [19] that can be visualized, for example, as maps of prediction errors.

The association between M. perstans, W. bancrofti, and Plasmodium spp. was assessed using multivariate regression models on a single parasite species with all other parasite species as covariates. Demographic and cluster effects were accounted for at the unit of the school.

Model Formulation

We assumed that the M. perstans status Yij of child i at location si, which takes a value of 1 if the child was tested positive and 0 otherwise, follows a Bernoulli distribution Yij ~ Ber(pij), with pij measuring the infection risk at location si. The outcome can be related to its predictors via standard multivariate logistic regression analysis. This model is given by where βij = (β0, β1, β2, … βp) is the vector of regression coefficients and the intercept, and are the model covariates (the fixed part of the model), and εij is a location-level exchangeable random effect that accounts for clustering of individuals in schools. They are assumed to be independent, arising from a normal distribution (∼N(0, τ2)) where τ2 accounts for the non-spatial variation in the infection risk data.

The spatial relationship often found among parasitemia survey locations was considered by introducing spatially correlated random effects ϕi at every sampled location si, which is the standard way of incorporating geographical dependence in the model. The underlying spatial process was modeled by the residuals using the geostatistical design described in Diggle et al. (1998) [18] via a multivariate normal distribution, ϕ = (ϕ1, …. ϕn)T with variance-covariance matrix Σ. Moreover, an isotropic spatial process was assumed, i.e., Σij = σ2exp(−ρdij), where dij is the Euclidean distance between locations i and j, σ2 is the spatial variation (known as the sill), and ρ is a smoothing parameter controlling the rate of correlation decay with increasing distance. For the exponential correlation function, the minimum distance at which the spatial correlation between locations is less than 5% (range of spatial process) is calculated by 3/ρ for the exponential correlation structure.

To complete Bayesian model specification, independent normal prior distributions was assumed for the regression coefficients, with mean 0 and variance 100. For σ2, τ2, and ρ inverse gamma distributions with mean 1 and variance equal to 100 were adopted. We ran a single chain sampler with a burn-in of 5,000 iterations, followed by 100,000 iterations. Convergence was assessed by inspection of ergodic averages of selected model parameters and convergence was successfully achieved before the 100,000th iteration. The strength of correlations and significance of the co-variates was assessed by inspecting the estimated odds ratios (ORs) and their Bayesian credible intervals (BCI).

For appraisal of the best fitting multivariate model, the deviance information criterion (DIC) was applied [20]. The smaller the DIC, the better the model fit. Bayesian kriging was applied to generate smooth risk maps for M. perstans prevalence based on the parameter estimates of the best fitting model [18].

Results

Parasitological Findings

Children with M. perstans microfilaremia were observed in 47 out of the 76 study sites (61.8%), with prevalence ranging from 0.4% to 72.8%. The highest prevalence was observed at sites south of Lake Albert and north-west of Lake Victoria with prevalence decreasing toward zero when moving to the north-eastern and southern sites (Fig 1).

thumbnail
Fig 1. School survey locations and observed prevalence of M. perstans microfilaremia in Uganda.

http://dx.doi.org/10.1371/journal.pntd.0004319.g001

Of the 12,207 children examined for M. perstans microfilaremia, 11,606 were examined concurrently for infection with W. bancrofti (CFA). Co-infections with M. perstans and W. bancrofti were observed in 33 individuals (0.3%) in six schools (Fig 2).

thumbnail
Fig 2. Observed proportional distribution of mono- and co-infections (yellow) with M. perstans (green) and W. bancrofti filariasis (red).

Data from 11,606 pupils aged 5–19 years in 76 schools in Uganda (2000–2003).

http://dx.doi.org/10.1371/journal.pntd.0004319.g002

Four of these six schools were clustered together in the area north of Lake Kyoga that had relatively high W. bancrofti prevalence, but low M. perstans prevalence. However, the few children infected with M. perstans in these schools, also tested positive for W. bancrofti CFA. High levels of mono-infections with M. perstans were primarily found in the southern parts of Uganda, whereas mono-infections with W. bancrofti were restricted to the east-central northern areas of Uganda (Fig 2). Malaria has previously been found to be widespread in Uganda (see Stensgaard et al. (2011) [9] for more details).

Co-infections with M. perstans and Plasmodium spp were observed in 347 of 11,469 examined children (3.0%). Triple-infections with M. perstans, Plasmodium spp, and W. bancrofti were observed in only nine out of 11,267 (0.08%) examined children (in three schools). Co-infections and triple-infections were approximately equally distributed among age groups and sex.

Regression Analyses

The non-spatial bivariate logistic regression analyses revealed that most of the climatic and environmental predictors were significantly associated with M. perstans prevalence (Table 2).

thumbnail
Table 2. Parameter estimates based on bivariate logistic regression models for M. perstans microfilaremia in school-aged children in Uganda (2000–2003).

http://dx.doi.org/10.1371/journal.pntd.0004319.t002

Diurnal land surface temperature (LST) range (Tmax minus Tmin), which was negatively associated with M. perstans prevalence, showed the best fit among the temperature variables as measured by the AIC. Among the vegetation associated variables, NDVI and the percentage of forest cover around the schools showed positive associations with M. perstans infection status, with NDVI composited over the wet season showing the best fit to the data. Among the livestock density factors, only cattle densities showed a significant (and positive) association with M. perstans infection status. Furthermore, age was a significant risk factor, with three times as high odds of being infected in the oldest age group (14–19 years) as compared to the youngest age group (5–9 years). Sex, on the other hand, was not significant at the 5% significance level and thus not included in subsequent Bayesian models.

In the Bayesian multivariate regression analyses (Table 3), the introduction of exchangeable random effects (model B), improved model performance considerably based on DIC estimates (4,991 vs. 3,543).

thumbnail
Table 3. Factors associated with M. perstans microfilaremia in Ugandan school-aged children based on non-spatial and spatial logistic multivariate regression modeling of national survey data (2000–2003).

http://dx.doi.org/10.1371/journal.pntd.0004319.t003

The random effect had also an influence on the regression parameters of the covariates, but all covariates remained significant except forest cover. The introduction of location-specific random effect parameters into the model (model C) further increased model performance (DIC 3,543 vs. 3,389) suggesting that this is the best fitting model, while the covariate parameter estimates remained largely unchanged. Models 2 and 3 estimated approximately the same geographic variability σ2 (3.46 vs. 3.65). The estimated spatial range (above which spatial correlation drops below 5%) was 44.3, which is equivalent to about 7.5 km at the Equator.

Risk Mapping

A predictive M. perstans filariasis risk map for Uganda (Fig 3) was obtained based on the best fitting model, the spatial logistic regression model (model 3).

thumbnail
Fig 3. Geostatistical model-based predicted mean prevalence of Mansonella perstans in school-aged children in Uganda.

Smooth map of the predicted mean prevalence of M. perstans (a), and the corresponding map of the standard deviations of the predictions (b), highlighting areas of high/low uncertainty associated with the model predictions.

http://dx.doi.org/10.1371/journal.pntd.0004319.g003

Highest risks were predicted in the central areas, below Lake Kyoga, with highest prevalence (>20%) south of Lake Albert, and at the northern areas and western shores of Lake Victoria. Intermediate levels (10–20%) were predicted in a belt stretching from south of Lake Albert to the north-eastern shores of Lake Victoria, but with pockets of high risk in the far north-western and south-eastern parts of the country. In contrast, low prevalence estimates (≤1%) were predominantly predicted in the north-east and central-south of Uganda. When interpreting the maps in Fig 3, it should be noted that these are based on model-predictions, and that areas with few survey points may have relatively high levels of associated prediction error. Furthermore, the predictions are based on data from school children only, and thus not necessarily representative for the adult Ugandan population infection levels, which may be considerably higher given the relationship between age and infection risk.

Geographical Overlaps and Parasite-Parasite Associations

The smooth map of the predicted endemic areas (predicted prevalence >5%) of M. perstans from the present study was super-imposed with a map of predicted endemic lymphatic filariasis (>5% W. bancrofti CFA prevalence) and high risk malaria (>50% Plasmodium spp. prevalence) previously published [8] to delineate areas of co-endemicty (Fig 4).

thumbnail
Fig 4. Maps of the predicted geographic co-distribution of M. perstans and W. bancroftibased on A) a 10% prevalence threshold, and B) a 5% prevalence threshold.

The predicted areas are based on surveys of Ugandan school-aged children in 2000–2003 and Bayesian geostatistical model predictions of each (single) parasite infection. Predicted areas of high risk malaria (Plasmodium spp. infection prevalence >50%) is shown in hatch as an overlay.

http://dx.doi.org/10.1371/journal.pntd.0004319.g004

Results from parasite-parasite association inferred from multivariate logistic regression models revealed a significant positive association between M. perstans microfilaremia and testing positive for W. bancrofti (CFA), when clustering at the unit of schools was accounted for (Table 4). No other significant associations were found.

thumbnail
Table 4. Parasite-parasite associations as assessed by multivariate logistic regression models based on a national survey conducted in Uganda (2000–2003).

http://dx.doi.org/10.1371/journal.pntd.0004319.t004

Discussion

The present study provided countrywide, model-based prevalence maps for M. perstans in Uganda, at a high spatial resolution. To our knowledge this is the first study to apply rigorous Bayesian geostatistical risk mapping to national survey data of this neglected human parasitic infection. The study furthermore identified risk factors and displayed high prevalence areas, and thus provides new insights into the ecological preferences of the unknown vector (Culicoides spp.). The resulting maps were finally combined with geostatistical risk maps previously developed for bancroftian filariasis and malaria [9], to delineate overlapping areas (co-distributions) and to investigate levels of co-infection and parasite-parasite associations. Overall, the investigations provide a deeper understanding of the zoogeographical patterns of this widespread, yet little studied parasitic infection, of importance for integrated disease control planning and implementation [11].

An increasing number of geospatial applications now analyze the relationship between parasitic infections and environmental factors, to generate predictive risk maps, including uncertainty estimates [21,22]. The majority of these studies have pertained to malaria risk [2326], but more recently also to a number of NTDs, such as schistosomiasis [2730], lymphatic filariasis [9], loiasis [3132], and soil-transmitted helminthiasis [33,34]. Besides being useful for spatial targeting of control measures, surveillance, and measuring progress toward elimination, these studies can also give important new insights and clues about the ecology of the parasites and their vectors or intermediate hosts.

Environmental Drivers

Because of the limited number of studies of M. perstans epidemiology in Africa, the current knowledge about the climatic and other environmental factors that help drive transmission of this filarial parasite is very scarce [2]. Our results indicated that high prevalence of M. perstans in Uganda was associated with cooler areas with little diurnal temperature variation, and with high NDVI values, a surrogate variable for soil moisture and correlated with vegetation biomass. A positive association was also observed with forested areas (although not significant in the final model). An association with forested ecosystems was also observed in Gabon [35,36] and Cameroon [37], and historically by Low (1903) [38], who noticed that high prevalence was associated with tropical forests alternating with swamps and other large, open ground areas. Other reports from West Africa have related the common occurrence of M. perstans along the forest fringes between the rain forest and open land, to the particular species and density of the vectors [39,40], while studies from central/southern Africa similarly indicated high prevalence in or near dense forest [4144].

The association to forested areas, as well as banana plantations [45,46], has been linked to the importance of decomposing woody material, tree holes, and forest floor cover as breeding sites for the vector species, Culicoides spp. However, besides the affinity for moist substrates the some 1,400 described species of Culicoides show a highly diverse range of species habitat preferences, ranging from salt- and freshwater marshes, to animal dung, water logged pastures, and leaking irrigation pipes [5].

In Uganda, a total of 31 Culicoides species have been listed thus far [47], and several of them (e.g., C. grahamii), which have been identified as vectors of M. perstans in the Congo [48] and Cameroon [39], occur in areas predicted to be endemic for M. perstans [47]. Yet, no studies have been carried out to confirm the role of this or other Culicoides species in the transmission of M. perstans in Uganda.

Here, the identified environmental correlates of M. perstans give us important clues about the bionomics of the unknown vector species. Besides the climatic associations, areas of high prevalence of M. perstans were found to coincide with areas of high cattle densities (but not densities of other types of livestock). This is interesting and calls for further investigation, as M. perstans also has been shown to occur at high prevalence in Fulani nomads (cattle raising people) in northern Nigeria [49]. Possible explanations could be either a role for cattle in providing a steady source of blood meals for an opportunistic biting, or in creating habitats for larval development. Animal dung has been shown to provide important larval habitats for several Culicoides species [50,51] and, C. grahamii, for example, has been incriminated in blue tongue virus transmission among cattle in Kenya [52].

Parasite Co-Distributions and Associations

While models and maps of individual parasite infections are valuable, the distribution of these infections rarely occurs independently of each other. Multiple species are often found within populations (co-endemicity) and individuals (co-infection), and co-infections are increasingly being recognized to have important health consequences [5356]. Concomitant infections with helminths have, for example, been shown to increase susceptibility to infection with P. falciparum [57,58]. Delineating areas of geographic overlap, where co-infections might occur, is thus an important operational issue for integrated disease control planning, implementation, and evaluation.

In Uganda, several NTDs have been reported to be co-endemic [59], and also to be co-endemic with malaria [9]. Here we found that while perstans filariasis was widely overlapping with areas of high malaria risk (Fig 3), there were no significant associations between infections with Plasmodium spp and M. perstans and/or W. bancrofti at the individual level or at the unit of the school. A similar result was found by Nielsen et al. (2006) [60], in a study from north-eastern Tanzania, whereas Kelly-Hope et al. (2006) found a negative spatial association between W. bancrofti and P. falciparum malaria prevalence in West Africa [61].

In contrast, we observed a distinct pattern of geographic segregation between M. perstans and W. bancrofti, another filarial parasite of human health importance in Uganda (Fig 2). While M. perstans was mainly predicted in the central-to-southern parts of the country, W. bancrofti dominated in the central-northern parts. This pattern has been noted previously [10], but this is the first time a co-distribution map based on rigorous geostatistical modeling of individual infections is presented.

This very likely reflects the ecological distinctiveness of the M. perstans Culicoides spp. vector compared to that of the Anopheles mosquitoes transmitting W. bancrofti in Uganda. It is noted, for instance, that while high M. perstans infection risk is related to forested and densely vegetated areas, the opposite seems to apply for W. bancrofti, which showed a negative association with NDVI [9]. Similar contrasting epidemiologies have been shown for other filarial infections in Africa, i.e., between onchocerciasis and loiasis in the Democratic Republic of the Congo (DRC) [62], although L. loa and M. perstans have been found to coexist with high prevalence geographically in some African countries [35,37].

The limited geographical overlap observed in Uganda, explains the relatively low levels of overall filarial co-infection (0.8%). Yet, where the two parasites did overlap in space, in a high prevalent W. bancrofti foci in central Uganda at the northern geographical range margin of M. perstans plus a location south of Lake Albert, with high M. perstans prevalence (Figs 2 and 3), a positive association was observed between the two species. This finding warrants further investigation of potential risk factors for co-infection at the school and individual level in Uganda, and indicates that special attention should be paid to children living in geographically overlapping areas, even if these areas may be few.

In conclusion, this study adds further to our knowledge about the distinct zoogeography of filarial parasites [11] in Africa. The observed correlation between M. perstans prevalence and cattle density warrants further scientific inquiry, particularly the role played by livestock as either opportunistic blood meals (resource) for the Culicoides midges and/or the role of dung as larval habitats. Finally, we urge further studies based on geographically stratified field-collections of Culicoides, to clarify the identity, bionomics, and behavior of the vector species of M. perstans in Uganda and elsewhere in Africa, as this is a vital piece of the puzzle toward a fuller understanding of the transmission cycle and epidemiology of M. perstans infections.

Supporting Information

S1 Appendix. Table of surveyed schools with names, prevalence and geographical coordinates.

doi:10.1371/journal.pntd.0004319.s001

(XLS)

Author Contributions

Conceived and designed the experiments: ASS AWO PES EMP. Analyzed the data: ASS PV. Wrote the paper: ASS PV AWO JU TKK PES. Designed and carried out the parasitologic field surveys: AWO PES EMP. Read and approved the final manuscript: ASS PV AWO PES EMP JU TKK.

References

  1. 1. Nelson GS (1965) Filarial infections as zoonoses. J Helminthol 39: 229–250. pmid:5328829 doi: 10.1017/s0022149x00020630
  2. 2. Simonsen PE, Onapa AW, Asio SM (2011) Mansonella perstans filariasis in Africa. Acta Trop 120 Suppl 1: S109–120. doi: 10.1016/j.actatropica.2010.01.014. pmid:20152790
  3. 3. Kelly-Hope LA, Cano J, Stanton MC, Bockarie MJ, Molyneux DH (2014) Innovative tools for assessing risks for severe adverse events in areas of overlapping Loa loa and other filarial distributions: the application of micro-stratification mapping. Parasit Vectors 4: 178. doi: 10.1186/1756-3305-7-307
  4. 4. Linley JR, Hoch AL, Pinheiro FP (1983) Biting midges (Diptera: Ceratopogonidae) and human health. J Med Entomol 20: 347–364. pmid:6312046 doi: 10.1093/jmedent/20.4.347
  5. 5. Linley JR (1985) Biting midges (Diptera: Ceratopogonidae) as vectors of nonviral animal pathogens. J Med Entomol 22: 589–599. pmid:3908679 doi: 10.1093/jmedent/22.6.589
  6. 6. Mellor PS, Leake CJ (2000) Climatic and geographic influences on arboviral infections and vectors. Revue Scientifique et Technique de l’Office International des Epizooties 19: 41–54.
  7. 7. Carpenter S, Groschup MH, Garros C, Felippe-Bauer ML, Purse BV (2013) Culicoides biting midges, arboviruses and public health in Europe. Antivir Res 100: 102–113. doi: 10.1016/j.antiviral.2013.07.020. pmid:23933421
  8. 8. Onapa AW, Simonsen PE, Baehr I, Pedersen EM (2005) Rapid assessment of the geographical distribution of lymphatic filariasis in Uganda, by screening of schoolchildren for circulating filarial antigens. Ann Trop Med Parasitol 99: 141–153. pmid:15814033 doi: 10.1179/136485905x19829
  9. 9. Stensgaard AS, Vounatsou P, Onapa AW, Simonsen PE, Pedersen EM, et al. (2011) Bayesian geostatistical modelling of malaria and lymphatic filariasis infections in Uganda: predictors of risk and geographical patterns of co-endemicity. Malar J 10: 298. doi: 10.1186/1475-2875-10-298. pmid:21989409
  10. 10. Onapa AW, Simonsen PE, Baehr I, Pedersen EM (2005) Rapid assessment of the geographical distribution of Mansonella perstans infections in Uganda, by screening schoolchildren for microfilariae. Ann Trop Med Parasitol 99: 383–393. pmid:15949186 doi: 10.1179/136485905x361990
  11. 11. Molyneux DH, Mitre E, Bockarie MJ, Kelly-Hope LA (2014) Filaria zoogeography in Africa: ecology, competitive exclusion, and public health relevance. Trends Parasitol 30: 163–169. doi: 10.1016/j.pt.2014.02.002. pmid:24636357
  12. 12. WHO (1987). Control of lymphatic filariasis. A manual for health personnel. Geneva: World Health Organization.
  13. 13. Calvete C, Estrada R, Miranda MA, Borras D, Calvo JH, et al. (2009) Ecological correlates of bluetongue virus in Spain: predicted spatial occurrence and its relationship with the observed abundance of the potential Culicoides spp. vector. Vet J 182: 235–243. doi: 10.1016/j.tvjl.2008.06.010. pmid:18667341
  14. 14. Beyer HL (2012) Geospatial modelling environment (version 0.7.2.1). Spatial Ecology, LLC.
  15. 15. Stensgaard A, Jorgensen A, Kabatereine NB, Malone JB, Kristensen TK (2005) Modeling the distribution of Schistosoma mansoni and host snails in Uganda using satellite sensor data and geographical information systems. Parassitologia 47: 115–125. pmid:16044680
  16. 16. Stensgaard AS, Jorgensen A, Kabatereine NB, Rahbek C, Kristensen TK (2006) Modeling freshwater snail habitat suitability and areas of potential snail-borne disease transmission in Uganda. Geospat Health 1: 93–104. pmid:18686235 doi: 10.4081/gh.2006.284
  17. 17. Akaike H (1973) Information theory and an extension of the maximum likelihood principle. In: Petrov B, Csaki F, editors. Second international symposium on information theory. Budapest: Akademiai Kiado. pp. 267–281.
  18. 18. Diggle PJ, Tawn JA, Moyeed RA (1998) Model-based geostatistics. J R Stat Soc Ser C Appl Stat 47: 299–326. doi: 10.1111/1467-9876.00113
  19. 19. Best N, Richardson S, Thomson A (2005) A comparison of Bayesian spatial models for disease mapping. StatMethods Med Res 14: 35–59. doi: 10.1191/0962280205sm388oa
  20. 20. Spiegelhalter DJ, Best NG, Carlin BR, van der Linde A (2002) Bayesian measures of model complexity and fit. J R Stat Soc Ser B Stat Methodol 64: 583–616. doi: 10.1111/1467-9868.00353
  21. 21. Simoonga C, Utzinger J, Brooker S, Vounatsou P, Appleton CC, et al. (2009) Remote sensing, geographical information system and spatial analysis for schistosomiasis epidemiology and ecology in Africa. Parasitology 136: 1683–1693. doi: 10.1017/S0031182009006222. pmid:19627627
  22. 22. Hürlimann E, Schur N, Boutsika K, Stensgaard AS, de Himpsl ML, et al. (2011) Toward an open-access global database for mapping, control, and surveillance of neglected tropical diseases. PLoS Negl Trop Dis 5: e1404. doi: 10.1371/journal.pntd.0001404. pmid:22180793
  23. 23. Gemperli A, Vounatsou P, Sogoba N, Smith T (2006) Malaria mapping using transmission models: application to survey data from Mali. Am J Epidemiol 163: 289–297. pmid:16357113 doi: 10.1093/aje/kwj026
  24. 24. Gosoniu L, Vounatsou P, Sogoba N, Maire N, Smith T (2009) Mapping malaria risk in West Africa using a Bayesian nonparametric non-stationary model. Comput Stat Data Anal 53: 3358–3371. doi: 10.1016/j.csda.2009.02.022.
  25. 25. Gosoniu L, Veta AM, Vounatsou P (2010) Bayesian geostatistical modeling of Malaria Indicator Survey data in Angola. PLoS One, 5: e9322. doi: 10.1371/journal.pone.0009322. pmid:20351775
  26. 26. Raso G, Schur N, Utzinger J, Koudou BG, Tchicaya ES, et al. (2012) Mapping malaria risk among children in Côte d'Ivoire using Bayesian geo-statistical models. Malar J, 11: 160. doi: 10.1186/1475-2875-11-160. pmid:22571469
  27. 27. Beck-Wörner C, Raso G, Vounatsou P, N'Goran EK, Rigo G, et al. (2007) Bayesian spatial risk prediction of Schistosoma mansoni infection in western Côte d'Ivoire using a remotely-sensed digital elevation model. Am J Trop Med and Hyg 76: 956–963.
  28. 28. Raso G, Matthys B, N'Goran EK, Tanner M, Vounatsou P, et al. (2005) Spatial risk prediction and mapping of Schistosoma mansoni infections among schoolchildren living in western Côte d'Ivoire. Parasitology 131: 97–108. pmid:16038401 doi: 10.1017/s0031182005007432
  29. 29. Schur N, Hürlimann E, Garba A, Traore MS, Ndir O, et al. (2011) Geostatistical model-based estimates of schistosomiasis prevalence among individuals aged ≤20 years in West Africa. PLoS Negl Trop Dis 5: e1194. doi: 10.1371/journal.pntd.0001194. pmid:21695107
  30. 30. Vounatsou P, Raso G, Tanner M, N'Goran EK, Utzinger J (2009) Bayesian geostatistical modelling for mapping schistosomiasis transmission. Parasitology 136: 1695–1705. doi: 10.1017/S003118200900599X. pmid:19490724
  31. 31. Zoure HG, Wanji S, Noma M, Amazigo UV, Diggle PJ, et al. (2011) The geographic distribution of Loa loa in Africa: results of large-scale implementation of the Rapid Assessment Procedure for Loiasis (RAPLOA). PLoS Negl Trop Dis 5: e1210. doi: 10.1371/journal.pntd.0001210. pmid:21738809
  32. 32. Diggle PJ, Thomson MC, Christensen OF, Rowlingson B, Obsomer V, et al. (2007) Spatial modelling and the prediction of Loa loa risk: decision making under uncertainty. Ann Trop Med Parasitol 101: 499–509. pmid:17716433 doi: 10.1179/136485907x229121
  33. 33. Karagiannis-Voules DA, Biedermann P, Ekpo UF, Garba A, Langer E, et al.(2015) Spatial and temporal distribution of soil-transmitted helminth infection in sub-Saharan Africa: a systematic review and geostatistical meta-analysis. Lancet Infect Dis 15: 74–84 doi: 10.1016/S1473-3099(14)71004-7 pmid:25486852.
  34. 34. Karagiannis-Voules DA, Odermatt P, Biedermann P, Khieu V, Schär F, et al. (2015) Geostatistical modelling of soil-transmitted helminth infection in Cambodia: do socioeconomic factors improve predictions? Acta Trop 141: 204–212. doi: 10.1016/j.actatropica.2014.09.001 pmid:25205492.
  35. 35. Akue JP, Nkoghe D, Padilla C, Moussavou G, Moukana H, et al. (2011) Epidemiology of concomitant infection due to Loa loa and Mansonella perstans in Gabon. Plos Negl Trop Dis 5: e1329. doi: 10.1371/journal.pntd.0001329. pmid:22022623
  36. 36. Van Hoegaerden M, Chabaud B, Akue JP, Ivanoff B (1987) Filariasis due to Loa loa and Mansonella perstans: distribution in the region of Okondja, Haut-Ogooue Province, Gabon, with parasitological and serological follow-up over one year. Trans R Soc Trop Med Hyg 81: 441–446. pmid:3479854 doi: 10.1016/0035-9203(87)90163-5
  37. 37. Mommers EC, Dekker HS, Richard P, Carcia A, Chippaux JP (1995) Prevalence of L. loa and M. perstans filariasis in Southern Cameroon. Trop Geogr Med 47: 2–5. pmid:7747326
  38. 38. Low G. C. (1903). Filaria perstans. British Medical Journal, i, 722–724. doi: 10.1136/bmj.1.2204.722
  39. 39. Kershaw WE, Keay RWJ, Nicholas WL, Zahra A (1953) Studies on the epidemiology of filariasis in West Africa, with special reference to the British Cameroons and the Niger Delta. IV The incidence of Loa-loa and Acanthocheilonema perstans in the rain-forest, the forest fringe and the mountain grasslands of the british Cameroons, with observations on the species of Chrysops and Culicoides found. Ann Trop MedParasit 47: 406–425.
  40. 40. Nicholas WL, Kershaw WE, Keay RWJ, Zahra A (1953) Studies on the epidemiology of filariasis in West Africa, with special reference to the british Cameroons and the Niger Delta. III The distribution of Culicoides spp biting man in the rain-forest, the forest fringe and the mountain grasslands of the british Cameroons. Ann Trop MedParasit 47: 95–111.
  41. 41. Clarke Vde V, Harwin RM, MacDonald DF, Green CA, Rittey DA (1971) Filariasis: Dipetalonema perstans infections in Rhodesia. Cent Afr J Med 17: 1–11. pmid:5103154
  42. 42. Udonsi JK (1986) The status of human filariasis in relation to clinical signs in endemic areas of the Niger Delta. Ann Trop Med Parasitol 80: 425–432. pmid:3539045
  43. 43. Vanhoegaerden M, Chabaud B, Akue JP, Ivanoff B (1987) Filariasis due to Loa loa and Mansonella perstans: distribution in the region of Okondja, Haut-Ogooue Province, Gabon, with parasitological and serological follow-up over one year. Trans R Soc Trop Med Hyg 81: 441–446. pmid:3479854 doi: 10.1016/0035-9203(87)90163-5
  44. 44. Noireau F, Carme B, Apembet JD, Gouteux JP (1989) Loa loa and Mansonella perstans filariasis in the Chaillu mountains, Congo: parasitological prevalence. Trans R Soc Trop Med Hyg 83: 529–534. pmid:2617607 doi: 10.1016/0035-9203(89)90280-0
  45. 45. Hopkins CA (1952) Notes on the biology of certain Culicoides studied in the British Cameroons, West Africa, together with observations on their possible role as vectors of Acanthocheilonema perstans. AnnTrop MedParasit 46: 165–&.
  46. 46. Nicholas WL (1953) The Dispersal of Culicoides grahamii and C. austeni from their breeding sites prior to their taking a blood-meal. Ann Trop Med Parasit 47: 309–323. pmid:13105263
  47. 47. Khamala CPM (1971) Ecological distribution of East African Culicoides latreille (Dipt, Ceratopogonidae) as shown by light-traps. Bull Ent Res 60: 549–557. doi: 10.1017/s0007485300042310
  48. 48. Noireau F, Itoua A, Carme B (1990) Epidemiology of Mansonella perstans filariasis in the forest region of South Congo. Ann Trop Med Parasit 84: 251–254. pmid:2222027
  49. 49. Anosike JC, Dozie INS, Onwuliri COE, Nwoke BEB, Onwuliri VA (2005) Prevalence of Mansonella perstans infections among the nomadic Fulanis of northern Nigeria. Ann Agric Environ Med 12: 35–38. pmid:16028864
  50. 50. Walker AR (1977) Seasonal fluctuations of Culicoides species (Diptera Ceratopogonidae) in Kenya. Bull Entomol Res 67: 217–233. doi: 10.1017/s0007485300011032
  51. 51. Thompson GM, Jess S, Murchie AK (2013) Differential emergence of Culicoides (Diptera: Ceratopogonidae) from on-farm breeding substrates in Northern Ireland. Parasitology 140: 699–708. doi: 10.1017/S0031182012002016. pmid:23363545
  52. 52. Walker AR (1977) Culicoides as potential vectors of viruses to livestock in Kenya. Mosquito News 37: 285–286.
  53. 53. Mwangi TW, Bethony JM, Brooker S (2006) Malaria and helminth interactions in humans: an epidemiological viewpoint. Ann Trop Med Parasitol 100: 551–570. pmid:16989681 doi: 10.1179/136485906x118468
  54. 54. Graham AL (2008) Ecological rules governing helminth-microparasite coinfection. Proc Natl Acad Sci USA 105: 566–570. doi: 10.1073/pnas.0707221105. pmid:18182496
  55. 55. Brooker S, Akhwale W, Pullan R, Estambale B, Clarke SE, et al. (2007) Epidemiology of Plasmodium-helminth co-infection in Africa: populations at risk, potential impact on anemia, and prospects for combining control. Am J Trop Med Hyg 77: 88–98. pmid:18165479
  56. 56. Eziefula AC, Brown M (2008) Intestinal nematodes: disease burden, deworming and the potential importance of co-infection. Curr Opin Infect Dis 21: 516–522. doi: 10.1097/QCO.0b013e32830f97fd. pmid:18725802
  57. 57. Hillier SD, Booth M, Muhangi L, Khihembo M, Mohammed K, et al. (2006) Malaria and helminth co-infection in a semi-urban population of pregnant women in Uganda. Am J Trop Med Hyg 75: 161–161.
  58. 58. Booth M (2006) The role of residential location in apparent helminth and malaria associations. Trends Parasitol 22: 359–362. pmid:16797235 doi: 10.1016/j.pt.2006.06.007
  59. 59. Kolaczinski JH, Kabatereine NB, Onapa AW, Ndyomugyenyi R, Kakembo ASL, et al. (2007) Neglected tropical diseases in Uganda: the prospect and challenge of integrated control. Trends Parasitol 23: 485–493. pmid:17826335 doi: 10.1016/j.pt.2007.08.007
  60. 60. Nielsen NO, Simonsen PE, Magnussen P, Magesa S, Friis H (2006) Cross-sectional relationship between HIV, lymphatic filariasis and other parasitic infections in adults in coastal northeastern Tanzania. Trans R Soc Trop Med Hyg 100: 543–550. pmid:16324731 doi: 10.1016/j.trstmh.2005.08.016
  61. 61. Kelly-Hope LA, Diggle PJ, Rowlingson BS, Gyapong JO, Kyelem D, et al. (2006) Negative spatial association between lymphatic filariasis and malaria in West Africa. Trop Med Int Health 11: 129–135. pmid:16451336 doi: 10.1111/j.1365-3156.2005.01558.x
  62. 62. Tekle AH, Zoure H, Wanji S, Leak S, Noma M, et al. (2011) Integrated rapid mapping of onchocerciasis and loiasis in the Democratic Republic of Congo: impact on control strategies. Acta Trop 120 Suppl 1: S81–90. doi: 10.1016/j.actatropica.2010.05.008. pmid:20525531