A spatial genetics approach to inform vector control of tsetse flies (Glossina fuscipes fuscipes) in Northern Uganda

Abstract

Tsetse flies (genus Glossina) are the only vector for the parasitic trypanosomes responsible for sleeping sickness and nagana across sub-Saharan Africa. In Uganda, the tsetse fly Glossina fuscipes fuscipes is responsible for transmission of the parasite in 90% of sleeping sickness cases, and co-occurrence of both forms of human-infective trypanosomes makes vector control a priority. We use population genetic data from 38 samples from northern Uganda in a novel methodological pipeline that integrates genetic data, remotely sensed environmental data, and hundreds of field-survey observations. This methodological pipeline identifies isolated habitat by first identifying environmental parameters correlated with genetic differentiation, second, predicting spatial connectivity using field-survey observations and the most predictive environmental parameter(s), and third, overlaying the connectivity surface onto a habitat suitability map. Results from this pipeline indicated that net photosynthesis was the strongest predictor of genetic differentiation in G. f. fuscipes in northern Uganda. The resulting connectivity surface identified a large area of well-connected habitat in northwestern Uganda, and twenty-four isolated patches on the northeastern margin of the G. f. fuscipes distribution. We tested this novel methodological pipeline by completing an ad hoc sample and genetic screen of G. f. fuscipes samples from a model-predicted isolated patch, and evaluated whether the ad hoc sample was in fact as genetically isolated as predicted. Results indicated that genetic isolation of the ad hoc sample was as genetically isolated as predicted, with differentiation well above estimates made in samples from within well-connected habitat separated by similar geographic distances. This work has important practical implications for the control of tsetse and other disease vectors, because it provides a way to identify isolated populations where it will be safer and easier to implement vector control and that should be prioritized as study sites during the development and improvement of vector control methods.

Keywords: landscape genetics; maximum entropy model; sleeping sickness; spatial genetics; tsetse fly; vector control.

Figure 1

Map showing the spatial context of the study in northern Uganda. Sampling sites used for the population genetic input data are indicated as black dots. Numbers are the same as in Table S1 (Appendix S1), where information on these sites is reported. The map also shows the distribution of the two Trypanosoma parasites, Trypanosoma brucei gambiense to the west and T. b. rhodesiense to the east (gray lines), responsible for the chronic and acute form of the HAT disease. Water bodies (rivers and lakes) are shown in light gray with the major ones identified by name. The map also reports the district names for the region

Figure 2

Flow diagram of the methodological pipeline. Inputs (I₁, I_2, and I₃) are shown as parallelograms, methods (M₁–M₆) as rectangles, and outputs (O₁–O₆) as ovals

Figure 3

Model outputs for the top‐scoring environmental variable, net photosynthesis (PSN) obtained using Circuitscape. (a) Map showing the resistance surface costs for the only environmental variable strongly correlated with genetic differentiation, net photosynthesis (PSN, Table 1). Resistance costs vary from dark green to dark red reflecting areas of low and high resistance to tsetse movement, respectively. (b) The output of Circuitscape analysis showing the current map of the modeled connectivity expressed as current density, varying for low (black) to high (white) connectivity

Figure 4

Map showing the connectivity surface based on net photosynthesis (PSN) and 317 presence data and using a univariate MaxEnt (Elith et al., 2011) analysis. The map also shows the location of discrete isolated patches in purple identified with tools implemented in R (see Figure 2 for details)

Figure 5

Habitat suitability maps for G. f. fuscipes in northern Uganda: (a) updated habitat suitability map obtained using 317 presence data, 12 environmental variable relevant to tsetse ecology (Table 1), and a canonical multivariate MaxEnt (Elith et al., 2011) analysis. This map also shows the twenty‐four isolated patches identified by the model (gray polygons), the three transects (black segments) used for the field survey, and the location of the tsetse sample from one of the isolated patches used to validate the method; (b) FAO habitat suitability map for G. f. fuscipes (Wint & Rogers, 2000). The legend to the right of each map explains the map color scheme, ranging from dark red (highly suitable habitat) to green (unsuitable habitat). Water bodies are shown in light blue

Figure 6

Histogram of genetic differentiation found between samples at geographic distances of 25–100 km. Pairwise estimates from within the main continuous habitat are shown in red, and pairwise estimates including the ad hoc sample from the model‐predicted isolated patch are shown in blue. F_ST was computed in ARLEQUIN (Excoffier & Lischer, 2010; Wright, 1951) adjusted for finite populations (Rousset, 1997) using the equation F_ST/(1 − F_ST)