In sub-Saharan Africa, African animal trypanosomoses (AAT), transmitted by tsetse flies (genus Glossina), are among the main constraints to the development of cattle farming (Itard et al., 2003). Demographic and climatic pressures have modified tsetse habitats, which are more and more fragmented (Van den Bossche et al., 2010). In the Mouhoun river basin, Burkina Faso, the cyclical vectors of AAT are riverine tsetse species, namely Glossina palpalis gambiensis Vanderplank (G.p.g.) and Glossina tachinoides Westwood (G. t.) (Diptera: Glossinidae). The fragmentation of tsetse habitats was studied repeatedly in this area, in order to understand its impact on tsetse distribution and densities (Bouyer et al., 2005; Guerrini et al., 2008), population structure and dispersal (Bouyer et al., 2007, 2009; Koné et al., 2011a).

The risk of animal trypanosomoses was also mapped, based on the number of infectious fly per trap per day as a risk indicator (Bouyer et al., 2006; Guerrini and Bouyer, 2007). This risk mapping revealed that the most dangerous river sections were located at the border of protected forests, due to a “border effect” leading to increased tsetse densities. No difference was, however, observed between the tsetse infection rates in the different landscapes and no environmental factor could be associated to this infection rate.

The goal of the present study was to explore the relationship between tsetse infection rate and various environmental factors at a local scale (the Mouhoun river basin).

Trypanosomosis is also an infection for tsetse, which in turn try to control it through pre-existing defenses and a immune gene expression that is induced upon feeding on trypanosomes (Dyer et al., 2013 # 1145). Many studies of vector manipulation by parasites demonstrated that parasites can impact vector survival, trophic behavior, fecundity (Lefevre and Thomas, 2008). Studies conducted in experimental conditions revealed that nutritional stress can increase their susceptibility to infection by trypanosomes (Akoda, 2009). For example, starvation increased the infection rates of Glossina morsitans morsitans Westwood with Trypanosoma congolense and the maturation rate of T. brucei brucei infections (Kubi et al., 2006). It was thereafter demonstrated that fly starvation can lead to a decreased expression of immune genes in newly hatched flies or a lack of immune responsiveness to trypanosomes in older flies, that can contribute to increased susceptibility of nutritionally stressed tsetse flies to trypanosome infection (Akoda et al., 2009b). In the case of tsetse infections with Trypanosoma brucei brucei, nutritional stress can also result in enhanced maturation of the trypanosome infection (Akoda et al., 2009a). It was also demonstrated that exposure of adult flies to high temperatures can increase the development rate of trypanosomes in adult flies (Leak, 1998). Moreover, flies emerging from pupae exposed to high temperatures are also more sensitive to infections by T. b. rhodesiense (Burtt, 1946) and T. congolense (Ndegwa et al., 1992). The impact of heat stress on tsetse immunity might thus be one of the explanations for the pattern observed in this study.

Under field conditions, a positive correlation was also observed between latitude, associated to increasing mean annual temperature and tsetse infection rates (Ford and Leggate, 1961). The authors stated that “at higher altitudes or, as in Zululand, a situation approaching the temperate zone, low infections occur as a response to cooler climates.” In that study, the regression of proboscis-type infections on mean annual temperature was much more marked than that of gut and/or gut and proboscis-type infections, which is surprising given that a decreased immune response associated to heat stress is probably involved.

In our study, we observed a similar pattern at a more local scale and between seasons, i.e., a strong increase of infection rates with mean monthly maximum temperature. Moreover, the regression of all infections together on mean monthly temperature was less marked than for that of gut and/or gut and proboscis-type infections (T. congolense). Since we used monthly average temperatures, both exposure of the pupae and the adult flies could have contributed to the observed pattern. Cattle rearing systems and farmer habits have been characterized in the study sites (Koné et al., 2012), together with a monitoring of AAT prevalence in sentinel herds (Métras et al., 2008). The sociological study showed that herders, particularly those of the Fulani ethnic group, tend to avoid the main river during the rainy season, when temporary ponds are available. Tsetse are, however, able to disperse outside the gallery forest during this season, up to 2km in the neighboring savannah (Cuisance et al., 1985; Vreysen et al., 2013). Overall, the AAT infection rates in cattle were similar between sites but showed a significant peak in July (middle of the rainy season), which is different from the peak of infection observed in tsetse (hot dry season, from April to May, when temperatures were highest). It is thus not likely that the pattern observed in this study was driven by an increased contact with infected hosts and this indirect effect would not give such a strong signal.

The impact of temperature on the maturation rate of T. congolense was not significant, which might be due to overall very high maturation rates observed. This maturation rate might have been overestimated by considering any fly positive in its proboscis as a mature infection.

Mapping vectorial diseases rely on mapping different risk indicators, like the vectorial capacity (Tran et al., 2005). The main elements of the vectorial capacity are the relative density of vectors to hosts, the survival of vectors, their feeding preferences and their infectious rate. Recent AAT risk modeling efforts based on satellite images in the Mouhoun river mainly relied on mapping tsetse catches per trap per day, which are considered to be correlated to their relative densities to the hosts (Bouyer et al., 2006; Guerrini and Bouyer, 2007). Even if the number of infectious fly per trap per day were considered, corresponding to the entomological inoculation rate, an indicator commonly used to map malaria risk for example, no environmental predictor of tsetse infection rate was evidenced. The integration of the results presented here in these AAT mapping efforts would probably lead to improved accuracies of the maps, and since the densities of tsetse are relatively stable over time (k demographic strategy) (Koné et al., 2011b), it might also allow making seasonal predictions. Of course, mechanical transmission cannot be neglected (Desquesnes et al., 2009) and can lead to a risk of trypanosomosis independent from tsetse, including outside the tsetse belt (Pagabeleguem et al., 2012). Moreover, high temperature might have contradictory impacts on different factors of transmission, for example by reducing tsetse lifespan and density (Van den Bossche et al., 2010).

In our study, it was surprising that air temperature measured from meteorological stations located quite far from the study sites better predicted tsetse infection rates than the MODIS LSTd measured in the 1^*1 km pixels where the tsetse were captured. This is probably due to discrepancies between LSTd and air temperature, in relation to land cover, whereas the maximum air temperatures in the meteorological stations were probably highly correlated to those of the study sites. It would be very interesting to conduct a similar study including the monitoring of air temperature and hygrometry using field meteorological stations. In our study, the correlation between LSTd and max temperature was very high (r² = 0.77, p < 10⁻³). LSTd was still highly correlated to the infection rate, even if less powerful than max temperature to predict it. LSTd is a MODIS composite synthetic variable. In Senegal, we, however, observed important discrepancies between air temperature measured on the ground and LSTd. Other indexes like NDVI and MIR could be combined to the latter to better predict air temperature, with an accuracy of ~1°c. Actually, these parameters are correlated to land cover, which in turn impacts air temperature. Efforts are presently underway in Senegal to build models of air temperatures using a network of field meteorological stations. The inverse relationship observed with monthly rainfalls in the case of T. congolense might be due to a mitigation of the evapotranspiration caused by high temperatures by an increase of relative hygrometry. Actually, tsetse survive longer at high temperatures when relative hygrometry is higher (Buxton, 1955).

A recent model predicted a large-range expansion of sleeping sickness in East and South Africa, caused mainly by a shift of up to 60 per cent of its geographical extent (Moore et al., 2012). This change was mainly anticipated from predicted changes in global temperatures impacting on the following entomological and epidemiological parameters: total tsetse population size, natural death rate of tsetse, parasite maturation rate in tsetse and tsetse biting rate. However, the potential impact of temperature on tsetse primary infection rate was not considered and might improve greatly model predictions. The raw data generated in this survey are made available to encourage modeling efforts (Supplementary file 1). Trap locations can also be requested from the corresponding author.