Learning Affects Host Preference in Tsetse Flies

Tsetse flies are very efficient cyclic vectors of African trypanosomosis. Since tsetse are generally infected by the first blood meal, as in the case of sleeping sickness for example, any propensity to feed on the same host a second time will improve transmission within this host species, whereas transmission between host species will decrease. To test this hypothesis we presented a monitor lizard and a cow in a stable to marked tsetse flies that had first fed on one of these two hosts. 80% of the teneral flies that fed did so on the cow when provided the choice. Among the flies having feeding experience, a disproportionately high number of flies that had fed on one host returned to this host for the second meal. We discuss the energetic advantages of such a learning behavior and its importance in sleeping sickness epidemiology. The findings are of relevance to the role played by such learning behavior in disease transmission by other insect vectors of zoonoses.


■ INTRODUCTION
Even though the tsetse fly vectors of African trypanosomosis are sometimes present at modest densities, any relaxation of sanitary measures results in recrudescence of the disease (14).There are therefore some underlying mechanisms serving to optimize transmission cycles.In West Africa, the main vectors of sleeping sickness are tsetse flies of the palpalis group (subgenus Nemorhina) that thrive in vegetation along rivers (3).These riparian subspecies have a well known host range of reptiles, ruminants and man (1,4,16).Local host availability contributes to significant differences in the origins of blood meals between populations of the same tsetse species (6).In hymenoptera, it is well known that foraging experience can affect host selection (8,13).The existence of such learning mechanisms in insect vectors is being elucidated and has tremendous epidemiological implications (9), but has not been shown, as yet, for tsetse flies.Since tsetse flies are generally infected by Trypanosoma brucei in the first blood meal, any propensity to feed on the same host will improve transmission within this host species.Correspondingly, transmission between host species will decrease.Here we show that the first host chosen by teneral tsetse influences them to feed on the same one for the next blood meal.

■ MATERIALS AND METHODS
All the experiments took place at CIRDES, Burkina Faso.To test the effect of the first blood meal on subsequent host choice, separate groups of laboratory-reared male Glossina palpalis gambiensis (palpalis group) were at first exposed to either a caged (mesh size 2.5 x 5 cm) monitor lizard or a tethered dwarf cow in a stable (10.4 x 4.0 x 2.0 m high) where mosquito netting formed the upper half of the four walls.These groups constituted random samples from a population of more than 100,000 flies bred for more than twenty years at CIRDES using random mating.The experiment was first carried out during the dry season (mean daily temperature of 28.5°C and relative humidity of 10-25%) in February 2002 and repeated during the wet season (mean daily temperature of 25.9°C and relative humidity of 63-94%) in July 2003, in the fly-proof stable where these ambient climatic conditions prevailed.Marked one-day post emergence flies (acrylic paint on the pronotum) were released into the stable with either the cow or the monitor lizard (no-choice situation; 120 flies on the cow and 100 on the monitor lizard in 2002, and 155 each on the cow and the monitor lizard in a repeat experiment in 2003).Engorged flies were captured and released two days later (minimal inter-blood meal period in nature) with a similar number of teneral flies (n = 56 in 2002 and 105 in 2003) into the same stable but now holding both hosts (choice situation).Replete flies were caught and dissected to determine the origin of the blood in the crop where it stays for about 30 min.Blood meals were colored (in 10% Giemsa) and twice examined blind by microscope -monitor lizard erythrocytes are oval and nucleated whereas those of bovids are round and anucleated.The monitor lizard (Varanus niloticus) was 68 cm long and the dwarf cow (Bos taurus, Baoule breed) was 3 years old and weighed 165 kg.

Summary
Tsetse flies are very efficient cyclic vectors of African trypanosomosis.Since tsetse are generally infected by the first blood meal, as in the case of sleep-ing sickness for example, any propensity to feed on the same host a second time will improve transmission within this host species, whereas transmission between host species will decrease.To test this hypothesis we presented a monitor lizard and a cow in a stable to marked tsetse flies that had first fed on one of these two hosts.80% of the teneral flies that fed did so on the cow when provided the choice.Among the flies having feeding experience, a dis-proportionately high number of flies that had fed on one host returned to this host for the second meal.We discuss the energetic advantages of such a learn-ing behavior and its importance in sleeping sickness epidemiology.The find-ings are of relevance to the role played by such learning behavior in disease transmission by other insect vectors of zoonoses.
Statistical analysis of fly preferences was made using contingency cross tables obtained from categorical variables characterizing the origin of the blood meals.Pearson's independency chi-square tests, relative risks and confidence intervals were calculated using R statistical software (7).

■ RESULTS
In the no-choice situations, 73 and 89% of the recaptured flies were found engorged during the dry and rainy season, respectively, on the cow.These percentages were 59 and 37%, respectively, for the flies exposed to the monitor lizard.
In the choice situations, over 80% of the teneral flies that fed did so on the cow when provided the choice between the cow and the monitor lizard (Table I).Despite this preference for the cow, the feeding preferences of previously fed flies turned out to be largely dependent on the source of their first blood meal (Table I).Disproportionately high numbers (in brackets) of the flies that had first fed on the monitor lizard returned to feed on this less favored host (48% in 2002 and 65% in 2003; Pearson χ2 test, p < 0.001).The probability for a fly that had first fed on the monitor lizard to choose this host for its second blood meal was 4.1 times greater (1.5-10.7,95% confidence interval) in 2002 and 4.8 times (2.3-10.0) in 2003 in comparison to teneral flies' propensity to feed on this animal.Because of the high preference for the cow, no significant difference was found between flies that had first fed on the cow and teneral flies.However, the probability for a fly that had first fed on the cow to choose this host for its second blood meal was 1.6 time greater (1.2-2.2, 95% confidence interval) in 2002 and 2.8 times (1.7-4.8) in 2003 in comparison to flies that had first fed on the monitor lizard.

■ DISCUSSION
A significantly higher percentage of flies engorged on the cow in the no-choice experiments.This cannot be attributed to a genetic preference for one host or the other on the part of some flies, for if such a preference existed in the fly groups taken here at random from the same population, then the proportions feeding on the monitor lizard and the cow in these no-choice situations should add up to 100%.The data differ significantly from such complementarity.Moreover, G. palpalis gambiensis is considered to be opportunistic but not host specific (2,16).
In the choice situations, teneral G. palpalis gambiensis showed a clear preference for the bovid over the monitor lizard, a fact that could be accounted for by the host size alone in the confines of the stable.Despite this, a disproportionately high number of flies that had fed on the monitor lizard for the first meal returned to this host for the second meal, even with the simultaneous presence of the cow in the stable.This cannot be ascribed to a density dependent factor since the total fly density bearing on any given fly on each host is the same, leaving no environmental factor but previous experience to influence the feeding preference.It has already been observed that tsetse flies (G.palpalis gambiensis) that had fed on goats in captivity persisted to exploit this host upon release into a forest where goats were very rare (2).The present findings are consistent with this observation and show that blood meal sourcing in tsetse flies is influenced by the first meal they take.
Within a given habitat host availability can change between successive generations of a tsetse species as well as between the habitats occupied by the same fly species.Learning permits a widening of the host range: the first choice is very opportunistic but, once a host has been found, tsetse flies become specialists, focusing on this available host species in a given habitat.Tsetse can thus learn to exploit an available, even less preferred, host rather than undertake energetically costly and potentially dangerous long-range flights in search of an unpredictable, though possibly preferred, alternative.The same phenomenon has been identified in hymenoptera (18) and in other insect vectors of disease (9).It is therefore not surprising that a species such as G. palpalis gambiensis can exploit a large range of habitats (from natural to man-made; 3) and show variable host preferences (6,12).
The present findings have a bearing on host-parasite coevolution and pathogenicity of trypanosomes by reducing the probability of inter-host transmission of these parasites (5).In fact, when a trypanosome is transmitted only in cattle, highly pathogenic strains disappear and the disease takes on an endemic character.Where spatial encroachment between cattle and game favors vector confusion this can result in the transmission of highly pathogenic strains of trypanosomes that lead to epidemic situations (14).
Moreover, learning behavior in tsetse flies may play a crucial role in the epidemiology of human sleeping sickness in that animal breeding practices and human habits may inadvertently affect successful exploitation of an alternative host by tsetse simply by its proximity to a targeted species.This may afford a route for trypanosomosis transmission to man when people living in the same place as an infected domesticated or wild host use a common resource such as a water course (17).In current two-host trypanosomosis models, the probability of a tsetse fly to choose one or the other host species is assumed to be constant during its entire life (10,11).This study shows that it is not the case.The transmission rate of parasites from one host species to another may thus be dependent on spatial encroachment between the host species and not only on innate preferences of the vector.Recently, Sané et al. wrote that "Everything leads one to conclude that the endemic and epidemic phases of sleeping sickness are more linked to vector opportunism than to its eclecticism.The number of hosts of the tsetse fly is less important than its tendency to alternate its bloodmeals between animals and man" (12).
It is necessary to integrate such learning behaviors in tsetse into epidemiological models for forecasting risk.The main vectors of sleeping sickness belong to the morsitans and palpalis groups in East and West Africa, respectively.The former group is a less  9) 98% (42)

Table I
Feeding preferences of three groups of tsetse flies (G.palpalis gambiensis) given a choice between a monitor lizard and a cow in the dry and wet seasons opportunistic feeder and shows rather innate preferences (16).As in certain hymenoptera (15), we believe that learning may be less important in the morsitans group, explaining why current tsetse fly feeding and disease transmission models fit well to East African epidemiological patterns.It is worth noting that whereas the two parasites, Trypanosoma brucei rhodesiense in East Africa and T. brucei gambiense in West Africa, are so closely related as to be hardly distinguishable genetically, the role of animal reservoirs is secondary in West Africa while preponderant in East Africa.This epidemiological difference between the regions cannot be explained by current models of disease transmission nor by trypanosome specificities since T. b. gambiense also occurs in a variety of wild and domesticated hosts.We believe that learning in G. palpalis gambiensis, which serves to confine the disease cycle to a given host reservoir, could provide an explanation for this epidemiological enigma.
Finally, the existence of this phenomenon in other vectors of zoonoses such as Rift Valley fever and West Nile virus (transmitted by mosquitoes) may have implications for human outbreaks and should be explored.