Background
Many plants and animals actively defend themselves from ectoparasites by chemical and/or behavioural responses [
1‐
3]. The role of host defensive mechanisms (e.g. such as the secretion of chemical compounds) in driving the host specificity of agricultural pest insects has been well documented [
1,
4,
5]. In contrast, relatively little is known about the importance of physical behaviours mounted by vertebrate hosts in generating selection for host specificity in insect disease vectors [
6]. Of all the insect vectors of human disease,
Anopheles mosquitoes are responsible for the greatest loss of life and morbidity through their role in malaria transmission [
7,
8].
The frequency with which mosquito vectors feed on humans, and adult mosquito survival, are key determinants of malaria transmission intensity [
7,
9]. Both of these phenomena may be influenced by host physical movements. Specifically hosts that are free to exhibit behavioural responses can prevent mosquitoes from biting [
10,
11], and/or interrupt their blood feeding [
12]. These host responses could limit parasite transmission by reducing host – vector contact rates and increasing the risk of mortality in host-seeking mosquitoes [
13]. Alternatively, host physical movements that do not kill mosquitoes but divert them to other hosts could enhance parasite transmission by increasing the number of hosts that vectors contact during their lifetime [
12,
14]. Consequently, host behavioural responses could have substantial impacts on the fitness of both malaria parasites and their vectors, and correspondingly generate selection for specificity on poorly defensive host types.
Studies of mosquitoes and other haematophagous insects have shown that their feeding success on vertebrates can be significantly reduced by host defensive behaviours [
12,
15,
16]. The effectiveness of host behavioural responses has been shown to vary between host species [
17], individuals [
18], and also in association with additional factors such as whether hosts are infected by parasites [
15], and the overall density of biting insects [
19,
20]. The consequences of such host behaviours may be non-linearly related to ectoparasite fitness. For example the feeding and reproductive success of fleas is significantly reduced by strong host behavioural and immune defenses [
21‐
23], but moderate host defensive behaviours can enhance their blood intake and survival [
23]. Consequently, the net impact of host behaviours on the feeding success of vectors may be context-specific, and ideally should be assessed against the background of host physical movements that they are most likely to encounter when foraging within natural settings.
The strong preference of African malaria vectors such as
Anopheles gambiae s.s for human-feeding has been speculated to be the result of the poor anti-mosquito defensive behaviours of people relative to other animal alternatives. A reason cited for why humans have been assumed to be weakly defensive hosts is that they are typically asleep during the night-time hours where malaria vector biting activity is concentrated. However, other host species which malaria vectors could feed upon (e.g. livestock, dogs) also sleep during these hours so this behaviour does not uniquely distinguish humans. While the associations between host-specific defensive behaviours and insect host preferences have been experimentally investigated in Diptera foraging on birds [
15], rodents [
24] and livestock [
25], testing this hypothesis on human malaria vectors has been restricted by ethical constraints arising from the requirement to monitor human behaviour in response to attack by potentially infected mosquito vectors. To overcome this obstacle, here experiments were conducted in an experimental semi-field-system (SFS) situated at the Ifakara Health Institute (IHI) in Tanzania in which vector – human interactions can be studied under relatively natural conditions using only mosquitoes that are guaranteed to be malaria-free.
Within this setting, the feeding success and subsequent fitness of the two most important African malaria vectors
Anopheles arabiensis and
An. gambiae s.s were compared under conditions when hosts were free to exhibit natural physical behaviours in response to mosquitoes, and when mosquitoes were directly applied to the skin of immobile hosts. The assumption was that most forms of host defensive behaviour towards mosquito biting are eliminated when mosquitoes are applied directly to the skin surface of immobile hosts. It is, however, possible that some subtle host behaviours, such as skin rippling, can still occur even under these conditions. Thus, the experimental design employed here allowed comparison of mosquito feeding and fitness in response to host behavioural limitation, but not necessarily its complete absence. The following hypotheses were tested: (1) mosquito feeding success and subsequent fitness is greater when host behaviour is restricted, and (2) the relative efficiency of host physical behaviours in preventing mosquito biting is correlated with the documented host preferences of these mosquito vectors (e.g. cattle for
An. arabiensis[
26], and humans for
An. gambiae s.s [
27]). In testing the latter hypothesis, the relative efficiency of host behaviour was estimated by the magnitude of difference in mosquito feeding success upon ‘free’ and behaviourally-restricted hosts. A specific prediction was that the host species which are least effective at repelling mosquito biting would be cows for
An. arabiensis, and humans for
An. gambiae s.s. By characterizing the fitness costs imposed by host physical movement, this study can shed light on the potential for malaria vectors to adapt to new host species in response to the mass coverage of public-health interventions that specifically protect humans.
Discussion
This study demonstrates the potential impact of host physical movements on the feeding success and subsequent fitness of the African malaria vectors An. arabiensis and An. gambiae s.s. Consistent with initial prediction, there was some evidence that allowing hosts to make physical movements in the presence of mosquito biting did limit Anopheles feeding success. With the exception of cow hosts, blood feeding success rates in An. arabiensis applied directly to the skin of immobile hosts were generally higher than when they attempted to feed on the same individuals under more natural semi-field conditions. However this was not true for An. gambiae s.s. where host feeding rates were generally similar under semi-field conditions where hosts were free to move as when they were behaviourally-restricted (for four out of six host types). This suggests that host physical defensive behaviours may have a differential impact on these two vectors. While limiting host movement did consistently increase the feeding success of at least one vector species (An. arabiensis), it did not improve the quality of blood meals obtained by those that succeeded in feeding in terms of blood meal size, reproduction and survival. In fact, the oviposition and survival of mosquitoes that foraged on hosts under more natural semi-field conditions was generally greater than in those that had fed on behaviourally restricted hosts. These results indicate that while host physical movements may have moderate, host-specific impacts on malaria vector feeding probability, they do not diminish the quality of blood meals of the mosquitoes who are able to feed from them.
A key prediction that this study aimed to test was whether the exhibition of physical movements in the host species that are preferred by these mosquito vectors are less effective at deterring biting than those made by other types that they rarely select. Confirmation of this prediction would support the hypothesis that host defensive behaviours may have influenced the evolution of host specificity in this system. However, evidence of this phenomenon was mixed.
Anopheles arabiensis had a higher feeding success on its naturally preferred cattle hosts under both feeding conditions. Furthermore this host type was the only one whose movements were not associated with a reduction in
An. arabiensis feeding success. This suggests that either cattle have poorer behavioural responses than the other host species assayed here [
19], or that
An. arabiensis has evolved strategies to more effectively evade the behavioural responses of this host type [
36]. In contrast, the feeding success of
An. gambiae s.s was unaffected by whether hosts were free to move or behaviourally restricted for both their naturally preferred human hosts, and some rarely exploited host types (e.g. dogs, goats and calves). In both vector species, the restriction of behaviour in chickens generated the greatest proportionate increase in mosquito feeding success. This finding matches results from other mosquito systems which indicate avian hosts have considerably more effective anti-mosquito behaviours than mammals, and may explain why this host type is rarely exploited by these vectors in nature [
37]. Consequently, variation in host physical behaviours may be generating selection on these malaria vectors to avoid certain host species (e.g. chickens), but cannot consistently account for their previously-documented host species preferences on the basis of experiments conducted here.
Contrary to prediction, mosquitoes feeding on hosts that were free to move during exposure either obtained larger (in the case of
An. arabiensis and some host types for
An. gambiae s.s.) or similarly sized blood meals than when allowed to feed from immobilized hosts. These results contrast with previous studies of the mosquito
Aedes aegypti and other ectoparasites that have shown host physical movements reduce their blood meal size [
12,
21‐
23]. Discrepancies between these and the current study may reflect genuine biological differences in the impact of host defensive behaviour on different haematophagous insects. Another possibility is that this discrepancy is due to differences in the timing and duration of the mosquito exposure period under conditions where they were free or behaviourally limited used here. Pilot investigations in this system indicated that after landing on a host, mosquitoes require less than six minutes to initiate feeding and complete a blood meal (Lyimo
et al. personal communication). Consequently, the 15 minute exposure period used in trials where mosquitoes were directly applied to host skin was deemed sufficient to allow mosquitoes to commence biting and feed to repletion. However, in semi-field experiments mosquitoes were exposed to freely moving hosts for a period of 12 hours (7 pm-7 am) in accord with the duration of their natural host seeking period. The enhanced feeding success of
An. arabiensis on unrestrained hosts could be a by-product of an increased, innate predisposition for feeding during the night time hours when semi-field experiments were conducted (host restraint experiments conducted during the day). Several of the host types assayed here including humans are also more likely to be sleeping during night time hours which may have minimized the impact of any anti-mosquito defensive behaviours they are capable of making. This phenomenon could explain a lack of difference between mosquito feeding success on free and behaviourally-restrained hosts, but not the increased blood meal size of
An. arabiensis reported here. An additional explanation for these results could be that
An. arabiensis exposed to freely moving hosts in semi-field conditions could feed repeatedly from hosts throughout the night to top up their blood intake beyond what could be acquired from the one contact permitted in trials with immobilized hosts.
It is also possible that when mosquitoes are given the opportunity to choose where on the host body they bite (as under the semi-field, but not host behavioural restriction conditions used here), they preferentially select sites from which blood can be more efficiently imbibed. This could include areas of skin that are relatively thinner and/or blood vessels more easily accessible [
38] than the locations where mosquitoes were applied under host behavioural limitation conditions. For example in experiments where host behaviour was limited, mosquitoes were applied to human forearms whereas under natural conditions they preferentially bite feet [
39]. Similarly there is some evidence that
An. arabiensis preferentially land and feed on cow legs [
36], whereas here they were exposed to a variety of sites on the cow body (e.g. flanks, and thigh muscles). However, previous studies have shown that the relative ‘attractiveness’ of particular biting sites on the bodies of hosts is highly dependent on their position (e.g.
An. gambiae s.s. preferentially bite human feet when people are sitting down, but this preference is not evident when people are lying down or have their feet in the air [
39]). This suggests that there may be no intrinsically ‘optimal’ biting sites on the body of hosts, and that the variation in mosquito feeding success observed here may not be explained by treatment-specific differences in where on the host body mosquitoes were allowed to feed from. Further experiments are required to test whether the enhanced blood meal sizes associated with foraging on freely moving rather than behaviourally restricted hosts as reported here could be explained by any of these additional factors. Finally, it is noted that although the semi-field conditions used here did permit relatively realistic interactions between mosquitoes and the host types they typically encounter, they may not be fully representative of natural field conditions in which these interactions take place against a more complex background of variation in environmental conditions, host and mosquito abundance and diversity. Direct field evaluations of these hypotheses are currently problematic given their requirement to allow (potentially-infectious) mosquitoes to feed on human subjects. However if risk-free methodologies for human exposure develop, further investigation of this phenomenon under field conditions are warranted to validate results presented here.
Bearing in mind the caveats to interpretation as discussed above, the results of this study suggest that physical defensive behaviours exhibited by common host species including humans do not impose substantial fitness costs on African malaria vectors. If this is the case, alternative explanations for the evolution of host specificity in this system are needed. One possibility could be the existence of haematological variation between host species in the key traits likely to influence blood resource quality to mosquitoes (e.g. haemoglobin levels and red cell density, [
40,
41]. Several haematological properties such as haemoglobin concentration, red blood cell density and amino acid composition are known to be vary within and between vertebrate species [
42,
43], and could account for some of the variation in the fitness of haematophagous insects [
9]. Laboratory experiments in which
An. gambiae s.s. were fed blood from different host species under standardized membrane feeding conditions suggest that haematological factors alone, in the absence of additional host behavioural, ecological or physiological factors, can generate some variation in mosquito fitness [
44]. However
in the host-specific variation in mosquito fitness observed under these laboratory conditions was only partially consistent with results obtained from the more natural semi-field conditions here, and similarly did not indicate that mosquito fitness was consistently highest on the blood of preferred species. Further investigations are ongoing to evaluate the role of naturally-occurring haematological variation on the fitness of malaria vectors under more natural semi-field condition and will help resolve this issue. Other than host defensiveness and/or haematological properties, other potential explanations for the evolution of host specialization in these malaria vectors include larger-scale properties of host ecology. For example, it could be specific properties of the preferred habitats of different host types (e.g. climatic suitability of human relative to animal dwellings) that drive selection towards anthrophily, rather than innate host biological properties. Other factors such as the relative abundance and aggregation of hosts across the landscape may also be responsible for generating selection towards the types that mosquitoes most frequently encounter. Further investigation into these and other potential hypotheses are encouraged to help resolve the nature of selection acting upon malaria vector-host interactions.
As mosquito blood meal size is strongly and positively correlated with their reproductive success [
45‐
47] and long-term survival [
48], host behaviours that limit blood intake are expected to reduce these fitness traits [
13]. However, the relatively larger blood meals that mosquitoes acquired from feeding on freely-moving hosts under semi-field conditions did not translate into correspondingly greater reproductive success. This may be because regardless of host species or behavioural manifestation, mosquitoes that succeeded in feeding always obtained the minimum volume of blood required to initiate oviposition [
49] and maximize egg production. Anopheline fecundity is known to be linearly related to blood meal volume only above a minimum threshold below which no eggs are produced, and below a maximum threshold above which no further eggs are produced [
47]. Although the larger mosquito blood meals obtained by mosquitoes under semi-field conditions were not associated with greater reproductive success, they were correlated with enhanced mosquito longevity. Blood resources are thought to be used primarily for mosquito reproduction [
50‐
52], however some studies indicate the longevity of mosquitoes and other ectoparasites increase with ingested blood meal size [
21,
23,
43]. Observation of a similar phenomenon here suggests these mosquito vectors also use blood proteins to synthesize energy reserves for survival [
50]. Finally, it is noted that the longevity effects that were measured here only represent post-feeding survival, and not direct mortality associated with host-seeking [
13]. Variation in the recapture rate of mosquitoes in semi-field trials could provide an indirect estimate of host-seeking associated mortality. However, as these recapture rates were generally similar across trials with different host species, it suggests there may be no significant variation in host-seeking mortality between host species. Previous work under the semi-field conditions suggest that mortality associated with host-seeking by these vectors is negligible [
29], but further work is required to confirm this under natural field conditions.
The current up-scaling of long-lasting insecticide treated nets (LLIN) [
53] and indoor residual spraying (IRS) [
54] throughout sub-Saharan Africa, improvements in housing [
55], and use of other protecting measures such as repellents [
56] means that the relative ‘defensiveness’ of humans to malaria vectors relative to other available host types is substantially increasing. This increased protection of humans is clearly providing immediate epidemiological benefits by reducing malaria transmission [
57], but may also provide longer-term ‘evolutionary’ advantages by generating selection for mosquito vectors to switch their host choice to less well defended host species [
44]. Such changes are most likely to occur when mosquitoes would receive a clear fitness advantage from shifting away from humans; a process that could be undermined by the existence of strong defensive behaviours in other potential host species such as domestic animals and livestock. With the exception of chickens, no evidence was found here to suggest that the physical movements of the other animal species most likely to be kept in and around households are more costly to malaria vectors than those of humans. In fact it appears that
An. arabiensis may encounter substantially less effective defensive behaviours when foraging on cattle than humans, and thus may do better to switch to the former host type especially if humans are universally covered with bed nets [
29]. It is thus hypothesized that variation in host physical defensive behaviours are unlikely to prevent malaria vectors from exploiting alternative host species (e.g. cows) when humans are unavailable.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
INL and HMF designed the research. INL, AAD, KFM and EMM performed the research. INL and HMF analyzed and interpreted the data. DTH and RR assisted with analysis. INL drafted the manuscript. DTH, RR and HMF commented on the manuscript. All authors read and approved the final manuscript.