This study was the first to examine the effect of entomopathogenic fungi on wild mosquito blood feeding in the field. In particular, the current study investigated more or less instantaneous impacts on feeding within a single feeding night (i.e. within a few hours of fungal exposure). The results show that B. bassiana treatments significantly reduced blood feeding, with B. bassiana alone able to inhibit 37% of blood feeding relative to the control. Permethrin was able to inhibit 43% of blood feeding, a higher percentage than observed in a previous study in the same study village where another pyrethroid, alphacypermethrin, reduced Cx. quinquefasciatus blood feeding by 27% . Given the results, it is unknown why no additive or synergistic effects were seen in the blood feeding inhibitions when the B. bassiana (BC) and permethrin (CP) treatments were combined into the B. bassiana and permethrin (BP) treatment, especially in light of recent laboratory findings , although the synergistic effects seen on mortality  may differ from any behavioural effects.
Both of the fungal species used in our study have previously shown a propensity to reduce mosquito blood feeding under laboratory conditions [18, 19]. This response may be linked to the down-regulation of genes controlling digestion in mosquitoes inoculated with B. bassiana  indicating that digestion and nutrient acquisition is not a priority for mosquitoes after fungal infection. Although these earlier studies looked at feeding over several days following infection and at different mosquito genera, they found a similar level of blood feeding reduction as found in our study [18, 19]. Relatively rapid changes in feeding behaviour after infection with M. anisopliae or B. bassiana have also been reported in many other insect types .
The mechanism behind the very rapid blood feeding inhibition observed in the current study is unknown but may be due to physiological and/or behavioural reasons. A mosquito may enter the huts several hours before blood feeding which would allow the fungus time to start germination and cuticle penetration. As far as we are aware no data have been published on the germination and penetration times on mosquito cuticles. However, in infected termites B. bassiana germination occurred between 6 and 12 hours post infection, with penetration between 12 and 24 hours . In this study, mosquitoes had to pass directly through the fungus-treated netting so some conidia could have got into the mosquito spiracles or at the base of the setae. This may decrease the fungal penetration time because the cuticle is thinner in these places . Even during pre-penetration growth of the conidia the wax layer of the insect cuticle is degraded  and insects use both cellular and humoral immune responses against fungal infections starting as early as cuticle degradation . Therefore, if the germination and pre-penetration times on mosquito cuticles is similar to that on termites, then it is feasible that the immune system could have been activated during the short time the mosquitoes and sleepers were in the huts. Alternatively, the mosquito antennae and maxillary palpi may have become covered in conidia, interfering with their ability to detect the human host. In addition, termites have been shown to groom after fungal infection which successfully removes conidia . This may also have taken place with our wild mosquitoes and could have interfered with their host seeking.
The results indicate that neither M. anisopliae nor B. bassiana repels foraging mosquitoes, as corroborated by a recent laboratory study . In addition, B. bassiana conidia can reduce blood feeding in Cx. quinquefasciatus. However, no significant mortality was found in wild-caught Cx. quinquefasciatus mosquitoes collected from the huts. Although previous findings have found that Cx. quinquefasciatus is susceptible to M. anisopliae [11, 22] it is important to note that adult Cx. quinquefasciatus mortality has not previously been measured following B. bassiana infection either in the laboratory or field. Even after discounting the M. anisopliae data due to the extremely low viability of the batch used in our study, the B. bassiana viability was within the range that could be used in the field in the future, but did not significantly impact wild mosquito mortality. There are three main reasons why this may be the case.
Firstly, the experimental method may have been ineffective at providing a sufficiently lethal dose to the wild mosquitoes, even though it was able to elicit a significant behaviour modification. Possible reasons include certain conditions affecting the conidia on the netting, and the short contact time of the mosquitoes. After one week under field conditions dry conidia were seen to be released from the window netting in the huts. This quick evaporation of ShellSol T and release of conidia has also been found in Tanzania (Matt Kirby, Pers. Comm.) and may lead to a lack of conidial protection from the field conditions, and a decrease in the effective concentration. Using other oil formulations  or encapsulation techniques may lead to higher conidial protection. Laboratory studies have shown that conidial viability is directly affected by the polyester netting [15, 41]. In addition, temperature and humidity can adversely affect fungal conidia [42–44], however, the climatic conditions were similar for the experimental hut study and cone bioassay experiments carried out at the same time and under the same conditions (Howard et al. Manuscript Submitted), so similar adverse effects would be expected. Nevertheless, fungal spores used in the cone bioassays were able to infect mosquitoes causing significant mortality (Howard et al. Manuscript Submitted), but those applied in the experimental huts could not.
Scholte et al.  found much higher levels of mortality in their field study where the An. gambiae s.l. mosquitoes were found resting on fungus-impregnated cotton cloths . The short contact time with the hut fungal netting, although not an issue for An. gambiae s.s. mosquitoes in the behaviour experiments in the laboratory, could have caused problems because a recent study has shown that longer exposure times can cause significantly quicker mortality rates . There appears to be a threshold number of conidia per unit surface area required for successful mosquito infection . This may be related to the up-regulated mosquito immune system being able to clear low-level fungal infections [34, 46]. If the proportion of viable conidia was decreased by the polyester netting/field conditions then the wild Culex may not have been receiving enough viable conidia to initiate a successfully lethal fungal infection. Other proposed application methods in the field include cotton resting targets , clay pots , and odour baited stations , all of which will ensure longer contact times but would target resting mosquitoes post-feeding, and so may not affect blood feeding in the same way as the method used in this study.
The second reason for the lack of fungus-induced mortality could be that even if a successful fungal infection was received by the hut-entering mosquitoes, then it is possible that the mosquitoes died of natural causes before any significant toxic effects of fungal infections could be seen because control mortality after 7 days was 54% and was not significantly different from the fungus and/or permethrin-exposed mosquitoes. This may have masked any effects of the fungus. The natural mortality could be quite high because the mosquitoes entering the huts were of an unknown age range, and insecticide resistance in Culex mosquitoes is known to be associated with fitness costs  that can lead to reduced survival rates .
Finally, the third possible reason for the lack of fungal-induced mortality is that the wild multi-insecticide-resistant Cx. quinquefasciatus mosquitoes in Benin may just not have been susceptible to fungal infection. As mentioned, Cx. quinquefasciatus adults have not been previously shown to be susceptible to B. bassiana either in the laboratory or field. A previous laboratory study comparing An. gambiae s.s. and Cx. quinquefasciatus found very few differences in susceptibility to M. anisopliae infection, with both male and female Cx. quinquefasciatus having significantly reduced life spans after continuous exposure to both dry and oil-formulated conidia . However, Scholte  speculates that wild Tanzanian insecticide-susceptible Cx. quinquefasciatus in the field had higher immunocompetence towards M. anisopliae infection than wild An. gambiae s.l. because the infection rates were 10% and 33% respectively. Wild Culex may be less susceptible to fungal infection due to interactions of their micro-flora , or because their insecticide resistance mechanisms protected them . Micro-flora interactions can protect insects from infection; Pseudomonas bacteria found in insecticide-resistant diamond-back moths showed antagonistic activity against M. anisopliae and B. bassiana . In addition, B. bassiana has been shown to be a poor competitor in the presence of M. anisopliae , and is therefore likely to compete poorly with the wild Culex gut flora. Insecticide-resistant Cx. quinquefasciatus in Sri Lanka were shown to adversely affect the development of the filarial worm Wuchereria bancrofti, thought to be due to elevated esterase activity . Serebrov et al.  found that infection of greater wax moth caterpillars with M. anisopliae caused elevated levels of esterases and GST, presumably as part of the immune response. If elevated esterase and GST levels are also an important immune response to fungal infection in Cx. quinquefasciatus, this would explain the low susceptibility of the wild mosquitoes in this study; in effect their immune system is already activated because they naturally have higher levels of these enzymes . Further work should be carried out to identify whether wild insecticide-resistant Culex mosquitoes can be killed using entomopathogenic fungi, and to identify for which reason this did not occur in the present study.
Although there were not enough malaria vectors to analyse the effect the fungi may have on these mosquitoes, it is also important to test new control tools on Culex mosquitoes. This is because in many areas Culex mosquitoes are often more numerous than Anopheles and as such personal protection methods such as ITNs are often bought to prevent the nuisance biting as much as for any other reason. Failure to control these nuisance mosquitoes can reduce the uptake of ITNs for malaria control [53, 54]. Therefore tools that can reduce the biting of insecticide-resistant Culex mosquitoes are also required. This is especially true in East Africa, India and South East Asia where Culex mosquitoes are the vectors of filariasis.
Reducing blood feeding is important in terms of disease control and the finding that B. bassiana can reduce blood feeding in wild mosquitoes so soon after they acquired a fungal infection is both unexpected and important, but further research in areas with filariasis transmission is required to monitor whether this behaviour modification could be used to help prevent filariasis transmission. In addition, work should also be carried out specifically targeting malaria vectors to substantiate whether this behaviour is also present in Anopheles mosquitoes, and if this could have any effect on malaria transmission. Because blood feeding was significantly affected so soon after acquiring a fungal infection it is suggested that future application techniques for fungi in the field should target host-seeking mosquitoes. If the fungi are deployed as post-feeding resting targets [17, 22, 23], then one of the main ways in which entomopathogenic fungi could help reduce disease transmission would be missed.