Malaria vector control relies primarily on the selective application of residual insecticides through either indoor residual house spraying (IRS) or insecticide-treated nets (ITNs). At high coverage, these approaches have proven highly effective in reducing malaria morbidity and mortality at an affordable cost 1 . However, the ever-increasing development of resistance to insecticides 2 3 is of great concern. Insecticide resistance in malaria vector populations covers all classes of insecticides currently used in public health and is widespread geographically 2 4 5 6 7 . It is, therefore, not surprising that interest in alternative non-chemical strategies has increased over the last decade.

Fungal pathogens commonly infect insects 8 and there has been extensive research on numerous species of Deuteromycete fungi (e.g. Culicinomyces spp., Beauveria spp., Metarhizium spp. and Tolypocladium spp.) for use as biological pest control agents in agriculture 9 10 11 12 13 . Although such fungi appear to have limited impact on mosquito populations under natural conditions 8 14 , there is increasing evidence supporting the potential use of isolates of Beauveria bassiana and Metarhizium anisopliae for control of adult mosquito vectors 15 16 17 18 19 20 21 22 23 24 25 .

Given the emerging problems of insecticide resistance, one of the key requirements for any new (bio) pesticide product for mosquito control is to have limited cross-resistance with existing chemical insecticides 26 27 . Clearly, if resistance to widely used insecticides, such as permethrin and DDT, confers resistance to fungal pathogens then potential novel biopesticide products will have limited utility either as replacements for insecticides, or in integrated strategies for insecticide resistance management 18 . The likelihood of cross-resistance occurring in mosquitoes, however, appears to be remote and in fact it would seem that in certain instances infection with fungi counteracts resistance that is based on metabolic mechanisms, at least in laboratory colonies 25 . One aim of the current study, therefore, was to compare the virulence of a candidate strain of the entomopathogenic fungus B. bassiana against insecticide-resistant and susceptible Anopheles arabiensis laboratory colonies, as well as wild collected adult mosquitoes to determine fungal susceptibility.

Additionally, because fungal pathogens are living organisms, the rate at which they penetrate and grow within an infected host is determined by temperature. Accordingly, the efficacy of certain fungal biopesticide products in agriculture has been shown to be strongly influenced by environmental temperature, in some cases further mediated by thermal behaviour of the target insect 28 29 30 31 . The effects can be such that under certain conditions, speed of kill is rapid and overall control very good, while under other conditions, speed of kill is very slow and control inadequate 30 31 . In a recent study, Blanford et al 22 demonstrated that Anopheles stephensi mosquitoes infected with B. bassiana or M. anisopliae did not exhibit any change in thermal behaviour that might affect speed of kill. Nonetheless, temperature remains an important environmental factor likely to affect fungal germination and growth rate inside mosquito hosts. With respect to malaria control, a critical factor is how the speed of kill (virulence) varies relative to the extrinsic incubation period (EIP) of the malaria parasite; if mortality is faster than the rate of parasite development then impact on transmission will be greater than if mortality is slower than the parasite rate of development 21 . Importantly, both the EIP and pathogen growth vary with temperature 31 32 . The second aim of the current study, therefore, was to explore the effect of temperature on virulence of B. bassiana against insecticide susceptible and resistant An. arabiensis. The daily average temperature measured inside traditional African houses between seasons in western Kenya is 23 ± 1.8°C 33 34 35 . Therefore, the impact of B. bassiana was assessed nder temperature regimes 2°C lower and higher than this average to capture the range of mean temperatures likely experienced in indoor resting sites.

Mosquitoes caught resting indoors north of Karonga on Lake Malawi consisted solely of An. arabiensis. Conditions during the collection period were dry and hot, favouring a preponderance of this member of the An. gambiae complex, which is generally more tolerant of such conditions 36 40 41 42 .

There are clear indications of insecticide resistance to all insecticides and their respective classes tested in one or more of the An. arabiensis samples used. The controlling mechanisms of these resistance phenotypes are likely to involve target site mutations such as kdr as well as metabolic detoxification 2 7 43 44 45 46 47 . Resistance to insecticides in major malaria vector species, coupled to the limited number of insecticides available for use in public health programmes, highlights the need to evaluate the potential efficacy of entomopathogens.

All laboratory-reared and wild-caught An. arabiensis lines were susceptible to B. bassiana. Though quantitative differences were detected between the two exposure temperatures in all colonies tested, Beauveria significantly reduced mosquito longevity at both temperature regimes with no evidence for enhanced resistance to fungal infection due to insecticide resistance. Where mortality rate was apparently slowed due to DDT resistance, this effect was due to enhanced overall survival in the selected lines relative to the baseline colonies, rather than any significant reduction in susceptibility to fungus per se. Why these resistant mosquitoes survived better than controls in the absence of insecticide exposure is unclear. The DDT resistance in these lines is linked to higher levels of expression of glutathione S-transferases (GST) and esterases 47 . Conceivably these generic detoxifying enzymes could enhance survival in the laboratory environment, although trade-offs against other traits and fitness measures might be expected in other environments 48 49 50 .

Beauveria killed mosquitoes significantly quicker at 25°C than at 21°C. This result is consistent with the known temperature-growth profile for this isolate, which indicates a temperature optimum of around 26°C (unpublished data). As suggested previously, to interpret the possible significance of this effect it is important to consider not just absolute speed of kill, but speed of kill relative to the length of the EIP. According to the classic day-degree model of Detinova 51 , the EIP of P. falciparum is 12.3 days at 25°C and 22.2 days at 21°C. Thus, if mosquitoes became infected with malaria and fungus more or less simultaneously (as would happen if mosquitoes contacted fungus on a treated surface following an infectious blood feed), no mosquitoes would have survived long enough to transmit malaria in any of the colonies held at 25°C. At 21°C, fungal infection of the recently derived Malawi colony would have reduced the percentage of mosquitoes potentially able to transmit malaria (i.e. comparing percent alive at day 22 in control and treated populations) from 64% to 4%, representing a 92% reduction. In the two longer-lived DDT resistant colonies, the equivalent figures are 64% to 17% and 55% to 20%, representing reductions in transmission potential of approximately 70%. Thus, although still contributing to substantial reductions in transmission potential, the fungus appears to work less well at 21°C.

Fully extrapolating these results to potential impact in the field is difficult as mortality schedules could potentially differ markedly between lab and field environments for a variety of reasons. Moreover, the current study considers only one dose and it is likely that higher doses could help compensate for the apparent thermal constraint at 21°C. Studies exploring higher fungal doses and different bioassay exposure techniques have shown the potential for much more rapid mosquito mortality than observed in the current study [e.g. 15 52 ], and studies with other insect hosts indicate 'dose x temperature' interactions whereby effects of lower doses are magnified at sub-optimal temperatures 53 . Of course, selection of a different fungal isolate that is less temperature sensitive, or combining isolates with different temperature optima could overcome the constraint completely 18 . Furthermore, the growth rate of the malaria parasite slows exponentially as temperatures decrease further towards 18°C 51 , whereas the decline in fungal growth rate appears more linear over this range (unpubl. data) so it is likely that at slightly cooler temperatures still, the relative efficacy of the current isolate would recover. In addition, sub-lethal effects of infection such as impact on malaria parasite development 16 18 and reduced feeding propensity 54 55 can reduce mosquito vectorial capacity irrespective of speed of kill 21 . Nonetheless, with potential for considerable variation in both mean conditions and diurnal temperature ranges across different transmission environments 56 , understanding the effects of temperature on biopesticide performance is an important area for further research [see also 57 ].

Overall, the results of the current study demonstrate that relative susceptibility of Anopheles arabiensis to a candidate fungal biopesticide strain is not affected by resistance to insecticides (see also 25 ), that wild-caught mosquitoes are equally susceptible to fungal infection and that although there was temperature-dependent variation in fungal virulence, fungal infection led to substantial reductions in malaria transmission potential in conditions typical of local African houses (at least in western Kenya). These empirical data add support to recent modelling studies suggesting that as long as coverage is high (a goal of most conventional vector control operations), slow acting biopesticides can deliver substantial reductions in malaria transmission across a range of conditions 21 23 58 59 . Of particular relevance here is the study of Koella et al 58 , who demonstrated that the level of control (whether biological or conventional) necessary to reduce or even prevent malaria transmission depends on the background transmission intensity. In areas of low to moderate transmission, <50% reduction could provide substantial control, whereas in areas of very high transmission, even 90% reduction might not be sufficient to deliver any benefit due to the strongly saturating relationship between malaria prevalence and transmission 58 60 . Thus the significance of the temperature-dependence in biopesticide performance needs to be considered in relation to the local epidemiology context.

Additionally, the efficacy and impact of a biopesticide will depend on ultimate use strategy. For example, in a recent theoretical study, Hancock 59 demonstrated that under conditions of intense transmission, high single coverage of either ITNs or IRS with a fungal biopesticide might not substantially reduce malaria prevalence in the human population, whereas intermediate coverage of both interventions simultaneously could. This conclusion, together with the empirical data demonstrating that resistance to insecticides does not confer resistance to fungi, highlights the potential for development of novel integrated control strategies combining insecticide and biopesticide interventions. Understanding such interactions, together with the local environmental context, are important areas for future research to define possible limits to biopesticide performance and identify isolates, doses and potential delivery systems to optimise control strategies across time and space.

The authors declare that they have no competing interests.

CKK carried out wild mosquito collections, species identification, insecticide and fungal susceptibility tests, data analysis, interpretation of results, and drafted the first version of the manuscript. BDB supervised all the laboratory experiments and contributed to the subsequent writing of the manuscript. BGJK obtained funding for the project and contributed to the editing of the manuscript. LLK supervised the insecticide susceptibility assays and species identification. MF assisted with the statistical analyses and contributed to the editing of the manuscript. RHH organised the field trip to Malawi and was involved in wild mosquito collections, identification and rearing. MBT provided fungal spores, was involved in methodology of fungal experiments and contributed to editing the manuscript. MC was involved in project design and contributed to the final editing of the manuscript. All authors read and approved the final version of the manuscript prior to submission.