Infection of malaria (Anopheles gambiae s.s.) and filariasis (Culex quinquefasciatus) vectors with the entomopathogenic fungus Metarhizium anisopliae

Background Current intra-domiciliary vector control depends on the application of residual insecticides and/or repellents. Although biological control agents have been developed against aquatic mosquito stages, none are available for adults. Following successful use of an entomopathogenic fungus against tsetse flies (Diptera: Glossinidae) we investigated the potency of this fungus as a biological control agent for adult malaria and filariasis vector mosquitoes. Methods In the laboratory, both sexes of Anopheles gambiae sensu stricto and Culex quinquefasciatus were passively contaminated with dry conidia of Metarhizium anisopliae. Pathogenicity of this fungus for An. gambiae was further tested for varying exposure times and different doses of oil-formulated conidia. Results Comparison of Gompertz survival curves and LT50 values for treated and untreated specimens showed that, for both species, infected mosquitoes died significantly earlier (p < 0.0001) than uninfected control groups. No differences in LT50 values were found for different exposure times (24, 48 hrs or continuous exposure) of An. gambiae to dry conidia. Exposure to oil-formulated conidia (doses ranging from 1.6 × 107 to 1.6 × 1010 conidia/m2) gave LT50 values of 9.69 ± 1.24 (lowest dose) to 5.89 ± 0.35 days (highest dose), with infection percentages ranging from 4.4–83.7%. Conclusion Our study marks the first to use an entomopathogenic fungus against adult Afrotropical disease vectors. Given its high pathogenicity for both adult Anopheles and Culex mosquitoes we recommend development of novel targeted indoor application methods for the control of endophagic host-seeking females.


Introduction
Malaria and bancroftian filariasis rank amongst the world's most prevalent tropical infectious diseases. An estimated 300-500 million people are infected with malaria annually, resulting in 1.5-3 million deaths [1]. Lymphatic filariasis is probably the fastest spreading insect-borne disease of man in the tropics, affecting about 146 million people [2]. The use of residual insecticides for anopheline vector control, either through indoor house spraying [3] or for bednet impregnation [4], has proven highly effective in various parts of Africa, but is not without obstacles. Emergence and spread of insecticide resistance in anophelines [1, [5][6][7], environmental pollution [8], and unresolved issues pertaining to their toxicity to humans and non-target organisms [3,9,10] hamper progressive use and broad acceptance of these tools. Limited susceptibility and rapid build-up of resistance to synthetic pyrethroids by culicine filariasis vectors [11], combined with the availability of effective anti-filarial drugs [12], is causing a gradual shift from vector control to mass-chemotherapy, though resurgence of transmission in the absence of vector control remains problematic [13,14]. A continued search for appropriate vector control strategies to augment this limited arsenal of tools is called for [15], and includes biological control methods [16,17].
Only a handful of studies have evaluated biological control agents/methodologies to control adult stages of tropical disease vectors. Soarés [31] infected adult Ochlerotatus sierrensis with the deuteromycete Tolypocladium cylindrosporum, resulting in 100% mortality after 10 days, whereas Clark et al. [26] showed in a laboratory study that adult mosquitoes of Culex tarsalis, Cx. pipiens, Aedes aegypti, Ochlerotatus sierrensis, Ochlerotatus nigromaculis, and Anopheles albimanus were susceptible to Beauveria bassiana. Recently, Scholte et al. [32] reported that adult An. gambiae is susceptible to B. bassiana, a Fusarium spp., and Metarhizium anisopliae.
Spores (conidia) of M. anisopliae have been known for some time to be infectious to adults and emerging pupae of tsetse flies (Diptera: Glossinidae) [41,42] and novel devices for exposing wild-caught specimens have been developed [43]. Following our initial observation that the same fungus is pathogenic to An. gambiae [32], we report here on the effects of different exposure times to conidiatreated substrates (24, 48 hrs or continuous exposure), and dose-response experiments using oil-formulations of conidia. In addition, we report the effects of continuous exposure to conidia of adult Cx. quinquefasciatus.

Bioassays
Three different bioassays were conducted to study the effect of: 1) dry conidia of M. anisopliae on infection and survival of adult male and female An. gambiae s.s. and Cx. quinquefasciatus. In this bioassay exposure of mosquitoes to conidia was continuous throughout the experimental period; 2) limited exposure (24 or 48 hrs only) to dry conidia on infection and survival of male and female An. gambiae s.s.. Following these periods the source of conidia was removed to avoid further exposure; 3) different doses of conidia, formulated in sunflower oil on infection and survival of female An. gambiae s.s..
Bloodfed Cx. quinquefasciatus were collected daily from local houses in Mbita Point, western Kenya. For bioassays, only the F1 offspring from these wild mosquitoes was used. Larvae and adult mosquitoes were kept under similar conditions as the An. gambiae s.s. used in bioassay 1.

Fungus
Metarhizium anisopliae var. anisopliae (Metsch.) Sorokin, isolate ICIPE-30 (courtesy Dr. N. Maniania) was used for all three bioassays. The fungus was originally isolated in 1989 from a stemborer, Busseola fusca Fuller, near Kendu Bay, western Kenya. Prior to use, fresh conidia were stored in the dark at 4°C.

Experimental procedures
Bioassay 1 For each of three replicates, 30-50 male and female An. gambiae s.s. or Cx. quinquefasciatus mosquitoes, aged 1-2 days at the start of the experiment, were placed in a 30 cm cubic iron-framed cage covered with white mosquito netting. These mosquitoes were offered a 6% glucose solu-tion absorbed onto white laboratory filter-paper placed in a glass vial ( Figure 1). The paper was enveloped by a locally available hair-roller ('suspensor'), basically a polyethylene perforated tube (height 6.2 cm, diameter 2.5 cm) with protruding, short plastic hairs that enlarged its surface area. This suspensor was dusted with 100 mg of dry conidia using a small paintbrush and carefully placed over the filter-paper. Mosquitoes landing on the suspensor to consume glucose would thus be exposed to conidia through tarsal contact or on the head and thorax region when feeding through the holes in the suspensor. The suspensor remained in the cage until the end of the bioassay (a minimum of 8 days), resulting in continuous exposure of mosquitoes to the conidia whenever they consumed glucose. The control group was exposed to a similar suspensor, but without conidia. An estimation of conidia density was made using a haemocyte counter (Fuchsrosenthal ® , 0.2 mm depth), showing that 100 mg of conidia equalled approximately 6 × 10 8 conidia. Dead mosquitoes were removed from the cage daily, placed on moist filter-paper (distilled water) in a parafilm-sealed petri dish, and examined for fungal growth after a few days.

Bioassay 2
Experimental procedures were identical to those described above, except that the source of conidia was removed from the cages after either 24 or 48 hrs and replaced with a non-contaminated source of glucose.

Bioassay 3
Conidia were inoculated on oatmeal agar and placed in an incubator to grow for 2 weeks, after which fresh conidia were harvested using a 0.05% Triton-x solution and a glass rod. The solvent containing conidia was concentrated by removing the supernatant after centrifuging for 3 min at 5000 rpm. Dilutions were made using 0.05% Tween 20 to obtain conidia concentrations of 10 5 , 10 6 , 10 7 , and 10 8 conidia/ml. Sunflower oil was added to these solvents to obtain 10% oil-formulations for all solvents. For the bioassay, 1 ml of the various oil-formulations was pipetted evenly over a 8 × 6 cm piece of filter-paper and left to dry at 70% RH for 48 hrs, resulting in spore densities of 1.6 × 10 7 -1.6 × 10 10 conidia/m 2 . These papers were gently placed in cylindrical glass vials so that the paper coated the inside of the vial. Mosquitoes were tested individually by placing a 1 to 3 day-old female An. gambiae s.s. in a vial, which was then sealed with cotton netting material. For each dose, 10-15 mosquitoes were tested each time, and all of the above procedures replicated thrice. Mosquitoes had access to a 6% glucose solution by placing freshly soaked cotton wool pads on the netting daily. The filterpaper was removed from the vial after three days, essentially delivering tarsal exposure of mosquitoes to conidia for 72 hrs. Dead mosquitoes were removed from the vials The set-up used to contaminate mosquitoes with dry conidia of Metarhizium anisopliae Figure 1 The set-up used to contaminate mosquitoes with dry conidia of Metarhizium anisopliae. Dimensions are in cm.

2.5
Glass vial, containing 6% glucose solution Filter paper daily and placed in petri dishes containing moist filterpaper. These were then sealed off and placed in an incubator at 27°C for three days, after which the cadavers were checked for sporulating M. anisopliae using a dissection microscope.

Data analysis
Survival curves were analysed by Kaplan-Meier pair-wise comparison [47] and Cox regression analysis (using SPSS 11.0 and Genstat 5 software). For each trial, mosquito survival data were fitted to the Gompertz distribution model [48] (using Genstat 5) as described by Clements and Paterson [49], from which LT 50 values were calculated. LT 50 values of treated versus control groups were compared using paired sample t-tests.

Results
M. anisopliae showed to be pathogenic to adult male and female An. gambiae s.s., both when mosquitoes were exposed to oil-formulated or dry conidia. Cox regression analysis showed that this pathogenicity was not dependant on exposure time. Mosquitoes exposed to conidia for 24 or 48 hrs or continuously, died significantly faster (p < 0.001) than the untreated groups (Table 1). No significant differences in survival were found between groups exposed for either 24 or 48 hrs (p = 0.861), but there appeared to be a trend that survival was lower under continuous exposure, compared to 24 or 48 hrs, although this difference was also not significant (p = 0.092). Overall, males died faster than females (p = 0.020). Following continuous exposure, 100% mortality of both male and female An. gambiae was observed by day 7, at which time an average of 41.7 and 82.3% of the respective sexes in the control treatment were still alive.
In all bioassays where mosquitoes were exposed to dry conidia, fungal sporulation was observed in > 95% of the cadavers. For conidia in oil-formulations, percentage sporulation was positively correlated with conidial dose. Percentages ranged from 4.43 ± 4.4 % for mosquitoes that had been exposed to the lowest, to 83.70 ± 8.3 % for those that had been exposed to the highest dose (Table 2).
Mosquito survival data from all three bioassays closely fitted the Gompertz distribution model (variance accounted for ranged from 96.2-99.6%; Figure 2 and 3). Estimates of  daily survival rates derived from the Gompertz model [49] showed a dramatic reduction following exposure to conidia. In the dose-response experiment, daily survival rates were inversely related to the exposure dose. Table 3 shows estimated daily survival rates for An. gambiae at different ages for the doses tested.
Pair-wise comparisons of Kaplan-Meier data and survivalcurves for the different doses of exposure (Table 2, Figure  3) showed that mosquito survival in the untreated group was not significantly different from that observed for mosquitoes exposed to the lowest dose (1.6 × 10 7 conidia/m 2 ) (p= 0.448). Survival of mosquitoes exposed to the two intermediate doses (1.6 × 10 8 and 1.6 × 10 9 conidia/m 2 ) was significantly different from the lowest (1.6 × 10 7 conidia/m 2 ) (p = 0.014 and p < 0.001), as well as the highest (1.6 × 10 10 conidia/m 2 ) dose, (p = 0.001 and p = 0.014). LSD multiple comparison analysis of the LT 50 val-ues showed similar, but not equal cluster formation: the five groups appeared to be divided into two clusters (

Discussion
Epidemiological models on malaria and filariasis show that adult mosquito survival is the most sensitive component of vectorial capacity [50][51][52]. Notable reductions in the life-span of female An. gambiae and Cx. quinquefasciatus will therefore have a considerable impact on reducing the transmission risk of malaria and lymphatic filariasis. For An. gambiae, daily survival rates (p) have been calculated for field populations, either from mark-releaserecapture studies or physiological state ratios of wildcaught females. Takken et al. [53], using mark-releaserecapture methodology near Ifakara, southeast Tanzania, estimated p to be 0.78. Gillies [54], using physiological state methodology in the coastal lowlands of Tanzania, found that female mortality rate was almost constant during the 1-parous to 6-parous age period, with p being 0.85, but that it increased sharply thereafter. Killeen et al. [55] arrived at a mean p value of 0.90, reflecting an approximate median of estimates for four holoendemic sites in Africa, which equals our estimates from the Gompertz model.
The results from the current study show that the daily survival rates of M. anisopliae-infected adult mosquitoes at any given moment in the mosquito's life span, is lower than non-infected mosquitoes, and that their life span is reduced, provided that the conidial dose is high enough.
If this is translated from a laboratory situation to the field it could result (under favourable circumstances) in reduced vectorial capacities of these mosquitoes. For Cx. quinquefasciatus, life-expectancies of 30.82 ± 3.41 and 44.12 ± 4.19 days were found for caged male and females in Tanzania [56]. Oda et al. [57] found mean longevities of 39.8 (males) and 64.4 days (females), when kept in the laboratory at 25°C. In our study we recorded lower survival of uninfected specimens, but when 100% of the treated female Cx. quinquefasciatus had died, 90 ± 6 % of the untreated specimens were still alive, demonstrating that this species is highly susceptible to infection with M. anisopliae.
The correlation of increasing proportion of sporulation with exposure of An. gambiae to increasing conidial doses suggests that the mosquitoes' immune system is only able to defend against the fungal infection at low doses, diminishing effectiveness when exposed to increasing conidial doses. Mosquitoes may overcome a low infection level by melanization and/or encapsulation of blastospores [58]. These defences would however have a cost to the mosquito's fitness [59,60], which is a possible explanation for the positive correlation between conidial dose, mortality rate and proportion of sporulation found in bioassay 3.
Unfortunately, in the present study we did not examine mosquitoes for melanization.
For successful conidial attachment, fungal penetration and, in the end, killing of a mosquito, a threshold number of condia per unit surface area is required. In our doseresponse experiments the lowest dose resulting in a significant effect on mosquito survival was 1.6 × 10 8 conidia/ m 2 , equaling 160 conidia/mm 2 . Apart from the two claws at the end of the last tarsus, the legs/tarsae of mosquitoes are densely covered with hair-like structures ('feathers'), which make it difficult for conidia to attach to the cuticula of the tarsus. This was confirmed by observations under a light microscope at magnification 40×. Several mosquitoes were gently removed from the contamination cage, placed into a glass vial (they had been exposed for 24 hrs to 1 × 10 6 conidia/ml of oil-formulated conidia on impregnated filter-paper) and killed by adding a small droplet of chloroform in the vial. Many conidia were found attached to the 'feathers' of the last few tarsae, and several at the 'feathers'of the tibia, but only few were actually attached to the cuticle. The ones that appeared to be attached to the cuticle, were located near the end of the last tarsus, around the claws, and in the intersegmental areas of the tarsae, where very few 'feathers' are present. Apparently the effective conidial dose (i.e. conidia that actually attach to the mosquito's cuticle and subsequently invade the integument and haemocoel) is an unknown, but presumably it is a rather low fraction of the conidial dosage that attaches to the mosquito. There appeared to be considerable differences in LT 50 values between An. gambiae exposed to dry or to oil-formulated conidia. The dose of dry conidia that was used was estimated to be 6 × 10 8 conidia per suspensor. The suspensor has a surface area 48.7 cm 2 , but has a 'hairy' surface area, containing approximately 40 plastic 'hairs' (height 3.5 mm, diameter 0.5 mm) per cm 2 suspensor surface, which increases its surface area with an estimated 107 cm 2 to 155.7 cm 2 , resulting in a conidial dose per surface area of approximately 5.6 × 10 10 conidia/m 2 . This is 3.5 times higher than the highest oil-formulated dose used in bioassay 3 (1.6 × 10 10 conidia/m 2 ).
When comparing the effects of indoor application of residual insecticides, or bednets treated with synthetic pyrethroids, on mosquito mortality, M. anisopliae delivers lower overall mortality rates and speed of killing target insects. However, apart from its demonstrated pathogenicity to An. gambiae and Cx. quinquefasciatus, M. anisopliae exhibits several characteristics that make it an attractive agent for biocontrol of adult mosquitoes in sub-Saharan Africa. It can be mass-produced easily and cheaply [23] and has a considerable shelf-life if stored under proper conditions. The fungus is not harmful to either birds, fish, or mammals [34] and since the fungus is one of the most common entomopathogenic fungi, with a worldwide distribution, its use for biocontrol would not mean the introduction of a non-endemic organism into the African ecosystem.

Conclusions
The experiments clearly showed that the malaria vector An. gambiae s.s., and the filariasis vector Cx quinquefasciatus are susceptible to M. anisopliae. Their lifespan is greatly reduced if contaminated with an appropriate dose of conidia. As mosquito longevity is the single-most important parameter in the vectorial capacity equation, prospects for developing this adult mosquito control strategy are promising and may in due course be developed into an adult mosquito control tool. Questions regarding application methodology, with the aim to optimize exposure of vector mosquitoes to sources of fungal spores/ conidia, are currently being addressed.

Authors contributions
E-JS, BNN and RCS were directly involved in the experimental work, analysis of the data and drafting of the manuscript. BGJK conceived of the study, obtained funding for it in collaboration with WT and revised the manuscript prior to submission.