Although globally malaria mortality rates have fallen by 45% between 2000 and 2012, malaria is still killing an estimated 627,000 people each year [1]. An effective way to alleviate the burden of malaria is to control its vector (anopheline mosquitoes) using insecticides. This can be achieved either through the use of insecticide-treated bed nets (ITNs) or through indoor residual spraying of insecticides (IRS). However, because of rapidly expanding insecticide resistance, there is a need to find alternatives to control the mosquitoes [1]. Entomopathogenic fungi have been proposed as novel biological control agent to kill malaria mosquitoes [2–4]. Such fungi have already been used on a wide scale to control beetles [5], locusts [6] and other pest insects in agriculture.

Spores of hypocrealean entomopathogenic fungi are able to infect insects, including mosquitoes, via attachment to the insect’s epicuticle [7]. Spores will penetrate the insect’s cuticle by forming a germ tube and appressorium. The latter structure uses mechanical pressure and produces cuticle-degrading enzymes for further penetration [8, 9]. Once the fungal structures reach the haemocoel, they are able to use the insect nutrients and grow. If the fungi are able to overcome the host immune defences, the host will die and saprophytic growth starts subsequently from the dead host. Finally, sporulation of the fungus takes place a few days later, depending on environmental conditions [10].

More specifically, the hypocrealean entomopathogenic fungi Beauveria bassiana and Metarhizium anisopliae can reduce the lifespan of mosquitoes under laboratory and field conditions [11], and they are also equally effective in killing insecticide-resistant mosquitoes [12–14]. Besides the lethal effects, fungal infections reduce rodent Plasmodium sporozoite levels [15], female fecundity and feeding propensity [16, 17]. Most likely, the latter is the result of entomopathogenic fungi that reduce the host-seeking behaviour of mosquitoes by lowering the response of olfactory receptor neurons exposed to odour cues, as demonstrated for Anopheles stephensi[18].

Entomopathogenic fungi kill mosquitoes relatively slowly compared to insecticides (1–2 weeks vs 1–2 days). This is potentially beneficial for controlling malaria, because the Plasmodium development time in the mosquito is about ten to 14 days [19]. Because of the delayed mortality, fungal pathogens specifically kill those mosquitoes that are old enough to transmit the parasite. Therefore, the selective pressure for survival in mosquitoes is reduced, and thus, the probability of developing resistance against fungal infections is much lower [20].

Beauveria bassiana is a cosmopolitan fungus from which more than thousand isolates have been collected from different locations and different points in time worldwide [21]. Phylogenetic analysis has revealed at least 18 different clades (S A Rehner, pers comm). However, to date it is not clear how much these fungal isolates vary in their ability to kill mosquitoes. The majority of mosquito control studies has focused on one isolate, namely IMI 391510 [12, 17, 22, 23] (except Blanford et al.[15]). It is important to study the natural variation of virulence against mosquitoes because this has an impact on the choices for selecting the optimal fungal agents for controlling malaria vectors and other mosquitoes and their diseases. The natural occurring variation in virulence can also be used to uncover the mechanisms that underpin it, which will allow the estimation of the potential for resistance development in the vector. For example, the whole genome of isolates with contrasting variation in virulence could be sequenced, and pairwise polymorphisms, deletions and Single Nucleotide Polymorphisms (SNPs) in structural and regulatory parts of genes could be analysed. Furthermore, transcriptome analysis of contrasting isolates could be performed, and linking the DNA-sequence variation to variation in gene expression could provide unique data on the fungal genes involved in insect infection and the molecular genetic mechanisms which influence virulence. This, in turn, will increase the potential for the development of entomopathogenic fungi as a biological control method against mosquitoes.

On the other hand, previous studies have shown that morphological and physiological characteristics of fungi are related to their virulence, such as hyphal growth rate, conidial viability, conidia production, conidia size, enzyme secretion among other factors [24–28]. For instance, fungal isolates with rapid germination and high hyphal growth rate could be desirable because such fungi could infect the host more rapidly [24]. Such characteristics could be used as criteria for isolate selection in addition to their virulence towards mosquitoes.

Therefore, in the current study, the virulence of 29 isolates of B. bassiana was evaluated on the malaria mosquito Anopheles coluzzii (formerly: Anopheles gambiae sensu stricto). Phenotypic characteristics of the fungal isolates such as sporulation, spore size and growth rate were measured and their correlation with virulence analyzed.

Under laboratory conditions, all 29 tested isolates of B. bassiana killed An. coluzzii mosquitoes, although the rate at which this happened differed significantly among the isolates. The daily risk of death for mosquitoes when exposed to the most virulent isolate was around ten times higher than when exposed to the least virulent isolate. This observation is a substantial addition to a previous smaller scale study which tested six isolates of B. bassiana and two isolates of Metarhizium anisopliae against Anopheles stephensi mosquitoes [15]. In that study, of two isolates of M. anisopliae tested, only one was pathogenic to mosquitoes. Similarly, from six B. bassiana isolates, one was not pathogenic whereas the others had different rates of virulence.

Much progress has been made regarding the feasibility of using entomopathogenic fungi as a biological control agent for malaria mosquitoes. However, researchers have thus far overlooked the potential of selecting the most suitable isolate from the existing natural variation in fungal characteristics. From the study by Blanford et al.[15] onwards, isolate IMI391510 was selected in further experiments. This was based on practical considerations and not necessarily because it was the most virulent towards mosquitoes. For instance, this isolate was already evaluated and received regulatory approval for the control of other insects under field conditions. Follow-up studies mostly focused on isolate IMI391510 only or other isolates of M. anisopliae. These studies managed to standardize and improve their application under laboratory conditions [29], test it under field conditions [31], on insecticide-resistance mosquitoes [12–14], and on different substrates such as cloth, netting, wood, clay tiles, and mud walls [11, 17, 32]. In the present study, it has been shown the potential of selecting the most virulent isolate since this can result in an increased effectiveness of up to ten-fold. Such selection should be dependent on the ecological situation at the location of intended use, as using a highly virulent isolate may potentially lead to an increased selection pressure that can result in resistance in mosquitoes [3]. Additionally, genetic diversity of natural populations of mosquitoes have to be taken into consideration, since some genotypes could be more susceptible to fungal infection than others, and this could lead to an increased selective pressure as well [33]. Moreover, it has been shown that there is a dose-dependent effect of spore concentration [17, 29], so when comparing a highly lethal to a lethal isolate, the highly lethal isolate could have a similar effect using a lower concentration. Thus, the amount of spores can potentially be reduced, which would lower the application costs. Nevertheless, other factors need to be further considered since virulence is not the only variable that is important when using a fungus as biological control agent. Viability and persistence for example, can be limitations in tropical (malaria-endemic) climates.

It was hypothesized that phenotypic characteristics of B. bassiana potentially correlate with virulence in mosquitoes as previous studies have shown that morphological and physiological traits of B. bassiana are related to virulence in other insects [24, 25, 28]. For example, fungal isolates with rapid germination and high hyphal growth rate may be advantageous to use as a biological control agent, because such fungi could infect the host more rapidly [24]. Zhang et al.[28] found a positive correlation between growth rate, sporulation and germination rate of B. bassiana with virulence of red turpentine beetle larvae. Nevertheless, in the current study no relationship could be demonstrated between morphological characteristics and the virulence of B. bassiana isolates to An. coluzzii. This is consistent with earlier work of Talaei-Hassanloui et al.[26] who did not find any correlation between radial growth rate, spore size, germination or pigmentation and virulence on second instar larvae of Leptinotarsa decemlineata and Plutella xylostella. Ideally, one would like to be able to predict virulence of a fungus from its phenotypic characteristics without conducting bioassays that require many live insects and relatively long experimental duration before results are known. However, no such trend was observed in this study. Therefore, bio-assays remain the method of choice to unravel fungal virulence, at least for this fungal-insect system and with this selection of isolates.

The results suggest that other molecular and physiological mechanisms, such as the excretion of chitinases and the ability to avoid and/or counter the insect immune response, could be related to the variation in fungal virulence. Indeed, the wide variation observed in virulence offers a unique opportunity to understand the genetic, molecular and physiological mechanisms that underpin this variation. In B. bassiana several proteins have been characterized as relevant for virulence. For instance, proteins of the P450 family have been identified as important for cuticle degradation [34, 35], and Bbslt2, a novel mitogen-activated protein (MAP) kinase, was identified to have a crucial role in regulating fungal development, growth and pathogenicity [36]. As well as Bbgas1, a gene encoding a putative transferase (Glycosylphosphatidylinositol-Anchored β-1,3-Glucanosyltransferase) is involved in conidial thermotolerance and virulence [37]. Many more genes have also been linked to virulence, which hint that virulence is a complex process regulated by several pathways.

The results demonstrate that there is much natural variation in virulence of fungal isolates of B. bassiana that can be exploited by choosing the most suitable isolate for controlling malaria mosquitoes. This notion makes the use of entomopathogenic fungi an even more viable option, especially if used as part of an integrated vector management strategy. In addition, the natural variation observed in virulence offers the possibility to focus on the genetic mechanisms that determine this variation and would help in understanding the fungal mechanisms of B. bassiana towards malaria mosquitoes, and provide evidence that this could be an evolution-proof biological control agent. This is where the current research effort is focusing.