The Mosquito Melanization Response Is Implicated in Defense against the Entomopathogenic Fungus Beauveria bassiana

Mosquito immunity studies have focused mainly on characterizing immune effector mechanisms elicited against parasites, bacteria and more recently, viruses. However, those elicited against entomopathogenic fungi remain poorly understood, despite the ubiquitous nature of these microorganisms and their unique invasion route that bypasses the midgut epithelium, an important immune tissue and physical barrier. Here, we used the malaria vector Anopheles gambiae as a model to investigate the role of melanization, a potent immune effector mechanism of arthropods, in mosquito defense against the entomopathogenic fungus Beauveria bassiana, using in vivo functional genetic analysis and confocal microscopy. The temporal monitoring of fungal growth in mosquitoes injected with B. bassiana conidia showed that melanin eventually formed on all stages, including conidia, germ tubes and hyphae, except the single cell hyphal bodies. Nevertheless, melanin rarely aborted the growth of any of these stages and the mycelium continued growing despite being melanized. Silencing TEP1 and CLIPA8, key positive regulators of Plasmodium and bacterial melanization in A. gambiae, abolished completely melanin formation on hyphae but not on germinating conidia or germ tubes. The detection of a layer of hemocytes surrounding germinating conidia but not hyphae suggested that melanization of early fungal stages is cell-mediated while that of late stages is a humoral response dependent on TEP1 and CLIPA8. Microscopic analysis revealed specific association of TEP1 with surfaces of hyphae and the requirement of both, TEP1 and CLIPA8, for recruiting phenoloxidase to these surfaces. Finally, fungal proliferation was more rapid in TEP1 and CLIPA8 knockdown mosquitoes which exhibited increased sensitivity to natural B. bassiana infections than controls. In sum, the mosquito melanization response retards significantly B. bassiana growth and dissemination, a finding that may be exploited to design transgenic fungi with more potent bio-control activities against mosquitoes.


Introduction
Melanization is an immediate immune response in arthropods leading to the physical encapsulation of pathogens in a dense melanin coat, and to the generation of toxic metabolites that can harm certain pathogens. It is also a prominent wound healing process manifested by the blackening of the wound area in arthropods. Melanization is triggered by pattern recognition receptors (PRRs) that upon binding pathogen associated molecular patterns (PAMPs) activate a cascade of serine proteases culminating in the proteolytic cleavage and conversion of the prophenoloxidase (PPO) zymogen into active phenoloxidase (PO), the rate limiting enzyme in melanogenesis [1]. The protease cascade acting upstream of PPO involves often a modular protease and several downstream clip-domain serine proteases (CLIPs) [2,3]. This cascade is under tight temporal regulation by serine protease inhibitors (SRPNs). In the dipterans Drosophila and Anopheles gambiae, the absence of SPN43Ac, also called necrotic, [4] and SPRN2 [5], respectively, resulted in the appearance of spontane-ous melanotic pseudotumors in adult tissues and a reduced life span, suggesting that aberrant control of melanization imposes a fitness cost on the host. Further, this process is regulated spatially which ensures that melanin formation occurs exclusively on microbial surfaces minimizing collateral damage to the host.
Biochemical studies in Manduca sexta [6] and Tenebrio molitor [7] revealed that PPO activation is further controlled by the requirement of non-catalytic CLIPs [also known as serine protease homologs (SPHs)] as co-factors for prophenoloxidase activating enzymes (PPAE) to trigger proper processing of PPO into PO. SPHs have substitutions at one or more of the critical His/Asp/ Ser triad that renders them non-catalytic. Functional genetic studies in the malaria vector A. gambiae revealed that a clip domaincontaining SPH termed CLIPA8 is required for the melanization of Plasmodium berghei ookinetes in the mosquito midgut [8] and bacteria in the hemocoel [9]. While there is no evidence yet for the direct involvement of CLIPA8 in the processing of PPO, these studies provided a strong genetic evidence for the central role of SPHs in the melanization response in vivo.
Several reports have linked melanization to insect defense. In the crustacean Pacifastacus leniusculus, PO activity is required for defense against the bacterial pathogen Aeromonas hydrophila: RNAimediated silencing of PO was associated with increased susceptibility to A. hydrophila while silencing pacifastin, an inhibitor of the crayfish PO cascade, resulted in increased resistance to the bacterium [10]. The fact that Photorhabdus bacteria pathogenic to M. sexta [11], and polydnaviruses carried by female parasitoid wasps [12], evolved independent specific strategies to counteract the host melanization response is a further indication of the importance of this response in insect defense.
Previous genetic studies in the model dipteran Drosophila revealed that the melanization response does not seem to be critical for survival of flies after bacterial or fungal infections [13,14]; rather, melanization seems to enhance the effectiveness of subsequent immune reactions in the fly by weakening a microbial infection at an early stage [14]. However, a more recent study, employing a larger panel of bacterial species, demonstrated an important role for this immune process in modulating tolerance as well as resistance of the fly to specific bacterial infections [15]. Abolishing PO activity in the malaria vector A. gambiae, by silencing CLIPA8, did not affect mosquito survival after infections with Escherichia coli or Staphylococcus aureus [9]. Both bacterial species were cleared from CLIPA8-silenced mosquitoes as efficiently as from controls suggesting that melanization is not critical for antibacterial defense in the mosquito. In A. gambiae, the melanization response to P. berghei is also controlled by CLIPA8 [8], in addition to the complement-like protein TEP1 [16] and two leucine-rich immune proteins, LRIM1 [17] and APL1C [18,19]. The latter two proteins form an obligate disulfide-linked heterodimer in the mosquito hemolymph that interacts with and stabilizes a cleaved form of TEP1 [20,21]. In addition to triggering ookinete lysis in the basal labyrinth of the midgut epithelium [16][17][18], the TEP1/ LRIM1/APL1C complex (henceforth TEP1 complex) is also required for the melanotic response to ookinetes in refractory mosquito genotypes [16,17] as well as to bacteria injected directly into the hemolymph (unpublished data). Nevertheless, wildtype laboratory and field caught A. gambiae mosquitoes rarely melanize malaria parasites [22] and melanization does not seem to be important for A. gambiae anti-bacterial defense [9], questioning the role of this response in mosquito immunity.
Research on mosquito immunity has focused mainly on parasites, bacteria (reviewed in [23]) and lately viruses [24][25][26], whereas entomopathogenic fungi received little attention despite their ubiquitous nature and their route of infection which unlike other microbial classes does not require ingestion by the host. Rather, these fungi infect by direct penetration through the mosquito cuticle into the hemolymph. This mode of infection is particularly attractive for immunity studies because it does not require artificial injection of the microbe into the hemolymph. It also bypasses the midgut epithelium which was shown recently to engage in promoting complement-mediated ookinete lysis in the basal labyrinth of the midgut epithelium, by triggering intracellular nitration of ookinete surface proteins [27]. Here, we investigate the role of melanization in defense against natural infections with the entomopathogenic fungus B. bassiana and provide novel insights into the cellular and molecular mechanisms triggering fungal melanization in vivo.

Beauveria bassiana infection triggers the melanization response of Anopheles gambiae
Mosquito immune responses to entomopathogenic fungi such as B. bassiana remain poorly understood. We carried a meticulous microscopic analysis of B. bassiana development in adult A. gambiae mosquitoes to determine whether melanization is also triggered in this model infection system and against which stages. To facilitate detection of the fungus in dissected mosquito abdomens we utilized a GFP-expressing strain of B. bassiana [28]. Individual mosquitoes were injected intrathoracically with 200 freshly prepared conidia (spores) and fungal development was monitored in dissected abdomens at 1, 6, 12, 24 and 48 h post-infection (pi). In these assays, mosquitoes were infected by injecting conidia rather than by the natural route (tarsal contact with spores), because in the former mycelial growth was frequently observed in dissected abdomens which was the not case in natural infections. Most conidia were rapidly melanized at 1 hr pi; these appeared black since GFP was masked by the melanotic capsule ( Figure 1A). Only rare nonmelanized conidia were observed at that time point. At 6 h pi all conidia were melanized ( Figure 1B), however, at later time points, some melanized conidia exhibited an enlarged size indicating that germination was taking place within the melanotic capsules ( Figure 1C). Indeed, at 24 h pi, germ tubes started emerging from melanized conidia concomitant with melanin formation around their walls ( Figure 1D). Two days pi, melanin was also detected on several hyphae that emerged from germ tubes, while few hyphae remained non-melanized ( Figures 1E and 1F). Altogether, our data revealed that conidia, germ tubes and hyphae were efficiently melanized in A. gambiae mosquitoes, yet this immune reaction was not sufficient to abort completely the development of the mycelium. However, we noticed that hyphal bodies, yeast-like single-cells that differentiate from growing hyphae, were rare and sometimes absent in abdomens at 48 h post conidia injection, whereas these stages were commonly present in the hemolymph of other insect species at the same time point [29,30]. This suggests that melanization might be retarding the growth of the fungus in the mosquito.

Author Summary
Melanization is an important immune response and wound healing mechanism in arthropods that leads to melanin formation and deposition on microbial and wound surfaces, respectively. In the Anopheles gambiae mosquito that transmits the malaria parasite Plasmodium, melanization is dispensable for parasite killing. Further, we have shown that in Anopheles gambiae this immune process does not seem to play a role in defense against bacterial infections, which questions the role of melanization in mosquito immunity and the microbial pressure that drove its evolutionary path. Here, we infected mosquitoes with the entomopathogenic fungus Beauveria bassiana to study the role of melanization in anti-fungal defense. We show that mosquito blood cells, as well as specific immune proteins present in the mosquito blood participate in the melanization response to Beauveria bassiana. Our results also reveal that melanization does not abort the growth of the fungus but rather retards significantly its proliferation. The hyphal body stages, which are freely circulating single cells that can disseminate the infection appear earlier in mosquitoes exhibiting a compromised melanization response than in control mosquitoes. These findings provide novel insights into Beauveria bassiana-Anopheles gambiae interactions, which may be exploited to design transgenic fungi with enhanced bio-control potential against mosquitoes.

Cellular and humoral melanotic responses elicited against B. bassiana in adult mosquitoes
The melanization of P. berghei in certain refractory A. gambiae mosquito genotypes is a humoral response dependent on CLIPA8 [8], and TEP1 complex [16,17,20,21]. In this model system, the midgut basal lamina constitutes a physical barrier that inhibits direct contact between hemocytes and ookinetes residing in the basal labyrinth. Bacteria injected into the hemolymph also elicit a humoral melanotic response dependent on CLIPA8 [9] and TEP1 complex (unpublished data), suggesting that these are core proteins in the mosquito melanization response. To address whether they exhibit similar roles in infections with B. bassiana, TEP1 and CLIPA8 were silenced in adult female mosquitoes by RNAi knockdown (kd) [31], then individual mosquitoes were injected with 200 conidia of GFP-expressing B. bassiana. Mosquito abdomens were dissected two days after spore injection in order to score fungal melanization. Western blot analysis of hemolymph extracts confirmed that TEP1 and CLIPA8 were efficiently silenced four days after injection of their corresponding double-stranded RNAs (Figure 2A). In LacZ kd control mosquitoes, a thick melanin coat covered the majority of the growing mycelium as expected ( Figure 2B). In contrast, hyphal melanization was completely abolished in CLIPA8 and TEP1 kd mosquitoes, suggesting that these proteins are indeed core regulators of the melanization response ( Figures 2C and 2D, respectively). Intriguingly, in these two genotypes, only the base of the growing mycelium from which hyphae emerged was still melanized as efficiently as in LacZ kd controls. These findings were unexpected since the same gene knockdowns completely abolished the melanotic response to P. berghei [8,16] and bacterial infections ( [9] and unpublished data).
We hypothesized that two distinct mechanisms are driving the melanotic response to the fungus. The first is hemocyte-mediated and targets the early stages of fungal development in the mosquito, including the germinating spores and germ tubes. The second is humoral, dependent on TEP1 and CLIPA8 functions, and targets the hyphae that develop later. To address this point, abdomens dissected from wildtype mosquitoes, at 12 and 48 h after conidia injection were immunostained with polyclonal antibody against PPO6, which is known to be expressed in hemocytes [32,33]. Abdomens dissected at the earlier time point clearly showed a circular arrangement of hemocytes around enlarged conidia that were apparently germinating within the melanotic capsule ( Figure 3A). Most of these hemocytes showed absence of, or a faint PPO signal possibly because they have exhausted their PPO Alternatively, some of these hemocyte may not express PPO. A hemocyte strongly expressing PPO was resting on top of two other hemocytes that are in direct contact with the conidium (Figures 3A), suggesting that hemocytes recruited to the germinating spore may form more than one layer around it attempting to abort its growth, pretty much similar to nodule formation in larger insects. At 48 h after infection, no hemocytes were detected in close proximity to melanized hyphae supporting the humoral nature of this response ( Figure 3B). The hyphal tips from where growth occurs exhibited a thin, often barely detectable layer of melanin, but a strong PPO signal ( Figure 3B), suggesting that melanin biosynthetic reactions were still particularly active at these foci. Yet, the whole was taking place in the absence of hemocytes from the hyphal tips.
The microscopic analysis described above indicates that the early melanotic response to conidia injection requires the direct participation of hemocytes while that triggered against growing hyphae is humoral and dependent on TEP1 and CLIPA8. To investigate further this point, we measured the temporal dynamics of hemolymph PO activity in CLIPA8, TEP1 and LacZ kd mosquitoes at 24, 48 and 72 h after spraying with a suspension of 1610 8 conidia/ml. PO activities in both CLIPA8 ( Figure 4A) and TEP1 kd mosquitoes ( Figure 4B) were similar to that in LacZ kd controls at 24 h post-challenge, when the fungus has just invaded the cuticle. However, at later time points, the activity dropped significantly in both CLIPA8 and TEP1 kd mosquitoes while it remained relatively unchanged in controls, which further supports the humoral nature of the melanotic response to late fungal stages.

PPO recruitment to hyphae is dependent on both TEP1 and CLIPA8
Initiation of the melanization reaction requires limited proteolytic cleavage of zymogen PPO into active PO, the rate limiting enzyme in melanogenesis. The mechanisms which trigger PPO recruitment to microbial surfaces remain unclear. Here, we analyzed PPO localization to hyphae in TEP1, CLIPA8 and LacZ kd (control) mosquitoes at 48 h after conidia injection, using confocal microscopy. In the control group, PPO staining was observed on mycelial structures coated with a thick melanin capsule (data not shown) as previously reported in Figure 3B. Additionally, PPO was also detected along the length of hyphae on which melanin deposition was barely detectable or even absent ( Figure 5A), as if a lag phase existed between PPO recruitment and melanogenesis on these hyphal surfaces. PPO staining was often detected around the branching points of established hyphae ( Figure 5A). Interestingly, silencing CLIPA8 or TEP1 completely abolished PPO localization to hyphae, and consequently none of these structures was melanized ( Figures 5B and 5C, respectively). However, in these genotypes, PO was still detected on the melanized base of the mycelium from which hyphae emerged, corroborating our previous conclusion that the melanotic response against the early fungal stages is independent of TEP1 and CLIPA8 functions. Interestingly, the rare hyphal bodies detected in control mosquitoes at that time point, were not labelled with PPO ( Figure 5A) suggesting that these stages might escape melanization.
TEP1 binds to bacteria enhancing their phagocytosis by a hemocyte-like cell line [34] and to Plasmodium ookinetes, as they egress from midgut epithelial cells into the basal labyrinth, leading to their lysis. The fact that TEP1 binds to evolutionary distant microbial surfaces and that PPO localization to hyphae is TEP1dependent, prompted us to study whether TEP1 associates with hyphal surfaces to trigger downstream events culminating in PPO activation and subsequent melanin formation. Abdomens dissected from control (LacZ kd) mosquitoes at 48 h after conidia injections revealed strong TEP1 localization on melanin-free hyphal surfaces ( Figure 6A); where melanin had previously formed, TEP1 signal was either faint or absent, possibly because it was masked by the thick melanotic capsule. Also, the rare hyphal bodies detected in these abdomens were labelled with TEP1. Thus, TEP1 localization to hyphal surfaces clearly precedes melanin formation; the tips from where hyphae grew were always TEP1 positive but melanin negative ( Figure 6A). In TEP1 kd mosquitoes, the melanization of hyphae was completely abolished ( Figure 6B). Interestingly, hyphal bodies were more common in these mosquitoes relative to controls, suggesting rapid fungal growth. In summary, our data revealed that TEP1 association with hyphal surfaces is a prerequisite for the initiation of a local melanotic reaction against B. bassiana.

The mosquito melanization response protects against B. bassiana infection
The microscopic observation of melanotic capsules around entomopathogenic fungi has been reported earlier in several insect species including Chironomus [35], the leafhopper Empoasca fabae [36] and the grasshopper Melanoplus sanguinipes [37]. However, the relative contribution of this immune response to anti-fungal defense remains poorly understood. In A. gambiae, melanization is dispensable for defense against bacterial infections, despite the fact that bacteria trigger PPO activation in the hemolymph [9]. Additionally, field caught A. gambiae mosquitoes [22] as well as most laboratory strains rarely melanize Plasmodium ookinetes suggesting that melanization is dispensable for defense against these parasite stages. The fact that mosquitoes mounted a strong melanotic response to B. bassiana prompted us to test the relevance of this response to anti-fungal immunity. To address this point, TEP1, CLIPA8 and LacZ kd adult female mosquitoes were naturally infected with a wildtype B. bassiana strain (80.2) either by spraying with a suspension of 1610 8 condia/ml or by gentle dragging over a lawn of spores on a potato dextrose agar plate. Mosquitoes were then incubated at 27uC at 90% humidity and their survival scored on a daily basis. Survival assays revealed a significant increase in susceptibility of CLIPA8 and TEP1 kd mosquitoes to B. bassiana over controls, whether gentle dragging ( Figure 7A) or spraying ( Figure 7B) was used to establish an infection. Interestingly, the TEP1 kd group succumbed more quickly to infection than the CLIPA8 kd, suggesting that TEP1 might be controlling more than one anti-fungal effector mechanism.
We then scored hyphal body colony forming units in CLIPA8, TEP1 and LacZ kd mosquitoes four days after spraying with 5610 7 conidia/ml, to determine whether the compromised survival in the two former genotypes is due to increased fungal proliferation. Our data revealed that CLIPA8 and TEP1 kd mosquitoes contained indeed significantly higher numbers of hyphal bodies than controls which contained none at that time point; the median values were 15, 50 and 0, respectively ( Figure 7C). Here, it is worth mentioning that, in general, hyphal bodies appeared more quickly in wildtype mosquitoes when fungal infection was established through injection rather than the natural route. This explains why these stages were sometimes detected in whole mounts of abdomens at 48 h post conidia injection ( Figures 5A and 6B), but were absent from mosquitoes even at day four after natural infection ( Figure 7C). Hyphal bodies were significantly more abundant in TEP1 (P = 0.0017) than in CLIPA8 kd mosquitoes, which explains the increased sensitivity of the former genotype to B. bassiana challenge.

Discussion
Melanization is an important immune response in insects that is triggered against diverse microbial classes including parasites [38], bacteria [9,15,39] and fungi [35][36][37]. Functional genetic studies in several insect species revealed an important role for this response in insect immunity to bacterial infections [10,11,15]. Here, we investigated the role of melanization in A. gambiae anti-fungal defense using B. bassiana as a model. Our work has been prompted by early electron microscopy studies showing the formation of melanotic capsules around pathogenic fungi invading the hemocoel of other insect species [36,37], and by the fact that entomopathogenic fungi employ a different route for mosquito invasion compared to bacteria and Plasmodium parasites. The latter two, naturally infect through the oral route and traverse the midgut epithelium in order to gain access into the hemocoel, whereas pathogenic fungi breach the cuticle reaching directly into the hemocoel using a combination of mechanical pressure and an array of cuticle-degrading enzymes [40].
Results obtained from temporal analysis of the melanotic response to B. bassiana developmental stages in adult mosquitoes, using fluorescent microscopy, are in line with early reports showing that this response did not prevent the germination of B. bassiana conidia in the hemolymph of other insect species [37,41]. Melanization occurred rapidly on injected conidia, then progressed over the germ tubes as well as hyphae that constitute the bulk of the mycelium (Figure 1). Only in rare cases was the mycelium completely melanized, rather hyphae almost always succeeded to break through the melanotic capsule. Our results revealed that the mosquito mounts a potent melanotic response against the fungus, with melanized hyphae sometimes measuring more than one millimeter in length (data not shown). This response, however, is not sufficient to kill the fungus. A possible explanation could be the depletion of hemolymph PPO later during infection, due to the continuous triggering of the response by the rapidly growing fungus; however, western blot analysis excluded such possibility since PPO levels remained relatively unchanged in the hemolymph up to five days post-infection ( Figure S1). Nevertheless, we provided, for the first time, tangible evidence that melanization retards significantly the growth of the fungus in the mosquito. This is reflected in the absence of hyphal bodies in control (LacZ kd) mosquitoes four days after spraying with a conidial suspension, compared to their presence in CLIPA8 and TEP1 kd mosquitoes processed at the same time ( Figure 7C).  The delay in the differentiation of hyphal bodies in control mosquitoes is probably imposed by the strong melanotic response triggered against hyphae. This is supported by the detection of PPO and TEP1 staining not only on hyphae but also around the branching points where new hyphae and possibly hyphal bodies emerged ( Figures 5A and 6A, respectively). Delaying or inhibiting hyphal body differentiation may limit fungal dissemination, since these single cell stages proliferate in the hemolymph ultimately establishing their own mycelia. It was previously reported that melanin exhibits anti-fungal properties in vitro against Aphanomyces astaci [42] and Metarhizium anisopliae [43], however, the mechanism by which it interferes with fungal growth is still not clear. A plausible explanation is the ability of melanin to bind and inhibit the activity of a wide range of proteins [44] including lytic enzymes produced by microbes, such as chitinases, which are involved in fungal cell wall remodeling during cell division [45]. Hence, melanin might slow down fungal growth by interfering with the synthesis of new cell wall material during that process. TEP1 signal overlapped with a thin layer of melanin (Ai and Aii, arrows with open heads). TEP1 was not detected on heavily melanized hyphal surfaces probably because it was masked by melanin (Ai and Aii, arrows with filled heads). (B) TEP1 staining of hyphae was completely abolished in TEP1 kd mosquitoes (compare Aii with Bii, and Aiii with Biii). Consequently, melanin did not form over these hyphal surfaces but was present only on early fungal stages including germinating spores and germ tubes (Bi, open arrowheads). Note the abundance of hyphal bodies in TEP1 kd (Biii, arrows) compared to controls. TEP1 (red) GFP-expressing B. bassiana (green), nuclei (blue). All scale bars are 50 mm. doi:10.1371/journal.ppat.1003029.g006 The rare hyphal bodies detected in control mosquitoes at 48 h after conidia injection were not melanized nor exhibited a PPO signal ( Figure 5A), suggesting that they escape melanization. The evasion of host defense by these in vivo stages have been proposed earlier and was attributed to their minimal cell wall which lacks immuno-stimulatory carbohydrates [29]. A more recent study based on lectin-mapping revealed that B. bassiana developmental stages exhibit differences in the composition of surface carbohydrates, in particular hyphal bodies which seem to shed most carbohydrate epitopes from their surface [30]. This minimal cell wall, however, did not prevent TEP1 association with the surface of hyphal bodies ( Figure 6A). TEP1 recruitment to GFPexpressing P. berghei ookinetes triggers parasite lysis as reflected by the loss of cytoplasmic GFP signal and membrane blebbing [16]. TEP1 labelled hyphal bodies were still expressing GFP and did not show an aberrant morphology, however, it is difficult to conclude at that stage whether they are live or not without detailed electron microscopy analysis. Nevertheless, the fact TEP1 kd mosquitoes exhibited significantly higher numbers of hyphal bodies and increased sensitivity to natural B. bassiana infections compared to CLIPA8 kd, inform an additional, melanizationindependent role of TEP1 in limiting fungal growth, that remain to be defined. TEP1 is known to be an important anti-bacterial factor [34,46], which raises the possibility that the rapid death observed in fungal infected TEP1 kd mosquitoes could be due to the proliferation of opportunistic bacterial infections rather than fungal proliferation. However, this possibility was excluded because fungal infection triggered a similar survival pattern in TEP1 kd mosquitoes pre-treated with a cocktail of antibiotics ( Figure S2).
Using immunohistochemistry and confocal microscopy, we observed a PPO-positive signal on hyphae that exhibited either minimal or no melanin formation (at least within the resolving power of light microscopy) in control mosquitoes ( Figure 5A). On these hyphae melanogenesis appeared lagging behind fungal growth, whereby apical parts of hyphae exhibited strong PPO staining but minimal or no melanin formation, while the basal parts showed PPO staining around thick melanotic capsules. This is the first time mosquito PPO is detected on microbial surfaces that do not exhibit clear signs of melanin formation. In A. gambiae-P. berghei model system, a PPO signal was always detected around dead parasites confined in a dense melanotic capsule [16]. A plausible explanation for this unusual pattern of PPO localization to hyphae is that the rapidly growing fungus probably exhausts the mosquito melanotic response. This is supported by the finding that PO activity in control mosquitoes did not change significantly between 24 and 72 h post infection, suggesting continuous PPO activation, at least during that period (Figure 4).
The depletion of TEP1 or CLIPA8 from mosquito hemolymph completely abolished PPO localization to hyphae and their subsequent melanization, suggesting that PPO recruitment to fungal surfaces is an indirect process, that depends, most likely, on the prior assembly of an immune protein complex on microbial surfaces. Whether this complex includes TEP1 and CLIPA8, and whether these two proteins recruit PPO directly or indirectly to microbial surfaces remain to be elucidated. Our data also revealed that, in addition to their role as cofactors for PPO activation [6,47], clip-domain serine protease homologs seem to be required for PPO recruitment to microbial surfaces. It was previously reported that TEP1 binding to ookinetes in a melanotic refractory strain of A. gambiae, triggered their lysis and subsequent melanotic encapsulation. In that study, the authors proposed that PPO activation and recruitment was triggered by dead parasites, already killed by TEP1, and not by TEP1 itself, suggesting that the melanotic response is reminiscent of wound healing and does not represent an immune defense reaction per se [16]. It is difficult to reconcile our findings with those of the above study for the following reasons. First, in our infection model, PPO does not seem to be recruited to the surface of a dying fungus since melanized hyphae were not killed and were still growing at their tips, often elaborating lateral branches. Second, even though we did not assay directly the co-localization of TEP1 and PO (both antibodies were produced in the same host species), the fact that both were able to bind hyphae exhibiting minimal or no melanin formation, suggests that they might co-localize on hyphal surfaces.
The A. gambiae melanization response can be triggered by bacteria [9], Plasmodium ookinetes [17,48], Sephadex beads [49] and fungi (according to this report). Further, certain immune proteins like TEP1 ( [16,49] and unpublished data) and CLIPA8 [8,9] are required in all these melanotic events; of note, the role of CLIPA8 in bead melanization has not been addressed but is expected to be also essential. Altogether, these findings indicate that the molecular mechanisms that underlie the mosquito melanization response to foreign bodies with distinct biochemical surface characteristics are controlled to a certain extent by the same genetic loci. They also raise intriguing questions as to the nature of the upstream molecular recognition process that triggers the melanotic response to each of these foreign surfaces, especially that Sephadex beads, which are inanimate bodies, are still efficiently melanized in the hemolymph.
In this report, we also showed that the mosquito elicits both cellular and humoral melanotic responses against B. bassiana. The fact that neither TEP1 nor CLIPA8 kd abolished melanization of the early fungal stages (Figures 2 and 6B) despite being essential proteins in this response [8,9,16], and the detection of a layer of hemocytes around germinating spores ( Figure 3A), suggested an early cellular melanotic response. These findings challenge the current belief that melanization in insect stages with limited numbers of hemocytes, such as adult mosquitoes, occurs in a humoral manner without the direct participation of hemocytes [50]. It is possible that these cellular responses were missed because they are rare events elicited in specific cases against particular pathogens. Surprisingly, no hemocytes were observed close to hyphae later during infection, indicating that the melanotic response to these stages is humoral. There are two plausible explanations for this phenomenon which are not necessarily mutually exclusive. First, lectin mapping assays revealed that different developmental stages of B. bassiana display diverse surface carbohydrates [30]. Since sugars play important roles in non-self recognition, distinct sugar signatures may elicit different immune responses. Second, B. bassiana may interfere, at some point during its development, with the migration of hemocytes, as previously reported in the larvae of Spodoptera exigua infected with this fungus [51]. We would like to point out however, that the melanotic response elicited against hyphae, although humoral in nature, still depends on an indirect role of hemocytes being the main producers of many immunity proteins including TEP1 [34] PPO and CLIPA8 [33].
In summary, the interactions between the mosquito melanotic response and B. bassiana are evocative of an ''arms race'' where the fungus is almost always the winner. This does not mean that melanization is not protective; we have provided evidence that this immune response retards significantly fungal growth, and might severely compromise or even completely abrogate the growth of fungi that are less virulent than B. bassiana. In fact, by successfully adapting to a particularly wide host range, B. bassiana must have evolved strategies to overcome insect immune defenses and enhance its pathogenesis. This is supported by recent insights from the B. bassiana genome which revealed species-specific expansions of gene families encoding toxins, proteases and putative effector proteins that may be associated with B. bassiana host flexibility and pathogenesis [52].
Based on our findings, we propose that transgenic B. bassiana strains designed to incapacitate the mosquito melanotic response once in the hemolymph may prove to be more potent biocontrol agents than wildtype strains. Finally, the observed delay between PPO localization to hyphal surfaces and melanogenesis renders B. bassiana a tractable model to study the yet poorly understood molecular interactions that culminate in PPO tethering and activation on microbial surfaces; these studies would be difficult to perform in other infection models where the microbe becomes quickly melanized upon contact with the hemolymph, as in the case of Plasmodium ookinetes. Anopheles gambiae rearing, antibiotic treatment, and Beauveria bassiana strains All experiments were performed with Anopheles gambiae G3 strain which was reared as previously described [53]. Briefly, A. gambiae mosquitoes were maintained at 27uC and 80% humidity with a 12 h day-night cycle. Larvae were reared on tropical fish food. Adult mosquitoes were maintained on 10% sucrose and given mice blood (Mice were anaesthetized with ketamine) for egg production. For antibiotic treatment, freshly emerged mosquitoes were maintained on a 10% sucrose solution containing gentamycin (15 mg/ml), penicillin (10 units/ml) and streptomycin (10 mg/ ml) for six days prior to infection with B. bassiana, and isolated from the rest of the colony in closed plastic containers. Fresh antibiotics solution was provided every 12 h. This antibiotic regimen was continued for an additional 24 h after fungal infection then stopped. Wildtype B. bassiana strain 80.2 (a kind gift of D. Ferrandon) and a GFP-expressing strain (242-GFP; a kind gift of M. Bidochka) were cultured on potato dextrose agar (PDA) plates at 25uC and 80% humidity. Conidia (spores) used for mosquito challenges were harvested from 3-4 weeks old cultures by adding 10 ml ddH 2 O to each PDA plate and scraping the surface of the mycelia with sterile cell scrapers. Conidia were separated from other mycelial structures over a sterile funnel packed with autoclaved glass wool, washed two times with ddH 2 O by centrifugation at 4000 rpm, counted and diluted to the appropriate concentration. Freshly prepared conidia were used for all experiments.

Gene silencing by RNA interference
Double-stranded RNAs (dsRNA) for LacZ (control), TEP1 and CLIPA8 were synthesized as previously described [9,16,21], respectively. In vivo gene silencing by RNA interference was performed as previously reported [31] and efficiency of gene silencing was confirmed by immunoblotting.

Mosquito survival and B. bassiana proliferation assay
Freshly emerged adult mosquito females were injected with dsRNAs (3 mg/ml) for LacZ, TEP1 and CLIPA8, left four days to recover, then challenged with B. bassiana (strain 80.2). For survival assays, fungal challenges were conducted in two ways. A batch of fifty cold-anaesthetized mosquitoes per genotype were either sprayed with a suspension of 1610 8 conidia/ml in 0.05% Tween-80, using glass atomizers purchased from sally@ AccessoriesforFragrances.com, or dragged gently over a lawn of conidia in PDA cultures. Mosquito survival was scored on daily basis over a week. Three biological experiments were performed for each treatment. The Kaplan-Meier survival test was used to calculate the percent survival. Statistical significance of the observed differences was calculated using the Log-rank test.
For the fungal proliferation assay, LacZ, TEP1 and CLIPA8silenced mosquitoes were sprayed with a suspension of B. bassiana (strain 80.2) containing 5610 7 conidia/ml in 0.05% Tween-80. Four days post-infection, approximately 15 batches of 2 mosquitoes each per genotype were grinded in 400 ml ddH 2 O containing 0.05% Tween-80, then 50 ml of the homogenate was spread on B. bassiana selective medium [PDA containing 1 mg/ml yeast, 50 mg/ ml Gentamycin, 50 mg/ml Penicillin, 50 mg/ml Streptomycin, 5 mg/ml Crystal violet and 250 mg/ml Dodine (Sigma)]. Plates were incubated at 25uC at 80% humidity and hyphal body colony forming units were scored six days later. The experiment was repeated twice. Statistical significance was calculated using the Mann-Whitney test; medians were considered significantly different if P,0.05.

Immunohistochemistry and confocal microscopy
Four days post-gene silencing, individual mosquitoes were injected with approximately 200 conidia of GFP-expressing B. bassiana (strain 242-GFP). Abdomens were dissected at the indicated time points, fixed in 4% formaldehyde for 50 minutes, washed 3 times in 16phosphate buffered saline (PBS) and blocked for 1 h at room temperature in blocking buffer (16PBS containing 2% BSA and 0.05% Triton X-100). Then, abdomens were incubated overnight at 4uC with rabbit polyclonal a-TEP1 (1/350) or rabbit polyclonal a-PPO6 (1/500) diluted in blocking buffer. Following incubation, abdomens were washed three times with 16 PBS containing 0.05% Triton X-100, then incubated with Alexa-546 conjugated a-rabbit secondary antibody (Molecular Probes) diluted 1/800 in blocking buffer. After washing, nuclei were stained with Hoechst (1/10000) and abdomens mounted in Vectashield mounting medium (Vector Laboratories). Images were collected using on a Zeiss LSM 710 META confocal microscope.

PPO enzymatic assay
CLIPA8, TEP1 and LacZ kd mosquitoes were sprayed with a suspension of 1610 8 conidia/ml of B. bassiana strain 80.2. Hemolymph was collected at 24, 48 and 72 h after challenge in ice-cold phosphate buffered saline (PBS) containing protease inhibitors and protein concentration was determined using the Bradford Reagent (Fermentas). PO enzymatic assay was performed as previously described [9] using approximately 5 mg hemolymph proteins per reaction. Absorbance at 492 nm was measured in a Multiskan Ex microplate reader (ThermoLabsystems) after incubation with L-DOPA at room temperature for 50 min.