Two Proteases Defining a Melanization Cascade in the Immune System of Drosophila*

The melanization reaction is used as an immune mechanism in arthropods to encapsulate and kill microbial pathogens. In Drosophila, the serpin Spn27A regulates melanization apparently by inhibiting the protease that activates phenoloxidase, the key enzyme in melanin synthesis. Here, we have described the genetic characterization of two immune inducible serine proteases, MP1 and MP2, which act in a melanization cascade regulated by Spn27A. MP1 is required to activate melanization in response to both bacterial and fungal infection, whereas MP2 is mainly involved during fungal infection. Pathogenic bacteria and fungi may therefore trigger two different melanization cascades that use MP1 as a common downstream protease to activate phenoloxidase. We have also shown that the melanization reaction activated by MP1 and MP2 plays an important role in augmenting the effectiveness of other immune reactions, thereby promoting resistance of Drosophila to microbial infection.

The melanization reaction is used as an immune mechanism in arthropods to encapsulate and kill microbial pathogens. In Drosophila, the serpin Spn27A regulates melanization apparently by inhibiting the protease that activates phenoloxidase, the key enzyme in melanin synthesis. Here, we have described the genetic characterization of two immune inducible serine proteases, MP1 and MP2, which act in a melanization cascade regulated by Spn27A. MP1 is required to activate melanization in response to both bacterial and fungal infection, whereas MP2 is mainly involved during fungal infection. Pathogenic bacteria and fungi may therefore trigger two different melanization cascades that use MP1 as a common downstream protease to activate phenoloxidase. We have also shown that the melanization reaction activated by MP1 and MP2 plays an important role in augmenting the effectiveness of other immune reactions, thereby promoting resistance of Drosophila to microbial infection.
Drosophila utilizes multiple defense mechanisms against microbial pathogens (1). A major mechanism is the synthesis of antimicrobial peptides that kill pathogens in the hemolymph. Genetic studies have revealed that the Toll and Imd signaling pathways regulate antimicrobial peptide expression. The Toll pathway is important for the defense against Gram-positive bacteria and fungi (2). Upon infection, secreted pattern recognition receptors recognize microbial cell wall components and then trigger cleavage of the Spätzle protein to generate the ligand that activates Toll signaling in the fat body, an immuneresponsive tissue. GNBP1 and PGRP-SA are two pattern recognition receptors that act upstream of Toll and recognize Grampositive bacteria (3)(4)(5). Toll activation by fungal infection is independent of GNBP1 and PGRP-SA but involves the serine protease Persephone (6) and regulation by the serpin-type protease inhibitor Necrotic (7). Antimicrobial peptides against Gram-negative bacteria are induced under the control of the Imd pathway (8). The Toll and Imd pathways are homologous to the mammalian Toll-like receptor and tumor necrosis factor pathways, respectively, reflecting the evolutionary conservation of the innate immune systems between insects and mammals (1).
The melanization reaction is the most immediate immune response against invading pathogens in Drosophila (9). Although melanization is apparently not used as a defense mechanism in mammals, it serves as a major immune response against the malaria parasite in the mosquito (10). Melanization is visible by the blackening of a wound site or the surface of pathogens, which results from the synthesis and deposition of melanin. In addition to being important for wound healing, melanin can encapsulate and sequester pathogens, and the reaction intermediates appear to be directly toxic to microbes as well (11). During melanization, the enzyme phenoloxidase (PO) 4 catalyzes the oxidation of phenols to quinones, which then polymerize non-enzymatically to form melanin. In Drosophila and other arthropods, PO is produced and released into the hemolymph as an inactive zymogen called prophenoloxidase (PPO). Biochemical studies in larger insect species have identified proteases involved in melanization, including PPO-activating enzyme (PPAE) that directly cleaves and activates PPO at the end of a cascade upon microbial infection (12)(13)(14). These studies have also identified serpins as being involved in regulating the melanization cascade.
Recent genetic studies in Drosophila have provided new insights into the mechanisms that regulate the melanization reaction. The pattern recognition receptor called PGRP-LE is involved in inducing melanization in response to Gram-negative bacteria (15). In addition, the serpin Spn27A inhibits activation of the melanization reaction (16,17). The target of Spn27A is thought to be PPAE, the protease that cleaves PPO, as recombinant Spn27A inhibits the beetle PPAE in vitro (16). However, the endogenous target of Spn27A in Drosophila has remained unknown.
Here, we have described the identification of two proteases named MP1 and MP2 (melanization protease) that define a melanization cascade regulated by Spn27A, which is activated by microbial infection in Drosophila. Our data suggest the existence of two melanization cascades, one involving MP2 activated primarily by fungal infection and the other involving an analogous protease activated by bacterial infection, which converge on MP1 as a common activator of phenoloxidase. Our data also suggest that the melanization reaction, although not essential for survival following microbial infection, plays a critical role in enhancing the effectiveness of other immune reactions in Drosophila.

MATERIALS AND METHODS
Genetics and Molecular Biology-Oregon R was used as the wild-type strain unless otherwise indicated. The spz rm7 (2), rel E20 (18), and spn27A 1 (16) mutations, as well as the UAS-Spn27A (17), UAS-dFadd RNAi (19), and UAS-Dif RNAi (20) lines have been described earlier. The Bloomington Stock Center was the source of all other stocks.
The cDNAs GH02921 and LP12377 for MP1 and MP2, respectively, were obtained from the Drosophila Genomics Resource Center. To make constructs for overexpression of MP1 and MP2, DNA fragments encoding full-length MP1 and MP2 protein or just the catalytic domain of these proteases were generated by PCR using cDNA templates and cloned into the pUASP vector (21). The Easter signal sequence was attached to the catalytic domains of MP1 and MP2 to make the preactivated forms of these proteases (22). To make the MP1 and MP2 RNAi constructs, PCR fragments of ϳ600 bp of the appropriate cDNAs were cloned as inverted repeats into the pWIZ vector (23). Results were obtained with at least two independently generated transgenic lines bearing the MP1 and MP2 RNAi constructs. The Spn27A RNAi construct was made using the pUAST-R57 vector essentially as described earlier (20). Quantitative real-time PCR was performed essentially as described previously (5).
Microbial Infection and Survival Measurement-For septic injury, adult flies were pierced in the lateral side of the thorax with a tungsten needle of 0.1 mm diameter previously dipped into a concentrated microbial culture (24). The flies were photographed ϳ24 h after infection. For Beauveria bassiana infection, anesthetized flies were placed onto a plate containing a carpet of the fungus, agitated until covered by spores, and incubated at 29°C for further analysis (24). To measure survival after microbial infection, the flies were transferred to fresh vials every 2-3 days after infection, and the number of live insects was recorded (24).
PO Activity Assay-PO activity was assayed 4 h after septic injury or 24 h after natural fungal infection, essentially as described previously (25). The hemolymph of 30 adult flies was collected with a microcapillary and then squeezed into 20 l of phosphate-buffered saline containing protease inhibitors on ice. Protein concentration was measured using the Bio-Rad Protein Assay. Five g of total hemolymph proteins in 40 l of phosphate-buffered saline containing protease inhibitors was mixed with 120 l of phosphate-buffered saline saturated with L-3,4-dihydroxiphenylalanine. After incubation at room temperature for 30 min, the absorbance at 490 nm of the samples was measured. Each experiment was repeated at least three times.

Identification of Proteases MP1 and MP2 in a Screen for
Suppression of the spn27A Melanization Phenotype-The serpin Spn27A regulates the melanization reaction in Drosophila, as loss of spn27A function leads to constitutive melanization (Fig. 1A) and semilethality (16,17). In the melanization reaction, Spn27A is presumed to inhibit PPAE, the terminal protease that activates PO (Fig. 1B). Spn27A is also maternally required for embryonic dorsoventral patterning (26,27), where its target is Easter, the terminal protease in a cascade leading to activation of Toll signaling (Fig. 1B). Loss of maternal Spn27A results in a ventralized embryo phenotype, which can be partially rescued by the reduction in gene dosage of Easter (26). By analogy, we thought that the suppression of the constitutive melanization phenotype caused by the loss of spn27A function could be used as a genetic strategy to uncover the target of Spn27A and other proteases in the melanization cascade. Using standard crosses, we generated flies homozygous for the null mutation spn27A 1 and heterozygous for one of 12 chromosomal deficiencies that uncovered 13 different serine proteases of the clip family. We focused on clip proteases, which contained an N-terminal clip domain in addition to a C-terminal catalytic domain (Fig. 1C), because Easter is a clip protease and all known PPAEs identified biochemically in arthropods are also clip proteases (9). Of the 12 deficiencies screened, two showed significant (5-10-fold) suppression of the spn27A melanization phenotype (see "Materials and Methods"). These two deficiencies uncovered the genes CG1102 and CG3066, which encoded two clip proteases, hereafter called MP1 and MP2, respectively. In the catalytic domain, MP1 and MP2 were 56 and 44% identical to Easter, respectively, and similar to Easter, had sequence features (such as codon usage for the catalytic serine) shared by serine proteases that apparently define the most evolutionarily ancient class and occupy the most downstream positions within cascades ( Fig. 1C) (28). Interestingly, MP1 and MP2 were shown by microarray analysis to be up-regulated upon bacterial and fungal infection in a Toll-dependent manner (29,30).
To test that suppression of the spn27A melanization phenotype was due to functional reduction of MP1 or MP2 instead of another gene within the two deficiencies that uncover these proteases, we wanted to specifically lower the expression level of MP1 and MP2. We decided to use RNAi to knock down their expression and therefore generate transgenic flies that could express double-stranded RNA targeting either MP1 or MP2 for RNAi under the control of the UAS sequence element and the Gal4 transcriptional regulator (31). The act-Gal4 or da-Gal4 "driver," in which Gal4 protein is broadly expressed under the actin or daughterless gene promoter, induces the expression of UAS-coupled genes throughout development in virtually all tissues. To drive such ubiquitous activation of MP1 or MP2 RNAi, we generated flies carrying the act-Gal4 driver and the UAS construct encoding double-stranded RNA targeting MP1 or MP2. These MP1 and MP2 RNAi flies were viable to the adult stage, except for ϳ50% of the insects that died at the late pupal stage or during eclosion. Reverse transcription-PCR analysis indicated that MP1 and MP2 expression was specifically knocked down in each case ϳ10-fold from the wild-type level (data not shown). In addition, the lethality and constitutive melanization caused by loss of normal Spn27A function was completely suppressed in flies in which either MP1 or MP2 RNAi was simultaneously induced with Spn27A RNAi using the da-Gal4 driver (data not shown). These results suggest that MP1 and MP2 are involved in the melanization reaction that is negatively regulated by Spn27A. While we were preparing this work for publication, MP2 was reported to be involved in activating melanization (32,33).
MP1 and MP2 Are Required to Activate Melanization and PO Activity upon Microbial Infection-To test directly a requirement for MP1 and MP2 in the melanization reaction, we examined the ability of MP1 and MP2 RNAi flies to activate melanization upon microbial infection. Flies were pierced with a tungsten needle previously dipped into a concentrated solution of the Gram-negative bacterium Escherichia coli and the Gram-positive bacterium Micrococcus luteus. A few hours after piercing, significant melanization around the wound site was seen in control flies ( Fig. 2A) but not in flies in which MP1 or MP2 RNAi was induced with act-Gal4 (Fig. 2, B and C).
To examine whether MP1 and MP2 are required to activate PO, we assayed PO enzymatic activity in hemolymph extracted from adult flies (Fig. 2D). In uninfected wild-type flies, only a low level of PO activity was detected; however, 4 h after septic injury with E. coli and M. luteus, a dramatically higher level of PO activity was seen. Overexpression of Spn27A blocked the induction of PO activity by bacterial infection. When MP1 RNAi flies were similarly infected, no significant PO activity was detected in the hemolymph 4 h after infection. In MP2 RNAi flies after bacterial infection, PO activity was significantly lowered but not completely blocked. Similar results were obtained by infecting MP2 RNAi flies with E. coli and M. luteus separately (data not shown).
We also checked hemolymph PO activity after natural infection by the fungus B. bassiana (Fig. 2E). In wild-type flies, significant PO activity was induced 24 h after B. bassiana infection, although the level was lower than that after bacterial septic injury. MP1 RNAi flies again showed only a background level of PO activity after fungal infection. Interestingly, no significant PO activity was seen in MP2 RNAi flies after fungal infection, in contrast to the lowered activity seen after bacterial infection.
Altogether, these data indicate that MP1 and MP2 are required to activate melanization and PO after microbial infection. They also suggest that MP1 is involved in the response to bacteria and fungi, whereas MP2 is more specifically involved in the response to fungi.
MP1 or MP2 Activity Is Sufficient to Induce Constitutive Melanization and PO Activity-To check whether MP1 or MP2 activity is sufficient to induce constitutive melanization and PO activity, we generated transgenic flies in which the UAS/Gal4 system could be used to overexpress a preactivated form of MP1 or MP2 consisting of just the catalytic domain after signal sequence cleavage. Larvae in which preactivated MP1 was induced with act-Gal4 developed more slowly than wild-type larvae. Approximately 85% of these larvae showed constitutive melanization and eventually died with melanization visible over the whole body (Fig. 3B). The remaining 15% developed to the pupal and adult stages with no apparent defect. When we used the c564-Gal4 driver expressed in the fat body and hemocytes to induce the overexpression of preactivated MP1, most of the insects were able to survive to or beyond the pupal stage; however, 10% of the pupae and 30% of the adults showed constitu-FIGURE 1. Screening for suppressors of the spn27A melanization phenotype identifies two serine proteases, MP1 and MP2. A, the null mutation spn27A 1 for the serpin Spn27A results in constitutive melanization, as seen here on the wing (arrow), and in semilethality. B, in melanization, Spn27A is presumed to inhibit the protease PPAE that cleaves PPO to generate PO, a key enzyme in melanin biosynthesis. Spn27A also regulates embryonic dorsoventral patterning, where its target serine protease is Easter (EA). GD and SNK are serine proteases that act sequentially in a cascade to activate EA, which cleaves the Spätzle protein (SPZ) to generate the Toll ligand (SPZ*). C, MP1 and MP2 proteases contain a clip domain in the N-terminal region and a C-terminal catalytic domain. Cleavage (indicated by scissors) is required to activate clip protease. Chymotrypsin-like serine proteases can be phylogenetically classified based on sequence patterns that serve as evolutionary markers (28). The marker for the most primordial class of serine proteases is Ser 195 :TCN/ Ser 214 :TCN/Pro 225 , where the catalytic serine (Ser 195 ) and the highly conserved serine adjacent to the active site (Ser 214 ) are encoded by a TCN (N ϭ any base) codon and the residue 225 (chymotrypsinogen numbering) is Pro. MP1, MP2, EA, and SNK belong to this class, whereas GD belongs to a modern class one evolutionary transition removed from the primordial class, with residue 225 being Tyr that, unlike Pro, enhances catalysis caused by Na ϩ binding. The GenBank TM accession numbers for the protein sequences of GD, SNK, EA, MP1, and MP2 are AAF48122, AAF54897, AAF55170, AAF52151, and AAF54143, respectively. tive melanization. The ubiquitous overexpression of preactivated MP2 seemed to have a weaker effect. Approximately 10% of the larvae showed constitutive melanization (Fig. 3C), and most larvae developed into pupae, of which ϳ40% showed constitutive melanization (Fig. 3E). Most of these melanized pupae died. Among ϳ50% of the insects that survived to adulthood, ϳ5% showed a constitutive melanization phenotype (Fig. 3G). Most of the melanization appeared to be under the epithelia and in the hemolymph.
We also checked PO activity in the hemolymph of flies overexpressing preactivated MP1 or MP2 (Fig. 3H). Consistent with the melanization phenotype described above, a high level of hemolymph PO activity was detected in these flies even without any microbial infection. This level was nearly comparable with the levels seen in wild-type flies 4 h after microbial infection or in spn27A 1 flies. These results indicate that preactivated MP1 or MP2 is sufficient to induce the activation of melanization and PO activity.
Flies ubiquitously overexpressing the full-length form of either MP1 or MP2 were completely viable. Only a few (2-3%) displayed some melanization spots, and none produced significant constitutive PO activity (data not shown). These results indicate that MP1 and MP2 must be activated by zymogen cleavage to induce melanization and PO activity.
The induction of melanization and PO activity by overexpression of preactivated MP1 or MP2 was blocked by simultaneous overexpression of Spn27A (Fig. 3H ). This result indicates that MP1 and MP2 cannot act downstream of the direct target of Spn27A, which is presumed to be the terminal protease in the melanization cascade that activates PO.
MP1 and MP2 Function Sequentially in Activating Melanization and PO Activity-Lethality and melanization caused by the loss of Spn27A function are completely suppressed by RNAi knockdown of either MP1 or MP2. Moreover, MP1 and MP2 are both required, whereas preactivated MP1 or MP2 is each sufficient to induce melanization and PO activity. These observations suggest that the two proteases act in a common pathway rather than in independent pathways. To test whether MP1 and MP2 act sequentially, as in a zymogen activation cascade, we examined whether RNAi of one protease could block the induction of melanization and PO activity caused by overexpression of the preactivated form of the other protease. We found that melanization and PO activity induced by overexpression of preactivated MP2 was completely blocked by MP1 RNAi (Fig. 3H). In the reciprocal experiment, MP2 RNAi did not block the melanization of larvae induced by overexpression of preactivated MP1 (as seen in Fig. 3B). PO activity in adult hemolymph could not be assayed in this case, as simultaneous induction of MP2 RNAi and preactivated MP1 overexpression led to complete lethality, compared with 85% lethality from preactivated MP1 overexpression alone (see above) before the adult stage. These results are consistent with MP2 functioning genetically upstream of MP1 to activate melanization and PO. RNAi flies (C ) showed significantly reduced melanization. The extent of melanization is outlined. D, PO activity in adult hemolymph was assayed 4 h after bacterial septic injury. In wild-type (wt) flies, PO activity was significantly induced by infection (0 and 4 h). As a control, when Spn27A was overexpressed (actϾSpn27A), PO activity was comparable with the un-induced level. In MP1 RNAi flies, PO activity was also at the un-induced level, whereas in MP2 RNAi flies, PO activity was significantly reduced but with greater variability. E, hemolymph PO activity was assayed 24 h after natural infection by the fungus B. bassiana. In wild-type flies, PO activity was significantly induced, although it was not as high as seen after bacterial septic injury (wt, 0 and 24 h). In either MP1 or MP2 RNAi flies, PO activity after natural fungus infection was comparable with the un-induced level. RNAi was induced with act-Gal4. Bars indicate S.D. of at least three independent experiments.

MP1 and MP2 Do Not Require Toll Signaling to Activate
Melanization-An earlier study suggests that melanization requires activation of the Toll signaling pathway to deplete Spn27A (17). To investigate whether MP1 and MP2 activate melanization through Toll, we overexpressed preactivated MP1 or MP2 in flies homozygous for spz rm7 , a loss-of-function mutation in the spätzle gene encoding the Toll ligand. We found that spz rm7 did not block constitutive melanization (as seen in Fig. 3B) and larval lethality caused by overexpression of preactivated MP1. Indeed, spz rm7 enhanced the partial lethality caused by preactivated MP1 overexpression alone (see above), causing complete lethality before the adult stage that precluded the assay of PO activity in adult hemolymph. However, we determined that spz rm7 did not block the high PO activity produced by overexpression of preactivated MP2 (Fig. 3H ). Thus, MP1 and MP2 appear to regulate melanization and PO activation, in large part, independently of the Toll signaling pathway.
Melanization Is Important for Enhancing the Effectiveness of Other Immune Reactions-As no significant PO activity is detectable in MP1 RNAi flies after bacterial or fungal infection (Fig. 2), we used these flies to address whether the melanization reaction is important for resistance to microbial infection. Flies were challenged with a Gram-negative bacterium (Erwinia carotovora), a Gram-positive bacterium (Enterococcus faecalis) or the fungus Candida albicans, or naturally infected with the entomopathogenic fungus B. bassiana. The survival rate of MP1 RNAi flies was not significantly different from wild-type flies after all types of infection (Fig. 4, A-C, and data not shown). Similar results were obtained for MP2 RNAi flies after infection with bacteria (Fig. 4, A and B) or the fungus C. albicans (data not shown). These results suggest that the melanization reaction is largely dispensable for resistance to microbial infection. A similar conclusion was reached from a recent independent study of MP2 (33). Interestingly, however, we found that the melanization reaction was important for resistance to microbial infection in flies impaired in the Toll or the Imd pathway controlling antimicrobial peptide synthesis. For example, after infection with the Gram-negative bacterium E. carotovora, the survival rate was moderately lower for flies in which MP1 (or MP2) and dFadd in the Imd pathway were both knocked down by RNAi compared with flies in which just dFadd was knocked down (Fig. 4D). A similar decrease in the survival rate of flies from RNAi knockdown of MP1 (or MP2) and Dif in the Toll pathway, compared with knockdown of just Dif, was observed after infection with the fungus C. albicans (Fig. 4F). Such a distinct synergistic effect was less discernible after infection with the Gram-positive bac- terium E. faecalis, perhaps because of the rapid death of the flies (Fig. 4E). We note that the differences in relative survival rates described here, although modest, were reproducible despite the variability in absolute survival rates between experiments. These observations suggest that the melanization reaction, although not essential for resistance to microbial infection, can play an important role in enhancing the effectiveness of other immune reactions.
Interestingly, compared with MP1 RNAi flies, MP2 RNAi flies were less viable after natural infection with the fungus B. bassiana (Fig. 4C). The pronounced difference in susceptibility to fungal infection made us wonder whether MP2 might have a role in activating the Toll antifungal pathway in addition to its role in activating melanization. Consequently, we monitored in MP1 RNAi and MP2 RNAi flies the expression of Drosomycin, an antifungal peptide gene regulated mainly by the Toll pathway, and of Diptericin, an antibacterial peptide gene regulated by the Imd pathway. As measured by quantitative real-time reverse transcription-PCR (Fig. 4, G-I) and reporter gene expression (data not shown), the knockdown of MP1 or MP2 had no significant effect on the induction of Diptericin after infection by E. carotovora, indicating that MP1 and MP2 do not regulate the Imd pathway and that the fat body, the site of antimicrobial peptide expression, is still functional (Fig. 4G). However, MP2 RNAi (but not MP1 RNAi) adults showed reduced Drosomycin expression, ϳ50% of the wild-type level, after infection with the fungus B. bassiana (Fig. 4I). MP2 RNAi did not similarly reduce Drosomycin expression after challenge by the Grampositive bacterium M. luteus (Fig. 4H). These results suggest that MP2 has a role in activating Drosomycin expression in response to natural fungal infection.

DISCUSSION
Melanization is a conserved host defense reaction in insects and other arthropods, such as in the mosquito, where it is a critical determinant of resistance to the malarial parasite. A recent key finding revealed that Spn27A, a serpin-type protease inhibitor, is a negative regulator of melanization in Drosophila. In this study, we have used suppression of the spn27A melanization phenotype as a genetic strategy to identify two Drosophila proteases, MP1 and MP2, having essential roles in activating melanization in response to microbial infection.
Our data are consistent with MP2 acting genetically upstream of MP1 and thus with MP2 activating MP1 in a protease cascade leading to melanization (Fig. 5). Interestingly, the application of evolutionary markers suggests that MP1 and MP2 belong to the primordial class of serine proteases that tend to function most downstream in a protease cascade, as in the case of Easter and its direct activator involved in activating Toll during development (Fig. 1C). However, we have not been able to demonstrate that MP2 directly activates MP1, as MP2 failed to cleave the zymogen form of MP1 when co-expressed in transfected Drosophila S2 cells, thereby suggesting that another protease acts in between MP2 and MP1. As the protease down-stream of MP2 that acts either in parallel to or upstream of Spn27A, MP1 is a candidate to be PPAE, the terminal protease in the melanization cascade that cleaves PPO as well as the putative direct target of Spn27A. We have not detected cleavage of PPO by MP1 when assayed by co-expression in transfected S2 cells. 5 However, biochemical studies in the beetle and tobacco hornworm have identified a non-enzymatic cofactor required for PPO cleavage by PPAE to generate PO activity in vitro (13,34), which may indicate that such a cofactor is required to demonstrate PPO cleavage by MP1.
Our data suggest that another melanization cascade exists in Drosophila besides the one that we have defined involving MP1 and MP2. In activating melanization and PO activity, MP1 is essential during both bacterial and fungal infection, whereas MP2 is essential during fungal infection and partially required during bacterial infection ( Fig. 2D and E). Thus, another protease may function analogously to MP2 in activating MP1 during bacterial infection (Fig. 5). The convergence of two different melanization cascades on MP1 is consistent with the idea that MP1 is the shared terminal protease of both cascades that activates phenoloxidase.
We presume that MP1 and MP2 activate melanization in the hemolymph. As both MP1 and MP2 have an N-terminal signal sequence for secretion, they may be secreted by the fat body and/or blood cells into the hemolymph. This possibility is consistent with the detection of MP1 and MP2 in extracted hemolymph when epitope-tagged versions of the full-length proteases are overexpressed with act-Gal4 (data not shown) and with constitutive melanization induced by overexpression of preactivated MP1 using the c564-Gal4 driver, which is expressed in the fat body and hemocytes (see "Results").
MP1 and MP2 define a protease cascade distinct from the one that may activate the Toll pathway, as they activate melanization independently of this pathway (Fig. 3). However, we found that MP2 (but not MP1) is important for the induction of Drosomycin expression and for viability of Drosophila following natural fungal infection (Fig. 4). One explanation is that MP2 activates two distinct pathways, one leading to melanization and the other leading to the induction of Drosomycin expression (Fig. 5), and that this dual role is important for resistance to natural fungal infection. Preliminary experiments indicate that overexpression of preactivated MP2 does not induce Drosomycin expression in non-infected adult flies, unlike in the case of the Persephone protease (6), thereby suggesting that MP2 does not induce Drosomycin expression by directly activating the Toll pathway. Nonetheless, there may exist cross-talk between melanization and the Toll pathway mediated by MP2 involving an as yet undefined mechanism.
Surprisingly, the melanization reaction does not appear to be critical for survival of Drosophila after bacterial or fungal infection ( Fig. 4; 33). However, we observed that the inability to activate melanization is detrimental when flies are also defective in the Toll or the Imd pathway controlling antimicrobial peptide synthesis (Fig. 4). Melanization is an immediate immune response that temporally precedes the induction of 5 H. Tang and C. Hashimoto, unpublished data. FIGURE 5. Model depicting MP1 and MP2 proteases in multiple immune pathways. MP1 and MP2 act in the melanization cascade to activate PO, the key enzyme in melanin synthesis. This cascade is triggered primarily by fungi and to a lesser degree by bacteria. Another melanization cascade involving an unidentified protease and MP1 may be more specifically induced during bacterial infection. MP2 is also involved in inducing antimicrobial peptide expression by an unknown mechanism potentially involving cross-talk with the Toll pathway. Arrows represent genetic relationships and not necessarily direct biochemical interactions.
antimicrobial peptide synthesis, which requires gene transcription. Consequently, melanization may play a crucial role in weakening a microbial infection, thereby enhancing the effectiveness of subsequent immune reactions. Having a single switch, such as MP2, to activate a temporal sequence of immune reactions would therefore seem to be an advantageous mechanism for ensuring a potent defense against a microbial pathogen.
In conclusion, we have identified two serine proteases among the large set of serine proteases encoded in the Drosophila genome as being essential components of a melanization cascade activated by microbial infection. A major goal in future studies will be to delineate the entire cascade from the pattern recognition receptor that triggers the cascade to the putative cofactor required for phenoloxidase activation.