Background

The constant interactions of infectious microbes with their hosts explain the emergence of complex immune systems. Comparative immunology provides one of the best approaches to understanding the logic of metazoan host defense mechanisms and their diversification throughout evolution. This approach reveals that despite the apparently extreme diversity of immune reactions, similar mechanisms are used across the animal kingdom to cope with microbes. Phagocytosis, epithelial production of antimicrobial peptides, sequestration of iron or mucus barriers are found in many organisms, indicating either an ancient origin or their recurrent emergence by convergent evolution. Other immune modules are specific to a limited number of organisms. Emergence, loss and diversification of these immune modules are thought to reflect the evolutionary trajectory of metazoan lineages facing various selective pressures from pathogens. Immune diversification has often involved the expansion of particular gene families. For instance, the number of genes encoding Toll-like receptors, peptidoglycan recognition proteins and C-type lectins markedly varies among phyla [1]. A better understanding of the metazoan immune system requires an analysis of the rules underlying such immune gene amplification. The recently developed CRISPR/Cas9 genome editing approach offers a new strategy for generating mutations in a quasi-systematic manner. This allows, for the first time, the question of gene family diversification to be tackled, by generating mutations in single and multiple genes belonging to the same family. We previously reported the functional analysis of two of the three Drosophila prophenoloxidases (PPOs), PPO1 and PPO2, using single- and double-mutant analysis [2]. We subsequently realized that the PPO1,PPO2 double-mutant stock used in our previous study carries a mutation in the PPO3 gene. Here, we investigate the function of PPO3, and perform a single-, double- and triple-mutant analysis of the three Drosophila PPOs. Our study extends our previous results and attributes specific and complementary functions to each of the three PPOs, providing an insight into how this protein family has evolved in Drosophila.

Melanization is a major immune module found in arthropods but not in vertebrates [3, 4]. It involves the rapid synthesis of a black pigment named melanin, at the site of infection or injury. A key enzyme in melanin biosynthesis is phenoloxidase (PO), which catalyzes the oxidation of phenols to quinones, which subsequently polymerize into melanin. PO is usually synthesized as a zymogen called prophenoloxidase (PPO), which is cleaved to generate the active PO enzyme upon activation of a proteolytic cascade. Several roles have been ascribed to the melanization reaction in insects [5–8]. PO activity contributes to wound healing by forming a scab at the epithelial injury site. By-products of PO activity are reactive oxygen species, which are thought to contribute to the killing of microbes and pathogens. Finally, melanization participates in the encapsulation reaction against parasites. Deposition of melanin on the parasite forms a physical barrier, allowing the localized and confined production of toxic compounds while ensuring the protection of the host. Melanization is clearly an arthropod-specific mechanism, but its functions in providing a barrier and generating microbicidal reactive oxygen species are usually carried out by other mechanisms in other species.

The Drosophila melanogaster genome contains three PPO genes, all on the second chromosome [2]. PPO1 and PPO2 are produced in specialized hemocytes (blood cells), called crystal cells. Crystal cells represent 5 % of the hemocyte population in larvae [9, 10]. Upon injury, they rupture and release PPOs into the hemolymph (the insect blood), where they are activated by a cascade of serine proteases (SPs) [11]. Using null mutations in PPO1 and PPO2, we recently showed that PPO1 and PPO2 are responsible for all the PO activity in the hemolymph [2]. While PPO1 is involved in the rapid early delivery of PO activity, PPO2 present in the crystals of crystal cells provides a storage form, which can be deployed in a later phase [2]. This and other studies also revealed an important role for PPO1 and PPO2 in survival after infection with Gram-positive bacteria and fungi, underlining the importance of melanization in insect host defense [4, 12–14]. However, previous work left open the question of the function of the third PPO, PPO3, which could not be addressed because no mutant was available. The expression pattern of PPO3 was not clear either. Some reports suggested that PPO3 is expressed in crystal cells [15, 16], while others proposed it is expressed in lamellocytes [17, 18]. Lamellocytes are a type of larval hemocytes that are induced upon injury or wasp infection and play a key role in the encapsulation of foreign bodies and parasites [19]. Interestingly, while PPO1 and PPO2 require proteolytic cleavage to be activated, PPO3 is thought to be produced in an active form, although a putative cleavage site is present [18, 20].

The starting point of the present study came with the serendipitous observation that the PPO1 Δ ,PPO2 Δ double mutant that we previously used to analyze the role of PPO1 and PPO2 [2], also carries a cryptic null deletion in PPO3. This raised the possibility that the phenotype initially attributed to PPO1 and/or PPO2 could be caused by the absence of PPO3. To clarify this question, we separated the PPO3 mutation from the PPO1 Δ and PPO2 Δ mutations and additionally generated an independent PPO3 mutant using the CRISPR/Cas9 approach [21]. Using single, double and triple mutants of the three PPO genes, we confirm that PPO1 and PPO2 are the only source of hemolymphatic PO upon septic injury. Our study also reveals a role for PPO3 in the encapsulation process, in association with PPO2.

Results

Identification of a cryptic PPO3 deletion in the PPO1 Δ,PPO2 Δ double-mutant stock

By testing primers for the PPO3 gene, we serendipitously discovered the presence of a deletion of 336 bp, removing 112 (residues 105–217) of the 683 amino acids of the PPO3 protein (Fig. 1a). This deletion is referred to as PPO3 1, and we use parentheses around this mutant to refer to previous results where this mutant was present in a cryptic state. As the deletion does not change the reading frame, a residual protein is expected to be produced, lacking part of a conserved copper-binding domain found in all PPOs (Fig. 1b). The presence of this cryptic mutation raised the possibility that some of the phenotypes observed in the PPO1 Δ ,2 Δ ,(3 1 ) flies could be due to the absence of PPO3. To clarify the function of the three PPOs, we separated PPO3 1 from PPO1 Δ and PPO2 Δ by meiotic recombination. We also induced a null mutation in PPO3 using CRISPR/Cas9, referred to as PPO3 SK3, which is caused by a frameshift mutation in the first exon of the gene (Fig. 1a). Both the PPO3 1 and PPO3 SK3 mutations, which were generated in two distinct genetic backgrounds, show the same phenotype in all the experiments described below, indicating that PPO3 1 is also a null mutation. As expected, PPO3 mutants are perfectly viable and do not exhibit any overt developmental or pigmentation defect. Figure 1c shows that PPO1 Δ ,2 Δ and PPO1 Δ ,2 Δ ,3 1 but not PPO3 1 flies have a reduced life expectancy, confirming that the simultaneous presence of PPO1 and PPO2 is required for optimal fly longevity [2].

Fig. 1
figure 1

Molecular characterization of two novel PPO3 mutations. a Schematic representation of the PPO3 gene locus and PPO3 coding sequence. The gene map was adapted from FlyBase. The PPO3 1 mutation was present in the PPO1 Δ ,2 Δ double-mutant stock while the PPO3 SK3 mutation was generated by CRISPR/Cas9. Shown are the protein domains of PPO3 and a partial sequence where the deleted nucleotides are marked in red. b Structural modeling shows that the PPO3 1 mutation (orange portion) deletes most of the catalytic pocket, including copper-coordinating residues of the copper-binding domain A (light blue). Protein domains are colored as in (a). c Lifespan analysis of unchallenged flies reveals an increase in mortality rate of PPO1 Δ ,2 Δ and PPO1 Δ ,2 Δ ,3 1 mutant flies but not PPO3 1 mutants. Each survival curve corresponds to three independent experiments with 20 flies each. Chr chromosome, UTR untranslated region, wt wild type

Full size image

PO1 and PO2 are the sole POs contributing to hemolymph injury-mediated melanization in larvae and adults

Using a needle to injure wild-type larvae or adults induces a melanization spot at the wound site, the extent of which is usually proportional to the injury size. This blackening reaction results from de novo melanin synthesis catalyzed by PO and is further enhanced by the presence of microbial products [22]. Our previous study attributed all hemolymphatic PO activity to PPO1 and PPO2 due to the absence of melanization in the PPO1 Δ ,2 Δ ,(3 1 ) stocks, while single PPO1 or PPO2 mutants showed only reduced or almost normal melanization, respectively. Consistent with this finding, no melanization spot on the cuticle of injured PPO1 Δ ,2 Δ and PPO1 Δ ,2 Δ ,3 1 mutants was observed (Fig. 2a, b). In contrast, PPO3 mutants show a wild-type level of melanization in injured larvae and adults. We next measured PO activity via the L-DOPA assay in hemolymph samples extracted from larvae. We found no significant difference in PO activity between PPO3 1 and wild-type hemolymph samples. In contrast, hemolymph samples of PPO1,2 deficient larvae show no PO activity, indicating that PPO3 alone is not sufficient to produce PO activity in hemolymph (Additional file 1: Figure S1). This confirms that PPO1 and PPO2 together produce all injury-mediated melanization in larvae and adults [2]. Survival analyses had shown that PO is required for resistance to microbial infection, notably to Gram-positive bacteria and fungi [2]. Survival analyses using the new fly lines now show that the strict PPO1 Δ ,2 Δ double mutation recapitulates all the phenotypes previously described using PPO1 Δ ,2 Δ ,(3 1 ): PPO1 Δ ,2 Δ flies are more susceptible to septic injury with the Gram-positive bacteria Staphylococcus aureus, Enterococcus faecalis and Bacillus subtilis and exhibit a mild susceptibility to natural infection with the fungus Beauveria bassiana (Fig. 3). In contrast, PPO3 SK3 flies exhibit a wild-type level of resistance upon challenge with the same microbes. These experiments confirm the important role of melanization in fighting infection by Gram-positive bacteria and fungi, and are consistent with our previous results indicating that PPO1 and 2 are the sole sources for hemolymphatic PO. Additional data (Fig. 4a, b) show that PPO3 1 does not markedly affect the Toll and Imd pathways in adults as revealed by the wild-type inducibility of Diptericin and Drosomycin, their respective read-out genes, in this mutant.

Fig. 2
figure 2

Both PPO1 and PPO2 but not PPO3 contribute to injury-related melanization in adults and larvae. Melanization of adults (a) and larvae (b) after clean injury is abolished only in the simultaneous absence of PPO1 and PPO2. A normal melanization spot was observed in the two PPO3 mutants. Arrows indicate the pricking site. Adults and larvae were wounded with a tungsten needle and blackening of the wound was recorded 1 h later in larvae and 16 h later in adults. A representative picture is shown for each genotype

Full size image
Fig. 3
figure 3

Contribution of PPO1,2 and PPO3 to host defense against bacteria and fungi. Survival rates of flies following septic injury with Gram-positive lysine-type bacteria (Staphylococcus aureus and Enterococcus faecalis), Gram-positive DAP-type bacterium (Bacillus subtilis) and natural infection with the entomopathogenic fungus Beauveria bassiana. Flies lacking the Toll ligand spätzle were used as immune-deficient controls. The x-axis is the time post-infection in days and the y-axis is the percentage of living flies. PPO1 Δ ,2 Δ and PPO1 Δ ,2 Δ ,3 1 flies are less resistant to infections with S. aureus (P < 0.0001), Enterococcus faecalis (P < 0.0001), Bacillus subtilis (P < 0.0001) and Beauveria bassiana (P < 0.0005) compared to wild-type flies. Differences between PPO3 SK3 and w 1118 do not reach statistical significance. Data were analyzed by a log-rank test

Full size image
Fig. 4
figure 4

PPO3 is not required for Toll and Imd pathway activities. a Expression of Diptericin (Dpt) in PPO mutant flies. Total RNA was extracted from animals either uninfected or collected 6 h and 24 h after septic injury with Gram-negative bacteria Erwinia carotovora. Shown are the relative expression levels of Dpt in relation to Rp49. PPO3 1 flies show normal induction of Dpt expression levels. The Imd pathway mutant Relish was used as an immune-deficient control. b Expression of Drosomycin (Drs) in PPO3 1 mutant flies 24 h and 48 h after septic injury with the Gram-positive bacteria Micrococcus luteus shows that PPO3 1 mutant flies have a normal induction of Drs. The Toll pathway mutant spätzle rm7 was used as an immune-deficient control. Shown are the relative expression levels of Drs in relation to Rp49. On the graphs, 100 % corresponds to the Dpt (Drs) expression level of wild-types flies 6 h (24 h) after septic injury. Data were analyzed using the Mann–Whitney test. Values represent the mean ± standard deviation of at least three independent experiments. n.s. not significant, UC unchallenged

Full size image

PPO3 is specifically expressed in lamellocytes

Drosophila larvae have two types of hemocytes in the unchallenged state: plasmatocytes, which are macrophage-like cells, and crystal cells, which produce PPO1 and contain crystals of PPO2 [2, 19, 23, 24]. A third ‘inducible’ cell type, the lamellocytes, is produced only upon wasp infection or injury. Lamellocytes are large flat cells that differentiate in the lymph gland or from circulating plasmatocytes and adhere to foreign objects or aberrant tissue in a process called encapsulation. Capsules are usually melanized in a secondary step. The expression pattern of PPO3 has been a source of conflicting observations. Irving et al. [17] reported its localization in lamellocytes using a microarray approach. In contrast, several studies [15, 16] report expression of PPO3 in crystal cells. The observation that ubiquitous silencing of PPO3 by in vivo RNAi suppresses melanization induced by the melanotic tumor hop Tuml mutation, led Nam et al. [18] to hypothesize that PPO3 is indeed expressed in lamellocytes. To distinguish between these two possibilities, we generated a reporter gene in which the yeast transcription factor GAL4 is under the control of 1.6 kb of PPO3 upstream sequences. This transgenic line was crossed with uas-GFP flies to reveal the expression profile of PPO3 using green fluorescent protein (GFP). We found no expression of PPO3-Gal4 in plasmatocytes or crystal cells from naive larvae (data not shown). We then analyzed the expression pattern of PPO3-Gal4;uas-GFP in larvae upon infestation with the parasitoid wasp Leptopilina boulardi, which induces a massive production of lamellocytes. Figure 5a, b shows that the PPO3 reporter was not expressed in plasmatocytes or crystal cells in parasitized larvae, but was strongly expressed in all lamellocytes. High numbers of GFP positive lamellocytes were observed in circulation or around the wasp egg (Fig. 5c, d). No other tissue or cell expressed the reporter and no expression was detected in embryos and adults, consistent with the absence of lamellocytes at these two stages.

Fig. 5
figure 5

PPO3 is specifically expressed in lamellocytes. Differential interference contrast (DIC) and GFP fluorescence micrographs of hemocytes from larvae infested with Leptopilina boulardi expressing a PPO3-Gal4,UAS-GFP construct show that PPO3 is specifically expressed in all lamellocytes (L) either in circulation (a, b) or in the capsule surrounding the wasp egg (c). Plasmatocytes (arrows) and crystal cells (dashed line) do not express the reporter gene. GFP expression is green and DAPI staining is shown in blue. An overview of the whole infested larvae is shown in (d). The arrow indicates the site of the wasp larvae. Note that lamellocytes are found both around the egg and in circulation. e Quantitative reverse-transcription PCR (qRT-PCR) shows that PPO3 gene expression is higher in larvae infested with L. boulardi compared to unchallenged larvae. In the graph, 100 % corresponds to PPO3 expression levels of naive larvae. DAPI 4′,6-diamidino-2-phenylindole, GFP green fluorescent protein, UC unchallenged

Full size image

Consistent with these observations and previous reports [25, 26], qRT-PCR analysis indicated that the level of PPO3 transcripts is low in unchallenged wild-type larvae but markedly up-regulated in larvae collected 52 h after wasp infestation, when lamellocytes have become abundant (Fig. 5e). While PPO1 and PPO2 are produced by crystal cells, our data clearly demonstrate that PPO3 is specifically expressed in lamellocytes. Since lamellocytes are absent in adults, this observation also explains why PPO3 does not contribute to hemolymphatic PO activity upon injury during this stage.

Both PPO2 and PPO3 contribute to capsule melanization

A surprising result of Binggeli et al. was the observation that capsules around parasitoid wasp eggs were not melanized in PPO1 Δ ,2 Δ ,(3 1 ) larvae [2]. This led us to conclude prematurely that PPO3 was not essential for the melanization process during encapsulation. The discovery of the cryptic PPO3 1 mutation in the original stock and the specific expression of PPO3 in lamellocytes prompted us to analyze further the role of PPO3 in the encapsulation process. Wild-type, PPO3 1, PPO1 Δ ,2 Δ and PPO1 Δ ,2 Δ ,3 1 second-instar larvae were infested by Leptopilina boulardi, and the presence of a melanized wasp egg was subsequently analyzed. Figure 6 shows the presence of melanized L. boulardi eggs in wild-type, PPO1 Δ ,2 Δ double-mutant and PPO3 1 single-mutant third-instar larvae. In contrast, no melanized capsules were ever observed in the infested triple PPO1 Δ ,2 Δ ,3 1 mutants (the presence of intact wasp eggs was checked by dissecting the larvae). Contradicting our previous conclusion, this result indicates that PPO3 does play a role in encapsulation that can be masked by the presence of PPO1 and/or PPO2. We then generated PPO1 Δ ,3 1 and PPO2 Δ ,3 1 double mutants to determine which of PPO1 or PPO2 contributes to capsule melanization together with PPO3. The result was unambiguous: we never found melanized capsules in L. boulardi infested PPO2 Δ ,3 1 larvae while PPO1 Δ ,3 1 larvae were phenotypically indistinguishable from PPO3 1 (Fig. 6). Thus, two PPOs from different origins, PPO2 from crystal cells and PPO3 from lamellocytes, contribute to melanization that accompanies encapsulation.

Fig. 6
figure 6

Both PPO3 and PPO2 contribute to melanization around a wasp capsule. Representative photos showing infested larvae containing eggs from a Leptopilina boulardi parasitoid wasp. Larvae mutant for PPO3 1 or PPO1 Δ ,2 Δ show a melanized egg. No melanized capsule was found in PPO2 Δ ,3 1 and PPO1 Δ ,2 Δ ,3 1 infested mutant larvae. The presence of a non-melanized egg in PPO2 Δ ,3 1 and PPO1 Δ ,2 Δ ,3 1 mutant larvae was confirmed by subsequent dissection. Arrows indicate the position of the melanized wasp egg

Full size image

Heating larvae at 65 °C for 10 min induces the spontaneous activation of PPO [9]. As a consequence of this treatment, the population of sessile crystal cells that are attached underneath the integument can easily be visualized through the cuticle as black dots [27]. We used this method to investigate the respective contribution of the three PPOs to the melanization reaction. Consistent with the notion that black dots after heating are caused by the presence of PPO2 in crystal cells, PPO1 Δ and PPO3 1 mutant larvae had melanized dots corresponding to the sessile crystal cells, but PPO2 Δ mutants did not (Fig. 7a). We then subjected third-instar larvae previously infested by L. boulardi to the heat treatment. In addition to the black dots corresponding to the spontaneous activation of PPO2 in crystal cells, wild-type larvae infested with L. boulardi show large black spots that were not observed in unchallenged larvae (Fig. 7b). The same treatment (infestation followed by heating) applied to the various PPO mutants suggests that the large black spots are due to PPO3 activity in lamellocytes, as they are totally absent in PPO3 1 larvae. Our conclusions were further strengthened by the observation that silencing PPO3 by RNAi in lamellocytes alone using the PPO3-Gal4 driver phenocopied the PPO3 1 phenotype. Furthermore, a knot mutant (also called collier), which cannot produce lamellocytes [28], shows a melanization pattern similar to the PPO3 1 mutants following wasp infestation.

Fig. 7
figure 7

Phenoloxidase activity in PPO mutant larvae upon wasp infestation as revealed by heat treatment. Representative photos showing unchallenged larvae (a) and larvae infested by Leptopilina boulardi parasitoid wasps after heat treatment (b). Black dots on the cuticle correspond to melanized crystal cells due to PPO2 activity (black arrows, absent in PPO2 mutants). Large black patches correspond to melanized lamellocytes due to PPO3 activity (red arrows, absent in PPO3 mutants). Knot Col-1 larvae have no lamellocytes [28]

Full size image

Melanization linked to encapsulation is not dependent on Hayan or Sp7

It has been proposed that PPO3 is produced in its active form, while PPO1 and PPO2 are synthesized as zymogens, which are cleaved by SP activity to generate the active form [18, 20]. This cleavage is thought to be mediated by a clip-domain SP named Hayan [22]. Hayan also exists as a zymogen that is itself stimulated through a stepwise process involving other SPs. One upstream clip-domain SP is SP7 (also called MP2), which is specifically expressed in crystal cells [12–14, 29]. Indeed, while a null mutation in Hayan totally abolishes hemolymphatic PO activity in adults [22], a partial Sp7 loss-of-function, Sp7 PAE1, slightly reduces it [14]. To date, the roles of Hayan and Sp7 in the encapsulation reaction in larvae have not been fully clarified.

To address the role of SP cleavage in larval and adult melanization, we first generated using CRIPR/Cas9 two null mutations in Hayan and Sp7 and analyzed their contribution to the melanization reaction using the same techniques as described above. We confirmed that both Sp7 SK6 and Hayan SK3 mutations affect melanization at an injury site, with slight differences according to the stage of the animal. As expected, the newly generated null Sp7 mutant, Sp7 SK6, induces a much stronger phenotype than the previously described hypomorphic SP7 PAE1 allele: Sp7 SK6 mutant adults show a very weak melanized spot at the injury site compared to SP7 PAE1 (Fig. 8a). As expected [22], the Hayan SK3 mutation totally abolishes melanization at the injury site in adult flies. Interestingly, injured Hayan SK3 and Sp7 SK6 larvae had opposite phenotypes compared to mutant adults. Whereas Sp7 Δ larvae do not develop any melanization at the injury site, Hayan SK3 larvae still display a very weak blackening reaction (Fig. 8b). However, both Sp7 SK6 and Hayan SK3 larvae showed a wild-type pattern of melanized black dots upon heating at 65 °C (Fig. 8c). This result indicates that heat treatment can induce spontaneous activation of PPO in crystal cells in the absence of upstream SPs. Finally, we observed that both Sp7 SK6 and Hayan SK3 mutant larvae can produce melanized capsules around L. boulardi eggs (Fig. 8d). This indicates that Hayan and Sp7 are not mandatory for the melanization around the capsule, although a minor contribution of these SPs cannot be excluded. In summary, the loss of Hayan or Sp7 both mimic the PPO1 Δ ,2 Δ phenotype, pointing to a major role of these SPs following septic injury. The formation of wild-type melanized capsules in both SP mutants is consistent with the observation that PPO3 does not need cleavage to become active.

Fig. 8
figure 8

Hayan and SP7 do not contribute to melanization of a wasp capsule. Melanization of adults (a) after clean injury is abolished only in the Hayan SK3 mutant. A slightly reduced melanization spot was observed in the Sp7 SK6 mutant. In contrast, after clean injury of larvae (b), melanization is abolished only in Sp7 SK6 whereas Hayan SK3 shows residual melanization. Arrows indicate the pricking site. Adults and larvae were wounded with a tungsten needle and blackening of the wound was recorded 1 h later in larvae and 16 h later in adults. After heat-treating larvae (c), black dots, corresponding to crystal cells, appeared in Hayan SK3 and Sp7 SK6 comparable to wild-type controls. Infested larvae containing eggs of Leptopilina boulardi parasitoid wasps (d) show the presence of a melanized egg in Hayan SK3 and Sp7 SK6. Arrows indicate the position of the melanized wasp egg. A representative picture is shown for each genotype

Full size image

PPO3 arose by gene duplication from PPO2 during radiation of the D. melanogaster group

To place the functional differences among Drosophila PPO genes in an evolutionary context, we reconstructed the phylogeny of the PPO gene family using genome sequences from 23 species of Drosophila [30, 31]. We confirmed previous observations by Salazar-Jaramillo et al. [26] that the PPO1 and PPO2 genes are conserved across all Drosophila genomes sequenced, whereas PPO3 is restricted to the D. melanogaster group (Additional file 2: Figure S2). By analyzing a larger panel of species in the D. melanogaster group than in previous work [26], we found that a canonical D. melanogaster-like PPO3 gene can only be found in the D. melanogaster subgroup (containing D. melanogaster, D. simulans, D. sechellia, D. erecta and D. yakuba) and D. eugracilis, but not in other members of the D. melanogaster group or species outside this group. Evidence for a partial PPO3 gene can be found in D. takahashii (Additional file 2: Figure S2), which may indicate that a PPO3-like gene was present in the ancestor of these species. All Drosophila species with canonical (or partial) PPO3 genes are able to encapsulate parasitoid wasps melanotically and to produce lamellocytes [26, 32, 33]. However, additional species in the D. melanogaster group (i.e., D. ficusphila) are also able to cellularly encapsulate wasps and produce lamellocytes [34], suggesting that the acquisition or retention of PPO3-based melanization in lamellocytes occurred after the cellular basis for encapsulating parasitoid wasps evolved.

Phylogenetic evidence for all PPO genes in Drosophila shows the PPO3 clade clusters more closely with PPO2 than with PPO1 (Additional file 3: Figure S3; see also [8]), consistent with the functional overlap between PPO2 and PPO3 reported here. Moreover, both PPO3 and PPO2 have a four-exon gene structure, whereas PPO1 has six exons. The PPO gene tree also shows that all PPO3 genes cluster together outside the PPO2 clade, which is consistent with two hypotheses for the timing of the PPO3 duplication event: (i) the PPO3 gene arose prior to the diversification of the Drosophila and Sophophora subgenera and was subsequently lost in multiple lineages, or (ii) the PPO3 gene arose in the D. melanogaster group and underwent a transient period of accelerated sequence evolution shortly after duplication. Assuming the species tree in Ometto et al. [35], an ancient origin is unlikely because it requires one gain and at least seven independent losses in different Drosophila lineages, while simultaneously invoking selective maintenance of an ancestral PPO3 prior to the evolution of the lamellocyte cell type on the lineage leading to the D. melanogaster group. A recent origin is more parsimonious and only requires a single evolutionary event followed by an accelerated rate of evolution in one of the genes after duplication (in this case PPO3), a pattern that has been observed for duplicate genes in yeast and mammals [36]. Thus, we conclude a PPO2-like ancestral gene was likely the source for the PPO3 gene duplication event, and that this event most likely occurred recently during the radiation of the D. melanogaster group species.

The maximal genomic extent of the PPO3 duplication (inferred by alignment with D. ananassae, the closest outgroup species that lacks this locus and has a high-quality genome assembly) does not fully contain the sequence in our PPO3-Gal4 reporter construct that is needed to give lamellocyte expression (Additional file 2: Figure S2). In fact, the region of our reporter that extends the regulatory region reported by Ferjoux et al. [16] and likely contains lamellocyte-specific regulatory elements lies within the neighboring gene CG9890, which is conserved in D. ananassae. Thus, it is likely that regulatory sequences responsible for lamellocyte expression were not a part of the ancestral locus that gave rise to PPO3, but rather arose during or after the PPO3 duplication event.

After the gain of PPO3, D. sechellia subsequently lost the ability to produce lamellocytes and melanotically encapsulate parasitoids [26, 32, 33]. Salazar-Jaramillo et al. [26] found an accelerated rate of evolution in the D. sechellia PPO3 gene (Dsec\GM15980), which they proposed reflected relaxation of selective constraints on PPO3 associated with the loss of the melanotic encapsulation phenotype. We found no evidence for an increased rate of evolution of the D. sechellia PPO3 gene in our data, and tracked this discrepancy to a gene model error in FlyBase that fuses incomplete versions of PPO3 and a neighboring gene (CG44252) in D. sechellia that was used by Salazar-Jaramillo et al. [26]. We did, however, find evidence for an inactivating mutation in the D. sechellia PPO3 gene at amino acid position 48, which converts the terminal glutamine residue of the propeptide region to a stop codon, and is predicted to generate a truncated version of the PPO3 protein. This loss-of-function coding-sequence mutation together with loss of PPO3 expression in D. sechellia [26] supports the general model that PPO3 functions specifically in lamellocytes, and that the proper cellular context for its function in melanotic encapsulation is required to maintain selective constraint on this locus.

Discussion

Gene duplication is recognized as an important process in evolution. About 40 % of the 13,601 Drosophila genes are duplicates of other genes [37]. Duplication is often associated with sub-functionalization in which each of the daughter genes adopts part of the function of the parental gene or neo-functionalization in which duplication is at the origin of a novel function [37]. Gene duplication can lead to the formation of a large gene family. Both animals and plants harbor large families of genes devoted to immune defense. Immune genes encoding effector or recognition molecules often exist in multiple copies, while genes encoding signaling pathway components are typically single copy [38, 39]. It is expected that a large repertoire of immune molecules confer broader recognition and effector capacity as well as more regulatory flexibility in the use of this repertoire. Thus, the higher number of genes encoding peptidoglycan recognition proteins or antimicrobial peptides in the fruit fly Drosophila compared to the pea aphid can easily be understood by their respective environments: aphids feed on sap, a rather sterile diet, while Drosophila feed on rotting fruits swarming with microbes [40]. Besides these general statements, the raison d’être of the size of an immune gene family remains speculative.

Phenoloxidases are rate-limiting enzymes, which determine the extent of melanization, and as such can be considered as effector immune molecules [8]. The number of PPO genes is variable among insects, ranging from one in the bee Apis mellifera to ten in the mosquito Aedes aegypti [8]. In this work and in Binggeli et al. [2], we have generated loss-of-function mutations in the three PPO genes of Drosophila and analyzed the contribution of each of them to immune reactions. Our gene deletion and double-mutant analysis allowed us to ascribe specific functions to each of the three PPOs. Our study confirms that two crystal cell PPOs, PPO1 and PPO2, contribute to the bulk of melanization induced by injury, with PPO1 immediately available and PPO2 being deployed later [2]. Melanization is one of the most rapid immune responses, as the dark spot is visible as soon as 10 min after injury. This suggests a key role for this mechanism in the early steps of wound healing and pathogen control. This could explain why PPOs are rather stored as ready-to-use proteins and not regulated at the transcriptional level. The existence of crystal cells, whose sole reported function so far is melanization, provides Drosophila with an efficient way to store and quickly release this key enzyme. Recent studies have shown that crystal cells either derive from progenitor blood cells or can differentiate from plasmatocytes [15, 27, 41, 42]. Thus, Drosophila can indirectly modulate the amount of PPO1 and PPO2 by regulating crystal cell differentiation. The observation that the PPO1 deletion does not affect crystals in crystal cells led us to propose that PPO2 is the main source of crystalline PPO and that PPO1 is either localized in the cytoplasm of crystal cells or continuously secreted into the hemolymph [2]. The extreme fragility of crystal cells did not allow us to clarify this point.

Our work here demonstrates that PPO3 is restricted to lamellocytes and contributes to the encapsulation of a wasp egg. This explains why the PPO3 transcripts are present at barely detectable levels in unchallenged larvae, which contain few or no lamellocytes. An intriguing observation is that two PPOs, PPO2 and PPO3, contribute to the melanization of the capsule formed around a wasp egg. This means that it is achieved by phenoloxidases derived from two hemocyte types: crystal cells and lamellocytes. Lamellocytes are large flat cells with adhesive properties that can bind to wasp eggs and form multilayers. Our study suggests that crystal cells could also be guided to the capsule where they would release PPO2 crystals close to the encapsulated egg. It is still unknown whether crystal cells have the ability to stick to non-self elements or if they need the assistance of other hemocytes to reach their target as suggested by [27]. The absence of any marked role of PPO1 in encapsulation is puzzling. We speculate that localization of PPO1 in the hemolymph, which remains to be assessed, could explain why PPO1 is not involved in encapsulation. PPO in circulation would not have the ability to be directed to a foreign element such as a wasp egg, but could be immediately activated by the presence of oxygen at a wound site.

Combining the various mutations allowed us to show that, while each single PPO mutant has a specific phenotype, knocking out two genes out of three is required to abolish fully a specific function. Thus, Drosophila PPOs have partially overlapping functions to optimize melanization in at least two stress conditions, injury and parasitization (Fig. 9). Future studies should investigate the role of these PPOs in the melanization of organs such as the gut, salivary gland and trachea, which is sometimes observed upon oral infection or in a tumorous-like state [43, 44]. The selective pressure exerted by parasitoid wasps on D. melanogaster has probably led to the development of dedicated cellular immune defenses, with the emergence of inducible adhesive cells with PPO3 (lamellocytes), storage cells (crystals cells with PPO2) and hematopoietic organs (the lymph gland and sessile islets). All these elements are activated upon parasitization and contribute collectively to capsule formation by mechanisms that are still poorly understood [45].

Fig. 9
figure 9

Model of phenoloxidase activation after injury and wasp infestation. Both PPO1 and 2 are synthetized by crystal cells. In the unchallenged condition, PPO2 is stored in the crystal cells while the localization of PPO1 is still unknown (here shown in the hemolymph) (middle). After a clean or septic injury, PPO1 and PPO2 are the only source of hemolymphatic PO activity (left). After parasitic infestation, PPO3 is produced by lamellocytes and contributes with PPO2 to the melanization around the wasp egg during encapsulation (right). UC unchallenged

Full size image

Most Drosophila species possess two PPO genes, PPO1 and PPO2, except for some species from the D. melanogaster group (D. melanogaster, D. simulans, D. sechellia, D. erecta, D. yakuba and D. eugracilis), which have three (Additional file 2: Figure S2). Since the PPO3 gene sequence, structure and function are more similar to those of PPO2 than of PPO1, we propose that a duplication of an ancestral PPO2-like gene gave rise to an ancestral PPO3-like gene in the common ancestor of the lineage leading to D. eugracilis and the D. melanogaster subgroup. This was followed by a functional diversification with PPO2 maintaining expression in crystal cells and PPO3 becoming expressed only in lamellocytes. This scenario is consistent with a period of accelerated evolution of the PPO3 lineage after duplication (Additional file 3: Figure S3). Future studies in other Drosophila species might shed light on whether this gene duplication event led to sub-functionalization (the ancestral gene was expressed in both crystal cell and lamellocyte lineages) or neo-functionalization (the ancestral gene was only expressed in crystal cells and the duplicate gene acquired a new expression pattern in lamellocytes). The observation that the existence of a third PPO is restricted to the D. melanogaster group suggests that massive production of PO emerged as a crucial mechanism to reinforce the defense against parasitoids in this lineage. The same selective pressure has probably led to other anti-parasite defenses. Recent studies have shown that wasp encapsulation in D. ananassae (a species outside the D. melanogaster subgroup) is mediated by another mechanism involving multinucleated giant hemocytes, that are formed by the aggregation of hemocytes [46]. The capsule around the wasp is not melanized in this species [26, 33, 46]. The melanization of the capsule points to a critical role of melanization against parasites. This is also supported by studies showing that virulent wasps have suppressive mechanism against melanization, such as the injection of serpins [47]. We performed survival analysis with PPO mutants using a highly virulent wasp species, L. boulardi, and a weakly virulent one, L. clavipes. The results did not reveal any striking effect although there was a trend for a lower resistance of larvae to L. clavipes infestation in the absence of PPO (Additional file 4: Figure S4). Consistent with this, we noticed that it is easier to cultivate the wasp Asobara tabida using PPO1,2,3 flies compared to wild type. Taking possible effects of the genetic backgrounds into consideration, the results of our survival analysis in the laboratory should be taken with caution. It is possible that the optimum level of melanization for successful encapsulation changes in an ongoing arms race requiring a very tight balance, as melanization is also toxic for flies. This would explain why PPO3 mutants tend to be more susceptible than PPO1,2,3 deficient flies upon infection with L. clavipes (Additional file 4: Figure S4). Thus, the relevance of PPO in wasp encapsulation requires further analysis that should take into account the influence of the genetic background and the use of a more natural setting of infection.

The mechanisms of activation of PPO in Drosophila are not yet fully understood. Previous studies suggested that PPO1 and PPO2 require cleavage by SPs while PPO3 would be constitutively active. Indeed over-expression of PPO3 is sufficient to induce melanization of tissues as diverse as the eye or the salivary gland [18]. Here, we show that mutations in Hayan or SP7 strongly reduce melanization due to septic injury but do not affect melanization around a wasp capsule. This suggests that either PPO3 does not require proteolytic cleavage, as suggested by molecular modeling and in vitro testing of PPO3 mutants [20], or its maturation requires distinct SPs. Interestingly in this context, heat treatment of wasp-infected larvae produced melanized cell aggregates, which corresponds to PPO3 activity in lamellocytes. This indicates that PPO3 is present in lamellocytes in an inactive form that can be activated by heat. Thus, the possibilities are that PPO3 needs an additional step to be activated, it is produced together with an inhibitor, or it is inactive due to the absence of its substrate. The mechanism by which PPO3 is activated only in lamellocytes around a capsule remains to be deciphered.

This analysis started with the serendipitous discovery of a PPO3 deletion in the PPO1,PPO2 double-mutant fly stock. Since this deletion was also found in a white control stock, we assume that this mutation was introduced when backcrossing PPO1. Since fly stocks in the laboratory are cultivated in the absence of parasitoid pressure, it cannot be excluded that this gene quickly pseudogenizes as it has no role beyond encapsulation. We were fortunate that the presence of this mutation does not affect the main conclusion of Binggeli et al. [2], which states that PPO1 and PPO2 are responsible for all hemolymphatic PO activity. The absence of melanization around the wasp egg in PPO1 Δ ,2 Δ ,(3 1 ) larvae led us to miss the important contribution of PPO3 to encapsulation. This underlines that one of the main issues in generating an extensive characterization of the Drosophila immune system is that of the genetic background and namely what we refer to as ‘wild type’. Fly geneticists have usually a number of ways to assess the effect of the genetic background by analyzing the phenotype caused by mutations in different contexts. This can be done by placing the mutation over a deficiency or by extensive backcrossing to generate isogenic stocks as recently described in [48]. An important task should be to analyze how the studied phenotype varies in diverse backgrounds. In this study, the phenotypes of each of the PPO single mutations were clear enough and in accordance with their expression patterns, so that we could be confident about their respective functions. The striking phenotypes of PPO1 Δ ,2 Δ and PPO2 Δ ,3 1 double mutants, which fail to melanize upon injury and wasp infection, respectively, reinforced our conclusions, which would have been more difficult to draw from only single-mutant analysis. Thus, generation of multiple mutations in combination is an adequate approach to assess better the function of gene families, notably those involved in effector mechanisms. Similar studies on other immune gene families should provide insights into the organization of the Drosophila immune system and that of other organisms.

Conclusions

We conclude that differences in spatial localization, immediate or late availability, and mode of activation underlie the functional diversification of the three Drosophila PPOs, with each of them having non-redundant but overlapping functions.

Methods

Insects stocks and mutant generation

Unless indicated otherwise, w 1118 or y 1 w 1118 flies were used as wild-type controls. The PPO1 Δ, PPO2 Δ, PPO1 Δ ,2 Δ ,(PPO3 1 ), Relish E20 (Rel E20), spätzle rm7 (spz rm7) and SP7 PAE1 lines were described previously [2, 10, 14, 49]. kn col-1 lamellocyte deficient larvae were obtained by using the w*; kn col-1 ;P{col5-cDNA.C} fly line that carries the lethal mutation knot and a knot transgene with a restricted expression pattern rescuing this lethality [28]. w*; P{10XUAS-mCD8::GFP}attP2 (UAS-GFP) lines was obtained from the Bloomington Stock Center. The parasitoid wasp L. boulardi (kindly provided by M Crozatier) was reared on PPO1 Δ ,2 Δ ,3 1 triple-mutant fly stocks at room temperature. After emergence, wasps were kept at room temperature and provided with honey until used for experiments. The PPO3 SK3, Hayan SK3 and Sp7 SK6, mutant lines were generated by CRISPR/Cas9 as described in [21] (Additional file 5: Figure S5). The uas-PPO3-IR (ch II) was obtained from VDRC (50737). The PPO3-Gal4 line was generated by cloning a 1.6 kb sequence upstream of the PPO3 gene (PPO3-Gal4 forward 5′-TTGAGGGCGGTGAAGTTGTC-3′, reverse 5′- GGAGGACCTTTAGCGAGCAG-3′) into pBPGUw vector [50] followed by PhiC31 integrase-mediated transgenesis (BestGene Inc). The PPO1 Δ ,2 Δ, PPO1 Δ ,3 1 and PPO2 Δ ,3 1 lines were generated by reintroducing the corresponding wild-type alleles into the PPO1 Δ ,2 Δ ,3 1 mutant by meiotic recombination. Drosophila stocks were maintained at 25 °C on standard fly medium.

Microorganism culture and infection experiments

The bacterial strains used and their respective optical density of the pellet (OD) at 595 nm were: the Gram-negative bacteria Erwinia carotovora (Ecc15, OD 200); the DAP-type peptidoglycan (PGN) containing Gram-positive bacteria Bacillus subtilis (OD 5); and the Lys-type PGN containing Gram-positive bacteria Micrococcus luteus (M. luteus, OD 200), Staphylococcus aureus (S. aureus, OD 0.5) and Enterococcus faecalis (OD 0.5). Strains were cultured in Luria broth at 29 °C (Erwinia carotovora and M. luteus) or 37 °C (other species). Spores of entomopathogenic strains Beauveria bassiana 802 were grown on malt agar plates at 29 °C for approximately 3 weeks until sporulation [49]. Systemic infections (septic injury) were performed by pricking adult females in the thorax with a thin needle previously dipped into a concentrated pellet of bacteria. Natural infections were initiated by shaking anesthetized flies in a Petri dish containing a sporulating culture of entomopathogenic fungi Beauveria bassiana. Infected flies were subsequently maintained at 29 °C (Erwinia carotovora, M. luteus and Beauveria bassiana) or at 25 °C (all other bacteria). At least three tubes of 20 flies were used for each survival experiment and survival was scored daily. For lifespan experiments, flies were kept on normal fly medium and were flipped every 2 days.

Wounding experiment

A ‘clean’ injury was an injury performed with a sterilized needle. A low level of bacterial contamination is still possible since the surface of the insect was not sterilized. For imaging of the melanization reaction upon pricking, the thorax of the animal was pricked (as described in infection experiments) using a sterile needle (diameter ~5 μm). Pictures were taken 16 h post-pricking. Third-instar larvae were pricked dorsally near the posterior end, using a sterile needle (diameter ~5 μm). Pictures of melanized larvae were taken 1 h post-injury. Pictures were captured with a Leica DFC300FX camera and Leica Application Suite. For publication purposes, brightness and contrast were increased on some images.

Live imaging and immunofluorescence

For hemocyte imaging, larvae were cleaned in phosphate-buffered saline (PBS) and dissected on silane-coated microscope slides (Tekdon, Inc) into a drop of 1 % bovine serum albumin-PBS. Hemocytes were allowed to settle for 30 min, then fixed with 4 % paraformaldehyde-PBS for 15 min and rinsed with PBS.

For staining of capsules from wasp-infested larvae, capsules were dissected 3 days after infection, fixed with 4 % paraformaldehyde-PBS and rinsed with PBS. Samples were observed for fluorescence with an Axioplot imager Z1 and Axiocam mRM camera (Zeiss).

PO activity

Hemolymph was collected by dissecting larvae in 4 °C PBS. The protein concentration was adjusted after a Bradford test. Sample volumes were adjusted in 20 μl of 5 mM CaCl2 solution. After addition of 80 μl L-DOPA solution (20 mM, pH 6.6), the samples were incubated at 29 °C in the dark. The OD at 492 nm was then measured. An L-DOPA solution without hemolymph was used as a blank. Each experiment was repeated three times.

Quantitative reverse-transcription PCR

For quantification of mRNA, whole flies or larvae were collected at indicated time points. Total RNA was isolated from 10–15 adult flies or larvae using TRIzol reagent and dissolved in RNase-free water. Next, 1 μg of total RNA was reverse-transcribed in 10 μl reaction volume using PrimeScript RT (TAKARA) with random hexamer and oligo dT primers. Quantitative PCR was performed on a LightCycler 480 (Roche) in 96-well plates using the LightCycler 480 SYBR Green I master mix or on a LightCycler 2.0 (Roche) in capillaries using dsDNA dye SYBR Green I (Roche). Primers were as follows: Diptericin forward 5′- GCTGCGCAATCGCTTCTACT-3′, reverse 5′-TGGTGGAGTGGGCTTCATG-3′; Drosomycin forward 5′-CGTGAGAACCTTTTCCAATATGAT-3′, reverse 5′- TCCCAGGACCACCAGCAT-3′; Rp49 forward 5′-GACGCTTCAAGGGACAGTATCTG-3′, reverse 5′-AAACGCGGTTCTGCATGAG-3′; and PPO3 forward 5′-GGCGAGCTGTTCTACT-3′, reverse 5′- GAGGATACGCCCTACTG-3′.

Wasp infestation and quantification of fly survival to wasp infestation

For wasp infections, 30 synchronized second-instar wild-type or mutant larvae were placed on a pea-sized mound of fly food within a custom-built wasp trap in the presence of three female wasps for 2 h (L. boulardi) or 72 h (L. clavipes). For survival experiments, parasitized larvae were kept at room temperature and scored daily for flies and wasps. The difference between enclosed flies and wasps to the initial number of larvae was set as dead larvae/pupae. For imaging of lamellocytes or wasp eggs, infested larvae were dissected 72 h after being exposed to wasps. For qRT-PCR of PPO3, the total RNA of larvae was isolated 52 h after wasp infection.

Heating of larvae to induce spontaneous melanization

Larvae were heated in PBS at 67 °C for 20 min to induce the spontaneous activation of PPO within hemocytes [9, 27].

PPO sequence analysis

Predicted orthologs of PPO genes from 23 species in the Drosophila genus were identified by extracting and concatenating exonic regions from the UCSC Genome Bioinformatics Database multiz27way insect alignments [51]. Genomic alignments extracted from UCSC for each gene were cross-referenced with orthology predictions for 12 Drosophila genomes from OrthologDB [52] and in three cases where UCSC genomic alignments lacked a gene model for a particular species (PPO1 from D. mojavensis, PPO1 from D. grimshawi and PPO2 from D. ananassae), the FlyBase gene model indicated by OrthologDB was added to the alignment [30, 53].

Sequences for all three PPO genes were combined and aligned using Clustal Omega [54] using default settings in SeaView 4.0 [55]. A preliminary phylogenetic tree was produced using the BioNJ algorithm [56] in SeaView. PPO sequences clustering in the incorrect gene or species clade were inspected in the UCSC Genome Browser. Three predicted PPO genes based on the UCSC alignments (PPO1 for D. miranda, PPO1 for D. kikkawai and PPO3 for D. kikkawai) clustered incorrectly in the PPO2 clade. Inspection of global alignment nets and chains in the UCSC browser revealed these to be alignment artifacts. Likewise, systematic inspection of nets and chains for all D. melanogaster group species revealed that a predicted PPO3 sequence from D. takahashii based on UCSC whole-genome alignment was composed of two non-syntenic fragments, and there was no evidence for a PPO3 sequence in the two species most closely related to D. takahashii (D. biarmipes and D. suzukii). Thus, these four sequences were removed from the final alignment before producing the final phylogenetic tree using RAxML version 8.0.26 [57]. Maximum-likelihood tree searches were conducted using a general time-reversible model of nucleotide substitution with Γ rate heterogeneity, with all model parameters estimated by RAxML. Trees were inferred using a combined approach, with an initial 100 bootstrap replicates and a full maximum-likelihood search for the best-scoring tree, using the rapid bootstrap algorithm [58]. The best-scoring maximum-likelihood tree was visualized and annotated in FigTree version 1.4.2 [59].

The PPO3 gene overview was adapted with FancyGene [60]. The protein structure of PPO3 wild type and PPO31 was predicted by Phyre2 [61] and visualized using the VMD suite [62]. All structural illustrations were rendered in VMD.

Statistical analysis

Each experiment was repeated independently a minimum of three times (unless otherwise indicated). Error bars represent the standard deviation of replicate experiments (unless otherwise indicated). Statistical significance of survival data was calculated with a log-rank test, and P values are indicated in figure legends. Statistical significance of PO activity was calculated with two-way ANOVA with Tukey correction. Otherwise, statistical significance was calculated using the Mann–Whitney test and P values of <0.05 (indicated by *), <0.005 (indicated by **) and <0.0005 (indicated by ***) were considered significant.