Pore-Forming Toxin-Like Proteins in the Anti-Parasitoid Immune Response of Drosophila Open Access

Parasitoid wasps exert strong selective pressure on their host populations [1, 2]. In response, insect hosts have evolved myriad strategies to overcome parasitoid wasp attack. Drosophila species in the ananassae subgroup are highly resistant to parasitoid wasps compared to a diverse panel of drosophilids [3‒5]. Contrary to the melanogaster subgroup species and other related (“oriental”) subgroups, which use lamellocytes to encapsulate the invaders [6‒8], species in the ananassae subgroup evolved a novel immune cell type, the multinucleated giant hemocyte (MGH) [4]. In addition, whereas in Drosophila melanogaster the melanization of the capsule defines the anti-parasitoid defense response [9], melanization never occurs in the species of the ananassae subgroup, and the gene encoding for prophenoloxidase 3 (PPO3), a key enzyme required for capsule melanization, is absent in their genomes [8]. These observations imply that a distinct and highly efficient anti-parasitoid immune defense mechanism must operate in the ananassae subgroup.

Horizontal transfer of toxin-encoding genes from bacteria or bacteriophages to the genomes of several insect species, including those of the ananassae subgroup, has been previously reported [10‒13]. These genes of prokaryotic origin encode homologs of two bacterial toxins, cytolethal distending toxin B (cdtB) and apoptosis inducing protein of 56 kDa (aip56), which are both absent in the genomes of the melanogaster subgroup species. In prokaryotes, CdtB is the active subunit of a tripartite holotoxin, which causes cellular distension, cell cycle arrest, and cell death in eukaryotic cells via its nuclease activity [14‒17]. AIP56 was first isolated from the mariculture pathogen Photobacterium damselae subsp. piscicida. As a prokaryotic AB toxin, it exhibits proteolytic activity by cleaving NF-kB p65, resulting in apoptosis by interfering with the regulation of anti-apoptotic genes [18, 19]. Recently, we reported that the horizontally transferred cdtB and cdtB::aip56 fusion genes of D. ananassae are integrated into the innate immune system and act as novel humoral factors, essential modules in the highly efficient anti-parasitoid defense [12].

Single cell transcriptomic analysis of D. ananassae hemocytes [20] revealed 14 hemolysin E-like genes expressed in MGHs, cells that differentiate exclusively after parasitoid wasp infection [4]. Additional analysis led to the discovery of 23 additional hemolysin E-like genes in the genome of these species. Hemolysin E proteins (HlyE, also known as cytolysin A or ClyA) are pore-forming toxins in the Enterobacteriaceae family and can lyse a variety of mammalian cells [21, 22]. They are secreted by bacteria through outer membrane vesicles, where they oligomerize to form active pore assemblies. A HlyE monomer is composed of five helical regions and a short β-hairpin called the β-tongue, which has a primary role in the insertion into the host membrane [23, 24]. Oligomerization of these proteins occurs in a specialized environment that is dependent upon an altered redox status in the outer membrane vesicles [25]. The hydrophobic β-tongue transforms into a helix-turn-helix that first embeds into the lipid bilayer [26]. This step is followed by massive structural rearrangements in the molecule, that lead to stable pore assembly and drive the loss of cytoplasm and consequent death of the target cell.

Here we used phylogenetic approaches, bioinformatical and expression analyses to reveal the origin and characteristics of the bacterial toxin-like molecules in eukaryotes using D. ananassae as a model organism. Our findings suggest that the Hemolysin E-like proteins may represent a novel family of humoral factors, acquired by horizontal gene transfer, that are important elements of the anti-parasitoid defense reaction in innate immunity of several insect species.

D. ananassae wild type (14024-0371.13) was obtained from UC San Diego Drosophila stock center. Leptopilina boulardi (G486) parasitoid wasps were kindly provided by Prof. Todd Schlenke (University of Arizona, Tucson, AZ, USA) and were maintained on D. melanogaster wild type Oregon-R, obtained from the Bloomington Drosophila Stock Center. Each strain was kept at 25°C on standard yeast-cornmeal food.

Seventy second instar D. ananassae larvae were exposed to 17 female L. boulardi G486 parasitoids for 6 h. One day after the wasp attack, parasitized larvae with a melanotic spot on the cuticle, the site of oviposition, were placed in vials with standard yeast-cornmeal food until the required developmental stage for sampling.

RNA was prepared with RNeasy mini kit (Qiagen) according to manufacturer’s instruction. One microgram RNA was used to generate 25 µL cDNA with the RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific) and the Oligo(dT)18 Primer. For Quantitative RT-PCR, 2 µL of 10 times diluted cDNA was used with PerfeCTa SYBER Green SuperMix (Quanta bio) in a Rotor-Gene Q (Qiagen) qPCR platform. Oligonucleotide primers used in the work for qRT-PCR are listed in online supplementary Table S1 (for all online suppl. material, see https://doi.org/10.1159/000542583). Reaction conditions were the following: 95°C for 2 min, 45 cycles at 95°C for 10 s, 57°C for 45 s, and 72°C for 15 s. Two independent experiments were carried out with duplicates, except for the tissue-specific expression analysis, where we had triplicates. A Rotor-Gene Q Series Software and Q-Rex 1.0 were used for data analysis. To interpret gene expression levels, the ∆∆Ct was calculated. Obtained cycle threshold (ct) values were related to those of the housekeeping gene GF23239 (LOC6505882), which encodes a ribosomal protein, the homolog of rp49 of D. melanogaster.

The significance of differences was determined either by the Student’s t test, or one-way ANOVA, and Tukey’s honestly significant difference (HSD) tests of the delta Ct values. The performed tests are indicated at each figure legend.

Coding sequences of hemolysin 6 (hL6) (Dana\GF22667) or hemolysin 16 (hl16) (excluding the predicted N-terminal signal peptide) (Dana\GF21479) from D. ananassae, were PCR amplified using Q5 High-fidelity DNA polymerase (New England Biolabs, Ipswich, MA, USA). The amplification was performed from D. ananassae cDNA library using gene-specific oligonucleotide primers listed in online suppl. Table S2. Hemolysin E (HlyE) (Q68S90_ECOLX) was PCR amplified from genomic DNA of the wild type Escherichia coli (SzMC 0582) (Szeged Microbial Collection, University of Szeged, Hungary) using gene-specific oligonucleotide primers listed in online suppl. Table S2. PCR products were digested with restriction enzymes (hl6: EcoRI-SalI; hl16: BamHI-HindIII; hlyE: BamHI-HindIII) and cloned into the pETDuet-1 (Merck Millipore, Burlington, MA, USA) plasmid for bacterial expression.

For the expression of C-terminal 3 × Flag-tagged HL6, HL16 (including the predicted N-terminal signal peptide), or HlyE proteins in D.Mel-2 insect cells (ATCC), the CDS were PCR amplified using gene-specific oligonucleotide primers listed in online suppl. Table S2. The PCR products were inserted into the pDONR221 Gateway entry vector (Thermo Fisher Scientific, Waltham, MA, USA) by BP reaction and subsequently cloned into the pMT-CoHygro-DEST-3 × Flag destination expression plasmid (made in house, see the map and sequence in online online suppl. Dataset S1). This plasmid provides a copper-inducible promoter (pMT) and a C-terminal 3 × Flag tag. All constructs were validated by DNA sequencing.

Recombinant HL6, HL16, and HlyE proteins were expressed in SixPack E. coli strain [27] as follows: cells were cultured in 500 mL standard Luria-Bertani broth supplemented with 100 µg/mL carbenicillin. Protein expression was induced when the cell density reached 0.6 at OD 600 nm, using 0.5 mm IPTG for 20 h at 16°C. Proteins were purified from inclusion bodies using a single freezing-thawing method [28]. In brief, cells were harvested and lysed by sonication in phosphate-buffered saline (pH = 8.0) followed by high-speed centrifugation at 4°C, 30 min, 21,000 g. Inclusion bodies were completely resuspended (washing step) in 20 mL buffer containing 20 mm Tris (pH = 8.0), 300 mm NaCl, 1 mm EDTA, 1% Triton X-100, and 1M urea, and centrifuged at 4°C, 20 min, 12,000 g. Washing was repeated twice. Finally, inclusion bodies were resuspended in PBS supplemented with 2M urea and stored at −20°C for 24 h. The frozen samples were thawed slowly at room temperature and centrifuged at 4°C, 20 min, 12,000 g. Supernatants containing the resolubilized recombinant proteins were collected and dialyzed over PBS supplemented with 0.65M urea at 4°C for 24 h, followed by a second dialysis in PBS at 4°C for 16 h. After centrifugation at 4°C, 10 min, 5,000 g, supernatants were collected and filter-sterilized, and proteins were concentrated using an Amicon Ultra centrifugal device with 10 kDa MWCO (Merck Millipore, Burlington, MA, USA) at 4°C, 60 min, 4,000 g. Proteins were collected, flash-frozen in liquid nitrogen and stored at −80°C before use for immunization.

3 × Flag-tagged HL6, HL16 and HlyE proteins were expressed in stably transfected Schneider’s Drosophila Line 2 (D. Mel-2, ATCC CRL-1963, Manassas, VA, USA) cell lines. Cells were maintained in Insectagro DS2 Serum-Free medium (Corning, Corning, NY, USA) supplemented with 2 mm stable l-glutamine (Biosera, Nuaille, France) and 1× PenStrep (Biosera, Nuaille, France) at 25°C. Cells were transfected with the appropriate plasmid DNA using ExpiFectamine Sf transfection reagent (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer, and selected in the presence of 300 µg/mL hygromycin B (Serva, Heidelberg, Germany). Stably transfected cell lines were grown in T175 tissue culture flasks and the expression of 3 × Flag-tagged HL6, HL16, or HlyE were induced by CuSO₄ treatment (0.5 mm final concentration) for 24 h before harvesting. Cells were lysed in EB buffer containing 50 mm HEPES pH = 7.6, 150 mm NaCl, 0.5 mm EGTA, 2 mm MgCl₂, 0.1% NP40, 5% glycerol, 0.5 mm DTT, 1 mm PMSF, 1× EDTA-free protease inhibitor cocktail (Roche, Basel, Switzerland), 0.1 µL/mL benzonase nuclease (Merck Millipore, Burlington, MA, USA) using Ultra-Turrax T8 homogenizer (IKA-Werke, Staufen im Breisgau, Germany). Cell lysates were clarified by centrifugation (4°C, 20 min, 12,000 g), and supernatants were subjected to immunoprecipitation using anti-FLAG-M₂ magnetic beads (Sigma Aldrich, St. Louis, MO, USA) for 2 h at 4°C. HL16 harbors an endogenous signal peptide; therefore, the cell culture supernatant containing the secreted protein was also collected and used for protein purification. After binding, matrices were washed with EB, followed by extensive washing with Tris-buffered saline (TBS: 10 mm Tris-HCl pH = 8.0, 150 mm NaCl). Bound proteins were eluted in TBS supplemented with 200 ng/µL 3 × Flag peptide (Sigma-Aldrich, St. Louis, MO, USA). Eluted proteins were concentrated, flash-frozen and stored at −80°C before use.

BALB/C mice were injected subcutaneously with 1 μg purified recombinant Hemolysin E-like proteins produced in E. coli in Complete Freund Adjuvant (DIFCO). Immunization was repeated twice at 3-week intervals with proteins in Incomplete Freund Adjuvant (DIFCO). Antisera were prepared and screening was done by standard enzyme-linked immunosorbent assay (ELISA). Wells of ELISA plates (Corning) were coated with 100 µLs of 100 ng/mL recombinant proteins overnight at 4°C. Sheep anti-mouse immunoglobulin G (IgG), Horseradish peroxidase-linked whole antibody (GE Healthcare, UK) (1:10,000), and o-Phenylenediamine (Sigma-Aldrich) were used for detection. The polyclonal sera were further characterized by WB analysis on crude protein extracts of D. ananassae samples and the expected 42 kDa (HL6), 35 kDa (Hl16) bands were obtained.

Multiple sequence alignment was performed with the T-Coffee Web Server [29], and visualized with Jalview 2.11.3.2. For three-dimensional analysis of D. ananassae Hemolysin E proteins, either the AlphaFold database or the ColabFold v1.5.5: AlphaFold2 using MMseqs2 colab notebook (https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb) was used, which predicts protein structures via a machine learning approach solely based on the amino acid sequence [30]. We performed the analysis with the colab notebook with default parameters (template_mode: none; mas_mode: mmseqs2_uniref_env; pair_mode: unipaired_paired; model_type: auto; num_recycles: 3; recycle_early_stop_tolerance: auto; relax_max_iteration: 200; pairing_strategy: greedy; max_msa: auto; num_seeds: 1; use_dropout: off). Signal peptides were predicted with the SignalP 6.0 algorithm [31].

Crude protein extracts were prepared in sample buffer (63 mm Tris pH = 6.8, 9% glycerol, 0.18 mg/mL bromophenol blue, 2.3% SDS, 0.11 mm 2-mercaptoethanol), using a homogenizer, and centrifuged at 18,000 g for 5 min. Protein concentrations were determined by Amido Black assay. Extracts containing 100 µg protein were loaded per well and fractionated on 12% SDS PAGE, blotted to polyvinylidene difluoride (PVDF) membrane (Merk Millipore), blocked with 5% nonfat milk in TBS (10 mm Tris pH = 7.5, 150 mm NaCl), and incubated with primary antibodies (polyclonal serum in the following dilutions: 1:10,000 for anti-HL6 and 1:6,000 for anti-HL16) for 1 h. For loading control α-Tubulin was detected, using the anti α-Tubulin DM1α (Sigma, T6192) monoclonal antibody at 1:10,000 dilution. Membranes were washed 3 times (10 min each) with TBS containing 0.1% Tween 20, incubated with Polyclonal Goat Anti-Mouse Immunoglobulins/horseradish peroxidase (HRP) (Dako) (1:10,000 diluted in TBS containing 0.1% Tween 20 and 1% bovine serum albumin [BSA]), and washed 3 times. Reactions were visualized with Immobilon Western Chemiluminescent HRP Substrate (Merck Millipore).

D. melanogaster (Oregon-R) larvae were infected with L. boulardi. Fourty-eight hours following the infection parasitoids were isolated from the host larvae into Schneider’s medium (Lonza) complemented with 5% fetal bovine serum (Gibco) and 1 nm 1-phenyl 2-thiourea (Sigma) (CSM). Then 100 parasitoid wasp larvae in 500 µL CSM were incubated at room temperature for 1 h in 500 µL CSM with 100 µL (10 ng/µL) purified FLAG-tagged recombinant D. ananassae HL6 or HL16 proteins in TBS buffer. As control, wasp larvae were incubated with 100 µL Hemolysin free TBS buffer containing 200 ng/µL FLAG peptide. Larvae were washed 3 times (each for 5 min) in PBS and fixed with 2% paraformaldehyde for 10 min, then subjected to indirect immunofluorescence assay as described below.

Larvae were dissected in CSM. Fat bodies and parasitoid wasps were removed and fixed with 2% paraformaldehyde for 10 min, washed 3 times in PBS (5 min each), and blocked with 0.1% BSA in PBS supplemented with 0.1% Triton X-100. Hemocytes were adhered for 1 h on microscope slides, fixed with acetone for 6 min, air dried, and blocked with 0.1% BSA in PBS for 20 min. Samples were incubated with the primary antibodies (polyclonal serum in 1:500 dilution) for 2 h and normal mouse serum was used as control. For the binding experiments the primary antibody was a monoclonal mouse anti-FLAG M2 (Sigma) used at a dilution of 1:1,000. Samples were washed three times in PBS, incubated with the Alexa Fluor 488 goat anti-mouse IgG secondary antibody (Invitrogen, 1:1,000), containing DAPI (2.5 μg/mL) for 45 min, washed three times in PBS, mounted in Fluoromount G medium, and analyzed with an Olympus FV1000 confocal laser scanning microscope.

Second instar larvae were wounded on the dorsal part of the 5th segment using sterile minutien pins of 100 μm diameter. For septic injury, before wounding the minutien pins were dipped in a 1:1 mixture of Gram-positive (Bacillus subtilis, SzMC 0209) and Gram-negative (E. coli, SzMC 0582) bacteria (obtained from the Szeged Microbial Collection, University of Szeged, Hungary) or a 50% suspension of Beauveria bassiana entomopathogenic fungus spores (Kwizda Agro, Artis Pro) in sterile phosphate-buffered saline (PBS). Treated larvae were placed on vials with standard yeast-cornmeal food for 48 h, then RNA was isolated.

Hemolysin E homologs were found by conducting TBLASTN [32] searches using the amino acid sequence for the hemolysin E domain of all paralogs found in the D. ananassae genome as queries. A subsequent phmmer v 3.1b2 (http://hmmer.org/) search was run on the Refseq protein database (accessed on 19 August 2022) to find more divergent prokaryotic HlyE homologs. Hits from both of these searches were pared down to retain only hits with E-values <0.01. To eliminate redundant sequences from the alignment, CD-HIT version 4.8.1 [33] was used with a 97% similarity cutoff. Sequences were aligned using MUSCLE v 5.1 [34]. The alignment was visually inspected and remaining redundant, short, or poorly aligned sequences were removed. Additionally, hits from eukaryotic taxa that were on short contigs without any bona fide eukaryotic genes were removed from the alignment. The N-terminal signal peptide was manually trimmed and sequences were realigned with MUSCLE. This alignment was then trimmed using the ClipKIT smartgap algorithm [35]. Following all alignment quality control, there were 254 sites and 134 sequences, 17 of which were of prokaryotic origin. The best tree inference model was selected using IQTree [36] as implemented on the CIPRES server. Maximum likelihood gene trees were constructed using RAxML [37] using the JTT+gamma model as implemented on the CIPRES server [38]. One representative sequence of each bacterial taxon was selected as the outgroup in each tree estimation to determine whether outgroup choice affected topology. All bacterial sequences tested as outgroup resulted in similar topologies with equal log likelihoods. Nodes with <50% bootstrap support were collapsed using the di2multi package form ape v 5.6.2 [39]. Phylogenies were visualized using the ggtree package [40].

In a preliminary screen, we identified insect homologs in the refseq_genomes and wgs databases to a Drosophila willistoni Hemolysin E protein (which appears to evolve more slowly than those of D. ananassae), using the tblastn algorithm on the NCBI web server. We then repeated the screen, now adding two query sequences that were retrieved in the first screen: one from Drosophila rhopaloa and one from the nematoceran Exechia fusca and the threshold was set to E <0.15. At this stage, we limited the search to Drosophilidae. All overlapping hits on the same DNA strand, plus those that were less than 200 bp apart were combined and counted as a single hit. To estimate how many drosophilid sequences were available in total, we searched the same databases for homologs to the Rel homology domain in the Relish protein of D. melanogaster, a 297 amino acid region that we expected to be present in every species. This search retrieved 318 Drosophilidae species in the wgs database (online suppl. Table S4), all except three with at least two hits per species, corresponding to the expected Relish and Dif/Dorsal homologs.

To test toxicity of the D. ananassae Hemolysin E-like proteins, U937 pro-monocytic cells grown in RPMI-1640 medium (Gibco) supplemented with 0.137 mm streptomycin, 0.175 mm penicillin, 4 mm glutamine, and 5% FBS (Gibco) were washed, and 5 × 10⁴ cells in 100 µL fresh complete RPMI medium were incubated with 2 µg purified recombinant HL6, HL16, and HlyE proteins in 60 µL TBS. The recombinant proteins were produced either in E. coli or in D.Mel-2 cells. Photographs were taken using an Alpha XDS-1T inverted microscope and a Nikon D5300 DSLR camera. Besides visual observations, lysis of the cells was further tested with Lactate Dehydrogenase (LDH) Cytotoxicity Assay Kit (Abcam) according to the manufacturer’s instructions.

Hemolysin E proteins (HlyE, also known as cytolysin A or ClyA) are pore-forming toxins and important virulence factors of bacteria. Transcriptomic analysis of D. ananassae blood cells revealed the expression of 14 orphan genes in the MGHs (denoted as hl1-14, online suppl. Table S3), which encode for predicted Hemolysin E-like proteins [20]. A further analysis of annotated genes in the D. ananassae genome revealed the presence of another 23 additional hemolysin E-like genes (hl15-hl38), each with at least one intron (online suppl. Table S3). These genes have no homologs in the melanogaster subgroup species. Out of the 37 hemolysin E-like genes, 23 are localized on the left arm of the X chromosome (XL), and the others are localized on the 2L, 3R, 3L, and XR chromosomes (Fig. 1). The encoded proteins are between 33 and 48 kDa, each with a single conserved hemolysin E-like domain. Exceptions are hl36, which contains additional keratin-like repeats, and hl8 and hl38, which appear to be truncated pseudogenes (see below). Of the 14 MGH-specific genes, 10 are localized within a tight, 30 kb, cluster on XL (Fig. 1). Further analysis revealed that 15 of the encoded proteins carry predicted N-terminal signal peptides, suggesting their secretion or membrane localization. Notably, all 14 MGH-specific genes lack such export signals (online suppl. Table S3).

Three of the predicted hemolysin E-like genes, hl4, hl38, and hl8, appear to be pseudogenes. The hl38 gene encodes a short protein that is 94% identical to the first 118 amino acids of HL3, and the neighbor gene, hl4, encodes the missing 221 amino acids of a full-length Hemolysin E-like protein, which is 92% identical to the C-terminal part of HL3. The two gene halves are separated by a 9,619 bp LTR retrotransposon, which is also represented in at least 8 other places in the D. ananassae genome, and there is no evidence of consensus splice signals that could rescue this split gene. The predicted hl38 transcript runs into a stop codon in the transposon after a few nucleotides, while hl4 is predicted to initiate inside the long terminal repeat of the transposon. Furthermore, the hemolysin E-homologous region in the hl8 gene is interrupted by a 32 bp deletion, causing a frameshift. The annotated hl8 gene product lacks 82 amino acids from the N-terminal region of a Hemolysin E-like protein. Thus, while the hl4, hl38, and hl8 may be transcribed, they are unlikely to retain their original function.

Multiple sequence alignment of the D. ananassae Hemolysin E-like proteins and the E. coli HlyE revealed some degree of evolutionary conservation, as it is noteworthy that at several positions the amino acids are identical (online suppl. Fig. S1). Although the alignment did not result in a clear consensus sequence, pairwise similarities with the bacterial protein indicate the possibility of a similar function of the insect proteins.

The NCBI Conserved Domain Search revealed that each Hemolysin E-like protein of D. ananassae carried a hemolysin E-like domain, which in bacteria is composed of five α-helical regions and a β-hairpin named as β-tongue. Furthermore, AlphaFold analysis revealed the predicted three-dimensional structures of the D. ananassae Hemolysin E-like proteins, which were compared to those of the E. coli HlyE. The comparisons revealed that the insect proteins were highly similar to that of the E. coli HlyE (online suppl. Fig. S2).

We sought to determine whether insect-encoded hemolysin E genes arose from one or more horizontal gene transfer events from microbes to insects or even between insects. To address this, we first conducted a tblastn search using the hemolysin E domains of all the D. ananassae paralogs against the nr database. This search recovered few hits to prokaryotic hlyE genes, but many hits to genes encoding uncharacterized or hypothetical proteins in other Drosophila and parasitoid wasp species (online suppl. Fig. S3). Importantly, none of the eukaryotic hits were identified as homologs to other bona fide eukaryotic gene families. To find more distantly related Hemolysin E homologs, we then ran the same set of D. ananassae hemolysin E domains in a phmmer search against the Refseq Protein database. This recovered more prokaryotic hits, though some were from other gene families, such as YadA. These hits often did not align well with sequences annotated as Hemolysin E and were subsequently removed from the alignment. The final alignment used in tree generation contained 254 sites and 134 sequences, 17 of which were of prokaryotic origin (online suppl. Fig. S3). Notably, there are 0 conserved sites among all 134 sequences (i.e., no site has 100% conservation), indicating high sequence variation among the hemolysin E gene family.

Homologs from bacterial taxa were chosen arbitrarily as outgroups to root the tree. Log likelihoods of each tree were the same (−32790.294571) regardless of which bacterial taxon was selected as the outgroup. In all cases, insect homologs of Hemolysin E formed a monophyletic clade on a long branch with 100% bootstrap support (online suppl. Fig. S4). Relationships between insect homologs of Hemolysin E are poorly resolved, likely due to the number of duplication events that occurred within the insect lineages. Interestingly, there are Hemolysin E homologs from Nasonia vitripennis and Trichomalopsis sarchophagae, two hymenopteran parasitoids of flies, interspersed among the drosophilid Hemolysin E homologs on the tree (online suppl. Fig. S3). This suggests there may have been inter-insect HGT events that facilitated the incorporation of hemolysin E genes into diverse insect genomes or repeated donation of closely related hlyE genes from bacteria into these insects.

We were surprised by the erratic presence and absence of hemolysin E genes among different insect species and therefore investigated this phenomenon in more detail, with a focus on flies in the Drosophilidae. As the quality and completeness may vary between genome projects, we chose to conduct a thorough de novo tblastn search for hemolysin E homologies in available genome sequences. Figure 2 shows the number of relevant hits in a selection of extensively sequenced genomes, and a full set of hemolysin E homologs in all accessible drosophilid genomic sequences is listed in online suppl. Table S4. Each hit corresponds to a DNA segment with significant sequence similarity to one of the characterized hemolysin E genes, including pseudogenes and in some cases individual hemolysin E exons (if they are separated by more than 200 bp, see Materials and Methods).

Strikingly, we could only identify hemolysin E-homologous regions in 45 of the 335 sequenced species (13%), but they were distributed across the drosophilid species tree (Fig. 2; online suppl. Table S4). They were relatively well represented within the subgenus Sophophora, not least in the subgroup ananassae (including D. ananassae), in which all 14 sequenced species harbored large families of hemolysin E homologs with up to 113 members (online suppl. Table S4). More modest numbers (2–12) were found in some other members of the larger melanogaster species group. Specifically, we identified several hemolysin E homologs in most members of the rhopaloa, elegans and ficusphila subgroups, while none were found in the melanogaster, eugracilis, suzukii, takahashii, oshimai, or montium subgroups, nor in the obscura group. Furthermore, most species in the willistoni and saltans groups have a few hemolysin E homologs (but 22 in Drosophila sturtevanti), while no copies were found in the genus Lordiphosa, which is nested together with the willistoni and saltans groups in the phylogenetic tree of the Drosophila subgenus Sophophora.

The only investigated member of subgenus Dorsilopha, Drosophila busckii, had eight hemolysin E-homologous regions, while the hemolysin E homologs were more sparsely distributed in the subgenus Drosophila (here defined to include the nested genera Zaprionus, Liodrosophila, Hypselothyrea, Hirtodrosophila, and Scaptomyza, Fig. 2 and online suppl. Table S4). Within this large subgenus, we found hemolysin E homologs only in a single Scaptomyza species and in 9 of the 26 investigated species of the repleta group. These species typically had only a single homolog each (Fig. 3; online suppl. Table S4). Finally, out of 19 sequenced genomes from drosophilids outside the genus Drosophila (defined as above), we found hemolysin E homologs in one species only, Chymomyza caudatula, with two homology regions (online suppl. Table S4).

We considered the possibility that the absence of identified hemolysin E homologs in many species could be due to incompleteness of available genomic sequences. However, at least two Rel homology domains, presumably corresponding to the expected Rel and Did/Dorsal homologs, were found in all 335 species listed here (see Methods), including the 45 where hemolysin E homologs were detected. This suggests that few hits, if any, were lost due to incompleteness of genomic sequences.

As species of the ananassae subgroup are considerably more resistant to parasitoid wasps than D. melanogaster, we analyzed the expression of the hemolysin E-like genes in naïve and L. boulardi-infected D. ananassae larvae. Total RNA was isolated from infected (72 h after infection) and age-matched naïve larvae, and the gene expression profiles of the infected larvae were normalized to those of the naïve animals. All genes, except for hl2, hl4, hl8, hl28, hl35, and hl38, were significantly induced following parasitoid attack (Fig. 3). Owing to the very high sequence similarities between the hl13 and hl14 and also the hl22, hl24, and hl25 gene transcripts, no discriminative oligonucleotide primer sets could be used in the qRT-PCR. Therefore, the expression and induction of these genes could not be monitored individually. The transcription of the hl15, hl29, hl32, and hl33 genes could not be detected (online suppl. Table S3), though several primer sets were tested (online suppl. Table S1).

We next examined the tissue-specific expression of each of the hemolysin E-like genes in parasitized larvae. The majority of these genes (hl1, hl2, hl3, hl7, hl8, hl9, hl10, hl11, hl12, hl13, hl14, hl16, hl17, hl18, hl19, hl20, hl21, hl22, hl23, hl24, hl25, hl27, hl30, hl31, hl34, hl35, hl36, and hl38) showed the highest expression in the fat body. hl5, hl6, and hl37 were most highly expressed in blood cells, while hl4 and hl28 genes were most highly expressed in the rest of the body (Fig. 4, online suppl. Fig. S5). For further investigations, two candidates were selected on the basis of their characteristic expression: hl6 (Dana\GF22667), which showed high expression in blood cells, and hl16 (Dana\GF21479), which expressed specifically in the fat body.

We next analyzed whether other stimuli besides parasitoid wasp infection could trigger induction of the hl genes. We challenged larvae by sterile wounding or septic wounding using a 1:1 mixture of Gram-negative (E. coli) and Gram-positive (B. subtilis) bacteria, or spores of B. bassiana, an entomopathogenic fungus. Neither sterile nor septic injury significantly induced expression of the hl6 and hl16 (Fig. 5), but their expression increased after L. boulardi infection, consistent with our previous findings (Fig. 3, 5, 6).

To further examine the expression patterns of hl6 and hl16, we surveyed the transcript and protein levels of these genes over the course of development. In naïve samples, expression of the hl6 and hl16 genes had slight variations, but parasitoid wasp infection caused strong induction in the gene expression profile, which persisted up until the adult stages (Fig. 6a).

We then analyzed the expression pattern of the HL6 and HL16 proteins using anti-HL sera. Developmental analysis revealed induction and a high expression of both the HL6 (42 kDa) and the HL16 (35 kDa) proteins after parasitization, especially in larval and pupal stages. Moreover, an elevated expression of the HL16 protein could still be detected in adults that survived infection (Fig. 6b).

To reveal the tissue-specific expression of the HL6 and HL16 proteins indirect immunofluorescence assays were carried out in wasp-infected larvae. The HL6 protein was expressed in MGHs, in a subpopulation of spherical hemocytes and in the fat body (Fig. 7). The HL16 protein was detected in the fat body and also in vesicles localized in the hemolymph (Fig. 7).

While our expression data suggest the hemolysin E-like genes play a role in parasitoid response, we next wanted to determine whether the encoded proteins were interacting with parasitoid wasps in vivo. To this end, we isolated L. boulardi larvae from D. ananassae 72 h post-infection, then performed indirect immunofluorescence analysis using anti-HL6 and anti-HL16 sera. Both the HL6 and HL16 proteins were detected on the parasitoids (Fig. 8a). To test whether recombinant Hemolysin E-like proteins can bind directly to the parasitoid surface, we incubated under in vitro conditions L. boulardi larvae isolated from D. melanogaster with purified FLAG-tagged recombinant HL6 and HL16 proteins. Indirect immunofluorescence analysis showed that both proteins bound to the parasitoid (Fig. 8b).

To assess whether D. ananassae Hemolysin E-like proteins, similarly to their prokaryotic counterparts, function as pore-forming toxins, HL6, HL16, and E. coli HlyE were generated in recombinant forms, in E. coli, and in the D.Mel-2 cell line, purified, and incubated with U937 pro-monocytic cells, a known target for the bacterial HlyE [44]. Microscopic imaging and the lactate dehydrogenase (LDH) cytotoxicity assay were used to assess the damage of cellular membranes and lysis of the target cells. Sampling was done at 2, 4, 6, and 24 h following incubation of the U937 cells with the HL6, HL16, and HlyE proteins. The E. coli HlyE protein produced in E. coli exhibited toxicity just 1 h after incubation. However, the HlyE generated in D.Mel-2 cells did not show toxic activity, even 24 h following the treatment (online suppl. Fig. S6). Toxic activity of the D. ananassae Hemolysin E-like proteins generated in either prokaryotic or eukaryotic cells could not be detected (online suppl. Fig. S6).

The fruit fly D. ananassae is particularly resistant to parasitoid wasp attack when compared to congeners. Intriguingly, D. ananassae and several other Drosophila species lack both encapsulating lamellocytes and PPO3, a key enzyme responsible for melanotic capsule formation and parasitoid wasp death in D. melanogaster. This suggests that outside the melanogaster subgroup another, heretofore unknown parasitoid wasp-killing mechanism operates.

Some of these particularly parasitoid-resistant species, including D. ananassae, have evolved a new class of encapsulating blood cells called MGHs, which contribute to the highly efficient elimination of developing parasitoid wasps [4, 5, 20, 45]. In addition, the presence of toxin gene families of microbial origin, encoding for homologs of CdtB and AIP56 were reported in several insect species, including those in the ananassae subgroup of Drosophilidae [10, 11]. These genes were captured by drosophilid flies through horizontal gene transfer ca. 20 million years ago, and we found that in D. ananassae they are necessary for the resistance to a diverse array of parasitoids [12].

Here, in the very same species, we describe the presence of an orphan gene family of 37 members that encode predicted proteins carrying hemolysin E domains characteristic of prokaryotic pore-forming toxins [20], and investigated the origin, expression, and characteristics of the D. ananassae hemolysin E-like genes and proteins. We found that several drosophilids that have evolved MGHs also carry cdtB, aip56, and hemolysin E-like genes (Fig. 2), and these genes have no homologs in D. melanogaster. We hypothesized that insect-encoded hemolysins originate from an ancient horizontal gene transfer event from bacteria, followed by several duplication events in insects and possibly later inter-insect horizontal gene transfer events. However, we cannot rule out the possibility that insect hemolysins arose convergently through evolution, for a number of reasons. First, although visual inspection of the alignment reveals apparently shared, derived amino acid substitutions (synapomorphies) between insect and prokaryotic homologs of the gene, it is possible that these characters are instead homoplastic owing to shared selective pressures. Given the possibility of the latter, it is therefore possible that some insect genes were annotated as hemolysins due to structural similarity, rather than shared ancestry among them. Second, unlike in previously characterized microbe-to-insect HGT events, there is no clear insect-associated bacterial donor lineage represented in the phylogeny we reported. The lack of a clear donor lineage could be the result of sampling bias or underrepresentation of these sequences in available databases, ancient origins of these genes, rapid divergence after horizontal gene transfer events, or a combination of these.

The scattered distribution of hemolysin E-like genes among flies, or indeed among other insects, further obscures the nature of the origin of these genes. It could be interpreted as a result of frequent and relatively recent horizontal gene transfer events between bacteria and insects. However, that seems unlikely, given that the genes harbor introns and appear to be relatively well adapted to the eukaryotic environment. It is also possible, as alluded to above, that once a horizontal transfer event occurred from bacteria to an insect and the insect copy gained introns, this intron-bearing hemolysin E-like gene was horizontally transferred from one insect lineage to others. Such a scenario may have given rise to the cdtB gene of Drosophila biarmipes and of Myzus persicae and M. cerasi aphids, which each has the same splice junctions as the cdtB copies in D. ananassae [10]. Finally, another alternative hypothesis is monophyly. However, if the insect hemolysin E-like genes indeed originated from a single ancient horizontal transfer event from bacteria to insects, we must then infer that the genes were subsequently independently lost along many parallel lineages and retained in a handful of others. In that scenario, we would also expect the phylogenetic relationships among the hemolysin E genes to mirror that of the phylogeny of the insects that harbor them (the gene tree to match the topology of the species tree), which is clearly not the case. Therefore, we conclude that the working hypothesis best explained by the data is one or a few initial horizontal transfer events followed by the origin of introns and subsequent inter-insect transfer.

The expression of D. ananassae hemolysin E-like genes and proteins are induced after parasitoid wasp infection (Fig. 3, 5, 6) primarily in the blood cells and the fat body (Fig. 4, online suppl. Fig. S5), the main immune tissues of Drosophila [46]. Single cell transcriptomic analysis showed high expression of the hl6 gene in MGHs [20], and the HL6 protein was also detected in these cells (Fig. 7). Bacteria secrete the HlyE molecules through an outer membrane vesicle system, which provides the conditions for oligomerization and formation of active pore assemblies [21, 25]. The strong HL6 expression in a respective subpopulation of spherical hemocytes (Fig. 7) indicates that these cells could also possess a significant role in the anti-parasitoid immune defense. Expression of most of the D. ananassae hemolysin E-like genes was enriched in the fat body of wasp-infected larvae (Fig. 4, online suppl. Fig. S5), the main site of anti-microbial peptide production. Fifteen of the predicted Hemolysin E-like proteins possess N-terminal signal peptides, suggesting that they could be secreted into the hemolymph. We hypothesize that HL1-HL14 proteins, which lack N-terminal signal peptides (online suppl. Table S3), could be cargo for the large number of exosomes and microvesicles released by MGHs [20]. The hl4 and hl28 genes showed the highest expression in the rest of the organism (Fig. 4, online suppl. Fig. S5), but these two genes were in general expressed at an extremely low level in both naïve and infected animals (Fig. 3). The hl4, the hl8, and the hl38 genes are all localized on the chromosome XL, where most of the hemocyte-specific genes reside (Fig. 1), and we could detect a corresponding transcript for them (Fig. 3, 4, online suppl. Fig. S5). However, due to a transposon insertion (hl4 and hl38) and a frameshift deletion (hl8), they can be regarded as pseudogenes, which could also explain their low expression profile and the lack of their induction following parasitoid wasp infection (Fig. 3).

Most of the D. ananassae hemolysin E-like genes are localized on the L arm of the X chromosome (Fig. 1). Their adjacent localization suggests the possibility of tandem gene duplications over evolutionary time in this lineage. Following integration into the eukaryotic genome, the genetic elements of prokaryotic origin must undergo a eukaryotic domestication process to become incorporated into the existing genetic networks [47] as they may acquire new regulatory elements or eukaryotic motifs such as introns [10]. Each hemolysin E-like gene in D. ananassae carries at least one intron (online suppl. Table S3). Furthermore, there is a salient relationship between the chromosomal location and the regulation of these genes. For example, the hl1-hl14 genes which are expressed in MGHs [20], are located relatively close to each other on the L arm of the X chromosome (Fig. 1). This further supports the hypothesis for tandem duplications of the hemolysin E-like gene regions within certain chromosomal areas, or suggests that neighboring genes might be under the control of the same regulatory elements [48]. Moreover, transcription of the hl15, hl29, hl32, hl33 genes, which are localized relatively far from the other elements of the family (Fig. 1) could not be detected (online suppl. Table S3, S1; Fig. 3). Cryptic feature of these hemolysin E-like genes might be due to the absence of their regulatory elements or due to the possible accumulation of mutations in their regulatory sequences over the course of evolution that have changed their expression profiles.

The endogenous D. ananassae Hemolysin E-like proteins attached to the parasitoid wasps (Fig. 8a) and the purified recombinant proteins, under “in vitro” conditions, were also bound to the parasitoids isolated from D. melanogaster (Fig. 8b). This evidence is consistent with the direct binding of the proteins to the parasitoids. Toxicity assays revealed that recombinant bacterial HlyE generated in E. coli was highly toxic to the eukaryotic cell line. However, neither E. coli and D.Mel-2-derived D. ananassae Hemolysins nor D.Mel-2-derived HlyE were lytic. To explain this, we hypothesize that these toxins require a specific cell type for their production with a specific subcellular environment that enables proper folding and processing of these proteins to become functionally active. We further hypothesize that only specialized D. ananassae cells, such as those of the fat body or relevant hemocyte subpopulations, could provide the proper environment or the specific conditions necessary to form active pore-forming assemblies.

Our findings highlight the importance of the horizontal gene transfer in the evolution and adaptation of insect immune responses. The horizontally acquired genes may provide new functional modules that can be integrated into pre-existing gene networks. The large number and the high similarity of these genes forecasts limitations on the possible assays for further functional investigations. Moreover, as the available protein expression systems could not generate functional toxins, it is still unclear how the Hemolysin E-like molecules are processed in D. ananassae, and what conditions are required to provide the proper intracellular microenvironment for generation of functional protein complexes. It is also unknown how these toxins can be directed against eukaryotic invaders without harming the eukaryotic host. The elucidation of these mechanisms could provide the chance for alternate therapies based on pore formation and membrane damage of targeted cells using these virulence factors captured and adopted from bacteria.

Finally, in addition to the lamellocyte encapsulation and melanization response against parasitoid wasps found in D. melanogaster and other insects our study shows that in species lacking these traits, an alternative strategy has evolved that relies on the evolution of novel cell types like MGHs and humoral factors of prokaryotic origin that confer resistance to parasitoid wasps.

We are grateful to Prof. Todd Schlenke for the L. boulardi parasitoids. We thank Olga Kovalcsik, Anita Balázs, and Katalin Ökrös for their technical assistance.

This study protocol was reviewed and approved by the Government Office of Csongrád-Csanád County, Department of Food Chain Safety and Animal Health, Hungary, Approval No. XVI/966/2023.

The authors have no conflicts of interest to declare.

In this research, the NKFI K135877 (I.A.) grant from the Hungarian National Science Foundation, the National Laboratory for Biotechnology Program Grant No. (2022-2.1.1-NL-2022-00008), and the Hungarian Academy of Sciences (Lendület Program Grant No. [LP2017-7/2017]) (Z.L.) provided costs of the RNA and protein works, costs of antibody production, the charges for the microscopy and other laboratory equipment required for the work, salaries, and publication fee. The grant 2018-05114 from the Swedish Research Council (D.H.) and the National Institute of General Medical Sciences of the National Institutes of Health (award No. R35GM119816) (N.K.W.) provided the technical support for bioinformatical analysis and support of supply. The Institute of Genetics, HUN-REN Biological Research Centre, Szeged, Hungary provided salary (G.C.). The Doctoral School of Biology, University of Szeged, Hungary (L.B.M.) supplied scholarship.

L.B.M. and G.C.: conceptualization, experimental design, plasmid constructs, RNA isolation, qRT-PCR, Western blot analysis, IIF experiments, toxicity assays, data and bioinformatical analysis, and preparation of the manuscript. I.A.: conceptualization, experimental design, antibody generation, IIF experiments, toxicity assays, data analysis, and preparation of the manuscript. V.V.: immune induction, RNA isolation, and qRT-PCR. E.Á. and Z.L.: experimental design, plasmid constructs, generation of recombinant proteins, Western blot analysis, and preparation of the manuscript. D.H., R.L.T., and N.K.W.: bioinformatical analysis, data analysis, and preparation of the manuscript.

The data that support the finding of this study are included in this article and its online supplementary material files. Further inquiries can be directed to the corresponding author.