Horizontal gene transfer underlies the painful stings of asp caterpillars (Lepidoptera: Megalopygidae)

Significance

Animal venoms are widely acknowledged as rich sources of new molecules with applications in medicine, agriculture, and science. Whereas the venoms of snakes and arachnids have been extensively studied, the venom of many insect taxa remains entirely uncharacterized. Asp caterpillar venoms cause extreme pain, but nothing is known about their production, composition, or mode of action. This study reveals the anatomy of venom production and delivery, the composition of the venom, and its mechanism of inducing pain. The central insight of this paper is that the toxins responsible for causing pain are encoded by genes that have been acquired by horizontal transfer from bacteria, highlighting the role of horizontal transfer in the evolution of animal venoms.

Abstract

Larvae of the genus Megalopyge (Lepidoptera: Zygaenoidea: Megalopygidae), known as asp or puss caterpillars, produce defensive venoms that cause severe pain. Here, we present the anatomy, chemistry, and mode of action of the venom systems of caterpillars of two megalopygid species, the Southern flannel moth Megalopyge opercularis and the black-waved flannel moth Megalopyge crispata. We show that megalopygid venom is produced in secretory cells that lie beneath the cuticle and are connected to the venom spines by canals. Megalopygid venoms consist of large aerolysin-like pore-forming toxins, which we have named megalysins, and a small number of peptides. The venom system differs markedly from those of previously studied venomous zygaenoids of the family Limacodidae, suggestive of an independent origin. Megalopygid venom potently activates mammalian sensory neurons via membrane permeabilization and induces sustained spontaneous pain behavior and paw swelling in mice. These bioactivities are ablated by treatment with heat, organic solvents, or proteases, indicating that they are mediated by larger proteins such as the megalysins. We show that the megalysins were recruited as venom toxins in the Megalopygidae following horizontal transfer of genes from bacteria to the ancestors of ditrysian Lepidoptera. Megalopygids have recruited aerolysin-like proteins as venom toxins convergently with centipedes, cnidarians, and fish. This study highlights the role of horizontal gene transfer in venom evolution.

Larvae of moths of the genus Megalopyge (Fig. 1 A and B) are referred to as puss or asp caterpillars, reflecting their thick coverage of long hairs and the possession of painful venoms. Envenomations by asp caterpillars are notoriously painful, having been described as “the worst pain a patient has ever experienced,” “being hit with a baseball bat,” or “walking on hot coals” (1, 2). The pain builds from the time of injury, radiates to other parts of the body, and may persist for hours or days. The algogenic effects of asp caterpillar envenomations are typically accompanied by erythema and edema, and sometimes nausea, tingling, tachycardia, chest pains, fever, headache, and vomiting. Skin at the site of injury develops white spots that turn red, eventually forming a signature dark grid-like pattern that mirrors the distribution of venom spines underneath the long, fine hairs on the surface of the caterpillar (Fig. 1 C and D).

Fig. 1.

Venom apparatus of asp caterpillars. (A) Megalopyge opercularis late instar. (B) Megalopyge crispata late instar. (C) Clusters of droplets of venom (blue arrow) harvested from the tips of spines with parafilm, showing a grid-like pattern. (D) Characteristic grid-like pattern of the lesion after natural envenomation by M. opercularis. Image by Dirk M. Elston (3) reproduced with permission. Panels E–I show scanning electron micrographs of an M. opercularis exuviae. (E) Cluster of venom spines (vs) among plumose hairs (ph). (Scale bar, 100 µm.) (F) Tips of two spines showing characteristic longitudinal striations. (Scale bar, 10 µm.) (G) Base of the venom spine (bvs) showing connection to the cuticle. (Scale bar, 50 µm.) (H) Venom spine broken near the tip showing trilobed cross-section due to longitudinal striations. (Scale bar, 25 µm.) (I) venom spine broken closer to base showing circular cross-section. (Scale bar, 50 µm.) Panels J–N show light micrographs of slices through formalin-fixed M. opercularis. (J) Longitudinal section through pigmented plumose hairs showing socketed connection at base. (Scale bar, 100 µm.) (K) Near-longitudinal section showing slender spine joining bulbous base or cone that merges with cuticle without a socket structure. Arrows indicate the transitional ring of the spine shaft where the cellular surround of the central lumen terminates. (Scale bar, 50 µm.) (L) Cross-section through venom spines and plumose hairs close to tips. (Scale bar, 100 µm.) (M) Cross-section through bases of plumose hairs (circled yellow) and venom spines (vs). “Dots” (orange arrow) indicate possible neuronal insertions. (Scale bar, 50 µm.) (N) Cell mass beneath scolus, showing canals (dark red circles) and large elongate secretory cells (gray circles). (Scale bar, 50 µm.) Panels O–R show µ-CT data. (O) Transverse slice through M. crispata late instar, showing canal-like structures through cuticle (blue arrows) and end-apparatuses of secretory cells (yellow arrows). (Scale bar, 1 mm.) (P) Three-dimensional reconstruction showing exocuticle with venom spines (red), endocuticle (green), and secretory cell end-apparatuses (yellow). Three spines have been labeled pink, their venom canals dark blue, and the secretory cell end-apparatuses they attach to light blue. (Q) View of three labeled spines within surrounding structures. (R) Closeup view of three labeled spines with surrounding structures.

Megalopygidae is a family within the lepidopteran superfamily Zygaenoidea with >250 described species found in North, South, and Central America (4). Recently, we reported the proteinaceous composition of venom produced by another family of zygaenoid caterpillars, the Limacodidae (5). In that study, we found that limacodid venom consisted mainly of peptides <10 kDa, including disulfide-rich knottins, modified neurohormones, and immune-derived cecropin-like peptides, with the latter responsible for the pain caused by stings from these caterpillars. Although Megalopygidae and Limacodidae are closely related families within Zygaenoidea (4), it remains uncertain whether venom use evolved independently in the two families (6) or was inherited from a common venomous ancestor with subsequent losses in related taxa including the nonvenomous Limacodidae (7). One source of information that may inform these evolutionary possibilities is the composition and mode of action of megalopygid venom. Like limacodid venoms, megalopygid venoms reported to display strong hemolytic effects, proinflammatory and procoagulant effects, and have weak hyaluronidase activity (8, 9). However, there are some indications that megalopygid venoms are different from those of the Limacodidae. A previous investigation into a venom spine extract from Megalopyge [Lagoa] crispata (10) demonstrated that the pain-causing component(s) eluted early from a size exclusion column, were easily inactivated by proteases, heat, or storage at 4 °C, and, therefore, were likely high molecular mass proteins. A venom spine extract of the megalopygid caterpillar Podalia ca. fuscescens also contained mostly high molecular mass proteins >10 kDa (11).

Peptides and proteins follow diverse evolutionary paths to become venom toxins (12–17). The recruitment of endogenous peptides and proteins with innately “toxin-like” properties into venom is thought to be a key process. This may be followed by duplication and specialization of the daughter gene products (12, 15), although many toxins remain single-copy gene products (18, 19). While evolution acts on genes that are transferred between generations of the same species through normal reproduction (vertical transfer), it also acts on those genes that are more rarely transferred between individuals of different, often distantly related species (horizontal transfer). Horizontal gene transfer increases the genetic diversity that may be recruited to produce venom toxins, and multiple venom toxins encoded by horizontally transferred genes have now been reported (20–23).

Here, we describe the anatomy of the megalopygid venom system. We show that a major component of megalopygid venoms are pore-forming aerolysin-like proteins (megalysins) which are responsible for the pain associated with envenomation. These megalysins were recruited as venom toxins in the Megalopygidae from genes that were transferred horizontally from bacteria to the ancestors of Lepidoptera. This study reveals the detailed workings of the megalopygid venom system and highlights the role of horizontal gene transfer in the evolution of venom toxins.

Results

Megalopygids Possess a Unique Venom Apparatus.

Applying parafilm supported on a wire loop to the backs of M. opercularis (Fig. 1A) and M. crispata (Fig. 1B) resulted in venom droplets deposited in a grid-like pattern (Fig. 1C) that closely resembles the pattern of erythema characteristic of envenomations by asp caterpillars (Fig. 1D). We observed that freshly collected venom was clear or light brown in color, but unrefrigerated venom samples developed a dark brown or black color.

Megalopyge caterpillars are covered with long and slender nonvenomous hairs that conceal shorter, stouter venom spines (11, 24–26). To further investigate the anatomy of the venom apparatus, we examined spines and the surrounding tissues using scanning electron microscopy and light microscopy of slices through resin-embedded, formalin-fixed venom scoli. Scanning electron micrographs show clusters of smooth venom spines alongside branched, plumose hairs (Fig. 1E). Venom spines are up to 30 μm in diameter, tapering to dagger-like tips <5 μm thick (Fig. 1F). At the base, venom spines show a bulbous and wrinkled appearance and merge with the cuticle without any socket structure (Fig. 1G). At the tip, each venom spine has three longitudinal striations or grooves that yield a trilobed cross-section (Fig. 1H) but are hollow with a circular cross-section for most of their length (Fig. 1I). Light micrographs show that the plumose hairs are ~8 μm in diameter, covered with thin spines ~30 μm long that are directed upward from the shaft (Fig. 1J). The venom reservoir within the venom spines is continuous with a bulb-like chamber at the base (Fig. 1 K and L), whereas plumose hairs are connected to the cuticle with a complex articulate socket structure (Fig. 1M and also see Fig. 1 G and J). The venom spine bulb is continuous with a canal that traverses the thick scolus cuticle and terminates under the epidermis, surrounded by a densely packed mass of secretory cells (Fig. 1N). Immediately beneath the cuticle, small cuboidal cells dominate, whereas more deeply, the canals lie adjacent to a secretion associated with large secretory cells (~30 μm) with densely staining nuclei (Fig. 1N). Micro-computed tomography (µ-CT) data show these canals as striations traversing the cuticle and terminating in X-ray dense structures below the cuticle that likely represent the end-apparatuses of the secretory cells (Fig. 1 O–R and Movie S1). Together, these multicellular secretory units have the characteristic features of type III insect secretory units (27) and are the likely secretory source of venom in Megalopyge caterpillars. These data are consistent with the morphological study of M. crispata venom spines by Lamdin et al. (25) and reveal numerous differences between the venom apparatus of asp caterpillars and both the limacodid Doratifera vulnerans (5) or the saturniid Lonomia obliqua (28).

Asp Caterpillar Venoms Are Rich in Aerolysin-Like Proteins as Well as Peptides.

To determine the composition of M. opercularis and M. crispata venoms, we combined venom proteomics with transcriptomics of venom scoli and the tissue immediately below them. To identify venom-encoding transcripts, we performed liquid chromatography–tandem mass spectrometry (LC-MS/MS) of each venom. Since the venom we analyzed was extruded from the tips of the venom spines rather than being obtained by extract or digestion, we are confident that the identified proteins are venom proteins, despite our transcriptomes representing both venom-producing and non-venom-producing tissue (29). For each species, we analyzed native, reduced and alkylated, and reduced and alkylated and trypsinized venom samples by LC-MS/MS. By comparing tandem mass spectra with translated ORFs from the transcriptome of the corresponding species, we identified 45 and 41 toxin-encoding transcripts for M. opercularis and M. crispata, respectively (Dataset S1).

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and LC-MS/MS results show that M. opercularis venom consists of substantial amounts of both proteins >10 kDa and peptides <10 kDa (Fig. 2 A and B). A total of 32 (71%) detected polypeptides were proteins >10 kDa, accounting for 38% of the abundance of all transcripts encoding venom toxins (normalized transcripts per million, TPM), and 35% of venom MS counts (Fig. 2C). The remaining 13 (29%) of polypeptides detected were peptides <10 kDa, which accounted for 62% of transcript abundance and 65% of venom MS counts.

Fig. 2.

Venom proteome of M. opercularis. (A) SDS-PAGE of whole venom, with proteins identified by LC-MS/MS. (B) Total ion chromatogram of whole venom with peptides identified by LC-MS/MS. (C) Breakdown of venom composition by family (number of polypeptides detected, Left), with abundance measures for transcriptomic data (normalized TPM encoding venom polypeptides, Center) and proteomic data (spectral counts from venom proteomics used for quantitation, Right). (D) Amino acid sequences of major venom peptides. Cys residues are highlighted in gray. Z, pyroglutamate; # indicates C-terminal amidation.

Megalopyge venoms contain large proteins, the most abundant and diverse of which are twelve related proteins of 29 to 32 kDa grouped as “Family 8” that share high sequence similarity with bacterial pore-forming toxins of the aerolysin family (30) (Fig. 2C). For example, U-MPTX₈-Mo12 is annotated by the HMMER algorithm as belonging to the protein family (Pfam) “Insecticidal Crystal Toxin p42” with E value 8.4e⁻³, and U-MPTX₈-Mo15 is annotated by HMMER to belong to the related Pfam “Clostridium Epsilon Toxin ETX/Bacillus Mosquitocidal Toxin MTX” with E value 2.2e⁻². We refer to these proteins hereafter as megalysins for megalopygid aerolysin-like proteins. Aside from megalysins, M. opercularis venom contained six members of the pheromone/odorant binding protein family, some of which are closely related to the thaumetopoein Tha p 1, an allergen present in irritative setae of the processionary caterpillar Thaumetopoea pityocampa (Notodontidae) (31). Other proteins in the venom are closely related to venom components of Hymenoptera, including two cystine-rich secretory proteins and a hyaluronidase (Dataset S1). The melanization and immune enzyme phenoloxidase were also detected, as well as an S1 protease closely related to phenoloxidase-activating factor, which may explain our observation that the venoms spontaneously darken in color when left at room temperature. Other detected proteins include two with MD2-related lipid recognition (ML) domains, two aldo-keto-reductases, a serpin, and two proteins without detectable similarity to any described proteins. The same techniques applied to M. crispata venom yielded similar results, identifying many homologous proteins differing usually by a few amino acid substitutions (SI Appendix, Fig. S1). In comparison to M. opercularis venom, M. crispata venom was found to lack family 2 peptides but included lysozyme and a peptide predicted by HMMER to have a serine protease inhibitor I29 domain (32).

The most abundant venom peptides (Fig. 2D) included a 9.1-kDa peptide with eight cysteine residues annotated by HMMER as a single domain von Willebrand factor peptide (U-MPTX₁-Mo1); two peptides of 6.8 to 7.3 kDa in mass with six cysteine residues and annotated by HMMER as Kazal domain peptides (U-MPTX₃-Mo3 and -Mo4); two peptides of 2.7 to 3.4 kDa with four cysteines and unknown structure (U-MPTX₂-Mo2 and -Mo9), a 3.0-kDa peptide with two cysteine residues (U-MPTX₉-Mo13); and six peptides without cysteine residues. Posttranslational modifications were evident in some peptides, including C-terminal amidation, N-terminal cyclization of glutamine to pyroglutamate, and removal of C-terminal lysine residues (Dataset S1).

The composition of asp caterpillar venoms is markedly different to that of a limacodid caterpillar venom that we recently studied using similar techniques (5), despite both families being closely related within the lepidopteran superfamily Zygaenoidea. While the venom of the limacodid caterpillar D. vulnerans is rich in short linear neuropeptides such as cecropins, adipokinetic hormone/corazonin-related peptide, and inhibitor-cystine-knot-like peptides, these families were absent from the megalopygid venoms. Similarly, while aerolysin-like proteins were abundant and diverse in the megalopygid venoms, they were not detected in the venom of D. vulnerans. Indeed, a quantitative comparison of venom proteome similarity between asp caterpillars, nettle caterpillars, and three independently evolved insect venoms (ant, robber fly, and assassin bug) shows that megalopygid venom is no more similar to limacodid venom than it is to the venoms of insects of other orders (SI Appendix, Table S1).

Asp Caterpillar Venoms Activate Mammalian Neurons by Membrane Permeabilization.

Megalopygid caterpillars use their venoms for defense, and envenomation in mammals causes immediate and sustained pain. To investigate the nature and mode of action of the pain-causing toxin(s), we tested the bioactivities of M. crispata venom that was either untreated or pretreated to inactivate particular classes of molecules. Protease treatment (50 ng/µL trypsin and pronase for 2 h) was applied to inactivate both peptides and proteins, while heat (60 °C, 1 h) or acetonitrile (50% for 1 h) was applied to denature larger proteins but leave peptides intact. In each case, the untreated or treated venom was lyophilized before resuspension in water. These venom samples were tested on the human SH-SY5Y neuronal cell line using a duplex fluorescence assay (33) in which we simultaneously monitored intracellular calcium ([Ca²⁺_i]) as well as exposure of intracellular nucleic acids to propidium iodide, which is present extracellularly and has entered the cell due to membrane permeabilization (Fig. 3A). Application of M. crispata venom caused a robust and sustained concentration-dependent increase of [Ca²⁺]_i and exposure of DNA with EC₅₀ estimates of 92 ± 47 pg/mL and 107 ± 48 pg/mL, respectively. The sustained increase of [Ca²⁺]_i caused by M. crispata venom was distinct from that of the positive control 0.025% Triton-X100, where the increase in [Ca²⁺]_i was transient due to nonspecific permeabilization of the cell membrane allowing passage of larger molecules including the fluorescence quencher present in the extracellular media. Venom that was pretreated with heat, acetonitrile, or proteases was inactive in this assay (Fig. 3A).

Fig. 3.

Algogenic, membrane-permeabilizing, insecticidal, and anticoagulant effects of asp caterpillar venoms. (A) Activation of SH-SY5Y cells by membrane disruption induced by M. crispata venom, visualized with [Ca²⁺]_i-sensitive dye and propidium iodide (Left). Membrane permeabilization and calcium influx were concentration-dependent, but all activity was lost after treatment by heat, exposure to acetonitrile, or proteases (Right, P < 0.001 in each case, unpaired Student’s t test, each point n = 3). (B) Activation of DRG cells by application of venoms (50 µg/mL) visualized with [Ca²⁺]_i-sensitive dye. Graphs on the right show data for all cells, median (dark blue) and interquartile range (light blue). (C) Application of 1 µg/mL M. crispata venom to DRGs was ameliorated by treating venom with acetonitrile, heat, or protease (P < 0.001 in each case, Kruskal–Wallis test, n = 100 cells per sample at 250 s). (D) Pain behavior, erythema, and edema induced by injection of M. crispata venom (100 ng) into mouse hindpaw were inhibited by prior treatment of venom with acetonitrile. *P < 0.05, unpaired Student’s t test. **P < 0.005, paired (within mouse) or unpaired (between mouse) Student’s t test, n = 3. In the Middle panel, red and blue arrows indicate the injection of untreated and acetonitrile-treated venoms, respectively, and “i” and “ui” indicate injected and uninjected paws, respectively. (E) Fractionation of 50 µg M. crispata venom by size exclusion chromatography (Top) followed by testing of each individual fraction (second panel) located potent membrane-permeabilizing activity in fractions 5 and 6 (datapoints intersected by dotted lines connected to fractions 5 and 6 in the first panel). Numbers indicate fractions as in Top panel; PI, propidium iodide. Lower panels, concentration–response curves obtained by diluting these fractions. (F) Insecticidal activity of asp caterpillar venoms (Top), each point n = 8 flies; and anticoagulant of M. crispata venom on human plasma (Bottom), **P < 0.005 Student’s t test, n = 3.

These results were mirrored in cultures of mouse dorsal root ganglion (DRG) cells, which include neurons responsible for peripheral nociception. Application of untreated M. crispata venom to DRG cultures caused a strong and sustained concentration-dependent increase in [Ca²⁺]_i in all cells, neuronal and non-neuronal (Fig. 3B). Treatment with heat, acetonitrile, or proteases destroyed the ability of venom to activate DRG cells (Kruskal–Wallis test, P < 0.001 in each case; Fig. 3C). Similarly, while intraplantar injection of M. crispata venom (100 ng lyophilized and resuspended in water) into the hindpaw of mice caused sustained spontaneous pain behaviors, swelling of the injected paw, and erythema, acetonitrile-treated venom lost these activities (Fig. 3 D, Upper panel, P < 0.05 at indicated timepoints).

Together, these data reveal that the pain-causing agent in M. crispata venom is a protein (rather than a peptide or small molecule) that activates mammalian neuronal cells via an increase in [Ca²⁺]_i and propidium iodide–permeable pore formation. Consistent with this result, none of the eight major peptides of M. opercularis venom, which we produced using solid phase peptide synthesis (SPPS) or recombinant expression (SI Appendix, Fig. S2), had any effect on mouse DRG cells, up to a concentration of 10 µM. To further investigate the active components of M. crispata venom, size exclusion chromatography was used to separate 50 µg of venom into 15 fractions (Fig. 3 E, Upper). Fractions 4 to 10 displayed membrane-permeabilizing activity (Fig. 3E, second panel), and concentration–response curves (Fig. 3 E, Lower) revealed that, of these, fractions 5 and 6 were the most potent, with the highest dilution tested (1:729) producing approximately half the effect of the undiluted fractions (SI Appendix, Fig. S3). These fractions eluted between molecular weight standards at 44 and 17 kDa, which corresponds well will the predicted sizes of mature megalysin monomers (29 to 32 kDa). We analyzed the content of each fraction using LC-MS/MS, finding that fractions 5 and 6 contain multiple megalysin proteins and coincide with the peak abundance of Mc16 and Mc8, respectively (SI Appendix, Fig. S3).

Asp caterpillar venoms also have insecticidal effects when injected into vinegar flies (Drosophila melanogaster). The LD₅₀ values for these insecticidal effects measured at 2 h after injection were 13.5 ± 4.5 µg/g and 5.6 ± 1.7 µg/g for M. opercularis and M. crispata respectively, and at 24 h after injection they were 5.2 ± 2.0 µg/g and 3.3 ± 0.7 µg/g, respectively (Fig. 3 E, Left). Since asp caterpillar venom has been reported to have procoagulant activity (8) and coagulotoxicity is the key bioactivity that results in human fatalities after envenomation by taturana caterpillars (Lonomia spp.) (34), we also tested whether M. crispata venom would affect clotting of human plasma. We observed that 100 µg/mL M. crispata venom increased the spontaneous clotting time from 300.3 ± 13.8 s to 434.4 ± 32.0 s (mean ± SD; P < 0.005, Student’s t test; Fig. 3 E, Right). These data suggest that asp caterpillar venom possesses anticoagulant activity rather than procoagulant activity as previously reported (8). However, the insecticidal and anticoagulant activities observed occur at orders of magnitude higher concentration than the observed nociceptive effects.

Megalysins Are Pore-Forming Toxins.

Among proteins detected in venom, the megalysins are good candidates to mediate the membrane-permeabilizing activity observed in Asp Caterpillar Venoms Activate Mammalian Neurons by Membrane Permeabilization. These proteins are closely related (up to 33.3% global identity) to many proteins from Lepidoptera and bacteria (Fig. 4 and Dataset S2), several of which have been characterized structurally and functionally. Among proteins with tertiary structures in the Protein Data Bank (PDB), those with the highest sequence similarity are pore-forming toxins from bacteria such as parasporin-2, a Cry toxin from Bacillus thuringiensis (PDB 2ztb; up to 14.3% global identity), and the epsilon toxin from Clostridium perfringens (PDB 6rbp; up to 10.5% global identity).

Fig. 4.

Megalysins are pore-forming toxins. (A) Monomeric structures of four M. opercularis megalysins predicted by Alphafold 2. (B) Crystal structures of monomeric aerolysin-like proteins from bacteria and vertebrates (from left PDB 2ztb, 1uyj, 5zu4, 3c0o). (C) Monomeric structure of Mo12 predicted by Alphafold 2, rainbow display from the N terminus (blue) to C terminus (red). (D) Crystal structure of monomeric B. thuringiensis parasporin-2 (PDB 2ztb, left, rainbow display as in C with α-helices and β-strands marked. The three domains are indicated (Center), and a ribbon diagram is shown (Right). (E) Crystal structure of heptameric aerolysin (PDB 5jzt) showing alternating hydrophobic (yellow) and hydrophilic (green) residues that allow formation of a β-barrel pore. The alignment and hydropathy plot (Lower) demonstrate conservation of this sequence feature in megalysins. (F) Domain 1 aromatic patches of parasporin-2 (PDB 2ztb, Top Left) and Mo12 (Top Right). Alignment shows conservation of this sequence feature, which is thought to function in initial membrane binding, in megalysins.

To investigate whether megalysins have sequence features consistent with being pore-forming proteins, we analyzed their primary structures and modeled their tertiary structures using the Alphafold 2 algorithm (35). The predicted structures of Mo10, Mo12, Mo15, and Mo26 are highly similar (Fig. 4A), and in each case, the closest match to the modeled structure in the PDB, as judged by DALI (z score > 15.0 in each case), was parasporin-2 (PDB 2ztb), the monomeric form of a B. thuringiensis Cry toxin, which is toxic to a variety of human cells but lacks insecticidal activity (36, 37). The megalysins and parasporin-2 share key sequence features with aerolysin and other bacterial pore-forming proteins that are known to mediate pore formation (36, 37). These features include a series of alternating hydrophilic and hydrophobic residues on β-strands S8 and S9 (Fig. 4 D, Left) which, after oligomerization and structural rearrangement, form the pore β-barrel, with hydrophilic and hydrophobic residues facing into and out of the pore, respectively (Fig. 4E). A hydropathy plot of the equivalent region in each of the megalysin sequences shows strong conservation of this feature (Fig. 4 E, Lower). Another functional region of bacterial pore-forming proteins is a patch in domain 1 that is rich in aromatic residues, many of which are solvent-exposed, which is thought to mediate initial binding of monomers to the lipid bilayer (36, 37). This feature is also conserved in megalysins (Fig. 4F). We also note that the internal diameter of the aerolysin pore (13.7 Å) is of a size suitable to conduct both ions and small molecules such as propidium iodide (13.2 × 10.1 × 3.9 Å). Thus, the predicted tertiary structures of the megalysins are consistent with them functioning as pore-forming proteins that underlie the activity described in Asp Caterpillar Venoms Activate Mammalian Neurons by Membrane Permeabilization.

Megalysins Are Encoded by Genes Transferred Horizontally from Bacteria to Lepidoptera.

The megalysins are part of the aerolysin toxin family, which is widespread among bacteria, fungi, plants, and animals (38). This protein family has been recruited into the venoms of multiple animal groups including centipedes, cnidarians, and fish, and it is thought to have been transmitted between different lineages of organisms through horizontal gene transfer at least seven times (20, 21). Aerolysin-like proteins have also been reported to be produced by nonvenomous lepidopterans, including protein expression in egg yolk (39, 40). To investigate the evolutionary origin of the megalysins and their placement in the aerolysin family, we retrieved proteins with similarity to the twelve M. opercularis and eight M. crispata megalysins from the National Center for Biotechnology Information (NCBI) nonredundant database using basic local alignment search tool (BLAST) searches and supplemented these with aerolysin-like sequences from diverse taxa (38) to yield a database of 379 proteins from Lepidoptera (140), bacteria and archaea (184), nonlepidopteran arthropods (15), vertebrates (18), plants (5), and other organisms (22) (Dataset S2, Sheet 1).

The megalysins belong to a group of lepidopteran proteins whose closest relatives are bacterial pore-forming aerolysin proteins. The megalysins themselves show up to 33.3% global identity with aerolysin-like proteins from nonvenomous Lepidoptera, and up to 29.6% global identity with sequences from bacteria and archaea (Dataset S2, Sheet 2) but <25% global identity with the next most similar sequences from algae, and <22% global identity with sequences originating from other insects. When all lepidopteran aerolysin-like proteins are used as queries in a BLASTp search against all nonlepidopteran sequences, the top hit is between a sequence from the crambid moth Ostrinia furnacalis (XP_028165651.1) to “Clostridium epsilon toxin ETX/Bacillus mosquitocidal toxin MTX2 family pore-forming toxin” from Dickeya sp. gammaproteobacteria (WP_038918640.1). Comparison of this pair of lepidopteran and bacterial sequences shows that they share 43.6% global identity and 67.0% global similarity and were matched with BLAST E value of 1e⁻¹⁰⁷ within the context of this search (Dataset S2, Sheet 3). Indeed, the top 951 hits (global identity from 43.6% to 23.1%, E values ranging from 1e⁻¹⁰⁷ to 2e⁻⁷³) were all between lepidopteran sequences and those from bacteria and archaea, whereas top hits by lepidopteran sequences to those of nonbacterial taxa were lower (global identity <23% and E <e⁻⁷⁴ in each case; Fig. 5A and Dataset S2, Sheet 3). Notably, megalysins are only very distantly related to aerolysin family proteins that have been recruited as venom toxins by centipedes (β-pore-forming toxins; E = 2e⁻¹²) or fish (natterins; E = 2e⁻⁶), and even more distantly related to aerolysin itself (E > 0.05). They are also not closely related to another group of aerolysin-like proteins in insects, the insect “natterin-like” proteins (E > 0.05) (41). For this reason, we refer to closely related proteins from nonvenomous lepidopterans as “megalysin-like proteins” in the following results.

Fig. 5.

Evolution of megalysin-like proteins. (A) Distribution of megalysin-like proteins across the tree of life. Numbers in ovals show maximum global identity with lepidopteran megalysin-like proteins. Dark blue circles indicate no known amino acid sequences with global identity >20%. Numbers in parentheses show (number of nucleotide hits with E < 1e⁻²⁰ in NCBI’s whole genome shotgun contigs database: number of WGS databases searched). Pink lines show putative horizontal gene transfer events. (B) Molecular phylogeny of megalysins and related sequence reconstructed using Bayesian inference. Numbers indicate posterior probabilities for selected nodes. The pink oval indicates posterior probability for the sister group of lepidopteran megalysin-like proteins occurring in Gammaproteobacteria.

To further investigate the evolution of megalysins and related lepidopteran proteins, we performed Bayesian inference of phylogeny on a smaller set of 271 sequences comprising the megalysins, related lepidopteran proteins, and their closest relatives, with Aeromonas hydrophila aerolysin as the outgroup (Fig. 5B). All megalysins from M. opercularis and M. crispata venom formed a single clade (posterior probability (P) = 0.95), with multiple orthologues between the two species (P = 1). All lepidopteran sequences formed a single clade (P = 0.80) containing multiple subclades that include sequences from diverse families. This topology, with multiple paralogues in a single species, suggests a single origin of megalysin-like proteins in Lepidoptera followed by gene duplication and divergence. The sister group of all lepidopteran sequences contains six members, all of which originate from the order Enterobacterales within the Gammaproteobacteria (P = 0.79). Other metazoan sequences were more distantly related.

There is evidence for the presence of genes encoding aerolysin-like proteins in multiple ditrysian Lepidoptera genomes. Genomic evidence of aerolysin-like genes has been reported from 13 diverse lepidopteran superfamilies including Geometroidea, Noctuoidea, Pyraloidea, Gelechioidea, Papilionioidea, and the early-diverging ditrysian superfamilies Yponomeutoidea and Tineoidea (Dataset S2, Sheet 4). Thus, such genes are widespread across the clade Ditrysia, which contains the vast majority of lepidopteran species. Many species possess multiple paralogues; for example, searching the Silkbase database (42) using the megalysins as queries reveals three different genes on chromosome 24 of the silkworm Bombyx mori that encode aerolysin-like proteins. For this reason, megalysins and related proteins reported from nonvenomous Lepidoptera represent products of genes within the lepidopteran nuclear genome rather than bacterial symbionts or bacterial contamination of genome sequencing.

Together, these data strongly suggest that a gene encoding an aerolysin-like protein was horizontally transferred from bacteria to an ancestral lepidopteran, was retained (and in some lineages, duplicated), due to serving an unknown nonvenom role. Subsequently, in an ancestral asp caterpillar, the product of this gene was recruited once as a venom toxin (followed by duplication and diversification). The exact donor and recipient of this gene transfer are unknown. We searched for, but did not find, related genes in the two available genomes from nonditrysian lepidopterans, Adela reaumerurella (WYDE00000000.1) and Neomicropteryx cornuta (43), although this might also have resulted from gene loss. A handful of sequences among nonlepidopteran holometabolan insects possess moderate sequence similarity to aerolysin-like proteins, which could suggest the transfer recipient was an ancestor of Holometabola rather than just Lepidoptera. However, the most parsimonious hypothesis that fits the available data is horizontal transfer from a bacterium, possibly a gammaproteobacterium, to an ancestor of the ditrysian Lepidoptera. If this occurred shortly after the divergence of Ditrysia, it is likely to have occurred approximately 175 Mya (44).

Discussion

A diverse array of toxic defenses has evolved independently across the Lepidoptera (13, 45). These include subcuticular chambers containing cyanogenic glucosides in the zygaenid caterpillar Zygaena filipendulae, (46); “spicule-like” urticating and toxin-bearing structures in the processionary notodontid Ochrogaster lunifer and the erebid Euproctis chrysorrhoea, (47, 48); and in many other caterpillars including many megalopygids, limacodids, and saturniids, spines that inject liquid venoms. We found that the anatomy of venom production in M. opercularis is different from both that of the limacodid D. vulnerans and the saturniid Lonomia obliqua, both of which possess a single large specialized secretory cell within individual venom spines. By contrast, in M. opercularis, large cells resembling type III secretory units (27) located underneath the cuticle secrete venom into a canals continuous with the reservoir of venom spines. The association of the type III secretory units with what appear to be hollow, modified sensory hairs is an example of organule-like structures that combine the attributes of sensory hairs and secretory structures (49).

The composition of Megalopyge venoms also differs markedly from those of previously investigated caterpillars. The most extensively studied lepidopteran venom comes from the taturana (Saturniidae: Lonomia spp.) of South America. These highly dangerous venoms are notorious for their ability to cause consumptive coagulopathy and hemorrhagic syndrome that may lead to death. Research on Lonomia venom toxins has focused on two proteins purified from spine extracts that likely contribute to the coagulopathy: Lopap, a tetramer-forming 21-kDa lipocalin family protein with enzymatic activity that catalyzes the conversion of prothrombin to thrombin (50), and Losac, a 45-kDa hemolin family protein that proteolytically activates Factor X (51). While extensive research on Lonomia venoms has yielded multiple other putative venom toxins from cDNA libraries and bristle extract proteomics (52–54), it is worth noting that a global venom proteome of the type determined in this study of Megalopyge venoms has never been reported for Lonomia spp. To the best of our knowledge, Lonomia venom studies have always employed bristle extracts rather than venom harvested from the tips of spines and hence include house-keeping and cuticle proteins (29). The use of SDS-PAGE bands and spots for protein analysis rather than whole venom has likely also led to the omission of venom peptides and minor venom components. In any case, hemolin and lipocalin family proteins are not present in Megalopyge venoms, and there is likely to be little overlap between the venom composition of Megalopyge and Lonomia venoms. It is perhaps more surprising that Megalopyge venoms show little overlap with those of their more closely related relatives within Zygaenoidea, the Limacodidae. None of the major classes of toxins we found in Megalopyge spp. venoms were found in the venom of the limacodid Doratifera vulnerans (5) and vice versa. Taken together, our findings that both venom system anatomy and venom composition differ markedly between Limacodidae and Megalopygidae suggest that the two families evolved venom independently, as has been previously suggested (6), although we cannot rule out the possibility that they have diverged widely since descent from a shared venomous ancestor (7).

Despite the dissimilarity in chemical composition, the venoms of Limacodidae and Megalopygidae nevertheless show convergent activity, with both venoms causing pain in mammals by permeabilizing cell membranes. In Limacodidae, this is due to the action of small (<10 kDa) cecropin-like peptides (5), whereas in Megalopygidae, the membrane-permeabilizing proteins are larger (>20 kDa) megalysins. The evolution of the same venom activity but by recruitment of distinct classes of innately toxin-like endogenous molecules illustrates the diverse endogenous substrates from which venom toxins may evolve. Notably, however, there are indications that the mechanism of membrane permeabilization by megalysins is different from that of peptides such as cecropin-like toxins. Limacodid cecropin-like venom toxins and many hymenopteran membrane-permeabilizing venom toxins typically produce increases in [Ca²⁺]_i that are transient in nature (5, 55), which we attribute to formation of large pores or drastic disruption of the cell membrane that allows quencher in the extracellular medium to enter the cytoplasm and extinguish fluorescence. In contrast, asp caterpillar venoms induce sustained increases in [Ca²⁺]_i even in the presence of extracellular quencher, suggesting formation of smaller, discrete pores through the membrane that allow passage only of ions and small molecules.

Asp caterpillar venoms display insecticidal and anticoagulant activities. Our finding that the venom is anticoagulant rather than procoagulant as previously reported (8) may reflect the use of a mechanical coagulometer in this study compared to absorbance-based methods used in the previous investigation. However, the insecticidal and anticoagulant activities observed occurred at moderately high concentrations (µg/g and µg/mL), i.e., six orders of magnitude higher than concentrations able to activate mammalian neurons (pg/mL). These results are consistent with asp caterpillar venoms being primarily adapted to cause pain and therefore deter predators, rather than inducing death or incapacity in invertebrates, or coagulotoxicity in vertebrates (1, 2, 10).

Our data suggest that the membrane-permeabilizing activity of Megalopyge venoms is due to large pore-forming proteins, the megalysins. The megalysins have been recruited from aerolysin-like proteins which occur in Lepidoptera due to horizontal gene transfer from bacteria. Aerolysin family proteins including the megalysins have been recruited convergently into the venoms of megalopygid caterpillars, centipedes (20), cnidaria, and fish (38). In the case of centipedes, it has been suggested based on phylogenetic analysis that the aerolysin family β-pore-forming toxins have been transferred not once but twice from bacteria to centipedes (20). The evolution of fish aerolysin family proteins, the natterins, is complex but also likely to involve horizontal gene transfer events (41, 56). Notably, megalysins are not closely related to centipede β-pore-forming toxins, fish natterins, or natterin-like proteins detected in insect genomes (56). Aerolysin-like proteins are typically categorized as toxins, so it is perhaps surprising that we found that genes encoding them were horizontally transferred early in lepidopteran evolution and retained and duplicated in nonvenomous Lepidoptera for millions of years prior to their recruitment as venom toxins in the Megalopygidae. Indeed, this finding is consistent with the broad distribution of the aerolysin protein family in both venomous and nonvenomous organisms (21, 38) and the innately toxin-like properties of many endogenous proteins (15), particularly in some functional classes such as immune effectors. The megalysins highlight the contribution of horizontal gene transfer to the evolution of venom toxins.

Materials and Methods

Detailed materials and methods are provided in SI Appendix, Materials and Methods. In brief, light microscopy and μ-CT of fixed tissues were used to investigate the structure of the venom apparatus. RNA-Seq of the venom scoli combined with bottom–up and top–down MS of venom was used to resolve a venom proteome. Functional characterization was carried out on either untreated venom or venom treated with heat, proteases, or acetonitrile. Peptides produced by recombinant expression or SPPS were also tested. Functional assays included calcium imaging of primary peripheral sensory neurons and the neuroblastoma cell line SH-SY5Y, as well as insecticidal and anticoagulation assays. Sequence analysis of megalysins was carried out using Bayesian phylogenetics and by comparison to known sequences.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Movie S1.

Animation showing sweep through Z-stack of transverse μ-CT slices through a venom scolus of larval M. crispata. Note the venom canals transversing the cuticle and terminating at the X-ray dense end-apparatuses of the secretory cells.

Data, Materials, and Software Availability

Amino acid and nucleotide sequences for Megalopyge sp. venom proteins and their nucleotide coding sequences were submitted to GenBank and allocated the accession numbers OP514844–OP514929 (57). Raw sequencing reads were submitted to GenBank’s Sequence Read Archive with accession PRJNA884172 (58), and assembled transcriptomes were submitted to GenBank’s Transcriptome Shotgun Assembly database with accessions GKBX00000000 (M. opercularis) (59) and GKBZ00000000 (M. crispata) (60). The versions described in this paper are the first versions, GKBX01000000 and GKBZ01000000. The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD037610 (61).