Are Fireworms Venomous? Evidence for the Convergent Evolution of Toxin Homologs in Three Species of Fireworms (Annelida, Amphinomidae)

Abstract

Amphinomids, more commonly known as fireworms, are a basal lineage of marine annelids characterized by the presence of defensive dorsal calcareous chaetae, which break off upon contact. It has long been hypothesized that amphinomids are venomous and use the chaetae to inject a toxic substance. However, studies investigating fireworm venom from a morphological or molecular perspective are scarce and no venom gland has been identified to date, nor any toxin characterized at the molecular level. To investigate this question, we analyzed the transcriptomes of three species of fireworms-Eurythoe complanata, Hermodice carunculata, and Paramphinome jeffreysii-following a venomics approach to identify putative venom compounds. Our venomics pipeline involved de novo transcriptome assembly, open reading frame, and signal sequence prediction, followed by three different homology search strategies: BLAST, HMMER sequence, and HMMER domain. Following this pipeline, we identified 34 clusters of orthologous genes, representing 13 known toxin classes that have been repeatedly recruited into animal venoms. Specifically, the three species share a similar toxin profile with C-type lectins, peptidases, metalloproteinases, spider toxins, and CAP proteins found among the most highly expressed toxin homologs. Despite their great diversity, the putative toxins identified are predominantly involved in three major biological processes: hemostasis, inflammatory response, and allergic reactions, all of which are commonly disrupted after fireworm stings. Although the putative fireworm toxins identified here need to be further validated, our results strongly suggest that fireworms are venomous animals that use a complex mixture of toxins for defense against predators.

Keywords: Amphinomidae; Annelida; fireworms; transcriptomics; venom toxins; venomics.

Fig. 1.

—Phylogeny of the family Amphinomidae and venomous annelid representatives. Phylogenetic tree of the family Amphinomidae with the genera that include the three fireworm species investigated here, highlighted in bold. Cladogram based on phylogenetic reconstruction from Borda et al. (2015). Images of the fireworm species focus of this study are shown to the right: Paramphinome jeffreysii (A); Eurythoe complanata (B); and Hermodice carunculata (C). Fireworm species are grouped with earthworm (D) Eisenia foetida based on their defensive use of toxins and highlighted in yellow. Representatives of species that use venom for predation are also shown, highlighted in blue: bloodworm Glycera capitata (E) and leech Helobdella europaea (F). Images courtesy of Arne Nygren (A), Denis Riek (B), Arthur Anker (C), and Alexander Semenov (E). Remainder images (F and D) are publicly available under Creative Commons License.

Fig. 2.

—Venomics pipeline for the identification of toxin homologs. Diagram of the bioinformatics pipeline followed to identify putative venom components in fireworms. The different steps are highlighted with colored ovals and the corresponding software used indicated below each step. After RNA extraction and sequencing, the first step is data processing (green) and de novo assembly of the transcriptomes. Subsequently, a gene prediction and filtering step (gray) is performed, in which contigs are translated into amino acids, and open reading frames and signal sequences are predicted. The following step is a homology search (blue) using three different search strategies based on BLAST, HMMER-sequence and HMMER-domain. The results are merged and the putative toxins validated (maroon) through BLAST and phylogenetic reconstructions to generate a final list of candidate toxins (yellow).

Fig. 3.

—Diversity of toxin homologs identified in each fireworm species. Pie charts show the diversity of toxin homologs identified in each transcriptome. Toxin homologs are grouped by toxin class and numbers inside the pie chart indicate the relative abundance of each toxin class, as the percentage of the total number of toxins (i.e., 261 in Hermodice carunculata, 438 in Eurythoe complanata, and 169 in Paramphinome jeffreysii). The number of transcripts corresponding to each toxin class is indicated in parenthesis. Abbreviations are as follows: CLEC, C-type lectin; M12, metalloproteinase M12; PEP S1, peptidase S1; PEP S10, peptidase S10; PL, phospholipase; ShK, ShKT-domain containing peptides; AChE, acetylcholinesterase; SMase, sphingomyelinase. Other, groups a variety of less diverse toxin homologs (see supplementary files S2–S4, Supplementary Material online).

Fig. 4.

—Orthologous toxins in fireworms and associated functions. Alluvial diagram showing the 253 orthologous toxins identified in the three fireworm species in the first node, the toxin class to which they belong in the second node, and the associated biological function in the third node. The relative abundance of toxin homologs corresponding to each category (species, toxin class, and function) is shown below the name as a percentage of the total. The number of transcripts that correspond to each category is shown in parenthesis. Abbreviations are as follows: CLEC, C-type lectin; M12, metalloproteinase M12; PEP S1, peptidase S1; PEP S10, peptidase S10; PL, phospholipase; ShK, ShKT-domain containing peptides.

Fig. 5.

—Phylogenetic tree of cystatin sequences. Cystatin phylogeny including sequences from venomous and nonvenomous taxa and fireworm homologs. Sequences from nonvenomous taxa are indicated with closed circles and fireworm homologs are highlighted in bold. Tree reconstructed with RAxML-HPC-PTTHREADS v8.2.10 and support values estimated through a rapid bootstrap algorithm and 1,000 pseudoreplicates. Bootstrap support values are given for all nodes and clade names are indicated by colored squares.

Fig. 6.

—Phylogenetic tree of lipocalin sequences. Lipocalin phylogeny including sequences from venomous and nonvenomous taxa and fireworm homologs. Sequences from nonvenomous taxa are indicated with closed circles and fireworm homologs are highlighted in bold. Tree reconstructed with RAxML-HPC-PTTHREADS v8.2.10 and support values estimated through a rapid bootstrap algorithm and 1,000 pseudoreplicates. Bootstrap support values are given for all nodes and clade names are indicated by colored squares.

Fig. 7.—

Phylogenetic tree of Kazal-type protease inhibitors. Kazal-type inhibitor phylogeny including sequences from venomous and nonvenomous taxa and fireworm homologs. Sequences from nonvenomous taxa are indicated with closed circles and fireworm homologs are highlighted in bold. Tree reconstructed with RAxML-HPC-PTTHREADS v8.2.10 and support values estimated through a rapid bootstrap algorithm and 1,000 pseudoreplicates. Bootstrap support values are given for all nodes and clade names are indicated by colored squares.

Fig. 8.

—Phylogenetic tree of metalloproteinase M12 sequences. Metalloproteinase M12 phylogeny including sequences from venomous and nonvenomous taxa and fireworm homologs. Sequences from nonvenomous taxa are indicated with closed circles and fireworm homologs are highlighted in bold. Tree reconstructed with RAxML-HPC-PTTHREADS v8.2.10 and support values estimated through a rapid bootstrap algorithm and 1,000 pseudoreplicates. Bootstrap support values are given for all nodes and clade names are indicated by colored squares.

Fig. 9.

—Phylogenetic tree of ShKT-domain containing peptides. ShKT-domain containing peptide phylogeny including sequences from venomous and nonvenomous taxa and fireworm homologs. Sequences from nonvenomous taxa are indicated with closed circles and fireworm homologs are highlighted in bold. Tree reconstructed with RAxML-HPC-PTTHREADS v8.2.10 and support values estimated through a rapid bootstrap algorithm and 1,000 pseudoreplicates. Bootstrap support values are given for all nodes and clade names are indicated by colored squares.

Fig. 10.

—Multiple sequence alignment of turripeptide-like sequences. Multiple sequence alignment of turripeptides and fireworm homologs generated with MAFFT v7.310. Conserved residues are colored, including the conserved cysteines, which are highlighted in yellow. The predicted signal sequence is delimited by a yellow square and the Kazal domain is delineated by a purple square.