Crown‐of‐thorns sea star (CoTS) outbreaks are a main cause of hard coral cover decline across the Indo‐Pacific, posing a major threat to the resilience of coral reefs. However, the drivers underlying CoTS feeding on preferred (e.g.,
Crown-of-thorns sea stars (CoTS) are found throughout the Indian and Pacific Oceans, as well as the Red Sea and the Gulf of Oman (Pratchett et al. 2014). Throughout these locations, four to five CoTS species have been identified, which comprise the Acanthaster species complex: including the Northern Indian Ocean species Acanthaster planci sensu strictu; the Red Sea species Acanthaster benziei; and the Southern Indian Ocean species Acanthaster mauritiensis (Vogler et al. 2008;Haszprunar et al. 2017;Wörheide et al. 2022;Foo et al. 2024;Uthicke et al. 2024). The taxonomy of the Pacific species is uncertain with Acanthaster cf. solaris as a potential name for the western Pacific species and Acanthaster cf. ellisii in the eastern Pacific (Uthicke et al. 2024). These sea stars are corallivorous (review Foo et al. 2024). At low densities (1-10 sea stars per hectare; Pratchett et al. 2014;Dumas et al. 2016), CoTS are argued to be beneficial to reefs, as they preferentially prey on faster-growing coral species (e.g., Acropora species), creating more space for slower-growing species (e.g., Porites species), and thus, increase coral diversity (Done and Potts 1992;Bellwood et al. 2024). In contrast, high-density 'outbreak' populations of CoTS where densities can reach up to 15,000 sea stars per hectare can decimate reefs (Dumas et al. 2022). Each CoTS is able to consume 5-12 m 2 of coral surface annually (Chesher 1969;Pearson and Endean 1969;Dana and Wolfson 1970), and outbreak populations spread at speeds of up to 60 km per year (Reichelt et al. 1990;Vanhatalo et al. 2017). These outbreaks drive significant declines in coral cover (Osborne et al. 2011;De'Ath et al. 2012), compounding the impacts of heatwave-induced bleaching, which has now become the primary driver of coral mortality (Bozec et al. 2022;Byrne et al. 2025).
Early post-settlement, CoTS juveniles are herbivores with a preference for crustose coralline algae (CCA) and under favourable conditions, they undergo an ontogenetic diet transition to coral prey when they reach around 8 mm diameter (Yamaguchi 1974). After this ontogenetic diet transition, their size rapidly increases (Zann et al. 1987;Deaker et al. 2020). In order to locate their coral prey, CoTS use olfactory sensing (Ormond et al. 1976;Ling et al. 2020;Webb et al. 2024), with the densities of preferred coral prey shown to influence the behaviour of CoTS to either vacate or stay on a reef (Ling et al. 2020). Where Acropora and pocilloporid corals (e.g., Seriatopora and Stylophora species) are available, CoTS exhibit a preference for these as prey (De Bruin 1972;Ormond et al. 1973;Keesing 1990Keesing , 2021;;De'Ath and Moran 1998;Johansson et al. 2016;Foo et al. 2024). They also grow faster on a diet of these corals (Keesing and Halford 1992; J. K. Keesing 2021). Porites species are among the least preferred prey of CoTS (De'Ath and Moran 1998;Pratchett 2007;Kenyon and Aeby 2009;Millican et al. 2024) and CoTS have minimal growth rates on a diet of these corals (Keesing 2021). Even at very low abundances of preferred coral prey (Acropora and pocilloporid corals), CoTS still show a preference for these species, albeit with increased consumption of non-preferred coral taxa (Keesing et al. 2019;Keesing 2021).
It is still unclear why CoTS show feeding preferences for certain coral taxa. Hypotheses include: nutritional differences between corals (Ormond et al. 1976;Keesing 1990); accessibility to coral tissue based on coral morphotype (e.g., tabular vs. branching) and ease of tissue digestion (Keesing 1990); coral symbionts that repel CoTS (shrimps, gobies, trapeziid crabs) (Glynn 1980;Pratchett 2001; also see Montano et al. 2017); and, coral nematocyst and venom defence (Barnes et al. 1970;Ormond et al. 1976;Deaker et al. 2021). Recent studies are starting to unveil the interplay between CoTS and coral prey choice. For instance, a laboratory study concluded that CoTS' preference for Acropora species is not driven by growth form or nutritional content, but rather by prior exposure to that coral species during rearing (Johansson et al. 2016). This suggests that feeding preference may involve a conditioning process, potentially akin to an immune-type response. Furthermore, as CoTS mature they increase their ability to feed on a wider range of corals, including those that can cause damage to them as juveniles (Johansson et al. 2016). These findings support the hypothesis that CoTS feeding preferences are influenced by their ability to tolerate or adjust to coral venom defences as a result of previous exposure. This is particularly relevant given the dose-dependent nature of venom bioactivity, which may explain why larger juveniles can diversify their diet (Deaker et al. 2021), as venom is less likely to cause lethal injuries at this stage.
Coral venom and toxins provide protection against predation (Gochfeld 2004). For instance, a toxin isolated from the soft coral Sarcophyton glaucum, sarcophine, deters predatory corallivorous fishes (Ne'eman et al. 1974). Moreover, the coral Echinopora lamellosa has shown to cause 100% mortality of CoTS post-ingestion (Johansson et al. 2016). Although it is unclear how E. lamellosa invoked CoTS mortality, venom toxins may contribute to a coral's defence, causing lethal and sub-lethal injuries to CoTS during contact and ingestion. Furthermore, CoTS exhibit a humped posture during feeding to protect their tube feet, which are more vulnerable to damage from coral nematocysts than the stomach (Barnes et al. 1970). On smaller coral colonies specifically, this humped posture allows the sensitive tube feet to remain on a non-nematocyst surface (e.g., calcareous rock), whilst exposing the less-sensitive stomach to the coral (Barnes et al. 1970). This hypothesis is further supported by observations of discharged nematocysts on CoTS tube feet after exposure to coral (Moore unpublished, as cited in Moore and Huxley 1976). Furthermore, venom defence from other cnidarian taxa such as hydrozoans may provide protection for the coral against CoTS; for example, neighbouring Millepora species have been shown to produce the strongest aversion reaction in CoTS (Moore and Huxley 1976) and may provide refuge for Acropora species, against CoTS predation (Kayal and Kayal 2017). It has also been suggested that the venom from Acropora digitifera has a defensive role against predators due to the low diversity and abundance of toxins present in the venom (Gacesa et al. 2015). This contrasts with predatory venoms, which have a larger suite of toxins due to high evolutionary selection pressure acting upon the venom proteins allowing them to be utilised for predation across a wide variety of prey taxa (Gacesa et al. 2015). These studies indicate that investigating the venom toxin repertoire of CoTS prey may be imperative to understanding their resilience against CoTS attack.
Scleractinian venoms are understudied compared with those of other venomous cnidarians such as jellyfish and sea anemones, for which the biochemistry and molecular biology of the bioactive peptides are well understood (Gacesa et al. 2015;Jouiaei et al. 2015;Schmidt et al. 2019;Klompen et al. 2022). For instance, the venom of sea anemones contains one of the most diverse arrays of toxin polypeptides found across the animal kingdom (Castañeda and Harvey 2009;Frazão et al. 2012;Orts et al. 2013;Finol-Urdaneta et al. 2020). Studies of scleractinian venoms show that they are all rich in biologically active peptides (Gacesa et al. 2015;Garcia-Arredondo et al. 2016;Ben-Ari et al. 2018;Yosef et al. 2020;Drake et al. 2021;Schmidt et al. 2022), with significant similarities in the major classes of toxins present across Cnidaria (Rachamim et al. 2015;Gacesa et al. 2015). For instance, some commonly found toxins and peptide/protein families in cnidarians include: pore-forming toxins (PFTs; e.g., aerolysins, actinoporins and Membrane Attack Complex/Perforin toxins [MAC-PF]), neurotoxins (e.g., K V /Na V channel neurotoxins and the peptidic neurotoxic small cysteine-rich proteins [SCRiPS]), and cytotoxic enzymes (e.g., peptidases, metalloproteases and phospholipase A2s [PLA2]). Therefore, in this study we used known cnidarian toxins (from the literature and the cnidarian filtered UniProtKB/Swiss-Prot Tox-Prot database; Jungo et al. 2012) to BLAST against our genomic databases. This method allowed us to investigate the abundance and diversity of toxins in the genomes of four species of CoTS preferred prey (A. digitifera, A. hyacinthus, A. millepora and A. tenuis) and four species of non-preferred prey (P. australiensis, P. compressa, P. lutea and P. rus). We also included one species from each genus (A. cervicornis and P. astreoides) that are native to the Caribbean, where CoTS are absent, to increase our dataset for toxins in each genus and also investigate geographic differences. Our aim was to determine if there are differences in the toxin repertoire of CoTS preferred and non-preferred coral prey species to clarify the natural defences that corals have access to during a CoTS outbreak. Knowledge of scleractinian venoms could be used to inform the design of effective management measures to combat CoTS outbreaks for example, planting more venomous corals across at-risk reefs. Recently, a study has already shown that molecular biology methods can be harnessed to produce biotools that could be utilised during CoTS outbreaks (Harris, Hillberg, et al. 2025). Indeed, CoTS outbreaks are relatively amenable to control in contrast to climate warming, sparking considerable interest in developing intervention tools to suppress this species (Babcock et al. 2020;Harris, d'Artagnan, et al. 2025).
Cnidarian genome databases were acquired from publicly available sources (Table 1). One species from each of the two coral genera included in this current study was native to the Caribbean where CoTS is absent (A. cervicornis and P. astreoides; Table 1). These Caribbean species were included to observe any patterns between geographic regions where corals are exposed and not exposed to CoTS. We also investigated the Echinopora lamellosa transcriptome (NCBI BioProject: PRJNA602211; Quek et al. 2020), however this returned no significant hits to any toxins, venom proteins or the UniProtKB/ Swiss-Prot Tox-Prot database, and thus, was not included in this study.
Toxins were identified following described methods (Brinkman et al. 2015;Gacesa et al. 2015;Rachamim et al. 2015). Briefly, a literature review of cnidarian toxins was conducted and from this, known toxin and venom peptide/protein families from cnidarians were acquired from the NCBI and UniProt databases (when available) and from the relevant literature. The characterised toxins were then searched against InterProScan for their conserved domains and the peptides were split into groups based on their function and conserved domain sequence (Data S1: Table A). Toxin orthologs in each group were then used as queries to search against the cnidarian databases using BLASTp or
TABLE 1 | Genomic databases used for toxin investigations. Taxa shown in descending order from the preferred (green) prey of CoTS to the not preferred (blue).
tBLASTn searches. Hits from the Acropora and Porites genomes were kept if they met the quality criteria (query cover ≥ 70% and e-value < 1 × 10 -5 ) and passed manual quality control (see Section 2.4). Hits that passed quality control were subsequently searched against the InterProScan database to find conserved domains and also BLASTed against the NCBI nr database where the top hit to a cnidarian protein was collected (Tables S1-S10). The hit collected was based on the hit scoring the lowest e-value that also harboured a query cover ≥ 70% that was not annotated as a predicted or hypothetical protein.
In addition to searching the known cnidarian venom constituents and toxins, we filtered the UniProtKB/Swiss-Prot Tox-Prot database to toxins and venom proteins only found in Cnidaria. This equated to 341 venom/toxin proteins at the date of access (accessed: 20.05.25). We then BLASTed this database against the coral genomes and collected the top hit, corresponding to the lowest e-value. Only hits with an e-value of 1 × 10 -5 and > 70% query cover were retained. Hits then underwent quality control (see Section 2.4). After passing quality control, gene ontology terms were collected from the homologues from the UniProtKB/ Swiss-Prot Tox-Prot database and hits were searched against the InterProScan database to find conserved domains (Tables S11 and S12).
The ToxCodAn-Genome package (Nachtigall et al. 2024) was used to annotate toxin CDS in the coral genomes (Table S14). Briefly, the annotation database was constructed using the pre-constructed anthozoan database in ToxCodAn-Genome, in addition to manually adding several cnidarian-specific toxins that were not already present in this database (e.g., CFXs and SCRiPS; Data S1: Table A). Sequences from the NCBI Entrez database that contained both the strings 'cnidarian' and 'toxin' were also added to the annotation database. This resulted in an annotation database containing 1620 sequences (Data S2). Any annotation hits from the coral genomes to a toxin CDS were also BLASTed to the cnidarian-filtered UniProtKB/Swiss-Prot Tox-Prot database. Strict and relaxed parameters were tested during annotation (strict: 80% protein identity; 50 bp minimum length; and 50 bp minimum gene size; relaxed: 20% protein identity; 20 bp minimum length; and 20 bp minimum gene size).
Sequences that contain stop codons are detailed in the supplemental results tables as potential pseudogenes (Tables S1-S10). Additionally, sometimes duplicated hits from the same genome were returned. Therefore, duplicated hits were removed if they had 100% identity in the mature toxin region or > 98% identity in the whole toxin sequence. It should be noted however, that duplicate copies of nearly identical toxin genes and proteins may in fact highlight their importance during envenomation. For instance, as a single nematocyst must be replaced after discharging its venom, duplicate copies of a toxin gene may speed up the production of the toxin (Moran et al. 2009).
Phylogenetic trees were constructed for the collected putative toxins from Acropora species and Porites species in the actinoporin, MAC-PF, CFX and SCRiP families. Our putative toxins were aligned with previously confirmed toxins in these families (Data S1: Table A) using MUSCLE in Geneious Prime (version 2024.0.7). Alignments were then manually edited and trimmed to the corresponding domain or conserved region (see Data S1 for further information). Alignments were then submitted to IQ-tree (Trifinopoulos et al. 2016) to obtain the correct model of amino acid evolution via ModelFinder (Kalyaanamoorthy et al. 2017). Using the corrected Akaike information criterion (AICc) the best-fit model of evolution was obtained; maximum likelihood trees were generated in IQ-tree using 1000 ultrafast bootstrap replicates (UFBoot) and SH-aLRT for branch support. Trees were rooted with sister proteins from the same superfamily or homologues from distantly related species (see Data S1 for further information). Trees were edited and visualised in Interactive Tree of Life (iTOL; version 6; Letunic and Bork 2024).
Jellyfish JFT-1b homologues in Porites and the true MAC-PF homologues across Acropora and Porites were investigated for episodic positive selection. Briefly, nucleotide sequences were retrieved from the true MAC-PF and JFT-1b clades constructed from phylogenetic analysis. Nucleotide alignments were aligned with the true nucleotide sequences of confirmed MAC-PF or JFT-1b proteins in other organisms (sea anemones and jellyfish, respectively) using MUSCLE in Geneious Prime (version 2024.0.7). Alignments were then submitted to HyPhy (version 2.2.4; Kosakovsky Pond et al. 2005) to clean the stop codons. Once stop codons were removed from the alignment, the alignment (and its subsequent phylogenetic tree) was inputted into a GTR mutation model using the MG94xREV model in HyPhy to conduct a maximum likelihood analysis. Maximum likelihood analysis estimated and optimised branch lengths and subsequently calculated the global alignment ω value. Following this, a Branch-site Unrestricted Statistical Test for Episodic Diversification (BUSTED) analysis (Murrell et al. 2015) was conducted to test for gene selection (Table S13) using the DataMonkey server (Weaver et al. 2018).
Both our custom BLAST search and discovery bioinformatics methods revealed unique proteins that were not found using the alternative method (Tables S1-S11). The venom-specific proteins/toxins that returned the most paralogs across Acropora and Porites species using the discovery bioinformatics pipeline were Kunitz-type neurotoxins, actinoporins containing the anemone cytotox domain, MAC-PFs, and SCRiPS (Figure 1; Table S12). Metalloproteinases and PLA2s also returned a high number of paralogs but these enzymes may harbour non-toxic functions; thus caution must be taken interpreting the abundance of these protein families until functional studies can confirm which are involved in envenomation (Figure 1, Tables S9 and S10). Porites compressa was the species observed to have the most diverse venom proteins (using the genomic resources available), only lacking homologues to two of the venom protein groups-homologues with an aerolisin/ETX pore-forming domain and a homologue to a potential neurotoxin with the zinc finger and iron-sulphur binding domain (Table S12). Contrastingly, Porites astreoides had the least diverse putative venom of the species we investigated, only harbouring homologues to actinoporins, MAC-PF and Kunitz-type neurotoxins (Table S12).
The venoms between Acropora and Porites species were largely made up of protein families with similar abundance. The only venom protein family that differed significantly between these genera was SCRiP neurotoxins, with Acropora species harbouring significantly more SCRiPS in their venom than Porites species (Student's t-test, t(6.3083) = 2.539, p = 0.042; Figure 1). However, the large amount of potential pseudogenes in Acropora species skewed this result, as when the potential SCRiPS pseudogenes across all species were omitted, there was no difference in the number of SCRiPS toxins between Acropora and Porites species (Mann-Whitney-U test, U = 14, Z = 0.9439, p = 0.3452). Despite the similarity in the abundance of venom proteins, there were several differences in the presence/absence and phylogeny of venom protein families between and within Acropora and Porites species (Figures 23456).
In addition to SCRiPS neurotoxins, we also found several potential pseudogenes across our actinoporins, MAC-PFs, metalloproteases, kunitz-like neurotoxins and SCRiPS (Tables S1, S3, S5, S6 and S10) that harboured premature introns. Toxin pseudogenes have been reported for other venomous cnidarians for example, sea anemones (Sachkova et al. 2019) and represent a type of venom evolution through birth-and-death mechanisms where toxin genes lose their function (Nei and Rooney 2005;Sachkova et al. 2019). Most of the potential pseudogenes we found across our coral species were in the SCRiPS neurotoxin gene family (Table S6). In snake venoms, gene function loss via pseudogenes is hypothesised to be from shifts in ecological niches and contributes to the species-specific venom diversity (Dowell et al. 2016). Whether the potential pseudogenes found in this study correspond to functional diversity across scleractinian venoms is interesting and requires future attention.
Our ToxCodAn-Genome analysis (both strict and relaxed parameters) only returned a hit to one cnidarian toxin gene (C0H691; SCRiP2 from Acropora millepora) in the A. hyacinthus, A. millepora and A. digitifera genomes. This contrasts with our custom BLAST search and discovery bioinformatics methods using the toxin protein sequences. The difference in results likely reflects the divergence of toxin genes at the nucleotide level across different species in scleractinian venoms. This further emphasises the need for combining genome sequencing with tailored venom gene expression studies for each scleractinian species.
Our custom BLAST searches and discovery bioinformatics found 70 putative actinoporins across the Acropora and Porites species FIGURE 1 | Putative coral venom protein families found across each coral species identified by custom BLAST searches and discovery bioinformatics (Tables S1-S11).
s n i e t o r p f o r e b m u N Actinoporins Cytolysins MAC-PFs Aerolysins CFXs Kunitz-type Damicornin-like Defensin-like SCRiPS Metalloproteases Phospholipase A2 Unique toxins 0 50 100 150 200 250 Acropora cervicornis Acropora digitifera Acropora hyacinthus Acropora millepora Acropora tenuis Porites astreoides Porites australiensis Porites compressa Porites lutea Porites rus investigated (Tables S1 and S11, Figure 3). Phylogenetic analysis revealed that these putative actinoporins were present across seven of the ten actinoporin phylogenetic groups (Figure 3). Three actinoporin hits exclusively from Porites species (one hit from each of P. compressa, P. lutea and P. australiensis) hit to Nematostella vectensis tereporin-like actinoporin (XP_001633907) from the UniProt ToxProt database and all were missing the anemone cytotox domain (PFAM ID: PF06369) but contained the bryoporin (PANTHER ID: PTHR40388) and cytolysin/lectin (SUPERFAMILY ID: SSF63724) domains. Unsurprisingly, these same proteins grouped with Nematostella vectensis tereporinlike actinoporin in our phylogenetic tree making a distinct actinoporin clade we named 'Cnidarian tereporin-like actinoporins' (99.8/100% SH-aLRT/UFBoot support; Figure 3). This contrasts with the other 67 putative actinoporins, which all contained the anemone cytotox domain (PFAM ID: PF06369; Table S1). None of our coral species' putative actinoporins were related to the confirmed scleractinian haemolytic actinoporin, Δ-Pocilopotoxin-Spi1 from Stylophora pistillata (Ben-Ari et al. 2018; 'Scleractinia actinoporin group 1' Figure 3). However, two putative actinoporins (one from A. tenuis and A. millepora) formed a sister group to 'Scleractinia actinoporin group 1' (89.7/81% SH-aLRT/UFBoot support; Figure 3). These same proteins met the quality criteria for hydralysins and they contained the aerolysin/ETX pore-forming domain (SUPERFAMILY ID: SSF56973; Table S2). This suggests that these actinoporins were distantly related to hydralysins. Thus, a total of 72 actinoporins were found in our study across Porites and Acropora species. One putative actinoporin from A. hyacinthus did not group phylogenetically with any other actinoporins we used in our tree (group represented by '∆' in Figure 3) but did contain the anemone cytotox domain and when BLASTed against the NCBI nr database, hit to the DELTA-actitoxin-Aeq1b-like isoform X3 (XP_067055194.1) in Acropora muricata. Black boxes represent the presence of protein found in that coral species. Actinoporin, MAC-PF, CFX and SCRiP group terminology on the y-axis is informed by phylogenetic analysis conducted in our current study . Asterisks (*) refers to actinoporin phylogenetic groups with low support (blue font; Figure 3). Hashed boxes represent proteins that were found in a previous study (Barroso et al. 2024) but not in our current study. Graph created in BioRender (Gorman 2025; https://BioRender.com/ hbygxgm) and figure edited in InkScape. Images of CoTS and coral species downloaded from the media library at the Integration Network University of Maryland Centre for Environmental science (https:// ian. umces. edu/ media -libra ry/ ). FIGURE 3 | Legend on next page. 6681| Acropora palmata KAK2570228.1| Acropora cervicornis XP_044167121.1| Acropora millepora XP 044166976.1| Acropora millepora BLAZ01000183.1| Acropora tenuis BLFC01000653.1b| Acropora digitifera CM035876.1| Acropora hyacinthus BLFC01000152.1| Acropora digitifera 9179| Platygyra carnosus 9182| Platygyra carnosus 4876| Montastraea faveolata 34536| Montastraea cavernosa 60790| Anthopleura elegantissima BLAZ01001350.1| Acropora tenuis BLFC01000653.1a| Acropora digitifera CM035869.1a| Acropora hyacinthus XP 029209891.2| Acropora millepora BLFC01000770.1| Acropora digitifera BLAZ01000569.1f| Acropora tenuis KAK2571155.1| Acropora cervicornis CM035869.1b| Acropora hyacinthus XP_044182610.1| Acropora millepora BLAZ01000569.1d| Acropora tenuis BLAZ01000569.1c| Acropora tenuis BLAZ01000569.1e| Acropora tenuis BLAZ01000569.1a| Acropora tenuis BLAZ01000569.1b| Acropora tenuis 1589034| Equinatoxin II Actinia equina ACTPC_ACTFR| Δ-actitoxin-Afr1a/Fragaceatoxin C Actinia fragacea AAD39836.1| Equinatoxin IV Actinia equina ACTP5_ACTEQ| Δ-actitoxin-Aeq1b/Equinatoxin V Actinia equina 10190| Anthopleura elegantissima ACTP1_ANTAS| Δ-actitoxin-Aas1a/Bandaporin Anthopleura asiatica ACTP1_URTCR| Δ-actitoxin-Ucs1a/urticinatoxin Urticina crassicornis ACTP1_STIHL| Δ-stichotoxin-She4a/Sticholysin I Stichodactyla helianthus ACTP2_STIHL| Δ-stichotoxin-She4b Stichodactyla helianthus ACTPA_HETCR| Δ-stichotoxin-Hcr4a Heteractis crispa ACTPG_OULOR| Δ-actitoxin-Oor1b Oulactis orientalis ACTP1_PHYSE| Δ-alicitoxin-Pse1a/PsTX-20A Phyllodiscus semoni ACTP2_ACTVL| Δ-thalatoxin-Avl1b Actineria villosa 9088| Favia sp ACTP1_SAGRO| Δ-sagatoxin-Srs1a/Src-1 Sagartia elegans BOPM01000008.1| Porites australiensis ENA|OKRP01000879| Porites rus 9870| Stylophora pistillata 91581| Seriatopora hystrix 5837| Pocillopora damicornis 52714| Madracis auretenra 9725| Δ-Pocilopotoxin-Spi1 Stylophora pistillata 30437| Fungia scutaria 24700| Pocillopora damicornis BLAZ01000454.1| Acropora tenuis XP 044182936.1| Acropora millepora CM035873.1| Acropora hyacinthus ACTL1_HYDVU| HALT-1 Hydra vulgaris ACTL6_HYDVU| HALT-6 Hydra vulgaris ACTL4_HYDVU| HALT-4 Hydra vulgaris ACTL2_HYDVU| HALT-2 Hydra vulgaris ACTL3_HYDVU| HALT-3 Hydra vulgaris ACTL5_HYDVU| HALT-5 Hydra vulgaris HIv1_RNAseq.g20438.t1| Porites compressa plut2.m8.24321.m1| Porites lutea BOPM01001304.1| Porites australiensis 242539| Nematostella vectensis XP_001633907| tereporin-Ca1 Nematostella vectensis AAV65396| Physcomitrin Physcomitrium patens Tree scale: 1 100/100 53.6/83 89.7/99 56.3/80 87.5/84 23.7/90 27.7/57 0/66 99.9/100 89.7/81 92.9/100 80.4/81 59.3/99 91.7/99 80.6/77 99.6/99 81.5/99 99.8/100 66.2/95 56.3/65 57.6/58 100/100 99/100 92.4/98 91.6/100 94.8/98 65.6/53 90.7/87 35.8/53 61.1/60 99/100 85.7/80 96.1/61 94/100 100/100 97.3/99 68.8/63 98.5/99 0/56 83.5/98 68.2/41 61.9/94 0/46 0/45 61.8/77 0/87 100/100 96.5/99 82.2/39 91.2/100 56.4/98 84.7/100 43.7/46 90.4/84 95.2/100 99.7/100 86/50 32.1/93 6.4/84 98.1/56 91.5/38 92.3/92 82.6/97 99.7/100 80.3/42 63/95 99.9/100 78.6/92 85.1/86 96.2/99 46.7/91 84/97 78.3/86 92.4/98 0/56 29.4/45 87.4/99 76.9/80 94.6/100 99.2/100 100/100 76.7/100 99/100 99.8/100 90.4/89 0/59 92.2/95 92.3/78 76.6/99 89.2/97 25.7/90 99.8/100 92.3/99 71.8/99 78.7/100 79.2/95 75.8/47 85/100 83.7/100 83.6/100 98.5/100 99.9/100 60.3/83 82.1/99 87.7/100 71.9/50 81.5/98 Cnidarian tereporin-like actinoporins Scleractinian actinoporin group 1 Anthozoan actinoporin group 1 Acropora-specific actinoporin group 1 Anthozoan actinoporin group 2 Scleractinian actinoporin group 2 Scleractinian actinoporin group 3 Porites-specific actinoporin group 1 Porites-specific actinoporin group 2 Hydralysins Δ * † † 1365294x, 2026, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mec.70202 by University Of Rhode Island, Wiley Online Library on [01/03/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License FIGURE 3 | Maximum likelihood tree of actinoporins across Cnidaria. Putative actinoporins from the current study are denoted in green (Acropora species) and brown (Porites species). Model of evolution and tree constructed in IQ tree (Trifinopoulos et al. 2016) and tree visualised and edited in Interactive tree of life (iTOL; Letunic and Bork 2024). SH-aLRT/Ultrafast bootstrap (UFBoot) support values are shown on branches. Branches are coloured based on SH-aLRT support (maximum support-light green; minimum support-purple). Blue font represents groups that have a < 70% SH-aLRT support. The asterisk '*' group denotes two Acropora actinoporins containing the aerolisin/ETX pore-forming domain (SUPERFAMILY ID: SSF56973), whilst the delta '∆' group represents a putative actinoporin in Acropora hyacinthus grouping separately from all other actinoporins investigated. The dagger ' †' symbol indicates potential pseudogenes. FIGURE 4 | Maximum likelihood tree of Membrane Attack Complex/Perforin toxins (MAC-PF) across Cnidaria. Putative MAC-PFs from the current study are denoted in green (Acropora species) and brown (Porites species). Model of evolution and tree constructed in IQ tree (Trifinopoulos et al. 2016) and tree visualised and edited in Interactive tree of life (iTOL; Letunic and Bork 2024). SH-aLRT/Ultrafast bootstrap (UFBoot) support values are shown on branches. Branches are coloured based on SH-aLRT support (maximum support-light green; minimum support-purple). The dagger ' †' symbol indicates potential pseudogenes. 79.8/92 100/100 100/100 78.5/80 99/82 TX60A_PHYSE| Δ-alicitoxin-Pse2a Phyllodiscus semoni TX60A_ACTVL| Δ-thalatoxin-Avl2a Actineria villosa TX60B_PHYSE| Δ-alicitoxin-Pse2b Phyllodiscus semoni BLAZ01001365.1| Acropora tenuis jamg1.model.Sc0000077.36| Porites lutea XP_029181532.2| Acropora millepora BOPM01000137.1| Porites australiensis plut2.m8.5176.m1| Porites lutea HIv1_RNAseq.g12695.t1| Porites compressa ENA|OKRP01000152| Porites rus BLAZ01000786.1| Acropora tenuis BLAZ01000479.1| Acropora tenuis BLFC01000595.1| Acropora digitifera ENA|OKRP01001149| Porites rus JAIFHZ010000092.1a| Acropora hyacinthus JAIFHZ010000092.1b| Acropora hyacinthus BLAZ01000787.1| Acropora tenuis XP_044164399.1| Acropora millepora BLFC01000482.1| Acropora digitifera HIv1_RNAseq.g39994.t1| Porites compressa jamg1.model.Sc0000037.126| Porites lutea SRR19144704.142.1| Porites astreoides ENA|OKRP01001779b| Porites rus ENA|OKRP01001779a| Porites rus ENA|OKRP01000292| Porites rus HIv1_RNAseq.g40042.t1| Porites compressa jamg1.model.Sc0000037.104| Porites lutea BOPM01000004.1c| Porites australiensis ENA|OKRP01001054b| Porites rus BOPM01000004.1a| Porites australiensis jamg1.model.Sc0000037.106| Porites lutea ENA|OKRP01001054a| Porites rus BOPM01000004.1b| Porites australiensis HIv1_TS.g33838.t1| Porites compressa jamg1.model.Sc0000037.105| Porites lutea BLFC01000929.1b| Acropora digitifera XP 029200908.2| Acropora millepora CM035871.1b| Acropora hyacinthus KAK2573128.1c| Acropora cervicornis KAK2573130.1| Acropora cervicornis BLFC01000610.1a| Acropora digitifera XP_029200884.2| Acropora millepora BLAZ01000336.1c| Acropora tenuis XP_029200973.2| Acropora millepora XP 029200976.2| Acropora millepora BLAZ01000721.1| Acropora tenuis CM035871.1c| Acropora hyacinthus CM035871.1d| Acropora hyacinthus XP_029200674.2| Acropora millepora KAK2573127.1| Acropora cervicornis KAK2573128.1a| Acropora cervicornis BLFC01000929.1a| Acropora digitifera BLAZ01000336.1b| Acropora tenuis BLFC01000610.1b| Acropora digitifera KAK2573128.1b| Acropora cervicornis CM035871.1a| Acropora hyacinthus XP_029200616.2| Acropora millepora KAK2573125.1| Acropora cervicornis BLFC01000632.1| Acropora digitifera BLAZ01000336.1a| Acropora tenuis AAV65396| Physcomitrin Physcomitrium patens † † Tree scale: 1 75.6/47 33.4/40 0/96 100/98 98.1/98 96.5/97 87.4/72 99.3/100 93.4/100 43.7/87 97.6/100 43.3/46 Acropora-specific MAC-PF group 1 Porites-specific MAC-PF group 1 True MAC-PF group 1 Acropora-specific MAC-PF group 2 Acropora-specific MAC-PF group 3 Acropora-specific MAC-PF group 4 79.8/96 95.9/99 100/100 98.4/99 99.9/100 87.4/96 100/100 100/100 98.7/99 87.2/78 87.3/81 100/100 87/98 98.2/99 68.6/99 100/100 93.6/86 95.2/97 100/100 88.2/99 85.6/96 88.3/100 100/100 37.3/93 85.8/90 81.1/97 92.9/92 99.8/100 99.9/100 78.6/100 100/100 0/67 95.9/92 89.5/86 77.5/87 93.9/94 0/73 99.9/100 99/100 100/87 98.5/81 1365294x, 2026, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mec.70202 by University Of Rhode Island, Wiley Online Library on [01/03/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
In addition, we also constructed a phylogenetic tree for another superfamily of PFTs, the Membrane Attack Complex/ Perforin (MAC-PF) toxins as they were found across all Acropora and Porites species. A total of 57 MAC-PFs were found across all species investigated and these MAC-PFs were spread across six distinct phylogenetic groups (Tables S3 and TR108646_c0_g1_i1.p1| Craspedacusta sowerbyi (JFT-2a) P0A377| Cry2Aa Bacillus thuringiensis 73.1/72 99.8/100 61.7/92 44.2/75 27.6/82 99.8/100 98.9/100 99.6/100 95.6/99 100/100 50.9/67 75.8/84 99.9/100 92.8/99 99.9/100 97.8/100 54.3/81 55.5/84 99.8/100 100/100 99.2/100 99.4/100 82.7/99 99.9/100 78.9/80 85.3/82 83.1/99 77.9/98 89.1/81 27.2/18 100/100 100/100 100/100 55.5/95 96.7/98 99.8/100 86.7/33 92.4/84 99.1/100 78.4/83 97.1/97 91.5/98 100/100 98.7/100 100/100 96.6/100 94.1/99 99.7/100 14.3/81 95.6/99 98.3/89 100/100 95.7/99 97.7/100 98.4/100 96.8/100 100/100 16.5/55 100/100 25.6/92 96.6/98 91.8/98 97.4/89 82.7/87 63.7/89 93.6/60 15.1/71 100/100 92.5/99 92.5/99 99.7/100 Tree scale: 1 JFT-1a JFT-1b; type I JFT-1b; type II JFT-1c JFT-2-like JFT-2b JFT-2a FIGURE 6 | Legend on next page. † 96.7/85 90.9/97 85.6/84 89.4/74 84.8/87 94.1/88 68.1/97 88.3/97 99.8/90 87.7/100 87.2/100 0/59 84.9/56 71.7/60 55.5/98 73.5/93 0/63 83/69 95/99 90.3/100 plut2.m8.5907.m1| Porites lutea HIv1_RNAseq.g8934.t1| Porites compressa BLFC01000745.1| Acropora digitifera CM035875.1_1| Acropora hyacinthus plut2.m8.18399.m1_1| Porites lutea BOPM01000032.1_2| Porites australiensis ENA|OKRP01001695| Porites rus BOPM01000032.1_1| Porites australiensis ENA|OKRP01013356| Porites rus BOPM01000032.1_3| Porites australiensis plut2.m8.7843.m1| Porites lutea CM035872.1| Acropora hyacinthus BLFC01000524.1| Acropora digitifera KAK2556033.1| Acropora cervicornis XP_044176998.1| Acropora millepora Tree scale: 1 75/88 94.8/75 0/39 98.9/100 92.2/80 77.1/79 77/79 99.3/79 95.5/100 90.6/71 61.6/67 74.8/100 85.1/77 90.5/86 93/96 84.9/94 0/54 96.9/93 79.6/86 0/45 76.2/94 28.5/50 87.8/69 78.8/98 99.7/98 0/21 83/74 72.7/44 84.5/44 95.3/100 90.5/81 92.5/84 76.3/32 31.3/63 77.3/55 87.1/85 67.7/88 91.2/67 84.5/99 80.3/95 76.7/88 47.6/80 89.7/92 80.1/87 87.8/92 0/49 85.9/99 60/75 98.4/98 SCRiP-β SCRiP-γ SCRiP-α † † † † † † † † † † † † † † 91.8/98 88.3/100 100/100 0/91 0/64 31.4/84 78.5/84 0/55 0/68 0/73 92.4/84 85.4/93 76.8/80 0/47 87.9/99 97.6/98 6.1/83 0/30 0/45 0/94 91.5/86 93.2/98 87/97 93.2/99 87.2/80 82.2/99 73.8/20 68.6/8475.7/24 83.6/65 93.7/56 92/53 90.8/98 90.8/100 91.3/94 96.4/93 99.9/100 15.3/49 1365294x, 2026, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mec.70202 by University Of Rhode Island, Wiley Online Library on [01/03/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License S10, Figure 4). Four of these six phylogenetic groups only contained putative MAC-PFs from Acropora species ('Acroporaspecific MAC-PF group 1-4'; Figure 4), whilst one group was a Porites-specific group ('Porites-specific MAC-PF group 1'; Figure 4) and the final group was the true MAC-PFs ('True MAC-PFs'; Figure 4), which contained MAC-PFs previously characterised in sea anemones (Nagai et al. 2002;Satoh et al. 2007) that grouped with putative MAC-PFs discovered in this current study from both Acropora and Porites species (87.2/78% SH-aLRT/UFBoot support). Interestingly, the true MAC-PF homologues were only present in the Indo-Pacific coral species we investigated. Thus, selection analysis was conducted to understand whether these proteins were under episodic diversifying selection. Selection analysis revealed that the true MAC-PF clade was under positive selection (mean ω = 2.223; p < 0.001; Table S13).
Custom BLAST searches used eight jellyfish CFX proteins as queries (Data S1). No Acropora species returned hits to CFXs that met the quality criteria (Table S4). Contrastingly, three putative CFX proteins which met the quality criteria were found in three Porites species-one in P. australiensis, P. compressa and P. rus, respectively (Table S4). None of these putative CFX proteins contained any recognised conserved domains. All putative CFX-like toxins from Porites species grouped with CFXs from other anthozoans for example, Ricordea yuma, Gorgonia ventalina, Exaiptasia pallida and Eunicella cavolini (100/100% SH-aLRT/UFBoot support; Figure 5). This group formed a sister clade to JFT-1b type I toxins (93.6/60% SH-aLRT/UFBoot support; Figure 5). This contrasts with other stony corals Stylophora pistillata and Madracis auretenra that formed a sister group to the JFT-1b type II toxins (86.7/33% SH-aLRT/UFBoot support; Figure 5). None of the putative CFXs from Porites species grouped with JFT-1a,c or JFT-2a,b toxins (Klompen et al. 2021; Figure 5).
To understand whether these Porites-specific JFT-1b type I homologues are under the influence of positive episodic selection (similar to other JFT-1b toxins (Klompen et al. 2021)), a selection analysis was conducted across the clade. The selection analysis revealed there was evidence of positive episodic selection across the full Porites JFT-1b type I clade (mean ω = 1.753; p < 0.001; Table S13).
For our custom BLAST bioinformatics, we identified neurotoxins from a range of cnidarians and divided them into groups based on their conserved domain regions identified by InterProScan (ATX_III; defensin-like; ShK; SCRIPs; Kunitztype; and potassium type 5 and 6 neurotoxins; Data S1). We found putative neurotoxins with similarities to kunitz-domain neurotoxins (Tables S5 and S11); SCRiPS (Tables S6 and S11); in our custom and discovery bioinformatics methods. For defensinlike toxins, only one putative protein in P. compressa (Table S8) and P. rus (Table S11) met the quality criteria; however only the P. rus hit contained the canonical defensin-like domain.
No proteins were returned for ShK neurotoxins; however two coral species (A. cervicornis and P. compressa) harboured homologues to ShK-related antimicrobial peptides AmAMP1 from A. millepora (XP_067033458.1) and damicornin (F1DFM9.1) from Pocillopora damicornis (Table S7). We also retrieved a hit from A. millepora to the true AmAMP1. Thus, in total, three ShK-related antimicrobial peptides were found across the coral species investigated (Table S7). P. compressa harboured a homologue most closely related to damicornin whereas unsurprisingly, A. cervicornis harboured a homologue most closely related to AmAMP1 from A. millepora. No putative potassium type 5, type 6 or neurotoxins containing the ATX_III were found in Acropora or Porites species (Figure 2).
Altogether, 41 putative SCRiPS were found across Acropora and Porites species which were returned from both custom BLAST searches and discovery bioinformatics (Tables S4, S6 and S11). Our SCRiP phylogenetic data show that the majority of Porites (P. rus, P. lutea and P. australiensis) and Acropora (A. millepora, A. cervicornis, A. hyacinthus and A. digitifera) species harboured putative SCRiPS that grouped with the SCRiP-δ clade (Figure 6). The SCRiP-α and SCRiP-ϒ clades were only present across Acropora species (A. tenuis, A. digitifera, A. hyacinthus and A. millepora; Figures 2 and 6). However, P. compressa and P. lutea did form a distinct sister clade to SCRiP-ϒ and SCRiP-δ (94.1/88% SH-aLRT/UFB support; Figure 6), referred to as 'SCRiP-ϒ/δ-like'. Neither Acropora nor Porites species contained any SCRiP-β.
Custom BLAST searches returned putative metalloprotease hits with the astacin domain as well as hits which contained the peptidase M10, peptidase M34 or the hemopexin domains (Tables S10 and S12). The metalloproteinases that contained the peptidase M10 and astacin domains contained the highly conserved cnidarian zinc-binding motif (HExxHxxxxxH; Hwang et al. 2022). Our discovery bioinformatics returned a hit to a putative metalloproteinase called 'Nematocyst expressed protein 6' (UniProt ID: K7Z9Q9) across four coral species (A. cervicornis, A. millepora, P. compressa and P. lutea). These same putative metalloproteinases were also found with our custom BLAST searches due to the presence of the astacin domain (Tables S10 and S11).
FIGURE 6 | Small Cysteine-Rich Proteins (SCRiPs) maximum likelihood tree. Putative SCRiP homologues from this current study are denoted in green (Acropora species) and brown (Porites species). Other sequences were taken from Barroso et al. (2024). Groups named accordingly with those named in Barroso et al. (2024). Model of evolution and tree constructed in IQ tree (Trifinopoulos et al. 2016) and tree visualised and edited in Interactive tree of life (iTOL; Letunic and Bork 2024). Groupings named as in Klompen et al. (2021). SH-aLRT/Ultrafast bootstrap (UFBoot) support values are shown on branches. Branches are coloured based on SH-aLRT support (maximum support-light green; minimum support-purple). The dagger ' †' symbol indicates potential pseudogenes.
In addition to metalloproteases, custom BLAST searches and the discovery of bioinformatics revealed homologues to phospholipase A2s (PLA2s). Custom BLAST searches only revealed 27 phospholipase A2 homologues across genome databases that had been annotated (A. cervicornis, A. millepora, P. compressa and P. lutea), and these same PLA2s were also found by discovery bioinformatics (Tables S9 and S11). All PLA2 homologues discovered harboured the Phospholip_A2_1 (PFAM ID: PF00068) domain.
Discovery bioinformatics revealed homologues to unique toxins/proteins present across Acropora and Porties species. For instance, homologues to sea anemone epidermal growth factorlike (EGF) toxins (Toxin Bcs III in Bunodosoma caissarum and OMEGA-stichotoxin-Shd4a in Stichodactyla haddoni) containing the EGF (PFAM ID: PF00008) or EGF_CA (PFAM ID: PF07645) domains were found in A. cervicornis, P. compressa and P. lutea (Table S11). The EGF-like toxin homologues from A. cervicornis, P. compressa and P. lutea shared 27.42%, 25.58%, 31.71% with human EGF (Q6QBS2) in the mature toxin region, respectively (Data S1: Figure B).
Despite most venom protein families being found across all coral species we investigated, some proteins displayed species-specific affiliations (Table S11, Figure 2). For example, only Porites australiensis contained a homologue to the Delta-actitoxin-Amc1a in the sea anemone Antheopsis maculata (P69929), which harboured a zinc-finger (zf-C2H2, PFAM ID: PF00096) and iron-sulphur binding (Fer4_4, PFAM ID: PF12800) domain.
When BLAST searched against the NCBI clustered-nr database, the closest characterised protein was the zinc finger protein 709-like in scleractinian Montipora foliosa (XP_068690159.1). Likewise, only Porites compressa harboured a homologue to the precursor protein U-actitoxin-Ugr1a-c in Urticina grebelnyi (R4ZCU1; Table S11). This toxin homologue contains the lowdensity lipoprotein receptor domain class A (Ldl_recept_b; PFAM ID: PF00058).
We also discovered homologues to unique toxins with unknown functions during envenomation that were also present across Acropora and Porties species for example, A. cervicornis, A. millepora, P. compressa and P. lutea harboured a homologue to an insulin-like peptide in a sea anemone Oulactis sp. (P0DY19) containing the insulin domain (PFAM ID: PF00049) (Table S11, Figure 2).
Our study has shown the efficiency of using both custom BLAST searches in addition to a discovery bioinformatics method to discover both conserved and novel putative venom proteins across scleractinian venoms. We found that the putative venom of both Acropora and Porites species was richest in Kunitz-type neurotoxins, pore-forming toxins (e.g., actinoporins and MAC-PFs) and neurotoxic SCRiPS. We also discovered unique toxins across both coral genera, including an insulin-like peptide and homologues to the anemone neurotoxins delta-actitoxin-Amc1a in Antheopsis maculata and U-actitoxin-Ugr1a-c in Urticina grebelnyi. Although we found large overlaps between the protein families present across our coral species, we did find several notable differences between and within coral genera. Notably, only Porites species harboured homologues to jellyfish CFXs that grouped phylogenetically with JFT-1b type I toxins, whilst only Acropora species harboured PFTs with the aerolysin domain, highlighting that each genus has access to different families of toxic proteins. We also observed that certain protein groups were absent from Caribbean coral species that have no contact with CoTS compared with Indo-Pacific species that are in regular contact with CoTS. These same protein families showed evidence of positive episodic selection, highlighting that their environment likely impacts the evolution of these toxins.
Although our study is a non-targeted bioinformatics study utilising publicly available genomes, our results are similar to those of venom-targeted laboratory studies of cnidarian species. The identification of these venom protein families agrees with previous studies, including venom-targeted laboratory studies, on hydrozoans (Jaimes-Becerra et al. 2019;Hérnández-Elizárraga et al. 2023), sea anemones (Castañeda and Harvey 2009;Frazão et al. 2012;Orts et al. 2013;Jaimes-Becerra et al. 2019;Finol-Urdaneta et al. 2020); jellyfish (Weston et al. 2013), scleractinian corals (Schmidt et al. 2019) and staurozoans (Jaimes-Becerra et al. 2019), that have all shown these protein families make up the majority of venom proteins from these species.
Unsurprisingly, we found pore-forming toxin (PFT) homologues (actinoporins, MAC-PFs, CFXs) in all the species we studied, as PFTs represent a major class of toxins in cnidarian venoms (Frazão et al. 2012;Rachamim et al. 2015;Podobnik and Anderluh 2017). Although all the coral species we investigated contained these classes of proteins, we observed interesting differences. For example, only Porites species (P. compressa, P. australiensis and P. rus) harboured CFX homologues. These homologues are grouped with the jellyfish JFT-1b type I CFXs in our phylogenetic analysis. Jellyfish are known to contain sublethal and lethal toxin repositories (Nisa et al. 2021) and as Porites species contain homologues of these proteins, it may imply that Porites species harbour toxins that are more effective at causing injuries to CoTS and therefore deterring CoTS predation. Specifically, JFT-1b type I CFXs have been shown to have different bioactivity to type II, with type I toxins being cardioactive whereas type II are predominantly haemolytic (Brinkman et al. 2014;Klompen et al. 2021). Furthermore, type II showed no cross-reactivity with type I antibodies, emphasising their structural differences (Brinkman et al. 2014). Interestingly, in our phylogenetic tree (Figure 5) a CFX homologue from Stylophora pistillata, whose venom is hypothesised to contain several haemolytic molecules (Ben-Ari et al. 2018), grouped with the haemolytic JFT-1b type II CFXs. This reflects that the phylogenetic grouping that reflects the bioactivity of these toxins in jellyfish may also be similar for scleractinians. Other studies have found CFX-like proteins in anthozoan venoms (Jaimes-Becerra et al. 2019;Klompen et al. 2020Klompen et al. , 2021)). However, it is currently debated whether anthozoans truly contain CFX-like toxins, or rather an actinoporin sister family related to CFXs, as no CFX homologues have been successfully isolated from anthozoan venoms (Klompen et al. 2021).
Distinct Pore-Forming Toxins Not Found in the Other Genera
Three putative actinoporin homologues across three Porites species lacked the anemone actinoporin domain and were most closely related to the tereporin-like actinoporin in Nematostella vectensis (Table S1, Figure 3). Similarly, two Acropora species returned hits from BLAST searching hydralysins and both hits harboured a distinct aerolysin/ETX pore-forming domain (SUPERFAMILY ID: SSF56973; Table S2, Figure 3). This finding is unsurprising since hydralysins and aerolysins harbour several conserved motifs (Sher et al. 2005). Whether these diverse families of actinoporins contributed to a diversified biological activity during envenomation requires future work. An actinoporin from S. pistillata venom, Δ-Pocilopotoxin-Spi1, elicited haemolytic activity on human erythrocytes 600 times greater than that of other scleractinian venoms (Ben-Ari et al. 2018). This shows that a coral's repertoire of PFTs likely contributes to the bioactivity of its venom. However, it has recently been shown that MAC-PF paralogs in the cnidarian Nematostella vectensis have undergone duplications and have been incorporated into endomesodermal cells, where they likely have non-venomous functions (Surm et al. 2024), emphasising the importance of combining functional studies with bioinformatics to characterise the true venom of scleractinians.
Unsurprisingly, enzymes and neurotoxins were a major feature of Acropora and Porites species venoms (e.g., kunitz-type neurotoxins, SCRiPS, PLA2s and metalloproteinases). However we found no evidence of neurotoxins containing the Shk or ATX_ III domain nor potassium type 5 or 6 neurotoxins. This agrees with a previous study, stating that there is an absence of Shklike neurotoxins in scleractinian corals (Schmidt et al. 2019). Despite no coral species returning hits to Shk neurotoxins, we did find homologues to Shk-related scleractinian antimicrobial toxins across three of the coral species we investigated. These represented homologues to AmAMP1 from A. millepora or damicornin from Pocillopora damicornis (Table S7). AmAMP1 harbours antimicrobial activity to both Gram-positive and Gram-negative bacteria (Mason et al. 2021), whereas damicornin has been shown to elicit activity against Gram-positive bacteria (Vidal-Dupiol et al. 2011). In sea anemones, Shk neurotoxins have evolved from a gene duplication event from Shk-like neuropeptides and were subsequently recruited into the venom (Sachkova et al. 2020). Shk-like neuropeptides show similar bioactivity to Shk neurotoxins (Sachkova et al. 2020). Thus, although AmAMP1 and damicornin are Shk-related and show bioactivity against bacteria, it cannot be ruled out that they (and their homologues) may also be found in the venom and harbour other toxic functions (Mason et al. 2021). Indeed, we found several conserved residues when we aligned Shk-toxins that show voltage-gated potassium channel blocking activity with our Shklike homologues in Porites compressa and Acropora cervicornis (Data S1: Figure A). We also found two defensin-like neurotoxin homologues in P. compressa and P. rus, which are utilised as neurotoxins in sea anemones but have antimicrobial activity in other animals (Finol-Urdaneta et al. 2020). Thus, understanding the breadth of bioactivity and function of these toxins is an important goal for future research.
In contrast to the absence of Shk neurotoxins, all coral species that we investigated harboured homologues to cnidarian kunitz-type neurotoxins. A previous study investigating gastropods, found a kunitz-type neurotoxin in the transcriptome of the CoTS predator Charonia tritonis, which was shown to be a homologue to the cone snail conkunitzin-S1 (Zhang et al. 2022).
The authors synthesised this conkunitzin-S1 homologue (Ctkunitzin) and investigated its bioactivity against mice and CoTS in laboratory assays (Zhang et al. 2022). In mice, Ct-kunitzin suppressed pain recognition and reduced their locomotory activity, whilst in CoTS, Ct-kunitzin destroyed the tube feet (Zhang et al. 2022). In the black mamba snake venom (Dendroaspis polylepis) kunitz-type toxins (known as dendrotoxins) are among the most potent venom components, exhibiting exceptionally high lethality (Laustsen et al. 2015). Thus, kunitz-type neurotoxins may elicit damaging bioactivity against CoTS predation in both Acropora and Porites species and their bioactivity warrants further investigation.
Similarly to kunitz-type neurotoxins, we found the scleractinianspecific neurotoxins, SCRiPS (Sunagawa et al. 2009;Jouiaei et al. 2015), across all Acropora and Porites species. Furthermore, the phylogeny of our Acropora SCRiPS agrees with that of a recent study (Barroso et al. 2024). However, this same study did not find homologues to SCRiPS in P. rus and P. australiensis genomes or a P. lutea transcriptome (Barroso et al. 2024), contrasting with our results from both our custom BLAST and discovery pipelines. We found large expansions of SCRiPS homologues in Porites and Acropora species in the SCRiP-δ clade. As SCRiP-δ was the only SCRiP clade detected in sea anemones and had the largest purifying selection (toxin potency preserved after initial expansion) and a SCRiP-δ from A. millepora had the highest potency versus a SCRiP-α (Jouiaei et al. 2015), it was hypothesised that SCRiP-δ may be the main SCRiP functioning in predator defence (Barroso et al. 2024). Our data would agree with this, emphasised by the large expansion of the SCRiP-δ families in both Porites and Acropora species, emphasising that SCRiP-δ must play an important role in the biology of these corals. Future studies should investigate the data provided by this current study and the previous studies on SCRiPS (Jouiaei et al. 2015;Barroso et al. 2024) to investigate the role and function of SCRiPS (especially proteins from the SCRiP-δ family) in predator defence of coral species against CoTS.
Enzymes are an important component of venoms, functioning not only in direct toxicity but also in the spread of the venom, activating other venom components and helping to preserve the venom (Nevalainen et al. 2004;Gonçalves and Costa 2021;Yu et al. 2021;Delgado-Prudencio et al. 2022). We found PLA2s and metalloproteinases from our custom BLAST and discovery bioinformatics searches across our coral species. In certain cnidarian species for example, the jellyfish Chrysaora quinquecirrha, venom metalloproteinases constitute the main toxicity (Yang et al. 2024). Functional studies will be needed to elucidate which of the enzymes found in this current study are specifically involved in envenomation, since enzymes have a plethora of functions and may not be involved in envenomation or present in the venom itself. Indeed, harbouring a large repertoire of different families of neuroactive peptides would allow efficient paralysis of prey and/or predators.
Two Porites (P. compressa and P. lutea) and two Acropora (A. cervicornis and A. millepora) species had a protein that had a hit to an insulin-like peptide (ILP) in the sea anemone Oulactis species. This insulin-like peptide from Oulactis species has been shown to elicit weak activity against sodium and potassium channels but had no effect on human insulin or insulin-like growth factor receptors (Mitchell et al. 2021). Contrastingly, a study conducted on ILPs from the venom of the sea anemone Exaiptasia pallida showed that one synthesised ILP elicited hypoglycemia and abnormal locomotory function in zebrafish and also successfully bound to human insulin through the formation of three hydrogen bonds (Guo, Tang, et al. 2024). ILPs have been shown to play a role in the envenomation elicited by other venomous animals by disrupting the metabolic equilibrium in victims for example, cone snails (Guo, Huang, et al. 2024) and snakes (Matkivska et al. 2023). Thus, although it is unclear whether these ILPs discovered in Acropora and Porites species play a role during envenomation, future functional studies are warranted to investigate their role.
We found homologues to sea anemone EGF-like toxins in A. cervicornis, P. compressa and P. lutea. Gigatoxin I from the sea anemone Stichodactyla gigantea was the first EGF-like peptide discovered in sea anemones and shares 31%-33% homology to mammalian EGFs (Shiomi et al. 2003). Gigatoxin I is potently paralytic to crabs (Shiomi et al. 2003). Likewise, in the sea anemone Heteractis crispa, an EGF-like toxin has a similar 3D structure to human EGF (Guo, Fu, et al. 2024). The EGF-like toxin homologues we found in this current study from A. cervicornis, P. compressa and P. lutea shared a 25%-31% homology with human EGF (Data S1: Figure B). This similarity to vertebrate EGFs is also true in other animal venoms for example, in ant venom, where EGF-like toxins have both structurally and molecular mimicry to vertebrate EGF and this facilitates their bioactivity by causing long-lasting hypersensitivity in victims (Eagles et al. 2022).
Species-specific affiliations to unique toxins were also observed in P. australiensis and P. compressa, both harbouring a homologue to a sea anemone toxin, Delta-actitoxin-Amc1a in Antheopsis maculata and U-actitoxin-Ugr1a-c in Urticina grebelnyi, respectively. Similarly, these toxins contained unique domains not found in any of the other coral species toxins we investigated in this study (zinc finger and iron-sulphur binding domain in the Delta-actitoxin-Amc1a homologue and the low-density lipoprotein receptor b in the U-actitoxin-Ugr1a-c homologue). Delta-actitoxin-Amc1a has been shown to have toxicity to crabs in bioassays and had no known homology with any other toxins when it was discovered (Honma et al. 2005). U-actitoxin-Ugr1a harbours a structural β-defensin-fold and shows activity on human sodium-selective acid-sensing ion channels (ASIC) and in mice models produced an analgesic effect (Osmakov et al. 2013;Finol-Urdaneta et al. 2020). Ugr1a shows a similar bioactivity and structure to another sea anemone β-defensin-fold toxin, APETx2 from Anthopleura elegantissima, that also targets ASIC channels (Chagot et al. 2005).
Although it is unknown whether any of these unique toxins are utilised against CoTS during predation, the presence of a rich venom toxin arsenal with diverse bioactivity and structures present across both Acropora and Porites species, highlights the extensive venom deterrent that these species have access to during CoTS predation.
In this study, we included two scleractinian species that were endemic to the Caribbean and are therefore 'naïve' to CoTS (i.e., A. cervicornis and P. astreoides) and eight species in Pacific reefs with a history of CoTS exposure, to examine whether venom differences may have evolved through selective pressure through predation exposure from CoTS (Figure 2). Interestingly, homologues of the true MAC-PF toxins in Actiniaria were only found across CoTS-exposed coral species (Figures 2 and 4), being absent from both A. cervicornis and P. astreoides. Furthermore, we only saw CFX homologues in our Porites-CoTS exposed species, although not exclusively so, with CFX homologues absent in P. lutea.
Indeed, animal toxins have been shown to diversify rapidly under adaptive selection pressures (Kordiš and Gubenšek 2000). In cnidarians specifically, the majority of toxin gene families have been shown to occur under negative selection (e.g., actinoporins, JFTs and SCRiPS neurotoxins), where variations in venom genes accumulate rapidly under short bursts and beneficial genes get fixed in the population and accumulate for long periods of time (Jouiaei et al. 2015). These toxin families under negative selection are hypothesised not to be influenced by the chemical arms race that occurs between predator and prey due to their long accumulation times and high conservation of amino acid sequences (Jouiaei et al. 2015). However, selection forces can differ between genes and it has been shown that certain clades of JFTs have undergone episodic positive selection (Klompen et al. 2021). Both true MAC-PFs homologues in Acropora and Porites in addition to Porites JFT-1b homologues were under episodic positive selection in our study. Future research is needed to investigate whether these proteins are indeed involved in exposure to predators such as CoTS.
Apart from venom components, many other factors need to be considered to understand the complete repertoire of coral defence strategies against predation. These factors include (but are not limited to) a coral species' nematocyst community and the plasticity of this community to changing predation pressures and environmental conditions, in addition to the metabolites and volatiles produced from corals and their associated taxa (i.e., the holobiont) which may aid in deterring predators (David et al. 2008;Ben-Ari et al. 2018;Modolon et al. 2020;O'Hara et al. 2021;Klompen et al. 2022). These variables highlight the multitude of factors that must be considered when investigating the role of coral venom in defence against CoTS. Future research should aim to profile gene expression of venoms and cnidomes from different polyp types of various coral species, sampled over a wide range of temperatures, locations and CoTS predation pressures, which would enable rigorous comparisons of venoms between coral species (Rachamim et al. 2015).
Overall, our study provides new insights into the similarities and differences in the venom profiles between CoTS-preferred (Acropora species) and CoTS-non-preferred prey (Porites species), shedding light on the toxin repertoires of two major reefbuilding genera in Moorea, French Polynesia that are exposed to CoTS. Critically, our study emphasises the need for future research combining high-quality coral genomes with tailored gene expression of scleractinian nematocysts and investigating the function of extracted toxins. The foundation laid by our study will help us to investigate the evolution, loss and duplication of true scleractinian toxin genes as well as their function.
While many protein families were found across all coral species (kunitz-type neurotoxins, actinoporins, MAC-PFs and SCRiPS neurotoxins) the phylogenies of these protein families and individual proteins within these families differed between coral genera. We also observed a difference in the presence/absence of homologues to the true-MAC-PF toxins in CoTS-exposed corals compared to corals not exposed to CoTS. Similarly, only CoTS-exposed Porites species harboured a JFT-1b type I homologue compared to Porites astreoides which is ies not exposed to CoTS. Moreover, both these protein homologues were observed to be undergoing positive episodic selection. Additionally, some proteins were found exclusively in Porites species venom (e.g., JFT-1b type I CFXs and tereporin-like actinoporins) or Acropora species venom (e.g., SCRiP-α and PFTs harbouring the aerolysin domain). These exploratory findings suggest that, despite similarities in the overall toxin repertoire, there are distinct molecular signatures that may influence CoTS feeding behaviour. It is critical to further characterise the influence of coral venom on the feeding preferences of CoTS to inform management strategies to conserve reefs during CoTS outbreaks. For instance, more venomous corals could be strategically distributed/planted throughout reefs, potentially providing a bioactive shield for more vulnerable corals. Overall, our results demonstrate a wide repertoire of toxins in corals that provide a molecular perspective on coral susceptibility to CoTS predation events on coral reefs.
1365294x, 2026, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/mec.70202 by University Of Rhode Island, Wiley Online Library on [01/03/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License