Crown-of-thorns starfish spines secrete defence proteins

Animals

COTS were collected under permit G17/38293.1 by the Association of Marine Park Tourism Operator divers working on the Great Barrier Reef, Cairns region, Northern Queensland. Starfish (∼25 cm in diameter) were transported by air freight to the University of the Sunshine Coast Aquarium facility (Sippy Downs), where they were housed for up to two weeks in a recirculating protein skimmed saltwater system of approximately 1,500 L capacity.

COTS venom extraction for lethality bioassay

Two COTS venom extracts, ‘Spine Extract’ (SpE) and ‘Stress Water Extract’ (StWE), were prepared and their cytotoxicity tested. The SpE was prepared from 20 aboral spines collected from three different adult COTS (n = 60, unknown sex). Spines were combined and placed into 50 ml of Milli-Q water, then agitated for 1 h to facilitate venom release. A 14 ml aliquot was separated through a 10-kDa cut-off Amicon ultra-15 centrifugal filter unit (Merck, Australia) under centrifugation (4000 xg for 10 min) at 4 °C. Fractions which contained <10 and >10 kDa proteins (designated as <10-kDa SpE and >10-kDa SpE, respectively) were separately collected into fresh tubes and used immediately in a brine shrimp lethality bioassay. To obtain StWE, two adult COTS (unknown sex) were placed into individual glass tanks filled with 1 L of aerated filtered seawater (FSW). Animals were intermittently mechanically agitated by physical tapping with blunt forceps for 1 h. The conditioned seawater (2 × 1 L) was drained from each tank and strained through a 100-µm mesh to remove any particulate matter. The two samples were then subjected to a mobile phase filtration through a 0.45-µm membrane filter (Waters, Milford, MA, USA) under vacuum. Each filtered sample (32 ml; designated as StWE) was subsequently fractionated through a 10-kDa cut-off Amicon ultra-15 centrifugal filter unit (Merck, Rahway, NJ, USA) by centrifugation (4000 × g for 10 min) at 4 °C. After centrifugation, the solution retained on the membrane filter (>10-kDa StWE) and the eluted solution (<10-kDa StWE) were collected. Samples were immediately used in the brine shrimp lethality bioassay. All protein concentrations were quantified based on the absorbance at 280 nm using the Nanodrop 2000 instrument (Thermo Fisher Scientific, Waltham, MA, USA) after filtrations.

Brine shrimp lethality bioassay

A brine shrimp lethality bioassay was performed according to Meyer et al. (1982), with some modification. Brine shrimp eggs (Artemia salina) were hatched in a hatching container filled with 300 ml of aerated FSW for 48 h at room temperature. At 48 h post-hatch, brine shrimp larvae were transferred via a pipette into a 96-well plate filled with 20 µl/well of aerated FSW (N = 14  ± 3 individuals per well). Brine shrimp larvae were then exposed to the different extracts, added at various concentrations (5 replicates per dose per extract). For each extract, the higher concentrations used in the brine shrimp lethality assay were prepared at 1:2 of the original concentration of a given extract in each replicate (which was obtained from our preparation procedure described in the COTS venom extraction methods section). Two additional dilutions (1:10 and 1:100 of the upper concentration) were also tested for LD50 calculations. For negative controls, FSW (20 µl) was added into the well.

Artemia larvae were counted at 0, 24, and 48 h post-treatment to calculate percent mortality. Data was visualised by applying a log10 function to the COTS venom concentrations (x-axis) and plotting against the average shrimp mortality rate converted into probits (y-axis) (Finney, 1952). The lethal dose required to cause 50% mortality in Artemia (LC50) was calculated via a 3-point linear regression. Statistical analysis of data was performed using the IBM SPSS software. A one-way ANOVA (univariate analysis) and subsequent Scheffe post-hoc tests were used to establish significance at a confidence threshold of 95% (P ≤ 0.05). Where data did not contain an equal variance of error according to Levene’s test (P ≤ 0.05), the Games-Howell post-hoc test was used to evaluate significant differences between different groups at a confidence threshold of 95% (P ≤ 0.05).

Histology and scanning electron microscopy (SEM)

At least four spines, removed from COTS using surgical scissors, were immediately placed into 4% paraformaldehyde and left overnight at 4 °C before being transferred and stored in 70% ethanol (EtOH). Spines were rehydrated in 50% EtOH (two washes; 30 min each) followed by autoclaved distilled water (three washes; 20 min each). Samples were decalcified by soaking in Morse’s solution (10% sodium citrate and 50% formic acid) overnight before being washed with autoclaved distilled water to quench the decalcification reaction. Samples were then processed through increasing ethanol concentrations (70%–100%) and 100% xylene, before being embedded in paraffin wax. Embedded tissue blocks were sectioned at 8–12 µm-thick using a Leica microtome, and sections placed onto SuperFrost slides (Thermo Fisher Scientific, Waltham, MA, USA) and dried overnight at 37 °C. Hematoxylin and eosin (H&E) staining was completed following a routine protocol with modifications (Smith et al., 2018). Finally, sections were cleared using xylene solution, before being mounted with a coverslip using dibutylphthalate polystyrene xylene (DPX)-based mounting medium. Slides were viewed under a Leica DM550 microscope equipped with a Leica camera.

For scanning electron microscopy (SEM), at least 4 aboral spines were removed using surgical scissors and immediately fixed with 10% formalin for overnight at 4 °C. After initial fixation, the spines were additionally fixed using a glutaraldehyde and paraformaldehyde buffer (pH 7.4) for 4 h. Spines were then cleaned 3 times (10 min each) using 0.2 M Millonig’s buffer (pH 7.4) before and after immersion in osmium tetroxide buffer for 1 h. The spines were then dehydrated in a graded series of aqueous ethanol (EtOH; 50% to 100%), dried using a CO₂ critical point dryer, and platinum-coated using an Eiko IB-5 Sputter Coater onto an SEM mount. Four specimens were viewed using a JEOL 6300 field emission SEM at 15 kV.

COTS spine protein extraction and preparation for mass spectrometry (MS)

To maximise the proteins identified from COTS spine, spines from both relaxed and stressed individuals were investigated to account for any potential differences in secretion. Three adult COTS were held in isolation in aerated tanks (approximately 40 L in volume; one individual per tank) with artificial saltwater (salinity 35 ppt) to 30% capacity (12 L). Animals were acclimatised over a 6 h period with hourly 50% water changes using FSW. Acclimation was complete when surface foam and mucus were no longer visible (animals in this condition were considered ‘relaxed’). Spines from each individual relaxed COTS (n = 4 spines per individual) were removed from the aboral side of the body using surgical scissors, placed in 10 ml of FSW, then gently mixed for 30 min. The same three adult COTS were stressed by mechanical agitation for 5 min, after which observable indicators of stress (including mucus production, formation of white foam on the water surface and increased movement), were observed. Spines (n = 4 spines per stressed individual) were then collected, immersed in 10 ml of FSW, then gently mixed for 30 min. Following mixing, the spine-conditioned FSW was acidified with trifluoroacetic acid (TFA) to a final concentration of 0.1%, then stored at −80 °C until preparation for mass spectrometry.

Acidified samples of both relaxed and stressed COTS spine-conditioned water were individually semi-purified using a Sep-Pak Vac 20cc (5 g) C18 cartridge (Waters, Milford, MA, USA), according to the manufacturer’s instructions. Following sample loading and washes, samples were eluted using 5 ml 60% ACN/0.1% TFA. The eluent was dried in a speedvac (Savant, Barnstable, MA, USA) and stored at −20 °C. The extracts were reconstituted in 200 µl aqueous 0.5% formic acid (FA_(aq)) in MilliQ water and an in-solution trypsin digest performed using trypsin solution (Promega, Madison, WI, USA), following the method described previously (Ni et al., 2018).

Micro-high-pressure liquid chromatography-tandem quadrupole time-of-flight mass spectrometry (µHPLC qTOF-MS/MS) analyses and protein identification

The tryptic peptides were analysed as described by Hall et al. (2017). In brief, µHPLC-MS/MS on a ExionLC liquid chromatography system (AB SCIEX, Concord, Canada) coupled to a qTOF X500R MS (AB SCIEX, Concord, Canada) equipped with an electrospray ion source. Each sample (20 µl) was injected onto a 100 mm × 1.7 µm Aeris PEPTIDE XB-C18 100 µHPLC column (Phenomenex, Sydney, Australia) equipped with a security guard column. Peptides were eluted using solvent A, 0.1% FA_(aq), and solvent B, 100% ACN:0.1% FA_(aq), at 400 µL/min flow rate. A linear gradient of 5–35% solvent B over 10 min was applied followed by a steeper gradient from 35% to 80% solvent B in 2 min and 80% to 95% solvent B in 1 min. Solvent B was held at 95% for 1 min to wash the column then returned to 5% solvent B for equilibration prior to the next sample injection. The ionspray voltage was set to 5500 V, declustering potential 100V, curtain gas flow 30, ion source gas 1 40, ion source gas 2 50 and spray temperature at 450 °C. MS data was acquired in the Information Dependant Acquisition mode. Full scan qTOF-MS data was acquired over the mass range 350–1400 m/z; product ion MS/MS data was acquired over 50–1800 m/z. Ions observed in the qTOF-MS scan exceeding a threshold of 100 cps and a charge state of +2 to +5 were set to trigger the acquisition of product ion. The data was acquired and processed using SCIEX OS software (AB SCIEX, Framinham, MA, USA). Biological triplicates were used for the analysis.

The µHPLC qTOF-MS/MS data were imported to the PEAKS studio (Bioinformatics Solutions Inc., Waterloo, ON, Canada, version 7.0) with the assistance of MS Data Converter (Beta 1.3, http://sciex.com/software-downloads-x2110). Peptides were analysed using PEAKS v7.0 (BSI, Canada) against the protein database assembled from genome data (http://marinegenomics.oist.jp) (Hall et al., 2017). De novo sequencing of peptides, database searches and characterisation of specific post-translational modifications were used to analyse the raw data; false discovery rate was set to ≤ 1%, and [−10 * log (p)] was calculated accordingly where p is the probability that an observed match is a random event. The PEAKS used the following parameters: (i) precursor ion mass tolerance, 0.1 Da; (ii) fragment ion mass tolerance, 0.1 Da (the error tolerance); (iii) tryptic enzyme specificity with two missed cleavages allowed; (iv) monoisotopic precursor mass and fragment ion mass; (v) a fixed modification of cysteine carbamidomethylation; and (vi) variable modifications including lysine acetylation, deamidation on asparagine and glutamine, oxidation of methionine and conversion of glutamic acid and glutamine to pyroglutamate.

RNA-seq, gene expression and annotation

For initial relative expression analysis, transcriptome data for each of the tissues represented in http://marinegenomics.oist.jp/cots was downloaded and heatmaps constructed using R (version 3.1.1) (https://www.r-project.org/) based on Z-score values using the scale function. More in-depth RNA-seq gene expression quantification was performed using COTS radial nerve cord, tube foot and sensory tentacle derived from Roberts et al. (2017) and Smith et al. (2017). In addition, COTS were randomly selected for aboral spine and stomach RNA-seq. Spines were cut three mm from the middle section along the aboral arm at the base of the spine, and then immediately transferred to RNAlater (Thermo Fisher Scientific, Waltham, MA, USA) before storing at −20 °C. Stomach tissue was obtained by inducing stomach eversion using small (<5 cm) coral fragments, when eversion was visually confirmed, ∼200 mg of stomach tissue was cut and immediately put into cold RNAlater (Thermo Fisher Scientific, Waltham, MA, USA) before storage at −20 °C. Approximately 100 mg of the individual tissues from spine and stomach, was later used for RNA extraction using TRIzol™ Reagent (Thermo Fisher Scientific, Waltham, MA, USA), as per manufacturers protocol, then total RNA was quantified using a Nanodrop 2000 (Thermo Fisher Scientific, Waltham, MA, USA) and integrity was assessed with a Bioanalyzer RNA 6000 Nano mRNA kit (Agilent Technologies, Santa Clara, CA, USA). Only samples with a RIN value >4 were deemed suitable for RNA-seq. Total RNA from aboral spine (n = 6) and stomach (n = 3) was sent to Novogene (Hong Kong) for Illumina 2500 sequencing using their standard workflow (https://en.novogene.com/).

Full-length precursor sequences of proteins identified in the MS analysis of spine samples were retrieved from the protein database assembled from the available COTS transcriptomes/genome data, and subsequently used for the prediction of putative secreted proteins. Putative secreted proteins were predicted based on the presence of signal peptide and absence of transmembrane domain in the protein precursor sequences, using SignalP (V.5.0) and TMHMM (V.2.0) webtools (Almagro Armenteros et al., 2019; Krogh et al., 2001). Tissue expression of genes encoding identified putative secreted proteins was investigated using the publicly available transcriptomes from the COTS genome browser (http://marinegenomics.oist.jp) (Hall et al., 2017). This comprised of COTS tube foot, mouth, spine, sensory tentacle, radial nerve cords (male and female), testis, ovary, egg, mid-gastrula tissues (Hall et al., 2017; Roberts et al., 2017). Gene expression in TPM (transcripts per kilobase million) values was transformed to the z-scores before being visualized in a heat map using the ClustVis webtool (Metsalu & Vilo, 2015).

General functions of identified putative proteins were investigated based on the Gene Ontology (GO) annotation. In brief, full-length sequences of putative secreted proteins were annotated against a non-redundant database on the National Center for Biotechnology Information (NCBI) using a BLASTp search on the Omicsbox software (BioBam). GO annotation, limited to the biological process category, was then performed. The top-40 most abundant GO terms were displayed in a WordCloud format where the front size represented the node score calculated according to the following formula: score = ∑_GOsseq × α^dist, where ‘seq’ is the number of different sequences annotated at a child GO term and ‘dist’ the distance to the node of the child (BioBam).

Based on GO and literature, plancitoxin, phospholipase A2 and peptidase inhibitor 16-like were recognised as toxins. Relative expression of plancitoxin, phospholipase A2 and peptidase inhibitor 16-like genes in tissues was determined using the CLC Genome Workbench (version 20) based on transcripts per kilobase million (TPM), utilizing the COTS genome Version 1.0 (http://marinegenomics.oist.jp/cots), as reference. The statistical power of this experimental design, calculated in RNASeqPower was 0.725.

Phylogenetic analysis

Echinoderm non-COTS sequences used for phylogenetic analysis were curated from the NCBI protein database. Deduced protein sequences were aligned in MEGA v.7 and v.11 using the CLUSTAL algorithm and phylogenetic trees constructed using the maximum likelihood method with 1,000 bootstrap replicates. iTOL online tool (Letunic & Bork, 2021) was used to assist in phylogenetic tree illustration.

Toxicity of COTS venom extracts

COTS aboral spines (∼3 cm in length; Fig. 1A) were collected and processed for venom collection (SpE), in addition to stress water extract (StWE). Brine shrimps were exposed to varying concentrations of crude and semi-purified extracts, then activity was observed at 48 h post-exposure (Fig. 1B). Brine shrimp exposed to FSW (as negative control) showed a mortality of 3.48 ± 3.28% (mean ± SD). Crude SpE at 110 µg/mL induced significantly higher mortality (52.93 ± 14.63%) when compared with the control group, however, dilutions of 1:10 and 1:100 did not. Fractionation of SpEs (<10 kDa and >10 kDa), demonstrated that only >10 kDa at 93.3 µg/ml contained significant toxicity (51.33 ± 13.13%), comparable to that of crude SpE. Crude StWE induced significant mortality at 107 and 10.7 µg/ml (88.21 ± 9.64% and 50.08 ± 11.45%, respectively), while fractionated StWE at >10 kDa caused significant mortality (81.52 ± 9.81%) at 98.2 µg/ml. Although <10 kDa StWE did induce mortality at the highest concentration tested (10.0 µg/ml), the data showed an unequal variance of error according to Levene’s test (Games-Howell post-hoc test was considered insignificantly different to the control group (P = 0.063)). The LC50 of each extract and the extract toxicity are summarized in Table 1. All extracts, except the <10 kDa SpE, were considered toxic based on Meyer’s criterion. Following Clarkson’s criterion, the SpE and >10 kDa SpE exhibited medium toxicity, whereas all StWE treatments were considered highly toxic.

Histology of COTS spines and proteomic analyses of spine-conditioned water

SEM images of COTS aboral spines provided a view of the basic ultrastructure of the outer surface of COTS spines with progressive magnifications (Figs. 2A and 2B), revealing a topography of craters and pores (2–10 µM in diameter) covering the entirety of the spine. A SEM image of a longitudinal sectioned spine revealed the underlying components, including an inner core consisting of a porous three-dimensional meshwork of endoskeleton plates (trabeculae), which aligned longitudinally and parallel to the spine’s length, and an outer epithelial layer (Fig. 2C). This was similarly observed through histological section analysis (Fig. 2D), and further demonstrating an outer columnar epithelium with a thick basement membrane. A well-defined nucleus was present and granular-like dots could be observed, dispersed throughout the basement membrane.

Aboral spine proteins were extracted from combined relaxed and stressed COTS individuals to ensure the full repertoire of potential toxin proteins were identified following mass spectrometry. In total, 157 protein sequences were identified that contained at least one MS peptide match (Table S2), from which there were 56 different proteins. GO annotation indicated that identified proteins were mainly associated with cellular process, lipid transport, cytoskeleton organisation and modulation of process of another organism (e.g., induces hemolysis in another organism) (Fig. 3A). Sixteen proteins were consistently identified within the majority of spine extracts (both relaxed and stressed COTS spines) and predicted to encode a signal peptide (Fig. 3B, Table 2). Besides oki.111.24 which was annotated as an uncharacterised protein, other secreted proteins had high confidence matches to known proteins, including for example, ependymin-related proteins, plancitoxin-1, phospholipase A2, cysteine-rich secretory protein, pancreatic triacylglycerol lipase-like protein and vitellogenins. A cluster analysis based on gene expression pattern, showed two distinct groups of genes, dependent on whether highly expressed in the spine (Fig. 3B), which were further investigated below.

COTS spine excretory-secretory proteins

Plancitoxins

The COTS genome models predict three plancitoxin genes, two with full-length (oki.27.33 and oki.27.35) and one with partial length (oki.27.34), that encode proteins with deoxyribonuclease (DNase) II domain organisation (Fig. 4A). The deduced full-length forms were 356 amino acids in length which include the signal peptide sequences. Oki.27.33 shared highest similarity to oki.27.35 and to the previously identified COTS plancitoxin I (84.12%; GenBank accession number: BAD13432), whereas oki.27.34 shared ∼36–38% similarity to oki.27.33 and oki.27.35. A DNase II domain (domain ID: IPR004947), catalytic domain repeats 1 and 2 (CD09120 and CD09121), and seven conserved cysteine residues were recognised in the full-length COTS plancitoxins (Fig. 4A). Gene expression of all forms were relatively abundant in the spine, although less, plancitoxins are also detected in other adult COTS tissues such as tube foot, sensory tentacle, RNC and stomach (Fig. 4B). Within the sensory tentacle, expression of oki.27.34 was notably higher than other forms. Only oki.27.33 and oki.27.35 were identified from aboral spine protein preparations. Phylogenetic analysis showed that COTS plancitoxins share similarity with plancitoxins reported in other echinoderms and molluscs, yet formed a distinct clade from vertebrate DNase II-alpha proteins and their homologous proteins in nematodes (Fig. 4C).

Phospholipase A2

The COTS genome contained 8 genes encoding secreted-type PLA2 (sPLA2) proteins: oki.8.260, oki.8.261, oki.8.262, oki.8.264, oki.8.268, oki.8.269, oki.8.270, and oki.36.73. All genes except oki.36.73 were located on scaffold 8. Deduced proteins from sPLA2 genes on scaffold 8 shared ∼50–58% similarity, whereas oki.36.73 showed highest similarity (41.29%) to oki.8.268. All COTS sPLA2 precursors (159–175 amino acids in length) displayed a signal peptide sequence, a PLA2 domain (domain ID: IPR016090), a catalytic PLA2 histidine active site (PROSITE patterns ID: PS00118), and 14 conserved cysteine residues (Fig. 5A). The deduced PLA2 from oki.8.261 and oki.8.262 shared 98.73% and 100% identity, respectively, to the previously reported COTS PLA2-I and -II (GenBank: BAE46765 and BAE46766) (Ota et al., 2006). Relative gene expression of COTS sPLA2 indicated that the oki.8.260, oki.8.261, oki.8.262, and oki.8.264 were abundant in the spine, whereas oki.8.268 was predominant in the sensory tentacle, tube foot, and radial nerve cords (both sexes), but also detected in the spine tissue (Fig. 5B). However, gene expression of oki.8.269, oki.8.270, were very low or not detected in any tissues investigated, besides oki.36.73, which showed some expression in stomach tissue. Among all sPLA2 identified, only protein products from oki.8.261, oki.8.262, and oki.8.264 were detected from COTS aboral spine sample preparations (Fig. 3). Phylogenetic tree analysis of COTS sPLA2 proteins with different groups/types of sPLA2 in other species showed that echinoderm sPLA2 (including those of COTS) formed a distinct cluster, which was then rooted with PLA2 groups I (e.g., type IA PLA2 toxin in a cobra and type IB PLA2 in humans) and IX (i.e., PLA2 venom in a marine snail, Conus magus) (Fig. 5C). However, oki.36.73, which formed a separate root from other proteins, appeared to be distinct from other secreted PLA2 proteins. Another distinct clade (including humans) contained a group of sPLA2 proteins from various species which had been classified as PLA2 groups II, III, V and X.

Peptidase inhibitor 16-like

Oki.208.17 was annotated as a peptidase inhibitor 16-like protein (PI16) (Table 2), sharing greatest similarity to a PI16-like protein from Patiria miniata (accession number: XP038075979; 63.32% identity) and a PI16 found in in Asterias rubens (accession number: XP033625744; 52.36% identity). The oki.208.17 precursor was 220 amino acids in length and contained a signal peptide, cysteine-rich secretory protein (CRISP) domain (IPR001283), a venom allergen 5-like domain (IPR002413) and an allergen V5/testis-specific protein (Tpx-1)-related conserved site (IPR018244), corresponding to the ‘D₁₄₂HYTQLVWAKS₁₅₂’ motif (Fig. 6A). Ten cysteine residues were present, eight of which were within the CRISP domain. Expression of oki.208.17 was high in the spine tissue (almost 700 TPM), whereas low expression (<10 TPM) was also detected in other tissues (Fig. 6B). Phylogenetic tree analysis indicated that COTS oki.208.17 was most closely related to echinoderm PI16s, which formed a sister clade with vertebrate PI16/PI16-like proteins (Fig. 6C). Snake venom-type CRISPs and insect venom allergen 5, were more distantly related.

Other excretory-secretory proteins identified from COTS aboral spine extracts

Multiple ependymin-related proteins (EPDRs) were identified from COTS spine sample preparations (Table 2; oki.11.66 and oki.141.73) were detected from spine extracts, and with their gene expression highest in the spines (Fig. 3B). Similarly, a pancreatic lipase-related protein (oki.162.19) and an uncharacterized protein (oki.111.24) are highly expressed in spines. The uncharacterized protein did not match to any other species in the NCBR nr database or had any recognised domains, however, the predicted mature protein was rich in cysteines (18 cysteine residues). Although identified from spine extracts, single DMBT1-like (oki.83.49) and kielin/chordin-like proteins (oki.258.16), as well as three Vtg-like proteins (oki.36.13, oki.36.14, and oki.185.68) that contained a ‘vitellogenin_N superfamily’ and ‘VWD superfamily’ motifs, were most highly expressed in the mouth and sensory tentacles, while spine expression was low (Fig. 3B). The secreted DMBT1-like protein contained 3 domains (Table 2): the F5/8 Type C domain (also known as discoidin domain), a calcium-binding EGF-like domain, and a scavenger receptor Cys-rich (SRCR) domain, which were repeated throughout the entire protein.