Injury affects coelomic fluid proteome of the common starfish, Asterias rubens

Extracellular body fluids, such as blood plasma in vertebrates and hemolymph and coelomic fluid (CF) in invertebrates, serve as a transport system for metabolites, nutrients, signaling and defense molecules in virtually all animals. Extracellular body fluids contain soluble factors that are constitutively secreted as well as those that are released under specific physiological conditions from different sources throughout the entire body, providing an opportunity to develop an overall profile of the biological status of the organism. Various soluble factors are directly or indirectly involved in different physiological processes, such as wound healing, inflammation, tissue remodeling and cell migration. These soluble factors will leak into the extracellular body fluids, serving as biomarkers of their corresponding processes.

Echinoderms have a well-developed perivisceral coelom (Lawrence, 1987) in which CF circulates through the whole body, bathing internal tissues and organs. The volume of CF constitutes a significant fraction of the whole body mass, reaching 20% in starfish and exceeding 40% in sea urchins (Giese, 1966). Cellular elements found in CF at concentrations of 10⁵–10⁶ cells ml⁻¹ are collectively referred to as coelomocytes (Chia and Xing, 1996). Their morphology in starfish is well described (Kanungo, 1984; Sharlaimova et al., 2014). Coelomocytes are not only mediators of innate immune responses (Smith et al., 2010), but are also actively involved in the early repair phase of starfish arm regeneration (Ben Khadra et al., 2017; Ferrario et al., 2018), providing wound closure and ‘hemostatic’ activity. The number of circulating coelomocytes in starfish rapidly increases in the first hours after arm tip amputation (Pinsino et al., 2007) or an immune challenge (Coteur et al., 2002; Holm et al., 2008), indicating the importance of cellular components in the early post-traumatic period.

Previously, two injury models have been established in Asterias rubens (Kozlova et al., 2006; Sharlaimova et al., 2014) to study the dynamics of the CF cell population during the early regeneration period: puncture wound (PW) and blood loss (BL). The PW model is characterized by making a single puncture into the body wall, leading to slight bleeding, whereas BL involves arm tip amputation, excessive bleeding and artificial washing of the coelomic cavity, effectively reducing the CF cell pool by up to 90% (Sharlaimova et al., 2014). These injuries are clearly different in their consequences: PW induces a rapid and transient increase in the number of circulating coelomocytes in the few hours after injury with its cellular composition remaining unchanged (Gorshkov et al., 2009; Kozlova et al., 2006), whereas BL is followed by changes in the cellular compositions of both CF and coelomic epithelium, upregulation of protein synthesis in coelomocytes, and migration of small, poorly differentiated cells from the coelomic epithelium into the CF (Kozlova et al., 2006; Sharlaimova et al., 2014). Involvement of small, poorly differentiated cells in the renewal of CF cell populations after BL injury has been proposed (Sharlaimova et al., 2014).

Echinoderms, having an outstanding capability to regenerate body parts and even complete individuals from a fragment following self-induced or traumatic amputation processes, represent valuable models in regeneration research (Ben Khadra et al., 2017; Candia-Carnevali et al., 2009; Dolmatov, 1999). Recent high-throughput studies of regeneration in sea cucumbers (Dolmatov et al., 2018; Zhang et al., 2017) have enabled discovery of new players involved in this process; unfortunately, nothing is known about the protein composition of the CF in these animals. Previous proteomic studies of the echinoderm CF have generally focused on immune-challenge- and age-related changes in sea urchins (Bodnar, 2013; Dheilly et al., 2011; Dheilly et al., 2012; Dheilly et al., 2013). These studies analyzed whole CF without thorough removal of coelomocytes, which apparently hampered identification of scarce extracellular components in favor of highly abundant intracellular proteins. Although several proteomic studies have been performed on starfish analyzing mucous secretions (Hennebert et al., 2015), nervous system regeneration (Franco et al., 2014) and coelomocytes (Franco et al., 2011), none have addressed the protein composition of CF or its injury-related changes.

Here, we present a proteomic analysis of cell-free CF samples from A. rubens. This echinoderm provides a suitable model for proteomic analysis of the CF for the following reasons. First, regenerating specimens of A. rubens are frequently found in the wild, suggesting the existence of well-developed healing mechanisms. Second, A. rubens has well-developed perivisceral coelom in terms of both volume and size, permitting multiple rounds of collection of CF. Third, there are three A. rubens transcriptomic datasets available (Hennebert et al., 2015; Reich et al., 2015; Semmens et al., 2016), enabling the creation of a species-specific protein database. Forth, A. rubens is a common and easily obtained species of starfish in the White Sea in Russia.

Essentially, the main question we ask is: if an animal with high regenerative capacity and well-developed coelom is subjected to injury, can we identify the soluble proteins involved and are these related to regeneration? To address this question, we examined CF in the course of the early repair phase in two well-established injury models. The overall aims of this study were to (1) provide an overview of the starfish cell-free CF proteome and (2) describe the impact of injury on CF composition and reveal the injury-related soluble factors involved.

Steps were taken to ensure that animals did not suffer unnecessarily during any stage of the experiments. The animals were returned alive to their natural environment after experimentation.

Experiments were performed at the Biological Station of the Zoological Institute, Russian Academy of Sciences, on Cape Kartesh (Kandalaksha Bay, White Sea) in September 2016. Adult individuals of the common starfish, Asterias rubens Linnaeus 1758, were collected off Fettakh Island (66°20′05″ N, E33°39′07″ E). Asterias rubens is not a protected or endangered species. Animals were kept in cages at a depth of 5–6 m throughout the experimental period. We also included a 3-day acclimation and starvation period prior to experimentation in order to minimize possible perturbations of CF caused by the animal feeding and capture procedure. The starfish used had a radius ranging from 5 to 10 cm, measured from the largest arm to the center of the oral disc. Two types of experimental traumatic treatment were used: puncture wound (PW) and blood loss (BL). The experimental workflow is shown in Fig. 1. The PW (n=12 individuals, mean radius=8.3 cm) was inflicted by puncturing the aboral surface of the arm with a scalpel. The puncture resulted in an incision of approximately 10 mm and minor bleeding. The BL injury (n=8 individuals, mean radius=5.4 cm) was performed as follows: the arm tip was cut off, and the CF was drained off as completely as possible. Then, the coelomic cavity was washed with four 4 ml injections of 0.22 µm-filtered seawater into each arm tip, and the injected seawater was drained off to remove residual CF and coelomocytes. The injured animals were returned to their cages.

Two pilot studies were performed before analysis of injury responses. Firstly, individual samples of CF from untreated, recently captured starfish (n=23, mean radius=8.3±1.1 cm) were analyzed by reverse-phase HPLC to assess individual variability. Descriptive statistics are presented as means±s.d. Secondly, the effect of repeated sampling of CF on the protein profile of CF was examined by HPLC (n=5, mean radius=6.4±0.5 cm). Group differences across time in peak areas (proportions from total peak area) were evaluated using repeated-measures ANOVA, with sampling (before sampling and 6 h after sampling) as the within subjects factor. Arcsine square root transformation was used to stabilize variance and normalize proportional data.

To follow the changes occurring in response to PW and BL, we utilized a repeated-measures design to reduce the effect of biological variance among individuals. Repeated measures were realized as repeated sampling of CF from the same animals at different time points. The experimental workflow is shown in Fig. 1. A total of 12 animals were taken for PW, and eight animals for BL. Two time points were taken for PW (before injury and 6 h after injury), and three time points for BL injury (before injury, and 6 and 72 h after injury). Samples collected before injuries were used as controls for protein abundance changes caused by injuries. Samples from the same injury type and time point were pooled together to provide sufficient protein material. Four independent LC-MALDI-TOF/TOF acquisitions and one HPLC analysis of each pooled sample were performed. Therefore, two traumatic treatment (PW and BL) and five experimental groups (CPW, PW6h, CBL, BL6h and BL72h) each with one pooled biological replicate (pool of 12 samples in PW, and pool of eight samples in BL) were analyzed. Thus, our inference for each experimental group is based on comparisons of single pooled biological replicates and four LC-MALDI technical replicates complemented by HPLC analysis of undigested CF sample.

The samples of CF were collected 6 h after puncture wound (PW6h) and 6 and 72 h after blood loss (BL6h and BL72h, respectively). Samples collected just before injury served as controls (CPW and CBL). The CF was obtained from the coelomic cavity by puncturing the aboral epidermis at the arm tip with a 21 G double-ended needle (Fig. 2A) and collecting 1.5 ml of the CF by gravity flow into a microtube containing 45 μl of 0.5 mol l⁻¹ EDTA, pH 7.5. The arm tip opposite to the wounded one was used for sampling. Two low-speed centrifugations were performed to minimize damage to the fragile cells from excessive centrifugation and minimize potential contamination of the CF with non-secretory cellular proteins. Firstly, samples were centrifuged at 120 g for 10 min in a bucket rotor. Then, the supernatant was collected and subjected to centrifugation at 430 g for 10 min in a bucket rotor, followed by filtration with Spin-X centrifuge tube filters (0.45 μm, cellulose acetate membrane, Corning Costar) at 732 g for 5 min in a MiniSpin plus centrifuge (Eppendorf, Hamburg, Germany). The flow-through aliquots of 500 μl from independent samples of each experimental group were pooled (n=12 for PW, n=8 for BL), resulting in 6 and 4 ml of cell-free CF, respectively. The samples were snap-frozen in liquid nitrogen and stored at −80°C pending analysis. Each pooled sample was analyzed by HPLC.

Analysis of pooled CF samples was performed on a microbore HPLC system (MiLiChrom A-02, EcoNova, Novosibirsk, Russia). A sample volume of 100 μl was loaded onto a Jupiter C5 reversed-phase column (2×100 mm, 5 μm, 300 Å, Phenomenex, Torrance, CA, USA) and separated using a linear gradient of 15–80% B over 22 min at a flow rate of 200 μl min⁻¹. The mobile phases used were A, 0.125% (v/v) trifluoroacetic acid (TFA) in water, and B, 0.125% (v/v) TFA in acetonitrile. The column was maintained at 45°C. Absorbance was monitored at 206, 250 and 280 nm, with a UV detector slit width of 5 nm. Total protein concentration was estimated from the total peak area using A_205 nm=31 ml mg⁻¹ cm⁻¹ (Scopes, 1974). Major peaks were collected and in-solution digested with trypsin/Lys-C mix and subjected to MALDI TOF/TOF analysis.

The coelomocyte count data were obtained for each individual sample by counting formalin-fixed cells using a hemocytometer. The data were compared by non-parametric Wilcoxon matched pairs test. Data are presented as means±s.d.

Pooled CF samples were subjected to solid-phase extraction using Oasis HLB CC (Waters, Milford, MA, USA) and Strata C18-T 100 mg (Phenomenex) solid-phase extraction tubes sequentially connected and equilibrated with 10% v/v acetonitrile and 0.1% v/v TFA. Unbound compounds were removed by washing with 5 ml of equilibration buffer, and proteins were eluted with 1.5 ml of 90% v/v acetonitrile and 0.1% v/v TFA independently from both tubes. Elutes were mixed and dried with a rotor vacuum evaporator. The samples were dissolved in 100 μl of aqueous 5% v/v sodium deoxycholate (DOC) and incubated at 95°C for 5 min. In-solution digestion was performed according to the DOC protocol (Proc et al., 2010). Denatured samples were subsequently reduced with 100 μl of 20 mmol l⁻¹ dithiothreitol in 50 mmol l⁻¹ ammonium bicarbonate for 30 min at 60°C and alkylated with 100 μl of 50 mmol l⁻¹ iodoacetamide in 50 mmol l⁻¹ ammonium bicarbonate in the dark for 30 min at room temperature. The samples were diluted by the addition of 300 μl of 50 mmol l⁻¹ ammonium bicarbonate, and digested for 18 h at 37°C with 0.5 μg of trypsin/Lys-C mix (Promega, Madison, WI, USA). The digestion was stopped and the DOC was precipitated by acidifying the sample with 50 μl of 1% TFA and vortexing. Digested samples were centrifuged at 14,000 g for 10 min with a MiniSpin plus centrifuge (Eppendorf) to pellet the DOC. All digests were then desalted and concentrated with solid-phase extraction tips packed with 20 mg of Strata C18-T sorbent (Phenomenex). The samples were eluted with 600 μl of 90% v/v acetonitrile and 0.1% v/v TFA and dried with a rotor vacuum evaporator. After rehydration with 0.1% TFA, the samples were filtered and divided into four 50 μl aliquots that were used as technical replicates in the LC-MALDI-MS/MS analyses.

Peptides were separated with a Jupiter Proteo C12 reversed-phase column (1×50 mm, 4 μm, 90 Å, Phenomenex) on a microbore HPLC system (MiLiChrom A-02, EcoNova). A sample volume of 50 μl was injected and separated using a linear gradient of 10–35% B over 54 min followed by 35–90% B for 6 min at a flow rate of 50 μl min⁻¹. The mobile phases used were A, 0.125% (v/v) TFA in water, and B, 0.125% (v/v) TFA in acetonitrile. The column was maintained at 45°C. The effluent from the HPLC column was mixed with α-cyano-4-hydroxycinnamic acid (CHCA) matrix (12 mg ml⁻¹ in 95% acetonitrile) at a flow rate of 15 μl min⁻¹ via a micro tee. An in-house constructed Arduino-powered micro-fraction collector was used to deposit a total of 912 fractions of 0.5 μl in a 24×38 array on an LC-MALDI plate (SCIEX, Darmstadt, Germany). The column was washed with a saw-tooth gradient (15–80% for 4 min followed by 80–15% for 2 min, repeated eight times) and equilibrated to 10% B for 10 min before subsequent injections.

The fractionated samples were analyzed with a TOF/TOF 5800 System (SCIEX) instrument operated in the positive ion mode. The MALDI stage was set to continuous motion mode. MS data were acquired at 2400 laser intensity with 1000 laser shots/spectrum (200 laser shots/sub-spectrum) and MS/MS data were acquired at 3300 laser intensity with a DynamicExit algorithm and a high spectral quality threshold or a maximum of 1000 laser shots/spectrum (250 laser shots/sub-spectrum). Up to 35 top precursors with signal to noise ratio >30 in the mass range 850–4000 Da were selected from each spot for MS/MS analysis.

Four independent LC-MALDI-TOF/TOF acquisitions of each pooled sample were performed and processed together in one run with the Protein Pilot 4.0 software (SCIEX). The Paragon algorithm 4.0 was used in thorough mode with biological modifications and substitutions enabled. Carbamidomethyl cysteine was set as a fixed modification. The subject of Paragon searches was a pooled protein database comprising protein-coding open reading frames (ORFs) predicted from three A. rubens transcriptome shotgun assembly datasets: ovary (Reich et al., 2015), tube foot (Hennebert et al., 2015) and radial nerve (Semmens et al., 2016). Each transcriptome shotgun assembly dataset was subjected to two iterative reassemblies with iAssembler (Zheng et al., 2011). Then, ORFs were predicted with TransDecoder-v5.0.2 (Haas et al., 2013) at a minimum ORF length of 70 amino acids and using the homology option. Searches with HMMSCAN (Eddy, 2011) against Pfam-A version 31.0, and with BLASTP against a set of Echinodermata sequences downloaded from UniProtKB on 12 December 2017 were used for the homology option. Redundant sequences were then removed using CD-HIT (Li and Godzik, 2006) at a 97% identity threshold, local alignment and 80% alignment coverage for shorter sequences (-c 0.97, -G 0 and -aS 0.8 options), producing 24,796 ORFs. These were then enriched with a set of 947 non-redundant secretory and single-pass transmembrane protein-coding sequences predicted from the ovary transcriptome shotgun assembly of Asterias forbesi as described below. The resulting set was filtered with CD-HIT, producing a final database of 25,725 protein-coding sequences. The database also incorporated a list of common contaminants. False discovery rate (FDR) analysis was done by analysis of reversed sequences using the embedded PSEP tool. The mass spectrometry proteomics data and search database have been deposited at the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD010228.

where N_observed is the number of PSMs and N_observable is the number of expected peptides for the protein. If non-tryptic cleavages were detected, resulting in overrepresentation of the number of observed peptides, count data were corrected by addition of the number of observed non-tryptic cleavage sites to the number of expected peptides in Eqn 2. The list of expected peptides for the protein was obtained by in silico digestion with PeptideMass (Wilkins et al., 1997) with no missed cleavages, after removal of the signal peptide, and filtered by both a mass range of 850–4000 Da and a hydrophobicity index range of 4.9–32.2 a.u. calculated with SSRCalc (Krokhin, 2006).

The number of secretory soluble and single-pass transmembrane proteins in our database was estimated using Phobius (Käll et al., 2004). A fraction of sequences with signal peptide only, and sequences with one transmembrane region starting ≤30 amino acid residues from the N-terminus was accepted as a reasonable estimation of secretory soluble proteins. The fraction of single-pass transmembrane proteins was estimated as the sum of sequences having a signal peptide and one transmembrane region, sequences with one transmembrane region starting >30 amino acid residues from the N-terminus, and sequences containing two transmembrane regions starting ≤30 amino acid residues from the N-terminus.

Similarity searches in the UniProtKB database were performed using BLASTP implemented in Blast2GO tool (Conesa et al., 2005) at an E-value threshold of 10⁻⁵. TBLASTN searches were performed against the TSA and EST databases available at the NCBI server and restricted to Echinodermata. Sequences with undetectable similarity were submitted to HHblits at the MPI Bioinformatics Toolkit server (Zimmermann et al., 2018) under default settings, and the resulting multiple sequence alignments with E-values lower than 10⁻⁵ were entered into the HHpred server to search against the Pfam-A version 31.0 and SMART version 6.0 databases. Domain organization was analyzed with a SMART web tool (Letunic and Bork, 2018). Protein domain architectures with either different domain counts or different order were computed as independent domain combinations, even if they had the same domain types. Functional annotation was based on sequence similarity, conserved domain searches and similarity of domain organization with well-characterized proteins. If no functional terms were assigned via the BLAST search, putative terms were suggested based on conserved domain gene ontology (GO) terms and functions of proteins with similar domain organization. Glycosylphosphatidylinositol (GPI) anchoring was predicted with PredGPI (Pierleoni et al., 2008). Signal peptides were predicted with SignalP 4.0 (Petersen et al., 2011), and transmembrane regions were identified as a consensus of Phobius and TMHMM (Krogh et al., 2001) predictors.

To assess the CF composition variability, we examined individual samples from 23 untreated, recently captured starfish, with the use of reversed-phase HPLC. Substantial inter-individual differences were observed (Fig. 2B). The number of detected peaks varied from 11 to 19, with mean value of 15 peaks. Total peak area varied from 11.8 to 145.1 a.u. µl, with mean value of 46.6±29.7 a.u. µl, and a coefficient of variation (CV) of 63.9%. Total protein concentration was roughly estimated as 15.0±9.6 µg ml⁻¹. No correlation between total peak area and starfish size was observed. We then estimated variability by the eight most abundant peaks occupying 91.9±6.3% of total peak area. All peaks showed a right-tailed distribution with high variance and a CV ranging from 55% to 204% (Fig. 2C; Table S1). An interesting finding was a strong, positive correlation between the WAPL and KUTZ peaks (R=0.82, P=0.0001).

The observed inter-individual variability motivated us to use a repeated-measures experimental design as an effective way to exclude the effects of individual differences. Then, we examined the effect of repeated sampling of CF. No significant differences between HPLC profiles of CF samples collected 6 h apart were observed (Fig. 2D,E). Thus, starfish appear to be insensitive to the CF sampling method used.

Within the four experimental groups, we identified in total 119 proteins with a local FDR of 5%. The identification list was then filtered against a stringent threshold, leading to a final list of 91 accepted protein identifications (Table S2). Of these, 23 sequences were putative secretory proteins identified with only one high-scoring PSM (Figs S1–S23), retained so as not to exclude potential targets that may have been near the limit of detection. The position of the signal peptide cleavage site was confirmed for 20 proteins by identification of a corresponding peptide with an N-terminal non-tryptic cleavage site (Table S2). Processing of CRH-type peptide and progranulin into mature forms was confirmed by detection in the CF of fragments originating only from the predicted mature peptides. In spite of our database containing a partial coding sequence for progranulin, we detected four mature peptides corresponding to human granulins 3, 4, 5 and 6 (Fig. S24).

As expected, computational analysis based on SignalP, TMHMM, Phobius and PredGPI predictions showed that 77 of 91 proteins identified were predicted to be extracellular or bound to the plasma membrane. The proteins were grouped into four cellular location categories (Fig. 3A). Putative functions of identified proteins were assigned by searching for homologous sequences in the UniProt KB database using the Blast2GO tool. A relatively high number of sequences (59) showed similarity (at an E-value threshold of 10^–5) with annotated proteins, whereas 16 sequences showed homology to uncharacterized Strongylocentrotus purpuratus sequences and 16 sequences did not return any hits (Table S2). We then performed an analysis of protein domains and domain architectures derived from identified proteins (Figs 4 and 5). As a result, we obtained 68 unique domain architectures composed of 68 distinct domain types. EGF-like and SRCR domains occurred most frequently in unique architectures, six and five times, respectively. Significant portions of domain architectures were composed of only one domain: there were 29 single-domain, eight two-domain and 30 multi-domain architectures. The most complex architectures were observed in HYR/EGF/CCP (HEC) domains containing protein, consisting of 53 domains and six distinct domain types; and in coelomotrypsin (CLTR), consisting of 14 domains and six distinct domain types. Combinations of sequence similarity search and domain analysis enabled us to classify proteins into 12 categories according to their putative molecular function terms (Fig. 3B). The exceptions were nine proteins that could not be classified and were grouped as ‘uncharacterized secretory soluble’.

There were four uncharacterized secretory soluble proteins without any detectable similarity and recognizable domains: CHUPA (CHU), KARTESH (KASH), wound associated starfish peptide (WASSP) and TripleWG protein. CHU, KASH and WASSP are short polypeptides comprising predicted N-terminal signal peptides. Searches of available transcriptomes of Echinodermata using TBLASTN identified close homologs in starfish (Asteroidea) and sea urchins (Echinoidea) for CHU (Fig. 6A), whereas KASH and WASSP appear to be specific to Asteroidea (Fig. 6B,C). Both CHU and KASH peptides share some structural similarity: the N-terminal regions contain stretches of acidic residues followed by dibasic convertase cleavage sites, which indicates that this part may be an acidic spacer peptide, and both peptides bear highly conserved four-cysteine residue motifs, but with different cysteine spacing. All of these peptides were identified only in the echinoderms, and their relationships with peptides identified in other phyla are unclear.

TripleWG is a 1001-residue protein named on account of two predicted domains arranged in tandem and repeated three times. These domains are: ∼130 amino acid residues with a conserved WYR motif, and ∼190 amino acid residues with a conserved GW motif. Several remote homologs were detected for each predicted domain among invertebrates and lower chordates (Fig. S25); however, the observed domain organization appears to be unique to echinoderms.

To discriminate putative proteins with injury-related production, the list of all LC-MALDI identified CF proteins was split up over the separate injury types and time points (Fig. 7A). The PW treatment yielded 48 identifications before injury (CPW) and 62 identifications 6 h post-injury (PW6h). The BL treatment yielded 42 identifications before injury (CBL), 62 identifications 6 h post-injury (BL6h) and 37 identifications 72 h post-injury (BL72h). Comparing these sets, eight proteins were common to both PW6h and BL6h, while 20 proteins were exclusively associated with BL6h, and only six with PW6h. Of these 34 proteins with injury-related production, 13 were identified with a single PSM. The most apparent case of a protein identified uniquely and abundantly in both types of trauma was the ZP domain containing protein-1 (ZPD1). There was only one unique protein specific to the BL72h group when compared with both PW and BL trauma, namely the chupa-2 peptide (CHU2). It is noteworthy that WASSP was identified in both PW6h and BL72h, indicating delayed upregulation of this peptide following blood loss. Most of the proteins identified in control groups were within an overlap of 27 proteins across all groups. Thus, our data indicate that BL injury caused more prominent qualitative changes in CF composition, revealed by the identification of three times more unique proteins than in PW.

We used a G-test to compare proportions of observed and expected spectral counts for proteins overlapping between control and experimental groups. Differences were significant comparing PW6h with CPW (G=139.1, d.f.=43, P<0.001), BL6h with CBL (G=146.3, d.f.=36, P<0.001) and BL72h with BL6h (G=133.6, d.f.=33, P<0.001). Subsequently, we used the McNemar test for determination of differentially abundant proteins and illustrated injury-related changes by means of a heat map depicting both PAI and putative functional terms (Fig. 7B). We identified four differentially abundant proteins in the PW6h group and eight proteins in the BL6h group, with an overlap of two proteins. Of note, by 72 h post-injury, the BL group had downregulated six proteins and 18 of 20 proteins unique to the BL6h group had disappeared. Comparison of BL72h with CBL (no difference, G=30.7, d.f.=27, P=0.28) revealed that tendency to return to control values was observed by 72 h post-injury, with the exception of three unique proteins – WASSP, KASH2 and CHU2 – which actually showed delayed upregulation. In addition to the components showing differential regulation, there were abundant proteins unaffected across all injury types. The most apparent cases of these are mucin-related protein (MURP), peptidoglycan recognition protein (PGRP), the NlpC/P60 domain containing protein (NLPC) and Ly6-like protein-4 (LYSTAR4).

Both PW and BL sets of pooled cell-free CF samples showed nine to 10 major protein peaks, clearly identifying the most abundant CF proteins (Fig. 8A,B). The 10 most abundant HPLC peaks were collected and subjected to digestion and MS/MS analysis. Of these, nine protein peaks were identified: MURP, Asterias large peptidase inhibitor (ALPI), PGRP, Kunitz-type serine protease inhibitor (KUTZ), WAP-like protein (WAPL), CTL1, TIMP-related protein (TIRP), NLPC and vitellogenin-1 (VTG1). As expected, these proteins were among those with high PAI values determined by the quantitative proteomic approach (Fig. 7B).

HPLC profiles of control and injured starfish were substantially different in peak intensities (Fig. 8A,B). Injury resulted in 1.8- and 2.4-fold increases of total peak area for PW and BL, respectively. Injury-related changes in both the PW and BL sets, monitored by abundant components, showed similar trends in accordance with the quantitative proteomic data. Thus, HPLC analysis showed a good agreement with data obtained by a quantitative proteomic approach, independently revealing the most abundant CF proteins and their relative changes.

In order to correlate proteomic and cellular responses, we examined the effect of PW and BL injury on the concentration of circulating coelomocytes. Time-dependent changes of the coelomocytes’ concentration in response to injury are summarized in Fig. 8C. At 6 h after PW, the coelomocytes' concentration showed a significant 5-fold (P=0.002) increase from 1.52±1.12×10⁶ ml⁻¹ to 7.50±4.78×10⁶ ml⁻¹. Notably, at 6 h after BL, the coelomocytes' concentration significantly increased 4.2-fold (P=0.012), from 0.44±0.36×10⁶ ml⁻¹ to 1.86±1.28×10⁶ ml⁻¹, followed by a 3.5-fold decrease (P=0.012) to 0.52±0.90×10⁶ ml⁻¹ by 72 h post-injury. These data show that proteomic responses and increases in cell concentration are synchronous events, suggesting a functional relationship between these processes.

Here, we provide an overview of the proteins comprising cell-free CF of the common starfish, A. rubens. To gain insight into the CF proteome, we paid special attention to isolation of cell-free CF. We utilized needle sampling without aspiration, gentle cell centrifugation, and filtration of the supernatant to avoid contamination by intracellular proteins. By this approach we aimed to obtain as accurate as possible a set of extracellular proteins that are actually present in the CF.

The main goal of this study was to analyze early responses of CF to injury utilizing two traumatic treatments. Changes in CF composition revealed in injured starfish at both the cellular and proteomic levels for both types of injuries suggest the importance of both cellular and humoral components in early responses to injury. We found that, in spite of the vast loss of whole CF after BL injury, the proteomic response was more pronounced in BL6h than in PW6h, as revealed by both the LC-MALDI and HPLC approaches. We suppose that a vast loss of the entire CF and depletion of the coelomocyte depot are major factors responsible for the observed proteomic changes after BL injury. Significant post-traumatic increases in coelomocyte concentrations for both types of injury indicate that a large number of coelomocytes are withdrawn from circulation in normal state, and imply the existence of mechanisms effectively regulating recruitment of cells into the circulation.

Through this work, we showed that PW and BL may share a common core of injury-responsive proteins. Identification of proteins common to both types of injuries is important, as they suggest underlying general mechanisms of response to injury, whereas identification of proteins specific to a particular injury may suggest activation of different physiological processes contributing to functional discrimination between PW and BL. In the following sections we will discuss key proteins influenced by injury and suggest hypotheses about their potential roles in the physiology of starfish.

The largest fraction of the CF proteome is represented by pattern recognition receptors. As echinoderms, starfish lack an acquired immunity, relying on the presence of pattern recognition receptors for discriminating and eliminating pathogens (Smith et al., 2010). The carbohydrate binding proteins and C-type lectins in particular are key molecules of the echinoderm immune system. We detected four C-type lectins and fucolectin-related protein to be upregulated after injury. The identified C-type lectins share significant similarities with N-acetylgalactosamine-specific lectins from the starfish Asterina pectinifera (Kakiuchi et al., 2002). Notably, only CTL1 showed downregulation in both injury types. It is tempting to assume that CTL1 is consumed during cellular clot formation in response to injury. Induction of cell aggregation by C-type lectin in tunicates has been demonstrated (Matsumoto et al., 2001). It is noteworthy that amassin, a CF protein mediating calcium-dependent clot formation and occupying approximately 1% of total plasma protein in sea urchins (Hillier and Vacquier, 2003), was not detected in our samples. The absence of amassin is remarkable, because clotting in starfish is a calcium-dependent process (Kanungo, 1982), and this may indicate that starfish and sea urchins follow different pathways in clot formation.

Lytic polysaccharide monooxygenase-related protein (LPMO) is unique to BL injury. LPMOs are primarily bacterial and fungal cellulose degrading enzymes (Forsberg et al., 2014). Some of the LPMO-related proteins have lost their enzymatic activity but retained their carbohydrate binding capacity, such as chitin-binding protein from Streptomyces (Schnellmann et al., 1994), suggesting a possible role in carbohydrate recognition for starfish LPMO.

Other important components of the echinoderm immune system are SRCR-domain containing proteins, which are also well represented in the starfish CF proteome. Sea urchin coelomocytes express a highly complex set of SRCR genes with pronounced variability in expression levels (Pancer, 2000). In A. pectinifera, membrane and shed soluble forms of the SRCR-related protein ApSRCR1 play the role of opsonins in innate immunity (Furukawa et al., 2012). FASR proteins in addition to SRCR domains also contain a coagulation factor 5/8 C-terminal domain that is implicated in cell adhesion and carbohydrate binding (Baumgartner et al., 1998).

The EGF and GUB domain-containing protein ECU2 was affected only by BL injury. This protein resembles bone morphogenic protein 1 (BMP1) owing to the presence of both EGF and GUB domains. A role of BMP1 in intestine regeneration in the sea cucumber Holothuria glaberrima has been demonstrated (Mashanov et al., 2012). The CUB domains are represented in extracellular and developmentally regulated proteins, where they perform roles in protein recognition (Bork and Beckmann, 1993). EGF repeat-containing protein ERP resembles fibropellins due to the presence of a long stretch of EGF repeats. Fibropellins are components of the extracellular matrix that play a role in sea urchin embryo development (Bisgrove et al., 1991).

An interesting finding was a pair of ZP domain-containing proteins – ZPD1 and ZPD2 – sharing only 29% identity to each other. Remarkably, ZPD1 is a soluble protein, whereas ZPD2 is a transmembrane receptor. The ZP domain acts as a module for polymerization of extracellular proteins into filaments or matrices (Jovine et al., 2005), suggesting a function for identified ZPD proteins in recognition and cell adhesion. Thus, a rich repertoire and differential regulation of pattern recognition receptors may indicate activation of non-self-recognition during early response to injury.

Interestingly, only single-pass transmembrane proteins without any intracellular signaling domains were identified in the CF proteome. We propose that these proteins are released into circulation from the cell surface by the proteolytic mechanism known as ectodomain shedding, liberating biologically active soluble forms (Hayashida et al., 2010). The functions of these proteins are probably related to cell adhesion, pathogen recognition and phagocytosis, whereas signal transduction activity appears to be missing. Only ZPD2 of eight detected transmembrane proteins was upregulated in response to injury, indicating increased shedding rates of this receptor, which in turn may imply regulation of cell adhesion and migration in the early response to injury.

An interesting finding was the identification of novel collagen-related protein (CORP), a single-pass transmembrane type II protein containing two C-terminal type IV collagen domains targeted to the extracellular space. The putative function of CORP may be related to cell adhesion, similarly to vertebrate transmembrane collagens (Gordon and Hahn, 2010) and collagen IV in Drosophila (Dai et al., 2017).

We identified four proteins related to the large lipid transfer protein (LLTP) superfamily (Smolenaars et al., 2007), which contribute to discrimination between BL and PW injuries. The roles of LLTPs extend beyond the energy store and nutrient functions. Vitellogenins and apolipophorins are recognized as lipopolysaccharide (LPS)-responsive acute-phase proteins acting in innate immunity through pattern recognition receptor activity and direct bactericidal action (Zdybicka-Barabas and Cytryńska, 2013; Zhang et al., 2011). Recent studies have demonstrated that recombinant DUF1943 and vWFD domains of LLTPs possess pattern recognition, opsonizing and antibacterial activities in zebrafish (Sun et al., 2013a), a bivalve mollusk (Wu et al., 2015) and a scleractinian coral (Du et al., 2017). Thus, upregulation of LLTPs in starfish may imply its involvement in defense and further indicates activation of innate immunity in response to injury. However, in Drosophila, apolipophorins are involved in signaling as vehicles for lipid-linked morphogens Wnt and Hedgehog (Panáková et al., 2005), whereas Wnt genes are implicated in intestine regeneration of sea cucumbers (Mashanov et al., 2012; Sun et al., 2013b) and in early development of the starfish Patiria miniata (McCauley et al., 2013). Therefore, a specific transport role of LLTPs in an early response to injury cannot be excluded.

We also highlight the role of signal transducers in response to injury. Eight proteins were identified as signal transducers circulating in the CF, of which only serum amyloid A was upregulated, contributing to discrimination between BL and PW injuries. Serum amyloid A is a major vertebrate acute-phase protein with chemokine-like properties induced in response to infection, inflammation and injury. Potential functions include induction of migration and chemotactic recruitment of immune cells to the site of injury, and induction of proinflammatory cytokines and extracellular matrix-degrading enzymes (Uhlar and Whitehead, 1999). Close associations between intestinal regeneration and expression of serum amyloid A have been demonstrated in the sea cucumber H. glaberrima (Santiago et al., 2000). Moreover, LPS administration to sea cucumbers induces both a cellular response and upregulation of intestinal expression of serum amyloid A (Santiago-Cardona et al., 2003). Thus, upregulation of serum amyloid A in starfish CF may indicate activation of cell migration and regenerative processes in early responses to PW injury.

An important finding was identification of progranulin processed into four granulin-like peptides. Granulins are small 6 kDa peptides derived from one large precursor. Granulins have multiple biological roles: they stimulate proliferation and angiogenesis in wounds, support tumor growth and are involved in early embryogenesis (Bateman and Bennett, 1998; Bateman and Bennett, 2009). Granulin-like growth factor secreted by the human liver fluke induces angiogenesis and accelerates wound healing of mammalian host tissues in vivo (Smout et al., 2015). Against this background, it will be of great interest to investigate the physiological role of echinoderm granulin-like peptides in the course of response to injury.

An interesting finding was the diversity of Ly6-like proteins (LYSTAR), relatively abundant in CF, but unaffected by injury. The LYSTARs are remotely similar to the Ly6/uPAR superfamily. This superfamily includes secreted signaling proteins and receptors with diverse functions, including protection of cells from complement-mediated lysis (Fletcher et al., 1994), modulation of neuronal excitability targeting nicotinic acetylcholine receptors (Lyukmanova et al., 2016; Tsetlin, 2015), cell adhesion and migration (Alapati et al., 2014; Hänninen et al., 1997), induction of chemotaxis of hematopoietic stem cells (Selleri et al., 2006), and antimicrobial activity (Liu et al., 2017). Because acetylcholine is the major excitatory neurotransmitter in echinoderms (Devlin, 2001) and affects mechanical properties of mutable collagenous tissue (Wilkie, 2002), it is tempting to assume that LYSTARs may act as endogenous modulators of acetylcholinergic transmission in starfish.

We also report identification of mature corticotropin releasing hormone-type peptide (CRH), previously discovered in the A. rubens radial nerve transcriptome (Semmens et al., 2016). In spite of the involvement of CRH in the stress response in vertebrates (Koob, 1999), it was unaffected by injury, suggesting it may play another role in echinoderms.

Three collagen-like proteins were identified 6 h after PW and BL injury. Collagen is an important component of starfish tissues, playing a major role in determining the mechanical properties of the body wall (Blowes et al., 2017). However, extracellular matrix (ECM) components should not normally be present in the CF, and its identification may be indicative of ECM degradation. In sea cucumbers, degradation of the fibrous collagen by matrix metalloproteinases (MMPs) occurs during the early stages of intestine regeneration (Lamash and Dolmatov, 2013; Miao et al., 2017; Quiñones et al., 2002). Cell migration through the ECM is regulated by membrane-anchored MMPs, and some of them possess collagenase activity (Hotary et al., 2000; Quaranta, 2000). Of particular interest is that the appearance of collagens was correlated with an increase in coelomocyte concentrations after both injury types. In A. rubens, consecutive bleedings have been shown to cause a release of coelomocytes from the coelomic epithelium (Vanden Bossche and Jangoux, 1976) and migration of a specific sub-population of small epitheliocytes into the coelomic cavity (Sharlaimova et al., 2014). Degraded collagens detected in human serum during cancer progression indicate ECM degradation and cell invasion (Kehlet et al., 2016). Therefore, we propose that the appearance of collagens in CF in response to injury may indicate ECM degradation occurring during cell migration into the coelomic cavity.

Another remarkable group includes six proteins identified as hydrolases. Of these, only cathepsin B and coelomotrypsin are proteases that potentially could be involved in ECM degradation. Cathepsin B is a lysosomal cysteine protease, and in vertebrates it is involved in tumor cell proliferation, invasion and angiogenesis (Aggarwal and Sloane, 2014). Increased activity of MMPs through degradation of tissue inhibitors of matrix metalloproteinases (TIMPs) by cathepsin B has been demonstrated (Kostoulas et al., 1999).

Coelomotrypsin is a novel multidomain serine protease consisting of an N-terminal set of recognition and protein binding domains and a C-terminal trypsin-like domain. Proteins with a similar domain architecture have been identified in the P. miniata skeletal proteome by a proteomic approach (Flores and Livingston, 2017). The only characterized proteins that have a somewhat similar domain organization are vertebrate secreted proteases such as neurotrypsin, plasminogen activator and plasminogen. All of these are secreted as inactive zymogens and subjected to proteolytic activation, and are involved in various processes: neural plasticity (Stephan et al., 2008), ECM remodeling, cell invasion, adhesion and migration (Irigoyen et al., 1999).

Both lysozyme (Sana et al., 2004) and peptidoglycan recognition protein have been shown to have antimicrobial and bacteriolytic activity because both degrade peptidiglycans, major components of the bacterial cell wall (Callewaert and Michiels, 2010; Coteur et al., 2007). Antimicrobial activity can also be proposed for NlpC/P60 domain containing protein (NLPC), which shares a remote similarity with the NlpC/P60 superfamily of phage-associated bacteriolytic enzymes and bacterial peptidoglycan hydrolases involved in cell wall biogenesis (Anantharaman and Aravind, 2003). Remarkably, the N-terminus of NLPC is highly similar to the only sequenced N-terminal fragment of interleukin-1-like protein (PIR: A61273) purified from CF of A. forbesi (Beck and Habicht, 1991). Strong evidence of IL-1-like activity has been demonstrated in vertebrate assay systems, but no analyses have been performed on echinoderms (Beck and Habicht, 1991).

An interesting finding was the identification of PTX in CF, the major lethal factor from venomous spines of the crown-of-thorns starfish, Acanthaster planci (Shiomi et al., 2004). PTX is a toxic DNase II able to enter into the cell and induce apoptosis through DNA degradation (Ota et al., 2006). The protective role of PTX in mucous secretions of A. rubens (Hennebert et al., 2015) is obvious, whereas the presence of PTX freely circulating in the CF may suggest its involvement in internal defense against eukaryotic pathogens.

In striking contrast to the scarcity of proteases in CF was the high number and abundance of peptidase inhibitors. Our CF proteome contained 13 proteins predicted to function as peptidase inhibitors, including Kunitz, Kazal, WAP, TIL, TIMP, agrin N-terminal and BMSP domains. The WAP-like protein is a novel member of the whey acidic proteins (WAP) identified in echinoderms for the first time. It is noteworthy that WAP-like protein is a major and injury-responsive protein of starfish CF, whereas WAP is a major milk protein in some mammals. WAP domain proteins are pleiotropic molecules and, in addition to peptidase inhibitor activity, they have demonstrated antimicrobial, antiviral and anti-inflammatory activities (Scott et al., 2011). WAP domain proteins are also associated with tumor progression and have been recognized as markers for several cancers (Bouchard et al., 2006).

Novel multidomain protein ALPI includes EGF, vWFC, thyroglobulin and Kunitz-type domains, which define its putative peptidase inhibitor and protein binding activities.

The proteome of CF also contains two metalloproteinase inhibitors: TIMP and TIRP. The latter contains a single agrin N-terminal domain, structurally related to TIMPs. The function of TIMPs is related to the regulation of the MMPs activity discussed above. Thus, although MMPs were not identified in our study, their presence is suggested by the identification of TIMPs.

An important protein distinguishing PW from BL injury is β-microseminoprotein (BMSP), also known as prostate secretory protein PSP94. The BMSP is present in different human mucous secretions (Weiber et al., 1990), is involved in inhibition of tumor growth (Sutcliffe et al., 2014), possesses antifungal activity (Edström et al., 2012) and is overexpressed in early stages of intestine regeneration in the sea cucumber A. japonicus (Zhang et al., 2017). However, the exact molecular function of this important protein is not yet clearly understood.

We propose that a rich repertoire of peptidase inhibitors circulating in the CF are required for wound healing and regeneration, probably by regulating cell migration and ECM remodeling by orchestrating protease activity. However, an involvement in innate immunity through direct antimicrobial action is also possible.

This category includes several important proteins that further contribute to differentiation of BL from PW injury. Ependymin-like proteins are present in vertebrates, invertebrate deuterostomes, and protostomes (Suárez-Castillo and García-Arrarás, 2007). These proteins have been associated with long-term synaptic plasticity (Shashoua, 1991), are overexpressed in human colorectal tumor cells (Nimmrich et al., 2001), are involved in intestine regeneration in the sea cucumber (Suárez-Castillo et al., 2004) and are downregulated in vertebrate hematopoietic progenitor cells with the onset of proliferation and differentiation (Gregorio-King et al., 2002). Calcium-dependent involvement of vertebrate ependymin in cell-matrix adhesion has also been proposed (Hoffmann, 1994).

The only characterized protein having a domain organization similar to the identified TSR-containing proteins is vertebrate properdin, which has seven thrombospondin-type repeats. Properdin in involved in innate immunity through activation of the complement alternative pathway, pattern recognition receptor activity, and recognition and clearance of self-apoptotic and malignant cells (Kemper et al., 2010). It is important to note that no components of the complement system previously reported in echinoderms (Smith et al., 2010) were detected in our samples. Therefore, the complement system appears to be unaffected or expressed at undetectable levels under the conditions that we tested.

The importance of vWFC-related protein (VREP) in early responses to BL arises from its domain architecture similarity with vertebrate brorin, namely two von Willebrand factor C domains. Brorin, a member of the chordin family, is involved in neurogenesis in vertebrates as a BMP antagonist (Miyake et al., 2017), whereas chordin shapes the BMP morphogen gradient during sea urchin embryo development (Lapraz et al., 2009).

High affinity of vertebrate avidin to biotin is generally thought to inhibit bacterial growth, whereas the avidin-like domain of sea urchin fibropellin does not bind to biotin and was adopted for oligomerization (Itai et al., 2005). However, tissue-specific expression of avidin-like genes during early and late stages of intestine regeneration in the sea cucumber A. japonicus has been reported (Ba et al., 2015), suggesting an important role for starfish avidin-like protein in response to BL injury.

Asterlysin is a novel starfish protein with a bacterial aerolysin toxin domain. Aerolysin is a cytolytic pore-forming toxin from a gram-negative bacterium Aeromonas hydrophyla that targets eukaryotic cells (Howard et al., 1987), whereas eukaryotic aerolysins serve as defense molecules in Biomphalaria snails (Galinier et al., 2013) or assist in osmotically driven prey disintegration in hydra (Sher et al., 2008).

The group of actin-binding proteins was also upregulated after injury and further contributes to differentiation of BL from PW injury. Intracellular proteins are often considered to be contaminants or tissue leakage proteins in studies of extracellular proteomes. Because our CF isolation procedure was intended to produce non-traumatic sampling and gentle cell removal, an alternative interpretation is that these proteins may be genuine components of the CF, playing a role distinct from their intracellular function. Indeed, release of actin upon cell death and formation of actin filaments in body fluids are detrimental, and an actin scavenger system exists in vertebrates to sequester actin and facilitate its clearance from circulation (Lee and Galbraith, 1992). Thus, cofilin-like protein with actin filament depolymerisation activity may be a principal component of actin scavenger systems in starfish, like gelsolin in vertebrates, stabilizing extracellular actin in a globular form.

In contrast, there is strong evidence of direct involvement of actin in invertebrate innate immunity. In Drosophila, extracellular actin triggers a response associated with wounding and dead cell clearance (Srinivasan et al., 2016). Actin-derived antimicrobial peptides have been identified in A. rubens (Maltseva et al., 2007), and extracellular actin has been shown to mediate antimicrobial defense in the mosquito Anopheles gambiae (Sandiford et al., 2015). Interestingly, bactericidal activity has been shown to be more profound for globular actin, and our identification of the cofilin-like protein is in line with these data.

High-throughput transcriptomic and proteomic approaches targeted on non-model invertebrates have often revealed novel proteins that lack database matches or conserved domains that would allow for speculation about their functions. Indeed, approximately two-thirds of differentially expressed genes identified from transcriptomic analyses in echinoderms are ‘unknown genes’ not included in GO analysis (Fuess et al., 2015; Sun et al., 2013b). In starfish CF, we identified eight uncharacterized ‘orphan’ proteins, of which the most intriguing are injury-responsive Chupa, Kartesh and WASSP peptides. It is noteworthy that novel Chupa and Kartesh peptides are present in several isoforms, suggesting some functional divergence. Physiological roles of these peptides currently cannot be hypothesized; however, probable functions of small extracellular cysteine-rich proteins are predominantly related to signaling (hormones, growth factors) and binding (enzyme inhibitors, toxins, defensins). Identification of novel proteins with taxonomically restricted distribution is especially exciting because these might be responsible for adaptations specific for starfish, and can provide a unique opportunity to investigate the physiological roles of these ‘orphans’ in early responses to injury.

The data that we present here provide an overview of the coelomic proteins of the starfish A. rubens, expanding our knowledge of echinoderm CF proteomes. However, our work clearly has some limitations. A major source of unreliability is the use of single pooled biological replicates in proteomic analyses. Given that our quantitative proteomic findings are based on a single replicates, the results from such analyses should be treated with caution. Nevertheless, our data yielded valuable insights into the process of the response to injury, encouraging further investigations using larger sample sizes and repeated measures over a longer period of time, targeting more confident and precise delineation of the injury-related changes.

Through this work, we revealed a list of injury-responsive proteins potentially involved in defense, cell migration and wound healing, and demonstrated dramatic effects of injury on the CF proteome. Injury-responsive proteins, such as KUTZ, WAPL, TIRP and VTG1, both abundant and easily detected by HPLC, seem to be suitable candidate biomarkers of injury in starfish. Injury-responsive proteins with no clear homology, such as ALPI, VREP, ECU2 and other ‘orphans’, are also of outstanding interest for functional studies.

Although our findings suggest that BL is distinguished from PW by a set of proteins, the predicted functional terms are too general, hindering recognition of injury-specific physiological processes. This is a second limitation of our work and a general challenge of studies on non-model invertebrates. Therefore, further experiments on recombinant protein expression, tertiary structure determination and functional activity testing are needed to clarify the physiological roles of these proteins. Nevertheless, an overlap of both injury-responsive proteins and general functional terms between injuries could suggest five common processes playing a role in early responses to injury: (1) activation of innate immunity involving defense and non-self-recognition; (2) regulation of proteolysis by a rich repertoire of peptidase inhibitors; (3) activation of proteolysis and degradation of collagenous ECM; (4) regulation of cell adhesion and migration; and (5) activation of regenerative processes. Our assumptions about the activation of regenerative processes is supported by an overlap of injury-responsive proteins with those involved in regeneration in sea cucumbers, namely β-microseminoprotein, serum amyloid A, and ependymin-like and avidin-like proteins. Observed associations of cancer-related terms with such proteins as WAPL, LYSTARs, GRAN, BMSP, CATB, CLTR and ENDL is not surprising, because carcinogenesis involves both cell migration and transdifferentiation, which are thought to be important mechanisms of regeneration in echinoderms (Kalacheva et al., 2017). Therefore, our dataset may represent an important tool for discovery of novel proteins involved in regeneration in echinoderms, suggesting important targets for future studies. Moreover, observed signs of innate immune response call for further studies comparing responses to immune challenge and wounding, and targeting more precise discrimination of regeneration-related components.

In summary, despite the limitations of this study, our CF proteome represents an important starting point to understand the underlying mechanisms of early response to injury in echinoderms, highlights the importance of both ‘orphan’ proteins and proteins conserved throughout the deuterostomian lineage, and provides the opportunity to investigate the physiological roles of these proteins.

We are grateful to the staff of the White Sea Biological Station ‘Kartesh’ of Zoological Institute of the Russian Academy of Sciences for the provided facilities and assistance. We thank Drs Maurice Elphick and Patrick Flammang for sharing assembled transcriptome datasets.