A GM1b/asialo‐GM1 oligosaccharide‐binding R‐type lectin from purplish bifurcate mussels Mytilisepta virgata and its effect on MAP kinases

A 15‐kDa lectin, termed SeviL, was isolated from Mytilisepta virgata (purplish bifurcate mussel). SeviL forms a noncovalent dimer that binds strongly to ganglio‐series GM1b oligosaccharide (Neu5Acɑ2‐3Galβ1‐3GalNAcβ1‐4Galβ1‐4Glc) and its precursor, asialo‐GM1 (Galβ1‐3GalNAcβ1‐4Galβ1‐4Glc). SeviL also interacts weakly with the glycan moiety of SSEA‐4 hexaose (Neu5Acα2‐3Galβ1‐3GalNAcβ1‐3Galα1‐4Galβ1‐4Glc). A partial protein sequence of the lectin was determined by mass spectrometry, and the complete sequence was identified from transcriptomic analysis. SeviL, consisting of 129 amino acids, was classified as an R(icin B)‐type lectin, based on the presence of the QxW motif characteristic of this fold. SeviL mRNA is highly expressed in gills and, in particular, mantle rim tissues. Orthologue sequences were identified in other species of the family Mytilidae, including Mytilus galloprovincialis, from which lectin MytiLec‐1 was isolated and characterized in our previous studies. Thus, mytilid species contain lectins belonging to at least two distinct families (R‐type lectins and mytilectins) that have a common β‐trefoil fold structure but differing glycan‐binding specificities. SeviL displayed notable cytotoxic (apoptotic) effects against various cultured cell lines (human breast, ovarian, and colonic cancer; dog kidney) that possess asialo‐GM1 oligosaccharide at the cell surface. This cytotoxic effect was inhibited by the presence of anti‐asialo‐GM1 oligosaccharide antibodies. With HeLa ovarian cancer cells, SeviL showed dose‐ and time‐dependent activation of kinase MKK3/6, p38 MAPK, and caspase‐3/9. The transduction pathways activated by SeviL via the glycosphingolipid oligosaccharide were triggered apoptosis. Database Nucleotide sequence data have been deposited in the GenBank database under accession numbers MK434191, MK434192, MK434193, MK434194, MK434195, MK434196, MK434197, MK434198, MK434199, MK434200, and MK434201.

A 15-kDa lectin, termed SeviL, was isolated from Mytilisepta virgata (purplish bifurcate mussel). SeviL forms a noncovalent dimer that binds strongly to ganglio-series GM1b oligosaccharide (Neu5Acɑ2-3Galb1-3Gal-NAcb1-4Galb1-4Glc) and its precursor, asialo-GM1 (Galb1-3GalNAcb1-4Galb1-4Glc). SeviL also interacts weakly with the glycan moiety of SSEA-4 hexaose (Neu5Aca2-3Galb1-3GalNAcb1-3Gala1-4Galb1-4Glc). A partial protein sequence of the lectin was determined by mass spectrometry, and the complete sequence was identified from transcriptomic analysis. SeviL, consisting of 129 amino acids, was classified as an R(icin B)-type lectin, based on the presence of the QxW motif characteristic of this fold. SeviL mRNA is highly expressed in gills and, in particular, mantle rim tissues. Orthologue sequences were identified in other species of the family Mytilidae, including Mytilus galloprovincialis, from which lectin MytiLec-1 was isolated and characterized in our previous studies. Thus, mytilid species contain lectins belonging to at least two distinct families (R-type lectins and mytilectins) that have a common b-trefoil fold structure but differing glycan-binding specificities. SeviL displayed notable cytotoxic (apoptotic) effects against various cultured cell lines (human breast, ovarian, and colonic cancer; dog kidney) that possess asialo-GM1 oligosaccharide at the cell surface. This cytotoxic effect was inhibited by the presence of anti-asialo-GM1 oligosaccharide antibodies. With HeLa ovarian cancer cells, SeviL showed dose-and time-dependent activation of kinase MKK3/6, p38 MAPK, and caspase-3/9. The transduction pathways activated by SeviL via the glycosphingolipid oligosaccharide were triggered apoptosis.

Introduction
Many marine invertebrates possess lectins (glycanbinding proteins) with various glycan-binding properties [1][2][3]. In the differentiation of phylogeny, lectinmediated interactions between glycans and proteins were adapted into various kinds of key pathways involved in a variety of fundamental biological processes, including embryonic development, immune responses, and cell growth regulation [4][5][6]. During this functional diversification, marine invertebrates developed an unusually large number of lectins, many having convergent structures that facilitate binding to specific glycan structures exposed on the surface of target cells. This combination of functional divergence and structural convergence has resulted in many unique sequences and unusual glycan-binding specificities among lectins isolated from marine invertebrates [7][8][9][10][11].
Progress in 'omics' studies of mussels and other bivalve mollusks during the past decade has greatly enhanced our understanding of their genetics and molecular biology, leading to major advances in basic and applied scientific research. Mussels are a traditional seafood consumed heavily in Europe and increasingly in other parts of the world and are widely used as 'sentinel' organisms for biomonitoring [20]. Molecular studies have revealed the essential role of lectins as pattern recognition receptors (PRRs) for microbe-associated molecular patterns (MAMPs) in the innate immune systems of mussels [21,22]. A more complete understanding of these lectins will therefore provide a useful basis for improved mussel breeding practices and prevention of infections. Physiological processes and the immune system in mussels are strongly correlated with exposure to biotic and abiotic stress factors [23,24]. Numerous mussel immune system molecules including lectins were recently shown to be functionally modulated by pathogen exposure and ocean acidification [25,26], so that bivalve lectins are suggested to be important molecules which respond to the marine environment. The large and highly diverse lectin repertoire of mytilids [27,28], which probably includes several components yet to be identified, will facilitate effective new approaches for monitoring health status of mussel species, associated organisms, and their marine environments.
Since no R-type lectin had been biochemically purified from the family Mytilidae, mytilectins were considered for some time to be the only b-trefoil lectins present in mytilids, and it was speculated that their natural function related to the innate immune response [12,[15][16][17]. However, the isolation of a novel lectin from the purplish bifurcate mussel (Mytilisepta virgata) in this study suggests the possibility of greater diversification among lectins in this family. Furthermore, the transcriptome of this species revealed a lack of mRNAs encoding mytilectins and instead revealed the expression of multiple distinct mRNAs encoding proteins characterized by the presence of a ricin B-chain domain, typical of R-type lectins [29]. SeviL found from M. virgata shows characteristics of sugar chain binding and cell toxicity unlike any lectin reported to date. SeviL activated intracellular signaling pathways that resulted in cell death of mammalian carcinoma cells expressing asialo-GM1, whereas MytiLec-1 binds to Gb3 glycan. This is the first report that two different b-trefoil lectin families (each with its own glycan-binding specificity) coexist in the same animal species.

Purification of lectin (SeviL) from M. virgata
Supernatant 'Sup 1'(see Materials and methods 'Lectin purification') from homogenized M. virgata tissues displayed hemagglutination activity despite the absence of MytiLec-1 from this species. Repeated homogenization of precipitates yielded supernatants with successively reduced activity (data not shown). The precipitates were homogenized again with 50 mM lactose to obtain supernatants. The hemagglutinating activity was recovered by dialyzing the supernatant 'Sup 2' (see Materials and methods 'Lectin purification'). Sup 1 and 2 were applied to a lactosyl-agarose column, and the new lectin could be eluted with TBS containing 50 mM lactose (Fig. 1A). The lectin was characterized as a single polypeptide with molecular mass 15 kDa by SDS/ PAGE under both reducing and nonreducing conditions ( Fig. 1A) and was termed 'SeviL'. Purification from 400 g fresh tissue yielded 6.5 mg SeviL (Table 1). Hemagglutination activity of SeviL was required to the addition of calcium chloride (Fig. S1A), indicating that the activity was dependent on divalent cations such as Ca 2+ . Analytical ultracentrifugation revealed that SeviL was a tightly but noncovalently bound dimer (Fig. 1B).

Sugar-binding specificity of SeviL
Sugar-binding specificity of SeviL is summarized in Table 2. Hemagglutination activity was weakly inhibited by addition of monosaccharides such as D-Gal (25 mM), D-GalNAc (25 mM), and D-Fuc (25 mM) and of disaccharides such as melibiose (25 mM) and lactose (25 mM). These findings suggest that the chirality of the C3 and C4 carbons in galactose is essential for proteinglycan interaction. Hemagglutination activity was inhibited by administration of bovine submaxillary mucin (0.125 mgÁmL À1 ), but not porcine stomach mucin or fetuin, even at concentrations > 1 mgÁmL À1 (Table 2). These findings suggested that SeviL does not bind to porcine stomach mucin or fetuin, possibly because these glycoproteins have clusters of GlcNAc or sialyllactosamine at the reducing end [30].

Deduced primary structure of SeviL
The cDNA sequence of SeviL was identified using a combination of de novo peptide sequencing and our earlier transcriptomics results [29]. The peptide sequence obtained from trypsin digestion of SeviL (m/ z 685.81 (MH 2 ) 2+ ) was LDYN(M/T/S/C) GDLVANK ( Fig. S1B), which we compared with the mRNA sequences of five M. virgata tissues determined previously. A single sequence match was found with one protein product including the sequence 104 LDYNGGDLVANK 115 ( Fig. 2A). The complete 129 amino acid residue sequence was classified as an R-type lectin by the Pfam protein database (http://pfa m.xfam.org/) but is unrelated to MytiLec-1. At least two distinct lectin types (R-type lectins and mytilectins) having the b-trefoil fold structure are evidently present in the family Mytilidae. SeviL has a triple tandem-repeat structure with three subdomains, each consisting of~40 amino acids, with 13-21% sequence similarity, consistent with a b-trefoil fold (Fig. 2B). A QxW motif is conserved in each subdomain of R-type lectins, and SeviL shows a similar pattern at residues 40-42 (QxW), 79-81 (TxW), and 121-123 (ExW). The SeviL sequence included only one Cys (C) residue, which does not form a disulfide bond (Fig. 1A). A second sequence, highly homologous to SeviL and named SeviL-2, was identified in the M. virgata transcriptome. The two sequences only differ at 10 out of 129 amino acid residues (Fig. S1C). Due to the high heterozygosity of mussels, SeviL and SeviL-2 may either represent allelic variants of the same locus or the product of distinct orthologue genes, but the absence of a reference genome for this species hampered an in-depth investigation. All the experiments carried out and reported in this paper refer to SeviL-1, as the analysis of RNA-sequencing data [29] revealed that this variant was expressed > 4-fold higher than SeviL-2 in all tissues.

Comparison of SeviL orthologues among mytilid species
One or more SeviL orthologues are found in several other members of the family Mytilidae, including M. galloprovincialis, M. edulis, M. californianus, and M. trossulus, P. purpuratus and L. lithophaga, with a level of interspecies sequence conservation ranging from 25% to 99%, depending on the species considered (Fig. S2A). Like the aforementioned case of M. virgata, some species (i.e., P. purpuratus, M. galloprovincialis, and L. lithophaga) display two sequence variants, which were characterized more in detail in the Mediterranean mussel genome [32], revealing a two exons/one intron gene organization, with the ORF entirely contained within exon 2 (Fig. S2B). The three Trp residues of SeviL (residues 43, 81, 123; Fig. 3) are notably conserved in all of Mytilidae SeviL-like proteins.
The distribution of SeviL-like R-type lectins and mytilectins in the different mytilid subfamilies is variable and partially overlapping (Fig. 4). Transcriptome analysis suggests that R-type lectins (SeviL orthologues) are present in the subfamilies Brachidontinae (Mytilisepta, Perumytilus) and Lithophaginae (Lithophaga). On the other hand, both R-type lectins and mytilectins were found in the transcriptomes of Mytilinae (Mytilus, Perna), and neither of the two lectin families were detected in the genomes of Modiolinae, Bathymodiolinae (deep-sea mussels), or Arcuatulinae. While no SeviL-like R-type lectins could be found in the genomes of nonmytilid bivalves, mytilectin genes are present in Pectinidae (scallops). Overall, these observations reveal a a Total activity is shown by Titer 9 volume; b Specific activity was shown by titer per mg of protein; c Purification ratio was shown by comparing the value of specific activity on the crude extract vs. purified lectin; d Recovery of activity was revealed by comparing the value of total activity on the crude extract vs. purified lectin. Expression of SeviL mRNA in M. virgata tissues RNA-seq mapping graphs based on various tissues collected from a pool of adult mussels (Fig. 5A) show high expression of SeviL (TPM > 10) in gills and mantle rim. SeviL expression levels were much lower in the digestive gland and posterior adductor muscle and barely detectable in foot (TPM < 1). The high expression of SeviL in mantle rim was confirmed by qRT/PCR on individual mussels (Fig. 5B). The specificity of expression of SeviL in tissues (gills and mantle) that are constantly exposed to the external environment suggests the possibility that this lectin is involved in recognition of glycans found on parasitic or symbiotic microorganisms.

Asialo-GM1 oligosaccharide-dependent apoptosis
Possible triggering of ganglioside-dependent signals by SeviL was examined. Anti-asialo-GM1 pAb was applied to HeLa, MCF7, BT474, Caco2, and MDCK cells (see Materials and methods 'Mussels, cell lines, and reagents') and caused surface staining of each cell line except BT474 (Fig. 7A). Next, cells (10 5 mL À1 ) were incubated with various concentrations of SeviL for 48 h, and cell viability and proportions of living cells were determined by WST-8 assay. Increasing the SeviL concentration from 25 to 100 lgÁmL À1 resulted in apoptosis (cell death) for HeLa, MCF7, Caco2, and MDCK (Fig. 7B), but not for BT474. Cotreatment with anti-asialo-GM1 pAb blocked the cytotoxic effect of SeviL (Fig. 7B, 'SeviL + pAb'). These findings Perumytilus purpuratus. Names with numbers (e.g., mytvir1, mytvir2) indicate lectin variants from the same organism. Note that the sequence perpur1 and perpur2 display incomplete N-terminal and C-terminal regions, respectively. The sequence alignment within the polypeptide is analyzed by using MUSCLE program [31].

Localization of SeviL in M. virgata tissues
SeviL signals detected by the antiserum indicated its specific presence in the outer part of the mantle rim and gills (Fig. 10A,C), but not in the foot (Fig. 10E). This localization pattern reflected the transcriptional levels of SeviL-encoding mRNA coding in mussel tissues (Fig. 5A). The signals detected by the anti-GM1 pAb showed the same pattern of distribution as the expression of SeviL (Fig. 10B,D). These findings suggest that the main sites of expression of SeviL in M. virgata match with the location of detection of the antigens detected by the asialo-GM1 pAb. Since SeviL was obtained by the elution with the sugar-containing buffer from the mantle and gills of the mussels during the purification, it seems reasonable for the lectin to be found in these tissues with binding its ligands (Table 1).

Discussion
Over the past decade, glycobiological studies of nontraditional model organisms (such as bivalve mollusks) have revealed an unexpected diversity of lectins in various taxonomic groups. In this study, we have demonstrated the presence of an R-type lectin (SeviL) in Mytilisepta virgata, a member of the family Mytilidae. This lectin family is characterized by a b-trefoil fold structure and occurs across a wide range of animals, from microorganisms to humans. R-type lectins have been reported previously from the invertebrate phyla Porifera [34], Annelida [35], and Echinodermata [36]. SeviL was assigned to the R-type lectin family on the basis of sequence similarities to the prototypical ricin Bchain domain, but it displays features not found in other members of this family. First identified in a plant, Rtype lectins are found in a wide variety of organisms from bacteria to mammals, and numerous structures from this group have been analyzed in detail. Although a QxW motif is conserved in each of the three subdomains, no overall consensus sequence is found like that of C-type lectins or galectins. The primary structure of SeviL shares less than 20% similarity with other invertebrate R-type lectins (Fig. S2C), but there is much greater similarity (40-90%) among mytilid proteins (Fig. S2A). Both acidic and basic amino acids are found throughout the sequence of SeviL, in contrast to Myti-Lec-1, which has acidic amino acids only on the C-terminal side of each subdomain [12]. Surprisingly, SeviL also possesses 6 hydrophobic amino acids at the C terminus, as found with MytiLec-1 [12]. In the case of MytiLec-1, these residues were essential for dimerization [13], but structural analysis will be needed to determine whether the same is true of SeviL. Several similar sequences in other Mytilidae species besides M. virgata (Fig. 3) define a cluster of orthologues that we have named 'SeviL-like R-type lectins'. Curiously, the taxonomic spread of these lectins only partially overlaps that of mytilectins [18]. While no mytilectin was detected in the transcriptome of M. virgata, some mussel species (such as Mytilus galloprovincialis) possess both types of lectin, and others (such as Modiolus philippinarum) have neither (Fig. 4).
The transcription of M. virgata SeviL-like lectin and M. galloprovincialis MytiLec-1 was similarly confined to mantle and gills in both species (Fig. 5), suggesting that these lectins have similar roles in mussel physiology. However, while only the R-type lectin family is expressed in M. virgata, both members of the R-type family and the mytilectin family are encoded by the genome of M. galloprovincialis. It is presently unknown whether these two lectin families display an overlapping pattern of expression and are coregulated in this species. Determining how the expression of mussel lectins is modulated in response to external stimuli may bring new insights into the molecular ecology of these proteins, and helping to understand the role of lectin-glycan interactions in the notable capacity of bivalve mollusks to adapt to new environments.
Exposure to SeviL led to increased metabolism and induction of apoptosis in mammalian cells bearing asialo-GM1 oligosaccharide (Fig. 7), indicating the SeviL is a dimer like MytiLec-1 (Fig. 1B), whose selfassociation is essential for cytotoxicity. The use of an anti-asialo-GM1 polyclonal antibody (pAb), which blocked the access to the target of SeviL, completely abrogated the effect of the lectin (Fig. 7B). The cell regulatory mechanisms triggered by the interaction between SeviL and GM1b oligosaccharides will also be clarified by using a specific anti-GM1b antibody in future. SeviL activated various metabolic pathways (including MKK3/6, ERK 1/2 , p38, and caspase-3/9). Both MytiLec-1 and SeviL therefore can potentially regulate the growth of human cancer cells by binding their respective ligands and activating similar metabolic pathways (Figs 8 and 9). The mytilectin family is known to play roles not only in the regulation of cell death [44] and also cell proliferation [45], through the activation of kinases. SeviL and the other R-type lectins isolated from Mytilidae may similarly have multiple activities, and it is possible that their expression may be regulated by signals external to the organism.
b-Trefoil lectins of mussels have been proposed to be involved in defense against pathogenic microorganisms typically encountered by bivalves due to their filterfeeding habits [16,18,46]. By using immunohistochemistry techniques, the expression of SeviL was detected in tissues which are in direct contact with the internal and external environment (Fig. 10). The comparison between the transcriptional levels of SeviL-like lectin and mytilectin in mussels grown in different environments may clarify the specific role of these lectins in immune defense against invading microorganisms. The spatial overlap between the signals detected with the anti-asialo GM1 polyclonal antibody and the anti-SeviL antibody (Fig. 10) suggests the presence of similar or identical antigens with asialo-GM1 in the mussel tissues. The autoantibody which recognizes GM1b raised in patients affected by the Guillain-Barr e syndrome arises from infection with Campylobacter, because  . This evidence may support the hypothesis that the glycans recognized by SeviL may be present both in invading microorganisms and in the tissues of species pertaining to the family Mytilidae. It has been recently reported that the lectin subunit of the cholera toxin, which is known to bind GM1a, also binds to lipooligosaccharides and is capable of inhibiting the growth of genus Campylobacter [47]. The glycan structures of most invertebrates and microorganisms remain to be thoroughly investigated [48,49], and the application of structural glycobiology approaches to these phyla will provide a significant improvement in the knowledge on this subject. Besides the immune response of bivalves in response to infection by pathogenic microorganisms, the study of the infection and defense mechanisms enacted by these marine bivalves against neoplasia has also met a considerable interest in recent years. The group of Goff showed that horizontal transmission of cancer cells in bivalves resulted from activation of the retrotransposon gene 'Steamer' [50]. Such neoplastic cells may be propagated from one individual and transmitted to others through sea water [51]. In recent, the group of Metzger elucidated how the cancer cells of mussels were transferred across the Atlantic and Pacific Oceans and between the Northern and Southern hemispheres [52]. By knowing this situation, we will have more interest in how the cancer cells transfer into the mussels in the Asian area. In order to better elucidate the physiological role of SeviL-like lectins, one of our next goals will be to investigate whether their administration to tumor cells derived from bivalves may have a significant effect on cell growth regulation.
SeviL and MytiLec-1 bind to band a-galactosides, respectively, but it is not yet certain whether the natural ligand of these proteins is found within the organism itself or in the surrounding environment. The physiological roles of these proteins will remain unclear, however, until their target glycans are identified. Certain mollusks have characteristic glycosphingolipids with Gal or Gal-NAc at their glycan termini [53,54]. Such glycans are potential ligands for mytilectins, and similar ligands may exist for SeviL as well. Although the protein appears from its sequence to be a b-trefoil, it shows limited conservation to other such proteins at the sugarbinding sites (Fig. S4). Characteristics of the molecular

Lectin purification
Mussel mantles and gills were homogenized with 10 volumes (w/v) 150 mM NaCl containing 10 mM Tris/HCl, pH 7.5 (TBS) with 10 mM CaCl 2 . Supernatant (Sup 1) was collected by centrifugation at 27 500 g for 1 h at 4°C as described in our previous study [12], with some modification. Precipitate was homogenized with 10 volumes (w/v) TBS containing 50 mM lactose, and supernatant (Sup 2) was collected as above. Sup 2 was dialyzed extensively against TBS. Both Sup 1 and Sup 2 were applied to lactosyl-agarose column (5.0 mL), and the column was washed with TBS until absorbance at 280 nm (A 280 ) of effluent reached baseline level. Lectin was eluted with TBS containing 50 mM lactose.
Hemagglutination assay and sugar-binding specificity assay Hemagglutination assay was performed in 96-well V-shaped plates as described previously [55]. Twenty microliters of twofold dilution of purified lectin in TBS was mixed with 20 lL of a 1% suspension (with TBS; v/v) of trypsinized, glutaraldehyde-fixed rabbit erythrocytes, TBS, or TBS with 0.2% Triton X-100. Plates were incubated for 1 h at room temp, and formation of a sheet (agglutination-positive) or dot (agglutination-negative) was observed and scored as lectin titer. For sugar-binding specificity assay, 20 lL of sugar solution (200 mM) was serially diluted with TBS and mixed with 20 lL of lectin solution (adjusted to titer 16), trypsinized/glutaraldehyde-fixed rabbit erythrocytes, or TBS containing 1% Triton X-100. Plates were incubated for 1 h at room temp, and minimal inhibitory sugar concentration was determined.

Protein quantification and molecular mass determination
Protein was quantified using a protein assay kit (Thermo Fisher/ Pierce) based on the principle of bicinchoninic acid for colorimetric detection [56,57], using ovalbumin as standard. SDS/PAGE [58] was performed in 15% (w/v) acrylamide gel under reducing or nonreducing conditions, and gels were stained by Coomassie Brilliant Blue R-250. run, the rotor was kept stationary at 293 K in vacuum chamber for 1 h for temperature equilibration. A280 scans were performed at 10-min intervals during sedimentation at 201 600 g and analyzed using the continuous distribution (c (s)) analysis module in SEDFIT [59]. Frictional ratio (f/f 0 ) was allowed to float during fitting. c(s) distribution was converted to molar mass distribution c(M). Partial specific volume of protein, solvent density, and solvent viscosity were calculated from standard tables using the program SEDNTERP [60].

Analytical ultracentrifugation
Determination of primary structure of SeviL by mass spectrometry The partial peptide sequence of SeviL was derived by Proteomics International (Nedlands, WA, Australia). 200 lg lectin was dialyzed extensively against distilled water to remove salt, lyophilized, and digested by trypsin, and peptides were extracted by standard techniques [61]. Peptides were analyzed by electrospray ionization mass spectrometry using Prominence nano HPLC system (Shimadzu, Kyoto, Japan) coupled to 5600 triple time-of-flight (TOF) mass spectrometer (AB Sciex, Framingham, MA, USA). Tryptic peptides were loaded onto Zorbax 300SB-C18 column, 3.5 mm (Agilent, Santa Clara, CA, USA) and separated on a linear gradient of water/ acetonitrile/ 0.1% formic acid (v/v). MS/MS spectra were analyzed using PEAKS Studio software platform v. 4.5 SP2 (Bioinformatics Solutions, Waterloo, ON, Canada) with manual interpretation.
Transcriptomic analysis of full-length cDNA cDNA sequence of SeviL obtained as above was used to screen de novo-assembled transcriptome data of M. virgata obtained from four tissues (i.e., gills, mantle rim, posterior adductor muscle, and digestive gland) from a pool of mussels collected at the seashore of Saikai city [29]. The assembled contig corresponding to the putative mRNA lectin sequence was identified by tBLASTn (e-value threshold was set at 0.05). The partial peptides sequences obtained as described in the previous section were used as queries for BLAST searches against the transcriptome. Correct assembly of the consensus transcript was confirmed by back-mapping RNA-seq reads to the sequence and by assessment of uniform and homogenous mapping along the entire coding sequence. The expression level of SeviL in various tissues was calculated in silico as TPM (transcripts per million) using CLC Genomics Workbench v.10 RNAseq mapping tool (Qiagen, Hilden, Germany), setting length fraction parameter to 0.75, similarity fraction parameter to 0.98, and match/mismatch/deletion penalties to 3/3/3. RNA-seq datasets from the tissues mentioned above were used for this analysis [29].
To further confirm the tissue specificity of SeviL, qRT/ PCR analyses were carried out on three individual mussels, as described in [29]. In this case, the sequence-specific primers designed for SeviL are (5 0 -> 3 0 ): AATTTGGGGCG TAAAGACCT (forward primer) and GGACTCTCTTCC GAGGGTG (reverse), aiming at the amplification of a 111-bp target sequence.

Sequence data availability
The cDNA sequence of SeviL, SeviL-2, and the orthologue sequences identified in publicly available transcriptomes of other mytilid species have been deposited in the GenBank repository, under the accession numbers MK434191-MK434201. The sequence alignments among each orthologue in the different species or the subdomains in the polypeptide are analyzed by using the MUSCLE program (http://www.drive5.com/muscle) [31].

Glycan array analysis
Glycan array analysis was performed by Sumitomo Bakelite Co. (Tokyo, Japan). SeviL was fluorescence-labeled (k ex/em 555/570 nm) using HiLyte Fluor 555 labeling kit-NH 2 as per the manufacturer's instructions. A wide range of 52 glycans including N-glycans, O-glycans, Lewis glycans, lactosamine, blood-type glycans, gangliosides, and globosides were immobilized on wells. Fluorescence-labeled SeviL at concentrations ranging from 0 to 100 lgÁmL À1 were incubated overnight at 4°C with shielding from light. SeviL-binding glycans were detected by using Bio-REX Scan 300, an evanescent fluorescence scanner (Rexxam Co. Ltd., Osaka, Japan). Wavelength of laser light was used for Cy3, and the exposure time was 300 ms [62].

Cell viability and cytotoxicity assays
Cells were maintained in RPMI 1640 supplemented with heat-inactivated FBS 10% (v/v), penicillin (100 IUÁmL À1 ), and streptomycin (100 lgÁmL À1 ) at 37°C. Cytotoxic effects and cell growth following treatment with SeviL at concentrations ranging from 0 to 100 lgÁmL À1 were determined using Cell Counting Kit-8 containing WST-8 [12]. Cells (2 9 10 4 , in 90 lL solution) were seeded into 96-well flatbottom plates and treated with 10 lL lectin for 24 h at 37°C. To evaluate glycan-inhibitory effects, anti-asialo-GM1 oligosaccharide pAb (50 lgÁmL À1 ) was co-incubated with cells in addition to lectin for 24 h and then applied to the assay system. For assay of effect on cell growth, each well was added with 10 lL WST-8 solution and incubated 4 h at 37°C. Cell survival rate was determined by measuring A 450 (reference: A 600 ) with a microplate reader (model iMark; Bio-Rad, Tokyo, Japan).
Immunocytochemical analysis of asialo-GM1 oligosaccharide expression Cells (1 9 10 6 ) were fixed with 4% paraformaldehyde in PBS for 15 min at room temperature, washed 39 with PBS, blocked with 1% BSA in PBS for 30 min at room temperature. They were washed 39 with PBS, incubated with or without 100 lL anti-asialo-GM1 oligosaccharide pAb (dilution 1:200 with PBS) at 4°C for 30 min, washed 39 with PBS, treated with 100 lL Alexa Fluor Ò 488-tagged goat anti-rabbit IgG (dilution 1:200 in PBS) at 4°C for 30 min. Cells were placed onto low fluorescence glass slides, mounted with 50% glycerol solution, and examined by confocal microscopy. Confocal images were obtained using FV10i FLUOVIEW (Olympus, Tokyo, Japan).

Immunohistochemistry of SeviL expressions on the mussel tissues
Mussel organs (gill, mantle rim, and foot) were cut into around 1-cm-square pieces, embedded in the Tissue-tek compound and frozen in isopentane, cooled in by liquid nitrogen. The frozen tissue block was sliced on 6-lm-thick with Leitz cryostat (Leica Instruments, Nussloch, Germany), placed on silicon-coated glass slides. Sections were sequentially fixed in PBS containing 4% paraformaldehyde for 15 min at room temperature, incubated in blocking solution containing 0.05% saponin and 1% BSA in PBS for 30 min, incubated with or without anti-SeviL or antiasialo-GM1 oligosaccharide pAb (dilution 1:100 with blocking solution) at room temperature for 1 h. After washing the tissues with PBS, they were treated with Alexa Fluor Ò 568-labeled goat anti-rabbit IgG or Alexa Fluor Ò 488-labeled goat anti-rabbit IgG (dilution 1:100 in blocking solution) at room temperature for 1 h, washed with PBS, mounted with 50% glycerol solution, and observed by confocal microscopy FV10i FLUOVIEW. Nuclei were stained by DAPI (364/454 nm) [64].

Statistical analysis
Experiments were performed in triplicate, and results presented as mean AE standard error (SE). Data were subjected to one-way analysis of variance (ANOVA) followed by Dunnett's test, using SPSS STATISTICS software package, v. 10 (www.ibm.com/products/spss-statistics). Differences with P < 0.05 were considered significant.   .  Table S1. List of 52 oligosaccharides used for the glycan-array analysis.