High Affinity Interaction between a Bivalve C-type Lectin and a Biantennary Complex-type N-Glycan Revealed by Crystallography and Microcalorimetry*

Codakine is an abundant 14-kDa mannose-binding C-type lectin isolated from the gills of the sea bivalve Codakia orbicularis. Binding studies using inhibition of hemagglutination indicated specificity for mannose and fucose monosaccharides. Further experiments using a glycan array demonstrated, however, a very fine specificity for N-linked biantennary complex-type glycans. An unusually high affinity was measured by titration microcalorimetry performed with a biantennary Asn-linked nonasaccharide. The crystal structure of the native lectin at 1.3Å resolution revealed a new type of disulfide-bridged homodimer. Each monomer displays three intramolecular disulfide bridges and contains only one calcium ion located in the canonical binding site that is occupied by a glycerol molecule. The structure of the complex between Asn-linked nonasaccharide and codakine has been solved at 1.7Å resolution. All residues could be located in the electron density map, except for the capping β1–4-linked galactosides. The α1–6-linked mannose binds to calcium by coordinating the O3 and O4 hydroxyl groups. The GlcNAc moiety of the α1,6 arm engages in several hydrogen bonds with the protein, whereas the GlcNAc on the other antenna is stacked against Trp108, forming an extended binding site. This is the first structural report for a bivalve lectin.

Lectins are multivalent carbohydrate-binding proteins that play important roles in the social life of cells. A growing repertoire of lectins has been identified in invertebrates (1), where these molecules are involved in self/nonself recognition (2). For example, lectins play a role in aggregation mechanisms in corals and sponges (3) or in sperm-egg recognition in oysters (4). Lec-tin mediation of symbiosis with algae or bacteria has been observed in coral (5) and nematodes (6). Nevertheless, the most common function assessed for lectins in marine invertebrates is their role in innate immunity by specific binding of polysaccharide-coated pathogenic bacteria (7,8).
Different lectins have been identified in bivalves and they most frequently belong to the C-type lectin family. Proteins from this group of calcium-dependent lectins have been reported in oysters (9,10), scallops (11), and clams (12,13). C-type lectins are characterized by a carbohydrate recognition domain (CRD) 3 with a conserved fold and the involvement of a calcium ion in carbohydrate binding (14). The crystal structure of mannose-binding protein was the first one to be described (15). The CRD belongs to a larger family sharing a common fold and is referred to as the C-type lectin-like domain (16).
Codakine is a 14-kDa C-type lectin purified from the gill of the tropical clam Codakia orbicularis (Linné 1758) by affinity chromatography on a mannose-agarose column (17). It forms homodimers and heterodimers with isoforms 1 (NCBI accession number AAX19697) and 2 (NCBI accession number ABQ40396) (13). A 19-amino acid peptide signal suggests that the lectin travels through a secretory pathway. The 129-amino acid sequence of the mature protein has significant sequence similarities to various fish lectins (33%). Sequence homologies have, moreover, been observed between codakine and mermaid nematodes (18). These results point to the probable role for the gill-located codakine in either antibacterial protection or in the recognition of sulfur-oxidizing bacteria, symbionts that are needed for the survival of C. orbicularis in sandy anaerobic environments (19).
In this paper, we present data on the specificity and affinity of codakine for various monosaccharides and oligosaccharides using inhibition of hemagglutination, glycan microarrays, and microcalorimetry. The native crystal structure displays a new covalent dimerization mode. The structure of the complex with a high affinity biantennary N-glycan displays a new oligosaccharide binding mode where both antennae are in contact with the protein.

EXPERIMENTAL PROCEDURES
Protein Purification-The previously described protein purification protocol (13) was used with minor modifications. Briefly, about 5 g of nitrogen frozen gill tissue of C. orbicularis was crushed with a pestle in 50 ml of buffer composed of 20 mM Tris-HCl, 100 mM NaCl, 100 M CaCl 2 , pH 7.4 (T buffer). After a 10-min centrifugation at 10,000 rpm, the supernatant was dialyzed overnight at 4°C against fresh T-buffer four times. Insoluble matter was pelleted by centrifugation as described above. The 0.25-m filtered supernatant was loaded onto a mannose-agarose column pre-equilibrated with T-buffer. After washing with T-buffer containing 1 M NaCl, codakine was eluted by 0.1 M EDTA in T-buffer. The eluted fractions were pooled and dialyzed extensively during 2 days at 4°C against T-buffer. The electrophoretic profile of eluted fractions was checked on 15% SDS-polyacrylamide gel (20). The molar extinction coefficient and optical density at 280 nm were used to determine the concentration of codakine.
Hemagglutination Assays-Hemagglutination tests were performed using microtiter plates with U-bottom wells by the 2-fold serial dilution method. 25 l of rabbit erythrocytes (2% in NaCl; Biomérieux) were mixed with serially diluted codakine in T-buffer (described above), starting with a concentration of 1 mg/ml. After a 30-min incubation at 37°C, the plates were read. One unit of hemagglutination activity was defined as the highest dilution of lectin giving a complete hemagglutination (HA unit, g/ml). For the inhibition tests, a dilution of codakine equivalent to four units of HA, corresponding to 15.6 g/ml, was used. Inhibition of hemagglutination was assayed by 2-fold serial dilutions of the following sugars in T-buffer: D-mannose, D-galactose, D-glucose, N-acetylglucosamine, L-fucose, N-acetylneuraminic acid, and D-rhamnose, each starting at a concentration of 100 mM.
Glycan Microarray Analysis-Purified codakine (1 mg/ml) was labeled with the Alexa Fluor 488 protein labeling kit (Invitrogen) according to the instructions of the manufacturer. The purification of Alexa-labeled codakine was performed by mannose-agarose chromatography, as described above. Glycan microarray tests were carried out by the standard procedure of Core H of the Consortium for Functional Glycomics (21).
Titration Microcalorimetry-Thermodynamic parameters were estimated by isothermal titration calorimetry (ITC) using a Microcal (Northampton, MA) VP-ITC microcalorimeter. All experiments were carried out at 298 K. Ligands and lectins were prepared in T-buffer (described above). Titration of codakine binding was performed in a cell with a volume of 1.447 ml by 30 injections of 10 l of ligand with 5-min intervals while stirring at 310 rpm. Blank titration with buffer was used as a reference. The experimental data were fitted to a theoretical titration curve using software supplied by Microcal with ⌬H (enthalpy change), K a (association constant), and n (number of binding sites/monomer) as adjustable parameters, from the classical relationship (24). For each ligand, experiments were repeated two or three times.
Crystallization and Data Collection-Native codakine was concentrated to 15 mg/ml in 20 mM Tris-Cl, pH 8, 5 mM CaCl 2 , and 5 mM ␣-methyl-mannoside on Vivaspin 5 kDa (Vivascience). All crystals were obtained by the hanging drop vapor diffusion method using 2-l drops containing a 50:50 (v/v) mix of protein and reservoir solution at 20°C. Initial crystallization screening was performed using the JCSGϩ suite (Qiagen). After three months, crystals were obtained with a solution containing 0.2 M diammonium citrate, pH 5, 20% (w/v) polyethylene glycol 3350. A range of related conditions were used for co-crystallization of the complex between codakine (22 mg/ml in the same buffer) and nona-Asn (10 mM). One large cube-SCHEME 1. Fmoc deprotection.

Crystal Structure of C. orbicularis Lectin
shaped crystal was obtained with 0.2 M lithium sulfate, 100 mM sodium acetate, pH 7.5, and 15% polyethylene glycol 4000 from a well with a broken lamella.
Glycerol was added to 20% (v/v) to the crystallization solution as cryoprotectant prior to freezing crystals in a gaseous nitrogen stream at 100 K. Native and complex data were collected at the European Synchrotron Radiation Facility on an ADSC Q4R CCD detector (Quantum Corp.) on beamline ID14-1 and an ADSC Q315R detector on beamline ID29, respectively. Diffraction images were processed using MOSFLM (25). All further computing was performed using the CCP4 suite (26), unless otherwise stated. Data processing statistics are presented in Table 1.
Molecular Replacement and Structure Refinement-Attempts to solve the crystal structure of codakine by molecular replacement using the C-type lectin structures available in the Protein Data Bank were unsuccessful. A homology model was then constructed using FUGUE and ORCHESTRAR (Tripos Inc., St. Louis, MO). Among profile families identified by the FUGUE program, four templates with more than 25% identity with codakine were selected: human E-selectin (27) (Protein Data Bank code 1ESL), human lithostathine (28) (Protein Data Bank code 1QDD), human tetranectin (29) (Protein Data Bank code 1TN3), and Hemitripterus americanus antifreeze protein (30) (Protein Data Bank code 2AFP). Structurally conserved regions were built by ORCHESTRAR, and loops were modeled by an ab initio approach. The conserved disulfide bridges and calcium ions were incorporated in the model that has been deposited in the Protein Model Data Base with accession number PM0074967.
This model was then used for solving the codakine structure by molecular replacement with the program Phaser (31). The program Acorn (32) was subsequently used to improve the electron density. A few correctly placed segments were chosen from the molecular replacement solution and used as starting coordinates for Acorn phasing. Among the smallest fragments tested, Acorn was able to phase the structure starting from the positions of eight sulfur atoms. The very high resolution (1.3 Å) of the native crystals allowed the program to calculate an excellent electron density map where ARpWarp (33) built the complete model. In the carbohydrate binding site, clear density could be observed for one calcium atom. Two glycerol molecules and one citrate molecule were also included in the model.
The structure in complex with the nona-Asn was solved by molecular replacement with Molrep (34), using the native codakine structure as a search model. After the addition of a calcium ion and solvent molecules, clear residual electron density was visible for at least five sugar monomers. The oligosaccharide was docked manually into the electron density according to its chemical structure; no ambiguity at the glycosidic linkages was observed.
For the refinement of each structure, 5% of the observations were immediately set aside for cross-validation analysis (35) and were used to monitor various refinement strategies. Manual corrections of the models using Coot (36) were interspersed with cycles of maximum likelihood refinement with REFMAC (37). The two models were validated with the WhatIf suite (38) and deposited in the Protein Data Bank with accession numbers 2vuv and 2vuz for the native codakine and the nonasaccharide complexes, respectively. The refinement statistics are listed in Table 1.
Miscellaneous Methods-Areas of buried surfaces were calculated with PISA (39). All figures were drawn with the PyMOL Molecular Graphics System (DeLano Scientific LLC, San Carlos, CA) unless otherwise stated.

RESULTS
Overall Structure of Codakine-In the presence of ␣-methylmannoside (␣MeMan), codakine crystallized in rock-shaped crystals of space group C2 that diffract to 1.3 Å resolution. The asymmetric unit contains one monomer, and a disulfide bond dimer is generated by the 2-fold axis. The monomer consists of a typical C-type lectin CRD with two parts: the lower part contains the two helices and three strands, and the upper part is composed of four strands and long loops (Fig. 1A) (40). As predicted previously by bioinformatics, the mature protein lacks the 19-amino acid signal peptide present in the nucleotide sequence. The first two ␤-sheets and the last one run antiparallel, and a salt bridge is observed linking the termini of the peptide chain. For the C-terminal amino acid, clear density for the backbone was found but not for the side chain. Apart from this, all amino acids could be modeled in the density map. Additional density was observed near C␥ of Pro 91 , which appears to be a 4-hydroxyproline.
Three intramolecular disulfide bonds are observed (Cys 2 -Cys 13 , Cys 30 -Cys 124 , Cys 103 -Cys 116 ). The last two disulfide bridges are conserved in C-type lectins, linking the first helix to the last strand and one strand to a loop, respectively. The first bridge is present in long form C-type lectin-like domain (41). One free cysteine, Cys 53 , is observed on the ␣2-helix. Cys 44 , the residue involved in the unique dimerization mode observed in this C-type lectin, is located at the beginning of the same helix. The dimer is further stabilized by hydrogen bonds from His 51 and Glu 55 to Asp 11 and Gln 8 on the other chain. Hydrophobic interactions were found for His 51 stacking against Leu 10 from the other chain and Leu 10 against Phe 9 . Through the 2-fold symmetry, this creates a chain of contact between the 6 hydrophobic residues that extends to 20 Å; this results in a dimer interface area of about 400 Å 2 .
C-type lectins usually contain two calcium ions in conserved binding sites referred to as site 1 and site 2 (40). In the present structure, only one calcium ion is present in site 2 that corresponds to the monosaccharide binding site. The calcium is coordinated to the acidic oxygen of Glu 93 , Glu 101 , and Asp 113 ; to the oxygen atoms from amide Asn 95 and Asn 112 ; and to the main chain oxygen of Asp 113 . Analysis of the electron density indicates that ␣MeMan is not present in the site but is instead replaced by a well ordered glycerol molecule. Additional density was also observed and attributed to another glycerol molecule and a citrate molecule.
Analysis of the glycerol located in the carbohydrate binding site indicates that two oxygen atoms establish the contacts classically observed for the carbohydrate ligand. These two oxygens complete the coordination sphere of the calcium. The first oxygen shows hydrogen bonds to the side chains of Asn 95 , Glu 93 , and Asn 112 , the second oxygen to the side chains of Glu 101 and Asn 112 and to the backbone oxygen of Asp 113 (Fig. 1C).
Affinity and Specificity for Monosaccharides-Hemagglutination inhibition assays were performed using rabbit erythrocytes. Codakine shows a hemagglutination activity for a minimum of 3.9 g/ml (one unit of HA). Inhibition of interaction between the lectin and the red blood cells was assayed with a range of monosaccharides. The highest inhibition power (using 4 HA units of codakine) was observed for D-mannose and L-fucose) (25 mM). Glucose and GlcNAc have a weaker effect (100 mM), and no inhibition was observed for D-galactose, D-rhamnose, D-neuraminic acid even at the higher concentration. Codakine can therefore be confirmed as a mannose/fucosespecific lectin, as was predicted from the EPN peptide signature in the carbohydrate binding site (40).
For a complete characterization of the interaction between codakine and the most efficient monosaccharides used in the hemagglutination assay, affinity and thermodynamic parameters were measured using ITC (Table 2). Mannose and GlcNAc have a rather low affinity with a dissociation constant of 0.27 and 0.46 mM, respectively. No affinity could be measured for galactose. The dissociation constant obtained for ␣MeMan is 52 M, which is about 20 times more potent than the interaction with mannose, an increase frequently observed in lectincarbohydrate interactions. The ␣-methyl-fucoside also appears to be a good inhibitor, as observed in hemagglutination assays. In all cases, the association is driven by enthalpy. The entropy term is weak and slightly favorable except for ␣MeMan. In this latter case, the strong enthalpy contribution (⌬H ϭ Ϫ33.6 kJ/mol) is partly opposed by an entropy barrier (T⌬S ϭ Ϫ9.1 kJ/mol), as classically observed in protein-carbohydrate interactions (42).
Affinity and Specificity for Oligosaccharides-The specificity of the fluorescence-labeled lectin was evaluated by binding to the 320 oligosaccharides present on the glycan array available at the Consortium for Functional Glycomics (Fig. 2). Only 13 oligosaccharides are recognized with high affinity; they all represent complex-type biantennary N-glycans (complete data available as supplemental material). The shortest ligand is the core pentasaccharide, consisting of a chitobiose and a trimannoside with ␣1-3 and ␣1-6 linkage. Larger N-glycans elongated by the successive addition of GlcNAc, galactose, and sialic acid are also recognized by the lectin. Fucosylation at position 3 of the antennary GlcNAc moieties (Lewis x) is tolerated, but not fucosylation of the chitobiose core.  In order to characterize the thermodynamic and structural basis of this very narrow specificity, a biantennary nonasaccharide N-glycan linked to Asn (Scheme 1) was prepared from egg yolk. Codakine dissociation constants were measured for this compound and also for trimannoside in order to evaluate the influence of the internal chitobiose moiety on the high affinity (Fig. 3). The affinity for the trimannoside (K d ϭ 80 M) is of the same order of magnitude as the one measured for ␣MeMan ( Table 2). In contrast, the affinity for nona-Asn is submicromolar (K d ϭ 432 nM). To our knowledge, this is the highest affinity reported for a C-type lectin, excluding multivalency effects. The 200-fold increase observed for nona-Asn compared with the trimannoside could be due to either the terminal N-acetyllactosamine on each antenna or to the presence of internal chitobiose. Since the core pentasaccharide Man3GlcNAc2 is efficiently recognized in the glycan array, it could be concluded that the internal chitobiose has a crucial effect on the affinity.
When analyzing the thermodynamic contributions, the binding of codakine to nona-Asn shows a higher enthalpy contribution (⌬H ϭ Ϫ45.3 kJ/mol) than to trimannoside (⌬H ϭ Ϫ41.9 kJ/mol). However, the difference in free energy (and therefore in affinity) is mainly due to the different entropy terms. Both oligosaccharides display an unfavorable entropy contribution, but the entropy for binding the trimannoside (T⌬S ϭ Ϫ18.5 kJ/mol) is much stronger than for the nonasaccharide (T⌬S ϭ Ϫ8.9 kJ/mol).
Crystal Structure of Codakine⅐Nona-Asn-Cocrystallization of codakine with nona-Asn yielded one rock-shaped crystal diffracting to 1.7 Å in the C222 1 space group. The overall shape of each monomer is similar to what is observed in the structure of codakine complexed with glycerol. The intermonomer disul-fide bridge also occurs, albeit with a longer S-S bond distance (2.33 Å instead of 2.05 Å). Such a deviation from normality might be due to radiation damage. This variation, associated with rotational flexibility of this linkage, induces a slightly different orientation of the second monomer with respect to the first one (supplemental Fig. S2).
Electron density could be readily identified for one calcium ion and 5 carbohydrate residues. After building the ␤GlcNAc12␣Man13(␤Glc-NAc12␣Man16) Man pentasaccharide, the two extra GlcNAc could be added at position 1 of the central ␤Man. The galactose residues capping the two antennae and the Asn moiety could not be located in the electron density (Fig. 4A).
The ␣Man of the 1-6 branch is located on the calcium ion in an orientation previously reported in DC-SIGN (43), with O3 and O4 coordinating the calcium ion. The OH3 group makes hydrogen bonds to the side chains of Glu 93 , Asn 95 , Glu 101 , and Asn 112 . The OH4 group binds to the Glu 101 and Asn 112 side chains and to the Asp 113 main chain oxygen, and the O6 hydroxyl binds to the Arg 115 side chain ( Table 3). The GlcNAc residue of the 1-6 branch interacts with Glu 93 via OH6, with Asn 95 via the ring oxygen O5 and with Ser 97 via the acetyl oxygen. The branching ␤Man establishes two hydrogen bonds between O4 and the Asn 107 and Trp 108 side chains. The ␣Man of the 1-3 branch is not in contact with the protein, but the antenna folds back, and the terminal GlcNAc is perfectly stacked onto Trp 108 , creating a strong hydrophobic interface. The two GlcNAc moieties of the chitobiose core do not make direct contact with the protein, but the one linked to ␤Man interacts with the Ser 61 side chain via a water molecule and establishes two intramolecular hydrogen bonds with the ring oxygens of ␤Man and ␣Man in the 1-6 branch (Fig. 4B).
The overall shape of the oligosaccharide is rather folded. The conformations of all of the glycosidic linkages lie in energy minima that were identified previously in energy maps (44,45). In solution, the most flexible one would be the ␣Man1-6Man linkage that could generate several different overall shapes of the biantennary N-glycan. In the present case, the constraints created by the rather deep binding pocket result in a folded conformation, with the ␣Man of the 1-6 arm close to GlcNAc2 of the chitobiose core.

DISCUSSION
Comparison with Other C-type Lectins-The family of C-type lectins is characterized by low sequence similarities despite a well conserved fold. Indeed, structural comparison by secondary structure matching (46) indicates high structural similarity with a range of vertebrate C-type lectin-like domains, such as   (Fig. 1B) illustrates the high similarities in secondary structure and calcium site 2. Codakine lacks calcium site 1, due to differences in the loop connecting strands ␤3 and ␤4. The loop is shorter and lacks acidic amino acids, such as Glu 324 , in DC-SIGN. More importantly, Asp 320 at the end of the DC-SIGN ␤3-strand is replaced by a lysine (Lys 70 ) that extends into the groove. The positively charged side chain occupies the calcium location in site 1 and establishes the equivalent interactions with conserved amino acids (Asp 102 and Asn 96 side chains and Glu 101 backbone oxygen). Langerin is the only other structure of CRDs lacking calcium in site 1, and it also contains a lysine residue occupying the place of calcium. However, in langerin, the connecting loop is longer, and the resulting larger groove acts as a secondary carbohydrate binding site (48).

Molecular Basis for the High Affinity toward Oligosaccharides-
Codakine displays millimolar affinity for monosaccharides (K d ϭ 0.27 mM for mannose), which is in agreement with ITC data previously obtained for other C-type lectins, such as mannose-binding protein with mannoside (49) and tunicate lectin TC14 with fucose (50). However, the strong affinity observed between codakine and the biantennary N-glycan (K d ϭ 0.432 M) is unique when compared with other C-type lectin/oligosaccharide interactions. The dissociation constant for E-selectin interacting with Sialyl-Lewis X has been estimated at 120 M by fluorescence polarization (51), whereas DC-SIGN interaction with Man 9 GlcNAc 2 was measured with a K d of 26 M by competition assays (52).
Indeed, the high affinity appears to be due to the extended binding site with numerous contacts between the oligosaccharide and the protein side chain. The stacking between Trp 108 and the GlcNAc of the 3 arm also seems to be very favorable. It results in a buried protein surface of 116 Å 2 , almost as large as the surface buried by the mannose at the main binding site (133 Å 2 ). However, there are some discrepancies between this observation and the results derived from the glycan array data, since the core pentasaccharide Man3GlcNAc2, which lacks the GlcNAc moieties on the antennae, is also a high affinity ligand of codakine.
When comparing the codakine⅐nona-Asn complex with the DC-SIGN⅐pentasaccharide ␤GlcNAc12␣Man13(␤Glc-NAc12Man16)Man (43), only the ␤GlcNAc12Man disaccharide segments on the calcium sites are similar (Fig. 4C). In DC-SIGN, the rest of the pentasaccharide does not establish much contact with the protein, and the binding site appears to be more open. The deeper character of the codakine binding site is mainly due to the longer loop between ␣2-helix and ␤3-strand (Fig. 1D).
The inner chitobiose core seems to play a crucial role for the high affinity binding to codakine. From the glycan array data, the presence of these residues of the core is strictly needed. However, no direct contact is observed between the two Glc-NAc and the protein. Interestingly, for DC-SIGN, a similar effect is observed, since Man 9 GlcNAc 2 has a 2-3-fold increase in affinity when compared with Man 9 . It was proposed that the inner GlcNAc residues restrict the conformation of nearby sugar groups, resulting in lower entropy of binding. This hypothesis is confirmed by our work, since the biantennary glycan with the chitobiose core has a significantly lower entropy barrier (Ϫ8 kJ/mol) than the trimannose compound (Ϫ18 kJ/mol). This is also correlated with the structural data, since the GlcNAc residue linked to ␤Man establishes one hydrogen bond to the ␣Man residue on the 1-6 arm, which should significantly stabilize this otherwise very flexible linkage.
Structure-Function Relationship in Invertebrate Lectins-Codakine is the first mannose-specific invertebrate lectin to be structurally characterized. Crystal structures were previously obtained for tunicate (Polyandrocarpa misakiensis) lectin TC14 complexed with galactose (50) and sea cucumber (Cucumaria echinata) CEL-I lectin complexed with GalNAc (53). In all cases, the lectins adopt a dimeric arrangement allowing for opposite face presentation of carbohydrate binding sites that would be more appropriate for aggregating cells than for avidity binding (Fig. 5). All dimers are generated by 2-fold (or pseudo-2-fold) symmetry. TC14 is a noncovalent dimer with interaction through the ␣2-helix, whereas codakine and CEL-I dimerization involves an intermolecular disulfide bridge.
The biological role and the natural ligands are not yet defined for the three lectins that have been crystallized. Invertebrate lectins are involved in innate immunity and are thought to bind to polysaccharides present at the surface of bacteria and parasites. Recently, the structural basis for the interaction between lung surfactant protein D, a C-type lectin, and heptose residues present in bacterial lipopolysaccharide was demonstrated (54). It would be of interest to determine which bacterial lipopolysaccharide fragments could mimic the N-glycan conformation observed in the present structure and which could be recognized by codakine.
Additionally, codakine could play a role in the recognition of symbiotic sulfur-oxidizing bacteria (19). The presentation of opposite face binding sites resulting from dimeric association could result in the cross-linking of microorganisms to the invertebrate cell surface. GlcNAc-terminated biantennary glycans have been identified in invertebrates (55), and codakine would therefore attach to the gill surface through this ligand. Further investigation of the polysaccharides present on the sulfur-oxidizing bacteria should provide important clues regarding the role of invertebrate lectins in symbiotic associations that are often of primary importance in marine organisms.