Galactosyl-binding Lectins from the Tunicate Didernnurn candidurn PURIFICATION AND PHYSICOCHEMICAL CHARACTERIZATION*

The plasma of the ascidian Didemnum candidum possesses lectin activity directed toward galac- tosyl moieties. We report the purification by affinity chromatography, the physicochemical properties, amino acid composition, and partial N-terminal amino acid sequence of two galactosyl-binding lectins D. candidum lectins I and I1 (DCL-I and DCL-11) from the plasma of this protochordate species. Both lectins were purified by affinity chromatography (on acid-treated Sepharose 4B and asialofetuin conjugated to Sepharose 4B) to homogeneity as judged by immunoelec- trophoresis, size exclusion chromatography on high performance liquid chromatography, and polyacryl- amide gel electrophoresis. Isoelectric focusing in polyacrylamide gels revealed that DCL-I focuses as a fam- ily of bands at pH 3.8-5.2, while DCL-I1 focuses at pH 9.2-10.2. Gas chromatography analyses of alditol ace- tate derivatives indicated that no carbohydrate components are associated with the lectins. Approximate subunit molecular weights estimated by polyacrylam- ide gel electrophoresis and size exclusion chromatography on high performance liquid chromatography in 6 M guanidine HCl under reducing conditions were 13,400-14,500 for DCL-I and 14,500-15,500

and the horseshoe crab Carcinoscorpius rotundicauda, rabbit C-reactive protein, and lamprey and carp immunoglobulin p chains. DCL-I1 showed amino acid composition and similarities with several fish immunoglobulin light chains, immunoglobulin-related molecules isolated from mouse and marmoset T cells, and carp and goldfish immunoglobulin heavy chains. DCL-I N-terminal amino acid sequence showed up to six identities in a stretch of 19 residues with immunoglobulin-related molecules and acute phase proteins (Creactive protein and serum amyloid P component). DCL-I cross-reacted in enzyme-linked immunosorbent assays with antibodies made against human C-reactive protein.
Although the exact functions of invertebrate lectins are unknown, the findings that lectins of diverse specificities occur both in the serum (1) and on the surface of phagocytic cells (2) suggest that these molecules might play an essential role in self/non-self discrimination. Within the invertebrates, the tunicate group is considered to be evolutionarily the closest group to vertebrates, and, therefore, study of specific recognition molecules in this subphylum should provide clearest evidence for relatedness to vertebrate recognition molecules such as immunoglobulins, complement components, and acute phase proteins. We have previously assessed the distribution of plasma lectins in 10 species of North American tunicates and have found that all species possess detectable lectins and that lectins of distinct specificities occur in the individual species (3). The ascidian Didemnum candidum possessed powerful lectin activity directed toward galactosyl moieties. In this paper, we report the purification by affinity chromatography of two galactosyl-binding lectins from the plasma of D. candidum, We also describe their physicochemical properties and compare their amino acid compositions and the partial N-terminal amino acid sequence of DCL-I' to those of other invertebrate and lower chordate lectins, vertebrate immunoglobulin chains, and other putative recognition molecules, such as acute phase proteins. The sequence and composition comparisons were made in order to investigate the possible relationships of DCL-I and DCL-I1 to other animal lectins and to the extended family or "superfamily" of immunoglobulin-related recognition molecules (4).

RESULTS
Purification of D. candidum Latins-We compared the efficiency of three different galactosyl-containing immunoadsorbents: Sepharose 4B, acid-treated Sepharose 4B, and asialofetuin coupled to Sepharose 4B. D. candidum BF-S was applied continuously to the thoroughly washed equilibrated columns and 2-ml fractions were collected and tested for agglutination of Pronase P-treated human B red blood cells. When titers equaled those of D. candidurn BF-S, columns were washed to the baseline (absorbance: 0.008) with Trisbuffered saline, and the elution was accomplished with a galactose concentration gradient from 0-50 mM. Two peaks (peak I and peak 11) were eluted from the asialofetuin Sepharose column (Fig. I), while only one peak was obtained either from the acid-treated or untreated Sepharose column (Fig. 2). Selected fractions associated with each peak were dialyzed against Tris-buffered saline, concentrated in a Savant Speed Vac Concentrator, and titrated with Pronase P-treated human B red blood cells. Fractions with high activity from each peak were pooled and dialyzed against Tris-buffered saline, and protein concentrations were determined for every pool. Polyacrylamide gel electrophoresis (Fig. 3) showed that peaks I and I1 obtained from asialofetuin Sepharose chromatography correspond to two lectins respectively) with different subunit apparent molecular weights 14,500;15,500). The single peaks obtained on untreated Sepharose and acid-treated Sepharose corresponded to the lower molecular weight component DCL-I. Table I shows the saturation capacities of every immunoadsorbent and the protein yield for peak I which indicates that acidtreated Sepharose was the most effective matrix for the purification of lectin DCL-I. In further purifications, D. candidum BF-S was applied to an acid-treated Sepharose column until depleted from DCL-I, and the unbound material was applied to an asialofetuin Sepharose column. Both columns were washed to baseline absorbance and lectins were eluted as described before. Table I1 shows the results of the purification procedure which was followed. The increase in specific activities of the DCL-I and DCL-I1 preparations with respect to the starting material (S-BF) were 38-and lz-fold, respectively (Table 11). It should be considered that the specific activity of the S-BF represents the combined specific activity of both lectins in the S-BF. Since DCL-I represents at least about 11.5% of the total proteins in the S-BF, the high increase in specific activity of the purified DCL-I could be due to the presence of inhibitors, possibly glycoconjugates bearing galactosyl residues in the S-BF, that are separated during the purification procedure. Homogeneity of the DCL-I and DCL-I1 preparations which was confirmed by SDS-PAGE under reducing conditions (Fig. 3), by immunoelectrophoresis (Fig. 4), and by size exclusion HPLC in 6 M GdnHCl of the reduced and alkylated lectins (Fig. 5). Immunoelectrophoresis shows that DCL-I and DCL-I1 migrate with opposite charge at pH 8.6 and precipitate as single arcs with the rabbit anti-whole D. candidum BF-S (Fig. 4). Size exclusion HPLC in 6 M GdnHCl yielded single symmetrical peaks for reduced * Portions of this paper (including "Experimental Procedures," Tables I-VII, and Figs. 1-9) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are available from the Journal of Biological Chemistry, 9650 Rockville Pike, Bethesda, MD 20814. Request Document No. 86M-391, cite the authors, and include a check or money order for $9.60 per set of photocopies. Full size photocopies are also included in the microfilm edition of the Journal that is available from Waveriy Press. and alkylated DCL-I and DCL-I1 (Fig. 5).
Physicochemical Properties of DCL-I and DCL-11-Both species of D. candidum galactosyl-binding lectin exhibited a single sedimenting boundary during sedimentation velocity. Measurements on each species were confined to protein concentrations less than 1 mg/ml, and, under these conditions, a slightly positive dependence of the sedimentation coefficient on protein concentration was observed. Such behavior is suggestive of a tendency of the protein to associate at higher concentrations. This is in keeping with the rather low solubility routinely noted for each of the lectin species. The native molecular weights estimated for each of the lectin species by sedimentation equilibrium, 56,600 and 57,500, suggest that they are of similar molecular weight (Table 111). Furthermore, the frictional ratio estimated for each species from its molecular weight and hydrodynamic behavior suggests each is a globular protein.
As shown in Table 111, approximate subunit molecular weights estimated from semilog plots of molecular weight versus mobility in SDS-PAGE under reducing conditions (Fig.  6A) were different for DCL-I and DCL-11: 14,500 and 15,500, respectively. As shown in Table I11 and Fig. 6B, results obtained through gel filtration in 6 M GdnHCl of the reduced and alkylated subunits were slightly lower, but still showed a molecular weight difference of about 1,000 between DCL-I and DCL-I1 subunits. Thus, both empirical molecular weight estimations of the constituent polypeptide chains of each of the lectin species are consistent with each protein containing four equal-sized polypeptide chains.
As shown in Fig. 7, the far UV circular dichroism spectrum of DCL-I is qualitatively suggestive of a large proportion of @-structure in the protein. Analysis of the circular dichroism data according to the method of Siege1 et al. (22) suggests 29% a-helix and 37% @-structure in the protein. The fluorescence spectra of DCL-I (40 pg/ml in Tris-buffered saline) yielded wavelengths of maximum excitation and emission at 288 nm and 330 nm, respectively.
Carbohydrate analysis by gas chromatography of alditol derivatives from DCL-I showed that only traces of mannose and galactose were present in the samples, probably less than 0.01% each. Such a small amount of galactose could be residual sugar still bound to the combining site of the lectin. A significant glucose peak (1.12%) observed in the analysis of native DCL-I dropped to 0.09% for DCL-I samples that were subject to gel filtration in 6 M GdnHCl prior to carbohydrate analysis. This, of course, suggests that the glucose was a contaminant rather than a part of any putative oligosaccharide covalently attached to the DCL-I molecule.
Although size homogeneity was suggested by the behavior of DCL-I and DCL-I1 on SDS-PAGE or size exclusion chromatography in 6 M GdnHCl, isoelectric focusing in a pH 3-10 range in 6 M urea demonstrated a family of bands focusing at pH from 3.65 to 5.21 for DCL-I (Fig. &I) and a much less complicated profile for DCL-I1 (Fig. 8B).
A comparison of the amino acid compositions of DCL-I and DCL-I1 (Table IV) indicated that the two lectin forms differed mainly in their amounts of aspartic acid, serine, glutamic acid, alanine, cysteine, valine, phenylalanine, and histidine. For DCL-I, the total amount of aspartic and glutamic acid was 32, while histidine, lysine, and arginine totaled 13. For DCL-11, the total amount of aspartic and glutamic acids was 31, while histidine, lysine, and arginine was 16. This suggests that most of DCL-I1 acidic residues are in amidated form since the lectin focuses in the pH range from 9.2 to 10.2. It is noteworthy that methionine was absent in both DCL-I and DCL-11, while there are relatively high proportions of glycine, Galuctosyl-binding Lectins from Tunicate D. candidum alanine, serine, and aspartate in both lectins.
The amino acid compositions of DCL-I and DCL-I1 were compared to each other and to our data base by the S A Q method (20). Our data base consists of about 250 proteins including immunoglobulins, myeloma proteins, Pz-microglobulin, acute phase proteins, lectins, Thy-1, complement components, and totally unrelated proteins such as a-and Bhemoglobins, HLA antigens, trypsinogen, nerve growth factors, actin, serum albumin and prealbumin, glycophorin A, melanoma antigens, viral glycoproteins, lysozyme, a-fetoproteins, p-endorphin, etc. The parameter S A Q and related derivatives have been shown to provide a preliminary estimate of relatedness among proteins, although we emphasize that amino acid sequence data are necessary to confirm the identification. The methods, nevertheless, have been very useful, especially in predicting the conservation of mammalian immunoglobulin p chains (23), which was subsequently confirmed by amino acid sequence analysis and for the comparison of the structure of a Thy-1 homolog isolated from a tunicate species with Thy-1 isolated from a vertebrate species (24). Based upon comparisons of unrelated and related proteins carried out by Marchalonis and Weltman (20) and , values lower than 80 S A Q units suggest a high likelihood of relatedness. The frequency of unrelated proteins having values smaller than 100 would be less than 2%. The data of Table V indicate that the galactose-binding protein DCL-I is very similar in amino acid composition to the sialic acid-specific lectin of the tunicate, Halocynthia pyriformis. It also shows similarities to the rabbit C-RP, lamprey serum hemagglutinin and Ig p chain, stingray Ig light chains, carp Ig p chain, lectin "carcinoscorpin," a lectin from the Indian horseshoe crab Carcinoscorpius rotundicauda, and to heavy chain variable regions. DCL-I differs from DCL-I1 in 109 S A Q units, suggesting that they are marginally related. DCL-I1 shows similarities to carp, goldfish, and trout immunoglobulin light and p chains, murine T cell line products, shark and stingray immunoglobulins, and y chain variable regions. Higher S A Q values suggesting lower degrees of similarity were obtained comparing DCL-I and -11 to lectins from Botrylloides, lamprey (Petromyzon marinus) eggs, Limulus, slug Limax flavus (26), oyster Crassostrea virginica, and the so-called "immune protein" of the silk moth Cecropia.
We performed N-terminal amino acid sequence analysis on purified DCL-I and a comparison of the first 21 amino acids of DCL-I is made to the N-terminal sequence of mammalian and fish immunoglobulins and immunoglobulin variable regions (Table VI). Identities in up to five positions were found between D. candidum lectin DCL-I and various immunoglobulin variable regions. These included the aspartic acid ( D ) at position 1, the valine (V) at positions 2 and 3, 4, 12, and 19, and the serine (S) at position 7. In addition, we compared the first 21 residues of D. candidum lectin DCL-I with N-terminal amino acid sequences of two acute phase proteins, mammalian C-RP (27) and SAP (28), and the lectin from the horseshoe crab Limulus polyphemus (Table VII). Six identities were found between DCL-I and C-RP and three with SAP. No identities occurred with limulin, but an interesting overlapping was observed between C-RP, SAP, and limulin in the same stretch where identities with DCL-I, C-RP, and SAP occurred. These data support the results of the amino acid composition analysis (Table V) which suggested that the D. candidum lectin shows a relationship to immunoglobulin variable regions and C-RP. It is interesting that in the short segment which was compared, DCL-I shows highest sequence homology with the same regions of those proteins that exhibit low values of S A Q with respect to DCL-I.
Finally, we also examined the possible serological crossreactivities between Didemnum lectins and mammalian C-RP by enzyme-linked immunosorbent assay. Our results (Fig.  9, A and B ) showed substantial cross-reactivity of DCL-I with C-RP of human origin. Other lectins tested, such as Halocynthia lectins HPYL-I1 and I11 and bovine serum albumin, gave no cross-reaction.

DISCUSSION
We have isolated two lectins (DCL-I being the major lectin and DCL-I1 the minor lectin) from the plasma of the tunicate D. candidurn. Both lectins are present in the body fluids from individual colonies, as well as pooled fluids. Our previous report on the screening for lectins on 10 species of American tunicates (3) showed that multiple specific lectins are present in all species except Styela plicata in which only sialic acidbinding lectins could be detected. Results reported elsewhere (29) suggest that at least three different specific lectins that bind galactose and lactose are present in plasma of the tunicate Botrylloides leachii. We believe that this is the first reported isolation and detailed characterization of a lectin from a protochordate species.
Purification by affinity chromatography on three different galactosyl-containing immunoadsorbents (Sepharose 4B, acid-treated Sepharose 4B, and asialofetuin conjugated to Sepharose 4B) showed that both lectins, although galactosylspecific, bound differently to the three immunoadsorbents. DCL-I bound to the three immunoadsorbents (the best yield was achieved with acid-treated Sepharose 4B), while DCL-I1 could only be isolated from the asialofetuin column. Both lectins, however, were effectively eluted from the affinity columns with galactose gradients, with DCL-I1 requiring a higher galactose concentration for elution than did DCL-I. Isolated lectins DCL-I and DCL-I1 were homogeneous by SDS-PAGE immunoelectrophoresis and size exclusion HPLC in 6 M GdnHCl of the reduced and alkylated molecules. Based on these results, routine purification of DCL-I and DCL-I1 was accomplished by the use of two columns: first an acidtreated Sepharose 4B column was used to isolate DCL-I, and, subsequently, DCL-I1 from the flow-through was isolated on an asialofetuin Sepharose 4B column. This procedure resulted in a recovery of 69% of the total agglutination units processed.
A physicochemical analysis of both D. candidum lectins showed that DCL-I and DCL-I1 are slightly different in both subunit and native molecular weights in that both lectins are composed of four equal-sized subunits of approximately 14,000 and 15,000, respectively. Other properties such as &, , , f/fmin, and i (sedimentation coefficient, frictional ratio, and partial specific volume) are similar, although not identical, which suggests that DCL-I and DCL-I1 are globular proteins of molecular weights about 56,600 and 57,500, respectively. Thus, D. candidum lectins are relatively small molecules if compared with lectins isolated from other invertebrate species such as L. polyphemus (Mr = 400,000) (30), Tridacnu maxima (Mr = 470,000) (31), and Halocynthia roretzi (Mr = 600,000) (32).
Amino acid compositions of DCL-I and DCL-I1 are different enough to consider them distinct lectins ( S A Q value is 109 units) and each one might be constituted by isolectins as judged by isoelectric focusing profiles. Although homogeneous in size, both proteins DCL-I and DCL-I1 exhibit a certain degree of heterogeneity in charge i.e. they focus in gels as a family of multiple bands, DCL-I being more heterogeneous than DCL-11. The amino acid sequence analysis of the DCL-I gave no evidence of heterogeneity in the primary structure of the first N-terminal 21 residues and, of course, heteroge-neity may occur in other regions of the sequence in order to account for the multiple peaks observed in isoelectric focusing. These isoelectic focusing patterns are reproducible, which suggests that they are not the product of deamination of asparagine or glutamine during the purification procedure. Distinct isolectins have been demonstrated to be present in invertebrate species such as Helixpomatia (garden snail) (33), as well as plants such as Arachis hypogaea (peanut) (34).
Circular dichroism analysis indicates that DCL-I contains a high proportion of @-structure (37%) and also a significant amount of a-helix (29%). It resembles the galactosyl-binding lectin from Tridacna maxima (31) in the high content of pstructure (40%); however, T. maxima lectin only contains about 10% of a-helix. Most lectins from animal and plant sources such as Helix (35) and concanavalin A (36) contain a large amount of @-structure. Exceptions, however, include wheat germ lectin which only contains 10% of @-structure (37) and limulin 111, the lectin purified by Roche and Monsigny (38) from the horseshoe crab, which appears to lack any a-helix or @-structure. Antibodies, on the other hand, contain a high proportion of @-structure and very little a-helix.
DCL-I appears to possess no covalently attached oligosaccharide side chains, and, in this respect, differs from other invertebrate lectins that have been shown to contain large amounts of carbohydrates bound to the protein. These include limulin (39) (24%), Geodia cydonium (40) lectin (9.9%), and T. maxima lectin (31) (-7.0%). In this respect, DCL-I resembles the lectins I and I1 from the sponge Axinella polypoides (41), which contains only 0.5% carbohydrate, and some plant lectins such as peanut agglutinin, concanavalin A, and garden pea lectins that lack carbohydrate (42).
In addition to the classical immunoglobulin system from vertebrates, several other molecules or families of molecules such as complement components, acute phase proteins (C-RP, SAP, etc.), major histocompatibility complex products, Thy-1 antigens, some humoral bactericidal substances, and humoral or cell-bound invertebrate and lower chordate lectins have been thought to participate in non-self-recognition mechanisms. However, the question of homology among most of these molecules remains open. Although Sir Macfarlane Burnet long ago suggested that invertebrate lectins might be related to the early precursors of immunoglobulins from vertebrates, little subsequent information has been obtained to support this hypothesis. Recent evidence of possible relationships between invertebrate lectins and certain putative recognition molecules from vertebrate include reports on the similarity of combining sites for galactose between myeloma proteins and Tridacna gigas lectins (43), the "lectin properties" ascribed to certain acute phase proteins and complement components that were found to precipitate with carbohydrates (44), and the sharing of short stretches of amino acid sequence between putative recognition molecules (21). Work by Kaplan et al. (45) on the amino acid sequence of limulin showed no sequence homology with vertebrate immunoglobulins. However, Robey and Liu (46) found not only that limulin binds phosphorylcholine as C-RP does, but the two proteins actually possess a short stretch of amino acid sequence with a high per cent homology. We have reported (21) that limulin and C-RP cross-react with anti-idiotypic monoclonal antibodies made against the myeloma protein TEPC 15, which also binds phosphorylcholine, and that the three molecules share short stretches of amino acid sequence that may account for the described cross-reactivity. However, at this point, it is impossible to discriminate if this is the product of convergence of the two molecules.
Since data on the primary structure of D. candidum lectins are incomplete, we compared their amino acid compositions by the use of the parameter S A Q (20) in order to infer tentative relationships with other putative recognition molecules. Although less rigorous than the comparison of amino acid or gene sequences, these statistical methods ( S A Q of Marchalonis and Weltman (20), SAn of Cornish-Bowden (26)) have given results equivalent to those resulting from the comparison of amino acid sequences. Considering that the frequency of unrelated proteins having values smaller than 100 S A Q units would be less than 2% and taking 100 S A Q units as the upper limit of possible relationship, we found that DCL-I might be structurally related to the sialic acidbinding plasma lectin of the tunicate H. pyriformis. It also shows a considerable degree of relatedness with an egg lectin from the lamprey (P. marinus), with lamprey and carp p chains, with carcinoscorpin (the sialic acid-binding lectin from the Indian horseshoe crab C. rotundicauda), and with rabbit C-RP. A marginal degree of relatedness occurs with DCL-11. DCL-I1 shows amino acid composition similarities with several fish immunoglobulin light chains, with immunoglobulin-related molecules isolated from mouse and marmoset T cells, and with carp and goldfish immunoglobulin heavy chains.
Although the only rigorous quantitative criterion for assessing homology among molecules and demonstrating the existence of families is the comparison of their complete amino acid or gene sequences, the comparison of partial Nterminal sequence can also supply important information provided that a considerable number of identities exists between the molecules in question. In our comparison of DCL-I N-terminal amino acid sequence with immunoglobulin-related molecules, some interesting identities are observed. The comparison with mammalian acute phase proteins is encouraging in that several identities are found between DCL-I and C-RP and SAP in a region that overlaps with the sequence stretches that exhibit several identities with limulin.
Our data on the comparison of amino acid compositions and N-terminal sequences of Didemnum lectin I with acute phase proteins were supported by results obtained through a serological approach: enzyme-linked immunosorbent assay experiments showed antibodies made against human C-RP cross-reacted with DCL-I. The absence of significant binding to other lectins such as Halocynthia lectins HPYL-I1 and -111 and to bovine serum albumin suggests a certain specificity in the binding to DCL-I. It is possible that the identities in several positions of the amino acid sequence observed in the N-terminal stretch compared on the DCL-I and C-RP molecules may account for the cross-reactivities observed. We reported elsewhere (21) that anti-idiotypic monoclonal antibodies, made against the myeloma protein TEPC 15, crossreact with limulin, the sialic acid-binding lectin of the horseshoe crab L. polyphemus and C-RP (21). However, the anti-idiotypic monoclonal antibodies did not cross-react with DCL-I. This was not surprising since DCL-I binds galactosyl residues, while the other three molecules, TEPC 15, C-RP, and limulin, bind phosphorylcholine. Polyclonal antibodies made against C-RP cross-react with TEPC 15 and limulin, but also with DCL-I. This suggests that although TEPC 15, C-RP, and limulin might share common determinants related to their binding sites for phosphorylcholine, determinants showed by DCL-I and C-RP are probably located in other regions of the molecule. Moreover, it is relevant to our studies that C-RP has also been reported to bind galactosyl residues and to precipitate with galactans which suggest that C-RP would exhibit more than one specificity (44).
Although very diversified in morphological and ecological features are probably divergent and far removed from the sults reported here are encouraging in this respect.

13.
14. 15.  ranging from 20 to 60 grams) were pmehased from Gulf Specimen Company, he., Panacea. FL. The Didemnum mdidum body fluids; Twenty-eight colonies of D. Eandidum (single colony wet weights colonies were deared from adherent meterial, sliced with a ~caipel, and the body fluids (BF), mainiy plasma, were colleeted on ice. The BP was cleared from debts and the dark pigment p r t i d e s by centrifugation at 2,500 rpm a t 4'C fa 30 mi". Sodium seide was added to the Supernatant ( B P I ) to e final mncentration of 0.04% (wlv) and the BF-S was Stored at -25'C. Scheidemr (11) in 1.2% a~amre in barbital buffer DH 8.6 containing 20 mM ~alaetcse. EleetroDhoresis

MINI-PRINT SUPPLEMENT
Immmoelectrophorasisz Immmoeleftrophoresis was performed according to the technique of was cariced aut a t 120 V iw 60 mi". Migration i k p monitwsd by-BSA-Bromophenol blue dye marker. Gels were dsvelopsd overnight with rabbit anti-D candidum 8F-S (100 111 per trowh) fixed and stained with 0.2% Amido Black in 2% a c e t i c a c i d a n d % s t a i n e d t h 2% acetic acid. Whn(pre0ipltation with electrophoresed 5 described befwe, but in the absence of galactose in the gels. Gels were developed glycopteins and polysaccharides was exemined, Didemnum lectin samples (I mg/ml) were destained as described bel-.