The Primary Structure and Carbohydrate Specificity of a β-Galactosyl-binding Lectin from Toad (Bufo arenarum Hensel) Ovary Reveal Closer Similarities to the Mammalian Galectin-1 than to the Galectin from the Clawed Frog Xenopus laevis

The detailed characterization of a galectin from the toad (Bufo arenarum Hensel) ovary in its primary structure, carbohydrate specificity, and overall biochemical properties has provided novel information pertaining to structural and evolutionary aspects of the galectin family. The lectin consists of identical single-chain polypeptide subunits composed of 134 amino acids (calculated mass, 14,797 daltons), and its N-terminal residue, alanine, is N-acetylated. When compared to the sequences of known galectins, the B. arenarum galectin exhibited the highest identity (48% for the whole molecule and 77% for the carbohydrate recognition domain (CRD)) with the bovine spleen galectin-1, but surprisingly less identity (38% for the whole molecule and 47% for the CRD) with a galectin from Xenopus laevis skin (Marschal, P., Herrmann, J., Leffler, H., Barondes, S. H., and Cooper, D. N. W. (1992) J. Biol. Chem. 267, 12942-12949). Unlike the X. laevis galectin, the binding activity of the B. arenarum galectin for N-acetyllactosamine, the human blood group A tetrasaccharide and Galβ1,3GalNAc relative to lactose, was in agreement with that observed for the galectin-1 subgroup and those galectins having “conserved” (type I) CRDs (Ahmed, H., and Vasta, G. R. (1994) Glycobiology 4, 545-549). Moreover, the toad galectin shares three of the six cysteine residues that are conserved in all mammalian galectins-1, but not in the galectins from X. laevis, fish, and invertebrates described so far. Based on the homologies of the B. arenarum galectin with the bovine spleen galectin-1 and X. laevis skin galectin, it should be concluded that within the galectin family the correlation between conservation of primary structure and phylogenetic distances among the source species may not be a direct one as proposed elsewhere (Hirabayashi, J., and Kasai, K. (1993) Glycobiology 3, 297-304). Furthermore, galectins with conserved (type I) CRDs, represented by the B. arenarum ovary galectin, and those with “variable” (type II) CRDs, represented by the X. laevis 16-kDa galectin, clearly constitute distinct subgroups in the extant amphibian taxa and may have diverged early in the evolution of chordate lineages.

Galectins (Barondes et al., 1994a), comprise S-type ␤-galactosyl binding lectins, present in both homeotherm and poykilotherm vertebrates and invertebrates, that require a reducing environment but do not require divalent cations for their binding activity (Hirabayashi and Kasai, 1993;Barondes et al., 1994b). The primary structures of a considerable number of galectins are currently available (Hirabayashi and Kasai, 1993;Gitt et al., 1995;Hadari et al., 1995;Madsen et al., 1995;Hirabayashi et al., 1996). Furthermore, the three-dimensional structures of a limited number of galectins (galectins-1 (Bourne et al., 1994;Liao et al., 1994) and galectin-2 (Lobsanov et al., 1993)) have been reported recently, providing definitive information about their folding patterns, the amino acid residues that interact with the ligand and determine the architecture of the binding site, and the nature of the interactions that are established. Among the amino acid residues that are substantially conserved among various galectins, His 44 , Asn 46 , Arg 48 , Asn 61 , Trp 68 , Glu 71 , and Arg 73 (residue numbers are those of bovine spleen ) are recognized as critical for sugar binding (Lobsanov et al., 1993;Bourne et al., 1994;Liao et al., 1994) whereas Ser 29 , Phe 30 , Asn 33 , His 44 , Asn 46 , Arg 48 , Asn 61 , and Arg 111 interact with each other to provide the architecture of the CRD 1 .
Based on selected features such as their primary structure and subunit architecture, the galectin family has been subdivided in several subgroups ("proto," "tandem," and "chimera" (Hirabayashi and Kasai 1993); galectin-1-8 ( Barondes et al., 1994a)). Although all members of the galectin family bind lactose/N-acetyllactosamine, a limited diversity exists in the carbohydrate specificity (Leffler and Barondes, 1986;Abbott et al., 1988;Ahmed et al., 1990;Marschal et al., 1992;Oda et al., 1993;. Based on the differences in specificity and the conservation of amino acid residues that interact with the carbohydrate ligands, we classified galectins into two types: "conserved" (type I) and "variable" (type II) . This classification may reflect not only common features of their carbohydrate specificity, but possibly evolutionary aspects of their recognition functions. Most of the lectins under galectin-1 group have "conserved" CRDs and exhibit a very similar carbohydrate specificity (Leffler and Barondes, 1986;Abbott et al., 1988;Ahmed et al., 1990. The CRDs of lectins grouped as galectin-2, -3, -4, -5, -7, and -8 may have deletions or replacements in the relevant aforementioned amino acid sequence positions and are different from the conserved group in their carbohydrate specificities (Marschal et al., 1992;Oda et al., 1993;. Among the few amphibian species examined (subclass Anura: Bufo arenarum Hensel , Rana japonica (Sakakibara et al., 1979), Rana pipiens (Roberson and Armstrong, 1980), Rana tigerina (Shet and Madaiah, 1989), Rana catesbeiana (Ozeki et al., 1991), and Xenopus laevis (Marschal et al., 1992); subclass Urodela: Ambystoma mexicanum ), a number of distinct lectin activities have been identified in various adult tissues, including skin, muscle, and gonad. Among these, members of the galectin family have been isolated and characterized Shet and Madaiah, 1989;Ozeki et al., 1991;Marschal et al., 1992). However, a detailed biochemical and structural characterization, including the complete primary structure and carbohydrate specificity, has been accomplished only for the X. laevis galectin (Marschal et al., 1992). Yet the position of this lectin remains to be settled in the most recent classification (Barondes et al., 1994a). Because of subunit size and architecture, carbohydrate specificity, and primary structure of the CRD, this galectin has been classified as a "proto" type member (Hirabayashi and Kasai, 1993) carrying a "variable" (type II) CRD .
In contrast with X. laevis, where galectins are mainly confined to adult skin (Marschal et al., 1992) and have not been detected in embryos (Marschal et al., 1994), in the toad B. arenarum galectins are expressed in oocytes and further postfertilization stages such as single cell, two-cell, blastula, gastrula, and somite development . Differences in the location and developmental expression of galectins in the two amphibian species raise the possibility that, although present in phylogenetically closely related taxa, they might mediate mechanisms associated with substantially distinct biological functions, such as host defense in X. laevis (Marschal et al., 1992) or developmental processes in B. arenarum . In mouse embryogenesis, galectin-1 and -3 are detected in the trophectoderm of the blastocyst as early as just before its implantation in the uterine wall, and they are believed to play a role in attachment by binding lacto-N-fucopentaose I (Colnot et al., 1996). Following gastrulation, galectin-1 is first expressed in somite myotomes, whereas galectin-3 is confined to the notochord. However, unlike the mammalian galectin-1 or galectin-3, the activity of the B. arenarum galectin can be detected in every stage prior to blastula, suggesting further differences in its putative function(s) proposed for mammalian galectins, such as embryo implantation and development. Therefore, the toad B. arenarum may constitute a suitable model for the elucidation of the biological function(s) of galectins in the embryogenesis of poykilotherm vertebrates.
As a first step of a systematic initiative aimed at the investigation of galectin structure-function relationships, we report herein the biochemical characterization, primary structure, and carbohydrate specificity of a galectin from B. arenarum ovary and examine its primary structure similarities with those reported elsewhere from homeotherm and poykilotherm vertebrates. Surprisingly, our results suggest that the toad ovary galectin is closer, both in primary structure and carbohydrate specificity, to the bovine galectin-1 than to the galectin from a related amphibian species, X. laevis, indicating that, within this lectin family, structural and functional divergence may have occurred early in vertebrate evolution.

EXPERIMENTAL PROCEDURES
Reagents-The protein assay reagent was from Bio-Rad. Ampholine PAGplate and gel permeation chromatography molecular weight standards and ribonuclease A were purchased from Pharmacia Biotech Inc. The peroxidase substrate diammonium 2,2Ј-azinobis(3-ethylbenzothiazoline-6-sulfonate) (ABTS) was from Kirkegaard & Perry Laboratories. Sequencing grade reagents and solvents for protein sequencing, amino acid analysis, and HPLC were from Applied Biosystems (Division of the Perkin-Elmer Corp.). Sequencing grade endoproteinase Asp-N, endoproteinase Glu-C, and trypsin were from Boehringer-Mannheim, and lysyl endopeptidase was from Wako Bioproducts. Horseradish peroxidase (HRP) was from Sigma. All other reagents were of the highest grade commercially available.
Toad Ovaries-Toad (B. arenarum Hensel) ovaries weighing approximately 50 g each were dissected from human chorionic gonadotrophinstimulated females as reported elsewhere  and stored at Ϫ80°C.
Purification of Toad Ovary Galectin-The galectin was purified through an improved protocol as reported elsewhere (Ahmed et al., 1996). Briefly, toad ovaries (500 g) were homogenized with cold (4°C) phosphate-buffered saline (diluted 1:10), 0.01 M 2-mercaptoethanol, 0.1 M lactose (PBS(1:10)/ME/lactose) containing 0.1 mM phenylmethylsulfonyl fluoride (2 ml buffer/g wet tissue). After centrifugation, the clear supernatant was mixed with DEAE-Sepharose (10 ml of supernatant/1 of ml resin) preequilibrated with PBS(1:10)/ME. After gentle mixing for 1 h at 4°C, the slurry was transferred to a fritted glass funnel, and the resin was washed with 10 bed volumes of cold PBS(1:10)/ME to remove lactose and unbound protein. The bound proteins were eluted with 500 ml of PBS/ME, 0.002 M EDTA, 0.5 M NaCl, and the eluate was adsorbed on a column of lactosyl-Sepharose preequilibrated with PBS/ME, 0.002 M EDTA, 0.5 M NaCl. The column was thoroughly washed with equilibrating buffer followed by 5 bed volumes of PBS(1:10)/ME and the bound protein was eluted with 0.1 M lactose in PBS(1:10)/ME. The fractions containing protein were pooled, and aliquots were absorbed on DEAE-Sepharose columns (0.5 ml bed volume), overlaid with 50% glycerol in eluting buffer, and stored at Ϫ20°C until use.
Agglutination Tests-Agglutination tests were carried out in BSAcoated 96-well Terasaki plates (Robbins Scientific, Mountain View, CA) as reported earlier ). An equal volume of a 0.5% suspension of glutaraldehyde-fixed protease-treated rabbit red blood cells was added to 5 l of 2-fold dilutions of galectin in PBS/ME (pH 7.5). The plates were gently vortex-mixed and incubated at room temperature for 1 h. Agglutination was read under a microscope and scored from 0 (negative) to 4. The reciprocal of the highest dilution of galectin showing an agglutination score of 1 was recorded as the titer. The specific activity of the lectin was defined as the titer/mg of protein/ml. Controls were carried out by adding PBS/ME instead of lectin.
Protein Determination-Protein concentrations were determined on 96-well flat bottom plates with the Bio-Rad protein assay following a modification of the manufacturer's protocols, using BSA as a standard. To 100 l of 5-40 g/ml protein standard solutions and samples, 100 l of Coomassie Blue dye reagent prediluted 2.5-fold with water were added. After 5 min, the reactions were read in a Molecular Devices Plate Reader at 595 nm, and the data were analyzed through the Softmax program.
Polyacrylamide Gel Electrophoresis-Analytical polyacrylamide slab gel electrophoresis in the presence of sodium dodecyl sulfate (2%) was carried out on 15% (w/v) acrylamide gels under reducing conditions as reported elsewhere (Laemmli and Favre, 1973).
Gel Permeation Chromatography-Gel permeation chromatography of the galectin was carried out in a Pharmacia Superose 6 (1 ϫ 30 cm) column equilibrated with PBS/ME, 0.25 M NaCl, 0.01 M lactose (pH 7.5) in a high performance liquid chromatography (HPLC) system that consisted of a Beckman-116 pump, a Beckman programmable detector module-166 (280 nm), and a Panasonic KX-1080i recorder, at a flow rate of 0.4 ml/min. Gel permeation chromatography of the horseradish peroxidase (HRP)-conjugated galectin was carried out with PBS (azide-free), 0.25 M NaCl, 0.01 M lactose (pH 7.5) in the similar way as described above. BSA (66 kDa), ovalbumin (43 kDa), carbonic anhydrase (29 kDa), and ribonuclease A (13.7 kDa) were run as standards. Apparent molecular weights were determined from a plot of M r versus K av , where K av is defined as Isoelectric Focusing-Analytical isoelectric focusing of the galectin (20 g) was carried out on a thin (1 mm) layer (5% polyacrylamide) Ampholine PAGplate (Pharmacia) (pH range, 4.0 -6.5) in an EC 1001 electrophoresis unit (EC Apparatus Corp.) according to the manufacturer's instructions. The pH gradient was determined by measuring the pH of the supernatant solutions obtained by grinding 5-mm slices of the gel in 1 ml of distilled water. The gel was fixed with sulfosalicylic acid-trichloroacetic acid-water, stained with 0.1% Coomassie Brilliant Blue R 250 in ethanol-acetic acid-water, and destained with ethanolacetic acid-water. The gel was scanned densitometrically with a BioImage Gel Scanner (Millipore). The pI was determined from a plot of pI values of markers versus distance from the cathode of pI standard markers. The pI markers (Sigma) used were glucose oxidase (pI 4.2), trypsin inhibitor (pI 4.6), ␤-lactoglobulin A (pI 5.1), and carbonic anhydrase II (pI 5.4 and 5.9).
Thermal Stability-The temperature stability of the toad ovary galectin was determined by incubating 100-l samples in PBS/ME (3 g/ml) at various temperatures for 30 min, cooling them on wet ice, and titrating them with glutaraldehyde-fixed protease-treated rabbit red blood cells .
Stability in a Nonreducing Environment-The purified toad ovary galectin (100 g) was absorbed on 1 ml each lactosyl-Sepharose and asialofetuin-Sepharose, and each matrix was thoroughly washed with aerated PBS (20 ml/1 ml matrix) and stored at room temperature. Control matrices contained the same amount of lectin in PBS/ME. After 8 days, the lactosyl-Sepharose and asialofetuin-Sepharose columns were eluted with 2 ml of PBS/ME, 0.1 M lactose, and the eluates were dialyzed with PBS/ME in the presence of 2 mg of BSA. The hemagglutinating activity was measured with glutaraldehyde-fixed proteasetreated rabbit red blood cells. In another set of experiments, 20 g of active lectin (in 100 l of PBS) were dialyzed with aerated PBS at room temperature for 18 h. The same amount of lectin was blocked with 0.1 M lactose in 100 l of PBS and dialyzed with aerated PBS, 0.1 M lactose at room temperature for the same period. The control had the same amount of lectin in 100 l of PBS/ME and was dialyzed with PBS/ME at room temperature for the same period of time.
Characterization of the Carbohydrate Specificity and Preparation of the Galectin-HRP Conjugate-The purified toad ovary galectin was carboxamidomethylated with iodoacetamide on a solid phase under mild conditions in the presence of excess ligand, yielding carboxamidomethylated galectin (CAM a -G). Unlike conventional methods (see below: CAM b -G in "Primary Structure Analyses"), this carboxamidomethylation procedure at lower temperature and shorter treatment time was optimized  in order to maximize the retention of the lectin's carbohydrate binding activity in nonreducing environments. Briefly, a DEAE-Sepharose column (0.5 ml bed volume) containing toad ovary galectin was washed with PBS(1:10), and immediately the column was overlaid with 1 ml of 0.1 M iodoacetamidecontaining 0.1 M lactose. After incubation at 4°C for 1 h in the dark, the column was washed with azide-free PBS(1:10) to remove excess reagent and lactose, and the CAM a -G was eluted with PBS (azide-free), 0.5 M NaCl. The CAM a -G was conjugated to HRP through glutaraldehyde coupling as follows. To a mixture of CAM a -G (0.7 mg) and HRP (2.0 mg) in 1.3 ml of PBS (azide-free), 0.5 M NaCl, 0.1 M lactose, 160 l of 1% glutaraldehyde were added. After overnight incubation at 4°C, the conjugation mixture was diluted 40-fold with cold water and adsorbed onto DEAE-Sepharose (0.5 ml) preequilibrated with azide-free PBS(1: 10). The column was washed to remove lactose, and the conjugate was eluted with 2 ml of PBS (azide-free), 1 M NaCl and purified by affinity chromatography on lactosyl-Sepharose as indicated above. Finally, the conjugate was separated from the unreacted galectin by gel permeation chromatography on a Superose 6 column as described above. The purified galectin-HRP conjugate was stored in 1% BSA, 50% glycerol at Ϫ20°C.
Optimal pH for Binding-To determine the optimal pH for toad ovary galectin binding, 24 ng of galectin-HRP conjugate in 60 l of water containing 0.1% Tween 20 were mixed with 60 l of various buffers (0.2 M), and 100 l of this mix were used in triplicate in the binding assay as described above. The buffers used were citrate-phosphate, pH 4.0 -6.0; phosphate, pH 6.5-8.0; and carbonate-phosphate, pH 8.5-9.5.
Solid Phase Binding Inhibition Assay-Binding of the toad ovary galectin to asialofetuin and its inhibition by sugars were determined on polystyrene wells of microtiter plates (Dynatech Laboratories). For that purpose, asialofetuin (0.5 g/100 l/well) in 0.1 M Na 2 CO 3 , 0.02% NaN 3 (pH 9.6) was adsorbed onto the wells of microtiter plates and incubated at 37°C for 3 h. After aspirating the residual asialofetuin solution, fixation was carried out with 2% formaldehyde in PBS at 37°C for 30 min. The plates were washed three times with PBS (azide-free), 0.05% Tween 20 and incubated with the galectin-HRP conjugate (for binding assays) or with a preincubated mixture of conjugate and test ligands (for binding inhibition assays). The preincubation of the galectin-HRP conjugate (24 ng in 60 l of azide-free PBS-Tween 20 buffer) for binding inhibition assays was carried out by mixing equal volumes conjugate and the test ligand at varying concentrations. After 1 h at 4°C, the conjugate-ligand mixture (100 l) was added to the wells in duplicate, and the plates were incubated for 1 h at 4°C. The plates were washed with ice-cold azide-free PBS-Tween 20 buffer, and the bound peroxidase activity was assayed with ABTS. To quantitate the amount of galectinperoxidase conjugate that bound to ASF, at the time the plate was developed, equal volumes containing increasing amounts of galectinperoxidase conjugate were added to uncoated wells and developed with ABTS under conditions (time and temperature) identical to the binding assay. From the absorbance value of each point, a standard curve was drawn.
Primary Structure Analyses-The toad ovary galectin was reduced with dithiothreitol and exhaustively carboxamidomethylated in solution. The toad ovary galectin (1.5 mg) in 3 ml of PBS/ME, 0.5 M NaCl was dialyzed with 0.01 M ammonium hydrocarbonate and freeze dried. The freeze-dried galectin was dissolved in 400 l of 8 M deionized urea, 0.05 M Tris-HCl, pH 8.3, reduced with 0.045 M dithiothreitol (40 l) at 50°C for 30 min, and carboxamidomethylated with 0.1 M iodoacetamide (80 l) in a nitrogen atmosphere at room temperature for 2 h. The carboxamidomethylated galectin (CAM b -G) was dialyzed against 0.01 M ammonium hydrocarbonate and freeze dried.
HPLC Purification of Peptides-The peptides were separated on a microbore RP-HPLC system consisting of Applied Biosystems model 140A pumps and model 1000S diode-array detector (2.3-l flow cell, 0.0025-inch inside diameter tubing). Fractionation of the peptides was performed either on a Zorbax-SB C-18 silica column (0.1 ϫ 15 cm, d p ϳ5 m, 300 Å pore size, Microtech Scientific, Saratoga, CA), or on an Aquapore ODS-300 C-18 silica column (0.1 ϫ 25 cm, d p ϳ 7 m, 300 Å pore size, Applied Biosystems) equilibrated at room temperature in 0.1% aqueous trifluoroacetic acid, and eluted at flow rates of 50 -80 l/min using linear gradients of acetonitrile/water/trifluoroacetic acid (80:20:0.1, v/v). The column effluent was monitored at 215 nm, the UV absorption spectra of the absorbing material were determined, and the eluate was manually collected and stored at Ϫ20°C prior to further analysis. Rechromatography of some fractions in the second RP-HPLC elution solvent system consisting of 2-propanol/acetonitrile/water/trifluoroacetic acid (70:20:10:0.1, v/v) was required prior to sequencing.
Sequence and Amino Acid Analysis-Automated Edman degradation of the peptides (Hewick et al., 1981) was performed on an Applied Biosystems model pulsed-liquid 477A/120A sequencing system. The PTH-amino acids were separated and identified as described (Pohl, 1994). The O ␥ -carboxamidomethyl ester derivative of PTH-Glu, PTH-Glu(CAM), which was identified during sequencing of several CAM b -G peptides, co-eluted with PTH-Thr. About 20% of PTH-Glu(CAM) was converted into PTH-Glu during conversion of its PTC derivative. Automated amino acid compositional analysis of the samples was done in an Applied Biosystems model 420A/130A derivatizer/PTC-amino acid analyzer equipped with an on-line hydrolysis unit assembly. The standard manufacturer's protocol was used to carry out both acid hydrolysis and derivatization with phenylisothiocyanate. The PTC-amino acids were identified and quantitated using the Dynamax chromatography software (Rainin Instruments). Under the elution conditions used, PTC-Cys(CM) was eluted between PTC-Asp and PTC-Glu.
Acyl N 3 O Shift on Blocked N-terminal Fragments, Ac-(1-13) and Ac-(1-21)-The purified peptides (150 -200 pmol) were dried in polypropylene test tubes, dissolved in dry trifluoroacetic acid containing 0.001% dithiothreitol (50 l, sequencer reagent R3), and sealed under argon. The N 3 O acyl shift on the serine and threonine residues (Fontana and Gross, 1986) was induced by incubating the trifluoroacetic acid solution at 54°C for 8 h or 12 h. The peptides were directly loaded onto the sequencer glass fiber disc (precycled with 3 mg of polybrene) and were subjected to Edman degradation.
Peptide Synthesis-Ac-(1-13), a 13-residue N ␣ -acetylated model peptide, was synthesized by solid-phase synthesis on the Rainin Instruments model Symphony/Multiplex peptide synthesizer, using the standard synthesis protocol utilizing the Fmoc (N-(9-fluorenyl)methoxycarbonyl)/tert-butyl protection strategy. The peptide was purified by preparative RP-HPLC; its mass and amino acid composition were confirmed.
Mass Spectrometry Analysis-The MALDI mass spectrometry analyses of fragments of CAM b -G were performed on a Kratos Instruments model KOMPACT MALDI III mass spectrometer or on a Bruker Instruments Protein TOF mass spectrometer. The HPLC fractions (0.3 l) were spotted on a target site of a 20-sample slide, followed by addition of 0.3 l of matrix (saturated ␣-cyano-4-hydroxycinnamic acid, Aldrich) dissolved in ethanol/water (1:1, v/v). The sample matrix was allowed to dry at room temperature for 5 min, and each sample was desorbed with 50 laser shots, each giving a spectrum. The shots were averaged to give the final spectrum. The instrument was operated in the linear mode and was calibrated using the external standard peptides. All ESI mass spectrometry analysis and tandem mass spectrometry analysis was performed at the Emory University Mass Spectrometry Center on a JEOL SX102/SX102B five sector mass spectrometer (configuration BE-BEE) using a JEOL Generation 3 ESI source. All peptides were dissolved in either 5:5:1 or 5:15:1 water/methanol/acetic acid and introduced by flowing the solution into the ESI source at a flow rate of 1 l/min at concentrations of 1-20 pmol/l. The full scan spectra were 50 -150 scans signal averaged at a resolving power of 1000 or 3000. Tandem mass spectrometry experiments were performed by mass selecting the ion of interest with MS1(B1E1) and examining the metastable products using link B/E scans (B2E2). The narrow range tandem mass spectrometry experiment was 1353 spectra signal averaged. The full range tandem mass spectrometry experiments were over 273 scans signal averaged over the m/z range of 550 -700. These conditions provided the best signal to noise ratio for the ions of interest.

RESULTS
Galectin Purification and Homogeneity-The improved protocol optimized for a mammalian galectin-1 (Ahmed et al., 1996) was equally satisfactory for purification of the toad ovary galectin. The use of DEAE-Sepharose prior to affinity chromatography allowed the partial isolation of the galectin and the removal of lactose present in the extraction buffer. The high salt eluate from DEAE-Sepharose was immediately loaded on the lactosyl-Sepharose column, the resin was washed to the base line, and the bound protein eluted with lactose. The yield of the purified protein was 8 -10 mg/kg of wet ovaries and had a specific activity of 2-5 ϫ 10 4 mg Ϫ1 ml. The purified galectin showed a single polypeptide corresponding to approximately 14.5 kDa on SDS-polyacrylamide gel electrophoresis under reducing conditions (Fig. 1C). However, gel permeation chromatography under nondenaturing conditions suggested that the galectin subunits may undergo a concentration-dependent dimerization (Fig. 1A). At the loading concentration of 1.4 mg/ml, the dimer:monomer ratio was 48:1, whereas at a concentration of 15 g/ml, the equilibrium shifted toward the monomeric species, with a dimer:monomer ratio of 3:1. Molecular mass estimates from gel permeation chromatography were 14.5 and 32 kDa for the monomer and dimer, respectively (Fig.  1B). On isoelectric focusing, the galectin exhibited nine isoforms with pI values ranging from 4.49 to 4.92 ( Fig. 2) with prevalent components at pI 4.73 (44%), 4.78 (19%), and 4.8 (17%).
Stability-The toad ovary galectin retained full agglutinating activity when maintained at temperatures up to 37°C, but a gradual decrease of activity occurred between 42 and 70°C. The activity was completely abolished when the protein was heated at 85°C for 30 min (Fig. 3). The stability of the lectin binding activity in nonreducing environments was examined under two different conditions: adsorbed on affinity matrices and in solution, with and without its ligand. The lectin, adsorbed on lactosyl-Sepharose or asialofetuin-Sepharose in a buffer containing no reducing agent, retained full activity after 8 days at room temperature. The lectin also retained full activity in the absence of a reducing agent if maintained in solution in the presence of excess ligand (0.1 M). The activity of the lectin, however, was reduced almost 40-fold in solution, in the absence of both reducing agent and ligand, and could not be restored by the addition of the reducing agent (data not shown).
sponding to approximately 71 kDa, represents the galectin-HRP conjugate, most likely resulting from equimolar crosslinking of galectin (29.6 kDa) and HRP (approximately 44 kDa), and exhibits both peroxidase and hemagglutinating activity. The fractions containing the 71-kDa species were pooled and used for binding assays. About 20% of the affinity-purified protein was recovered in the active galectin-HRP conjugate pool. To optimize the solid phase assay, the wells of the microtiter plates were coated with varying concentrations (0.015-16 g/well) of ASF, followed by fixing and extensive washing, and finally, the addition of varying concentrations (5-40 ng/well) of galectin-HRP conjugate. Fig. 5a shows the binding profiles for each conjugate concentration tested: increased binding was observed with increasing amounts of conjugate added for all concentrations of the coating ASF. Except for the highest conjugate concentration tested (40 ng/well), maximum binding was achieved at approximately 0.5 g of ASF/well, with absorbance values reaching a plateau beyond that concentration. Fig. 5b shows the binding of variable amounts of conjugate to ASFcoated wells (0.5 g/100 l/well) as a function of time. For all conjugate concentrations tested, the binding was approxi-mately linear during incubation times up to 21 min. From these preliminary studies, the optimal amount of ASF for coating the plates was established as 0.5 g/100 l/well, and the amount of galectin-HRP conjugate to be added for the binding and binding inhibition experiments, as 20 ng/100 l/well. The optimal substrate incubation time was determined to be 20 min. Approximately 2 ng of the galectin-HRP conjugate was bound to ASF under the optimal conditions. No further blocking of the ASFcoated wells was required, but the addition of Tween 20 to binding and washing buffers substantially reduced any background absorbance (results not shown). Under the optimal conditions established, increasing concentrations of lactose inhibited the binding of galectin-HRP conjugate to ASF in an approximately sigmoid profile reproducible on several experiments (Fig. 5c). The lactose concentration required for 50% inhibition of the galectin-HRP conjugate binding to ASF varied from 60 to 90 M. The binding activity of toad ovary galectin to ASF was optimum at pH 7.5-8.5 (Fig. 5d). The activity declined drastically below pH 7.0. All further experiments were carried out at pH 7.5. As optimized, this assay for determining galectin's carbohydrate-binding specificity is simple, sensitive, and rapid, since it was completed in less than 3 h after the plates were coated with ASF.
Carbohydrate Specificity-The carbohydrate-binding specificity of toad ovary galectin was determined by analyzing the binding of the galectin-HRP conjugate to ASF in the presence of several saccharides through the solid phase assay, under the optimal conditions established. Preliminary studies applying this method to selected carbohydrate ligands yielded a relative inhibitory activity profile identical to that obtained through hemagglutination-inhibition assays with the unmodified toad ovary galectin , indicating that the specificity of the lectin was not modified through the conjugation procedure. This validation of the solid phase method developed allowed comparisons of the relative inhibitory activities of carbohydrates tested with those estimated through other well established methods that use unmodified (Ahmed et al., 1990) or radiolabeled (Leffler and Barondes, 1986) galectins. For each test saccharide, a complete inhibition curve was determined, and the molar concentrations that inhibited the binding of the lectin conjugate to ASF by 50% (I 50 ) were calculated and nor- malized with respect to the lactose included in each plate as a standard (Table I). The inhibitory potencies of several test oligosaccharides and methyl glycosides relative to the I 50 value of galactose, the nonreducing terminal monosaccharide common to all, are summarized in Table II. The binding of the toad ovary galectin with Gal␤1,3(4)GlcNAc and thiodigalactoside were about 4-fold higher than that with lactose. The binding of the galectin to the human blood group A tetrasaccharide Fuc␣1,2[GalNAc␣1,3] Gal␤1,4Glc was about 9-fold weaker than to lactose. The binding to Gal␤1,3GalNAc was negligible. Therefore, the overall binding inhibition pattern of the toad ovary galectin suggests a specificity very similar to those observed in mammalian 14-kDa lectins exhibiting conserved (type I) carbohydrate-recognition domains , but rather different from that of the clawed frog X. laevis, for which N-acetyllactosamine and thiodigalactoside were 2-and 5-fold less effective than lactose, and the human blood group A tetrasaccharide almost as effective as lactose (Marschal et al., 1992).
Primary Structure Analyses-Edman degradation of 100 -200 pmol of the native lectin or CAM b -G did not yield an N-terminal sequence, indicating that the protein N terminus is not accessible to sequencing, i.e. "it is blocked." CAM b -G was then cleaved with trypsin and lysyl endopeptidase, and the resulting peptides were purified by RP-HPLC. All expected fragments from both digests were recovered and completely sequenced (Fig. 6); however, it was noted that cleavage of several trypsin and/or lysyl endopeptidase sensitive bonds (e.g. Lys 13 -Pro 14 , Lys 29 -Gly 30 , Lys 64 -Glu 65 , and Arg 74 -Glu 75 ) proceeded slowly and led to substantial material losses during HPLC purification. For this reason, the protein was also cleaved using a mixture of trypsin and lysyl endopeptidase. These fragments were recovered in high yield on RP-HPLC, were all sequenced, and had their masses confirmed by MALDI. The overlaps for the tryptic and lysyl endopeptidase cleavage sites were obtained by sequencing the Asp-N and Glu-C peptides of CAM b -G (Fig. 6). The core of the amino acid sequence was thus elucidated; however, the structure of the blocked N terminus remained to be determined.
The N-terminal Structure-The tryptic and lysyl endopeptidase digests of CAM b -G contained peptides, Ac-(1-21) and Ac-(1-13), respectively, that were eluted as distinct entities on RP-HPLC and that did not yield an unambiguous sequence on Edman degradation (see below), suggesting that these are the blocked N-terminal peptides of CAM b -G. 2 Cleavage of Ac-(1-21) with pepsin afforded five fragments: three yielded a unique sequence, VAVTNL, VTNL, and NLKPGHCVEIK. The other two fragments were N-terminally blocked; their MALDI mass   spectrometry analysis was unsuccessful. The peptic peptide of Ac-(1-21), NLKPGHCVEIK (observed average molecular mass: 1295 daltons, calculated: 1295 daltons), identifies Ac-(1-21) as the N-terminal fragment of the protein (Fig. 6). Sequencing of both Ac-(1-21) and Ac-(1-13) yielded a low PTH signal in the first several cycles. The PTH signal for several amino acids was, however, onset and leveled off in the later cycles as is demonstrated for Ala, Gly, and Val in Table  III. It was reasoned that the delayed onset of the sequencing signal was caused by an incremental removal of the N ␣ -protective group, or by a partial cleavage of the selected peptide bonds. Such modifications would likely occur during the acid cleavage step of Edman degradation (Fontana and Gross, 1986;Weller et al., 1990). To test this possibility, the peptides were treated with trifluoroacetic acid at 54°C for 8 h or 12 h prior to sequencing. Indeed, Edman degradation of the acid-treated peptides proceeded efficiently from the first cycle: two sequenceable fragments were found in each fraction, and the C-terminal sequence of the longer fragment, TNLNLK in Ac-(1-13), and HCVEIK in Ac-(1-21), was identified (Table III). These C-terminal sequences align with the peptic fragments of Ac-(1-21) (see Fig. 6). Since Ser and Thr were the only Nterminal residues identified in the acid-treated material, it was concluded that acid treatment induced the acyl N 3 O shift on these two residues. Clearly, the N 3 O shift on the Thr residue opened the TNLNLK sequence. By inference, therefore, the N 3 O shift on the Ser residue opened the SAGVAVTNLNLK sequence. Hence, the partial sequence of Ac-(1-21) was identified as XSAGVAVTNLNLKPGHCVEIK (X being the blocking group).
The identity of the blocking group was determined by amino acid analysis and mass spectrometry analysis of Ac-(1-13). The difference between the amino acid composition of Ac-(1-13), Asx 2 , Ala 3 , Ser 1 , Gly 1 , Thr 1 , Val 2 , Leu 2 , Lys 1 (Table IV), and its partial sequence determined by Edman degradation, SAGVAVTNLNLK, is one alanine residue. The difference between the determined monoisotopic molecular mass of the peptide (1298.7 daltons) and its monoisotopic molecular mass based on its amino acid composition (calculated: 1256.7 daltons) is 42.0 daltons, suggesting that X is N ␣ -acetyl-alanyl. To further determine that the N-terminal sequence of Ac-(1-13) was indeed C 2 H 3 O-Ala-Ser-, tandem mass spectrometry was performed. The peptide should yield a distinct y 12Љ (nomenclature of Roepstorff and Fohlman (1984)) fragment of 1186.68 daltons. The doubly charged species (m/z ϭ 650) was chosen for the tandem mass spectrometry. Experiments with synthetic Ac-(1-13) showed that the proposed sequence produced a doubly charged y 12ٞ 2ϩ . Therefore, a narrow range metastable mass spectrum of the isolated peptide was obtained. Indeed, the y 12ٞ ϩ ion (m/z ϭ 593.8) that was produced corresponds to the C 2 H 3 O-Ala-Ser-sequence (data not shown). Attempts to obtain full scan range tandem mass spectra to prove by mass spectrometry the entire sequence were unsatisfactory because of the amount of the sample available. However, the observed peaks in the full scan range tandem mass spectra were consistent with the proposed sequence (not shown).

Molecular Mass and Amino Acid Composition of Galectin-
The ESI mass spectra of the full-length protein are in complete agreement with the determined sequence. The ESI of the native protein found ions that correspond to average masses of 14,791, 14,867, 14,943, and 15,020 daltons. These correspond to the pure native protein and to its mono-, di-, and tri-disulfidelinked adducts with 2-mercaptoethanol (calculated: 14,797, 14,873, 14,949, 15,015). 3 CAM b -G, carboxamidomethylated at its four cysteine residues, gave an ion that corresponds to an average molecular mass of 15019 daltons (calculated: 15,025 daltons). A minor species of CAM b -G (average molecular mass: 15,078) that was also identified likely corresponds to CAM b -G esterified at Glu-37 (calculated: 15,082) by iodoacetamide. 4 The amino acid composition of CAM b -G (Table IV) is in excellent agreement with the determined sequence.
amphibian, the clawed frog X. laevis was 48%. The percent similarities of the toad ovary galectin were 82 and 79% with bovine spleen galectin-1 and X. laevis galectin, respectively. We compared the predicted secondary structures of toad ovary galectin, bovine spleen galectin, and X. laevis galectin, calculated according to the method of Chou and Fasman (1978) (Fig.  8). The three-dimensional structure of the bovine spleen galectin-1 (Liao et al., 1994) showed a correlation with its predicted secondary structure in 9 of 11 ␤-strands. Although the overall secondary structures of all three galectins (Fig. 8, a-c) were similar, within the CRDs (Fig. 8d), the toad and bovine galectins showed almost identical profiles, whereas significant differences between those and the X. laevis galectin were observed at positions corresponding to residues 52, 53, 63-70, and 73 (residue numbers are those of bovine spleen galectin). The hydropathy profiles of the above three galectins (Fig. 9) also indicated that the CRDs of toad ovary and bovine spleen galectins similar (Fig. 9, a and b), but both considerably different from that of X. laevis galectin (Fig. 9c).

DISCUSSION
The concentration-dependent monomer-dimer equilibrium observed for the purified toad ovary galectin is noteworthy and similar to that reported for the galectin-1 from CHO cells (Cho and Cummings, 1995). This noncovalent dimerization also resembles observations for other lectins that yield higher order aggregates as the protein concentration is increased . Furthermore, experimentally cross-linked galectins are more effective than dimeric galectins in mediating cell adhesion to plastic surfaces . Because most galectins have one binding site per subunit, it becomes clear that properties, such as agglutination, glycan precipitation, or the cross-linking between cells and extracellular matrix, would be mediated by the dimer rather than the monomeric form. Although the monomer has carbohydrate-binding capability which is independent of dimer formation, at very low concentrations the galectin may exist mostly as a monomer and as such, remain "inactive" in mediating any biological events requiring cross-linking of carbohydrate moieties. The dimer:monomer ratio (48:1) at the highest concentration (1.4 mg/ml, equivalent to 47 M) of toad ovary galectin was comparable to that observed for CHO galectin-1 at 80 M (Cho and Cummings, 1995). However, the concentration range of dimer-monomer equilibrium for the toad ovary galectin was broader than that of CHO galectin-1. At 800 nM CHO galectin-1 exists mostly   and  The N-terminal peptides of CAM b -G, Ac-(1-21) and Ac-(1-13), were subjected to Edman degradation yielding no PTH-amino acid signal above the background level signal in the first cycle, as demonstrated for PTH-Ala, -Gly, and -Val (data for Ac-(1-13) are not shown). Following exposure of Ac-(1-21) and Ac-(1-13) to neat TFA, two PTH-amino acids were identified in each sequencing cycle for both peptides, beginning with PTH-Ser and -Thr in the first cycle.
b The PTH derivatives of dehydro-Ala and its adducts with dithiothreitol were present but not quantitated. c The PTH derivatives of dehydroaminoisobutyric acid and of its adducts with dithiothreitol were present but not quantitated.

TABLE IV Amino acid composition of CAM b -G and Ac-(1-13)
The experimental values are the averages of three independent determinations; the numbers in parentheses are derived from the CAM b -G sequence. as monomer (80%) (Cho and Cummings, 1995), whereas most (75%) of the toad galectin was in dimer form at 15 g/ml (500 nM). The concentration at which most of the toad ovary galectin was present as a monomer was not established in the present study, but it would probably be below 20 ng/ml, because the lectin still agglutinated rabbit erythrocytes at that concentration (specific activity 2-5 ϫ 10 4 mg Ϫ1 ml).
As reported for most highly purified galectin preparations from diverse animal sources, the presence of reproducible isoform profiles was observed in multiple preparations of the toad ovary galectin. The range of pIs of the nine lectin bands was higher than the one reported earlier , and the pI values of the three major components (pI 4.73-4.80) were close to the theoretical value (pI 4.92) calculated from the primary structure. Peptide sequencing of the purified protein failed to yield alternate residues at any position that could provide a structural basis for the presence of isoforms, and therefore, it is unlikely that point mutations are the mecha-nism responsible. For galectin-3, the unequal phosphorylation of serine residues was shown to be the cause of isolectin formation (Huflejt et al., 1993), and although it is possible that such derivatization of amino acids is the source of the heterodispersity observed in most galectins, this has not been demonstrated yet for galectin-1.
The analysis of the optimal binding activity and stability of the toad ovary galectin under a variety of experimental conditions reveals that this protein can remain fully active for a long period of time in the absence of reducing agents, only if the cysteines are carboxamidomethylated or if the intact protein is stored in the presence of soluble or solid phase-bound ligand, in this case lactosyl-and ASF-Sepharose. Similar results were obtained with the CHO galectin-1 when bound to laminin-Sepharose (Cho and Cummings, 1995). For the electric eel galectin, lactose protects the lectin against inactivation as assessed by prevention of the loss of fluorescence by oxidation (Levi and Teichberg, 1981). It has been proposed that the FIG. 7. Sequence alignment showing amino acid sequence identities of various galectins. Identical residues to those of toad ovary galectin are shown by asterisks. Deletions (-) and open spaces are introduced for maximum homology and unknown amino acids, respectively. The amino acid residues that participate in interaction with N-acetyllactosamine  are shown as ϩ (hydrogen bonding), ⌬ (hydrophobic interaction), and * (water-mediated interaction). The amino acid residues (shown as #) interact with each other to provide architecture of the CRD. BH/S, bovine spleen and heart galectins-1 (Abbott et al., 1989;Liao et al., 1994); H14-I, human lung galectin-1 (Hirabayashi et al., 1987); R14, rat lung galectin-1 (Clerch et al., 1988); M14, mouse galectin-1 (Wells and Malluci, 1991); C14, chicken skin galectin-1 (Hirabayashi et al., 1987); C16, chicken liver galectin-1 (Sakakura et al., 1990); Rc, R. catesbeiana oocyte galectin (Ozeki et al., 1991); XL, X. laevis skin galectin (Marschal et al., 1992); E, electric eel galectin (Paroutaud et al., 1987); CE, conger eel skin mucus galectin (Muramoto and Kamiya, 1992); N16, nematode 16-kDa galectin, sequence shown from residue no. 16 ; N32-II, nematode galectin domain II (Hirabayashi et al., 1992); SP-I, sponge galectin (Pfeifer et al., 1993). ligand, lactose or N-acetyllactosamine, maintains the galectin in the active conformation (Lobsanov et al., 1993;Bourne et al., 1994;Liao et al., 1994), probably by preventing oxidative inactivation due to formation of intramolecular disulfide bonds (Tracey et al., 1992). It is interesting to note that the 14-kDa ␤-galactoside-binding lectin from the eggs of another amphibian, R. catesbeiana, does not require a reducing agent for preservation of its binding activity, although the structural basis of this observation has not been elucidated (Ozeki et al., 1991). Among the six cysteine residues that are conserved in all galectins-1 characterized in higher vertebrates, none is conserved in the galectins from X. laevis (Marschal et al., 1992), the conger eel Conger myriaster (Muramoto and Kamiya, 1992), electric eel (Paroutaud et al., 1987), nor in those isolated from the nematode Caenorhabditis elegans (Hirabayashi et al., 1992) and the sponge Geodia cydonium (Pfeifer et al., 1993). In contrast, the toad ovary galectin constitutes the first example within the poykilotherm vertebrate lineages that exhibits three of six cysteine residues conserved in mammalian galectins-1 (see Fig. 7). The pH range for the optimal binding activity of the toad ovary galectin was rather narrow (7.5-8.5) and overlapped with that of its natural physiological environment. Like other galectins examined with regard to their thermal stability, the toad ovary galectin was irreversibly inactivated by exposure to high temperatures during a relatively short period of time. Therefore, in this and all the aforementioned biochemical aspects examined, the toad ovary lectin closely resembles those galectins isolated from mammalian sources, such as that from bovine spleen (Ahmed et al., 1996), possibly reflecting a conservation of biological roles of galectins from such phylogenetically distant taxa.
The carbohydrate specificity of the toad ovary galectin was also similar to the bovine spleen galectin and other mammalian galectins, but surprisingly different from that of the clawed frog X. laevis, another amphibian species. On the basis of differences in carbohydrate-binding patterns and the conservation of critical amino acid residues in the CRD, we categorized galectins into two groups, type I (conserved) and type II (variable) . The relative inhibitory efficiencies of four key oligosaccharide structures, lactose, N-acetyllactosamine, Gal␤1,3GalNAc, and the human blood group A-tetrasaccharide ( Fig. 10a-d) can provide the preliminary information required for the assignment of a galectin to either group. The hydroxyls (Fig. 10, boldface letters) at C-4Ј and C-6Ј of galactose residue are critical for both conserved and variable CRDs. Neither epimerization (for OH at C-4Ј) nor substitutions (for both OH) are possible without considerable changes in binding affinity. The 3-OH (shadowed background) of Glc/Glc-NAc (in Gal␤1,4Glc or Gal␤1,4GlcNAc) or the 4-OH of GlcNAc (in Gal␤1,3GlcNAc) cannot be epimerized or substituted for the conserved CRDs, but for variable CRDs the epimerization is allowed, since Gal␤1,3GalNAc (Fig. 10c) is an equally potent inhibitor as Ga␤1,4Glc in RI36-I (Oda et al., 1993;. The substitution of 2-OH (outlined letters) of Glc residue by NHAc in Gal␤1,4GlcNAc (Fig. 10b) promoted binding 5-10-fold compared to Gal␤1,4Glc for conserved CRDs, but for variable CRDs, various degrees of binding (negligible to 11-fold better) were observed . The substitution of OH (italic letters) at C-2Ј or C-3Ј individually or both, did not affect the binding dramatically. The 2-3-fold weaker binding in the case of conserved CRD was observed probably because of steric hindrance between the bulky substituents and the interacting amino acids of the protein. Interestingly, for some galectins having variable CRDs, the substitutions at C-2Ј and C-3Ј (Fig. 10d) promoted binding 10 -32-fold (Leffler and Barondes, 1986;Sparrow et al., 1987;Oda et al., 1993), probably due to deletions in the CRD domain that yield a structure that can accommodate the bulky substitutions . In summary, the analyses of sugar In d, numbers are those from bovine spleen galectin-1  aligned to each relevant amino acid starting from residue 44: q, toad ovary; E, bovine spleen; ϫ, X. laevis skin. binding specificity suggest that the relative binding affinity of the conserved (type I) CRDs for the oligosaccharides in question would be in the following decreasing order: N-acetyllactosamine Ͼ lactose Ͼ A-tetrasaccharide Ͼ Gal␤1,3GalNAc. Both the galectin-1 from bovine spleen (Ahmed et al., , 1996 and the toad ovary galectin (this study) showed the above order of specificity.
The results of peptide sequencing described above (Fig. 6) established that the 14.8-kDa lactose-binding lectin from the B. arenarum ovary is a member of the S-lac lectin family and very close to the galectin-1 group (Barondes et al., 1994a). Within the galectin-1 group, it most closely resembled the galectin-1 from bovine heart or spleen (48% overall identity) (Fig. 7), and contrary to what might be expected from consideration of the phylogenetic distances, the overall identity was 38% with the 16-kDa galectin from another amphibian species X. laevis. Interestingly, the carbohydrate specificities of the two amphibian and mammalian galectins reflected their structural relationships. Studies on three-dimensional structures of galectin-1 from bovine spleen and heart Bourne et al., 1994) and galectin-2 from human (Lobsanov et al., 1993) suggest that the CRD comprises the amino acid residues 44 -73 (residue numbers are those of bovine spleen ). Within the CRD, the relevant amino acids responsible for sugar binding (His 44 , Asn 46 , Arg 48 , His 52 , Asp 54 , Asn 61 , Trp 68 , Glu 71 , and Arg 73 ) are identical among the sequences of toad ovary galectin and the galectins-1 from bovine spleen/heart, human lung and rat lung. The X. laevis galectin CRD also contains all the above amino acids except for positions 52 (Ser instead of His) and 73 (Lys instead of Arg).
Although all three galectins from toad ovary, bovine spleen and X. laevis apparently showed an overall similarity in secondary structure (Fig. 8a-c) as calculated according to the method of Chou and Fasman (1978), it is clear that the profile for the toad ovary galectin is almost identical to that of bovine spleen, and both are less similar to the X. laevis galectin. Within the CRDs, the ␤-sheet content profiles (Fig. 8d) for toad ovary (closed symbol) and bovine spleen (open symbol) are considerably different from that of X. laevis (cross symbol) in residues 52, 53, 63-70, and 73 (residue numbers are based on the bovine spleen galectin amino acid sequence). Among those, Arg 73 is directly involved in the binding of bovine spleen galectin-1 to N-acetyllactosamine hydroxyl on C-3 of GlcNAc, Trp 68 is involved in hydrophobic interactions with the Gal pyranose ring, whereas the main chain carbonyl of His 52 and side chain of Arg 73 participate in water-mediated interactions with the N atom of the N-acetyl group . X. laevis lacks His 52 and Arg 73 in its CRD and also shows a decrease of binding toward N-acetyllactosamine as compared to that for lactose. The hydropathy profile of the toad ovary galectin CRD also closely resembled that of the bovine spleen galectin CRD, but both differ from the X. laevis profile (Fig. 9). Therefore, it is likely that the differences in relevant amino acid residues (Ser instead of His 52 and Lys instead of Arg 73 ) in the CRD of X. laevis constitute the structural basis for the differences in carbohydrate specificity observed between this galectin and that from toad ovary.
In the present study we have determined the primary structure of the toad ovary galectin that reveals the presence of all relevant amino acids identical to mammalian galectin-1 and exhibits a type I (conserved) CRD, also found in other mammalian examples such as bovine spleen galectin-1. As can be predicted from the conserved CRD, the toad ovary galectin exhibits carbohydrate-binding specificities very similar to that described in bovine spleen galectin-1 (Ahmed et al., 1996). Therefore, our results suggest that galectins carrying conserved (type I) CRDs, such as the B. arenarum ovary galectin and those with variable (type II) CRDs and represented by the X. laevis 16-kDa galectin, are clearly distinct subgroups in the extant amphibian taxa and may have diverged early in the evolution of chordate lineages. Whether a 14-kDa galectin similar to the toad ovary galectin and to those included in the galectin-1 subgroup, such as the bovine galectin-1, is present in X. laevis ovary remains to be demonstrated. A preliminary assessment on X. laevis eggs and embryos by conventional methodology (Nishihara et al., 1986;Milos et al., 1990) and by sensitive immunological and RNase protection assays (Marschal et al., 1994) reported elsewhere, failed to reveal the presence of any galectin activity. Based on the homologies among the three galectins calculated either for their full-length amino acid sequences (48% between toad ovary and bovine spleen and 38% between toad ovary and X. laevis) or for their CRDs alone (77% between toad ovary and bovine spleen and 47% between toad ovary and X. laevis) it should be concluded that the correlation of their phylogenetic distances among the source species may not be as direct for the galectin family members as proposed elsewhere (Hirabayashi and Kasai, 1993;Kasai and Hirabayashi, 1996). A likely explanation for this observation may be that mutation rates would not be constant within the galectin family and the galectin-1 subgroup may have evolved under low mutation rates due to functional constraints. If this is the case, the galectin-1 would be a considerably homogeneous category within this lectin family and the high conservation of residues in the type I CRD that interact directly with the ligand, relative to those in the Type II CRD  would buttress this idea.
Acknowledgments-We thank M. C. Sullards for the ESI mass spectrometry analysis of the whole protein, Dr. A. Woods for help with MALDI mass analysis of several peptides, and Drs. M. S. Quesenberry and F. Hubalek for critically reading the manuscript.