The Mushroom Marasmius oreades Lectin Is a Blood Group Type B Agglutinin That Recognizes the Galα1,3Gal and Galα1,3Galβ1,4GlcNAc Porcine Xenotransplantation Epitopes with High Affinity*

A blood group B-specific lectin from the mushroomMarasmius oreades (MOA) was investigated with respect to its molecular structure and carbohydrate binding properties. SDS-PAGE mass spectrometric analysis showed it to consist of an intact (H; 33 kDa) and truncated (L; 23 kDa) subunit in addition to a small polypeptide (P; 10 kDa). Isolation in the presence of EDTA produced only the H subunits, indicating that the latter two are formed by metalloprotease cleavage of the intact H subunit. Tryptic digestion of the H, L, and P polypeptide chains followed by mass spectral analysis supports this view. The lectin strongly precipitated blood group type B substance, was nonreactive with type A substance, and reacted weakly with type H substance. Carbohydrate binding studies reveal a high affinity for Galα1,3Gal (but not for the isomeric α1,2-, α1,4-, and α1,6-disaccharides); Galα1,3Galβ1,4GlcNAc; and the type B branched trisaccharide. MOA also reacts strongly with murine laminin from the Engelbreth-Holm-Swarm sarcoma and bovine thyroglobulin, both of which contain multiple Galα1,3Galβ1,4GlcNAc end groups. This linear B trisaccharide is a component of porcine tissues and organs, preventing their transplantation into humans. MOA also shares carbohydrate recognition of this trisaccharide with toxin A elaborated by Clostridium difficile.

Blood group-specific lectins have served as serological reagents for typing blood for over 65 years (1)(2)(3). Among the useful reagents are Dolichos biflorus lectin for human type A 1 erythrocytes (4,5), the B 4 Griffonia simplicifolia isolectin for type B erythrocytes (6), and the Ulex europeus I (7,8) and eel (Anguilla anguilla) serum agglutinins for type O(H) erythrocytes (1,2). Of all the blood group-specific lectins studied, group B-specific lectins are probably fewest in number.
Presently, the fruiting bodies of mushrooms are being studied as sources for lectins with novel carbohydrate binding activity. Several surveys have been published that classify these proteins based on their ability to agglutinate human (and rabbit) erythrocytes according to blood type (9 -12).
Crude extracts of the mushroom Marasmius oreades (Tricholomataceae) were first reported to exhibit human type B erythrocyte agglutinating activity in 1951 (9,13). A lectin from this source was isolated by affinity chromatography on a matrix of ␣-galactosyl-polyacrylamide gel by Horejsi and Kocourek (14). It was reported to agglutinate type B erythrocytes six times more avidly than type A erythrocytes and was stated to be a heterodimeric protein of 50 kDa, with subunits of 33 and 23 kDa. No mono-or oligosaccharides (25 mM) tested were found to inhibit agglutination of human type B erythrocytes.
We have undertaken a more detailed investigation of this potentially important B-specific lectin with interesting results. Quantitative precipitation studies using a highly purified preparation with A, B, and O(H) blood group substances show absolute specificity for B versus A substance, with slight Obinding activity. Binding studies by the techniques of hapten inhibition of agglutination and precipitation, and hapten binding in solution by isothermal titration calorimetry established that the lectin possesses an extended binding site specific for the Gal␣1,3Gal structure, in contrast to other B-specific lectins having more general affinity toward ␣-galactosides. A companion paper (15) reports the cloning, sequencing, expression, and characterization of the recombinant lectin and indications of its homology to other important carbohydrate-binding proteins.
Quantitative Precipitation and Hapten Inhibition Assay-Quantitative microprecipitation assays and inhibition of precipitation by sugar haptens were performed as described previously (17). For quantitative precipitation assays, 20 g of MOA and increasing amounts of glycoprotein or polysaccharide were used in a total volume of 120 or 200 l, all in PBS plus 0.15 M NaCl. Quantitative carbohydrate hapten inhibition assays were conducted by adding increasing amounts of haptens to a mixture of MOA (20 g) and type B blood group substance (5.0 g). The reactions were incubated at 37°C for 1 h and then at 4°C for 48 h followed by centrifugation of the precipitates, washing, and determination of protein by the Lowry method (18).
Amino acid analysis, automated amino acid sequencing, and MALDI-TOF mass spectrometric analyses of the native lectin, gel-isolated subunits, and tryptic peptides thereof were performed at the University of Michigan Biomedical Core Facilities.
Isothermal Titration Calorimetry-Isothermal titration calorimetry was carried out as previously described (19) except that 1.0 ml of lectin solutions containing 0.08 -0.12 mM subunits were used. Since this volume underfills the titration cell, variable volume calculations were applied, using the Bindworks software installed in the instrument.
Hemagglutination Assays-Hemagglutination assays were performed using formaldehyde-treated erythrocytes in V-well microtiter plates as previously described (17). Titers were recorded as the greatest dilution of the lectin solution yielding visible agglutination. Agglutination of Ehrlich ascites tumor cells (grown in in vitro culture, kindly supplied by Dr. D. K. MacCallum, University of Michigan) was observed microscopically.
Fluorescein-labeled MOA Staining-MOA was fluorescently labeled using fluorescein isothiocyanate (Sigma) at pH 9.5 in the presence of 0.2 M lactose, followed by exhaustive dialysis. A solution of fluoresceinlabeled MOA (8 g/ml) in PBS containing 2% goat serum and 0.2% Triton X-100, in the absence or presence of 20 mM linear B6 trisaccharide, was applied to a cryostat section (10 m, fixed in paraformaldehyde) of porcine skeletal muscle. After standing at room temperature for 2 h, the staining solution was removed, and the section was washed with PBS and examined under a fluorescent microscope.
Preparation of Lectin-Fruiting bodies of M. oreades mushrooms were harvested in June and from August to October 2000 from grassy plots in Ann Arbor, Michigan, or were purchased from American Mushroom Hunter Corp. (Middle Town, NJ). Caps and undamaged stems were cleaned of soil and debris and chopped into small pieces. Initially, fresh tissue (45 g) was homogenized at 4°C in 250 ml of PBS containing 10 mM thiourea, 1 g/liter ascorbate, 50 mg/liter phenylmethylsulfonyl fluoride, and 1-2 g of insoluble polyvinylpolypyrrolidone (Sigma) in a Waring Blendor at high speed. The homogenate was stirred 3-4 h in the cold, strained through four layers of cheesecloth, and centrifuged at 13,000 ϫ g for 20 min. The supernatant solution was made 20% saturated with (NH 4 ) 2 SO 4 , and the resulting solution was centrifuged to remove a small amount of precipitate. This supernatant solution was adjusted to 80% saturation with solid (NH 4 ) 2 SO 4 and stirred overnight, and the precipitate was collected by centrifugation, redissolved in 20 ml of PBS, and dialyzed. An affinity column (2.5 ϫ 15 cm) of melibiose-Sepharose gel, prepared using divinyl sulfone coupling, was loaded with the dialyzed fraction and washed with PBS until the absorbance of the effluent at 280 nm became Ͻ0.1. Lactose (0.1 M) in PBS was then used to displace bound protein. The protein solution so eluted was dialyzed against PBS and passed through a second affinity column, Synsorb-B, consisting of type B trisaccharide (Gal␣1,3(L-Fuc␣1,2) Gal␤O(CH 2 ) 8 CONH) linked to diatomaceous earth. The bound lectin was eluted with 20 mM diaminopropane containing 0.15 M NaCl, pH ϳ11. The "pass-through" and displaced protein fractions were combined separately and assayed against a panel of human erythrocytes. Finally, the protein fraction displaced from the Synsorb B column was passed through a column of Synsorb A, which contains covalently bound type A trisaccharide. A very small amount of material was bound (Ͻ5%); it was eluted with 20 mM diaminopropane, 0.15 M NaCl. Both fractions were assayed against types A, B, and O erythrocytes.
It was later found (see below) that proteolysis and possibly oxidation was taking place during the isolation and purification steps. Subsequently, the extraction procedure was modified by including a protease inhibitor mixture (product P8215; Sigma) at the level of 1 ml/liter of extract buffer in place of phenylmethylsulfonyl fluoride, eliminating Ca 2ϩ and including 1.25 mM EDTA in all buffers, and carrying out the extraction and ammonium sulfate precipitations under an atmosphere of argon.

RESULTS
By employing a series of chromatographic affinity columns, we have isolated a lectin that strongly agglutinates human B erythrocytes with a concomitant low activity against O cells. Successive passage through Synsorb B and Synsorb A columns removed a slight activity against type A erythrocytes that was present in some preparations. The hemagglutinating activity against types A, B, and O cells and several other cell types is presented in Table I. Hemagglutination activity was unchanged by extensive dialysis of the lectin and assay in metalfree buffer containing EDTA, indicating the absence of a divalent metal ion requirement (data not shown). Subsequently, metal-free buffer containing EDTA was used routinely to prevent matalloprotease degradation of the lectin.
SDS-PAGE analysis of the lectin revealed the presence of two major bands at 33 and 23 kDa (designated H and L, respectively), as was previously reported (14). A significant band, not noted previously, was also observed at ϳ10 kDa, designated P (Fig. 1). Upon heavy loading of the gel, several minor bands migrating between the L and P bands were also observed. The absence of 2-mercaptoethanol in the sample preparation buffer did not alter the pattern of bands, indicating the absence of interchain disulfide links. Samples prepared without heating also showed a band at ϳ55 kDa, streaking between the bands, and lesser amounts of the other bands, suggesting that bands H, L, and P are associated into a heteromeric structure that is dissociated slowly by SDS at room temperature.
The native protein was subjected to size exclusion chromatography by HPLC, wherein it migrated as a single, symmetric peak having the same mobility as Pisum sativum and Xanthosoma sagittifolium lectins, both known to be of ϳ50 kDa (20,21). In the absence of a haptenic sugar, all three lectins appeared to be of much lower molecular mass based on calibration of the column with cytochrome c, ovalbumin, and bovine serum albumin. In the presence of 0.1 M lactose, but not methyl ␣-mannoside, MOA migrated slightly faster than ovalbumin (Fig. 2), at an apparent mass of 49.5 kDa, suggesting that the lectin interacted weakly with the column matrix, despite its being a silica-based noncarbohydrate matrix. Lectins frequently are observed to migrate through size exclusion matrices at slower rates than expected by their molecular masses, either because of weak interaction with the matrix (generally polysaccharide-based matrices) or because of their compact structures, usually with a high amount of ␤-structure. 2 Interestingly, neither sugar used altered the migration of the other two lectins, both of which bind mannose. These results do not rule out the possibility that native MOA is somewhat larger than 49.5 kDa, but it is clear from the HPLC results it exists in solution as a single molecular weight species composed of at least three different sized subunits.
The native lectin was subjected to MALDI-TOF mass spectrometric analysis, wherein mass ions corresponding to the three bands were detected (data not shown). Pairs of masses corresponding to the 23-and 32-kDa subunits, separated by ϳ180 and 260 Da, respectively, were observed (probably due to binding of the matrix), whereas the 10-kDa subunit exhibited a single molecular mass of ϳ9800 Da. Within the limits of the mass calibration, the sum of the two smaller molecular masses (corresponding to L and P subunits) approximates the larger molecular mass of band H. Mass spectrometric analysis of tryptic peptides from the various bands and total amino acid analysis (data not shown) also support the conclusion that the two lighter bands, L and P, are fragments of the intact band H.
No significant amounts of any amino acids were obtained during several cycles of automated amino acid sequencing of the native protein or of the H and L bands, indicating that they possess blocked N-termini. The P band released small amounts of several amino acid derivatives at some of the cycles (similar to traces seen in the native protein also), but no single sequence was detected, suggesting that it is also largely blocked and may be heterogeneous.
Proceeding from the observation that recombinant MOA is a homodimer consisting of two identical subunits of 32 kDa (15) whose solutions are completely colorless (no absorbance above 320 nm), we set out to attempt to isolate a similar protein from the mushroom, taking care to inhibit any protease activity that could give rise to the "clipped" lectin (the 23-and 10-kDa polypeptide chains) as well as prevent possible oxidation of aromatic side chains. The modified procedure described above indeed led to the isolation of a homodimer similar to the recombinant lectin in lacking any 23-and 10-kDa subunits and having no absorbance above 320 nm (data not shown).
To further study the blood group specificity of the M. oreades lectin, we conducted quantitative precipitation assays of the lectin with soluble cyst blood group substances. Fig. 3 shows that MOA reacts strongly with human blood type B substance, not at all with type A substance, and rather weakly with type H substance. The explanation for these reactions will be discussed in terms of the specificity studies described below.
Sugar ligand binding to MOA was conducted using three approaches: inhibition of type B hemagglutination, hapten inhibition of MOA-type B substance precipitation, and isothermal titration calorimetry. All three techniques gave approximately the same results, with the calorimetric data being the most precise. Being a blood type B agglutinin, MOA was assayed primarily against D-galactosyl-terminated sugars and oligosaccharides. As shown in Fig. 4 and Table II, lactose, N-acetyllactosamine, and melibiose (Gal␣1,6Glc) were very poor ligands. Methyl ␣-galactopyranoside was similarly very poor. The blood group B disaccharide, Gal␣1,3Gal, was an excellent ligand, with K a ϭ 6.0 ϫ 10 3 M Ϫ1 , whereas the isomeric disaccharides Gal␣1,2Gal, Gal␣1,4Gal, and Gal␣1,6Gal bound poorly or not at all. The addition of a GlcNAc group to the reducing end of Gal␣1,3Gal to give Gal␣1,3Gal␤1,4GlcNAc increased the binding by ϳ50% to K a ϭ 9.7 ϫ 10 3 M Ϫ1 . Similarly, adding an L-fucosyl group to the disaccharide to afford the blood group B branched trisaccharide (Gal␣1,3(L-Fuc␣1,2)Gal) enhanced its affinity to MOA 4-fold (K a 3.6 ϫ 10 4 M Ϫ1 ). Finally, we assayed the trisaccharide L-Fuc␣1,2Gal␤1,4Glc (fucosyllactose), related to the blood group H trisaccharide; it had a K a of 548 M Ϫ1 by isothermal titration calorimetry. It is apparent that the L-fucosyl residue makes a significant contribution to the binding affinity of the essentially inactive lactose (K a ϭ 185 M Ϫ1 ). This also is probably the reason that MOA recognizes and agglutinates human O erythrocytes to a limited extent and gives a weak precipitin curve with blood group H substance.
On the basis of MOA's recognition of blood type B disaccha-2 W. J. Peumans, personal communication. rides and linear trisaccharides, we believed that the lectin should precipitate both with laminin from the Engelbreth-Holm-Swarm murine sarcoma and with bovine thyroglobulin. Indeed, as shown in Fig. 5, MOA reacted strongly with laminin, which we have shown contains Gal␣1,3Gal␤1,4GlcNAc end groups (22,23), as well as with bovine thyroglobulin, which has the same determinants (24). Significantly, the lectin did not give a precipitin reaction with pigeon ovalbumin, which contains multiple Gal␣1,4Gal end groups (25), thus demonstrating its specificity for Gal␣1,3Gal groups. Also of significance, MOA did not recognize the blood group type A disaccharide (GalNAc␣1,3Gal), as shown by its failure to precipitate with this disaccharide-polyacrylamide glycoconjugate. MOA readily agglutinated Ehrlich ascites tumor cells, which contain the same epitopic end groups in their cell membranes (26).
Fluorescein-labeled MOA stained porcine striated skeletal muscle (Fig. 6), endothelial cells lining the capillaries being the significant structures stained by the lectin. Incubation of the staining solution in the presence of 20 mM linear B6 trisaccharide essentially abolished staining (data not shown), indicating specific binding of the fluorescein-labeled lectin to the porcine tissue. DISCUSSION Results presented in this paper indicate that the M. oreades lectin, as isolated previously and in the initial stages of this work, was a dimer composed of an intact and a truncated subunit in addition to a small polypeptide presumably generated by proteolytic cleavage of the intact H subunit. Both the intact H and truncated L subunits are blocked at their N termini. Subsequently, we isolated an intact form of the lectin and an intact recombinant lectin (15), which have very similar binding properties, indicating that cleavage is due to the presence of contaminating metalloprotease activity and is not relevant to protein function.
The most interesting aspect of this study concerns the high specificity of the lectin for Gal␣1,3Gal end groups; the ␣1,2-, ␣1,4-, and ␣1,6-linked disaccharides do not bind to the lectin. These data indicate that MOA has an extended binding site that accommodates the Gal␣1,3Gal disaccharide. To the best of our knowledge, this is the first lectin shown to exhibit this exclusive specificity. Additional contributions to the binding energy are made by the addition of ␤1,4GlcNAc and ␣1,2-Lfucosyl groups to the reducing Gal unit of Gal␣1,3Gal. These additions account for the blood group B activity of the lectin. The structures of these oligosaccharides, related to human blood group B substance and assayed for their activity in this study, are presented in Fig. 7.
It is instructive to compare MOA with the Griffonia simplicifolia I-B 4 isolectin (GS I-B 4 ), a highly specific ␣-D-galactosylbinding lectin, with respect to the size of their combining sites. A great deal of binding data indicate that the GS I-B 4 isolectin has a very restricted binding site that does not appear to extend significantly beyond the monosaccharide unit (i.e. the ␣-D-galactosyl group) (27,28). This concept of a restricted site was substantially verified by the recently completed x-ray crystallographic structure of the B 4 isolectin complexed with Gal␣1,3Gal in which it is shown that only the nonreducing ␣-galactosyl group makes contact with the lectin (29). Moreover, Kirkeby and Moe (30), using an enzyme-linked lectin assay, showed that the ␣-galactosyl group and the ␣1,2-, ␣1,3-, and ␣1,4-galactobiosyl groups as well as the linear B-trisaccharide (Gal␣1,3Gal␤1,4GlcNAc) linked to human serum albumin were very similar in their binding affinity to GS I-B 4 . On the contrary, MOA, which binds only very weakly or not at all to methyl ␣-D-galactopyranoside or to the ␣1,2-, ␣1,4-, and ␣1,6linked galactobioses, has an extended binding site, which accommodates the blood type B disaccharide and linear trisaccharide (Gal␣1,3Gal and Gal␣1,3Gal␤1,4GlcNAc, respectively) and the branched chain blood group B trisaccharide determinant with high affinity.
Taking advantage of these carbohydrate binding properties of MOA, we demonstrated that it gave a very strong precipitation reaction with laminin from the Engelbreth-Holm-Swarm murine sarcoma. This highly glycosylated glycoprotein contains the linear type B blood group trisaccharide at many of its chain ends (22,23). It also recognizes and precipitates bovine thyroglobulin, which also contains the linear B trisaccharide (24). Interestingly, this lectin will not precipitate the galactomannan from Cassia alata that contains multiple ␣-galactosyl end groups, illustrating its lack of reactivity with single such sugar residues (Fig. 5). Nor did MOA bind to cross-linked guaran, also a galactomannan. Conversely, both the galactomannan (not shown) and pigeon ovalbumin precipitated strongly with GS I isolectins (Fig. 5). MOA does not recognize the blood group type A disaccharide (GalNAc␣1,3Gal), as shown by its failure to precipitate with this disaccharide-polyacrylamide glycoconjugate, whereas the GS I-A 4 isolectin, which is specific for ␣GalNAc end groups (27), reacted strongly with it (Fig. 5).
MOA should be a valuable reagent for the glycobiologist for the detection and preliminary characterization of glycoconjugates containing Gal␣1,3Gal disaccharide and Gal␣1,3Gal␤1,4GlcNAc trisaccharide on cell surfaces and in solution, since these epitopes occur in many biologic sources. They have been found, for example, in the basement membranes of mice, rats, and rabbits (31); the surface of 3T3 cells (32); calf thyroid plasma membranes (33); and the plasma membranes of Ehrlich ascites tumor cells (26,34). It also has been known for many years that porcine tissues and organs contain the Gal␣1,3Gal␤1,4GlcNAc epitope, the socalled Galili trisaccharide (35), which prevents their use for transplantation into humans. Indeed, fluorescein-labeled MOA does stain porcine striated skeletal muscle as shown in Fig. 6, in which endothelial cells lining the capillaries bind the lectin. This epitope is present in most cells of nonprimate mammals and New World monkeys but not in humans, apes, or Old World monkeys. It is also possible that this carbohydrate antigen is present on the surface of the malaria parasite, Plasmodium falciparum (36). Another interesting application of this lectin could be its use in the possible competition with toxin A elaborated by Clostridium difficile, which is responsible for antibiotic-induced diarrhea (37). Both MOA and toxin A recognize the Gal␣1,3Gal␤1,4GlcNAc trisaccharide epitope. FIG. 5. Quantitative precipitation assay of MOA with glycoconjugates. Varying amounts of glycoconjugates, ranging from 0 to 100 g were incubated with 20 g of lectin in a total volume of 150 l of PBS (pH 7.2). After 48 h at 4°C, the amounts of protein precipitated were quantified. q, laminin; ϩ, bovine thyroglobulin; f, C. alata galactomannan; OE, pigeon ovalbumin; E, GalNAc␣1,3Gal␤Ϫpolyacrylamide; Ⅺ, GS I interaction with pigeon ovalbumin; , GS I interaction with GalNAc␣1,3Gal␤-polyacrylamide.