Characterization of the Structural Determinants Required for the High Affinity Interaction of Asparagine-linked Oligosaccharides with Immobilized Phaseolus vulgaris Leukoagglutinating and Erythroagglutinating Lectins*

The carbohydrate binding specificities of the leu- koagglutinating phytohemagglutinin (L-PHA) and erythroagglutinating phytohemagglutinin (E-PHA) lectins of the red kidney bean, Phaseolus vulgaris, have been investigated by lectin-agarose affinity chromatography of Asn-linked oligosaccharides. High affinity binding to E-PHA-agarose occurs only with biantennary glycopeptides containing 2 outer galactose resi- dues and a residue of N-acetylglucosamine linked p1,4 to the /3-linked mannose residue in the core. This species is not retarded on L-PHA-agarose. In contrast, tri- and tetraantennary glycopeptides containing outer galactose residues and an a-linked mannose residue sub- stituted at positions C-2 and C-6 are specifically re- tarded on L-PHA-agarose. Triantennary glycopeptides containing outer galactose residues and an a-linked mannose residue substituted at positions C-2 and C-4 are not retarded on L-PHA-agarose. Additionally, the presence of outer sialic acid residues or a core fucose residue does not influence the behavior

glutinating; L-PHA, leukoagglutinating phytohemagglutinin; E-PHA, erythroagglutinating phytohemagglutinin; ConA-Sepharose, concanavalin A-Sepharose; PHA-P, phytohemagglutinin P. * For convenience, the lectin species with one or more E subunits have usually been termed E-PHA, whereas the L4 species is termed L-PHA. However, in this paper, L-PHA-agarose refers to the 4 species coupled to agarose and E-PHA-agarose refers to the E4 species coupled to agarose. toward these cells (4). These observations demonstrate that the carbohydrate binding specificities of the E4 and lectins are different even though the lectins are similar in molecular weight and amino acid composition (1, 3, 5 ) . Several studies have been published on the carbohydrate binding specificity of the erythroagglutinating forms of the lectin (6-8). We previously reported that the binding site for these lectins on human erythrocytes was an Asn-linked complex-type oligosaccharide and that the galactose residues of this oligosaccharide were important determinants of binding (6). More recently, Irimura et al. (7), using affinity chromatography on E-PHA-agarose, determined that the biantennary complex-type Asn-linked oligosaccharide of human erythrocyte glycophorin interacts with high affinity with the lectin. These workers also confiied that galactose residues are required for this interaction. However, Irimura et al. (7) did not investigate the possibility that immobilized E-PHA might interact with other structurally related glycopeptides.
In contrast, very little is known ahout the carbohyrkate binding specificity of L-PHA. Several investigators have found that high concentrations of N-acetylgalactosamine (>25 mM) can dissociate L-PHA, as well as E-PHA, bound to cells or glycoproteins (6, 9, 10). However, the relationship of this phenomenon to the carbohydrate binding specificity of L-PHA is obscure. For example, L-PHA does not bind significantly to blood group A erythrocytes (1) which contain oligosaccharides with terminal N-acetylgalactosamine residues, but does bind to thyroglobulin (9-11) which lacks N-acetylgalactosamine residues (12).
Since L-PHA is a widely used lectin, we felt that it would be useful to better characterize its carbohydrate binding specificity and to compare this to the binding specificity of E-PHA. To accomplish this, we have immobilized L-PHA and E-PHA on agarose supports and have studied the interaction of a panel of glycopeptides of known structure with these immobilized lectins. Our results indicate that L-PHA and E-PHA interact with different Asn-linked oligosaccharides and that the difference is related to the substitution pattern of the mannose residues of the oligosaccharides.

EXPERIMENTAL PROCEDURES
Materials-Concanavalin A-Sepharose was obtained from Pharmacia Fine Chemicals and. PHA-P was purchased form P-L Biochemicals. Affi Gel 10 was obtained from Bio-Rad. ['HIAcetic anhydride (100 mCi/mmol) was purchased from New England Nuclear. Sulfopropyl-Sephadex, Sephadex G-150, Sephadex G-25, bovine epididymal a-L-fucosidase, and bovine thyroglobulin were obtained from Sigma. Vibrio cholera neuraminidase and pronase (B grade) were obtained from Calbiochem-Behrlng. P-Galactosidase, fl-hr-acetylglucosaminidase, and a-mannosidase were prepared from jack bean meal by the method of Li and Li (13).
Sources of Test Glycopeptides-Desialized fibrinogen glycopeptide was obtained from a preparation of the glycoprotein supplied by Drs. J. Miletich and G. Broze of the Washington University School of Medicine. The structure of this glycopeptide is shown in Table I as   IgA-11-C glycopeptide (Table I, In containing 2 terminal NeuAc residues was prepared (14) and labeled with r3H]acetic anhydride (15). IgA-11-A glycopeptide was also prepared (14) and labeled with r3H]acetic anhydride (15). This glycopeptide has the structure of glycopeptide III in Table I. Mouse IgM glycopeptide I-A (labeled with [2-'H]mannose) and fetuin glycopeptide were generously provided by Dr. Rosalind Kornfeld and Dr. Jacques Baenziger, respectively, of the Washington University School of Medicine. These glycopeptides have related structures (16-18), but differ in the number and linkage of sialic acid residues. The structure of the fetuin glycopeptide is shown in Table I as V. A glycopeptide with the structure of VI (Table I) was prepared from the mouse lymphoma cell line (BW5147), labeled in [2-3H]mannose, as described previously (19).
Glycopeptides with the Structures of IZ, IV, VZZ, and VZZZ in Table  1 were prepared from bovine thyroglobulin by the following technique. Crude bovine thyroglobulin (1 g) was dissolved in 5 ml of 0.1 M Tris-HCl, 2 mM CaC12, pH 8.0, and digested with pronase (40 mg) for 24 h in a toluene atmosphere at 60 "C. Additional pronase (5 mg) was then added and the incubation continued for 6 h. The digest was boiled for 5 min and desalted on a column of Sephadex G-25 (1 X 50 cm) in 7% 1-propanol in H20. The void fractions were pooled and the solvent evaporated. The residue was dissolved in 0.1 M Tris-HC1,l m~ CaC12, 1 mM MgCI2, 0.02% NaN3, pH 8.0, and one-tenth of the sample was applied to a column of ConA-Sepharose (IO-ml volume). The column was eluted sequentially with 10 mM a-methylglucoside and 100 mM a-methylmannoside as described previously (19). This procedure was repeated until all the material had been fractionated by ConA-Sepharose chromatography. and dissolved in 1.0 ml of PBS/NaN3 (6.7 mM KHzP04, 0.15 M NaCI, The glycopeptides not bound by ConA-Sepharose were desalted 0.02% NaN3, pH 7.4) and applied to the column of L-PHA-agarose (see below). The column was washed with PBS/NaN3 and the unbound glycopeptides were recovered, as well as the retarded (bound) glycopeptides in the trailing edge. The unbound glycopeptides were desalted by passage over Sephadex G-25 and reapplied to the L-PHAagarose. These steps were repeated (approximately 15 times) until no more glycopeptides were retarded in their elution from the L-PHAagarose column. From these data, the capacity of the L-PHA-agarose column was found to be approximately 10 nmol of glycopeptide/ml of gel. The glycopeptides not bound by Cod-Sepharose or L-PHA-agarose are designated 1-1, whereas the glycopeptides not bound by Cod-Sepharose but bound by L-PHA-agarose are designated 1-2. Structural analyses on fractions 1-1 and 1-2 were done in collaboration with Dr. L. Dorland and Dr. J. F. G. Vliegenthart (University of Utrecht) using a 500-MHz 'H-NMR spectrometer. Fraction 1-1 was found to be a mixture of triantennary glycopeptides and some biantennary glycopeptides (resulting from overloading of the ConA-Sepharose columns). These glycopeptides have the structures of glycopeptides ZI and IV, respectively, in Table I. The 1-2 fraction was found to be a mixture of triantennary glycopeptides and tetraantennary glycopeptides. These compounds have the structures of glycopeptides VI1 and VIII, respectively, in Table I. In addition, glycopeptides IZ, IV, VlI, and VZII (Table I), from bovine thyroglobulin, contain some outer galactose residues linked a1,3 to a penultimate galactose as indicated. Further details concerning this finding will be published later by Dr. Dorland and Dr. Vliegenthart.
Purification of L-PHA a n d E-PHA-L-PHA was purified by the procedure of Weber (I), using PHA-P as the starting material. The yield of L-PHA was 11 mg/250 mg of PHA-P. The purified L-PHA was devoid of hemagglutinating activity in the assay described (20) at a concentration of 1 mg/ml. The lectin was stored at -10 "C in 0.05 M KHzPO4, pH 6.0. Approximately 50 p g of L-PHA was electrophoet al. (II), which separates L-PHA and E-PHA subunits. A single resed in a nondenaturing polyacrylamide gel, as described by Felsted band of stained protein was observed.
E-PHA was purified from PHA-P following the combined procedures of Weber (1) and Leavitt et al. (3). Following elution of the hemagglutinating activity from SP-Sephadex ( l ) , the E-PHA preparation was reapplied to SP-Sephadex. In this case, the E-PHA was eluted with a linear gradient of 0-0.188 M NaCI, as described by Leavitt et al. ( 3 ) . The last peak of eluted material, having the highest '' R. Kornfeld, unpublished data. specific activity of hemagglutination, has been shown by Leavitt et al. (3) to be pure E,-PHA, devoid of L-PHA subunits. This preparation of E-PHA (11 mg) had a hemagglutinating activity of 1130 units/ mg and was stored under conditions similar to the L-PHA as described above.
Protein was determined by the method of Lowry et al. (21) using bovine serum albumin as a standard.
Coupling of L-PHA and E-PHA to Affi Gel 10-L-PHA (19 mg) was dissolved in a final volume of 6 ml of buffer containing 0.1 M NaHC03, 0.2 M N-acetylgalactosamine, pH 8.0. A slurry of Affl Gel 10 (10 ml) was washed as described by the manufacturer, and the moist cake was added quickly to the solution of L-PHA. The solution was mixed gently for 18 h at 4 "C and the gel was then allowed to settle. Uncoupled L-PHA was recovered in the supernatant, and the gel was resuspended in 0.1 M NaHC03, pH 8.0, containing 0.1 M ethanolamine to block any remaining coupling sites. After 2 h a t 4 "C, the L-PHAagarose was placed in a column (0.5 X 30 cm) at room temperature and washed with PBS/NaN,?. Coupling efficiency was estimated by determining the amount of protein remaining in the supernatant after the coupling reaction. L-PHA was coupled to Affi Gel 10 with approximately 60% efficiency, and the amount of coupled protein was estimated to be 1.8 mg/ml of gel.
E-PHA was coupled to Affi Gel 10 using the procedure described above, except that 7 mg of E-PHA was used in the coupling reaction. Coupling efficiency was estimated to be 50%, and the amount of coupled lectin was approximately 0.6 m g / d of gel. T h e E-PHAagarose was placed in a column (0.5 X 30 cm) and washed at room temperature with PBS/NaN3.
Both columns were routinely stored in PBS/NaNS at 4 "C and were allowed to stand at room temperature approximately 1 h before use.
We have found recently that commercial preparations of these lectins coupled to agarose (E-Y Laboratories, Inc.) give similar results in the fractionation of Asn-linked oligosaccharides as our own preparations.
Lectin Affinity Chromatography-Glycopeptides to be chromatographed on E-PHAor L-PHA-agarose were dissolved in 0.5 ml of PBS/NaN3 and were applied to the columns a t room temperature. Fractions (1 m l ) were collected at 10 d / h , and aliquots were removed for either counting radioactivity or determining total neutral hexose. Neutral hexose was determined by phenol-sulfuric acid assay (22).
For liquid scintillation counting, 0.4 ml of sample was mixed with 4.0 ml of 3a70 scintillation mixture (Research Products International Corp.) and counted in a Beckman LS-7000 counter. Recovery of glycopeptides was routinely 90-100%. The columns are extremely stable and have been used repeatedly for over a year without detectable changes in behavior.
GZycosidase Treatments-Glycosidase treatments of glycopeptides were done as described by Kornfeld et al. (19). Table  I lists the various Asn-linked oligosaccharides which were tested for their ability to interact with L-PHA-agarose. An example of the behavior of three of these glycopeptides on the L-PHA-agarose column is shown in Fig. 1. Glycopeptide IV is not retarded on the column ( A ) and is therefore scored as noninteracting (Table I) Table I. Of the glycopeptides listed in Table I, only    residues of the Asn-linked oligosaccharides in the interaction with L-PHA-agarose was determined by treating the glycopeptides with various glycosidases and reapplying the partially digested samples to the lectin-agarose column. Treatment of glycopeptide VI with neuraminidase has no effect on the interaction of the compound with L-PHA-agarose. However, the subsequent removal of the galactose residues with Pgalactosidase abolishes the high affinity interaction of glycopeptide VI with L-PHA-agarose (data not shown). These results indicate that a second determinant necessary for the interaction of Asn-linked oligosaccharides with L-PHA is the presence of outer galactose residues. We have been unable to obtain homogeneous preparations of glycopeptide VI containing only 1 or 2 galactose residues, and, therefore, the positional effects, if any, and the minimal number of galactose residues required for the interaction of glycopeptides with L-PHA are unknown.

Structures ofglycopeptides tested for their ability to interact with either L-PHA-agarose or E-PHA-agarose
Removal of the internal fucose residue of glycopeptide VI by treatment with a-L-fucosidase does not affect the behavior of the glycopeptide on the L-PHA column. However, this treatment did abolish binding of glycopeptide VI to pea lectin-Sepharose, confirming studies by Kornfeld et al. (19) which demonstrated that core fucose is a requirement for binding of glycopeptides to this lectin.
Interaction of Glycopeptides with E-PHA-agarose- Fig. 2 shows an example of the chromatography of two glycopeptides on E-PHA-agarose. As shown in A , glycopeptide 1 1 is not retarded, whereas glycopeptide III ( B ) is significantly retarded on E-PHA-agarose. The results of the chromatography of the other Asn-linked oligosaccharides are indicated in Table   I. The only glycopeptide that is retarded on the column is glycopeptide III. Since this is the only glycopeptide with a residue of N-acetylglucosamine linked P1,4 to the P-linked mannose, we conclude that this residue is an important determinant for high affinity binding to E-PHA-agarose. This is best illustrated by comparing the structures of glycopeptide I (which fails to interact with E-PHA) and glycopeptide III.
The only difference between these two glycopeptides is the presence of the "bisecting" N-acetylglucosamine residue on glycopeptide III, as well as the presence of a residue of Nacetylneuraminic acid on the glycopeptide. Since removal of the N-acetylneuraminic acid from glycopeptide III has no effect on the interaction with E-PHA-agarose (see below), the difference in the behavior of the two glycopeptides must be due to the bisecting N-acetylglucosamine residue.
Effect of Outer Chain Sugar Residues on Glycopeptide Binding to E-PHA-agarose-Removal of the N-acetylneuraminic acid from glycopeptide III with neuraminidase has no effect on the interaction of the glycopeptide with E-PHA-agarose. However, the removal of l galactose residue from the glycopeptide by treatment with P-galactosidase abolishes the high affinity interaction with E-PHA-agarose (data not shown). These results indicate that the 2 outer galactose residues of the glycopeptide constitute a second determinant in the lectin-oligosaccharide interaction.

DISCUSSION
The data presented in this paper demonstrate that E-PHA and L-PHA have similar but distinct carbohydrate binding specificities. Both lectins are capable of high affinity interactions with Asn-linked complex-type oligosaccharides and, in both instances, the galactose residues of these oligosaccharides are important in this interaction. On the other hand, there are significant differences in the binding specificities of these lectins. E-PHA interacts with highest affinity only with galactosylated glycopeptides containing an N-acetylglucosamine residue linked to the P-linked mannose residue of the core. Only biantennary glycopeptides with this type of substitution were available in this study. It is possible that tri-or tetraantennary glycopeptides with this substitution pattern can also interact strongly with E-PHA-agarose. Irimura et al. (7) and Yoshima et al. (24) have shown that the Asn-linked oligosaccharide on glycophorin is a biantennary oligosaccharide with a bisecting N-acetylglucosamine residue. Thus, the data available indicate that E-PHA binds to glycophorin molecules on human erythrocytes (6,23) and that this binding results from the high affinity interaction of E-PHA with the bisected biantennary oligosaccharide of glycophorin.
L-PHA, which does not bind significantly to human erythrocytes, fails to interact with the bisected biantennary glycopeptide with high affinity. L-PHA interacts best with galactosylated triantennary and tetraantennary glycopeptides that have at least 1 of the a-linked mannose residues substituted at positions C-2 and C-6 with P-linked N-acetylglucosamine residues. The other a-linked mannose may be singly substituted at C-2 (as in the triantennary species) or doubly substituted at C-2 and C-4 (as in the tetraantennary species). Neither of these tri-or tetraantennary glycopeptides interacts with high affinity with E-PHA.
The carbohydrate binding specificity of L-PHA also has some similarities to that reported for the pea and lentil lectins (19). These latter lectins bind triantennary glycopeptides that have an a-linked mannose residue substituted a t C-2 and C-6, but fail to bind triantennary species that have a-linked mannose residues substituted a t C-2 and C-4. However, they also require the presence of an a-linked fucose residue in the core, whereas L-PHA binding is not influenced by this fucose residue. The pea and lentil lectins also differ from L-PHA in that they fail to bind tetraantennary glycopeptides (such as glycopeptide VI11 of Table I), but do bind biantennary glycopeptides that contain the inner fucose residue (glycopeptide II of Table I).
In this study, we have not tested glycopeptides with 0linked oligosaccharides for their ability to interact with L-PHA-or E-PHA-agarose. It is unlikely, however, that either of these lectins interact strongly with 0-linked oligosaccharides. Previous studies have shown that only Asn-linked oligosaccharides interact strongly with E -PHA (6, 23). The finding that L-PHA fails to bind to human erythrocytes, even after neuraminidase treatment, suggests that 0-linked oligosaccharides do not serve as binding sites for this lectin since human erythrocytes contain high amounts of 0-linked oligosaccharides.
It is important to note that our study has utilized E4or L4-PHA coupled to agarose, and we have not investigated the specificity of the tetrameric lectins L1E3-, L2E,-, and L3E1-PHA coupled to agarose. It is possible that these latter lectins may have unusual differences from either L4-PHA-or E4-PHA-agarose in their binding requirements or affinities for glycopeptides.
Previous reports have demonstrated the value of lectin affinity chromatography for the fractionation of glycopeptides (12, 19, 25-29). Kornfeld et al. (19) most recently demonstrated the usefulness of serial lectin affinity chromatography for the fractionation of cell membrane Asn-linked oligosaccharides. A comparison of our findings with L-PHA-and E-PHA-agarose with those previously reported with Concanavalin A-, pea lectin-, and lentil lectin-Sepharose indicate that the use of all these lectins in the correct sequence can provide an extremely simple, rapid, and sensitive technique for the fractionation of Asn-linked oligosaccharides. In the following paper, we present a scheme for serial lectin affinity chromatography and illustrate the usefulness of this approach for the separation of Asn-linked oligosaccharides of cultured cell lines (30).
Note Added in Proof-A recent report (Hammarstrom, S., Hammarstrom, M. L., Sundblad, G., Arnarp, J., and Lonngren, J. (1982) Proc. Natl. Acad. Scc. U. S. A. 79, 1611-1615) described a study of the specificity of L-PHA using the techniques of quantitative precipitation and precipitation-inhibition by a number of different glycopeptides and synthetic oligosaccharides. They concluded that the most complementary structure to the binding site of L-PHA was a pentasaccharide composed of two N-acetyllactosamine disaccharides linked through the GlcNAc residues to a mannose residue at both positions C-2 and C-6. This finding is consistent with our data indicating that glycopeptides with this structural feature are selectively retarded in their elution from a column of L-PHA-agarose.