Lectin affinity high-performance liquid chromatography. Interactions of N-glycanase-released oligosaccharides with Ricinus communis agglutinin I and Ricinus communis agglutinin II.

The structural determinants required for interaction of oligosaccharides with Ricinus communis agglutinin I (RCAI) and Ricinus communis agglutinin II (RCAII) have been studied by lectin affinity high-performance liquid chromatography (HPLC). Homogeneous oligosaccharides of known structure, purified following release from Asn with N-glycanase and reduction with NaBH4, were tested for their ability to interact with columns of silica-bound RCAI and RCAII. The characteristic elution position obtained for each oligosaccharide was reproducible and correlated with specific structural features. RCAI binds oligosaccharides bearing terminal beta 1,4-linked Gal but not those containing terminal beta 1,4-linked GalNAc. In contrast, RCAII binds structures with either terminal beta 1,4-linked Gal or beta 1,4-linked GalNAc. Both lectins display a greater affinity for structures with terminal beta 1,4-rather than beta 1,3-linked Gal, although RCAII interacts more strongly than RCAI with oligosaccharides containing terminal beta 1,3-linked Gal. Whereas terminal alpha 2,6-linked sialic acid partially inhibits oligosaccharide-RCAI interaction, terminal alpha 2,3-linked sialic acid abolishes interaction with the lectin. In contrast, alpha 2,3- and alpha 2,6-linked sialic acid equally inhibit but do not abolish oligosaccharide interaction with RCAII. RCAI and RCAII discriminate between N-acetyllactosamine-type branches arising from different core Man residues of dibranched complex-type oligosaccharides; RCAI has a preference for the branch attached to the alpha 1,3-linked core Man and RCAII has a preference for the branch attached to the alpha 1,6-linked core Man. RCAII but not RCAI interacts with certain di- and tribranched oligosaccharides devoid of either Gal or GalNAc but bearing terminal GlcNAc, indicating an important role for GlcNAc in RCAII interaction. These findings suggest that N-acetyllactosamine is the primary feature required for oligosaccharide recognition by both RCAI and RCAII but that lectin interaction is strongly modulated by other structural features. Thus, the oligosaccharide specificities of RCAI and RCAII are distinct, depending on many different structural features including terminal sugar moieties, peripheral branching pattern, and sugar linkages.

The common castor bean, Ricinus communis, contains two lectins, Ricinus communis agglutinin I (RCAI)' and Ricinus communis agglutinin I1 (RCAII). RCA1 (RCA-120) is a tetrameric hemagglutinin with a molecular weight of 120,000, consisting of two a (Mr = 29,500) and two p (Mr = 37,000) subunits. RCAII (RCA-60) is a highly toxic, dimeric protein with a molecular weight of 60,000, consisting of a single a (Mr = 29,500) and a single @ (Mr = 34,000) subunit. For both lectins, only the @ subunits display carbohydrate binding activity. Thus RCAII, which is essentially monovalent with only one / 3 subunit, is not able to agglutinate cells, whereas RCAI, which is divalent with two @ subunits, is a true agglutinin. The constituent polypeptide subunits of RCAI and RCAII have closely related antigenic features and amino acid sequences (for review see Refs. [1][2][3][4]. In an earlier study, we systematically examined the affinities of RCAl and RCAII for glycopeptides bearing either Asnlinked or 0-glycosidically linked oligosaccharides (3). RCAl and RCAII were found to each display a high degree of specificity for oligosaccharide structural features. Subsequent studies from a number of laboratories using various techniques have further established the specificities of these two lectins (2,(4)(5)(6). We have recently developed a technique of lectin affinity high-performance liquid chromatography (HPLC) for analyzing and separating oligosaccharides and have established the utility of such a method for defining lectin specificity (7). Utilizing homogeneous N-glycanase-released, reduced oligosaccharides of known structure, we have examined the oligosaccharide specificities of RCAI and RCAIl by lectin affinity HPLC. These studies have provided significant new insights about the specificity of RCAI and RCAII. The elution position of individual oligosaccharides during RCAl and RCAll affinity HPLC is reproducible and highly characteristic of specific structural features. Therefore, lectin affinity HPLC using RCAl and RCAII is a useful adjunct for the fractionation and characterization of oligosaccharide structures.

RESULTS
Purified N-glycanase-released, reduced oligosaccharides with the structures shown in Table  I and products derived from these oligosaccharides were examined by lectin affinity HPLC on silica-bound RCAI and RCAII. Results of such analyses are shown in Figs. 1-8, along with schematically illustrated structures of the oligosaccharides examined. Although reproducible elution times characteristic of specific oligosaccharide structures were obtained with both RCAI-and RCAII-silica columns, we have presented the data in terms of elution within discrete regions (I-V for RCAr and I-IV for RCArl). In each case, region I corresponds to the column void, i.e. complete lack of interaction with the immobilized lectin. Polysaccharides which are not recognized by either RCAl or RCAIr and are structurally unrelated to Asn-linked oligosaccharides but of similar size, for example (Gl~NAcp1,4)~, elute in region I. Oligosaccharides which interact with RCAI-or RCArr-silica are retarded to varying extents by the lectins and in some cases require highly stringent conditions for elution.
Analysis of Oligosaccharides by RCAI Affinity HPLC-011gosaccharides elute within five major regions (I-V) during RCA, affinity HPLC. As noted above, region I corresponds to the column void, while regions 11-IV reflect a progressive increase in interaction with RCAI that results in retardation but not binding of oligosaccharides. Structures bound by RCAI-silica remain tightly associated with the lectin without evidence of elution for >60 min of chromatography in PBS/ NaN,. Bound oligosaccharides were eluted at 45 min (region V) with 200 mM lactose. Neither high concentrations of other sugars, including Gal, GlcNAc, and GalNAc, nor acidic pH (1.0 N acetic acid) were effective a t eluting bound oligosaccharides.
Removal of sialic acid from dibranched complex-type oligosaccharides, which exposes the underlying B1,Clinked Gal residues, results in binding to RCAI. Thus, asialo Di-1 (data not shown) and asialo Di-2 ( Fig. 1C) bind to the lectin and   (Fig. 1D). Thus, the presence of terminal @1,4-linked Gal promotes binding of oligosaccharides to RCAI. The presence of a2,6-linked sialic acid markedly reduces oligosaccharide-RCAI interaction, resulting in retardation but not binding of dibranched oligosaccharides to the lectin, whereas the presence of a2,3-linked sialic acid or absence of @1,4-linked Gal abolishes interaction of dibranched structures with RCAI. The core S t I " e , consisting of Man3GlcNAc2, also does not interact with RCA1. Other oligosaccharides devoid of Gal and structurally distinct from complex-type oligosaccharides, for example high mannosetype structures (HM-I), also do not interact with RCAI (Fig.  1H).
Di-1 and Di-2 oligosaccharides differ both in the linkage of their sialic acid residues (a2,3 wrsus a2,6) and in the presence of Fuc on a large fraction of structures constituting Di-2 but not Di-1. The presence of Fuc attached to the reduced core GlcNAc does not appear to affect interaction of Di-2 with RCAI, since 1) roughly half of the Di-2 oligosaccharides contain Fuc, yet analysis of Di-2 or structures derived from Di-2 by RCAI affinity HPLC consistently yields a single uniform peak, 2) desialylated forms of Di-1 and Di-2 coelute during RCAl affinity HPLC, yet only asialo Di-2 contains Fuc, and 3) removal of Fuc by a-fucosidase does not alter the interaction of Di-2 or structures derived from Di-2 with RCAI.
Dibranched complex-type oligosaccharides bearing a single terminal B1,Chnked Gal differ in their interaction with RCAI, depending upon the location of the branch bearing the peripheral Gal. Oligosaccharides with a single terminal @1,4linked Gal on the branch arising from the al,a-linked core Man (asialo Di-3) elute in region I V (Fig. 1E). In contrast, structures with a single terminal @l,Clinked Gal on the branch arising from the a1,G-linked core Man (Di-4) elute in region I11 (Fig. 1F). Removal of Gal from either asialo Di-3 or Di-4 abolishes interaction with RCA,, resulting in elution of the agalacto structures in region I. Thus, the affinity of RCAI for oligosaccharides bearing a single terminal @1,4linked Gal is greater when this moiety is present on the branch attached to the a1,3as compared to the al,6-linked core Man. Of note, dibranched oligosaccharides containing a single terminal B1,Clinked Gal (asialo Di-3 and Di-4) are retarded but not bound by RCAI-silica, whereas dibranched structures bearing 2 terminal @1,4-linked Gal residues (asialo Di-1 and asialo Di-2) are tightly bound by the lectin.
Dibranched complex-type oligosaccharides containing @1,4linked GalNAc in place of @1,4-linked Gal do not significantly interact with RCAI. Di-6, which bears two branches terminating in @l,4-linked S04-GdNAc, elutes in region I during RCAI affmity HPLC (data not shown). Desulfated Di-6, which is identical to asialo Di-2 aside from the presence of terminal &1,4-linked GalNAc in place of @1,4-Iinked Gal, is slightly retarded by RCAr (Fig. 1G). The latter result indicates that the presence of an N-acetyl-rather than an HO moiety at the C-2 position of terminal ,81,4-linked Gal results in a profound decrease in interaction with RCAI.  Table I. The interaction of RCAI with dibranched complex-type oligosaccharides also depends upon the linkage of terminai Gal moieties. Di-8 is identical to asialo Di-2 except for the presence of a terminal 81,3-rather than a 81,4-linked Gal on the branch arising from the alP-linked core Man. Like asialo Di-2, Di-8 binds to RCAI and is eluted in region V with 200 mM lactose ( Fig. 2A). Removal of the terminal B1,4-linked Gal (on the b m c h arising from the al,6-linked core Man) completely abolishes interaction of Di-8 with RCAI (Fig. 2B). Thus, in contrast to the presence of a single terminal @1,4linked Gal residue (Fig. lE), the presence of a single terminal B1,3-linked Gal residue on the branch arising from the a1,3linked core Man does not &t in detectable interaction with RCAI. Di-9, which bears two terminal P1,3-linked Gal residues, is slightly retarded during RCAI affinty HPLC, eluting at the beginning of region I1 (Fig. 2C). h i d o Di-2, desulfated Di-6, and Di-9 share identical oligosaccharide &ructurea except for the terminal moieties on the peripheral branches, bearing &1,4-linked Gal, @1,4-linked GalNAc, and Pl,&linked Gal, respectively. RCAI discriminates between these closely related oligosaccharides, binding the .Erst structure and differentially retarding the second and third. The interaction with RCAr appears to be slightly greater for terminal m,3linked Gal than for terminal @1,4-linked GalNAc moieties.
The presence of a bisecting GlcNAc residue attached to the @-linked core Man does not appear to alter the behavior of *E. D. Green, and J. U. Baenziger, submitted for publication.

RCAI and RCA, Affinity HPLC
dibranched complex-type oligosaccharides during RCAI affinity HPLC. Di-7, which contains two a2,6-linked sialic acid residues, elutes in region 11, whereas asialo Di-7 binds to RCAI and is eluted in region V with 200 mM lactose (data not shown). Agalacto Di-7 elutes in region I. Thus, the analogous forms of Di-1 and Di-7 display the same interactions with RCAI, despite the presence of a bisecting GlcNAc on Di-7 but not Di-1. Tetra-1 and Tri-1 each consist of a heterogeneous mixture of oligosaccharides which contain the same number of sialic acid residues (4 and 3, respectively) and share the identical underlying carbohydrate structure (Table I). Tetra-1 and Tri-1 oligosaccharides differ in the linkage (a2,3 versus a2,6) of sialic acid residues and in the distribution of a2,3-and a2,6linked sialic acid moieties among the peripheral branches (8). Both Tetra-1 (Fig. 3A) and Tri-1 (Fig. 3B) separate into two peaks during RCAl affinity HPLC. In each case, a major oligosaccharide fraction binds to the lectin and is eluted in region V, while the remainder elutes in region I or between regions I and 11. These fractionations reflect the linkage and/ or distribution of differently linked sialic acid moieties on Tetra-1 and Tri-1 oligosaccharides, since 1) analysis of each isolated Tetra-1 and Tri-1 fraction by differential digestion with Newcastle disease virus versus C. perfringens neuraminidase (as described in Table 2 of Ref. 8) reveals that oligosaccharides eluting in region I or between regions I and I1 bear relatively more a2,3-and less a2,6-linked sialic acid, while structures eluting in region V contain relatively less a2,3-and  Fig. 1, with detailed structures shown in Table I. more a2,6-linked sialic acid. For example, roughly 53% of the sialic acid residues on the Tri-1 oligosaccharides in the unbound peak were released by Newcastle disease virus neuraminidase, whereas only 20% of the sialic acid residues on structures in the bound peak (region V) were released by this enzyme (8). Sialic acid residues were quantitatively released from oligosaccharides in both peaks by digestion with C. perfringens neuraminidase. Similar results were obtained by analysis of the earlier versus later eluting Tetra-1 oligosaccharides. Therefore, Tri-1 and Tetra-1 oligosaccharides bound by RCAI contain relatively more a2,6-linked sialic acid (Newcastle disease virus neuraminidase resistant) than those structures not bound by RCAI; and 2) removal of sialic acid (Fig.  3C) or both sialic acid and Gal (Fig. 3 0 ) from Tetra-1 and Tri-1 results in elution of the oligosaccharides as single, uniform peaks. Asialo Tetra-1 and asialo Tri-1 bind tightly to RCAI (Fig. 3C), whereas agalacto Tetra-1 and agalacto Tri-1 elute unretarded in region I (Fig. 30).
The sialylated forms of Tetra-1 and Tri-1 oligosaccharides not bound by RCA, differ slightly in elution position (region I versus between regions I and 11). This may reflect differences in linkage of sialic acid residues, distribution of distinctly linked sialic acid moieties among the peripheral branches, number of peripheral branches (four versus three), and/or the presence of a branch arising from the C-6 position of the a1,6linked core Man on Tetra-1 but not Tri-1. Due to the limited knowledge of the detailed structures of these heterogeneously sialylated oligosaccharides, these possibilities cannot at present be discriminated. Finally, in contrast to sialylated dibranched structures, which are at best retarded but not bound by RCAI (Fig. lA), the presence of three or four sialylated branches is in some cases sufficient for binding of oligosaccharides to RCAI (Fig. 3, A and B ) . Thus, it appears that RCAr discriminates among penultimate sugar moieties, with GlcNAc being preferred over reduced GlcNAc or reduced Glc. We would predict that lactose would interact with RCAI to a greater degree than reduced lactose. Furthermore, the ability of lactose but not Gal to elute bound oligosaccharides from RCAI-silica can be readily understood.
Analysis of Oligosaccharides by RCA, Affinity HPLC-Oligosaccharides elute within four major regions (I-IV) during RCAII affinity HPLC. Region I again corresponds to the column void, while region I1 reflects oligosaccharides retarded but not bound by RCAII. Regions I11 and IV correspond to differing elution positions of bound oligosaccharides. Structures bound by RCAII remain tightly associated with the lectin, with no evidence of elution for >60 min of chromatography in PBS/NaN3. Bound oligosaccharides were eluted at 30-40 min with 100 mM GalNAc in 1.0 N acetic acid. These conditions were the only ones found to effectively and quantitatively elute RCAII-bound oligosaccharides. For example, high concentrations of GalNAc, Gal, GlcNAc, or lactose in the absence of acetic acid, acetic acid alone, or high concentrations of sugars other than GalNAc in acetic acid did not result in elution of bound oligosaccharides from the lectin.
The presence of a2,3-versus a2,6-linked sialic acid does not appear to differentially affect oligosaccharide interaction with RCAII. Di-1 and Di-2, which contain two terminal a2,6and a2,3-linked sialic acid moieties, respectively, coelute in region I1 during RCAlI affinity HPLC (Fig. 4, A and B ) . The asialo forms of Di-1 and Di-2, each bearing two terminal @1,4linked Gal residues, bind to the lectin and elute between regions I11 and IV following addition of 200 mM GalNAc in 1.0 N acetic acid (Fig. 4C). Removal of Gal from these dibranched structures reduces but does not abolish interaction with RCAII. Agalacto Di-1 and agalacto Di-2 elute in region 11, slightly earlier than the sialylated forms of these oligosaccharides (Fig. 40). Removal of terminal GlcNAc from the  Fig. 1, with detailed structures shown in Table I. agalacto structures abolishes interaction with RCAII, resulting in elution of the MansGlcNAcp core in region I (Fig. 4E). The presence of Fuc attached to the reduced core GlcNAc does not appear to affect interaction of dibranched oligosaccharides with RCAII, for the same reasons cited above for RCAI.
Dibranched oligosaccharides bearing the peripheral sequence S04-GalNAcB1,4GlcNAc~l,2 display the same interaction with RCAII as dibranched structures bearing the peripheral sequence SiaaGal~l,4GlcNAcB1,2. Di-6, which contains two sulfated peripheral branches, coelutes in region I1 with Di-1 and Di-2, both of which contain two sialylated peripheral branches (Fig. 7A). Similarly, desulfated Di-6, bearing two terminal BlP-linked GalNAc residues, coelutes between regions I11 and IV with asialo Di-1 and asialo Di-2, both of which contain two terminal B1,4-linked Gal moieties Symbols for the schematically illustrated oligosaccharides are the same as in Fig. 1, with detailed structures shown in Table I. oligosaccharides (5000 cpm each) were analyzed by RCAIl affinity HPLC. Bound oligosaccharides were eluted with 100 mM GalNAc and 1.0 N acetic acid in PBS/NaNs at 30 min (arrow). Elution regions I-IV are indicated by brackets. Symbols for the schematically illustrated oligosaccharides are the same as in Fig. 1, with detailed structures shown in Table I. (Fig. 7 B ) . Hybrid-type oligosaccharides bearing a single terminal pl,4-linked GalNAc residue on the branch arising from the al,3-linked core Man (desulfated Hyb-1) elute in region 111, slightly earlier than complex-type oligosaccharides bearing two terminal /31,4-linked GalNAc moieties (Fig. 70). Removal of the two peripheral Man residues attached to the a1,G-linked core Man does not alter binding of desulfated Hyb-1 to RCAII (data not shown). Thus, the interaction of RCAII with oligosaccharides containing a single branch, which is attached to the al,3-linked core Man, reflects the type and linkage of the branch's terminal moiety; i e . interaction with &&linked GalNAc (Fig. 7 0 and data not shown) > /31,4linked Gal (Fig. 5C) > @1,3-linked Gal (Fig. 6C).
The affinity of RCAII for dibranched oligosaccharides bearing a terminal sulfate on the branch attached to the a1,3linked core Man depends upon the composition of the branch attached to the a1,G-linked core Man. The sulfated/sialylated structure Di-5, which contains a typical sialylated branch attached to the al,g-linked core Man, coelutes with Di-1, Di-2, and Di-6 in region I1 (Fig. 7 E ) . Similarly, asialo Di-5 coelutes with asialo Di-1, asialo Di-2, and desulfated Di-6 between regions I11 and IV (Fig. 7F). Removal of the Gal residue from asialo Di-5, leaving only the /31,2-linked GlcNAc attached to the al,6-linked core Man, results in elution of the oligosaccharide in region I1 (Fig. 7G). In contrast, Hyb-1 (Fig.   7C), bearing two Man residues attached to the al,6-linked core Man, and asialo Di-5 treated with Diplococcal /3-galactosidase and P-N-acetylhexosaminidase to remove the peripheral branch attached to the al,G-linked core Man (Fig. 7 H ) , coelute in region I. Of note, Di-3, which is analogous to this latter form of Di-5, also elutes in region I (Fig. 5A).
Analysis of tri-and tetrabranched complex-type oligosaccharides by RCAIl affinity HPLC yielded more complicated separations than those obtained with dibranched structures. Whereas Di-1 and Di-2 are retarded but not bound by RCAII, Tri-1 oligosaccharides bind to the lectin and are eluted in region IV (Fig. 8 A ) . Removal of sialic acid from Tri-1 does not appear to significantly alter interaction with RCAII, since asialo Tri-1 is eluted in region IV, slightly later than asialo Di-1 and asialo Di-2 (Fig. 8B). In contrast to the agalacto forms of Di-1 and Di-2, agalacto Tri-1 binds to RCAII and is eluted in region I11 (Fig. 8C).
The agalacto forms of Tri-1, Tri-2, and Tetra-1 were also digested with Diplococcal P-N-acetylhexosaminidase, which selectively removes peripheral GlcNAc moieties (27). This enzyme releases both /31,2-linked GlcNAc residues from agalacto Tri-1, but only the Pl,Z-linked GlcNAc residue attached to the al,3-linked core Man from agalacto Tri-2 and agalacto Tetra-1. The Diplococcal 8-N-acetylhexosaminidase-digested forms of agalacto Tri-1, agalacto Tri-2, and agalacto Tetra-1 all elute in region I during RCAII affinity HPLC (data not shown). In the case of Tri-1, this indicates that 1 or both  Table I (GalPl,4GlcNAc), which also elutes in region I1 (15 min). The difference in interaction between UDP-GalNAc, which contains GalNAc in an 01 linkage, and free GalNAc or lactosamine suggests that the p anomer of GalNAc or Gal is essential for interaction with RCAII.

DISCUSSION
In this study, we have demonstrated that RCAI and RCAII affinity HPLC can be effectively utilized for the fractionation and characterization of N-glycanase-released, reduced oligosaccharides. Like RCAI and RCAII, we have found that leukoagglutinating phytohemagglutinin (L-PHA), erythroagglutinating phytohemagglutinin (E-PHA), Datura stramonium agglutinin, and Vicia villosa agglutinin display high degrees of specificity for oligosaccharides and that these lectins have a greater capacity to discriminate between closely related oligosaccharides than had previously been appreciated (7,8). As a result, oligosaccharides elute a t characteristic and reproducible positions during lectin affinity chromatography. Furthermore, our ability to detect subtle aspects of oligosaccharide-lectin interaction has been enhanced with lectin-silica affinity HPLC (7), as compared to lectin-agarose affinity chromatography. Thus, lectin affinity HPLC can be used for the separation of heterogeneous mixtures of oligosaccharides, for the characterization of oligosaccharide structures, and for the definition of lectin specificities. In addition, in this study we have found that silica-bound RCAI and RCAII are capable of discriminating between smaller structures, such as monoand disaccharides, which has not previously been possible with agarose-bound forms of these lectins. For example, lactosamine, reduced lactosamine, and reduced lactose are separated during RCAI affinity HPLC. Similarly, GalNAc and lactosamine are retarded during RCAII affinity HPLC, whereas GlcNAc and Gal are not.
Studies examining the oligosaccharide specificities of RCAI and RCAll have been performed in this (3) and other (1, 2, 4, 5 ) laboratories utilizing various techniques. Using '251-labeled glycopeptides of known structure, we previously characterized the interaction of peptide-bound oligosaccharides with RCAI and RCAII using precipitation assays (3). Our current approach for analyzing RCAI and RCAII specificity differs significantly from that used previously. In the present study, we have analyzed N-glycanase-released, reduced oligosaccharides rather than Asn-linked structures and have examined oligosaccharide specificity by lectin affinity HPLC rather than by solution-binding assays. Nonetheless, in virtually all respects, the results presented in this study are in agreement with our previous conclusions about RCAl and RCA11 specificities (3). This is true for weak as well as strong oligosaccharide-lectin interactions. For example, we previously found that the presence of a2,6-linked sialic acid decreases but does not abolish binding of oligosaccharides to RCAI. We now find that structures bearing a2,6-linked sialic acid are retarded during RCAI affinity HPLC, whereas those with a2,3-linked sialic acid are not. Although we have detected differences in the specificity of some lectins, such as L-and E-PHA (8), for Asn-linked versus N-glycanase-released oligosaccharides, this does not appear to be the case for either RCAI or RCAII.
If the elution position of an oligosaccharide during lectin affinity HPLC is considered indicative of the strength of its interaction with that lectin, then a number of new conclusions about the specificities of RCAI and RCAll can be made. Although the carbohydrate-binding / 3 subunits of these lectins may be highly homologous (1-3), their specificities for mono-, di-, and oligosaccharides differ in several respects. Both lectins have a high affinity for lactosamine (Galpl,4Glc-NAc), consistent with their common ability to interact with complex-type oligosaccharides bearing one or more peripheral lactosamine moieties. The penultimate sugar residue is of major significance for interaction with RCAI, since reduced lactose and reduced lactosamine are less retarded during RCAr affinity HPLC than unreduced lactosamine, and Gal elutes unretarded. Similarly, oligosaccharides bound by RCAI can be eluted with lactose but not with Gal. In the case of RCAII, GalNAc interacts with the lectin almost as strongly as lactosamine, but free Gal does not interact. In addition, GalNAc but not Gal can be used to elute bound oligosaccharides bearing either terminal Gal or GalNAc from RCAII. These differences in RCAI and RCAIl specificity can be utilized to distinguish larger structures containing terminal GalNAcpl,4GlcNAc from those bearing terminal Galp1,4-GlcNAc: the former interact with RCAII but not RCAI, whereas the latter interact with both lectins. For oligosaccharides bearing both types of terminal sequences, analysis by Vicia uillosa agglutinin affinity HPLC can be used to confirm the presence of terminal GalNAc, since this lectin does not interact with terminal Gal (7).
The interaction of oligosaccharides with RCAl and RCAll is influenced by the location and number of peripheral branches bearing Galpl,4GlcNAc as well as by other structural features of the peripheral branches themselves. For example, RCAl and RCAII display different specificities for oligosaccharides bearing a2,3-uersus a2,6-linked sialic acid moieties. The presence of a2,3-linked sialic acid abolishes interaction of dibranched complex-type oligosaccharides with RCAI, whereas the presence of a2,6-linked sialic acid on such structures reduces but does not eliminate lectin interaction. In contrast, the presence of either a2,3-or a2,6-linked sialic acid on dibranched complex-type oligosaccharides reduces but does not eliminate interaction with RCAII. As discussed above, these results are in agreement with our previous analyses using soluble lectins and precipitation assays (3). Of note, the presence of a2,3-linked sialic acid markedly enhances interaction with L-and E-PHA, whereas the presence of a2,6linked sialic acid abolishes interaction with these lectins (8).
RCAl and RCAll display opposite preferences for the location of peripheral lactosamine-containing branches (Figs. l and 5). Therefore, lectin affinity HPLC with RCAI and RCAIl can be used to determine the attachment site (a1,3-or a1,6linked core Man) of a peripheral branch. RCAI and RCAll can also be used to distinguish between terminal ,81,3-uersus p1,4linked Gal moieties. For example, RCAl displays little affinity for oligosaccharides bearing either 1 or 2 terminal pl,3-linked Gal residues, whereas RCAlI binds such structures, requiring GalNAc for elution. Both lectins strongly interact with oligosaccharides bearing terminal pl,4-linked Gal residues.
The desialylated forms of tri-and tetrabranched complextype oligosaccharides are bound by RCAI. Subsequent removal of Gal abolishes interaction of these structures with RCAI. In contrast, RCAll discriminates between various agalacto triand tetrabranched oligosaccharides, depending upon the attachment sites of their peripheral GlcNAc residues. Tribranched oligosaccharides devoid of sialic acid and Gal continue to interact with RCAII as long as there is not a pl,6linked GlcNAc residue attached to the al,6-linked core Man. The presence of such a GlcNAc residue abolishes interaction of agalacto oligosaccharides with RCAlI. Therefore, RCAII does not require peripheral Gal or GalNAc for interaction with certain complex-type oligosaccharides. Even in the presence of terminal Gal, tribranched oligosaccharides with a peripheral branch attached to the C-6 position of the a1,6linked core Man (asialo Tri-2) do not interact as strongly with RCAII as tribranched structures without this peripheral branch (asialo Tri-1). Thus, the presence of a substituent attached to the C-6 position of the a1,6-linked core Man has a detrimental effect on interaction with RCAII.
In addition to providing a useful method for fractionating and analyzing oligosaccharide structures, our studies using lectin affinity HPLC have further demonstrated the exquisite specificity displayed by lectins for oligosaccharide structural features. Such a high degree of specificity could be essential for lectins in higher organisms, which may play a role in binding and transport of soluble glycoconjugates and/or in cell-cell recognition. The ability of plant lectins, such as RCAI, RCAII, L-PHA, E-PHA, Datura stramonium agglutinin, and Vicia uillosa agglutinin (7, 8), to distinguish among a large number of closely related oligosaccharides emphasizes the importance of considering lectin specificity within the context of oligosaccharide structures as well as competitive monoand disaccharide haptens. Our observations also indicate that without detailed knowledge of the carbohydrate structures under study, one must interpret the binding of lectins to intact glycoproteins and cell surfaces with considerable caution.