@-Linked N-Acetylgalactosamine Residues Present at the Nonreducing Termini of 0-Linked Oligosaccharides of a Cloned Murine Cytotoxic T Lymphocyte Line Are Absent in a Vicia uillosa Lectin-resistant Mutant Cell Line*

The 0-linked oligosaccharides of the cloned, murine cytotoxic T cell line B6.1.SF.1 were compared with the corresponding oligosaccharides from a Vicia vil- losa lectin-resistant mutant of B6.1.SF.1 called VV6 (Conzelmann, A., Pink, R., Acuto, O., Mach, J.-P., Dolivo, S., and Nabholz, M. (1980) Eur. J. Immunol. 10, 860-868). The W 6 mutant cells are deficient in binding sites for this GalNAc-specific lectin. Cells were grown in the presence of [3H]glucosamine and [%I galactose to label the glycoproteins, and the desialyzed, alkaline borohydride-released oligosaccharides were isolated and characterized. The W 6 cells contained a series of 0-linked oligosaccharides ranging in size from a disaccharide to a pentasaccharide. These were composed of galactose, N-acetylglucosamine, and N- acetylhexosaminitol, the latter sugar being derived from the reducing terminus. The predominant oligosaccharide had the partial structure

3 Recipient of a fellowship of the Swiss National Foundation.
In uitro generated murine CTL' express a surface glycoprotein termed T145 which is absent on other types of lymphocytes and lymphomas (1, 2). It has been found that T145 interacts with high affinity with a GalNAc-specific lectin from Viciu villosa and this property has been used to separate CTLs from other lymphocytes (2). Previously, one of us (A. C.) reported that cloned murine CTL lines which are independent of feeder cells but dependent on Interleukin 2 have high levels of binding sites for V. uillosa lectin (3). These cell lines express 7'145 and several other surface glycoproteins which bind to V. villosa lectin-Sepharose and are specifically eluted with the haptene sugar GalNAc. V. uillosa lectin was shown to be highly cytotoxic for these cloned cell lines and this property formed the basis for selecting V. uihsa resistant mutants from a mutagenized culture of a parent CTL-line called B6.1. SF.1 (3). Based on the degree of resistance to the lectin, the mutants could be subdivided into two subclasses, one with intermediate resistance (30-100-fold) and the other with high resistance (1000-fold). The clones with intermediate resistance had a partial loss of V. uillosa lectin-binding sites whereas a clone with high resistance had a 100-fold decrease in lectinbinding sites compared to the parental line. Interestingly, both classes of mutants retained susceptibility to the cytotoxic effects of Helix pomatiu lectin, another GalNAc-specific lectin.
These findings suggested that the parental clones of CTLs expressed an unusual type of oligosaccharide that contained GalNAc residues and that the V. uillosa-resistant clones had specific defects in the assembly of this oligosaccharide. In this paper we analyze the GalNAc-containing oligosaccharides made by the parental CTL-line B6.1.SF.1 and one of the highly resistant mutants (W6).
Enzyms-j3-Galactosidase and j3-N-acetylhexosaminidase from jack bean meal were prepared as described (5). a-N-Acetylgalactosaminidase from Charonia lumpas was from Miles. It contained 0.8% j3-N-acetylgalactosaminidase activity relative to its activity toward the a-anomer. a-N-Acetylgalactoeaminidase from Aspergillus niger was prepared as described (6). j3-Galactosidase from Escherichia coli was from Sigma. Pronase was from Calbiochem-Behring Corp.
Maintenance and Radiolabeling of Tissue Culture Lines-B6.1.SF.l is a murine CTL-line derived as described (7) and subsequently subcloned several times. W 6 is a V. uibsa lectin-resistant mutant line derived from B6.1.SF.1 (3). All cell lines were routinely grown in a-minimal essential medium containing 5 mM glucose, penicillin, streptomycin, and 5% fetal calf serum. Since the cell lines are dependent on Interleukin 2, the culture medium was supplemented with 25% conditioned medium prepared from ConA-stimulated rat spleen cells (7). The cells were radiolabeled in a-minimal essential medium containing 5 mM glucose, penicillin, streptomycin, 216 mg/liter L-glutamine, 10% fetal calf serum, and 25% rat spleen cell conditioned medium. The fetal calf serum and the conditioned medium were dialyzed against Dulbecco's modified Eagle's medium prior to use.
The cells were plated at 5 X lo' or 2 X 106 cells/ml and grown to confluency which took 4 and 2 days, respectively, in the presence of 5-15 pCi/ml of tritiated monosaccharides. The cells were then harvested and washed twice with phosphate-buffered saline.
Preparation of Glycopeptides-Glycopeptides were prepared from labeled, washed cells by extensive Pronase digestion as described (8) except that the Pronase was preincubated for 30 min at 37 "C to inactivate potential contaminating glycosidases.
Preparation of 0-Linked Oligosaccharides-The Pronase-digested material was desalted by gel filtration on Sephadex G-25 and the glycopeptides were then heated at 80 "C for 1 h in 10 mM HC1 to remove sialic acid residues. The reaction mixtures were lyophilized, dissolved in 0.9% NaC1, 10 mM KPO,, pH 7.4, 1 mM MgCI2, 1 mM CaC12 and applied to a ConA-Sepharose column. The material which passed through the ConA-Sepharose was subjected to gel filtration on a Bio-Gel P-6 column (1.4 X 100 cm) equilibrated with 0.1 M NH4HC03. The fractions containing the larger glycopeptides were pooled as noted in Fig. 1, desalted, and incubated at 45 "C for 16 h in the presence of 0.05 M NaOH, 1 M Na borohydride to release the 0linked oligosaccharides (9). The samples were then adjusted to pH 5.0 with 1 N acetic acid, brought back to pH 8.0 with NaOH, and reapplied to the same Bio-Gel P-6 columns as described above. The various fractions were pooled as shown in Fig. 1 and desalted on Sephadex G-25.
High-performance Liquid Chromatography for Separation of Reduced Oligosaccharides-This was performed using an AX-5 (Micropak) column (4 X 300 mm, Waters Associates) on a Varian 5000 high performance liquid chromatography apparatus. The method of Mellis and Baenziger (10) was followed with the following modifications: elution was performed with H,O/acetonitrile (35:65), followed by a linear increase of Hz0 to 50% at 30 min, a further linear increase of H20 to 100% over the next 5 min, 100% H20 for another 5 min, then a linear increase of 100 mM potassium phosphate, pH 6.8 from 0 to 100% over 15 min, and a final 5 min of 100% phosphate buffer. Prior to analysis, the samples were desalted on small columns of Amberlite MB-3 (0.5 X 2 cm) in H20. The Amberlite columns were prewashed with a drop of 50 mM N-acetylglucosamine and water to minimize nonspecific adsorption of oligosaccharides. The samples were then passed through 0.22-p Millipore filters, taken to dryness and dissolved in 50-100 pl of the starting solvent.
Methods of Oligosaccharide Structural Analysis-Mono-and oligosaccharides were separated by descending paper chromatography on Whatman No. 1 paper using one of the following solvent systems: solvent A, ethyl acetate/pyridine/acetic acid/water (5:5:1:3) in a preequilibrated tank for 15 h; solvent B, ethyl acetate/pyridine/water (82:1), in a chromatography tank, not pre-equilibrated, solvent C, 1butanol/pyridine/water (643) in a tank not pre-equilibrated, The sample lanes were cut into 1-cm sections and the distributions of radioactivity were determined by scintillation counting. Standards were located by the silver nitrate dip assay (11).
Strong acid hydrolysis of oligosaccharides from [3H]glucosaminelabeled cells was performed in 3 N HCl at 100 "C for 4 h. The samples were then taken to dryness by rotary evaporation and the amino sugars were reacetylated by adding sequentially 1 ml of deionized H20, 0.1 ml of saturated NaHC03, and 0.15 ml of freshly diluted 2% acetic anhydride in H20. After 5 min at room temperature, the samples were taken to dryness in a rotary evaporator, desalted by passage over an Amberlite MB-3 column (0.5 X 6 cm) and spotted on Whatman No. 1 paper that had been impregnated with metaborate as described (12). The paper was developed for 24 h in solvent system C and the distribution of radioactivity was determined.
The [3H]galactose-labeled oligosaccharides were methylated by the method of Hakomori (13). The methylated species obtained after acid hydrolysis were separated by thin layer chromatography on plates of Silica Gel G using the solvent system acetone:water:concentrated ammonia (2503:1.5) (14). 0.5-cm segments were scraped off the plate and used for liquid scintillation counting. The methylated standards were located by spraying the plate with 5% concentrated sulfuric acid in ethanol and heating at 110 "C.
Exoglycosidase treatmenta with jack bean j3-galactosidase and j3-N-acetylhexosaminidase were performed in 50 mM sodium citrate, pH 4.6, in a final volume of 30 pl. Treatment with E. coli j3-galactosidase was done in 50 mM sodium phosphate buffer, pH 7.3, containing 4 mM MgC12 in a final volume of 30 pl. Digestion with a-Nacetylgalactosaminidase from A. niger was performed in 50 mM sodium citrate, pH 4.0, in a final volume of 100 pl. Digestion with a-N-acetylgalactosaminidase from c. hmpas was carried out in 50 mM sodium citrate, pH 4.6, in a final volume of 50 pl, under a toluenesaturated atmosphere. The amount of enzyme and the time of incubation are given with the individual experiments. After incubation at 37 "C, the samples were passed through a small Amberlite MB-3 column (0.5 X 2 cm) before spotting onto paper.

B6.1.SF.l and Mutant VV6
Cells-Based on the specificity of the V. vilbsa lectins, it was anticipated that the lectin-binding sites present on the parental cells would contain GalNAc (16). Consequently parent and mutant cells were grown in the presence of [3H]glucosamine which serves as a precursor for N-acetylgalactosamine as well as for N-acetylglucosamine and sialic acid (17). The labeled cells were digested with Pronase and the glycopeptides were desalted and subjected to mild acid treatment in order to remove sialic acid residues. This was done to simplify the subsequent workup. At this stage the labeled material consisted of N-linked glycopeptides, 0-linked glycopeptides, and proteoglycans. The material was then passed over a column of ConA-Sepharose which retained some of the N-linked glycopeptides. The run-through material (85% of the starting radioactivity), which contained tri-and tetraantennary N-linked oligosaccharides as well as the 0-linked oligosaccharides, proteoglycans, and free sialic acid, was further resolved by gel filtration on Bio-Gel P-6 (Fig. l, A and D). Most of the radioactivity eluted in the void volume of the column (pool A) or was slightly included (pool B). The A and B fractions were subjected to alkaline borohydride treatment to release 0-linked oligosaccharides and rerun on the Bio-Gel P-6 columns ( Fig. 1, B

-C, E-F).
Approximately 13% of the radioactivity was released from both parent and mutant material as small oligosaccharides and these fractions were pooled as noted on the figure. The oligosaccharides released from the pool A material were used for most of the subsequent experiments.
Amino Sugar Content of the A Fractions-Aliquots of fractions A1-A4 were hydrolyzed in strong acid, and the resulting monosaccharides were separated by paper chromatography in solvent C. The results are summarized in Table I  Aliquots of the indicated fractions were hydrolyzed, reacetylated, and separated by paper chromatography. Numbers in the first column indicate the percentage of radioactivity recovered in each fraction as compared to the total amount of radioactivity in the original Pronase digest of the starting material; numbers in the other columns indicate the percentage of radioactivity recovered in each fraction as compared to total recovery in the paper chromatogram. though the percentage was low in the A1 fractions. Since GalNAc is the predominant N-acetylhexosamine linked to serine or threonine in established structures, it i s likely that GalNAcol is being recovered in the N-acetylhexosaminitol region. However, we cannot exclude the presence of GlcNAcol. The fractions also contained GalNAc and GlcNAc but the ratio of GalNAc to N-acetylhexosaminitol and to GlcNAc was decreased in fractions A%A4 in the mutant cells relative to the parental cells.
Since these cells synthesize considerable amounts of chondroitin and chondroitin sulfate A,' it seemed likely that most of the GalNAc in fraction A1 was derived from these proteoglycans. Approximately 85% of the oligosaccharide material in parental fractions A2 and A3 and mutant fraction A2 bound to Amberlite MB-3 even after reacetylation. This charged material was considered to be most likely derived from the proteoglycans and was not analyzed further. The material in fraction A4 from parental cells and fractions A3 and A4 from the mutant cells did not bind to Amberlite MB-3. The oligo-* Andreas Conzelmann, unpublished observation. saccharides in these fractions were further analyzed by HPLC, using conditions which fractionate oligosaccharides mainly on the basis of size (10). The parental fraction A4 contained oligosaccharides ranging in size from a disaccharide to a hexasaccharide with the dominant peaks being the pentasaccharide and the hexasaccharide (Fig. 2 A ) . No free N-acetylhexosaminitol was detected. In contrast, fractions A3 and A4 from the mutant contained material corresponding to Nacetylhexosaminitol and oligosaccharides ranging in size from a dito a pentasaccharide with the tetrasaccharide being the major peak (Fig. 2, E and C). No material was detected in the position expected of a hexasaccharide.
The amino sugar content of these various fractions is shown in Fig. 3. Each fraction contained N-acetylhexosaminitol as would be expected if these species were derived from 0-linked oligosaccharides. However, no GalNAc was detected in the tetra-and pentasaccharides from the mutant cells whereas this sugar was a prominent component of the equivalent fractions from the parental cells. GalNAc was also present in the tri-and hexasaccharides derived from the parental cells.
Isolation of [3H]Galactose-labeled Oligosaccharides-The protocol described for the preparation of [3H]glucosaminelabeled oligosaccharides was also used to prepare the analogous fractions from [3H]galactose-labeled cells. The behavior of alkaline borohydride-released oligosaccharides on HPLC is shown in Fig. 4. As in the case of [3H]glucosamine-labeled material, the major oligosaccharide from the mutant cells migrated as a tetrasaccharide and no hexasaccharide was detected. In contrast, the predominant oligosaccharide in parental cell material was a pentasaccharide and a significant hexasaccharide peak was also present.
Structure of A4b from Parental Cells-Parental fraction A4b from the HPLC column migrated as a trisaccharide and yet it contained N-acetylhexosaminitol, GalNAc, GlcNAc, and galactose. This indicated that it most likely consisted of a mixture of trisaccharides. The [3H]glucosamine-and the [3H]galactose-labeled fractions were therefore subjected to paper chromatography in solvent C which separated the mixture into three components, termed a, @, and y (Fig. 5). The y peak contained most of the [3H]galactose. In addition, hydrolysis of the [3H]glucosamine-labeled y peak revealed the presence of N-acetylhexosaminitol and GalNAc in a ratio of 1:0.85 and only a trace of GlcNAc (Table 11). Taken together, these data indicate that the y fraction consists of a trisaccharide containing N-acetylhexosaminitol, galactose, and GalNAc. The a and (3 fractions did not appear to be pure, so they were not analyzed further (Table 11).
Methylation of [3H]galactose-labeled A4by followed by acid hydrolysis and separation of the methylated monosaccharides using thin layer chromatography gave rise to a single species which migrated with the 2,3,64rimethylgalactose standard (Fig. 6). This indicates that the trisaccharide contains a single galactose residue substituted at C-4 by GalNAc. Since 2,3,6trimethylgalactose does not separate well from 2,4,6-trimethylgalactose in this TLC system, an independent approach was used to confirm the C-4 substitution. An aliquot of [3H] galactose-labeled material was subjected to periodate oxidation which would destroy the galactose if the substitution is C-4 but not destroy it if the substitution is C-3. As shown in Fig. 7, periodate oxidation completely destroyed the [3H] galactose.
Aliquots of the [3H]galactose-labeled trisaccharide were also digested with a series of exoglycosidases as shown in Fig.  8. The trisaccharide was resistant to E. coli @-galactosidase (Fig. SB), but treatment with jack bean (3-N-acetylhexosaminidase converted the trisaccharide to a disaccharide which migrated with the mobility of Gal(31,3GalNAcol (Fig.  8, C and F ) . This disaccharide can be acted upon by E. coli @galactosidase which released [3H]galactose (Fig. 8C). These  -labeled ( B ) fraction A4b from B6.1.SF.1 cells were subjected to paper chromatography in solvent C for 120 h. The entire sample was either counted ( A ) or the strips were eluted into water (B) and one-sixth of each fraction was used for scintillation counting; the remaining five-sixths was pooled as indicated, filtered, and used to determine the amino sugar content (Table 11).

Amino sugar content of A46 fractions from B6.l.SF.l cells
The A4b fraction from [3H]glucosamine-labeled parental B6.1.SF.1 cells was separated into three fractions by paper chromatography as shown in Fig. 5B. The samples were eluted from the paper and aliquots were subjected to strong acid hydrolysis followed by paper chromatography in solvent C. The radioactivity recovered in the Nacetylhexosaminitol area was set to 1. results confirm that the GalNAc is at the nonreducing end of the trisaccharide and indicate that it is linked to the galactose. Further evidence for the &linkage was obtained by demonstrating that the GalNAc is not released by the a-Nacetylgalactosaminidases from A. niger and C. lampas (Fig. 8,   D and E, respectively).
Partial Structure of A4e from Parental Cells"A4e eluted from the HPLC column in the position expected of a hexasaccharide. The ratio of N-acetylhexosaminito1:GalNAc:Glc-NAc was determined to be 1:1.5:0.6 (see Fig. 3). This composition indicated that the fraction consisted of a mixture of oligosaccharides with at least one of these containing more than one GalNAc residue. The oligosaccharides also contain galactose since the A4e material was isolated from [3H]galactose-labeled cells (Fig. 4). The material gave rise to a single, symmetrical peak on paper chromatography in solvent systems A and B, so it was not possible to separate the various components that were presumed to be present in the mixture. Digestion of [3H]glucosamine-labeled A4e material with P-N-acetylhexosaminidase released 47% of the radioactivity as free GalNAc (Fig. 9). The ratio of N-acetylhexosamini-to1:GalNAc:GlcNAc in the residual oligosaccharides was (data not shown). The disaccharide was resistant to jack bean ,?-galactosidase which is known to be unable to cleave Gal/31,3GalNAc linkages. These data indicate that the disaccharide in A4a is probably Gal/31,3GalNAcol.
Fraction A3c gave rise to a single symmetrical peak when analyzed by paper chromatography in solvents C and A (Fig.  1OA). Treatment of [3H]galactose-labeled material with jack bean @-galactosidase released 56% of the radioactivity as free galactose and the residual oligosaccharide migrated in the position expected of a trisaccharide (Fig. 10B). Treatment with both jack bean 8-galactosidase and p-N-acetylhexosaminidase released the same amount of galactose (54%), but now the residual oligosaccharide co-migrated with the Galpl,3GalNAcol standard (Fig. 1OC). Methylation of intact, [3H]galactose-labeled A3c gave rise to tetramethylgalactose as the sole methylated species, indicating that all of the galactose residues are unsubstituted (Fig. 6). Analysis of fraction A3c from [3H]glucosamine-labeled cells revealed the presence of N-acetylhexosaminitol and GlcNAc (Fig. 3B). Taken together, the most likely structure of A3c is Galfll,4Glc- NAc@1,6(Ga1~1,3)N-acetylhexosaminitol. The presence of the Gal~1,3-N-acetylhexosaminitol sequence would explain why jack bean @-galactosidase only released one-half of the galactose from the oligosaccharide.

DISCUSSION
The major finding in this study is that a cloned, murine cytotoxic T cell line with numerous binding sites for the V. villosa lectin contains 0-linked oligosaccharides with GalNAc residues at their nonreducing termini. The smallest of these is a trisaccharide with the sequence GalNAc@1,4Gal/3-N-acetylhexosaminitol. Since these oligosaccharides were desialyzed in order to facilitate the workup, it is probable that the native oligosaccharides also contain sialic acid residues. In sharp contrast, a V. uiUosa lectin-resistant clone that fails to bind this lectin contains 0-linked oligosaccharides that are devoid of the GalNAc residues at the nonreducing termini. This finding suggests that these oligosaccharides on the parental cells may serve as the V. villosa binding sites.
N-Acetylgalactosamine has previously been found as a constituent of cell surface glycoproteins in a variety of structural configurations. GalNAc or sialyl-GalNAc linked O-glycosidically to serine or threonine (a structure classically found on mucins) has been described on rat AH66 ascites hepatoma cells (19), on epiglycanin, which is the major surface glycoprotein of the TA3-Ha murine mammary carcinoma cell line (20), and on human Tn' erythrocytes (21). Other sugars may be added so that a family of oligosaccharides results, as in AH66 cells, epiglycanin, glycophorin (22), the W3/13 glycoprotein of rat thymocytes (23) and the low density lipoprotein receptor (24). a-linked GalNAc also occurs as a constituent of the blood group A determinant, GaiNAcal,3[Fucal,2] Gal@1, which is located at the nonreducing termini of oligo-P. Stoffen, personal communication.
saccharides linked either Nor 0-glycosidically to peptides or to glycolipids (25). This determinant, apart from being the antigen of blood group A-positive human red blood cells, occurs in other species such as hog (26) and rat (27). By contrast, oligosaccharides containing 8-linked GalNAc residues are quite unusual. Recently, such residues have been found to be a constituent of the very rare blood group determinant called Cad (28). Glycophorin from erythrocytes of Cad-positive individuals was shown to contain 0-linked oligosaccharides with the structure GalNAcp1,4(NeuAca2,3)-Galj31,3(NeuAca2,6)GalNA~-Ser/Thr which differs from the 0-linked structures on glycophorin from Cad-negative individuals only by having the additional @-linked GalNAc residue. Erythrocytes from Cad-positive individuals react strongly with antibodies against another blood group determinant, Sda, which is an antigen of varying strength present on erythrocytes of more than 90% of Caucasians. This finding led to the suggestion that Sda and Cad might be related (29).
In fact, a difference in GalNAc content of the Tamm-Horsfall urinary glycoprotein between Sda-positive and negative individuals has been reported (30) and treatment of Sda-positive Tamm-Horsfall glycoprotein with Escherichia freundii endo-@-galactosidase released a pentasaccharide that strongly inhibited the agglutination of Sda-positive erythrocytes by human anti-Sda antiserum (31). The structure of this pentasaccharide was found to be GalNAcj31,4(NeuAc2,3)Ga1/31, 4GlcNAc@l,3Gal (31). In addition to erythrocytes and the Tamm-Horsfall glycoprotein, Sda activity has been found in urinary mucin (32) and various tissues from other species (33). @-linked GalNAc has also been detected in fish egg glycoproteins (34)(35)(36). For instance, the oligosaccharide GalNAc~1,4Gal-~1,4(NeuNGc2,3)GalNAc@1,3Gal@1,3Gal-NAcol is present in trout egg glycoproteins. In this sequence the galactose substituted @1,4 by the GalNAc is not substituted by sialic acid as it is in the Cad-and Sda-positive oligosaccharides. The fact that the trisaccharide isolated from the parental cytotoxic T cells also contains the GalNAc@l,4Gal@ sequence is of particular interest. Since this material was desialyzed prior to its isolation, it is possible that it contained sialic acid in its native state. This similarity suggested to us that the Nacetylgalactosaminyltransferase responsible for the synthesis of the Cad and Sda determinants may also be involved in the synthesis of the 0-linked oligosaccharides on the cytotoxic T cells and that the mutant VV6 cells may be deficient in this enzyme activity. This proved to be the case as documented in the following paper (37).
In addition to the relationship of the GalNAc-containing oligosaccharides to the Cad and Sda determinants, our data suggest that these oligosaccharides may serve as the main binding sites for the the V. villosa lectin. In a binding assay with purified V. villosa B4 lectin, we obtained a linear Scatchard plot and calculated that the parental cell line contained 9 X lo6 binding sites/cell and bound the B4 lectin with an association constant of 1.5 X IO7 M" (data not shown). Moreover, although the lectin preparation used to select mutants contained both A and B subunits: pure B4 lectin was as cytotoxic for the parental cell line as the less pure prepa-S. Tollefsen, personal communication. The lane containing the treated sample was cut in 1-cm strips which were individually placed in vials with water. One aliquot was used to determine the distribution of radioactivity. The rest was pooled as noted (a, @, and y in B ) and subjected to strong acid hydrolysis to determine the amino sugar content. The arrow indicates the position of the GalNAc standard.  (Fig. 4B) was subjected to preparative paper chromatography in solvent A and gave rise to a single symmetrical peak. Aliquots of this material were treated as follows: A, no treatment; 8, digestion with 0.1 unit of jack bean @-galactosidase for 48 h at 37 "C; C, digestion with 0.1 unit of jack bean 8-galactosidase plus 0.05 unit of jack bean @-N-acetylhexosaminidase for 48 h at 37 "C. The samples were then desalted on Amberlite MB-3 and subjected to paper chromatography in solvent A. The standards are: 1, Gal@l,3GalNAcol; 2, galactose.
ration. While GalNAcal-Ser/Thr sequences can act as binding sites for the B4 lectin (16), the parental cytotoxic T cells contained very little of this structure relative to the larger 0linked oligosaccharides with the 8-linked GalNAc. The absence of @-linked GalNAc-containing 0-linked oligosaccharides in the lectin-resistant VV6 cells provides the strongest evidence that these residues serve as B4 lectin-binding sites in this cell type. Recently, Tollefsen and Kornfeld have shown that V. villosa B4 lectin binds to Cad-positive erythrocytes whereas it is unable to bind to Cad-negative cells.5 This finding provides direct evidence that the Cad determinant serves as a V. villosa lectin-binding site. These findings demonstrate that the B, lectin is capable of binding to both aand @-linked GalNAc residues. This is compatible with the observation of Tollefsen and Kornfeld that p-nitrophenyl aand @-GalNAc are equally potent as inhibitors of the binding of t h e B4 lectin to Tn erythrocytes (16).
A previous analysis of desialylated 0-linked oligosaccharides on the surface of murine lymphocytes revealed GalP1,SGalNAc-Ser/Thr as the predominant structure (38). The larger 0-linked oligosaccharides detected in the present study were not found.
In the following paper, we characterize the glycosyltransferase responsible for the synthesis of the @-linked GalNAc residues present on the 0-linked oligosaccharides and demonstrate the absence of this enzyme in the mutant VV6 cells.