LEC18, a dominant Chinese hamster ovary glycosylation mutant synthesizes N-linked carbohydrates with a novel core structure.

The dominant Chinese hamster ovary cell glycosylation mutant, LEC18, was selected for resistance to pea lectin (Pisum sativum agglutinin (PSA)). Lectin binding studies show that LEC18 cells express altered cell surface carbohydrates with markedly reduced binding to 125I-PSA and increased binding to 125I-labeled Datura stramonium agglutinin (DSA) compared with parental cells. Desialylated [3H]Glc-labeled LEC18 cellular glycopeptides that did not bind to concanavalin A-Sepharose exhibited an increased proportion of species that were bound to DSA-agarose. Most of these glycopeptides bound to ricin-agarose and were unique to LEC18 cells. This fraction was purified from approximately 10(10) cells and shown by 1H NMR spectroscopy and methylation linkage analysis to contain novel N-linked structures. Digestion of these glycopeptides with mixtures of beta-D-galactosidases and N-acetyl-beta-D-glucosaminidases gave core glycopeptides that, in contrast to cores from parental cells, were mainly not bound to concanavalin A-Sepharose or to PSA-agarose. 1H NMR spectroscopy, matrix-assisted laser desorption ionization/time of flight mass spectrometry, electrospray mass spectrometry, and collision-activated dissociation mass spectrometry showed that the LEC18 core glycopeptides contained a new GlcNAc residue that substitutes the core GlcNAc residues. Methylation linkage analysis of the parent compound provided evidence that the GlcNAc is linked at O-6 to give the following novel, N-linked core structure. [formula: see text]

I-PSA and increased binding to 125 I-labeled Datura stramonium agglutinin (DSA) compared with parental cells. Desialylated [ 3 H]Glc-labeled LEC18 cellular glycopeptides that did not bind to concanavalin A-Sepharose exhibited an increased proportion of species that were bound to DSA-agarose. Most of these glycopeptides bound to ricin-agarose and were unique to LEC18 cells. This fraction was purified from ϳ10 10 cells and shown by 1 H NMR spectroscopy and methylation linkage analysis to contain novel N-linked structures. Digestion of these glycopeptides with mixtures of ␤-D-galactosidases and N-acetyl-␤-D-glucosaminidases gave core glycopeptides that, in contrast to cores from parental cells, were mainly not bound to concanavalin A-Sepharose or to PSA-agarose. 1 H NMR spectroscopy, matrix-assisted laser desorption ionization/time of flight mass spectrometry, electrospray mass spectrometry, and collision-activated dissociation mass spectrometry showed that the LEC18 core glycopeptides contained a new GlcNAc residue that substitutes the core GlcNAc residue. Methylation linkage analysis of the parent compound provided evidence that the GlcNAc is linked at O-6 to give the following novel, N-linked core structure.
The N-linked carbohydrates of mammalian cells are initially synthesized on dolichol-phosphate, transferred en bloc to an Asn-X-(Ser/Thr) sequon, trimmed by glycosidases, and matured via the sequential action of a series of glycosyltransferases (1). The core region of a mature N-linked structure is the Man␣ (1,6)[Man␣ (1,3)]Man␤ (1,4)GlcNAc␤ (1,4)GlcNAc, originally synthesized on dolichol. In mammalian cells, only four types of core have been described: the Man 3 GlcNAc 2 Asn described above; Man 3 GlcNAc 2 (Fuc)Asn, where Fuc 1 is linked O-6 to the Asn-linked GlcNAc; and cores, with or without Fuc, that contain a bisecting GlcNAc (2,3). By contrast, certain plant glycoproteins have N-linked carbohydrates with Xyl attached at O-2 of the ␤-1,4-linked Man and/or Fuc attached at O-3 of the Asn-linked GlcNAc (4,5). A glycoprotein from honeybee has been reported to have two Fuc residues substituting the Asn-linked GlcNAc at O-3 and O-6 (6). No substitutions of the latter types have been reported to date in mammalian glycoproteins (1)(2)(3). However, an altered N-linked core might be expected to be the basis of cellular resistance to the toxicity of pea lectin (PSA), because PSA binds specifically to the fucosylated core region of biantennary and ␤-1,6-branched, triantennary, N-linked carbohydrates (7).
LEC18 Chinese hamster ovary (CHO) cells are rare mutants that were isolated following a selection for resistance to PSA (8). They are 39-fold more resistant to PSA compared with parental cells. In addition, LEC18 cells are 16-fold more resistant to Lens culinaris agglutinin (LCA), a lectin with a very similar binding specificity to PSA (9). The properties of LEC18 CHO cells are dominant in somatic cell hybrids formed with parental cells (8), showing that their lectin resistance arises from a gain-of-function mutation. Other gain-of-function CHO glycosylation mutants include LEC10, LEC11, LEC12, LEC29, and LEC30 (10). Each of these mutants synthesizes cell surface carbohydrates carrying a sugar residue that is not synthesized by parental CHO cells and that confers a distinctive pattern of lectin resistance. In this paper, we show that LEC18 cells also synthesize a species of N-linked carbohydrate that is not present in parental cells, nor in any of the previously described dominant CHO mutants. These carbohydrates have a novel GlcNAc substitution in their core region, which markedly alters the conformation of the trimannosyl core and is likely to be the reason that LEC18 cells are highly resistant to both PSA and LCA.  and Sephadex G-25 were from Pharmacia Biotech Inc.; L-PHA-agarose,  RCA I -agarose, RCA II -agarose, tomato-agarose, DSA and DSA-agarose, and PSA and PSA-agarose were from Vector; Bio-Gel P-2 (45-90 mesh), the Bradford protein reagent, and Chelex-100 (200 -400 mesh) were from Bio-Rad. Trifluroacetic acid and constant boiling hydrochloric acid (6 N) were from Pierce; ␤-D-galactosidase (Diplococcus pneumoniae and bovine testis) and N-acetyl-␤-D-glucosaminidase (D. pneumoniae) were from Boehringer Mannheim; ␤-D-galactosidase (Jack bean) was from Oxford GlycoSystems; D(ϩ)-Glc, D(ϩ)-Gal, D(ϩ)-Man, D(ϩ)-Fuc, D(ϩ)-GlcN, D(ϩ)-GalN, methyl ␣-D-mannoside, methyl ␣-D-glucoside, Triton X-100, Nonidet P-40, bovine serum albumin, and Amberlite MB-3 were from Sigma; Pronase (Streptomyces griseus) was from Boehringer Mannheim. Ecolume was from ICN Biomedicals, CA; acetonitrile (high performance liquid chromatography grade), 50% (w/v) sodium hydroxide, and other reagent grade chemicals were from Fisher.
Cell Lines and Cell Culture-CHO cells were grown in suspension at 37°C in complete ␣ medium (Life Technologies, Inc.) containing 10% fetal calf serum. Independent LEC18 isolates Pro Ϫ LEC18.21B and Gat Ϫ LEC18.14F were obtained from a selection for mutants resistant to the toxicity of PSA (8). Pro Ϫ 5 parental CHO cells, Pro Ϫ Lec1.3C, and Pro Ϫ Lec13.6A CHO glycosylation mutants were described previously (14).
Binding of 125 I-PSA and 125 I-DSA to Whole Cells-Binding experiments were performed as described previously (15). PSA (15 g) or DSA (30 g) were iodinated to a specific activity of 5.6 ϫ 10 7 and 0.2 ϫ 10 7 cpm/g protein, respectively using the chloramine-T method. 125 I-Lectins were mixed with unlabeled lectins to obtain a range of 2 ng/ml to 200 g/ml (including preparations of identical concentrations but different specific activities) in phosphate-buffered saline (pH 7.4) containing 2% bovine serum albumin (PBS/BSA). Exponentially growing cells, washed 3 times with PBS/BSA, were added to 125 I-lectin in a final volume of 100 l. After 1 h at 4°C, cell-bound and free lectin were separated by centrifugation through 15% BSA and counted separately in a LKB ␥-counter.
Preparation of Radiolabeled Cell Surface Glycopeptides-Suspension cultures of Pro Ϫ 5 and Pro Ϫ LEC18.21B cells were established at a density of 1.5 ϫ 10 5 cells/ml in complete medium, and 100 Ci of [ 3 H]Glc was added to each 10-ml culture. Following a 48-h incubation at 37°C, the cells were washed 3 times with 10 ml of PBS containing 0.01 M Ca 2ϩ , 0.01 M Mg 2ϩ , 0.02% azide (PBS 2ϩ ). Following extraction in 5 mM Tris-HCl, pH 7.8 containing 1% Nonidet P-40 for 5 min at 4°C and 5 min at room temperature, nuclei were removed by centrifugation. The supernatant was incubated with 1 mg/ml Pronase in 5 mM Tris-HCl (pH 7.8), 3 mM CaCl 2 at 55°C under toluene. Additional Pronase was added every 24 h for 3 days. After boiling for 10 min, the sample was centrifuged for 10 min at 12,000 ϫ g, desalted on Bio-Gel P-2, and stored at Ϫ20°C.
Samples of glycopeptides were applied to L-PHA-agarose, RCA I -agarose, RCA II -agarose, tomato-agarose, or DSA-agarose columns in PBS 2ϩ (pH 7.4), and the columns were washed with Ͼ8 column volumes of the same buffer. The addition of sugar was not required for elution of glycopeptides from the L-PHA-agarose column. Buffer containing 100 -200 mM lactose was used to elute bound glycopeptides from RCA Iagarose and RCA II -agarose columns. The bound glycopeptides from tomato-agarose and DSA-agarose columns were eluted using a mixture of N,NЈ-diacetylchitobiose and N,NЈ,NЉ-triacetylchitotriose (7-10 mg/ ml) in PBS 2ϩ (pH 7.4). Chromatography was performed either at room temperature or at 4°C at flow rates of 6 -10 ml/h.
Purification of Glycopeptides for Structural Analysis-Approximately 10 10 Pro Ϫ LEC18.21B cells were pelleted, washed twice with PBS 2ϩ , and incubated with 1 mM Tris-HCl buffer (pH 7.2) containing 1.0% Nonidet P-40 on ice for 30 min. After centrifugation at room temperature for 30 min at 3000 rpm, the supernatant was incubated with 1 mg/ml Pronase in 1 mM Tris-HCl (pH 7.2), 3 mM CaCl 2 at 55°C under toluene. Fresh Pronase was added every 24 h for 5 days. After boiling for 10 min, the sample was centrifuged for 2 h at 30,000 rpm in a type 50 Ti rotor. The supernatant was loaded onto a column of Sephadex G-25 (5 cm ϫ 65 cm). The column was eluted with glass distilled water, and fractions of 10 ml were assayed with phenol-sulfuric acid for neutral hexoses (16). Positive fractions were pooled, concentrated by rotoevaporation to 5 ml, and diluted with an equal volume of 2ϫ ConA buffer. Insoluble material was removed by centrifugation at 3000 rpm for 10 min, and the supernatant was chromatographed on a column of ConA-Sepharose (1.5 cm ϫ 22 cm). Unbound glycopeptides were desalted on a Sephadex G-25 column (5 cm ϫ 65 cm), lyophilized, resuspended in 2 ml water, and adjusted to 10 mM HCl by adding 10 l of aqueous 2 N HCl. After heating at 80°C for 1 h to hydrolyze bound sialic acid, the hydrolyzate was evaporated to dryness, co-evaporated with methanol to remove traces of acid, and passed through a column of Sephadex G-25. The desialylated glycopeptides, dissolved in 1 ml of PBS 2ϩ were applied to tomato-agarose (0.5 cm ϫ 20 cm). The unbound fraction was fractionated on DSA-agarose (0.5 cm ϫ 20 cm). The DSAretarded fraction was desalted on Bio-Gel P-2 (1.5 cm ϫ 70 cm), lyophilized, dissolved in 1 ml of PBS 2ϩ , and chromatographed on a column of RCA II -agarose (0.5 cm ϫ 20 cm). The column was eluted with 200 ml of PBS 2ϩ followed by 100 ml of 100 mM lactose in PBS 2ϩ . Pooled fractions were desalted on Bio-Gel P-2. A portion of each glycopeptide fraction was subjected to monosaccharide analysis. The remainder was used to record 1 H NMR spectroscopy and was subjected to methylation linkage analysis or treated with exoglycosidases to isolate core glycopeptides as described below.
Methylation Linkage Analysis-Glycopeptides (ϳ100 g) were permethylated according to Hakomori (18). The permethylated products were purified using a Sep-Pak C 18 cartridge (Waters), hydrolyzed with 2 M trifluoroacetic acid, reduced with sodium borodeuteride, acetylated with pyridine-acetic anhydride, and analyzed by gas-liquid chromatography-mass spectrometry (GLC-MS) as described (19). GLC-MS was performed with a Hewlett Packard 5890A gas liquid chromatograph using a SP-2330 column (Supelco) for analysis of neutral monosaccharide derivatives and a DB-1 column (J&W Scientific) for the analysis of hexosamine derivatives, coupled to a Hewlett-Packard 5970 MSD mass spectrometer.
Mass Spectrometry-Electrospray-mass spectrometry (ES-MS) and MS/MS analysis of glycopeptides were performed on an API-III triplequadrupole mass spectrometer (PE-SCIEX, Ontario, Canada) using the SCIEX IonSpray interface with nitrogen as the nebulizer gas, an ion spray voltage of 3300, and the orifice at 35 or 70 V. The sample was infused into the mass spectrometer at 2 l/min using a Harvard apparatus syringe pump after diluting 1:1 with 50% acetonitrile/H 2 O containing 0.1% trifluoroacetic acid. Collision-activated dissociation mass spectrometry (MS/MS) was performed using argon as the collision gas at a collision gas thickness of ϳ1 ϫ 10 10 molecules/cm 2 and a collision energy of 30 -35 eV. Under these collision-activated dissociation conditions, daughter ion spectra of selected parent ions were obtained.
Matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF-MS) was performed on a Voyager RP Biospectrometry work station (Perspective Biosystems, Framingham, MA), using ␣-cyano-4-hydroxycinnamic acid as matrix. A laser power of 190 attenuator (nitrogen laser at 337 nm) was used, and an average of 50 scans were taken.
Enzyme Digestions-Jack bean ␤-D-galactosidase digestion was carried out in 100 l of sodium citrate buffer (pH 3.5) at 37°C under toluene for 24 -72 h with a total of 1 unit enzyme. Bovine testis ␤-Dgalactosidase and Jack bean N-acetyl-␤-D-glucosaminidase digestions were performed in 100 l of sodium citrate buffer (pH 4.5) at 37°C under toluene for 24 -96 h with a total of 100 munits enzyme. D. pneumoniae ␤-D-galactosidase and N-acetyl-␤-D-hexosaminidase digestions were performed in 50 -100 l of sodium citrate buffer (pH 5.0) at 37°C under toluene for 24 -96 h, with a total of 100 milliunits of enzymes. Each reaction was started with 25 milliunits enzyme, and an additional 25 milliunits were added at 24-h intervals.

RESULTS
Binding of 125 I-PSA and 125 I-DSA to LEC18 Cells-The increased resistance of LEC18 cells to PSA and LCA (8) is expected to be due to the expression of altered cell surface carbohydrates. As predicted, LEC18 mutants exhibited a marked (2-4-fold) reduction in 125 I-PSA binding when compared with parental CHO cells over a broad range of PSA concentrations (Fig. 1). In the linear region of the binding curves, the differences in 125 I-PSA binding were ϳ4 -5-fold when the assay was performed at room temperature (data not shown). Two independently isolated LEC18 mutants exhibited a similar reduction in PSA binding. Lec13 CHO mutant cells were used as a negative control, as they synthesize N-linked carbohydrates that lack a core Fuc residue and hence do not bind to PSA (20).
Most CHO glycosylation mutants exhibit an increased sensitivity to and binding of certain lectins (21). Such lectins, once identified, can be useful for affinity purification of novel glycopeptides. In binding experiments with a series of 125 I-lectins, DSA was identified as binding better to LEC18 cells than to parental CHO cells (Fig. 1). DSA binds to poly-N-acetyllactosamine chains (22,23), and it can be seen that LEC18 cells bound significantly more 125 I-DSA than parental cells. Both independent LEC18 mutants exhibited a similar increase in DSA binding. Lec1 CHO cells were used as a negative control, because they do not synthesize complex, N-linked carbohydrates (24).
A Glycopeptide Fraction Unique to LEC18 Cells-The increased binding of 125 I-DSA by LEC18 cells would be consistent with increased affinity or increased synthesis of poly-N-acetyllactosamine chains (23). To identify carbohydrates unique to LEC18 cells, detailed studies with metabolically labeled cellular glycopeptides were carried out by serial lectin affinity chromatography. ConA-Sepharose chromatography (25) of [ 3 H]Glc-labeled glycopeptides from parental and LEC18 CHO cells showed no significant differences in profile. About 85% of the label passed through the column, 5% bound and eluted in the biantennary fraction, while 10% bound and was eluted in the hybrid and oligomannosyl fraction. The glycopeptides that were not bound to the ConA-Sepharose column should include branched N-glycans with poly-N-acetyllactosamine units that would bind to DSA (22,23). These glycopeptides were subjected to mild acid hydrolysis to remove sialic acid residues and fractionated on a tomato-agarose column to identify any difference in branched poly-N-acetyllactosamine content (26). In both cases, ϳ85% of the glycopeptides passed through the column, while ϳ15% were bound and eluted with N,NЈ,NЉ-triacetylchitotriose. However, a difference between parent and LEC18 glycopeptides was uncovered when the species that passed through tomato-agarose were fractionated on DSA-agarose. This glycopeptide fraction from LEC18 was largely retarded on (loosely bound to) DSA-agarose, whereas the same glycopeptide fraction from parental cells was largely unbound to the column (Fig. 2). Monosaccharide analysis of [ 3 H]Glc-labeled DSA-retarded glycopeptides by high performance anion-exchange chromatography with pulsed amperometric detection showed evidence for the presence of only the expected N-linked sugar residues (Man, Gal, GlcNAc, Fuc) in the LEC18 glycopeptide fraction (data not shown).
About 65% LEC18 glycopeptides were retarded on the DSAagarose column, but ϳ15% parental glycopeptides were also present in this pool. In order to separate glycopeptides unique to LEC18 cells, further fractionation of the DSA-retarded glycopeptides was performed on a series of lectin columns. A dramatic difference in elution profile was observed on RCA IIagarose (Fig. 2). LEC18 glycopeptides separated into late re- tarded (45%) and bound (55%) species, whereas parent glycopeptides eluted in the unbound (35%) or early retarded (65%) fractions, indicating that the glycopeptides derived from LEC18 cells were indeed unique to the mutant.
LEC18 Glycopeptides Have an Altered N-Linked Core Region-To determine whether the glycopeptides unique to LEC18 cells were altered in their core region, as might be expected from the PSA resistance of LEC18 cells, DSA-retarded glycopeptides of both parental and LEC18 cells (Fig. 2) were treated exhaustively with a mixture of bovine testis ␤-Dgalactosidase and Jack bean N-acetyl-␤-D-glucosaminidase. The digestion products fractionated on Bio-Gel P-2 into two major glycopeptide pools (V o in the excluded volume and V e in the included volume) and two monosaccharide pools (Gal and GlcNAc) (Fig. 3). As the V o fraction was complex and probably contained some partially digested material, the V e fraction was choosed for further studies.
Fractionation on ConA-Sepharose and PSA-agarose revealed a significant difference between V e glycopeptides of LEC18 and parent cells. As would be expected, ϳ95% of parental CHO V e glycopeptides did bind to ConA-Sepharose, indicating that they had been converted to trimannosyl cores (Fig. 4). By contrast, ϳ65% of LEC18 V e glycopeptides did not bind to ConA-Sepharose (Fig. 4). These results suggested that the core glycopeptides from LEC18 cells have a modified structure. Similarly, for core glycopeptides obtained following exhaustive digestion with a mixture of ␤-D-galactosidase and N-acetyl-␤-D-glucosaminidase from D. pneumoniae (27), a much greater proportion of LEC18 V e glycopeptides did not bind to ConA-Sepharose, compared with parental glycopeptides.
A trimannosyl, fucosylated core structure would be expected to bind to PSA-agarose (7). This was found to be the case for a preparation of fucosylated core glycopeptides from parental CHO cells (Fig. 4). However, the V e glycopeptides from LEC18 cells did not bind to PSA-agarose (Fig. 4), despite the presence of almost one full Fuc equivalent (Table I). The monosaccharide compositional analysis of the V e fraction of LEC18 and parental CHO cells in Table I also showed that neither LEC18 nor parent core glycopeptides had any Gal residues. In addition, parent cores had only approximately two GlcNAc equivalents showing that the glycosidase digestion was complete. However, LEC18 core glycopeptides contained an additional GlcNAc residue that had resisted N-acetyl-␤-D-hexosaminidase digestion. This GlcNAc was postulated to be the reason most LEC18 core glycopeptides did not bind to ConA-Sepharose or PSA-agarose (Fig. 4).

FIG. 2. Lectin affinity chromatography of 3 H-Glc-labeled glycopeptides. [ 3 H]
Glc-labeled glycopeptides that passed through ConA-Sepharose were desialylated, and fractionated on tomato-agarose. The unbound fraction was subsequently fractionated on DSA-agarose (upper panel). In the lower panel, the DSA-retarded fraction from parental CHO or LEC18 cells was fractionated on RCA II -agarose.

Structural Analysis of LEC18 Glycopeptides By 1 H NMR Spectroscopy, Methylation Linkage Analysis, and Mass Spectrometry-The studies on [ 3 H]Glc-labeled glycopeptides sug-
gested that an N-linked glycopeptide fraction unique to LEC18 cells gave rise to a core structure with an additional GlcNAc residue. In order to locate the linkage position, cellular glycopeptides from ϳ10 10 LEC18 cells were isolated, and the DSAretarded/RCA II -bound fraction (Fig. 2) was prepared. The 1 H NMR spectrum of these LEC18 glycopeptides is shown in Fig.  5. It is evident that the glycopeptides are a mixture of N-linked structures containing poly-N-acetyllactosamine and Fuc and a small amount of sialic acid. Hydrazinolysis of this preparation revealed multiple species eluting at the position of neutral oligosaccharide standards on GlycopreP 1000 (Oxford Glyco-Systems, Oxford, UK). Furthermore, Bio-Gel P-4 chromatography of the released oligosaccharides indicated that they are high molecular weight complex carbohydrates with Ͼ Ͼ24 Glc units (data not shown).
The 1 H NMR spectrum in Fig. 5 is unusually complex in the 4.9 -5.2-ppm region, and no resonances corresponding to known H 1 or H 2 chemical shifts for core Man residues in previously reported structures were present in the NMR carbohydrate data base (Sugabase, version 1.05; 3). The resonance at 2.028 ppm is typical of -NAc groups from GlcNAc residues in poly-N-acetyllactosamine chains (28). Thus the glycopeptides are a mixture of complex N-glycans containing poly-Nacetyllactosamine chains.
After further acid hydrolysis to remove the small amount of sialic acid, a portion of these glycopeptides was subjected to methylation linkage analysis. From the data in Table II, it is evident that the LEC18 glycopeptides have terminal Gal (3), GlcNAc (1), and Fuc (0.7) residues and substituted Gal, Glc-NAc, and Man residues. The substitutions of the Man residues are consistent with a mixture of tri-and/or tetraantennary N-linked carbohydrates. The presence of Gal substituted at O-3 is consistent with the presence of poly-N-acetyllactosamine chains, as is the 4-substituted GlcNAc. The most remarkable finding from this analysis was the presence of a single, unsubstituted GlcNAc residue and of 1.75 molar residues of 4,6substituted GlcNAc. The only 4,6-substituted GlcNAc in the usual N-linked structure arises from the presence of Fuc at the O-6 of the Asn-linked GlcNAc. In the LEC18 glycopeptides, Fuc linkage accounts for ϳ0.7 of the 4,6-linked GlcNAc, leaving one full residue of 4,6-substituted GlcNAc to be accounted. In fact, it will be shown subsequently that this residue is located in the altered core region of LEC18 glycopeptides.
The methylation linkage analysis also showed the presence of O-4 substituted Gal residues (Table II), a substitution not observed previously for N-linked carbohydrates from mammalian cells (1-3). However, the presence of this residue in glycolipids and in fish egg N-glycans has been observed previously (29,30). Since no direct linkage analysis has previously been reported for CHO-derived poly-N-acetyllactosamine containing N-linked carbohydrates, the O-4-substituted Gal may be a feature of some N-glycans in all Pro Ϫ CHO cells. Whatever its origin, this residue was susceptible to the exoglycosidases used, because Gal residues were absent after digestion (Table I), indicating that 4-substituted Gal residues are part of the poly-N-acetyllactosamine chains and are linked by ␤-glycosyl residues. Furthermore, the methylation linkage analysis showed the presence of small amounts of 6-substituted, 4,6-substituted, and 3,6-substituted Gal residues (Table II). Finally, the  (Fig. 2) were treated exhaustively with a mixture of bovine testis ␤-D-galactosidase and Jack bean N-acetyl-␤-D-glucosaminidase, and the products were purified by Bio-Gel P-2 chromatography to obtain the V e fraction (Fig. 3). This core glycopeptide fraction was hydrolyzed with 2 M trifluoroacetic acid at 100°C for 3-4 h and subjected to HPAEC-PAD, and fractions were counted as described under "Experimental Procedures." Monosaccharide V e core glycopeptides Approximately 500 g of the RCA II -bound fraction (Fig. 2) were prepared from 10 10 LEC18 cells. They were extensively desalted on Bio-Gel P-2, passed through Chelex, exchanged with D 2 O, and subjected to 1 H NMR spectroscopy at 500 MHz.

TABLE II
Methylation linkage analysis of DSA-retarded/RCA II -bound glycopeptides unique to LEC18 cells The RCA II -bound glycopeptides unique to LEC18 cells (Fig. 2) were subjected to methylation analysis as described under "Experimental Procedures." The partially methylated sugar derivatives present in known N-linked structures are shown. The sample also contained small amounts (Յ0.1-0.2 residue) of sugars such as Glc and Ara, which are likely to be contaminants, as no evidence for these moieties was apparent in 1 H NMR spectra (Figs. 5 and 9) nor in compositional analyses performed by HPAEC-PAD (Table I) analysis showed no evidence for 3,4,6-substituted GlcNAc that would be predicted if the Asn-linked GlcNAc was disubstituted, for 3,4,6-substituted Man that would be predicted for the presence of a bisecting GlcNAc, nor for 2,3,6-substituted Man or for a significant molar proportion of any unusual sugar like Xyl. Structural Analysis of LEC18 Core Glycopeptides-Because our studies of [ 3 H]Glc-labeled glycopeptides identified the core region as the location of a novel modification in LEC18 Nlinked carbohydrates ( Fig. 4 and Table I), the unlabeled glycopeptides shown in Fig. 5 were treated exhaustively with a mixture of ␤-D-galactosidases and N-acetyl-␤-D-glucosaminidases from D. pneumoniae, bovine testis, and Jack bean. The V e fraction was obtained by Bio-Gel P-2 chromatography and shown to elute between Man 3 GlcNAc 2 Asn and Man 5 GlcNAc 2 Asn standards by high performance anionexchange chromatography with pulsed amperometric detection (data not shown). Fig. 6 shows the MALDI-TOF-MS of these LEC18 V e core glycopeptides. The mass spectrum contained one major molecular ion (MH ϩ ) at 1374.4 atomic mass units, which is the mass expected for a glycopeptide with the composition Man 3 GlcNAc 3 (Fuc)Asn. Fig. 6 also shows the ES-MS (lower panel) of LEC18 V e core glycopeptides, which was recorded at 35 V. Again, only one major MH ϩ ion was observed at 1374.4 atomic mass units.
When ES-MS spectra were recorded at 70 V, a strong MH ϩ ion at 1374.4 atomic mass units along with some fragment ions were observed (Fig. 7). Attempts were made to obtain daughter ions by collision-activated dissociation (MS/MS). The MS/MS spectrum in Fig. 8 shows that the MH ϩ ion at 1374.4 atomic mass units gave daughter ions inter alia at atomic mass units 1228.4, 1171.4, 889.2, and 686.2. The interpretation of these ions is shown in Scheme 1. Briefly, the ions at 1228.4 and 1171.4 atomic mass units were derived from the MH ϩ ion with a loss of a Fuc or GlcNAc residue, respectively. However, the ion at 889.2 atomic mass units corresponds to a loss of three Man residues from MH ϩ and has a composition of GlcNAc 3 (Fuc)Asn, confirming the presence of an extra GlcNAc residue in the core region. This ion loses a GlcNAc residue to give the ion at 686.2 atomic mass units.
Based on this information, it was possible to interpret many of the fragment ions observed in the ES-MS spectrum of Fig. 7. Four fragmentation pathways could be deduced (Scheme 2). It is evident that the glycopeptide undergoes fragmentation both from the nonreducing end and from the Asn end to give fragment ions that are interpretable in structural terms. The most critical ions observed were at atomic mass units 889.2, 686.2, and 483.2 and were composed of GlcNAc 3 (Fuc)Asn, GlcNAc 2 (Fuc)Asn, and GlcNAc 1 (Fuc)Asn, respectively. The latter shows that the Asn-linked GlcNAc is substituted with Fuc. This sequence of ions and the MS/MS data provide strong evidence that the new GlcNAc is attached to the core GlcNAc residue and not to the Asn-linked GlcNAc. 1 H NMR spectra of the LEC18 V e core glycopeptides shown by mass spectrometry in Fig. 6 to contain one major molecular species are shown in Fig. 9. It is apparent that the preparation contained a single glycopeptide species with three GlcNAc residues, three Man residues, and one Fuc residue. Spectra were recorded at 23 and at 42°C. At the higher temperature, the HOD peak shifted to reveal certain resonances obscured at 23°C. When the chemical shifts of H 1 , H 2 , -NAc, and Fuc regions were entered into Sugabase version 1.05 (3), no structure was given, consistent with the evidence that the LEC18 V e core is a new structure. Since only four other fucosylated core structures are known to exist in mammalian N-linked carbohydrates (1-3) and since the ES-MS, MS/MS, and GLC-MS data provide strong evidence for the proposals in Schemes 1 and 2, the spectra in Fig. 9 were tentatively assigned (Table  III). Several reasons argue that these assignments will prove correct once larger quantities of LEC18 core glycopeptides can be subjected to chemical analysis. Whereas the presence of the bisecting GlcNAc causes H 1 and H 2 resonances for each core Man residue to shift, only the H 1 of the Man␤1,4-resonance was significantly changed in the LEC18 core; whereas the presence of a bisecting GlcNAc does not alter the H 1 or -NAc resonances of the Asn-GlcNAc, both are changed in the LEC18 core. Most interestingly, the H 1 and -NAc resonances of the LEC18 core GlcNAc are different from any other core, including those from nonmammalian sources. Importantly, none of the latter cores show the resonance at 5.211 ppm. This is assigned to the new GlcNAc of the LEC18 core because it is very similar to the H 1 at 5.144 ppm of the GlcNAc in N,NЈ-diacetylchitobiose-p-nitrophenyl (33) and to the N-acetyl of this GlcNAc, which occurs at 2.012 ppm compared with 2.018 ppm for the novel LEC18 GlcNAc residue. The chitobiose disaccharide was synthesized by a new GlcNAc-GlcNAc-transferase from snail, and the LEC18 GlcNAc, also in GlcNAc-GlcNAc linkage, behaves very similarly, although not identically. The 5.211-ppm resonance at both 23 and 42°C has a J value of 5.5 Hz, a value intermediate between the J values observed for residues in ␣-linkage (J ϭ 3-4 Hz; 3) and residues in ␤-linkage (J ϭ 6.5-8.0 Hz; Ref. 3). The resonance assigned to the novel Glc-NAc is not present in an equimolar amount, in the same manner that the Asn-GlcNAc is not. However, there was no evidence for 3,4,6-substituted GlcNAc in the linkage analysis (Table II) the novel core region of LEC18 N-linked carbohydrates is shown in Scheme 3.

DISCUSSION
In this paper, a series of experiments on highly purified [ 3 H]Glc-labeled and unlabeled glycopeptides from LEC18 CHO cells have provided evidence for a new N-linked core structure that has not previously been observed in N-glycans from any source. The novel core structure was found on branched, polylactosamine-containing, N-glycans from LEC18 cells that were not found in parent CHO cells and that bound to both DSA and RCA II (Fig. 2). Parent CHO glycopeptides were present in the DSA-retarded fraction, but they were shown to have the expected trimannosyl cores (Ͼ95% bound to ConA) after exhaustive digestion with ␤-D-galactosidases and N-acetyl-␤-Dglucosaminidases ( Fig. 4 and Table I). By contrast, identically treated LEC18 glycopeptides remained largely unbound to ConA and PSA (Fig. 4). Composition analysis revealed that the LEC18 core glycopeptides had an extra GlcNAc compared with core glycopeptides from parent cells ( Table I).
Proof of an extra GlcNAc in LEC18 core glycans was obtained by preparing ϳ500 g of the unique LEC18 species that binds to DSA and RCA II . It was shown by 1 H NMR spectroscopy to be a mixture of branched, poly-N-acetyllactosamine-containing species with a core fucose residue (Fig. 5). This material was too complex to analyze further by 1 H NMR, as no corresponding spectrum is present in Sugabase (3). Exhaustive exoglycosidase digestion of this material gave a discreet core preparation of ϳ45 g. MALDI-TOF-MS analysis revealed a major MH ϩ ion at 1374.4 atomic mass units corresponding to Hex 3 HexNAc 3 dHex 1 Asn (or Man 3 GlcNAc 3 (Fuc 1 )Asn). Fragment ions generated during ES-MS revealed the critical ion at 889.2 atomic mass units that corresponded to a species of HexNAc 3 dHex 1 Asn (or GlcNAc 3 (Fuc 1 )Asn). MS/MS analysis of the 1374.4 atomic mass units ion confirmed the generation of this critical fragment ion (Scheme 1). In addition, fragment ions generated by ES-MS confirmed the results of MS/MS. These ions were generated when the sample was bombarded at 70 V under conditions normally used to obtain the molecular weight of proteins. Interpretation of the ES-MS data revealed a fragment ion corresponding to breakage at every glycosidic linkage and the occurrence of four fragmentation pathways (Scheme 2). Although no examples in the literature describe fragment ions obtained from glycopeptides using ES-MS (34), our experience shows that it is possible to obtain sequence information on underivatized, native glycopeptides by recording ES-MS using two different conditions (one at 35 V and the SCHEME 3. Proposed structure for the LEC18 V e core glycopeptides.

TABLE III
Chemical shift assignments for the LEC18 V e core glycopeptide fraction Assignments for LEC18 resonances are taken from the spectra in Fig. 9 and are based on those of related fucosylated core structures typical of parent and LEC10 CHO cells (Campbell et al. (39)). The chemical shifts of the most relevant fucosylated structures available in the literature are given for comparison. The data for core A were obtained from Grey et al. (31), and those for core B were from Alonso et al. (32). NA, not assigned because of complexity in the H 2 region. other at 70 V). This is very useful when dealing with small amounts of samples that are difficult to derivatize.
Taken together, the 1 H NMR and MS data provide proof for the existence of a novel core that is characterized by GlcNAc substitution of the core GlcNAc. The evidence that the new GlcNAc is linked at O-6 of the core GlcNAc was derived from GLC-MS of the parent compound (Table II). However the configuration of the new linkage is uncertain because the coupling constant of the H 1 for the novel GlcNAc is 5.5 Hz (Fig. 9), a value intermediate between known ␣and ␤-linked residues (3). The combined data provide strong evidence for the core structure showed in Scheme 3.
If the new GlcNAc in the LEC18 core is in ␤-linkage, as seems likely based on its similarity to the terminal nonreducing GlcNAc in GlcNAc␤1,4GlcNAc-p-nitrophenol (33), it was nevertheless not hydrolyzed by two N-acetyl-␤-D-hexosaminidases. This may be because the new GlcNAc residue is internal and hence crowded and not accessible to the enzymes. An analogous situation exists in the case of ganglioside GD 1a and also in lipopolysaccharides from Campylobacter jejuni, which mimic the GD 1a structure. These molecules contain two, nonreducing, terminal sialic acid residues, one linked to the penultimate Gal, and the other linked to the internal Gal. Known sialidases are able to remove only the sialic acid residue that is linked to the penultimate Gal, but the sialic acid linked to the internal Gal remains resistant (35).
The combined data presented in this paper strongly suggest that the addition of the new GlcNAc residue is the result of the dominant mutation in LEC18 cells. The novel core is in species of LEC18 glycopeptides that are not synthesized by parent CHO cells; the novel core is fucosylated, gives a new NMR spectrum, and does not bind to ConA or PSA (Fig. 3); and the novel core correlates with the high degree of resistance of LEC18 cells to PSA and LCA. Other dominant CHO mutants have analogous properties: LEC11, LEC12, LEC29, and LEC30 CHO mutants add ␣(1,3)-Fuc residues to lactosamine chains on N-glycans (36 -38) while LEC10 CHO mutants add the bisecting GlcNAc to N-glycans (39). Parent CHO cells do not synthesize either ␣(1,3)-Fuc or bisecting GlcNAc-containing N-glycans. In each dominant mutant, the new carbohydrates are the result of the action of a glycosyltransferase activity that is newly expressed and is not present in parent CHO cells. Each newly expressed transferase is one that is developmentally regulated in mammals. Previous studies have provided biochemical evidence for the absence of these specialized transferase activities in CHO cells (36 -39), but more recently we have shown by Northern analysis that CHO cells also lack RNAs for the relevant glycosyltransferase, whereas the dominant mutants possess the expected RNA species. 2 Based on our experience with previous dominant CHO mutants, it was expected that LEC18 cell extracts would have a transferase activity that adds the new GlcNAc residue to a suitable acceptor and that such an activity would be absent from parent CHO cells. In fact, transfer of a GlcNAc residue from UDP-GlcNAc to a biantennary, N-linked glycan that terminates in GlcNAc has been achieved with LEC18 cell extracts, 3 whereas no such activity is detected in parent cell extracts (36). It will now be necessary to determine optimal conditions for assaying this transferase activity and to characterize its product. It is predicted to be a medial Golgi GlcNAc-transferase that is developmentally regulated in vivo.