Novel Poly-GalNAcβ1–4GlcNAc (LacdiNAc) and Fucosylated Poly-LacdiNAc N-Glycans from Mammalian Cells Expressing β1,4-N-Acetylgalactosaminyltransferase and α1,3-Fucosyltransferase*

Glycans containing the GalNAcβ1–4GlcNAc (LacdiNAc or LDN) motif are expressed by many invertebrates, but this motif also occurs in vertebrates and is found on several mammalian glycoprotein hormones. This motif contrasts with the more commonly occurring Galβ1–4GlcNAc (LacNAc or LN) motif. To better understand LDN biosynthesis and regulation, we stably expressed the cDNA encoding the Caenorhabditis elegans β1,4-N-acetylgalactosaminyltransferase (GalNAcT), which generates LDN in vitro, in Chinese hamster ovary (CHO) Lec8 cells, to establish L8-GalNAcT CHO cells. The glycan structures from these cells were determined by mass spectrometry and linkage analysis. The L8-GalNAcT cell line produces complex-type N-glycans quantitatively bearing LDN structures on their antennae. Unexpectedly, most of these complex-type N-glycans contain novel “poly-LDN” structures consisting of repeating LDN motifs (-3GalNAcβ1–4GlcNAcβ1-)n. These novel structures are in contrast to the well known poly-LN structures consisting of repeating LN motifs (-3Galβ1–4GlcNAcβ1-)n. We also stably expressed human α1,3-fucosyltransferase IX in the L8-GalNAcT cells to establish a new cell line, L8-GalNAcT-FucT. These cells produce complex-type N-glycans with α1,3-fucosylated LDN (LDNF) GalNAcβ1–4(Fucα1–3)GlcNAcβ1-R as well as novel “poly-LDNF” structures (-3GalNAcβ1–4(Fucα 1–3)GlcNAcβ1-)n. The ability of these cell lines to generate glycoprotein hormones with LDN-containing N-glycans was studied by expressing a recombinant form of the common α-subunit in L8-GalNAcT cells. The α-subunit N-glycans carried LDN structures, which were further modified by co-expression of the human GalNAc 4-sulfotransferase I, which generates SO4-4GalNAcβ1–4GlcNAc-R. Thus, the generation of these stable mammalian cells will facilitate future studies on the biological activities and properties of LDN-related structures in glycoproteins.

There is a growing appreciation of the important roles of glycoconjugates in modulating a wide variety of biological processes. A critical limiting factor in exploring glycan functions is the difficulty of obtaining many glycans either in a purified chemical form or within a relatively homogenous biological system. A common nonreducing terminal modification of glycans in mammalian glycoproteins is the N-acetyllactosamine (LN) 1 disaccharide Gal␤1-4GlcNAc-R, which may be sialylated, fucosylated, sulfated, or modified by other sugars to generate a wide range of terminal structures. However, another type of nonreducing terminal glycan structure that is less well understood but now appears to also be expressed by many organisms, including mammals, is based on the LacdiNAc (LDN) sequence GalNAc␤1-4GlcNAc-R, which can also occur within 4-O-sulfated, ␣1,3-fucosylated, or ␣2,6-sialylated derivatives. LDN-type glycans play vital roles in regulating the circulatory half-life of pituitary glycoprotein hormones (1)(2)(3) and other glycoproteins (4,5), including tenascin-R produced by oligodendrocytes and small interneurons in the hippocampus and cerebellum (6). Other glycoproteins containing LDN-type glycans include human glycodelin, a human glycoprotein with potent immunosuppressive and contraceptive activities (7,8), and zona pellucida glycoproteins from murine eggs (9). Many human pathogens synthesize LDN and fucosylated LDN glycans, such as GalNAc␤1-4(Fuc␣1-3)Glc-NAc-R, termed LDNF. Both LDN and LDNF as well as other modified LDN-type glycans are important determinants recognizable by our adaptive immune system and various carbohydrate-binding proteins within the innate immune system, such as DC-SIGN (10 -17).
Although some LDN-based structures have been synthesized in vitro, there is no convenient biological system available to produce these structures in vivo to facilitate studies of their biological properties, and little is understood about the relationship of LDN expression to protein glycosylation in general. We recently identified a cDNA encoding the Caenorhabditis elegans ␤1,4-N-acetylgalactosaminyltransferase (Ce␤4GalNAcT) (18) and found that the recombinant Ce␤4GalNAcT was active and capable of promoting LDN synthesis in vitro. Here we have explored whether mamma-lian cells stably expressing the Ce␤4GalNAcT can be used to generate LDN-type N-glycans on cellular and recombinant glycoproteins. To this end, we generated a cell line derived from CHO Lec8 cells, termed L8-GalNAcT, that expresses Ce␤4GalNAcT and produces complex N-glycans containing LDN structures on their antennae. CHO Lec8 cells were chosen because they lack a functional UDP-Gal transporter and synthesize complex N-glycans with terminal ␤-linked GlcNAc residues (19,20). Unexpectedly, L8-GalNAcT cells also produced novel poly-LDN-type structures with the repeating LDN structure (-3GalNAc␤1-4GlcNAc␤1-) n , indicating that at least one member of the mammalian ␤1,3-Nacetylglucosaminyltransferase family, which is responsible for poly-N-acetyllactosamine synthesis through the addition of ␤3-linked GlcNAc to terminal ␤4-linked Gal residues, also recognizes terminal ␤4-linked GalNAc to allow formation of the poly-LDN structures. Furthermore, we found that the poly-LDN backbone was efficiently fucosylated by recombinant human ␣1,3-fucosyltransferase IX (FucT 9) when it was stably co-expressed with the Ce␤4GalNAcT in a new cell line, termed L8-GalNAcT-FucT. The ability to generate such novel LDN-related glycans should aid in future studies to define the biological roles of LDN-containing glycans in a variety of biological systems, including bioactive glycoproteins in humans, such as pituitary hormones and fertility-related glycoproteins like glycodelin as well as in human immune responses to parasitic infections.

EXPERIMENTAL PROCEDURES
Materials-All chemicals and reagents used in this study, unless otherwise indicated, were from Sigma. FuGENE 6 and Complete Protease Inhibitor Mixture were from Roche Applied Science. Geneticin and zeocin were from Invitrogen. HighSignal West Pico Chemiluminescent Substrate was from Pierce. Radiolabeled nucleotide sugars were from PerkinElmer Life Sciences. GalNAc␤1-4GlcNAc␤-S-pNP was synthesized from GlcNAc␤-S-pNP using Ce␤4GalNAcT as previously described (18). The plasmid encoding full-length human GalNAc 4-sulfotransferase I was a kind gift from Dr. Jacques Baenziger.
Establishment of Cell Lines-CHO Lec8 cells (19,20) were transfected with a plasmid encoding the complete open reading frame of Ce␤4GalNAcT under the control of the CMV promoter; this plasmid also encoded a Geneticin resistance gene. The transfection was carried out using FuGENE 6 according to the manufacturer's instructions, and the cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal calf serum and 600 g/ml Geneticin to select for stably transformed cells. After 4 weeks in culture, a cell line expressing the Ce␤4GalNAcT activity (L8-GalNAcT) was cloned from a single cell by limiting-dilution culturing.
Next, L8-GalNAcT cells were transfected, as above, with a plasmid encoding the complete open reading frame of human FucT 9 under the control of the CMV promoter; this plasmid also encoded a zeocin resistance gene. The cells were then cultured in Dulbecco's modified Eagle's medium containing 10% fetal calf serum, 600 g/ml Geneticin, and 400 g/ml zeocin to select for stably transformed cells. After 4 weeks in culture, a cell line expressing GalNAcT and FucT activities (L8-Gal-NAcT-FucT) was cloned from a single cell as above.
Glycosyltransferase Assays-Confluent cells grown in T-25 culture flasks were rinsed twice with phosphate-buffered saline and extracted in 1.25 ml/flask of 100 mM sodium cacodylate, pH 7.0, 1.5% Triton X-100, 1ϫ Complete Protease Inhibitor Mixture (EDTA-free). The extracts were centrifuged at 21,000 ϫ g for 4 min, and the supernatants were used for further analyses. Glycosyltransferase assays were carried out for 1 h at room temperature in duplicate using extracts equivalent to 15,000 cells/reaction in a final volume of 50 l of 50 mM sodium cacodylate, pH 7.0, 0.15% Triton X-100, 5 mM ATP. GalT and GalNAcT assays contained 20 mM MnCl 2 , 1 mM GlcNAc␤-S-pNP, 50 M UDP-Gal or UDP-GalNAc, and 0.1 Ci/reaction of UDP-[ 3 H]Gal or UDP-[ 3 H]Gal-NAc. Fucosyltransferase assays contained 1 mM GalNAc␤1-4GlcNAc␤-S-pNP, 50 M GDP-Fuc, and 0.1 Ci/reaction of GDP-[ 3 H]Fuc. The reactions were then diluted with 1 ml of water, and the products were isolated using 1 cm 3 Sep-Pak C-18 cartridges (Waters) and assayed for incorporation of radioactivity by liquid scintillation as described (21).
SDS-PAGE and Western Blot Analysis-Cell extracts were mixed with loading buffer, resolved by SDS-PAGE (4 -20% gradient), and transferred to a nitrocellulose membrane. For Western blot analysis, the membrane was blocked with 5% bovine serum albumin in a buffer of 20 mM Tris-HCl, pH 7.2, 150 mM NaCl, 2 mM CaCl 2 , 0.05% Tween 20 for 5 h at 4°C. It was then incubated with primary antibody (mouse monoclonal anti-LDN IgM SMLDN1.1 or anti-LDNF IgM SMLDNF (11)) in the same buffer (without bovine serum albumin) for 1 h at room temperature. The membrane was then washed in the same buffer, and incubated with the secondary antibody (horseradish peroxidase-conjugated, goat anti-mouse IgM) as before. The membrane was then washed again, incubated in HighSignal West Pico Chemiluminescent Substrate for 2 min at room temperature, and exposed to a BioMax film (Eastman Kodak Co.) for 1 min. The film was then developed using a processing machine (Konica SRX-101). For Ponceau S staining, the membrane was rinsed in water and stained for 3 min with 0.2% Ponceau S in 3% trichloroacetic acid. The membrane was then rinsed again in water and air-dried. Preparation of N-Glycans and MALDI-TOF MS Analysis-N-Glycans were prepared from 3 ϫ 10 7 cells of each cell line as previously described (22). Briefly, proteins extracted from washed cell pellets were reduced and carboxymethylated and then digested with L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated trypsin. The tryptic peptides were then treated with peptide N-glycosidase F, and the released N-glycans were purified and permethylated. MALDI data were acquired using a Perseptive Biosystems Voyager-DETM STR mass spectrometer in the reflectron mode with delayed extraction. Derivatized glycans were dissolved in 10 l of methanol, and 1 l of dissolved sample was premixed with 1 l of matrix (2,5-dihydrobenzoic acid) before loading onto a target plate.
GC-MS Linkage Analysis-Partially methylated alditol acetates were prepared from permethylated samples for gas chromatography-MS linkage analysis. Linkage analysis was carried out on a PerkinElmer Clarus 500 instrument fitted with an RTX-5 fused silica capillary column (30 m ϫ 0.32-mm internal diameter; Restek Corp.). The sample was dissolved in hexanes and injected onto the column at 65°C. The column was maintained at this temperature for 1 min and then heated to 290°C at a rate of 8°C/min.
Expression and Analysis of the Glycoprotein Hormone ␣-Subunit-A plasmid encoding an HPC4-tagged version of the human glycoprotein hormone ␣-subunit under the control of the CMV promoter was constructed. A PCR-amplified DNA fragment starting at bp 72 of the open reading frame of the human ␣-subunit and extending beyond the stop codon was subcloned into the BamHI site of the pcDNA 3.1(ϩ)-TH vector. The resulting vector (pHPC4-␣-subunit) encodes a fusion protein consisting of a signal peptide at the N terminus, followed by an HPC4 linear peptide epitope (23,24), which is followed by the mature peptide of the ␣-subunit (beginning at Ala 25 , the first amino acid after the endogenous, cleavable, signal peptide). The mature HPC4-tagged ␣-subunit protein (after the removal of the signal peptide) has a calculated molecular mass of 11,785.01 Da, but the mature protein contains two N-glycosylation sites and migrates anomalously on reducing SDS-PAGE as a glycoprotein with an apparent molecular mass of ϳ25 kDa (25). CHO Lec8 and L8-GalNAcT cells were transfected with pHPC4-␣-subunit, with or without co-transfection with a plasmid encoding full-length human GalNAc 4-sulfotransferase I (26). The cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal calf serum for 36 h post-transfection. The ␣-subunit was then affinity-purified from the extracellular medium and analyzed by SDS-PAGE and Western blotting using the HPC4 monoclonal antibody, as previously described (18).

Establishment of L8-GalNAcT and L8-GalNAcT-FucT Cell
Lines-CHO Lec8 cells were chosen as founder cells for a new cell line to explore the expression of glycans with the LDN motif GalNAc␤1-4GlcNAc␤1-R. The CHO Lec8 cells lack a functional UDP-Gal transporter; thus, they are expected to synthesize complex N-glycans with terminal ␤-linked GlcNAc residues at the nonreducing termini of their antennae (19,20). The recombinant Ce␤4GalNAcT has ␤1,4-N-acetylgalactosaminyltransferase (GalNAcT) activity toward GlcNAc-terminating glycans (18). CHO Lec8 cells were transfected with a plasmid encoding Ce␤4GalNAcT under the control of the CMV promoter as well as a Geneticin resistance gene. A stable, clonal cell line (L8-GalNAcT) was established from these cells by antibiotic selection and limiting-dilution culturing as described under "Experimental Procedures." We also wanted to explore the possibility that the Lec8-GalN-AcT cells could be used for stably producing N-glycans containing the LDNF motif GalNAc␤1-4(Fuc␣1-3)GlcNAc␤1-R, which is a common sequence found in parasite-derived glycans and is recognized by the human dendritic cell lectin DC-SIGN (14). To this end, we generated an additional cell line, L8-GalNAcT-FucT, by transfecting L8-GalNAcT cells with a plasmid encoding human FucT 9 under the control of the CMV promoter as well as a zeocin resistance gene. FucT 9 acts efficiently on the acceptor glycans with the LN sequence Gal␤1-4GlcNAc-R to generate the Lewis x motif Gal␤1-4(Fuc␣1-3)GlcNAc-R (27). Recombinant FucT 9 also shows activity in vitro toward the LDN acceptor GalNAc␤1-4GlcNAc␤1-R to generate the LDNF motif GalNAc␤1-4(Fuc␣1-3)GlcNAc␤1-R (28).
Both these cell lines retain their newly integrated genetic material even after several months of culturing in the absence of antibiotics, as judged by their enduring resistance to the reintroduction of these selection agents and by the retention of the introduced glycosyltransferase activities.
Glycosyltransferase Activities in L8-GalNAcT and L8-GalN-AcT-FucT Cells-The newly established L8-GalNAcT and L8-GalNAcT-FucT cell lines were assayed to assess whether they effectively express the glycosyltransferase activities of the introduced cDNAs. Cell extracts from the new cell lines and from the parental CHO Lec8 cells were assayed for GalT and GalNAcT activities acting on the acceptor GlcNAc␤1-pNP and for fucosyltransferase (FucT) activity acting on the acceptor GalNAc␤1-4GlcNAc␤1-pNP. All three cell lines have comparable levels of GalT activity (Fig. 1). It should be noted that CHO Lec8 glycans lack Gal residues due to a deficiency in UDP-Gal transport and not in GalT activities (19,20). By contrast, CHO Lec8 cells lack GalNAcT activity, whereas both L8-GalNAcT and L8-GalNAcT-FucT cells have GalNAcT activity levels substantially higher than their endogenous GalT activity. Furthermore, L8-GalNAcT-FucT cells, but not the other cell lines, contain substantial FucT activity in addition to GalNAcT activity.
The Presence of LDN and LDNF Structures on Endogenous Glycans of L8-GalNAcT and L8-GalNAcT-FucT Cells-We examined whether the new cell lines produce N-glycans containing LDN and LDNF on endogenous glycoproteins. Cell extracts from the two cell lines and from the parental CHO Lec8 cells were resolved by SDS-PAGE and probed for the presence of LDN and LDNF determinants by Western blot analysis using monoclonal antibodies that specifically bind to each of these structures (Fig. 2). In agreement with the results of the glycosyltransferase activity assays and with previous observations (18,29), CHO Lec8 glycoproteins lack detectable levels of LDN or LDNF structures. By contrast, we observed strong staining of extracts from L8-GalNAcT cells with anti-LDN but only trace staining with the anti-LDNF antibody, demonstrating that these cells express LDN structures on many glycoproteins.
In contrast to the results with L8-GalNAcT cells, anti-LDNF strongly stained glycoproteins from L8-GalNAcT-FucT cells, and there was a decrease in staining with anti-LDN. These results suggest that the LDN structures found in L8-GalNAcT cells were largely converted to LDNF structures in L8-GalN-AcT-FucT cells and that the LDNF antigen is expressed on many different glycoproteins.
We expect that the LDN and LDNF structures in the new cell lines should be expressed on N-rather than O-glycans, as we have previously demonstrated in transiently transfected CHO-Lec8 cells (18). This is to be predicted, since CHO-Lec8 cells produce O-glycans consisting primarily of GalNAc␣1Ser/Thr (30), which may or may not be ␣2-6-sialylated (31).
Mass Spectrometric Analysis of N-Glycans Produced by L8-GalNAcT and L8-GalNAcT-FucT Cells-To obtain a more detailed view of the carbohydrate structures produced by the cell lines in this study, MALDI-TOF-MS analyses were carried out on N-glycans released from the endogenous glycoproteins of these cells ( Fig. 3 and Table I) Taken together, these data demonstrate that expression of GalNAcT in CHO-Lec8 cells allows for the production of complex N-glycans with LDN and poly-LDN antennae and that co-expression of GalNAcT and FucT allows for the production of N-glycans with LDNF and poly-LDNF antennae (Fig. 4).

Expression of a Recombinant ␣-Subunit of the Glycoprotein Hormones and GalNAc 4-Sulfotransferase in L8-GalNAcT
Cells-We then asked whether the recombinant GalNAcT in L8-GalNAcT cells could act on the common ␣-subunit of human glycoprotein hormones, which normally contains two complex-type N-glycans. These N-glycans from pituitary hormones are typically characterized by the presence of the terminal sulfated sequence SO 4 -4GalNAc␤1-4GlcNAc␤1-2Man␣-R (32,33). The addition of the ␤1-4-linked GalNAc to the GlcNAc␤1-2Man␣-R acceptor to generate the LDN determinant may be catalyzed by a GalNAcT that utilizes UDP-GalNAc (34) and may specifically recognize N-glycans on glycoprotein hormones. The LDN sequence is then sulfated by a glycoprotein hormone GalNAc-4-sulfotransferase that utilizes the donor phosphoadenosine phosphosulfate (26,35).
We transiently transfected both parental CHO Lec8 cells and L8-GalNAcT cells with a cDNA encoding an N-terminal HPC4 epitope-tagged form of the ␣-subunit, as described under "Experimental Procedures." It has been shown that the free ␣-subunit expressed in CHO cells acquires an O-glycan, whereas this is not normally observed when the ␣-subunit is co-expressed with specific ␤-subunits of glycoprotein hormones (36). Medium from the transfected cells was recovered and the secreted HPC4-tagged ␣-subunit was purified and analyzed by Western blot. The recombinant ␣-subunit synthesized by parental CHO Lec8 cells migrated as a broad band with an apparent approximate molecular mass of ϳ25 kDa (Fig. 5), which is similar in mobility to the free ␣-subunit observed for recombinant ␣-subunit expressed in insect cells (25), which have some similarity to Lec8 CHO cells in being unable to terminally glycosylate either N-or O-glycans. By contrast, the recombinant ␣-subunit expressed in L8-GalNAcT cells migrated as a broad band with higher apparent molecular mass extending to ϳ32 kDa (Fig. 5). Western blotting with anti-LDN showed that the ␣-subunit from parental CHO Lec8 cells lacks LDN determinants as expected, whereas the ␣-subunit expressed in L8-GalNAcT cells is well recognized by anti-LDN (data not shown). Thus, the ␣-subunit can be modified by the GalNAcT in L8-GalNAcT cells to generate the LDN determinant. The increase in apparent M r of the recombinant ␣-subunit from L8-GalNAcT cells suggests that multiple sugar residues were added, consistent with the presence of poly-LDN structures, such as those observed in the analysis of the total N-glycans. We further explored the N-glycosylation and potential sulfation of the ␣-subunit, by co-expressing the ␣-subunit in L8-GalNAcT cells with a recombinant form of the human GalNAc-4-sulfotransferase (GalNAc4-ST) that recognizes terminal LDN sequences on the N-glycans of pituitary hormones (26,35). Co-expression of the GalNAc4-ST with the ␣-subunit in L8-GalNAcT cells caused a dramatic decrease in its apparent M r and sharpening of its mobility on SDS-PAGE (Fig. 5), which is probably due to a truncation of poly-LDN structures by sulfation of terminal GalNAc residues within LDN. Only a small amount of recombinant ␣-subunit was generated, and further extensive studies will be required to purify the protein from multiple cell lines and sequence the N-glycans. The studies demonstrate that the N-glycans of the common ␣-subunit of pituitary hormones can be modified with LDN and possibly poly-LDN sequences and that co-expression with the GalNAc4-ST alters the biosynthesis of the LDN-containing N-glycans. DISCUSSION The analyses of the N-glycans from L8-GalNAcT and L8-GalNAcT-FucT cells indicate that the terminal GlcNAc residues found in CHO Lec8 cells are converted into LDN or LDNF structures in L8-GalNAcT and L8-GalNAcT-FucT cells, respectively. Unexpectedly, we found that L8-GalNAcT and L8-Gal-NAcT-FucT cells contain N-glycans bearing repeating poly-LDN or poly-LDNF units, respectively. N-Glycans from CHO Lec8 cells are approximately equally distributed between bi-, tri-, and tetraantennary species (Figs. 3A and 4A), represented by the glycans with m/z of 1836.5, 2081.6, and 2326.6, respectively. The largest N-glycan mass detected in L8-GalNAcT cells indicates that it contains at least seven repeating LDN units, if the glycan is assumed to be a tetraantennary structure, and up to nine repeating units, if the glycan is assumed to be a biantennary structure. The largest N-glycan mass detected in L8-GalNAcT-FucT cells indicates 4 -6 repeating LDNF units, depending on the assumed number of antennae. To our knowledge, this is the first reported occurrence of poly-LDN or poly-LDNF glycan structures biosynthesized by cells.
The presence of repeating units in both cell lines is remarkable and demonstrates that the endogenous ␤1,3-N-acetylglucosaminyltransferase, or i-extension enzyme(s) (iGn-T), can act on LDN structures in vivo. Several ␤1,3-N-acetylglucosaminyltransferases capable of elongating LN units and generating poly-LN structures have been identified, and these include iGn-T (37) and two different enzymes termed ␤3Gn-T1 and ␤3Gn-T2 (38). All of these enzymes can act in vitro and in vivo on terminal LN units to generate poly-LN structures. However, none of these studies on recombinant forms of the ␤1,3-N-acetylglucosaminyltransferases examined their activity toward acceptors with nonreducing terminal LDN rather than terminal LN structures. Several independent ␤3Gn-T gene families have been described (39), and the role of individual members or even as yet unidentified ␤3Gn-Ts in modification of terminal LDN versus terminal LN acceptors is not known. At least one of the ␤1,3-N-acetylglucosaminyltransferases may act on terminal LDN structures, as shown by studies on a human serum-derived preparation of ␤1,3-N-acetylglucosaminyltransferase(s). This preparation demonstrated activity toward acceptors with terminal LDN structures (40), although the activity was not directly compared between the LDN versus LN acceptors. In a recent study (40), it was shown that the serum-derived ␤1,3-Nacetylglucosaminyltransferase(s) along with bovine ␤4GalT could be used in vitro to synthesize the pentasaccharide GalNAc␤1-4GlcNAc␤1-3GalNAc␤1-4GlcNAc␤1-3Gal␤1-Omethyl. It was also demonstrated that a partially purified recombinant form of human FucT 4 could add ␣1,3-linked fucose residues to internal GlcNAc residues within the LDN structure (40). Clearly, the potential roles of ␤3Gn-T gene family members in modifying LDN and/or LN and allowing for poly-LDN and poly-LN structures need to be further investigated.
The likely pathways for LDN, poly-LDN, and poly-LDNF synthesis in the newly created cell lines are depicted in Fig. 6 (top), whereas the contrasting biosynthetic pathway for LN, poly-LN, and poly-Le x is depicted in Fig. 6 (bottom). This scheme is also supported by previous studies on enzyme activities in cell-free systems (28). The LN and/or LDN sequences could be generated on N-glycans, as shown here, or on O-glycans. In either case, fucosylation could occur on a newly synthesized LN or LDN unit, which can be elongated by the sequential actions of ␤3Glc-NAcT(s) and either ␤4GalT (Fig. 6, bottom) or ␤4GalNAcT (Fig. 6,  top), respectively. As the chain grows through the action of these two sets of enzymes, the chain is fucosylated, thereby allowing the production of large sized polyfucosylated structures, such as poly-Le x or poly-LDNF.
In most cells, the addition of terminal Gal residues by endogenous GalTs supercedes the activity of endogenous Gal-NAcT(s), which is generally low to undetectable in most mammalian cell lines and tissues (41). Thus, whereas expression of GalNAcT may allow some generation of LDN-containing structures, the stronger expression of GalT promotes the synthesis of predominantly LN-containing glycans. An exception could be the GalNAcT that recognizes pituitary hormones, which might allow a shift to LDNtype glycans in this subset of glycoproteins. In any case, the balance between the LDN versus LN pathways may be strongly dependent on the activity levels of the GalTs versus GalNAcTs. This general idea of a possible dichotomy in biosynthetic pathways in LDN versus LN has been suggested previously in studies on N-glycosylation pathways in invertebrates (42,43).
Most studies on mammalian glycoproteins have reported that the LDN-type glycans are modified to contain the terminal sequence NeuAc␣2-6GalNAc␤1-4GlcNAc-R or SO 4 -4Gal-NAc␤1-4GlcNAc␤1-2Man␣-R, generated by an ␣2,6-sialyl-transferase or a GalNAc4-ST, respectively (42,44). Thus, sialylation or sulfation of newly generated LDN structures would probably act as chain terminators and block glycan elongation by ␤1,3-N-acetylglucosaminyltransferase(s), thus limiting the generation of poly-LDN-type glycans. CHO cells do not express either the ␣2,6-sialyltransferase (45) acting on LN acceptors or a GalNAc4-ST acting on terminal GlcNAc acceptors (present study); thus, there are no competitive chain termination events to block poly-LDN synthesis in these cells. Interestingly, the bovine ␣2,6-sialyltransferase has been shown to act on both terminal LN and LDN acceptors (46,47). We speculate that the recombinant ␣-subunit expressed in L8-GalNAcT cells acquires poly-LDN structures, like other common glycoproteins in these cells, causing its apparent M r to be anomalously high, whereas co-expression with the recombinant GalNAc4-ST caused a truncation and termination of poly-LDN synthesis, causing a decrease in the apparent M r of the recombinant ␣-subunit (Fig.  5). Studies are in progress to define all of the N-glycans structures from the ␣-subunit expressed in these cells and the precise effects of 4-O-sulfation on LDN and LDNF biosynthesis. The presence of poly-LN in animal cell glycoproteins is well described and can occur in both N-and O-glycans (48) as well as in glycosphingolipids (49,50). However, there have so far been no reports of poly-LDN-type structures synthesized by animal cells, except for the structures defined here. There may be several reasons for why these structures have thus far failed to be detected. Structural analyses of any polysaccharide, and especially poly-LDN, can be especially difficult. A common technique to characterize large sized glycans in animal cells is to probe for the presence of poly-LN sequences using the bacterial endo-␤-galactosidase from Bacillus fragilis (51) or Escherichia freundii (52), both of which cleave poly-LN structures at internal Gal␤1-4 linkages to release fragments with a reducing Gal residue. However, the actions of these enzymes toward poly-LDN-containing glycans are not well defined. Only one study has reported such an analysis, where it was demonstrated that the semisynthetic glycan Gal␤1-4GlcNAc␤1-3GalNAc␤1-4Glc-R, which has an internal LDN motif, can be cleaved at the internal GalNAc residue by the B. fragilis endo-␤-galactosidase, although the glycan was clearly less sensitive to the enzyme compared with a conventional poly-LN structure Gal␤1-4GlcNAc␤1-3Gal␤1-4Glc-R (40), where cleavage at the internal Gal residue was quantitative. It is now possible, using N-glycans from the cell lines described here to perform detailed studies on the sensitivities of different endo-␤-galactosidases to poly-LDN and poly-LDNF structures.
One of the major cell types used in studies on LDN-type glycans is the human 293 embryonic kidney cell line (HEK-293), whereas CHO cells are often used as cells expressing only LN and not LDN-type glycans. Both HEK-293 and CHO cells have demonstrable ␤4GalNAcT activity toward acceptors with terminal, nonreducing GlcNAc (8,29), although HEK-293 cells have a somewhat higher level of activity (8). However, HEK-293 cells also have much higher levels of ␤4GalT activity toward the same acceptors, yet surprisingly some recombinant glycoproteins expressed in HEK-293 cells acquire significant amounts of LDN-related structures (8,53). Similar results were seen for canine kidney Madin-Darby canine kidney cells, where secreted glycoproteins contain appreciable amounts of LDN-related glycans (54). Furthermore, several specific glycoproteins were found to contain glycans with both LN and LDN motifs (7-9, 54 -57). Interestingly, in such studies so far the glycans examined lack either poly-LN or poly-LDN-type structures, which suggests that the cells generating those glycoproteins may lack iGn-T activities or that the glycans are not accessible to the iGn-T activities, either through compartmentalization or protein folding. In addition, there is evidence that the mammalian ␤4GalNAcTs may exhibit preference for certain glycan branches (34,58) or be glycoprotein-specific, as shown for the pituitary ␤4GalNAcT acting on pituitary-specific glycoprotein hormones (59), thereby providing a mechanism for limiting LDN-type structures to a few glycoproteins. By contrast, the C. elegans ␤4GalNAcT appears to lack acceptor specificity in vitro (18) and act on all branches of N-glycans. Clearly, further studies are needed to define the acceptor and glycoprotein specificity of the different mammalian ␤4GalNAcTs. As our study demonstrates, CHO Lec8 cells are a useful system in which to explore the specificity of ␤4GalNAcTs, since there is no acceptor competition with endogenous ␤4GalTs.
Previous studies on recombinant human FucT 9 showed that it acts on terminal LN units to generate a terminal Lewis x type structure and on terminal LDN units to generate the LDNF structure (60). However, the enzyme does not efficiently recognize internal GlcNAc residues within acceptors containing poly-LN motifs, thus indicating that FucT 9 strongly prefers the terminal, nonreducing end (28). Thus, based on this observation, it appeared unlikely that human FucT 9 could participate in the generating of polyfucosylated oligosaccharides. However, it was suggested that human FucT 9 may allow expression of polyfucosylated poly-LN by fucosylating the penultimate GlcNAc residues within growing poly-LN units, where fucosylation precedes the addition of an extending ␤1-3 GlcNAc to the terminal Gal residue (28). No previous studies have examined the specificity of human FucT 9 toward poly-LDN-type acceptors. Our results suggest that FucT 9 efficiently acts on LDN-type structures to allow formation of long chain poly-LDNF. This observation might suggest that FucT 9 fucosylates internal GlcNAc residues at the nonreducing end of growing poly-LDN chains, as predicted for the poly-LN chains (28).
Although there are a few cell lines capable of generating glycans containing LDN and LDNF structures (8,29,53,54,61), the glycosylation is highly heterogeneous and mixed structures of both LDN and LN are generated. By contrast, L8-GalNAcT and L8-GalNAcT-FucT cells are the first cell lines to predictably and exclusively produce LDN or LDNF structures on their complex N-glycans. The availability of these novel cell lines should greatly enhance the investigation into the biological roles of LDN-and LDNF-based N-glycan structures by providing a valuable tool to produce endogenous or recombinant glycoproteins bearing them. In addition, the cells themselves, fixed or otherwise, could be used as antigens for vaccination to produce anti-LDN/LDNF antibodies either for experimentation or as protection against LDN/LDNF-bearing pathogens (12).