Expression of human H-type alpha1,2-fucosyltransferase encoding for blood group H(O) antigen in Chinese hamster ovary cells. Evidence for preferential fucosylation and truncation of polylactosamine sequences.

The human H(O) blood group is specified by the structure Fuca1–2Galb1-R, but the factors regulating expression of this determinant on cell surface glycoconjugates are not well understood. To learn more about the regulation of H blood group expression, cDNA encoding the human H-type GDPFuc:b-D-galactoside a1,2-fucosyltransferase (a1,2FT) was stably transfected into Chinese hamster ovary (CHO) cells. The new cell line, designated CHO(a1,2)FT, expressed surface neoglycans containing the H antigen. The structures of the fucosylated neoglycans in CHO(a1,2)FT cells and the distribution of these glycans on glycoproteins were characterized. Seventeen percent of the [H]Gal-labeled glycopeptides from CHO(a1,2)FT cells bound to the immobilized H blood group-specific lectin Ulex europaeus agglutinin-I (UEAI), whereas none from parental CHO cells bound to the lectin. The glycopeptides from CHO(a1,2)FT cells binding to UEA-I contained polylactosamine [3Galb1–4GlcNAcb1-] with the terminal sequence Fuca1–2Galb1– 4GlcNAc-R. Fucosylation of the polylactosamine sequences on complex-type N-glycans in CHO(a1,2)FT cells caused a decrease in both sialylation and length of polylactosamine. Unexpectedly, only small amounts of terminal fucosylation was found in diantennary complex-typeN-glycans. TheO-glycans and glycolipids were not fucosylated by the H-type a1,2FT. Two major high molecular weight glycoproteins, one of which was shown to be the lysosome-associated membrane glycoprotein LAMP-1, preferentially contained the H-type structure and were bound by immobilized UEA-I. These results demonstrate that in CHO cells the expressed H-type a1,2FT does not indiscriminately fucosylate terminal galactosyl residues in complex-type N-glycans, but it favors glycans containing polylactosamine and dramatically alters their length and sialylation.

The A and B blood group antigens and the Lewis B and Y antigens are formed by the action of specific glycosyltransferases, which act on a precursor oligosaccharide containing the H blood group structure Fuc␣1-2Gal␤1-3(4)GlcNAc␤1-R (1,2). This precursor H-substance can be synthesized by either of two GDPFuc:␤-D-galactoside ␣1,2-fucosyltransferases (␣1,2FT). 1 One enzyme, designated the H-type, is the product of the human H blood group locus (3,4) and the other enzyme, designated the SE-type, is a product of the Secretor locus (5,6).
Although H blood group determinants are expressed on some glycoproteins, such as band-3 glycoprotein of erythrocytes (7) and respiratory mucins (8 -10), the expression of H antigens on glycoproteins appears to be restricted, since relatively few glycoproteins contain significant amounts of the determinant. Because the H-type ␣1,2FT can add fucose to a wide variety of simple acceptors in vitro, including phenyl-␤-D-galactose (11) and small glycans terminating in the lactosamine sequence Gal␤1-3(4)GlcNAc-R (12), it would seem possible for H antigens to occur indiscriminately on glycoproteins, since lactosamine sequences are commonly found on mammalian glycoproteins (13). There has not been a systematic study, however, about the specificity of the H-type ␣1,2FT toward complex glycoconjugates and the consequences of the ␣1,2-fucosylation on overall glycoconjugate biosynthesis in cells.
Construction of CHO(␣1,2)FT Cells-Plasmid pCDM7 (␣1,2)FT (3) was linearized by digestion at a unique NheI site within the M13 origin of replication in the vector pCDM7. The linearized DNA was transfected with the G418 resistant plasmid pSV2Neo, with a 10:1 molar ratio of linearized pCDM7 (␣1,2)FT and pSV2Neo, using the calcium phosphate transfection procedure of Chen and Okayama (21). Clonal cell lines were derived from within the G418 resistant transfectant population using cloning cylinders. Clonal transfectant cell lines analyzed by flow cytometry for expression of cell surface blood group H antigen, using a monoclonal anti-H antibody and procedures described previously (22,23). A representative clonal line that stably expressed cell surface blood group H-determinants was termed CHO(␣1,2)FT. Construction of a control CHO cell line (CHO-V cells) stably transfected with the vector pCDM7 has been described previously (23). CHO cells lacking expressed ␣1,2FT were designated parental CHO cells.
Assay for ␣1,2FT-Extracts were prepared from CHO cell lines by homogenizing washed cell pellets in 1% Triton X-100, as described previously (24). Reaction mixtures for assaying ␣1,2FT activity contained cell extracts and 3 M GDP-[ 14 C]fucose, 25 mM phenyl-␤-Dgalactoside, and 5 mM ATP, in 25 mM sodium phosphate, pH 6.1. Reactions were conducted at 37°C for a period of time allowing a linear rate of incorporation of radiolabeled fucose, which was generally less than 1% of the counts incorporated into the added artificial acceptor (22).
Flow Cytometry-CHO-V or CHO␣(1,2)FT cells were subjected to flow cytometry analysis, using mouse monoclonal IgM antibodies (22,25). Cells were stained with a monoclonal anti-H antibody (Chembiomed, 10 g/ml) or a control anti-Le x monoclonal antibody (anti-SSEA-1, Citation, 1:25 dilution). The cells were then washed, and stained with fluorescein conjugated goat anti-mouse IgM (Sigma, 40 g/ml), and subjected to analysis by flow cytometry on FACScan (Becton-Dickinson, Mountain View, CA), as described previously (22,26). Cell staining was measured in arbitrary fluorescent intensity units and displayed on a four-decade log scale.
Lectin Agglutination Assays-Cells were removed from dishes by treatment with 0.25% trypsin solution containing 2 mM EDTA and the cells were washed three times in Tris buffer (50 mM Tris, 1 mM CaCl 2 , 1 mM MgCl 2 , pH 8.0). Cell suspensions for analyses were prepared by gently suspending cells to a concentration of 3 ϫ 10 6 cells/ml. Aliquots (10 l) of cell suspensions were placed on microscope slides, and an equal volume of different concentrations of lectin solutions (1 mg/ml or 0.1 mg/ml) were added. The microscope slides were gently shaken and agglutination was determined by direct and microscope observation 3 min after addition of the lectin. Immediate agglutination was defined as 3ϩ, an intermediate level of agglutination was 2ϩ, perceptible but weak level of agglutination was 1ϩ, and lack of agglutination was 0 .
Preparation of Metabolically Radiolabeled Glycoconjugates from Cultured Cells-Plates of cells (60 mm) at 50% confluency were incubated in 2 ml of the corresponding medium containing 1 mCi of [ 3 H]galactose or [ 3 H]glucosamine for 48 h. Medium was removed from incubation dishes and the cells were gently washed in the plates three times with 1-ml portions of cold PBS. Radiolabeled cells were scraped from the culture plates with a rubber policeman, and cell suspensions in cold PBS were gently aspirated with sterile pipettes and transferred to 1-ml centrifuge tubes. After washing three times in cold PBS, the cell pellets were collected by centrifugation. Metabolically labeled glycopeptides and glycolipids from the same cell pellets were separately obtained essentially according to the method of Finne and Krusius (29), modified as described previously (15).
Glycolipids were applied to aluminum-backed Silica Gel 60 10 ϫ 10-cm plates (Merck); lower phases were developed in CHCl 3 :CH 3 OH: H 2 O (65:25:4) and upper phase glycolipids in CHCl 3 :CH 3 OH:0.25% KCl (5:4:1). Radiolabeled glycolipids were detected on thin layer chromatograms by autoradiography (15). The pelleted residues obtained after extraction of radiolabeled glycolipids were washed with ethanol to remove excess organic solvents and resuspended in 0.1 M Tris, 1 mM CaCl 2 , pH 8.0, containing Pronase (10 mg/ml). After incubation for 18 h at 60°C in a toluene atmosphere, the sample was boiled 5 min and desalted by passage over a column (1 ϫ 50 cm) of Sephadex G 25-80 in 7% n-propanol. The radiolabeled glycopeptides were recovered from the void factions, pooled, and dried in a shaker bath evaporator under reduced pressure.
Chromatography of Glycoconjugates-Radiolabeled glycopeptides were fractionated on ConA-Sepharose as described (30). Ulex europaeus agglutinin-I was immobilized on Affi-Gel-10 according to the manufacturer's instructions using 4 mg/ml free fucose in the coupling reaction to protect activity of the lectin. The final coupling density was approximately 10 mg/ml. For affinity chromatography of glycopeptides or free oligosaccharides, a column (0.5 ϫ 8 cm) of UEA-I-agarose was prepared and equilibrated in TBS-NaN 3 (10 mM Tris, pH 8.0 containing 1 mM CaCl 2 , 1 mM MgCl 2 , and 0.02% NaN 3 ). For analysis of total cell extracts, the columns were equilibrated in TBS-NaN 3 containing 0.1% Triton X-100 and 10 mg/ml phenylmethylsulfonyl fluoride. Fractions (0.5 ml) were collected and material bound by the lectin were eluted with buffer containing 4 mg/ml fucose. HPLC of oligosaccharides was carried out on a Beckman model 334 system using an Alltech carbohydrate column (5 m, 4.1 ϫ 30 cm), as described previously (31). Initial mobile phase consisted of a mixture of 70% acetonitrile in water decreasing to a 50% acetonitrile during 40 min. Fractions (500 l, 0.5 min/fraction) were collected, and aliquots were assayed for radioactivity.
SDS-Polyacrylamide Gel Electrophoresis of Metabolically Radiolabeled Glycoproteins-After lectin affinity chromatography of extracts from metabolically radiolabeled cells, the column fractions were collected for further experiments. One-third of each fraction was precipitated in 10% trichloroacetic acid. The resulting protein pellets were centrifuged and washed twice with 500 l of chilled acetone. The labeled glycoproteins were resuspended in 20 -40 l of sample buffer and 1-5 l of 0.1 M Tris, pH 8, were added to insure basic conditions. SDS-polyacrylamide gel electrophoresis (SDS-PAGE) was carried out according to the methods of Laemmli (32) using 7.5% acrylamide at 20 A for approximately 3.5 h. The resulting gels were immersed in fixing solution (ethanol, acetic acid, water) overnight. Fixing solution was removed and the gels were immersed in EN 3 HANCE (Dupont) for 30 min with gentle agitation. Excess enhancer was removed, and the gels were dried and exposed to Kodak X-Omat x-ray film (Sigma) for 2-7 days. LAMP-1 was immunoprecipitated from the UEA-I-bound glycoproteins using methods described by Do et al. (33), and the immunoprecipitate was analyzed by the above procedures.
Glycosidase Treatments-Metabolically radiolabeled glycopeptides were treated with 500 milliunits of endo-␤-galactosidase in 100 l of 0.05 M sodium acetate buffer, pH 5.5, for 18 h at 37°C in a toluene atmosphere. The products were separated by descending paper chromatography in a solvent system of ethyl acetate:pyridine:glacial acetic acid:water (5:5:1:3). After the chromatograms were dried, 1-cm segments were transferred to 12 ϫ 75-mm test tubes containing 1 ml of 0.1 M pyridine acetate buffer, pH 5.4, to elute radiolabeled oligosaccharides, and aliquots were assayed for radioactivity by liquid scintillation counting. The separated oligosaccharides were pooled. Aliquots were dried in 1-ml microcentrifuge tubes for ␣-L-fucosidase digestions. Beef kidney ␣-L-fucosidase (Boehringer Mannheim) was added (200 -400 milliunits) in 100 l of 0.05 M sodium acetate buffer, pH 5.0. The mixtures were incubated for 18 h under a toluene atmosphere. The samples were boiled for 5 min, and oligosaccharides were analyzed by HPLC, as described above.
Methylation Analysis-The [ 3 H]Gal-labeled glycopeptides and oligosaccharides were methylated by the method of Hakomori (34), hydro-lyzed, reduced, acetylated, and analyzed by condensational gas chromatography, essentially as described by Cummings and Kornfeld (35). The partially methylated alditol acetates obtained in this way from glycopeptides were separated by gas chromatography using a flame ionization detector with a detection temperature of 250°C. The radioactive combustion product of tritiated water was collected in chilled scintillation vials for periods of 10 or 20 s/collection between 10 and 20 min after injection. Each condensate was mixed with 2.5 ml of scintillation fluid and 250 l of water to determine the retention time of the radiolabeled partially methylated derivative. The partially methylated alditol acetate standards of galactose were prepared according to the procedure of Doares et al. (36), and separated using Hewlett Packard 5890 gas chromatograph fitted with a SP2330 column (Supelco Inc.). Each peak was identified by its mass spectrum (electron impact) fragmentation pattern on a Hewlett Packard 5970 mass selective detector operated at an ionization potential of 70 eV. The retention times for each of the partially O-methylated, alditol acetates were consistent with previously reported values (36).

Construction of a Chinese Hamster Ovary Cell Line Express-
ing the Human H-type ␣1,2-Fucosyltransferase-The isolation and characterization of a cloned cDNA that encodes a human ␣1,2-fucosyltransferase (␣1,2FT) has been described previously (22,24). The catalytic properties of the enzyme encoded by this cDNA and its chromosomal localization strongly suggest that this cDNA corresponds to the human H blood group locus. Recent studies analyzing the structure and function of this gene in H blood group 8-deficient individuals (Bombay and paraBombay) are consistent with this assignment (4). We sought to further define the biochemical nature of oligosaccharide products constructed by this enzyme by expressing it in CHO cells. The oligosaccharide precursors synthesized by CHO cells have been extensively characterized previously (14 -17), and represent a prototypical cell line with which to explore the in vivo acceptor substrate utilization properties of the human H blood group ␣1,2FT.
To construct such cells, a human cDNA encoding this enzyme was cloned into the mammalian expression vector pCDM7 and was stably transfected into a parental CHO cell background, using a calcium phosphate transfection procedure and the G418 selection plasmid pSV2Neo, as described under "Experimental Procedures." As we have noted previously (23), approximately 25% of the resulting G418-resistant transfectants stably expressed the corresponding cell surface H blood group oligosaccharide determinant, as assessed by flow cytometry. A clonal cell line that exhibited stable expression of cell surface blood group H-determinants was isolated from this population and designated CHO(␣1,2)FT. Flow cytometry analyses demonstrate that virtually 100% of the cells derived from this clone exhibit bright fluorescence when stained with a monoclonal anti-H antibody (Fig. 1). As expected, the cell line also expressed substantial amounts of ␣1,2FT activity. Cell extracts were prepared from CHO(␣1,2)FT cells and assayed for ␣1,2FT activity using phenyl-␤-galactoside, an artificial acceptor specific for ␣1,2FT (11). Extracts prepared from CHO(␣1,2)FT cells contain an ␣1,2FT activity (specific activity of 99.8 pmol/mg h). By contrast, the control CHO-V cell line contains no detectable ␣1,2FT activity, nor do these cells express any detectable amounts of cell surface H blood group determinants (Fig. 1).
Cell Agglutination with Fucose-binding Lectins-To aid in characterizing the carbohydrate structures containing the H blood group antigen synthesized by CHO(␣1,2)FT cells, we sought to identify a fucose-binding lectin that bound with high affinity to the cells. As controls, we assessed the agglutinability of the parental CHO cells and a CHO cell clone, designated Clone 3, which stably expresses the murine UDPGal:Gal␤1-4GlcNAc ␣ 1,3 galactosyltransferase and synthesizes surface glycans with the terminal sequence Gal␣1-3Gal␤1-4Glc-NAc␤1-R (15). Agglutination was measured at both 1.0 and 0.1 mg/ml amounts of the lectins and was visually scored. The fucose-binding lectins tested with reported reactivity to H blood group determinants were E. europaeus agglutinin (37), A. anguilla agglutinin (38 -40), Ulex europaeus I (UEA-I) (41)(42)(43), and T. purpureas agglutinin (44,45). The CHO(␣1,2)FT cells were agglutinated by all the fucose-binding lectins at 1.0 mg/ ml, but UEA-I was the best agglutinin of cells at 0.1 mg/ml (Fig.  2). E. europaeus agglutinin also agglutinated CHO Clone 3 cells, which is consistent with the ability of this lectin to also bind terminal galactosyl residues (37). The parental CHO cells were not agglutinated by any of these lectins. The CHO Clone 3 cells were also agglutinated by G. simplicifolia I-B 4 isolectin, which is specific for terminal ␣-galactosyl residues (46). Since the CHO(␣1,2)FT cells were highly agglutinated by UEA-I, this lectin was selected for use in the lectin affinity chromatographic separations of glycans from these cells.
Glycoproteins from CHO(␣1,2)FT Containing H Blood Group Antigens-To identify whether glycoproteins in the CHO-(␣1,2)FT cells contain the H-blood group antigen, both parental CHO cells and CHO(␣1,2)FT cells were metabolically radiolabeled with [ 3 H]galactose or [ 3 H]glucosamine, as described previously and under "Experimental Procedures." Extracts of the cells in 0.1% Triton X-100 were passed over an affinity column of immobilized UEA-I. Approximately 5% of the total [ 3 H]Gallabeled material (this includes radiolabeled glycolipids and glycoproteins) from the CHO(␣1,2)FT cells bound to the column and was eluted with fucose, whereas very little material from the parental cells was bound (Fig. 3). A similar result was obtained with [ 3 H]GlcN-labeled material, except that approximately 30% of the radiolabeled material was bound by UEA-Iagarose (data not shown). The samples from raphy (Fig. 4). A number of glycoproteins were not bound by the lectin, whereas only two predominant glycoproteins in CHO(␣1,2)FT cells of Ϸ100 kDa and Ϸ120 kDa were bound by immobilized UEA-I. A similar result was obtained with [ 3 H]Gal-labeled material (data not shown).
To confirm that the column chromatography on UEA-I-agarose was efficient for all those glycoproteins containing the H antigen structure, the gel lanes containing the [ 3 H]Gal-labeled material UEA-I unbound and bound material were sliced into 0.5-cm sections and treated with Pronase to generate total glycopeptides. The pooled, radiolabeled glycopeptides from each lane were then re-applied to a column of UEA-I-agarose. Approximately 40% of the radiolabeled glycopeptides from the UEA-I-bound glycoproteins was bound by the immobilized lectin, but the glycopeptides derived from the originally unbound glycoproteins did not bind to the column (data not shown). These results demonstrate that UEA-I-agarose is effective in binding those glycoproteins containing H antigen and that the bound glycoproteins are enriched for glycans containing this determinant. As discussed below, the UEA-I-agarose column contains a high density of lectin and can bind simple oligosaccharides containing a single terminal ␣1,2-fucosyl residue. Thus, the lack of binding of the majority of glycoproteins to the UEA-I-agarose column is due to their lack of H antigen.
We have previously shown that in CHO cells polylactosamine sequences are enriched on two major glycoproteins, one identified as lysosome-associated membrane glycoprotein 1 (LAMP-1), and the other, which has the properties of LAMP-2 (33). In CHO cells, LAMP-1 has a size of Ϸ120 kDa and LAMP-2 has a size of Ϸ100 kDa. A polyclonal antibody reactive against murine LAMP-1, which cross-reacts with hamster LAMP-1, was used to assess whether the ϳ120-kDa glycoprotein was indeed LAMP-1. The UEA-I-bound material in the [ 3 H]GlcN-labeled material was immunoprecipitated with the anti-LAMP-1 antibody and analyzed by SDS-PAGE. A single protein was immunoprecipitated migrating with a mass of ϳ120 kDa (Fig. 4). These results demonstrate that fucosylation is preferential to glycoproteins containing polylactosamine sequences and that LAMP-1 is the ϳ120-kDa glycoprotein containing the H antigen. Approximately 17% of the total [ 3 H]Gal-labeled glycopeptides from CHO(␣1,2)FT cells bound to the column and were eluted with fucose, but no material from parental CHO cells was bound (Fig. 5). When the diantennary complex-type N-glycans from CHO(␣1,2)FT cells bound by ConA-Sepharose were analyzed for binding to UEA-I-agarose, very little material (Ͻ5%) was bound and eluted with fucose. This material was not analyzed further.
The glycopeptides from CHO(␣1,2)FT not bound by ConA-Sepharose and bound by UEA-I-agarose were isolated and reapplied to the column either before or after treatment with ␣-L-fucosidase. As expected, untreated glycopeptides quantitatively rebound to the column, whereas the glycopeptides treated with ␣-L-fucosidase did not bind (data not shown). These results demonstrate that a specific subset of glycopeptides derived from CHO(␣1,2)FT contain terminal fucosyl residues recognizable by UEA-I-agarose.
Fucose on CHO(␣1,2)FT Cells Is Present on Polylactosaminecontaining Glycopeptides-The complex-type N-glycans of CHO cells are well defined and are mixtures of di-, tri-, and tetraantennary structures, some of which contain polylactosamine [-3Gal␤1-4GlcNAc␤1-] n (14 -17). The O-glycans of CHO cells have no polylactosamine sequences and are simple core 1 structures (Gal␤1-3GalNAc␣1-Ser/Thr-R) that are either mono-or disialylated (19). The glycopeptides bound by UEA-I-agarose (Fig. 5) were all large sized on gel filtration on Sephadex G-25, in contrast to O-glycans, which have an intermediate elution on this column (data not shown) (49). This indicates that O-glycans are not present in the material bound by UEA-I-agarose.
To determine whether fucosylation occurred on polylactosamine, the [ 3 H]Gal-labeled glycopeptides from both parental CHO and CHO(␣1,2)FT cells not bound by ConA-Sepharose were treated with endo-␤-galactosidase, an enzyme capable of cleaving polylactosamine at internal unsubstituted galactosyl residues (50). The resulting fragments were separated by descending paper chromatography (Fig. 6). It has been demonstrated previously that endo-␤-galactosidase treatment of [ 3 H]Gal-labeled glycopeptides from parental CHO cells produces a sialylated tetrasaccharide (NeuAc␣2-3Gal␤1-4GlcNAc␤1-3Gal), a trisaccharide (Gal␤1-4GlcNAc␤1-3Gal), and a disaccharide (Gal␤1-4GlcNAc) (14, 16). Endo-␤-galactosidase released 67% of the radioactivity from parental CHO cells, and the released material migrated as the three expected products (Fig. 6A). In contrast, endo-␤-galactosidase treatment of total [ 3 H]Gal-labeled glycopeptides from CHO(␣1,2)FT cells released only 48% of the radioactivity and the pattern of the fragments produced was different from that obtained from parental cells (Fig. 6B). There was a marked reduction in the sialylated tetrasaccharide and a large reduction in the trisaccharide and disaccharide material. A new fragment migrating slightly slower than the trisaccharide was present and was designated peak a. This peak a fragment was enriched in the glycopeptides from CHO(␣1,2)FT cells bound by UEA-I-agarose compared to those unbound ( Fig. 6C and D). We predicted that peak a is the tetrasaccharide Fuc␣1-2Gal␤1-4GlcNAc␤1-3Gal. To confirm this, peak a was isolated following preparative descending paper chromatography and first analyzed for size on amine adsorption HPLC. The fragment elutes in the position of a tetrasaccharide and treatment of the fragment with ␣-Lfucosidase converts it to a trisaccharide (Fig. 7). In addition, the majority of radioactivity in this tetrasaccharide is bound by UEA-I-agarose (data not shown) and eluted with 4 mg/ml fucose. The results indicate that the tetrasaccharide component identified in glycans released from CHO(␣1,2)FT cells by endo-␤-galactosidase is the tetrasaccharide Fuc␣1-2Gal␤1-4GlcNAc␤1-3Gal. Methylation analysis, described below, confirms that galactose residues in glycopeptides from CHO-(␣1,2)FT cells bound by UEA-1-agarose are substituted at the C-2 position by fucose.
The results of endo-␤-galactosidase treatment and HPLC analysis for the [ 3 H]Gal-labeled material are compiled in Table  I. From this table it is evident that the lengths of polylactosamine in glycopeptides from CHO(␣1,2)FT cells are shorter than those from the parental cells, based on the recovery of disaccharide Gal␤1-4GlcNAc and the total amount of released material. The disaccharide provides a measure of polylactosamine chain length, since long polylactosamines release  Fig. 5 were treated with endo-␤-galactosidase, and the resulting material was analyzed by descending paper chromatography, as described under "Experimental Procedures." The migration positions of standards are indicated. The peak of radioactivity not migrating with known standards was designated as peak a. A, total parental CHO-derived glycopeptides; B, total CHO␣(1,2)FT glycopeptides; C, CHO␣(1,2)FTderived glycopeptides bound by UEA-I-agarose; D, CHO␣(1,2)FT-derived glycopeptides not bound by UEA-I-agarose. more releasable disaccharide than shorter ones (16). In addition, the presence of terminal fucosylation of glycopeptides in CHO(␣1,2)FT cells occurs at the expense of sialylation, indicating a competition reaction between ␣1,2-fucosylation and ␣2,3-sialylation.

FIG. 6. Descending paper chromatography of fragments produced following endo-␤-galactosidase treatment of [ 3 H]Gallabeled glycopeptides derived from parental CHO cells or CHO␣(1,2)FT cells. The [ 3 H]Gal-labeled glycopeptides obtained in
Methylation Analysis of Glycopeptides-To determine the substitution pattern of galactosyl residues, the [ 3 H]Gal-labeled glycopeptides form parental CHO and CHO(␣1,2)FT cells were methylated before and after treatment with neuraminidase and the methylated species analyzed by gas chromatography. The results are compiled in Table II. Only two species of methylated galactosyl residues are obtained from the methylation of [ 3 H]Gal-labeled glycopeptides from parental CHO cells. The two species are 2,3,4,6-tetra-O-methylgalactose, which arises from terminal galactosyl residues, and 2,4,6-tri-O-methylgalactose, which arises from 3-substituted galactosyl residues within both sialylated sequences NeuAc␣2-3Gal-R and polylactosamine sequences [-3Gal␤1-4GlcNAc␤1-] n . No galactose residues in parental CHO cells are substituted at the C-2 position, as evidenced by the lack of 3,4,5-tri-O-methylgalactose. In contrast, 28% of the galactose residues in glycopeptides from CHO(␣1,2)FT cells are substituted at C-2 and there is a decrease in the other two methylated species. The 2-substi-tuted galactose residues is expected to arise by fucosylation at C-2 by the ␣1,2FT in the CHO(␣1,2)FT cells. Methylation analysis of the glycopeptides from CHO(␣1,2)FT cells after ␣-L-fucosidase treatment shows a decrease in the 3,4,6-tri-Omethylgalactose and an increase in the 2,3,4,6-tetra-O-methylgalactose (data not shown). The decrease in 3-substituted galactosyl residues in the CHO(␣1,2)FT cells is consistent with a decreased sialylation and reduction in polylactosamine chain length. As expected, the amount of 2-substituted galactose residues is enriched in glycopeptides bound by UEA-I-agarose (Table II). These results demonstrate that a high percentage of the total galactosyl residues in glycopeptides derived from CHO(␣1,2)FT cells are substituted at the C-2 position by fucose and that there is an overall decrease in sialylation and polylactosamine chain length in these cells compared to parental CHO cells. These results demonstrate that there is preferential fucosylation of glycoproteins in CHO cells by the ␣1,2FT and that this preferential fucosylation is occurring on polylactosamine.
The Glycolipid Profiles of CHO(␣1,2)FT and Parental CHO Cells Are Identical-Metabolically radiolabeled glycolipids from both CHO(␣1,2)FT and parental CHO cells were compared by thin layer chromatography analysis, as described previously (15). The detectable [ 3 H]Gal-labeled glycolipids in both lower and upper phases fractions from both parental CHO and CHO(␣1,2)FT cells were identical (Fig. 8, A and B). These results indicate that the ␣1,2FT does not efficiently use endogenous lactosylceramide as an acceptor and does not compete for sialylation of this glycosphingolipid. DISCUSSION The results of our study demonstrate that the human H-type ␣1,2FT is effective in fucosylating terminal galactosyl residues within polylactosamine resulting in both a shortening of the chain length and a decrease of ␣2,3-sialylation of terminal galactosyl residues. Because of its preference for polylactosamines, the ␣1,2-fucosylation is preferentially restricted to two major glycoproteins. It has been shown previously that in CHO cells polylactosamine is enriched on only two major glycoproteins, LAMP-1 and another glycoprotein tentatively identified as LAMP-2 (33). The preferential presence of polylactosamine on these glycoproteins in CHO cells correlates well with studies showing that polylactosamine also occurs preferentially on LAMPs in human granulocytes (51,52). In CHO cells the H-type ␣1,2FT does not efficiently act on terminal galactosyl residues of diantennary complex-type N-glycans, nor is it efficient in fucosylating either O-glycans or lactosylceramide.

FIG. 7. Amine-adsorption HPLC of oligosaccharides released by endo-␤-galactosidase from [ 3 H]Gal-labeled glycopeptides derived from CHO␣(1,2)FT cells. The [ 3 H]
Gal-labeled glycans corresponding to oligosaccharides released by endo-␤-galactosidase treatment were obtained by preparative descending paper chromatography from CHO␣(1,2)FT cells, as shown in Fig. 6C, and analyzed by amine adsorption HPLC, as described under "Experimental Procedures". The peak a material was treated with ␣-L-fucosidase and reanalyzed on HPLC, as described under "Experimental Procedures". Ⅺ-Ⅺ, Glc-NAc␤1-3Gal; E-E, Gal␤1-4GlcNAc␤1-3Gal; q --q, peak a material shown as Fuc␣1-2Gal␤1-4GlcNAc␤1-3Gal; q-q, peak a material treated with ␣-L-fucosidase and re-analyzed by HPLC. Ordinate indicates scale only; backgrounds were 10 -30 cpm. though Fuc␣1-2Gal␤1-4Glc-Cer has been described in rat intestine (56). The broad acceptor specificity of the H-type ␣1,2FT supports a prediction that terminal fucosylation by this enzyme could be indiscriminate and could occur on many different types of N-and O-glycans, and thus be present on many different glycoproteins. The results of our study indicate otherwise. Within CHO cells during glycoprotein biosynthesis, the ␣1,2FT is highly selective in fucosylating glycoproteins and demonstrates a marked preference for polylactosamine. There is very little information about the expression of H blood group on glycoproteins resulting from the action of the human H-type ␣1,2FT, and often, it is unclear whether expression is due to the SE-type or H-type ␣1,2FT. In general, however, the evidence indicates the H antigen expression is highly limited to certain glycoproteins. For example, band 3 glycoprotein (the anion transporter) is the major ABO(H)-containing glycoprotein on human erythrocytes and it carries H antigens on its polylactosamines (7,57,58). In hamster pancreas ductal adenocarcinomas, it was found that blood group A antigen (and by inference the H blood group precursor) occur on only three major glycoproteins of 120, 135, and 150 kDa, and the antigen is present on highly branched N-glycans (62). A small number of surface glycoproteins of 140, 120, and 80 kDa from human endothelial cells bind to UEA-I-agarose (60). In a high tumorigenic clone of rat colon carcinoma cells, H antigen expression was essentially limited to splice variants of the CD44 molecule (61). Although most plasma glycoproteins lack either polylactosamine sequences or blood group antigens, there is recent evidence that at least three glycoproteins contain minor levels of H blood group antigen. Factor VIII (62) and von Willebrand factor (63) contain a small amount of H antigen on di-and triantennary N-glycans lacking polylactosamine. Human plasma ␣ 2 -macroglobulin contains H-antigen on N-glycans, but only about 10% of the glycoprotein is precipitable anti-H antibody (64). Transgenic mice expressing the human H-type ␣1,2FT under the control of a mammary gland-specific promoter produce milk containing very high levels of 2Ј-fucosyllactose, but only a single glycoprotein contains the H antigen (80). Although a systematic survey of glycoproteins expressing H antigen in human cells and tissues under the control of the H-type ␣1,2FT has not been conducted, the evidence strongly suggests that H antigen expression by this enzyme is limited to certain glycoproteins, and even when present on glycoproteins lacking polylactosamine sequences, the quantitative amounts of the antigen are low.
The human epidermoid carcinoma cell line A431, from which the cDNA encoding the human H-type ␣1,2FT was originally isolated, synthesizes the epidermal growth factor receptor, which contains polylactosamine sequences (66) and is a major carrier of blood group A and fucose determinants (and by inference H antigen) (66 -68). These blood group determinants are on N-glycans and not O-glycans, based on two observations. First, the epidermal growth factor receptor lacks O-glycans (68) and the well characterized O-glycans on the human low density lipoprotein receptor, synthesized by the same cells, lack any fucosyl residues (69). These observations, and the fact that in the CHO␣(1,2)FT cells O-glycans are not modified by the Htype ␣1,2FT, raise the question whether this enzyme can utilize O-glycans on glycoproteins as acceptors. In some cases, such as that observed for the respiratory mucins, O-glycans contain the H antigen, but the occurrence of this structure is due to the SE-type ␣1,2FT and not the H-type ␣1,2FT (9,70).
There are several possible explanations for the observation that the H-type ␣1,2FT exhibits preferential fucosylation of glycoproteins. One possibility is that the enzyme has better access to polylactosamine sequences on glycoproteins than to other carbohydrate acceptors. Such a mechanism has been proposed to explain the branching of some diantennary Nglycans in specific glycoproteins by N-acetylglucosaminyltransferase V (GlcNAcT-V) (71,72). In this case GlcNAcT-V acts on diantennary complex-type N-glycans but only if they are accessible on the intact glycoprotein. Thus, glycoprotein conformation has an important role in allowing accessibility of glycans to enzymes. Surprisingly, most diantennary N-glycans are inaccessible to this GlcNAcT-V, and this was demonstrated to be the case for glycoproteins in Chinese hamster ovary cells (72). The important role of accessibility and protein conformation has also been demonstrated for the UDPGlc:glycoprotein glucosyltransferase, which transfers glucose residues to high mannose-type N-glycans of glycoproteins in a partially folded state and does not act efficiently on high mannose glycans in native glycoproteins (73). It is thus possible, that the human H-type ␣1,2FT acts on terminal ␤-galactosyl residues in glycoproteins to which it has access, and polylactosamine-containing glycans may be the most accessible.
It is also possible that the human H-type ␣1,2FT may recognize some peptide determinant in specific glycoproteins as a prelude to the fucosylation reaction. Precedent for specific rec-

FIG. 8. Autoradiography of a thin layer chromatogram of the [ 3 H]Gal-labeled glycolipids extracted from parental CHO cells and CHO␣(1,2)FT cells. Lipid extracts of the [ 3 H]Gal-labeled cells
were separated into upper and lower phases and impurities were removed as described under "Experimental Procedures." A, the lower phase neutral glycosphingolipids material (100,000 cpm) were applied to Silica Gel-60 aluminum-backed plates and separated by chromatography in the solvent system CHCl 3 :CH 3 OH:H 2 O (65:25:4). The radiolabeled glycolipids were visualized by autoradiography. Standards indicated are as follows: 1, NeuAc␣2-3Gal␤1-4Glc␤1-ceramide; 2, Glc␤1ceramide; 3, Gal␤1-4Glc␤1-ceramide. (The radiolabeled material migrating slower than lactosylceramide is not a glycosphingolipid and is insensitive to ceramidase.) B, the upper phase ganglioside fraction (25,000 cpm) was also subjected to thin layer chromatography as in A, but using the solvent system CHCl 3 :CH 3 OH:0.25% KCl (5:4:1). Standards indicated are GM 3 and GM 2 . ognition mechanisms of glycosylation have been described for lysosomal enzymes, which acquire GlcNAc␣-1-phosphate on mannosyl residues by the lysosomal phosphotransferase (74), and pituitary glycoprotein hormones, which acquire terminal ␤1-4-linked GalNAc residues through the action of a specific N-acetylgalactosaminyltransferases that recognizes peptide sequences within the acceptors (75). At present we do not know the basis for this preferential fucosylation by the H-type ␣1,2FT. Future experiments will also explore the ␣1,2-fucosylation of glycoproteins when the enzyme is expressed in other cell types and the specificity of the recombinant enzyme for specific glycoprotein acceptors in vitro.
The reduction in polylactosamine chain length and decrease in sialylation we observed in CHO␣(1,2)FT cells is consistent with proposals about importance of competition reactions between glycosyltransferases in regulating glycoconjugate biosynthesis (76,77). Related studies on glycoconjugates synthesized in the presence and absence of expression of the SE ␣1,2FT suggest that fucosylation by this enzyme alters surface expression of Lewis antigens, probably by competing with sialyltransferases acting on terminal Gal residues of the lactosamine unit and ␣1,3-fucosyltransferases acting on GlcNAc residues of the lactosamine unit (78,79). Recent studies on respiratory mucins demonstrate that expression of the SE ␣1,2FT has profound effects on the structures of O-glycans, presumably through competition reactions with other glycosyltransferases (9,10,70). Thus, fucosylation of lactosamine units by the human H-type ␣1,2FT "commits" an oligosaccharide to a defined pathway preventing sialylation of terminal lactosamine units and precluding further elongation of polylactosamine [-3Gal␤1-4GlcNAc␤1-] n .
In previous studies we demonstrated that CHO cells transfected with the murine UDPGal:Gal␤1-4GlcNAc (Gal to Gal) ␣1,3-galactosyltransferase (␣1,3GT), contain polylactosamine modified by the addition of terminal ␣1-3 linked galactosyl residues (15). ␣-Galactosylation by this enzyme decreases sialylation of polylactosamine sequences, but has much less effect on length of polylactosamine than is observed in the present study with the human H-type ␣1,2FT. The biochemical basis for this difference is unclear. Perhaps the H-type ␣1,2FT localizes in CHO cells to a more proximal Golgi compartment than the ␣1,3GT, and can more effectively compete for polylactosamine elongation compared to the ␣1,3GT. Such differential localization has recently been proposed as a possible explanation for differential effects on polylactosamine synthesis in CHO cells by two ␣1,3FTs (FTIII and FTIV) (20). Alternatively, the H-type ␣1,2FT and the ␣1,3GT may differ in their acceptor preference for polylactosamine length on acceptor glycoproteins.
The results of our study raise many questions about the specificity of the both the H-type and the SE-type ␣1,2FTs and suggest new directions for research on these enzymes. It would now be important to conduct a parallel study using CHO cells expressing the SE-type ␣1,2FT to determine whether that enzyme causes a different type of glycosylation than we have observed for the H-type ␣1,2FT. The specificity of the H-type and the SE-type ␣1,2FT for glycoprotein acceptors should now be studied more thoroughly in vitro using either purified or recombinant enzymes and glycoprotein acceptors containing complex-type N-glycans with and without polylactosamine sequences. It should be noted, however, that in some cases recombinant, soluble enzymes have altered specificity compared to native enzymes (65). In addition, the specificity of the ␣1,2FTs for O-glycans should be studied. The predilection of the H-type ␣1,2FT for polylactosamine during glycoprotein biosynthesis suggest that the level of expression of the enzyme could also have a major effect on limiting polylactosamine length and modification. Future studies should also address the effects of differential levels of ␣1,2FT expression in cells on overall sialylation and polylactosamine structure. It is also possible that in cells expressing very high levels of ␣1,2FT that the enzyme might begin to compete favorably for sialylation of those N-glycans lacking polylactosamine, and thus cause more pronounced effects on the terminal glycosylation (e.g. ␣2,3/6sialylation or ␣1,3-galactosylation) of glycans in cells.