Stable expression of the Golgi form and secretory variants of human fucosyltransferase III from BHK-21 cells. Purification and characterization of an engineered truncated form from the culture medium.

Stable BHK-21 cell lines were constructed expressing the Golgi membrane-bound form and two secretory forms of the human α1,3/4-fucosyltransferase (amino acids 35-361 and 46-361). It was found that 40% of the enzyme activity synthesized by cells transfected with the Golgi form of the fucosyltransferase was constitutively secreted into the medium. The corresponding enzyme detected by Western blot had an apparent molecular mass similar to those of the truncated secretory forms. The secretory variant (amino acids 46-361) was purified by a single affinity-chromatography step on GDP-Fractogel resin with a 20% final recovery. The purified enzyme had a unique NH2 terminus and contained N-linked endo H sensitive carbohydrate chains at its two glycosylation sites. The fucosyltransferase transferred fucose to the O-4 position of GlcNAc in small oligosaccharides, glycolipids, glycopeptides, and glycoproteins containing the type I Galβ1-3GlcNAc motif. The acceptor oligosaccharide in bovine asialofetuin was identified as the Man-3 branched triantennary isomer with one Galβ1-3GlcNAc. The type II motif Galβ1-4GlcNAc in bi-, tri-, or tetraantennary neutral or α2-3/α2-6 sialylated oligosaccharides with or without N-acetyllactosamine repeats and in native glycoproteins were not modified. The soluble forms of fucosyltransferase III secreted by stably transfected cells may be used for in vitro synthesis of the Lewisa determinant on carbohydrates and glycoproteins, whereas Lewisx and sialyl-Lewisx structures cannot be synthesized.

Construction of ␣1,3/4-Fucosyltransferase Variants-Mutants of the human ␣1,3/4-fucosyltransferase were generated by PCR-based sitedirected mutagenesis of a FucT-III cDNA (a gift from Dr. S. Gonski, Hoechst AG, Frankfurt, FRG) encoding the full length of the enzyme. PCR was performed using the Expand High Fidelity DNA-polymerasemixture (Boehringer) and the supplied buffer according to the manufacturer's protocol at standard concentrations of 0.3 M for each primer and 0.2 mM each deoxynucleotide. PCR conditions, if not otherwise stated, were a 2-min denaturation step at 94°C, followed by 30 cycles with 15 s of denaturation at 94°C, 20 s of annealing at 50°C, 2 min of elongation at 72°C, and final elongation for 8 min at 72°C. DNA fragments were cloned into the eukaryotic expression vector pCR3 using a TA cloning kit (Invitrogen, Leek, The Netherlands). Positive clones were identified by using standard techniques and were verified by using the CircumVent Thermal Cycle Dideoxy Sequencing kit (New England Biolabs, Schwalbach, Germany).
The mutant pCRFT3T2 encodes a full-length human FucT-III that is COOH-terminally elongated with a tripeptide-spacer GAG followed by the epitope FDKNYVANSGK derived from human cytomegalovirus glycoprotein ICP 36 (that is recognized by the monoclonal antibody A3C5). 2 The construct was generated in one step using FucT-III cDNA template, the primer FT6-N (forward) 5Ј-ACT CTG ACC CAT GGA TCC CCT-3Ј, and the mutagenesis primer FT-TAG2 (reverse) 5Ј-TCA CTT GCC GCT GTT TGC GAC GTA ATT TTT GTC GAA TCC AGC TCC GGT GAA CCA AGC CGC.
Mutant pCRS1FT3 encodes a FucT-III containing the human IL-2 signal peptide (17 amino acids using the second start codon) and the first two amino acids, Ala-Pro, of the mature IL-2 protein linked NH 2terminally to Val-36 of FucT-III. This construct was generated in a one-step PCR using the FucT-III cDNA template, the mutagenesis primer s1FT3 (forward) 5Ј-AGG ATG CAA CTC CTG TCT TGC ATT GCA CTA AGT CTT GCA CTT GTC ACA AAC AGT GCA CCT GTG TCC CGA GAC GAT-3Ј and primer FT6-C (reverse) 5Ј-CTC TCA GGT GAA CCA AGC CGC TAT GC-3Ј.
Mutant pCRS2FT3T2 encodes a FucT-III containing a 20-amino acid human IL-2 signal peptide (using the first start codon) fused NH 2terminally to Ala-47 of FucT-III and containing GAG linker and the A3C5 epitope at the COOH terminus as described above for mutant pCRFT3T2. The vector was generated in a two-step PCR procedure essentially as described (16,26). The first PCR was performed using human IL-2 cDNA as a template and the mutagenesis primers IL2N (forward) 5Ј-AAG ATG TAC AGG ATG CAA CTC C-3Ј and FT3N2IL  (reverse) 5Ј-CGG GAG GAC CCA CTA GGT GCA CTG TTT GTG-3Ј  (PCR conditions: 2 min at 94°C, 30 cycles with 15 s at 94°C, 20 s at  50°C, 20 s at 72°C, and a final 8 min at 72°C). One fifth of the first  reaction mixture was used in a second PCR using human FucT-III  cDNA as a template, fresh IL2N primer, and the mutagenesis primer  FT-TAG2 (shown above) at reduced primer concentrations of 0.04 M. The fragment of the expected length was recovered by agarose gel electrophoresis and was extracted from the gel using JETsorb reagents (Genomed, Bad Oeynhausen, Germany) prior to TA cloning.
Transfection of BHK-21 Cells-BHK-21 cells were grown as described previously (12) and transfected by the calcium phosphate precipitation method. Semi-confluent cells were typically transfected with 4 g of plasmid DNA (pCRFT3, pCRS1FT3, or pCRS2FT3T2) and 1 g of the plasmid DNA pSV2pac to confer resistance to puromycin. Puromycin-resistant cells were selected using several medium exchanges with DMEM containing 10% fetal calf serum (FCS) and 0.8 g ml Ϫ1 puromycin-dihydrochloride during a time period of 2-3 weeks. Confluent cells were grown for 2 days in the absence of FCS and the corresponding supernatants were tested for fucosyltransferase activity after a 5-10-fold concentration in a Speed-Vac or by ultrafiltration using Centricon 10 cartridges (Millipore, Eschborn, FRG).
Preparation of Cell Extracts-Cell extracts were freshly prepared prior to each assay. Cells were washed with 5 ml of DMEM, trypsinized, and spun at 1000 ϫ g. After resuspending in ice-cold extraction buffer (1 ml of 20 mM Mops-KOH buffer, pH 7.5, containing 2% Triton X-100/ 10 7 cells) cells were disrupted using a Potter-Elvhjem homogenizer at 0°C. If not otherwise stated, extracts were used within the same day of homogenization for fucosyltransferase activity determinations.
Enzyme Purification-The culture supernatant from stably transfected BHK-21 cells (1600 ml) was 3.5-fold concentrated by ultrafiltration and was applied on a 25-ml GDP-Fractogel column equilibrated with 20 mM Mes-KOH, pH 6.8, containing 0.02% NaN 3 and 1 mM dithioerythreitol at a flow rate of 1mlmin Ϫ1 at room temperature. The column was washed with 20 mM Mes-KOH buffer, pH 6.8, containing 50 mM NaCl, 0.02% NaN 3 , and 30% glycerol (120 ml) and subsequently with 100 ml of the same buffer containing 500 mM NaCl. Elution of the enzyme was performed at a flow rate of 1.7 ml min Ϫ1 with the same buffer containing 1.5 M NaCl (90 ml). The enzyme was concentrated by ultrafiltration using Centricon 10 cartridges to a final protein concentration of 30 g ml Ϫ1 and was stored in elution buffer containing 30% glycerol and 0.02% NaN 3 at Ϫ20°C without any loss of activity over a period of 6 months.
SDS-PAGE and Western Blot Analysis-SDS-PAGE was performed according to Laemmli (17) using 12.5% and 3% acrylamide in the resolution and stacking gels, respectively.For Western blot analysis, proteins were transferred to nitrocellulose (Millipore) in a semidry instrument (Bio-Rad). The membrane was blocked with Tris-buffered saline containing 10% horse serum and 3% bovine serum albumin for 1 h and incubated overnight with biotinylated anti-cytomegalovirus tag monoclonal antibody (A3C5) or anti-FucT-III antiserum in blocking buffer at 1:500 and 1:1000 dilutions, respectively, overnight. The second antibody, streptavidin or anti-rabbit immunoglobulin coupled to horseradish peroxidase, respectively, was used at a 1:1000 dilution. The blots were developed with Tris-buffered saline containing 0.5 mg ml Ϫ1 4-chloro-1-naphthol solubilized in methanol and 0.2% perhydrol. Endoglycosidase treatment of the enzyme with recombinant PNGase F and endoglycosidase H as well as mild acid treatment were performed as described (12).
Fucosyltransferase Assays-The fucosyltransferase activity with oligosaccharides, glycolipids, glycopeptide, and 8-methoxycarbonyloctyl glycoside acceptors type I and II was tested at 37°C in 60 l of the following reaction mixture using 20 -60 microunits of enzyme: 50 mM Mops/NaOH buffer, pH 7. permethylated. The reaction mixtures with glycolipid and 8-methoxycarbonyloctyl glycoside as acceptors were diluted with water to 1 ml and applied to Sep-Pak C 18 cartridges, which were washed with 5 ml of water. The products were then eluted with 1 ml of methanol. The incorporation of [ 14 C]fucose was determined by liquid scintillation counting. The glycopeptide reaction mixture was separated by reversedphase chromatography on a C 18 Vydac column as described previously (19). Absorbance was monitored at 202 nm, and fractions corresponding to individual peaks were analyzed by MALDI/TOF-MS.The fucosyltransferase activity toward glycoprotein substrates was tested in 20 l of the following reaction mixture: 20 mM Tris-Mes buffer, pH 6.8, 50 mM NaCl, 1 mM dithioerythritol, 25 mM MnCl 2 , 10 mM Fuc, 5 mM ATP, 0.05 mM GDP-[ 14 C]Fuc (1 nmol contained 300,000 cpm), and 2 mg ml Ϫ1 of glycoprotein acceptor (1 nmol of bovine asialofetuin, 1.7 nmol of ␤-trace protein, and 1.8 nmol of soluble form of the IL-4 receptor). After incubation the reaction mixtures were precipitated for 10 min with 1 ml of 1% phosphotungstic acid in 0.5 M HCl (0°C) and transferred under vacuum to glass microfiber filters (Whatman GF/C). Filters were washed with 1% tungstophosphoric acid in 0.5 M HCl and methanol, dried, and counted for radioactivity in a Beckmann LS 6000 LC scintillation counter. There was no incorporation of Fuc in reaction mixtures without enzyme or without acceptor. For a detailed characterization of the fucosylated N-linked oligosaccharides from asialofetuin, 10 mg of the protein were incubated with S2FT3T2 in the buffer described above containing 750 nmol of GDP-[ 14 C]Fuc for 24 h at 37°C. The resulting protein was dialyzed against 0.5% acetic acid, lyophilized, and the corresponding oligosaccharides were released by automated hydrazinolysis in a GlycoPrep instrument (NϩO mode, Oxford Glycosystems, UK). The oligosaccharides were then separated by amino-bonded phase high performance liquid chromatography (13), and the radiolabeled peaks analyzed by HPAE-PAD, MALDI/TOF-MS, and methylation analysis.
Enzyme kinetics of S2FT3T2 was performed in reaction mixtures as described above where the concentration of the type I 8-methoxycarbonyloctyl glycoside varied between 0.0095 and 0.91 mM. The apparent kinetic parameters were determined from a Michaelis-Menten curve fit to the experimental data using the least square method. One unit of enzyme activity was defined as the amount of enzyme catalyzing the transfer of 1 mol of fucose/min to the 8-methoxycarbonyloctyl glycoside type I substrate.
Methylation Analysis of Carbohydrates-For methylation analysis, oligosaccharides were permethylated according to Hakomori (18), purified on a Sephadex LH 2 O column, hydrolyzed, reduced, and peracetylated as described (14). Separation and identification of partially methylated alditol acetates was performed on a Finnigan gas chromatograph (Finnigan MAT Corp., San Jose, CA), equipped with a 30-meter DB5 capillary column, connected to a Finnigan GCQ ion trap mass spectrometer.
Analytical HPAE-PAD of Native and Desialylated Oligosaccharides-A Dionex BioLC System (Dionex, Sunnyvale, CA) equipped with a CarboPac PA1 column (4 mm ϫ 250 mm) was used in combination with a pulsed amperometric detector (detector potentials and pulse durations: E1 ϭ ϩ50 mV, T1 ϭ 480 ms; E2 ϭ ϩ500 mV, T2 ϭ 120 ms; E3 ϭ Ϫ500 mV, T3 ϭ 60 ms). Prior to HPAE-PAD analysis, N-glycans were desialylated by solubilizing in sialidase buffer (10 mM sodium acetate, 1 mM CaCl 2 , 0.02% sodium azide) and by incubating with 0.2 units ml Ϫ1 V. cholerae sialidase for 2 h at 37°C. The NeuAc/oligosaccharides mixtures were injected onto the column before or after desalting. Elution was performed by applying a 2-min isocratic run with 100% solvent A followed by a linear gradient from 0 to 20% solvent B over 38 min and a linear gradient to 100% solvent B within 10 min. The flow rate was 1 ml min Ϫ1 ; solvent A: 0.2 M NaOH, solvent B: 0.2 M NaOH containing 0.6 M sodium acetate. Separation of small oligosaccharides and their fucosylated products was performed by applying a linear gradient from 0 -20% 0.1 M NaOH over 40 min and to 100% 0.1 M NaOH containing 0.6 M sodium acetate within 10 min as described (19).
MALDI/TOF-MS-2,5-Dihydroxybenzoic acid was used as UV-absorbing matrix. 2,5-Dihydroxybenzoic acid (10 mg ml Ϫ1 ) was dissolved in 10% ethanol in water. For analysis by MALDI/TOF-MS, the solutions of the native or reduced and permethylated oligosaccharides were mixed with the same volume of matrix. 1 l of the sample was spotted onto a stainless steel tip and dried at room temperature. The concentrations of the Analyte mixtures were approximately 10 pmol l Ϫ1 .
Measurements were performed on a Bruker Reflex™ MALDI/TOF mass spectrometer using a N 2 laser (337 nm) with 3-ns pulse width and 107-108 watts/cm 2 irradiance at the surface (0.2 mm 2 spot). Spectra were recorded at an acceleration voltage of 20 kV using the reflectron for enhanced resolution.
Electrospray Ionization Tandem Mass Spectrometry (ESI-MS/ MS)-A Finnigan MAT TSQ 700 triple quadrupole mass spectrometer equipped with a Finnigan electrospray ion source was used for ESI-MS. The reduced and permethylated samples were dissolved in acetonitrile saturated with NaCl (about 10 pmol/l) and injected at a flow rate of 1 l/min into the electrospray chamber. A voltage of ϩ5.5 kV was applied to the electrospray needle. For collision-induced dissociation experiments, parent ions were selectively transmitted by the first mass analyzer and directed into the collision cell (argon was used as collision gas) with a kinetic energy set around minus 60 eV.

Construction and Isolation of Stable BHK-21 Cells
Expressing Human ␣1-3/4-Fucosyltransferase Activity-BHK-21B cells were transfected with plasmids encoding the membranebound form of FucT-III with a tag at its carboxyl terminus (FT3T2), and two soluble forms of FucT-III containing the human IL-2 leader peptide sequence, at amino acid 36 (S1FT3) or amino acid 47 (S2FT3T2, with a tag at its COOH terminus) (Fig. 1). After transfection, cells were selected in medium containing 0.8 g ml Ϫ1 puromycin hydrochloride. Resistant cell clones were isolated, and, after reaching confluence, supernatants and cellular extracts were assayed for fucosyltransferase activity with Gal␤1-3GlcNAc-O(CH 2 ) 8 COOMe as an acceptor substrate ( Table I). The highest activity of secreted FucT-III was measured in the supernatant of stable BHK-21 cells transfected with pCRS2FT3T2. Secretion of enzyme activity increased for up to 40 h in confluent cultures supplemented with fresh medium. This clone (S2FT3T2) was used for the production and purification of the recombinant enzyme (see below). For the cell clones expressing secretory forms of FucT-III, some fucosyltransferase activity was detected in the cellular extracts (10 -20% of the secreted form, see Table I) presumably representing the enzyme fraction transported along the secretory pathway of cells. For cells transfected with the membranebound form of FucT-III, most of the activity was found to be associated with cellular extracts; however, considerable amounts (about 40% of the total activity measured with Gal␤1-3GlcNAc-O(CH 2 ) 8 COOMe as a substrate) were detected in the culture supernatant of confluent monolayers (viability higher than 95% based on trypan blue exclusion) that were cultivated for a 24-h period in fresh medium. The Western blotting analysis of concentrated supernatants from the three cell lines using the anti-FucT-III antibody showed bands of similar molecular masses around 40 kDa. This indicates a significant shedding of the membrane-bound form (FT3T2), presumably after proteolytic cleavage within the stem region.
Purification and Characterization of the Soluble Form of the Recombinant Human FucT-III-For the purification of S2FT3T2, confluent recombinant monolayers were grown for 4 weeks in DMEM containing 2% FCS, the medium being harvested every 2 or 3 days and being frozen at Ϫ20°C until further use. The enzyme was purified from 1600 ml of culture supernatant by affinity chromatography on a GDP-Fractogel column using the one-step procedure as described under "Experimental Procedures." The eluate containing the S2FT3T2 was concentrated by ultrafiltration and was stored in the presence of 30% (v/v) glycerol. The activity of the pooled supernatants after thawing decreased by approximately 2.5-fold. The purified S2FT3T2 preparation (4.5 ml) had a protein concentration of 30 g ml Ϫ1 and an activity of 33 units/liter with the type I 8-methoxycarbonyloctyl glycoside as a substrate). Thus, the final recovery of recombinant S2FT3T2 was approximately 20%.
Upon SDS-PAGE analysis followed by Coomassie staining, the enzyme preparation was found to consist of a closely spaced doublet of about 40 -42 kDa molecular mass, which was also detected with the anti-FucT-III antibody as well as with the anti-tag monoclonal antibody A3C5 (Fig. 2). Gas-phase sequencing of the purified enzyme yielded the expected 20 aminoterminal amino acids ( 1 APSGSSRQDTTPTRPTLLIL 20 ) that are deduced from the S2FT3T2 cDNA. This result indicates that the difference in molecular mass between the two S2FT3T2 forms seen after SDS-PAGE is not due to proteolytic degradation of the polypeptide.
N-Glycosylation of the Recombinant Human FucT-III-The recombinant S2FT3T2 contains two potential N-glycosylation sites (Asn-154 and Asn-185, according to the numbering of the FucT-III wild-type sequence). Incubation of the purified recombinant enzyme with PNGase F gave rise to a 3-4-kDa decrease in apparent molecular mass, resulting in two clearly distinguishable bands. The mass shift observed suggests that both glycosylation sites of S2FT3T2 are occupied by N-linked oligosaccharides. The lower molecular mass band after PNGase digestion had an apparent molecular mass of approximately 38 kDa (predicted molecular mass: 38,357). After incubation with Endo H, a small part of S2FT3T2 was not sensitive to the enzyme, suggesting the presence of small amounts of complextype glycans. However, most of the enzyme showed a mobility shift in SDS-PAGE similar to that obtained with PNGase F, indicating that the majority of the oligosaccharides are of the oligomannosidic or hybrid type (Fig. 2). Furthermore, mild acid hydrolysis of the enzyme before and after PNGase F or Endo H treatment did not produce any detectable shift in molecular weight (not shown), suggesting the absence of NeuAc in N-or, if present, O-linked glycans.
Substrate Specificity of the Soluble Form of the Recombinant Human FucT-III-The S2FT3T2 catalyzed the fucosylation of the type I but not of the type II 8-methoxycarbonyloctyl glyco-side. The apparent kinetic parameters determined with the type I acceptor using saturating concentrations of the GDP-Fuc were V ϭ 0.83 Ϯ 0.07 pmol min Ϫ1 ml Ϫ1 , and K m ϭ 0.54 Ϯ 0.08 mM, assuming a Michaelin behavior for the enzyme.
The specificity of the S2FT3T2 toward the small oligosaccharide acceptors LNFP, LNT, LNnT, LST-a, and SLN was analyzed after incubations of 2, 4, and 21 h. The formation of the fucosylated products was monitored by HPAE-PAD (Fig. 3), and the molecular masses of the products were determined by MALDI/TOF-MS (Table II). It was found that the reaction was linear for at least 4 h, so the activities shown in Table II were calculated based on 2-h incubation values. The S2FT3T2 activity with the type I acceptor (LNT) is 1.5-fold higher than with the type II acceptor (LNnT). Substitution of the terminal monosaccharide residue of the LNT with ␣2,3-linked sialic acid causes a 1.3-fold increase in activity of the S2FT3T2. Substitution of the LNT with ␣2-linked Fuc causes a 3.3-fold increase in activity of the S2FT3T2 (Table II). To identify the linkage position of fucose residues, after reduction and permethylation, the oligosaccharides were analyzed by collision-induced decomposition mass spectrometry and by methylation analysis (Fig. 4 and Table III Table II). No fucosylation was detected in the trisaccharide SLN when incubated under identical conditions.
S2FT3T2 activity toward glycolipids was tested with Gal␤3GlcNAc␤3Gal␤4Glc␤1-1ceramide as a substrate. It was found that 63 pmol/min/ml of Fuc transferred to 50 nmol of the acceptor.
Complex-type bi-, tri-, and tetraantennary structures with zero to four N-acetyllactosamine repeats and with or without ␣2,3/6-linked NeuAc residues were incubated with the enzyme for 4 and 24 h. Analyses by HPAE-PAD and MALDI/TOF-MS revealed that no fucosylation had occurred irrespective of the antennarity, number of lactosamine repeats or sialylation degree after a 24-h incubation period.  The glycoproteins, bovine asialofetuin, native fetuin and bovine thyroglobulin, human ␤-trace protein from hemofiltrate (20), recombinant human ␤-trace protein, recombinant human antithrombin III, and recombinant human IL-4 receptor from CHO cells, were tested as acceptors for S2FT3T2 (Table IV) by determination of [ 14 C]fucose incorporation. Only very low incorporation of fucose was achieved. The enzyme showed the highest activity with asialofetuin as a substrate. The activity with recombinant ␤-trace protein from BHK-21 cells was higher than with its natural counterpart isolated from hemofiltrate (15). No fucosylation of bovine thyroglobulin or recombinant antithrombin III from CHO cells was detected. Fucosylated glycoproteins were subjected to SDS-PAGE, and, following subsequent autoradiography, radiolabeled bands were detected at migration positions corresponding to the molecular masses of the untreated glycoproteins.
N-Glycan Structures of in Vitro Fucosylated Bovine Asialofetuin-For the determination of the linkage position of fucose in asialofetuin oligosaccharides, 10 mg of the glycoprotein were incubated with S2FT3T2 in the presence of 750 nmol of GDP-[ 14 C]Fuc (2 ϫ 10 5 cpm) for 18 h. The oligosaccharides from the unmodified and modified glycoprotein were released by automated hydrazinolysis and subjected to HPAEC-PAD (Fig. 5). Three major oligosaccharide peaks were obtained for the glycan mixture from unmodified asialofetuin: biantennary, trianten-nary 2,4-branched, and triantennary 2,4-branched with one Gal␤1-3GlcNAc antenna in a ratio of 10:55:35. The glycan mixture from in vitro fucosylated asialofetuin yielded a new peak eluting at 15.5 min (C1) with a concomitant decrease of peak C (C1, A, B, C in ratio of 27:11:53:9). The glycan mixture from S2FT3T2-treated asialofetuin was subjected to separation on NH 2 -bonded phase. Two major peaks were obtained, which were not completely separated (data not shown). Two molecular ions corresponding to reduced and permethylated triantennary N-glycans (m/z ϭ 2537) and a fucosylated triantennary structure (m/z ϭ 2710), respectively, were detected after MALDI/TOF-MS. The native material obtained after NH 2bonded phase was subjected to preparative HPAEC-PAD yielding peaks C1 and B (Fig. 5, panel 3). Methylation analysis revealed the presence of only 4-substituted GlcNAc in peak B and a mixture of 4-substituted and 3,4-disubstituted GlcNAc as well as terminal fucose in peak C1. Peak C1 eluted at 15.5 min in HPAEC-PAD and was completely converted to a structure eluting at the position of peak C upon mild acid treatment (Fig.  5, panel 4), whereas peak B was not affected. Thus, the results indicate that the N-linked oligosaccharide of asialofetuin with one type I antenna is modified by S2FT3T2 with fucose linked to position 4 of GlcNAc in the type I motif. No indication of the presence of fucosylated O-linked glycans was obtained in our experiments.

DISCUSSION
The construction of a soluble form of FucT-III through the replacement of amino acids 1-35 and 1-46 by the signal sequence of human IL-2 (constructs S1FT3 and S2FT3T2, respectively) produced catalytically active secreted forms of the enzyme when expressed from stably transfected BHK-21 cell lines.  which is more convenient and can also be used for the purification of other recombinant human fucosyltransferases. 3 It is also advantageous over other methods, where part of the protein A polypeptide sequence is linked to the amino terminus of the FucT-III, resulting in chimeric enzyme forms that can be purified by adsorption on and elution from IgG-Sepharose. However, in the procedure described in the present work, no fusion with unrelated bulky protein moieties that might alter enzyme specificity toward different substrates (9) is required. For the purified enzyme described here, two closely spaced bands with a unique amino-terminal sequence were detected. Apparent molecular masses of about 42 kDa were calculated from their mobility in SDS-PAGE and Coomassie staining as well as in Western blotting analysis using an antibody that recognizes the tag sequence fused to the COOH terminus of the enzyme. Since the expected NH 2 terminus of the S2TF3T2 polypeptide was unequivocally detected upon gas-phase sequencing of the product, the observed difference in apparent molecular mass is not due to proteolytic degradation at any part of the recombinant enzyme. The results obtained after the incubation with Endo H or PNGase F and neuraminidase/mild acid treatment indicated that the two glycosylation sites of S2FT3T2 are occupied with oligomannosidic or hybrid-type glycans that are not decorated with significant amounts of NeuAc. Based on binding studies with concanavalin A (8), occupancy of N-glycosylation site(s) has been suggested previously for an ␣1-3/4-fucosyltransferase purified from the culture medium of the A431 carcinoma cells.
The BHK-21 cells stably transfected with the membrane-bound wild-type form of the human FucT-III unexpectedly secreted considerable amounts of a soluble form of the enzyme into the culture supernatant (about 40% of the total activity that was measured after a 24-h production period of confluently growing and more than 95% viable BHK-21 cells). The detection of enzyme activity in supernatants of transiently transfected COS cells has been previously reported by Kukowska-Latallo (10) using expression plasmids encoding the human wild-type FucT-III. The detection of significant enzyme activity in the supernatants of stably transfected BHK-21 (Ͼ95% viable) cell clones in the present study suggests a common mechanism of endoproteolytic cleavage of human fucosyltransferase III within the stem region inside Golgi/endoplasmic reticulum compartments, since the apparent molecular weight detected for the secreted form of FT3T2 was very similar to that observed for the truncated genetically engineered secretory variants S2FT3T2 and S1FT3. This would also explain the high amounts of this enzyme that are detected in human milk and supernatants of A431 cells.     (22) indicate that amino acid deletions in this region do not affect enzyme activity. The S2FT3T2 efficiently fucosylates small oligosaccharides containing the type I and type II structures. Fuc was transferred preponderantly to the O-4 position of GlcNAc in LNT (70%), LNFP (100%), and LST-a (80%), whereas virtually all the Fuc was transferred to the O-3 position of Glc in LNnT. These results are in agreement with those described for enzyme preparations purified from human milk and the medium of A431 cells (7,8). However, in our studies with S2FT3T2 and LNT or LST-a as acceptors, additional monofucosylation at the reducing Glc occurred, and bifucosylated products that contained fucose at the O-4 and O-3 positions of GlcNAc and Glc were detected. These structures were not detected in previously reported work. SLN was not fucosylated by S2FT3T2, in contrast to what was described for FucT-III from milk or secreted from A431 cells.
All glycoproteins tested, except for asialofetuin, were found to be very poor substrates for S2FT3T2 based on [ 14 C]fucose incorporation studies. The asialofetuin used in the present work contained triantennary glycans with terminal Gal␤1-3 and Gal␤1-4 linked to GlcNAc as described by Rice et al. (23), but no branched O-linked oligosaccharide described by Edge and Spiro (24) was detected. This probably is due to the different commercial origin of the protein preparation used here. The fucosylated N-linked oligosaccharide of asialofetuin was identified as a 2,4-branched triantennary structure containing one Gal␤1-3GlcNAc antenna. This type of oligosaccharide is present, if any, in only very small amounts in the other glycoproteins which were used in this work as substrates for S2FT3T2. The observed substrate specificity of the enzyme enables us to suggest that human FucT-III provides a useful tool for detection of type I structures in N-glycosylated glycoproteins.
Based on antibody binding studies, COS-1 cells start to express Le x , sLe x , Le a , and sLe a structures at their surface after transfection with the FucT-III gene (1,10); some of the fucosylated molecules are glycoproteins (e.g. PSGL-1) (25). However, the secreted FucT-III produced in BHK-21 cells only had the capability of in vitro modifying type I structures, resulting in Le a and sLe a type motives in small oligosaccharides, glycolipid, glycopeptide, and glycoproteins. In contrast to previously reported work (10), in our study no structure could be detected, indicating that the enzyme recognizes the GlcNAc in type II N-glycans as a substrate; therefore, a biosynthetic involvement of FucT-III in the formation of Lewis x or sialyl-Lewis x motifs on glycoconjugates seems questionable. This difference in specificity might be due to the truncation at the amino terminus of the enzyme, or it may result from differences in the intracellular environment/compartmentalization and the in vitro assay conditions applied. However, it should be emphasized that solubilization of membrane-bound glycosyltransferases after disruption of cells in the presence of detergents will frequently result in a mixture of proteolytically cleaved and intact forms, which makes it difficult to unequivocally assess the substrate specificity of the native Golgi enzymes by in vitro assays. Finally, this problem could be solved by coexpression experiments using glycosyltransferase genes and, e.g., secretory model glycoproteins, which must be modified properly by the recipient cell line and must be structurally characterized thoroughly with respect to their carbohydrates. In preliminary studies of our laboratory using coexpression of human erythropoietin and FucT-III in BHK cells, no peripheral fucosylation of the secreted erythropoietin could be detected. This supports our view that the enzyme acts as a ␣1-4-fucosyltransferase also in vivo.