Purification, cDNA Cloning, and Expression of GDP-l-Fuc:Asn-linked GlcNAc α1,3-Fucosyltransferase from Mung Beans*

Substitution of the asparagine-linked GlcNAc by α1,3-linked fucose is a widespread feature of plant as well as of insect glycoproteins, which renders the N-glycan immunogenic. We have purified from mung bean seedlings the GDP-l-Fuc:Asn-linked GlcNAc α1,3-fucosyltransferase (core α1,3-fucosyltransferase) that is responsible for the synthesis of this linkage. The major isoform had an apparent mass of 54 kDa and isoelectric points ranging from 6.8 to 8.2. From that protein, four tryptic peptides were isolated and sequenced. Based on an approach involving reverse transcriptase-polymerase chain reaction with degenerate primers and rapid amplification of cDNA ends, core α1,3-fucosyltransferase cDNA was cloned from mung bean mRNA. The 2200-base pair cDNA contained an open reading frame of 1530 base pairs that encoded a 510-amino acid protein with a predicted molecular mass of 56.8 kDa. Analysis of cDNA derived from genomic DNA revealed the presence of three introns within the open reading frame. Remarkably, from the four exons, only exon II exhibited significant homology to animal and bacterial α1,3/4-fucosyltransferases which, though, are responsible for the biosynthesis of Lewis determinants. The recombinant fucosyltransferase was expressed in Sf21 insect cells using a baculovirus vector. The enzyme acted on glycopeptides having the glycan structures GlcNAcβ1–2Manα1–3(GlcNAcβ1–2Manα1–6)Manβ1–4GlcNAcβ1–4GlcNAcβ1-Asn, GlcNAcβ1–2Manα1–3(GlcNAcβ1–2Manα1–6)Manβ1–4GlcNAcβ1–4(Fucα1–6)GlcNAcβ1-Asn, and GlcNAcβ1–2Manα1–3[Manα1–3(Manα1–6)Manα1–6]Manβ1–4GlcNAcβ1–4GlcNAcβ1-Asn but not on, e.g. N-acetyllactosamine. The structure of the core α1,3-fucosylated product was verified by high performance liquid chromatography of the pyridylaminated glycan and by its insensitivity to N-glycosidase F as revealed by matrix-assisted laser desorption/ionization time of flight mass spectrometry.

The most characteristic features of asparagine-linked oligosaccharides from plants are the substitution of the core pentasaccharide by xylose and ␣1,3-linked fucose (1,2). The result-ing heptasaccharide "MMXF 3 " (Fig. 1) very often constitutes the main oligosaccharide species on a plant glycoprotein (3,4). According to their biosynthesis, these structures are classified as complex-type N-glycans, even though the terms paucimannosidic or truncated N-glycans appear to be more justified. The ␣-mannosyl residues may, however, be substituted by GlcNAc and these GlcNAc residues may be further decorated by galactose and fucose to form the same structure as the human Lewis a epitope ( Fig. 1) (5,6).
The antigenicity of "paucimannosidic" plant N-glycans is well documented (7)(8)(9)(10)(11). Since both xylose and core ␣1,3-fucose are not seen in mammalian glycoproteins they may form the key component of epitopes for carbohydrate-reactive antibodies (9,10,12). There is, however, evidence that the ␣1,3-linked fucosyl residue is the predominant antibody binding structural element (3,8,11,13). Due to the ubiquitous occurrence of such paucimannosidic N-glycans throughout the plant kingdom, they are responsible for the frequently observed cross-reactivity of antibodies raised against plant glycoproteins and are therefore termed "cross-reactive carbohydrate determinants" (12,14,15). Anti-cross-reactive carbohydrate determinants antibodies of the IgE class have been found in sera of many allergic patients (8,11,13,14,16,17). While the clinical role of cross-reactive carbohydrate determinants remains controversial, they are suspected to obscure (at least in vitro) allergy diagnosis. Anti-cross-reactive carbohydrate determinants antibodies will also react with many insect glycoproteins such as honeybee venom phospholipase A 2 or neuronal membrane glycoproteins from insect embryos because insects, like plants, are capable of synthesizing the core ␣1,3-fucose epitope (3, 11-13, 18, 19).
In contrast to the blood group-related fucosyltransferases which act on the nonreducing terminus of N-glycans, O-glycans, or glycolipids (20), core fucosyltransferases have received little attention. Only recently, the molecular cloning of GDP-L-Fuc 1 :Asn-linked GlcNAc ␣1,6-fucosyltransferase (core ␣1,6-fucosyltransferase, Fuc-T C6, Fuc-T VIII) from porcine brain and from human gastric cancer cells has been reported (21,22). As regards core ␣1,3-fucosyltransferase (Fuc-T C3), a first charac-terization of the enzyme from mung bean seedlings revealed its dependence on the presence of nonreducing terminal GlcNAc (23). In this paper, we report the purification to homogeneity of Fuc-T C3 from mung bean seedlings, the cloning of its cDNA by a PCR-based approach, and the expression of active recombinant Fuc-T C3 in baculovirus infected insect cells.
SV Total RNA Isolation System, avian myeloblastosis virus reverse transcriptase, and Taq Polymerase were purchased from Promega. Lipofectin, fetal calf serum, 3Ј-5Ј RACE System (version 2.0) for rapid amplification of cDNA ends, and degenerate oligonucleotides were purchased from Life Technologies, Inc., whereas Fuc-T C3 specific primers were synthesized by Vienna Biocenter Genomics. The TA Cloning Kit was obtained from Invitrogen.
Purification of Core ␣1,3-Fucosyltransferase-All purification steps were performed at 4°C. Mung bean seedlings were homogenized with a kitchen blender using 0.75 volumes of extraction buffer per kg of beans. The extraction buffer consisted of 0.5 mM dithiothreitol, 1 mM EDTA, 0.5% polyvinylpyrrolidone, 0.25 M sucrose, and 50 mM Tris-HCl buffer, pH 7.3. The resulting homogenate was filtered through two layers of cloth and the filtrate was centrifuged at 30,000 ϫ g for 40 min. The supernatant was discarded and the pellet was extracted with solubilization buffer consisting of 0.5 mM dithiothreitol, 1 mM EDTA, 1.5% Triton X-100, and 50 mM Tris-HCl, pH 7.3, by stirring overnight. Subsequent centrifugation at 30,000 ϫ g for 40 min yielded the Triton X-100 extract which was further purified as follows.
Step 2: the sample was applied to a column (2.5 ϫ 32 cm) of Affi-Gel Blue (Bio-Rad) equilibrated with buffer A. After washing the column with this buffer, adsorbed protein was eluted with buffer A containing 0.5 M NaCl. Step 3: following dialysis of the eluate from step 2 against buffer B (25 mM sodium citrate buffer, pH 5.3, containing 0.1% Triton X-100 and 0.02% NaN 3 ) it was loaded onto a column (1.5 ϫ 18 cm) of S-Sepharose equilibrated with the same buffer. Bound protein was eluted with a linear gradient from 0 to 0.5 M NaCl in buffer B. Fractions containing Fuc-T C3 were pooled and dialyzed against buffer C (25 mM Tris-HCl buffer, pH 7.3, containing 5 mM MnCl 2 and 0.02% NaN 3 ).
Step 4: the dialyzed sample was applied to a column (0.5 ϫ 4.5 cm) of GnGn-Sepharose previously equilibrated with buffer C. Elution of the bound protein was accomplished with buffer C containing 1 M NaCl instead of MnCl 2 .
Step 5: the enzyme was then dialyzed against buffer D (25 mM Tris-HCl, pH 7.3, containing 10 mM MgCl 2 , 0.1 M NaCl, and 0.02% NaN 3 ) and subsequently loaded onto a column (0.5 ϫ 4.5 cm) of GDPhexanolamine-Sepharose. After washing the column with buffer D, Fuc-T C3 was eluted by substituting MgCl 2 and NaCl with 0.5 mM GDP. Active fractions were pooled, dialyzed against 20 mM Tris-HCl buffer of pH 7.3, and lyophilized.
Electrophoretic Methods-SDS-PAGE was performed in a Bio-Rad Mini Protean Cell on gels containing 12.5% acrylamide and 1% bisacrylamide. Gels were either stained with Coomassie Brilliant Blue R-250 or silver. Isoelectric focusing of Fuc-T C3 was carried out on precast gels with a pI range from 6 to 9 (Servalyt precotes 6 -9, Serva) and gels were silver stained according to the manufacturers instructions. For twodimensional electrophoresis, lanes from the focusing gel were excised, treated with S-alkylation reagents and SDS, and subject to SDS-PAGE as described previously (30).
Amino Acid Sequencing and Mass Spectrometric Peptide Mapping-Protein bands were excised from Coomassie-stained SDS-polyacrylamide gels, carboxyamidomethylated, and digested with sequencing grade trypsin according to the in-gel digestion procedure described previously (31). The tryptic peptides were separated by reverse phase HPLC on a 1.0 ϫ 250-mm Vydac C18 at 40°C with a flow rate of 0.05 ml/min using a HP 1100 apparatus (Hewlett-Packard). Isolated peptides were sequenced with a Hewlett-Packard G1005A protein sequencing system according to the manufacturer's protocol. In addition, the peptide mixture obtained by in-gel digestion was analyzed by MALDI-TOF MS (see below).
Reverse Transcriptase-PCR and cDNA Cloning of Core ␣1,3-Fucosyltransferase-Total RNA was isolated from 3-day-old mung bean hypocotyls using the SV Total RNA Isolation System from Promega according to the supplier's instructions. To achieve first strand cDNA synthesis, total RNA was incubated for 1 h at 48°C with avian myeloblastosis virus reverse transcriptase and oligo(dT) primer using the Reverse Transcription System (Promega). First strand cDNA was subjected to PCR using as the sense primer 5Ј-GCIGARTAYTAYGCIG-ARAAYAAYATHGC-3Ј (S1) and as the antisense primer 5Ј-CRTADA-TRTGRTAIACIGTYTC-3Ј (S2) or 5Ј-TADATISWYTCCATYTCRAA-3Ј (S3), where I stands for inosin; R for G ϩ A; Y for T ϩ C; H for T ϩ C ϩ A; D for T ϩ G ϩ A; S for G ϩ C; and W for A ϩ T. PCR was performed on 10 l of the reverse transcriptase reaction in a volume of 50 l containing 0.1 mol of each primer, 0.1 mM dNTPs, 2 mM MgCl 2 , 10 mM Tris-HCl buffer of pH 9.0, 50 mM KCl, and 0.1% Triton X-100. After an initial denaturation step at 95°C for 2 min, 40 cycles of 1 min at 95°C, 1 min at 49°C, and 2 min at 72°C were run. The final extension step at 72°C was carried out for 8 min. PCR products were subcloned into pCR2.1 vector using the TA Cloning Kit (Invitrogen) and sequenced.
On the basis of the sequence of PCR product(s), the missing 5Ј and 3Ј regions of the cDNA coding for Fuc-T C3 were obtained by 5Ј-and-3Ј-rapid amplification of cDNA ends (RACE) using the RACE kit from Life Technologies, Inc. according to the manufacturer's recommenda-tions. 3Ј-RACE was performed with hemi-nested PCR using as antisense primer the universal amplification primer supplied with the kit and as sense primers at first 5Ј-CTGGAACTGTCCCTGTGGTT-3Ј and then 5Ј-AGTGCACTAGAGGGCCAGAA-3Ј. Likewise, 5Ј-RACE was performed by means of hemi-nested PCR using as the sense primer the abridged anchor primer supplied with the kit and as antisense primers either 5Ј-GAATGCAAAGACGGCACGATGAAT-3Ј and then 5Ј-TTC-GAGCACCACAATTGGAAAT-3Ј or PCR was performed with an annealing temperature of 55°C under conditions otherwise as described above. Both 5Ј-and 3Ј-RACE products were subcloned into pCR2.1 vector and sequenced.
DNA Sequence Analysis-Sequences of subcloned fragments were determined by the dideoxynucleotide chain termination method using an ABI PRISM Dye Terminator Cycle Sequencing Ready reaction Kit and an ABI PRISM 310 Genetic analyzer (Perkin-Elmer). T7 and M13 forward primers were used for sequencing the PCR products cloned in pCR2.1. Sequencing of both strands of the complete coding region was performed by the Vienna VBC Genomics-Sequencing Service using the cycle sequencing method with infrared labeled primers (IRD700 and IRD800) and a LI-COR Long Read IR 4200 sequencer (Lincoln, NE).
Expression of Recombinant Fuc-T C3 in Insect Cells-The coding region of the putative Fuc-T C3 cDNA including the cytoplasmic and the transmembrane regions was amplified using the forward primer 5Ј-cggcggatcCGCAATTGAATGATG-3Ј and the reversal primer 5Ј-ccggct-gcaGTACCATTTAGCGCAT-3Ј by means of the Expand High Fidelity PCR System (Roche Molecular Biochemicals). The PCR product was double digested with PstI and BamHI and subcloned into alkaline phosphatase-treated baculovirus transfer vector pVL1393 previously double digested with PstI and BamHI. To allow homologous recombination, the transfer vector was co-transfected with BaculoGold viral DNA (PharMingen, San Diego, CA) into Sf9 insect cells in IPL-41 medium containing Lipofectin. After 5 days of incubation at 27°C, various volumes of supernatant containing recombinant virus were used for infection of Sf21 insect cells. After incubation for 4 days at 27°C in IPL-41 medium containing 5% fetal calf serum, the Sf21 cells were harvested and washed twice with phosphate-buffered saline. The cells were resuspended in 25 mM Tris-HCl buffer of pH 7.4 containing 2% Triton X-100 and disrupted by sonication on ice. This homogenate as well as the culture supernatant were assayed for Fuc-T C3 activity. Mock infections were performed with recombinant baculovirus encoding tobacco GlcNAc transferase I (28).
Analysis of the Transferase Product-Dabsylated GnGn-hexapeptide (2 nmol) was incubated with insect cell homogenate containing recombinant Fuc-T C3 (0.08 milliunit) in the presence of non-radioactive GDP-L-fucose (10 nmol) under conditions otherwise identical to those described for determination of transferase activity (see above). A control experiment was performed with homogenate from mock-infected insect The remaining half of the sample containing the Fuc-T C3 product was digested with N-glycosidase A. The resulting oligosaccharides were pyridylaminated and analyzed by reverse phase HPLC (8,32,33). The transferase product was degraded using N-acetyl-␤-glucosaminidase and again analyzed by HPLC. Pyridylaminated GnGnF 6 derived from human IgG, MMF 3 from honeybee venom phospholipase A 2 , GnGn, GnM, MGn, and MM from bovine fibrin were used as reference substances (23,26,34). Additionally, the mass of the pyridylaminated product was determined by MALDI-TOF MS (see below).
MALDI-TOF Mass Spectrometry-Mass spectrometry was performed on a DYNAMO (Thermo BioAnalysis, Santa Fe, NM), a linear MALDI-TOF MS capable of dynamic extraction (a synonym for delayed extraction). Two types of sample-matrix preparations were employed. Peptides and dabsylated glycopeptides were dissolved in 5% formic acid and aliquots were spotted on the target, air dried, and overlaid with 1% ␣-cyano-4-hydroxycinnamic acid. Pyridylaminated glycans, reducing oligosaccharides, and underivatized glycopeptides were properly diluted with water, spotted on the target, and air dried. After addition of 2% 2,5-dihydroxybenzoic acid, the samples were immediately dried by application of vacuum.
Protein Concentration-Protein concentrations were determined by the bicinchoninic acid method (Pierce) or, at the final steps of enzyme purification, by amino acid analysis (35).

RESULTS
Purification of Core ␣1,3-Fucosyltransferase-Fuc-T C3 was purified from mung bean seedlings by Triton X-100 extraction of a crude microsomal preparation and several chromatographic steps including cation exchange and two types of affinity chromatography. The typical elution profile of activity from the S-Sepharose is shown in Fig. 2. Conveniently, this step provided separation from N-acetyl-␤-glucosaminidase which otherwise would have degraded the ligand of the subsequent affinity chromatography on GnGn-Sepharose (Fig. 2). After the last purification step on GDP-hexanolamine-Sepharose, the final yield was 18 g of protein from 5 kg of mung beans (Table  I). SDS-PAGE revealed two bands, a major band at 54 kDa and one at 56 kDa (Fig. 3). In order to check whether the two polypeptides are distinct or just different forms of the same enzyme, the bands were compared by MALDI-TOF MS of tryptic peptides obtained by in-gel digestion. The mass spectra of the 54-and 56-kDa band were indistinguishable indicating that both bands represent the same enzyme, putatively Fuc-T C3.
Isoelectric focusing of purified Fuc-T C3 revealed several isoforms with pI values ranging from 6.8 to 8.2 (Fig. 3). The enzymatic activity of the different bands was verified by loading two lanes with 1 g each of enzyme. One lane was silver stained, the other was used for measurement of enzymatic activity. For this purpose, the lane was cut into 4-mm pieces.
After sonication in the presence of buffer A (see "Experimental Procedures"), Fuc-T C3 activity was determined in the supernatant. All gel pieces corresponding to stained bands gave high Fuc-T C3 activity, even the intensity of bands correlated with activity suggesting that all bands represented active transferase. The three major bands with pI values of approximately 6.8, 7.1, and 7.6 were accompanied by faint satellite bands. In two-dimensional electrophoresis, these satellite bands migrated slightly slower than the major species thus representing the 56-kDa band from normal SDS-PAGE (Fig. 3). Apparently, Fuc-T C3 occurs in at least 7 isoforms.
Partial Amino Acid Sequences and cDNA Cloning of Fuc-T C3-The major protein band of apparently 54 kDa was digested in-gel with trypsin. Four peptides were isolated by reverse phase HPLC and sequenced to yield the following peptide sequences: peptide 1, KPDAxFGLPQPSTAS; peptide 2, PETVY-HIYVR; peptide 3, MESAEYYAENNIA, and peptide 4, GRFEMESIYL. Attempts to sequence the N terminus of intact Fuc-T C3 failed. On the basis of peptides 1, 2, and 3, degenerate oligonucleotides were synthesized and used as primers for reverse transcriptase-PCR as described under "Experimental Procedures." The PCR products obtained with primers S1-A2 and S1-A3 consisted of 744 and 780 bp, respectively, both sharing the same 5Ј end. The 780-bp fragment included the coding sequence for peptide 2. To obtain the full-length cDNA, 3Ј-and 5Ј-RACE was performed using the primers described under "Experimental Procedures." Full-length cDNA consisted of 2.2 kilobases and contained an open reading frame of 1530 bp which included the coding regions for all four peptides derived from the purified enzyme (Fig. 4).  Due to the lack of information about the natural N terminus of Fuc-T C3, the possible N terminus can only be deduced from potential initiation codons between the putative transmembrane region (see below) and the first stop codon toward the 5Ј end. The open reading frame that starts with the first Met residue located right beneath a stop codon encodes a protein of 510 amino acids, a molecular mass of 56.8 kDa, and a calculated pI of 7.51 (Fig. 4). A theoretical tryptic peptide map of the deduced complete amino acid sequence exhibited significant similarity to the map of purified Fuc-T C3 (data not shown).

Expression and Characterization of Recombinant Fuc-T C3-
The coding region of Fuc-T C3 was engineered into a baculovirus transfer vector. Various amounts of progeny virus from the co-transfection were used to infect Sf21 insect cells. In the best batch obtained, total fucosyltransferase activity of cells and supernatant was about 30 times higher than in the mock infected control batch. The endogeneous activity measured in the absence of recombinant transferase arises, however, essentially from insect Fuc-T C6 and only to a marginal extent from Fuc-T C3 (32, 36). Thus, the increase in Fuc-T C3 caused by the recombinant baculovirus is well above 100-fold.
Similar to the natural enzyme, the recombinant transferase displayed a broad maximum of activity around pH 7.0 when measured in Mes-HCl buffer and the presence of divalent cations, in particular of Mn 2ϩ , enhanced its activity. Among the acceptors employed, GnGn-peptide gave the highest incorporation rates under standard assay conditions, closely followed by GnGnF 6 -peptide and M5Gn-Asn (Table II). The apparent K m values for the acceptor substrates GnGn-peptide, GnGnF 6 -peptide, M5Gn-Asn, and for the donor substrate GDP-fucose were estimated to be 0.19, 0.13, 0.23, and 0.11 mM, respectively. No transfer was observed to MM-peptide which lacks the terminal GlcNAc residue on the 3-linked mannose regarded to be a structural requirement for core fucosyltransferases (1, 2, 37). By the standard assay, fucosyl transfer to GalGal-peptide could not be observed. However, a low rate of incorporation was demonstrated by MALDI-TOF MS (see later). Recombinant Fuc-T C3 was inactive toward common acceptors used for the determination of blood group ␣1,3/4-fucosyltransferases which transfer fucose to GlcNAc residues at the nonreducing termini of oligosaccharides (Table II). Thus, with regard to substrate specificity and kinetic properties, the recombinant transferase performed comparable to its natural counterpart (23).
Analysis of the Deduced Amino Acid Sequence of Fuc-T C3-Analysis of Fuc-T C3 by "TMpred" (provided by EMBnet, Switzerland) suggested a transmembrane region between Asn-36 and Gly-54 which, remarkably, contains a glutamic acid residue (Fig. 5). The C-terminal, major part of the enzyme most likely comprises the catalytic domain and can therefore be assumed to face the lumen of the Golgi apparatus. Thus, mung bean Fuc-T C3 appears to be a type II transmembrane protein like all hitherto analyzed glycosyltransferases involved in glycoprotein biosynthesis (38).
A BLASTP search (with deactivated filter) of all data bases accessible via NCBI showed similarity of mung bean Fuc-T C3 to essentially all known mammalian, protozoan, and bacterial ␣1,3/4-fucosyltransferases with probability values ranging from 7.4 ϫ 10 Ϫ17 for Chinese hamster Fuc-T to 10 Ϫ6 for a putative Vibrio cholerae fucosyltransferase. Remarkably, exon II alone was sufficient to retrieve by BLASTP all these ␣1,3/4fucosyltransferases with even higher probability values. In contrast, data bank searches with the individual exons I, III, and IV (see below) did not reveal a similarity to any known protein or nucleotide sequence. Alignments of mung bean Fuc-T C3 exon II with fucosyltransferases responsible for Lewis epitope synthesis revealed four regions of significant homology, (Fig. 6) as will be discussed later.
Exon-Intron Organization of Mung Bean Core ␣1,3-Fucosyltransferase-Three fragments covering the whole open reading frame as overlapping pieces were amplified from genomic DNA by PCR. While the genomic PCR product covering cDNA region 392 to 944 bp had the size expected from cDNA, the fragments covering bp Ϫ174 to 522 and 890 to 1550 appeared considerably larger indicating them to contain introns. Therefore, these fragments were subcloned, sequenced, and the sequences were analyzed for potential donor and acceptor splice sites using NetPlantGene V2.0 (39). The suggested splice sites were between base pairs 384/385 (CAG g . . . cag G), 1049/1050 (AG gt . . . ag GT), and 1277/1278 (AG gt . . . ag GT) and agree with the cDNA sequence. Thus, the open reading frame of mung bean Fuc-T C3 gene is interrupted by three introns. The four exons therefore encode for amino acid residues 1-128, 129 -350, 351-426, and 427-510.
Structural Characterization of the Fucosylated Product-The structure of the product generated by the recombinant putative core ␣1,3-fucosyltransferase was analyzed in two ways. For both strategies, a dabsylated glycopeptide having the "GnGn" structure was used as acceptor. The sample (2 nmol) was incubated overnight with homogenate from baculovirus-infected Sf21 cells. An initial analysis by MALDI-MS revealed an al-most complete fucosylation of the substrate in contrast to an approximately 5% conversion to fucosylated product in the mock infected control sample (data not shown). As was shown later by HPLC analysis of the pyridylaminated glycan, most of this product from endogenous insect cell fucosyltransferase was ␣1,6-fucosylated (see below).
The first analytical strategy to prove the identity of the product of the recombinant enzyme made use of the inability of N-glycosidase F to hydrolyze substrates with ␣1,3-fucose attached to the Asn-linked GlcNAc (27). Aliquots of the putative GnGnF 3 -peptide were mixed with similar amounts of substrate (dabsyl-GnGn-peptide) and the mixtures were incubated with either N-glycosidase F or N-glycosidase A. The extent of hydrolysis was determined by MALDI-TOF MS after 2 and 20 h. Only N-glycosidase A was able to digest both substrate and fucosylated product. In contrast, N-glycosidase F, although having completely hydrolyzed the substrate ([M ϩ H] ϩ ϭ 2262.3) after only 2 h, did not hydrolyze the core ␣1,3-fucosylated glycopeptide ([M ϩ H] ϩ ϭ 2408.4) even after 20 h of incubation.
As a second proof, the remaining 1 nmol of Fuc-T C3 product in sample A was digested with N-glycosidase A and pyridylaminated and the fluorescent derivatives were subjected to reverse phase HPLC. The significant reduction of elution time of the product compared with the substrate implies ␣1,3-fucosylation  Fig. 1  of the reducing terminal GlcNAc (Fig. 7) (4,36,40). No other known structural feature of N-linked oligosaccharides exerts such a strong and characteristic effect on the retention time of pyridylamino glycans (40). The compounds mass of 1564.5 agreed exactly with the mass expected for the sodium adduct of the pyridylamino glycan. To exclude any possibility of fucose being linked to a nonreducing terminal GlcNAc and to allow comparison of retention time with a reference oligosaccharide analyzed previously by independent methods, i.e. pyridylaminated MMF 3 from honeybee venom phospholipase (Fig. 1) (26), the product was digested with N-acetyl-␤-glucosaminidase. Indeed, the putative MMF 3 coeluted with the reference glycan.
FIG. 6. Conserved regions of core and blood group ␣1,3-fucosyltransferases. Four regions of apparent homology between mung bean Fuc-T C3 and most currently known ␣1,3/4-fucosyltransferases are shown. Conserved residues are represented by white letters on black background or, if common to only a few transferases, on gray background. The blocks B and D represent the highly conserved regions I and II previousyl described (43,44). In the case of the putative fucosyltransferase from Dictyostelium discoideum and the EST from S. japonicum, the gene products have not yet been analyzed. A lysine residue shown to be essential for activity of human Fuc-T V and VII is marked by an arrow (45). The number of residues between the depicted partial sequences are given in brackets. Transferases are identified by SwissProt (square brackets) or, if not applicable, by GenBank accession numbers.
The results from both experiments provide convincing evidence that the recombinant enzyme transfers fucose in ␣1,3-linkage to the reducing terminal GlcNAc of a complex N-glycan, thus being a core ␣1,3-fucosyltransferase.
The above described analytical strategy was applied to investigate fucosyl transfer to ␤1,4-diand monogalactosylated N-glycans. By MALDI-TOF MS, the transfer rate to dabsylated GalGal-peptide was found to be about 0.7% of that with dabsylated GnGn-peptide. However, when a mixture of GnGn-, GalGn-, GnGal-, and GalGal-peptide (prepared by partial enzymatic degalactosylation of dabsylated GalGal-peptide) was used as the substrate, about half of the monogalactosylated species were readily fucosylated just as the GnGn-peptide (Fig.  8, upper panel). To determine which of the two monogalactosylated isomers had been fucosylated, the oligosaccharides were analyzed by HPLC. To allow comparison of elution times with reference oligosaccharides, the mixture was digested with N-acetyl-␤-glucosaminidase and, after heat denaturation of this enzyme, with ␤-galactosidase. The fact that the positions of core fucose as well as of terminal GlcNAc residues strongly influence the elution positions of pyridylaminated N-glycans on reverse phase (4,32,34,40) allowed to identify the four oligosaccharides obtained by the above described procedure (Fig. 8,  lower panel). The glycans MMF 3 , GnMF 3 , MGn, and GnGn are regarded as the glycosidase digestion products of GnGnF 3 , GalGnF 3 , GnGal, and GalGal, respectively. Thus, Fuc-T C3 had acted on ␤1,4-monogalactosylated N-glycans if Gal was located on the 6-arm (GalGn) but not if it was on the 3-arm (GnGal).

DISCUSSION
As the first enzyme of its kind, mung bean core ␣1,3-fucosyltransferase has been purified, cloned, and heterologously expressed. Partly based on previous work (23), a purification scheme was established which gave an almost million-fold purification. Nevertheless, SDS-PAGE revealed two bands of apparently 54 and 56 kDa. Mass spectrometric peptide mapping indicated these bands to represent isoforms of the same enzyme. Zeng et al. (41) who observed a similar pattern for soybean xylosyltransferase suggested limited proteolysis as an explanation. Different glycosylation site occupancy may likewise account for the small mass difference. Whatever their difference might be, the ratio of the two isoforms differed dra- FIG. 7. HPLC analysis of pyridylaminated fucosyltransferase product. Glycopeptides obtained by incubation of dabsylated substrate (GnGn-peptide) with insect cell homogenates were digested with Nglycosidase A and the oligosaccharides were pyridylaminated and subjected to reverse phase HPLC. A, in the control experiment with mockinfected insect cells, a small peak which, according to its elution time and to previous work (36), represents GnGnF 6 , can be seen in addition to residual substrate; B, in the sample prepared with recombinant enzyme, almost the entire substrate was converted to a product (P) wich exhibits the very low retention indicative of core ␣1,3-fucosylation; C, isolated transferase product, putative GnGnF 3 ; D, transferase product after digestion with N-acetyl-␤-glucosaminidase; E, MMF 3 from honeybee phospholipase A 2 .
FIG. 8. Fucosyltransfer to ␤1,4-monoand digalactosylated substrate. A mixture of dabsylated glycopeptides having the structures GnGn, GalGn, GnGal, and GalGal was prepared by limited digestion with ␤-galactosidase and used as the substrate for Fuc-T C3. After 20 h, the transferase reaction mixture was analyzed by MALDI-TOF MS (upper panel). All of the GnGn was converted to GnGnF 3 (peak 1: calculated [M ϩ H] ϩ ϭ 2408.4), whereas only about half of the isobaric pair GalGn/GnGal (peak 2, 2424.5) gave rise to a fucosylated product (peak 3, 2570.6), and only a minute amount of GalGal (peak 4, 2586.6) had been fucosylated (peak 5, 2732.8). Other peaks in the spectrum mainly represent sodium adducts. While the structure of the fucosylated GalGal (5) has not been analyzed, the monogalactosylated and fucosylated glycan was assigned the structure GalGnF 3 (see matically between mung bean seedlings from two suppliers. Based on partial peptide sequences, a cDNA putatively encoding Fuc-T C3 was cloned. A first confirmation of the authenticity of this cDNA came from comparison of the theoretical tryptic map of the translated sequence with the peptide masses obtained from purified Fuc-T C3. Unfortunately, no matching peptides could be found for a large portion at the N terminus comprising the putative transmembrane and cytoplasmic domain. While suppression effects are not uncommon in MALDI-MS, this might also indicate that the purified enzyme was a truncated form lacking these domains. Expression of the cloned cDNA in the baculovirus-insect cell system finally confirmed that it encoded Fuc-T C3. The recombinant transferase used the acceptors GnGn-, M5Gn-, and the "mammalian" GnGnF 6peptide with similar efficiency. It is, to our knowledge, the first time that the biosynthetic intermediate M5Gn is shown to be a potential acceptor for fucose. Remarkably, GnGnF 6 which certainly does not occur in plants, appeared to be the best acceptor in kinetic terms. A ␤1,4-linked Gal residue on the 3-arm inhibits the action of Fuc-T C3. This explains the reduction of fucosylated glycans in plant cells expressing recombinant ␤1,4galactosyltransferase (42). Incorporation of fucose by the recombinant enzyme rendered a glycopeptide resistant against N-glycosidase F which is in keeping with the inability of this glycosidase to act on core ␣1,3-fucosylated substrates (27). In addition, the product was analyzed by reverse phase HPLC using the authentic reference glycan MMF 3 from honeybee venom phospholipase A 2 .
Considering the evolutionary distance between Fuc-T C3 and mammalian Lewis blood group ␣1,3/4-fucosyltransferases and their different acceptor substrates, we did not expect to find sequence homologies to this enzyme family. Indeed, the amino acid sequence of Fuc-T C3 displays an insignificant overall homology of 18 -21% when compared with these fucosyltransferases. However, a large part of exon II (residues 154 to 350) exhibits, e.g. 31% identical residues with chimpanzee Fuc-T VI. The conserved residues are found clustered in four regions as depicted in Fig. 6. Two of these clusters constitute the highly conserved regions identified previously by Breton et al. (42). Especially, region D (region II in Refs. 43 and 44) appears to be highly conserved between mammalian Lewis and plant core fucosyltransferases. This region also contains a Lys residue identified to be essential for acitivity of human Fuc-T V and Fuc-T VII (identified by an arrow in Fig. 6) (45). Mammalian Fuc-T contain in region B (region I in Refs. 43 and 44) a DSD-motif suggested to be part of the catalytic site (46). In mung bean Fuc-T C3, a SSD-motif is found at this site. Remarkably, bacterial and protozoan fucosyltransferases exhibit a lower degree of homology than most mammalian Fuc-Ts with the possible exception of a putative Schistosoma japonicum Fuc-T of which only an EST exists (44). Although exon III on its own is not significantly homologous to the Lewis blood group ␣1,3/4-fucosyltransferases, when coupled to exon II, its first part (residues 351 to 384) can be tentatively aligned with, e.g. human Fuc-T VI or V to reveal the conserved motifs Arg-Trp-(Arg/Lys) (with Trp-Arg being found in all mammalian Fuc-Ts) and Cys-X-Y-Cys, where X very often is a basic residue. In contrast, a Cys residue which is located between the conserved regions A and B and which has been shown to be involved in binding of GDP-fucose by human Fuc-T III, V, and VI (47), is not seen in Fuc-T C3. In other animal fucosyltransferases, this Cys residue is replaced by Ser or Thr. However, the only hydroxy amino acid found at this site of Fuc-T C3 is Tyr. It shall be noted, that no sequence similarities of residues 385-510 from mung bean Fuc-T C3 with animal ␣1,3/4-fucosyltransferases are to be expected because these much shorter enzymes do not contain a comparable region. Despite a similar substrate specificity, the recently cloned porcine and human core ␣1,6fucosyltransferases do not exhibit any obvious sequence similarities with mung bean core ␣1,3-fucosyltransferases (21,22).
Sequencing of the genomic region containing the open reading frame of Fuc-T C3 predicted three introns dispersed between four exons. Exon II with its conserved regions and the cytoplasmatic region on exon I are separated by a large intron of 771 bp. Introns interrupting the coding region have also been found in mouse FTVII (Q11131) (48) and in Caenorhabditis elegans CEFT-1 (Q21362) (49), the latter containing nine introns. In contrast, in many other ␣1,3-fucosyltransferases the entire coding sequence is contained within a single exon (44,50,51).
Mung bean Fuc-T C3 is the first plant fucosyltransferase and the first core ␣1,3-fucosyltransferase which has been cloned and sequenced. Our designated abbreviation "Fuc-T C3" takes into account the transfer of fucose into the 3 position of the core GlcNAc. Following the nomenclature of other fucosyltransferases, the enzyme may also be designated Fuc-T X. More significantly, it will now be possible to express large quantities of Fuc-T C3, thus enabling the in vitro synthesis of a variety of core ␣1,3-fucosylated N-glycans or N-glycopeptides from acceptors which are derived from mammalian glycoproteins. These "vegetabilized" structures will aid in the further elucidation of the role of core ␣1,3-fucose in the immunogenicity of plant and insect glycoproteins.