Identification of Core α1,3-Fucosylated Glycans and Cloning of the Requisite Fucosyltransferase cDNA fromDrosophila melanogaster

For many years, polyclonal antibodies raised against the plant glycoprotein horseradish peroxidase have been used to specifically stain the neural and male reproductive tissue ofDrosophila melanogaster. This epitope is considered to be of carbohydrate origin, but no glycan structure fromDrosophila has yet been isolated that could account for this cross-reactivity. Here we report that N-glycan core α1,3-linked fucose is, as judged by preabsorption experiments, indispensable for recognition of Drosophila embryonic nervous system by anti-horseradish peroxidase antibody. Further, we describe the identification by matrix-assisted laser desorption-ionization time-of-flight mass spectrometry and high performance liquid chromatography of two Drosophila N-glycans that, as already detected in other insects, carry both α1,3- and α1,6-linked fucose residues on the proximal core GlcNAc. Moreover, we have isolated three cDNAs encoding α1,3-fucosyltransferase homologues from Drosophila. One of the cDNAs, when transformed into Pichia pastoris, was found to direct expression of core α1,3-fucosyltransferase activity. This recombinant enzyme preferred as substrate a biantennary core α1,6-fucosylated N-glycan carrying two non-reducingN-acetylglucosamine residues (GnGnF6;K m 11 μm) over the same structure lacking a core fucose residue (GnGn; K m 46 μm). The Drosophila core α1,3-fucosyltransferase enzyme was also shown to be able to fucosylate N-glycan structures of human transferrinin vitro, this modification correlating with the acquisition of binding to anti-horseradish peroxidase antibody.

Glycoproteins from plants and invertebrates are often highly immunogenic and many antibodies (both IgG and IgE) directed against them bind to core ␣1,3-fucose and/or ␤1,2-xylose residues of their N-linked oligosaccharides (1-7); since these modifications are present across plant species and one or the other modification is present in a number of invertebrates, these antibodies are highly cross-reactive. Indeed, polyclonal anti-bodies to horseradish peroxidase (anti-HRP) 1 have been used for nearly two decades to specifically stain neurons, and their growth pathways, in Drosophila melanogaster (8 -11); similar staining has also been described in grasshopper (11), whereas in Caenorhabditis elegans 10% of neurons are stained by this antibody (12). A number of proteins in Drosophila, such as an Na ϩ ,K ϩ ATPase christened Nervana, a receptor tyrosine phosphatase, and cell adhesion molecules (e.g. fasciclin I and II, neurotactin and neuroglian), have been found to bind anti-HRP (9 -11, 13, 14); however, no data on their glycosylation have been described. A number of Drosophila nac (neurally altered carbohydrate) mutants have been described, which are defective in anti-HRP staining of adult flies and display wing morphological defects (15) and, when under cold stress, some minor behavioral and eye developmental abnormalities (9). Neither the affected gene(s) nor the underlying biochemical defect(s) have been identified.
As compared with the amount of information known on the genetics and genome of Drosophila, relatively little is known about the structures of its glycoconjugates; there exists one report from 1991 on the N-linked oligosaccharides of larval membrane glycoproteins in which only data on oligomannosidic and core ␣1,6-fucosylated N-glycans and no data on glycan structures that could explain anti-HRP binding were presented (16). However, subsequently, core ␣1,3-linked fucosylated glycans were found in insects (e.g. on bee venom glycoproteins) (17)(18)(19). Thus, since ␤1,2-xylose has to date not been found in this class of organism, we still considered it reasonable to hypothesize that anti-HRP binding in Drosophila was due to core ␣1,3-fucosylation of N-linked oligosaccharides.
In the present paper, we report the specific inhibition of anti-HRP staining of Drosophila embryonic nervous system by a neoglycoconjugate carrying core ␣1,3-fucose. We also describe the first analysis of the N-glycans of adult flies to be published and demonstrate the presence of N-glycans carrying both core ␣1,3and ␣1,6-linked fucose. Furthermore, we describe the isolation of a Drosophila cDNA encoding a fucosyltransferase activity that creates in vitro an epitope for anti-HRP antibody. 2

EXPERIMENTAL PROCEDURES
Immunostaining of Whole-mount Drosophila Embryos-D. melanogaster embryos (Canton S) were collected and staged at 25°C. After dechorionation in 50% bleach for 3 min, the embryos were thoroughly rinsed in water and fixed for 20 min in 4% paraformaldehyde in PBS/ n-heptane (1:1), followed by devitellinization in n-heptane/methanol (1:1). Fixed embryos were washed three times with methanol and then equilibrated in 0.1% Triton X-100 in PBS. The embryos were blocked at 4°C overnight in 0.1% Triton X-100, 0.1% BSA, and 5% normal goat serum in PBS and incubated with rabbit anti-HRP (1 g/ml, Sigma) or anti-bee venom antibodies (20 g/ml, Sigma) for 90 min at room temperature. After vigorous washing, embryos were probed with FITCconjugated (1:200, Sigma) or biotin-conjugated goat anti-rabbit IgG (1:300, Vector Laboratories, CA). In the latter case, the embryos were stained using a Vectastain Elite peroxidase kit with diaminobenzidine as a substrate (0.07 mg/ml) in the presence of 0.012% hydrogen peroxide, mounted in 70% glycerol (v/v) in PBS, and visualized by light microscopy. FITC-labeled samples were observed by using a Bio-Rad MRC 600 confocal scanning microscope with 488 nm excitation.
For anti-HRP staining inhibition studies, fucosylated and chemically defucosylated bromelain glycopeptides cross-linked to bovine serum albumin (BSA-MUXF 3 and BSA-MUX), 2 respectively, were prepared as described (6) and estimated to contain 0.1 nmol of GlcNAc/g of conjugate (equivalent to an average of 3.5 glycans incorporated/BSA molecule). The chemical defucosylation to yield MUX glycopeptide was estimated by sugar content analysis to have been 95% effective. These neoglycoconjugates in the concentration range 0.004 -40 M in terms of GlcNAc (or 0.04 -400 g/ml) were allowed to preabsorb the rabbit anti-HRP antibodies (1 g/ml) for 1 h at room temperature before applying as primary antibody in whole-mount embryo stainings.
Preparation of N-Glycans from Adult Flies-Three grams of frozen whole flies (Canton S strain) were denatured in 20 ml of stirred boiling water (10 min). After cooling, formic acid (to attain a final concentration of 5% (v/v)) and 1 mg of pepsin were added. Proteolysis was allowed to proceed overnight at 37°C. The resulting extract was centrifuged; the pellet was washed and the washings added to the first supernatant. The supernatant was applied to a Dowex AG50Wϫ2 column (1.5 ϫ 50 cm), and the column was washed with 50 ml of 2% (v/v) acetic acid. Glycopeptides were eluted with 0.4 M ammonium acetate, pH 6, and orcinolpositive fractions were pooled and concentrated prior to gel filtration (Sephadex G25 medium; elution with 1% (v/v) acetic acid). The resultant glycopeptide fractions were lyophilized and dissolved in 250 l of 100 mM citrate-phosphate, pH 5, and incubated at 95°C for 10 min to inactivate any remaining proteases. The N-glycans were released using peptide:N-glycosidase A (37°C, 24 h). The sample was then acidified by adding two volumes of 5% (v/v) acetic acid and applied to a 2-ml Dowex AG50Wϫ2 column. The column was washed with 2% (v/v) acetic acid, and the unretained orcinol-positive fractions were subject to gel filtration (Sephadex G15; elution with 1% (v/v) acetic acid) and then to reversed-phase chromatography (200 l of Lichroprep; elution with 5% (v/v) acetic acid). The material retained on the 2-ml Dowex column was subject to a second round of peptide:N-glycosidase A digestion, in order to release remaining anti-HRP binding components. Lyophilized Nglycans were dissolved in water and subject to either MALDI-TOF analysis (see below) or pyridylamination (17,20,21). The final yield of N-glycans was estimated as being 10 -15 nmol, as judged by determination of the amino sugar content (22) and assuming two GlcNAc residues/N-glycan.
Enzyme-linked Immunoassays of Drosophila Peptides-An aliquot (1%) from each stage of the (glyco)peptide and glycan purification was removed and diluted 1:100 in water. Equal amounts of diluted (glyco)peptide or glycan (0.5 mg/ml bromelain glycopeptides were used as a control) and a freshly prepared aqueous solution containing 0.5 mg/ml each of 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide and N-hydroxysucciminde were added to wells of Cova-Link ELISA strips (Nunc; 50 l of each solution/well). The strips were incubated for 90 min at 37°C and then blocked and incubated with antibodies as previously described (6), except that 1:2500 (ϳ7 g/ml) anti-HRP or anti-bee venom were used as primary antibodies.
Matrix-assisted Laser Desorption-Ionization Mass Spectrometry-Aliquots of 0.8 l of underivatized or pyridylaminated N-glycans were applied to a flat sample platen and dried immediately under mild vacuum, followed by addition of 1 l of matrix (2% (w/v) 2,5-dihydroxybenzoic acid in 30% (v/v) acetonitrile or 0.03 M 1-hydroxyisoquinoline, 0.1 M 2,5-dihydroxybenzoic acid in 50% (v/v) acetonitrile) and drying once more (23,24). MALDI-TOF mass spectra were acquired on a Dynamo (Thermo BioAnalysis, Hemel Hempstead, United Kingdom) linear time-of-flight mass spectrometer with a dynamic extraction setting of 0.1. External mass calibration was performed with pyridylaminated N-glycans or with a partial dextran hydrolyzate. On-plate ␣-fucosidase digestion was performed using 0.2 milliunits of bovine kidney fucosidase in 10 mM ammonium acetate, pH 5, with incubation in a "wet chamber" at 37°C.
EcoRI-KpnI-digested PCR products for FucTA and FucTC and the PstI-KpnI-digested PCR fragment for FucTB were cloned into pPICZ␣ expression vectors (Invitrogen, The Netherlands). The integrity and reading frames of the constructs, as well as 5Ј and 3Ј ends of full-length RT-PCR products, were confirmed by DNA sequencing using an ABI PRISM Big Dye Terminator Sequencing Ready Mix and ABI PRISM 310 genetic analyzer (PerkinElmer Life Sciences, Applied Biosystems). Nucleic acid sequences were analyzed using the DNAStar suite of programs and NCBI BLAST and NCBI BLAST2 programs (www.ncbi.nlm.nih.gov).
Assay of Fucosyltransferase Activity-Culture supernatants were concentrated 10-fold using Vivaspin concentrators (Vivascience, Gloucestershire, United Kingdom), the buffer replaced with 50 mM MES, pH 6.0, 5 mM MnCl 2 and then assayed for core ␣1,3-fucosyltransferase activity using dabsyl-GnGn (see Fig. 1 for an explanation of oligosac-charide structures), which is a dabsylated tetrapeptide with the sequence Gly-Glu-Asn-Arg derived from Pronase digestion of asialoagalacto bovine fibrin. The standard 20-l assay mixture, based on that described previously (25), consisted of 5 l of 10-fold concentrated supernatant and final concentrations of 100 mM MES, pH 6.0, 1 mM AMP, 50 mM N-acetylglucosamine, 1 mM GDP-Fuc, 10 mM MnCl 2 , and 0.05 mM dabsyl-GnGn. After incubation for 5 or 24 h at 37°C, an aliquot of the assay mixture was diluted 10-fold with water. Then, 0.8 l of diluted assay mixture was mixed on the sample platen with 0.8 l of 1% (w/v) ␣-cyanohydroxycinnaminic acid in 70% (v/v) acetonitrile and allowed to dry. MALDI MS spectra were then acquired with external mass calibration performed with a mixture of tryptic peptides derived from the peanut allergen Ara h 1.
The enzyme was also assayed using dansyl-GnGn or dansyl-GnGnF 6 , which are dansylated dipeptides (Asn-Ser) derived from IgG carrying an asialoagalacto-N-glycan with or without prior fucosidase digestion (26). Samples were injected after incubation onto an RP-HPLC column (5-m Hypersil ODS, 4 ϫ 250) and eluted isocratically in 9% acetonitrile buffer in 50 mM potassium phosphate, pH 2. The glycopeptide substrate and product were detected by fluorescence. Elution positions were compared with those of dansyl-GnGn left untreated, incubated with recombinant Arabidopsis core ␣1,3-fucosyltransferase (27) (generating standard GnGnF 3 ), or incubated with an extract of chicken heart containing ␣1,6-fucosyltransferase activity (prepared as described in Ref. 28 in order to generate GnGnF 6 ). The exact concentration of acceptor substrate was determined by analysis of the GlcNAc and amino acid content (22).
In addition, dansyl glycopeptides derived from IgG were galactosylated in vitro with bovine milk ␤1,4-galactosyltransferase to generate GalGal and GalGalF 6 . Dansylated MM and MMF 6 glycopeptides were generated by hexosaminidase treatment of GnGn and GnGnF 6 derived from IgG. The subsequent fucosyltransferase assays were performed as follows; 2 l of concentrated FucTA supernatant was incubated with 0.2 nmol of dansylated glycopeptide and 10 nmol of GDP-Fuc in a final volume of 20 l. For K m determinations, dansyl glycopeptides derived from IgG (GnGn and GnGnF 6 ) were used with 2 l of concentrated FucTA supernatant (16-h incubation time).
Preparation and Analysis of Fucosylated Transferrin-Human transferrin (5 mg; Sigma) was digested with 0.2 units of sialidase from Clostridium perfringens (Sigma) in 1 ml of 50 mM sodium acetate, pH 5.0, at 37°C overnight. After addition of 2 units of ␤-galactosidase from Aspergillus oryzae (29), the sample was incubated overnight to yield "GnGn-transferrin." Fucosylation of the glycomodified transferrin was performed as described above for glycopeptide, except that 10 g of GnGn-transferrin was used as the substrate and Triton X-100 was added to a final concentration of 0.08% (w/v). An aliquot containing ϳ3 g of partially ␣1,3-fucosylated transferrin (termed "GnGnF 3 -transferrin") was then incubated with 4 l of 50 mM sodium citrate, pH 5.0, and 1 milliunit of ␤-N-acetylglucosaminidase from Streptococcus pneumoniae at 37°C over night to obtain MMF 3 -transferrin. In all reactions a chicken heart extract possessing ␣1,6-fucosyltransferase activity, either heat-denatured (as a control) or native (to produce, e.g. MMF 6transferrin), was added.
The N-glycans of the variously modified transferrin samples were analyzed by mass spectrometry. Briefly, 4 g of protein were incubated with 0.2 g of pepsin in 20 l of 5% (v/v) formic acid overnight at 37°C. The digest was dried in a Speed-Vac evaporator. Enzymatic release, clean-up, and analysis by MALDI-TOF MS of N-glycans were performed as described (23).

Re-evaluation of the Drosophila Anti-HRP Carbohydrate
Epitope-Previously published data on the characterization of plant epitopes reacting with anti-HRP antibody (3,6), as well as data indicating that the plant glycoprotein bromelain inhibited anti-HRP binding to Drosophila (9), led us to hypothesize that core ␣1,3-fucose is the key element recognized by anti-HRP in neural tissue of the fruit fly. To test this we used neoglycoconjugates of bromelain glycopeptide cross-linked to bovine serum albumin, BSA-MUXF 3 and BSA-MUX (6), to compete for binding to anti-HRP on Drosophila whole-mount Core ␣1,3-Fucosylation in Drosophila embryo preparations. The core ␣1,3-fucosylated neoglycoconjugate BSA-MUXF 3 had the capacity to inhibit binding of anti-HRP to the complete nervous system of Drosophila embryo (Fig. 2B), whereas the (at least 95%) defucosylated form (BSA-MUX) showed no or little inhibitory effect (Fig. 2C). The total abolition of neural staining with the anti-HRP antibody (1 g/ml) was achieved at a concentration of 40 g/ml BSA-MUXF 3 (i.e. 4 M in terms of GlcNAc).
ELISA was also used to examine the interaction of adult Drosophila extract with anti-HRP. The antibody was preincubated with 2 M (in terms of GlcNAc) of either BSA-MUXF 3 or BSA-MUX; the percentage of blank-corrected inhibitions were respectively 85% and 30%. (In considering these figures, it should be borne in mind that 5% residual fucose was still present on BSA-MUX.) In addition, since free glycans are not able to cross-link to the Cova-Link strips, the demonstrated lack of an ELISA reading after peptide:N-glycosidase A digestion of Drosophila glycopeptides was concluded to indicate that the anti-HRP epitope was indeed present on the N-glycans prepared for the later structural analyses.
Since the N-glycan structures of two honeybee venom components previously elucidated (17,19) indicated the presence of core ␣1,3-fucose (but not ␤1,2-xylose) and since an anti-bee venom phospholipase antiserum has been previously shown to be partly directed against N-glycan components cross-reacting with plant glycoproteins (4), we decided to also probe Drosophila whole-mount embryos with anti-whole bee venom antiserum. The immunofluorescence pattern of the nervous system is indistinguishable from that of anti-HRP, although somewhat less intense (Fig. 2, panels D and E). Also, ELISA indicated that the interaction of anti-bee venom with crude adult Drosophila extract, as well as peptic peptides, was 70% inhibited by 20 g/ml BSA-MUXF 3 (i.e. 2 M in terms of GlcNAc content), whereas the interaction of anti-HRP with these extracts was 70% inhibited by 25 g/ml bee venom phospholipase (also 2 M in terms of GlcNAc).
Taken as a whole, these results suggest that core ␣1,3-fucose is a sugar moiety indispensable for the recognition of the Drosophila neural carbohydrate epitope by anti-HRP antibodies. This observation is strongly supported by the data reported below on the detection of core ␣1,3-fucosylated N-glycan structures in Drosophila and isolation of the requisite fly fucosyltransferase cDNA, the recombinant enzyme having the potential to modify in vitro N-glycan substrates lacking core ␣1,3-fucose and so create de novo the epitope recognized by anti-HRP antibody.
Analysis of Drosophila N-Glycans-In order to test whether the fly contains core ␣1,3-fucosylated structures, we prepared N-glycans from adult Drosophila. Previously, the only published data on Drosophila N-glycans were acquired after hydrazinolysis of membrane glycoproteins from third instar larvae and indicated a full range of oligomannosidic structures, as well as some MMF 6 (16). Our MALDI-TOF MS analysis of the glycans from adults ( Fig. 3A and Table I) also shows the presence of oligomannose structures (Man 9 , for instance, accounts for 20% of the total N-glycans), but in this case a simple glycan (m/z 1080.0) with one deoxyhexose residue (presumably fucose) is the major structure (36% of the total); low quantities of glycans with one additional Glc-NAc residue, with and without fucose, were also detected. Most significantly, however, about 1% of the total N-glycans were found to carry two deoxyhexose residues. To examine their nature further, the N-glycans were subject to a number of analyses, including glycosidase digestion either of the entire underivatized N-glycan pool (with subsequent MALDI-TOF MS analysis) or of the entire pyridylaminated N-glycan pool (with subsequent RP-HPLC analysis), as well as fractionation by HPLC of the pyridylaminated N-glycans (Fig. 3B) and MALDI analysis of each HPLC fraction (see Table I).
The major RP-HPLC fraction (fraction 12) contained a species with an m/z value indicative of a monofucosylated Man 3 structure. Digestion with bovine ␣-fucosidase of the complete pool of pyridylaminated Drosophila glycans showed a shift in the RP-HPLC retention time of the major peak from 13 to 7.5 g.u. (the former, relatively late, retention time being in agreement with previous analyses of core ␣1, 6-fucosylated glycans (17)(18)(19), the latter being the same retention time as MM, the expected product). Similarly, on-plate ␣-fucosidase digestion of the underivatized whole N-glycan pool with subsequent analysis by MALDI indicated a shift in the m/z value of the major peak from 1080.0 to 933.8, suggestive of the removal of one fucose residue. The MALDI, RP-HPLC, and digestion data, therefore, indicate that MMF 6 is the major N-glycan in whole adult flies. Among the oligomannosidic structures, MALDI-TOF MS of the HPLC fractions demonstrated there were single isomers of Man 3 , Man 4 , Man 9 , and a probable GlcMan 9 and two or three isomers each of Man 6 , Man 7 , and Man 8 . The peaks containing Man 5-9 were sensitive to endoglycosidase H, as judged by digestion of the whole pyridylaminated N-glycan pool (data not shown).
As determined by MALDI-TOF MS, HPLC fraction 9 contained four structures with m/z values characteristic of Man 3 , Man 4 , and difucosylated forms of Man 2 and Man 3 . Fraction 9 was therefore re-applied to reversed phase before and after ␣-fucosidase digestion. From the undigested sample, it was possible to separate subfractions 9A and 9B (Fig. 4A, upper  chromatogram), whereas, similarly to the results of ␣-fucosidase digestion of the endoglycosidase-H treated entire pyridylaminated N-glycan pool, ␣-fucosidase digestion caused a  Table I for (Fig. 3A) or pyridylaminated glycans in spectra of the complete N-glycan pool were determined by external calibration that was refined by internal calibration relative to the MMF 6 and Man 9 peaks. Retention times and fraction numbers refer to the results of RP-HPLC (Fig. 3B). Percentage occurrences were calculated based on peak areas from MALDI-TOF or RP-HPLC data. ND, peak detected by MALDI-TOF MS not found in any individual RP-HPLC fraction. shift in the retention time of the peak corresponding to subfraction 9B from 8.0 g.u. to 5.0 and 5.2 g.u (Fig. 4A, lower  chromatogram). The species at 5.0 and 5.2 g.u. have the same retention time as MMF 3 and MUF 3 , structures previously analyzed by the same method in this laboratory from bee venom and ragweed pollen glycoproteins (17,19,21 (17,19), we conclude that ϳ1% of the N-glycans from adult Drosophila carry two core fucose residues (see Table I).
Characterization of Recombinant Drosophila Core ␣1,3-Fucosyltransferase-Searching the newly sequenced Drosophila genome with mammalian Lewis and plant core ␣1,3-fucosyltransferases indicated the presence of three homologous genes. The probable reading frames were analyzed and primers for RT-PCR were designed, and the entire open reading frames of FucTA, FucTB, and FucTC were isolated. Subsequently, cDNA fragments encoding the soluble forms of these proteins (i.e. excluding the predicted cytoplasmic and transmembrane domains) were expressed in the yeast P. pastoris. The media of the yeast cells were then assayed for fucosyltransferase activity by MALDI-TOF MS using dabsylated glycopeptides carrying GnGn (specific for core fucosyltransferases) or GalGal (i.e. GnGn with two non-reducing terminal ␤1,4-galactose residues; also suitable for Lewis x -generating ␣1,3-fucosyltransferases) N-glycans as substrates (see Fig. 1 for structure of GnGn). In these studies, in which the appearance of new peaks 146 m/z units larger than substrate peak and its laser-induced break-down product was considered indicative of the transfer of fucose, GnGn was found to be the most suitable substrate for one of the three novel ␣1,3-fucosyltransferase fly homologues, named FucTA (Fig. 5A). Some activity of FucTA toward GalGal was also detected; subsequent sequential ␤-galactosidase and ␤-N-acetylhexosaminidase digestion indicated that the fucose was transferred to the core and not to the antennae (data not shown). On the other hand, no activity was found in the media of FucTB or FucTC transformants with either of the aforementioned substrates. Interestingly, prefucosylation of GnGn with chicken heart extract (GnGnF 6 ) resulted in a superior conversion of this substrate by FucTA than with GnGn (Fig. 5B), whereas no activity was detected for any of the three Drosophila fucosyltransferase homologues when disaccharides lacto-Nbiose (Gal␤1,3GlcNAc; substrate for Lewis a -generating enzymes) or N-acetyllactosamine (Gal␤1,4GlcNAc; substrate for Lewis x -generating enzymes) were used as substrates in a radioactive assay using GDP-L-[ 14 C]fucose. The activity in the medium of FucTA-transformed Pichia was ϳ10 milliunits/liter, as judged by assays using GnGnF 6 as substrate.
To examine the substrate preference of FucTA further, an RP-HPLC method using fluorescently-labeled dansylated (rather than dabsylated) GnGn and GnGnF 6 glycopeptides derived from IgG as acceptors was employed. Under this method, as shown by a previous study (26) and by control experiments with recombinant Arabidopsis thaliana core ␣1,3-fucosyltransferase, the appearance of a new peak of lower retention is indicative of the addition of an ␣1,3-linked fucose residue (Fig.  5, C-F; compare chromatograms C and E with the respective chromatograms D and F after incubation with the recombinant Drosophila enzyme). Again, since the concentration of the glycopeptide acceptors was identical, the results of these assays suggested a preference of the enzyme for GnGnF 6 over GnGn (compare F with D). More exactly, when the acceptor concentration was varied and the results analyzed by means of a Hanes' plot, the K m values for GnGnF 6 and GnGn were found to be 11 and 46 M, respectively, suggestive of a preference for the ␣1,6-fucosylated acceptor (see also Table II). In addition, a variety of dansylated galacto-, agalacto-, and ahexosaminylglycopeptides from IgG were tested as substrates. Galactosylated variants were either generated from asialofibrin or by galactosylation in vitro of IgG glycopeptides to yield GalGal and GalGalF 6 , and the percentage conversion was compared by analysis of RP-HPLC peak areas. The comparisons show that ␣1,6-fucosylated glycans were better, and ␤1,4-galactosylated glycans worse, substrates. On the other hand, glycopeptides lacking any terminal GlcNAc residues (i.e. carrying MM or MMF 6 ) were not substrates at all. It was also demonstrated that, for GnGn or GalGal, the nature of the peptide portion (IgG or fibrin) of the dansylated glycopeptide had no effect on the percentage conversion (data not shown).
Complete sequencing of the cDNA encoding FucTA indicated that this enzyme is a protein with a predicted length of 503 amino acids (Fig. 6) with a putative N-terminal transmembrane domain (residues 11-28), a feature well known in other Golgi glycosyltransferases. A comparison with the annotation of Flybase entry CG6869 indicates that our FucTA cDNA is shorter than predicted, due to the presence of an extra miniintron of 66 nucleotides. Other small changes encountered between the nucleotide sequences of the experimentally obtained and predicted cDNAs may of course be due to differences in the strain of Drosophila used. Compatible with the enzymatic data, alignments (Fig. 7) indicate that FucTA is probably more closely related to plant core ␣1,3-fucosyltransferases (e.g. as judged by the presence of a AXF(I/V)SNCXARNXRLQ motif and its greater size, 503 residues) than FucTB and FucTC (which are characterized also by shorter sequences of 444 and 425 residues, respectively, and may be closer to the mammalian Lewis-type ␣1,3/4-fucosyltransferases). The overall identity of Drosophila homologues with mammalian and plant ␣1,3-fucosyltransferases is not significantly high (about ϳ 20%); however, comparisons of the regions of homology displayed in Fig. 7 revealed, on average, 32% amino acid identity. Typical ␣1,3-fucosyltransferase motifs (I/V)DXYG, YKFX-LAFENS, DY(I/V)TEK, and CXXC as described previously (30,31) were detected in all aligned sequences. Interestingly, FucTA has a unique N terminus (up to residue 163 of 503) displaying no homology to FucTB, FucTC, plant, or mammalian fucosyltransferases. In contrast, other than the cytoplasmic, transmembrane, and stem regions, which tend to be divergent anyway, it is the C-terminal region of plant core fucosyltransferases that accounts for their greater length.
Anti-HRP Binding of ␣1,3-Fucosylated Glycoproteins-In order to probe anti-HRP binding of the MMF 3 F 6 structure more closely, we decided to create de novo the same structures as found in the course of this study in fly in a protein-bound form suitable for Western blotting. Human transferrin was chosen as a carrier for generation of the epitope in vitro, since it has predominantly simple biantennary N-glycans attached at two different asparagine residues, which after desialylation and degalactosylation are suitable substrates for core fucosyltransferase, and since it is a protein found neither in plants nor insects. Even though the properties of anti-HRP have been much studied in the past using plant or insect glycoproteins (6,32), neoglycoconjugates (6), or, at relatively high concentration, pyridylaminated HRP oligosaccharides (3,5), all these studies have relied on use of chemical or enzymatic treatments to destroy the epitope rather than show that a recombinant enzyme can create it.
Using recombinant Arabidopsis core ␣1,3-fucosyltransferase and chicken heart extract (possessing ␣1,6-fucosyltransferase activity) as controls, we could demonstrate that human asialoagalacto-transferrin (GnGn-transferrin) is modified in vitro by the Drosophila enzyme to generate singly fucosylated (GnGnF 3 ) and, in combination with chicken heart extract, dou-TABLE II Substrate specificity of Drosophila FucTA Dansylated glycopeptides were derived from IgG and modified as described, prior to their use in fucosyltransferase assays. Relative conversion was determined by integration of peak areas in comparison to the data for GnGnF 6 as substrate. ND, not detected; *, examined by radioactive assay using GDP-L-[ 14 C]fucose. bly fucosylated N-glycans (GnGnF 3 F 6 ). With subsequent ␤-hexosaminidase digestion, the same difucosylated structure as one of those found in adult Drosophila, i.e. MMF 3 F 6 , was created de novo. All preparations of modified human transferrin were checked by analysis of its N-glycans before and after incubation with fucosyltransferase.
Control and modified transferrin samples were subject to SDS-polyacrylamide gel electrophoresis and immunoblotting with anti-HRP antibody. We could clearly demonstrate that core ␣1,3-fucosylation with either the Drosophila or Arabidopsis enzymes (Fig. 8, lanes 5-8), and not ␣1,6-fucosylation (Fig.  8, lane 4), of the human transferrin enabled its recognition by anti-HRP antibodies. The prior hexosaminidase treatment was found to greatly enhance the reactivity of anti-HRP antibodies in a Western blot (data not shown); thus, the preference of the antibody reflects the essential lack of terminal GlcNAc residues on both horseradish peroxidase (3,33) and, as is shown in the present study, Drosophila glycoproteins. DISCUSSION In the present study we have applied various approaches to examine the possible molecular basis of neural anti-HRP staining in Drosophila. We have demonstrated that (a) core ␣1,3fucosylated neoglycoconjugates are able to specifically abolish the anti-HRP staining of the nervous system in whole-mount Drosophila embryos; (b) staining of embryos with anti-bee venom antiserum, which contains antibodies binding to ␣1,3fucosylated glycoproteins (4, 6), yields the same pattern as anti-HRP staining; (c) adult flies contain core ␣1,3-fucosylated N-glycans of a type also found in other insect glycoproteins (bee venom and Mamestra brassicae cells) (17)(18)(19) and share the feature of core ␣1,3-fucosylation with Schistosoma (34), Haemonchus (35), and plants (24,33); (d) the fly genome encodes a core ␣1,3-fucosyltransferase that would be, presumably, responsible for the formation of the detected core ␣1,3-fucosylated glycans; (e) the recombinant fly core ␣1,3-fucosyltransferase can be used to modify human transferrin in order to create de novo a core ␣1,3-fucosylated N-glycan structure found in adult flies (i.e. MMF 3 F 6 ), a modification correlating with the acquisition of binding to anti-HRP. Our data, therefore, strongly suggest that N-glycan core ␣1,3-fucosylation is the key post-translational process modifying Drosophila neural glycoproteins and creating the epitope recognized by the anti-HRP antiserum. Studies by others also indicate that the cross-reaction to fly neuronal tissue is due to carbohydrate (as suggested by experiments showing abolition of binding after periodate oxidation) and moreover due to an N-glycan (as indicated by the inhibition observed after preabsorption of anti-HRP with HRP glycopeptides or with inactivated bromelain) (9, 11). How-ever, we are the first to study this inhibition using neoglycoconjugates with both native and defucosylated forms of the bromelain glycopeptide.
The following points also suggest that it is relatively unlikely that any other glycomodification is responsible for anti-HRP binding in Drosophila. First, ␤1,2-linked xylose (the only other obvious candidate for binding to anti-HRP) has not been found on the N-glycans from any insect, in this or any other study; neither is there any detectable homologue of the plant ␤1,2xylosyltransferase in fly (although, of course, it is theoretically possible that a functional evolutionary convergence has occurred). Second, our results indicate that peptide:N-glycosidase A treatment of the glycopeptide preparation removed all the anti-HRP cross-reactive material as determined by ELISA. Third, the low level (ϳ1%) of ␣1,3-fucosylation in adult flies is compatible with previous reports (8 -11), indicating that the anti-HRP epitope is restricted to a small fraction of the cells (i.e. only neurons and male reproductive tissue). Fourth, use of other methods (i.e. hydrazinolysis) on Drosophila larvae, which also exhibit anti-HRP staining, failed to result in the detection of any other structure (not even difucosylated) that could conceivably bind anti-HRP (16). Fifth, two mutations that eliminate anti-HRP binding of Drosophila embryos (in the deficiency mutant Brd 15 and the TM3 chromosomal balancer; Refs. 9, 11, 36, and 37) map closely to the same region as the FucTA gene. Indeed, the location of the estimated breakpoints in Brd 15 (71A1-2 and 71C1-2) would suggest that the FucTA gene (mapped to 71B2) is absent in this mutant, whereas in TM3 (which has multiple breakpoints including 71C) it is, at least, a remarkable coincidence that anti-HRP binding is abolished. Certainly, we will be examining these chromosomal aberrations in our ongoing studies.
The substrate preference of FucTA is of interest, since relatively few N-glycans are present in the preparation that carry non-reducing terminal GlcNAc or Gal residues; indeed, like many N-glycans from plant and invertebrate sources, the vast majority of the N-glycans have non-reducing terminal mannose residues, even if the glycans are core fucosylated, and so could not act as substrates for the FucTA core ␣1,3-fucosyltransferase. Indeed, all core ␣1,3and ␣1,6-fucosyltransferases described to date do not utilize mannose-terminating N-glycans, whether they be of insect, plant, or mammalian origin (25,38,39). This raises the possibility that, even though the prior MMF 3 ) or Arabidopsis core ␣1,3-fucosyltransferase and native chicken heart extract (lane 8; MMF 3 F 6 ). After incubations with native and/or denatured fucosyltransferase preparations, lanes 3-8 were treated with ␤-N-acetylhexosaminidase as described under "Experimental Procedures." Note that only transferrin modified by core ␣1,3-fucosyltransferases was recognized by anti-HRP antibody. action of, at least, GlcNAc-TI is required, the Drosophila glycans have become processed in the Golgi in vivo or degraded during extraction. The former scenario is very likely since a membrane-bound Golgi ␤-hexosaminidase has been previously found in insects (40), whereas the latter is more unlikely since, in contrast to native fly extracts, the boiled fly extract used for the glycan preparation contained no detectable ␤-hexosaminidase activity. The other aspects of the substrate preference of FucTA are also worthy of comment; the enzyme can significantly, uniquely in comparison to other core fucosyltransferases (25,27,38), utilize glycopeptides, which have both antennae galactosylated, whereas it has a bias toward ␣1,6fucosylated N-glycans. These substrate preferences may, of course, reflect that galactosylated glycans are only barely present in the fly and possibly became irrelevant during evolution as a 'NO-GO' signal, whereas ␣1,6-fucosylated glycans account for about 40% of the total and perhaps became a positive signal for the enzyme. Indeed, all ␣1,3-fucosylated glycans in the fly are also ␣1,6-fucosylated; in this connection, it is also noteworthy that the insect and mammalian ␣1,6-fucosyltransferase activities previously characterized cannot utilize core ␣1,3-fucosylated acceptors (41).
Many genes encoding glycosyltransferase homologues can be identified in Drosophila, including core ␣1,6-fucosyltransferase and GlcNAc-TII; however, considering that the GlcNAc-TI has recently been described (42), the core ␣1,3-fucosyltransferase described in the present study is, therefore, only the second Drosophila Golgi glycosyltransferase to be characterized. The cloning of its cDNA, the localization of the corresponding gene and the determination of the structure of ␣1,3-fucosylated glycans gives us both genetic and biochemical means to define further its potential role in the biosynthesis, expression, and biological significance of the anti-HRP epitope in Drosophila.