Molecular Basis of Anti-horseradish Peroxidase Staining in Caenorhabditis elegans*

Cross-reactivity with anti-horseradish peroxidase antiserum is a feature of many glycoproteins from plants and invertebrates; indeed staining with this reagent has been used to track neurons in Drosophila melanogaster and Caenorhabditis elegans. Although in insects the evidence indicates that the cross-reaction results from the presence of core α1,3-fucosylated N-glycans, the molecular basis for anti-horseradish peroxidase staining in nematodes has been unresolved to date. By using Western blots of wild-type and mutant C. elegans extracts in conjunction with specific inhibitors, we show that the cross-reaction is due to core α1,3-fucosylation. Of the various mutants examined, one with a deletion of the fut-1 (K08F8.3) gene showed no reaction to anti-horseradish peroxidase; the molecular phenotype was rescued by injection of either the K08F8 cosmid or the fut-1 open reading frame under control of the let-858 promoter. Furthermore, expression of fut-1 cDNA in Pichia and insect cells in conjunction with antibody staining, high pressure liquid chromatography, and matrix-assisted laser desorption ionization time-of-flight mass spectrometry analyses showed that FUT-1 is a core α1,3-fucosyltransferase with an unusual substrate specificity. It is the only core fucosyltransferase in plants and animals described to date that does not require the prior action of N-acetylglucosaminyltransferase I.

From the ‡Department fü r Chemie, Universitä t fü r Bodenkultur, A-1190 Wien, Austria, §Institut fü r Biochemie, Justus-Liebig Universitä t, D-35292 Giessen, Germany, and ¶Institut fü r Botanik, Universitä t Wien, A-1030 Wien, Austria Cross-reactivity with anti-horseradish peroxidase antiserum is a feature of many glycoproteins from plants and invertebrates; indeed staining with this reagent has been used to track neurons in Drosophila melanogaster and Caenorhabditis elegans. Although in insects the evidence indicates that the cross-reaction results from the presence of core ␣1,3-fucosylated N-glycans, the molecular basis for anti-horseradish peroxidase staining in nematodes has been unresolved to date. By using Western blots of wild-type and mutant C. elegans extracts in conjunction with specific inhibitors, we show that the cross-reaction is due to core ␣1,3-fucosylation. Of the various mutants examined, one with a deletion of the fut-1 (K08F8.3) gene showed no reaction to anti-horseradish peroxidase; the molecular phenotype was rescued by injection of either the K08F8 cosmid or the fut-1 open reading frame under control of the let-858 promoter. Furthermore, expression of fut-1 cDNA in Pichia and insect cells in conjunction with antibody staining, high pressure liquid chromatography, and matrix-assisted laser desorption ionization time-of-flight mass spectrometry analyses showed that FUT-1 is a core ␣1,3-fucosyltransferase with an unusual substrate specificity. It is the only core fucosyltransferase in plants and animals described to date that does not require the prior action of N-acetylglucosaminyltransferase I.
The structures of the oligosaccharides expressed by different organisms and tissues or on individual glycoproteins vary greatly. Over the years, it has become obvious that the glycosylation of invertebrate glycoproteins is significantly different from that of mammals, but although organisms such as Caenorhabditis elegans have been studied to a great extent in terms of molecular and developmental biology, an exact understanding of their glycosylation has lagged behind. Recently, however, a number of studies have indicated that the glycoconjugates of C. elegans have a number of unusual features (1)(2)(3)(4)(5)(6). None of the studies published so far agree absolutely as to the types of N-linked oligosaccharides, but the consensus would appear to be that this organism has a range of N-glycans carrying one or many fucose residues, with or without methyl or phosphorylcholine groups, as well as the standard oligomannose structures (7). Multiantennary structures may also exist, but there are no sialylated complex glycans of the types found in vertebrates. Glycolipids of the arthro-series carrying phosphorylcholine (8) as well as unusual fucosylated O-glycans have also been described (9). Some of these features have already been found in parasitic nematodes and are a source of immunomodulatory, immunogenic, or allergenic activity (10).
Still, however, the exact structures of many of the N-glycans from the worm are unknown; in particular, the nature of the linkages of the multiple fucose residues is not resolved, although RP-HPLC 1 evidence (2) indicates that both core ␣1,3and core ␣1,6-fucosylation of N-glycans occur. The former is also a feature of plant glycoproteins and is the major epitope of many antibodies raised against plant glycoproteins, such as anti-horseradish peroxidase (anti-HRP) (11); this antibody also recognizes neuronal tissue in C. elegans (12,13). Because of the known specificity of anti-HRP, this cross-reaction could be inferred to be due to the presence of core ␣1,3-fucose; however, the actual molecular origin of the cross-reaction observed remained unresolved, particularly due to a lack of knowledge about the exact glycosylation capacity of C. elegans.
Because the glycosylation machinery of any organism is genetically determined, examination of genomes is useful in modern glycobiology. Indeed, a large number of glycosyltransferase homologues can be identified in the genome of C. elegans. However, in only a few cases has their activity been determined. In particular, N-acetylgalactosaminyltransferases with roles in O-glycan (14) and, potentially, N-glycan and glycolipid biosynthesis (specifically, the ␤1,4-N-acetylgalactosaminyltransferase BRE-4) (15) have been expressed in recombinant form. An ␣1,2-fucosyltransferase with an unknown biological function (16), one N-acetylglucosaminyltransferase V homologue (17), one glucosyltransferase that potentially modifies O-glycans (18), and three genes encoding proteins with Nacetylglucosaminyltransferase I activity have also been described (19). The presence of the latter three is intriguing because plants, insects, and mammals would appear to only have one such gene. In mammals, N-acetylglucosaminyltransferase I is necessary for the generation of complex N-glycans and is requisite for survival in utero (20,21). Furthermore, the prior action of this enzyme is necessary in plants, insects, and mammals for both core ␣1,3and core ␣1,6-fucosylation (22)(23)(24). However, the biosynthetic pathways required for the gen-eration of the more unusual fucose modifications of C. elegans N-glycans have not been studied, although the initial assumption would be that core fucosylation would occur in a similar way as in other eukaryotes. Although there is putatively only one core ␣1,6-fucosyltransferase, there are five ␣1,3-fucosyltransferases (25) which in theory could be responsible for the core ␣1,3-fucosylation and the anti-HRP binding observed.
Following our previous work on the anti-HRP epitope of Drosophila (26), we have now re-examined the binding of anti-HRP to C. elegans proteins. By using glycan analysis, Western blotting, and fucosyltransferase assays of extracts of both wild-type and mutant nematodes, we conclude that the cross-reaction is actually due to core ␣1,3-fucose. Furthermore, we have shown that one of the ␣1,3-fucosyltransferase homologues is indeed a core ␣1,3-fucosyltransferase responsible in vitro and in vivo for generation of the anti-horseradish peroxidase epitope.

EXPERIMENTAL PROCEDURES
Preparation of C. elegans Extracts-C. elegans, whether wild-type or mutant, were routinely maintained at 16°C on a lawn of Escherichia. coli OP50 on NGM agar plates containing 250 units/ml nystatin. Each of the mutant strains VC182, VC212, VC378, RB511, and RB706 has a deletion in a different gene encoding an ␣1,3-fucosyltransferase homologue (Table I); deletions were confirmed by reverse transcription-PCR  and genomic PCR. For larger scale preparations, wild-type N2 or mutant C. elegans were grown for 4 days in liquid culture at room temperature with E. coli OP50 and separated from bacteria and debris by 30% (w/v) sucrose gradient centrifugation. Pellets were stored at Ϫ80°C before use. C. elegans extracts were prepared from a total of 1 g of nematodes (wet weight) homogenized in 3 ml of 50 mM MES, pH 7, containing protease inhibitor cocktail (EDTA-free, Roche Applied Science) and 1% (w/v) Triton X-100. After incubating for 30 min on ice, the debris was removed by centrifugation at 10,000 ϫ g for 10 min at 4°C. The supernatant was aliquoted and stored at Ϫ20°C before use. These extracts were used for both enzyme assays and Western blotting.
Preparation of Transgenic Worms-Cosmids encoding fucosyltransferase homologues within their normal genetic environment were obtained from The Sanger Centre, Cambridge, UK. For constitutive expression of the fut-1 gene, the entire open reading frame was obtained by reverse transcription-PCR using the primers K08F8.3/1/KpnI, GG-GGTACCATGACTGCAAGAAGCATCAA, and K08F8.3/2/NheI, CTAG-CTAGCTAATCTAACGGAATAGAATC. Both fragment and plasmid pPD118.25 (L3786), which contains an apparent internal enhancer from let-858, the sequences upstream of the let-858 translation start site, a multiple cloning site, a coding sequence for gfp, and the let-858 3Ј-region 2 were digested with KpnI and NheI and ligated. Recombinant clones in which the gfp sequence was replaced by the fut-1 open reading frame were selected; one of these clones was designated pPD118.25/ fut-1. VC378 mutants were injected with mixtures of pRF4 (carrying a dominant rol-6 allele) with either cosmid DNA or pPD118.25/fut-1 at a ratio of 16:1 (27), and resultant rolling lines were selected. Worms injected with pRF4 alone were used as controls.
For the preparation of genomic DNA from transformed lines, five singly picked adult rollers in 2 l of water were incubated for 1 h at 60°C with 20 l of lysis buffer (50 mM KCl, 10 mM Tris-Cl, pH 8.3, 2.5 mM MgCl 2 , 0.45% Nonidet P-40, 0.45% Tween 20, and 60 g/ml proteinase K); the extract was then heated for 15 min at 95°C. From this mixture, 0.2 l were used for PCR using MasterMix (Promega) and the primers K08F8.3/1/KpnI and K08F8.3/2/NheI (35 cycles with an annealing temperature of 56°C). For protein extracts, 50 singly picked adult rollers were lysed in 1ϫ SDS-PAGE buffer prior to electrophoresis and Western blotting. Genomic DNA and protein extracts were also prepared from singly picked controls.
Western Blotting-Proteins were separated by SDS-PAGE on 12.5% gels and transferred to nitrocellulose using a semi-dry blotting apparatus. After blocking with 0.5% (w/v) BSA, membranes were incubated with either rabbit anti-horseradish peroxidase, rabbit anti-bee venom, or rat monoclonal YZ1/2.23 at the dilutions described in the relevant figure legends. After washing, alkaline phosphatase conjugates of either goat anti-rabbit or anti-rat (1:2000) were used with subsequent color detection with SigmaFAST TM 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium. In some cases, and where indicated in the figure legends, the primary antibodies were preincubated for 1 h with the BSA neoglycoconjugates BSA-MUXF 3 and BSA-MUX carrying, respectively, native and trifluoroacetic acid-defucosylated bromelain glycopeptides, as described previously (11). The degree of defucosylation resulting in BSA-MUX was estimated as being at least 95% as judged by monosaccharide analysis.
Glycan Analyses-N-Glycans were prepared from mixed populations of either wild-type and mutant C. elegans as described previously (1). Glycans were then analyzed by both average and monoisotopic MALDI-TOF MS using either the ThermoBioanalysis Dynamo with dihydroxybenzoic acid (20 mg/ml) as matrix or the Bruker Ultraflex TOF-TOF with 6-aza-2-thiothymine (5 mg/ml) as matrix, respectively. Internal calibration was performed using theoretical values for Hex 3 HexNAc 2 and Hex 9 HexNAc 2 .
The partial reading frame of fut-1 suitable for soluble expression in Pichia using pPICZ␣ B, either in the form purchased from the supplier (Invitrogen) or in a form in which the vector was modified by inverse PCR to include a region encoding a FLAG tag just upstream of the multiple cloning site but downstream of the region encoding the Ste13 signal cleavage site, was generated by PCR using Expand polymerase a The names fut-1 and fut-4 are already listed in Wormbase; fut-2 is allocated to CE2FT-1 (EGAP9.3), an ␣1,2-fucosyltransferase; fut-8 is allocated to the core ␣1,6-fucosyltransferase. The names fut-3, fut-5, and fut-6 are proposed by us and have been approved by Wormbase. fut-3 and fut-4 are also currently listed in Wormbase as tag-8 and tag-20, respectively (i.e. temporarily assigned genes).
b The "CEFT" style names follow the nomenclature of Cummings and co-workers as listed in Refs. 42 and 50; the "FucT" style names follow the nomenclature of Oriol et al. (25).
c The listed GenBank TM /EMBL accession numbers are those of the cDNAs sequenced during the present study. d Length of the predicted protein.
with C. elegans cDNA and the primers K08F8.3/3/PstI, AACTGCAGC-AAAATCTGAACAAAAGGATTG, and K08F8.3/2/XbaI, GCTCTAGAC-TAATCTAACGGAATAGAATC. The PCR product was purified from the PCR mix using the GFX DNA purification kit (Amersham Biosciences). Both fragment and vector were digested with PstI and XbaI prior to ligation, transformation, selection, and sequencing. The expression vector was transformed into Pichia pastoris GS115, and colonies were selected and expression performed with methanol induction at 16°C as described previously (28). Concentrated supernatants were assayed either directly or after Affi-Gel Blue-Sepharose chromatography as described previously for Drosophila ␣1,4-N-acetylgalactosaminyltransferase (28). The N-terminally FLAG-tagged form was detectable after Western blotting by using anti-FLAG M2 monoclonal antibody (1:10,000) and alkaline phosphatase-conjugated anti-mouse IgG (1:2,000).

Assays of Fucosyltransferases by HPLC and MALDI-TOF MS-Dab
syl-and dansyl-glycopeptides were prepared as described previously (26,29). In the case of the dabsyl-glycopeptide from fibrin, the primary glycopeptide product after desialylation was dabsyl-GalGal (see Scheme I for a definition of the N-glycan nomenclature used), whereas after desialylation and degalactosylation the primary dansyl-glycopeptide product from IgG was dansyl-GnGnF 6 ; ␤-galactosidase or ␣-fucosidase digestion provided the GnGn forms. Incubation with jack bean ␤-Nacetylhexosaminidase was performed to generate dabsyl-MM, dansyl-MM, or dansyl-MMF 6 from the relevant GnGn or GnGnF 6 form, whereas a recombinant C. elegans hexosaminidase, 3 shown to cleave GlcNAc only from the ␣1,3-arm of GnGn, was used to specifically generate dabsyl-GnM. A dansylated glycopeptide carrying Man 5 GlcNAc 2 was prepared after Pronase digestion of Aspergillus oryzae amylase.
Dabsyl-or dansyl-glycopeptide (0.25 nmol) was typically incubated at room temperature (23°C) overnight in the presence or absence of GDP-Fuc (5 nmol) in a final concentration of 40 mM MES, pH 6.5, and 10 mM MgCl 2 or MnCl 2 with 2 l of enzyme solution (final volume 5 l). In the case of C. elegans extracts, swainsonine was added to a final concentration of 12 mM. The incubation mixture was then subject to either MALDI-TOF MS (dabsyl) or RP-HPLC (dansyl) as described previously (26).
N-Glycans from the VC378 strain (one-tenth of the glycans prepared from 3 g of worms) were incubated with FUT-1 and GDP-Fuc for 48 h at room temperature. The glycans were then purified by use of a minicolumn (Lichroprep RP18, AG3, AG50) as described previously (30).
After overnight incubation, the assay mixtures were analyzed by RP-HPLC with tomato Lewis-type fucosyltransferase used as a positive control (31).
Analysis of the Enzyme Product-Larger scale incubations of dansyl-MM and dansyl-MMF 6 (2 nmol) with FUT-1 were subject to RP-HPLC. The purified products (putatively dansyl-MMF 3 and dansyl-MMF 3 F 6 ) were analyzed by MALDI-TOF MS and by methylation analysis. For the latter, the glycopeptides were permethylated, hydrolyzed, reduced, and peracetylated prior to analysis of the obtained partially methylated alditol acetates by capillary gas-liquid chromatography/mass spectrometry by using the instrumentation and microtechniques described elsewhere (32,33).
Expression in Insect Cells-Drosophila S2 cells grown at 27°C were transiently transfected using TransFectin Lipid Reagent (Bio-Rad), according to the manufacturer's protocol for adherent cells, with either empty pIZT/V5-His or pIZT/V5-His constructs containing complete open reading frames of either one of the five C. elegans fucosyltransferase homologues or Arabidopsis thaliana FucTA. Transfected cells were lysed 3 days post-transfection, and equal amounts of extracted proteins were resolved by 12% SDS-PAGE, transferred to a nitrocellulose membrane, and probed with anti-HRP antibodies as described above. Successful transfection was confirmed by the fluorescence of the gfp-zeo R gene product encoded by the vector.
Incubation of Fucosyltransferase with a Neoglycoconjugate-A neoglycoconjugate, BSA-MM, consisting of a glycopeptide from bovine fibrin modified to remove sialic acid, galactose, and N-acetylglucosamine residues cross-linked to bovine serum albumin using dinitrodifluorobenzene (11), was incubated with FUT-1 overnight in the absence and presence of GDP-Fuc prior to SDS-PAGE and Western blotting as described above. In parallel, the band corresponding to BSA-MM was excised from a Coomassie-stained gel and subjected to trypsin and peptide:N-glycosidase A digestion as described (30).

RESULTS
Antibody Binding Studies-In initial studies we screened antibodies that recognize core ␣1,3-fucose with C. elegans extracts. Whereas anti-HRP is a polyclonal antibody shown previously to bind both core ␣1,3-fucose and ␤1,2-xylose (11, 34 -36), anti-bee venom antisera contain a proportion of antibodies recognizing core ␣1,3-fucose on insect and plant glycoproteins (37), and the monoclonal antibody YZ1/2.23 has been shown to bind core ␣1,3-fucose (11,34,38). In the present study, previous Western blotting results (39) were confirmed that showed that anti-horseradish peroxidase binds wild-type C. elegans glycoproteins; furthermore, anti-bee venom and YZ1/2.23 also showed cross-reactivity (Fig. 1A). The reactivities of both anti-horseradish peroxidase and anti-bee venom were greatly reduced in the presence of 0.5 M of a neoglycoconjugate consisting of bromelain glycopeptides attached to bovine serum albumin (BSA-MUXF 3 ) (Fig. 1A). Even 10 times more of the corresponding defucosylated conjugate (BSA-MUX) did not generate the same inhibitory effect; this is in contrast to the results with Schistosoma mansoni egg extracts with which the defucosylated conjugate also showed significant inhibitory effect (data not shown). On this basis we concluded that the cross-reaction of anti-horseradish peroxidase to C. elegans is due to core ␣1,3-fucose, whereas in the case of S. mansoni both xylose and fucose would appear to play a role. Indeed, to date, no xylose has been detected on the N-glycans of any nematode, whereas this monosaccharide is a known component of trematode N-glycans (40).
In order to determine the genetic basis of this cross-reaction, we examined homozygous mutants that have deletions in genes encoding proteins with homology to ␣1,3-fucosyltransferases. It was obvious that only the VC378 strain, which carries a deletion in the fut-1 (K08F8.3) gene, did not display anti-horseradish peroxidase staining in Western blots (Fig. 1B). Most interestingly, a strain in which all three N-acetylglucosaminyltransferase I genes are mutated still displayed anti-horseradish peroxidase staining (Fig. 1B, lane 7). Indeed, extracts of the gly-14; gly-12 gly-13 worm show no N-acetylglucosaminyltransferase I activity, and MS analysis of its N-glycans generates data compatible with the absence of its activity (41). The blotting data with this mutant was, therefore, a first indication that the core ␣1,3-fucosyltransferase of the nematode does not require the prior action of N-acetylglucosaminyltransferase I. N-Glycan Analyses-As a follow up to these data, we examined the N-glycans of wild-type and mutant nematodes and found that the VC378 strain lacks a subset of N-glycans (compare Fig. 2, A and B; see also Table II There is also, qualitatively at least, a large increase in the intensity of the Fuc 1-2 Hex 5-6 HexNAc 2 peaks (m/z 1403, 1549, and 1711) and Fuc 1-2 Hex 4 -5 HexNAc 2 Me 1 (m/z 1255, 1401, 1417, and 1563) and a decrease in the intensity of Fuc 2 Hex 3 HexNAc 2 (m/z 1225) as compared with neighboring ones. Some of these differences are only revealed with the monoisotopic spectra.
Of the other four fucosyltransferase mutants examined, only RB706 (with a deletion in the fut-6 gene) appeared to lack some species with the same m/z as those absent in VC378 (Table II), specifically those of m/z 1679, 1693, 1695, 1841, 1855, 1857, and 2017. On the other hand, no obvious non-wild-type species were present in RB706. The other three fucosyltransferase mutants gave spectra considered more or less equivalent to wild type (Table II). Our own data (not shown) with the triple knock-out gly-14;gly-12 gly-13 showed the presence of fucosyl- ated N-glycans of the form Fuc 1-2 Hex 4 -7 HexNAc 2 Me 0 -1 in this strain even in the absence of N-acetylglucosaminyltransferase I activity (41). Based on the anti-HRP binding evidence presented in Fig. 1B, we assume some of the fucosylated glycans in the latter mutant are core ␣1,3-fucosylated.
Assays of Fucosyltransferases in Nematode Extracts-The data from Western blotting and N-glycan analysis suggested that the VC378 strain lacks core ␣1,3-fucosylation and that the relevant enzyme does not require the prior action of N-acetylglucosaminyltransferase I. On the other hand, enzymatic assays with wild-type extracts showed that there was one type of fucosyltransferase activity, probably core ␣1,6, that used a biantennary glycan with two terminal N-acetylglucosamine residues (GnGn), and no difucosylation, or addition of solely a core ␣1,3-fucose, to this glycan was observed in vitro. However, preliminary data indicated the presence of a fucosyltransferase activity capable of modifying a trimannosyl N-glycan core (MM). 4 We followed up this latter result by using MALDI-TOF MS and RP-HPLC assays of wild-type and mutant C. elegans. Wild-type extracts displayed activity toward dabsyl-MM as shown by an increase, in the presence of GDP-Fuc, in the m/z corresponding to the mass of one fucose. In contrast, VC378 extract did not display this activity (Fig. 3), although control experiments with dabsyl-GnGn demonstrated that this extract did contain a putative core ␣1,6-fucosyltransferase activity. In order to gain evidence that the fucose transferred by C. elegans extracts to MM is indeed core ␣1,3-linked, the assays were also performed using dansyl-MM and dansyl-MMF 6 (derived from human IgG). As shown previously with Drosophila and mung bean core ␣1,3-fucosyltransferases with dansyl-GnGn and dansyl-GnGnF 6 , core ␣1,3-fucosylation results in a reduction in retention time of dansylated IgG glycopeptides upon RP-HPLC (23,26,29), whereas ␣1,6-fucosylation results in an increase in retention time. The former effect was apparent for dansylated MM and MMF 6 upon incubation with C. elegans N2 extract in the presence of GDP-Fuc, whereas no shift was apparent with VC378 extract (Fig. 4), indicating a lack of core ␣1,3-fucosyltransferase activity in the mutant. This result is consistent with the absence of anti-horseradish staining in this strain.
Cloning of ␣1,3-Fucosyltransferase Homologues-The five ␣1,3-fucosyltransferase homologue cDNAs were cloned to generate the full-length reading frames. 5Ј-and 3Ј-primers for isolation of the fut-1, fut-3, and fut-6 cDNAs were designed on the basis of EST information. In the case of fut-1 and fut-6, EST entries with SL1 spliced leaders are present in the Gen-Bank TM /EMBL data base, whereas one fut-3 EST has an SL2 spliced leader. For fut-4 and fut-5, the exact location of the 5Ј-end was unknown from previous cDNA sequences. Indeed for fut-5, there are no corresponding ESTs, and the initially predicted T05A7.5 reading frame included both the fut-5 and fut-6 sequences (the two genes being in fact separated by ϳ10 kb), whereas the initially predicted K12H6.3 (fut-4) reading frame encoded no transmembrane domain. Thus, homology searching and visual inspection of the translated genomic sequences to search for conserved motifs and potential transmembrane domains was performed prior to primer design for both fut-4 and fut-5. The lengths of the predicted proteins range from 378 to 433 residues. For fut-3, fut-5, and fut-6, the respective predicted protein sequence lengths are the same as given for CEFT-3, CEFT-2A, and CEFT-2B as indicated in Ref. 42. As judged from the EST data bases, FUT-1 is expressed in mixed populations, in L1 and L2 larvae, and in hermaphrodite embryos. The genes encoding all five ␣1,3-fucosyltransferase homologues are present on chromosome II.
The predicted FUT-1 protein sequence is the longest of the five ␣1,3-fucosyltransferase homologues, the major difference compared with the others being in the putative stem region (Fig. 5). FUT-1 shares 33-35% identity with the other four ␣1,3-fucosyltransferases over 260 -310 residues, and it shares 78% identity over the whole sequence with the Caenorhabditis briggsiae predicted protein CBP14511. Ignoring CBP14511, the closest hits on performing a BLAST search are Drosophila melanogaster FucTC (identities 101/291), S. mansoni FucTA (102/302), Glossinia mortisans FucTC (99/296), Xenopus laevis Lewis 2 ␣1,3/4-fucosyltransferase homologue (110/337), and D. melanogaster FucTA (103/341), all of these being more highly ranked matches than any of the other C. elegans fucosyltransferases; however, of these homologues, only the D. melanogaster core ␣1,3-fucosyltransferase FucTA has proven enzymatic activity. In the case of C. elegans FUT-3, the potential N-glycosylation site residue 152, conserved in all five C. elegans and in many animal ␣1,3-fucosyltransferases (e.g. Asn-194 of FUT-1), has been shown as part of a survey of glycoproteins to be actually glycosylated (43). On the other hand, the DXD or SXD motif present in many ␣1,3/4-fucosyltransferases, claimed in many glycosyltransferases to have a role in metal co-factor or donor substrate binding, is changed to DTP (residues 202-204) in the case of FUT-1.
FUT-1 Generates the Anti-HRP Epitope in Vivo-The constructs encoding full-length forms of all five ␣1,3-fucosyltransferases were transfected into Schneider cells. As Schneider cells do not express the anti-horseradish peroxidase epitope, they represent a suitable null background for expression of core ␣1,3-fucosyltransferases. As controls, constructs carrying either a known core ␣1,3-fucosyltransferase, from A. thaliana (44), or no insert were used. Of the five C. elegans ␣1,3-fucosyltransferases, only cells transfected with fut-1 gained antihorseradish peroxidase reactivity (Fig. 6A), a result that is in accordance with the Western blotting and glycan and enzymatic data from the corresponding mutant VC378 strain. The gain of the epitope was indeed even more pronounced with the C. elegans than with the A. thaliana enzyme.
In order to show that fut-1 was capable of "rescuing" the anti-HRP defect of the VC378 strain, complementation was performed by co-injection into VC378 mutant worms of either the cosmid K08F8 or the pPD118.25/fut-1 plasmid with the vector pRF4 carrying the rol-6 marker. In the case of the cosmid, seven rolling lines were isolated, and two of them displayed anti-HRP staining (Fig. 6B, lanes 3 and 4). Only the "rescued" lines gave rise to PCR products corresponding in size to both the wild-type and mutant forms of the fut-1 gene; the nonrescued lines contained only the mutant-sized gene. With the constitutive pPD118.25/fut-1 vector, three rolling lines were isolated, which all displayed anti-HRP binding (Fig. 6B,  lanes 6 -8). Injection of pRF4 alone did not rescue staining (Fig.  6B, lane 5). All complemented lines have a similar anti-HRP staining pattern as wild-type worms, with constitutive expression resulting in a higher intensity of staining as compared with wild-type and cosmid-rescued lines. As both experiments upon transfection of insect cells and C. elegans were successful, we concluded that FUT-1 is the core ␣1,3-fucosyltransferase necessary for the acquisition of the anti-HRP epitope in wildtype C. elegans. For the subsequent experiments it was therefore decided to focus on the properties of FUT-1.
Assays of Recombinant FUT-1-Constructs encoding soluble forms (i.e. lacking the cytoplasmic and transmembrane regions) of C. elegans FUT-1, with or without an N-terminal FLAG tag, were transformed into P. pastoris. The culture supernatants were assayed for fucosyltransferase activity toward dabsylated MM, GnGn, GalGal, and ␤GN␤GN. Supernatants from yeast transformed with both forms of the fut-1 construct only displayed activity toward dabsyl-MM (Fig. 7A), an activity absent from control transformants. Furthermore, FUT-1 was also tested with glycans from the corresponding mutant, VC378; a reduction in the Hex 3 HexNAc 2 (m/z 934) peak and an increase in the Fuc 2 Hex 3 HexNAc 2 (m/z 1226) species was observed, compatible with the conversion of putative MM and MMF 6 to MMF 3 and MMF 3 F 6 respectively, assuming that the species with m/z 1080 consists of MMF 3 after incubation with FUT-1, rather than MMF 6 , as it would before such treatment (compare Fig. 7, C and D). Tests were also performed using pyridylaminated forms of lacto-N-tetraose and lacto-N-neotetraose at the same concentration as used with the glycopeptide substrates (i.e. 50 M) and at 400 M; however, no shifts in retention time indicative of conversion to fucosylated forms of these tetrasaccharides were observed at either concentration. Furthermore, whereas a dansylated peptide carrying Man 5 GlcNAc 2 was not a substrate for recombinant FUT-1, a dabsylated peptide carrying GnM was an acceptor (data not shown). The latter result would suggest that some substitution of the ␣1,6-arm may be tolerated by FUT-1, although certainly the ␣1,3-linked mannose must be free.
The activity of recombinant FUT-1 was also confirmed in an HPLC-based assay. As in the case of wild-type C. elegans extracts, a shift to lower retention time was observed when either dansyl-MM or dansyl-MMF 6 was used as a substrate, with a higher conversion being apparent with the latter substrate (Fig. 8). The HPLC-purified products were confirmed by MALDI-TOF MS to have m/z values compatible with fucosylation; furthermore, methylation analysis of the enzymatic products demonstrated that the two fucose residues of the putative MMF 3 F 6 were indeed on the same GlcNAc residue, because a fully methylated GlcNAc, indicative of 3,4,6-trisubstitution, was detected by gas-liquid chromatography/mass spectrometry. On the other hand, the putative MMF 3 product contained a GlcNAc with a nonsubstituted 6-hydroxyl, a feature absent from the MMF 3 F 6 sample (data not shown).
The recombinant FUT-1 could be eluted with 0.6 M NaCl from Affi-Gel-Sepharose and subsequent SDS-PAGE indicated the presence of a protein of M r 55,000 (as compared with a theoretical value of 46,000 for the soluble form of FUT-1). Tryptic digestion of the nontagged form indicated the presence of 10 peptides compatible with the theoretical map of FUT-1, whereas Western blotting of the FLAG-tagged form showed strong reactivity also around M r 55,000 (data not shown). Subsequent characterization of the activity of FUT-1 was per- formed on these preparations of the enzyme. The optimal pH was found to be between pH 6.5 and 8, whereas at least 2-fold higher activity was found in the presence of Mg(II) ions as compared with Co(II), Mn(II), Ca(II), Fe(II) or EDTA; no activity was detected in the presence of either Cu(II), Ni(II), Zn(II), or Fe(III) ions.
Finally, in order to show that FUT-1 indeed generates the anti-HRP epitope in vitro, a neoglycoconjugate of bovine serum albumin carrying an MM-glycopeptide was incubated with purified FUT-1 in the absence and presence of GDP-Fuc. As predicted from the other substrate specificity experiments, the thereby modified BSA-MM conjugate was recognized by anti-HRP (Fig. 9A, lane 2). MALDI-TOF MS analysis of the glycans of the conjugate incubated in the presence of GDP-Fuc showed the presence of a peak with a composition compatible with an MMF 3 structure (Fig. 9C), whereas this peak was absent from the control (Fig. 9B). DISCUSSION The fucosylation of C. elegans glycoconjugates is potentially highly complex due to the presence of around 20 ␣1,2-fucosyltransferases, five ␣1,3-fucosyltransferases, and one ␣1,6-fucosyltransferase homologue. To date the exact structures of only some of the fucosylated N-and O-glycans are known (2,9), whereas no fucosylated glycolipids have as yet been detected. There is also only fragmented data on the fucosylation of other nematodes: up to three fucose residues on the core of N-glycans (45,46) and the presence of a Lewis-type fucosyltransferase (47) in Haemonchus, fucosylation of artho-series glycolipids from Ascaris suum (48), and the presence of Lewis epitopes in Dictyocaulus viviparus (49). In C. elegans, the single obvious ␣1,6-fucosyltransferase homologue (FUT-8) has been expressed in an active recombinant form, 4 and the determination of the enzymatic activity of only one ␣1,2-fucosyltransferase (FUT-2 or CE2FT1) has been published (16) but left its biological substrate unrevealed. However, the activity of most of the ␣1,2-fucosyltransferase homologues is still unknown, whereas published and preliminary data suggested that the ␣1,3-fucosyltransferase homologues were Lewis-type enzymes (42,50,51). Thus even if Lewis-type glycans and fucosylated glycolipids have not been found in C. elegans, the presence of enzymes capable of generating such structures in vitro suggests there is the potential that Lewis or related epitopes occur in small amounts in the worm, depending on the presence of suitable in vivo precursors, even though there are data indicating no binding of Lewis a or Lewis x antibodies in Western blots (52). The presence of core ␣1,3-fucose, on the other hand, was not accounted for at all at the genetic level.
Considering that we had determined previously that core ␣1,3-fucosylation is the probable basis for anti-horseradish peroxidase binding in D. melanogaster (26) and that xylose had not been found to be a constituent of C. elegans N-glycans, we hypothesized that one or more core ␣1,3-fucosyltransferases were responsible for generation of the epitope in the nematode. Initial results with inhibitors suggested that core ␣1,3-fucose was responsible for the cross-reaction, but we failed to detect core ␣1,3-fucosylation by C. elegans extracts in vitro when using dabsyl-or dansyl-GnGn as substrates. These substrates are utilized by core ␣1,3-fucosyltransferases from plants and insects (26,44), as well as by the core ␣1,6-fucosyltransferase of C. elegans. 4 It was only when we wished to rule out that the native core ␣1,6-fucosyltransferase could utilize MM that we detected a core fucosyltransferase in wild-type C. elegans extracts capable of transferring to this structure. As shown in Fig. 4A, dansyl-MMF generated upon incubation with wildtype C. elegans extracts had the elution characteristics of a core ␣1,3and not of a core ␣1,6-fucosylated structure.
In parallel to the inhibition Western blots and enzyme assays with wild-type worm extracts, we surveyed the anti-horseradish peroxidase binding characteristics and N-glycan spectra of the five publicly available ␣1,3-fucosyltransferase knockouts. Of these, only VC378 lacked the epitope (Fig. 1B). This was a surprising result, considering that VC378 has a 597-bp deletion in the fut-1 gene, 5 resulting in the absence of residues 46 -190 of the predicted protein sequence. Previously, mammalian cells transfected with fut-1 cDNA were found to express a Lewis-type enzyme activity (50).
However, encouraged by the results with the VC378 strain, we initiated our own studies on the expression of fut-1; indeed, compatible with the various results with the mutant, transfection of fut-1 into yeast, C. elegans, or insect cells resulted in expression of an enzyme with the ability to generate the antihorseradish peroxidase epitope (Figs. 6 and 9). In studies with the yeast-produced enzyme only MM-fucosylating activity was detected (Figs. 7 and 8). Also, blotting analysis of the relevant mutants and transfection into insect cells indicated that the other four ␣1,3-fucosyltransferases are not involved in expression of the anti-HRP epitope (Figs. 1B and 6A). Thus, there are a number of key results (lack of epitope, enzyme activity, and certain glycans in the VC378 strain, complementation of the molecular phenotype with either the cosmid or the open reading frame, generation of the epitope in insect cells, and, in vitro, demonstration of fucosyltransferase activity in transformed yeast and linkage analysis of the enzymatic product) that demonstrate that FUT-1 is the core ␣1,3-fucosyltransferase that synthesizes the anti-HRP epitope in C. elegans.
The different conclusions from our work and the earlier study on the enzymology of FUT-1 have a number of potential origins. First, the previous study did not apparently involve testing of N-glycan substrates. Second, a relatively high lacto-N-neo-tetraose concentration (5 mM) was used in the earlier study (50), which gave rise to less than 0.003% conversion of acceptor to product/h. Third, a different expression system could theoretically affect the substrate specificity. In our experiments with dabsyl-and dansyl-MM, we used 50 M acceptor substrate and attained 5-10% conversion/h. We also examined whether FUT-1 displayed activity to dabsyl-GalGal and lacto-N-neo-tetraose at 50 M (the latter also at 400 M), but we observed no transfer. Thus, it may be that the Lewis-type activity is indeed a property of FUT-1, but only at very high substrate concentrations. Furthermore, all three expression systems we used (C. elegans, insect, and yeast) showed production of the anti-HRP epitope upon transfection into a null background, and thus the core ␣1,3-fucosylating property is reproduced independent of the expression system used.
The newly defined properties of FUT-1, particularly its preference for MM as opposed to GnGn, indicate that the biosynthesis of core difucosylated N-glycans differs between the nematode and insects. This may mean that during evolution either the core ␣1,3-fucosyltransferase of the nematode has changed its properties or that a functional convergence has occurred that the same glycan product can be generated by a different series of biosynthetic steps. Indeed, it would appear that most, if not all, glycans that carry core ␣1,3-fucose in D. melanogaster are also core ␣1,6-fucosylated (26), and this may be true also in C. elegans (2). However, there are more solely core ␣1,6-fucosylated N-glycans in both species. This would suggest that the core ␣1,6-fucosyltransferase acts first, after N-acetylglucosaminyltransferase I. In insects, however, the core ␣1,3-fucosyltransferase still requires the presence of nonreducing terminal N-acetylglucosamine residues, and we assume that the action of core ␣1,6-fucosyltransferase is followed by that of core ␣1,3fucosyltransferase (FucTA, in the case of D. melanogaster), and then a Golgi ␤-hexosaminidase, such as that present in Sf9 cells (53), can act to generate MMF 3 F 6 . In C. elegans, however, the core ␣1,3-fucosyltransferase does not act when a nonreducing terminal GlcNAc substitution is present on the ␣1,3-arm, and so one may hypothesize that in wild-type C. elegans, the action of the N-acetylglucosaminyltransferase I isoforms is followed by those of core ␣1,6-fucosyltransferase, ␣-mannosidase II, and by a Golgi ␤-hexosaminidase (the latter having already been described in C. elegans (54)). The thereby generated 5 Details of the deletion are given under the on-line address: aceserver.biotech.ubc.ca/cgi-bin/generic/allele?classϭAllele;nameϭgk183. MMF 6 -carrying glycoproteins would then be hypothesized to act as substrates for FUT-1, which synthesizes MMF 3 F 6 . It is probable, however, that the in vivo situation is more complicated. First, there are low amounts of glycans with the same composition as MMF 3 F 6 (i.e. Fuc 2 Hex 3 HexNAc 2 ) still present in VC378. Second, there is a series of tri-and tetrafucosylated N-glycans present in the wild-type absent from the VC378 mutant as well as mono-and difucosylated species in the gly-14;gly-12 gly-13 triple knock-out.
With respect to the first point, a number of isomers of Fuc 2 Hex 3 HexNAc 2 are conceivable. In addition to MMF 3 F 6 , which is probably present in the wild-type as judged from previous chromatographic evidence (2), the second residue of the core chitobiose could also be modified by ␣1,3-fucose, such as observed in Haemonchus contortus (45,46). Considering the findings with the RB706 mutant, which also lacks some of the more highly fucosylated structures but not the anti-HRP epitope, C. elegans FUT-6 is certainly an interesting candidate for future studies on nematode N-glycosylation pathways.
The second point, based on the presence of tri-and tetrafucosylated structures, raises the question whether MM is the only substrate of FUT-1 in vivo. On the other hand, in vitro FUT-1 only obviously fucosylates glycans that are probably MM and MMF 6 among a mixture of VC378 N-glycans (Fig. 7C). Thus, it may be that some larger glycans cannot function as substrates for FUT-1 because of subsequent glycosylation events resulting in "NOGO" signals; this would explain why adding FUT-1 to VC378 glycans does not generate the full series of wild-type glycans. Further complicating the picture is the finding that worms lacking GlcNAc-TI possess fucosylated glycans (Fuc 1-2 Hex 5-7 GlcNAc 2 ) not found in wild-type worms, while lacking MM, as judged by the absence of Hex 3 GlcNAc 2 (41), proteins from these worms bind anti-HRP and so therefore can be assumed to contain N-glycans resulting from the action of FUT-1. Thus, one can surmise that further substrates, yet to be defined (other than MM, which is more or less absent from the gly-14;gly-12 gly-13 triple knock-out), may be used by FUT-1 in vivo and/or that unknown hexosyl-and fucosyltransferases can act before and after FUT-1. Further characterization of potential substrates, as well as of the tri-and tetrafucosylated N-glycans in Caenorhabditis, is therefore still required.
As part of our studies, we performed various N-glycan analyses and confirmed the complexity we and others have reported previously. Our data agree well with those of Zhu et al. (41) in terms of the high degree of fucosylation of wild-type glycans as well as the presence of mono-and difucosylated glycans in the gly-14; gly-12 gly-13 triple knock-out. However, the results with wildtype worms differ in that Zhu et al. (41) also detected glycans of the form Me 0 -1 Fuc 2-4 Hex 7-8 HexNAc 2 which we did not, but they did not find phosphorylcholine, which we did detect. This may be because of the fact that they released their glycans by hydrazinolysis, which may remove phosphorylcholine, while enabling the release of that proportion of glycans hypothesized to have a high degree of galactosylation on the core fucose residues (55). Some of the structures with four or more hexose residues, three or more fucose residues, and 2-O-methylation of fucose, observed by us and others, are still unique to C. elegans and have yet to be discovered in other nematodes; however, the presence of galactose attached to core fucose residues has already been observed on squid rhodopsin (56).
Although the VC378 mutant appears to be viable under normal laboratory conditions, we assume that core ␣1,3-fucosylation has some, possibly subtle, function in nematode biology. In C. elegans, expression of this epitope appears to be restricted to 10% of the neurons (12), and indeed fut-1 is ap-parently only expressed in specific neural cells (16). However, from the viewpoint of the wider importance of nematodes as parasites, more significant is that core ␣1,3-fucose is a known IgE epitope (39,57), and so the presence of this epitope on excretory-secretory antigens, for example (as shown previously by antibody binding for H. contortus (39)), may have a role in the interactions of parasitic nematodes with their hosts. C. elegans also shares carbohydrate epitopes with H. contortus that induce protective immunity (58). Other tests on protective immunity to H. contortus conferred by an antibody specific for excretory-secretory antigen suggest ␣1,3-fucosylated LacdiNAc has a role (59), although structural evidence for this epitope in C. elegans and H. contortus is lacking. On the other hand, the H. contortus H11 glycoprotein, which also confers protection, is known to be core ␣1,3-fucosylated (46). Furthermore, glycans of parasitic nematodes generate Th2 responses, a fact that also has been found to hold for C. elegans glycans (even though this is not of direct pathogenic significance, because C. elegans is not an animal parasite); and even more significant is preliminary data indicating that fucosylated C. elegans glycans are responsible for the observed Th2 response (60). Therefore, C. elegans fucosylation mutants should be interesting models, not just for biosynthetic and glycomics studies but also for future immunological investigations.