Array-assisted Characterization of a Fucosyltransferase Required for the Biosynthesis of Complex Core Modifications of Nematode N-Glycans*

Background: The chitobiose region of nematode N-glycans can be modified with three fucose residues. Results: Glycan arrays and other analytical techniques facilitated the definition of the biologically relevant activity of Caenorhabditis FUT-6. Conclusion: The concerted action of Caenorhabditis FUT-1, FUT-6, and FUT-8 is required for trifucosylation of worm N-glycan cores. Significance: New approaches for studying glycans from parasitic nematodes are now possible. Fucose is a common monosaccharide component of cell surfaces and is involved in many biological recognition events. Therefore, definition and exploitation of the specificity of the enzymes (fucosyltransferases) involved in fucosylation is a recurrent theme in modern glycosciences. Despite various studies, the specificities of many fucosyltransferases are still unknown, so new approaches are required to study these. The model nematode Caenorhabditis elegans expresses a wide range of fucosylated glycans, including N-linked oligosaccharides with unusual complex core modifications. Up to three fucose residues can be present on the standard N,N′-diacetylchitobiose unit of these N-glycans, but only the fucosyltransferases responsible for transfer of two of these (the core α1,3-fucosyltransferase FUT-1 and the core α1,6-fucosyltransferase FUT-8) were previously characterized. By use of a glycan library in both array and solution formats, we were able to reveal that FUT-6, another C. elegans α1,3-fucosyltransferase, modifies nematode glycan cores, specifically the distal N-acetylglucosamine residue; this result is in accordance with glycomic analysis of fut-6 mutant worms. This core-modifying activity of FUT-6 in vitro and in vivo is in addition to its previously determined ability to synthesize Lewis X epitopes in vitro. A larger scale synthesis of a nematode N-glycan core in vitro using all three fucosyltransferases was performed, and the nature of the glycosidic linkages was determined by NMR. FUT-6 is probably the first eukaryotic glycosyltransferase whose specificity has been redefined with the aid of glycan microarrays and so is a paradigm for the study of other unusual glycosidic linkages in model and parasitic organisms.

Glycan arrays have begun to revolutionize the way in which we study carbohydrate-protein interactions (1) and, in combination with modern glycoanalytical and chemical glycobiological approaches (2), have transformed the experimental tools available for modern structural and functional glycobiology. However, in comparison with the examination of the binding of antibodies or lectins with glycan arrays, the determination of enzyme activities, especially of glycosyltransferases, with these platforms is not so well established, and generally only previously studied enzymes have been assessed (3)(4)(5)(6)(7). One exception is a recent study on the specificity of "new" glycosyltransferases from bacteria toward simple saccharide structures (8); nevertheless, the actual in vivo substrates for these enzymes remain unknown.
To date, because the focus of glycan arrays has been on mammalian glycans (9), a huge number of possible glycan structures in nature, especially non-mammalian glycans (e.g. those of model organisms or parasites) are underrepresented on existing platforms. These organisms also have or are predicted to have a variety of glycosyltransferases that have previously been unstudied or only incompletely studied; thereby, non-mammalian glycomes and enzymes represent an untapped resource as well as an underestimated challenge. One of these model organisms with a particularly rich glycomic potential is the nematode Caenorhabditis elegans (10); this species has not only become well established as a model for studies of developmental biology and innate immunity but is also related to parasitic nematodes that represent a biological burden for millions of human beings and livestock worldwide as well as for plant crops. Because nematode glycoconjugates have immunomodulative properties (11) or are relevant in attempts to produce vaccines (12), there is a need for new approaches to study biosynthesis, binding partners, and functions of these molecules in order to identify new therapeutic targets.
Indeed, the core region of nematode N-glycans is particularly unique due to the range of so-called complex core modifications (13,14), which represent a set of targets for lectins and potential therapeutics; not just ␣1,6-fucosylation of the reducing terminal (proximal) asparagine-bound N-acetylglucosamine occurs as in mammals, but also ␣1,3-fucosylation of both GlcNAc residues (proximal and distal) of the chitobiose core ( Fig. 1). Thereby, nematodes express both the core ␣1,3-fucose on the proximal residue as also present in plants, slime molds, and other invertebrates (15) as well as a novel form of ␣1,3fucosylation of the distal N-acetylglucosamine. Furthermore, core fucose residues can be capped with hexose; for instance, substituted and unsubstituted galactose is found linked ␤1,4 to the core ␣1,6-fucose (GalFuc) 2 and is also present in planaria and cephalopods (16), whereas ␣1,2-galactosylation of the distal fucose has been recently discovered in nematodes (17). These unusual modifications occur not just in C. elegans but also in a number of parasitic worms, such as Ascaris suum, Hemonochus contortus, and Oesophagostomum dentatum (17,18), and represent epitopes particular to a subset of nematodes.
These N-glycan core motifs have been found to be targets of one endogenous C. elegans galectin (LEC-6) and two nematoxic lectins (CCL2 and CGL2) derived from fungi (19 -21). However, because nematode glycans are poorly represented by current glycan arrays, misleading results can be obtained regarding which glycans are the true binding partners. For instance, CCL2 was observed to bind mammalian glycans of types (fucosylated on the antennae) absent from most nematodes (20), whereas CGL2 bound mammalian glycans with (sub)terminal galactose (22); in contrast, the biological data indicated that the respective in vivo targets are core ␣1,3-fucosylated glycans and GalFuc epitopes actually found in nematodes (20,21).
The definition of these three enzymes (FUT-1, FUT-8, and GALT-1) still left the molecular basis for the other core modifications unsolved. Neither the fucosyltransferase modifying the distal N-acetylglucosamine nor other glycosyltransferases modifying ␣1,3-linked fucose or the galactosylated core ␣1,6fucose have, to date, been studied. There were clues from glycomic studies of mutant worms, because not only fut-1 and fut-8 mutants are deficient in certain fucosylated N-glycans (21,28), but also fut-6 mutant worms have an altered glycomic profile (28). Therefore, we suspected that FUT-6 may have a role in N-glycan biosynthesis independent of its ability to generate Lewis-type epitopes in vitro. Using a mixture of array-and glycomic-based approaches, we now show that FUT-6 is the enzyme that generates the distal Fuc␣1,3GlcNAc unit in C. elegans, a result then exploited in the chemoenzymatic synthesis of a complete trifucosylated N-glycan core. In the process, we were able to recreate the biosynthetic pathway leading in nematodes to a multiply fucosylated core recently shown to have relevance to nematoxic lectin binding.

EXPERIMENTAL PROCEDURES
Enzyme and Lectin Preparation-HisFLAG-tagged soluble forms of C. elegans FUT-1, FUT-6, and FUT-8 were expressed in Pichia pastoris. The constructs were prepared directly from RT-PCR fragments in the case of FUT-1 and FUT-6, also known as CEFT1 and CEFT3 (24), or by reamplification from a previously described expression vector in the case of FUT-8 (28,31) into a reconstructed form of pPICZ␣ vectors named pPICZ␣HisFLAG. First, the modified expression vector was obtained after two rounds of inverse PCR using KOD polymerase (Takara) to incorporate a region encoding a His tag and a FLAG tag between the region encoding the ␣-factor signal sequence and the ClaI, PstI, and EcoRI restriction sites. Truncated open reading frames for the three fucosyltransferases (excluding the cytoplasmic and transmembrane domains) were then isolated after PCR using the following forward and reverse primers for FUT-1 (AACTGCAGAAATCTGAACAAAAGGATTGG with GCTCTAGACTAATCTAACGGAATAGAATC), FUT-6 (AAC-TGCAGAGGAGTAAACATAAAGATTCC with GCTCTAG-ACAACTACAAATATTTCGAAGC) and FUT-8 (TCTGGA-AAAAGAAAGACAAGAAC with CGGGTACCTAATCTA-AAAGAGCTTCG). The PCR products were cut with the relevant restriction enzymes and then ligated into the pPICZ-␣HisFLAG vector. Linearized constructs were integrated into the Pichia genome by electroporation. All recombinant FUTs were expressed at 18°C for 96 h prior to His tag purification using nickel affinity chromatography; purified FUTs were rebuffered and stored in 25 mM Tris-HCl, 150 mM NaCl, pH 7.0, at 4°C. Expression of the enzymes was verified by SDS-PAGE as well as Western blotting with an anti-FLAG antibody (for FUT-6, a single band of ϳ45 kDa was observed upon Coomassie Blue staining; data not shown). C. elegans GALT-1 (30) and Coprinopsis cinerea lectin CCL2 (20) were the kind gifts of Dr. Markus Künzler.
Glycan Microarray-Microarrays were printed on Nexterion H N-hydroxysuccinimide-activated glass slides (Schott AG, Mainz, Germany) employing a robotic non-contact spotter, sciFLEXARRAYER S11, from Scienion AG (Berlin, Germany). Droplets (2.5 nl) of each glycan solution (50 M; glycans 1-17, Fig. 3) in sodium phosphate buffer (300 mM, pH 8.5, 0.005% Tween 20) were spatially arrayed with a spot pitch of 550 m. Each glycan was spotted in six replicates (2 different glycans/row), producing a 12 ϫ 11 spot array, which was printed in 14 copies onto each slide. After printing, the slides were placed in a 75% humidity chamber (saturated NaCl solution) at 25°C for 18 h. Unreacted N-hydroxysuccinimide groups were quenched with 50 mM ethanolamine in 50 mM sodium borate buffer, pH 9.0, for 1 h. The slides were washed with PBST (PBS solution containing 0.5% Tween 20), PBS, and water and dried in a slide spinner.
Fluorescence was measured using an Agilent G265BA microarray scanner system (Agilent Technologies, Santa Clara, CA) and quantified with ProScanArray Express software (PerkinElmer Life Sciences), employing an adaptive circle quantitation method from 50 m (minimum spot diameter) to 300 m (maximum spot diameter). Average relative fluorescent unit values with local background subtraction of six spots and S.D. were reported as histograms.
the second (Hypersil ODS, Agilent). Selected purified and characterized glycans were among the substrates tested with FUT-6.
MALDI-TOF MS-Mass spectra were recorded using either UltrafleXtreme III or Autoflex Speed MALDI-TOF mass spectrometers, respectively, equipped with a pulsed N 2 laser (337 nm) or a SmartBEAM solid state laser and controlled by Flex-Control version 3.3 software (Bruker Daltonics). MS/MS was performed by laser-induced dissociation of selected parent ions.
In-solution Assays-For natural pyridylaminated glycans or chemically synthesized compounds with the alkylamine linker, the reaction was carried on in a minimal mixture (2.5 l) of glycan candidate, MnCl 2 (20 mM), MES buffer (80 mM, pH 6.5), GDP-Fuc (2 mM), and recombinant C. elegans FUT-6, at room temperature or 37°C overnight. Aliquots of the reaction mixture were examined by MALDI-TOF MS using 6-aza-2-thiothymine as matrix (3 mg/ml), and MS/MS (laser-induced dissociation) was performed on the products to assign the structure. As required these compounds were preincubated with FUT-1, FUT-8, and/or GALT-1 at room temperature with the requisite nucleotide sugar and Mn(II) and/or with jack bean ␤-hexosaminidase at 37°C.
Pyridylaminated lacto-N-neotetraose (6 pmol) was incubated with FUT-6 under similar reaction conditions prior to FIGURE 3. Glycan array screening for FUT-6 substrates. A, N-glycan microarray after enzymatic reaction with bovine milk ␤1,4-galactosyltransferase followed by C. elegans ␣1,3-fucosyltransferase (FUT-6) to generate Lewis x type epitopes (boxed inset) on the antennae of complex and hybrid N-glycans. a, nonmodified N-glycan structures originally printed on the microarray. b, N-glycan microarray images after galactosylation incubated with the galactose-recognizing lectin from R. communis (RCA-555). c, images of the galactosylated N-glycan microarray after FUT-6 reaction and incubation with the fucose-recognizing lectin from A. aurantia (AAL-555). d, fluorescence intensities of the galactosylated array after incubation with RCA-555. e, fluorescence intensities with AAL-555 after GalT and FUT-6 incubation. B, N-glycan microarray after enzymatic reaction with CeFUT-6 without preincubation with galactosyltransferase. a, N-glycan microarray images after incubation with fucose-recognizing lectins (AAL and CCL2) and anti-HRP-555 antibody. b, fluorescence intensities after incubation with AAL-555. c, fluorescence intensities after incubation with CCL2-647. Compounds 19 and 21, already containing fucose, were included as controls; however, only the latter (LDNF) reacted with CCL2. Each histogram represents the average relative fluorescent unit values for six replicates with the S.D. Red triangles, fucose; yellow circles, galactose; blue squares, N-acetylglucosamine; green circles, mannose.
NMR-Compounds were freeze-dried and dissolved in deuterium oxide (D 2 O) for recording 1 H NMR spectra. Nuclear magnetic resonance experiments were acquired on a Bruker 500-MHz spectrometer, and chemical shifts (␦) are given in ppm relative to the residual signal of the solvent used (D 2 O 4.79 ppm).

Array-based Screening of Fucosyltransferase Specificities-
As initial glycomic evidence from single mutants (28) suggested that the C. elegans ␣1,3-fucosyltransferase FUT-6 might have an activity other than its known ability to synthesize Le x as well as fucosylated LacdiNAc (LDNF) in vitro (24), a finding that was reproduced by our own assays (Fig. 2), we considered new approaches to reveal the in vivo specificity of this enzyme. We have previously tested a new N-glycan array with two fucosyltransferases with known specificity, the core ␣1,3-fucosyltransferase FucTA from Arabidopsis thaliana, and the core ␣1,6fucosyltransferase FUT-8 from C. elegans (7). On-array fucosylation by both these enzymes was easily assessed using the fucose-specific lectin from A. aurantia (AAL). Furthermore, human asialotransferrin fucosylated on the LacNAc antennae in vitro by FUT-6 was previously shown to bind AAL (33). Therefore, we investigated the impact of His tag-purified recombinant FUT-6 on AAL binding to the array with and without prior incubation with ␤1,4-galactosyltransferase (Fig.   3) because pregalactosylation is a requirement for synthesis of Le x ; thereby, FUT-6 was incubated with galactosylated and non-galactosylated forms of the N-glycan array, which contained paucimannosidic glycan cores lacking non-reducing GlcNAc as well as mono-, bi-, tri-, and tetra-antennary N-glycans, the latter representing also galactosyltransferase substrates. The efficiency of pregalactosylation was assessed using the galactose-specific lectin from R. communis (RCA).
As expected from the previous in vitro data on FUT-6, galactosylated structures 7-17 (whose galactosylation status was proven by RCA binding) gained AAL reactivity upon incubation with FUT-6 ( Fig. 3A). Furthermore, non-galactosylated paucimannosidic structures (compounds 1, 2, and 5 on the arrays either with or without preincubation with galactosyltransferase; Fig. 3, A and B) lacking the ␣1,6-arm were surprisingly also AAL-positive. On the non-galactosylated array, additional FUT-6 substrates were those with non-reducing GlcNAc on the ␣1,3-arm but lacking the ␣1,6-mannose linked to the core ␤1,4-mannose (8 and 10; Fig. 3B). The spot corresponding to galactosylated N-glycan 18, included as a positive substrate control on an otherwise non-galactosylated array, was recognized by AAL after incubation with FUT-6 due to the expected formation of antennal Le x . The multiantennary non-galactosylated glycans and the hybrid-like structures with an ␣1,6-mannose were not modified by FUT-6. None of the products of FUT-6 were bound by anti-HRP or by the fungal CCL2 lectin, both of which can recognize core ␣1,3-fucose; only the preformed LDNF trisaccharide 21 was recognized by CCL2. Therefore, we concluded that FUT-6 not only generates Le x epitopes on the antennae of glycan substrates in vitro but transfers fucose to another position on selected N-glycans. However, it cannot form the anti-HRP epitope, which is dependent on ␣1,3-fucosylation of the proximal core GlcNAc.
Glycomic Analysis of Fucosyltransferase Mutants-Additional clues regarding the specificity of FUT-6 were expected by analyzing C. elegans mutants lacking this enzyme. The fut-6 single mutant was previously shown to lack tetrafucosylated glycans in the peptide:N-glycanase A-released pool but still possessed trifucosylated glycans (28). Therefore, various double mutants lacking two fucosyltransferases each were also prepared: fut-1;fut-6 (previously shown to be completely resistant to CCL2 (20)), fut-6;fut-8, and fut-1;fut-8. Glycomic analyses indicated that the latter two mutants had maximally two fucose residues on their N-glycans (Fig. 4), whereas the fut-1;fut-6 still had traces of trifucosylated N-glycans. This is a further suggestion of an N-glycan-modifying activity of FUT-6.
In-solution Assays with Various Substrates-Although the in vitro lectin-based assay was a first indication of the unique specificity of FUT-6 and the glycomic analyses were in accordance with a role of this enzyme in the modification of N-glycans in vivo, further verification was required in order to localize the new glycosidic linkage formed by this enzyme. Therefore, a range of substrates suitable for "in-solution" studies were examined: specifically, pyridylaminated forms of glycans isolated from C. elegans, remodeled dabsylated glycopeptides, and chemically synthesized glycans of the form (i.e. functionalized with an alkylamine spacer) also used for printing the N-glycan array.
A working hypothesis was that FUT-6 transfers fucose to the distal GlcNAc of N-glycans; this type of fucose modification was presumed missing from fut-1;fut-6 mutants but was previously shown to be overrepresented in a hex-2;hex-3 strain (17). Therefore, N-glycans from these mutants presumed to be biosynthetic intermediates were incubated with purified FUT-6. In particular, two isomers of core ␣1,6-fucosylated glycans with the compositions Hex 1 HexNAc 2 Fuc 1 and Hex 3 HexNAc 3 Fuc 1 as well as two structures of the form Hex 2-3 HexNAc 2 Fuc 1 (one of which is an ␣1,3-mannosidase digestion product of a natural glycan) were tested (Fig. 5, A-G). Only four of these seven glycans were actual acceptor substrates for the in vitro enzymatic reaction; the common element in these substrates was an absence of an ␣1,6-mannose, but the presence of the ␤1,4-mannose, on the core region. Unfortunately, MS/MS of these difucosylated products did not result in the appearance of a key fragment for the transferred fucose (data not shown). Notably, a glycan treated with ␣1,3-mannosidase but retaining the ␣1,6mannose is not accepted by FUT-6 ( Fig. 5E).
Based on these data, a dabsylated glycopeptide derived from bovine fibrin was remodeled by degalactosylation and core ␣1,6-fucosylation followed by removal of the antennal GlcNAc residues and of the ␣1,3/6-mannose residues (Fig. 5, H-L). The resultant glycopeptide carrying a Man 1 GlcNAc 2 Fuc 1 glycan was a substrate for recombinant FUT-6 (Fig. 5M); MS/MS showed a low intensity fragment of m/z 512, indicative that the transferred fucose is associated with the Man␤1,4GlcNAc region but not with the reducing terminal GlcNAc (data not shown).
For exact analysis of the position of the transferred fucose, a selection of glycans with an alkylamine linker on the reducing end (used also for printing the glycan arrays) was employed for reactions in solution. In total, four substrates (compounds 1, 2,  8, and 10), chosen on the basis of a positive result with the glycan array (Fig. 3), were tested, and all were found to be fuco- sylated by FUT-6 in vitro as shown by mass spectral data (Fig. 6). MS/MS of the simplest structure resulted in a set of fragments that were more informative than those of the pyridylaminated and dabsylated products; thereby, the localization of the transferred fucose to the distal GlcNAc rather than to the ␤-linked mannose was more certain. In particular, the simultaneous appearance of Hex 1 HexNAc 1 Fuc 1 (m/z 511 as [M ϩ H] ϩ and m/z 533 as [M ϩ Na] ϩ ) and alkylaminated HexNAc 2 Fuc 1 (m/z 655 as [M ϩ H] ϩ and m/z 677 as [M ϩ Na] ϩ ) ions as fragments of the fucosylated "monoantennary" compound 1 was the most promising piece of evidence that the fucosylation by FUT-6 took place on the distal GlcNAc (Fig. 6, J and L).
Characterization of an Enzymatic Product by NMR-In order to more definitively verify the linkage of the transferred fucose, the chemically synthesized trisaccharide 1 (Man 1 GlcNAc 2 ) was fully fucosylated by FUT-6 (Scheme 1) and the fucosylated product 23 analyzed by 1 H NMR; the chemical shifts for each proton of both compounds are listed in Table 1. The introduction of one fucose moiety was confirmed by the appearance of a new anomeric proton at ␦ 5.16 ppm with a coupling constant of J H-1 , H-2 ϭ 3.9 Hz, characteristic of an ␣-glycosidic bond. The characteristic signal of H-6 protons from the newly introduced fucose appeared as a doublet at ␦ 1.08 ppm with a coupling constant of 6.6 Hz and an integration for three protons. The complete assignment of the 1 H NMR spectra of the fucosylated product and the trisaccharide 1 was achieved by one-and twodimensional NMR experiments, using 1 H-1 H COSY, 1 H-13 C heteronuclear single quantum correlation spectroscopy, total correlation spectroscopy, and NOESY. A closer look at the protons of each sugar (Scheme 1, A-G) showed that the H-3 signal of the distal GlcNAc was shifted from ␦ 3.69 ppm (GlcNAc B) in trisaccharide 1 to ␦ 3.96 ppm (GlcNAc E) in the fucosylated product 23, indicating fucosylation at this position. Furthermore, an NOE effect was observed between the H-1 of the fucose at 5.16 ppm and the H-3 of the GlcNAc E at 3.96 ppm, confirming the ␣1,3 linkage.
Biosynthetic Pathways for Complex Core Modifications-To investigate the basis for the multifucosylated core, the chemically synthesized compound 10 was first ␣1,6-fucosylated by C. elegans FUT-8 and sequentially remodeled using various recombinant glycosyltransferases from C. elegans as well as jack bean hexosaminidase. In conclusion, two biosynthesis pathways were revealed, leading to the final formation of a trifucosylated N-glycan core: (a) hexosaminidase 3 FUT-1 3 FUT-6 and (b) FUT-6 3 hexosaminidase 3 FUT-1 (Fig. 7, A and B).
FUT-1 only worked on hexosaminidase-processed substrates lacking non-reducing terminal GlcNAc, whereas FUT-6 did not have this restriction. The results also showed that there was no specific order of ␣1,3-fucosylation on the two core GlcNAcs.
Considering that many core fucose residues are also capped with galactose, it was also of interest to examine the point at which the core ␣1,6-fucose can be galactosylated, using the only proven nematode galactosyltransferase, C. elegans GALT-1 (30). Galactose was only transferred by GALT-1 to the core ␣1,6-fucose (Fig. 7, C and D); no further galactosylation appeared to occur on any of the ␣1,3-linked fucose residues. Furthermore, the action of GALT-1 was prevented by preincubation with either FUT-1 or FUT-6. The downstream ␣1,3fucosylation by FUT-1 and FUT-6 on substrates carrying the GalFuc epitope followed pathways similar to those of the nongalactosylated glycans as described above.
Formation of a Trifucosylated N-Glycan Core in Vitro-One of the biosynthetic pathways toward the formation of the trifucosylated N-glycan core (FUT-8 3 FUT-6 3 hexosaminidase 3 FUT-1; Fig. 9) was performed on a larger scale, starting with compound 10 (0.9 mg, 0.87 mol). The sequential introduction of the fucose residues was easily monitored by following the NMR chemical shifts of the H-1 and H-6 protons of the fucose residues with different linkages (Fig. 9 and Table 2). Initially, the pentasaccharide 10 was treated with FUT-8 and GDP-Fuc to introduce the core ␣1,6-fucose, the fucosylated product 24 was purified, and 1 H NMR was recorded. The H-1 of the fucose appeared as a doublet at ␦ 4.90 ppm with a coupling constant of J H-1,H-2 ϭ 3.7 Hz. The characteristic doublet corresponding to H-6 of the newly introduced fucose appeared at ␦ 1.22 ppm.

H NMR data for fucosylated tetrasaccharide 23
Data were collected after incubation of trisaccharide 1 with FUT-6; the letters A-G refer to the residues as annotated in Scheme 1. Highlighted in boldface type are the alterations in the chemical shift value for H3 of the distal GlcNAc, indicating the introduction of the fucose at this position. ␦   trifucosylated structure are in agreement with those reported previously for a similar compound prepared by chemical synthesis (34).

DISCUSSION
Fucosyltransferase Substrate Screening-The substrate specificities of glycosyltransferases can be very subtle, and apparently small changes to glycan structures distant to the site of glycosylation can have an impact on whether a glycan is an acceptor or not; whether or not a protein-linked glycan is modified depends also on factors such as accessibility on the protein surface, glycosyltransferase expression levels, and the concentrations of the nucleotide sugar donors. The traditional view (35) was that for each glycosidic linkage, there is a specific enzyme ("one linkage-one enzyme"). However, it later became obvious that, in many circumstances, multiple enzymes can form the same linkage, or one enzyme may be able to form multiple related linkages. This scenario is shown by the activities of the six proven human ␣1,3-fucosyltransferases forming Lewis epitopes, one of which (Fuc-TIII) can form either ␣1,3 or ␣1,4 linkages, dependent on the substrate (36). In the past, screening of glycosyltransferase specificities was often unsystematic and reliant on the availability of natural sources of glycosylation precursors; generally, even for invertebrate enzymes, such as FUT-6, previously used substrates were those based on acceptors for mammalian enzymes, which led to misleading results.
However, the development of glycan arrays opens up new possibilities for examining these enzymes but to date has been (in terms of eukaryotic systems) limited to studying rather well characterized examples, such as mammalian fucosyl-and sialyltransferases or enzymes involved in plant cell wall biosynthesis (37)(38)(39)(40). On the other hand, glycosyltransferases have been useful in synthesis of glycan libraries subsequently printed onto arrays (41).
Recently, some of us have developed a systematic array of N-glycans and N-glycan-like molecules on the basis of printing alkylamine-modified chemically synthesized oligosaccharides onto glass surfaces. These have been successfully appraised also using glycosyltransferases (a galactosyltransferase, a sialyltransferase, and two core fucosyltransferases) of known specificities, employing lectins and antibodies as detection reagents (6,7,42). A particular challenge, therefore, was to examine a glycosyltransferase from a model organism with an in vitro substrate specificity apparently not matching its role in vivo.
As summarized above, the ␣1,3-fucosyltransferase homologue FUT-6 from C. elegans can act as a Lewis-type enzyme in vitro, but a deletion in the fut-6 gene results in a loss of tetrafucosylated non-Lewis-type N-glycans in vivo; among the double fucosyltransferase mutants, no more than two fucose residues are present in the fut-6;fut-8 mutant. Although suggestive of a role for FUT-6 in N-glycan processing in vivo, our data resulting from on-chip fucosylation are the first to show its unique  Fig. 7, A and B), whereas spectra G-K display those of the galactosylated glycans Hex 3 HexNAc 2-3 Fuc 0 -3 -(CH 2 ) 5 NH 2 with m/z 1367, 1513, 1310, and 1456 (see Fig. 7, C and D). All the ions annotated are [M ϩ Na] ϩ , and predicted structures of the key ions are shown in Consortium for Functional Glycomics format; the alkylamine linker, -(CH 2 ) 5 NH 2 , is represented by a short vertical bar at the right of N-acetylglucosamine. 475, HexNAc 1 Fuc 1 -(CH 2 ) 5    These data were then confirmed in assays with a variety of substrates in solution, followed by product characterization by MALDI-TOF MS/MS and NMR. The common element in both the N-glycan and Lewis-type acceptors for FUT-6 appears to be a Hex␤1,4GlcNAc unit, where the hexose can be either mannose or galactose. Interestingly, FUT-6 can transfer two fucoses to a galactosylated monoantennary N-glycan, one to the distal GlcNAc and one to the antenna (data not shown). However, only the transfer to the distal GlcNAc is meaningful in terms of the glycome of C. elegans, and, of those structures tested, only monoantennary N-glycans, lacking non-reducing terminal galactose as well as the ␣1,6-mannose of the trimannosyl core, are biologically significant substrates for FUT-6. The unusual specificity of this enzyme for such N-glycans correlates with the structures of the distally fucosylated monoantennary N-glycans observed in the hex-2;hex-3 mutant (17); our data also suggest a role for a Golgi-localized mannosidase activity removing the core ␣1,6-mannose during the biosynthesis of FUT-6-modified C. elegans N-glycans.
N-Glycan Core Modifications in Nematodes-With the knowledge that FUT-6 can fucosylate the distal GlcNAc of the N-glycan core, it is easier to understand the enzymatic biosynthesis pathways of highly fucosylated N-glycan cores in C. elegans (13,43). Thereby, the two fucosyltransferases (FUT-1 and FUT-8) are defined to solely fucosylate the proximal GlcNAc, whereas FUT-6 is the third core-modifying fucosyltransferase; thus, the order of fucosylation was of interest. Because the other ␣1,3-fucosyltransferase, FUT-1, can act on products of the core ␣1,6-fucosyltransferase FUT-8, but FUT-8 cannot act on FUT-1 products (31), the same rule might apply to FUT-6. Therefore, the order of fucosylation was tested on FUT-6 and FUT-8. Indeed, FUT-6-processed glycans (products of 1, 2, 8, and 10) cannot be modified by FUT-8 (data not shown), but FUT-8 products are acceptors for FUT-6. On the other hand, the action of GALT-1, which synthesizes the ligand for two galectins (the endogenous nematode LEC-6 and the fungal nematoxic CGL2), is inhibited when either distal or proximal ␣1,3-fucose is present. These reactions are summarized in Fig. 10; the substrate status of tested compounds is also shown in Table 3.
However, fucosylation of the core is not the end of the story. FUT-6-processed substrates are then available for the action of a putative ␣1,2-galactosyltransferase-generating ␣-galactosylated fucose (Gal␣1,2Fuc␣1,3) epitope (17), which differs from the proximal GlcNAc-linked GalFuc (Gal␤1,4Fuc␣1,6) epitope FIGURE 10. Predicted biosynthetic routes for core N-glycan modifications in nematodes. Tested biosynthetic routes (solid lines) involved in the formation of the core chitobiose modifications are summarized based on our experimental data, including reactions of pyridylaminated sugars, dabsylated glycopeptides, and chemically synthesized compounds; predicted GO or NO-GO reactions (broken lines) are judged on substrate specificities of relevant glycoenzymes; short broken lines with a double bar are "dead ends." FT, fucosyltransferase; Hex, jack bean hexosaminidase; GT1, galactosyltransferase GALT-1. Red triangles, fucose; yellow circles, galactose; blue squares, N-acetylglucosamine; green circles, mannose.

TABLE 3 Substrate status of N-glycans
Proven substrates and non-substrates for recombinant C. elegans FUT-6 are shown using the nomenclature of the Consortium for Functional Glycomics. The boxed region surrounding one antenna of one N-glycan indicates the region modified by the Lewis-type activity; the other substrates are modified on their core regions.
found in C. elegans as well as in planaria or mollusks. It is possible that an appropriate mix of approaches (analysis of mutants and use of substrates first identified by array screening) will be necessary for the identification of further glycosyltransferases required for the modification of glycans in C. elegans and other "lower" model or parasitic organisms.
Indeed, the postulated core modification pathways are not just applicable to the model organism C. elegans but also take place in parasitic nematodes, such as A. suum, H. contortus, and O. dentatum (17,18). Therefore, based on glycomic data, we presume that the distal fucosylation reaction performed by FUT-6 is specific to a subset of nematodes. However, trifucosylation has not been detected to date in a number of other nematode species, including Trichinella spiralis (44) and Onchocerca volvulus (45); this appears to correlate with the phylogeny of nematode fucosyltransferases because obvious FUT-6 orthologues are lacking in the trichinellid and filarial species (data not shown).
Perspectives-Parasites have a high impact on quality of life as well as on agricultural productivity; on the other hand, because the immune systems evolved while being subject to the selection pressure of helminth infections, the absence of parasites is possibly associated with the huge increase in allergies and autoimmunity (46). Indeed, whereas glycosylation may play a role in the efficacy of vaccination against nematodes in farm animals (47), nematode glycans have been implicated in the effects of these organisms on mammalian immune systems (48) and may be relevant to the effects of controversial therapies against autoimmune diseases, in which patients ingest nematode eggs (49). It is of interest that trifucosylated glycans, similar to those prepared during the present study in vitro, are present on the H11 glycoprotein of H. contortus, a known vaccine antigen candidate (18). Thus, the definition of parasite glycan modification pathways and the utilization of the relevant enzymes may not only aid the identification of vaccine targets or the preparation of recombinant vaccine antigens with a more natural glycosylation pattern but also facilitate the production of immunomodulatory factors; furthermore, synthesis of trifucosylated glycan structures required for definition of the natural specificity of endogenous carbohydrate-binding proteins in nematodes or of nematoxic lectins becomes feasible. Indeed, because a fut-1;fut-6 double mutant is completely resistant to nematoxic CCL2, whereas fut-1 and fut-6 single mutants are either only partially or not resistant (20,21), the FUT-6-mediated modification of nematode N-glycans is (either on its own or in combination with other residues) an interesting target for anthelminthic agents.