Bisecting Galactose as a Feature of N-Glycans of Wild-type and Mutant Caenorhabditis elegans*

The N-glycosylation of the model nematode Caenorhabditis elegans has proven to be highly variable and rather complex; it is an example to contradict the existing impression that “simple” organisms possess also a rather simple glycomic capacity. In previous studies in a number of laboratories, N-glycans with up to four fucose residues have been detected. However, although the linkage of three fucose residues to the N,N′-diacetylchitobiosyl core has been proven by structural and enzymatic analyses, the nature of the fourth fucose has remained uncertain. By constructing a triple mutant with deletions in the three genes responsible for core fucosylation (fut-1, fut-6 and fut-8), we have produced a nematode strain lacking products of these enzymes, but still retaining maximally one fucose residue on its N-glycans. Using mass spectrometry and HPLC in conjunction with chemical and enzymatic treatments as well as NMR, we examined a set of α-mannosidase-resistant N-glycans. Within this glycomic subpool, we can reveal that the core β-mannose can be trisubstituted and so carries not only the ubiquitous α1,3- and α1,6-mannose residues, but also a “bisecting” β-galactose, which is substoichiometrically modified with fucose or methylfucose. In addition, the α1,3-mannose can also be α-galactosylated. Our data, showing the presence of novel N-glycan modifications, will enable more targeted studies to understand the biological functions and interactions of nematode glycans.

Nematodes represent, along with arthropods, one of the largest groups of animals to exist on the planet; 25.000 species are described, but the existence of up to one million has been estimated (1,2). They have various ecological niches and include free-living "worms" in the soil, fungivorous, entomopathogenic, and necromenic species as well as parasites of plants and mammals, which share the basic conserved body plan (more-or-less a digestive tube surrounded with muscle, whether larger or smaller). There are five major clades (Rhabditina, Enoplia, Spirurina, Tylenchina, and Dorylaimia) (2), yet the glycosylation of only a few nematode species has been studied with an inevitable focus on the model nematode Caenorhabditis elegans and parasitic species (3). Thereby, the use of C. elegans mutants has been highly valuable in dissecting aspects of nematode N-glycan biosynthesis and revealing the in vivo substrates for certain glycosyltransferases (4).
As many nematodes are parasites, their interactions with the immune systems of their hosts have attracted attention; particularly, there are relationships between autoimmunity, allergy, vaccination, and helminth infections. The "old friends" hypothesis seeks to understand the evolutionary factors that have shaped the immune system and to explain correlations between lifestyles in the developed world and modern "epidemics," which are due to immunological misbalance (5)(6)(7). Promising data have suggested that "worm therapy" may bring advantages to some patients with Crohn's disease or allergies (8,9); however, such approaches are controversial. Nevertheless, crude extracts even of Caenorhabditis elegans were shown to induce a glycan-dependent Th2 response (10), whereas the excretory-secretory products of some nematodes also have immunomodulatory activity (11). Furthermore, the native glycoproteins of some nematodes have proven effective in vaccination trials, whereas recombinant forms are not, which is suggestive that post-translational modifications may have a role in an efficacious immune response (12).
As at least some of the molecules relevant to nematode immunomodulation or vaccination are glycoproteins, a proper understanding of nematode glycosylation is of biomedical and veterinary relevance. Over the years, it has become apparent that the core chitobiosyl region of nematode N-glycans is subject to a range of modifications, with up to three core fucose residues being present (␣1,3and ␣1,6-linked on the reducing-terminal "proximal" GlcNAc and ␣1,3-linked on the second "distal" GlcNAc). However, up to four fucose residues have been detected on C. elegans N-glycans and the exact nature of the linkage of the fourth fucose has remained obscure despite work in our own and other laboratories (3,(13)(14)(15). Combined with the latest knowledge regarding the specificity of C. elegans core fucosyltransferases (13,16,17) as well as our recent data regarding the exact structures of N-glycans from the C. elegans double hexosaminidase mutant and other nematodes (18 -20), we concluded that some models for the tri-and tetrafucosylated N-glycans were incorrect. By preparing a triple mutant unable to core fucosylate its N-glycans, we generated a C. elegans strain containing maximally one fucose residue on the N-linked oligosaccharides. Thereby a pool of unusual mannosidase-resistant N-glycans was identified and, using mass spectrometry (MS) and NMR, we reveal their modification with bisecting galactose frequently capped with fucose or methylfucose.

EXPERIMENTAL PROCEDURES
Preparation of the C. elegans Triple Mutant-Wild-type C. elegans (N2) and single mutants fut-1(ok892), fut-6(ok475) and fut- 8(ok2558) were obtained from the Caenorhabditis Genetics Centre (CGC), University of Minnesota, USA. All C. elegans strains were cultured under standard conditions at 20°C (21). A fut-1;fut-6 double mutant was generated by standard crossings (22). Briefly, the ok892 single mutant was first crossed with N2 wild-type males in order to produce male progeny for subsequent fertilization of the ok475 single mutant. Hermaphrodites from the F2 generation were isolated and allowed to produce eggs prior to examination of the maternal genotypes by PCR. F3 progeny from the heterozygotes carrying both mutations were isolated in a large amount and their genotypes were also examined 48 h after isolation. Additional isolations of heterozygotes and genotyping PCR were carried out until homozygotes with both mutations appeared. To generate the triple mutant, progeny of the ok2558 single mutant and N2 were crossed to fut-1;fut-6 double mutants prior to genotype screening. Following the workflow mentioned above, homozygous mutations in the three fucosyltransferase genes were accumulated over a few generations until a genome of a triple mutant was obtained. Genotypes of each generation were followed using the primers described below.
N-glycan Preparation and MALDI TOF MS Analysis-C. elegans were grown in liquid culture with E. coli OP50 in standard S complete medium, harvested after cultivation at room temperature (20°C) for 4 -6 days and purified by sucrose density centrifugation (in two independent preparations, the yield was 5 and 9 g of worms respectively). N-glycans were released from worm peptic peptides using peptide/N-glycosidase F as previously described, with a subsequent digestion of remaining glycopeptides using peptide/N-glycosidase A (23). The N-glycome of the mutant was profiled by MALDI-TOF MS (Autoflex Speed, Bruker Daltonics, Bremen, Germany) in positive ion mode using FlexControl 3.4 software. Free glycans were labeled with 2-aminopyridine prior to fractionation by normal phase high pressure liquid chromatography (NP-HPLC) 1 and reversed-phase HPLC (RP-HPLC; see below). All the HPLC peaks were collected and examined by matrix-assisted laser desorption ionization-time-of-flight (MALDI-TOF) MS, using 6-aza-2-thiothymine as matrix; tandem MS (MS/MS) to confirm the composition of all proposed structures was performed by laser-induced dissociation (precursor ion selector was generally set to Ϯ0.6%). The detector voltage was generally set at 1977 V for MS and 2133 V for MS/MS; 1000 -3000 shots from different regions of the sample spots were summed. Spectra were processed with the manufacturer's software (Bruker Flexanalysis 3.3.80) using the SNAP algorithm with a signal/noise threshold of six for MS (unsmoothed) and three for MS/MS (four-times smoothed). In total ϳ1500 MS and MS/MS spectra were manually interpreted on the basis of the mass, fragmentation pattern, and results of chemical and enzymatic treatments; isobaric structures present in different RP-HPLC fractions were defined on the basis of comparisons in the aforementioned parameters. At least five MS/MS fragment ions were used to aid definition of each of the structures.
NMR Spectroscopy-The isolated oligosaccharides have been lyophilized, dissolved in D 2 O (99.996%; Sigma-Aldrich) in concentrations of ϳ150 g in 600 l and transferred into 5 mm high precision NMR sample tubes (Promochem, Wesel, Germany). All spectra have been recorded on a Bruker AV III-600 AVANCE spectrometer (Bruker, Rheinstetten, Germany) at 600.13 MHz ( 1 H) equipped with a Cryo-Probe™ Prodigy and were performed using the Bruker Topspin 3.1 software. The 1D proton spectra were recorded with presaturation, acquisition of 32 k data points and a relaxation delay of 1.0 s. After zero filling to 64 k data points and Fourier transformation spectra were performed with a range of 7200 Hz. 2D homonuclear DQF-COSY and TOCSY (100 ms mixing time) spectra have been measured with standard Bruker programs; 128 experiments, each with 2048 data points, were recorded with an appropriate number of scans. Linear forward prediction to 256 data points in the f 2 dimension and sinusoidal multiplication in both dimensions and Fourier transformation led to 2D-spectra with a range of 6000 Hz in both dimensions. All measurements have been made at 298.1 K and chemical shifts were referenced to external acetone (␦ H 2.225 ppm).
LC-MS/MS of HPLC-purified PA-labeled N-glycans-PA-labeled N-glycans were analyzed by LC-MS/MS using a 10 cm ϫ 150 m I.D. column, prepared in-house, containing 5 m porous graphitized carbon (PGC) particles (Thermo Scientific, Waltham, MA). Glycans were eluted using a linear gradient from 0 -40% acetonitrile in 10 mM ammonium bicarbonate over 40 min at a flow rate of 10 l/min. The eluted N-glycans were detected using a LTQ ion trap mass spectrometer (Thermo Scientific) in negative-ion mode with an electrospray voltage of 3.5 kV, capillary voltage of Ϫ33.0 V and capillary temperature of 300°C. Air was used as a sheath gas and mass ranges were defined dependent on the specific structure to be analyzed. The data were processed using the Xcalibur software (version 2.0.7, Thermo Scientific). Glycans were identified from their MS/MS spectra by manual annotation; the nomenclature of Domon and Costello for fragment annotation was employed (25).

Impact of a Triple Fucosyltransferase Knock-out on the C.
elegans Glycome-Previously, N-glycans with up to four fucose residues have been detected in C. elegans (13, 15, 26 -30), whereas only three fucosyltransferases (FUT-1, FUT-6 and FUT-8) required for the trifucosylation of the core chitobiosyl region of N-glycans in C. elegans have been identified (13,16,17). With the goal of restricting the N-glycome of this organism to glycans with maximally one fucose, a triple mutant with deletions in the three corresponding genes (fut-1; fut-6;fut-8) was constructed. Deletions of all three genes were confirmed by genomic PCR (supplemental Fig. S1), but no overt major phenotypic defects were detected under laboratory conditions. The N-glycans were prepared after largescale liquid cultivation of the worms and were subsequently fluorescently labeled with 2-aminopyridine; the overall mass spectrometric profile indicated the presence of a range of N-glycans (m/z 665-1961 as [MϩH] ϩ ; Fig. 1A and Table I). More exact examination of the spectra showed mass differences of 2 or 14 Da (e.g. m/z 1135, 1149 and 1151 or 1297, 1311 and 1313 as [MϩH] ϩ ); thereby, some ⌬m/z values between glycans were of 146 (deoxyhexose), 160 (methyl and deoxyhexose), or 162 (hexose). This was an indication for the presence of fucose on a number of glycans and suggested compositions of Hex 3-4 HexNAc 2 Fuc 1 Me 0 -1 -PA. In addition to the fucosylated and/or methylated glycans, masses corresponding to oligomannosidic and phosphorylcholine-modified oligosaccharides were detected (Table I); the latter modification is defined because of the diagnostic m/z 369 MS/MS fragments (HexNAc 1 PC 1 ; supplemental Fig. S2). However, as such glycans have been previously described in nematodes (14, 15, 18 -20), their analysis was not the focus of the current study.
Upon incubating the complete N-glycome with jack bean ␣-mannosidase, the disappearance of the oligomannosidic glycans was accompanied by a large increase in the peak at m/z 665 (Hex 1 HexNAc 2 -PA). Nevertheless, a number of glycans were not shifted by this treatment, including the one at m/z 1151, which is predicted to have the composition of Hex 4 HexNAc 2 -PA. On the other hand, the glycan with the composition Hex 5 HexNAc 2 -PA ([MϩH] ϩ ion of m/z 1313) was now absent (Fig. 1B). Comparable data regarding mannosidase-resistance of some glycans of the mutant strain was obtained by RP-HPLC ( Fig. 1C and 1D), which is well-established to also separate isomers of many N-glycans. Compatible with the overall mass spectrometric profiles, the major HPLC peak of 7.2 g.u. containing Man 3 GlcNAc 2 -PA and Man 5 GlcNAc 2 -PA was replaced, after mannosidase digestion, by a new peak of 6.5 g.u. containing a glycan of m/z 665 corresponding to Man 1 GlcNAc 2 -PA; the retention time for this "trimmed-down" product of the oligomannosidic glycans is in keeping with literature values (31). More significantly, a set of glycans eluting between 4.0 and 5.5 glucose units (g.u.) were seemingly resistant to this treatment because their elution position did not shift after incubation with the mannosidase. MALDI-TOF MS of the relevant fractions indicated that they contained glycans with the composition of Hex 3-4 HexNAc 2 Fuc 0 -1 Me 0 -2 -PA ([MϩH] ϩ ions of m/z 989 to m/z 1311). As expected from having deletions in all three known core fucosyltransferase genes, MS/MS of these glycans strongly suggested that core fucosylation (which would be shown by a fragment of m/z 446, i.e. Fuc 1 GlcNAc 1 -PA) was absent, whereas in general MS/MS spectra of the various glycans primarily confirmed composition (Fig. 2). Treatment of the glycopeptides remaining after PNGase F digestion with PNGase A merely resulted in release of residual glycans of the same m/z as those in the PNGase F pool and that also displayed no sign of core fucosylation as judged by MS/MS (data not shown).
Structure of Mannosidase-resistant Paucihexosidic N-glycans-As a number of glycans in the triple mutant were resistant to ␣-mannosidase digestion regardless of compositions normally associated with paucimannosidic N-glycans present in many invertebrates, we sought to examine these more closely, including the fucosylated forms. First, a fraction eluting at 4.2 g.u. (see Fig. 1C) containing glycans of Hex 3-4 HexNAc 2 -PA (m/z 989 and 1151) and one of Hex 4 HexNAc 2 Fuc 1 -PA (m/z 1297) was analyzed. All three glycans were confirmed to be mannosidase-resistant, but the two glycans lacking fucose lost one hexose upon treatment with recombinant Aspergillus ␤-galactosidase (Fig. 3). Subsequent incubation with ␣1,2/3-specific mannosidase resulted in primary products of Hex 1-2 HexNAc 2 -PA. Considering previous GC-MS data that 3,4,6-trisubstituted mannose exists in C. elegans (27), whereas on the other hand the relevant GlcNAc-TIII has no homolog in nematodes and that bisecting GlcNAc is known to cause mannosidase-resistance of the two core ␣-mannose residues (32, 33), we postulated that a bisecting position (C-4) for ␤-galactose prevented removal of the nonsubstituted ␣-mannose residues. A C-2 modification of the core ␤-mannose is unlikely as glycans carrying a substitution on the C-2 position, such as ␤1,2-xylose, display a different mannosidase sensitivity to the glycans studied here (34).
In contrast, the fucose-containing glycan of m/z 1297 (m/z 1319 as [MϩNa] ϩ ) in the 4.2 g.u. fraction was resistant to both ␣-mannosidase and ␤-galactosidase. A closer examination of the MS/MS spectrum shows that the fucose is two hexose residues distant from the distal GlcNAc as shown by the fragment ions at m/z 973 (or 987 in case of the methylated structures; Fig. 2). To explore the structure of Hex 4 HexNAc 2 -Fuc 1 -PA more thoroughly, we performed selected treatments on a 2D-HPLC fraction containing solely this glycan. Initially, we tried a number of fucosidase treatments, but no removal of fucose was observed (data not shown). However, the glycan is partially sensitive to hydrofluoric acid and the defucosylated portion was sensitive to ␤-galactosidase, but resistant to ␣-galactosidase (Fig. 4). Subsequently, one mannose could be released from the ␤-galactosidase-sensitive portion when incubating the glycan with ␣1,2/3-mannosidase; because of the initial galactosidase resistance, the pattern of digestion after hydrofluoric acid treatment suggests that the glycan has a fucose cap on the ␤-galactose of a structure the same as the aforementioned Hex 4 HexNAc 2 -PA. Regarding the type of linkage, we found in control experiments that hydrofluoric acid can remove ␣1,2-fucose residues, but not so efficiently as it cleaves ␣1,3-fucose (supplemental Fig. S3); however, longer incubations resulted in artifacts. Therefore, considering also previous GC-MS data showing the presence of 2-substituted galactose (27) and the multiplicity of ␣1,2-fucosyltransferase homologs (35) in C. elegans, we propose that the Hex 4 HexNAc 2 Fuc 1 -PA contains an ␣1,2-fucose linked to the bisecting ␤-galactose. Further evidence came from NMR and LC-MS n experiments (see below).
Methylated Fucose as a Component of Paucihexosidic Nglycans-We also investigated the nature of N-glycans putatively containing methyl groups. The fraction of 5.2 g.u (see Fig. 1C) contained, other than traces of Man 7-9 GlcNAc 2 , a glycan with [MϩH] ϩ ions at m/z 1149, which could correspond to Hex 3 HexNAc 2 Fuc 1 Me 1 -PA. MS/MS showed a fragment ion at m/z 987, which suggested the presence of methylfucose (Fig. 2F). Thus considering the aforementioned experiments, we treated this fraction with hydrofluoric acid, which resulted in the expected loss of 160 Da (i.e. methylfucose) to yield a glycan of m/z 989 (Hex 3 HexNAc 2 -PA; Fig. 5). This glycan was resistant to ␣-mannosidase (as judged by the ratio of m/z 989 to m/z 1149 remaining constant; the presence of m/z 665 and 827 products was assigned to digestion of the oligomannosidic glycans also present in this fraction). The hydrofluoric acid-treated fraction was then incubated sequentially with ␤-galactosidase and ␣1,2/3-mannosidase, which resulted in products of m/z 827 and 665. The portion of Hex 3 HexNAc 2 Fuc 1 Me 1 -PA, which had not lost the methylfucose residue was, however, resistant to both hexosidases. Among other fractions, we examined the ␣-galactosidase and hydrofluoric acid sensitivity of glycans of m/z 1135 (Hex 3 HexNAc 2 Fuc 1 -PA) and 1163 (Hex 3 HexNAc 2 Fuc 1 Me 2 -PA) as well as two isomers of m/z 1311 (Hex 4 HexNAc 2 Fuc 1 -Me 1 -PA) separated in the two fractions of 4.8 and 5.5 g.u.; the earlier fraction also contains an m/z 989 structure, which is a putative reducing-terminal epimer of standard Man 3 GlcNAc 2 (epimerization of up to 10% of the reducing-terminal GlcNAc to ManNAc has been previously reported (36)). Whereas the m/z 1311 glycan (Hex 4 HexNAc 2 Fuc 1 Me 1 -PA) of 5.5 g.u. was sensitive to ␣-galactosidase, the isobaric structure eluting at 4.8 g.u. was resistant to this treatment ( Fig. 6B and 6E). On the other hand, all four fucosylated structures in the two fractions were sensitive to hydrofluoric acid as indicated by the incomplete loss of 146 or 160 Da (fucose or methylfucose; Fig. 6C and 6F). Notably, the Hex 3 HexNAc 2 Fuc 1 Me 2 -PA glycan (m/z 1163) appears to be modified with methyl groups on both the fucose and the ␣1,3-mannose residues. This is compatible with the MS/MS spectrum indicating an m/z 841 fragment (putatively Man 2 GlcNAc 2 Me 1 -PA; Fig. 2G). Due to their similar fragmentation and digestion properties, it is concluded that the basic structures of the Hex 3-4 HexNAc 2 Fuc 1 and Hex 3-4 HexNAc 2 Fuc 1 Me 1 glycans are the same. Indeed, the fragmentation patterns of the isobaric hydrofluoric acid digestion products of m/z 1297 and m/z 1311 were highly similar suggestive of a common basic structure (supplemental Fig.  S4C and S4D).
NMR Analysis Indicative of Bisecting Galactose-As the data from chemical and enzymatic digestions were indicative for an unusual location of galactose residues, which blocked the action of ␣-mannosidase, we sought further confirmation for our model of bisecting galactose being a feature of the mannosidase-resistant glycans of C. elegans. Because of the relatively low amount of glycan material available, we considered use of a nondestructive method of glycan analysis and so turned to NMR to yield further insights. To aid definition of the novel features of the putatively bisected glycans from the C. elegans, 1 H NMR and homonuclear 2D NMR spectra of "classical" Man 3 GlcNAc 2 -PA and a pool of mannosidaseresistant paucihexosidic glycans from the triple mutant were compared (Fig. 7). Data on related N-glycan structures were used for comparative analysis of the interglycosidic bonds (37)(38)(39).
For the pool of putatively bisected glycans, the structural variants led to broadening or doubling of some proton NMR signals; certainly, the modification of the core ␤-D-Manp (residue 3; see Fig. 7), as compared with standard Man 3 GlcNAc 2 -PA, can be surmised by the alterations in a number of the chemical shifts for this residue (e.g. 3.77, 3.80 and 3.85 ppm rather than 3.82, 3.69 and 3.77 ppm for the H-3, H-4 and H-6b protons; see Table II). The strong shift for H-4 can be taken as confirmation for a bisecting residue on C-4 of the core ␤-man-  Fig. 3 (obtained by NP-HPLC followed by RP-HPLC) was incubated with green coffee bean ␣-galactosidase, but no hexose residue was removed (B); hydrofluoric acid treatment converted ca. 40% of the structure into a defucosylated form with m/z 1151.5 (C). Afterward, the product was treated with either ␣-galactosidase (D) or ␤-galactosidase; only ␤-galactosidase resulted in a loss of a hexose residue (E). Finally, ␣1,2/3-mannosidase further trimmed the ␤-galactosidase product and yield Hex 2 HexNAc 2 -PA with m/z 827.3 (F). nose and so shares a trend observed with data on a bisected mammalian glycan (40), whereas the data are not compatible with a C-2 modification. The presence of ␤-D-Galp (residue 4Ј) and some ␣-d-Galp (residue C) linked to mannose as well as some ␣-L-Fucp linked to galactose is shown by relevant chemical shifts typical for such residues (4.6 -4.3 (38, 40), 5.5-5.1 (41), and 5.3-5.2 ppm (42) respectively for the anomeric H-1); for the latter, it is noteworthy that older NMR data was interpreted as showing ␣1,2-fucosylation of mannose (43), but our data overall indicate the presence of a bisecting Fuc␣1,2Gal motif. Indeed, the proton signals of the ␤-D-Galp 4Ј show the most pronounced shift variations, which may be caused by the fucose or methylfucose bound in substoichiometric amounts to this unit, whereas the ␣-galactose is assumed to be linked to position 2 of the ␣-D-Manp (residue 4). In conclusion, the NMR data support the presence of 3,4,6-trisubstituted bisected core ␤-mannose as well as the presence of galactose and fucose residues.
LC-MS Analyses Confirm the Fuc␣1,2Gal␤1,4 Modification-As a final analytical method, purified N-glycans were also applied to LC-MS/MS in negative ion mode (Fig. 8). All MS/MS spectra were dominated by an ion resulting from 2,4 A cross-ring cleavage of penultimate GlcNAc, which are diagnostic for N-glycans. In the case of the singly-charged precursors, we observed Y 3  This ␣-mannosidase resistant structure was separated by RP-HPLC, eluting at 5.2 glucose unit (A). HF treatment was first applied to the fraction and resulted in a partial conversion to Hex 3 HexNAc 2 -PA (m/z 989.4) (B). This product was incubated with either jack bean ␣-mannosidase or ␤-galactosidase, but only the latter resulted in loss of a hexose residue (C and D). Further digestion of the ␤-galactosidase product using ␣1,2/3-mannosidase resulted in removal of the "lower arm" mannose and formation of a final product with m/z 665.3. tensity MS 2 ions, though, were more informative in terms of the linkage of the bisecting galactose and of the fucose residues.
FIG. 6. Structural characterization of two isomeric N-glycans by ␣-galactosidase digestion and HF treatment. Two different RP-HPLC fractions containing Hex 4 HexNAc 2 Fuc 1 Me 1 -PA with m/z 1311.6 (4.8 g.u. and 5.5 g.u.; A and D) were treated with ␣-galactosidase (B and E); complete loss of one hexose was observed only for the late eluting fraction, resulting in a product of m/z 1149.5. Hydrofluoric acid treatment resulted in partial removal of either fucose or methylfucose (FMe) residues from Hex 3-4 HexNAc 2 Fuc 1 Me 0 -2 -PA (C and F); thus, the fucose is not ␣1,3-linked (which would be fully removed by this treatment) and not ␣1,6-linked (which would be resistant), but the degree of release is compatible with the proposed Fuc␣1,2 linkage.  Table II.   TABLE II  1 Figure 7; numbers in brackets refer to the corresponding chemical shifts for the Man 3 GlcNAc 2 structure. Significant differences in the chemical shifts (0.08 -0. 12   This was further confirmed by the presence of Z 3␣ /Z 3␤ -CH 2 O ions at m/z 597, which indicated a C-3 and C-4 bi-substitution of the ␤-Man (Fig. 8A, right panel); this is similar to the Z i /Z i -CH 2 O fragmentation pattern of a Lewis type structure (45). In addition, fragment ions at m/z 443 were concluded to result from 0,2 A cleavage of ␤-Man, which would be absent if the C-2 of the core ␤-mannose is substituted; together with the ␤-galactosidase sensitivity and NMR data, the MS/MS results are compatible with this glycan carrying a bisecting  (45). Thus, consistent with the partial sensitivity toward hydrofluoric acid, we conclude that the bisecting Gal is modified with ␣1,2-linked Fuc.

H NMR chemical shifts (parts per million) of the mannosidase-resistant N-glycan pool. The nomenclature for the numbering of the saccharide units is according to that of Halbeek et al. (37). Structures and spectra are shown in
Galactose as a Component of Paucihexosidic N-glycans in Wild-type C. elegans-Considering the results with the triple mutant, we sought for glycans of the same RP-HPLC retention time (4.0 -5.5 g.u.) as the mannosidase-resistant forms in other strains. As part of an earlier study, extensive 2D-HPLC fractionation (normal phase followed by reversed phase) had been performed on the pmk-1 strain, which is defective in a p38 MAP kinase homolog and is hypersensitive to some fungal galectins (46), but in glycosylation terms is "pseudowild-type". We found one 2D-HPLC fraction containing a glycan with the composition Hex 5 HexNAc 2 -PA (m/z 1313), which did not co-elute with standard Man 5 GlcNAc 2 -PA (7.2 g.u.) on a standard RP-HPLC column, but instead eluted at 4.6 g.u., i.e. within the elution range for the mannosidaseresistant glycans from the triple mutant. The MS/MS spectrum of this structure was also different as compared with that of the standard Man 5 GlcNAc 2 (i.e. the dominance of the m/z 827 fragment in the latter was not apparent; supplemental Fig.  S4E and S4F).
Considering the hexosidases available and data indicative for the presence of galactose on C. elegans N-glycans, we attempted both ␣-mannosidase and ␣or ␤-galactosidase treatments of this glycan. Indeed, the Hex 5 HexNAc 2 -PA from the pmk-1 strain was resistant to ␣-mannosidase, but sensitive to the recombinant fungal ␤1,4-galactosidase (Fig. 9). After this treatment, a further hexose could then be removed with jack bean ␣-mannosidase, but further digestion was still not possible. We observed that ␣-galactosidase could remove one hexose and that a final hexose is cleaved by an ␣1,2/3-specific mannosidase. Thus, the Hex 5 HexNAc 2 -PA glycan was concluded to have the composition Gal 2 Man 3 -GlcNAc 2 ; the digestion data is compatible with a bisecting position for ␤-galactose and ␣-galactosylation of the ␣1,3mannose residue. Furthermore, in the wild-type N2 embryo N-glycome, an Hex 4 HexNAc 2 glycan (4.2 g.u.) displayed a similar ␣-mannosidase resistance, but ␤-galactosidase sensitivity, as for the co-eluting m/z 1151 glycan from the triple mutant (data not shown).

DISCUSSION
Studies over the past 15 years have revealed a huge diversity in the N-glycome of the "simple" multicellular nematode Caenorhabditis elegans (3,14). Some of the features are shared with other parasitic and nonparasitic nematode species (18 -20, 47, 48). Examples include trifucosylation of the core chitobiose (including fucosylation of the distal core GlcNAc), galactosylation of core fucose residues and the antennal modification with phosphorylcholine. On the other hand, although antennal ␣1,3-fucosylation (Lewis and Lewistype epitopes) is known from some parasitic nematode species (49), this feature is lacking from C. elegans. However, a fourth fucose residue is apparent in a subset of N-glycans of C. elegans (13, 15, 26 -30) and structural models have suggested that this is attached either directly or indirectly (via galactose) to ␣-mannose residues of the trimannosylchitobiosyl core (15,26,50,51). However, tetrafucosylation of a trimannosylchitobiosyl core is a structural proposition contradicting not only the known specificity of the enzyme (the FUT-6 ␣1,3-fucosyltransferase), which modifies the distal core GlcNAc (17), but also GC-MS data indicating that a portion of the core ␤-mannose residues is 3,4,6-trisubstituted and that 2-substituted galactose is also present in wild-type C. elegans (27).
Mutant C. elegans strains have been highly valuable in verifying glycobiosynthetic pathways in this organism: in our laboratory we have analyzed single fucosyltransferase mutants (13), double fucosyltransferase mutants (17), a Golgi mannosidase II mutant (52), and hexosaminidase single and double mutants (18,53). Also, the N-glycomes of worm strains with other glycosyltransferase, fucosylation and methylation defects have been analyzed (28,51,54,55). Very often one observes not just the loss of N-glycans but the apparent gain of others.
In order to simplify the N-glycome, we constructed a triple mutant featuring deletions in the genes known to encode fucosyltransferases modifying the core chitobiose unit of C. elegans N-glycans. Indeed, as predicted the N-glycans of the triple mutant contained maximally one fucose residue. Focusing on a range of HPLC-enriched N-glycans, which possess up to five hexose residues, but which were ␣-mannosidase resistant (hence "paucihexosidic"), we reveal, using chemical and enzymatic treatments in conjunction with off-line MALDI-TOF MS and on-line LC-MS as well as NMR, that the fucose residue on such glycans is ␣1,2-linked via a bisecting ␤1,4galactose to the core ␤-mannose. Thereby, some of these glycans either lack an ␣1,6-mannose or carry an ␣-galactose on the ␣1,3-mannose; variants of these glycans are also methylated.
Preliminary data, as well as the aforementioned "old" GC-MS data (27), lead us to believe that the antennal fucose is also significantly present on core fucosylated glycans in the wild-type. Furthermore, we have previously shown that the "GalFuc" epitope can also carry a fucose residue (18), which yields a fifth attachment point for fucose on C. elegans N-glycans; nevertheless, only maximally four fucose residues have ever been detected on oligosaccharides of this organism. The difficulty in digesting these bisected N-glycans with glycosidases may be explained by predicted 3D-conformations suggesting that the bisecting modification "folds back" onto the core of the glycan (supplemental Fig. S5).
Thus, after some fifteen years of work on C. elegans glycans by various laboratories, we offer a reinterpretation of a number of previously-published studies regarding the N-glycome of this organism, which may have also repercussions for the understanding of data regarding glycan-binding nematotoxic proteins. As a double hexosaminidase worm mutant, which apparently lacks bisecting galactose (18), is completely or partially resistant to tectonin and MpL (51,56), it would be attractive to propose that the binding sites on N-glycan antennae for these nematotoxins are indeed the methylated or nonmethylated forms of the ␣1,2-fucose attached to the bisecting galactose. Armed with the exact structure of C. elegans N-glycans, the binding specificities of such proteins can be reassessed; furthermore, such glycomic knowledge is a prerequisite to fine-tune approaches to understand the biological significance of the seemingly endless variation of the C. elegans glycome in order to make this otherwise wellunderstood organism into a system truly suitable for examination of glycobiological paradigms. □ S This article contains supplemental Figs. S1 to S5. ‡ ‡ Current address: Helmholtz-Institut fü r Pharmazeutische Forschung Saarland, 66123 Saarbrü cken, Germany.