Null Mutations in Drosophila N-Acetylglucosaminyltransferase I Produce Defects in Locomotion and a Reduced Life Span*

UDP-GlcNAc:α3-d-mannoside β1,2-N-acetylglucosaminyltransferase I (encoded by Mgat1) controls the synthesis of hybrid, complex, and paucimannose N-glycans. Mice make hybrid and complex N-glycans but little or no paucimannose N-glycans. In contrast, Drosophila melanogaster and Caenorhabditis elegans make paucimannose N-glycans but little or no hybrid or complex N-glycans. To determine the functional requirement for β1,2-N-acetylglucosaminyltransferase I in Drosophila, we generated null mutations by imprecise excision of a nearby transposable element. Extracts from Mgat11/Mgat11 null mutants showed no β1,2-N-acetylglucosaminyltransferase I enzyme activity. Moreover, mass spectrometric analysis of these extracts showed dramatic changes in N-glycans compatible with lack of β1,2-N-acetylglucosaminyltransferase I activity. Interestingly, Mgat11/Mgat11 null mutants are viable but exhibit pronounced defects in adult locomotory activity when compared with Mgat11/CyO-GFP heterozygotes or wild type flies. In addition, in null mutants males are sterile and have a severely reduced mean and maximum life span. Microscopic examination of mutant adult fly brains showed the presence of fused β lobes. The removal of both maternal and zygotic Mgat1 also gave rise to embryos that no longer express the horseradish peroxidase antigen within the central nervous system. Taken together, the data indicate that β1,2-N-acetylglucosaminyltransferase I-dependent N-glycans are required for locomotory activity, life span, and brain development in Drosophila.

The first committed step toward synthesis of Glc 3 Man 9 GlcNAc 2pyrophosphate-dolichol is catalyzed by UDP-GlcNAc:dolichylphosphate GlcNAc-1-phosphate transferase. This step is essential for the synthesis of all N-glycans. Tunicamycin, a GlcNAc analogue that is a competitive inhibitor of GlcNAc-1-phosphate transferase, is toxic to yeast (24) and to mammalian cells in culture (25). Tunicamycin prevents normal mouse embryogenesis (26,27). GlcNAc-1-phosphate transferase-null mouse embryos complete preimplantation development but die 4 -5 days after fertilization; neither trophoblast nor embryonic endodermal lineages derived from these early embryos survive in culture in vitro indicating that N-glycosylation is needed for the viability of early embryonic cells (28). In contrast, Chinese hamster ovary cells suffer no obvious phenotypic abnormalities in the absence of GlcNAcTI (29). Mgat1 null mice, however, die at embryonic stage E9.5 days (30,31). The data indicate that the oligomannose N-glycans made in the first two phases of N-glycan synthesis are essential to the survival of both unicellular and multicellular animals, whereas GlcNAcTI-dependent N-glycans are needed for normal vertebrate development but not for the survival of individual cells. Here we present evidence that GlcNAcTIdependent N-glycans are also needed for the normal development of D. melanogaster.
Vertebrate glycan function has been studied by analysis of mice and humans with mutations in genes required for glycosylation (32)(33)(34). Such studies are complicated by the fact that synthesis of glycans requires a complex multienzyme system acting on a large number of protein targets. In the expectation that the relatively primitive N-glycan synthesis pathway in Drosophila and C. elegans may be more amenable to functional analysis than the vertebrate pathway, we have initiated studies on Mgat1 null mutations in these invertebrates. We have shown that C. elegans Mgat1-deficient mutants are viable and have an apparently normal phenotype when grown under standard laboratory conditions but show decreased survival times when exposed to pathogenic bacteria (35). We have reported the cloning and expression of the Drosophila Mgat1 gene (36). Here we show that null mutations in Drosophila Mgat1 give rise to viable adults with dramatically altered N-glycans that result in pronounced defects in locomotion, a severely reduced life span, and abnormal brain development.

EXPERIMENTAL PROCEDURES
Fly Stocks and Generation of Mutants-All fly stocks were maintained at room temperature on standard cornmeal agar media. The line y 1 w 67c23 ; Mgat1 KG02444 (stock BL-13222; Bloomington Stock Center) contains a P-element insertion at the 5Ј end of Mgat1 and was used to generate both the precise excision line Mgat1 ϩ9 and the imprecise excision allele Mgat1 1 . Because Mgat1 ϩ9 was generated at the same time as the Mgat1 1 allele and is in the same genetic background, this line was used as a control for all of our subsequent analyses (GlcNAcT1 assays, mass spectrometry, immunohistochemistry, locomotion, and life span) and behaved as wild type. y w; Sp/CyO; Sb⌬2-3/TM6, Ubx was used as a source of transposase. The excision breakpoints were determined by PCR analysis and sequencing using the primers F1 5Ј-CCGATTGGGG-TAGGTAAAT and R3 5Ј-CTGAGAGTGGCACACTTTC. The P-element line l(2)05510 05510 cn 1 /CyO; ry 506 contains a lethal mutation in the gene immediately upstream of Mgat1 (stock BL-12192).
GlcNAcTI Enzyme Assay-Fourteen adult flies from each group were homogenized in 0.2 ml of 25 mM MES buffer, pH 6.5, containing 1% Triton X-100 and protease inhibitor mixture. GlcNAcTI activity was measured using 0.6 mM Man␣1-6[Man␣1-3]Man␤1-O-n-octyl (Toronto Research Chemicals, Toronto, Canada) as acceptor substrate and 1.2 mM UDP-[ 3 H]GlcNAc (100,000 dpm/nmol) as donor substrate. The assay mixture also contained 3.0 mM AMP, 60 mM GlcNAc, 20 mM MnCl 2 , in 0.05 M MES buffer, pH 6.5, and 0.01 or 0.02 ml of enzyme in a total volume of 0.04 ml. Time of incubation was 60 min at 37°C. The assays were carried out as described previously (36). The rate of product formation was proportional to enzyme volume.
Mass Spectrometric Analysis of N-Glycans of Wild Type and Mutant Drosophila-Adult flies (ϳ0.5 g of Mgat1 ϩ9 /Mgat1 ϩ9 , Mgat1 1 /CyO-GFP, Mgat1 1 /Mgat1 1 ) were anesthetized with carbon dioxide, suspended in 0.3 ml of water, and boiled for 10 min. The preparation of N-glycans was performed as described previously (11,37). Proteins were extracted in lysis buffer (35 mM Tris, 8 M urea, 4% CHAPS, 65 mM dithiothreitol, pH 8.0), and centrifuged at 10,000 ϫ g for 15 min. The protein content of the supernatant was determined using the Bradford assay (38). Protein in the supernatant was precipitated with 15% trichlo- The names assigned to these structures are used in the synthetic schemes shown in Fig. 2 (M ϭ Man; Gn ϭ GlcNAc; F ϭ Fuc). Oligomannose N-glycans have from 5 to 9 Man residues; only M9Gn2 and M5Gn2 are shown. M3Gn2 and M4Gn2 are paucimannose N-glycans. The remaining structures have a GlcNAc␤1-2Man␣1-3 moiety and are therefore dependent on prior GlcNAcTI action. The R group is defined in the figure; there are four R variants depending on the absence or presence of core ␣1-3and ␣1-6-linked Fuc residues (designated as F 3 and F 6 , respectively).  Table 1 and on Paschinger et al. (44). The names of the N-glycans are defined in Fig. 1. Enzyme names are in italics. Reactions shown by continuous arrows have been established by in vitro assays, and broken arrows are based on other evidence (44). Arrows crossed with double lines indicate reactions that do not occur. The figure shows the conversion of oligomannose N-glycans (M9-5Gn2) to hybrid (GnM5-3Gn2), paucimannose (M4-3Gn2), and roacetic acid on ice for 1 h. The protein pellet was washed with 1 ml of acetone (three times) and 1 ml of chloroform/methanol/water (10:10:1) (three times), dried under nitrogen, and stored at Ϫ20°C. The protein pellet was lyophilized prior to treatment with protein N-glycanase F (PNGase F). The released glycans were purified and subjected to matrixassisted laser desorption/ionization-time of flight mass spectrometry (MALDI-TOF-MS). The glycans were reduced with NaBH 4 (0.2 ml of 10 mg/ml NaBH 4 in 10 mM NaOH) overnight at room temperature. Borate was removed by adding 2 drops of acetic acid on ice, followed by co-evaporation with 3 ml of ethanol, 3 ml of 1% acetic acid in methanol (five times), and 1 ml of toluene (three times). The reduced glycans were desalted, permethylated (39), and analyzed by MALDI-IT-TOF (where IT indicates ion trap) MS for confirmation. Collision-induced dissociation was performed using MALDI-QIT MS (where QIT indicates quadrupole ion trap) (Kratos-Shimadzu Biotech) with 2,5-dihydroxybenzoic acid as matrix (40). All the structures reported here were confirmed by derivatization and MALDI-IT-MS.
Locomotory Activity-The locomotory activity of adult male and female flies was measured using a slightly modified open field test (41). Briefly, individual flies from each genotype were placed in a covered Petri dish (15 mm) and allowed to adapt to their environment for 5 min. The length of time adult flies were moving over a period of 3 min was then measured. Twenty five flies of each genotype were used for each experiment.
Life Span Determination-The life span of Mgat1 1 /Mgat1 1 adults was compared with Mgat1 1 /CyO-GFP and Mgat1 ϩ9 /Mgat1 ϩ9 flies. Briefly, an overnight egg collection was obtained from Mgat1 1 /CyO-GFP flies and aged for 24 -36 h. Upon hatching into first instar larvae, Mgat1 1 / Mgat1 1 homozygotes were sorted from Mgat1 1 /CyO-GFP heterozygotes by the absence (Mgat1 1 /Mgat1 1 ) or presence (Mgat1 1 /CyO, GFP) of a GFP marker. The sorted larvae were then transferred to vials containing standard medium and allowed to develop. Adult flies were removed as soon as they eclosed and were placed into fresh vials (10 flies/vial). The starting population size for each genotype was 100. The males and females were kept in separate vials. Dead flies were scored, and vials were changed every 3 days. For statistical analysis, the mean and maximum life span of each strain was calculated from the time in days when survival reached 50 and 10%, respectively, of the starting population in each of the 10 cohorts of each strain.

Generation of Mgat1
Mutants-As a first step to identify the in vivo function of Mgat1, we characterized the genomic structure of the Mgat1 gene from Drosophila. Unlike C. elegans, which contains three Mgat1 genes, the Drosophila genome contains a single Mgat1 gene that is contained within 3.2 kb of genomic DNA and is flanked at the 5Ј end by the gene l(2)05510 and at the 3Ј end by the gene CG13424 (43). The predicted intron/exon structure of Drosophila Mgat1 is illustrated in Fig. 3. As described previously (36), Mgat1 gives rise to a 2.8-kb cDNA predicted to encode a 458-amino acid protein with 52% amino acid sequence identity to human GlcNAcT1. To determine the functional requirement for Mgat1 in flies, we then generated a series of Mgat1 mutants by imprecise excision of a transposable P-element, KG02444, located within the first exon of Mgat1, 545 bp 5Ј of the start ATG ( Fig.  3). In total, we generated three independent deletions that removed various portions of the Mgat1 gene and failed to complement each other. More importantly, all of these mutants complemented the mutation in the gene located immediately 5Ј to Mgat1, l(2)05510, demonstrating that our excisions are specific to Mgat1. All of the Mgat1 complex (Gn2M3Gn2) N-glycans. The major structure in wild type flies is M3Gn2F 6 (box with a thick continuous line) (14,44); other structures present in large amounts (32-68% of M3Gn2F 6 ) are boxed with thin continuous lines. The remaining boxed structures (discontinuous lines) are present in amounts less than 10% of M3Gn2F 6 . Unboxed structures have not been detected by MS but are included in the figure on the basis of other evidence (44). GlcNAcTI adds GlcNAc in ␤1-2 linkage to the Man␣1-3 arm of M5Gn2 to form the hybrid N-glycan GnM5Gn2 (36,69). Two Man residues are removed from GnM5Gn2 by the action of ␣3,6-mannosidase II (MaseII) to form the truncated hybrid N-glycans GnM4Gn2 and GnM3Gn2 (6). A specific ␤-N-acetylglucosaminidase not found in vertebrates (Gnase (9, 49)) removes the GlcNAc added by GlcNAcTI to form M4Gn2 and M3Gn2 paucimannose N-glycans. GlcNAcTII (70) acts on GnM3Gn2 to initiate the synthesis of complex N-glycans; this is a minor pathway in plants, insects, and C. elegans because Gnase competes more effectively for substrate than GlcNAcTII. The substrates, products, and reactions of the core ␣1,6-FucT (6FucT) and ␣1,3-FucT (3FucT) are shown (44). Both core ␣1,6-FucT (FucT6) (44) and ␣1,3-FucT (FucTA) (45) in Drosophila are dependent on prior GlcNAcTI action. FucT6 cannot act on structures with a core ␣1-3-linked Fuc and must therefore act before FucTA to make the small amounts of M3Gn2F 3 F 6 in wild type Drosophila (14,44). GlcNAcTI-null flies make small amounts of fucosylated M3Gn2 (M3Gn2F X ) ( Table 1) indicating a GlcNAcTIindependent path to this structure; neither of the previously reported core FucTs are responsible because both enzymes require prior GlcNAcTI action (44). The site and linkage of the Fuc on M3Gn2F X are unknown. GlcNAcTI-null flies make relatively large amounts of M3Gn2 and M4Gn2 (Table 1) suggesting that a GlcNAcTI-independent ␣-mannosidase (Mase) acts on M5Gn2 upstream of GlcNAcTI; such a mannosidase has been reported in Spodoptera frugiperda (71) but not in Drosophila. B, N-glycan synthesis in Mgat1-null flies. The figure is based on MS analysis of the N-glycan structures in GlcNAcTI-null flies ( Table 1) and was obtained by removing all arrows dependent on the action of GlcNAcTI from the wild type fly scheme (see A). The definitions of names and arrows are as for A. The amounts of M5-M9Gn2 (boxes with continuous lines) are increased by 50 -250% in the null flies. M3Gn2F 6 , M3Gn2, and M4Gn2 (boxes with bold discontinuous lines) are reduced by 100, 59, and 19%, respectively. A small amount of M3Gn2F was observed in GlcNAcTI-null flies (Table 1); this M3Gn2F is neither M3Gn2F 3 nor M3Gn2F 6 (the respective FucTs require prior GlcNAcTI action) and has been designated as M3Gn2F X (not boxed). Structures boxed with thin discontinuous lines are present in low amounts in wild type flies but are absent in the mutant flies. The arrows attached to M4Gn2F X indicate that the structure may be an intermediate in the pathway to M3Gn2F X . C, N-glycan synthesis in fdl-null flies. The figure is based on MS analysis of the N-glycan structures in the fdl and Df(2R)achi 2 fly strains; fdl flies have a hypomorphic mutation in the fdl gene, and Df(2R)achi 2 flies have a null mutation in fdl but also have mutations in five other genes (49). The fdl gene was recently cloned and shown to encode the ␤-N-acetylglucosaminidase (Gnase) that removes GlcNAc incorporated by GlcNAcTI (49). The definitions of names and arrows are as for Fig. 2A. GnM3Gn2F 6 (box with bold continuous lines), present in small amounts in wild type flies, is the major N-glycan in fdl-null flies. The structures in boxes with thin continuous lines show moderate increases in fdl-null flies. M5Gn2 (box with thin discontinuous lines) is decreased by 54%, and M3Gn2 and M3Gn2F 6 (boxes with bold discontinuous lines) are decreased by 73 and 83%, respectively, in fdl-null flies. The small amount of fucosylated M3Gn2 (M3Gn2F X , unboxed) in fdl-null flies (49) cannot be due to either of the previously described core FucTs because both routes require Gnase action; suggested routes to M3Gn2F X are shown. The other unboxed structures are either products of Gnase (absent in fdl-null flies), or structures not detected by MS, or not reported by Lé onard et al. (49). mutants were homozygous viable, although the adults appeared sluggish (see below). To identify potential null mutants in Mgat1, we used PCR analysis to map the breakpoints and found one line, Mgat1 1 , that deleted most of the first and all of the second exon, including the translational start site. At the same time, we also identified a precise excision of the P-element, Mgat1 ϩ9 , which restored the Mgat1 locus and was subsequently utilized as a wild type, genetic control for all of our remaining experiments.
To confirm that Mgat1 1 was in fact a null allele, we then measured GlcNAcT1 activity from extracts derived from Mgat1 1 /Mgat1 1 flies and compared these to Mgat1 1 /CyO-GFP and the wild type precise excision line Mgat1 ϩ9 /Mgat1 ϩ9 . We found that homozygous mutant flies had no detectable GlcNAcT1 activity, whereas heterozygotes exhibited intermediate levels (105 Ϯ 29 pmol/h/mg protein) compared with the wild type controls (259 Ϯ 40 pmol/h/mg protein).
Comparison of N-Glycans from Wild Type and Mgat1 Mutant Flies-To determine whether the synthesis of oligomannose, hybrid, complex, and paucimannose N-glycans was affected in the Mgat1 1 /Mgat1 1 mutants, we examined the levels of N-glycans released by PNGase F using MALDI-TOF mass spectrometric analysis (40). The N-glycan patterns obtained from wild type (Mgat1 ϩ9 /Mgat1 ϩ9 ) ( Table 1 and (Table 1 and Fig. 4B) flies were compared. As described previously (11,13,14), the dominant N-glycan structures found in wild type adult flies were paucimannosidic (M3Gn2F 6 , M3Gn2, M4Gn2) and oligomannosidic (M5Gn2) N-glycans (Table 1 and Fig. 2A). In contrast, Mgat1 1 /Mgat1 1 flies showed a dra-matic decrease in the amount of M3Gn2F to almost undetectable levels ( Table 1 and Fig. 2B). The amount of M3Gn2F 3 in wild type flies is very small (14,44) indicating that almost all the decrease in M3Gn2F in mutant flies is because ofM3Gn2F 6 . Furthermore, because both core ␣1,  and core ␣1,6-FucT (44) require the prior action of GlcNAcTI, the very small MS peak for M3Gn2F seen in GlcNAcTI-null fly extracts (Table 1) is because of a different as yet uncharacterized FucT (Fig. 2B, product M3Gn2F X ). M4Gn2F, GnM3Gn2, and GnM3Gn2F were not detected in Mgat1 1 /Mgat1 1 flies, and M3Gn2 and M4Gn2 were moderately decreased ( Table 1). The mutant flies also showed a significant accumulation of M5Gn2 and small increases in M6Gn2, M7Gn2, M8Gn2, and M9Gn2 (Table 1); these structures are synthesized upstream of the GlcNAcTI block. The values for the heterozygous flies were intermediate between the wild type and mutant values (Table 1). M3Gn2F 3 F 6 (with both an ␣1-3and ␣1-6-linked Fuc residue on the same Asn-linked core GlcNAc), Gn2M3Gn2, and Gn2M3Gn2F ( Fig. 2A) have been reported in the extracts of flies (14) and in cultured Drosophila cells (45); however, these structures were not observed in the analysis shown in Table 1 probably because we used PNGase F to release N-glycans, and Fabini et al. (14) and Rendic et al. (45) used PNGase A. Biosynthetic pathways based on the structural analyses in Table 1 and the work of others (14,45) are shown in Fig. 2    a This peak has been identified by derivatization and MALDI-IT-TOF as a (Hex) 8  sterility, we examined the testes from Mgat1 1 /Mgat1 1 males and found that they produced normal levels of mobile sperm, indicating that the sterility was not because of defects in spermatogenesis (data not shown). We then determined whether Mgat1 1 /Mgat1 1 males were able to mate with and fertilize females. In wild type flies, females store sperm within structures called spermathecae after mating. Although we were able to detect mobile sperm within the spermathecae of females mated to Mgat1 ϩ9 /Mgat1 ϩ9 or Mgat1 1 /CyO-GFP males, we could not detect any sperm in females mated to Mgat1 1 /Mgat1 1 males (data not shown). This suggests that the sterility defect is associated with a failure of Mgat1 1 / Mgat1 1 males to mate.

Mgat1 1 /Mgat1 1 Mutants Exhibit Defects in Locomotory Activity and a Reduced
Life Span-Consistent with the observation that Mgat1 1 / Mgat1 1 males did not mate with wild type females, we also found that Mgat1 1 /Mgat1 1 adults appeared sluggish and slower in their movements. To quantify any potential locomotory defects, we measured the amount of time mutant and control flies spent moving during a 3-min period using an open field test. We found that Mgat1 1 /Mgat1 1 adults showed Ͼ95% reduction in movement compared with heterozygous and wild type controls (Table 2). Specifically, we found that In addition to the locomotory defects, Mgat1 1 /Mgat1 1 mutants also appeared to die earlier than wild type flies. Of note, mutant adults were recovered only when mutants were isolated as first instar larvae and allowed to develop at low density in the absence of wild type larvae. Under these conditions, approximately two-thirds of the mutants eclosed as pupae. The remaining third appeared to die throughout larval development. Mutant adults that emerged showed a normal external morphology but appeared to die earlier than their wild type or heterozygous counterparts.
To determine whether there was a significant difference in life span between homozygous mutants and controls, we determined the mean and maximum life span for both males and females (Table 3). We found that Mgat1 1 /Mgat1 1 males and females had a severely reduced life span compared with both Mgat1 1 /CyO-GFP and Mgat1 ϩ9 /Mgat1 ϩ9 controls. Specifically, Mgat1 1 /Mgat1 1 males had a mean life span of 12.8 Only one peak at m/z 1336.9 corresponds both to a fly food peak and to a potential N-glycan (Hex3HexNAc4; see Table 1). Derivatization and MALDI-IT-MS (40) showed this ion to be a (Hex) 8  . Taken together, these data clearly demonstrate that Mgat1 is required for locomotory activity and survivorship in flies.
The Central Nervous System of Maternal-null Mgat1 1 /Mgat 1 Embryos Does Not Bind Anti-HRP-Antibodies raised against the plant glycoprotein HRP have been used to specifically label Drosophila and C. elegans neurons. A specific ␣1,3-fucosyltransferase (␣1,3-FucT) that adds an ␣1-3-linked fucose to the proximal N-glycan core is essential for synthesis of the HRP epitope in both C. elegans (46) and Drosophila (14,45). In vitro studies of the ␣1,3-FucT (FUT-1) required for HRP epitope synthesis by C. elegans have shown that the enzyme does not require the prior action of GlcNAcTI (46). Furthermore, a C. elegans strain with null mutations in all three GlcNAcTI genes (37) displayed normal staining with anti-HRP in the complete absence of GlcNAcTI enzyme activity (46). In contrast, the Drosophila ␣1,3-FucT (FucTA) required for synthesis of the HRP epitope acts only on substrates that require the prior action of GlcNAcTI (14,45). However, we initially observed that Mgat1 1 /Mgat1 1 mutant and Mgat1 1 /CyO-GFP control embryos exhibit similar levels of staining with anti-HRP (data not shown). A possible explanation for this discrepancy is the presence of maternally derived Mgat1 mRNA in the Mgat1 1 /Mgat1 1 mutant embryos. Indeed, retention of maternally derived Mgat1 mRNA has been demonstrated in pre-implantation Mgat1 Ϫ/Ϫ mouse embryos (47). Consistent with this possibility, we found that unlike Mgat1 1 /CyO-GFP, Mgat1 ϩ9 /Mgat1 ϩ9 , or other wild type brains, Mgat1 1 /Mgat1 1 adult fly brains do not show anti-HRP staining (data not shown). To prove that the anti-HRP immunoreactivity observed in Mgat1 1 /Mgat1 1 embryos was because of maternal contribution, we examined embryos obtained from Mgat1 1 / Mgat1 1 females mated to Mgat1 1 /CyO-GFP males. The resulting Mgat1 1 /Mgat1 1 embryos (which lack any maternal contribution) were negative for anti-HRP staining, whereas sibling Mgat1 1 /CyO-GFP embryos (which also lack any maternal contribution) exhibited normal anti-HRP staining (Fig. 5). This finding suggests that maternally derived Mgat1 mRNA is responsible for the presence of the HRP epitope in null mutant embryos consistent with the observation that Drosophila FucTA requires prior GlcNAcT I action (14,45).
The major Drosophila N-glycan structure with two Fuc residues on the Asn-linked GlcNAc of the core is M3Gn2F 3 F 6 ( Fig. 2A), although it represents only about 0.4 -1.0% of the total N-glycans in wild type flies (14); however, a cultured Drosophila cell line has been reported with a 19% content of M3Gn2F 3 F 6 (45). M3Gn2F 3 F 6 is at least partly responsible for the staining of fly neurons with anti-HRP antibody. Mass spectrometric analysis shows a small amount of M3Gn2F in the adult Mgat1 1 /Mgat1 1 mutants (Figs. 2B and 4B), indicating the presence of GlcNAcTI-independent FucTs not involved in synthesis of the HRP epitope.
Mgat1 1 /Mgat1 1 Flies Exhibit a Fused Lobe Phenotype-Because Drosophila Mgat1 1 /Mgat1 1 mutants did not stain with anti-HRP, we could not use this marker to determine whether the locomotory defects

/Mgat1 1 mutants exhibit a severely reduced life span
Adult flies of each genotype were maintained at 25°C in shell vials (10 flies per vial) containing standard cornmeal agar medium. The starting population size for each genotype was 100. Flies were transferred to fresh medium and scored for survivorship every 3 days. The mean (50% survival) and maximum (10% survival) life span for each genotype is shown in days. observed in our mutants were because of defects in adult brain structures. To circumvent this problem, we therefore immunostained whole mount brains from control Mgat1 ϩ9 /Mgat1 ϩ9 flies and Mgat1 1 /Mgat1 1 mutants with anti-FasII, which labels a subset of axons (including the mushroom bodies) within the central nervous system. Although we did not observe any gross morphological defects using either light microscopy or FasII staining, we did find that Mgat1 1 /Mgat1 1 mutants exhibit a fused lobe phenotype (Fig. 6). Specifically, we observed Ͼ50% reduction in the separation of the ␤ lobes in all of the brains that we examined, with full fusion observed in 40% of the samples. This phenotype is similar to that observed in fused lobe (fdl) mutant flies (48). fdl mutants were first identified in an enhancer trap screen for genes that are expressed during late larval development in structures that will give rise to the central complex of the adult brain. In the case of fdl, insertion of the enhancer trap element also results in mildly penetrant defects in the adult brain consisting of fused ␤ lobes in the mushroom body (48).
Interestingly, the fdl gene is highly homologous to hexosaminidase genes in other species suggesting that it may encode the ␤-N-acetylglucosaminidase ( Fig. 2A, Gnase) that removes the GlcNAcTI-dependent GlcNAc residue to form paucimannose N-glycans in Drosophila. Indeed it has been reported recently (49) that the fdl gene encodes a hexosaminidase with the same substrate specificity as Gnase. Léonard et al. (49) studied two Fdl-deficient Drosophila lines. The fdl fly has a hypomorphic mutation in the fdl gene and shows the fused lobe phenotype; no other genes are mutated in this fly. Df(2R)achi 2 flies have a null mutation in the fdl gene, but there are five other mutant genes in this fly strain; however, none of these genes appear to be involved in glycan metabolism (49).
Comparison of MS analyses of wild type flies (Table 1 and Fig. 2A), Mgat1 1 /Mgat1 1 flies (Table 1 and Fig. 2B), and Df(2R)achi 2 flies (Fig. 2C) (49) shows that M3Gn2F 6 , the major N-glycan in wild type flies, is decreased by 83% in Df(2R)achi 2 flies and by over 98% in Mgat1 1 /Mgat1 flies relative to wild type flies. The only other structure that is significantly decreased in both mutant flies relative to wild type (73 and 59%, respectively) is M3Gn2, the precursor of M3Gn2F 6 . Analysis of Mgat1 1 /Mgat1 1 flies shows that M4Gn2 is decreased by 19%, and M4Gn2F (a very minor component of wild type flies) is not detected in the Mgat1 null flies; no data were reported for these two structures in the fdl-deficient lines (49). M6Gn2, M7Gn2, M8Gn2, and M9Gn2 are increased by 50 -250% in both Mgat1 1 /Mgat1 1 and Df(2R)achi 2 flies relative to wild type flies (Fig.  2C). M5Gn2 is increased by 47% in Mgat1 null and decreased by 54% in fdl-null flies. GnM3Gn2F 6 is present in low amounts in wild type flies and absent in Mgat1 1 -null flies but is the major N-glycan in fdl-null flies (Fig. 2C). Wild type, Mgat1 null, and fdl-null flies make little or no GnM3Gn2, Gn2M3Gn2, and Gn2M3Gn2F. It is concluded that the absence of M3Gn2 and M3Gn2F 6 plays a major role in generating the fused lobe phenotype because these glycans are significantly decreased in both Mgat1 null and fdl-null flies; a role for other glycans, however, cannot be ruled out on the basis of the above analysis.

DISCUSSION
Formation of the sugar-amino acid linkage occurs throughout the entire phylogenetic spectrum (archaea, eubacteria, and eukaryotes) and involves 13 monosaccharides, 8 amino acids, and at least 41 linkages (50). Three N-glycan and 20 O-glycan linkages have been reported in eukaryotes (50). It has been estimated that about 0.5-1.0% of the translated mammalian genome participates in oligosaccharide production and function (51). Protein-bound glycans have many functions (32,33,52), e.g. cell adhesion, control of the immune system, embryonic development and differentiation, and have been implicated in diseases such as metastatic cancer.
N-Glycans occur primarily on secreted and membrane-bound proteins. Oligomannose N-glycans (Fig. 1, Man 5-9 GlcNAc 2 ), found in both unicellular and multicellular eukaryotes, are ancient structures essential for the viability of all cells. GlcNAcTI-dependent N-glycans (Fig. 2) appeared in evolution at about the same time as multicellular organisms and are essential for normal mouse (30,31) and human (53,54) embryonic development consistent with a major function for these N-glycans in cell-cell and cell-environment interactions. Over 20 genes encoding enzymes involved in N-glycosylation have been inactivated by null mutations in mice (32)(33)(34). Eighteen genes encoding enzymes involved in N-glycosylation have been implicated as causes of congenital disorders of glycosylation (CDG) in humans (34,55); the CDGs are a family of genetic multisystemic disorders with severe nervous system involvement. Both mice (56) and humans (53,54) have been reported with null mutations in Mgat2, the gene encoding GlcNAcTII downstream of GlcNAcTI ( Fig. 2A). Human Mgat2 deficiency is named CDG-IIa and is characterized by severe psychomotor retardation.
Although the GlcNAcTI-dependent structures in vertebrates are complex N-glycans with antennary extensions of the Man␣1-6(Man␣1-3)Man␤1-4GlcNAc␤1-4GlcNAc core (Fig. 2), plants, insects, and C. elegans synthesize predominantly GlcNAcTI-dependent Man 3-4 GlcNAc 2 paucimannose N-glycans (Fig. 2) instead of complex N-glycans. Although the paucimannose N-glycans in Drosophila are modified by fucosylation (11,14,44), these structures lack the antennary branches typical of vertebrate complex N-glycans. We have determined the phenotypic effects of a null mutation in Mgat1 in flies in the expectation that the functions of GlcNAcTI-dependent N-glycans can be more readily analyzed in this organism than in vertebrates. The Mgat1 1 /Mgat1 1 adult flies are null mutants because extracts showed no GlcNAcTI enzyme activity, and mass spectrometric analysis showed dramatic changes in N-glycans compatible with lack of GlcNAcTI enzyme activity. Mutant adults were recovered only when animals were removed from the vial at the larval stage and allowed to develop at low density. Under these conditions, mutant pupae eclosed normally, and adults showed a normal external morphology but had a significantly reduced life span. Mutant flies grown under sterile conditions showed the same marked reduction in life span indicating that infection by pathogenic microorganisms was not responsible for the reduced life span (data not shown). Mgat1 1 /Mgat1 1 mutants were also significantly more sluggish than wild type flies. Moreover, this defect in locomotion is likely responsible for the observed male sterile phenotype because mutant males produced motile sperm but were unable to mate with either mutant or wild type females. Although we did not observe any gross morphological defects in the brains of adult mutants that could account for the locomotory phenotype, we did find that the brains of adult mutant flies did not react with an antibody to HRP. Furthermore, microscopic examination of the FIGURE 6. Mgat1 1 /Mgat1 1 mutants exhibit a fused lobe phenotype. Whole mount brains from Mgat1 ϩ9 /Mgat1 ϩ9 (A) and Mgat1 1 /Mgat1 1 (B) adults were immunolabeled with anti-FasII antibodies to detect mushroom body lobes. Only the Mgat1 1 /Mgat1 1 brains exhibited a fused ␤ lobe phenotype. The arrow points to a gap between the mushroom body ␤ lobes that is present in wild type and Mgat1 1 /CyO-GFP brains but reduced by Ͼ50% in all Mgat1 1 /Mgat1 1 brains and is completely absent in 40% of mutant brains. mutant brains showed the presence of fused ␤ lobes within structures called mushroom bodies. Interestingly, many studies have shown that mushroom bodies are required for learning and memory in Drosophila. At present, the severe locomotory defects observed in Mgat1 1 /Mgat1 1 mutants preclude us from determining whether the fused lobe phenotype would result in defects in learning and memory. Nonetheless, our data clearly indicate that GlcNAcTI-dependent N-glycans are required for normal development of the nervous system of the fly. Future studies, involving the identification of additional alleles or the ability to rescue the locomotory defects, may allow us to determine whether N-glycans are also required for higher processes such as those involved in learning and memory.
Mammalian cells in culture suffer no obvious phenotypic abnormalities when deprived of GlcNAcTI (29), which is consistent with the hypothesis that these cells do not require the kind of cell-cell interactions provided by vertebrate N-glycans. Mgat1 Ϫ/Ϫ C. elegans are viable and have an apparently normal phenotype when grown under standard laboratory conditions but show altered survival times when exposed to pathogenic bacteria (35), suggesting that worms developed GlcNAcTIdependent paucimannose N-glycans to cope with a hostile bacterial environment. This study shows that a null mutation of Mgat1 in Drosophila resulted in severe developmental abnormalities and supports the hypothesis that the fly uses GlcNAcTI-dependent N-glycans to mediate some cell-cell interactions during development. Mgat1 Ϫ/Ϫ mice die in utero at embryonic day 9.5 (30, 31) indicating a need for GlcNAcTI-dependent complex N-glycans during development. This graded change in importance and functions of GlcNAcT I during evolution is probably related to the differences in the types of N-glycans synthesized downstream of GlcNAcT I by the various organisms.
The dystrophin glycoprotein complex is an assembly of proteins spanning the sarcolemma of vertebrate skeletal muscle cells. An O-mannosyl glycan (sialyl␣2-3Gal␤1-4GlcNAc␤1-2Man␣1-O-Ser/Thr) (57, 58) on ␣-dystroglycan, one of the components of the dystrophin glycoprotein complex, has been identified as a receptor for laminin and other extracellular ligands. Defects in this O-mannosyl glycan have been associated with a distinct group of autosomal recessive congenital muscular dystrophies (59,60). Protein O-mannosyl ␤1,2-N-acetylglucosaminyltransferase 1 (POMGlcNAcT1), a homologue of GlcNAcTI, catalyzes the synthesis of the GlcNAc␤1-2Man␣1-O-Ser/Thr moiety on ␣-dystroglycan. POMGlcNAcT1 is deficient in patients with muscle-eyebrain disease, a congenital muscular dystrophy (61,62). Drosophila has functional genes encoding homologues of ␣-dystroglycan and the two protein ␤-O-mannosyltransferases (POMT1 and POMT2) that incorporate O-Man residues into ␣-dystroglycan (63). However, BLAST analysis of the Drosophila genome with either the vertebrate Mgat1 or POMGlcNAcT1 nucleotide sequence identifies Mgat1 as the only homologous gene (36,63). The enzyme encoded by Drosophila Mgat1 is unable to catalyze the transfer of GlcNAc to the Man␣-O-Ser/Thr moiety on ␣-dystroglycan (63). It can therefore be concluded that the phenotype observed in the Mgat1 1 /Mgat1 1 null flies is because of abnormal N-glycan structures and is not related to ␣-dystroglycan function.
It has been estimated that as many as 50% of all proteins in humans and mice may be N-glycosylated (66). It is probable that many proteins are also N-glycosylated in Drosophila. If a mutant animal with a defect in the N-glycosylation pathway shows an abnormality, it is essential to identify the protein or proteins targeted by the mutation. Although this is usually a difficult task, it has been achieved in some cases (67). The phenotype of Drosophila Mgat1 1 /Mgat1 1 mutants is probably due to several different GlcNAcTI targets. However, the absence of the HRP epitope in mutant neurons indicates that the protein or proteins that carry the ␣1-3-fucosylated N-glycan associated with the HRP epitope are among these targets. Polyclonal anti-HRP antibodies have been used to purify Drosophila proteins carrying the HRP epitope (68). Neurotactin, fasciclins I and II, neuroglian, and three receptor protein-tyrosine phosphatases were identified, and many other bands were seen on Western blotting with anti-HRP antibody (68). These proteins can be used to study the roles of N-glycans in protein function in Drosophila.