The Drosophila melanogaster brainiac Protein Is a Glycolipid-specific β1,3N-Acetylglucosaminyltransferase *

Mutations at the Drosophila melanogaster brainiac locus lead to defective formation of the follicular epithelium during oogenesis and to neural hyperplasia. Thebrainiac gene encodes a type II transmembrane protein structurally similar to mammalian β1,3-glycosyltransferases. We have cloned the brainiac gene from D. melanogastergenomic DNA and expressed it as a FLAG-tagged recombinant protein in Sf9 insect cells. Glycosyltransferase assays showed thatbrainiac is capable of transferringN-acetylglucosamine (GlcNAc) to β-linked mannose (Man), with a marked preference for the disaccharide Man(β1,4)Glc, the core of arthro-series glycolipids. The activity of brainiactoward arthro-series glycolipids was confirmed by showing that the enzyme efficiently utilized glycolipids from insects as acceptors whereas it did not with glycolipids from mammalian cells. Methylation analysis of the brainiac reaction product revealed a β1,3 linkage between GlcNAc and Man, proving thatbrainiac is a β1,3GlcNAc-transferase. Human β1,3GlcNAc-transferases structurally related tobrainiac were unable to transfer GlcNAc to Man(β1,4)Glc-based acceptor substrates and failed to rescue a homozygous lethal brainiac allele, indicating that these proteins are paralogous and not orthologous tobrainiac.

The importance of glycoconjugates in regulating developmental processes is continually being supported by studies performed in various model organisms like Caenorhabditis elegans (1), Drosophila melanogaster (2), and the mouse (3). The Drosophila genes sugarless, sulfateless, pipe, tout-velu, and dally participate in the formation of proteoglycans. Loss of function mutations in some of these genes produce polarity phenotypes mechanistically connected to incorrect diffusion of the signaling proteins wingless and hedgehog (4 -6). The rotated abdomen locus, whose disruption is associated with a helical rotation of the body, has been found to encode a potential O-mannosyltransferase (7), and fringe, which modulates the interaction of the Notch receptor with its ligands (8), has recently been demonstrated to be a ␤1,3N-acetylglucosaminyltransferase (GlcNAcT) 1 (9,10).
The Drosophila gene brainiac (brn) (11) encodes a protein that shares structural motifs with ␤1,3glycosyltransferases (12,13). The brn gene is localized on the X chromosome. brn was shown to cooperate with the epidermal growth factor receptor and one of its ligands, the Drosophila TGF␣ homologue gurken (11) during oogenesis. Mutant brn alleles exhibit altered morphology of the follicular epithelium (11), female sterility (14), and germ line loss (15). Furthermore, brn embryos develop neural hyperplasia and epidermal hypoplasia (11) as encountered with Notch hypomorphic alleles and other neurogenic mutants, suggesting implications of brn in Notch signaling (16,17).
While the relationships between brn and specific signaling pathways have been examined genetically, the nature of these interactions remained elusive as long as the biochemical function of brn was unclear. In the present study, we show that brn has a ␤1,3N-acetylglucosaminyltransferase (GlcNAcT) activity directed toward the Man(␤1,4)Glc core structure of arthroseries glycolipids.

EXPERIMENTAL PROCEDURES
Cloning and Expression-The brn gene was amplified by PCR from D. melanogaster OregonR genomic DNA during 30 cycles at 95°C for 45 s, 55°C for 30 s, 72°C for 60 s using the primers 5Ј-TTTGGATCC-GTCGCCATGCAAAGT-3Ј and 5Ј-CCTGTTCTAGATGCTACGCGTAA-T-3Ј. The resulting 1.0-kb fragment was digested with BamHI and XbaI and subcloned into the pFastbac-FLAG(a) vector (Invitrogen) linearized at the BamHI and XbaI sites. The FLAG-tagged brn gene was expressed as a recombinant baculovirus in insect cells as described previously (13). Infected cells (10 7 ) were lysed at 72 h post-infection in 600 l of 50 mM Tris/HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, and a protein inhibitor mixture (complete, EDTA free, Roche Diagnostics) on ice. Post-nuclear supernatants were incubated with 240 l of anti-FLAG M2-agarose beads (Sigma) under rotation for 2.5 h at 4°C. Beads were washed three times with Tris-buffered saline and used as enzyme source for assays.
Glycolipid Extraction-D. melanogaster Schneider 2 cells, Spodoptera frugiperda Sf9 cells, and human colon carcinoma Caco-2 cells were washed three times in phosphate-buffered saline and extracted in isopropanol:hexane:H 2 O (55: 25:20). Extracts were spun twice at 500 x g, and supernatants were dried under N 2 . Phospholipids were removed by saponification in 0.2 M NaOH in methanol for 24 h at 37°C. After neutralization with HCl, the extracts were expanded to theoretical upper phase (methanol:water:chloroform, 47:48:3), applied on C 18 Sep-Pak cartridges, and eluted with 5 ml of methanol. Eluates were dried under N 2 and resuspended in 500 l of methanol. The procedure yielded about 120 g of mannose equivalents for 10 8 S2 and Sf9 cells and 20 g of mannose equivalents from 10 7 Caco-2 cells as determined by the phenol sulfuric acid assay (19).
Structural Analysis-A mixture of substrate and of 10 nmol of product was separated by reversed phase HPLC on a 3 ϫ 250 mm column filled with 5 M ODS Hypersil (Shandon) at a flow rate of 0.6 ml/min. The column was eluted with a linear gradient from 6 to 24% of methanol during 18 min in 0.1 M ammonium acetate, pH 4.0. p-Nitrophenylglycosides were monitored at 245 nm. The mixture was also analyzed after incubation with N-acetyl-␤-hexosaminidase from jack beans (Sigma) (25). The fraction of interest was collected in a screw capped glass vial and dried in a speed-vac concentrator. A small aliquot was used for matrix assisted laser desorption mass spectrometry as described elsewhere (25). The sample was dried over phosphorus pentoxide in vacuo and permethylated using NaOH (26). Partially permethylated alditol acetates were prepared using NaBD 4 as the reducing agent and analyzed by gas chromatography/mass spectrometry using a 60 m SP2330 (Restek) (27) and a Finnigan Ion Trap ITD800. Derivatives of terminal and 3-substituted galactose served to compare retention times with the data given by Doares et al. (27).

RESULTS
We have cloned the D. melanogaster brn gene by PCR amplification and expressed it as an N-terminally FLAG-tagged full-length protein in Sf9 insect cells. The recombinant brn protein was bound to anti-FLAG-agarose beads, and cellular contaminants such as possible endogenous acceptor substrates were washed out before assaying for enzymatic activity. A GlcNAcT activity was only detected toward the Man(␤1-OpNP) acceptor when monosaccharide substrates were assayed (Table  I). Highest activity was measured toward the disaccharide acceptor Man(␤1,4)Glc(␤1-OpNP), whereas a slight activity was also detected toward Gal(␤1,4)Glc(␤1-OpNP) ( Table I). The Man(␤1,4)Glc structure represents the core of arthro-series glycolipids found in nematodes (28) and insects (29) among others.
In Drosophila, the arthro-series Man(␤1,4)Glc core is elongated with a ␤1,3-linked GlcNAc (30), suggesting that brn may represent the enzyme catalyzing this step. To test this hypothesis, we have isolated neutral glycolipids from Drosophila S2 and Spodoptera Sf9 cells and assayed these glycolipids as acceptors for the anti-FLAG beads-bound brn enzyme. A signifi-cant GlcNAc-transferase activity was detected when incubating brn together with insect glycolipids, whereas only a low activity was measured with glycolipids extracted from mammalian Caco-2 cells, likely reflecting the low specificity of brn for lactosylceramide. The reaction products were separated by TLC and plates were autoradiographed, revealing a [ 14 C]Glc-NAc-labeled band at the size of a trihexoside ceramide in S2 and Sf9 cells (Fig. 1).
The nature of the linkage between GlcNAc and the underlying ␤-linked Man residue was investigated by methylation analysis of the brn reaction product GlcNAc-Man(␤1-OpNP). In reversed phase HPLC, the presumed disaccharide product eluted slightly ahead of the substrate Man(␤1-OpNP). The disaccharide peak disappeared upon incubation with N-acetyl-␤-hexosaminidase ( Fig. 2A). The purified fraction corresponding to the disaccharide peak exhibited a pseudomolecular ion of m/z 513.5. Linkage analysis of the GlcNAc-Man(␤1-OpNP) disaccharide product gave a peak at the relative retention time of 0.597, which suggests a 2-or a 3-substituted mannosyl residue (27). The fragment spectrum clearly identified the derivative as substituted in the 3-position (Fig. 2B), thus confirming the identity of brn as a ␤1,3 GlcNAcT.
The human ␤3GnT transgenes and a brn transgene were expressed in flies carrying the allele brn 1.6P6 , which causes lethality at the late pupal stage. The transgenes were expressed ubiquitously using armadillo GAL4 transactivator lines. The brn transgene did rescue brn 1.6P6 mutant males from their hemizygous late pupal lethality, whereas the human ␤3GnT transgenes did not (Table II). The rescue of brn 1.6P6 males was confirmed by detection of the forked marker, whose gene is located besides the brn 1.6P6 allele on the X chromosome. Control crosses of females carrying brn 1.6P6 with yellow white males did not yield any living brn 1.6P6 forked/Y males either. The inability of human ␤3GnT enzymes to compensate for the loss of brn activity in mutant flies suggested that the former enzymes cannot elongate the arthro-series glycolipid core in vivo. This was confirmed in vitro by showing that the human ␤3GnT enzymes did not exhibit significant activity toward the Man(␤1,4)Glc(␤1-OpNP) acceptor (Table II). DISCUSSION We have shown that Drosophila brn, a member of the ␤1,3 glycosyltransferase family, encodes a ␤1,3 GlcNAcT enzyme with a specificity for the Man(␤1,4)Glc disaccharide found in arthro-series glycolipids (29). Several mammalian enzymes structurally related to brn have been suggested to represent homologues (35)(36)(37). However, the specificity of brn for Man(␤1,4)Glc, a disaccharide that has never been described in vertebrates, rather indicates that brn and mammalian ␤1,3 glycosyltransferases are paralogous proteins derived from a common ancestor gene.
The functional disparity between the ␤1,3 GlcNAcT brn and mammalian ␤1,3 GlcNAcT enzymes is further supported by the inability of the latter to complement the lethal phenotype of the mutant allele brn 1.6P6 in Drosophila. The specificity of brn toward Man␤1,4Glc-Cer suggests the presence of functional homologues only in organisms harboring arthro-series glycolipids, whose core structure is GlcNAc(␤1,3)Man(␤1,4)Glc-Cer. A protein structurally related to brn has recently been described in C. elegans (38), which express arthro-series glycolipids (28). The loss of that gene, named bre-5 (39), renders the animal resistant to high doses of Bacillus thuringiensis Bt toxin. Since Bt toxin binds to arthro-series glycolipids (40), it is possible that bre-5 participates in the formation of this class of glycolipids in C. elegans and thereby represents a true orthologue of brn.
brn mutations affect follicle cell-germ line interactions and lead to neurogenic phenotypes in Drosophila embryos. Considering the involvement of brn in glycolipid biosynthesis, one can envision that arthro-series glycolipids may regulate cell adhesion, proliferation, and differentiation via carbohydrate-lectin interactions. On the other hand, arthro-series glycolipids may modulate specific signaling proteins in a way similar to gangliosides affecting the epidermal growth factor receptor (41,42), insulin receptor (43), and platelet-derived growth factor receptor (44) signaling cascades. The notion that brn glycolipid products interact with adhesion or signaling proteins implies The mixture was prepurified over Sep-Pak C 18 cartridges and subjected to reversed phase chromatography (trace a). A small product peak (P) eluted ahead of the substrate (S). The product disappeared upon incubation with ␤-N-acetylhexosaminidase (trace b) which indicates it to contain a ␤-linked GlcNAc residue. B, methylation analysis of p-nitrophenyl disaccharide. The electron impact mass spectrum of the partially methylated monodeuterated alditol acetate derived from the mannosyl residue of the disaccharide product shows several fragments indicative of a substitution in position 3 as depicted by the insert. Especially the presence of mass 118 and the absence of mass 190 exclude a 2-substitution, which could not be ruled out from the retention time alone (27).  that other mutant genes with phenotypes similar to those encountered in brn mutant flies may encode partner lectin/ signaling proteins. Along this line, Drosophila egghead mutants have similar and non-additive phenotypes to brn (17). Experiments aimed at characterizing the biochemical and functional relation between brn products and the egghead protein should reveal the mechanisms how arthro-series glycolipids regulate morphogenic events during Drosophila development.