Egghead and brainiac are essential for glycosphingolipid biosynthesis in vivo.

The Drosophila genes, brainiac and egghead, encode glycosyltransferases predicted to act sequentially in early steps of glycosphingolipid biosynthesis, and both genes are required for development in Drosophila. egghead encodes a beta4-mannosyltransferase, and brainiac encodes a beta3-N-acetylglucosaminyltransferase predicted by in vitro analysis to control synthesis of the glycosphingolipid core structure, GlcNAcbeta1-3Manbeta1-4Glcbeta1-Cer, found widely in invertebrates but not vertebrates. In this report we present direct in vivo evidence for this hypothesis. egghead and brainiac mutants lack elongated glycosphingolipids and exhibit accumulation of the truncated precursor glycosphingolipids. Furthermore, we demonstrate that despite fundamental differences in the core structure of mammalian and Drosophila glycosphingolipids, the Drosophila egghead mutant can be rescued by introduction of the mammalian lactosylceramide glycosphingolipid biosynthetic pathway (Galbeta1-4Glcbeta1-Cer) using a human beta4-galactosyltransferase (beta4Gal-T6) transgene. Conversely, introduction of egghead in vertebrate cells (Chinese hamster ovary) resulted in near complete blockage of biosynthesis of glycosphingolipids and accumulation of Manbeta1-4Glcbeta1-Cer. The study demonstrates that glycosphingolipids are essential for development of complex organisms and suggests that the function of the Drosophila glycosphingolipids in development does not depend on the core structure.

Invertebrates, Caenorhabditis elegans and Drosophila melanogaster, have recently attracted considerable attention as model organisms for deciphering specific biological roles of complex carbohydrates. One elegant example of this was a number of studies leading to the identification of a series of glycosylation genes critical for vulval invagination in C. elegans, which were all shown to affect a common biosynthetic pathway for the assembly of the O-linked oligosaccharide linker region common for all proteoglycans (1). Another example was the role of the O-linked fucose glycosylation pathway on the Notch receptor function (2). The Drosophila neurogenic genes brainiac and egghead encode glycosyltransferases essential for epithelial development during oogenesis and in the embryo (3,4). egghead and brainiac mutants display similar, non-additive defects, which has led to the proposal that they act in the same pathway (3). In previous reports we demonstrated that brainiac encodes a UDP-N-acetylglucosamine: ␤Man ␤1,3-N-acetylglucosaminyltransferase (␤3GlcNAc-transferase), and egghead encodes a GDP-mannose:␤Glc ␤1,4mannosyltransferase, with putative functions in sequential steps in the biosynthesis of the core structure of arthro-series glycosphingolipids (GlcNAc␤1-3Man␤1-4Glc␤1-Cer) as predicted by in vitro analysis ( Fig. 1) (5)(6)(7). Loss of either gene is predicted to abrogate glycosphingolipid biosynthesis at the dior monosaccharide-ceramide step.
Insect, nematode, and vertebrate glycosphingolipids share a common element consisting of Glc␤1-ceramide, after which they differ markedly in structure and complexity (Fig. 1A). Insect and nematode glycosphingolipids are built on Man␤1-4Glc␤1-ceramide (MacCer) 1 predicted to be catalyzed by Egghead, while vertebr␣te glycosphingolipids are built on Gal␤1-4Glc␤1-ceramide (LacCer) catalyzed by the ␤4galactosyltransferases, ␤4Gal-T5 and -T6 (8,9). Despite considerable differences in overall structures of glycosphingolipids among insects and vertebrates, it is clear that homologous glycosyltransferase genes conserved throughout evolution catalyze most biosynthetic steps. Egghead is perhaps the only exception suggesting that MacCer-based glycosphingolipids represent a specific functional basis for the diversification of the underlying biosynthetic pathways. Importantly, vertebrate glycosphingolipids based on the LacCer core diverge at the third biosynthetic step to form different classes of structures (Fig. 1A), which are differentially expressed in cells and are differentially expressed during development and differentiation (10). The vertebrate glycosphingolipid lacto-series is initiated by addition of ␤1,3GlcNAc to LacCer by brainiac orthologs designated ␤3GnTs (11)(12)(13)(14)(15). Interestingly, Drosophila brainiac functions both on the invertebrate and vertebrate precursor substrate MacCer and LacCer, while the vertebrate orthologs appear to only act on LacCer (5).
In this report we present direct evidence that Egghead and Brainiac do function in vivo in the glycosphingolipid pathway and are essential for glycosphingolipid biosynthesis in vivo. Furthermore, we demonstrate that despite the fundamental difference in the structure of core glycosphingolipid, the Drosophila egghead mutant can be rescued by introduction of the corresponding enzyme from the human glycosphingolipid biosynthetic pathway. In contrast the fly glycosphingolipid biosynthetic pathway is not elongated in vertebrate cells. The results show that glycosphingolipids are essential for development of complex organisms and suggest that the function of Drosophila glycosphingolipids in development does not depend on the core structure.
Extraction of Glycosphingolipids from Mature Flies, Larvae, and CHO Cells-glycosphingolipids were extracted and fractionated by methods similar to those described previously (16). Freeze-dried flies (ϳ5-10 g) were homogenized (Polytron, Kinematica AG, Luzern, Switzerland) in ϳ5 volumes of solvent A (2-isopropanol/hexane/water, 55:25:20, v/v/v, upper phase discarded), centrifuged, and the supernatant removed; this step was repeated with ϳ5 volumes of solvent B (chloroform/methanol, 1:1, v/v) and then sequentially again with ϳ5 volumes each of solvent A and solvent B. The four solvent extracts were combined, dried in a rotary evaporator, and treated with 10 ml of methanol/water/1-butanol (4:3:1, v/v/v) containing 25-30% methylamine at 55°C for 4 h (flask tightly stoppered), with occasional agitation and sonication. The reagents were again reduced to dryness by rotary evaporation, resuspended in a minimal volume of solvent C (chloroform/ methanol/water, 30:60:8, v/v/v), and applied to a column of DEAE-Sephadex A-25 (Ac-form). Neutral glycosphingolipids were eluted with 5 volumes of solvent C. Acidic glycosphingolipids were eluted with 5 volumes of 0.8 M sodium acetate in methanol, which was reduced to dryness, subjected to exhaustive dialysis (3,500 Da cutoff) against deionized water to remove salts, and recovered by lyophilization prior to HPTLC analysis.
CHO cells (ϳ1-2 ml packed volume) were subjected to a similar protocol, except that extensive sonication (20 -30 min per step) was used instead of Polytron homogenization to break up cells during extraction, and the order was solvent B (1ϫ), solvent A (2ϫ), and solvent B (1ϫ).
Frozen larvae (50 -150 mg) were subjected to a similar but truncated micro-scale extraction procedure with the following differences: (i) solvent volumes were 3 ml each, and the order was solvent B (1ϫ), solvent A (2ϫ), and solvent B (1ϫ); (ii) larvae were macerated with a glass rod and sonicated for 20 min at each step; (iii) 20 -25% methylamine reagent volume was 1 ml; (iv) drying steps were carried under N 2 stream at 35-40°C; (v) the DEAE-Sephadex anion exchange fractionation was omitted, but instead total lipids were subjected to a solid-phase extraction (SPE) cleanup step to remove as much non-lipid carbohydrate material as possible. SPE was carried out on 0.5-g octadecyl-silica cartridges (Honeywell/Burdick & Jackson, Muskegon, MI), applying lipids sonicated thoroughly in 0.5 N NaCl (1 ml). The pass-through was collected and re-applied two times and the SPE cartridge then washed sequentially with 0.5 N NaCl (2 ml) and deionized water (4 ml). Lipids were eluted with methanol (2 ml), followed by solvent B (2 ml), combining both and drying under N 2 stream prior to HPTLC analysis.
High Performance Thin Layer Chromatography-Analytical HPTLC was performed on silica gel 60 plates (E. Merck, Darmstadt, Germany) using chloroform/methanol/water (60:35:8, v/v/v; solvent D) as mobile phase for neutral or total lipids and chloroform/methanol/water (50:47: 14, v/v/v, containing 0.038% (w/v) CaCl 2 ; solvent E) as mobile phase for acidic lipids. Detection was made by Bial's orcinol reagent (0.55% orcinol (w/v) and 5.5% H 2 SO 4 (v/v) in ethanol/water 9:1 (v/v); the plate is sprayed and heated briefly to ϳ200 -250°C). Preparative HPTLC was carried out on neutral lipids using solvent D as mobile phase, streaking crude fraction lengthwise on 10 ϫ 20-cm plates; separated glycosphingolipid bands were visualized under UV after spraying with primulin (Aldrich; 0.01% in 80% aqueous acetone). Bands were marked by pencil and individually scraped from the plate. Glycosphingolipids were then isolated from the silica gel by repeated sonication in solvent B followed by centrifugation. Following concentration of the extract, primuline was removed by passage through a short column of DEAE-Sephadex A-25 in solvent C, which was removed under N 2 stream prior to analysis by NMR spectroscopy.
1 H NMR Spectroscopic Analysis of Glycosphingolipids-Individual glycosphingolipids isolated by preparative HPTLC were deuteriumexchanged by repeated addition of CD 3 OD, sonication, and evaporation under nitrogen, then dissolved in 0.5 ml of Me 2 SO-d 6 , 2% D 2 O (0.03% tetramethylsilane as internal chemical shift reference) for NMR analysis (17). One-dimensional 1 H NMR spectra were acquired on a Varian Inova 500 MHz spectrometer at 35°C. Spectra were interpreted by comparison to those of authentic standards and published data (5,17,18).
Generation of Monoclonal Antibody Recognizing Man␤1-4Glc␤1-Cer-For production of the anti-MacCer monoclonal antibody BALB/c mice were immunized three times with 10 g of purified MacCer isolated from High Five TM cells as described (5). Hybridomas were selected by immunocytology on air-dried, acetone-fixed CHO cells stably transfected with full-length egghead as well as by ability to differentially recognize MacCer in enzyme-linked immunosorbent assay (20).
Stable Expression of Egghead in Chinese Hamster Ovary Cells-The 1.37-kb egghead-Myc-full fragment used for baculo constructs was cloned into the BamHI/XbaI sites of pcDNA3(ϩ)Zeocin. Chinese hamster ovary (CHO-K1) cells were stably transfected with the pcDNA3-egghead-Myc-full as described previously and clones selected with anti-Myc antibodies (Invitrogen) (6). Two rounds of screening and cloning were performed by limiting dilution cloning using immunoreactivity with anti-Myc monoclonal antibody.
Immunolabeling-CHO cells were grown to subconfluence in the appropriate media as recommended by American Type Culture Collection and fixed in 3% paraformaldehyde, permeabilized with 0.1% Triton X-100, and incubated with undiluted anti-MacCer hybridoma supernatants for 18 h at 4°C and detected with fluorescein isothiocyanateconjugated rabbit anti-mouse immunoglobulin (F261, Dako). Immunostaining with soluble mannan binding lectin (MBL) was performed on non-fixed cells using purified MBL from human serum detected with an anti-MBL monoclonal antibody (generous gift from P. Garred P, Copenhagen University Hospital, Copenhagen, Denmark) and fluores-cein isothiocyanate-conjugated rabbit anti-mouse immunoglobulin. Drosophila ovaries were dissected, fixed in 4% formaldehyde in phosphate-buffered saline, blocked in 0.1% bovine serum albumin, 0.05% Tween 20 in phosphate-buffered saline, and incubated with undiluted anti-MacCer antibody and detected with Cy5 anti mouse antibodies from Jackson ImmunoResearch Laboratories. Ovaries were mounted in 80% glycerol. DAPI was included in the washes to reveal nuclei.
Fly Strains-Armadillo-Gal4 (II), actin-Gal4, and tubulin-Gal4 are described in flybase (fly.bio.Indiana.edu/gal4.htm). Brn 1.6P6 is described in Goode et al. (21) and egh mutations in Goode et al. (3). We further characterized three egh mutant alleles at the molecular level. For isolation of genomic DNA and characterization of the enzymatic activity of egh and brn mutants, animals of the correct genotype were identified as follows: egh and brn alleles were balanced over a GFP-expressing FM7 balancer chromosome. Larvae were sexed, and mutant males were picked on the basis of their lack of GFP expression. In the case of the egh 9PP4 and brn 1.6P6 alleles, the cuticular marker yellow (y) present on these chromosomes was also used to identify mutants by the color of the head skeleton.
Genetic Mosaic Analysis-brn and egh mutant alleles were recombined onto FRT18 and mitotic recombination clones were induced in adult females by heat shock for 60 min at 38°C. The genotypes used are as follows: y w brn 1 Rescue of the egghead Mutant Flies-pUAS-␤4GalT6 was constructed by cloning full-length cDNA into pUAST. The same construct was used for the in vitro glycosylation assay described above verifying the activity of ␤4GalT6. pUAST-egh was constructed by cloning the full-length coding sequence of egh into the BglII/XbaI sites of pUAST. Both constructs were used to generate transgenic flies. Stocks carrying a ubiquitous driver (armadillo-Gal4 or actin-Gal4) and a UAS line were established (four independent UAS insertions were tested). Their ability to rescue the lethality of egh 9PP4 mutant was assayed by scoring the male progeny when these stocks were mated to heterozygous y egh 9PP4 f females.

RESULTS
egghead and brainiac Mutants Produce Truncated Glycosphingolipids-In vitro studies predicted that the enzymes encoded by the egghead and brainiac genes would be required for glycosphingolipid biosynthesis in vivo. To confirm this, we tested four different egghead (egh) mutants. As a first step, we sequenced three of the egh alleles to determine the nature of their molecular lesions. egh 62d18 resulted from an 11 nucleotide deletion which caused a frameshift at amino acid 97 and deletion of most of the coding sequence. This allele is expected to cause a complete loss of enzymatic activity as the active site has been deleted. egh 9PP4 resulted from a 15-base pair deletion that removed amino acids 113-117, of which two are conserved. egh 7 resulted from a single nucleotide change that changes the conserved methionine at position 308 to lysine (M308K). Extracts were prepared from larvae mutant for these alleles as well as egh 64h6 and tested for mannosyltransferase activity with n-octyl glucoside (Fig. 1B). All four mutants were devoid of significant detectable mannosyltransferase activity. We next asked whether brainiac (brn) and egghead mutants were blocked in glycosphingolipid biosynthesis in vivo, as would be predicted on the basis of their in vitro enzymatic functions, if no redundancy in these enzyme functions or alternate biosynthetic pathways exist. Characterization of glycosphingolipids from mutant larvae by thin layer chromatography showed accumulation of the truncated product Glc␤1-Cer in all four egh mutants, whereas MacCer accumulated in the brn mutant (Fig.  1C). We produced a monoclonal antibody that specifically recognizes MacCer but not further elongated glycosphingolipids, to provide a tool to visualize this biosynthetic intermediate in vivo. The specificity of the antibody was tested by immunostaining of glycosphingolipids separated by thin layer chromatography. The antibody detected MacCer but not LacCer or GlcNAc␤1-3Man␤1-4Glc␤1Cer (data not shown). The antibody was then used to test for the presence of MacCer in clones of cells lacking Egghead or Brainiac activity in the Drosophila ovary. Clones of cells lacking Brainiac activity, which accumulate MacCer, showed strong labeling (Fig. 1D). In contrast, cells lacking Egghead activity, which we expect to be blocked at the Glc␤1-Cer step, showed no labeling above background with this antibody (Fig. 1D). This indicates that Egghead and Brainiac are present and active in the follicular epithelial cells of egg chambers. The anti-MacCer antibody produced only background levels of labeling in the wild-type cells adjacent to the clones presumably reflecting low level of expression of the immediate precursor substrate for Brainiac (and subsequent enzymes), as co-expression of multiple intermediate species is a common feature found for glycosphingolipids. Taken together, these observations confirm the predicted functions of Egghead and Brainiac as enzymes required for sequential elongation steps of glycosphingolipid biosynthesis in vivo. egghead Mutants Are Rescued by Vertebrate ␤4Gal-T6 -The finding that egh mutants lack MacCer synthase activity provided an opportunity to assess the significance of the core structure in glycosphingolipid function. We therefore asked whether expression of the human glycosyltransferase, ␤4Gal-T6, which functions as a LacCer synthase (Fig. 1A), could restore glycosphingolipid biosynthesis in egghead mutant animals. That this might be possible was plausible because Brainiac can elongate both a Man␤1-4Glc␤1 and a Gal␤1-4Glc␤1 substrate in vitro ( Fig. 2A, lanes 1 and 4; see also Ref. 5). We first verified that the lethality of egh 9PP4 mutant could be rescued by ubiquitous expression of an egghead cDNA in transgenic flies. Egh 9PP4 actin-Gal4 UAS-egh flies were recovered at 66% of the frequency of FM7 actin-Gal4 UAS-egh control flies. We next tested whether expression of the human ␤4Gal-T6 cDNA (full coding Golgi-retained form) also rescued egh 9PP4 mutant flies to viability. For four different UAS-␤4Gal-T6 transgenes, egh 9PP4 armadillo-Gal4;UAS-␤4Gal-T6 flies were recovered at frequencies ranging from 76% to Ͼ100% of control males (note that FM7 males are weak and are recovered at less than the expected mendelian ratio, so it is possible for a healthy rescued egh genotype to be recovered at over 100% of the level of the FM7 controls in this experiment). For UAS line III-1, which gave 80% recovery with armadillo-Gal4, recovery of mutant animals increased to Ͼ100% with the stronger tubulin-Gal4 driver, indicating that the level of transgene expression can affect the degree of rescue obtained.
Rescued egh 9PP4 armadillo-Gal4;UAS-␤4Gal-T6 animals were normal in appearance, and both males and females were fertile, indicating that the transgene also rescued the female sterility caused by egh mutants (3,4). Therefore we were able to establish a homozygous viable strain of rescued animals. Glycosphingolipids were isolated from the rescued flies and their composition was analyzed by high performance thin layer chromatography and 1 H NMR. Analysis of the glycosphingolipids from the rescued animals by thin layer chromatography showed accumulation of a disaccharide glycosphingolipid, as the intensity of the ceramide dihexoside (CDH) band was elevated compared with wild-type animals, and some of the trisaccharide (CTH) form was also seen (Fig. 2B). 1 H NMR demonstrated that the CDH from rescued animals is exclusively the vertebrate glycosphingolipid core Gal␤1-4Glc␤1-Cer (Fig. 2C). We were unable to isolate sufficient material to demonstrate the presence of the elongated structure GlcNAc␤1-3Gal␤1-4Glc␤1-Cer by NMR. However, bands found migrating in the trisaccharide region (CTH, Fig. 2B), indicated that some elongation of the disaccharide ceramide structure had occurred. Further structural studies of elongated glycosphingolipids are required to fully assess the impact the Gal for Man substitution imposes. To further confirm that the in vivo function of the ␤4Gal-T6 rescue depends on elongation by Brainiac, we tested whether we could rescue egh 9PP4 brn 1.6P6 double mutants with ␤4Gal-T6. In the absence of Brainiac no rescue was obtained, indicating that elongation of the LacCer core by Brainiac is FIG. 1. egghead mutants lack mannosyltransferase activity and elongated glycosphingolipids. A, glycosphingolipid core biosynthetic pathways of Drosophila and human. Glc␤1-Cer is a common precursor for most extended glycosphingolipids. Glc␤1-Cer is synthesized by a single homologous glucosyltransferase. Invertebrate elongation is initiated with a mannose residue catalyzed by Egghead followed by elongation with GlcNAc catalyzed by Brainiac and additional monosaccharide residues catalyzed by distinct glycosyltransferases. Phosphoethanolamine linked to GlcNAc residues are found only in invertebrates (data not shown). Vertebrate elongation is initiated with a galactose residue, catalyzed by two homologous ␤4-galactosyltransferases (␤4Gal-T5 and ␤4Gal-T6) forming LacCer (Gal␤1-4Glc␤1-Cer) (8). After this step, elongation branches into three pathways: ganglio-series catalyzed by the GM2 synthase, globo-series catalyzed by ␣4Gal-T1, and lacto-series catalyzed by multiple homologous ␤3GlcNAc-transferases. Highly conserved homologous enzymes carry out most biosynthetic steps in the glycosphingolipid pathways of vertebrates, insects, and nematodes. One exception is the addition of mannose in invertebrate glycosphingolipids catalyzed by Egghead. This saccharide linkage is not found in vertebrate glycosphingolipids, and no homologue of the egghead gene is found in vertebrates. B, mannosyltransferase activity (counts/min) measured in larvae extracts of wild-type, brainiac 1 2. Human ␤4Gal-T6 converts egh mutants to the vertebrate glycosphingolipid biosynthetic pathway. A, Brainiac transferred GlcNAc to both the vertebrate core Gal␤1-4Glc␤1-octyl and the insect core Man␤1-4Glc␤1-octyl (lanes 1 and 4, respectively). ␤3GnT2 and ␤3GnT5 only elongated on Gal␤4Glc␤1-octyl (lanes 5 and 6) but not Man␤4Glc␤1-octyl (lanes 2 and 3). Glycosylation products were separated by HPTLC and stained with orcinol. B, high performance thin layer chromatography of glycosphingolipids from samples of the ␤4Gal-T6 rescued egh 9PP4 flies. C, 1 H NMR analysis of CDH region glycosphingolipids isolated by preparative HPTLC of samples from the ␤4Gal-T6 rescued egh 9PP4 flies. The sample was deuterium-exchanged and dissolved in Me 2 SO-d 6 /2% D 2 O; the spectrum was acquired at 35°C. Arabic numerals refer to ring protons of residues designated by Roman numerals or capital letters in the corresponding structures. The presence of LacCer (Gal␤1-4Glc␤1-Cer) was detected (peaks labeled I-1, I-2, II-1, and II-4). The asterisk marks the chemical shift position of the H-1 of ␤-Man in an NMR spectrum of MacCer (Man␤1-4Glc␤1-Cer) acquired under identical conditions (compare Fig. 3B); no comparable signal is detectable in this spectrum. Attempts to analyze glycosphingolipid from the triglycosyl-ceramide region were unsuccessful; GlcNAc␤1-3Gal␤1-4Glc␤1-Cer could not be identified. required for glycosphingolipid function in the rescued egh mutant flies. These experiments indicate that replacing the Mac-Cer core with LacCer is compatible with glycosphingolipid function in the fly.
Introduction of Egghead into Mammalian Cells-Three pathways of vertebrate glycosphingolipid biosynthesis are defined by the nature of third residue added to the LacCer core (Fig. 1A). Three different enzymes are responsible for the defining steps. The neo/lacto-series contains GlcNAc in a ␤3 linkage to LacCer. This resembles insect glycosphingolipids, which have GlcNAc in a ␤1,3 linkage to MacCer. However, the two mammalian enzymes that add GlcNAc in a ␤1,3 linkage to LacCer cannot elongate a MacCer substrate in vitro ( Fig. 2A,  lanes 2 and 3). CHO-K1 cells mainly express the ganglioside glycosphingolipid, GM3. Stable transfection of CHO-K1 cells to express egghead resulted in accumulation of MacCer, visualized by TLC (Fig. 3A). The intensity of the CDH band was higher in egghead-expressing cells, and this was shown by NMR to reflect accumulation of MacCer without further elongation (Fig. 3B). Egghead transfected cells, but not control CHO cells, bound the MBL (Fig. 3D), consistent with the prediction that they produce MacCer. These observations indicate that expression of Egghead can to a considerable extent override endogenous glycosyltransferases in mammalian cells and lead to synthesis of a truncated MacCer product.
Anti-MacCer antibody strongly labeled CHO cells transfected to express Egghead but not control cells transfected to express the vertebrate ␤3GnT2 enzyme (Fig. 3D). This indicates that Egghead can redirect mammalian glycosphingolipid biosynthesis to produce MacCer, which is not further elongated. Egghead therefore may serve as a competitive biosynthetic inhibitor of early glycosphingolipid synthesis in vertebrates. Although the Egghead-expressing CHO cells contain a low level of residual GM3, the reduced production of normal glycosphingolipids did not appear to cause defects in the growth, survival, or morphology of these cells in culture. DISCUSSION In this report we have presented direct evidence that Egghead and Brainiac are enzymes essential for glycosphingolipid biosynthesis in vivo, and thus demonstrate that glycosphingolipids are essential for Drosophila development. Furthermore, we show that substituting Egghead for the vertebrate LacCer synthase can provide the essential glycosphingolipid functions required to support development of egghead mutants, despite the fact the core structure of vertebrate and insect glycosphingolipids are different. In contrast introduction of the Drosophila glycosphingolipid pathway into mammalian cells can interfere with the normal biosynthesis because the vertebrate enzymes cannot elongate the MacCer insect glycosphingolipid core.
Drosophila lacking zygotic activity of the egh and brn genes die as pupae. Drosophila lacking maternal and zygotic activity of these enzymes are devoid of elongated glycosphingolipids and have a more severe defect, dying as embryos with a defect in correct specification of neural and epidermal cell types. Elongated glycosphingolipids also appear to be required for normal development of the mouse embryo. Mice mutant for the glucosylceramide synthase enzyme controlling the ultimate glycosphingolipid precursor die during gastrulation due to apoptosis in all germ layers but particularly in ectoderm (22). The mouse embryos lacking glucosylceramide synthase die at an earlier stage of embryogenesis than Drosophila egh and brn mutants. This may reflect a difference in the position at which truncation occurs. For example, elevated ceramide levels are known to be pro-apoptotic (23). Accordingly, knockdown of the Glc␤1-Cer synthase in Drosophila by RNAi also leads to increased apoptosis, thought to be due to elevated ceramide levels (24). Whether ceramide levels are responsible for apoptosis in the mouse mutants in vivo remains to be determined.
Egghead and Brainiac are expressed and required during oogenesis (3,4,21,25). In the absence of their function, development of the ovarian follicles is defective. We note that earlier reports suggested that the activity of these genes was limited to the germ line, because phenotypes were not observed in somatic mutant clones in the follicular epithelia. Using the Mac-Cer antibody on genetic mosaics we show that Egghead and Brainiac are both present and active in the follicular epithelia.
Interestingly, the orthologs of brn and egh do not appear to be essential for development of the nematode C. elegans (26). Instead, both genes are required for susceptibility to the crystal (5B) toxin from the bacterium Bacillus thuringiensis. It therefore appears that Drosophila has acquired functions for glycosphingolipids that are not shared among all invertebrates and that Drosophila presents an excellent model for studies of such functions in vivo. The phenotypes associated with brn and egh mutants initially suggested a role of these in Notch receptor modulation similar to but distinct from fringe. Given the demonstrated function of Brainiac and Egghead in glycosphingolipid biosynthesis it is tempting to suggest that extended glycosphingolipids in Drosophila might play a direct role in modulation of receptor functions in a manner similar to the effects of GM3 on the epidermal growth factor receptor (27)(28)(29). Alternatively, extended glycosphingolipids might play an indirect role on signaling by virtue of their contribution to the formation of lipid rafts and the recruitment of receptors to rafts. Another appealing possibility is that glycosphingolipids influence the cleavage of membrane-bound ligands, such as the activation of the epidermal growth factor receptor ligands Spitz, Gurken, and Keren by the Rhomboid family of secretases (30). Of special interest in this context is the possibility that glycosphingolipids could influence Rhomboid-2 cleavage of Gurken in oogenesis. Likewise in the case of Notch signaling, glycosphingolipid could affect the ␥-secretase, which is organized in lipid rafts and cleaves the intracellular tail of Notch (31). The availability of Drosophila lacking elongated glycosphingolipids will provide an opportunity to investigate the functions of glycosphingolipids in cell signaling in vivo. In considering possible modes by which glycosphingolipids may act, it is intriguing that they can do so apparently normally when their core structure has been altered by replacing MacCer with LacCer. This observation provides a starting point for further humanization of the biosynthetic pathway by further replacement of Brainiac with enzymes responsible for the next steps in the mammalian lacto-, ganglio-, or globo-series biosynthetic pathways. Conversely, vertebrate cells with MacCer-based glycosphingolipids provide a unique genetic tool to address structure-function relationships for glycosphingolipids.