Structural Analysis of Glycosaminoglycans in Animals Bearing Mutations in sugarless, sulfateless, and tout-velu DROSOPHILA HOMOLOGUES OF VERTEBRATE GENES ENCODING GLYCOSAMINOGLYCAN BIOSYNTHETIC ENZYMES*

Mutations that disrupt developmental patterning in Drosophila have provided considerable information about growth factor signaling mechanisms. Three genes recently demonstrated to affect signaling by members of the Wnt, transforming growth factor-b, Hedgehog, and fibroblast growth factor families in Drosophila encode proteins with homology to vertebrate enzymes involved in glycosaminoglycan synthesis. We report here the biochemical characterization of glycosaminoglycans in Drosophila bearing mutations in sugarless, sulfateless, and tout-velu. We find that mutations in sugarless, which encodes a protein with homology to UDP-glucose dehydrogenase, compromise the synthesis of both chondroitin and heparan sulfate, as would be predicted from a defect in UDP-glucuronate production. Defects in sulfateless, a gene encoding a protein with similarity to vertebrate N-deacetylase/N-sulfotransferases, do not affect chondroitin sulfate levels or composition but dramatically alter the composition of heparin lyasereleased disaccharides. N-, 6-O-, and 2-O-sulfated disaccharides are absent and replaced entirely with an unsulfated disaccharide. A mutation in tout-velu, a gene related to the vertebrate Exostoses 1 heparan sulfate co-polymerase, likewise does not affect chondroitin sulfate synthesis but reduces all forms of heparan sulfate to below the limit of detection. These findings show that sugarless, sulfateless, and tout-velu affect glycosaminoglycan biosynthesis and demonstrate the utility of Drosophila as a model organism for studying the function and biosynthesis of glycosaminoglycans in vivo.

Until recently, the analysis of glycosaminoglycan biosynthesis and structure has focused on vertebrate tissues and cultured cells. These studies have produced a wealth of information about the enzymes required for glycosaminoglycan synthesis and the varied structures of these polymers in different cell types and tissues (1). However, glycosaminoglycans are also found in invertebrates, including the fruit fly, Drosophila melanogaster, and the nematode Caenorhabditis elegans. Recent studies have shown that both of these genetically tractable organisms make chondroitin and heparan sulfate (2,3). Dro-sophila is notable for a structurally complex heparan sulfate, containing the principal sulfated species represented in vertebrates (2). Both Drosophila and C. elegans therefore provide the means to study the genes required for glycosaminoglycan synthesis in vivo, and the role of these polymers in developmental patterning.
Genetic screens for mutations in Drosophila affecting growth factor signaling pathways have revealed, thus far, three genes that bear homology to genes encoding vertebrate glycosaminoglycan biosynthetic enzymes. sgl (4 -6), sfl (7), and ttv (8), encode proteins with homology to UDP-glucose dehydrogenase, N-deacetylase/N-sulfotransferase, and EXT1, 1 respectively. sgl mutations compromise signaling mediated by Wg (4 -6), a member of the Wnt family, Dpp (4), a transforming growth factor-␤/bone morphogenetic protein-related protein, and the FGF receptors breathless and branchless (7). sfl likewise affects Wg, Hh, and FGFR signaling (7,9). ttv is unique among these three genes in that Hh signaling is selectively compromised, with Wg-directed events intact in ttv embryos and third instar larvae (9). Analysis of the vertebrate homologue of ttv, EXT1, has shown that EXT1 encodes a heparan sulfate co-polymerase (10,11). Recent studies of Drosophila ttv mutants have provided two lines of evidence that Ttv is also a heparan sulfate co-polymerase. First, analysis of glycosaminoglycans from ttv mutant larvae revealed a dramatic reduction in the levels of heparan sulfate but not chondroitin sulfate (2). Second, immunohistochemical staining with an antibody that recognizes an epitope of heparan sulfate generated by cleavage with heparin lyase is significantly reduced in embryos lacking both maternal and zygotic ttv function (9).
It is now well established that glycosaminoglycans and their associated proteoglycans are required for signaling mediated by a number of growth factors that pattern tissues during development. Indeed, defects in a Drosophila glypican encoded by the dally gene disrupt both Wg and Dpp signaling in a tissue-selective manner (12)(13)(14)(15). The ability of Dally to affect different signaling pathways in distinct cells suggests the hypothesis that the nature of the glycosaminoglycan attached to the protein core influences its biological activity. Structurally diverse forms of heparan sulfate show tissue-specific distributions (16), indicating that control of glycosaminoglycan fine structure could well be important in governing patterning and cell-cell communication. The distinct developmental defects found in mice bearing a gene trap mutation in heparan sulfate 2-O-sulfotransferase argues that distinct modifications of heparan sulfate are indeed crucial for particular patterning events (17). One way to directly investigate the biological functions of specific glycosaminoglycan structures in vivo is to examine patterning in animals bearing mutations in genes affecting distinct biosynthetic activities. A critical step in this approach is to perform structural analysis of glycosaminoglycans from mutants defective for particular genes to establish the molecular changes resulting from the loss of this biosynthetic activity. In earlier work we reported methods for structural analysis of glycosaminoglycans in Drosophila and C. elegans using HPLC separation of disaccharides generated by lyase digestion (2). We have now applied these procedures to the detailed analysis of glycosaminoglycans isolated from animals bearing mutations in three genes proposed to serve roles in glycosaminoglycan biosynthesis on account of their homology to vertebrate genes. Based upon the molecular phenotype of these mutants, we can draw conclusions as to the signaling activities of different classes of glycosaminoglycans during developmental patterning. Determination of Unsaturated Disaccharides by HPLC-The determination of unsaturated disaccharides was performed by fluorometric post-column HPLC as reported previously (2).

Materials
Preparation of Glycosaminoglycans from Drosophila-For the analysis of sgl, second instar larvae were used on account of the considerable lethality observed between the second and third instar stages. Studies of sfl and ttv were conducted using late third instar larvae, and both second and third instar larvae of the Oregon-R strain were used as wild-type controls. Approximately 20 mg of wild-type, sgl, sfl, or ttv homozygous larvae were lyophilized and extracted in 1.0 ml of 0.5% SDS, 0.1 M NaOH, 0.8% NaBH 4 as described previously (2). The resulting crude glycosaminoglycan was dissolved in 250 l of 50 mM sodium phosphate buffer (pH 6.0), and applied on a Ultrafree-MC DEAE, which had been equilibrated with 50 mM sodium phosphate buffer (pH 6.0). The eluted fractions of 0.15 M and 1.0 M NaCl in the phosphate buffer were collected, desalted with Biomax-5, evaporated, and resuspended in 12 l of water.
Analysis of Heparan Sulfate in Drosophila-For the analysis of Drosophila heparan sulfate, 5 l of 0.1 M acetate buffer (pH 7.0) with 10 mM calcium acetate and 10 l of an aqueous solution containing heparin lyase mixture, heparinase (1 mIU), heparitinase I (1 mIU), and heparitinase II (1 mIU) were added to a 5-l portion of sample solution. The mixture was incubated at 37°C for 16 h, and an 8-l aliquot was loaded onto the high performance liquid chromatograph. Using known amounts of standards we have determined that the detection limit of our system is 1.5 ng of heparan sulfate-derived disaccharides.
Analysis Genetic Stocks and Mutant Alleles-Two different alleles of sgl were used for our analysis, sgl P1731 and sgl A31 . Chromosomes with these mutations were maintained over a Green Fluorescent Protein gene bearing third chromosome balancer, TM3 P[wϩ, Act-GFP], Ser 1 . Larvae homozygous for sfl l(3)03844 were employed for the analysis of sfl function, and maintained over the TM3 balancer as above. tout-velu l(2)00681 mutants used in this study were maintained as a stock over the second chromosome balancer, CyO P [wϩ, ubq-GFP]. In all cases, homozygous larvae were identified by the lack of GFP fluorescence detectable under UV illumination in a dissecting microscope.

RESULTS
The analysis of glycosaminoglycans in sgl, sfl, and ttv larvae consisted of four assays. In the first two, the levels of both chondroitin and 4-O-sulfated chondroitin disaccharides were determined by HPLC separation of chondroitinase-treated material eluted from DEAE membrane under either low (0 -0.15 M NaCl) or high salt (0.15-1.0 M NaCl) conditions. Similarly, two assays of disaccharides generated by lyase digestion of heparan sulfate were conducted, one for material retained on DEAE membrane under low salt conditions (0 -0.15 M NaCl), and a second for material that bound to DEAE membrane at salt concentrations between 0.15 and 1.0 M. These experiments therefore provided a profile of chondroitin, low sulfated chondroitin 4-sulfate (Fig. 1, A and B), N-acetyl heparosan (a nonsulfated precursor of heparin), and heparan sulfate (Fig. 2, A  and B). The material that elutes prior to the first disaccharide peak for profiles of both chondroitin sulfate and heparan sulfate-derived disaccharides (Figs. 1 and 2) is found in enzyme blank controls, and represents other molecules found in the crude glycosaminoglycan preparation, such as glycoproteins. These molecules react with the fluorophore, 2-cyanoacetamide,  Tables I and III). but do not interfere with the analysis of the glycosaminoglycanderived disaccharides.
Analysis of Wild-type-Separation of glycosaminoglycans released from DEAE membrane with low or high salt washes showed these polymers can be subdivided into distinct classes. The low salt fraction of chondroitinase-sensitive polymers, which represent approximately 20% of the total, was comprised of ⌬Di-OS only, whereas the high salt-released fraction showed a mixture of ⌬Di-OS and ⌬Di-4S (Fig. 1, Table I). Likewise, distinct classes of heparan sulfate polymers were found in the low and high salt DEAE fractions (Fig. 2, Table II). Approximately 10% of the heparin lyase-sensitive material was found in the low salt fraction and it contained only ⌬UA-GlcNAc. The high salt fraction was a complex mixture of disaccharide species, dominated by mono-, di-, and trisulfate forms (all sulfated forms, 67%). These profiles of wild-type glycosaminoglycans, done in parallel with the analysis of sgl, sfl, and ttv mutants, provided a basis for determining the effects of these genes on glycosaminoglycan biosynthesis.
sugarless-sgl encodes a protein with homology to bovine UDP-glucose dehydrogenase (4 -6), an enzyme required for generating UDP-glucuronate, the substrate for glucuronate addition to glycosaminoglycan chains. If sgl were the only gene capable of providing UDP-glucose dehydrogenase activity, the levels of chondroitin, chondroitin sulfate, and heparan sulfate would be expected to be dramatically reduced, but not necessarily eliminated, in sgl mutant second instar larvae (see below). This prediction was borne out; we were unable to detect any chondroitin or chondroitin sulfate in the fraction of glycosaminoglycans from sgl mutants retained on DEAE under high salt conditions (Fig. 1, Tables I and III). A trace amount of ⌬Di-OS was detected in chondroitinase-sensitive glycosaminoglycans from sgl larvae that bound to DEAE under low salt conditions (Fig. 1C). sgl mutations also had a profound effect on heparan sulfate biosynthesis. Trace amounts of ⌬UA-GlcNAc were found only in the low salt preparation (Fig. 2C, Tables II  and III).
sgl, like sfl, and ttv, is expressed in the ovary and maternal contributions of mRNA and protein to the oocyte are essential for early embryonic development. Embryos that are sgl/sgl receive sufficient maternal contributions of sgl function from their sgl/ϩ mothers to go through early development normally and maternally provided activity typically wanes in late larval and early pupal development. For this reason, glycosaminoglycans are not expected to be completely absent in second or third instar larvae homozygous for sgl. We observed approximately 34-and 12-fold reductions in total disaccharides derived from chondroitin sulfate and heparan sulfate, respectively, in sgl mutants (Tables I-III). These findings show that sgl is critical for both chondroitin sulfate and heparan sulfate biosynthesis, and is likely the only gene providing UDP-glucose dehydrogenase activity in Drosophila.
sulfateless-The amino acid sequence of sfl predicts this gene encodes an N-deacetylase/N-sulfotransferase. If sfl does indeed provide this activity in Drosophila, sfl mutants should retain normal levels and composition profiles of chondroitin and chondroitin sulfate. Our analysis of sfl mutants documents that this is indeed the case; sfl mutants have normal levels of chondroitin and low sulfated chondroitin 4-sulfate chondroitin in both low and high salt DEAE fractions (Fig. 1, A and B, Table  I). sfl mutants did, however, show dramatic changes in their heparan sulfate profiles (Fig. 2, E and F and Table II). First, the five different forms of sulfated disaccharides detected with our methods were completely absent from the high salt DEAE fraction obtained form sfl homozygous larvae (Fig. 2F, Table  II). A trace amount of ⌬UA-GlcNAc was found in the high salt fraction from sfl mutants, although much reduced compared with wild-type larvae. Second, there was a large increase in the levels of ⌬UA-GlcNAc in the low salt DEAE fraction of sfl mutants compared with wild-type (Fig. 2E) and this disaccharide represents all the material derived by exhaustive heparin lyase digestion of glycosaminoglycans from this mutant (Table  III). It is important to note that the total level of heparan sulfate-derived disaccharide from sfl larvae is nearly the same  Tables II and III). ttv mutations reduce the levels of all heparan sulfate-derived disaccharides to below detectable limits (G and H, see also Tables II and III). as wild-type (12.3 versus 13.9 ng/mg for sfl and wild-type, respectively), indicating that differences in recovery do not account for the changes in composition we observe. These findings show that sfl: 1) affects heparan sulfate synthesis and not chondroitin or chondroitin sulfate generation, 2) is required for the sulfation at the N-position of GlcNAc, the 2-O-position of either glucuronic or iduronic acid, and the 6-O-position of Glc-NAc and GlcNS; and 3) results in the accumulation of N-acetyl heparosan, a nonsulfated precursor of heparin, at the expense of sulfated forms of disaccharides. tout-velu-Earlier studies showed that ttv, a gene related to vertebrate EXT1, affects the levels of heparan sulfate in vivo (2,9). These finding were consistent with ttv encoding a heparan sulfate co-polymerase. We have extended these experiments to determine precisely what glycosaminoglycans remain in ttv mutant larvae. Chondroitin sulfate-derived disaccharides are largely unchanged, with a modest reduction in levels while retaining a wild-type composition in both the low and high salt fractions ( Fig. 1 and Table I). All heparan sulfate-derived disaccharides, however, showed a marked reduction, essentially below our detectable limits of 1.5 ng (Fig. 2, G and H, and Tables II and III). The selective and substantive effect of ttv on heparan sulfate levels in vivo indicates that ttv is critical for the bulk of heparan sulfate biosynthesis.

Roles of sgl, sfl, and ttv in Glycosaminoglycan Biosynthesis-
We have examined the structures of glycosaminoglycans in animals bearing mutations in genes encoding proteins with homology to three known glycosaminoglycan biosynthetic enzymes. In all three cases, the changes in whole animal glycosaminoglycan profiles are consistent with the Drosophila gene providing the enzymatic activity predicted from the relatedness to the vertebrate genes (Table III, Fig. 3).
sgl mutants show dramatic reductions in both chondroitin and heparan sulfate levels, consistent with a defect in UDPglucuronate generation catalyzed by a UDP-glucose dehydrogenase. These findings support earlier studies showing that modification of Dally (14) and Syndecan (4) is altered in sgl mutants. The residual levels of chondroitin and heparan sulfate-derived disaccharides remaining in sgl mutants most likely results from maternal contributions of wild-type sgl mRNA to the embryo, which provides sufficient activity for normal embryonic development. The trace amounts we detect indicate that zygotic sgl function is required for continued glycosaminoglycan biosynthesis once maternal stores are exhausted.
sfl, a Drosophila gene encoding a protein related to verte-   brate N-deacetylase/N-sulfotransferase, also had a profound effect on heparan sulfate levels in vivo. While chondroitin levels and composition were unaffected by loss of sfl function, sfl mutants showed changes in heparan sulfate composition without appreciably affecting the total levels of heparan sulfate. We could not detect any sulfated disaccharides generated by heparin lyase digestion of glycosaminoglycans from sfl homozygous larvae. This finding provides in vivo support for earlier work using in vitro preparations (18) or tissue culture cells (19,20) demonstrating that N-deacetylation/N-sulfation is an early step in heparan sulfate modification required for subsequent sulfotransferase reactions at the 2-O position of iduronate or glucuronate, and the 6-O-position of glucosamine. N-Deacetylase/N-sulfotransferase activity is likely to be required to provide the template for a variety of later modification steps.
The analysis of ttv reported here extends earlier studies and documents that heparan sulfate synthesis is significantly compromised by loss of ttv function. Heparin lyase-sensitive glycosaminoglycans were reduced to below detectable limits in ttv mutant larvae. ttv did not, however, effect the levels or composition of chondroitinase-sensitive glycosaminoglycans. These findings show that ttv, a gene with the greatest homology to vertebrate EXT1 (8), supports the bulk of heparan sulfate synthesis during post-embryonic development.
The analysis of glycosaminoglycans in sgl, sfl, and ttv mutants has shown that the glycosaminoglycan biosynthetic apparatus is largely conserved between Drosophila and vertebrates. The structural forms of chondroitin and heparan sulfate represented in Drosophila are also those found in vertebrate species. In addition, the core proteins for several families of proteoglycans are found in Drosophila, including glypican (13) and syndecan (21). These findings demonstrate the utility of Drosophila as a model organism for studying the biosynthesis and function of glycosaminoglycans and proteoglycans in vivo.
Glycosaminoglycans in Developmental Patterning and Growth Factor Signaling-Our analysis of sgl, sfl, and ttv allows us to make some conclusions concerning the functions of various glycosaminoglycan forms in growth factor signaling and morphogenesis. Of these three mutations, sgl and sfl would appear to have the broadest effects on signaling, compromising both wg and FGFR-directed events. Furthermore, there is good evidence that sgl is required for dpp signaling in imaginal tissues (4,22). sfl apparently is also required for normal Hh signaling in the developing wing. While it is not known if sgl and sfl influence Wg, Dpp, FGF, and Hh signaling equally, their comparable participation in Wg and FGF signaling suggests that loss of both chondroitin and heparan sulfate in sgl mutants produces no more deleterious effect than a failure to add sulfate modifications to heparan sulfate. It would seem therefore, that at least for FGF and Wg signaling, heparan sulfate is the critical glycosaminoglycan in vivo.
In contrast to the multiple growth factor pathways affected by sgl and sfl, ttv selectively influences Hh signaling, leaving wg and FGFR-mediated patterning untouched in both the embryo and in the imaginal discs (8,9). The large reductions in heparan sulfate levels that accompany loss of ttv function might lead one to predict that ttv would seriously compromise signaling mediated by several growth factors known to require sulfated heparan sulfate, Wg and FGF in particular. Yet this is not the case. In embryos lacking both maternal and embryonically derived Ttv activity, FGFR and Wg signaling remain (9).
Several explanations for the selective participation of ttv in Hh signaling are possible in light of these findings. First, we have examined glycosaminoglycans in larvae, and the studies where ttv function has been completely removed were primarily performed using embryos. It is possible that there are higher levels of heparan sulfate in ttv mutant embryos compared with larvae, presumably derived from other EXT-related genes that could provide co-polymerase activity. Second, the experiments looking at ttv function in larvae made use of methods to induce recombination events in ttv/ϩ animals midway through development that produce clones of ttv/ttv cells. Sufficient Ttv may remain in these clones to provide significant levels of heparan sulfate co-polymerase activity, levels sufficient for Wg and FGF signaling, but not for Hh. ttv mutants may also contain small amounts of specific proteoglycans that are heparan sulfatemodified normally and permit normal FGF and Wg signaling (9). This explanation would require Ttv to have core protein selectivity, placing heparan sulfate chains on a distinct set of proteoglycans. Finally, it is possible that in the absence of any appreciable heparan sulfate co-polymerase activity, proteoglycan protein cores are modified with chondroitin sulfate, and that this sugar polymer is capable of supporting Wg and FGF signaling, but not Hh signaling or transport. In this scenario, sfl mutations have such a broad affect because N-acetyl heparosan chains are added to their normal protein cores, but do not function because of their lack of sulfation. The presence of heparan chains would preclude their substitution with chondroitin sulfate, preventing rescue of proteoglycan function in Wg and FGF signaling by chondroitin sulfate modification. FIG. 3. Role of sgl, sfl, and ttv in the biosynthesis of proteoglycans in Drosophila. Glycosaminoglycan synthesis is initiated by addition of xylose to specific serine residues of the protein core, with UDP-xylose serving as the sugar donor. The nucleotide-sugar donors are listed on the left of the figure, with the sequence of synthesis for the tetrasaccharide linker and glycosaminoglycan chain going from top to bottom. After the tetrasaccharide linker is made (GlcA-Gal-Gal-Xylserine), heparan sulfate synthesis requires addition of GlcNAc, whereas chondroitin proceeds with GalNAc addition. Our data support Sgl as a UDP-glucose dehydrogenase, the enzyme that catalyzes the conversion of UDP-glucose to UDP-glucuronate, a nucleotide-sugar donor required for all glycosaminoglycan biosynthesis. Ttv is related to EXT1, a heparan sulfate co-polymerase that adds alternatively GlcNAc and GlcA to the growing polymer. The biochemical defects of ttv mutants are consistent with ttv encoding the principal heparan sulfate co-polymerase in Drosophila. Finally, our data support the assignment of Sfl as an N-deacetylase/N-sulfotransferase required for replacing the acetyl group on GlcNAc with a sulfate. Loss of sfl function also prevents 2-O-sulfation of uronic acid and 6-O-sulfation of glucosamine, suggesting that N-sulfation is essential for subsequent modification steps in Drosophila. Abbreviations: Xyl, xylose; Gal, galactose; GlcA, glucuronic acid; GalNAc, N-acetylgalactosamine.