The Putative Tumor Suppressors EXT1 and EXT2 Are Glycosyltransferases Required for the Biosynthesis of Heparan Sulfate*

Hereditary multiple exostoses, characterized by multiple cartilaginous tumors, is ascribed to mutations at three distinct loci, denoted EXT1–3. Here, we report the purification of a protein from bovine serum that harbored thed-glucuronyl (GlcA) andN-acetyl-d-glucosaminyl (GlcNAc) transferase activities required for biosynthesis of the glycosaminoglycan, heparan sulfate (HS). This protein was identified as EXT2. Expression of EXT2 yielded a protein with both glycosyltransferase activities. Moreover, EXT1, previously found to rescue defective HS biosynthesis (McCormick, C., Leduc, Y., Martindale, D., Mattison, K., Esford, L. E., Dyer, A. P., and Tufaro, F. (1998) Nat. Genet. 19, 158–161), was shown to elevate the low GlcA and GlcNAc transferase levels of mutant cells. Thus at least two members of the EXT family of tumor suppressors encode glycosyltransferases involved in the chain elongation step of HS biosynthesis.


Hereditary multiple exostoses, characterized by multiple cartilaginous tumors, is ascribed to mutations at three distinct loci, denoted EXT1-3. Here, we report the purification of a protein from bovine serum that harbored the D-glucuronyl (GlcA) and N-acetyl-D-glucosaminyl (GlcNAc) transferase activities required for biosynthesis of the glycosaminoglycan, heparan sulfate (HS). This protein was identified as EXT2. Expression of EXT2
Heparan sulfate (HS) 1 proteoglycans, ubiquitously distributed on cell surfaces and in the extracellular matrix, consist of sulfated glycosaminoglycan chains that are covalently bound to various core proteins. HS polysaccharide, increasingly implicated in physiological processes such as cell adhesion, cytokine action, and regulation of enzymic catalysis, owes its biological properties to interactions with various proteins, mediated by specific saccharide sequences. Biosynthesis of HS chains involves the formation of an initial, simple polysaccharide, composed of alternating D-glucuronic acid (GlcA) and N-acetyl-D-glucosamine (GlcNAc) units, joined by 134 linkages. This polymer is subsequently modified through a series of reactions, which involves partial N-deacetylation and N-sulfation of Glc-NAc units, C-5 epimerization of GlcA to L-iduronic acid residues, and O-sulfation at various positions (1). The GlcA transferase (GlcA-T) and GlcNAc transferase (GlcNAc-T) reactions required to generate the initial HS polysaccharide precursor have been associated with a single protein (2), hereafter referred to as "HS-polymerase" (HS-POL). Partial purification of proteins from bovine serum revealed a ϳ70-kDa component with both activities (3). We now report the molecular cloning of this protein and demonstrate that it is 94% identical to human EXT2, a member of the EXT family of tumor suppressors. Also EXT1, another member of the same family (4 -7), is implicated with similar catalytic activities. Mutations of EXT genes have been associated with the development of hereditary multiple exostoses (HME), the most frequent of all skeletal dysplasias. These findings suggest that alterations in the formation of the HS precursor polysaccharide may be involved in tumor formation and further point to an important role for HS in control of bone growth.

MATERIALS AND METHODS
Isolation of HS-POL-The polymerase was isolated from bovine serum using an extension of the protocol described previously (3). Briefly, the procedure involved successive chromatographies through the following matrices: Red-Sepharose, concanavalin A-Sepharose connected to Red-Sepharose (recirculation for 48 h), phenyl-Sepharose, Superdex 200 (gel chromatography), UDP-Sepharose, Mono Q (anion-exchange chromatography), and Mono P (chromatofocusing). The product was finally separated by preparative SDS-PAGE and stained with Coomassie Blue. The implicated ϳ70-kDa component was digested with trypsin in the gel, and the resultant peptides were separated and sequenced as described (8).
cDNA Library Screening and DNA Sequencing-The cDNA probe used for screening was derived from a human EST clone (826 bp) from Soares fetal liver spleen library (GenBank accession no. U13869 (IM-AGE Consortium)). The clone was excised from vector pT7T3D (Amersham Pharmacia Biotech) with PacI and EcoRI and was labeled with [␣-32 P]dCTP, using a random-priming kit (Boehringer Mannheim). Nitrocellulose replicas of plaques from the bacteriophage gt10 bovine kidney cDNA library (catalog no. BL3001a; CLONTECH) were hybridized with the labeled probe according to the instructions of the manufacturers.
The nucleotide sequences of cDNAs were determined by repeated sequencing of both strands of alkaline-denatured plasmid DNA using the Cy5 AutoRead sequencing kit (Amersham Pharmacia Biotech). Nucleotide sequences were labeled using a Cy5-dATP labeling mix, and the sequencing reactions were performed using T7 DNA polymerase. Sequences were determined on an ALFexpress system (Amersham Pharmacia Biotech) and analyzed using the DNA-Star (DNASTAR Inc., Wisconsin) program. The nucleotide and protein sequences were applied to data base screening using BLAST search, NCBI (Internet address: http://www.ncbi.nlm.nih.gov/).

Transient Expression of HS-POL in COS-7 Cells-
The 2884-bp cDNA insert recovered from the bovine kidney cDNA library was cleaved with restriction enzyme BseRI to generate a 2256-bp fragment (corresponding to nucleotides 185-2441), which was then treated with Klenow fragment to generate blunt ends. This product was ligated into a pcDNA3 expression vector (Invitrogen), modified to introduce a His/ FLAG (MGGSHHHHHHDYKDDDDK-) tag at the N terminus. COS-7 cells were cultured in Dulbecco's modified Eagle's medium-F12 (catalog no. 31330-038, Life Technologies, Inc.) supplemented with 50 units/ml penicillin, 50 g/ml streptomycin, and 10% (v/v) heatinactivated (56°C, 30 min) fetal calf serum at 37°C and 7.5% CO 2 . For electrotransfection 70% confluent cells in a 175-cm 2  mM MgCl 2 , pH 7.2. The cells were resuspended in 500 l of washing buffer, and 30 g of plasmid cDNA was added along with 50 g of fish sperm carrier DNA (Boehringer Mannheim). Electrotransfection was carried out in a 0.4-cm cuvette (BTX) at 360 V and 500 microfarads. Following transfection the cells were resuspended in culture medium containing 2% Me 2 SO, transferred to a 10-cm culture dish, left at room temperature for 20 min, and finally incubated at 37°C for 72 h.
SDS-PAGE and Immunoblotting-Protein was analyzed on 10% polyacrylamide gel in SDS, on a Bio-Rad MiniProtean unit, according to the manufacturer's instruction. For Western detection, separated proteins were transferred to a polyvinylidene difluoride membrane (Millipore) in a Bio-Rad Trans-Blot semidry electroblot system, using 10 mM CAPS, 5% MeOH, pH 11, as transfer buffer at 6 V for 40 min. The membrane was blocked with PBS, 0.1% Tween 20, and 15% bovine serum and was then incubated with the anti-FLAG M2 antibody (Kodak) in the same solution. After washing, the His/FLAG-tagged HS-POL was detected using a chemiluminescence kit (ECL; Amersham Pharmacia Biotech), and the signal was recorded on a Bio-Rad G525 phosphoimaging device.
Assay of Cellular Glycosyltransferase Activities-Cell lines analyzed for GlcA-T and GlcNAc-T activities included clone 1D from Lmtk Ϫ mouse fibroblasts, mutant gro2C derived from the same cells (9), COS-7 cells, and transfected variants as indicated. After washing with PBS cells were scraped off in 50 mM Hepes, 0.15 M NaCl, 1% Triton X-100, pH 7.2, and lysed by incubation with gentle agitation at 4°C for 1 h. The lysates were centrifuged at 16,000 ϫ g for 10 min, and supernatants were subjected to glycosyltransferase assays as described before (3). Briefly, GlcA-T activity was measured by incubating lysates with UDP-

RESULTS
Cloning of HS-POL-The putative HS-POL isolated previously from bovine serum (3) was subjected to further purification through a series of chromatography steps (see "Materials and Methods"). The GlcA-and GlcNAc-T activities remained associated throughout this procedure. Final separation by SDS-PAGE yielded a ϳ70-kDa protein, which was isolated, and four tryptic peptides were sequenced. One of the peptides, residues 129 -147 in Fig. 1, matched a human EST cDNA containing a 257-amino acid residue open reading frame. The EST clone was used to screen a bovine kidney cDNA library, which yielded a 2884-bp cDNA with a coding region of 2154 bp, corresponding to a protein of 718 amino acids (Fig. 1). This cDNA was identified as EXT2. Sequence analysis of the predicted protein suggested that it adopts a type II configuration typical of glycosyltransferases (10) with a short N-terminal cytoplasmic tail, a transmembrane region, and a large lumenal domain with two potential N-glycosylation sites, as has been shown for EXT1 (11). The calculated M r is 81,900, somewhat larger than the apparent M r of the purified protein. It appears that the purified bovine HS-POL is a truncated form that has lost its transmembrane domain and been subsequently released from the cell, as is well established for glycosyltransferases (10).
Expression of HS-POL and Relation to EXT Proteins-Direct evidence that the cloned bovine HS-POL cDNA encodes a protein with both GlcA-and GlcNAc-T activities was obtained by expressing the His/FLAG fusion protein in COS-7 cells. Western blots of cell lysates using anti-FLAG antibodies showed a protein product of the appropriate size (Fig. 2). The fusion protein was recovered on an anti-FLAG affinity gel and assayed for GlcA-and GlcNAc-T activities. The apparent activities were ϳ4and ϳ10-fold elevated, respectively, compared with mock-transfected controls (Fig. 3). The transferase activities displayed by control cells were probably because of endogenous enzymes committed to HS biosynthesis.
A recent report shows that defective HS biosynthesis in a mutant mouse fibroblast cell line (gro2C) could be partially rescued by transfection with EXT1 (11). We therefore decided to analyze the gro2C cells as well as the corresponding wildtype L cells for the two glycosyltransferase activities, before and after transfection with EXT1. Both activities were 5-10fold lower in the mutant than in the wild-type cell lysates (Table I), indicating that the abrogated HS biosynthesis in gro2C cells is probably caused by defects in HS-POL. Transfection of gro2C cells with EXT1 showed a ϳ2-fold increase in both enzymatic activities (Table I), which suggested that EXT1, like EXT2, harbors HS-POL activities.
By contrast, transfection of control L cells with EXT1 led to a decrease in HS-POL activities (Table I). A similar decrease in polymerase activities was noted after transfection of COS-7 cells with bovine HS-POL/EXT2, thus explaining why expression of recombinant enzyme was detectable in terms of catalytic activity only after recovery of fusion protein by immunoabsorption. These results suggested that overexpression of either EXT1 or EXT2 proteins interfered with the apparent activity in normal cells. The HS produced by such transfected cells showed a somewhat lower apparent negative charge density than corresponding control HS, as demonstrated by ion-ex- change chromatography of HS from control and EXT1-transfected L cells (Fig. 2 in Ref. 11) and HS from control and HS-POL-transfected human kidney epithelial 293 cells (Fig. 4). DISCUSSION HME is characterized by the formation of cartilage-capped tumors (exostoses), which are derived from the growth plate of endochondral bone (12) and may lead to skeletal abnormalities and short stature. Malignant transformation into chondrosarcomas (13,14) or osteosarcomas (15, 16) has been described. Three genes have been associated with this autosomal dominant disorder, EXT1 on 8q24.1, EXT2 on 11p11-13, and EXT3 on 19p (5,7,17,18). Recently, several novel genes have been identified that share significant sequence homology with the EXT genes (19 -21). Although none of these have been linked with HME, their chromosomal localizations suggest association with other forms of cancer. The findings in this report show that EXT1 and EXT2 both encode a HS-POL. It is tempting to speculate that other members of the EXT family are similarly involved in the biosynthesis of glycosaminoglycans.
The precise defect in HS biosynthesis in HME is unclear. Complete elimination of HS-POL activities would result in total absence of the polysaccharide. Partial loss of activity might lead to the formation of fewer and/or shorter chains. However, it seems likely that the polymerase interacts with one or more of the enzymes that catalyze the various modification reactions through which the nonsulfated precursor po-lysaccharide is converted into the mature, sulfated product (1,22). A mutation in the appropriate EXT protein might affect such interaction as well as the initial polymerization reaction itself, with presently unpredictable effects on the structure of the final product. Indeed, many different types of tumors are associated with distinct changes in glycosaminoglycan, particularly HS, structure (23)(24)(25)(26)(27). Although the mechanisms behind these changes are generally unknown, the alterations may be expected to affect functional interactions with a variety of proteins that are potentially involved in neoplastic transformation. Examples of such proteins include a variety of growth factors that may be functionally dependent on HS fine structure (28), growth factor receptors, extracellular matrix macromolecules (29), and HS-degrading endoglycosidase(s) (heparanase) (30). Interestingly, a Drosophila homologue of EXT1 was recently implicated with the diffusion of Hedgehog, a presumably glycosaminoglycan-dependent process in embryonic development (31).
The present findings raise intriguing questions regarding the number of EXT type HS-POLs and the functional relation between these enzymes. Notably, deletion of either EXT1 or EXT2 causes disease, suggesting that these enzymes are not able to substitute for each other. We know that several of the polymer-modifying enzymes, acting further downstream in the process, have also recently been found to occur in genetically distinct isoforms (32). It has been proposed that HS chains with specifically tailored structure (designed for interactions with defined proteins) may be generated through the appropriate combination of such isoforms in biosynthetic assembly systems (32,33). Interactions involving EXT proteins at the cellular level are inferred from the consistent down-regulation of overall HS-POL activities because of transfection of cells with either EXT1 or EXT2. Understanding the role of HS biosynthesis in relation to HME, and possibly other types of neoplastic disease, will require detailed analysis of the expression of EXT/ HS-POLs in different cells and tissues, as well as of their interaction with other components of the HS biosynthetic machinery.   . Cells were transfected with bovine HS-POL (in pcDNA3 without the His/FLAG-tag), and stable expression was generated by selection with Geneticin (G418, Life Technologies, Inc.) as described (34). Proteoglycans were extracted from cell lysates, and HS chains (free from galactosaminoglycans) were purified as described (35) except that after digestion with chondroitinase ABC, HS chains were recovered from Superose-12 (Amersham Pharmacia Biotech).