Human tumor suppressor EXT family members EXTL1 and EXTL3 encode ␣1,4- N-acetylglucosaminyltransferases that likely are involved in heparan sulfate͞ heparin biosynthesis

Byung-Taek Kim*, Hiroshi Kitagawa*, Jun-ichi Tamura†, Toshiyuki Saito‡, Marion Kusche-Gullberg§, Ulf Lindahl§, and Kazuyuki Sugahara*¶

*Department of Biochemistry, Kobe Pharmaceutical University, Higashinada-ku, Kobe 658-8558, Japan; †Department of Environmental Sciences, Faculty of Education and Regional Sciences, Tottori University, Tottori 680-8551, Japan; ‡Genome Research Group, National Institute of Radiological Sciences, Anagawa 4-9-1, Inage-ku, Chiba 263-8555, Japan; and §Department of Medical Biochemistry and Microbiology, University of Uppsala, The Biomedical Center, S-751 23 Uppsala, Sweden

Communicated by Sen-itiroh Hakomori, Pacific Northwest Research Institute, Seattle, WA, April 17, 2001 (received for review February 22, 2001) The tumor suppressors EXT1 and EXT2 are associated with hered- xylose to specific Ser residues, followed by the addition of two itary multiple exostoses and encode bifunctional glycosyltrans- Gals and GlcA, each reaction being catalyzed by the respective, ferases essential for chain polymerization of heparan sulfate (HS) specific glycosyltransferase:xylosyltransferase, Gal transferases I and its analog, heparin (Hep). Three highly homologous EXT-like and II, and GlcA transferase I (GlcAT-I) (1, 8). Chain polymer- , EXTL1–EXTL3, have been cloned, and EXTL2 is an ␣1,4- ization for HS͞Hep is initiated once an ␣-GlcNAc is transferred GlcNAc transferase I, the key enzyme that initiates the HS͞Hep to this linkage-region tetrasaccharide core by the action of synthesis. In the present study, truncated forms of EXTL1 and ␣-GlcNAc transferase I (GlcNAcT-I) (9), whereas ␤-GalNAc EXTL3, lacking the putative NH2-terminal transmembrane transfer triggers CS͞DS synthesis (10, 11). The first hexosamine and cytoplasmic domains, were transiently expressed in COS-1 cells transfer, therefore, is considered to be the critical determining and found to harbor ␣-GlcNAc transferase activity. EXTL3 used not step in which the HS͞Hep and CS͞DS chains are selected. After only N-acetylheparosan oligosaccharides that represent growing attachment of the first ␣-GlcNAc residue, the resultant nascent ␤ ␤ HS chains but also GlcA 1–3Gal 1-O-C2H4NH-benzyloxycarbonyl pentasaccharide is elongated further by alternate additions of (Cbz), a synthetic substrate for ␣-GlcNAc transferase I that deter- GlcA and GlcNAc to form HS͞Hep backbones by the actions of mines and initiates HS͞Hep synthesis. In contrast, EXTL1 used only copolymerases, which exhibit dual catalytic activities of GlcA the former acceptor. Neither EXTL1 nor EXTL3 showed any glucu- transferase II (HS-GlcAT-II) and ␣-GlcNAc transferase II ronyltransferase activity as examined with N-acetylheparosan oli- (GlcNAcT-II) (8). Such HS͞Hep precursor chains subsequently gosaccharides. Heparitinase I digestion of each transferase-reac- are modified through a series of reactions of N-deacetylation͞ tion product showed that GlcNAc had been transferred exclusively N-sulfation of GlcNAc, epimerization of GlcA, and O-sulfation through an ␣1,4-configuration. Hence, EXTL3 most likely is in- at various positions of both residues (2), which confers the volved in both chain initiation and elongation, whereas EXTL1 structural diversity on HS͞Hep. possibly is involved only in the chain elongation of HS and, maybe, Recently, two independent studies revealed the connection Hep as well. Thus, their acceptor specificities of the five family between HS-synthesizing glycosyltransferases and an EXT gene members are overlapping but distinct from each other, except for family of tumor suppressors: (i) identification of EXT1 as the EXT1 and EXT2 with the same specificity. It now has been clarified gene that is capable of restoring HS synthesis in HS-deficient that all of the five cloned human EXT gene family harbor mutant cells (12) and (ii) identification of EXT2 by direct peptide glycosyltransferase activities, which probably contribute to the sequencing of HS copolymerase purified from bovine serum synthesis of HS and Hep. (13). EXT1 and EXT2 are associated with the development of hereditary multiple exostoses (HME), an autosomal dominant eparan sulfate (HS) is a glycosaminoglycan (GAG) found disorder characterized by aberrant bone formation most com- Habundantly on the surfaces of most cells and in the extra- monly originating from the juxtaepiphyseal regions of the long cellular space as proteoglycans (1). HS proteoglycans are known bones (14–16). More recent studies have established firmly that to function as cofactors in a variety of biological processes such recombinant forms of EXT1 and EXT2 indeed have both as cell adhesion, blood coagulation, angiogenesis, morphogen- GlcNAc and GlcA transferase activities (17) and form an esis, and regulation of growth factors and cytokine effects (2). EXT1͞EXT2 heterooligomeric complex in the Golgi apparatus Most, if not all, of their biological activities are assumed to be attributable to the HS side chains that interact with diverse ligands through specific saccharide sequences (3). Re- Abbreviations: GlcNAcT-I and GlcNAcT-II, ␣-GlcNAc transferases I and II; Cbz, benzyloxy- carbonyl; CHO, Chinese hamster ovary; CS, chondroitin sulfate; DS, dermatan sulfate; EXT, cent evidence indicates that HS chains are critically involved in hereditary multiple exostoses gene; GAG, glycosaminoglycan; GlcAT-I, glucuronyltrans- the development of Drosophila melanogaster by interacting with ferase I; Hep, heparin; HS-GlcAT-II, heparan sulfate glucuronyltransferase II; CS-GlcAT-II, signaling molecules, including Hedgehog, Wingless, and fibro- chondroitin sulfate glucuronyltransferase II; HME, hereditary multiple exostoses; HS, hepa- blast growth factor, and HS deficiency causes severe malforma- ran sulfate. tion in developing fly (4, 5) and mouse (6) embryos. ¶To whom reprint requests should be addressed at: Department of Biochemistry, Kobe Pharmaceutical University, 4-19-1 Motoyamakita-machi, Higashinada-ku, Kobe 658-8558, In addition to HS and its analog, heparin (Hep), chondroitin Japan. E-mail: [email protected]. sulfate (CS) and dermatan sulfate (DS) also are synthesized on ␤ ␤ The publication costs of this article were defrayed in part by page charge payment. This the common GAG–protein linkage region, GlcA 1–3Gal 1– article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. 3Gal␤1–4Xyl␤1-O-Ser (7), which is formed by the addition of §1734 solely to indicate this fact.

7176–7181 ͉ PNAS ͉ June 19, 2001 ͉ vol. 98 ͉ no. 13 www.pnas.org͞cgi͞doi͞10.1073͞pnas.131188498 Downloaded by guest on September 27, 2021 (17–19) to yield a more active, biologically relevant enzyme form in the fusion protein that had the cleavable insulin signal (17, 18). It should be noted that deficiency of either EXT1 or sequence for secretion and protein A for purification. EXT2 causes HME (15, 16). These findings indicate that both The cDNA fragment encoding a truncated form of EXTL3 EXT1 and EXT2 are essential glycosyltransferases for HS bio- lacking the NH2-terminal first 90 aa was amplified with the synthesis. previously cloned EXTL3 cDNA (23) as a template by PCR by However, three EXT-like genes homologous to EXT genes using a 5Ј primer (5Ј-CGAGATCTGAAGAGCTCCTG- have been isolated: EXTL1, EXTL2͞EXTR2, and EXTL3͞ CAGCTGGA-3Ј) containing an in-frame BglII site and 3Ј EXTR1 (20–23). Their chromosomal locations imply that they primer (5Ј-CGAGATCTTCTGCTCATCCTCTTCCCCAG-3Ј) also might be tumor suppressors (20, 21, 24, 25), although they containing a BglII site located 37 bp downstream of the stop are not linked with HME. EXTL2 protein is an N-acetyl- codon. PCR was carried out by using Pfu polymerase with 30 hexoaminyltransferase that transfers not only GalNAc but also cycles of 95°C for 30 sec, 56°C for 30 sec, and 72°C for 300 sec. GlcNAc to the common linkage-region core tetrasaccharide The amplified fragments were digested with BglII and cloned through ␣1,4-linkage and is a strong candidate for the key into the BamHI site of pEF-BOS͞IP. enzyme leading to HS͞Hep biosynthesis, segregating it from CS͞DS biosynthesis (26). Identification of EXTL2 as GlcNAcT-I Expression of the Soluble Forms of EXTL1 and EXTL3 in COS-1 Cells. for HS͞Hep synthesis suggested that other EXT-like genes also Each expression plasmid (6.7 ␮g) was transfected into COS-1 have some roles in HS formation. In the present study, soluble cells on 100-mm plates by using FuGENE6 (Roche Molecular forms of EXTL1 and EXTL3 were transiently expressed in Biochemicals, Tokyo) according to the instructions supplied by COS-1 cells and their glycosyltransferase activities were the manufacturer. Two days after transfection, 1 ml of the revealed. medium was collected and the secreted fusion protein was purified by incubation with 10 ␮l of IgG-Sepharose (Amersham Materials and Methods Pharmacia) for2hat4°C. The beads recovered by centrifugation Materials. UDP-[3H]GlcNAc (60 Ci͞mmol), UDP-[U-14C]GlcA were washed with each assay buffer described below, resus- (285.2 mCi͞mmol), and UDP-[3H]GalNAc (10 Ci͞mmol) were pended in the same buffer, and then tested for glycosyltrans- purchased from NEN. Unlabeled UDP-GlcNAc, UDP-GlcA, ferase activities. For GlcNAc transferase assays, the reaction and UDP-GalNAc were supplied by Sigma. Flavobacterium mixture of a total volume of 20 ␮l contained 10 ␮lofthe heparinum heparitinase I, jack bean ␤-N-acetylhexosaminidase, resuspended beads, 100 mM 2-(N-morpholino)ethanesulfonic ␤ and chondroitin were obtained from Seikagaku Kogyo (Tokyo). acid-NaOH (pH 6.5), 10 mM MgCl2, 1 mM ATP, GlcA 1– ␤ N-Acetylchondrosine was a gift from Keiichi Yoshida (Seika- 3Gal 1-O-C2H4NHCbz (250 nmol), or tetrasaccharide-serine gaku Kogyo). The linkage-region tetrasaccharide-serine, (1 nmol) for GlcNAcT-I assay (31) or N-acetylheparosan oligo- GlcA␤1–3Gal␤1–3Gal␤1–4Xyl␤1-O-Ser, was synthesized chem- saccharides with the nonreducing terminal GlcA (20 ␮g) for ically as described previously (27). N-Acetylheparosan oligosac- GlcNAcT-II assay (32) as acceptors, and 250 ␮M UDP- charide acceptors derived from the capsular polysaccharide [3H]GlcNAc (about 1.1 ϫ 106 dpm). Enzyme assays for GlcAT-I of Escherichia coli K5, GlcA␤1–4GlcNAc␣1-(4GlcA␤1– (33), HS-GlcAT-II (32), and ␣-GalNAc transferase (34), ␤- 4GlcNAc␣1-)n with the nonreducing terminal GlcA and GalNAc transferase for CS chain elongation (35), and GlcA GlcNAc␣1-(4GlcA␤1–4GlcNAc␣1-)n with the nonreducing transferase for CS chain elongation (CS-GlcAT-II) (8, 36) were terminal GlcNAc, were prepared as described previously carried out as described previously. ␤ ␤ (28). GlcA 1–3Gal 1-O-C2H4NHCbz was synthesized chemically (H. Shimakawa, Y. Kano, H.K., J.T., J. D. Esko, and Identification of the Enzyme-Reaction Products. The products from ϫ ␤ ␤ K.S., unpublished results). A Nova-Pak C18 column (3.9 150 the GlcNAcT-I reaction when using GlcA 1–3Gal 1-O- ͞ mm) and a Superdex peptide HR 10 30 column were supplied C2H4NHCbz were isolated by hydrophobic HPLC on a Nova-Pak by Waters and Amersham Pharmacia, respectively. C18 column in an LC-10AS system (Shimadzu) (31). The column ͞ was developed with H2O for 15 min at a flow rate of 1.0 ml min Construction of the Expression Vector pEF-BOS͞IP. An 894-bp cDNA at room temperature; thereafter, a linear gradient was applied to fragment containing the insulin signal sequence and the protein increase the methanol concentration from 0% to 100% over a A sequence was cleaved from pGIR201protA (29) by NheI 5-min period. The column then was developed isocratically with digestion and then inserted into the XbaI site of the expression 100% methanol for 20 min. The radioactive peak containing the vector pEF-BOS (30) to prepare pEF-BOS͞IP. A DNA insert of product was pooled and evaporated to dryness. The dried sample interest was ligated into the BamHI site derived from (about 40 pmol) was incubated with either 9 milliunits of pGIR201protA for expression as described below. ␤-N-acetylhexosaminidase in a total volume of 20 ␮lof50mM

sodium citrate buffer (pH 4.5) or with 10 milliunits of hepariti- BIOCHEMISTRY Construction of Soluble Forms of EXTL1 and EXTL3.The cDNA frag- nase I in a total volume of 50 ␮l of 20 mM sodium acetate buffer ment encoding a truncated form of EXTL1, lacking the NH2- (pH 7.0) containing 2 mM calcium acetate at 37°C overnight. terminal first 75 aa, was amplified with the first-strand cDNA Each enzyme digest was analyzed by using the same Nova-Pak ϩ transcribed from human adult brain poly (A) RNA (Biochain C18 column as noted above. Institute, San Leandro, CA) as a template by PCR by using a 5Ј GlcNAcT-II-reaction products of EXTL1 and EXTL3 were primer (5Ј-ATTGATCACATGGTGGCAGCTGCAACTG-3Ј) chromatographed individually by gel-filtration on a Superdex containing an in-frame BclI site and a 3Ј primer (5Ј- peptide HR 10͞30 column (10 ϫ 30 mm) by using 0.2 M Ј CGTGATCAGCTGGCTGACATGATAGGTT-3 ) containing NH4HCO3 as an eluant, and the fractions containing radioactive a BclI site located 115 bp downstream of the stop codon. PCR products were pooled and evaporated until the ammonium salt was carried out with Pfu polymerase by 30 cycles of 95°C for 30 became invisible with the intermittent addition of H2O. An sec, 52°C for 30 sec, and 72°C for 240 sec. A fragment of the aliquot (about 36 pmol) of each purified product was incubated expected size was excised, purified, and ligated into the pGEM-T with either 9 milliunits of ␤-N-acetylhexosaminidase in a total Easy vector (Promega), and then the recombinant vector was volume of 20 ␮l of 50 mM sodium citrate buffer (pH 4.5) or with introduced into the Subcloning Efficiency DM1 competent cells 10 milliunits of heparitinase I in a total volume of 100 ␮lof20 (Life Technologies, Grand Island, NY). The inserted EXTL1 mM sodium acetate buffer (pH 7.0) containing 2 mM calcium fragment was recovered by digestion with BclI and religated into acetate at 37°C overnight. Each digest was analyzed again by the BamHI site of the expression vector pEF-BOS͞IP, resulting using the same Superdex peptide column as described above.

Kim et al. PNAS ͉ June 19, 2001 ͉ vol. 98 ͉ no. 13 ͉ 7177 Downloaded by guest on September 27, 2021 Table 1. Acceptor substrate specificity of glycosyltransferases attributable to soluble, truncated forms of EXTL1 and EXTL3 EXTL1 EXTL3

Activity,* pmol͞ml Acceptor sugar [Donor] medium per h

GlcA␤1-3Gal␤1-O-C2H4NHCbz [UDP-GlcNAc] ND 746 GlcA␤1-3Gal␤1-3Gal␤1-4Xyl␤1-O-Ser ND ND [UDP-GlcNAc] N-Acetylheparosan oligosaccharides with the 44࿣ 219࿣ nonreducing terminal GlcA† [UDP-GlcNAc] N-Acetylheparosan oligosaccharides with the ND** ND** nonreducing terminal GlcNAc‡ [UDP-GlcA] Chondroitin§ [UDP-GalNAc] ND ND Chondroitin§ [UDP-GlcA] ND ND Gal␤1-3Gal␤1-4Xyl [UDP-GlcA] ND ND Fig. 1. Characterization of EXTL1 GlcNAcT-II-reaction products. GlcNAcT-II- N-Acetylchondrosine¶ [UDP-GalNAc] ND ND reaction products obtained with EXTL1 were digested with ␤-N-acetylhex- osaminidase (ᮀ) or heparitinase I (■), and each digested or undigested sample ND, not detected (Ͻ0.5 pmol͞ml medium per h). (E) was applied to a column of Superdex peptide HR 10͞30 (10 ϫ 30 mm) and *The values represent the averages of two independent experiments. chromatographed as described in Materials and Methods. The separated †GlcA␤1-4GlcNAc␣1-(4GlcA␤1-4GlcNAc␣1-)n. fractions (0.4 ml each) were tested for radioactivity. An arrow indicates the ‡GlcNAc␣1-(4GlcA␤1-4GlcNAc␣1-)n. elution position of free GlcNAc. §A mixture of GlcA␤1-(3GalNAc␤1-4GlcA␤1-)n and GalNAc␤1-(4GlcA␤1- 3GalNAc␤1-)n. Under the conditions, the chondroitin-synthesizing en- zyme system (11, 35, 36) as a positive control showed marked activities of filtration and subjected to digestion with heparitinase I and CS-GlcAT-II and GalNAcT-II. ␤-N-acetylhexosaminidase separately. The labeled products ¶GlcA␤1-3GalNAc. ࿣ were digested completely by heparitinase I, which cleaves only an When truncated EXT1 was used for the positive control, the measured activity ␣ was 539 pmol͞ml medium per h. 1,4-glucosaminyl glucuronate linkage (37, 38), and the original **When truncated EXT1 was used for the positive control, the measured radioactivity peak eluted near the void volume was shifted to the 3 activity was 111 pmol͞ml medium per h. free [ H]GlcNAc position, as shown in Fig. 1. ␤-N-Acetylhex- osaminidase did not release the radioactive monosaccharide at all. These findings indicate that GlcNAc had been transferred by Results EXTL1 protein through an ␣1,4-linkage to the nonreducing EXTL1 Is a Likely ␣1,4-GlcNAc Transferase for the Synthesis of the terminal GlcA of N-acetylheparosan oligosaccharides, repre- Repeating Disaccharide Region of HS. The soluble form of EXTL1, senting growing HS chains. Hence, EXTL1 encodes novel which lacks the putative transmembrane domain and is fused GlcNAcT-II, which likely is involved in the synthesis of the re- with the cleavable insulin signal and protein A sequences, was peating disaccharide region of HS͞Hep. transiently expressed in COS-1 cells. The expressed fusion protein was purified for separation from endogenous glycosyl- EXTL3 Protein Harbors Both GlcNAcT-I and GlcNAcT-II Activities, Which transferases by incubating with IgG-Sepharose, which specifi- Likely Are Required for HS Synthesis. The expression plasmid, which cally binds to protein A. Its sugar transferase activities were contains nucleotide sequences coding for the truncated EXTL3 examined by using various oligosaccharides as acceptors and fused with the insulin signal and protein A, was transfected into radioactive UDP-sugars as donors. A significant amount of COS-1 cells, and the secreted EXTL3 fusion protein was exam- GlcNAc was transferred from UDP-[3H]GlcNAc to N- ined for glycosyltransferase activities after purification with acetylheparosan oligosaccharides GlcA␤1–4GlcNAc␣1- IgG-Sepharose. Recombinant EXTL3 transferred GlcNAc effi- (4GlcA␤1–4GlcNAc␣1-)n (Table 1). In contrast, the enzyme did ciently to N-acetylheparosan oligosaccharides with the nonre- not transfer GlcA from UDP-[14C]GlcA to N-acetylheparosan ducing terminal GlcA (Table 1). To clarify whether it was a oligosaccharides GlcNAc␣1-(4GlcA␤1–4GlcNAc␣1-)n or to the copolymerase, N-acetylheparosan oligosaccharides with the linkage region trisaccharide Gal␤1–3Gal␤1–4Xyl, which is the nonreducing terminal GlcNAc were used as acceptors for testing acceptor for GlcAT-I (33). Neither GlcA␤1–3Gal␤1–3Gal␤1– HS-GlcAT-II activity. However, it showed no HS-GlcAT-II ␤ ␤ ␤ ␤ ␤ 4Xyl 1-O-Ser nor GlcA 1–3Gal 1-O-C2H4NHCbz, a substrate activity and no GlcAT-I activity toward Gal 1–3Gal 1–4Xyl for GlcNAcT-I (31), was used as an acceptor. Because mamma- (Table 1). lian chondroitin-synthesizing enzymes have not been cloned, The products from GlcNAcT-II reaction were characterized ␤-GalNAc and GlcA transferase activities (35, 36) were mea- by treatments with heparitinase I and ␤-N-acetylhexosamini- sured by using chondroitin as an acceptor and UDP-GalNAc or dase. On heparitinase I digestion, the radioactivity was released UDP-GlcA as a donor substrate, but no activity was detected. as free [3H]GlcNAc from the reaction products eluted near the N-Acetylchondrosine, GlcA␤1–3GalNAc, which is an acceptor void volume, as demonstrated by gel-filtration chromatography substrate for the unique ␣-GalNAc transferase activity of (Fig. 2A). In contrast, the 3H-labeled products were entirely EXTL2 (26, 34), did not serve as an acceptor either. The above insensitive to the ␤-N-acetylhexosaminidase treatment. These ␣-GlcNAc transferase activity of EXTL1 was confirmed by the findings indicate that EXTL3 transferred GlcNAc to N- control transfection experiment, where no transferase activity acetylheparosan oligosaccharides via the ␣1,4-linkage. was detected in the medium from COS-1 cells transfected with Surprisingly, EXTL3, unlike EXTL1, exhibited marked the empty expression vector without EXTL1. Thus, EXTL1 GlcNAcT-I activity as detected by using GlcA␤1–3Gal␤1- harbored GlcNAcT-II activity that can add GlcNAc to N- O-C2H4NHCbz. However, the linkage-region tetrasaccharide- acetylheparosan oligosaccharides, but no other catalytic activi- serine GlcA␤1–3Gal␤1–3Gal␤1–4Xyl␤1-O-Ser with the native ties were tested. structure did not serve as an acceptor. This apparent discrepancy To identify the EXTL1-reaction products, N-acetylheparosan was taken to indicate that the enzyme required an appropriately oligosaccharides labeled with [3H]GlcNAc were isolated by gel hydrophobic peptide sequence in the vicinity of the linkage

7178 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.131188498 Kim et al. Downloaded by guest on September 27, 2021 Table 2. Summary of glycosyltransferase activities detected in EXT family members EXT1* EXT2* EXTL1 EXTL2† EXTL3

GlcNAcT-I‡ ϪϪϪϩϩ GlcNAcT-II§ ϩϩϩϪϩ HS-GlcAT-II¶ ϩϩϪϪϪ ␣-GalNAcT࿣ ϪϪϪϩϪ

*Based on refs. 13, 17, and 18. †Based on refs. 26 and 31. ‡ GlcA␤1-3Gal␤1-O-C2H4NHCbz was used as the acceptor substrate. §N-acetylheparosan oligosaccharides with the nonreducing terminal GlcA were used as the acceptor substrate. ¶N-acetylheparosan oligosaccharides with the nonreducing terminal GlcNAc were used as the acceptor substrate. ࿣N-acetylchondrosine was used as the acceptor substrate.

GlcNAcT-I and GlcNAcT-II reactions, which indicates that EXTL3 most likely is involved in initiation and elongation of the HS͞Hep chain. Discussion In this study, both EXTL1 and EXTL3, like the other EXT members, were demonstrated to be novel ␣1,4-GlcNAc trans- ferases that likely are involved in the biosynthesis of HS and probably Hep. These findings are consistent with the observation that they have high in the C-terminal regions with the other family members, EXT1, EXT2, and EXTL2, all of which harbor ␣1,4-GlcNAc transferase activities (8), and that EXTL1 and EXTL3, like the other family members, have the DXD motif common to most glycosyltransferases (39). It is unlikely that the observed enzyme activities of EXTL1 and Fig. 2. Characterization of GlcNAcT-I- and GlcNAcT-II-reaction products EXTL3 were a result of endogenous EXT1 or EXT2, which may produced by EXTL3. (A) GlcNAcT-II-reaction products obtained with EXTL3 form a complex, as will be discussed below; HS-GlcAT-II were digested with ␤-N-acetylhexosaminidase (ᮀ) or heparitinase I (■), and activity, which might be expected for such complexes, was not each digested or undigested sample (E) was applied to a column of Superdex detected for either EXTL1 or EXTL3 upon their indi- peptide HR 10͞30 (10 ϫ 30 mm) and chromatographed as described in Mate- vidual overexpression. EXTL3 showed both GlcNAcT-I and rials and Methods. The separated fractions (0.4 ml each) were tested for GlcNAcT-II activities, which appear to be involved in the bio- radioactivity. An arrow indicates the elution position of free GlcNAc. (B) EXTL3 synthetic initiation and elongation, respectively, of HS͞Hep GlcNAcT-I-reaction products were subjected to digestion with ␤-N-acetylhex- osaminidase (ᮀ) or heparitinase I (■), and each digested or undigested sample precursor chains, whereas EXTL1 showed only GlcNAcT-II activity, which may be involved in HS chain elongation. Thus, the (E) was analyzed by HPLC on a Nova-Pak C18 column as described in Materials and Methods. The separated fractions (2 ml each) were measured for radio- gene products of all five EXT family members now have been activity. An arrow indicates the elution position of free GlcNAc. revealed to harbor glycosyltransferase activities, which can contribute to HS͞Hep biosynthesis, as summarized in Table 2 and Fig. 3. region, which can be substituted by an artificial aglycon such as Interestingly, the acceptor specificities of these glycosyltrans- C2H4NHCbz (see Discussion). Although EXTL3 resembles ferase activities are overlapping among multiple members, but EXTL2 in that it exhibited GlcNAcT-I activity toward the above distinct from each other. EXT1 and EXT2 are copolymerases linkage-region analog, EXTL3, unlike EXTL2, showed no with both GlcNAcT-II and HS-GlcAT-II activities for chain ␣-GalNAc transferase activity toward N-acetylchondrosine (26) elongation (13, 17) and form a heterooligomeric complex to as shown in Table 1, which clearly distinguishes EXTL3 from exhibit full enzyme activities of biological relevance (17–19). BIOCHEMISTRY EXTL2. Chondroitin was not used as an acceptor when either UDP-[3H]GalNAc or UDP-[14C]GlcA was included as a donor (Table 1). The observed GlcNAcT-I activity toward the synthetic substrate was not detected when using the culture medium obtained from the control transfection, confirming that the enzyme activity was attributable to the EXTL3 protein. The radioactivity of the GlcNAcT-I-reaction products, which had been retained on a C18 column, also was released by digestion with heparitinase I and eluted in the flow-through fraction as free [3H]GlcNAc, as demonstrated in Fig. 2B. ␤-N-Acetylhexosaminidase digestion released no radioactivity. These findings indicate that EXTL3 could transfer GlcNAc not only to N-acetylheparosan oligosaccharides but also to GlcA␤1– ␤ ␣ 3Gal 1-O-C2H4NHCbz via the 1,4-linkage. Thus, it was sug- gested that GlcNAc residues were transferred to the nonreducing termini of the acceptors exclusively via the ␣1,4-linkage in both Fig. 3. GAG biosynthesis and the EXT proteins.

Kim et al. PNAS ͉ June 19, 2001 ͉ vol. 98 ͉ no. 13 ͉ 7179 Downloaded by guest on September 27, 2021 Although they resemble each other in the acceptor specificity, and͞or Hep chains possibly might be initiated and polymerized they cannot compensate for the deficiency of the other and are efficiently on the tetrasaccharide core. However, such a not functionally redundant, as is evident from the finding that heterooligomeric complex formation has not been reported. deficiency of either one results in HME syndrome (15, 16). EXTL1 has only GlcNAcT-II activity and possibly is involved EXTL2 exhibits GlcNAcT-I activity that transfers only the first in the HS and͞or Hep chain elongation. However, it has been ␣1,4-GlcNAc residue to determine and initiate HS͞Hep chain reported that HS synthesis is abolished in a CHO cell mutant synthesis on the common GAG–protein linkage region in the pgsD-677 (28), which now has been identified as an EXT1- core proteins, although the enzyme activity can be demonstrated deficient mutant (51). A loss-of-function mutation in the mouse only by using a synthetic linkage-region analog, GlcA␤1– of EXT1 results in embryonic lethality (6), and HS synthesis is 3Gal␤1-O-naphthalenmethanol (26) or GlcA␤1–3Gal␤1-O- also abolished in EXT1Ϫ͞Ϫ ES cells and decreases to less than ϩ͞Ϫ C2H4NHCbz, as an acceptor (31), but not with the authentic 50% in cell lines. These findings appear to indicate that tetrasaccharide-serine (26, 31). EXTL3 showed both GlcNAcT-I EXT1 is primarily responsible for HS formation, especially in and GlcNAcT-II activities and, thus, is different from EXTL1, CHO cells and in early mouse development, and the HS synthesis which has only the latter enzyme activity (Table 1), and from the deficiency cannot be compensated by the other EXT members. GlcNAcT-I reported for Chinese hamster ovary (CHO) cells (9). Although the exact role in vivo of EXTL1 is unclear because of EXTL3 is also different from EXT1 and EXT2 in that it has the lack of mutants, it may function at later developmental no HS-GlcAT-II activity or polymerizing activity. It shares stages. Indeed, it recently has been demonstrated by whole- GlcNAcT-I activity with EXTL2, but is clearly distinct from the mount in situ hybridizations that EXT1, EXT2, and EXTL1 show latter as well because EXTL3 possesses GlcNAcT-II activity in nearly identical expression patterns at stages E10.5 and 11.5 addition to GlcNAcT-I activity. It also should be noted that during mouse embryogenesis, but some differences become EXTL3 does not harbor ␣-GalNAc transferase activity, which apparent at E13.5, and EXTL1 exhibits an expression profile EXTL2 possesses (26), although the physiological significance of distinct from that of EXT1 and EXT2 at E14.5 (52). Although this enzyme activity remains to be clarified. GlcNAcT-II enzyme activity is negligible in EXT1-deficient EXT genes have been well conserved between humans, D. CHO mutants including pgsD677 (28, 51), it is unknown to our melanogaster (ttv) (40), and C. elegans (rib-1, rib-2) (41). Recent knowledge whether EXTL1 as well as EXTL3 is expressed in studies have shown that HS and CS are also present in these wild-type CHO cells. Unlike other EXT gene family members, genetically tractable model animals (42, 43), and a total amount which are expressed ubiquitously, human EXTL1 is being ex- of HS drastically decreases in ttv-defective D. melanogaster (44), pressed in limited areas such as skeletal muscles, brain, and heart which indicates that EXT homologs contribute to HS biosynthe- (20). The expression is strongly detected in the liver as well as in sis and are indispensable in these animals also. Recently, we the above tissues in the mouse but not in human Northern blots found that rib-2 harbors both GlcNAcT-I and GlcNAcT-II (52). Thus, EXTL1 may play its roles in a tissue-specific and activities, and its acceptor specificity resembled that of EXTL3 spatiotemporally regulated manner. It also should be noted that (31). In view of the higher homology of rib-2 to EXTL3 (58%) there is substantial evidence that a gene encoding a putative than EXT2 (44%), rib-2 appears to be a C. elegans EXTL3 tumor suppressor lies in the chromosomal locus (1p36.1) of ortholog. Because rib-1 and rib-2 are the only genes identified EXTL1, and consistent deletions of the region have been de- as EXT homologs in C. elegans, mutations in the rib-2 gene tected in various tumors (see references in ref. 20), which also would cause severe developmental defects as ttv does in D. possibly indicates indispensable physiological functions of melanogaster. EXTL1. EXTL3, like EXTL2, transfers the first ␣1,4-GlcNAc residue It remains to be clarified how the GlcNAcT-II activity of to initiate HS͞Hep chains on the common linkage region EXTL1 or the similar GlcNAcT-II activity of EXTL3 plays its segregating HS and Hep synthesis from CS and DS synthesis. A respective role in chain polymerization of HS͞Hep, where it possible explanation for the existence of these two distinct requires HS-GlcAT-II as a partner. It is unknown which one of GlcNAcT-I molecular species for this key enzyme activity, which EXT1 or EXT2 cooperates with EXTL1 and͞or EXTL3 for the is critical for the sorting mechanisms, may be that they initiate synthesis of the repeating disaccharide-region or chain polymer- HS and͞or Hep chains on different core proteins by discrimi- ization or whether an as yet unidentified HS-GlcAT-II with no nating the amino acid sequences. This hypothesis is consistent homology exists in addition to EXT1 and EXT2. Human EXT1 with the notion of the importance of amino acid sequences, and EXT2 form a stable complex that accumulates in the Golgi which was proposed previously by Esko et al. (45) based on the apparatus (17–19) to exhibit substantially higher glycosyltrans- specificity of GlcNAcT-I in CHO cells. ␤-D-Xylosides with ferase activities than EXT1 or EXT2 alone (17, 18). It was once lipophilic aglycons containing two fused aromatic rings such as reported that neither EXTL3 nor EXTL2 interacts with EXT1 estradiol and naphthalene efficiently initiate HS biosynthesis, or EXT2 upon concomitant overexpression and the intracellular whereas CS chains are initiated predominantly on less lipophilic locations of the latter proteins do not change from the endo- xylosides (46, 47). Acidic amino acids close to the HS attachment plasmic reticulum to the Golgi apparatus in contrast to the sites are essential for HS glycanation (48), and a nearby hydro- reported translocation of EXT1 and EXT2 upon their complex phobic residue and repetitive Ser-Gly sequences work as en- formation (18). However, a possibility should be reexamined in hancers for HS attachment (49). These observations altogether light of the revealed catalytic activities of EXTL1 and EXTL3 indicate the critical roles of amino acid sequences in the prox- whether they interact with EXT1 and͞or EXT2. A possibility of imity of HS and Hep attachment sites of the core protein for multimeric complex formation also exists. In this context, it HS͞Hep chain initiation. Our observations that EXTL2 and should be remembered that HS chain polymerizing enzyme EXTL3 transferred a GlcNAc residue to chemically synthesized activity has not been demonstrated in vitro for an EXT1͞EXT2 disaccharides with a hydrophobic aglycon, but not to GlcA␤1– heterooligomeric complex. Alternatively, the single GlcNAc 3Gal␤1–3Gal␤1–4Xyl␤1-O-Ser, are consistent with this con- transfer catalyzed by EXTL1 and EXTL3 possibly breaks up the cept. In addition, the importance of distant regions from the elongation process and, in effect, serves as a chain-termination GAG attachment sites also has been emphasized recently in the mechanism. preferential addition of HS chains (50): the cysteine-rich glob- Although the three EXT-like gene products, all of which are ular domain of glypican-1 is as much indispensable as the GAG likely to be involved in some way in HS͞Hep synthesis, it remains attachment region to direct HS synthesis. If either EXTL3 or to be established how they share the respective roles and EXTL2 becomes associated with either EXT1 or EXT2, HS cooperate in development, pathophysiology and tumorigenesis,

7180 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.131188498 Kim et al. Downloaded by guest on September 27, 2021 and how their expressions are regulated at the transcriptional, the Science Research Promotion Fund of the Japan Private School translational and enzyme levels. Promotion Foundation and the Uehara Memorial Foundation (to H.K.), a grant-in-aid for Scientific Research on Priority Areas 10178102 (to K.S.) from the Ministry of Education, Science, Sports, We thank Dr. Kazunori Tsuchida for preparation of the vector and Culture of Japan, Grant 13401 from the Swedish Medical Re- pEF-BOS͞IP and Dr. Tomoya Ogawa (The Institute of Physical and search Council, and by the Human Frontier Science Program. B.-T.K. Chemical Research, Japan) for the synthetic tetrasaccharide-serine is a recipient of a postdoctoral fellowship from the Japan Society for and encouragement. This work was supported in part by grants from the Promotion of Science.

1. Rode´n, L. (1980) in The Biochemistry of Glycoproteins and Proteoglycans, ed. 26. Kitagawa, H., Shimakawa, H. & Sugahara, K. (1999) J. Biol. Chem. 274, Lennarz, W. J. (Plenum, New York), pp. 267–371. 13933–13937. 2. Bernfield, M., Go¨tte, M., Park, P. W., Reizes, O., Fitzgerald, M. L., Lincecum, 27. Tamura, J., Neumann, K. W. & Ogawa, T. (1996) Liebigs Ann. 1239–1257. J. & Zako, M. (1999) Annu. Rev. Biochem. 68, 729–777. 28. Lidholt, K., Weinke, J. L., Kiser, C. S., Lugemwa, F. N., Bame, K. J., Cheifetz, 3. Salmivirta, M., Lidholt, K. & Lindahl, U. (1996) FASEB J. 10, 1270–1279. S., Massague´, J., Lindahl, U. & Esko, J. D. (1992) Proc. Natl. Acad. Sci. USA 4. Perrimon, N. & Bernfield, M. (2000) Nature (London) 404, 725–728. 89, 2267–2271. 5. Lander, A. D. & Selleck, S. B. (2000) J. Cell. Biol. 148, 227–232. 29. Kitagawa, H. & Paulson, J. C. (1994) J. Biol. Chem. 269, 1394–1401. 6. Lin, X., Wei, G., Shi, Z., Dryer, L., Esko, J. D., Wells, D. E. & Matzuk, M. M. 30. Mizushima, S. & Nagata, S. (1990) Nucleic Acids Res. 18, 5322. (2000) Dev. Biol. 224, 299–311. 31. Kitagawa, H., Egusa, N., Tamura, J., Kusche-Gullberg, M., Lindahl, U. & 7. Lindahl, U. & Rode´n, L. (1972) in Glycoprotein, ed. Gottschalk, A. (Elsevier Sugahara, K. (2001) J. Biol. Chem. 276, 4834–4388. Science, Amsterdam), pp. 491–517. 32. Lind, T., Lindahl, U. & Lidholt, K. (1993) J. Biol. Chem. 268, 20705–20708. 8. Sugahara, K. & Kitagawa, H. (2000) Curr. Opin. Struct. Biol. 10, 518–527. 33. Kitagawa, H., Tone, Y., Tamura, J., Neumann, K. W., Ogawa, T., Oka, S., 9. Fritz, T. A., Gabb, M. M., Wei, G. & Esko, J. D. (1994) J. Biol. Chem. 269, Kawasaki, T. & Sugahara, K. (1998) J. Biol. Chem. 273, 6615–6618. 28809–28814. 10. Rohrmann, K., Niemann, R. & Buddecke, E. (1985) Eur. J. Biochem. 148, 463–469. 34. Kitagawa, H., Kano, Y., Shimakawa, H., Goto, F., Ogawa, T., Okabe, H. & 11. Nadanaka, S., Kitagawa, H., Goto, F., Tamura, J., Neumann, K. W., Ogawa, T. Sugahara, K. (1999) Glycobiology 9, 697–703. & Sugahara, K. (1999) Biochem. J. 340, 353–357. 35. Kitagawa, H., Tsutsumi, K., Ujikawa, M., Goto, F., Tamura, J., Neumann, K., 12. McCormick, C., Leduc, Y., Martindale, D., Mattison, K., Esford, L. E., Dyer, Ogawa, T. & Sugahara, K. (1997) Glycobiology 7, 531–537. A. P. & Tufaro, F. (1998) Nat. Genet. 19, 158–161. 36. Kitagawa, H., Ujikawa, M., Tsutsumi, K., Tamura, J., Neumann, K. W., Ogawa, 13. Lind, T., Tufaro, F., McCormick, C., Lindahl, U. & Lidholt, K. (1998) J. Biol. T. & Sugahara, K. (1997) Glycobiology 7, 905–911. Chem. 273, 26265–26268. 37. Yamada, S., Sakamoto, K., Tsuda, H., Yoshida, K., Sugahara, K., Khoo, K. H., 14. Solomon, L. (1964) Am. J. Hum. Genet. 16, 351–363. Morris, H. R. & Dell, A. (1994) Glycobiology 4, 69–78. 15. Hecht, J. T., Hogue, D., Strong, L. C., Hansen, M. F., Blanton, S. H. & Wagner, 38. Sugahara, K., Tohno-oka, R., Yamada, S., Khoo, K.-H., Morris, H. R. & Dell, M. (1995) Am. J. Hum. Genet. 56, 1125–1131. A. (1994) Glycobiology 4, 535–544. 16. Raskind, W. H., Conrad, E. U., Chansky, H. & Matsushita, M. (1995) Am. J. 39. Breton, C. & Imberty, A. (1999) Curr. Opin. Struct. Biol. 9, 563–571. Hum. Genet. 56, 1132–1139. 40. Bellaiche, Y., The, I. & Perrimon, N. (1998) Nature (London) 394, 85–88. 17. Senay, C., Lind, T., Muguruma, K., Tone, Y., Kitagawa, H., Sugahara, K., 41. Clines, G. A., Ashley, J. A., Shah, S. & Lovett, M. (1997) Genome Res. 7, Lidholt, K., Lindahl, U. & Kusche-Gullberg, M. (2000) EMBO Rep. 1, 282–286. 359–367. 18. McCormick, C., Duncan, G., Goutsos, K. T. & Tufaro, F. (2000) Proc. Natl. 42. Yamada, Y., Van Die, I., Van den Eijnden, D. H., Yokota, A., Kitagawa, H. Acad. Sci. USA 97, 668–673. & Sugahara, K. (1999) FEBS Lett. 459, 327–331. 19. Kobayashi, S., Morimoto, K., Shimizu, T., Takahashi, M., Kurosawa, H. & 43. Toyoda, H., Kinoshita-Toyoda, A. & Selleck, S. B. (2000) J. Biol. Chem. 275, Shirasawa, T. (2000) Biochem. Biophys. Res. Commun. 268, 860–867. 2269–2275. 20. Wise, C. A., Clines, G. A., Massa, H., Trask, B. J. & Lovett, M. (1997) Genome 44. Toyoda, H., Kinoshita-Toyoda, A., Fox, B. & Selleck, S. B. (2000) J. Biol. Chem. Res. 7, 10–16. 275, 21856–21861. 21. Wuyts, W., Van Hul, W., Hendrickx, J., Speleman, F., Wauters, J., De Boulle, 45. Esko, J. D. & Zhang, L. (1996) Curr. Opin. Struct. Biol. 6, 663–670. K., Van Roy, N., Van Agtmael, T., Bossuyt, P. & Willems, P. J. (1997) Eur. J. 46. Fritz, T. A., Lugemwa, F. N., Sarkar, A. K. & Esko, J. D. (1994) J. Biol. Chem. Hum. Genet. 5, 382–389. 269, 22. Van Hul, W., Wuyts, W., Hendrickx, J., Speleman, F., Wauters, J., De Boulle, 300–307. K., Van Roy, N., Bossuyt, P. & Willems, P. J. (1998) Genomics 47, 230–237. 47. Fritz, T. A., Agrawal, P. K., Esko, J. D. & Krishna, N. R. (1997) Glycobiology 23. Saito, T., Seki, N., Yamauchi, M., Tsuji, S., Hayashi, A., Kozuma, S. & Hori, 7, 587–595. T. (1998) Biochem. Biophys. Res. Commun. 243, 61–66. 48. Zhang, L. & Esko, J. D. (1994) J. Biol. Chem. 269, 19295–19299. 24. Wuyts, W., Spieker, N., Van Roy, N., De Boulle, K., De Paepe, A., Willems, 49. Zhang, L., David, G. & Esko, J. D. (1995) J. Biol. Chem. 270, 27127–27135. P. J., Van Hul, W., Versteeg, R. & Speleman, F. (1999) Cytogenet. Cell Genet. 50. Chen, R. L. & Lander, A. D. (2001) J. Biol. Chem. 276, 7507–7517. 86, 267–270. 51. Wei, G., Bai, X., Gabb, M. M. G., Bame, K. J., Koshy, T. I., Spear, P. G. & Esko, 25. Arai, T., Akiyama, Y., Nagasaki, H., Murase, N., Okabe, S., Ikeuchi, T., Saito, J. D. (2000) J. Biol. Chem. 275, 27733–27740. K., Iwai, T. & Yuasa, Y. (1999) Intl. J. Oncol. 15, 915–919. 52. Stickens, D., Brown, D. & Evans, G. A. (2000) Dev. Dyn. 218, 452–464. BIOCHEMISTRY

Kim et al. PNAS ͉ June 19, 2001 ͉ vol. 98 ͉ no. 13 ͉ 7181 Downloaded by guest on September 27, 2021