Demonstration of the Immature Glycosaminoglycan Tetrasaccharide Sequence GlcAβ1–3Galβ1–3Galβ1–4Xyl on Recombinant Soluble Human α-Thrombomodulin AN OLIGOSACCHARIDE STRUCTURE ON A “PART-TIME” PROTEOGLYCAN

Thrombomodulin (TM), a cell surface glycoprotein, is a critical mediator of endothelial anticoagulant defenses occurring both as a chondroitin sulfate proteoglycan (β-TM) and a protein (α-TM) unsubstituted by chondroitin sulfate (CS), hence its description as a “part-time” proteoglycan (PG) (Fransson, L. Å. (1987) Trends Biochem. Sci. 12, 406–411). Sugar analysis was performed on α-TM to investigate a possible biosynthetic mechanism for part-time PGs. Recombinant human α-TM, which was expressed in CHO-K1 cells, separated by anion-exchange chromatography from β-TM, and purified by immunoaffinity chromatography (Nawa, K., Sakano, K., Fujiwara, H., Sato, Y., Sugiyama, N., Teruuchi, T., Iwamoto, M., and Marumoto, Y. (1990) Biochem. Biophys. Res. Commun. 171, 729–737), was used for analysis. Preliminary sugar composition analysis after acid hydrolysis showed Xyl in addition to Gal, GalNAc, GlcNAc, Man, Fuc, and Glc.O-Glycosidically-linked oligosaccharides were liberated by mild alkaline treatment and purified. The isolated oligosaccharide fraction was derivatized with a fluorophore 2-aminobenzamide (2AB), resulting in two fluorescent components, a 2AB-oligosaccharide and a putative 2AB-Glc. Based on structural analysis by a combination of sequential exoglycosidase digestion and 500-MHz 1H NMR spectroscopy of the 2AB-oligosaccharide, the structure of the oligosaccharide was elucidated as GlcAβ1–3Galβ1–3Galβ1–4Xyl, which turned out to represent a glycosaminoglycan (GAG)-protein linkage region tetrasaccharide common to various PGs and was considered to be a biosynthetic intermediate of an immature GAG chain. The results may indicate that at least one class of the so-called part-time PGs bear the linkage tetrasaccharide at the GAG attachment sites and that the critical determining step or the rate-limiting step for PG biosynthesis is the transfer of the fifth sugar residue, the first hexosamine, rather than xylose.

Thrombomodulin (TM) 1 is an integral membrane glycoprotein expressed on endothelial cell surfaces, which binds to thrombin with high affinity, and acts as a critical mediator of endothelial anticoagulant defenses (1,2). Studies on the anticoagulant effects of TM demonstrated that it has cofactor activity for thrombin-dependent activation of protein C, direct inhibition of fibrinogen cleavage by thrombin, and indirect enhancement of the association of antithrombin III with thrombin. Although TM is a single polypeptide, it occurs naturally in two major glycoforms, one with (␤-TM) and the other without (␣-TM) a chondroitin sulfate (CS) chain (3,4), and hence has been termed "part-time" PG (5). Comparative studies on the anticoagulant effects of ␣-TM and ␤-TM revealed that the CS chain of TM plays an important role in regulating the anticoagulation functions. Indeed, the acceleration of the thrombin inhibition by antithrombin III is dependent on the presence of the CS moiety on TM. This moiety is not essential for protein C activation or enhanced inhibition by protein C inhibitor, but alters thrombin conformation (6), changes several kinetic parameters of protein C activation, and further enhances thrombin inhibition by the protein C inhibitor. On the other hand, platelet factor 4, which prevents complex formation between heparin and antithrombin III and thus functions as a procoagulant, binds to ␤-TM but not to ␣-TM, thereby promoting the generation of activated protein C (7). The CS moiety of TM also acts as the endothelial receptor for Plasmodium falciparum that infects red blood cells (8). The importance of the CS moiety of TM in such a variety of biological processes makes it imperative that we understand how the addition of the CS moiety is controlled.
Sulfated glycosaminoglycans (GAGs) including CS, dermatan sulfate, and heparin/heparan sulfate, are covalently bound to Ser residues in the core proteins through the common carbohydrate-protein linkage structure, GlcA␤1-3Gal␤1-3Gal␤1-4Xyl␤1-O-Ser. The biosynthesis of these GAGs is initiated by the addition of Xyl to specific Ser residues in the core protein and is followed by the addition of two Gal residues, a GlcA residue, and then alternating addition of N-acetylhexosamine and GlcA residues. The addition of each sugar is thought to be catalyzed by a specific glycosyltransferase, which uses the corresponding uridine diphosphate sugar as a donor substrate. It has been proposed that an acceptor consensus sequence for xylosyltransferase recognition exists around the serine resi-dues substituted by GAGs (9, 10). Mann et al. (11) proposed that the amino acids surrounding the serine residue generate a conformation whereby the GAG attachment site is bound poorly by xylosyltransferase, and consequently some molecules are substituted and others unsubstituted with GAG, resulting in a "part-time" PG. In addition, Lin et al. (4) indicated that the conformation well beyond the immediate site of GAG addition plays an important role in GAG attachment. Alternatively, another hypothesis for the formation of part-time PG involves competition for a serine residue between xylosyltransferase and N-acetylgalactosaminyltransferase (12). However, neither model satisfactorily explains the biosynthetic mechanism of part-time PG.
We recently identified a novel ␣-N-acetylgalactosaminyltransferase in fetal bovine serum, and also in mouse mastocytoma cells, which catalyzes the transfer of an ␣-GalNAc residue to the linkage tetrasaccharide-serine, GlcA␤1-3Gal␤1-3Gal␤1-4Xyl-1-O-Ser, derived from PGs, although its role in GAG biosynthesis remains unclear (13,14). In addition, Manzi et al. (15) reported that the ␣-GalNAc-capped pentasaccharide sequence bound to an artificial primer was secreted by human melanoma cells, Chinese hamster ovary cells, and several other human cancer cell lines when carbohydrate synthesis was stimulated using a primer 4-methylumbelliferyl-␤-D-xyloside. More recently, we found that the ␣-GalNAc-capped pentasaccharide serine GalNAc␣1-4GlcA␤1-3Gal␤1-3Gal␤1-4Xyl1-O-Ser, a reaction product of the ␣-GalNAc transferase, was not utilized as an acceptor for a glucuronyltransferase involved in CS biosynthesis (16). Therefore, we anticipated that the addition of an ␣-GalNAc residue to the tetrasaccharide core of the linkage region might serve as a stop signal that precluded further chain elongation, creating a part-time PG, and hence attempted to detect the ␣-GalNAc-capped pentasaccharide linkage structure GalNAc␣1-4GlcA␤1-3Gal␤1-3Gal␤1-4Xyl on part-time PGs. In the present study, we isolated and characterized the O-linked oligosaccharides on recombinant human ␣-TM. Structural analysis unexpectedly revealed the occurrence of an immature truncated GAG sequence GlcA␤1-3Gal␤1-3Gal␤1-4Xyl on the ␣-TM (47). The findings suggest that the critical determining step for PG biosynthesis is the transfer of the fifth sugar residue, the first hexosamine, rather than xylose.
Sugar Composition Analysis-␣-TM (600 g, equivalent to 9.2 nmol) was hydrolyzed in 4 N HCl, 4 M trifluoroacetic acid (1/1, v/v) at 100°C for 4 h, N-acetylated, and then pyridylaminated (20). The resulting pyridylaminated monosaccharides were identified and quantified by HPLC analysis on a PALPAK Type A column (Takara Shuzo Co., Kyoto, Japan). Elution was performed under isocratic conditions using acetonitrile, 0.7 M sodium borate (pH 9.0) (1/9, v/v) at a flow rate of 0.3 ml/min at 65°C. Eluates were monitored by fluorescence at an excitation wavelength of 310 nm and an emission wavelength of 380 nm. Similar analysis was carried out using an Ultrasphere ODS column (4.6 ϫ 250 mm; Beckman Instruments, Inc., Palo Alto, CA) instead of a PALPAK Type A column, and similar results were obtained from both analyses.
Isolation of the O-Linked Oligosaccharide Components of ␣-TM-␣-TM was dissolved in 0.5 M LiOH and kept at 4°C for 16 h to release O-linked saccharides from the core protein (19). After neutralization, the sample was applied to a column (2.5 ml bed volume) of AG 50W-X2 (H ϩ form, Bio-Rad). The flow-through fraction containing the O-linked oligosaccharide components was pooled and neutralized with 1 M NH 4 HCO 3 .
Derivatization of the Isolated Oligosaccharide with 2-Aminobenzamide (2AB)-Derivatization with 2AB of the oligosaccharide component of ␣-TM was performed as described (21). The isolated oligosaccharide fraction (0.6 nmol) was lyophilized in a microcentrifuge tube. A 5-l aliquot each of 0.25 M 2AB and 1.0 M NaCNBH 3 , both of which were prepared by dissolving in a glacial acetic acid/Me 2 SO mixture (3/7, v/v), were added to the sample, and the mixture was incubated at 65°C for 2 h. The excess reagent was removed by gel-filtration HPLC on a TSK-gel G2500PW column (Tosoh Co., Tokyo, Japan) using 20 mM CH 3 COONH 4 (pH 7.5) as eluent at a flow rate of 1 ml/min. Eluates were monitored by fluorescence at an excitation wavelength of 330 nm and an emission wavelength of 420 nm. The 2AB-derivatized oligosaccharide fraction (fraction 1, see Fig. 2) was pooled, evaporated to dryness, and used for the structural analysis described below.
Enzymatic Digestion-Enzymatic digestion with chondroitinases ABC or AC-II was carried out using 23 pmol of the 2AB-derivatized oligosaccharide (fraction 1) and 10 mIU of the enzyme in a total volume of 30 l of appropriate buffer at 37°C for 20 min according to Sugahara et al. (22). For successive enzymatic digestion, the 2AB-oligosaccharide (50 pmol) was first incubated with 10 mIU of ␤-glucuronidase in a total volume of 50 l of 0.05 M sodium citrate buffer, pH 4.5, at 37°C for 10 min as described (23), then boiled at 100°C for 1 min to terminate the reaction. One half of the sample was analyzed by HPLC, while the other half was mixed with 1.70 IU of ␤-galactosidase in 0.1 M Tris-HCl buffer, pH 7.2, containing 4 mM MgCl 2 and 0.25 M 2-mercaptoethanol at 37°C for 16 h (24).
HPLC-The 2AB-oligosaccharide (fraction 1) and its digests with ␤-glucuronidase or ␤-galactosidase were analyzed in an HPLC system LC10A connected to an RF-535 fluorometric detector (Shimadzu Corp., Kyoto, Japan) by two methods. Anion-exchange chromatographies were performed on a polyamine-bound silica PA03 column (4.6 ϫ 250 mm, YMC Co., Kyoto, Japan) using a linear gradient from 16 to 530 mM NaH 2 PO 4 over a 60-min period at a flow rate of 1.0 ml/min at room temperature. Gel filtration chromatographies were performed on a column (7.6 ϫ 500 mm) of Asahipak GS-320 (Asahi Chemical Industry, Kawasaki, Japan) using 50 mM CH 3 COONH 4 as eluent at a flow rate of 1 ml/min. In both analyses, eluates were monitored by fluorescence as described above. 500-MHz 1 H NMR Spectroscopy-The 2AB-oligosaccharide (fraction 1) was repeatedly exchanged in D 2 O with intermediate lyophilization.

Purity Assessment and Carbohydrate Composition Analysis of Recombinant Human
Soluble ␣-TM-The purified ␣-TM gave a single but slightly diffuse band of M r 65,000 -67,000 on SDS-PAGE under non-reducing conditions (Fig. 1). The apparent M r of ␣-TM was slightly reduced after N-glycosidase treatment or successive treatment with neuraminidase then Oglycosidase (data not shown). These results most likely indicate that ␣-TM has been modified with N-and O-linked sugar substituents. To analyze the carbohydrate composition, puri-fied ␣-TM was subjected to acid hydrolysis followed by Nacetylation and then pyridylaminated (20). The resulting pyridylaminated sugars were identified and quantified using a PALPAK Type A column or a Ultrasphere ODS column. The results revealed that ␣-TM contained GalNAc, GlcNAc, Glc, Man, Fuc, Gal, and notably Xyl in a molar ratio of 2:5:0.75:3: 1:6:1 (Table I). In the case of uronic acids as carbohydrate constituents, complete liberation is rarely possible without accompanying decomposition. When the reaction condition (4 N HCl, 4 N trifluoroacetic acid at 100°C for 4 h) was used for hydrolysis, destructive decarboxylation of uronic acid must have occurred (26). Thus, GlcA was not observed in this analysis even through ␣-TM contained GlcA as will be described below.
Isolation of the Oligosaccharides-O-Glycosidically-linked oligosaccharides were liberated from the purified ␣-TM by mild alkaline treatment and purified by passage through a cationexchange column as described under "Experimental Procedures." The isolated oligosaccharide fraction was derivatized with a fluorophore 2AB and fractionated by gel filtration HPLC, resulting in two fluorescent components, a 2AB-oligosaccharide (fraction 1) and a putative Glc-2AB (fraction 2) (Fig.  2). The latter peak was co-eluted with the authentic Glc-2AB when they were co-injected (data not shown). The 2AB-derivatized oligosaccharide was analyzed below, whereas the putative Glc-2AB was not analyzed further. Although ␣-TM seemed to contain O-glycosidase-sensitive oligosaccharides as revealed by the detection of GalNAc by the preliminary sugar composition analysis, O-linked oligosaccharides with a GalNAc residue at the reducing terminus were not detected. However, when 3 Hlabeled O-linked oligosaccharides released from ␣-TM by alkaline ␤-elimination with NaB 3 H 4 were fractionated by HPLC and used for the structural analysis, 3 H-labeled galactosaminitol as well as 3 H-labeled xylitol were detected indeed by acid hydrolysis of the fractions (data not shown). Therefore, Olinked oligosaccharides with a GalNAc residue at the reducing terminus were actually contained in ␣-TM but not released by mild alkaline treatment used prior to 2AB-derivatization in the present study.
Enzymatic Characterization of the Isolated 2AB-derivatized Oligosaccharide-Fraction 1 was subjected to HPLC on an amine-bound silica PA03 column. Only one peak was detected at the elution position of the authentic unsaturated linkage tetrasaccharide ⌬HexA␣1-3Gal␤1-3Gal␤1-4Xyl-2AB (Fig.  3A), and the molar ratio of the oligosaccharide to the core protein was calculated to be in the range of 0.77-1.33 based on the peak area on HPLC. The peak was shifted to a position corresponding to Gal␤1-3Gal␤1-4Xyl-2AB by ␤-glucuronidase digestion (Fig. 3B). In contrast, the peak was not shifted by digestion with chondroitinases ABC or AC-II (data not shown), demonstrating no repeating disaccharide unit. It was not sensitive to neuraminidase or ␣-N-acetylgalactosaminidase either, indicating that it is not covered with neuraminic acid or ␣-Gal-NAc. Fraction 1 was further analyzed by sequential enzymatic digestion using gel-filtration HPLC on a column of GS-320. One major peak was again observed at the elution position slightly ahead of the 2AB-derivatized authentic unsaturated linkage tetrasaccharide ⌬HexA␣1-3Gal␤1-3Gal␤1-4Xyl-2AB before the enzymatic treatment (Fig. 4A). The peaks marked by asterisks were attributable to the buffer salts. The peak was shifted to the position of Gal␤1-3Gal␤1-4Xyl-2AB upon ␤-glucuronidase digestion (Fig. 4B) and then to the position of Xyl-2AB upon subsequent exhaustive ␤-galactosidase digestion (Fig. 4C). These results altogether indicate that the structure of the compound in fraction 1 is most likely GlcA␤1-3Gal␤1-3-Gal␤1-4Xyl-2AB. 500-MHz 1 H NMR Spectroscopy-The structure of the compound in fraction 1 was further analyzed by 500-MHz 1 H NMR spectroscopy. The one-dimensional spectrum of fraction 1 recorded at 26°C is shown in Fig. 5. Signals at ␦ 4.4 -5.5 ppm were identified as H-1 resonances of the constituent saccharide residues by comparison with the NMR spectra of the nonsulfated linkage hexasaccharide-serine (27) and the corresponding hexasaccharide alditol (28 -30). Since H-1 resonances of the Xyl-1 residue, which is linked to 2AB, were not observed, the   spectrum of the parent oligosaccharide itself, which had been released by the alkali treatment but had not been derivatized with 2AB, was also recorded and is shown in the inset of Fig. 5. Proton signals in the one-dimensional spectrum were assigned below as described previously (27-30) using the two-dimensional COSY spectrum of fraction 1, which was measured at 26°C and is shown in Fig. 6. The NMR data of fraction 1 are summarized in Table II with those of the synthetic reference compound GlcA␤1-3Gal␤1-3Gal␤1-4Xyl-Ser (31).
Although the anomeric resonances of Gal-3 and GlcA-4 were clustered around ␦ 4.67 in the one-dimensional spectrum (Fig.  5), Gal-3 H-2 and GlcA-4 H-2 were readily discriminated based on the NMR data of the nonsulfated linkage hexasaccharide alditol (28 -30) and the tetrasaccharide-serine (31). Starting with the H-2 resonances (␦ 3.75) of the Gal-3 residue, other proton resonances of this residue including H-1 (␦ 4.670), H-2, and H-3 were assigned in the COSY spectrum. Accordingly, the remaining H-1 resonance at ␦ 4.672, which showed the connectivity with the well resolved H-2 in the COSY spectrum (Fig. 6), was assigned to GlcA-4. Hence, based on the NMR data of fraction 1 (Table II), the structure of the compound in fraction 1 was deduced as GlcA␤1-3Gal␤1-3Gal␤1-4Xyl-2AB. The ␤1-3 linkage between GlcA-4 and Gal-3 was strongly supported by the close similarity in chemical shift of Gal-3 H-3 to that of the corresponding signal of the reference compound. The deviation of the chemical shift of Gal-2 H-1 of fraction 1 from that of the reference compound is most likely due to the fact that the former is derivatized with 2AB and the terminal xylose is in a chain form, but the xylose residue in the latter is in a pyranose ring form. In fact, the chemical shift of Gal-2 H-1 is almost identical with those of the corresponding Gal-2 H-1 of the linkage hexasaccharide alditols (28 -30). DISCUSSION Protein species in which only some of the molecules are substituted with GAG have been termed "part-time" PGs (5). In addition to TM, collagen type IX (32), the invariant chain of the class II major histocompatibility complex (33), the high molecular weight receptor for transforming growth factor-␤ (34), and a lymphocyte homing receptor (35) are well known as part-time PGs. The addition of GAG chains on these proteins has generally profound effects on their various properties since it depresses the biological activity by blocking the active site on the protein or adds a new binding site to the molecule. Hence, it is of particular importance to investigate the biosynthetic mechanism generating part-time PGs and to understand how the addition of GAG chains is regulated. So far, the addition of GAG chains has been considered to occur in an all-or-none manner. Namely, no obvious intermediate oligosaccharide chains for the GAG biosynthesis have been found on core proteins of part-time PGs including ␣-TM. In this study, based on the structural analysis by a combination of enzymatic digestion and 1 H NMR spectroscopy, we demonstrated the following Olinked oligosaccharide containing a Xyl residue on recombinant human ␣-TM: GlcA␤1-3Gal␤1-3Gal␤1-4Xyl. This finding clearly indicates that the biosynthesis of GAG has actually FIG. 3. HPLC analysis of fraction 1 isolated from ␣-TM. Fraction 1 isolated from ␣-TM (see Fig. 2) was analyzed by HPLC on an aminebound silica column using a linear gradient of 16 -530 mM NaH 2 PO 4 over a 60-min period before (panel A) and after ␤-glucuronidase digestion (panel B) as described under "Experimental Procedures." The eluates were monitored by fluorescence at an excitation wavelength of 330 nm and an emission wavelength of 420 nm. The arrows denote the elution positions of the following authentic compounds: 1, Gal␤1-3Gal␤1-4Xyl-2AB; 2, ⌬HexA␣1-3Gal␤1-3Gal␤1-4Xyl-2AB. Note that a given unsaturated oligosaccharide and the corresponding saturated oligosaccharide, which differ only in the non-reducing terminal structure, were hardly separated under these conditions. FIG. 4. Sequential glycosidase digestions of fraction 1 isolated from ␣-TM. Fraction 1 isolated from ␣-TM (see Fig. 2) was analyzed by gel-filtration HPLC on a column of GS-320 (7.6 ϫ 500 mm), which was eluted with 50 mM CH 3 COONH 4 at a flow rate of 1 ml/min, as described under "Experimental Procedures." The eluates were monitored by fluorescence at an excitation wavelength of 330 nm and an emission wavelength of 420 nm. Panel A, before enzymatic digestion; panel B, after ␤-glucuronidase digestion; panel C, after ␤-glucuronidase digestion followed by ␤-galactosidase digestion. The peaks marked by asterisks were attributable to the buffer salts. The arrows denote the elution positions of the following authentic compounds: 1, ⌬HexA␣1-3Gal-NAc␤1-4GlcA␤1-3Gal␤1-3Gal␤1-4Xyl-2AB; 2, ⌬HexA␣1-3Gal␤1-3Gal␤1-4Xyl-2AB; 3, Gal␤1-3Gal␤1-4Xyl-2AB; 4, Xyl-2AB. been initiated on ␣-TM but is truncated leaving the polymerization incomplete. Therefore, we may have to revise the previous concept regarding the biosynthetic mechanism for generating a part-time PG.
Previous studies suggested that the transfer of a Xyl residue was the critical step determining GAG biosynthesis since xylosylation is the first step in the carbohydrate modification of the core protein (36). Mann et al. (11) proposed that the amino acids surrounding the serine residue generate a conformation whereby the GAG attachment site is bound poorly by xylosyltransferase and consequently only some of the molecules are substituted with GAG. They also showed using site-directed mutagenesis that the xylosyltransferase responsible for the initiation of the GAG chain can use a threonine residue for the substitution instead of a serine residue, but that such substitution is only partial, creating a part-time PG. Thus, it seems that the conformation of the GAG substitution site may be important for recognition by xylosyltransferase (11). On the other hand, Gerlitz et al. (12) presented the "acceptor consensus overlap" model involving glycosyltransferase competition for the serine residue to account for the expression of TM in its two distinct glycoforms, ␣and ␤-TMs, since there is an apparent overlap at Ser 474 in human TM in terms of the acceptor consensus sequences of xylosyltransferase and GalNAc transferase. In this context, several studies suggested that these two glycosyltransferases reside in the same subcellular compartment (cis-Golgi) (37-39), and it is therefore possible that direct   5. Structural-reporter-group regions of the 500-MHz 1 H NMR spectrum of fraction 1 recorded in D 2 O at 26°C. One-dimensional spectrum of fraction 1 is shown. The inset shows the spectrum of the oligosaccharides isolated from ␣-TM before derivatization with 2AB. The peaks marked by asterisks in the inset were attributable to impurities. The letters and numbers refer to the corresponding residues in the structure. competition for substrate could occur. Nonetheless, our present findings indicate that the xylosylation of the core protein occurs on both ␣and ␤-TMs.
Thus, xylosylation does not seem to be the critical step determining the conversion of a protein into a PG at least in the case of TM. Once a protein is xylosylated, initiation of GAG synthesis proceeds through the addition of the two Gal residues and one GlcA residue at least until the carbohydrate-protein linkage region is formed. However, in the next step, the first GalNAc residue is not always transferred to such a tetrasaccharide linkage region in the case of TM. In fact, culture medium or cell lysate of the malignant fibrous histiocytoma cell line MFH-7 (established by Dr. Hidetoshi Okabe, Department of Clinical Laboratory of Medicine, Shiga University of Medical Science) that abundantly synthesizes CS 3 was tested as an enzyme source of GalNAc transferase as described (13,40). However, the tetrasaccharide moiety of ␣-TM did not accept a GalNAc residue from UDP-GalNAc (data not shown). It is, therefore, likely that the amino acids surrounding the Ser residue for GAG substitution may generate a conformation whereby the substitution site is bound poorly by ␤-GalNAc transferase I that catalyzes the transfer of the first ␤-GalNAc residue to the linkage tetrasaccharide-serine for CS/DS biosynthesis. In this regard, it should be noted that ␣-GlcNAc transferase I that catalyzes the transfer of the first ␣-GlcNAc residue to the linkage tetrasaccharide-serine for heparin/heparan sulfate biosynthesis was reported to recognize the hydrophobic amino acids surrounding the GAG attachment site (41). On the other hand, we previously presented the hypothesis that C4 sulfation of the Gal adjacent to the GlcA residue in the common linkage region might drive the CS/DS biosynthesis (27,42,43). In addition, transient 2-O-phosphation of Xyl residues during the formation of the tetrasaccharide linkage region were recently reported by Moses et al. (44). Therefore, lack of characteristic structures such as 4-O-sulfation of Gal or 2-O-phosphation of Xyl residues of the linkage region in ␣-TM may prevent ␤-GalNAc transferase I from transferring a GalNAc residue to the linkage tetrasaccharide. In any case, ␤-GalNAc transferase I is regarded as a rate-limiting key enzyme to determine whether or not the chain elongation proceeds by the alternative addition of GlcA and GalNAc residues and the step determining generation of part-time PGs seems to involve the substrate recognition by ␤-GalNAc transferase I rather than xylosyltransferase.
Naturally occurring PGs with an immature GAG oligosaccharide have not been found on cell surfaces or in extracellular matrices. Therefore, maturation of GAG chains appears to be associated with intracellular transport and secretion of PGs. However, ␣-TM is normally transported and secreted even if the GAG modification of the core protein remains incomplete. Recently, subcellular localization of GlcA transferase I responsible for the synthesis of the linkage region was investigated by Sugumaran et al. (45). The linkage region GlcA transferase I activity was found in medial Golgi and trans-Golgi/trans-Golgi network fractions, and its distribution was distinct from those of linkage region Gal transferases I and II, which were found exclusively in the cis-Golgi fraction (45). Since the addition of the GlcA residue to the linkage region occurred on ␣-TM in this study, the core protein of ␣-TM is considered to have been transported normally from endoplasmic reticulum to medial Golgi or trans-Golgi/trans-Golgi network. Nevertheless, it seems to have escaped, by a certain unidentified mechanism, ␤-GalNAc addition.
In conclusion, our studies demonstrated the immature truncated GAG structure GlcA␤1-3Gal␤1-3Gal␤1-4Xyl on the recombinant human ␣-TM. The findings indicate that the anticoagulation functions of TM are regulated by the transfer of a GalNAc residue to the linkage tetrasaccharide rather than the transfer of a Xyl residue to the core protein and that ␤-GalNAc transferase I, which catalyzes the transfer of a ␤-GalNAc residue to the linkage tetrasaccharide, may play a critical role as the rate-limiting enzyme in the biosynthesis of part-time PGs. Further study of the carbohydrate structure on other part-time PGs is necessary to confirm the generality of the mechanism.