Irreversible Glucuronyl C5-epimerization in the Biosynthesis of Heparan Sulfate*

Glucuronyl C5-epimerase catalyzes the conversion of D -glucuronic acid to L -iduronic acid units in heparan sulfate biosynthesis. Substrate recognition depends on the N -substituent pattern of the heparan sulfate precursor polysaccharide and requires the adjacent glucosa-mine residue toward the non-reducing end to be N -sul-fated. Epimerization of an appropriately N -sulfated substrate is freely reversible in a soluble system, with equilibrium favoring retention of D - gluco configuration (Hagner-McWhirter, Å., Lindahl, U., and Li, J.-P. (2000) Biochem. J. 347, 69–75). We studied the reversibility of the epimerase reaction in a cellular system, by incubating human embryonic kidney 293 cells with D -[5- 3 H]galactose. The label was incorporated with glucuronic acid units into the heparan sulfate precursor polysaccharide and was lost upon subsequent C5-epi-merization to iduronic acid. However, analysis of oligosaccharides obtained by deaminative cleavage of the mature heparan sulfate chains indicated that all glucuronic acid units retained their C5- 3 H label, irrespective of whether they had occurred in sequences susceptible or resistant to the epimerase. All 3 H-labels of the final products resisted incubation with epimerase

groups along the HS chain (2). These features are largely established during HS biosynthesis, which appears to be regulated such that different types of cells, possibly also cells exposed to different stimuli, generate HS of different structures (3,4). Notably, however, further structural diversity may be introduced through endosulfatase action after completed biosynthesis (5).
HS biosynthesis, a Golgi process, is initiated by the generation of a linkage tetrasaccharide, GlcA␤1,3Gal␤1,3Gal␤1,4Xyl-, onto a serine residue in the proteoglycan protein core (6,7). N-Acetylglucosamine (GlcNAc) and D-glucuronic acid (GlcA) residues are then transferred from the corresponding UDPsugars to the non-reducing ends of nascent polysaccharide chains, concomitant with further modification of these polymers (3). N-Deacetylation and N-sulfation of GlcNAc units thus precede C5-epimerization of GlcA to L-iduronic acid (IdoA) residues, which, in turn, are followed by O-sulfation at different positions. Only a fraction of potential target units is attacked in each modification step, resulting in domain-type variability as well as more subtle structural diversity. The overall process is regulated in part by the substrate specificities of the various enzymes involved, all of which have now been cloned and expressed. However, additional regulatory mechanisms, still poorly understood, are required to explain the display of saccharide epitopes in HS chains. Our understanding of the regulation of HS biosynthesis has been hampered by the lack of information regarding the organization of the biosynthetic machinery. The enzymes involved are membrane-bound, with Nterminal transmembrane domains and C-terminal lumenal catalytic domains and are assumed to operate in a closely concerted mode. Still, little is known regarding their topological and functional interaction or the kinetics of the various steps of HS chain formation and modification. The process has been described in terms ranging from kinetically controlled competition between different enzymes for common polymer substrates (8) to more strictly molded assembly lines (9).
The present work was undertaken to gain further insight into the biosynthetic process by exploiting the properties of a unique step, i.e. the C5-epimerization of GlcA to IdoA residues. Solubilized (or recombinant) epimerase catalyzes a freely reversible reaction involving abstraction of the C5-hydrogen of a target hexuronic acid (GlcA or IdoA) followed by reinsertion (presumably of a proton) to yield the alternative C5-configuration (Fig. 1). Incubation with epimerase of an appropriate polysaccharide substrate containing C5-3 H-labeled hexuronic acid residues will release the label from GlcA as well as IdoA units; conversely, incubation of unlabeled substrate in 3 H 2 O leads to incorporation of label into both epimers (10). Is epimerization reversible also during HS biosynthesis in an intact cell? The answer to this question bears not only on the kinetics of epimerization in the Golgi, but may also provide clues to the overall organization of the HS biosynthetic apparatus. We have approached this problem by allowing cells to incorporate C5-3 H-labeled GlcA residues into HS chains under formation, followed by analysis of the product with regard to residual label. 14 C]Gal (293 mCi/mmol), D-[U-14 C]GlcN hydrochloride (311 mCi/mmol), and NaB 3 H 4 (55 Ci/mmol) were purchased from Amersham Biosciences. GlcA C5-epimerase was purified from bovine liver as described (11). Pre-packed PD-10 columns, DEAE-Sephacel, and Sephadex G-50 were obtained from Amersham Biosciences. Bio-Gel P-10 was purchased from Bio-Rad. Chondroitin ABC lyase (EC 4.2.2.4) and heparitinase I (EC 4.2.2.8) were obtained from Seikagaku (Japan). ␤-D-Glucuronidase from bovine liver (type B-10) was from Sigma.

Materials-D-[U-
Synthesis of [5-3 H]Gal-A 5-ketohexofuranose 1 used as starting material was generated from 1,2-O-isopropylidene-␣-D-galactofuranose (12) (75 mg, 0.34 mmol), following activation with bis(tri-n-butyltin) oxide and subsequent oxidation with Br 2 as described for the corresponding gluco-derivative (13) (Fig. 2). The 5-oxo derivative 1 (52 mg, 70%) was obtained after silica gel chromatography (CHCl 3 :MeOH, 19/ 1). NMR data were as follows: 13  [5-3 H]Gal was generated by reduction of 1 with NaB 3 H 4 followed by mild acid hydrolysis of the isopropylidene acetal (Fig. 2). A sample of 1 (2 mg dissolved in 40 l water) was added to 10 ml of methanol together with 50 mCi NaB 3 H 4 , and the mixture was incubated for 1 h at room temperature. After dropwise addition of 100 l of 1 M HCl to give pH 2-3 (in a fume hood), the mixture was incubated for another 30 min and was then heated to 55°C and evaporated to dryness. The addition of 20 ml of methanol followed by evaporation to dryness as above was repeated 5 times. The reduction products were dissolved in 3 ml of water and then analyzed by paper chromatography on Whatman number 1 paper in butanol/ethanol/water (10:3:5 by volume), along with D-[U-14 C]Gal as standard (running time 22 h). A major component (65% of total radioactivity) appeared at the migration position of Gal, whereas 19% migrated as altrose, another expected reaction product (L-enantiomer; Fig.  2). The yields of D-[5-3 H]Gal in two separate preparations were 5.4 mCi and 4.6 mCi. The two batches were combined before use.
Metabolic Labeling of Cells-Human embryonic kidney HEK 293 cells were cultured in Dulbecco's modified Eagle's medium with Glutamax-II (Invitrogen) using standard procedures at 37°C. The medium contained 4.5 g/liter glucose and was supplemented with 10% heatinactivated fetal calf serum (Invitrogen), 60 g/ml penicillin, and 50 g/ml streptomycin sulfate. When cells seeded in six 175-cm 2 flasks had reached ϳ80% confluency, the medium was replaced with 15 ml of medium supplemented as above but containing only 1 g/liter glucose and, in addition, Purification of Metabolically Radiolabeled HS-After labeling, the media were collected, and the cell layers were lysed by adding to each flask 10 ml of cold phosphate-buffered saline containing 1% Triton X-100. All samples were kept on ice on a rocking plate for 1 h. Urea was added to a final concentration of 8 M to both the media and cell lysates, and the mixtures were boiled for 15 min. The samples were then centrifuged at 11,000 ϫ g at 4°C for 20 min. Supernatants from media and cell lysates were mixed, and an equal volume of 0.1 M sodium acetate, pH 4, was added. The samples were applied on DEAE-Sephacel columns (10 ml) equilibrated with 50 mM sodium acetate buffer containing 2 M urea and 0.2 M NaCl, pH 4. After extensive washing (buffer as above), the columns were eluted with a linear gradient of 0.2-1 M NaCl in the same buffer (total volume 100 ml). Fractions containing proteoglycans eluted between 0.5 and 0.75 M NaCl were pooled, dialyzed against water, and lyophilized.
The proteoglycans were further purified by gel chromatography on a Sephadex G-50 column (1 ϫ 90 cm) in 0.2 M NH 4 HCO 3 . The proteoglycans, excluded from the gel, were pooled and desalted by lyophilization. The samples were treated with 1 unit/ml of chondroitinase ABC. The HS chains were then released from proteoglycan core proteins by ␤-elimination in 0.5 M NaOH for 16 h at 4°C followed by neutralization with 4 M acetic acid.
The free HS chains were recovered by absorption to a 1-ml DEAE-Sephacel column equilibrated in 0.2 M NH 4 HCO 3 . The columns were washed with 0.2 M NH 4 HCO 3 followed by 0.25 M NaCl and 0.2 M NH 4 HCO 3 , and were finally eluted with 2 M NH 4 HCO 3 . This purification procedure yielded pure HS as it was quantitatively degraded by treatment with heparitinase I (not shown).
HS Structural Analysis-Metabolically labeled HS chains (generally 400 ϫ 10 3 dpm of 3 H) were cleaved at the sites of N-sulfated GlcN units by treatment with nitrous acid at pH 1.5 followed by reduction with NaBH 4 (14). Under these conditions, GlcNSO 3 units are deaminated and converted to terminal 2,5-anhydromannitol residues, whereas Glc-NAc units remain intact (15). Deamination mixtures were fractionated on a column (1.3 ϫ 185 cm) of Bio-Gel P-10 eluted with 0.5 M NH 4 HCO 3 . Effluent fractions were analyzed for 3 H and 14 C. Fractions corresponding to di-and tetrasaccharides were pooled, desalted by lyophilization, and further analyzed separately. O-Sulfated disaccharides were separated by anion exchange HPLC on a Partisil-10 SAX column (4.6 ϫ 250 mm; Whatman), using a step gradient of KH 2 PO 4 (0.012, 0.028, and 0.168 M) at a flow rate of 1 ml/min. Fractions of 1 ml were collected and analyzed for radioactivity. O-Sulfated disaccharides were first isolated by preparative high voltage paper electrophoresis at pH 5.3 (0.083 M pyridine, 0.05 M acetic acid) and were then identified by paper chromatography (ethyl acetate/acetic acid/H 2 O, 3:1:1 by volume) (10).

RESULTS
The course of hexuronyl (HexA)-C5-epimerization in relation to subsequent polymer modification (O-sulfation) reactions in HS biosynthesis was investigated by monitoring the fate of C5-3 H atoms introduced with GlcA residues during the polymerization phase of the process. Incorporation of   The latter alternative would apply only if such blocking is favored over reversible C5-epimerization, because a GlcA unit generated by "back epimerization" of IdoA would be devoid of C5-3 H. Conversely, the occurrence of unlabeled GlcA units in primary epimerase target positions would point to a relatively looser organization of the biosynthetic process in which C5epimerization would be allowed to revert to D-gluco configuration before onset of intervening O-sulfation (Fig. 3).
Primary susceptibility of GlcA units in the HS precursor polysaccharide to enzymatic C5-epimerization is dictated by the adjacent N-substituent pattern. The GlcA residues in -Glc-NS-GlcA-GlcNS-and -GlcNS-GlcA-GlcNAc-sequences (reducing terminus to the right) thus are recognized as substrate by the epimerase, whereas those in -GlcNAc-GlcA-GlcNAc-and -GlcNAc-GlcA-GlcNS-sequences are not. These various structural constellations can be distinguished by deaminative cleavage of the polysaccharide with nitrous acid, which will attack N-sulfated GlcN units and break the corresponding glucosaminidic linkages, whereas N-acetylated GlcN residues are resistant (18). A HexA located between two GlcNS residues will thus be recovered in a disaccharide deamination product, whereas a -GlcNAc-GlcA-GlcNAc-structure will form part of a larger oligosaccharide. Tetrasaccharides derived from sequences of alternating N-acetylated and N-sulfated disaccharide units are particularly instructive, because the nonreducing terminal HexA, but not the internal GlcA unit, represents a potential epimerase substrate in the intact parent polysaccharide chain (Fig. 4). We therefore attempted to determine the distribution of the 3 H label between the variously sized deamination products. To enable such comparison, a 14 C reference label was introduced through the addition to the incubation medium of either [U- 14  GlcNS residues by treatment with HNO 2 , and the resultant diand oligosaccharides were separated by gel chromatography on a Bio-Gel P-10 column. The internal GlcA residues of HexA-(GlcNAc-GlcA) n -aMan R oligosaccharides correspond to units in the intact parent polymer that are not recognized as substrate by the epimerase, whereas the nonreducing-terminal HexA residues represent potential target units (Fig. 4) (17). Consistent loss of 3 H from such target units would result in progressively decreasing 3 H/ 14 C ratios for oligosaccharides of decreasing size. The results shown in Fig. 5 conform to this prediction, irrespective of whether the 14 C label was introduced through [ 14 C]GlcN (Fig. 5A) or [ 14 C]Gal (Fig. 5B). The largest (Ն18-mer) fragments were expected to contain labeled Gal residues of the HS-protein linkage region and were therefore excluded from further calculations. Instead, the 3 H/ 14 C ratio of the 16-mer was applied as a reference approximating complete retention of 3 H and used to calculate hypothetical ratios for the smaller oligosaccharides, assuming complete loss of nonreducing terminal 3 H. Comparison with the experimentally found ratios indicated significant loss of 3 H from all oligosaccharide species (Table  I). We therefore proceeded to analyze whether this lack of 3 H was restricted to IdoA units of the HS chain or also included GlcA units formed by back-epimerization of IdoA residues.
Analysis of Epimerase HexA Target Units-Disaccharide deamination products are derived from HS sequences composed of consecutive N-sulfated disaccharide units, hence con- taining HexA units that are initially all susceptible to epimerase attack (Fig. 4). The disaccharide fractions were separated by anion exchange HPLC (Fig. 6). As expected, all IdoA-containing species (Fig. 6, peaks 4, 5, and 6) had lost their C5-3 H label. By contrast, the 3 H/ 14 C ratios for the GlcA-containing disaccharides, GlcA2S-aMan R (peak 2) and GlcA-aMan R 6S (peak 3), were similar to that of the extended N-acetylated 16-mer reference structure, irrespective of whether the 14 C was introduced through [ 14 C]GlcN (Fig. 6A) or [ 14 C]Gal (Fig. 6B), thus demonstrating essentially quantitative retention of C5-3 H (Table II). Non-O-sulfated disaccharides (6.4% of total disaccharide units in N-sulfated domains) were isolated by paper electrophoresis and further separated into GlcA-aMan R and IdoA-aMan R species by paper chromatography (Fig. 7). Only the former component occurred in significant amounts, again with fully retained 3 H label (Table II). Principally similar results were obtained irrespective of whether the 14 C reference label had been introduced through [ 14 C]GlcN (Fig. 5A) or [ 14 C]Gal ( Fig. 5B; Table II).
Tetrasaccharide deamination products, isolated by gel chromatography (Fig. 5), were digested with ␤-D-glucuronidase to yield GlcA monosaccharide and GlcNAc-GlcA-aMan R trisaccharide, in addition to residual tetrasaccharide (Fig. 8). Most of the tetrasaccharides resisted ␤-glucuronidase digestion, indicating IdoA in a nonreducing terminal position. However, significant proportions were degraded to free GlcA and trisaccharide, 2 and analysis of these fractions showed that the total amounts of 3 H released as [ 3 H]GlcA by ␤-glucuronidase equaled those recovered in the trisaccharide digestion products (Fig. 8, A and B). These findings point to essentially complete retention of 3 H in GlcA residues that are potentially accessible to the epimerase in the intact polysaccharide substrate and yet retain D-gluco configuration and are recovered in nonreducing terminal position of tetrasaccharide deamination products (Figs. 4 and 9). This conclusion is further supported by the 3 H/ 14 C ratios of products generated by ␤-glucuronidase digestion of tetrasaccharide from HS labeled with  a Predictions are based on the assumption that the experimentally found ratio, 5.7, for the 16-mer fraction approximates the ratio for the epimerase-resistant -GlcNAc-GlcA-disaccharide unit. This reference ratio will be somewhat underestimated due to terminal IdoA formation in the largest oligosaccharides.
Susceptibility of GlcA units to the C5-epimerase during the initial phases of HS (and heparin) biosynthesis is dictated by positioning of these units in relation to N-acetyl and N-sulfate substituents. Taken together, the results presented here show that the potentially susceptible residues are either irreversibly converted to IdoA or escape an encounter with the enzyme altogether. DISCUSSION A wealth of information points to strict control in HS biosynthesis to yield polysaccharide chains of regulated composition and domain organization. HS preparations isolated from different mammalian organs thus differ in structure, whereas those obtained from the same organs of different individuals appear similar (19 -22). The HS generated in a single tissue (human aortic wall) shows gradual structural change with increasing age of the individual (21,22). Immunohistochemical application of various anti-HS antibodies demonstrate cellspecific expression of HS epitopes of different structure (23,24). Regulation presumably applies also to the GlcA-IdoA conversion in HS biosynthesis, judging from the variable content and domain distribution of IdoA units in HS chains. The potential extent of IdoA formation is restricted by the N-substitution pattern that is established through the GlcNAc N-deacetylation/N-sulfation step before (or along with) GlcA C5-epimerization (3). However, structural analysis of HS chains after completed biosynthesis indicates that the potential target GlcA units (in -GlcNS-GlcA-GlcNS-and -GlcNS-GlcA-GlcNAc-sequences) are only partly converted to IdoA. Thus, additional regulatory factors must come into play.
Reaction of the appropriate -GlcNS-HexA-GlcNS-substrate with C5-epimerase in a soluble system is freely reversible and yields an equilibrium product containing GlcA and IdoA in approximately a 2:1 ratio (10). Yet heparin, as well as Nsulfated domains in HS, shows extended sequences of consecutive IdoA-containing disaccharide units (25). It has been suggested that IdoA generation may be promoted by concomitant O-sulfation, coordinated to block back-epimerization to GlcA (8,26,27). A model of "kinetically controlled" polymer modification predicts that the regulation and, hence, the fine structure of the final HS product, be dictated by the relative amounts of the various enzymes involved (C5-epimerase and O-sulfotransferases) and their substrate specificities. In accordance with this proposal, the HexA 2-O-sulfotransferase shows a strong preference for IdoA over GlcA target units (28), and HexA2S  Table II.

H]Gal/ [ 14 C]Gal) HS
The various HexA-aMan R disaccharide species are derived from -Glc-NS-HexA-GlcNS-sequences in the intact polysaccharide. Disaccharides were isolated by gel chromatography (Fig. 5) and were further separated by anion exchange HPLC (Fig. 6), paper electrophoresis, and paper chromatography (Fig. 7). The values represent means of two independent analytical runs. The 16-mer fraction from Fig. 5 is included to provide a reference ratio essentially reflecting a (GlcNAc-GlcA-) n sequence.  (-GlcNS-GlcA-GlcNS-and -GlcNS-GlcA-GlcNAc-) as compared with non-target sequences (-GlcNAc-GlcA-GlcNS-and -Glc-NAc-GlcA-GlcNAc-). The 3 H/ 14 C ratios of variously sized oligomers indicated retention of potentially labile C5-3 H atoms (Table I). Moreover, detailed examination of disaccharide and tetrasaccharide fractions showed the same 3 H/ 14 C ratios for potential target and non-target GlcA units. Thus, all residual GlcA units fully retain the 3 H label, irrespective of position in the HS chain. These findings demonstrate that the GlcA C5epimerase reaction in HS biosynthesis is irreversible and does not approach equilibrium conditions. Similar conditions may apply to dermatan sulfate biosynthesis (31). The formation of IdoA units is the result of single encounters between susceptible GlcA residues and the epimerase, which are terminated once L-ido configuration has been attained. The (poorly understood) design of the biosynthetic "assembly line" rather than the amounts of the various enzymes is the key factor in control of polymer modification. Indeed, mouse embryos heterozygous with regard to the epimerase expressed about half the enzyme levels of wild-type littermates, yet produced HS of indistinguishable composition (14).
Incubation of the isolated, labeled HS with exogenous epimerase failed to release any significant amounts of 3 H from the polysaccharide, suggesting that all potential target sites generated during the N-deacetylation/N-sulfation phase of the biosynthetic process had been blocked by subsequent O-sulfation. Although this O-sulfation apparently occurred before any reversion of C5-epimerization, its regulatory role is unclear. Analysis of disaccharides derived from the N-sulfated domains indicated that most of the IdoA units were 2-O-sulfated, which could explain resistance toward back-epimerization (Fig. 6). The IdoA residues in domains composed of alternating N-acetylated and N-sulfated disaccharide units, which typically account for about half the total IdoA in HS chains, are rarely 2-O-sulfated, whereas adjacent 6-O-sulfation is common (20). Both GlcA-and IdoA-containing sequences are readily attacked by 6-O-sulfotransferases (32), and HexA C5-epimerization is precluded by adjacent 6-O-sulfate groups (17). O-Sulfotransferases, working in close association with the epimerase, thus may restrict the extent of epimerization. In view of the virtually irreversible GlcA C5-epimerization, it seems more difficult to explain how O-sulfation may promote IdoA formation.
Previous generation of microsomal heparin-related polysaccharide yielded sulfated products that released 3 H from [5-3 H]GlcA residues upon subsequent incubation with exogenous epimerase (29). Although this phenomenon was presumably an artifact of the experimental system (incomplete Osulfation), it nevertheless pointed to a sequestration of epimerase and substrate that apparently applies also to the intact biosynthetic apparatus. Following rapid interaction between the epimerase and its substrate, the polysaccharide chains are physically or functionally dislocated from the enzyme. The lack of reversibility in HexA C5-epimerization thus has bearing on general aspects of the HS biosynthetic apparatus and its organization.