Location of N-Unsubstituted Glucosamine Residues in Heparan Sulfate*

Functional properties of heparan sulfate (HS) are generally ascribed to the sulfation pattern of the polysaccharide. However, recently reported functional implications of rareN-unsubstituted glucosamine (GlcNH2) residues in native HS prompted our structural characterization of sequences around such residues. HS preparations were cleaved with nitrous acid at either N-sulfated or N-unsubstituted glucosamine units followed by reduction with NaB3H4. The labeled products were characterized following complementary deamination steps. The proportion of GlcNH2 units varied from 0.7–4% of total glucosamine in different HS preparations. The GlcNH2 units occurred largely clustered at the polysaccharide-protein linkage region in intestinal HS, also more peripherally in aortic HS. They were preferentially located within N-acetylated domains, or in transition sequences between N-acetylated andN-sulfated domains, only 20–30% of the adjacent upstream and downstream disaccharide units being N-sulfated. The nearest downstream (toward the polysaccharide-protein linkage) hexuronic acid was invariably GlcUA, whereas the upstream neighbor could be either GlcUA or IdoUA. The highly sulfated butN-unsubstituted disaccharide unit, -IdoUA2S-GlcNH26S-, was detected in human renal and porcine intestinal HS, but not in HS from human aorta. These results are interpreted in terms of a biosynthetic mechanism, whereby GlcNH2 residues are formed through regulated, incomplete action of an N-deacetylase/N-sulfotransferase enzyme.

teins, sometimes through highly specific sequence motifs in the HS chain (see also reviews in Refs. 7 and 8). The proteinase inhibitor antithrombin thus binds to a unique pentasaccharide sequence that contains a rare 3-O-sulfated D-glucosamine (GlcN) unit (9). Other such rare constituents are 2-O-sulfated D-glucuronic acid (GlcUA) and N-unsubstituted GlcN residues (2). 2 HS and the structurally related heparin are both synthesized through a non-sulfated precursor structure composed of alternating GlcUA and N-acetylated GlcN (GlcNAc) units (2,4,10,11). This precursor is modified through a series of enzymatic reactions, initiated by N-deacetylation and N-sulfation of Glc-NAc residues. The resultant N-sulfated GlcN (GlcNS) residues are prerequisite to subsequent modification, involving C5-epimerization of GlcUA to L-iduronic acid (IdoUA), O-sulfation at C2 of the hexuronic acid (HexUA, i.e. GlcUA or IdoUA) and O-sulfation at C6 of GlcNS or GlcNAc units, or, less common, at C3 of GlcNS. Heparin is more extensively modified, resulting in a saccharide structure highly enriched in IdoUA2S-GlcNS6S disaccharide units, whereas the modification of HS chains is more restrained. The HS structure is typically heterogeneous, with blocks of N-sulfated sequences (NS-domains) interspersed between unmodified, N-acetylated regions (NA-domains) and mixed NA/NS domains that consist of alternating N-acetylated and N-sulfated disaccharide units (12). O-Sulfate groups and IdoUA units occur in the NS-and NA/NS-but not in the NA-domains. The domain distribution, length, and modification pattern vary considerably between different HS species, depending on tissue of origin, developmental stage, and pathophysiological condition (12)(13)(14)(15)(16) (see also reviews in Refs. 4, 8, and 11).
Due to their varied structure HS molecules interact in distinct fashion with different proteins. The rare structural components presumably contribute to selective protein binding. The occurrence of N-unsubstituted GlcN (GlcNH 2 ) units was early recognized but largely ignored, since it was believed to reflect artificial loss of N-sulfate groups during handling of HS samples (17). However, findings of recent years not only revealed GlcNH 2 residues in native HS structures, but also implicated such components with important cell-biological and pathophysiological phenomena. A monoclonal antibody that recognized GlcNH 2 units in HS thus bound in selective fashion to extracellular tissue components in fresh-frozen rat kidney (18). The presence of GlcNH 2 residues was found to correlate with the ability of bovine and human endothelial HS to bind L-selectin (although binding was not critically dependent on the GlcNH 2 residues) (19). Moreover, GlcNH 2 units were identified as targets for a 3-O-sulfotransferase isoform (3-OST-3A) that introduces a sulfate substituent at C3 (20,21) and thereby * This work was supported by grants from the Swedish Foundation for Strategic Research, Grant QLK- CT-1999.00536 from the European Commission program, and from the Polysackaridforskning AB (Uppsala). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The overall contents of GlcNH 2 residues, generally low, vary between HS species. Values ranging from 1.2 to 7.5% of total GlcN were calculated for porcine and bovine HS preparations, based on reaction with o-phthaldialdehyde (26). The location of these units in relation to the various structural domains of HS chains has not been established. While some studies place GlcNH 2 residues in transition zones between modified (largely N-sulfated) and unmodified (largely N-acetylated) regions of the polymer (18, 26,27), the target GlcNH 2 unit for the 3-OST-3A O-sulfotransferase occurs downstream 3 of an adjacent 2-O-sulfated IdoUA unit (20,21), typically found in modified NS-domains (12). We here report a comprehensive structural analysis of GlcNH 2 -containing sequences in HS from human aorta and kidney and from porcine intestine.

EXPERIMENTAL PROCEDURES
Materials-HS preparations from human aorta and human kidney were isolated as described (13,15). HS from porcine intestinal mucosa was a gift from G. van Dedem (Diosynth, Oss, The Netherlands). Bovine lung heparin (The Upjohn Co.) was purified as described (28).
Deaminative Cleavage of Heparan Sulfate with Nitrous Acid and Radiolabeling of Products-Stock solutions of HS in water were prepared and concentrations (ϳ3 mg of saccharide/ml) were determined by the meta-hydroxydiphenyl method (29). Before HNO 2 treatment, the HS chains were reduced with unlabeled NaBH 4 to ascertain that subsequent radiolabeling was exclusively restricted to deamination products. The HS (ϳ2 mg/ml) was incubated with NaBH 4 (ϳ20 mg/ml) for 3 h at room temperature in a total volume of 200 l of water (pH adjusted to 8 -9 by NaOH). The reaction mixture was acidified with acetic acid to pH 4 and then immediately neutralized with NaOH.
Deamination of saccharide chains with nitrous acid was done at either pH 3.9 or pH 1.5, to effect cleavage at N-unsubstituted or Nsulfated GlcN residues, respectively, in strictly selective fashion (30). In both reactions the freshly prepared reagent was added to the dry HS/heparin sample in capped plastic tubes and incubated for 10 min at room temperature. The reaction was stopped by addition of 2 M Na 2 CO 3 to pH 8 -9, and the cleavage products were then either immediately radiolabeled by NaB 3 H 4 reduction, reduced with unlabeled NaBH 4 , or analyzed by gel chromatography in non-reduced form, as indicated (see Refs. 31 and 32 for methods). Radiolabeling was performed by adding 0.5 mCi of NaB 3 H 4 (64 Ci/mmol, Amersham Biosciences) to the reaction mixture (containing Յ150 g of HS), which was then incubated overnight at room temperature. The reaction was interrupted by addition of acetic acid to pH 4 (in a fume hood), followed by neutralization with NaOH (to pH 7-8). The radiolabeled HS oligosaccharides (Ն2-mers) were then generally separated from non-incorporated radiolabel on a 1 ϫ 55 cm gel filtration column of Sephadex G-15 (Amersham Biosciences) in 0.2 M NH 4 HCO 3 at a flow rate of 12 ml/h.
To determine the content of GlcNH 2 residues in the different HS species, 150 g of HS (reduced with unlabeled NaBH 4 ) was treated with 500 l of HNO 2 -pH 3.9 reagent, and the reaction products were reduced with NaB 3 H 4 as described above. The resultant radiolabeled oligosaccharides were recovered and quantified by scintillation counting, the 3 H incorporated indicating GlcNH 2 units in the HS starting material. Conversion of radioactivity into molar terms was achieved through use of known amounts of bovine lung heparin as a standard that was deaminated at pH 1.5 to achieve essentially complete degradation into disaccharides. The amount of labeled disaccharides formed upon reduction with NaB 3 H 4 provided a measure of specific activity, cpm of 3 H/mol disaccharide that was used to estimate the molar amounts of GlcNH 2 in HS samples.
Sequence Analysis Upstream and Downstream of GlcNH 2 Units-Structures upstream of GlcNH 2 residues were radiolabeled through deamination of HS (150 g) at pH 3.9, followed by reduction of the products with NaB 3 H 4 (see condensed outline in Fig. 2). The labeled cleavage products were recovered, separated by gel chromatography on columns (1 ϫ 200 cm) of Sephadex G-15, eluted with 0.2 M NH 4 HCO 3 at a flow rate of ϳ6 ml/h, and pooled into three size classes, Ն6-mer, 4-mer, and 2-mer. The Ն6-mer and 4-mer fractions were subjected to HNO 2 -pH 1.5 treatment, reduced with unlabeled NaBH 4 and again separated on Sephadex G-15. The secondary cleavage products obtained from the Ն6-mer fraction were pooled into new size classes designated Ն6Ј-mer, 4Ј-mer, and 2Ј-mer. Alternatively, the initial Ն6-mers were first separated according to size by Biogel P-10 (Bio-Rad) gel chromatography (1 ϫ 200 cm column, 0.5 M NH 4 HCO 3 at a flow rate of 2 ml/h). Each even-numbered oligomer (6 -12-mer) was then treated separately with HNO 2 -pH 1.5 and analyzed, in parallel with an untreated control, by gel chromatography on a Superdex 30 column (Amersham Biosciences) eluted with 0.5 M NH 4 HCO 3 .
Structures downstream of GlcNH 2 units (see outline in Fig. 5) were studied following deamination of 150 g of HS at pH 1.5 and end-group radiolabeling of products. Labeled oligosaccharides were recovered (Sephadex G-15, 1 ϫ 55 cm) and desalted. Oligomers Ն4-mer were further fractionated by Biogel P-10 gel chromatography, as described above. Each size class (4-to Ն18-mer) was pooled and desalted. The oligosaccharides were then separately treated with HNO 2 -pH 3.9 and analyzed, in parallel with the corresponding untreated controls, by Superdex 30 gel chromatography. To confirm that the first deaminative cleavage at pH 1.5 had gone to completion, each oligosaccharide was separately subjected to a second treatment with HNO 2 -pH 1.5 and analyzed on the Superdex 30 column.
Specificity of HNO 2 -pH 3.9 Procedure-To ascertain the cleavage specificity of the HNO 2 -pH 3.9 procedure, 150 g of full-length human aortic HS and ϳ1 ϫ 10 6 cpm of radiolabeled HS 10-mer (obtained by HNO 2 -pH 1.5/NaB 3 H 4 treatment of full-length human aortic HS) were N-acetylated essentially as described (33). Samples were dissolved in 200 l of 0.5% Na 2 CO 3 , 10% MeOH and put on ice. Acetic anhydride (Merck Eurolab AB) was then added in five 20-l portions over 1 h, while the pH was kept at 7 by alternating additions of solid Na 2 CO 3 and 10% MeOH. Saccharides were recovered after gel chromatography on Sephadex G-15 (1 ϫ 15 cm column) in 0.5 M sodium acetate/1 M NaCl and were then desalted on PD-10 columns (Amersham Biosciences). The N-acetylated samples were treated, in parallel with control samples not subjected to N-acetylation, with nitrous acid at pH 3.9. The unlabeled HS samples were reacted with NaB 3 H 4 , and the products were analyzed by gel filtration (Sephadex G-15, 1 ϫ 55 cm). The radiolabeled HS decamers were directly analyzed by Superdex 30 gel filtration.
Structural Characterization of Oligosaccharides-The nonreducingterminal HexUA of tetrasaccharides was identified by digestion with bovine liver ␤-glucuronidase (Oxford GlycoSciences, Abingdon, UK). Samples (ϳ5000 cpm of 3 H) were incubated for 16 h at 37°C with 80 units of enzyme in 25 l of 50 mM sodium acetate, pH 5.0, containing 100 g/ml bovine serum albumin. The products, and undigested controls, were analyzed by gel chromatography on Superdex 30. The specificity of the enzyme preparation was confirmed by incubation with GlcUA-[ 3 H]aMan R and IdoUA-[ 3 H]aMan R disaccharides, followed by anion-exchange HPLC of the incubation products (data not shown). 3 H-end-labeled tetrasaccharides containing an internal N-acetylated GlcN unit were cleaved by treatment with nitrous acid at pH 3.9 following N-deacetylation by hydrazinolysis (12), and labeled products were identified by the procedures indicated below.
O-Sulfated HexUA-[ 3 H]aMan R disaccharides were analyzed by anion-exchange HPLC on a Partisil-10 SAX column (Whatman), eluted with a stepwise gradient of KH 2 HPO 4 (0 -0.4 M) (31). The elution pattern was monitored either with a Radiomatic Flow Scintillation Counter (PerkinElmer Life Sciences) or by collecting 1-ml fractions that were analyzed by scintillation counting. Non-sulfated disaccharides were separated from mono-and disulfated disaccharides by high-voltage paper electrophoresis at pH 5.3 and were then separated by paper chromatography into GlcUA-and IdoUA-containing species that were quantified by scintillation counting as described (34).

Contents of N-Unsubstituted Glucosamine Residues in Heparan Sulfate Preparations
HS preparations from three sources were investigated, i.e. human aorta, human kidney, and porcine intestine. To deter-mine the contents of GlcNH 2 units, each HS species was treated with nitrous acid at pH 3.9 to specifically convert these units to terminal 2,5-anhydromannose residues. Reduction with NaB 3 H 4 afforded [ 3 H]aMan R residues, the incorporated radioactivity indicating the GlcNH 2 contents of the original HS samples. The molar proportions of GlcNH 2 residues were calculated based on the radioactivity incorporated into heparin standard disaccharides that had been reduced with the same batch of NaB 3 H 4 . The GlcNH 2 content varied between the HS preparations and was highest (4% of total disaccharide units) in human aortic HS, intermediate (2%) in human renal HS, and lowest (0.7%) in porcine intestinal HS. Approximate estimates of molecular size by gel chromatography gave peak values for the intact intestinal and aortic HSs of ϳ30 kDa and ϳ45 kDa, respectively, both preparations being quite polydisperse (Fig.  1). These data allowed us to estimate the average number of GlcNH 2 units per polysaccharide chain, ϳ0.5 for intestinal HS and ϳ4 for aortic HS.

TABLE I Characterization of products obtained by deamination and radiolabeling of N-unsubstituted glucosamine residues in heparan sulfates
HS samples were treated with HNO 2 at pH 3.9, and the products were reduced with NaB 3 H 4 . The resultant labeled oligosaccharides (2-, 4-, and Ն6-mers) were recovered by gel chromatography (Sephadex G-15; see Fig. 3A) and quantified. The amounts of GlcNH 2 residues were calculated using labeled disaccharides from heparin as standard. The 4-and Ն6-mers were then reacted with HNO 2 at pH 1.5 and the products (given an X'-designation) again separated on Sephadex G-15. close to either chain terminus. End-group-labeled deamination products obtained by reduction with NaB 3 H 4 contained 20 -30% each of di-and tetrasaccharides, the remaining ϳ50% being larger than 4-mers (gel chromatography data not shown; Table I). The size distribution of the latter components approximated a number-average representation of the intact chains ( Fig. 1A), suggesting a GlcNH 2 residue located toward the carbohydrate-protein linkage region. The formation also of smaller deamination products, di-and tetrasaccharides, points to the presence of two or more clustered GlcNH 2 residues in a small proportion of the chains, alternatively a single GlcNH 2 residue located very close to the non-reducing chain terminus. By contrast, gel chromatography of aortic HS following HNO 2 -pH 3.9 treatment showed polydisperse products ranging from ϳ40 kDa to Ͻ10 kDa in size, only partly overlapping the intact untreated polysaccharide (Fig. 1B). Notably, these elution patterns were based on colorimetric HexUA analysis, and the deamination products observed thus were derived from regions upstream 3 as well as downstream of any GlcNH 2 units in the chains. End-group-labeled deamination products fell largely into the low molecular weight range, more than half being smaller than 12-mers (Figs. 1B, 3; Tables I and II). Most of these oligosaccharides had presumably been generated by excision of saccharide domains between GlcNH 2 units. Each chain of aortic HS contained on average ϳ4 such residues, hence the appreciable depolymerization of the unlabeled HS chains upon deamination (Fig. 1B). Chemical N-acetylation of the HS precluded depolymerization as well as the formation of labeled oligosaccharides upon subsequent HNO 2 -pH 3.9/ NaB 3 H 4 treatment (data not shown).

Characterization of Saccharide Sequences Upstream of N-Unsubstituted Glucosamine Residues
N-Substituent Pattern-The N-substituent pattern of sequences upstream of GlcNH 2 was assessed by characterization of the labeled oligomers obtained after HNO 2 -pH 3.9/NaB 3 H 4 treatment of polysaccharide (procedure outlined in Fig. 2). The three HS samples all yielded labeled 2-mer, 4-mer, and Ն6mer, although in somewhat variable proportions (Table I; gel chromatogram shown for aortic HS in Fig. 3A). The disaccharides would represent a sequence of two adjacent GlcNH 2 units with an intervening HexUA residue (sequence a in Fig. 9), whereas the tetrasaccharides were derived from a (-GlcNH 2 -)HexUA-GlcNR-HexUA-GlcNH 2 -structure (where R is an Nacetyl or an N-sulfate group). 4 To identify the N-substituents of the internal GlcNR units, the tetrasaccharides were reacted with HNO 2 at pH 1.5 (cleavage at GlcNS residues), and the amounts of labeled disaccharide released were determined by gel chromatography on Sephadex G-15 (not shown). Thus 12-22% of the tetrasaccharides were found to contain an internal GlcNS unit, the remainder being N-acetylated (Table I; note Ј-designation to distinguish products from fragments generated in the initial HNO 2 -pH 3.9 deamination). The corresponding intact sequences in aortic HS are illustrated by structures b and c in Fig. 9.
The major fractions of labeled Ն6-mers recovered after HNO 2 -pH 3.9/NaB 3 H 4 treatment represent sequences of Ն5 monosaccharide units interspersed between GlcNH 2 residues in the intact HS chain. Deamination at pH 1.5 of the entire Ն6-mer fractions resulted in strikingly similar degradation patterns (not shown) for the three HS species, with increasing amounts of 2Ј-mer, 4Ј-mer, and residual Ն6Ј-mer, in the order FIG. 3. Gel chromatography of 3 H-labeled saccharides obtained by HNO 2 -pH 3.9/NaB 3 H 4 treatment of HS. HS from human aorta was cleaved at GlcNH 2 units and the products were radiolabeled (HNO 2 -pH 3.9/NaB 3 H 4 treatment) and separated by gel chromatography on Sephadex G-15 (A). The resultant Ն6-mers were separated further by gel chromatography on BioGel P-10 (B). Subfractions were recovered as indicated and separately deaminated at pH 1.5 (Fig. 4).

FIG. 4. Cleavage at GlcNS residues of 3 H-labeled saccharides obtained by HNO 2 -pH 3.9/NaB 3 H 4 treatment of HS.
Labeled oligomers formed on HNO 2 -pH 3.9/NaB 3 H 4 treatment of aortic HS were separated (Fig. 3) and 6-to 12-mer fractions were individually reacted with HNO 2 at pH 1.5. Each sample was analyzed by gel chromatography on Superdex 30 before (solid line) and after (broken line) HNO 2 -pH 1.5 treatment. Oligomer size is indicated above each peak. mentioned ( Table I). The yield of disaccharide defines the proportion of sequence d in Fig. 9, whereas the tetrasaccharides account for part of sequence e; about 25% of the GlcNH 2 residues in aortic HS are linked to an upstream -GlcNS-HexUA-GlcNAc-HexUA-structure. To obtain information about more remote upstream regions, the initial Ն6-mer fraction of aorta HS (HNO 2 -pH 3.9 cleavage) was separated further by gel chromatography on BioGel P-10 (Fig. 3B). A series of poorly resolved components was obtained, including an appreciable proportion of material excluded from the gel (Ն10 kDa; see also Fig. 1B). The parent HS chains thus contained regions with closely adjacent GlcNH 2 residues as well as extended regions lacking such residues. The clearly discernible 6-to 12-mer fractions were recovered and separately subjected to HNO 2 -pH 1.5 treatment (Fig. 4, A-D). A major portion of each end-labeled species was fragmented into smaller, even-numbered oligomers, indicative of N-sulfate groups. The oligosaccharide patterns locate GlcNS units to all potential positions upstream of GlcNH 2 residues. Nevertheless, a fraction of each labeled oligosaccharide retained the elution position of the initial HNO 2 -pH 3.9 deamination product, suggesting that two Glc-NH 2 units may be connected by fully N-acetylated -[HexUA-GlcNAc] n -HexUA-stretches of variable length. 4 The proportions of such deamination-resistant label were determined for each oligosaccharide in Fig. 4 and used to calculate the ratio between sequences fully N-acetylated (ϳ20%) and those containing one or more N-sulfate groups (ϳ80%) ( Table II; sequences f and e, respectively, Fig. 9).
Hexuronic Acid and O-Sulfate Residues-Substrate recognition by the GlcUA C5-epimerase that converts GlcUA to IdoUA residues requires that the adjacent upstream GlcN unit be N-sulfated, whereas the adjacent downstream GlcN unit may be either N-acetylated or N-sulfated (35). The HexUA in -Glc-NAc-HexUA-GlcNAc-or -GlcNAc-HexUA-GlcNS-sequences thus is invariably GlcUA, whereas that in -GlcNS-HexUA-GlcNAc-or -GlcNS-HexUA-GlcNS-may be either GlcUA or IdoUA. The influence of an N-unsubstituted GlcN residue on the C5-configuration of nearby HexUA units has not been in-

FIG. 5. Scheme of selective cleavage and radiolabeling procedures to define sequences downstream of GlcNH 2 units in HS.
Full-length HS was deaminated with HNO 2 at pH 1.5 to cleave chains at GlcNS units. Reduction of the resultant terminal 2,5-anhydromannose residues with NaB 3 H 4 yielded labeled (asterisk) [1-3 H]aMan R units. Radiolabeled oligomers were then deaminated at pH 3.9 to induce cleavage at the nearest upstream GlcNH 2 residue. Identity of HexUA and occurrence of O-sulfate groups are ignored in the scheme. The symbols are as in Fig. 2 (Fig. 3A), and fractions corresponding to Ն6-mers were pooled and further deaminated at pH 1.5. The labeled disaccharide products were isolated and analyzed by anion-exchange HPLC and paper chromatography (see "Experimental Procedures").

HS sample
GlcUA-aMan R   3A), and fractions corresponding to Ն6-mers were separated further by gel chromatography on BioGel P-10 (Fig. 3B). Fractions were pooled and deaminated with HNO 2 at pH 1.5, and the products were separately analyzed by gel chromatography (Fig. 4). The proportions of labeled components remaining at the elution position of the corresponding intact oligosaccharide were calculated and assumed to indicate the amounts of fully N-acetylated sequence. The degradation products derived from oligomers containing one or more N-sulfated GlcN units in various positions were quantified and used to calculate the proportion of partially N-sulfated sequence. [  a N.D., not determined. Oligosaccharides Ն14-mers were poorly resolved (Fig. 3B), and were therefore not subjected to deamination at pH 1.5. The proportions of fully N-acetylated and partially N-sulfated sequence were instead calculated assuming that 10% of these fragments resisted deamination. vestigated. We therefore identified the labeled disaccharide released by HNO 2 -pH 3.9/NaB 3 H 4 treatment of aortic HS, and found GlcUA-[ 3 H]aMan R as the only identifiable component (by anion-exchange HPLC and paper chromatography; data not shown), corresponding to a (-GlcNH 2 -)GlcUA-GlcNH 2 -sequence (a in Fig. 9) in the intact polymer.
The Ն6-mers obtained by HNO 2 -pH 3.9/NaB 3 H 4 treatment were deaminated at pH 1.5, and the resultant labeled 2Ј-mers were identified to provide information regarding the native -GlcNS-HexUA-GlcNH 2 -structures (Fig. 9d). Appreciable variability was observed (Table III), with both GlcUA and IdoUA residues immediately upstream of the GlcNH 2 unit, and Osulfate groups both at C2 of IdoUA and at C6 of GlcNH 2 . Notably, the relative amounts of the different disaccharides differed considerably between the HS species. Analysis of [ 3 H]disaccharides from sequence b in aortic HS gave a pattern (not shown) essentially similar to the corresponding disaccharide relating to sequence d (Table III). Labeled 4Ј-mers and Ն6Ј-mers ([HexUA-GlcNAc] Ն1 -HexUA-[ 3 H]aMan R oligosaccharides) obtained after HNO 2 -pH 1.5 treatment were analyzed to identify the radiolabeled reducing-terminal disaccharide unit. Cleavage of the oligosaccharides by HNO 2 -pH 3.9 treatment following hydrazinolysis yielded GlcUA-[ 3 H]aMan R exclusively (data not shown). The corresponding native sequence thus, as expected, is identified as -GlcNAc-GlcUA-GlcNH 2 - (Fig. 9, e and f).

Characterization of Saccharide Sequences Downstream of N-Unsubstituted Glucosamine Residues
N-Substituent Pattern-The N-substituent pattern of sequences downstream of GlcNH 2 in HS preparations was assessed by characterization of the labeled oligomers obtained after HNO 2 -pH 1.5/NaB 3 H 4 treatment of polysaccharide (procedure outlined in Fig. 5). Of the total 3 H label incorporated into deamination products from aortic and renal HS, 42 and 55%, respectively, appeared in disaccharides. Products Ն4-mer (58 and 45%, respectively, of the label) were separated by gel chromatography (BioGel P-10) into a series of distinct fractions, with a dominant tetrasaccharide peak (shown for aortic HS in Fig. 6). No significant change in elution pattern was observed following repeated HNO 2 -pH 1.5 treatment of each product and analysis by Superdex 30 gel chromatography (data not shown), indicating that cleavage at GlcNS units had been quantitative. Each fraction (4-to 14-mers) was separately treated with HNO 2 at pH 3.9 and subjected to Superdex 30 chromatography (Fig. 7, A-G). Distinct patterns of minor labeled degradation products were obtained, again with predominant 4-mer components. To ascertain that release of these products was indeed due to cleavage at GlcNH 2 units, labeled 10-mer was subjected to N-acetylation (treatment with acetic anhydride; see "Experimental Procedures") before reaction with HNO 2 -pH 3.9. No formation of smaller degradation products was observed (Fig. 7H).
The proportions of labeled oligosaccharides obtained upon

FIG. 6. Gel chromatography of 3 H-labeled saccharides obtained by HNO 2 -pH 1.5/NaB 3 H 4 treatment of HS.
HS from human aorta was cleaved at GlcNS units and the products were radiolabeled (HNO 2 -pH 1.5/NaB 3 H 4 treatment) and isolated by gel chromatography on Sephadex G-15 (not shown). Labeled Ն4-mers were isolated and separated by gel chromatography on BioGel P-10. Fractions were pooled as indicated and further reacted with HNO 2 at pH 3.9 (Fig. 7).

FIG. 7. Cleavage at GlcNH 2 residues of 3 H-labeled saccharides obtained by HNO 2 -pH 1.5/NaB 3 H 4 treatment of HS.
Labeled oligomers generated by HNO 2 -pH 1.5/NaB 3 H 4 treatment of aortic HS, and separated by BioGel P-10 chromatography (Fig. 6), were each reacted with HNO 2 at pH 1.5. The products were analyzed by gel chromatography on Superdex 30 (A-F). Solid line, untreated samples; broken line, HNO 2 -pH 3.9treated samples. Oligomer size is indicated above each peak. The 12-mer chromatograms are also shown in scale to fully include the peak of undegraded material (G). Separate chromatograms of Nacetylated 10-mer show the complete resistance of this material to deamination (H). The inset highlights the critical region of the chromatograms. HNO 2 -pH 3.9 treatment were used to assess N-substituent patterns of sequences downstream of GlcNH 2 units in the intact HS chains (Table IV). The results for aortic and renal HS were highly similar. The results for aortic HS are illustrated in Fig. 9 (where sequences a*, c*, and f*, not labeled in this approach, were deduced from the analysis of upstream structures a, c, and f, respectively). 4 While about one-third of the adjacent downstream GlcN neighbor residues were N-sulfated (Fig. 9, sequence g), most of the structures showed a GlcNAc unit in this position (sequences c*, h, i, f*). A major proportion of the following adjacent downstream GlcN residues were N-sulfated (sequence h), although smaller proportions of consecutive N-acetylated disaccharide units were also observed (i and f*).
Hexuronic Acid Residues-Labeled di-and tetrasaccharides released by deamination at pH 3.9 from the aortic HS 8-mer were analyzed in more detail to define the HexUA residues as well as any O-sulfate groups downstream the GlcNH 2 units. The disaccharide (corresponding to sequence g in Fig. 9) was identified as GlcUA-[ 3 H]aMan R , by anion-exchange HPLC and paper chromatography. No O-sulfate groups were detected (data not shown). Digestion with ␤-D-glucuronidase resulted in quantitative conversion of the tetrasaccharide (sequence h) into trisaccharide, thus identifying the terminal HexUA as GlcUA (Fig. 8). We conclude from these results that the HexUA immediately downstream of a GlcNH 2 residue is exclusively Gl-cUA and not IdoUA. DISCUSSION The aim of the present study was to provide comprehensive information regarding the location of GlcNH 2 residues in HS, of particular importance in view of recent reports on functional implications of such units (19 -23). A variety of techniques have been used in previous studies of GlcNH 2 residues in HS, and the conclusions have been diverse. Two studies located such units to transition zones between modified (N-sulfated) and unmodified (N-acetylated) saccharide regions (26,27). Moreover, in their study of glypican-1-linked HS produced by cultured endothelial cells Ding et al. (27) proposed that GlcNH 2 residues would be preferentially located close to the polysaccharide-protein linkage region. A GlcNH 2 unit was identified in a completely nonsulfated oligosaccharide epitope recognized by the monoclonal antibody 10E4 (23). By contrast, the -IdoUA2S-GlcNH 2 -sequence identified as target for the 3-OST-3A sulfo-transferase (20,21) would be located in the highly sulfated NS-domains of HS chains (12). Further, the increased formation of GlcNH 2 units attributed to inhibition of endogenous polyamine biosynthesis was found to occur primarily in highly modified, peripheral sections of HS chains (36).
We used procedures based on NaB 3 H 4 reduction of fragments generated by selective deaminative cleavage to detect and quantify GlcNH 2 units, and to define their structural context. The quantity of GlcNH 2 was indicated by the level of 3 H incorporation achieved following HNO 2 -pH 3.9 treatment of HS. The results ranged from 0.7 to 4% of total disaccharide units in three different samples, in fair agreement with previous reports based on other techniques (18,19,26). The specificity of our approach was ascertained by chemical N-acetylation of parent polysaccharide samples, which were found to preclude 3 H incorporation.
We attempted to locate GlcNH 2 units in HS chains by selective cleavage of the polymers at the sites of such residues, followed by size analysis of unlabeled (HexUA detection) and 3 H-reduced products. All three HS samples examined yielded distinct di-and tetrasaccharide labeled products in addition to FIG. 8. Digestion with ␤-glucuronidase of tetrasaccharide representing sequence immediately downstream of GlcNH 2 unit in aortic HS. The labeled 8-mer generated by HNO 2 -pH 1.5/NaB 3 H 4 treatment of aortic HS was deaminated at pH 3.9, and the products were separated by gel chromatography as shown in Fig. 7. The tetrasaccharide fraction was isolated and digested with ␤-glucuronidase, and the digest was analyzed by Superdex 30 gel chromatography (broken line); undigested control (solid line). The peak elution positions of diand tetrasaccharides from heparin are indicated in the figure.

TABLE IV
Characterization of domains downstream of N-unsubstituted glucosamine residues in aortic and renal heparan sulfate [ 3 H]Oligosaccharides obtained by HNO 2 -pH 1.5/NaB 3 H 4 treatment of aortic and renal HS were fractionated by gel chromatography on Sephadex G-15 (not shown), and fractions corresponding to Ն4-mers were separated further by gel chromatography on BioGel P-10 (Fig. 6). Distinct oligomers were pooled, deaminated with HNO 2 at pH 3.9, and the products were separately analyzed by Superdex 30 chromatography (Fig. 7). The amounts of secondary labeled deamination products (designated XЈ-mers) were calculated for each parent oligomer fraction. larger fragments. Notably, the proportions of labeled 2-and 4-mers were inversely related to the proportion of GlcNH 2 units, and thus accounted for half of the total 3 H incorporated into intestinal HS, (ϳ0.5 GlcNH 2 residue per chain), but only for 25% of the labeled fragments obtained from aortic HS (ϳ4 GlcNH 2 residues per chain) (Table I and Fig. 9). Moreover, the generation of small labeled oligosaccharides was not accompanied by any significant depolymerization of intestinal HS, as evidenced by gel chromatography of unlabeled deamination products (Fig. 1A). The labeled Ն6-mers largely covered the range of unlabeled deamination products ( Fig. 1A; numberaverage representation of the labeled species), suggesting the incorporation of 3 H at the reducing end of oligosaccharides similar in size to the initial HS chains. These findings have been rationalized in a model showing two (or more) GlcNH 2 units located close to the polysaccharide-protein linkage region of the intestinal HS chain (Fig. 10A), thus in agreement with the model suggested by Ding et al. (27). The labeled 2-/4-mers would be derived from saccharide residues located between the GlcNH 2 units, whereas the larger [ 3 H]oligomers represent the residual, major non-reducing portion of the initial chain. Release of small oligosaccharides from the nonreducing end of the HS chain cannot be excluded, but appears unlikely since it would not account for the formation of the large labeled products seen (Fig. 1A). Since the intestinal HS contains on average only one GlcNH 2 unit for every two chains, clustering of such units as proposed in Fig. 10A (upper sequence) implies a majority of chains without any GlcNH 2 residues (lower sequence). By contrast, most or all of the aortic HS chains contain GlcNH 2 residues. Whereas the short labeled oligosaccharides observed (Table I and Fig. 3A) may again be attributable to clustered GlcNH 2 units close to the polysaccharide-protein linkage region, the marked depolymerization of aortic HS upon HNO 2 -pH 3.9 treatment (Fig. 1B) clearly points to the occurrence of GlcNH 2 also in more peripheral positions of the chains (Fig.  10B). The highly polydisperse size distribution of labeled Ն6mers (Fig. 3B) indicates variable length of saccharide sequences interspersed between adjacent GlcNH 2 residues. No-tably, a minor fraction of the radiolabeled fragments would represent the nonreducing termini of the HS chains.
The structures identified upstream and downstream of GlcNH 2 residues in the aortic HS are highly diverse (Fig. 9). While the analytical data do not enable us to link upstream and downstream sequences around a given GlcNH 2 residue, we note that most of the GlcNH 2 residues are flanked on both sides by one or more N-acetylated disaccharide units. About 20 -30% of the adjacent downstream as well as upstream disaccharide units are N-sulfated. Similar proportions, ϳ30% of the upstream and downstream sequences, contain a single N-acetylated disaccharide unit between the GlcNH 2 reference point and the nearest N-sulfated unit. About 10% of the GlcNH 2 residues in aortic HS occur immediately adjacent to each other whereas ϳ25% are joined by fully N-acetylated saccharide of variable length. Apart from a higher proportion of consecutive FIG. 9. Overview of GlcNH 2 -containing structures in aortic HS. The frequencies of all identified upstream and downstream sequences were assessed as described in the text. The frequencies of downstream sequences are based on the data in Table IV, after adjustment for sequences a*, c*, and f* that are identical to sequences a, c, and f. The indicated frequencies of a*, c* and f* type sequences are maximal values, calculated on the assumption that each of these sequences are located between two GlcNH 2 residues. 4 Identity of HexUA and the positions of O-sulfate groups have not been considered in this scheme. The symbols are as in Fig. 2. The ϩ sign in sequence e indicates the obligatory presence of at least one Nsulfated disaccharide unit. N-unsubstituted disaccharide units, the data for renal HS appeared similar to that of aortic HS, with a predominance of N-acetylated disaccharide units flanking the GlcNH 2 residues on both sides (see Tables I and IV). Altogether, these findings indicate that GlcNH 2 units are preferentially located in NA-or NA/NS-domains but are scarce in NS-domains. Part of these residues thus may occur in transition zones between modified (N-sulfated) regions and unmodified (N-acetylated) saccharide regions as proposed (26,27). On the other hand, our results do not support the notion (26) that GlcNH 2 units are generally positioned in regions downstream of heavily modified (sulfated, IdoUA-rich) and upstream of less modified domains. Interestingly, the IdoUA2S-GlcNH 2 -sequence implicated as acceptor in the 3-OST-3A sulfotransferase reaction (20,22) could be identified in our study (as part of sequence d in Fig. 9; Table  III), though in minute quantity only.
The mechanism of GlcNH 2 formation during HS biosynthesis remains unclear. Several possibilities may be envisaged. A fraction of the GlcNAc residues initially incorporated into the precursor polysaccharide could be left N-unsubstituted in the concerted N-deacetylation/N-sulfation process. Residual N-acetyl groups could be selectively removed at a later stage, even after completed biosynthetic modification of the HS chain. Finally, N-sulfate groups could be selectively eliminated by an endo-sulfamidase, during or after the later stages of polymer modification. While there is presently no way of excluding any of these alternatives, we note that no endo-sulfamidase has yet been described (however, see Ref. 37). On the other hand, the two reactions of the N-deacetylase/N-sulfotransferase process are readily segregated in experimental systems involving either purified N-deacetylase/N-sulfotransferase enzyme and (GlcUA-GlcNAc-) n polysaccharide substrate (38), or microsomal preparations incubated with sugar nucleotides in the absence of sulfate donor (PAPS) (39). Moreover, four distinct N-deacetylase/N-sulfotransferase isoforms have been described, and one of these, NDST-3, has much higher N-deacetylase compared with N-sulfotransferase activity than any of the other NDST enzymes (40). While a role for NDST-3 in the generation of GlcNH 2 units would seem mechanistically attractive, it appears questionable in view of the relatively restricted expression pattern of this particular enzyme. It may further be noted that no significant increase in GlcNH 2 residues was found neither in a Chinese hamster ovary cell mutant deficient in N-sulfation (41), nor upon overexpression of an N-sulfotransferase-deficient NDST1 mutant in 293 cells. 5 Depletion of PAPS, on the other hand, induced by treatment of cells with chlorate led to formation of HS with accumulated GlcNH 2 residues (41,42).
How is the formation of GlcNH 2 residues related to other polymer modification reactions? The exclusive D-gluco configuration of the downstream HexUA neighbor indicates that this unit was not recognized as a substrate for the GlcUA C5epimerase, in accord with the notion that the GlcNH 2 residues have never been N-sulfated. By contrast, while the vast majority (Ͼ90%) of the HexUA residues immediately upstream of GlcNH 2 are GlcUA, these units may be subject to C5-epimerization and 2-O-sulfation (Table III). Also, the GlcNH 2 residue itself may be 6-O-sulfated (Table III), and even 3-O-sulfated by a specific 3-OST isoform (20,21). The remarkable functional implications of the latter reaction in relation to herpes simplex virus 1 infection (22) raise intriguing questions regarding the regulation of GlcNH 2 formation, in concert with other polymer modification reactions. Inhibition of endogenous polyamine synthesis in cultured endothelial cells, hence up-regulated polyamine uptake, led to formation of HS chains with increased proportions of GlcNH 2 units. Interestingly, these residues appeared to be preferentially associated with highly modified, peripheral sections of the polysaccharide (27,36), i.e. predicted target sites for the 3-OST-3A sulfotransferase reaction (20,21).