Heparin/Heparan Sulfate Biosynthesis

Heparan sulfate (HS) proteoglycans influence embryonic development as well as adult physiology through interactions with various proteins, including growth factors/morphogens and their receptors. The interactions depend on HS structure, which is largely determined during biosynthesis by Golgi enzymes. A key step is the initial generation of N-sulfated domains, primary sites for further polymer modification and ultimately for functional interactions with protein ligands. Such domains, generated through action of a bifunctional GlcNAc N-deacetylase/N-sulfotransferase (NDST) on a [GlcUA-GlcNAc]n substrate, are of variable size due to regulatory mechanisms that remain poorly understood. We have studied the action of recombinant NDSTs on the [GlcUA-GlcNAc]n precursor in the presence and absence of the sulfate donor, 3′-phosphoadenosine 5′-phosphosulfate (PAPS). In the absence of PAPS, NDST catalyzes limited and seemingly random N-deacetylation of GlcNAc residues. By contrast, access to PAPS shifts the NDST toward generation of extended N-sulfated domains that are formed through coupled N-deacetylation/N-sulfation in an apparent processive mode. Variations in N-substitution pattern could be obtained by varying PAPS concentration or by experimentally segregating the N-deacetylation and N-sulfation steps. We speculate that similar mechanisms may apply also to the regulation of HS biosynthesis in the living cell.

Heparan sulfate proteoglycans act as coreceptors for growth factors and cytokines and are important for maintenance of morphogen gradients during development (1). The biological activities generally depend on charge interaction of heparan sulfate (HS) 3 chains with proteins and thus on the distribution of sulfate residues along the chains. This structural trait is essentially established during HS biosynthesis (2) but may also be modified through postbiosynthetic action of recently identified endosulfatases (3).
HS is synthesized in the Golgi network through complex, concerted action of several distinct enzymes ( Fig. 1) (2). First, a glucuronic acid-galactose-galactose-xylose (GlcUA␤1-3Gal␤1-3Gal␤1-4Xyl␤1-) "linkage tetrasaccharide" is formed that will connect the HS chain proper with a serine unit of a proteoglycan core protein. The association of all modifications with N-sulfated regions demonstrates the key importance of the NDSTs in designing HS structure. Except for the C5-epimerase and the 2-O-sulfotransferase, which are encoded by single genes in mammals, the modification enzymes occur in several isoforms (4). Four NDSTs (NDST1-4) have been identified, NDST1 and NDST2 being the most widely distributed (5).
The mechanisms behind the non-random distribution of N-acetylated and N-sulfated disaccharide units in HS are unknown. The N-deacetylation and N-sulfation reactions could be experimentally segregated using a heparin-producing mastocytoma microsomal system (6); only after identification of bifunctional NDSTs was it realized that both processes should be attributed to a single enzyme (7)(8)(9). In the present study, recombinant NDSTs were used to define the mode of enzyme action, in the absence and presence of the sulfate donor, 3Ј-phosphoadenosine 5Ј-phosphosulfate (PAPS). N-Deacetylation of a fully N-acetylated precursor polysaccharide in the absence of PAPS was found to be an essentially stochastic process generating a limited proportion of randomly distributed, N-unsubstituted glucosamine (GlcNH 2 ) residues. By contrast, the addition of PAPS converted the enzyme to a virtually processive mode of action, resulting in formation of extended sequences of consecutive, N-sulfated disaccharide units. Our findings thus suggest that key structural features of HS chains may be ascribed to mechanisms inherent to the NDST enzymes.

EXPERIMENTAL PROCEDURES cDNA Construct and Transfection
His-tagged NDST1-Full-length cDNA of wild type mouse NDST1 cloned in pBluescript SK vector was used as template in a PCR to introduce an enterokinase cleavage site followed by a His 6 tag in the C-terminal end of the NDST1 protein. Forward and reverse primers were 5Ј-ccttcaagcacctggtg-3Ј and 5Ј-gagaattcctagtgatggtgatggtgatgcttatcgtcatcgtccctggtgttctggaggtcttcccgcagcca-3Ј (His tag is in bold), respectively. The PCR product was cleaved with BglII and EcoRI and used to replace the corresponding sequence of wild type NDST1 in pBluescript SK. The resulting NDST1-His sequence was then cloned into expression vector pBud4.1CE (Invitrogen) using XhoI and NotI.
His 6 -tagged NDST2-Full-length cDNA of wild type mouse NDST2 cloned in pUC119 vector was used as template in a PCR to introduce an enterokinase cleavage site followed by a His 6 tag in the C-terminal end of the NDST2 protein. Forward and reverse primers were 5Ј-ctcaggaacgaagccccct-3Ј and 5Ј-tgtctagactagtgatggtgatggtgatgcttatcgtcatcgtcgcccacactggaatgttgcaattcttcccg-3Ј (His tag in bold), respectively. The PCR product was cleaved with XbaI and Kpn21 and used to replace the corresponding sequence of wild type NDST2 in pUC119. The resulting NDST2-His sequence was then cloned into expression vector pCDNA3 using XbaI and EcoRI.

Isolation of 14 C-labeled K5 Polysaccharide
K5 polysaccharide was metabolically labeled and isolated essentially as described previously (10). Further purification of the 14 C-labeled K5 polysaccharide included ␤-elimination with 0.5 M NaOH, 18 h, 4°C, and neutralization by HCl followed by treatment with 12 units Benzonase (Merck) for 2 h in 37°C and

Processive Formation of Heparan Sulfate N-sulfated Domain
purification on a Sep-Pak Plus C-18 cartridge (Waters) primed in methanol. Finally, the polysaccharide was applied to DEAE-Sephacel, equilibrated in 50 mM Tris-HCl, pH 8.0, washed in 50 mM NaAc, pH 4.0, and eluted in 0.5 M NaCl. The purified 14 Clabeled K5 polysaccharide, molecular weight ϳ100,000 and specific activity 35 ϫ 10 3 cpm/g, was desalted on PD-10 column (GE Healthcare) before it was used in the enzyme incubations.

Incubations of E. coli K5 Polysaccharide with NDSTs
Equal amounts of NDST1 and NDST2, based on His tag content as assayed by Western blotting using penta-His-antibody (Qiagen), were used in the incubations with 14 C-labeled K5 capsular polysaccharide in 50 mM MES, pH 6.3, 1% Triton X-100, 10 mM MnCl 2 , and 0.25 g/ml Polybrene, in the presence or absence of PAPS. Standard incubations of 100 l contained 14 C-labeled K5 substrate (20,000 cpm) and enzyme corresponding to an N-deacetylase activity of 700 cpm (see above), with or without PAPS (Sigma-Aldrich; 0.1 mM if not otherwise indicated) and were maintained at 37°C for different periods of time. Additions of enzyme and PAPS were renewed each hour. Incubations with PAP were performed as described above, using PAP (Sigma-Aldrich) instead of PAPS.

Analysis of Products
Enzymatically modified K5 polysaccharide was examined for occurrence and distribution of GlcNH 2 and N-sulfated GlcNS residues by selective deaminative cleavage. Samples were treated with nitrous acid at pH 3.9 or at pH 1.5 to achieve chain cleavage at GlcNH 2 and GlcNS residues, respectively, as described previously (11). Deamination products were separated by gel chromatography on Superdex 30 (GE Healthcare) in 0.5 M NH 4 HCO 3 at a flow rate of 0.7 ml/min. Fractions were analyzed by scintillation counting. The extent of N-deacetylation or N-sulfation was calculated from the relative proportions of the various even-numbered oligosaccharide deamination products.
NS domains were prepared as described (12). Briefly, samples of modified K5 polysaccharide were first chemically N-deacetylated by treatment with 70% (w/v) aqueous hydrazine (Fluka) containing 1% (w/v) hydrazine sulfate at 96°C for 4 h (13) and were then cleaved by treatment with nitrous acid at pH 3.9. Products were separated by gel chromatography as described above.

RESULTS
The E. coli K5 capsular polysaccharide has the same [4GlcUA␤1-4GlcNAc␣1-] n structure as the unmodified endogenous substrate that NDST enzymes will encounter in the Golgi compartment of the cell. We applied metabolically radiolabeled K5 polysaccharide as target for recombinant NDSTs and aimed at elucidating the mode of action of the enzyme under different conditions of PAPS access. Recombinant NDSTs with an enterokinase-sensitive C-terminal His tag were produced in HEK 293 cells. The proteins are retained in the Golgi compartment through their membrane-spanning domain, and the cells were therefore solubilized in detergent-containing buffer before purification by Talon metal affinity chromatography. To ascertain that enzyme activity was not affected by the added His tag, N-deacetylase activity was measured before and after cleavage of purified NDSTs with enterokinase. As shown in supplemental Fig. 1, removal of the His tag did not affect enzyme activity. For further experiments, the enzymes were used without removal of the tag.
Polymer Modification by NDST1 and NDST2 in the Absence or Presence of PAPS-14 C-Labeled K5 polysaccharide was incubated with NDST1 or NDST2 in the absence of PAPS, and the products were cleaved at the generated N-unsubstituted GlcNH 2 units by deamination at pH 3.9. Gel chromatography of both samples revealed a disperse population of oligosaccharides, with tetramers and hexamers as the predominant species (Fig. 2, A and B). Notably, very small amounts of disaccharides were observed after cleavage, indicating that the enzyme rarely attacked GlcNAc residues in neighboring disaccharide units. The distribution of variously sized oligosaccharides pointed to an essentially random mode of enzyme attack for both NDST1 and NDST2.
Inclusion of PAPS in the incubations enabled the enzymes to perform both N-deacetylation and N-sulfation and resulted in a dramatically altered modification pattern (Fig. 2, A and B). Products were essentially resistant to deamination at pH 3.9, pointing to the absence of N-unsubstituted GlcNH 2 residues (data not shown). By contrast, deamination at pH 1.5 generated disaccharides as the predominant cleavage product, indicating degradation of consecutive N-sulfated disaccharide units. Access to PAPS thus altered the target selection pattern of the NDST enzyme, from random N-deacetylation to concerted N-deacetylation/N-sulfation of adjacent disaccharide units, as required to generate NS domains. Moreover, a greater proportion of the substrate was subject to modification, amounting to ϳ65% N-sulfation in the presence of PAPS and either NDST1 or NDST2, as compared with ϳ40% N-deacetylation in the absence of the sulfate donor (calculated from the chromatograms in Fig. 2, A and B). PAPS apparently affected the two NDST enzymes in a similar fashion and to the same extent. For further characterization of the influence of PAPS on enzyme action, we chose to study NDST2.
The effect of PAPS on the course of NDST action appeared strongly associated with the sulfate donor function but could conceivably also reflect a regulatory influence of the nucleotide portion. To test this alternative, 0.1 mM PAP rather than PAPS was included in the incubations, and the distribution of N-unsubstituted residues was determined. Judging from the increased susceptibility of the incubated polysaccharide to deamination at pH 3.9, PAP appeared to stimulate N-deacetylation (Fig. 2D). However, the same oligosaccharides were seen as after incubation in the absence of PAP, without any significant amounts of disaccharides. Binding of the nucleotide per se thus does not seem to modulate NDST action in a qualitative sense.
Processive Formation of NS Domains-To determine the size distribution of the NS domains generated by NDST2 in the presence of PAPS, the remaining unmodified GlcNAc units in the 14 C-labeled K5 substrate were deacetylated through hydrazinolysis and cleaved by reaction with nitrous acid at pH 3.9. Gel chromatography of the products showed extended NS domains, composed of Ն8 disaccharide units (Fig. 3). Hardly any oligosaccharides of smaller size (Ͻ8-mer) were obtained, suggesting that sulfation occurred in long, continuous stretches with only very few GlcNAc residues remaining unmodified between the NS domains. Such domains could be formed either by processive action of the NDST along the polysaccharide chain or by random, but efficient, coupled N-deacetylation and N-sulfation, eventually yielding extended N-sulfated stretches. To distinguish between these alternatives, incubation of the labeled substrate with NDST2 and PAPS was conducted for different periods of time, and the products were cleaved at N-sulfated residues by deamination at pH 1.5. The procedure resulted in progressive formation of disaccharides with increasing incubation time but gave very small amounts of larger oligosaccharides (Fig. 4), pointing to an essentially processive mode of N-sulfation.
Effects of PAPS on NDST Target Pattern-Authentic HS species contain NS domains and unmodified NA domains but also contain substantial proportions of NA/NS domains composed of alternating N-acetylated and N-sulfated disaccharide units (14). In our experimental setting, extended NS domains domi-    nated the N-sulfation pattern of the substrate when PAPS was included in the incubations (Figs. 2-4). We next decided to test whether altering the concentration of PAPS would influence the domain structure. At the lowest concentration tested, 0.1 M, no N-sulfation of the substrate was detected (Fig. 5). When increased to 1 M, di-and tetrasaccharides as well as larger cleavage products were obtained after treatment of the substrate with nitrous acid at pH 1.5 (Fig. 5). At a PAPS concentration of 5 M, the relative abundance of consecutive N-sulfated disaccharide units in the substrate was higher, as shown by the increased amounts of disaccharides obtained after nitrous acid treatment (Fig. 5). Incubations in the presence of 5 and 10 M PAPS gave identical results (data not shown). These data suggest that the N-substitution pattern of HS may be regulated through variation in PAPS concentration during biosynthesis. Since the enzyme attacked the substrate in an apparently random mode in the absence of PAPS, we wanted to see how uncoupling of the N-deacetylation and N-sulfation reactions would influence N-substitution patterns. 14 C-Labeled K5 polysaccharide was therefore incubated with NDST2 for 1 h in the absence or presence of PAPS followed by incubation for an additional hour with 0.1 mM PAPS. The formation of N-unsubstituted and N-sulfated GlcN units was monitored as before after selective deaminative cleavage of the polysaccharide. As in previous experiments, incubation without PAPS resulted in the generation of N-unsubstituted GlcNH 2 residues dispersed along the chain and separated by one, two, or more N-sulfated disaccharide units (Fig. 6A). An additional hour in the presence of PAPS only marginally lowered the amounts of N-unsubstituted residues in the substrate (Fig. 6A). However, during this second hour, N-sulfation had occurred, as indicated primarily by the formation of di-and tetrasaccharides after nitrous acid cleavage at pH 1.5 ( Fig. 6B) but also by the appearance of extended saccharide sequences resistant to deamination at pH 3.9 (fraction emerging between 38 and 41 ml in Fig. 6A). Taken together, these findings suggest that preformed N-unsubstituted GlcNH 2 units are not preferred targets for N-sulfation as compared with GlcNAc residues that undergo concerted N-deacetylation/N-sulfation. In chains containing such GlcNH 2 units the concerted, presumably processive modification reactions will primarily involve extended, uninterrupted N-acetylated regions but may then entail N-sulfation also of adjacent GlcNH 2 units (see scheme in Fig. 6).

DISCUSSION
Current information regarding HS structure points to extensive variability regarding level of sulfation as well as domain organization (14 -16). Such variability applies to samples derived from different tissues, cell types, or developmental stages of a particular species, whereas variation between individuals is low (at least in inbred mice (15)). These and similar observations support the view that HS biosynthesis is subject to stringent control. So far, little is known about the mechanisms of regulation. The lack of a template, such as DNA in protein synthesis, suggests that regulation is exerted largely through the availability of various HS biosynthetic enzymes with different activities and substrate specificities. The concerted action of these membrane-bound enzymes would in turn be controlled by their organiza- C-radioactivity (cpm)  tion in the Golgi network, possibly along with other proteins that lack catalytic activity but nevertheless assist the overall process. The "GAGosome" concept was introduced to define a physical complex of enzymes (and potentially other proteins) committed to the assembly of HS chains; it was speculated that product structure would be determined by the composition and stoichiometry of the complex, dependent on the relative abundancies of the various proteins (17). This concept has some support in experimental data. Complete polymer modification of a single chain during cell-free biosynthesis of heparin in a mouse-mastocytoma microsomal fraction thus occurred within seconds (6), in accord with the notion of a tightly knit enzyme complex catalyzing several reactions in rapid succession. Moreover, microsomal polymer formation, now known to be catalyzed by the EXT1-EXT2 complex (18,19), was found to be greatly promoted by concomitant sulfation (20). Actual physical association between enzymes has been demonstrated, e.g. between EXT2 and NDST1 (21) and between GlcUA C5-epimerase and L-iduronic acid 2-O-sulfotransferase (22). Moreover, we recently showed that the level of expression of EXT1 and EXT2 affected the amount of active NDST1 in the cell, in turn influencing HS structure (21).
Our present study, using recombinant NDSTs, PAPS as sulfate donor, and a radiolabeled bacterial polysaccharide as acceptor, shows that one of the key features of HS biosynthesis, i.e. the formation of extended NS domains, may be ascribed to mechanisms inherent to the NDST enzyme without any involvement of other proteins. The process is (not unexpectedly) slower than in a natural setting, yet efficient, yielding NS domains similar in size to those of heparin and considerably longer than those commonly occurring in HS (Fig. 3) (23)(24)(25). Results of incubations maintained for different periods of time (Fig. 4) pointed to a processive mode of action; i.e. once engaged, the enzyme will hold on to, and work its way along the substrate, substituting N-acetyl with N-sulfate groups. The process is radically different from the random target selection displayed by the same enzyme in the absence of PAPS (Fig. 2, A-C) and also noted when polysaccharide generated by mastocytoma microsomal enzymes from UDP-GlcUA and UDP-GlcNAc in the absence of sulfate donor was studied (6). However, we cannot presently distinguish between a bona fide processive mechanism, where the enzyme never lets go of the substrate, and a mechanism where the enzyme, after being released from the substrate, preferably rebinds to GlcNAc residues adjacent to already N-sulfated sequence. Indeed, previous experiments involving a crude microsomal enzyme fraction showed that N-deacetylation of a polysaccharide substrate was promoted by introduction of N-sulfate groups (26).
The observation of extended NS domains being generated by unaided NDST action again raises the crucial question of regulation. What factor(s) determine the length and distribution of such domains in authentic HS? Molecular modeling suggested that the sulfotransferase domains of the four NDST isoforms differ in structure, and the enzymes also differ in kinetics of N-deacetylation and N-sulfation (27). In fact, NDST2 has been associated with formation of the highly N-(and O-)sulfated heparin in mast cells (28,29), which show high expression of NDST2 but little or no NDST1 transcript (30). On the other hand, embryonic day 18.5 liver from NDST1 Ϫ/Ϫ mice also expresses NDST2 as the only NDST isoform yet produces a low sulfated HS (31). Conversely, recombinant NDST1 also yielded extended NS domains in the presence of PAPS (data not shown). Therefore, it seems unlikely that regulation be primarily dictated by availability of isoforms. Notably, the polysaccharide formed upon incubation of rat liver microsomes with UDP-sugar precursors and PAPS was much more N-sulfated than the actual HS produced by the corresponding hepatocytes (32), suggesting that appropriate regulation of polymer modification depends on factors associated with intact cells, which were lost in subcellular fractionation. Interestingly, overexpression in liver of heparanase, an HS-degrading endoglucuronidase, resulted in greatly increased turnover of heparan sulfate proteoglycan(s) with dramatically up-regulated N-sulfation and extended NS domains (33).
Variation in intracellular PAPS concentration could potentially be important for regulation of HS biosynthesis. We presently found that lowering PAPS concentration in vitro led to a relative decrease in disaccharide deamination products, suggesting that formation of NS domains had subsided (Fig. 5). More dramatic effects were noted by experimentally segregating NDST action in the absence and presence of PAPS, on the same polysaccharide substrate (Fig. 6). Although NS domains were generated in the presence of PAPS, saccharide regions containing the scattered GlcNH 2 residues introduced during preceding incubation in the absence of the sulfate donor remained essentially unaffected.
Several reports including immunolocalization studies (22,34,35) indicate that HS biosynthesis in several cell types occurs proximal to the trans Golgi/trans Golgi network, where chondroitin sulfate is synthesized and PAPS transporters are believed to reside (36,37). Hence, the first encounter between NDST enzymes and the growing HS chain may occur in Golgi subcompartments, where PAPS is absent or present at low concentrations. Notably, studies of NDST mutant cells suggested that N-deacetylation and N-sulfation can occur independently; overexpression in 293 cells of NDST1 lacking N-sulfotransferase activity resulted in elevated HS N-sulfation with no concomitant increase in GlcNH 2 residues (38). The more random modification in the absence of PAPS followed by the apparently processive synthesis of NS domains in the presence of the sulfate donor could thus represent one means to generate the cell specific domain structure of HS.