Cell-free Synthesis of Anticoagulant Heparan Sulfate Reveals a ...

THE JOURNAL OF BIOLOGICAL CHEMISTRY © 1996 by The American Society for Biochemistry and Molecular Biology, Inc.

Vol. 271, No. 43, Issue of October 25, pp. 27063–27071, 1996 Printed in U.S.A.

Cell-free Synthesis of Anticoagulant Heparan Sulfate Reveals a Limiting Converting Activity That Modifies an Excess Precursor Pool* (Received for publication, March 29, 1996, and in revised form, July 3, 1996)

Nicholas W. Shworak‡§, Linda M. S. Fritze‡§, Jian Liu‡, Lynne D. Butler‡, and Robert D. Rosenberg‡§¶ From the ‡Department of Biology, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139 and the §Department of Medicine, Harvard Medical School, Beth Israel Hospital, Boston, Massachusetts 02215

LTA cells synthesize a minor population of heparan sulfate proteoglycans (HSPGact) bearing anticoagulant heparan sulfate (HSact) with a specific monosaccharide sequence that accelerates the action of antithrombin (AT). LTA cells also synthesize a major population of heparan sulfate proteoglycans endowed with nonanticoagulant heparan sulfate (HSinact) lacking the AT-binding site. To investigate the pathway-specific features of HSPGact generation, we established a novel detergentcontaining cell-free system with unlabeled and labeled microsomes from wild-type and variant LTA cells, respectively. The unlabeled microsomes provide “HSact conversion activity” that requires 3*-phosphoadenosine 5*-phosphosulfate to convert [35S]HSPGinact into [35S] HSPGact, presumably by sulfation. The reaction kinetics demonstrate that the rate of HSact synthesis is constant over the first 4 h of incubation. During this time, the rate of HSact production is linearly dependent on the amount of unlabeled LTA microsomal protein over a range of 10 to 50 mg as well as on the level of [35S]HS substrate over a range of 0.4 to 4.0 mg, microsomal protein. Compared with labeled microsomes, equivalent or slightly greater levels of HSact were generated from 35S-labeled HSPG, microsomal HS, or cell surface HS, which demonstrates that HSinact is the minimal substrate and that large amounts of HSact precursor exit the Golgi apparatus. Indeed, extensive modification of wild-type LTA cell surface [35S]HS elevated HSact content from 9 to 35%. The hypothesis that microsomal HSact conversion activity predicts the cellular rate of HSact generation was tested with wild-type or variant LTA cells in which production of HSact has been significantly altered by mutagenesis or overexpression of core protein or growth conditions. The data demonstrate that microsomal HSact conversion activity accurately reflects the cellular rate of HSact synthesis over a very wide range of conditions. The possibility that the reduced HSact generation is due to an inhibitor was excluded by mixing experiments. The possibility that reduced HSact generation is caused by decreased levels of HSact precursor was excluded as equivalent levels of HSact were formed from wild-type and variant [35S]HS. Based upon the above data, the LTA cell microsomal HSact conversion activity contains one or more limiting components that kinetically regulate the * This work is supported in part by National Institutes of Health Grants HL-33014 and HL-41484. 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. ¶ To whom correspondence and reprint requests should be addressed: Massachusetts Institute of Technology, Bldg. 68-480, 77 Massachusetts Ave., Cambridge, MA 02139. Fax: 617-258-6553.

rate of cellular HSact generation and the levels of HSact precursor in HS greatly exceed HSact production.

The elucidation of the structure-function relationships for anticoagulant heparin, a glycosaminoglycan (GAG)1 of mast cells, has served as a model for discerning the role of anticoagulant heparan sulfate proteoglycans (HSPGact) in the vascular system. The circulating plasma protease inhibitor, antithrombin (AT), neutralizes proteolytic enzymes of the coagulation cascade by slow formation of a binary complex. However, in the presence of heparin, AT binds to a unique oligosaccharide structure and undergoes a conformational change that dramatically accelerates the rate of complex formation (1). The delineation of the critical monosaccharide sequence of sulfated and nonsulfated glucosamine and uronic acid residues that enhances neutralization of hemostatic enzymes was made possible by affinity purification of heparin fragments with AT and detailed structure-function studies of the resulting oligosaccharides (2–10). Although only 15–30% of heparin molecules are capable of specific interaction with protease inhibitor (2), this abundance is greater than would be predicted by the completely random assembly of the constituents of the AT-binding domain (11), which indicates that generation of this oligosaccharide sequence requires coordination of biosynthetic enzymes. The pharmacologic action of heparin exploits a natural anticoagulant mechanism of the blood vessel wall, which is based on the highly specific interaction described above (12). Endothelial cells produce HSPGact, a minor subpopulation of heparan sulfate proteoglycans (HSPGs) that accelerate neutralization of coagulation enzymes through the heparan sulfate (HS) component that contains the appropriate oligosaccharide structure to bind AT (HSact) (13, 14). The remaining HSPGs are nonanticoagulant (HSPGinact) and contain HS chains that lack the AT-binding site (HSinact). The demonstration of an HSPG that contains regions of defined monosaccharide sequence that mediate a specific biologic function suggests that a similar situation may exist for other biologic systems. However, unlike the generation of other biopolymers, the ordered assembly of distinct disaccharide units does not involve a simple template, and so it remains unclear how HSPGact is formed. The capacity of the HSPG biosynthetic enzymes to generate 1 The abbreviations used are: GAG(s), glycosaminoglycan(s); AT, antithrombin; BSA, bovine serum albumin; CS, chondroitin sulfate; HS(PG), heparan sulfate (proteoglycan); HS(PG)act, anticoagulant heparan sulfate (proteoglycan); HS(PG)inact, nonanticoagulant heparan sulfate (proteoglycan); PAGE, polyacrylamide gel electrophoresis; PBS, phosphate-buffered saline; PG(s), proteoglycan(s); PAPS, 39-phosphoadenosine 59-phosphosulfate; MES, 4-morpholineethanesulfonic acid.

27063

27064

Mechanism of Anticoagulant Heparan Sulfate Biosynthesis

HSPGact may stem exclusively from inherent abundance and interactive properties or alternatively could require additional specific interactions with either unique or non-unique core proteins or regulatory factors. However, Kojima et al. (15, 16) isolated two major endothelial cell HSPGs, ryudocan (syndecan-4) and syndecan-1, and found that both contained HSact and HSinact, which demonstrates that the generation of HSact does not require a single unique core protein. These findings were subsequently extended to fibroglycan, glypican, and perlican isolated from endothelial cells (17); however, it remained conceivable that purified PGs might contain an undetected component with a minor variation in primary sequence that directs generation of HSact. To resolve this issue, Shworak et al. (18) expressed epitope tagged rat ryudocan in LTA cells and early passage endothelial cells, both of which produce HSPGact, isolated the exogenous origin PGs, and demonstrated that a single core protein with a unique primary sequence initiates synthesis of both HSact and HSinact. Furthermore, the expression of mutated forms of epitope tagged ryudocan containing only a single functional GAG acceptor site demonstrates that each of the three acceptor regions initiates synthesis of both forms of HS to a similar degree. We have further examined the functioning of the Golgi biosynthetic machinery by overexpressing epitope tagged ryudocan and determining the extent of synthesis of HSact and HSinact (18). The generation of HSPGinact increased linearly as a function of ryudocan expression, which indicates that synthesis of this GAG is limited by the availability of core protein. In striking contrast, the production of HSPGact decreased as a function of ryudocan expression, which suggests that synthesis of this GAG is limited by factors other than core protein. The subsequent data strongly supported the view that increased intracellular levels of ryudocan are responsible for decreased generation of HSPGact. In particular, the reduction in HSPGact production occurred with overexpression of the ryudocan ectodomain, which is directly secreted into the medium, or by a ryudocan form incapable of augmenting the synthesis of GAGs due to mutated attachment sites. However, the suppression in HSPGact production did not appear to involve the direct inhibition of general HS biosynthetic enzymes by core protein as overexpression of ryudocan failed to significantly reduce the size of HSact or HSinact or perturb the structure of HSact and HSinact. On the basis of the above results, we concluded that the increased intracellular levels of ryudocan most probably saturate the capacity of a component that is specific to the HSact biosynthetic pathway. The existence of a pathway-specific component is also supported by the investigation of the chemically mutated LTA cell line, VI-7 (19). Although LTA VI-7 produces normal levels of HSinact, synthesis of HSact is 5–7-fold lower than for wild-type LTA cells. This abnormality is not due to a defect in core protein production as complementation is not observed when mutant LTA cells are transfected with the ryudocan core, which can initiate HSact production. Structural analyses revealed that HSact and HSinact from LTA VI-7 are indistinguishable from wild-type HSact and HSinact, respectively, which indicates that the general HS biosynthetic enzymes were not altered. Together, the analyses of LTA VI-7 and the core protein overexpressing cell line suggest that the production of HSact involves a pathway-specific component that limits the capacity of the general HS biosynthetic enzymes to generate the defined monosaccharide sequence of HSact. If the pathwayspecific component limits a single biosynthetic step, then the precursor for this step should be present in excess. The existence of such a precursor should permit the development of a cell-free system that would allow the identification of both the

limiting component and the structure of the required precursor. In the present report, we characterize a novel cell-free system that employs microsomal HSact conversion activity to generate HSact from HSinact, and we demonstrate that this microsomal HSact conversion activity predicts cellular HSact production. Thus, we have established an experimental approach for elucidating the pathway-specific features of HSact biosynthesis. The analysis of variant LTA cell lines reveals that HSinact contains an excess pool of molecules that function as a precursor for HSact synthesis. Moreover, the level of HSact production is limited by the activity of one or more microsomal sulfotransferases that modify this precursor population. EXPERIMENTAL PROCEDURES

Cell Lines and Cell Culture—LTA refers to our previously described clone 1A (20) of an adenine phosphoribosyltransferase-deficient variant (21) derived from LMTK2, a thymidine kinase-deficient mouse L cell line (22). We have previously described the generation from the LTA line of a transfection-derived line that overexpresses the ryudocan12CA5 cDNA, clone 33 (18), and a chemically mutated clone, VI-7 (20). The distantly related line LA9 refers to clone 3B, which was isolated by two successive rounds of limiting dilution from A9 (23), a mouse L cell line deficient in both adenine phosphoribosyltransferase and hypoxanthine phosphoribosyltransferase. All cell lines were maintained in logarithmic growth by subculturing biweekly in Dulbecco’s modified Eagle’s medium (Life Technologies, Inc.) containing 10% fetal bovine serum at 37 °C under 5% CO2 humidified atmosphere as described previously (24). Exponentially growing cultures were generated by inoculating 54,000 cells/cm2 and incubating for 2 days while postconfluent cultures were produced by inoculating 50,000 cells/cm2 and growing for 10 days with medium exchanges on days 4, 7, 8, and 9. Isolation and Analysis of Cellular HS—The methods for analyzing cellular HS have been previously described in great detail (24). In brief, proteoglycans were metabolically labeled for 1 h by incubating cultures in modified BME medium (24) supplemented with 1% Nutridoma-SP (Boehringer Mannheim) and 2 mCi/ml of Na235SO4 (carrier free, ICN) adjusted to 5000 Ci/mol. Monolayers were extensively washed with PBS and then solubilized with 0.75 ml of lysis buffer (1% Triton X-100, 150 mM NaCl, 50 mM Tris, pH 7.4, 1 mM EDTA, 1 mM MgCl2, 1 mM iodoacetate, 100 mM phenylmethylsulfonyl fluoride, 0.02% NaN3), for 20 min on ice. The cell extracts were collected, and labeled proteoglycans were isolated by DEAE chromatography. GAGs were prepared by b-elimination and then purified by phenol extraction and ethanol precipitation. Chondroitin sulfate was removed by digestion with chondroitinase ABC, and HS was purified by phenol extraction and ethanol precipitation. To examine HS size, samples were resolved by 15% SDSPAGE (25), and migration profiles were recorded with a Betascope 603 Blot Analyzer (Betagen, Waltham, MA), and the Mr values were determined by calibration against CS size standards as described previously (24). Isolation of Microsomes—The microsomal isolation procedure was derived from a previously described protocol (26), and all steps were performed on ice or at 4 °C with prechilled solutions. For either growth state, 11 150-mm plates were placed on ice, monolayers were washed extensively with PBS, and then cells were removed by scraping and were pelleted by centrifugation at 260 3 g for 5 min. Pellets were resuspended in 13.75 ml of 0.25 M sucrose and lysed in a PotterElvehjem homogenizer with 10 strokes. For exponentially growing cells, phase contrast microscopic examination of this initial homogenate revealed that .80% of nuclei were free of subcellular structures. Samples were centrifuged at 10,000 3 g for 20 min to produce an initial supernatant, and from the pellet a second homogenate was generated and centrifuged, as described above. The first and second supernatants were then pooled. For postconfluent cells, microscopic examination of the initial homogenate revealed that .95% of nuclei displayed adherent Golgi apparatus. Samples were centrifuged at 10,000 3 g for 20 min, and the initial supernatant was discarded. The Golgi containing pellet was subjected to two additional cycles of homogenization and centrifugation, as described above, and the second and third supernatants were pooled. Regardless of growth state, nuclei were free of visible subcellular structures after the final homogenization. The pooled supernatants were then layered over 5-ml cushions of 0.6 M sucrose in ;13-ml tubes and centrifuged at 100,000 3 g for 1 h. Each pellet of microsomes was resuspended in 100 ml of 0.25 M sucrose; protein concentration was determined by the procedure of Bradford (27) using BSA as a standard,

Mechanism of Anticoagulant Heparan Sulfate Biosynthesis and samples were snap-frozen with liquid nitrogen and stored at 280 °C. For both exponential and postconfluent cultures, this fractionation recovered ;10% of initial protein and .90% of initial starting microsomal HSact conversion activity. 35 S-Labeled microsomes were generated from 16 100-mm plates of exponentially growing clone 33 unless stated otherwise. Two plates were washed twice with PBS and then incubated at 37 °C for 10 min in 2 ml/plate of sulfate-free modified BME medium (24) containing 31 mM (50 mCi/ml) of carrier-free Na235SO4 (ICN) and 1% Nutridoma-SP. Monolayers were then washed with cold PBS and stored on ice while the recovered medium was sequentially transferred to and incubated with each subsequent pair of plates. Upon completion of labeling, monolayers were extensively washed with PBS, and cells were removed by scraping. Microsomes were prepared as described above except that 8 ml of 0.25 M sucrose was used for each of the two homogenization steps, and microsomes were pelleted without the 0.6 M sucrose cushion. Samples contain ;200,000 cpm/mg of which 95% was ethanol-precipitable, whereas ;65% was incorporated into GAG as determined by DEAE chromatography. Preparation of GAGs—Proteoglycans were prepared from cell-free reactions by preparative DEAE chromatography, as described previously (24). To prepare GAGs, the resulting 1-ml eluate was subjected to b-elimination by the addition of 10 ml of 8.88 M NaOH containing 0.89 M NaBH4 and ;16 h incubation at 46 °C. Samples were quenched with 60 ml of 8.54 M ammonium formate containing 1.7 M HCl and then chilled on ice and extracted against 0.75 ml of phenol/chloroform/ isoamyl alcohol (25:24:1). The aqueous phase was dried down by vacuum centrifugation; a slurry was generated by resuspension in 60 ml of water, and salt was removed by centrifugation for 2 min at 10,000 3 g. The supernatant was added to 200 ml of water, and GAGs were precipitated by the addition of 1.25 ml of absolute ethanol followed by centrifugation at 15,000 3 g for 20 min at 4 °C. The pellet was resuspended in 200 ml of 0.0005% Triton X-100 followed by removal of residual ethanol by incomplete vacuum centrifugation. Purification of Microsomal Proteoglycans and HS—To prepare 35Slabeled proteoglycans, 25 3 106 cpm (total radioactivity) of 35S-labeled microsomes were solubilized in 250 ml of lysis buffer and 750 ml of PLB (24), and samples were subjected to preparative DEAE chromatography, as described previously (24), except that the elution buffer contained 2.5 mg/ml glycogen (Boehringer Mannheim) as carrier instead of dextran sulfate, which inhibits cell-free HSact synthesis. Half of the eluate was mixed with 1 ml of 4 M guanidine-HCl, 50 mM Tris, pH 7.4, 1 mM EDTA, and the mixture was applied to a Centricon-100 ultrafilter, and the buffer was then exchanged four times with 2 ml of water. The resulting purified proteoglycans were recovered by soaking the membrane overnight in 100 ml of 30 mg/ml BSA, 0.03% Triton X-100, and 2 mg/ml glycogen at 4 °C and were used as substrate for cell-free HSact synthesis reactions. 35 S-Labeled GAGs were isolated from proteoglycans by subjecting the remaining half of the DEAE eluate to b-elimination and ethanol precipitation, as described above. To remove CS the GAG sample was incubated in 150 ml containing 10 mg of BSA, 2 mg of glycogen, 30 mM NaCl, 25 mM ammonium acetate, pH 7.0, and 0.02 units of chondroitinase ABC (Sigma, number C-2905) for 4 h at 37 °C. The reaction was extracted against 200 ml of phenol/chloroform/isoamyl alcohol (25:24:1); the organic phase was back-extracted against 50 ml of water, and the pooled aqueous phases were combined with 10 ml of 2 M NaCl and 1 ml of ethanol. The precipitated HS was then collected by centrifugation as described under “Preparation of GAGs.” Preparation of Cell Surface HS—Cell surface HS was prepared from a 75-cm2 flask of an exponentially growing culture inoculated with 38,000 cells per cm2. Two days postinoculation the monolayer was rinsed twice with PBS and then incubated for 16 h in 4 ml of sulfate-free modified BME medium (24) containing 50 mCi/ml carrier-free Na235SO4 (ICN), 200 mM Na2SO4 and supplemented with 10% fetal bovine serum at 37 °C. Monolayers were extensively washed with PBS and then incubated in 2 ml of trypsin-EDTA (Life Technologies, Inc.) containing 1 mM Na2SO4 at room temperature for 30 min followed by 37 °C for 10 min. The cell suspension was collected; the flask was rinsed with 0.5 ml of PBS, and this solution was added to the trypsinate along with 87 ml of 25 mg/ml trypsin inhibitor (Sigma). Cells were removed by centrifugation at 560 3 g for 15 min, and 2.5 ml of the supernatant was mixed with 6.5 ml of PLB. Proteoglycans were isolated by DEAE chromatography; GAGs were prepared by b-elimination, and CS was removed by digestion with chondroitinase ABC as described under “Purification of Microsomal Proteoglycans and HS.” Cell-free Synthesis of HSact and Quantitation of HSact—50-ml reactions were assembled on ice and unless stated otherwise contained 40

27065

mg (protein) of unlabeled microsomes isolated from postconfluent cultures of LTA cells, 400,000 total cpm (;272,000 cpm in HS) of microsomes (;2 mg) prepared from exponential cultures of clone 33, 35 mM NaCl, 5 mM MgCl2, 10 mM MnCl2, 1% Triton X-100, 50 mM MES, pH 7.0, 1 mM PAPS (Sigma), 24% nuclease-treated rabbit reticulocyte lysate (Promega, L4980), 50 mM NAD (Sigma), 50 mM ATP, 4 mM disodium phosphocreatine (Sigma, P6915), and 2 units of type I creatine phosphokinase from rabbit muscle (Sigma, C 3755). Reactions were incubated at 37 °C for 4 h and quenched by the addition of 250 ml of lysis buffer with 750 ml of PLB; 35S-labeled GAGs were isolated by preparative DEAE chromatography and b-elimination as described above and then analyzed for HSact content. The percentage of HS chains containing AT-binding sites was calculated as the product of the GAG HS content and the GAG HSact content. The relative contents of HS and CS were measured from representative samples of each preparation of 35 S-labeled microsomes as ethanol-soluble 35S counts generated by digestion with Flavobacterium heparitinase and chondroitinase ABC, respectively, as described previously (24). Typically Flavobacterium heparitinase degraded 85% of GAG lyase-digestible material. GAG HSact content was determined by AT-affinity microchromatography, as described by Shworak et al. (18). Analysis of HSact Content of Ryudocan12CA5—35S-Labeled clone 33 microsomes (2 3 106 cpm) were mixed with 32 mg of protein G-purified 12CA5 monoclonal antibody (Harvard Monoclonal Antibody/Cell Culture Facility) in a 13-ml volume containing 75 mM NaCl with 1% Triton X-100 and incubated on ice for 1 h. This mixture was then subjected to a standard reaction for the synthesis of HSact. Subsequently, the reaction was mixed with 4 ml of 10% SDS, 50 ml of NET (50 mM Tris, pH 7.4, 150 mM NaCl, 0.25% gelatin, 0.1% Triton X-100, 1 mM EDTA, 0.02% NaN3) as well as 40 ml of NET equilibrated (1:1 slurry) protein ASepharose (Pharmacia Biotech, Inc.), and then incubated for 1 h with shaking. The beads, which bound complexes of antibodyzepitope tagged ryudocan, were collected by centrifugation for 10 s at 3,000 3 g and washed in 1.25 ml of NET containing 0.1% SDS and then 1.25 ml of NET adjusted to 0.5 M NaCl. Beads were washed twice with BDT (0.1 mg/ml BSA, 10 mM dextran sulfate, 3.3 mM Tris, pH 7.4, 0.33% Triton X-100, 0.02% NaN3). Tagged ryudocan was then eluted by three sequential 100-ml incubations in BDT containing 24 mM of the 12CA5 epitope peptide YPYDVPDYA (Berkeley Antibody Co. Inc.). GAGs were then removed by b-elimination, purified, and analyzed for HSact content as described above. Determination of Precursor Pool Size—Cell surface HS (9.6 3 106 cpm) from exponential LTA cultures was incubated in standard HSact synthesis reactions for ;16 h and then 150 ml of 267 mM NaCl, 13.3 mg/ml glycogen was added, and samples were extracted against 750 ml of phenol/chloroform/isoamyl alcohol (25:24:1). The organic phase was then back-extracted against 100 ml of 200 mM NaCl. The pooled aqueous phases were combined with 750 ml of absolute ethanol, and samples were centrifuged at 10,000 3 g for 40 min. The HS pellet was resuspended in 50 ml of 0.0005% Triton X-100; ;40,000 cpm was removed for AT-affinity microchromatography, and the remainder was vacuumconcentrated to ;25 ml and subjected to additional modification and purification cycles. After the fifth cycle, the entire sample was subjected to AT-affinity microchromatography, as described above except that dextran sulfate was replaced with 2 mg/reaction of glycogen. The fraction that was not bound (HSinact) was extracted against phenol, ethanolprecipitated, and then processed by an additional modification cycle. RESULTS

Cell-free Synthesis of Anticoagulant Heparan Sulfate (HSact)—We have established a detergent-containing cell-free system to investigate the biosynthesis of HSact and to develop assays for identifying critical components in this process. This system employs “HSact conversion activity” from unlabeled microsomes to convert 35S-labeled microsomal HSPGinact into HSPGact. PGs are then isolated by DEAE chromatography, subjected to b-elimination to generate free GAGs, and the levels of [35S]HSact determined by AT-affinity microchromatography, as described under “Experimental Procedures.” Microsomes were chosen as the origin for both substrate and enzyme as this should ensure the presence of all required synthetic components. The microsomes that supply HSact conversion activity were isolated from postconfluent cultures of the clonal L cell line, LTA, which generates high levels of HSact (12% of HS) (18). The microsomal source of substrate was isolated after a

27066


10-min pulse labeling with Na235SO4 of an exponentially growing LTA derivative that overexpresses the ryudocan12CA5 cDNA, clone 33, and consequently produces low levels of HSact (;0.7% of HS) (18). Thus the [35S]HSinact component, or a discrete subpopulation thereof, of clone 33 HSPGs is utilized as a precursor for synthesis of HSact (the HSact precursor). The maturity of the substrate was examined by comparing purified microsomal HS to cellular HS isolated from exponential cultures of clone 33 after 1 h of labeling with [35S]Na2SO4. SDSPAGE analysis revealed that the microsomal HS and cellular HS both exhibit a modal size of 53,000 Da; thus the labeled microsomal HS is mature with respect to chain elongation. In contrast, the level of HSact in microsomal HS was 60% reduced (HSact was 0.29 6 0.012% (n 5 4) of microsomal HS and 0.69 6 0.14% (n 5 8) of cellular HS). The lower proportion of HSact suggests that microsomal HS is somewhat immature with respect to one or more critical modification(s) required for generation of HSact. This extent of immaturity ensures the existence of structures that are precursors for the formation of the ATbinding site. We have previously demonstrated that overexpression of ryudocan core protein can suppress HSact production and accordingly have proposed that the biosynthesis of HSact is regulated by a limiting factor (18), which raises the possibility that elevated levels of HSPGs from clone 33 microsomes might inhibit cell-free synthesis of HSact. To circumvent this potential problem, standard reactions contained a relatively low concentration of 35S-labeled HSPG (;15 pM HS provided from ;2 mg, by protein, of microsomes2) combined with a large excess of unlabeled microsomes (40 mg of protein). We presumed that a stoichiometric excess of the appropriate biosynthetic factors would be achieved which should maximize the fraction of HSinact converted to HSact and optimize assay sensitivity. When a standard HSact synthesis reaction mixture was constructed without unlabeled microsomes, only 0.31% of 35Slabeled HS was HSact; however, the inclusion of unlabeled microsomes elevated the level of HSact by 14-fold. To exclude the remote possibility that newly formed HSact arose by the transfer of radioactivity from clone 33 35S-microsomes (possibly through labeled PAPS) to the abundant HSact of unlabeled microsomes, we isolated [35S]ryudocan12CA5 (which is only present in clone 33 microsomes) from HSact synthesis reactions by immunopurification with the 12CA5 monoclonal antibody, as described under “Experimental Procedures,” and examined the contained HS. The addition of unlabeled microsomes augmented the HSact content of ryudocan12CA5 from 0.31 to 4.4%, which shows that HSinact of labeled microsomes is directly converted into HSact. Omission of PAPS, Triton X-100, or reticulocyte lysate from the reaction mixture prevented the formation of HSact (results not shown). The absolute requirement for PAPS suggests that HSact is generated by enzymatic sulfation of HSinact. It appears that this process involves the interaction of internal contents from the unlabeled and labeled microsomes, as suggested by the need for detergent. The dependence of the reaction on the presence of reticulocyte lysate is quite surprising. We suspect that this constituent supplies a cytosolic component(s) that stabilize(s) the concentration of the labile reagent PAPS. This assertion is supported by the observation that cell-free synthesis of HSact proceeds over several hours, as described below. The remaining components of the 2 Microsomal HS should have a specific activity of 2.86 3 1020 cpm/ mol given that the HS chains contain 109 disaccharides, with ;0.8 mol sulfate/disaccharide (18) and assuming the specific activity of cellular PAPS equilibrated to the external sulfate pool level of 1495 Ci/mmol. Standard reactions contained ;221,000 cpm of 35S-labeled HS that equates to 0.77 fmol of HS per 50-ml reaction.

FIG. 1. Identification of a minimal substrate for HSact synthesis. The capacity of different 35S-labeled substrates, purified from clone 33, to generate HSact was examined in otherwise standard cell-free reactions, as described under “Experimental Procedures,” except that samples were incubated for only 2 h. Reactions contained 400,000 cpm of labeled microsomes or 200,000 cpm each of PGs purified from standard pulse labeled microsomes by DEAE chromatography and ultrafiltration, microsomal HS purified from microsomal PGs by b-elimination and chondroitinase digestion, and cell surface HS generated from [35S]Na2SO4-labeled cultures of clone 33, as described under “Experimental Procedures.” The data are presented as the increased percentage of HSact generated by the addition of unlabeled microsomes. Accordingly, for each substrate the background HSact percentage, obtained from reactions incubated without unlabeled microsomes, was subtracted from the value obtained by the incubation of substrate with unlabeled microsomes. The various purified samples did not contain assay-perturbing contaminants as demonstrated by control mixing reactions. Specifically, equivalent levels of HSact were generated from reactions containing 2.3 3 106 cpm of labeled microsomes with or without the addition of 200,000 cpm of purified substrate. Results from duplicate samples (mean 6 range) are presented.

reaction mixture are much less critical as omission of any single element does not abolish production of HSact (results not shown). The above data demonstrate that HSinact from the clone 33 35S-microsomes contains one or more AT-binding site precursors and that postconfluent LTA cell microsomes possess HSact conversion activity that completes formation of the ATbinding site via one or more enzymatic sulfation reactions. We characterized the substrate requirements of our cell-free assay by comparing the amount of HSact generated from microsomes, microsomal proteoglycans, microsomal HS, and cell surface HS isolated from labeled clone 33 cells (Fig. 1). Almost equivalent levels of HSact were generated from unfractionated microsomes and purified microsomal proteoglycans, whereas ;1.5-fold greater levels of HSact were produced from microsomal HS. These results indicate that GAG chain, rather than


FIG. 2. Time course and dose response of HSact synthesis. HSact synthesis reactions were performed under standard conditions as described under “Experimental Procedures” except for varying either the incubation time (A), the mass of unlabeled microsomes (B), or the mass of labeled microsomes (C). A and B, results are presented as the average of duplicate samples and ranges were less than 5% of the mean. Similar data were obtained in two additional experiments. C, results were generated with 0.4 – 4 mg of labeled microsomes from three independent preparations. HS levels were calculated as Flavobacterium heparitinase-digestible GAG cpm recovered from reactions by DEAE chromatography, as describe under “Experimental Procedures.”

proteoglycan, is the essential substrate for HSact conversion activity and that core protein may partially inhibit HSact conversion activity. Furthermore, substantial amounts of HSact precursor must exit the Golgi to appear intact in mature cell surface HSinact as microsomal HS and cell surface HS support an identical extent of HSact production. Time course analysis (Fig. 2A) revealed that the rate of HSact synthesis was constant for the first 4 h of incubation, and this duration was chosen as the standard period of incubation.3 The rate of HSact production was linearly related to the amount of unlabeled LTA microsomal protein over a range of 10 –50 mg (Fig. 2B). Therefore, standard reactions were performed with 40 mg of unlabeled microsomes. The production of HSact was directly proportional to the amount of input [35S]HS as revealed by varying the level of labeled microsomes over a 10-fold 3 After this initial time period, the reaction rate gradually declines to 0 by 24 h. At this point the labeled HS can be recovered and will support further production of HSact when used as substrate in a subsequent cell-free reaction. Thus, the termination of HSact production by 24 h reflects an exhaustion of the reaction milieu.

27067

range (0.4 – 4.0 mg) (Fig. 2C). Therefore, the reaction conditions provide for first order kinetics with respect to substrate, and the substrate concentration must be well below the Km of HSact conversion activity for HS. Indeed, we have determined that the Km for cell surface HS purified from clone 33 is well above 30 nM (28), which greatly exceeds the routine concentration of [35S]HS provided by ;2 mg of substrate microsomes (15 pM). The maintenance of substrate concentration below the Km maximizes the fraction of HSinact converted to HSact and enhances assay precision as the percentage of HSact is independent of variations in substrate concentration. Together, the above data demonstrate that we have established cell-free conditions that allow for quantitation of HSact conversion activity. HSact Conversion Activity Predicts Cellular Phenotype—We next evaluated whether the microsomal HSact conversion activity predicts the cellular rate of HSact biosynthesis. This hypothesis was tested by examining both parameters for exponential and postconfluent cultures of wild-type LTA or LTA variants in which the production of HSact is altered by chemical mutagenesis, VI-7, or by overexpression of ryudocan12CA5 core protein, clone 33. We determined the cell-free activity of microsomes prepared from each culture type and then calculated the relative cellular rates of HSact synthesis from previously generated data in which pulse labeling was used to measure the extent of HS synthesis and the percentage of HSact (18, 19) (Table I). The data demonstrate that microsomal HSact conversion activity reflects the cellular rate of HSact production. Moreover, we have also examined a distantly related L cell clone, LA9, that does not generate HSact and found that LA9 microsomes completely lack HSact conversion activity (results not shown). The striking correlation between these two parameters implies that microsomal HSact conversion activity quantitates the effective level of biosynthetic components that define cellular HSact production. Analysis of L Cell Variants—The specificity of the cell-free assay enabled us to examine the molecular mechanisms that control HSact production. In particular, we investigated with mixing experiments whether the phenotype of LTA cell variants is due to the presence of an inhibitor. Given that the greatest difference in cellular HSact generation was observed in exponential cultures of wild-type and variant LTA cells (Table I), only microsomes from this growth state were analyzed. We assayed 4 mg of unlabeled LTA microsomes in the presence or absence of a 9-fold excess of unlabeled variant LTA microsomes (Fig. 3). The data revealed that unlabeled wild-type LTA microsomes with or without unlabeled LTA variant microsomes augmented the synthetic HSact level of HS an additional 0.5%. Thus, wild-type LTA microsomal HSact conversion activity is not inhibited by components of variant LTA cell microsomes. Moreover, additional mixing experiments demonstrate that HSact conversion activity from wild-type LTA microsomes is not inhibited by LA9 microsomes (results not shown). Together the data suggest that the HSact conversion activity of variant L cell microsomes is decreased because of the decreased levels of a positive acting component rather than the increased levels of a directly acting diffusable inhibitor. We next carried out cell-free reactions with 35S-labeled microsomes prepared from exponential cultures of wild-type and variant LTA cells to test whether HSinact from the different cell types differed in the capacity to function as a substrate for HSact production. The inherent level of microsomal HSact, as reflected by reactions that were performed without unlabeled LTA microsomes, was representative of the percentage of HSact generated by each cell type (compare Fig. 4 to Table I). However, for each cell type the microsomal HS contained a slightly lower percentage of HSact than cellular HS (16% reduced for

27068


TABLE I Comparison of microsomal HSact conversion activity and cellular HSact synthetic rate for LTA cell variants Exponential (exp) and postconfluent (pc) LTA cells were previously examined for the rate of HS synthesis and the percentage of HSact by 1-h pulse labeling with [35S]Na2SO4 (18, 19). The derived data are compiled for clarity with the relative rate of HS synthesis being normalized to the value for exponential LTA cultures, which was set at 11.1. The relative rate of cellular HSact synthesis was then calculated as the rate of HS synthesis multiplied by the percentage of HSact. Values were obtained from at least duplicate determinations, and standard deviations were ,10% of the mean. To obtain the microsomal relative rate of HSact synthesis, standard cell-free reactions were performed with 40 mg of unlabeled microsomes from the appropriate cell and culture type, and values were then normalized to the value for exponential LTA cultures, which was set at 1.0. Data for each sample were determined with least two independent isolates of microsomes and are presented as the mean 6 s. Cellular Cell line

a b

Growth phase

Relative HS synthetic rate

LTA

exp pc

11.1 5.9

3 3

Clone 33

exp pc

24.1 14.7

3 3

VI-7

exp pc

10.6 5.5

3 3

Microsomal

HSact contenta

Relative HSact synthetic rate

Relative HSact synthetic rate (n)b

5 5

1.00 0.71

1.00 6 0.25 (5) 0.67 6 0.07 (6)

0.78 2.58

5 5

0.19 0.38

0.14 6 0.04 (4) 0.39 6 0.04 (3)

1.2 2.7

5 5

0.13 0.15

0.15 6 0.02 (4) 0.16 6 0.01 (3)

9.0 12.2

Percentage of total HS. Mean 6 s, n is number.

LTA and 60% reduced for clone 33 and VI-7), which suggests that the microsomal level of HSact conversion activity may be limiting. The supplementation of reactions with the same amounts of wild-type LTA cell microsomal HSact conversion activity elevated the HSact content of the substrate HS to ;4.5%, independent of the source of the 35S-labeled microsomes. Thus, microsomal HSinact from wild-type or variant LTA cells functions as an equivalent substrate for HSact generation, which indicates that the biosynthetic defect of both clone 33 and VI-7 does not alter formation of the precursor structure(s) recognized by HSact conversion activity. Interestingly, the percentage of HSact generated from clone 33 and VI-7 microsomal HS in the cell-free assay was 4 –5-fold higher than the percentage of cellular HSact produced by the respective cell types (compare Fig. 4 to Table I). This observation indicates that the microsomal HSinact of LTA variants contains a large population of molecules that are capable of forming HSact (the “HSact precursor pool”) and that the levels of HSact precursor exceed HSact production. In a similar fashion, we also examined 35S-labeled microsomes from exponential cultures of the distantly related L cell line, LA9. The addition of 40 mg of unlabeled LTA microsomes augmented the level of 35S-labeled HSact from ,0.04% (the detection limit of the AT-microaffinity assay) to ;1% (n 5 2). Thus, LA9 cells synthesize an HSact precursor population despite the absence of HSact conversion activity. We note that ;4-fold less HSact was generated from 35S-labeled LA9 microsomes as compared with 35S-labeled wild-type or variant LTA microsomes. This difference suggests that the LA9 microsomal HSact precursor pool, as compared with the wild-type or variant LTA microsomal HSact precursor pool, is either 4-fold smaller in size or 4-fold more slowly converted. Alternatively, the maturity of labeled microsomal substrate from LA9 cells and the various LTA lines may differ. This possibility appears more likely since cell surface HS from LA9 cells, wild-type LTA, and LTA variant clone 33 contain similarly large pools of HSact precursor, as described below. Thus, the extent to which the various HSact precursor pools support HSact generation fails to correlate with the degree of cellular HSact production of the respective L cell variants. These data suggest that the L cell variants do not differ to a significant extent in the generation of HSact precursor but rather in the production of HSact conversion activity. The Substrate of HSact Synthesis Is a Large Subpopulation of HSinact—The observation that cell surface HS of clone 33 contains HSact precursor (Fig. 1) suggests that cell surface HS of wild-type LTA cells also contains HSact precursor. To deter-

mine the size of this pool, cell surface HS from wild-type cells was incubated in HSact synthesis reactions for ;16 h, recovered from the exhausted reaction, and subjected to additional cycles of prolonged incubation with fresh reagents. After three rounds of incubation, the HSact level reached a plateau of 35% (Fig. 5). To exclude the possibility that end product inhibition prevented further formation of HSact, the HSinact fraction was isolated after the fifth cycle and was subjected to an additional round of prolonged incubation, which yielded only a slight increase in HSact (Fig. 5). We note that incubation with crude microsomal extract may structurally alter HSact precursor; thus, the data provide only a minimal estimate of the size of the HSact precursor pool (26% of HS). While it is conceivable that the entire HSinact pool might function as HSact precursor, the testing of this hypothesis requires biochemically pure HSact conversion activity. The occurrence of a large HSact precursor pool is not limited to wild-type LTA cells since the HSact precursor pool is also a major constituent (33– 40%) of cell surface HS from clone 33 and the LA9 lines (results not shown). Cell surface HS from these cell lines contains ,1% HSact, compared with wild-type levels of 9%, which suggests that the higher precursor content of variant LTA cells results from underutilization of this subpopulation. Thus, for each examined cell line, the maximum level of HSact which could be generated from HS is very similar (i.e. HSact 1 HSact precursor 5 34 – 40% of HS). Furthermore, the high levels of HSact precursor in cell surface HSinact indicate that the Golgi generates levels of this species that greatly exceed HSact production and that the excess HSact precursor exits the Golgi without further modification. DISCUSSION

Our current investigation using a cell-free assay represents the first demonstration of the molecular mechanism that regulates HSPGact generation; the production of HSPGact is controlled by the level of a critical enzymatic activity acting upon an excess precursor pool. Our overall approach to this problem was based on the pioneering investigations of Silbert and Rice et al. (29, 30) who initially demonstrated that microsomal preparations from mast cells catalyze the transfer of [35S]SO4 from [35S]PAPS into microsomal heparin. We developed a cell-free assay for HSact conversion activity that requires the presence of PAPS to convert a subpopulation of labeled HSPGinact into HSPGact. In an attempt to closely mimic Golgi synthetic conditions, we initially chose detergent-solubilized microsomes as a source of labeled substrate; however, we subsequently observed that purified substrates (microsomal HSPG, microsomal HS, cell surface HS) all reacted slightly more efficiently. HS


FIG. 3. Microsomes of LTA cell variants lack a soluble inhibitor. Standard HSact synthesis reactions were performed as described under “Experimental Procedures.” Reactions contained either a unique source of microsomes (4 mg for LTA or 36 mg for LTA cell variants) or contained LTA microsomes (4 mg) mixed with microsomes from LTA cell variants (36 mg). Results from triplicate determinations (mean 6 s) were corrected for the HSact background of [35S]HS from reactions incubated without unlabeled microsomes and so represent the percentage of HSact generated by the addition of microsomes.

supported a marginally higher rate of HSact synthesis than purified HSPG, which indicates that the biochemical activity directly recognizes HS without a concomitant requirement for core protein. Due to the structural heterogeneity of the substrate, it is unclear whether the detected activity results from the action of a single enzyme or multiple enzymes. Despite this uncertainty, the comparison of microsomal HSact conversion activity and cellular HSact production of wild-type and L cell variants revealed a complete correspondence between these two parameters. Thus, the cell-free assay measures the enzymatic activity(ies) of the critical component(s) that ultimately defines cellular HSact generation. Based upon the present understanding of heparan biosynthesis, we presume that HSact conversion activity modifies the HSact precursor within the trans-Golgi; however, conclusive intracellular localization will require detailed immunelectron microscopic analysis. A substantial portion of mature cell surface HS (26 – 40%) is capable of functioning as a substrate for HSact synthesis (HSact precursor). The inability to convert all of HSinact into HSact under exhaustive cell-free modification conditions suggests that the HSact precursor is a structurally distinct subpopulation of HSinact. It is conceivable that formation of the HSact precursor is an inherent characteristic of the global HS biosynthetic machinery given that HS from LA9 cells contains HSact precursor despite the inability to generate HSact. Moreover we

27069

FIG. 4. Wild-type and LTA variants produce a functionally equivalent 35S-labeled substrate. 35S-Labeled microsomes were isolated from exponential cultures of clone 33, VI-7, and LTA, and 400,000 cpm was used as substrate in standard HSact synthesis reactions containing or lacking 40 mg of unlabeled microsomes from postconfluent LTA cultures, as described under “Experimental Procedures.” Results from quadruplicate samples (mean 6 s) are presented.

have recently observed that the identical situation exists for HS from Chinese hamster ovary cells.4 This surmise is also supported by the equivalency of labeled microsomal HSinact from the LTA variants to function as HSact precursor. We note that since HSact precursor comprises only a fraction of HSinact, it is plausible that the size of the precursor population may be controlled by a regulatory process. Given that HS isolated from Reichert’s membrane exhibits an extremely high proportion of HSact (.80% by mass) (31), it is even possible that for certain cell types all HSinact molecules could function as HSact precursor. The abundance of HSact precursor in cell surface HS demonstrates that within the Golgi only a portion of HSact precursor is converted to HSact. The incomplete processing of HSact precursor indicates that the level of HSact production is either governed by limiting the extent of HSact conversion activity within the Golgi apparatus or by limiting the amount of HSPGs entering a Golgi compartment containing excess HSact conversion activity. The former mechanism appears to be correct since detergent-solubilized microsomal HSact conversion activity mimics the extent of cellular HSact generation. A correlation between these two parameters would not be expected in the latter scenario as detergent solublization would destroy compartmentalization. Thus, our data support a kinetic model in which HS-modifying enzymes generate HSact precursor, which constitutes a subpopulation of HSinact. The HSact precursor contains a discrete oligosaccharide structure(s), the AT-binding 4

L. Zhang, unpublished data.

27070


FIG. 5. Determination of HSact precursor pool size. Cell surface [35S]HS from LTA cells was subjected to exhaustive HSact synthesis to reveal the size of the HSact precursor pool. GAGs were incubated in otherwise standard reactions for ;16 h, then were isolated by phenol extraction and ethanol precipitation, and subjected to four additional cycles of modification and isolation. After the fifth cycle the remaining HSinact fraction was isolated and then subjected to an additional modification cycle. The cumulative level of HSact, as determined by ATaffinity microchromatography, at the completion of each isolation is plotted as a function of the number of rounds of modification. Data are presented as the average of duplicate samples, and ranges were less than 5% of the mean; similar results were obtained in an additional experiment.

site precursor, which requires a limited number of modifications to generate the AT-binding site. HSact-producing cells possess a limiting activity, HSact conversion activity, which completes the modification of this precursor by at least a sulfation reaction to generate a functional AT-binding site. Thus, the cellular rate of HSact synthesis is defined by the functional level of the limiting activity. We note that special situations may exist, such as when HSact levels are very high, where production could potentially be regulated by controlling the HSact precursor population size. It has been proposed that 3-O-sulfation terminates the biosynthesis of anticoagulant heparin; however, it appears unlikely that a limitation in 3-O-sulfotransferase activity regulates the production of this GAG. Linhardt et al. (32) performed an extensive structural analysis of heparin oligosaccharides and concluded that an anticoagulant heparin precursor devoid of the critical 3-O-sulfate group does not occur in excess. Furthermore, Kusche et al. (33) found that anticoagulant heparin was not generated by the incubation of nonanticoagulant heparin with mastocytoma microsomal extract and PAPS. However, Razi and Lindahl (34) were able to produce anticoagulant heparin by a similar incubation of an octasaccharide fraction, isolated from nonanticoagulant heparin, which is enriched in the AT-binding site precursor. The authors proposed that in-

hibitory sequences within the heparin chain block access of the precursor site to the 3-O-sulfotransferase and suggest that the regulation of anticoagulant heparin production depends on the topological organization of the membrane-bound enzyme machinery. Given that the actual conditions of heparin synthesis in the Golgi apparatus remain unknown, it is possible that such competitive inhibition may limit AT-binding site formation. However, it is unlikely that this mechanism affects HSact biosynthesis, since the presence of a diffusable inhibitor is excluded by the negative mixing experiments, described above. In striking contrast to the above models of heparin biosynthesis, we propose that the HSact precursor contains an oligosaccharide structure corresponding to the AT-binding site devoid of the glucosamine 3-O-sulfate group and that this excess precursor is converted into HSact by limiting 3-O-sulfotransferase activity. Our kinetic model is supported by the observation that the disaccharide composition of HSact only differs from that of HSinact by the greater abundance of 3-O-sulfated glucosamine residues in the anticoagulant form (13, 16). This enrichment in 3-O-sulfated glucosamine residues, taken in conjunction with the absolute functional requirement for the presence of these species within the AT-binding site (4, 5, 7, 8), suggests that HSact conversion activity may possess 3-O-sulfotransferase activity. Indeed, we have calculated that the high levels of HSact generated by the exhaustive cell-free conversion of HSinact grossly exceed the preexisting abundance of 3-Osulfated glucosamine residues in the substrate.5 Moreover, HS from Chinese hamster ovary cells completely lacks 3-O-sulfated glucosamine residues but can be used to generate high levels of HSact (.10%).4 In the accompanying paper (28), we describe the purification of a component of HSact conversion activity and prove that it possesses 3-O-sulfotransferase activity but lacks other sulfotransferase functions. Furthermore, we demonstrate that the majority of HSact conversion activity exhibited by microsomes in the cell-free assay can be accounted for by purified 3-O-sulfotransferase. Thus, the predominant limiting component of HSact conversion activity is a 3-O-sulfotransferase. It is, therefore, likely that the three L cell variants exhibit reduced HSact production because of decreased activity of 3-O-sulfotransferase. This assertion is supported by cell-free complementation analyses that indicate that all three variants lie within the same complementation group.6 It is presently unclear whether the above phenotype directly affects a 3-Osulfotransferase required to generate HSact or indirectly modulates the activity of the enzyme by interfering with a critical regulatory component. We note that L cell variant data summarized in the accompanying communication (28) strongly indicate that 3-O-sulfation of HSPGact, as compared with HSPGinact, is independently controlled. The ultimate determination of the manner by which this process occurs awaits the molecular cloning and expression of a 3-O-sulfotransferase cDNA. Acknowledgments—We thank Dr. Arthur D. Lander for critiquing the manuscript and are grateful to members of the RDR laboratory for their insightful comments.

5 We have previously demonstrated that for clone 33, 3-O-sulfated residues comprise 9% (HSact) and 0.5% (HSinact) of sulfated disaccharides and that HSact contains four 3-O-sulfates per chain with three to four AT-binding sites per chain (18). Using our previously described AT protection assay (18), we have found that synthetic HSact also has three to four AT-binding sites per chain (J. Liu, unpublished data). Accordingly, if HSact conversion activity does not include 3-O-sulfotransferase activity then HSinact chains would contain about four 3-O-sulfates per molecule. Thus preexisting 3-O-sulfated residues would only allow for the conversion of 5.5% of HSinact into HSact. 6 The mixing in all possible permutations of microsomes from the variants fails to correct the respective defects in HSact conversion activity (N. W. Shworak, unpublished data).

Mechanism of Anticoagulant Heparan Sulfate Biosynthesis REFERENCES 1. Rosenberg, R. D., and Damus, P. S. (1973) J. Biol. Chem. 248, 6490 – 6505 2. Lam, L. H., Silbert, J. E., and Rosenberg, R. D. (1976) Biochem. Biophys. Res. Commun. 69, 570 –577 3. Rosenberg, R. D., and Lam, L. (1979) Proc. Natl. Acad. Sci. U. S. A. 76, 1218 –1222 4. Rosenberg, R. D., Armand, G., and Lam, L. (1978) Proc. Natl. Acad. Sci. U. S. A. 75, 3065–3069 5. Lindahl, U., Ba¨ckstro¨m, G., and Thunberg, L. (1983) J. Biol. Chem. 258, 9826 –9830 6. Lindahl, U., Ba¨ckstro¨m, G., Thunberg, L., and Leder, I. G. (1980) Proc. Natl. Acad. Sci. U. S. A. 77, 6551– 6555 7. Choay, J., Petitou, M., Lormeau, J. C., Sinay¨, P., Casu, B., and Gatti, G. (1983) Biochem. Biophys. Res. Commun. 116, 492– 499 8. Atha, D. H., Stevens, A. W., Rimon, A., and Rosenberg, R. D. (1984) Biochemistry 23, 5801–5812 9. Atha, D. H., Lormeau, J. C., Petitou, M., Rosenberg, R. D., and Choay, J. (1985) Biochemistry 24, 6723– 6729 10. Atha, D. H., Lormeau, J. C., Petitou, M., Rosenberg, R. D., and Choay, J. (1987) Biochemistry 26, 6454 – 6461 11. Casu, B. (1985) Adv. Carbohydr. Chem. Biochem. 43, 51–134 12. Marcum, J. A., and Rosenberg, R. D. (1989) in Heparin, Chemical and Biological Properties: Clinical Applications (Lane, D. A., and Lindahl, U., eds) pp. 275–294, Edward Arnold, London 13. Marcum, J. A., Atha, D. H., Fritze, L. M. S., Nawroth, P., Stern, D., and Rosenberg, R. D. (1986) J. Biol. Chem. 261, 7507–7517 14. Marcum, J. A., McKenney, J. B., and Rosenberg, R. D. (1984) J. Clin. Invest. 74, 341–350 15. Kojima, T., Shworak, N. W., and Rosenberg, R. D. (1992) J. Biol. Chem. 267, 4870 – 4877 16. Kojima, T., Leone, C. W., Marchildon, G. A., Marcum, J. A., and Rosenberg, R.

27071

D. (1992) J. Biol. Chem. 267, 4859 – 4869 17. Mertens, G., Cassiman, J.-J., Van den Berghe, H., Vermylen, J., and David, G. (1992) J. Biol. Chem. 267, 20435–20443 18. Shworak, N. W., Shirakawa, M., Colliec-Jouault, S., Liu, J., Mulligan, R. C., Birinyi, L. K., and Rosenberg, R. D. (1994) J. Biol. Chem. 269, 24941–24952 19. Colliec-Jouault, S., Shworak, N. W., Liu, J., de Agostini, A. I., and Rosenberg, R. D. (1994) J. Biol. Chem. 269, 24953–24958 20. de Agostini, A. L., Lau, H. K., Leone, C., Youssoufian, H., and Rosenberg, R. D. (1990) Proc. Natl. Acad. Sci. U. S. A. 87, 9784 –9788 21. Debenham, P. G., Kartner, N., Siminovitch, L., Riordan, J. R., and Ling, V. (1982) Mol. Cell. Biol. 2, 881– 889 22. Kit, S., Dubbs, D. R., Piekarski, L. J., and Hsu, T. U. (1963) Exp. Cell Res. 31, 297–312 23. Littlefield, J. W. (1963) Proc. Natl. Acad. Sci. U. S. A. 50, 568 –576 24. Shworak, N. W., Shirakawa, M., Mulligan, R. C., and Rosenberg, R. D. (1994) J. Biol. Chem. 269, 21204 –21214 25. Laemmli, U. K. (1970) Nature 227, 680 – 685 26. Fleischer, B., and Fleischer, S. (1970) Biochim. Biophys. Acta 219, 301–319 27. Bradford, M. M. (1976) Anal. Biochem. 72, 248 –254 28. Liu, J., Shworak, N. W., Fritze, L. M. S., Edelberg, J. M., and Rosenberg, R. D. (1996) J. Biol. Chem. 271, 27072–27082 29. Rice, L. I., Spolter, L., Tokes, Z., Eisenman, R., and Marx, W. (1967) Arch. Biochem. Biophys. 118, 374 –382 30. Silbert, J. E. (1967) J. Biol. Chem. 242, 2301–2305 31. Pejler, G., Ba¨ckstro¨m, G., Lindahl, U., Paulsson, M., Dziadek, M., Fujiwara, S., and Timpl, R. (1987) J. Biol. Chem. 262, 5036 –5043 32. Linhardt, R. J., Wang, H., Loganathan, D., and Bae, J. (1992) J. Biol. Chem. 267, 2380 –2387 33. Kusche, M., Torri, G., Casu, B., and Lindahl, U. (1990) J. Biol. Chem. 265, 7292–7300 34. Razi, N., and Lindahl, U. (1995) J. Biol. Chem. 270, 11267–11275