Molecular Cloning and Expression of Mouse and Human cDNAs Encoding Heparan Sulfate d-Glucosaminyl 3-O-Sulfotransferase*

The cellular rate of anticoagulant heparan sulfate proteoglycan (HSPGact) generation is determined by the level of a kinetically limiting microsomal activity, HSact conversion activity, which is predominantly composed of the long sought heparan sulfate d-glucosaminyl 3-O-sulfotransferase (3-OST) (Shworak, N. W., Fritze, L. M. S., Liu, J., Butler, L. D., and Rosenberg, R. D. (1996) J. Biol. Chem. 271, 27063–27071; Liu, J., Shworak, N. W., Fritze, L. M. S., Edelberg, J. M., and Rosenberg, R. D. (1996) J. Biol. Chem.271, 27072–27082). Mouse 3-OST cDNAs were isolated by proteolyzing the purified enzyme with Lys-C, sequencing the resultant peptides as well as the existing amino terminus, employing degenerate polymerase chain reaction primers corresponding to the sequences of the peptides as well as the amino terminus to amplify a fragment from LTA cDNA, and utilizing the resultant probe to obtain full-length enzyme cDNAs from a λ Zap Express LTA cDNA library. Human 3-OST cDNAs were isolated by searching the expressed sequence tag data bank with the mouse sequence, identifying a partial-length human cDNA and utilizing the clone as a probe to isolate a full-length enzyme cDNA from a λ TriplEx human brain cDNA library. The expression of wild-type mouse 3-OST as well as protein A-tagged mouse enzyme by transient transfection of COS-7 cells and the expression of both wild-type mouse and human 3-OST by in vitrotranscription/translation demonstrate that the two cDNAs directly encode both HSact conversion and 3-OST activities. The mouse 3-OST cDNAs exhibit three different size classes because of a 5′-untranslated region of variable length, which results from the insertion of 0–1629 base pairs (bp) between residues 216 and 217; however, all cDNAs contain the same open reading frame of 933 bp. The length of the 3′-untranslated region ranges from 301 to 430 bp. The nucleic acid sequence of mouse and human 3-OST cDNAs are ∼85% similar, encoding novel 311- and 307-amino acid proteins of 35,876 and 35,750 daltons, respectively, that are 93% similar. The encoded enzymes are predicted to be intraluminal Golgi residents, presumably interacting via their C-terminal regions with an integral membrane protein. Both 3-OST species exhibit five potentialN-glycosylation sites, which account for the apparent discrepancy between the molecular masses of the encoded enzyme (∼34 kDa) and the previously purified enzyme (∼46 kDa). The two 3-OST species also exhibit ∼50% similarity with all previously identified forms of the heparan biosynthetic enzymeN-deacetylase/N-sulfotransferase, which suggests that heparan biosynthetic enzymes share a common sulfotransferase domain.

The serine proteases of the intrinsic blood coagulation cascade are slowly neutralized by antithrombin (AT) 1 (reviewed in Ref. 1). This inhibition is secondary to the generation of 1:1 enzyme⅐AT complexes whose formation is dramatically enhanced by the mast cell product, heparin (2). Damus et al. (3) hypothesized that endothelial cell surface heparan sulfate proteoglycans (HSPGs) function in a similar fashion to accelerate coagulation enzyme inactivation by AT and therefore are responsible for the nonthrombogenic properties of blood vessels. We initially demonstrated that perfusion of the hind limbs of normal rodents and rodents deficient in mast cells with purified thrombin and AT leads to a greatly elevated rate of thrombin⅐AT complex formation and that the enzyme heparitinase as well as the natural heparin antagonist platelet factor 4 suppress the above acceleration (4,5). We subsequently showed that cultured cloned bovine macrovascular and rodent microvascular endothelial cells synthesize both anticoagulant HSPG (HSPG act ) and nonanticoagulant HSPG (HSPG inact ) (6 -8). HSPG act bear glycosaminoglycan (GAG) chains that bind tightly to AT and accelerate thrombin⅐AT complex generation (6 -8).
The biosynthesis of HSPG act requires generation of a core protein; assembly of a linkage region of four neutral sugars on specific serine attachment sites of the core protein; elongation of a GAG backbone composed of alternating N-acetylglu-cosamine and glucuronic acid residues; and modification of this homogenous copolymer by partial N-deacetylation with coupled N-sulfation of glucosamine residues, partial epimerization of glucuronic acid to iduronic acid residues, partial 2-O-sulfation of uronic acid residues, and partial 6-O-sulfation and partial 3-O-sulfation of glucosamine residues (reviewed in Ref. 9). This multienzyme pathway generates HSPG act with regions of defined structure that contain the primary AT binding domain sequence found in anticoagulant heparin: uronic acid 3 glucosamine (N-acetyl/N-sulfate) 6-O-sulfate 3 glucuronic acid 3 glucosamine N-sulfate 3-O-sulfate (6-O-sulfate) 3 iduronic acid 2-O-sulfate 3 glucosamine N-sulfate 6-O-sulfate (10 -17). These reactions also produce HSPG inact with regions of varying monosaccharide sequence that lack the primary AT-binding domain. The structure-function relationships of the AT binding domain have been elucidated with heparin/heparan sulfate oligosaccharides in association with fast reaction kinetics and equilibrium binding assays. The 6-O-sulfate group on residue 2 and the 3-O-sulfate group on residue 4 function in a thermodynamically linked fashion to supply half of the binding energy for interaction with AT and trigger a conformational event that accelerates neutralization of specific coagulation proteases (11,12). The amino and ester sulfate groups at residues 5 and 6 as well as carboxyl groups at other sites provide the other half of the binding energy for interaction with protease inhibitor (10,11). Furthermore, monosaccharide sequences outside the primary AT binding domain are essential in facilitating inhibition of coagulation proteases other than factor Xa (18,19).
During the past 8 years, several biosynthetic enzymes that generate HSPG act and HSPG inact have been purified. These proteins include the N-acetylglucosamine/glucuronic acid copolymerase (20), the N-deacetylase/N-sulfotransferase (NST) (21,22), the glucuronic acid/iduronic acid epimerase (23), the iduronic acid/glucuronic acid 2-O-sulfotransferase (24), the glucosamine 6-O-sulfotransferase (25), and the glucosamine 3-Osulfotransferase (3-OST) (26). However, the only enzymes that have also been molecularly cloned are two structurally and functionally distinct isoforms of the N-deacetylase/N-sulfotransferase (NST-1 from liver and NST-2 from mastocytoma) (27)(28)(29)(30)(31). The heparan biosynthetic enzymes must function in a coordinated manner to produce the AT binding domain, because the abundance of this sequence is much greater than predicted from a random assembly of constituents (32). The postulated regulatory mechanism must direct the biosynthetic enzymes to carry out the appropriate sequence of epimerization/sulfation reactions to generate the AT binding domain (33,34).
We have previously described a soluble cell-free system to investigate HS act generation and developed assays for defining critical enzymatic components (35). The investigations employing the above approach define a limiting HS act conversion activity that acts upon an excess precursor population to regulate cellular HSPG act biosynthesis (35). The major component of the limiting HS act conversion activity proved to be the long sought 3-OST, as documented by purification and characterization of this protein (26). The investigations utilizing the above technique also showed that HS act precursor (35% of the HS inact pool) is 3-O-sulfated to generate HSPG act and that HS inact precursor (65%) is 3-O-sulfated to produce HSPG inact . However, only a small fraction of either substrate is modified, with the remaining precursor population exiting the Golgi apparatus to appear on the cell surface. Thus, 3-OST constitutes a rate-limiting enzymatic activity that defines the level of cellular HSPG act generation. In contrast, the level of 3-OST activity does not appear to limit the mast cell formation of anticoagulant heparin, which contains a high proportion of molecules with the AT binding site (ϳ30%) (36); structural and biochemical analyses indicate that the precursor of anticoagulant heparin is present in minimal amounts (37,38).
Despite this biosynthetic difference, for both heparin and heparan, 3-O-sulfation does not guarantee the formation of anticoagulant GAG (33,34,39,40). This phenomenon may be secondary to alterations in the relative concentrations of two functional forms of 3-OST that differentially act upon HS act precursor versus HS inact precursor or changes in the relative levels of HS act precursor versus the HS inact precursor. The two functional forms of 3-OST may be due to two discrete gene products, posttranslational modification of a single gene product, or the presence of a regulatory factor that directs the enzyme to modify one or the other precursor. The relative levels of HS act precursor versus HS inact precursor are presumably controlled by earlier biosynthetic enzymes. In the current paper, we molecularly clone as well as express murine and human 3-OST, and we show that the expressed enzyme is able to 3-O-sulfate both HS act precursor and HS inact precursor. Furthermore, the deduced structure of 3-OST, when compared with NST, defines a heparan sulfate sulfotransferase domain and also suggests a novel mechanism for limiting the action of the enzyme.

Cell Lines and Cell Culture
We have previously described the clonal L cell line LTA (35,41), the generation of clone 33, an LTA transfectant that overexpresses the ryudocan 12CA5 cDNA (33), a rapidly growing revertant of clone 33, L-33 ϩ (26), and RFPEC, an immortalized line derived from rat fat pad endothelial cells (8). Primary mouse neonatal endothelial cells from the cardiac microvasculature of day 3-5 neonates (CME cells) were a generous gift from Dr. Jay Edelberg (MIT/Beth Israel Hospital), whereas COS-7 cells were obtained from the ATCC. Primary human umbilical vein endothelial cells (HUVEC) were maintained according to the supplier's (Clonetics Corp., San Diego, CA) protocol. Unless otherwise stated, 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, 100 g/ml streptomycin, and 100 units/ml penicillin at 37°C under 5% CO 2 humidified atmosphere, as described previously (42). Exponentially growing cultures were generated by inoculating 54,000 cells/cm 2 and incubating for 2 days, whereas postconfluent cultures were produced by inoculating 250,000 cells/cm 2 and allowing growth for 10 days with medium exchanges on days 4, 7, 8, and 9.

Peptide Purification and Sequencing
The purification of mouse 3-OST from L-33 ϩ has been previously described (26), and the final step 4 product was concentrated by reverse phase chromatography on an HP 1090 M system (Hewlett Packard) equipped with a C4 reverse phase HPLC column (250 ϫ 2.1-mm, 300-Å pore size, 5-m particle size) (Vydac, catalog number 214TP52) equilibrated in 1.6% acetonitrile (v/v), 0.1% trifluoroacetic acid (v/v). After application of the sample, the reverse phase matrix was washed with 60% acetonitrile, 0.1% trifluoroacetic acid, and bound species were eluted with 78.4% acetonitrile, 0.1% trifluoroacetic acid. Samples of 1.5 or 3 g, from two independent purifications, were digested with 0.15 or 0.3 g, respectively, of endopeptidase Lys-C (Waco) in a reaction volume of 100 l containing 1% RTX100 (Calbiochem), 10% acetonitrile, and 100 mM Tris-HCl, pH 8.0, at 37°C for ϳ16 h (43). Digestion products were chromatographed on a HP 1090 M system (Hewlett Packard) equipped with the above described C4 reverse phase HPLC column equilibrated in 98% buffer A (0.1% trifluoroacetic acid (v/v)), 2% buffer B (80% acetonitrile (v/v), 0.85% trifluoroacetic acid (v/v)). After application of digestion products, the reverse phase matrix was washed with 98% buffer A, 2% buffer B, and bound species were eluted with linear gradients of buffer B increasing to 37.5% over 60 min, to 75% over 30 min, and to 98% over 15 min (44). The eluate was monitored for absorbance at 210 and 280 nm, and peptide peaks were individually collected and analyzed with a model 477A/120A Protein Sequenator (Applied Biosystems). In addition, the NH 2 -terminal sequence of 1 g of concentrated 3-OST sample was directly determined.

Isolation of Mouse 3-OST Clones
Isolation of Cytoplasmic and Poly(A) ϩ RNA-Cytoplasmic RNA (17.5 mg) was isolated from postconfluent cultures of LTA cells (12 flasks of 175 cm 2 , ϳ1.6 ϫ 10 9 cells) by a modification of the procedure of Favaloro et al. (45). Monolayers were twice washed with PBS, cells were recovered by trypsinization and centrifugation (1000 ϫ g for 2 min), and cell pellets were washed by resuspension in PBS followed by centrifugation (1300 ϫ g for 4 min). Cells were lysed by vortexing for 30 s in 12 ml of ice-cold 50 mM Tris, pH 7.4, 140 mM NaCl, 5 mM EDTA, 1% Triton X-100, 5 mM vanadium ribonucleoside complexes (Life Technologies); samples were incubated on ice for 10 min and then vortexed for 1 min. Nuclei were pelleted by centrifugation at 6000 ϫ g for 10 min, the supernatant was mixed with an equal volume of 200 mM Tris, pH 7.4, 300 mM NaCl, 2% SDS, 25 mM EDTA, containing 200 g/ml proteinase K (Boehringer Mannheim), and the mixture was incubated at 65°C for 2 h. Samples were extracted twice against an equal volume of phenol/ chloroform/isoamyl alcohol (25:24:1), the aqueous phase was combined with 0.7 volumes of isopropyl alcohol, cytoplasmic RNA was pelleted by centrifugation at 3500 ϫ g for 10 min, and RNA was resuspended in 3.6 ml of 10 mM Tris, pH 7.4, 1 mM EDTA. Poly(A) ϩ RNA (59 g) was isolated from 16 mg of cytoplasmic RNA by two sequential purifications against 100 mg of oligo(dT)-cellulose (Life Technologies, catalog number 15939 -010) according to the manufacturer's specifications except that binding and wash buffers contained 0.1% SDS, and LiCl was substituted for NaCl. The final eluate (1.5 ml) was extracted against 1.5 ml of phenol/chloroform/isoamyl alcohol (25:24:1), the aqueous phase was then adjusted to 100 mM LiCl and 260 mM NaCl, an equal volume of isopropyl alcohol was added, the mixture was centrifuged at 15,000 ϫ g for 30 min, and the poly(A) ϩ RNA pellet was recovered in 40 l of diethyl pyrocarbonate-treated water.
PCR Cloning and Generation of a Mouse 3-OST Probe-Degenerate PCR primers (Fig. 1) were obtained from Biosynthesis. First strand cDNA was generated in a 50-l volume from 5 g of LTA poly(A) ϩ RNA primed with oligo(dT) using a reverse transcriptase-PCR kit (Stratagene) according to the manufacturer's specifications. Touchdown PCR (46,47) reactions (50 l) contained 1 l of first strand cDNA, 25 pmol of each primer, 0.25 l of AmpliTaq Gold (Perkin-Elmer), a 200 M concentration of each dNTP, and 1 ϫ GeneAmp PCR buffer. Two distinct sets of touchdown PCR conditions were required to obtain optimal yields of product. For amplification with primers 1S and 2A, reactions were heated to 95°C for 9 min, subjected to 20 cycles of 94°C for 30 s and 68°C for 1 min with a 0.5°C reduction per cycle, followed by 20 cycles of 94°C for 30 s and 58°C for 30 s with a 0.5°C reduction per cycle and 75°C for 30 s, and then 15 cycles of 94°C for 30 s, 55°C for 10 s, and ramping to 75°C over 50 s. Alternatively, for amplification with primers 1S and 3A or primers 2S and 3A, reactions were heated to 95°C for 4 min, subjected to 47 cycles of 95°C for 30 s and 69.5°C for 2 min with 0.2°C and 1 s reductions per cycle, followed by 25 cycles of 95°C for 30 s, 60°C for 15 s, and ramping to 75°C over 1 min. Amplification products were purified as the retentate from centrifugal ultrafiltration against a 30,000 molecular weight cut-off membrane (Millipore, catalog number SK1P343JO), and then 200 ng of DNA was end-polished with Pfu DNA polymerase and subcloned into pCR-Script Amp SK(ϩ) (Stratagene, catalog number 211188) according to the manufacturer's specifications. A resulting plasmid, pNWS182, contained the 1S/3A amplification product of 779 bp, which was released by digestion with EcoRI and SacII and isolated by low melting point agarose gel electrophoresis. A 32 P-labeled primer extension probe was then generated with a random primer labeling kit (Stratagene, catalog number 300385) by replacing the random primers with 5 M of primer 3A.
Construction and Screening of an L Cell cDNA Library-Using the manufacturer's recommended conditions an oligo(dT)-primed Zap Express cDNA library (Stratagene, catalog number 200451) was generated from 5 g of LTA poly(A) ϩ RNA that had been pretreated with methylmercury hydroxide. About 1.5 ϫ 10 6 primary recombinants were plaque-amplified by infection into Escherichia coli XL1-Blue MRFЈ. From the amplified library, 1.3 ϫ 10 6 plaques were transferred to a Colony/Plaque Screen (NEN Life Science Products) and screened with the above described 32 P-labeled probe specific for 3-OST. Hybridizations were performed at 42°C in 1.7 ϫ SSC, 8.3% dextran sulfate, 42% formamide, 0.8% SDS, and filters were washed twice with 2 ϫ SSC, 1% SDS for 30 min at 65°C. Positive clones were plaque-purified and then in vivo excised into pBK-CMV-based phagemids by infection with Ex-Assist helper phage followed by transduction of filamentous phage particles into E. coli XLOLR.

Isolation of Human 3-OST cDNA Clones
The National Center for Biotechnology Information data bank of the I.M.A.G.E. consortium (Lawrence Livermore National Laboratory) expressed sequence tag cDNA clones (48) was probed with the deduced mouse 3-OST amino acid sequence to reveal three partial-length species. I.M.A.G.E. consortium CloneID 220372 (accession numbers H86812 and H86876) was from the retinal library of Soares (N2b4HR), whereas clones 301725 (accession numbers N90867 and W16558) and 301726 (accession numbers N90856 and W16555) were from the fetal lung library of Soares (NbHL19W) and were obtained from the TIGR/ ATCC Special Collection (ATCC). The EcoRI/NotI insert of clone 220372 was 32 P-labeled by random priming and used to screen 5 ϫ 10 5 plaques from a TriplEx Brain cDNA library (CLONTECH), as described above. Positive plaques were purified, and TriplEx-based plasmids were in vivo excised according to the manufacturer's protocol and sequenced as described below.

Characterization of Mouse and Human 3-OST cDNA Clones
The 5Ј-and 3Ј-regions of all partial-and full-length clones were enzymatically sequenced from flanking primer sites of the respective cloning vectors. For full-length clones, the remaining sequence of both strands was obtained with internally priming oligonucleotides. Automated fluorescence sequencing was performed with Perkin-Elmer Applied Biosystems model 373A and 477 DNA sequencers. Each reaction typically yielded 400 -600 bases of high quality sequence. cDNA sequence files were aligned and compiled with the program Sequencher 3.0 (Gene Codes Corp.). All additional manipulations were performed with the University of Wisconsin Genetics Computer Group sequence analysis software package. Sequence comparison searches were performed on the data bases of GenBank TM , EMBL, DDBJ, PDB, Swiss-Prot, PIR, and dbEST.

Expression of 3-OST cDNAs
Construction of Expression Plasmids-The plasmid pCMV-3-OST contains the mouse 3-OST cDNA, an EcoRI/XhoI fragment from pNWS228 ( Fig. 2), inserted between the CMV promoter and the bovine growth hormone polyadenylation signal of EcoRI/XhoI-digested and phosphatase-treated pcDNA3 (Invitrogen). The plasmid pCMV-ProA3-OST is of similar structure, except the first 26 amino acids of 3-OST are replaced with 291 amino acids encoding a fusion protein of the transin leader sequence followed by protein A and a factor Xa cleavage site. pCMV-ProA-3-OST was generated by ligating a BamHI/SmaI fragment, containing the protein A region from pRK5F10PROTA (49), and an XmaI (end-filled with T 4 polymerase)/XhoI fragment, containing most of the mouse 3-OST cDNA from pNWS228, into BamHI/XhoIdigested and phosphatase-treated pcDNA3 (Invitrogen). The in vitro transcription plasmid, pNWS237, contains a T3 promoter site 5Ј of the human 3-OST cDNA and was constructed by inserting the complementary oligonucleotides 5Ј-dAATTATTAACCCTCACTAAAGGGAAG and 5Ј-dAATTCTTCCCTTTAGTGAGGGTTAAT (Biosynthesis) into the EcoRI site of the TriplEx based plasmid, pJL30.
Transient Expression of the Mouse 3-OST cDNA in COS-7 Cells-For each expression construct, three 175-cm 2 flasks were seeded with 3.6 ϫ 10 6 COS-7 cells, 6 h later the medium was exchanged with DMEM containing 10% Nu-Serum (Life Technologies) with 100 g/ml streptomycin and 100 units/ml penicillin, and cells were grown for an additional day. Monolayers were washed with PBS and then incubated at 37°C for 2.5 h with 10 ml/flask of freshly prepared DMEM containing 235 g/ml DEAE-dextran (molecular weight 500,000; Pharmacia Biotech Inc.), 9.5 mM Tris-HCl, pH 7.4, 0.9 mM chloroquine diphosphate (Sigma), and 3 g/ml of the appropriate pcDNA3-based expression plasmid. Monolayers were then exposed to freshly prepared 10% Me 2 SO in PBS for 1.5 min, washed twice with nonsupplemented DMEM, and fed 30 ml/flask of DMEM containing 10% fetal bovine serum, 100 g/ml streptomycin, and 100 units/ml penicillin; and cells were grown for an additional day. Monolayers were washed with PBS, and then cells were grown in 40 ml/flask of serum-free medium (DMEM containing 25 mM HEPES, pH 8.0, 1% Nutridoma SP (Boehringer Mannheim) (v/v), an additional 2 mM glutamine, 10 ng/ml biotin (Pierce), 100 g/ml streptomycin, 100 units/ml penicillin, and 1 ϫ a previously described Trace Metal Mix (26)) for 24 h. COS cell-conditioned serum-free medium was harvested, debris was removed by centrifugation at 1,000 ϫ g for 10 min followed by filtration through a 0.45-m membrane, and then samples were either immediately processed or were snap frozen with liquid nitrogen and stored at Ϫ80°C. Occasionally, conditioned medium from a second incubation of 8 -24 h was also collected.
Purification of Wild-type and Protein A-tagged Mouse Recombinantly Expressed 3-OST Enzyme (r3-OST)-Wild-type mouse r3-OST was purified, at 4°C, from 240 ml of freshly generated serum-free medium conditioned by COS-7 cells transfected with pCMV-3-OST. The medium was adjusted to pH 8.0, mixed with an equal volume of 2% glycerol, and then loaded (25 ml/h) onto a heparin-AF Toyopearl-650 M column (0.8 ϫ 5.7 cm) (TosoHaas, Montgomeryville, PA) equilibrated in 50 mM NaCl, 10 mM Tris-HCl, pH 8.0, 1% glycerol (v/v) (buffer C). The column was washed with 20 ml of buffer C at a flow rate of 0.8 ml/min and then with 20 ml of 150 mM NaCl, 10 mM Tris-HCl, pH 8.0, 1% glycerol (v/v) at a flow rate of 0.5 ml/min, and protein was eluted at a flow rate of 0.25 ml/min with a 20-ml linear NaCl gradient extending from 150 mM to 750 mM NaCl in buffer C. The fractions exhibiting HS act conversion activity (approximately 4 ml) were pooled, brought to a final concentration of 0.6% CHAPS (w/v) (Sigma) and dialyzed for 16 h against 4 liters of 25 mM MOPS (3-[N-morpholino]propanesulfonic acid) (Sigma), pH 7.0, 1% glycerol (v/v), 0.6% CHAPS (w/v) (MCG buffer) containing 50 mM NaCl. The dialysate was applied to a 3Ј,5Ј-ADP-agarose column (0.8 ϫ 1.2 cm, 3.7 mmol of 3Ј,5Ј-ADP/ml of gel) (Sigma) and eluted as described previously (26). The fractions containing HS act conversion activity were pooled (approximately 4 ml), aliquoted, frozen in liquid nitrogen, and stored at Ϫ80°C. Protein A-tagged mouse r3-OST was purified, at 4°C, from 155 ml of previously frozen serum-free medium conditioned by COS-7 cells transfected with pCMV-ProA3-OST. IgG-agarose beads (310 l of a 50/50 slurry) were gently stirred with the conditioned medium for 3 h, recovered by centrifugation at 2,000 ϫ g for 10 min, and washed twice with 1 ml of MCG containing 250 mM NaCl to remove nonspecifically bound protein. Protein A fusion protein was eluted from the beads with two sequential 30-min incubations in 100 l of 50 mM sodium acetate, pH 4.5, 150 mM NaCl, 0.6% CHAPS, and 1% glycerol. The pooled eluates were combined with an equal volume of 500 mM MOPS, pH 7.0, 0.6% CHAPS, and 1% glycerol and then aliquoted, frozen in liquid nitrogen, and stored at Ϫ80°C.
Cell-free Synthesis of Mouse and Human r3-OST-Synthetic capped mouse and human 3-OST mRNAs were generated from NotI-linearized pNWS228 (Fig. 2) and HinDIII-linearized pNWS237, respectively, using T 3 polymerase and m 7 G(5Ј)ppp(5Ј)G, as described previously (50). Unlabeled in vitro translation reactions (25 l) contained 0.25 g of synthetic mRNA, 1.8 l of canine pancreatic microsomal membranes (Promega), and 0.5 l each of amino acid mixture minus leucine and amino acid mixture minus methionine and were performed with nuclease-treated reticulocyte lysate (Promega), according to the manufacturer's specifications.
Measurement of HS act Conversion Activity-The HS act conversion activity (a 3-OST-catalyzed reaction that requires unlabeled PAPS to convert [ 35 S]HS inact into [ 35 S]HS act ) of crude and purified r3-OST samples was determined by comparison against a standard curve generated with 1-32 units of previously purified native 3-OST, as described previously (26). The [ 35 S]HS inact substrate was purified from metabolically labeled cell surface HS of exponentially growing clone 33 cells, as described previously (35). 35 S Labeling of HS by r3-OST- 35 S-Labeled HS was generated by incubating the various forms of r3-OST with [ 35 S]PAPS and unlabeled HS inact , which were prepared as described previously (26,35). Wild-type and protein A-tagged r3-OST (2500 units of HS act conversion activity) purified from COS cell-conditioned medium, were incubated in a 500-l reaction mixture, as described previously (26), for 2 h at 37°C, and 35 S-labeled polysaccharides were purified by DEAE-Sepharose chromatography as described previously (26). For cell-free synthesized r3-OST, 35 S labeling of HS was performed in a reticulocyte lysate-based reaction mixture (35) except that 100-l reactions contained 100 -300 units of in vitro translated r3-OST, 180 nM unlabeled HS inact , 5 M PAPS (60 ϫ 10 6 cpm), and samples were incubated at 37°C for 2 h. The reaction was quenched by the addition of 300 l of 267 mM NaCl, 13.3 g/ml glycogen and extraction against 600 l of phenol/chloroform/isoamyl alcohol (25: 24:1). 35 S-Labeled GAGs were ethanol-precipitated (35) and then isolated by DEAE chromatography as described previously (26).

Identification of Enzymatic Reaction Products
Identification of the Site of Sulfation on HS act and HS inact -The DEAE eluates containing 35 S-labeled polysaccharide were vacuum-concentrated to 1 ⁄5 volume and then desalted at a flow rate of 0.9 ml/min on TSK G3000 PW XL (0.78 ϫ 30 cm) and TSK G2500 PW XL (0.78 ϫ 30 cm) (TosoHaas) columns connected in series equilibrated in 0.1 M ammonium bicarbonate. The desalted product was then affinity-fractionated using AT/concanavalin A gel to obtain HS act and HS inact as described previously (26). Analysis of labeled products by treatment with GAG lyases and low pH nitrous acid were performed as described previously (42). In addition, the HS act and HS inact samples were each subjected to hydrazinolysis, high pH nitrous acid (pH 5.5), low pH nitrous acid (pH 1.5), and sodium borohydride reduction with the resultant disaccharides characterized on reverse phase ion pairing HPLC (RPIP-HPLC) as previously reported (33,34). The identification of [ 35 S]GlcA3 AMN-3-O-SO 3 and [ 35 S]GlcA3 AMN-3,6-O-(SO 3 ) 2 was confirmed by co-chromatography on RPIP-HPLC with the appropriate 3 H-labeled disaccharide standards, as described in prior publications (33,34).

Northern Blot Analysis
Total RNA from RFPEC and primary mouse CME cells was isolated by the method of Chomczynski and Sacchi (51), whereas poly(A) ϩ RNA was isolated from HUVEC cells as described above for LTA cells. Total RNA from the mast cell line CI.MC/C57.1 (C57.1) (52) was a generous gift from Dr. Stephen J. Galli (Beth Israel Hospital). Samples were resolved on 1.2% formaldehyde-agarose gels and subjected to Northern blot analysis as described previously (50). Mouse and human samples were hybridized with mouse or human probes, respectively, and washed as described for library screening, above, except hybridizations were performed at 60°C.

Peptide Sequencing and PCR Generation of a Mouse 3-O-Sulfotransferase (3-OST) Probe-The information necessary
for the molecular cloning of mouse heparan sulfate D-glucosaminyl 3-O-sulfotransferase (3-OST) was obtained by sequencing the amino terminus and Lys-C-generated peptides of the enzyme that we had previously purified from large quantities of serum-free tissue culture medium conditioned by an L cell line (26). These studies established the structures of 14 partially overlapping peptides which encompass 185 amino acid residues (see Fig. 3). Degenerate PCR primers were synthesized based on the sequence of the amino terminus (primer 1S) and two endopeptidase-derived fragments (primers 2S, 2A, and 3A) (Fig. 1A). When PCR was performed on an LTA first strand cDNA template, products of about 210 (primers 1S/2A), 780 (primers 1S/3A), and 610 (primers 2S/3A) bp were obtained (Fig. 1B), which suggests that all of the primer sites are contained within a single cDNA (Fig. 1C). To confirm this supposition, the two largest fragments were cloned into pCR-Script Amp SK(ϩ) and inserts were sequenced, which revealed that the 1S/3A product is 779 bp and contains the 611-bp 2S/3A product. The 779-bp insert encodes 12 of the sequenced peptide fragments and so was 32 P-labeled, as described under "Experimental Procedures," and used as a probe for cDNA library screening.
Isolation and Characterization of Mouse 3-OST cDNAs-We constructed an amplified Zap Express LTA cDNA library of 1.5 ϫ 10 6 primary recombinants and screened 1.3 ϫ 10 6 plaques with the above described probe, which revealed 40 positives that were plaque-purified and in vivo excised into plasmids. The cDNA inserts of each plasmid were characterized to eliminate duplicated recombinants due to library amplification. Size was determined by liberating cDNA inserts with digestion at flanking EcoRI and XhoI restriction sites followed by agarose gel electrophoresis; furthermore, the sequence at both ends of each insert was obtained from flanking vector primer sites. This analysis revealed 25 unique primary recombinants, which predominantly contained inserts of approximately 1.7, 2.3, or 3.3 kb. These different species were considered to reflect natural size variants of the mouse message, since Northern blots of LTA poly(A) ϩ RNA hybridized with 3-OST probe revealed the same three size categories of message (data not shown). The complete sequencing of 9 distinct primary recombinants, at least 2 from each size category, in conjunction with the partial sequencing of the remaining 16 clones showed that the size variants result from differences in the length of 5Ј-untranslated region due to the insertion of 0 -1629 bp at a single common internal point, the splice variant

3-O-Sulfation of Heparan Sulfates
site. 2 Most importantly, all clones shared identical protein coding regions, so for the sake of simplicity we present the characterization and analysis of the shortest species, the class 1 cDNA, which lacks additional sequence at the splice variant site.
Sequence data was obtained from two essentially full-length class 1 cDNAs and five partial-length cDNAs ( Fig. 2A) to create a composite cDNA structure (Fig. 3) of 1685 bp, excluding the 3Ј-poly(A) tract. The 5Ј-untranslated region is 322 bp, with the splice variant site occurring between residues 216 and 217. This region contains six ATG sites that do not conform to consensus initiation sites (53) and are followed by near in-frame termination codons. An open reading frame of 933 bp begins at position 323 with the first consensus initiation ATG (a purine occurs at Ϫ3) (53). The length of the 3Ј-untranslated region from all of the cDNA clones analyzed ranged from 301 to 430 bp ( Fig. 2A, and data not shown). Within this terminal 129 bp, five distinct polyadenylation sites were observed, and 13-18 bp upstream from each site is a variant of the consensus polyadenylation signal AATAAA (Fig. 3A). Poly(A) tails were most frequently observed at the first site (position 1556; ϳ50% of clones).
Isolation and Characterization of Human 3-OST cDNAs-Three clones containing partial-length human 3-OST cDNAs were identified by expressed sequence tag data base searching (48) and were obtained from the TIGR/ATCC Special Collection, as described under "Experimental Procedures." Sequencing of the insert ends revealed the clones to be essentially equivalent, since each contained the same 947-bp region of the human 3-OST cDNA (Fig. 2B). The insert of I.M.A.G.E. Consortium CloneID 220372 was 32 P-labeled and used to screen 5 ϫ 10 5 plaques from a TriplEx Brain cDNA library. Three positives were identified and isolated as TriplEx plasmids, and the largest cDNA (1.3 kb) was sequenced completely (Fig. 2B).
The nucleic acid sequences of mouse and human 3-OST cDNAs are ϳ85% similar. The largest isolated human clone contains 118 bp of 5Ј-untranslated region with two nonconsensus ATG sites (Fig. 4). The sequences of human and mouse cDNAs flanking the splice variant site on the 5Ј-limit are distinct (mouse, TTAAAG2; human, GCTCAG2), but on the 3Ј-limit they are identical (both have 2TAATTG), which raises the possibility that human 3-OST mRNA may also exhibit 5Ј splice variants. The first consensus ATG (with a purine occurring at Ϫ3 and a guanosine at ϩ4) (53) initiates an open reading frame of 921 bp. For all four human cDNA clones examined, only a single polyadenylation site was observed, resulting in a 3Ј-untranslated region of 266 bp, which is 26 bp less than the most frequently observed 3Ј-limit for the mouse cDNAs.

Predicted Protein Structures of Mouse and Human 3-OST-
The mouse and human cDNAs encode novel 311-and 307amino acid proteins of 35,876 and 35,750 daltons, respectively (Figs. 3 and 4), that exhibit 93% similarity. The deduced mouse primary structure contains regions corresponding to all 13 sequenced peptides and the amino terminus. For both types of 3-OST, the encoded protein is predicted to be an intraluminal resident. Kyte-Doolittle hydropathy analysis reveals only a single major hydrophobic region, which begins at the amino terminus and lacks sufficient length for a membrane-spanning domain. Moreover, the hydrophobic region differs from a membrane anchor in that it contains two glutamine residues and is FIG. 1. PCR isolation of a mouse 3-OST probe from LTA cDNA. A, peptides 1, 2, and 3 are sequences that were most favorable for the design of degenerate PCR primers (displayed), and are internal regions derived from sequencing the mouse 3-OST amino terminus and Lys-C products F2-55/F1-63 and F2-4 (Fig. 3), respectively. Primers 1S and 2S are of sense orientation, whereas primers 2A and 3A are of antisense orientation. The underlined Gly and Asp of peptides 2 and 3 differ from the respectively encoded cDNA residues of Pro and Trp (Fig. 3), which exhibit low molar recoveries on peptide sequencing. B, resolution of PCR products by 1.2% agarose gel electrophoresis with ethidium bromide staining; the positions of relevant DNA size markers are indicated. First strand cDNA from LTA cells was amplified with the indicated degenerate primers as described under "Experimental Procedures." Amplification with the 1S/2A primers was independently analyzed, since distinct conditions were required to generate a PCR product. Unconsumed primers from the 1S/2A reaction are visualized as a lower band. C, deduced arrangement of primer sites on the mouse cDNA based on PCR results (B) and confirmed by sequence analysis (Fig. 3). not flanked by cationic residues. Thus, the above stretch of 18 residues constitutes a hydrophobic leader signal, and this region is followed by a signal peptidase cleavage site between amino acids 20 and 21, as determined by the method of von Heijne (54). The possibility of signal peptidase cleavage is supported by the amino-terminal analysis of mouse 3-OST, which began with His 21 (Fig. 3). Given that heparan biosynthesis is considered to occur in the trans-Golgi, the above data suggest that the 3-OST is an intraluminal enzyme. Just past the signal peptidase cleavage site, the mouse 3-OST contains an extra 4 residues (Ala 24 -Pro 25 -Gly 26 -Pro 27 ) not found in the human form. Both 3-OST proteins exhibit five potential Nglycosylation sites, which account for the apparent discrepancy between the molecular masses of the predicted amino terminus-trimmed enzyme (ϳ34 kDa) and our previously purified enzyme (a broad band of 46 kDa was observed on SDS-poly-acrylamide gel electrophoresis) (26). 3 Only two cysteine residues are present, and these closely spaced residues are likely to 3 For peptides F2-53 and F2-49B (Fig. 3), sequencing detected an asparagine for Asn 196 , whereas no residue was recovered for Asn 141 , indicating that at least the latter site is likely to be N-glycosylated. Furthermore, in vitro transcription/translation of the 3-O-sulfotransferase cDNA yields a product that migrates on SDS-polyacrylamide gel electrophoresis as a sharp band with an M r of 35,700, but translation in the presence of canine pancreatic microsomal membranes results in five new products of M r 34,100, 36,800, 39,300, 41,900, and 44,500 (i.e. size progressively increased in increments of ϳ2.6 kDa). Endoglycosidase H treatment of translation products does not alter the size of the M r 35,700 band, but it reduces the larger bands to a single species with M r 34,100 (N. W. Shworak, unpublished data). Together, these data strongly suggest that the 3-OST is co-translationally modified by signal peptidase cleavage and is posttranslationally glycosylated with up to four N-linked moieties.

FIG. 2. Sequencing strategy for mouse and human 3-OST cDNA clones.
A schematic representation is shown of mouse class 1 (A) and human (B) 3-OST cDNA inserts contained within the indicated plasmids. The black box indicates the protein coding regions. The arrows summarize the length and direction of sequence obtained from various cDNA clones. Indicated are sequences obtained for essentially full-length clones, pNWS228 and pJL30, from vector primer sites (plain arrows) and 3-OST-specific synthetic oligonucleotides (closed circles). Also indicated are sequences of partial-length cDNAs obtained from vector primer sites (closed squares) and 3-OST-specific synthetic oligonucleotides (open squares). The location of the mouse PCR probes used for library screening is shown, whereas the human library was screened with the insert of clone 220372. The closed triangle indicates the location of the splice variant site as described under "Results." form a disulfide bond that generates a peptide loop of 10 amino acids. 4 Interestingly, the carboxyl-terminal 140-residue region is extremely basic (25% His, Lys, Arg; 12% Glu, Asp); however, this region does not exhibit previously recognized heparin binding motifs.
Expression of Mouse and Human r3-OST-Three distinct expression approaches were employed to confirm that the isolated cDNAs encode 3-OST enzyme. The resulting recombinantly expressed 3-OST enzyme was designated as r3-OST to distinguish this form from our previously purified native 3-OST enzyme. First, the vector pCMV-3-OST (a pcDNA3 derivative in which the CMV promoter transcribes the mouse 3-OST cDNA) was transiently expressed in COS-7 cells, and the resulting level of HS act conversion activity accumulated in serum-free medium over 32 h was measured, as described under "Experimental Procedures." HS act conversion activity is a 3-OST-catalyzed reaction that requires unlabeled PAPS to convert [ 35 S]HS inact into [ 35 S]HS act . Before or after pcDNA3 transfection, typically COS-7-conditioned serum-free medium contained a low but detectable amount of HS act conversion activity, whereas transfection by pCMV-3-OST elevated levels ϳ2,000-fold (Table I).
Second, to exclude the remote possibility that the expression of the mouse 3-OST cDNA indirectly induces, rather than directly encodes, HS act conversion activity, we analyzed a protein A/3-OST fusion protein. COS-7 cells were transiently transfected with pCMV-ProA3-OST, a pCMV-3-OST derivative in which the amino-terminal 26 residues of the mouse 3-OST are replaced with a protein A tag, and protein A-tagged mouse r3-OST was extracted with IgG-agarose beads from 155 ml of conditioned serum-free medium, as described under "Experimental Procedures." The affinity purification recovered undetectable and less than 0.5% of initial HS act conversion activity from control pcDNA3 and pCMV-3-OST transfection samples, respectively, whereas ϳ7,000 units (10% recovery) were extracted from pCMV-ProA3-OST transfection samples. Thus, the mouse 3-OST cDNA directly encodes HS act conversion activity.
Third, we examined the activities of cell-free synthesized mouse and human r3-OST. Synthetic capped mouse and human 3-OST mRNAs were generated by in vitro transcription and then in vitro translated with reticulocyte lysate in the presence and absence of canine pancreatic microsomal membranes, as described under "Experimental Procedures." HS act conversion activity was undetectable in the control in vitro translation reactions that lacked mRNA template, with or  (2), presumptive polyadenylation signals (single underline), and sites at which poly(A) tails were identified in different cDNA clones (double underline). The single underlined amino acids indicate segments of the predicted protein structure that match sequence information obtained from the amino terminus or HPLC-isolated fragments generated from Lys-C digestion of 1.5 and 3.0 g of 3-OST (F1 and F2 series fragments, respectively). Occasional HPLC peaks yielded sequence of two peptides (fragment designations ending in A or B). Potential N-glycosylation sites are in boldface type, and the predicted signal peptidase cleavage site is indicated (1). without microsomal membranes. A low level HS act conversion activity resulted from the addition of synthetic 3-OST mRNA templates to translation reactions lacking microsomal membranes (mouse, 0.86 Ϯ 0.028 units/l, n ϭ 3; human, 2.1 Ϯ 0.063 units/l, n ϭ 3); however, ϳ15-fold greater levels occurred when microsomal membranes were included in translation reactions (mouse, 14.3 Ϯ 0.27 units/l, n ϭ 3; human, 32.4 Ϯ 2.1 units/l, n ϭ 3). The apparent activation of nascent r3-OST by co-translational processing within microsomes may result from signal peptidase cleavage, N-linked glycosylation, and/or a facilitation of correct protein folding. The slightly greater production from the human 3-OST cDNA may reflect the more favorable context of the human initiation codon or the reduced length of the human 5Ј-untranslated region. Independent of the above considerations, the above data confirm that isolated mouse and human cDNAs encode HS act conversion activity.
We next examined the biochemical specificity of the HS act conversion activity generated from each expression approach by incubating crude or purified enzyme with [ 35 S]PAPS and unlabeled HS inact , recovering radiolabeled GAG by DEAE chromatography and characterizing the resultant products. The HS act conversion activity of the wild-type mouse r3-OST produced by transfecting COS-7 cells with pCMV-3-OST (1.35 ϫ 10 6 units in 240 ml of conditioned serum-free medium) was first purified away from potential contaminating sulfotransferase activities by heparin-AF Toyopearl chromatography followed by 3Ј,5Ј-ADP-agarose chromatography, which yielded ϳ1 g of protein containing 340,000 units (ϳ20,000-fold purification with 25% overall recovery), whereas the IgG agarosepurified protein A-tagged r3-OST and in vitro translation reactions of mouse and human 3-OST mRNA templates were directly analyzed, 5 as described under "Experimental Procedures." About 0.5-1 ϫ 10 6 cpm of product was generated with purified wild-type r3-OST, purified protein A-tagged r3-OST, and nonpurified in vitro translation reactions containing mouse and human r3-OST, respectively. Portions of each labeled product were incubated with purified heparitinase (0.5 units/ml) or chondroitinase ABC (0.5 units/ml), and HPLC-gel permeation chromatography analysis indicated that in all cases label was exclusively incorporated into HS. Portions of the labeled HS samples were also N-desulfated with nitrous acid at pH 1.5 and analyzed by P-2 polyacrylamide gel filtration to determine the amounts of liberated free [ 35 S]sulfate, as described under "Experimental Procedures." The results demonstrated no increased generation of free [ 35 S]sulfate (data not shown). Finally, portions of the labeled samples were AT affinity-fractionated, which revealed that in each case ϳ40% of the 35 S-label was incorporated in HS act and approximately ϳ60% of the 35 S-label was incorporated in HS inact . The labeled HS act and HS inact generated by the wild-type purified r3-OST were 5 When in vitro translation reactions lacking 3-OST mRNA templates were incubated with [ 35 S]PAPS and unlabeled HS under otherwise standard conditions, the radioactivity recovered in DEAE eluates was 0.032 Ϯ 0.0023% (translates without microsomes, n ϭ 3) and 0.044 Ϯ 0.0090% (translates with microsomes, n ϭ 3) of initial radioactivity, whereas similar incubation of translates containing mouse or human 3-OST mRNA templates resulted in the incorporation of 1-3% of initial radioactivity. Thus, direct analysis of in vitro translates should only reflect the sulfotransferase activity of the translated r3-OST.

3-O-Sulfation of Heparan Sulfates
chemically cleaved to disaccharides, appropriate 3 H-labeled disaccharide standards were added, and the 35 (6,8,55). Three size categories of rodent 3-OST mRNA (about 1.7, 2.3, and 3.3 kb) and a single size species of the human message (about 1.7 kb) (Fig. 6) are evident. As described above, the mouse forms arise from differential splicing within the 5Ј-untranslated region. Similar size categories are also expressed by rat (RFPEC) endothelial cells, suggesting a similar mechanism of origin. The abundance of each category varies with each cell line, which suggests that a mechanism exists to regulate such dif-ferential splicing. The immortalized mouse mast cell line, C57.1, expresses high levels of the same three size categories, which suggests that expression of a single 3-OST gene is required for the synthesis of both HS act and anticoagulant heparin.
The 3-OST Sequence Defines a Heparan Sulfotransferase Family-Extensive data bank searching revealed the 3-OST enzyme to be a previously unidentified protein; furthermore, the carboxyl-terminal 250 residues exhibit a low homology (ϳ30% similarity) to many previously identified sulfotransferases (which are typically ϳ300 residues in length) including chondroitin-, aryl-/phenol-, N-hydroxyarylamine-, alcohol-/hydroxysteroid-, flavonol-, and nodulation factor sulfotransferases (data not shown). We also observed a slightly greater homology (ϳ40% similarity) to a functionally unidentified open reading frame of 247 amino acids from Aeromonas salmonicida (GenBank TM accession number L37077). More importantly, the 3-OST protein exhibits ϳ50% similarity with all previously identified forms of the heparan biosynthetic enzyme NST (representatives shown in Fig. 7). In particular, extensive homology exists across the entire 250 -270 carboxyl-terminal residues of these enzymes. Thus, it appears that a common sulfotransferase structure is shared by two distinct types of heparan biosynthetic enzyme. Given that NST is a bifunctional enzyme, the above observation suggests that NST enzymes possess sulfotransferase activity within a ϳ270-residue carboxyl-terminal domain, whereas deacetylase activity would be contained within the remaining ϳ560 luminal residues. Interestingly, the region of consensus Lys 302 -Arg 323 , which encompasses the presumptive cystine-bridged peptide loop (described above), exhibits complete conservation for 12 of the 22 residues (including both cysteines) among all 3-OST and NST species. DISCUSSION We have previously demonstrated that the cellular rate of HSPG act generation is determined by the levels of a kinetically limiting microsomal activity, HS act conversion activity, and have proposed that 3-OST is the principal rate-limiting component of this microsomal activity (35). Based on this hypothesis, we purified the rate-limiting component of HS act conversion activity and confirmed its identity as 3-OST (26). During the current investigation, we have isolated and characterized the mouse and human 3-OST cDNAs. Three independent lines of evidence conclusively demonstrate that the murine and human cDNA clones encode the 3-OST from each species. First, the mouse cDNA exhibits all 13 sequenced peptides and the amino terminus of the purified enzyme; moreover, the primary  6. Expression of rodent and human 3-OST in different cell types. Northern blot analysis was performed on 10 g of rodent total RNA or 5 g of human poly(A) ϩ RNA isolated from various primary and immortalized cell types. Samples were hybridized with 32 P-labeled mouse or human probes, respectively, as described under "Experimental Procedures." Rodent samples are from an immortalized rat endothelial cell line (RFPEC), mouse primary cardiac microvascular endothelial cells (CME), and an immortalized mouse mast cell line (C57.1). Human poly(A) ϩ RNA was prepared from primary endothelial cells of umbilical vein (HUVEC). Sizes were determined by coelectrophoresis of mRNA size standards (Promega). amino acid sequence obtained from the enzyme corresponds to about 60% of the deduced coding region of the cDNA. It is also of significance to observe that the putative molecular weight of mouse 3-OST deduced from the cDNA, in conjunction with the 5 potential N-glycosylation sites, is consistent with the size of the previously purified enzyme as judged by SDS-polyacrylamide gel electrophoresis (a single broad band with an M r of 46,000) (26). Second, the expression of wild-type mouse r3-OST as well as protein A-tagged mouse enzyme by transient transfection of COS-7 cells and the expression of both wild-type mouse and human r3-OST by in vitro transcription/translation shows that the two isolated cDNAs directly encode HS act conversion activity. Third, transiently expressed wild-type and protein A-tagged mouse r3-OST isolated from COS-7 conditioned medium, and crude reaction mixtures from in vitro transcription/translation of mouse and human synthetic 3-OST mRNAs are all able to 3-O-sulfate heparan sulfate in a fashion identical to the previously purified enzyme (26). Specifically, the incubation of the recombinant proteins with unlabeled HS inact and [ 35 S]PAPS generated labeled HS act (40% of incorporation) and HS inact (60% of incorporation) due to transfer of We have previously demonstrated that HS inact is composed of two functionally distinct populations, the HS act precursor as well as the HS inact precursor (26,35). The HS act precursor exhibits an oligosaccharide structure corresponding to the ATbinding site devoid of the glucosamine 3-O-sulfate group, whereas the HS inact precursor does not possess this sequence, since 3-O-sulfation fails to produce the AT binding site (26). Our previous analyses of L cells in which HS act synthesis was perturbed by overexpression of ryudocan core protein or chemical mutagenesis suggest that 3-O-sulfation of HS act versus the HS inact precursor may take place in a differential manner; however, it is unclear whether differential sulfation of the two precursors involves multiple 3-OST gene products, a posttranslational modification of a single enzyme, or the presence of accessory biosynthetic factors (33,34). Although purified 3-OST modifies both populations with similar kinetic properties (26), it remained possible that the enzyme preparation contained two functionally distinct species with minor variations in primary sequence. However, the mouse and human r3-OST generated by either transient transfection of COS-7 cells or in vitro translation, sulfated both precursors in a fashion identical to the purified enzyme. Thus, a single 3-OST enzyme with a unique primary sequence can 3-O-sulfate both HS act precursor and the HS inact precursor. Accordingly, differential modification of these two substrates may involve a posttranslational modification, the presence of accessory biosynthetic factors, or possibly Golgi compartmentalization. This analysis, of course, does not exclude the potential existence of novel isoforms of the enzyme encoded by additional genes. The elucidation of mechanisms that regulate differential 3-O-sulfation may provide insight into the biological significance of the 3-O-sulfated glucosamine residues within HS inact and nonanticoagulant heparin (33,34,39,40).
Most previously isolated Golgi enzymes, including NST-1 and NST-2 (Refs. 27-29 and references therein) are type II membrane-bound proteins; however, 3-OST appears to be an intraluminal Golgi resident. Unlike type II membrane bound proteins, the 3-OST has no cytoplasmic domain. Furthermore, its hydrophobic leader sequence is not flanked by cationic residues, contains two hydrophilic residues (Gln), lacks sufficient length to span a membrane as an ␣-helix, and is followed by a signal peptidase cleavage site that is consistent with the aminoterminal analysis of purified 3-OST. Soluble intraluminal proteins have previously been identified as residents of the endoplasmic reticulum (56 -58); however, the recent isolation and characterization of Cab45 reveal a soluble resident of a postmedial Golgi compartment (59). Retention and localization of type II proteins involve residues adjacent to and within the transmembrane domain (60, 61), whereas endoplasmic reticulum retention of intraluminal proteins requires a specific carboxyl-terminal tetrapeptide, (H/K)DEL (56 -58). Consistent FIG. 7. Comparison of putative sulfotransferase domains. The program Pileup was used to align amino acid sequences from mouse 3-OST (residues 1-311), human 3-OST (residues 1-307), rat N-deacetylase/N-sulfotransferase 1 (rNST1, residues 516 -882, accession number M92042) (28), mouse N-deacetylase/N-sulfotransferase 2 (mNST2, residues 515-883, accession number U02304) (27), and Caenorhabditis elegans Ndeacetylase/N-sulfotransferase (ceNST, residues 459 -819, accession number U52002) (68). Consensus residues (shaded) are indicated for each position where at least four candidates exhibit identical or similar amino acids. Numeration is given for each enzyme and for a consensus sequence.

3-O-Sulfation of Heparan Sulfates
with this difference, the carboxyl-terminal tetrapeptide residues are conserved between mouse and human 3-OST but are different from the comparable NST residues (Fig. 7). Interestingly, Cab45 and 3-OST exhibit distinct residues at this site (HEEF and FDWH, respectively) (59), which suggests localization to different Golgi compartments. The high levels of 3-OST that accumulate in tissue culture medium testify to a leaky retention mechanism, which raises the possibility that modulation of enzyme retention could serve as a mechanism to control the cellular rate of HS act generation.
The sulfotransferases that act upon different substrates exhibit extensive structural diversity; indeed, similarity is greatest between members of this enzyme class that sulfate related substrates (62). Consistent with this observation, 3-OST and all known NST species possess a homologous carboxyl-terminal domain of ϳ260 residues that also exhibits homology to all known sulfotransferases. Given that this region constitutes Ͼ88% of the protein A-tagged r3-OST and so should contain the machinery for sulfation, we propose that a common domain structure is shared by heparan sulfate sulfotransferases or at least by heparan glucosaminyl sulfotransferases. Indeed, on the basis of this hypothesis, we have employed this homology to isolate cDNAs of additional heparan sulfotransferases. 6 The homologous region does not exhibit the motif GXXGXXK, which is found in many sulfotransferases (28) and is thought to participate in PAPS binding due to its homology with the consensus sequence of the glycine-rich phosphate binding loop (P-loop), GXXXXGK(S/T) (63,64). However, the presumptive cystine-bridged peptide loop region contains a sequence (Fig. 7, consensus residues 311-316) similar to GGKLEKC of aryl sulfotransferase IV. This region is thought to be positioned near the active site as indicated by affinity labeling of the contained Lys 65 and Cys 66 with ATP dialdehyde (65). Thus, the cystinebridged peptide loop of 3-OST may be located near the active site, which is consistent with the above observation. In this regard, we note that integrity of the disulfide bond may be required for enzymatic activity, since 3-OST is inactivated by incubation with dithiothreitol. 7 Interestingly, dithiothreitol also inhibits the activity of heparan sulfate 6-O-sulfotransferase but not chondroitin 4-O-, and 6-O-sulfotransferases (25,66). We note that endothelial cell synthesis of HS act is suppressed by exposure to the thrombogenic agent homocysteine and other reducing agents (67). Thus, it is possible that alterations in the cellular redox state may affect HS act synthesis by disrupting a critical disulfide bond in the 3-OST. The availability of the 3-OST cDNA now provides a critical tool for determining how endothelial cells regulate HSPG act generation and should permit an exploration of the molecular defects in this natural anticoagulant pathway that lead to vascular disorders.