Molecular cloning and expression of a glycosaminoglycan N-acetylglucosaminyl N-deacetylase/N-sulfotransferase from a heparin-producing cell line.

Heparin has a higher content of N-sulfated glucosamine and L-iduronic acid than heparan sulfate. Deacetylation of N-acetylglucosamine followed by N-sulfation may be important steps differentiating the biosynthesis of these glycosaminoglycans. We have cloned, by cross-hybridization with the cDNA from rat liver heparan sulfate N-deacetylase/N-sulfotransferase, a protein from a heparin synthesizing mastocytoma derived cell line called MST. This protein, which has both N-deacetylase/N-sulfotransferase activities, has a predicted amino acid sequence homology of 70% with the above rat liver enzyme and is unique for the following reasons. 1) It was found to be encoded by a 3.8-kilobase mRNA that was unique to heparin-producing cells; an 8.5-kilobase mRNA encoding the rat liver enzymes has been found to occur in all mammalian cells tested on the basis of nucleic acid cross-hybridization; 2) the protein overexpressed in COS cells in its full-length transmembrane form or as a soluble secreted protein A chimera displayed ratios of N-deacetylase to N-sulfotransferase activities that were 4-8-fold higher than that observed for the enzyme found in liver that is involved in the biosynthesis of heparan sulfate. These results suggest that the MST-derived enzyme is probably unique to the production of heparin in mast cells.

mRNA encoding the rat liver enzymes has been found to occur in all mammalian cells tested on the basis of nucleic acid cross-hybridization; 2) the protein overexpressed in COS cells in its full-length transmembrane form or as a soluble secreted protein A chimera displayed ratios of N-deacetylase to N-sdfotransferase activities that were "fold higher than that observed for the enzyme found in liver that is involved in the biosynthesis of heparan sulfate. These resulta suggest that the MST-derived enzyme is probably unique to the production of heparin in mast cells.
Heparin and heparan sulfate (HS)l are complex polysaccharides composed of alternating glucosamine and uronic acid residues some of which are sulfated. Avariety of biological functions have been proposed for heparin and HS, among them blood clotting (l), cell recognition (21, cell adhesions (3, 4) viral binding and infection (51, endocytosis (61, regulation of bFGF activity (7-ll), and developmental regulation of neural tissues (12). One of the reasons that could account for the broad range of functions of these molecules is their high degree of heterogeneity resulting from the different patterns and extent of sulfation.
Heparin biosynthesis occurs only in connective tissue mast cells, whereas that of HS seems to be ubiquitous (13). In addi- tion, heparin contains more N-sulfate groups, a higher total sulfate content and more iduronic acid than HS (14). These characteristics, which may be relevant to the biological functions of heparin and HS, also point to differences in their biosynthesis. Heparin and HS share biosynthetic pathways (14) beginning with polymerization of GlcNAc and G1cA followed by a series of modification reactions: N-deacetylation of N-acetyl-glucosam~e and further N-sulfation. This is a key step for the next set of reactions, epimerization of GlcA to L-iduronic acid, 2 -0 sulfation of r.-iduronic acid and 6-0 and 3-0 sulfation of glucosamine. Since the N-deacetylatio~N-sulfation of GlcNAc is an obligatory step for the subsequent reactions, changes in these activities could account for significant alterations in the sulfation pattern of heparin and HS. Recently, we have cloned, sequenced, and expressed a single protein from rat liver which catalyzes both N-sulfation and N-deacetylation during the biosynthesis of HS (15, 16). In mouse mastocytoma, which produces large amounts of heparin, Pettersson et al. (17) have reported that a protein possessing N-sulfotransferase activity acquires N-deacetylase activity upon addition of a crude factor, histones or polybrene (181, suggesting that in a heparin-s~thesizing tissue, there may be a different mechanism for controlling the N-deacetylation and N-sulfation of gIucosamine. Using the rat liver N D N S cDNA as a probe, we searched for homologous sequences in Northern blot analyses of poly(A+) RNA from MST cells, a stable heparin-producing cell line derived from the Furth murine mastocytoma (19). At low stringency, a message of 3.8 kb was detected, which was absent in non-heparin-producing cells. When the same probe was used to screen a nonamplified cDNA library from MST cells, two fulllength clones were isolated. Both have identical sequence and differ only in their length at both 5' and 3' ends. The MST cDNA contains a single open reading frame, predicting a type I1 membrane protein (as all other Golgi membrane proteins cloned SO far) which shares high identity with the rat liver NDNS. Upon transient transfection of COS cells with the MST cDNA, overexpression of both activities in vitro was found as well as higher specific incorporation in vivo of radiolabeled sulfate into heparan sulfate. Cell extracts of transfected COS cells and a purified soluble fusion protein containing the putative Golgi lumenal domain of the MST enzyme showed a 4-8-fold higher NDNS ratio compared with experiments with the rat liver enzyme; this suggests that increased N-deacetylation of GlcNAc followed by N-sulfation may be an important step and perhaps differently regulated in the biosynthesis of heparin compared with HS. Preparation of lbtal RNA and Poly(A+) RNA-Total RNA was extracted from MST cells using the method of Chomczynski et al. (20). To prepare Poly(A+) RNA, total RNA was loaded onto an oligo(dT) column (Life Technologies, Inc.) and fractionated following the manufacturer's instructions.

MATERIALS AND METHODS
RNA Hybridization-RNA was denatured at 65 "C for 10 min in a solution containing 34% (v/v) formamide, 4% formaldehyde, 13.5 m~ MOPS, pH 7.0,67 pg/ml ethidium bromide and then fractionated in a 1.2% agarose gel containing 6% (v/v) formaldehyde, 20 m~ MOPS, pH 7.0, 5 m~ sodium acetate, and 1 m~ EDTA. After treatment of the gel with 50 m~ NaOH, the gel was blotted to a Hybond N+ nylon membrane for 18 h. The membrane was then baked at 80 "C for 2 h and incubated at 65 "C for 2 h in a prehybridization solution consisting of 6 x SSPE, 5 x Denhardt's solution, 0.5% SDS, and 100 & m l of denatured salmon sperm DNA. Single strand probes were prepared as described by Hashimoto et al. (15). Probes were added to the prehybridization solution (4 x lo6 cpdml) and used to incubate the membrane for 15-18 h at 65 "C.
For low stringency washes, the membrane was incubated twice for 15 min at room temperature with a solution containing 2 x SSC, 0.1% SDS and then twice at 45 "C for 15 min. For high stringency washes, the membrane was initially washed twice with 2 x SSC, 0.1% SDS for 15 min at room temperature, and then twice with 0.1 x SSC, 0.1% SDS at 65 "C for 15 min. The membranes were then exposed at -70 "C using screens.
Construction of a hgtl0 Library-Poly(A+) RNA from MST cells was obtained as described above. The library was constructed with the Superscript Choice System (Life Technologies, Inc.) using 5 pg of poly(A+) and following the manufacturer's instructions. The cDNA was then ligated to A g t l O arms (Life Technologies, Inc.) and packaged in vitro using a ADNA packaging kit from Stratagene.
Screening of the &tlO Library-Phages obtained as described above and without further amplification were used to infect Escherichia coli C600 M and plated at 2-3 x lo4 plaque-forming uniteJ150-mm plate. 1.5 x lo6 plaques were screened as described by Sambmok et al. (211, using the hybridization conditions described previously for low stringency. After the primary screening, 26 positive signals were obtained.
Analysis of Positive Clones-Seven positive clones were treated with SalI, EcoRI, and NotI, followed by Southern analyses. The size of the inserts were 0.8, 1.7, 1.8, 2.0, 2.2, 3.6, and 3.7 kb. The last two were cloned into the pCMV5 vector as described below; the others were amplified by polymerase chain reaction, cut with EcoRI and cloned into the EcoRI site of M13mp18 and partially sequenced.
Construction of pCMST Plasmids-DNA of the A clones containing the 3.6 (MST-1)-and 3.7 (MST-3)-kb inserts were cut with EcoRI and ligated into the EcoRI site of the pCMV5 expression vector (22). DH-5 OL cells were transformed with the ligation mixture and plated in LB ampicillin plates. To distinguish the orientation of the inserts, recombinant plasmids were analyzed by restriction mapping. The plasmids were named pCMSTl and pCMST3.
DNA Sequence Analysis-DNA sequencing was done according to Sanger et al. (23) using deoxyadenosine 5' a-36S-triphosphate and Sequenase (United States Biochemical Corp.). Both strands of pCMST plasmids were sequenced according to the manufacturer's protocol using specific primers. The inserts cloned into M13mp18 were sequenced using the universal primer. These were found to be identical to the pCMST plasmids, demonstrating that they are partial clones of the same cDNA.
Detection of mRNAs Using cDNA of MST as Probe-Poly(A+) RNA from MST cells and rat liver was processed for Northern blot analysis as described above. To prepare the probe, the insert of MST5-2 ( Fig. 2) was amplified by polymerase chain reaction, separated on low melting point agarose gels, and purified using Magic polymerase chain reaction preparation from Promega; 50 ng of this DNA were labeled by random hexamer using an oligolabeling kit (Pharmacia LKB Biotechnology Inc.), followed by purification with a Nick column (Pharmacia). Hybridization was done as described above and the membrane was washed at low stringency conditions and exposed with a screen at -70 "C for 18 h.
Dansient Expression in COS-1 Cells-Transfection was done as described previously (15). Briefly, COS-l cells, grown on 1OO-mm plates, were transfected with different plasmids using the DEAE-dextran method (24). Five ml of DMEM containing 10% Nu serum (Collaborative Biomedical Products) were mixed with 0.2 ml of a solution ofphosphate-buffered saline containing 10 mg/ml DEAE-dextran plus 2.5 m~ chloroquine; DNA was added, and after vortexing, the mixture was immediately added to the cells. The cells were incubated for 4 h. The medium was replaced with 5 ml of 10% dimethyl sulfoxide in phosphate-buffered saline, incubated for 2 min at room temperature, and then changed to 25 ml of 10% (DMEM). Following 60 h, cells were scraped off the dish and the homogenate used to measure ND/NS activities.
Radiolabeling of COS Cells with Sulfate and Glucosamine-Wild type COS cells (3 x lo6 celldl0-cm plate) and mutant CM-15 COS cells (6 x lo6 celldlO-cm plate) were transfected using 5 pg of either the MST or rat liver heparan sulfate ND/NS cDNA. Cells were also transfected with the same amount of vector alone. Forty-eight hours after incubation, cells were radiolabeled for 24 h with 50 pCi/ml of HzS6S04 or 50 pCi/ml of [6-3H]glucosamine (DuPont NEN) in DMEM medium lacking sulfate (Life Technologies, Inc. formula 89-0053 AK) supplemented with 10% bovine fetal bovine serum (dialyzed against water) and 1 g/ml of glucose. The culture media was then removed and cells were rinsed twice with ice-cold phosphate-buffered saline and the washes combined with the original culture medium. Cells were then trypsinized, plates were washed twice with phosphate-buffered saline, and the cell suspension and phosphate-buffered saline washes were combined and centrifuged to pellet the cells. The supernate was then added to the culture medium pool. The glycosaminoglycans of this fraction and the cell pellet were purified by chromatography on DEAE-cellulose columns. Total incorporation of radiolabel was measured as the radioactivity remaining at the origin on a paper chromatogram following chromatography for 15 h in n-butanol, glacial acetic acid, 1 N NH40H (2:3:2). The content of heparan sulfate or chondroitin sulfate was determined as the difference between total radiolabel remaining at the origin compared with the label remaining following treatment of the samples with either chondroitinase ABC and chondroitinase AC 11. Construction and Expression of a Soluble Form of the MST Protein-Plasmid construction, protein expression, and purification of a soluble form of MST protein (a chimera of the Golgi lumenal domain and protein A) followed the procedures used for the rat liver-derived enzyme (16) with the exception of the primers utilized in the polymerase chain reaction which were:

5'GGTGGATTCCAAGGCCAAGGAACCCTT-GCC3' for the 5' region and S'GGTGGATTCTCAGCCCACACTGGAAT-
GTT3' for the 3' end region and the template which is the plasmid pCMST-1. In such way the protein A-MST fusion protein comprise residues 43-883 of the MST protein.
N-Deacetylase and N-Sulfotransferase Assays-The N-deacetylase activity in COS-1 cells extracts was measured by determining the release of r3H1acetate from N-PHIacetylated polysaccharide derived from E. coli K5-derived capsular polysaccharide with a specific activity of 400 cpm/ng (dry weight) after 1-h incubation as described by Bame et al. (25). The N-deacetylase activity of the purified soluble protein was measured using 69 ng of protein, and incubations were carried out for 30 min. N-Sulfotransferase activity in COS cells extracts was measured by determining the incorporation of 36s04 into heparan sulfate from [36Sladenosine 3"phosphate 5'-phosphosulfate (PAPS) as described by Brandan and Hirschberg (26). The N-sulfotransferase activity of the purified soluble protein was measured using 69 ng of protein and determining in 15 min the incorporation of 36S04 from [36S]PAPS into deacetylated K5 polysaccharide as described by Ishihara et al. (27). All reactions were linear up to 60 min and 280 ng of protein.

RESULTS
Northern Analyses ut Different Stringencies-Poly(A+) RNA from the heparin-producing mastocytoma-derived cell line MST was analyzed by Northern blots at low and high stringency using a single strand probe with an average size of 300 nucleotides derived from the open reading frame of the rat liver NDmS cDNA (Fig. lA). At high stringency, an 8.5-kb band was observed with poly(A+) RNA from MST cells. This mRNA coincides with the principal message of rat liver. In contrast to rat liver, no messages of 4.2 and 7.0 kb were observed. At low stringency, no significant differences were observed with rat liver RNA, but with MST, a very strong band of 3.8 kb was seen, suggesting a very abundant message, different from any homologous message seen in liver (Fig. lA) or other cells such as C O~O 320, SW-403 (human colon carcinoma) RLC (rat liver), COS-1 (green monkey kidney fibroblasts) or tissues such as human placenta and Ehrlich ascites tumor cells (not shown). However, in all these cells, a message of 8.5 kb was detected at high stringency. rat liver ND/NS, we decided to further investigate its significance. For this, a cDNA library from MST cells was constructed in A g t l O and screened at low stringency conditions. Twenty-six positive clones were identified, and seven of them were analyzed in detail (Fig. 2). Sequencing analysis showed that all of them were different cDNA clones from the same mRNA, five of them being partial lengths and two full length designated as MST-1 and MST-3. To determine which message gave rise to these cDNAs, a probe prepared from a portion spanning the 5' half of the MST cDNA was used in Northern analyses (Fig. 1B 1. The probe recognized a 3.8-kb message in poly(A+) RNA from MST; under the same conditions, no signal was detected at 8.5 kb in poly(A+) RNA from MST and none with poly(A+) RNA from rat liver (Fig. 1B).
Sequencing Analysis-The sequence revealed that MST-1 and MST-3 were identical except for a minor difference in the length given by a few residues on both ends. In addition, there is a point mutation a t nucleotide 2790 in the 3"untranslated region, perhaps an artifact generated during the construction of the library.
From the nucleotide sequence (Fig. 3), a single open reading frame predicts a protein of 883 residues with a molecular mass of 10,1196 Da. The ATG located at position 1 appears to be the correct initiation site, since it is the first methionine in the open reading frame and fits with the consensus sequence for initiation codons (28). Upstream from that ATG, there is a stop codon in the same reading frame at position -36, -96, and -102. Hydrophobicity plots of the protein showed a highly hydrophobic region between residues 19 and 41 which accounts for the only putative transmembrane domain in the predicted protein (consistent with a type I1 membrane protein). The protein has seven potential N-glycosylation sites (shown as black dots). Comparison with the Rat Liver N-DeacetylaselN-Sulfotransferase-The predicted protein sequence was compared with that from rat liver (Fig. 4). Both proteins are highly homologous and share 70.3% identity plus 10% amino acid similarity. There is less homology toward the amino terminus of both proteins. If the amino terminus of both proteins are compared (MST-protein from amino acids 1-85 with rat liver amino acids 1-84), only a 26% identity is found with an additional 30% similarity. If both proteins are compared removing the amino terminus, the identity increases to 74.8% and similarity to an additional 14.7%.
Although this sequence similarity suggested that the rat liver and MST protein may have similar functions, it may be significant that it is the transmembrane and stem region where they have lesser identity and homology; these regions are thought to be targeting domains of proteins to the Golgi apparatus (29). This raises the possibility that these proteins are not localized in the same (sub)compartment within the cell. This in turn may be relevant to the hypothesis that the MST protein is involved in the biosynthesis of heparin, whereas the one from liver of heparan sulfate (see below).
Expression of MST Clones in COS-1 Cells: the MST Protein Contains Both N-Deacetylase and N-Sulfotransferase Activities "In order to determine whether the MST clones encoded a protein containing ND/NS activities, both MST-1 and MST-3 were cloned in the expression vector pCMV-5 generating clones pCMST-1 and pCMST-3. Both clones and the vector containing the cDNA from rat liver (Fig. JA) were transfected into COS-1 cells. N-Deacetyiase and N-Sulfotransferase activities were measured in cell extracts 65 h after transfection. Both activities were overexpressed (Table I). Surprisingly, the ratios of N-deacetylaselN-sulfotransferase was 4-8-fold higher in the cell extracts of MST cDNA transfected cells compared with those with rat liver cDNA. The same lower ND/NS ratio was found when COS-1 cells were transfected with a vector containing a soluble form of the rat liver cDNA, suggesting, as a first approximation, that the differences in the activity ratios with rat liver cDNA are the results of the rat liver protein and not factors of the cell extracts. The MST N-sulfotransferase activity was significantly lower than that of rat liver.
To further substantiate the above hypothesis, a fusion protein was constructed using the putative lumenal portion of the MST protein fused, a t its amino terminus, to protein A, analogous to the above rat liver fusion protein (16). This generated a secreted soluble protein which was purified by its ability to bind IgG beads. SDS-gel electrophoresis of this preparation showed a single protein band with the expected size (not shown), which contained both N-deacetylase and N-sulfotransferase activity (see below).
The above results suggest that the MST-derived enzyme ex-     Table 11).  (35) who found activation using microsomes from mastocytoma. Ib further evaluate the differences between the rat liver and MST-derived enzymes, a detailed kinetic analysis was undertaken using purified fusion proteins. As a substrate, the polysaccharide capsule of E. coli K5 was utilized (301, since it contains a repeating backbone of glucuronic acid and N-acetylglucosamine, the hypothetical form of the endogenous substrate (see "Discussion").

~-s~f o~s f e r a~ N-Deacetytase
The most striking difference between the two proteins was their affinity for N-deacetylated K5 polysaccharide, the substrate for the N-sulfotransferase (Table I1 and Fig. 5, A and B ).
The apparent Km for the rat liver N-sulfotransferase was 33fold lower than that of the MST enzyme. Other kinetic parameters showed differences, but of lesser magnitude (Table 11).
The Km,app for PAPS was almost hdf for the MST enzyme compared with rat liver, whereas the V , , for N-sulfation was higher for the latter enzyme. The Km,app for the N-deacetylase was the same for both enzymes but the V , , for that of MST was higher than rat liver.
When the NDNS V , , ratios were compared between the rat liver and MST proteins, the latter were 4-fold higher (Table 111, suggesting that, based on kinetic behavior, the rat liver and MST enzymes are different.  Table 111, the total amount of radiolabeled sulfate incorporation into glycosaminoglycans in CM-15 cells increased by 25% compared with cells transfected with the vector alone; most of the increase was into heparan sulfate with only negligible increase into chondroitin sulfate. The increase in heparan sulfate radiolabeling was observed in the culture medium, extracellular matrix, and intracellular pool, suggesting that no particular pool of heparan sulfate was sulfated preferentially. Incorporation of [3H]glucosamine did not change significantly in cells transfected with the cDNA compared with controls strongly, suggesting no increase in polymer size had occurred and that the above cDNA encodes for a protein with NDNS activity in uiuo.

Incorporation of Radiolabeled Sulfate into Macromolecules
When sulfate incorporation was measured into wild type COS cells, no significant changes in sulfate incorporation, compared with cells transfected with the vector alone, were detected into either heparan sulfate or chondroitin sulfate, in agreement with previous studies using rat liver cDNA and wild type COS cells (34).

DISCUSSION
We have cloned, overexpressed, and characterized a protein fiom MST cells which has heparin-heparan sulfate N-deacety-1as~N-sulfotr~sferase activities. Several lines of evidence suggest that this protein may be involved primarily in the biosynthesis of heparin and not in that of heparan sulfate. 1) MST cells, the source of the mRNA encoding this protein, synthesize large quantities of heparin (19) as does mastocytoma tissue, from which the cell line was derived. 2) When Northern analyses were performed with different tissues, a very abundant message of 3.8 kb was detected in MST at low stringencies, using as probe the cDNA from the rat liver heparan sulfate  ND/NS. This message was not found in other tissues which do not synthesize heparin. At low and high stringencies, a message of 8.5 kb was detected in MST as well as other tissues, which was much less abundant than the above message. Because this message was also detected in rat liver, we believe it codes for the heparan sulfate ND/NS. 3) When a library from MST cells was probed with the rat liver heparan sulfate NDNS cDNA, a high number of positive clones was found (23 out of 150,000), suggesting that they were derived from an abundant message. 4) The predicted protein encoded by the above 3.8-kb message is a type I1 membrane protein; the sequence has several portions which are identical to that of rat liver NDNS, but shows significant difference in the amino terminus, the region which has been shown in other Golgi membrane proteins to be necessary for targeting to this organelle (29). We do not know whether this may result in targeting of the above MST-derived enzyme to an organelle other than the Golgi apparatus, perhaps a different subcompartment. 5) The ratio ofN-deacetylase to N-sulfotransferase activity is 7-8-fold higher in cell extracts of COS-1 cells transfected with the MST cDNA compared with those transfected with the rat liver cDNA. A similar difference in enzymatic ratio was also found with the purified fusion protein from MST.
Although we hypothesize that the MST enzyme is responsible for the biosynthesis of heparin in uiuo, we have been unable to demonstrate this directly. This could be due to several reasons, the principal one being that the COS system is heterologous and may lack the specialized machinery necessary for the biosynthesis of heparin; the synthesis of this molecule may require additional cellular factors not available in COS cells which presumably exist in MST cells. Some evidence for this exists by the fact that overexpression of both the rat liver and this MST ND/NS into wild type cells does not result in increased sulfation, presumably as a result of unknown regulatory factors in these cells (34).
Further evidence that MST and the rat liver protein may have different roles comes from a comparative analysis of the kinetic parameters of the enzymes, in particular, the very large difference in ICm for N-sulfotransfer~e activity using deacetylated K5 polysaccharide as substrate. This difference must, however, be taken with some caution: it clearly indicates that both enzymes are different and suggests that they may be involved in the biosynthesis of different glycosaminoglycans such as heparin and heparan sulfate. Detailed mechanistic interpretations are difficult to make even though the substrate used for the N-sulfotransferase reaction, N-deacetylated K5 polysaccharide, appears to be more closely related to the ZR vivo substrate than N-deacetylated heparin or heparan sulfate. N-Deacetylated K5 polysaccharide does not have late biosynthetic modifications such as L-iduronic acid and 0-sulfated groups (31) which occur in the above two substrates. N-Deacetylated