Decorin Core Protein Secretion Is Regulated by N-Linked Oligosaccharide and Glycosaminoglycan Additions*

Expression of decorin using the vaccinia virus/T7 expression system resulted in secretion of two distinct glycoforms: a proteoglycan substituted with a single chondroitin sulfate chain and N-linked oligosaccharides and a core protein glycoform substituted with N-linked glycans but without a glycosaminoglycan chain. In this report, we have addressed two distinct questions. What is the rate-limiting step in glycosaminoglycan synthesis? Is glycosylation with either N-linked oligosaccharides or glycosaminoglycan required for secretion of decorin? N-terminal sequencing of the core protein glycoform, the addition of benzyl-β-d-xyloside, and a UDP-xylose: core protein β-d-xylosyltransferase activity assay show that xylosylation is a rate-limiting step in chondroitin sulfate biosynthesis. Decorin can be efficiently secreted with N-linked oligosaccharides alone or with a single chondroitin sulfate chain alone; however, there is severely impaired secretion of core protein devoid of any glycosylation. A decorin core protein mutant devoid of N-linked oligosaccharide attachment sites will not be secreted by Chinese hamster ovary cells deficient in xylosyltransferase or by parental Chinese hamster ovary wild type cells if the xylosyltransferase recognition sequence is disrupted. This finding suggests that quality control mechanisms sensitive to an absence of N-linked oligosaccharides can be abrogated by interaction of the core protein with the glycosaminoglycan synthetic machinery. We propose a model of regulation of decorin secretion that has several components, including appropriate substitution with N-linked oligosaccharides and factors involved in glycosaminoglycan synthesis.

The extracellular matrix is a complex array of secreted molecules, including collagens, proteoglycans, glycosaminoglycans (GAGs), 2 and a number of glycoproteins (1,2). Maintenance of the structural and functional organization of extracellular matrix components requires the correct targeting of proteins to their destinations. This is achieved by the delivery of newly synthesized glycoproteins and proteoglycans via the secretory pathway. Newly synthesized proteins in the secretory pathway encounter several important components such as recognition elements for nascent protein insertion into the ER, molecular surveil-lance chaperones, endogenous enzyme for post-translational modifications, and routing molecules to deliver the finished protein to the cell exterior (3). Mutant proteins with abnormal conformations frequently fail to pass through the secretory pathway. Improperly folded proteins are either retained in the ER lumen until they become properly folded or destined for degradation (4). The N-linked oligosaccharide moieties of these proteins not only promote folding directly by stabilizing polypeptide structures but also promote folding indirectly by serving as specific recognition structures that allow glycoproteins to interact with a variety of lectins, glycosidases, and glycosyltransferases (4,5) that in turn affect the folding and secretion of the protein.
The biosynthesis of GAGs, including chondroitin sulfate/dermatan sulfate and heparin/heparan sulfate is initiated by the sequential addition of four monosaccharides, xylose-galactose-galactose-glucuronic acid, onto targeted serine residues in the core protein. Subsequently, the sugar chains are extended by the addition of two alternating monosaccharides, either N-acetylgalactosamine/glucuronic acid in chondroitin sulfate/dermatan sulfate or N-acetylglucosamine/glucuronic acid in heparin/heparan sulfate on the linker tetrasaccharides. A number of biological functions have been demonstrated for different GAGs through the combination of biochemical and genetic approaches (6 -8).
The core protein of the chondroitin/dermatan sulfate proteoglycan decorin has three consensus sites for N-linked oligosaccharides and a GAG attachment site. We have previously detected the secretion of recombinant glycosylated decorin but not of a nonglycosylated form using the vaccinia virus expression system (9). Decorin core protein devoid of a GAG chain has been isolated (10,11), although it has yet to be demonstrated whether this is due to postsecretory cleavage of the GAG attachment domain or to synthesis and secretion of core protein that by-passes the GAG synthetic machinery. The proportion of secreted decorin core protein devoid of GAG chain is cell type-dependent and inversely proportional to the endogenous activity of UDP-xylose:core protein ␤-D-xylosyltransferase (XT), the enzyme that catalyzes the first sugar transfer reaction initiating chondroitin sulfate polymerization on a core protein substrate.
Previous studies using canine renal tubule (Madin-Darby canine kidney) cells have suggested that distinct sets of secretory proteins are released from their apical and basolateral poles and that the chondroitin sulfate chain plays a role for apical sorting and secretion of chondroitin sulfate proteoglycans (12,13). This observation may suggest that the GAG chains contain structural determinants for sorting and/or that glycosyltransferases involved in the GAG biosynthesis may have a quality control function on proteoglycan secretion.
Structures in the core protein may also affect sorting of the proteoglycan. For example, studies from different groups have shown that a globular domain, G3, in the core protein of aggrecan affects glycosaminoglycan modification and product secretion (14 -17).
In this report, we have investigated the role of glycosylation in decorin secretion using inhibitors of glycosylation and site-specific deletion mutagenesis combined with overexpression of decorin and decorin mutants in CHO cell lines deficient in specific glycosylation steps. We suggest that xylosylation of the core protein is a rate-limiting step in GAG synthesis. Xylosylation not only is a function of the activity of XT but also affected by the overall structure of the core protein substrate. We also present data that are consistent with a role for either the GAG chain itself or the enzymes involved in GAG synthesis in controlling secretion. We propose a model of regulation of decorin secretion that has several components, including appropriate N-linked oligosaccharide substitution and GAG synthesis.
Mutagenesis of Human Decorin cDNA-The construction of plasmid pDCN1, which drives the expression of full-length human decorin as a polyhistidine-tagged fusion protein, was described previously (9). This protein has one GAG attachment site at Ser 34 and three potential N-linked oligosaccharide attachment sites at Asn 211 , Asn 262 , and Asn 303 . A sequential deletion strategy, whereby each site was deleted in a separate cloning step, was employed to produce the constructs used in the current study. Mutations were generated by specific base changes in the decorin cDNA by the method of strand overlap extension PCR (18,19), utilizing synthetic oligonucleotide primers and Pfu DNA polymerase (Stratagene). All constructs were verified by DNA sequencing. Further verification was done by in vitro transcription directly from the plasmid vector with T7 RNA polymerase and translation with a reticulocyte lysate system (Promega) in the presence of 40 Ci of TRAN 35 S-label (MP Biomedicals), and reaction products were analyzed by SDS-PAGE. pDCN3 is a GAG attachment site deletion. A PCR was done using primer P5 (5Ј-GGGGCAT-GCTAGAAGATGAGGCTGCTGGG-3Ј) containing a SphI site and a Ser 34 3 Ala mutation, and P2 (5Ј-CGGAAACTATAAGTAAT-TCTCTGCAGCCGCGGGGA-3Ј) containing a SacII site. Using pDCN1 as a template DNA, the resulting PCR product was directly cloned into pT-cam1 at SphI and SacII sites. pDCN4 harbors a single amino acid substitution at Asn 211 3 Gln. The first PCR step used pDCN1 as template, with primers P2 and P3 (5Ј-GCATTGCTGAT-ACCCAGATCACCAGCATTCC-3Ј). The second PCR step used primers P1 (5Ј-CGGAATCCGAATTCGCCATCGAGGGTAGGG-GCATG-3Ј) and P4 (5Ј-GGAATGCTGGTGATCTGGGTATCAG-CAATGC-3Ј). The products of the first and second PCRs were combined, and a final PCR product was generated with primers P1 and P2. The PCR product was cloned into pT-cam1 at EcoRI and SacII sites. pDCN5 contains a double mutation, Ser 34 3 Ala and Asn 211 3 Gln. The first and second PCRs were identical to that described for pDCN4. The final PCR product was generated with primers P5 and P2. pDCN6 contains a GAG site mutation (Ser 34 3 Ala) and two N-linked glycosylation site mutations (Asn 211 3 Gln and Asn 262 3 Gln). pDCN5 was used as a template for strand overlap extension PCR amplification. The first PCR step was done with primers P6 (5Ј-CACGAATTCATCATCGAGGGTAGG-3Ј) and P9 (5Ј-CGTG-TTGGCCAGAGAGCCCTGGTCAACAGCAGAATGC-3Ј) and the second with primers P7 (5Ј-CCCGCGGCTGCAGAGAATTACTT-ATAG-3Ј) and P8 (5Ј-GCATCTCTGCTGTTGACCAGGGCTCT-CTGGCCAACACG-3Ј). A third PCR was performed using equal amounts of these products, with primers P6 and P7, and the product was cloned into pT-cam1 at EcoRI and PstI sites. pDCN7 contains two N-linked glycosylation site mutations (Asn 211 3 Gln and Asn 262 3 Gln). pDCN6 was digested with SacI and PstI, and the resulting 524-bp fragment containing two N-glycosylation site mutations was isolated. pDCN1 was digested with SacI and PstI, and the deleted 524-bp fragment was replaced with the same fragment from pDCN6 containing the two N-linked glycosylation site mutations. pDCN8 has mutations in all glycosylation acceptor sites, Ser 34 3 Ala, Asn 211 3 Gln, Asn 262 3 Gln, Asn 303 3 Gln. The first PCR step used pDCN6 as the template, with primers P6 and P11 (5Ј-GATCCAACTACAGAGATCTGGTTGT-TATGAAGGTGAC-3Ј). In a second PCR, pDCN6 was amplified with primers P10 (5Ј-GTCACCTTCATAACAACCAGATCTCTGTAG-TTGGATC-3Ј) and P7. These products were combined and a final product amplified with primers P6 and P7. The PCR product was cloned into pT-cam1 using EcoRI and PstI sites. pDCN9 contains three N-linked glycosylation site mutations, Asn 211 3 Gln, Asn 262 3 Gln, and Asn 303 3 Gln. The 524-bp fragment from pDCN1 was deleted using SacI and PstI digestion and replaced with the 524-bp fragment from pDCN8 harboring the three N-linked glycosylation site mutations. pDCN1, pDCN8, and pDCN9 were used to generate recombinant vaccinia viruses for the investigation.
Recombinant Vaccinia Viruses-All recombinant vaccinia viruses expressing both wild type human decorin and mutant decorins were generated by homologous recombination between plasmid constructs and wild type vaccinia virus, as described previously (20). The recombinant vaccinia virus, vTF7-3, the wild type vaccinia virus strain (WR), and vT7-CP, a recombinant vaccinia virus engineered to express T7 bacteriophage T7 RNA polymerase in CHO cell lines (21), were generously provided by Dr. Bernard Moss (NIAID, National Institutes of Health).
Recombinant Decorin Expression and Purification-Recombinant decorin glycoforms were expressed in HT 1080 cells using the vaccinia/T7 bacteriophage expression system. In brief, confluent monolayers of HT-1080 cells were co-infected with vTF7-3 (10 pfu/cell) and vDCN1 (10 pfu/cell). Virus was allowed to adsorb for 2 h at 37°C. At 8 h postinfection, the virus innoculum was replaced with serum-free DMEM. Medium was harvested at 30 h postinfection. To express wild type or mutant decorin, a confluent monolayer of CHO cells (CHO WT, pgsA-745, pgsB-618, pgsG-224, and pgsC-605) was co-infected with vT7-CP (10 pfu/cell) and vDCN1, vDCN8, and vDCN9 (10 pfu/cell). At 8 h postinfection, the virus innoculum was replaced with 100 Ci/well of TRAN 35 S-label in methionine-and cysteine-free DMEM. The culture medium was harvested at 30 h postinfection, and the recombinant decorin was purified. To examine the kinetic effects of glycosylation on secretion, HT 1080 cells were co-infected with vTF7-3 (10 pfu/cell) and vDCN1 and vDCN9 (10 pfu/cell). At 8 h postinfection, cells were starved for 1 h in methionine-and cysteine-free DMEM. Cells were pulsed with TRAN 35 S-label at 100 Ci/ml for 30 min in the same medium and chased in serum-free DMEM for the indicated times. The media and cell layer were harvested and extracted with 4 M guanidine HCl, 50 mM sodium acetate, pH 5.8, 0.5% Triton X-100 overnight at 4°C.
Purification of Recombinant Decorin-Conditioned medium was applied to Sephadex G-50 columns (Amersham Biosciences) equilibrated and eluted with 5 mM imidazole, 0.5 M NaCl, 20 mM Tris-HCl, pH 8.0, 0.2% CHAPS to separate macromolecules from unincorporated radioactive precursors. The eluted macromolecular fraction was applied to a 1-ml column of iminodiacetic acid immobilized on Sepharose 6B (Sigma) that had been equilibrated with nickel chloride. After sample application, the column was washed with 5 column volumes of 30 mM imidazole, 0.5 M NaCl, 20 mM Tris-HCl, pH 8.0, 0.2% CHAPS, and bound material was eluted with 150 mM imidazole, 0.5 M NaCl, 20 mM Tris-HCl, pH 8.0, 0.2% CHAPS. Fractions were monitored by SDS-PAGE and Western blot. Fractions containing decorin were pooled and applied to Ultrafree-15 centrifugal filter devices (membrane size: Biomax 10; Millipore Corp.) to concentrate the sample, which was then applied to PD-10 columns (Amersham Biosciences) to exchange the solvent to phosphate-buffered saline, 0.2% CHAPS. Protein concentrations were determined by the molar extinction coefficient (22).
Expression of Decorin and Decorin Fragments in Escherichia coli-The cloning, expression, and purification of full-length decorin as an MBP fusion protein (MBP-decorin), and of a decorin N-terminal peptide as a GST fusion protein (GST-MD4) and the production of the isolated peptide (MD4) have been described previously (23).
SDS-PAGE-7.5% polyacrylamide gels were prepared in the buffer system of Laemmli (24), with a 3.5% polyacrylamide stacking gel. Gels were fixed in 40% (v/v) methanol containing 10% (v/v) acetic acid, and radioactive components were detected by fluorography using AMPLIFY (Amersham Biosciences) according to the manufacturer's instructions.
Tunicamycin Treatment-The addition of N-linked oligosaccharides was inhibited by treatment of cultures with 5 g/ml tunicamycin (Calbiochem), an inhibitor of N-linked glycosylation. Tunicamycin was dissolved in DMSO at high concentration, such that the final DMSO concentration in cultures did not exceed 0.5% (v/v). An equivalent concentration of DMSO was included in control cultures. CHO WT and pgsA-745 cells were co-infected with vT7-CP (10 pfu/cell) and vDCN1 (10 pfu/cell). At 2 h postinfection, either culture medium alone or medium containing 5 g/ml tunicamycin was added to each well. At 6 h postinfection, medium was replaced with 100 Ci/ml TRAN 35 Slabel (MP Biomedicals) with or without 5 g/ml tunicamycin. At 10 h postinfection, the medium and cell layer were harvested and extracted with 4 M guanidine HCl, 50 mM sodium acetate, pH 5.8, 0.5% Triton X-100 overnight at 4°C. The macromolecular fraction was separated from unincorporated radioactive precursors by chromatography on Sephadex G-50 columns equilibrated and eluted with 5 mM imidazole, 0.5 M NaCl, 20 mM Tris-HCl, pH 8.0, 0.2% CHAPS. The eluted macromolecular fraction was applied to a column of iminodiacetic acid immobilized on Sepharose 6B that had been equilibrated with nickel chloride. After sample application, the column was washed with the application buffer supplemented with 30 mM imidazole, and bound material was eluted with 250 mM imidazole in the same buffer.
N-terminal Sequencing of Recombinant Decorin Core Protein-An aliquot of purified recombinant decorin core protein was subjected to 7.5% SDS-PAGE and eletrotransferred to polyvinylidine difluoride membrane (Millipore Corp.). The core protein was visualized by Coomassie Blue staining, and the band was excised and subjected to N-terminal amino acid sequencing (Protein Chemistry Core Facility, Baylor College of Medicine, Houston, TX).
Determination of XT Activity-XT activity assays were performed following a protocol previously described with some modifications (25,26). In brief, HT1080 and pgsA-745 cells were cultured in medium supplemented with 10% fetal bovine serum in a humidified atmosphere of 5% CO 2 at 37°C until confluent. Cells were collected and resuspended in a sucrose buffer (0.25 M sucrose, 20 mM Tris-HCl, pH 7.4, 0.5 mg/ml pepstatin, 1 mg/ml leupeptin, 20 mM phenylmethylsulfonyl fluoride, 10 mM sodium azide). Cells were pelleted by centrifugation and resuspended in 0.5 ml of reaction buffer (50 mM MES buffer, pH 6.5, 50 mM potassium chloride, 10 mM magnesium chloride, 5 mM manganese chloride, 0.3% Triton X-100). Cells were extracted by agitation on ice for 10 min, and the protein concentrations of the extracts were determined by the BCA assay (Pierce) using the standard microtiter plate protocol. The conditions for the standard XT activity assay were determined by incubating 0.1 Ci of UDP-[ 14 C]xylose (PerkinElmer Life Sciences), 15 g of decorin core protein, and 500 g of tissue extract in a total volume of 100 l of reaction buffer, varying the temperature, time, and pH. In all activity assays, after incubation at 25°C for 1 h, the reaction mixture was placed on a glass microfiber filter membrane (GF/F; Whatman International Ltd.) with 100 l of reaction buffer and 100 l of 10% trichloroacetic acid and then allowed to dry completely at room temperature. The membranes were washed for 15 min with 10% trichloroacetic acid and then washed twice for 15 min with 5% trichloroacetic acid and dried at room temperature. Incorporated radioactivity was determined by liquid scintillation counting. To examine xylose substitution of decorin core proteins, 0.3 Ci of UDP-[ 14 C]xylose and 20 g of decorin core protein of DCN1 and DCN9 from different CHO mutant cells were added to the 300-l cell extracts containing XT activity. UDP-[ 14 C]xylose was added to the decorin core proteins without the cell extracts under the same conditions as a control. Cell extracts alone or 0.3 Ci of UDP-[ 14 C]xylose alone were used as controls. All samples were incubated at 25°C for 1 h, analyzed on SDS-PAGE, and visualized by fluorography.
RNA Isolation and Northern Blot Detection-CHO cells (CHO WT, pgsA-745, pgsB-618, pgsG-224, and pgsC-605) were co-infected with vT7-CP (10 pfu/cell) alone or in combination with recombinant viruses (10 pfu/cell). Total RNA was extracted at 24 h postinfection using TRIZOL reagent (Invitrogen), electrophoresed on a 1% (v/v) agarose gel, and transferred to a nylon membrane. Northern blot analysis was performed using a Northern blotting kit (Ambion). A 442-bp HincII fragment from the human decorin cDNA and 1076-bp mouse ␤-actin DNA were radiolabeled with [␣-32 P]dCTP (PerkinElmer Life Sciences) using a random primer labeling kit (Invitrogen) and used to probe membranes. Hybridization was detected by exposure to x-ray film overnight.

Overexpression of Recombinant Decorin Glycoforms and N-terminal
Sequencing-We have previously shown the expression of two distinct glycoforms of recombinant decorin by HT1080 cells using the vaccinia virus/T7 bacteriophage expression system (9). One glycoform is presented as a proteoglycan substituted with a single chondroitin sulfate chain and N-linked oligosaccharides. The second form was composed of a core protein substituted with N-linked glycans but devoid of chondroitin sulfate. The absence of detectable GAG on up to 40% of the secreted decorin could be due to several possibilities, including proteolytic cleavage of the N-terminal GAG attachment domain or lack of substitution of the Ser-Gly dipeptide attachment site. Cleavage of an N-terminal peptide would have resulted in a subtle change in apparent molecular weight of the core protein that would be difficult to discern on SDS-PAGE. We now have examined these possibilities by N-terminal amino acid sequencing of the two core protein glycoforms. Confluent HT1080 cells were infected with vTF7-3 and vDCN1 (encoding full-length human decorin) and incubated in the presence of TRAN 35 Slabel for 30 h. The secreted decorin glycoforms were purified, as described under "Experimental Procedures," and fractions enriched in decorin core protein glycoform were pooled (not shown). The decorin core protein pool was further purified by a second metal-chelating chromatography step that included extensive washing with solvent containing 80 mM imidazole to prevent binding of the proteoglycan glycoform, followed by elution of the bound core protein with 250 mM imidazole (Fig. 1a). The eluted decorin core protein migrated on 7.5% SDS-PAGE as a doublet of 45-48 kDa (Fig. 1b) that represents core protein substituted with 2-and 3-N-linked oligosaccharides, respectively (9). Five micrograms of purified decorin core protein was subjected to 7.5% SDS-PAGE and electrotransferred to polyvinylidine difluoride membrane. The recombinant decorin core protein doublet was visualized by Coomassie Blue staining (Fig. 1b), and each band of the doublet core protein was excised and subjected to N-terminal amino acid sequencing. The amino acid sequences of both core protein glycoforms were identical and matched exactly with the predicted sequence for intact recombinant protein (Fig. 1c), demonstrating that the N terminus had not been cleaved. Furthermore, the efficient sequencing through the Ser-Gly GAG attachment sequence indicates that the serine residue had not been xylosylated. This observation suggests that the lack of a GAG chain was due to the absence of GAG initiation at this site. In contrast, sequencing through the chondroitinase ABC-treated decorin proteoglycan yielded a blank cycle at the GAG attachment serine (data not shown).
GAG Synthesis in the Presence and Absence of Benzyl-␤-D-xyloside-Differential glycosylation of recombinant decorin may be due to the overexpression of core protein resulting in levels that exceed the capacity of the enzymes involved in initiation or polymerization of the GAG chain. Studies using ␤-D-xylosides as alternative acceptors for the assembly of GAG chains suggest that the enzymes involved in chain elongation are present in excess in the cell, and under normal circumstances, it is the availability of core protein that is rate-limiting (27)(28)(29)(30). Benzyl-␤-D-xyloside was used to assess the levels of the enzymes responsible for GAG polymerization in cells overexpressing recom-  35 Slabeled core protein with 250 mM imidazole. Recombinant proteoglycan and other contaminating labeled proteins were removed by extensive washing with solvent containing 80 mM imidazole. b, an aliquot of the core protein pool (5 g) from a was subjected to 7.5% SDS-PAGE, electrotransferred to a polyvinylidene difluoride membrane, and stained with Coomassie Blue. The arrowheads indicate the decorin core protein doublet, representing glycoforms containing two and three N-linked oligosaccharides. c, decorin core protein bands from b were excised and subjected to N-terminal amino acid sequencing, and both bands yielded the identical sequence. Decorin is expressed as a hexahistidine fusion protein His 6 , containing a Factor Xa cleavage sequence, upstream of the mature human decorin sequence (dashed arrow). The shaded box indicates the GAG attachment site.  binant decorin. HT1080 cells uninfected, infected with vTF7-3 alone, or co-infected with vTF7-3 and vDCN1 were used in these experiments. At 8 h postinfection, culture medium was replaced with sulfate-depleted medium containing [ 35 S]sulfate with or without 1 mM benzyl-␤-D-xyloside. The medium and cell layers were harvested and extracted at 24 h postinfection, and the macromolecular fractions were isolated as described under "Experimental Procedures." In the absence of benzyl-␤-D-xyloside (Fig. 2, open bars and  diagonally hatched bars), the total incorporation of [ 35 S]sulfate into GAGs was reduced in vaccinia virus-infected cells (bars 3, 5, 9, and 11) compared with uninfected cells (bars 1 and 7), which is probably due to virus-induced pathogenic effect. However, the addition of the ␤-D-xyloside (Fig. 2, solid bars and speckled bars) stimulated sulfate incorporation to similar levels in all cultures. In macromolecular fractions from the medium, the sulfate incorporation was increased 5-8-fold with ␤-D-xyloside compared with that of control cultures. This finding indicates that, first, the enzymes downstream of XT are present in excess and are capable of higher synthetic rates of GAG synthesis in the presence of excess substrate (in this case, ␤-D-xyloside) and, second, that the total GAG synthetic capacity is similar with or without overexpression of decorin core protein.
Substrate Requirements of XT-Based on the data presented above, it is likely that xylosylation is a critical rate-limiting step in the synthesis of decorin proteoglycan and that the secreted GAG-free decorin core protein is not xylosylated. Failure to transfer a xylose to the conserved serine residue may be due to incorrectly folded core protein that is unable to act as a substrate for XT. This possibility was investigated by using decorin core protein as a substrate in an XT assay. As described under "Experimental Procedures," recombinant decorin core protein was expressed in the XT-deficient CHO cell line, pgsA-745, to generate native decorin that had no modification at the Ser-Gly GAG attachment site (in separate experiments, not shown, this material appears identical to the GAG-deficient core protein from HT1080 cells shown in Fig. 1). Recombinant decorin core protein and core protein fragments were expressed as fusion proteins in E. coli, as previously described (23). Control peptides of MBP and GST alone were also purified. The different constructs used as substrates in the XT assay are shown schematically in Fig. 3a. Optimal conditions for the enzyme assay were determined before testing the decorin from different sources. 500 g of cell extracts, 10 g of decorin core protein from the CHO pgsA-745 cell line, and 0.05 Ci of UDP-[ 14 C]xylose were used to determine optimal pH, temperature, and incubation time for the activity of XT with respect to the decorin core protein acceptor (Fig. 3b). The standard enzyme assay was performed at pH 6.5 for 1 h at 25°C using a crude cell extract from HT1080 cells (speckled bars) as a source of XT activity, with controls being either no extract (open bars) or an extract of an identical number of cells from the pgsA-745 XT-deficient cells (solid bars) (Fig. 3c). For all substrates tested, the activity measured for the pgsA-745 extract was approximately the same and represented background. In the absence of an exogenous substrate (lane 3), there is measurable XT activity that represents transfer of xylose to endogenous core protein acceptor present in the crude extract. This activity could be reduced to background levels by preincubating cells in the presence of cycloheximide for 2 h prior to extraction (data not shown) and therefore represents the true background in the assay (dashed line). The XT activity measured with all of the E. coli expressed proteins (lanes 6 -15) was consistently low, with MBP-DCN essentially the same as MBP alone, and the MD4 peptide showing some activity (ϳ15%) above background. The native decorin core protein expressed by mammalian cells was a far more efficient acceptor of xylose transfer (lane 5), indicating that the XT activity is determined by factors other than primary sequence of the targeted substrate.
Expression of Decorin by Tunicamycin-treated Cultures-To explore the role of glycosylation in decorin secretion, cultures of CHO WT cells were incubated in the presence of tunicamycin, an antibiotic that blocks the first step in the pathway of N-linked oligosaccharide addition. Furthermore, to investigate the influence of N-linked oligosaccharide addition in the absence of GAG synthesis, we expressed decorin in a mutant CHO cell line, pgsA-745, deficient in XT activity (31). The level of decorin proteoglycan secreted by tunicamycin-treated parental CHO cells (CHO WT) is reduced relative to control by about 40% (Fig. 4, lanes  1 and 2). However, most noticeable secretion of the GAG-deficient core protein is almost entirely abolished. A small amount of nonglycosylated core protein at ϳ38 kDa (lane 2, arrowhead) can be detected. In the XT-deficient cells (pgsA-745; lanes 3 and 4), a similar amount of decorin core protein was secreted (lane 3) as in the parental cell line, and this was reduced to less than 10% in the presence of tunicamycin (lane 4, arrowhead). Expression of decorin glycoforms in the cell layer fractions of these cultures (lanes [5][6][7][8] showed that accumulation of core protein was similar across all treatments and cell lines. However, in the presence of tunicamycin, the core protein has a somewhat smaller size, consistent with a lack of N-linked oligosaccharides. Taken together, these results suggest that decorin can be effectively secreted either as a proteoglycan (PG) form without N-linked oligosaccharides (lane 2) or as a core protein form with N-linked glycan substitution (lane 3), but in the absence of GAG and N-linked glycans there is significant impairment of secretion (lane 4), and suggest that the nonglycosylated core protein accumulates in the cell.
Expression and Secretion of Mutant Decorin Proteins-To further examine the role of carbohydrate substitutions in decorin biosynthesis and secretion, we made a panel of decorin constructs with specific mutations. Asn 211 , Asn 262 , and Asn 303 were mutated to Gln residues to prevent N-linked glycosylation with minimal disruption of the protein secondary structure, to yield pDCN9 (Fig. 5a). The chondroitin sulfate attachment sequence was disrupted by creating a Ser 34 3 Ala mutation in pDCN9, to yield pDCN8 devoid of any carbohydrate attachment sites. These constructs were used to generate the recombinant vaccinia virus vDCN1, vDCN8, and vDCN9. The viruses were used to infect CHO WT cells, as described under "Experimental Procedures," and the secreted products were harvested and purified. As shown previously, DCN1 yielded a PG form and a doublet core protein (Fig. 5b, lane 1). DCN9 yielded a similar amount of a PG form (about 95% of DCN1 (Fig.  5, b (lane 3) and c (bar graph)) but significantly reduced secretion of core protein (asterisk). The core protein secreted by DCN9 appeared as one band with a size similar to that predicted for a nonglycosylated form of decorin. DCN8 showed a low level expression of a nonglycosylated core protein but no proteoglycan or N-glycosylated core protein (Fig. 5b,  lane 2). These data indicate that decorin can be secreted at levels comparable with wild-type decorin if substituted with a GAG chain, even in the absence of N-linked glycosylation. However, there is still a population that bypasses xylosylation and subsequent chondroitin sulfate chain elongation, but the proportion of nonglycosylated core protein present in the medium is less than 20% of core protein present in control cells. Taken together with the data in Fig. 5, where DCN1 expressed in the XT-deficient cell line yields N-glycosylated core protein secretion equivalent to that in wild-type cells, it appears that decorin can be secreted as a PG, with chondroitin sulfate and N-linked oligosaccharides or with chondroitin sulfate alone, or as a core protein, with N-linked oligosaccharides alone. However, in the absence of any carbohydrate substitution, secretion is severely impaired and may simply represent "leakage" of material targeted for intracellular degradation.
To compare the secretion rate of DCN1 (wild type) and DCN9 (N-glycosylation deficient) HT1080 cells were co-infected with vTF7-3 (10 pfu/cell) and vDCN1 and vDCN9 (10 pfu/cell) and then pulse-chase experiments were performed as described under "Experimental Procedures." The secretion rate of DCN9, a proteoglycan substituted with a single chondroitin sulfate chain but devoid of N-linked oligosaccharides, appears to be similar to that of DCN1, a proteoglycan substituted with a single chondroitin sulfate chain and N-linked oligosaccharides FIGURE 5. Site-directed mutagenesis of human decorin cDNA and overexpression in CHO WT cells. a, glycosylation attachment site mutants were generated by creating single amino acid changes in the decorin sequence using strand overlap extension PCR. The wild type decorin (pDCN1) clone has one GAG attachment site at Ser 34 and three potential N-linked oligosaccharide attachment sites at Asn 211 , Asn 262 , and Asn 303 . The pDCN8 contains four mutations, indicated by the crosses, and pDCN9 contains three mutations. b, CHO WT cells were co-infected with vT7-CP and vDCN1, vDCN8, and vDCN9, and at 6 h postinfection, the virus innoculum was replaced with TRAN 35 S-label, and incubation continued for 24 h. Secreted decorin fusion proteins were resolved by 7.5% SDS-PAGE and visualized by fluorography. The square bracket indicates the position of the intact decorin proteoglycan, double lines indicate the doublet core protein band, and the asterisk shows the position of the nonglycosylated core protein band. c, the PG smears were quantitated by excision from the gel and direct counting, and the data are shown in the bar graph, with the counts associated with DCN1 set at 100%. (Fig. 6a). PG secretions of DCN1 and DCN9 were compared by direct counting radioactivity of excised PG bands (Fig. 6b).
Expression of Decorin by Glycosyltransferase-deficient CHO Cells -To determine the stage in GAG biosynthesis required for secretion of the decorin core protein, we studied overexpression of DCN1, DCN8, and DCN9 in several CHO cell lines deficient in specific glycosyltransferases. pgsA-745 has no XT activity (31), pgsB-618 is defective in galactosyltransferase-I activity (32), and pgsG-224 is deficient in glucuronosyltransferase-I activity (33). pgsC-605 is a sulfate transporter-defective mutant (34) that affects the final step of sulfation of the chondroitin sulfate chain. The decorin glycoforms purified from conditioned media of CHO mutants after overexpression were resolved by SDS-PAGE (Fig. 7a). A PG glycoform was secreted by CHO WT parental cells and by the pgsC-605 (sulfate-deficient) mutant cells when expressing DCN1 (wild type) and DCN9 (N-glycosylation-deficient) but not DCN8 (GAG-deficient) proteins (Fig. 7, a and c). However, expression of the core protein glycoform was significantly reduced in all cell lines for both DCN8 (glycosylation-deficient) and DCN9. The expression of core protein glycoforms (both glycosylated doublet and nonglycosylated form) was quantitated by excision of the bands and direct counting of radioactivity (Fig. 7b). Compared with expression of the decorin core protein by DCN1 (arbitrarily set to 100% in the different cell lines), expression of core protein devoid of a chondroitin sulfate chain and N-linked oligosaccharides was less than 25% in all cases. The addition of a chondroitin sulfate chain (by CHO WT or pgsC-605 cells) was sufficient to rescue control levels of decorin PG (Fig. 7, a and c); however, the presence of XT alone (pgsB-618) or XT, galactosyltransferase-I, and galactosyltransferase-II (pgsG-224) was not sufficient to rescue the product of DCN9 core protein expression. Thus, despite having an intact xylose acceptor site, core protein secretion is reduced by more than 70%. The possibility that the mutant constructs were not transcribed with equivalent efficiency and thus caused decreased translation of the mutant proteins was tested by Northern blot analysis. Fig. 7d shows decorin mRNA expression in all CHO cell lines after infection with the wild-type and mutant decorins. There was no detectable decorin mRNA prior to infection (Fig. 7d, T7 lanes), and after infection, decorin mRNA levels were essentially identical across all cells and all constructs.
In assessing the requirements for transit of an N-linked oligosaccharide-deficient core protein through the secretory pathway in the CHO mutant cell lines, we assumed that the product of expression in pgsA-745 cells had no substitution at Ser 34 , and the product of the pgsB-618 and pgsG-224 was substituted with xylose or xylose-galactose-galactose, respectively. This possibility was examined by using the core protein products in an XT activity assay. We have shown that the wild-type decorin core protein (DCN1) secreted by pgsA-745 cells purified under native conditions is an excellent substrate for XT activity. Therefore, this assay can be used to test whether the core protein has a site available  DECEMBER 30, 2005 • VOLUME 280 • NUMBER 52 for transfer of xylose from a UDP-xylose donor. The core protein purified from the conditioned media of CHO WT, pgsA-745, pgsB-618, and pgsG-224 cells after overexpression of DCN1 (wild-type) and DCN9 (N-glycan-deficient) were used in the standard XT assay, as described under "Experimental Procedures." The products of the reaction to transfer [ 14 C]xylose from UDP-[ 14 C]xylose to core protein substrates were visualized by SDS-PAGE and fluorography (Fig. 8). DCN1 expressed in pgsA-745 cells was the most efficient substrate, as expected, given that 100% should be available for xylose transfer (Fig. 8,  lane 2). Nevertheless, a small amount of DCN1 expressed in CHO WT, pgsB-618, and pgsG-224 cells was also able to accept xylose, indicating that a proportion of the doublet core protein bypasses xylosylation in the cell but is effectively secreted by virtue of N-linked glycosylation. The DCN9 (N-linked glycosylation-deficient) core protein, which represents less than 20% of the wild-type core protein expression, was also an effective substrate when synthesized in pgsA-745 cells (Fig. 8, lane 6). However, DCN9 core protein from pgsB-618 and pgsG-224 cells showed no detectable transfer of xylose (Fig. 8, lanes 7 and 8), indicating that the substrate site was substituted at a minimum with xylose. This shows that xylose or xylose-galactose-galactose substitution alone is not b, the core protein bands indicated in a were excised from the gel and resolubilized, and radioactivity was quantitated relative to DCN1 (100%) by liquid scintillation counting. c, the PG protein bands from pgsC-605 indicated in a were excised from the gel and resolubilized, and radioactivity was quantitated relative to DCN1 (100%) by liquid scintillation counting. d, total RNA samples were extracted from CHO cells (CHO WT, pgsA-745, pgsB-618, pgsG-224, and pgsC-605), which were coinfected with vT7-CP and recombinant decorin viruses (as described in a). RNA was electrophoresed on a 1% agarose gel and transferred to a nylon membrane. Northern blot analysis was performed, and blots were probed for decorin and ␤-actin expression. sufficient to rescue the N-linked oligosaccharide-deficient core protein from intracellular retention. Taken together, we conclude that decorin secretion is regulated by N-linked oligosaccharides and GAG chain addition.

DISCUSSION
In previous studies, we have shown that overexpression of decorin and the structurally closely related proteoglycan biglycan in mammalian cells lead to the secretion of a proteoglycan form and a core protein glycoform (9,35). Furthermore, analysis of the chondroitin sulfate chains on the recombinant proteoglycan indicates that these all have a molecular mass of ϳ30 kDa, with no evidence of smaller or partial GAG chains. Therefore, it appears that once a commitment to synthesis of a GAG chain is made, this goes to completion. These observations suggest that subsequent to chain initiation, subsequent glycosyltransferases and associated substrates are available in excess. As the core protein transits the endoplasmic reticulum, it can interact with XT and has a complete GAG added, or it bypasses the XT and is secreted as a core protein with N-linked oligosaccharides but without a GAG chain.
In the current study, we have examined possible mechanisms for the secretion of a GAG-deficient core protein. Several lines of indirect evidence suggest that xylosylation of appropriate serine residues represents a critical control point of chondroitin sulfate proteoglycan biosynthesis (36,37). Furthermore, in vitro studies combined with surveying previous literature have led to designation of a putative consensus sequence for xylosylation of core protein that comprises a minimum of Ser-Gly-Xaa-Gly (38 -40). However, many proteins that contain putative GAG initiation sites escape xylosylation and are secreted as simpler glycoproteins, suggesting that there are requirements beyond the primary amino acid sequence in the vicinity of the xylosylation site that must be met before the protein is recognized by the XT in a productive manner. Thus, the substrate requirements for effective xylosylation are complex.
We have induced the secretion of GAG-deficient decorin core protein in cultured mammalian cells using the vaccinia virus/T7 phage system. In previous studies, we and others have suggested that the secondary structure of decorin is not only complex but critical to its function in many assays (9,35,(41)(42)(43)(44). Therefore, under circumstances of high level expression, it was possible that the GAG-free decorin represented a population of misfolded core that could not be recognized by XT. We investigated this possibility by establishing an assay of XT activity, utilizing XT present in crude cell extracts of HT1080 cells. Other investigators have utilized a variety of putative substrates to measure XT activity, including chemically deglycosylated core protein (45)(46)(47), synthetic peptides (25,48,49), silk (26,50,51), and recombinant bikunin (52). Our result showed that the recombinant decorin core protein isolated from mammalian cells (either HT1080 or the pgsA-745 cells) under native conditions was a significantly better substrate for XT than decorin produced by bacteria or a simple peptide spanning the GAG attachment domain. This showed that the decorin core protein glycoform was a good substrate for XT, and therefore its lack of a GAG chain was not a consequence of poor substrate recognition. Furthermore, these experiments showed that the requirements for effective xylosylation are more complex than that provided by a simple primary amino acid sequence. We showed activity with the decorin peptide (MD4), consistent with other investigators' observations that peptides will act as acceptors for xylose transfer; however, the activity seen with full-length native decorin was 5-10-fold higher.
Our studies demonstrate that xylosylation is a critical rate-limiting step in GAG biosynthesis, and the conformational information contained in the core protein is important for effective xylosylation. Both XT activity and UDP-xylose formation could be rate-limiting steps in xylosylation. Recently, the XT has been cloned from different mammals (53). XT shows two distinguishable features compared with other glycosyltransferases involved in GAG biosynthesis. The purified XT appears as a 120-kDa protein on SDS-PAGE (54), which is much larger than other glycosyltransferases (40 -70 kDa). Most of the XT activity was detected in the extracellular matrix or in the medium of cell cultures, although the XT is a transmembrane enzyme (53,55). The authors argue that XT has further regulatory functions in the GAG biosynthesis, and it has been previously suggested that the soluble XT can interact directly with membrane-bound galactosyltransferase-1, which in turn may be part of a membrane-bound enzyme complex (36).
Comparing and contrasting the fate of distinct decorin glycoforms in this study has led to the identification of a previously unrecognized regulatory role for glycosaminoglycan synthesis in proteoglycan secretion. Although the mechanism of this regulatory step remains unknown, selective use of CHO cell mutants and mutagenesis points to a role for the galactosaminoglycan chain and/or glycosyltransferases involved in chondroitin sulfate synthesis. The data presented in the current study may indicate that galactosaminoglycans and/or glycosyltransferases participating in chondroitin chain biosynthesis perform a chaperone-like function. In the presence of tunicamycin, earlier studies show secretion of decorin by human fibroblast or chondrocytes (56,57). In addition, undersulfated proteoglycans were secreted by cultured chondrocytes in the presence of either monesin, an inhibitor of protein secretion, or brefeldin A, an inhibitor of vesicular transport and secretion of macromolecules (58,59). These data suggested that proteoglycan can be secreted and aided by certain GAG and/or enzyme involved in GAG biosynthesis. Studies using Madin-Darby canine kidney cells have shown that the chondroitin sulfate proteoglycans were secreted apically, whereas most of the heparan sulfate proteoglycans were secreted into the basolateral medium (12,13). Protein-free chondroitin sulfate chains on hexyl-␤-D-thioxyloside were predominantly secreted to the apical medium in the results, indicating that the sorting information is localized to or associated with the chondroitin sulfate chain.
In general, glycoproteins that are destined for secretion or transit to the cell surface undergo surveillance by molecular chaperones in the lumen of the ER. This quality control mechanism ensures that globular proteins have achieved proper conformation prior to their progression Decorin core protein expressed from vDCN1 and vDCN9 in CHO cell lines, as indicated, was purified and used as substrate in a xylosyltransferase reaction. Transfer from UDP-[ 14 C]xylose to the core protein substrate was visualized by SDS-PAGE and fluorography. An equal amount of each core protein preparation was used in each reaction. DCN1 yields a doublet core protein (double lines), and DCN9 yields a nonglycosylated core protein band (asterisk).
in the routing process from the ER to the Golgi stacks. In recent years, a picture has emerged that describes how N-linked glycosylation, in combination with several independently acting enzymes, facilitates glycoprotein folding (4,5). For instance, in the absence of conformational maturation, UDP-glucose:glycoprotein glucosyltransferase functions as a folding sensor (60) that recognizes structural determinants common to nonnative glycoprotein structure (61). Protein-disulfide isomerase is another ER-based soluble enzyme that also appears to act as a chaperone (62)(63)(64), and lends support to the notion that many enzymes found in the ER may serve dual functions as catalysts and quality control sensors. As a rule, failure to attain conformational maturation following biosynthesis results in the selective elimination of misfolded polypeptides and unassembled protein complexes by a relatively stringent mechanism of conformation-based quality control.
To date, there is no direct evidence that any of the GAG and glycosyltransferases involved in GAG synthesis have chaperone or quality control function. We have shown that in the absence of N-linked glycosylation (via tunicamycin inhibition or deletion mutagenesis of critical Asn residues) and absence of XT activity (as in the CHO pgsA-745 cells) or XT acceptor site (Ser 34 3 Ala), there is a significantly reduced (Ͻ25%) secretion of core protein. Whether the core protein that does appear in the culture medium represents controlled secretion or is a consequence of virus-induced cell lysis remains unclear at present. However, the substantial retention and apparent intracellular degradation of the bulk of glycosylation-deficient core protein points to a significant role for both N-linked glycosylation and serine-linked GAG addition in regulating core protein transit and secretion. This may imply a dual and somewhat redundant mechanism of surveillance for correctly processed proteoglycan by systems sensitive to N-linked glycosylation and GAG addition, respectively. The mechanism by which the addition of a GAG chain is able to rescue an N-linked oligosaccharidedeficient core protein otherwise apparently destined for degradation is unknown. An understanding of the molecule bases of this activity will no doubt provide insight into the role of the GAG chain and glycosyltransferases in GAG biosynthesis.