Divergent secretory behavior of the opposite ends of aggrecan.

The proteoglycan, aggrecan has a globular domain, G1, at the N terminus and a different globular domain, G3, at the C terminus. Aggrecan produced by mutant nanomelic chickens is truncated due to a premature stop codon and consequently lacks G3 and a minor portion of its chondroitin sulfate domain (Li, H., Schwartz, N. B., and Vertel, B. M. (1993) J. Biol. Chem. 268, 23504-23511). The mutant protein is retained in the endoplasmic reticulum and fails to enter the Golgi stacks (Vertel, B. M., Walters, L. M., Grier, B., Maine, N., and Goetinck, P. F. (1993) J. Cell Sci. 104, 939-948). The homozygous mutant is lethal because of failure of chondrogenesis and osteogenesis, while the heterozygous mutant is dwarfed. To further elucidate the pathogenetic mechanisms underlying nanomelia and to determine if G1 and G3 are themselves secreted, we expressed them in transfected host cells. Expression was performed in wild type Chinese hamster ovary (CHO) cells and in mutant CHO cells which are unable to link glycosaminoglycan (GAG) chains to core proteins. We compared: (a) secretion of expressed G1 and G3 constructs containing contiguous GAG chain consensus sites and (b) GAG chain modification of the secreted proteins. We find that: 1) G3 is 24-100 times more rapidly secreted than G1; 2) secreted G3 contains contiguous chondroitin sulfate GAG chains, while secreted G1 lacks contiguous GAG chains; 3) G3 secretion is not coupled to xylosylation of contiguous GAG chain consensus sites. These results imply that G1 and G3 intrinsically differ in passage through the cell secretory route.

post-translational modifications, and (d) routing procedures to deliver the finished protein to the cell exterior.
Mutant proteins whose normal forms are ordinarily destined for cell export often fail to navigate the secretory pathway and are retained within cells (2). One proposed explanation for retention is that mutant proteins have abnormal conformations and fail to pass surveillance by molecular chaperones (3). The chaperones primarily reside in the ER lumen, where they are gatekeepers for secretory protein entry into the Golgi stacks. Such entry requires proper conformation of the globular domains of a core protein. Improperly folded proteins are either retained in the ER lumen until they become properly folded or they are degraded.
Aggrecan contains three globular domains, G1, G2, and G3, with contiguous consensus sites between G2 and G3 for attachment of keratan sulfate and chondroitin sulfate chains (4). Aggrecan is normally secreted into the extracellular matrix, where it is responsible for tissue resiliency. Its presence is vital for normal chondrogenesis and osteogenesis. G3 is absent in nanomelia due to a premature stop codon (5,6); the truncated aggrecan is not secreted and accumulates in the ER lumen (7). The homozygous mutation is lethal, while the heterozygote survives as a dwarf animal.
Members of a proteoglycan family comprised of aggrecan, versican, neurocan, and brevican share homologous G1 and G3 domains (8). G3 has three subdomains one of which, a C-lectin, is highly conserved and occurs in a wide variety of secreted and cell surface proteins (9). G3 is often absent from secreted aggrecan (10), suggesting that G3 may be dispensable in the extracellular space. In contrast, G1 is retained in secreted aggrecan and has the important role of binding to hyaluronan, producing an extracellular polymeric network (11,12). The present study was designed to focus on the behavior of G1 and G3 when they have been introduced into the secretory pathway of host cells, which normally produce proteoglycans. The results imply that G1 and G3 behave very differently from each other in the secretory pathway, suggesting that they may have separate roles in intracellular routing of aggrecan.

EXPERIMENTAL PROCEDURES
Construct Cloning-1) The G3 construct contains the endogenous avian aggrecan signal peptide (SP), endogenous avian glycosaminoglycan consensus sites (GAG), endogenous avian C-terminal G3 domain (G3) (13,14) and a 6-histidine (His 6 ) tag, abbreviated as SP-GAG-G3-His 6 . SP was amplified by polymerase chain reaction (PCR) using oligonucleotides 101 (5Ј-GACGTCAAGCTTATGACCACTCTACTA-3Ј, containing a HindIII site) and 102 (5Ј-CTGCAGTCTAGAT-GAGAGCTCTGCGGA-3Ј, containing an XbaI site). The GAG and G3 sequences were amplified by PCR using oligonucleotides 103 (5Ј-CTG-CAGTCTAGAGCCTTCCCTGAAATT-3Ј, containing an XbaI site) and 104 (5Ј-CGCCTACTCGAGCTAATGATGATGATGATGATGGGTGGG-TCTGTGCACGACACC-3Ј, containing an XhoI site, stop codon, and His 6 ). PCR products were first digested with XbaI and then ligated with T4 ligase. The ligated products were amplified by PCR, and the fulllength G3 insert was digested by HindIII and XhoI, then ligated into HindIII and XhoI sites of pcDNA3 vector (Invitrogen). 2) The G1 construct contains SP, endogenous N-terminal G1 domain (G1), GAG, and His 6 , abbreviated as SP-G1-GAG-His 6 . The SP, GAG, and His 6 domains are identical to those of construct G3. SP and G1 sequences were amplified by PCR using oligonucleotides 203 (5Ј-GACGTCGGATCCAT-GACCACTCTACTA-3Ј, containing a BamHI site) and 201 (5Ј-GAG-GCTCTGGTCCCAGGGGCCTTCCCTGAAATTAGC-3Ј), which overlaps * This research was initiated under National Institutes of Health Grant AR12683 (to M. L. T.) and supported in part by a grant (to M. L. T.) and graduate fellowships (to W. L. and J. Z.) from the University of Connecticut Health Center. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Cell Culture and Transfection-Chinese hamster ovary (CHO) cells and mutant cell lines were grown in minimal essential culture medium (Life Technologies, Inc.) supplemented with 5% fetal bovine serum and antibiotics (Life Technologies, Inc.; 100 units/ml penicillin, 100 g/ml streptomycin, 250 ng/ml amphotericin) at 37°C in humidified air with 5% CO 2 . 1.8 million cells/100-mm culture dish were seeded and cultured for 36 h prior to transfection. 15 g of DNA and 52.5 l of Lipofectamine reagent (Life Technologies, Inc.) were each added to 500 l of Opti-MEM (Life Technologies, Inc.), and mixed for 30 min. Meanwhile, a dish of cells was twice washed with Opti-MEM and cultured with 5 ml of Opti-MEM. The DNA/Lipofectamine mixture was then added to the cells, which were incubated for 6 h.
Radioactive Labeling and Nickel Chromatography-After transfection, the medium was aspirated. For 35 SO 4 labeling, 5 ml of fresh Opti-MEM and 250 l of [ 35 S]Na 2 SO 4 (DuPont NEN) were added to the cells, which were incubated for 16 h. For [ 35 S]Met labeling, 5 ml of methionine-and cysteine-free Opti-MEM medium and 125 l of 35 Sprotein labeling mix (Amersham) were added to the cells, which were incubated for 16 h. The medium was collected and protease inhibitors were added to final concentrations of 1 mM PMSF, 2 g/ml leupeptin, 0.4 g/ml antipain, 2 g/ml benzamidine, 2 g/ml aprotinin, 1 g/ml chymostatin, 1 g/ml pepstatin. Debris was removed by centrifuging at 1,000 rpm for 1 min. 900 l of medium was adjusted to 0.5% Triton X-100 concentration and added to 50 l of nickel-resin (Qiagen) for 1 h. Resin had been prepared by equilibration with 0.1 M NaCl, 44 mM NaHCO 3 , 1 mM imidazole, 0.5% Triton X-100. Sample-containing resin was spun at 14,000 g for 15 s and the liquid decanted. Resin was washed with buffer A (10 mM HEPES, 2 M NaCl, 10% glycerol, 0.1 mM PMSF, 2 mM imidazole, 0.5% Triton X-100, 6 M urea, 250 mM dithiothreitol, pH 8.0) for 30 min, then buffer B (10 mM HEPES, 750 mM NaCl, 10% glycerol, 0.1 mM PMSF, 10 mM imidazole, 0.5% Triton X-100, 1 mg/ml bovine serum albumin, pH 8.0) for 30 min, and buffer C (10 mM HEPES, 750 mM NaCl, 10% glycerol, 0.1 mM PMSF, 10 mM imidazole, pH 8.0) for 15 min, and eluted with 10 mM HEPES, 750 mM NaCl, 10% glycerol, 100 mM EDTA, 0.1 mM PMSF, 250 mM imidazole, pH 8.0, for 1 h. Cells were washed twice with Hanks' medium (Life Technologies, Inc.) and lysed with 1 ml of 0.5% Nonidet P-40 at 0°C for 10 min. The lysate was spun at 14,000 ϫ g for 5 min. The clear lysate was transferred to a new tube. 200 l of lysate ϩ 700 l of Opti-MEM were adjusted to 0.5% Triton X-100 concentration and added to 50 l of nickel-resin prepared as above. The resin was washed three times with buffer A for 30 min, buffer B for 30 min, and buffer C for 15 min, and then eluted as described above.
SDS-PAGE and Autoradiography-The eluates or digested products were electrophoresed in a 5-15% gradient SDS-PAGE gel, which was dried and exposed to x-ray film.
Kinetic Experiments and Radioactivity Quantitation-Samples were collected at 4, 8, and 12 h after adding radioactive isotope. The radioactive protein bands were measured by an Instant Imager (Packard); G1 and G3 signals were corrected by subtracting background due to proteins resulting from empty vector transfection. Samples loaded onto gels were known portions of total cell lysate or medium. Total amounts of radioactive G1 and G3 in each sample were calculated by normalizing to their initial volumes in the harvested samples. The secretion of G1 and G3 proteins is expressed as the total amount of G1 or G3 in the medium divided by the sum of the total amount of G1 or G3 in the medium plus cell lysate.

RESULTS
This study was designed to elucidate whether globular domains G1 and G3 are themselves able to transit through the secretory pathway. They were placed in constructs that reflect their arrangement in aggrecan itself. Fig. 1 shows the linear arrangement of the components and deduced amino acid sequences of the G1 and G3 constructs. Included are the endogenous avian aggrecan signal peptide, endogenous avian GAG chain consensus sites, and a His 6 tag, to enable purification of the expressed proteins. Avian G1 and G3 domains were included in their respective constructs. Constructs were assembled by ligation or by overlapping PCR (15); the finished DNA products were sequenced to determine fidelity and were inserted into a eukaryotic expression vector. In separate studies, the transfection system was optimized using chloramphenicol acetyltransferase inserted in the same vector. Enzyme protein expression, measured by enzyme-linked immunoabsorbent assay, was Ͼ500 times basal levels, per unit of CHO cellular protein (data not shown). Using the same conditions, G1 and G3 constructs were separately transfected into CHO host cells.
Transfected G3 was readily expressed as indicated by Northern analysis (data not shown), immunocytology (data not shown), and by detection of a neoproteoglycan secreted by the cells into the culture media (Fig. 2). The neoproteoglycan encompasses a broad band, ranging from 82 to 160 kDa, when its GAG chains or core protein are labeled with 35 SO 4 or [ 35 S]methionine, respectively. Comparison of Fig. 2 (b and c) indicates that the smaller components of the secreted neoproteoglycan population are not sulfated. Densitometry of lane 1 in Fig. 2c and lane 2 in Fig. 3a shows that about 80% of secreted G3 is in the broad band containing GAG chains. Fig. 2b shows that the neoproteoglycan GAG chains are primarily CS, with minor amounts of HS; the latter interpretation is based upon comparison of lanes 2 and 6, which shows heparitinase digestion of traces of chondroitinase-resistant material. In contrast to G3, G1 is not labeled by radioactive sulfate (Fig. 2a) and its se- creted form is slightly larger than intracellular G1 (Fig. 3); the estimated sizes are 103 and 91 kDa, respectively. This size difference is not altered following incubation of G1 in GAG chain hydrolases (data not shown). G1 has three consensus sites for N-glycosylation (Fig. 1), and it may be N-glycosylated to yield a larger molecule. G3 is unlikely to be N-glycosylated as its sole consensus site includes a consensus-abrogating proline.
Figs. 3 and 4 show that both G1 and G3 core proteins migrate anomalously in SDS-PAGE gels, each appearing about 2-fold larger than predicted by amino acid composition (Fig. 1). Intracellular G1 migrates near 91 kDa, while intracellular G3 migrates near 82 kDa. Transfection of two different mutant host cells, which are unable to add GAG chains to core proteins (17), confirmed the apparent sizes of G1 and G3 core proteins (Fig. 3). Core protein anomalous migration may be due to the His 6 tag, which seems to cause this behavior in other expressed proteins (18).
The pgsA and pgsB mutant cell lines are defective in xylosyl transferase and galactosyl transferase, respectively, the first two enzymes which catalyze xylose and galactose addition to core proteins. These two sugars form part of the linkage between core proteins and CS/HS chains. The G3 protein in the medium of these mutant cells lacks GAG chains and is similar in size to the smallest G3 components of the neoproteoglycan population produced by wild type cells (Figs. 2-4). Such small non-sulfated G3 neoproteoglycan species, secreted from wild type host cells, may have relatively few or incomplete GAG chains, perhaps due to saturation of one or more post-translational enzymes by excess core protein. Alternatively, they may have been secreted via a "shunt" pathway, bypassing the Golgi stacks.
Secretion was assessed by measuring total radioactivity in intracellular and extracellular G1 and G3 populations. After overnight labeling of wild type and mutant cells, G3 was only detectable in the medium, whereas G1 was discernible in cell extracts and medium (Fig. 3). G3 was detectable in cells after shorter labeling periods (Fig. 4); its level in lanes 4 and 5 is readily measured above the corresponding backgrounds in lanes 1 and 2. Kinetic analysis confirmed that G3 secretion from cells is greater than G1 secretion; the difference was 24-fold at 4 h, 48-fold at 8 h, and 100-fold at 12 h. This progressive increase probably reflects accumulation of G1 inside the cells. The rate of secretion of G3 is clearly quite different from that of G1.
It is not known whether aggrecan core protein requires xylosylation for exit from the ER to the Golgi stacks (19). Our results show that a partial aggrecan, based on G3, is readily secreted in the absence of xylosylation when transfected into the xylosyl transferase mutant, pgsA (Fig. 3). The secretion rate of non-xylosylated G3 is the same (100%) as that of the xylosylated G3 after 16 h. In contrast, a partial aggrecan based on G1 is poorly secreted in the presence or absence of xylosylation. Thus, G3 secretion is independent of xylose attachments, whereas xylose addition has no detectable effects on G1 secretion/retention.  -6), and G1 proteins (lanes 7-9), each at 4, 8, and 12 h, respectively. The arrows depict the locations of G1 and G3 intracellular proteins. The samples in panel a were exposed to film longer than the ones in panel b. c, percentage of G1 and G3 secreted by cells. DISCUSSION The difference in secretory behavior between G1 and G3 is consistent with retention of truncated aggrecan core protein, which lacks G3, in nanomelic chondrocytes. The two constructs used in this study represent partial aggrecan proteins, although the tandem arrangement of included domains corresponds to those of whole aggrecan itself. Not only is G1 secreted more slowly than G3 but it fails to become decorated with GAG chains. This latter observation suggests G1 either does not transit through the same secretory pathway as G3 or its consensus sites are inaccessible for GAG chain modification. In contrast, G3 is completely secreted within 16 h after transfection and a major portion of the protein is decorated with GAG chains, appearing in SDS-PAGE gels as a typical proteoglycan. Thus, a core protein encompassing G3 plus contiguous GAG chain consensus sites is able to move through the normal secretory pathway, mimicking the translocation of whole aggrecan protein; as in aggrecan protein itself, the GAG consensus sites serve as Golgi reporter groups.
In contrast, only small amounts of a protein encompassing G1 plus the same contiguous GAG chain consensus sites, which lacked GAG chains, appear in spent media. Perhaps this protein is released from dead or dying cells and had not traversed the normal secretory pathway. Alternatively, it might have traversed a "shunt" pathway, bypassing the Golgi stacks, if such exists in normal cells. G1 and G3 domains, in both aggrecan protein and the construct proteins, have opposite orientations to the contiguous GAG sites. Possibly, the tandem location site of a globular domain as C terminus versus N terminus to the GAG consensus sites may be important for traversing the secretory pathway. We do not consider this to be a likely possibility because earlier studies have shown that similar GAG chain consensus sites, flanked at the N terminus by a signal sequence and protein A, did transit through the Golgi stacks (20,21). Those studies also showed that a C-terminal His 6 tag did not affect GAG chain attachment.
Xylosylation of GAG chain consensus sites may be important for proteoglycan secretion (19,22). Our results show that xylosylation of contiguous GAG sites is not required for G3 protein secretion; non-xylosylated G3 protein is secreted in equal amounts to xylosylated G3 protein. This result also implies that GAG chain attachment per se is not required for core protein secretion. Furthermore, our results are consistent with the concept that ER molecular chaperones primarily recognize globular protein domains (3); non-globular domains such as GAG chain consensus sites may not be subject to chaperone surveillance.
In the case of a proteoglycan with multiple globular domains, such as members of the aggrecan family, each domain is potentially subject to chaperone surveillance. It is difficult to visualize, for an individual molecule, how such multidomain surveillance might be orchestrated and regulated. The observation that domain G3 is secreted normally, while domain G1 is not, suggests that one intramolecular domain may be preferably recognized by molecular chaperones. Thus, G3 may facilitate movement of the entire aggrecan core protein through the secretory pathway. If so, it is intriguing to speculate what molecular mechanisms may account for such behavior.