Post-translational modification of bone morphogenetic protein-1 is required for secretion and stability of the protein.

Bone morphogenetic protein (BMP)-1 is a glycosylated metalloproteinase that is fundamental to the synthesis of a normal extracellular matrix because it cleaves type I procollagen, as well as other precursor proteins. Sequence analysis suggests that BMP-1 has six potential N-linked glycosylation sites (i.e. NXS/T) namely: Asn(91) (prodomain), Asn(142) (metalloproteinase domain), Asn(332) and Asn(363) (CUB1 domain), Asn(599) (CUB3 domain), and Asn(726) in the C-terminal-specific domain. In this study we showed that all these sites are N-glycosylated with complex-type oligosaccharides containing sialic acid, except Asn(726) presumably because proline occurs immediately C-terminal of threonine in the consensus sequence. Recombinant BMP-1 molecules lacking all glycosylation sites or the three CUB-specific sites were not secreted. BMP-1 lacking CUB glycosylation was translocated to the proteasome for degradation. BMP-1 molecules lacking individual glycosylation sites were efficiently secreted and exhibited full procollagen C-proteinase activity, but N332Q and N599Q exhibited a slower rate of cleavage. BMP-1 molecules lacking any one of the CUB-specific glycosylation sites were sensitive to thermal denaturation. The study showed that the glycosylation sites in the CUB domains of BMP-1 are important for secretion and stability of the molecule.

BMP-1 consists of a prodomain that is cleaved by a furin-like enzyme in the trans-Golgi network, 2 an astacin-like zinc metalloproteinase domain (15), one epidermal growth factor-like domain, and three CUB domains. In other proteins, CUB domains mediate protein-protein interactions (16).
BMP-1 purified from mouse fibroblasts culture medium has been shown to be N-glycosylated (17). Sequence analysis reveals six potential N-glycosylation sites, one of which is located in the prodomain and five in the mature (active) molecule. However, no information is available on the structure and function of the glycosylation sites. Recent studies indicate that post-translational modification of proteins can have multiple roles including regulation of intracellular trafficking (18), stabilization of folded domains (19,20), protection from proteolytic degradation of the core protein (21), and modulation of enzyme/ hormone activities (22)(23)(24). Furthermore, high variability exists in the functions of N-glycan chains of proteins, which makes it difficult to predict with any confidence the functions of oligosaccharides on proteins (reviewed by Ref. 25).
In this study, we expressed FLAG-tagged BMP-1 in two different mammalian systems (HT1080 and 293-EBNA cells) and identified the type of glycosylation on BMP-1. By sitedirected mutagenesis, we also established the role of the Nglycosylation in folding, secretion, and C-proteinase activity of BMP-1.

EXPERIMENTAL PROCEDURES
Source of Materials-PCR products were purified with Qiaquick kits (Qiagen). Plasmids were extracted with Qiaprep spin miniprep kit (Qiagen). Prestained protein molecular weight standards (broad range) were from Bio-Rad. Full-length BMP-1 cDNA (GenBank TM accession number P13497) was cloned from a human placental cDNA library. The cDNA was inserted at the KpnI/XhoI sites of the expression vector pcDNA3 (Invitrogen), thereby placing it under the transcriptional control of a cytomegalovirus promoter. A FLAG tag amino acid sequence (DYKDDDDK) recognized by a mouse monoclonal anti-FLAG M2 antibody (Sigma) was introduced into the BMP-1 sequence (BMP-1F) immediately 5Ј of the stop codon. The cDNA encoding FLAG-tagged BMP-1 was subcloned into the episomal expression vector pCEP4 (Invitrogen) and pcDNA3, for heterologous protein expression studies in cultured cells. Previous studies have shown that the FLAG peptide at the C terminus of BMP-1 does not affect the procollagen C-proteinase activity of BMP-1 (2).
Site-directed Mutagenesis-Plasmids coding for the mutant BMP-1 proteins were produced by replacing wild-type fragments with the same fragments containing the desired mutations. These were generated by strand overlap PCR as described (26) using Pwo polymerase (Roche Molecular Biochemicals), a forward primer upstream and a reverse primer downstream, of unique restriction sites, respectively, and oligonucleotides containing the desired modification in both orientations (in bold, see below). For each mutation, the restriction sites used and their positions on the nucleotide sequence are indicated. Briefly, a DNA fragment was amplified using a forward primer and the antisense mutant primer and an overlapping fragment was amplified using the sense mutant primer and a downstream reverse primer. Both fragments were gel-purified (Qiagen), mixed, and re-amplified with the Pwo enzyme with the forward and reverse primers. The product was digested by the appropriate enzymes, gel-purified, and introduced in place of the corresponding wild-type fragment in BMP-1F. Mutagenic primers (mutation in bold) and restriction sites used to insert the mutant fragment into the wild-type sequence were as follows: N142Q, 5Ј-CATTGGGGGACAGTTCACTGGTA-3Ј, XcmI (position 383)/BlpI (position 913); N332Q, 5Ј-CAGCACAGGCCAGTTCTCCT-3Ј, BlpI (position 913)/PmlI (position 1416); N363Q, 5Ј-GATCATCCTGCAAT-TCACGTCCCT-3Ј, BlpI (position 913)/PmlI (position 1416); N599Q, 5Ј-CCTCACCAAGCTCCAAGGCTCCATCA-3Ј, BamHI (position 1391)/ XhoI (multicloning site of pcDNA3). For N91Q (5Ј-AAAAGCTGCAGT-TCCAGGACAGACTTCTAC-3Ј) and N726Q (5Ј-CCCCCCTCGAGTCA-CTTGTCATCGTCGTCCTTGTAGTCCTGGGGGGTCCGTTGTCTTT-TCTGCACT-3Ј), the mutagenic primers, which also contained a restriction site (underlined, N91Q), PstI (position 290), N726Q, XhoI (multicloning site of pcDNA3), and the FLAG tag (N726Q, italicized), were used in a single PCR reaction with reverse and forward primers, respectively. The two mutated PCR products were digested by PstI (position 251) and ApaI (position 586) (N91Q) or BamHI (position 1391) and XhoI (N726Q). Pwo DNA polymerase was used to minimize base misincorporation during the polymerase chain reactions. DNA sequencing (ABI) was used to verify the mutations, and to ensure that the cDNA clones were error-free.
Cell Culture and Transfection-Human fibrosarcoma HT1080 cells (ATCC CCL-121) and human embryonic kidney 293-EBNA cells (ECACC 85120602) were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Invitrogen) (complete DMEM) and for the 293 cells, with 0.25 mg/ml Geneticin (G418, Invitrogen) in a 37°C incubator with 5% CO 2 . The recombinant wild-type and mutant BMP-1F proteins were expressed in stably transfected HT1080 cells. Transfections were made with Lipofectin reagent (Roche Molecular Biochemicals) and 4 g of plasmid/T-25 flask. Cells were grown to ϳ50% confluency by overnight incubation in complete DMEM. After 2 rinses with Opti-MEM (Invitrogen), cells were transfected in serum-free Opti-MEM (Invitrogen) following the manufacturer's instructions and returned to the incubator. Twenty-four hours after transfection, medium was replaced by DMEM with 10% serum for a further 24 h. The cells were then trypsinized (Invitrogen) and diluted 1:15 for selection in 0.25 mg/ml Geneticin (G418, Invitrogen). 293-EBNA cells were transfected with the FLAG-tagged BMP-1 cloned into pCEP4 expression plasmid, following the procedure described above. Selection was applied by adding 0.25 mg/ml hygromycin (Invitrogen).
Preparation of the Medium-Stably transfected HT1080 or 293-EBNA cells were seeded in 100-mm dishes and grown to confluency. Cells were rinsed three times with phosphate-buffered saline (Invitrogen) and incubated in serum-free DMEM for 24 (HT1080) or 48 h (293-EBNA cells), unless otherwise stated. The tissue culture media were collected, cleared of cell debris by centrifugation at 1600 ϫ g for 10 min, and in the case of HT1080-transfected cells, concentrated to 100 l using Centriprep-30 and Microcon-10 concentrators (Amicon, Inc.). The samples were used immediately or stored at Ϫ80°C.
Assay of Procollagen C-proteinase-Recombinant BMP-1 was assayed for procollagen C-proteinase activity using human L-U-14 C-type I procollagen substrate (0.4 g). 14 C-Labeled type I procollagen was obtained and purified from the medium of human skin fibroblasts (27). Analysis of the cleavage products on SDS gels (7% separating, 3.5% stacking) was performed as described (3,28). The cleavage products were visualized by exposing dried gels to a phosphorimaging plate (Fuji, type BAS III) for phosphorimaging (Fujix BAS 2000). Bands corresponding to the pro-␣1(I) and pN-␣2(I) chains of type I procollagen and type I pN-collagen, respectively, were quantified using AIDA 2.0 software. The percentage of cleavage was calculated by multiplying the intensity of the pN-␣2(I), corrected for molecular mass, by the initial concentration of procollagen (3,28).
Electrophoresis and Western Blotting-Cells were rinsed once with PBS, and incubated on ice for 15 min with occasional shaking with 500 l of RIPA buffer (150 mM NaCl, 1% deoxycholate, 0.1% SDS, 10 mM Tris pH 7.6) containing 10 mM EDTA, and protease inhibitor mixture (Roche Molecular Biochemicals). Cells in RIPA buffer were scraped on ice and sonicated. Lysates were subjected to a 5-min centrifugation at 14,000 ϫ g at 4°C. Supernatants were retained and stored at Ϫ80°C until further analysis. The supernatants or cell lysates were resolved by electrophoresis on a 10% (w/v) SDS-Prosieve gel (Biowhittaker Molecular Applications) under reducing conditions and subjected to Western immunoblotting using the mouse monoclonal M2 antibody (Sigma) directed against the FLAG tag. Secondary antibody (anti-mouse peroxidase-conjugated IgG (Sigma)) was detected by the enhanced chemiluminescence method (SuperSignal West Dura extended duration, Pierce). The levels of BMP-1F were quantified by laser densitometry of enhanced chemiluminescence fluorograms exposed to pre-flashed films.
Glycan Trimming and N-Glycosylations Inhibitors-Stably transfected HT1080 or 293-EBNA cells expressing BMP-1F were grown to confluency on 100-mm dishes. Cells were preincubated in complete DMEM in the presence of 100 g/ml castanospermine (Calbiochem), or 5 g/ml swainsonine (Calbiochem) for 18 h. For tunicamycin (Sigma), cells were preincubated in the presence of 2 g/ml of the inhibitor, for 5 h. Cells were then rinsed three times with PBS, and reincubated in the presence of the inhibitors for 24 h in serum-free DMEM, and for only 18 h in the case of tunicamycin. Medium was collected and treated as described above.
Immunofluorescence Microscopy-pcDNA3, BMP-1F, and N4Qtransfected HT1080 were plated on glass coverslips in 6-well plates. After 24 h, cells were preincubated with or without 10 mM lactacystin (Calbiochem) for 1 h. Cells were rinsed three times with PBS and further incubated with or without lactacystin at the same concentration in serum-free medium for 4 h. Cells were washed three times with PBS and fixed and permeabilized with cold methanol (Ϫ20°C) for 5 min. Fixed cells were washed three times with PBS and incubated for 20 min at room temperature with anti-FLAG mouse antibody or with anticalreticulin rabbit antibody (Stressgen) in PBS supplemented with 1 mg/ml bovine serum albumin (Sigma). After washing, cells were incubated with fluorescein isothiocyanate-conjugated anti-mouse IgG (Sigma) or with rhodamine-conjugated anti-rabbit IgG (Santa Cruz Biotechnology). Cells were washed with PBS and coverslips were mounted in Mowiol 4-88 (Calbiochem) and observed with a Bio-Rad MRC1000 laser confocal microscope.

RESULTS
Characterization of the Post-translational Modification of BMP-1-We transfected cultured HT1080 human fibrosarcoma and 293-EBNA human embryonic kidney cells with cDNAs encoding C-terminal flagged BMP-1 (BMP-1F). We chose to express BMP-1 in two different cell types to minimize the risk of cell type-specific post-translational modification. BMP-1F was stably expressed in HT1080 and 293-EBNA cells and the secreted protein was incubated with endoF/N, which cleaves all N-linked structures, regardless of their complexity, by hydrolyzing the asparagine-oligosaccharide bond, or endoH, which cleaves specifically high mannose-type structures. The proteins were also digested with neuraminidase, which removes sialic acids, and O-glycosidase, which specifically cleaves O-linked glycans. Proteins were separated by SDS-PAGE and immunoblotted with anti-FLAG antibody. Fig. 1 shows that BMP-1F is correctly processed and secreted in both HT1080 and 293-EBNA cells (lanes 2 and 9). No immunoreactive bands were found in medium from HT1080 cells transfected with pcDNA3 empty vector (lane 1). Complete deglycosylation of BMP-1F by endoF/N produced a decrease in the molecular mass of the BMP-1F molecule from ϳ75 to ϳ60 kDa (lane 3), strongly suggesting that most, or all, of the 5 potential N-glycosylation sites present in the active form had been stripped of oligosac-charides. Digestion of BMP-1F with endoH did not produce a shift in mobility, suggesting that BMP-1F secreted by HT1080 cells (and 293-EBNA cells, data not shown) did not contain high mannose-type oligosaccharides (lane 4). Neuraminidase treatment (lanes 6 and 10) decreased the molecular mass of BMP-1F by ϳ5 kDa, showing that the BMP-1 contained sialylated sugars. However, further treatment with O-glycosidase had no effect (lane 8), suggesting that the BMP-1 lacked O-linked oligosaccharides.
Complex-type Glycosylation Is Not Required for Secretion and PCP Activity of BMP-1F-To determine the contribution of the complex-type N-linked oligosaccharides to secretion and enzymic activity of BMP-1F, two inhibitors of glycoprotein processing were used: castanospermine, which inhibits endoplasmic reticulum (ER) glucosidases I and II, thereby preventing the removal of the glucose residues of the Glc 3 Man 7-9 (Glc-Nac) 2 -N-linked glycoprotein (29), and swainsonine, which inhibits Golgi mannosidase II, therefore producing hybrid-type glycosylation (29). To do these experiments, stably transfected HT1080 cells were incubated in the presence and absence of castanospermine or swainsonine, and the secreted BMP-1F was digested with endoF/N and endoH. The digested proteins were examined by Western blotting, using the anti-FLAG antibody (Fig. 2). BMP-1F from cells cultured with castanospermine (lane 5) had the same electrophoretic mobility as BMP-1F from untreated cells (lane 2), whereas BMP-1F secreted in the presence of swainsonine (lane 8) had a faster mobility (ϳ70 kDa). The levels of secretion of BMP-1F were not affected by the presence of the inhibitors, compared with the control.
To characterize the oligosaccharides on BMP-1F secreted in the presence of the inhibitors, the proteins were incubated with endoF/N and endoH enzymes and the products were examined by SDS-PAGE and Western blotting. Whereas BMP-1F secreted in the absence of inhibitors was resistant to endoH (lane 4), BMP-1F secreted in the presence of either inhibitor was sensitive to this enzyme, indicating that the inhibitors had changed the nature of the oligosaccharides. By comparing the electrophoretic mobility of the bands obtained after endoH and endoF/N digestions, it could be observed that BMP-1F secreted in the presence of castanospermine contained only high-man-nose structures, because the two digestion products had the same electrophoretic mobility (lanes 6 and 7). In the case of BMP-1F secreted in the presence of swainsonine, the digestion product by endoH (lane 10) migrated slower than the one obtained with endoF/N (lane 9), denoting the presence of both complex-type and high mannose-type structures.
To investigate the effects of high mannose-and hybrid-type oligosaccharides on PCP activity of BMP-1, we assayed the enzymes using type I procollagen as substrate. No significant difference of PCP activity was observed between BMP-1F control, BMP-1F containing high-mannose structures (synthesized in the presence of castanospermine), and BMP-1F with hybridtype glycosylation (synthesized in the presence of swainsonine). Similar results were obtained with the 293-EBNA cells stably transfected with BMP-1F (data not shown). The results showed that complex-type oligosaccharide chains are not required for secretion and PCP activity of BMP-1.
N-Linked Oligosaccharides Are Necessary for Proper Folding and Secretion of BMP-1-To determine the role of N-linked sugars on folding and secretion of BMP-1, stably transfected 293-EBNA cells were incubated in the presence or absence of tunicamycin, which blocks transfer of N-acetylglucosamine onto the lipid carrier dolichol phosphate (30). Each medium and cell lysate were collected, the proteins were separated by SDS-PAGE and immunoblotted with anti-FLAG antibody (Fig.  3). Whereas BMP-1F was secreted from untreated cells (lane 1), BMP-1F could not be detected in the medium of tunicamycintreated cells (lane 2). Latent BMP-1F was present in the cell lysates of treated and untreated cells. However, latent BMP-1F in the tunicamycin-treated cells was ϳ70 kDa, and therefore smaller than in control samples. The smaller molecular mass confirmed the presence of N-linked oligosaccharides on BMP-1. tion of each site, we generated mutations in the BMP-1 sequence in which each of the asparagine residues was mutated, by site-directed mutagenesis, to glutamine. In addition to the 6 single mutants, we generated the double CUB1 mutant (N332Q and N363Q), as well as a molecule with Asn to Gln mutations of all 3 CUB sites and the C-terminal-specific site (N4Q), and a molecule in which all 6 potential sites were mutated (N6Q). Wild type and mutant constructs were stably transfected into HT1080 cells. Serum-free conditioned medium was analyzed by Western blotting using the anti-FLAG antibody (Fig. 5). All the single mutants (panel A, lanes 3-6, 8, and  9) and the double CUB1 mutant (panel B, lane 2) were secreted. The N4Q and N6Q mutants were not secreted (see lanes 10 and 11). The fact that the fully unglycosylated N6Q mutant was not secreted is in agreement with the result obtained with the cells treated with tunicamycin (Fig. 3). Noteworthy, the N4Q and N6Q mutants were not detected in cell lysates (data not shown).
The single mutants N142Q, N332Q, N363Q, and N599Q (panel A, lanes 4, 5, 6, and 8, respectively) migrated faster than the wild-type BMP-1F (lanes 2 and 7), indicating that these glycosylation sites are occupied by oligosaccharides in the wildtype molecule. In contrast, the N726Q mutant (lane 9) was indistinguishable from BMP-1F, suggesting that this site is not glycosylated in BMP-1. The prodomain mutant, N91Q, was secreted as the mature protein, which migrated with the same electrophoretic mobility as BMP-1F. This was in agreement with the fact that BMP-1F is secreted in its active form, and the absence of prodomain glycosylation did not block the cleavage of the prodomain or change the electrophoretic mobility of the mature enzyme. By analyzing the corresponding cell lysates (where BMP-1F exists in the latent form, panel C, lanes 2 and 3), the N91Q mutant was shown to migrate faster than BMP-1F, indicating that the Asn 91 site is glycosylated in latent BMP-1.
All the Single Mutants Cleave Type I Procollagen But CUB Domain Glycosylation Is More Important Than Catalytic Domain Glycosylation for PCP Activity-All the single mutants exhibited PCP activity, in that they cleaved 14 C type I procollagen to completion in 18 h (Fig. 6). In contrast, the double mutant N332Q/N363Q was a weak C-proteinase. Minor differences between the mutants were detected when type I procollagen was incubated with the mutants for 4 h (Fig. 7). No difference was observed between BMP-1F and prodomain N91Q mutant. The single CUB mutants cleaved procollagen slower than BMP-1F. Furthermore, as shown above, the CUB1 double mutant (N332Q/N363Q) was much slower at cleaving procollagen than control samples. These results showed that the absence of single glycosylation sites has a minor effect on PCP activity of BMP-1. However, absence of two or more glycosylation sites decreases the PCP activity of the molecule.
Sensitivity of the CUB Glycosylation Mutants to Heat Inactivation-To evaluate the ability of the N-linked glycosylation sites to stabilize the structure of BMP-1, we incubated the secreted forms of the mutated proteins at 55°C for up to 40 min, and assayed the molecules for PCP activity at 37°C. The results are shown in Fig. 8. N91Q, which had an unaltered mature sequence was used as the control. The secreted N142Q enzyme retained 20% of its activity after 40 min at 55°C, and  1 and 3 (C)) or presence of tunicamycin (2 g/ml) (lanes 2 and 4 (T)) in serum-free medium for 18 h. Proteins from the cell medium (lanes 1 and 2) or from the cell lysates (lanes 3 and  4) were separated in reducing conditions on a 10% SDS-PAGE gel, followed by Western blot using the anti-FLAG M2 antibody. was therefore fairly resistant to heat denaturation. However, the PCP activity of the CUB mutants N332Q/N363Q, N599Q, and N332Q decreased to zero in 5, 10, and 20 min, respectively. The N363Q mutant was partially resistant to heat denaturation. These results showed that the oligosaccharides in the CUB1 and -3 domains, and in particular those carried by Asn 332 and Asn 599 , are important for the thermal stability of BMP-1. The greater susceptibility of the CUB mutants to thermal denaturation was also observed at 45°C (data not shown).
The N4Q Mutant Is Degraded by the Proteasome-Recent work has shown that the ubiquitin-proteasome degradation pathway is not restricted to membrane proteins (31), but is also involved in the degradation of mutant secretory proteins (32,33). To determine the fate of the N4Q mutant, immunofluorescence was performed on N4Q expressing cells (Fig. 9) in the presence and absence of lactacystin, which is an inhibitor of the proteasome. This inhibitor specifically blocks the three peptidase activities of the proteasome (34,35), thereby preventing misfolded proteins from being deployed to distal compartments of the secretory pathway. In the absence of lactacystin, N4Q and calreticulin (an ER resident protein) did not co-localize, which indicated that N4Q was efficiently translocated out of the ER (Fig. 9A). However, in the presence of 10 M lactacystin, the N4Q mutant accumulated in the ER (Fig. 9, C and D), as shown by its croissant-shaped co-localization with calreticulin. These results are consistent with translocation of the N4Q mutant of the ER and subsequent degradation by the proteasome.

DISCUSSION
In this study we have shown that BMP-1 is N-linked glycosylated at five sites within the molecule and that these sites are important for protein secretion and stability. Three of the sites occur within the protein interaction CUB domains and are important in stabilizing the structure of the BMP-1 molecule. Previous studies had shown that mouse BMP-1 (17) and a BMP-1 homologue in sea urchin (36) were N-glycosylated, but little was known about which of the six potential sites in BMP-1 were glycosylated and the nature of the glycosylation. Our studies show that 5 of the 6 consensus Asn-X-Ser/Thr sites are glycosylated. The only site not glycosylated occurs in the BMP-1-specific domain at the C terminus of the molecule (Asn 726 -Arg-Thr-Pro). Presumably this asparagine is not glycosylated because proline occurs immediately C-terminal of the conserved threonine (37)(38)(39). We concluded that the oligosaccharide side chains were sialylated, regardless of whether the BMP-1 had been expressed in HT1080 cells or 293-EBNA cells. We also showed that BMP-1 does not contain O-linked oligosaccharides.
As recombinant BMP-1 contained complex type N-glycosylation, it was of interest to evaluate the impact of these particular glycan chains on secretion and activity of BMP-1. Castanospermine and swainsonine have been widely used in a number of studies to determine whether changes in the structure of the N-linked sugars affects glycoprotein functions. Our results indicate that the complex-type glycosylation was not required for secretion and activity of BMP-1. Indeed, BMP-1F with high mannose-type glycans created by castanospermine, or with hybrid-type structures generated by swainsonine, were secreted as efficiently as untreated BMP-1F, and were equally FIG. 6. Cleavage of type I procollagen by wild-type and mutant BMP-1F (18 h)(reducing conditions). 14 C-Labeled type I procollagen (0.4 g) was incubated at 37°C for 18 h with 5 l of concentrated medium from HT1080 cells that had been stably transfected with wildtype and mutant BMP-1F, or with empty vector (pcDNA3). Ϫ, procollagen alone. The proteins were separated in reducing SDS gels (7%) and detected using phosphorimaging. In samples containing wild-type and mutant BMP-1F, the procollagen was converted to pN-collagen, which is a normal intermediate in the conversion of procollagen to collagen containing the N-propeptides but not the C-propeptides. All the mutants were found to cleave completely and efficiently all the procollagen, except N332Q/N363Q.

FIG. 7. Cleavage of type I procollagen by wild-type and mutant BMP-1F for 4 h (reducing conditions).
HT1080 cells were transfected with vectors encoding wild-type and mutant BMP-1F. The proteins in the culture medium were concentrated, and the levels of BMP-1F were quantitated by Western blot analysis using the anti-FLAG antibody. The PCP activity of the BMP-1F mutants, normalized for BMP-1 concentration, was assayed by cleavage of 14 C-labeled type I procollagen for 4 h. Percentage of procollagen is shown from three separate experiments (ϮS.E.M.).

FIG. 8. Thermal stability of wild-type and mutant BMP-1F.
Serum-free medium from stably transfected HT1080 with wild-type and mutant BMP-1F constructs were preincubated at 55°C for 5, 10, 20, and 40 min. PCP activity in the samples was determined using 14 Clabeled type I procollagen, in a 4-h assay at 37°C. q, N91Q; f, N142Q; ࡗ, N332Q; OE, N363Q; ϫ, N599Q; E, N332Q/N363Q. active in cleaving type I procollagen. This situation is not unusual, as in most cases, swainsonine has been shown to have little effect on glycoproteins, which may indicate that a partial complex chain is sufficient for activity, and that the conformation is not altered (Ref. 30 and references therein,and Ref. 18). Similarly, complex-type glycosylation was not essential for targeting, correct folding, or activity of meprin A, which, like BMP-1, is a member of the astacin family of metalloproteinases (22). Effects of castanospermine have been shown to be dependent on the protein in question. Indeed, the inability of cells to remove glucose residues may have some effects on the transport, secretion, or activity of the protein. For example, glucosylation did not affect castanospermine-treated N-acetylglucosaminyltransferase III Golgi enzyme activity (40), but its localization. On the other hand, castanospermine has been shown to diminish human ␣1,3/4-fucosyltransferase III to 40% of the active enzyme (41). Glucose trimming (prevented by castanospermine) appears to be essential for a number of proteins, because the lectin chaperones calnexin/calreticulin have been shown to bind to one or more monoglucosylated N-linked glycans (reviewed by Ref. 42), to make them achieve their proper tertiary or quaternary structure (43). However, calnexin (44) and calreticulin (45) also have efficient oligosaccharideindependent chaperone activity in vitro.
Whereas the nature of the N-glycan chains (i.e. high-mannose or hybrid-type) was found not to be essential for secretion and activity of BMP-1, it appeared that some of the N-linked sugars are necessary for proper folding and secretion of BMP-1, as shown by the results obtained in the presence of tunicamycin, and with the multiple mutants N4Q and N6Q. In these experiments, BMP-1 was absent from the supernatant, and in tunicamycin-treated 293-EBNA cells, the inhibitor resulted in the intracellular trapping of the latent unglycosylated misfolded protein. However, the multiple mutants N4Q and N6Q could not be detected in the cell lysate of the transfected HT1080 cells either. This apparent contradiction could be explained by the fact that in tunicamycin-treated cells, all glycosylation on all glycoproteins are inhibited including those involved in the quality control and trafficking. In addition, tunicamycin has also been described as an inhibitor of protein synthesis, depending on the system used, some being more sensitive than others, and the nature of the proteins (30). In contrast, in the cells transfected with the multiple mutant constructs N4Q and N6Q, there is no change induced on proteins other than BMP-1F. Another possibility for this difference involves the level of BMP-1F production in HT1080 and 293-EBNA cells, which is much higher in 293-EBNA cells. Therefore, it is conceivable that the ER quality control could be overwhelmed in 293-EBNA cells, resulting in the accumulation of the protein inside the tunicamycin-treated cells, whereas the misfolded N6Q protein recognized as such in HT1080 cells would be rapidly degraded. Indeed, 293-EBNA cells have previously been suggested to have a reduced capacity to deal with improperly folded proteins (46).
Once we had established the implication of N-glycosylation in BMP-1 structure/function, our first question was to identify which sites were glycosylated, and which one(s) of these had a role in folding, secretion, and activity. We generated a series of single-glycosylation mutants of BMP-1, as well as multiple mutants. Our results showed that the individual elimination of any one site had no significant effect on the secretion of BMP-1F, except Asn 332 and Asn 599 from which the corresponding mutants N332Q and N599Q were slightly less secreted than the wild-type (72 and 86%, respectively). The double CUB1 mutant N332Q/N363Q was secreted to 60% of the control, indicating the significant contributions of the CUB glycan chains for efficient secretion, which was further confirmed by the absence of secretion of the multiple mutant N4Q. Furthermore, lactacystin allowed us to visualize this mutant accumulating in the ER, which showed that the N4Q mutant was degraded, at least in part, by the proteasome. Therefore, the N4Q mutant is recognized as misfolded by chaperones, which facilitate retrograde transport to the cytosol for degradation by the proteasome. This transport occurs through the Sec61p translocon (33). Numerous misfolded proteins have been described to be substrates of the proteasome, including a meprin A mutant lacking a MAM domain (47). Interestingly, a truncated meprin A mutant lacking all carbohydrates and expressed in 293-EBNA cells was not targeted for proteasomal degradation and was retained in the ER, implicating carbohydrates as essential factors in retrograde transport into the cytosol (22,32). These results are in agreement with our observation, as the N4Q mutant still retains two glycosylation sites. Taken together, these results suggest that N-linked oligosaccharides participate in BMP-1 folding, probably with additive contributions of the carbohydrates chains.
We also noted that none of the single mutants had altered PCP activity. On the other hand, the low PCP activity of the double CUB-1 mutant N332Q/N363Q indicated that N-glycosylation plays an important role in BMP-1 PCP activity, either directly in enzyme-substrate interaction, or indirectly in domain-domain interactions. The metalloproteinase (N142Q) mutant exhibited a similar cleavage rate as BMP-1F. Indeed, the position of the glycosylation site, between ␤-strand I and helix A, is far from the active site (Fig. 10A). As this glycosylation site is highly conserved in mouse (48), Xenopus (49), and sea urchin (36) BMP-1, as well as in human tolloid-like 1 and 2 (2), and Drosophila tolkin/tolloid-related-1 (50, 51), it is likely to have an important role in these proteins. This mutant was also found to be as resistant as N91Q/BMP-1F to thermal denaturation. This result is in contrast to previous work on meprin A, showing that the three metalloproteinase glycosylation site mutants, which were located close to the active site, were more sensitive to heat denaturation than the wild-type, and had a decreased enzymatic activity (22). CUB-1 N332Q and CUB-3 N599Q were both found to have a slower cleavage rate than BMP-1F (about 70% of the control), showing that these sites contribute to efficient PCP activity. To determine where these glycosylation sites were located, we used as a template the structure of boar seminal plasma PSPI/II (52) and bovine acidic seminal fluid protein aSFP (53) (Fig. 10B), which have been crystallized. These proteins are built on a single CUB architecture that resembles a jellyroll-type structure with two fivestranded ␤-pleated sheets, interconnected by nine loops. Multiple sequence alignments of PSP-I/II, aSFP, and CUB domains of other proteins showed highly conserved elements including the two disulfide bridges, and the hydrophobic internal core amino acids defining the CUB signature (52). By alignment, we localized the CUB-1 Asn 332 site on the ␤3 strand and Asn 363 on the ␤5 strand, and therefore on the same side of the "jellyroll." In contrast, the CUB3 Asn 599 was located on loop B.
The three CUB mutants N599Q, N332Q, and N363Q were all found to be more sensitive to thermal denaturation than N142Q/N91Q. The most sensitive to heat denaturation was the double CUB-1 N332Q/N363Q mutant, in agreement with the notion that glycosylated proteins are thermally destabilized by carbohydrate removal (54). Interestingly, these three glycosylation sites are highly conserved in mouse (48) and Xenopus (49) BMP-1, and Asn 332 and Asn 599 are also conserved in sea urchin BMP-1 (36) and Drosophila tolloid-related 1 (50). Furthermore, in Drosophila, tolloid (55) and tolkin/tolloid-related 1 (50, 51) possess an additional potential glycosylation site in their third CUB, located on the ␤5 strand, at exactly the same location as CUB1 Asn 363 in BMP-1. This suggests a very precise role for this particular glycosylation site across the tolloid family. It is known that glycosylation can affect the local secondary structure of proteins (56). Glycosylation can facilitate the formation of a key segment of secondary structure, and therefore potentially direct the protein-folding pathway, or provide rigidity to residues proximal to the glycosylation site. As a result, the overall stability of the protein might be enhanced (exemplified in the human CD2 (57)), or the protein is enabled to bind a ligand (suggested for the kallikrein N-glycosylated loop of neuropsin (58)). These data emphasize the importance of glycosylation in determining and stabilizing the structure of BMP-1.