Proteasomal Degradation of Unassembled Mutant Type I Collagen Pro-α1(I) Chains*

We have previously shown that type I procollagen pro-α1(I) chains from an osteogenesis imperfecta patient (OI26) with a frameshift mutation resulting in a truncated C-propeptide, have impaired assembly, and are degraded by an endoplasmic reticulum-associated pathway (Lamandé, S. R., Chessler, S. D., Golub, S. B., Byers, P. H., Chan, D., Cole, W. G., Sillence, D. O. and Bateman, J. F. (1995)J. Biol. Chem. 270, 8642–8649). To further explore the degradation of procollagen chains with mutant C-propeptides, mouse Mov13 cells, which produce no endogenous pro-α1(I), were stably transfected with a pro-α1(I) expression construct containing a frameshift mutation that predicts the synthesis of a protein 85 residues longer than normal. Despite high levels of mutant mRNA in transfected Mov13 cells, only minute amounts of mutant pro-α1(I) could be detected indicating that the majority of the mutant pro-α1(I) chains synthesized are targeted for rapid intracellular degradation. Degradation was not prevented by brefeldin A, monensin, or NH4Cl, agents that interfere with intracellular transport or lysosomal function. However, mutant pro-α1(I) chains in both transfected Mov13 cells and OI26 cells were protected from proteolysis by specific proteasome inhibitors. Together these data demonstrate for the first time that procollagen chains containing C-propeptide mutations that impair assembly are degraded by the cytoplasmic proteasome complex, and that the previously identified endoplasmic reticulum-associated degradation of mutant pro-α1(I) in OI26 is mediated by proteasomes.

The major fibrillar collagens (types I, II, and III) are the principal structural components of the extracellular matrix of many tissues, forming characteristic architecturally precise fibrils (1). They are synthesized as precursor molecules with a central triple-helical region containing a Gly-X-Y amino acid repeat motif, flanked by carboxyl-and amino-terminal propeptide globular domains (for review, see Ref. 2). Assembly of three individual pro-␣-chains to form a triple helix occurs within the endoplasmic reticulum (ER), 1 and is initiated by interactions between the C-propeptides. Triple helix folding then occurs sequentially from the COOH to the NH 2 terminus, and is essential for efficient secretion of the procollagen molecules (3). Mutations in the pro-␣1(I) and pro-␣2(I) chains of type I collagen which compromise initial chain association or disturb the folding of the triple helix result in the brittle bone disease osteogenesis imperfecta (OI) (4 -7) and one of the important biosynthetic consequences of these mutations is an increase in intracellular collagen degradation (7).
Intracellular degradation is an essential process for regulating the levels of many proteins and an important "quality control" mechanism which minimizes the accumulation within cells and the secretion of mutant or malfolded proteins. Several cellular compartments have been identified as sites for degradation, including the lysosomes which contain acid hydrolases, a post-Golgi non-lysosomal compartment, the ER, and the cytoplasm where the 26 S proteasome, a large catalytic protease complex, is responsible for the degradation (8,9). While the molecular basis of intracellular collagen degradation has not been fully defined, three of these four cellular compartments have been implicated as the site of collagen proteolysis. In fibroblasts, approximately 15% of normal procollagen is degraded intracellularly by a process that has been termed basal degradation (10,11), and degradation is significantly increased in cells synthesizing procollagens with structurally abnormal triple helical domains (7,12,13). Degradation of both normal and structurally abnormal procollagen molecules can be inhibited by NH 4 Cl (12)(13)(14), and brefeldin A (15), a drug which causes the cis-and medial-Golgi to fuse with the ER and prevents further intracellular transport. These results suggest that degradation occurs in the distal region of the secretory pathway after the brefeldin A block, in regions that are susceptible to NH 4 Cl inhibition such as the trans-Golgi and secretory vesicles or in the lysosomes. Recent studies have demonstrated directly the localization of procollagen I in the lysosome/endosome system (16) identifying this system as a site of collagen degradation, however, normal procollagen degradation in I-cell disease (Mucolipidosis II) fibroblasts, which are deficient in lysosomal hydrolases, suggests that trans-Golgi and secretory vesicles also represent sites of procollagen degradation (17).
Not all mutant collagen is degraded in the distal region of the secretory pathway. Procollagens with mutations in the pro-␣1(I) C-propeptide which compromise chain association are degraded by a process which is not prevented by brefeldin A, and was therefore assumed to occur within the ER (4). It has recently been shown that a number of soluble and integral membrane proteins that have been translocated into the ER and were thought to be degraded there, are in fact, degraded by the cytoplasmic proteasome complex (18 -23). This process requires reverse transport of protein back to the cytoplasm, a process which may be mediated by interaction of the protein with the Sec61 complex, one of the major constituents of the translocation apparatus (22), and could also involve molecular chaperones (24). In this study we explore the role of proteasomes in the ER-associated degradation of assembly impaired mutant type I procollagen. We examined two pro-␣1(I) C-propeptide OI mutations. The first mutation, an engineered frameshift mutation near the COOH-terminal end of the mouse pro-␣1(I) chain, was analagous to a mutation defined in a patient with OI type I (25). Fibroblasts from that patient contained both mutant and normal pro-␣1(I) mRNA, but mutant protein could not be detected in cells suggesting that it was rapidly and completely degraded prior to assembly (25). In addition, we examine fibroblasts from a patient (OI26) in which a heterozygous frameshift mutation impaired, but did not prevent subunit assembly (4). In both cases the use of specific proteasome inhibitors demonstrated a primary role for cytoplasmic proteasomes in the selective degradation of procollagen chains with mutations within the C-propeptide.

EXPERIMENTAL PROCEDURES
Cell Culture-Dermal fibroblast cultures from controls and a patient with the lethal perinatal form of OI (OI26) were established and maintained in culture as described previously (7). The molecular defect in OI26 is a heterozygous single base insertion in the final exon of COL1A1 (6). The codon reading frameshift generated by the mutation alters the amino acid sequence of the pro-␣1(I) C-propeptide and results in a chain which is 37 amino acids shorter than normal (Fig. 1B). Mouse Mov13 cells (26) were provided by Dr. R. Jaenisch (Whitehead Institute for Biomedical Research, Cambridge, MA). Mov13 cells synthesize no endogenous pro-␣1(I) chains since the transcription of both COL1A1 genes is blocked by a retroviral insertion in the first intron (26,27). Synthesis of pro-␣2(I) is unaffected.
Production of Stably Transfected Mov13 Cell Lines Expressing an Elongated Pro-␣1(I) Chain-The mutant mouse COL1A1 gene construct was a derivative of the previously described control expression construct pWTCI-IA (28). This control construct contains a functionally neutral Met 822 3 Ile substitution 2 within the triple helix that allows the protein to be distinguished from wild-type ␣1(I) by its altered CNBr cleavage pattern (29), and a silent Met 1199 3 Ala substitution within the C-propeptide. A clone was isolated that contained a 2-bp deletion within a ClaI site 69 bases upstream of the translation stop signal (data not shown). This deletion created a codon reading frameshift which would direct the synthesis of a pro-␣1(I) chain 85 residues longer than normal with an anomalous sequence covering the COOH-terminal 109 amino acids (Fig. 1A). The final reassembled 25-kilobase COL1A1 gene construct was named pWTCI-IAfs to indicate the amino acid substitutions in the protein product (the triple helical Met 822 3 Ile substitution, a silent C-propeptide Met 1199 -Ala alteration, and the 2-bp deletion creating a codon reading frameshift). This frameshift mutation is similar to one characterized in a patient with type I OI, where a 5-bp deletion predicted a pro-␣1(I) chain extended by 84 amino acids (25). The mouse gene, pWTCI-IAfs, was stably transfected into Mov13 cells along with pSV2neo. Neomycin-resistant transfected cells were selected in G418 (Life Technologies Inc.), and individual colonies isolated and expanded into cell lines as described previously (28,29).
Collagen Biosynthetic Labeling-Confluent cell cultures were labeled with [ 3 H]proline for 18 h as described previously (7). Following labeling the cell layer and medium fractions were treated separately (7, 32).
Briefly, after disruption of the cell layer by sonication, procollagens and collagens were precipitated from the cell and medium fractions with ammonium sulfate at 25% saturation. The precipitate was redissolved in 2 ml of 50 mM Tris-HCl, pH 7.5, containing 0.15 M NaCl and the protease inhibitors 5 mM EDTA, 10 mM N-ethylmaleimide, and 1 mM phenylmethylsulfonyl fluoride. Aliquots of procollagens were precipitated with 75% ethanol and either analyzed directly or subjected to limited pepsin digestion (100 g/ml pepsin in 0.5 M acetic acid, 4 o C, 16 h) to determine if the collagens had formed pepsin-resistant triple helices.
SDS-PAGE and Immunoblotting-Type I procollagen and collagen chains were analyzed by SDS-PAGE on 5% (w/v) polyacrylamide gels. Where indicated, samples were reduced before electrophoresis by the addition of dithiothreitol to a final concentration of 10 mM. Procollagen chains were also analyzed by two-dimensional gel electrophoresis (35) which resolves the chains on the basis of both charge and size. Radioactively labeled proteins were detected by fluorography. For immunoblotting, procollagen chains resolved by SDS-PAGE were electrophoretically transferred to nitrocellulose filters. Blots were incubated with LF-67 at a dilution of 1/10,000 and bound antibody detected using horseradish peroxidase-conjugated Protein A (Bio-Rad) and an enhanced chemiluminescence kit (ECL, Amersham Pharmacia Biotech).

RESULTS AND DISCUSSION
Stable Expression of the COL1A1 Frameshift Mutant Construct, pWTCI-IAfs, in Mov13 Cells-Mov13 cells are a unique model system in which to study intracellular collagen degradation. Expression of mutant COL1A1 genes in these cells allows the fate of the resultant mutant pro-␣1(I) chains to be easily followed without the complications of endogenous pro-␣1(I) expression (28,29). The mutant mouse COL1A1 frameshift construct, pWTCI-IAfs (Fig. 1), contains a mutation which is similar to one characterized in a patient with type I OI, where a 5-bp deletion predicted a pro-␣1(I) chain extended by 84 amino acids (25). While mutant mRNA was present within the OI cells and could be translated in an in vitro translation system, the protein was not detected in cell culture suggesting that the aberrant protein was rapidly degraded intracellularly. The predicted human and mouse proteins show extensive sequence homology, including a highly positively charged COOH terminus, and might be expected to share similar metabolic fates. To examine the biochemical consequences of the mouse frameshift mutation and address the question of procollagen subunit stability and the targeting of abnormal chains for intracellular degradation raised by the human type I OI mutation, Mov13 cells were transfected with the mutant construct, pWTCI-IAfs. Individual, stably transfected clones were selected in medium containing G418 then screened for expression of pro-␣1(I) mRNA by Northern blot analysis. In contrast to the untransfected Mov13 cells which produced no pro-␣1(I) mRNA (Fig. 2, lane 1), high levels of pro-␣1(I) mRNA of the correct size were apparent in two transfected cell lines, Mov13-IAfs4 and Mov13-IAfs10 (Fig. 2, lanes 5 and 6), and a third cell line, Mov13-IAfs2, contained low levels of pro-␣1(I) mRNA (Fig.  2, lane 4).

Mutant Frameshift Pro-␣1(I) Are Degraded Intracellularly in Transfected Mov13
Cells-To examine the ability of the frameshift mutant pro-␣1(I) mRNA to be translated and the mutant pro-␣1(I) chains to assemble into functional collagen molecules, stably transfected cells were biosynthetically labeled with [ 3 H]proline for 18 h and cell and medium fractions analyzed by SDS-PAGE after digestion with pepsin. Pepsin removes the NH 2 -and COOH-terminal globular domains but leaves the triple helical domain intact. Thus the presence of pepsin-resistant collagen indicates that stable collagen trimeric assembly has occurred. In untransfected Mov cells no pepsin-resistent collagen was present (Fig. 3, lanes 1 and 2) and in cells transfected with the wild-type collagen gene, high levels of pepsin- resistant collagen were present (Fig. 3, lanes 3 and 4). In cells expressing mutant mRNA, only minute amounts of pepsinresistant triple-helical collagen were present (Fig. 3, lanes  7-12). While these data indicate that the mutant procollagen is not able to assemble to form significant amounts of stable collagen trimer, it was important to demonstrate that this small amount of collagen trimer was derived from the transfected mutant gene and not from low-level transcription of the endogenous inactivated COL1A1. The mutant frameshift construct, pWTCI-IAfs, also carried the silent reporter Met 822 -Ile amino acid change within the helix. This marker allows normal and transfected mutant ␣1(I) chains to be distinguished because it deletes a cyanogen bromide cleavage site in the protein (28,29). The presence of the larger distinctive peptide in CNBrdigested pepsin-resistant collagen produced by cells transfected with the frameshift construct confirmed that the ␣1(I) chains were synthesized from the mutant gene and were not the result of low-level transcription of the inactivated endogenous COL1A1 in these clonal Mov13 cell lines (data not shown). From these results we conclude that although the mutant mRNA is transcribed at levels at least as high as that of the control COL1A1 transfectants which produce abundant collagen heterotrimers, only a very small proportion of the mutant pro-␣1(I) chains can associate with endogenous pro-␣2(I) and form triple-helical molecules which can be secreted from the cell.
Despite the accumulation of small amounts of type I collagen in these cultures over the 18-h labeling period, intracellular precursor pro-␣1(I) chains could not be detected by proline radiolabeling and fluorography (data not shown). This failure to detect intracellular frameshift mutant pro-␣1(I) chains using standard radiolabeling procedures indicated that degradation was rapid, almost complete, and the unassembled subunits did not accumulate within the cells. A more sensitive method of protein detection was thus required to allow the fate of the mutant protein to be investigated further. Cell layer proteins produced by untransfected Mov13 cells, and cells transfected with wild-type and mutant COL1A1 genes, were separated by SDS-PAGE, electrophoretically transferred to nitrocellulose, and the filter probed with antibody LF-67. Mov13-IAfs10 cells synthesized a pro-␣1(I) subunit that migrated more slowly than control pro-␣1(I) (Fig. 4, lane 3). This was consistent with the prediction that the mutant protein would be 85 amino acids larger than normal and may be substituted with an additional N-linked oligosaccharide group. The mutant pro-␣1(I) subunits migrated as monomers when analyzed without reduction (Fig.  4, lane 5), indicating that interchain disulfide bonds had not formed and suggesting that trimer assembly is severely impaired.
Proteasome Inhibitors Protect the Frameshift Mutant Pro-␣1(I) Chains from Intracellular Degradation-To determine the site of intracellular degradation, Mov13-IAfs10 cells were treated with several protease and vesicular traffic inhibitors and the relative levels of mutant protein compared. Cells were pretreated for 1 h with inhibitors, metabolically labeled with [ 35 S]methionine for 2 h, chased for 30 min in the presence of inhibitor, and the intracellular pro-␣1(I) chains immunoprecipitated with LF-67. In these untreated cells, mutant pro-␣1(I) was not detected, indicating that complete degradation of the mutant protein produced during the 2-h pulse had occurred within the 30-min chase period (Fig. 5A, lane 1). The mutant subunit was also not detected in cells treated with brefeldin A, which blocks protein transport into the Golgi and results in redistribution of most of the Golgi into the ER (36), monensin, a carboxylic ionophore which blocks intracellular traffic within the trans-Golgi, or NH 4 Cl, which raises the pH of acidic compartments such as the trans-Golgi and lysosomes (Fig. 5A,  lanes 2-4). Since these agents inhibit intracellular vesicular traffic, or inhibit lysosomal function, degradation of the mutant frameshift pro-␣1(I) does not occur in the distal regions of the secretory pathway. This suggests that, as in lethal OI patients with C-propeptide mutations (4), the degradation pathway may be "ER-mediated." Recently, it has become apparent that a number of mutant proteins which have been targeted to the ER and were thought to be degraded within that compartment are, in fact, degraded in the cytoplasm by the ubiquitin-proteasome pathway (18 -  (Fig. 5, lane 5). Incubation with this proteasome inhibitor prevented degradation of the frameshift mutant pro-␣1(I) chain. The identity of the protected protein was confirmed by immunoblotting (Fig. 5B,  lane 2).
To further examine the role of proteasomes in degradation of unassembled mutant procollagen, the effect of two other specific proteasome proteolytic activity inhibitors ZL 3 VS (33) and clasto-lactacystin ␤-lactone (37) were studied (Fig. 6). In these experiments cells were preincubated for 1 h with the inhibitor, labeled for 1 h (Fig. 6A) and chased for up to 1 h (Fig. 6B). Both these proteasome inhibitors also protected mutant procollagen from degradation (Fig. 6A, lanes 7 and 8) while in control cells transfected with wild-type pro-␣1(I) (Fig. 6A, lanes 1-4) similar procollagen band intensities in untreated and proteasome inhibitor-treated cells strongly suggested that normal collagen is not degraded by a proteasomal pathway. Interestingly, two bands which were not present in untreated samples were seen in the presence of the proteasome inhibitors and both were shown to be pro-␣1(I) chains by immunoblotting with LF-67 (Fig. 5B). The larger minor band migrated at the same position as the mutant chains previously detected by immunoblotting (Fig. 4), but the major protected pro-␣1(I) form was somewhat smaller. One possibility is that these two bands represent glycosylated and deglycosylated variants of the mutant pro-␣1(I) protein. Several glycoproteins have been shown to undergo deglycosylation prior to proteasomal degradation, such as the heavy chain of major histocompatibility complex class I molecules (22) and ribophorin I (38). There are two N-linked oligosaccharides addition sites within the mutant pro-␣1(I) C-propeptide (Fig. 1B) which offer possible targets for deglycosylation. However, treatment of the cells during the preincubation and labeling period with tunicamycin did not significantly alter the mobility of either band (data not shown) indicating that the deglycosylation cannot account for the difference in mobility of the two bands. In a pulse-chase experiment in the presence of clasto-lactacystin ␤-lactone the upper band representing the full-length frameshift mutant protein was converted to the smaller species almost completely after a 1-h chase period (Fig. 6B). These data suggest that the smaller form of the pro-␣1(I) is derived from the upper band by the action of a non-proteasomal ER or cytoplasmic protease, and it is the smaller form that is degraded by proteasomes since it is protected from degradation by proteasome inhibitors. Importantly, pro-␣2(I) chains did not coimmunoprecipitate with LF-67 even when the mutant pro-␣1(I) chains were protected from degradation (Fig. 6A, lanes 6 -8). This suggested either that the mutant C-propeptide sequence rendered the chains completely incompetent for assembly, or that the mutant pro-␣1(I) chains were no longer within the ER but had been transported to the cytoplasm by reverse translocation and were therefore unable to interact with pro-␣2(I). Together these data demonstrate that the vast majority of the frameshift mutant pro-␣1(I) chains have an impaired ability to assemble into trimers and are rapidly degraded in transfected Mov13 cells by proteasomes.
In fibroblasts isolated from a patient with type I OI, the pro-␣1(I) chain contains a COOH-terminal frameshift mutation (25), which is very similar to the engineered pro-␣1(I) mutant presented here. mRNA levels for the mutant chain were normal but mutant procollagen chains were not detectable suggesting that the mutant protein was rapidly degraded (25). The similarity of these OI mutant procollagen chains to the engineered pro-␣1(I) mutation expressed in Mov13 cells, suggested that degradation of the OI mutant may also be proteasome-mediated.
OI26 Mutant Pro-␣1(I) Is Also Protected by Proteasome Inhibitors-To determine if proteasomal degradation of procollagen chains with C-propeptide mutations is a general mechanism or is specific to the engineered mutation, we examined the degradation of mutant pro-␣1(I) chains in OI26 cells, derived from a patient with lethal OI. In OI26 a heterozygous frameshift mutation within the C-propeptide results in the synthesis of a truncated pro-␣1(I) containing an altered C-propeptide sequence (Fig. 1B) (4, 6). This mutation slowed, but did not completely prevent, assembly of the mutant chains into triple helical molecules. Mutant unassembled pro-␣1(I) chains were selectively degraded in OI26 cells and this degradation was not prevented by brefeldin A, suggesting an ER-associated degradation pathway (4). Because OI mutations are heterozygous it is normally not possible to discriminate the fates of normal and mutant allele products. However, the frameshift mutation in OI26 results in the synthesis of a more basic pro-␣1(I) chain, and the normal and mutant chains can be resolved by twodimensional gel electrophoresis, which separates the proteins on the basis of both charge and size (4,6). We were able to take advantage of this unique feature of the mutant protein to further explore its intracellular degradation. OI26 fibroblasts were incubated with the proteasome inhibitor ZL 3 al, labeled with [ 35 S]methionine, and the pro-␣1(I) chains immunoprecipitated with LF-67 and resolved by two-dimensional gel electrophoresis. In untreated cells only normal pro-␣1(I) chains were seen, however, mutant and normal pro-␣1(I) chains were present in comparable amounts in samples treated with the proteasome inhibitor (Fig. 7). These data demonstrate that, as in transfected Mov13 cells, pro-␣1(I) chains with a mutant Cpropeptide are degraded by proteasomes in OI26 fibroblasts.
Many plasma membrane and secretory proteins that fail to fold correctly within the ER are selectively degraded by a quality control process known as ER-associated degradation. The 26 S proteasome has been shown to be responsible for the ER-associated degradation of transmembrane proteins such as the cystic fibrosis transmembrane conductance regulator (21), connexin-43 (18), major histocompatibility complex class I heavy chains (22), insulin receptors (23), and T-cell receptor chains (19), as well as the secreted proteins ␣ 1 -antitrypsin (20) and apolipoprotein B (39). The experiments here demonstrate for the first time that proteasomes can also degrade extracellular matrix molecules during biosynthesis, and are involved in the ER-associated degradation of type I collagen pro-␣1(I) chains with mutant C-propeptide domains.
Although the mechanism of mutant procollagen chain removal to the cytoplasm is unknown several ER-resident chaperones have been implicated in mediating proteasomal degradation of abnormal proteins. Mutant insulin receptors associate with heat shock protein 90 prior to proteasomal degradation (23) and unassembled apolipoprotein B associates with calnexin prior to ubiquitination and subsequent proteasomal destruction (39), suggesting multiple components are involved in the recognition and export to the cytoplasm of proteins with abnormal conformations. Recent work implicates the ER-resident molecular chaperone BiP (40 -42), one of several proteins known to be associated with procollagen during fold-ing and assembly, in the process of "reverse translocation" (4,43). Investigation of 20 OI patient cell lines revealed that BiP was associated with type I procollagen in three lines containing C-propeptide mutations that impaired chain association but not with the remaining 17 lines that contained helix mutations and assembled normally (43). Furthermore, we have previously shown that BiP associates with mutant procollagen in OI26 cells and in a second OI cell line that contains a C-propeptide Trp to Cys amino acid substitution (4). In both these lines normal procollagen folding is disturbed and the formation of disulfide-linked trimers is retarded (4,6), suggesting that BiP specifically associates with procollagen chains that contain Cpropeptide mutations and have an impaired ability to assemble but not with chains with mutations in the triple helix which can assemble normally but have abnormal triple helices (43). Procollagen chains with abnormal triple helices are directed to lysosomes for degradation (12,13,16), and our results demonstrate that procollagens with an impaired ability to assemble as a consequence of C-propeptide mutations are degraded by proteasomes. BiP binding may thus play a critical role in recognizing and directing mutant procollagen to proteasomes for degradation. BiP may bind to the misfolded procollagen chains and then either escort them to the protein translocation pore for reverse translocation into the cytoplasm, or transfer them to other proteins that chaperone the abnormal procollagen to the ER membrane.