Protein disulfide isomerase acts as a molecular chaperone during the assembly of procollagen.

Protein-disulfide isomerase (PDI) has been shown to be a multifunctional enzyme catalyzing the formation of disulfide bonds, as well as being a component of the enzymes prolyl 4-hydroxylase (P4-H) and microsomal triglyceride transfer protein. It has also been proposed to function as a molecular chaperone during the refolding of denatured proteins in vitro. To investigate the role of this multifunctional protein within a cellular context, we have established a semi-permeabilized cell system that reconstitutes the synthesis, folding, modification, and assembly of procollagen as they would occur in the cell. We demonstrate here that P4-H associates transiently with the triple helical domain during the assembly of procollagen. The release of P4-H from the triple helical domain coincides with assembly into a thermally stable triple helix. However, if triple helix formation is prevented, P4-H remains associated, suggesting a role for this enzyme in preventing aggregation of this domain. We also show that PDI associates independently with the C-propeptide of monomeric procollagen chains prior to trimer formation, indicating a role for this protein in coordinating the assembly of heterotrimeric molecules. This demonstrates that PDI has multiple functions in the folding of the same protein, that is, as a catalyst for disulfide bond formation, as a subunit of P4-H during proline hydroxylation, and independently as a molecular chaperone during chain assembly.

Protein-disulfide isomerase is now firmly established as a multifunctional protein that both catalyzes the formation of disulfide bonds and acts as a subunit of prolyl 4-hydroxylase and microsomal triglyceride transfer protein (1). The function of PDI 1 as a component of these enzymes appears to be to maintain the catalytic subunits in a soluble form rather than directly participating in catalysis (2,3). In this respect, its function is independent of disulfide isomerase activity (4). More recently, PDI has been proposed to act as a molecular chaperone by binding to unfolded proteins, thereby preventing aggregation (5)(6)(7)(8). This proposal is based on the observation that PDI assists in the refolding of certain denatured proteins in vitro, but this activity appears to be substrate specific, with no activity or even negative (antichaperone) activity being observed with some protein substrates (9 -11). PDI also has been shown to interact with newly synthesized proteins (12,13) and with cysteine mutants of human lysozyme (14), but whether this interaction reflects chaperone activity or the binding of PDI to its substrate during disulfide bond formation still needs to be determined.
PDI is clearly a key cellular folding enzyme that is important for the maturation of several secreted and membrane associated proteins. This is particularly true for the folding and maturation of procollagen, where PDI is involved in a number of key stages. As the polypeptide chain is translocated across the membrane of the endoplasmic reticulum, intrachain disulfide bonds are formed within the N-propeptide and C-propeptide, and hydroxylation of proline and lysine residues occurs within the triple helical domain (15). Chains then associate via the C-propeptides to form homo-or heterotrimeric molecules. This allows the triple helical domain to form a nucleation point at its C-terminal end, ensuring correct alignment of the chains. The triple helix then folds in a C-to N-direction, with the N-propeptides finally associating and in some cases forming interchain disulfide bonds (16). PDI participates during proline hydroxylation as a subunit of prolyl 4-hydroxylase and also catalyzes the formation of both intra- (17) and interchain disulfide bonds (18).
Most of our understanding of how procollagen folds and assembles within the cell has come from studies of cells grown in culture, particularly either skin or tendon fibroblasts. Although this approach has provided us with a clear outline of the intracellular folding and assembly of procollagen, it does not lend itself to a more detailed analysis of the molecular recognition events occurring during assembly. To facilitate these studies, a semi-permeabilized cell system has been developed that reconstitutes the initial stages in the assembly and modification of procollagen as they would occur in an intact cell (19). Using this system, the translocation, disulfide bond formation, and assembly of procollagen into a correctly aligned triple helical molecule has been reconstituted, in a system that mimics the processes as they would occur within an intact cell (20). Here, we have extended these studies to investigate the role of PDI in the folding and assembly of procollagen. We have demonstrated that this protein not only participates in disulfide bond formation and proline hydroxylation but also acts as a molecular chaperone interacting specifically and independently with procollagen chains that remain monomeric, thereby preventing premature assembly or aggregation.

Construction of Recombinant Plasmids-Recombinant p␣1(III)⌬1
and p␣2(I)⌬1 have been described previously (21). Recombinant p␣1(I)⌬1 was generated from COL1A1-CMV (22) by excision of an internal 2.5-kb ApaI fragment and religation of the parental plasmid. An additional nucleotide was inserted by Pfu mutagenesis using the QuikChange mutagenesis kit (Stratagene Ltd., Cambridge, UK) at the ApaI cleavage site to preserve the correct reading frame. Recombinant plasmid constructs were generated by polymerase chain reaction overlap extension using the principles outlined by Horton (23). Polymerase chain reactions (100 l) comprised template DNA (500 ng), oligonucleotide primers (100 pmol each), in 10 mM KCl, 20 mM Tris-HCl, pH 8.8, 10 mM (NH 4 ) 2 SO 4 , 2 mM MgSO 4 , 0.1% (v/v) Triton X-100, 300 M each dNTP. Ten rounds of amplification were performed in the presence of 1unit Vent DNA polymerase (New England Biolabs, Beverly, MA). Recombinant C-propeptide-minus was generated using a 5Ј-oligonucleotide primer (5Ј GATTACGCCAAGCGCGCA 3Ј) complementary to the T3 promoter sequence upstream of the initiation codon in p␣1(III)⌬1 and a 3Ј-oligonucleotide primer (5Ј TCGCTAGGTACCCTAT-TATCCATAATACGGGGCAAAAC 3Ј) complementary to sequence in p␣1(III)⌬1 up to the C-proteinase cleavage site and incorporating a KpnI site to facilitate subsequent sub-cloning. Polymerase chain reaction yielded a 1200-bp fragment that was cut with HindIII and KpnI and subcloned into pBS-SK Ϫ (Stratagene Ltd., Cambridge, UK). Recombinant p␣1(III)⌬1:alt was generated using a 5Ј-oligonucleotide primer (5Ј AATGGAGCTCCTGGACCCATG 3Ј) complementary to a sequence 100 bp upstream of an XhoI site in p␣1(III)⌬1 and a 3Ј amplification primer (5Ј TCGCAGGGTACCGTCGGTCACTTGCACTGGTT 3Ј) complementary to a region 100 bp downstream of the stop codon in p␣1(III)⌬1. A KpnI site was incorporated to facilitate subsequent subcloning. Pairs of internal oligonucleotides, of which one included a 19 nucleotide overlap, were designed to generate a molecule with sequence coding for the B-G region from pro␣2(I) as described previously (25). Overlap extension yielded a product of approximately 1000 bp, which was purified, digested with XhoI and KpnI, and ligated into p␣1(III)⌬1 from which a 1080 bp XhoI-KpnI fragment had been excised.
Transcription in Vitro-Transcription reactions were carried out as described by Gurevich et al. (26). Recombinant plasmids were linearized and transcribed using T3 RNA polymerase (Promega, Southampton, UK). Reactions (100 l) were incubated at 37°C for 4 h. Following purification over RNeasy columns (Qiagen, Dorking, UK), the RNA was resuspended in 100 l of RNase-free water containing 1 mM DTT and 40 units of RNasin (Promega).
Posttranslational Incubations-After 60 min of translation, cycloheximide was added to 5 mM, and samples were incubated for a further time periods up to 60 min at 30°C in the presence or absence of 5 mM Fe(II) sulfate to allow hydroxylation to occur posttranslationally (20). SP cells were isolated by centrifugation at 13,000 ϫ g for 5 min. Pellets were resuspended in KHM buffer prior to subsequent analysis. Samples were prepared for electrophoresis and treated with proteases or the chemical cross-linkers BMH or DSP (Pierce and Warriner Ltd., Cheshire, UK).
Proteolytic Digestion-Isolated SP cells were solubilized in CT/T digest buffer (50 mM Tris-HCl, pH 7.4, containing 0.15 M NaCl, 10 mM EDTA, 1% (v/v) Triton X-100) and centrifuged at 13,000 ϫ g for 5 min to remove cell debris. The supernatant was recovered and then digested with a combination of chymotrypsin (250 g/ml) and trypsin (100 g/ml) (Sigma) for 1 min at 30°C. The reactions were stopped by the addition of soybean trypsin inhibitor (Sigma) to a final concentration of 500 g/ml and acidified by the addition of HCl to a final concentration of 100 mM. Samples were incubated with pepsin (100 g/ml) for 2 h at 30°C. The reactions were stopped by neutralization with Tris base (100 mM) and prepared for electrophoresis as described below.
Chemical Cross-linking-After translation, SP cell pellets were resuspended in a final volume of 50 l of KHM and chemical cross-linkers added from a 50 mM stock (prepared fresh in DMSO) to a final concentration of 1 mM for both DSP and BMH cross-linking experiments. Cross-linking of samples was performed for 10 min at 25°C followed by a further 10 min incubation after addition of 100 mM glycine or 5 mM DTT to quench the DSP or BMH reactions, respectively.
Immunoprecipitation-Cross-linked samples were denatured by boiling for 5 min in SDS/Nonidet P-40 denaturation buffer (25 mM Tris-HCl, pH 7.5, 150 mM NaCl containing 1% (w/v) SDS and 1% (v/v) Nonidet P-40). Insoluble material was removed by centrifugation at 13,000 ϫ g for 10 min, and the supernatant was adjusted to a final volume of 1 ml of immunoprecipitation buffer (50 mM Tris-HCl, pH 7.4, containing 0.15 M NaCl, 10 mM EDTA, 1% (v/v) Triton X-100). Immunoprecipitations were preincubated at 4°C for 40 min in IP buffer containing 50 l of protein A-Sepharose (10% (w/v) in PBS) (Zymed Laboratories Inc., San Francisco, CA), and the samples were centrifuged for 1 min at 10,000 ϫ g to remove protein A-binding components. Supernatants were recovered and made up to a volume of 1 ml with IP buffer. Immunoprecipitation of cross-linked products was carried out at 4°C in the presence of antibodies and 50 l of protein A-Sepharose (10% (w/v) in PBS). The polyclonal antisera to bovine PDI and rat P4-H ␣ subunit were used as described previously (2,27). Immune complexes were retrieved by brief centrifugation (13,000 ϫ g for 30 s) and washed twice in IP buffer, once in IP buffer containing 500 mM NaCl, and finally in IP buffer alone.
SDS-Polyacrylamide Gel Electrophoresis-Samples prepared for electrophoresis by the addition of SDS-PAGE loading buffer (0.0625 M Tris/HCl, pH 6.8, SDS (2% w/v), glycerol (10% v/v), bromphenol blue) in the presence or absence of 50 mM DTT and boiled for 5 min. After electrophoresis, gels were dried, processed for autoradiography, and exposed to Kodak X-Omat AR film.

Assembly of Procollagen Mini-chains in SP Cells-
The main aim of this study was to investigate the role of resident proteins within the endoplasmic reticulum in the folding and assembly of procollagen. To facilitate these studies we constructed a variety of different procollagen "mini-chains" that contain deletions within the triple helical domain. These deletions preserve the Gly-X-Y triplet consensus and are not predicted to alter the folding of the chains. We also prepared a C-propeptide-minus (CP-minus) construct that contains all of the pro␣1(III)⌬1 chain apart from the C-propeptide, the last amino acid being at the C-proteinase cleavage site (Fig. 1). Previous experiments have shown that procollagen mini-chains translated in the presence of SP cells are efficiently translocated, modified, and, in the case of pro␣1(III)⌬1, assembled into a correctly aligned triple helix (20,25).
To assay chain assembly, the procollagen constructs were transcribed in vitro, and the RNA transcript that was synthesized was translated in the presence of SP cells. Translation products were separated by SDS-PAGE under either reducing or nonreducing conditions (Fig. 2). Separation of procollagen chains under reducing conditions demonstrated that single translation products were produced for pro␣1(I)⌬1 and pro␣2(I)⌬1 (Fig. 2, lanes 2 and 3). However, the translation products from the pro␣1(III)⌬1 and the CP-minus RNA transcripts were separated into two polypeptides (Fig. 2, lanes 1   FIG. 1. Schematic diagram representing the domains present in the procollagen chains described in the text. The numbers refer to the number of amino acids in each domain. The numbers in parentheses indicate the sizes of the full-length triple helical regions. and 4). We have shown previously for pro␣1(III)⌬1 that these products represent unmodified and modified forms of the protein (20). The modifications that occur to the procollagen chain include proline and lysine hydroxylation and glycosylation of hydroxylysine residues, which can cause a retardation of electrophoretic mobility (28). The unmodified form is unlikely to have been properly translocated across the ER membrane because it was susceptible to digestion with externally added proteases (results not shown). Because the CP-minus construct contains the same triple helical domain as pro␣1(III)⌬1, the two different translation products are likely to be modified and unmodified forms of the protein. The lack of any apparent mobility change with the pro␣1(I)⌬1 and pro␣1(I)⌬1 chains could reflect the lower content of potential hydroxyproline and hydroxylysine residues within the truncated helical domains in these chains.
When the same samples were separated under nonreducing conditions, higher molecular weight molecules were detected with pro␣1(III)⌬1 (Fig. 2, lane 5) and to a certain extent with pro␣1(I)⌬1 (Fig. 2, lane 6) translation products. We have demonstrated previously that for pro␣1(III)⌬1 this represents a correctly folded triple helical molecule with interchain disulfide bonds stabilizing the structure in both the N-and C-propeptides (20). When a similar analysis was carried out on the pro␣2(I)⌬1 and CP-minus constructs, no interchain disulfide bonded trimers were seen (Fig. 2, lanes 7 and 8, respectively). However, a faster migrating polypeptide was observed (Fig. 2, compare lane 3 with lane 7 and lane 4 with lane 8). This is due to the formation of intrachain disulfide bonds, which result in a more compact structure with a faster mobility than the fully reduced polypeptide. These results are in agreement with previously published results for pro␣2(I)⌬1 (25) and are as expected for the CP-minus construct because the C-propeptide mediates the initial interaction of the procollagen chains. These and previous published results demonstrate that the pro␣1(III)⌬1 chain efficiently forms trimers, the pro␣1(I)⌬1 chain forms some homotrimers at low efficiency, and the pro␣2(I)⌬1 and CP-minus chains do not form trimers.
Interaction of Prolyl 4-Hydroxylase with Unfolded Procollagen Chains-Previous work on the substrate specificity of P4-H has demonstrated that the conformation of the procollagen triple helix determines enzyme activity (15). Thus, the enzyme will readily hydroxylate unhydroxylated procollagen chains as long as the chains have not formed a triple helix (29). It has also been shown that malfolded chains may be isolated as stable complexes with P4-H (30). This suggests that in addition to its enzymatic role, P4H may interact specifically with unfolded procollagen chains in a chaperone-like manner.
The approach we adopted to investigate this possibility was to allow the hydroxylation of newly synthesized procollagen chains to occur posttranslationally and assay the formation of a stable triple helix. We then determined whether P4-H or any other ER protein was associated with the folding chains. Translation reactions were carried out in the presence of ␣,␣Ј-dipyridyl, an iron chelator that is a potent inhibitor of prolyl 4-hydroxylase, for 60 min. Ferrous iron was added in excess of the chelator, thereby activating P4-H and allowing hydroxylation and triple helical formation to occur. We have used such a protocol previously and demonstrated that pro␣1(III)⌬1 chains can fold to form correctly aligned triple helices within 60 min under these conditions (20). The results presented here (Fig.  3A) clearly show that after 60 min of translation in the presence of ␣,␣Ј-dipyridyl, no modification to the pro␣1(III)⌬1 chain occurred, as judged by a lack of the appearance of a slower migrating polypeptide (Fig. 3A, compare lanes 1 and 2). How-  4 . SP cells were isolated by centrifugation, and translation products were separated on a 7.5% reducing gel. B, RNA encoding pro␣1(III)⌬1 was translated as described for A. After posttranslational incubation with iron (as described above) SP cells were resuspended in CT/T digest buffer and treated with a combination of chymotrypsin (250 g/ml) and trypsin (100 g/ml). The reactions were stopped by addition of soybean trypsin inhibitor and acidified by the addition of HCl to a final concentration of 100 mM. Samples were then incubated with pepsin (100 g/ml) for 2 h at 30°C and neutralized by addition of Tris base (100 mM). Digestion products were separated by SDS-PAGE under reducing conditions on a 12.5% gel. Lane 1 is the control sample incubated without ␣,␣Ј-dipyridyl; lanes 2 and 3 contain the samples incubated in the absence and presence, respectively, of FeSO 4 . ever, when the incubation was continued for 60 min in the presence of excess iron, chains were modified, as judged by the appearance of a polypeptide with a slower mobility (Fig. 3A,  lane 3). This posttranslational modification correlated with the formation of a protease resistant triple helical fragment (Fig.  3B, lane 3). This experimental system allows us to accumulate unhydroxylated procollagen chains, which can be hydroxylated posttranslationally. For pro␣1(III)⌬1 chains, these can fold to form triple helical molecules. Because these reactions take place in a functionally intact ER in the presence of the endogenous complement of ER resident proteins, we can determine which of these proteins are interacting with either unhydroxylated or triple helical chains during hydroxylation and folding.
To investigate which proteins were interacting with procollagen chains, we used the amine-specific, thiol-cleavable, bifunctional cross-linking agent DSP. We cross-linked procollagen chains to interacting proteins after 60 min of translation in the presence of ␣,␣Ј-dipyridyl and after various times up to 60 min after addition of exogenous iron. We also incubated the translation products in the absence of exogenously added iron and cross-linked to control for any degradation that may have occurred during posttranslational incubation. We then immunoprecipitated the cross-linked products with a variety of antibodies to known resident proteins of the ER, including Hsp47, PDI, Erp57, calnexin, calreticulin, and the ␣-subunit of P4-H. We then separated the immunoprecipitated products under reducing conditions; cross-linking to procollagen chains would be indicated by the appearance of a radiolabeled polypeptide after electrophoresis. We also confirmed that all the antibodies used for immunoprecipitation could still recognize their target protein after cross-linking (results not shown). Only two antibodies immunoprecipitated cross-linked products, PDI (␤-subunit of P4-H) and the ␣-subunit of P4-H antibodies (Fig. 4). All of the other antibodies tested did not immunoprecipitate crosslinked products (results not shown). Interestingly, the amount of cross-linked product immunoprecipitated with the P4-H antibodies was dependent upon the folding status of the procol-lagen chains. Thus, P4-H was bound most abundantly to pro␣1(III)⌬1 chains that were unhydroxylated (Fig. 4, lanes 1,  5, 6, and 10). Once the chains were hydroxylated and folded to form a triple helix, most of the P4-H dissociated (Fig. 4, lanes  1-4 and 6 -9). Such a transient association of P4-H with procollagen chains suggests an interaction with unhydroxylated chains followed by a dissociation of P4-H following hydroxylation and folding. The decrease in the interaction of P4-H with the procollagen chains was hydroxylation dependent and was not due to degradation because an equivalent amount of material was cross-linked and immunoprecipitated when the samples were incubated for 60 min posttranslationally in the absence of added iron (Fig. 4, lanes 5 and 10). These results clearly show that P4-H binds to the unhydroxylated chains and will, therefore, prevent the triple helical domain from interacting prior to the formation of a stable triple helix.
Interaction of PDI with Monomeric Chains-Having established that P4-H interacts with the triple helical domain of procollagen chains during their assembly, we then wanted to address the possibility that ER resident proteins may interact with the C-propeptide of chains that remain monomeric. For these experiments, we expressed the individual chains of type I procollagen, which remain monomeric prior to the formation of heterotrimeric molecules. We translated the pro␣1 and pro␣2 mini-chains of type I and the pro␣1 mini-chain of type III procollagen individually and then added the thiol-specific, noncleavable, bi-functional cross-linking reagent BMH. BMH was used instead of DSP for these studies to identify proteins interacting at regions of the protein other than the triple helical domain. When separated under reducing conditions, the appearance of a radiolabeled higher molecular weight product would indicate cross-linking of procollagen to the immunoprecipitated protein. Clear cross-linked products were seen for the pro␣2(I)⌬1 chain (Fig. 5, compare lanes 3 and 6). More diffuse cross-linked products were also seen for the pro␣1(III)⌬1 and pro␣1(I)⌬1 chains (Fig. 5, compare lane 1 with lane 4 and lane 2 with lane 5). After immunoprecipitation, only the PDI antibody was able to precipitate cross-linked products (Fig. 5, lanes  8 and 9). The cross-linked products from the pro␣1(I)⌬1 and pro␣2(I)⌬1 translations were resolved as distinct bands after immunoprecipitation (Fig. 5, lanes 8 and 9). The type III pro␣chain cross-linked products were not immunoprecipitated by the PDI antibody (Fig. 5, lane 7). We investigated this result further by expressing just the C-propeptide of type III procollagen (24) and cross-linking with BMH. Here again, no crosslinks to PDI were observed (results not shown). These results suggest that PDI was able to interact with type I procollagen chains but not type III procollagen chains.
It is not clear from these results whether PDI interacts independently with monomeric procollagen chains or as a subunit of P4-H. To address this point, we used BMH to cross-link pro␣-chains under conditions that result in P4-H binding, i.e. after translation in the presence of ␣,␣Ј-dipyridyl. It was clear from the results (Fig. 6) that even under conditions that result in binding of P4-H to pro␣1(III)⌬1, no immunoprecipitation of BMH cross-links occurred (Fig. 6, lanes 2 and 4). Cross-linking to the pro␣2(I)⌬1 chain was unaffected by the presence of ␣,␣Ј-dipyridyl demonstrating that the interaction of PDI is independent of hydroxylation. We also cross-linked the CPminus construct with BMH under conditions where P4H remains bound to this chain, but no cross-link products were observed (results not shown). These results suggest that the BMH cross-links that we observed are a consequence of PDI interacting with procollagen chains independently and not in a complex with P4-H.
There are two explanations for the lack of PDI cross-links to FIG. 4. Cross-linking of pro␣-chains and immunoprecipitation with antisera raised to PDI and P4-H ␣-subunit. RNA encoding the pro␣1(III)⌬1 chains was translated in the presence of SP cells. Samples were translated in the presence of ␣,␣Ј-dipyridyl and then incubated for various times as indicated with (lanes 1-4 and 6 -9) or without (lanes 1 and 10) addition of 5 mM FeSO 4 . Cells were isolated by centrifugation and resuspended in 50 l of KHM buffer, and DSP was added to a final concentration of 1 mM. Cross-linking was quenched by the addition of 100 mM glycine, and the samples were denatured by boiling in SDS/ Nonidet P-40 denaturation buffer. After removal of insoluble material and preclearing, samples were divided equally and immunoprecipitated overnight with antiserum raised to P4-H ␣-subunit or ␤-subunit (PDI). DSP cross-links were cleaved by addition of SDS-PAGE buffer containing 50 mM DTT to immune complexes. Pro␣1(III)⌬1 chains were resolved through 7.5% SDS-PAGE gels. pro␣1(III)⌬1 chains. Either the cross-links only occur with monomeric chains or there are sequence differences between the various chains that preclude cysteine specific cross-links to this chain. To investigate this point, we constructed a pro␣1(III)⌬1 chain that we predicted would remain monomeric. We have recently identified the sequence of amino acids within the C-propeptide that are responsible for selective association of procollagen chains (25). Hence, we replaced the pro␣1(III) sequence that directs homotrimer assembly with the corresponding sequence from the pro␣2(I) C-propeptide. The resulting molecule, which we called pro␣1(III)⌬1:alt, has exactly the same sequence as pro␣1(III)⌬1 apart from a stretch of 23 amino acids within the C-propeptide. As predicted, this chain, unlike pro␣1(III)⌬1, was unable to associate to form an interchain disulfide-bonded trimer when translated in the presence of SP cells (Fig. 7A, compare lanes 3 and 4). The synthesized product did show an increase in electrophoretic mobility when separated under nonreducing conditions, indicating that the intrachain disulfide bonds had formed. When the translation products were cross-linked with BMH and immunoprecipitated with antibody to PDI, clear PDI-procollagen products were formed (Fig. 7B, lane 3). Because the only difference between this chain and pro␣1(III)⌬1 is that this chain remains monomeric and has a slightly altered C-propeptide, we can deduce that PDI interacts at the C-propeptide either specifically with monomeric chains or via the specific altered sequence. This interaction may serve to chaperone the monomeric procollagen chains prior to their assembly into trimers. DISCUSSION The biosynthesis of multisubunit proteins entering the secretory pathway is regulated at the ER, where the individual subunits are synthesized and their assembly is coordinated. This regulation ensures that unassembled subunits are prevented from being transported out of the ER and are either degraded or maintained in an assembly competent state by interacting with ER resident proteins (31). The mechanism underlying this "quality control" appears to involve the binding of unassembled subunits to a variety of ER proteins until assembly occurs. The assembled complex is then released and can be transported from the ER. Such a mechanism has been likened to affinity chromatography, with the "matrix" being the resident proteins in the ER and the selective interactions occurring via oligosaccharide side chains (32), hydrophobic regions in the protein (33), of free thiol residues (34) or with specific molecular chaperones (35,36). This study has demonstrated that protein-disulfide isomerase also plays a role in this regulation by binding to procollagen chains either as a subunit of prolyl 4-hydroxylase or independently.
The procollagen trimer is folded and assembled through a series of distinct intermediates; the coordination of this assembly is crucial to produce a correctly folded, thermally stable triple helical molecule. One consequence of the sequence of events occurring during procollagen folding is that the individual chains have to be maintained in a soluble form prior to assembly occurring. The triple helical domain is inherently insoluble and must be prevented from self-association for several minutes, (37) because the chains associate at their Cpropeptides and the triple helical domain folds in the C to N direction (16). We show here that P4-H plays a key role in ensuring that the triple helical domain remains soluble by binding to unhydroxylated chains. The binding of P4-H to its substrate in the absence of the co-factors iron and ascorbate has been shown previously using purified proteins (38). Our observation that this interaction also occurs during biosynthe-  3 and 4) of ␣,␣Ј-dipyridyl. Cross-linking and immunoprecipitation was carried out with BMH as described, and samples were separated on a 7.5% reducing gel. sis within the ER provides convincing evidence that enzymes involved in posttranslational modification can under certain circumstances also play a crucial role in maintaining polypeptides in an assembly competent state. That P4-H can also bind to hydroxylated chains provides a clue to how procollagen chains hydroxylated at the N-terminal end of the triple helical domain are prevented from associating prior to folding and association of the C-propeptides. Interestingly, mutant procollagen chains, which contain a deletion in their triple helical domain and which form trimers with wild type chains, have also been shown to form a stable interaction with P4H (30). Such a stable interaction has been suggested to prevent secretion of these molecules by retention in the ER.
Monomeric procollagen chains that are destined to be incorporated into heterotrimers also need to remain soluble and be prevented from nonpermissive associations prior to assembly. When we expressed the type I pro␣-chains individually, we observed an interaction of the C-propeptides of these chains with PDI. The pro␣1 chain of type I procollagen has previously been shown to be able to form homotrimers at a low efficiency (39). This was also the case here, with the majority of the synthesized chains remaining monomeric. The pro␣1 chain of type III efficiently form homotrimers and, significantly, did not interact with PDI at its C-propeptide. However, when this chain was altered to prevent association by changing its recognition site to that of the pro␣2-chain of type I, an interaction with PDI could be detected. We and others have previously shown that point mutations within the C-propeptide can cause misfolding of this domain (21), which leads to binding to ER proteins, such as immunoglobulin heavy chain-binding protein (BiP) (40). However, in these cases, the pro␣-chains were unable to form correct intrachain disulfide bonds and migrated as a diffuse smear when separated under nonreducing conditions. The pro␣-chain we have constructed here with an altered recognition site was able to form intrachain disulfide bonds. The polypeptide synthesized migrated as a sharp band when separated under nonreducing conditions, which co-migrated with the wild type protein. This indicates that the C-propeptide folded correctly but was unable to assemble due to its altered recognition site. These results clearly demonstrate that PDI plays a crucial role in binding to the C-propeptide, thereby coordinating heterotrimer assembly.
A growing body of evidence is accumulating that suggests a key role for the collagen binding protein HSP47 in the biosynthesis of procollagen (36). Co-immunoprecipitation experiments have shown that procollagen within the ER is associated with HSP47 (41) and can dissociate upon transport to the Golgi apparatus (42). It has been reported that HSP47 binds to the N-propeptide of type I procollagen (43), yet other workers have also reported binding to the triple helical domain (42). During our studies, we were unable to immunoprecipitate our procollagen constructs with anti-HSP47 antibodies after cross-linking. It could be that our shortened triple helical domains do not contain the binding site for this molecule or could simply reflect inefficient cross-linking. The folding and assembly of procollagen is a complex process and may require a number of different chaperone proteins to ensure efficient folding, assembly, and intracellular transport. There is also likely to be redundancy in the involvement of accessory proteins, as has been illustrated by the successful expression and assembly of procollagen in insect cells (44), which are unlikely to contain HSP47. Clearly, PDI, P4H, and HSP47 may have overlapping functions as molecular chaperones during procollagen biosynthesis. The interactions described here provide the first direct evidence for a chaperone role for PDI (and P4-H) during assembly of procollagen chains by preventing their premature and hence nonproductive interaction. The interaction of procollagen intermediates with PDI may also, by virtue of its KDEL-retention sequence, provide a mechanism for retention of non-triple helical procollagen chains.