The Molecular Interactions of Heat Shock Protein 47 (Hsp47) and Their Implications for Collagen Biosynthesis*

Heat shock protein 47 (Hsp47) is a procollagen/colla-gen-specific molecular chaperone protein derived from the serpin family of proteins and essential for the early stages of collagen biosynthesis. In this paper, the results of an extensive biophysical analysis of mature recombinant mouse Hsp47 show the existence of both a struc-turally mesostable monomer with a 5-strand A-sheet and/or a hyperstable trimer; both states have biological activity. It is also demonstrated that Hsp47 is able to bind to a monomeric and partially folded conformation collagen mimic peptide (PPG) 10 . Upon peptide binding Hsp47 has the capacity to induce the peptide backbone to fold into a polyproline type II conformation. Induction of this conformation results in (PPG) 10 peptides associating into higher order assemblies with increased stability compared with the monomeric peptide alone. These assemblies are similar to those observed by others when the peptide is dissolved at high and A homology model of mrmHsp47 was constructed using the program MODELLER (41). A basis set of homologous structures was constructed with reference to alignment of the mrmHsp47 sequence (45) against a data base of known structures (ovalbumin (code 1OVA), human plasminogen activator inhibitor II (code 1BY7), and human (cid:1) 1 -antitrypsin (code 1QLP)). These structures were combined with multiple sequence alignment data obtain using ClustalW (42) in conjunction with the primary amino acid sequence of mrmHsp47. These data were then used by the MODELLER routines to produce a homology model.

The formation of collagenous fibrous tissue is a vital part of the process of wound healing. However, excessive collagen formation interferes with normal tissue function and can give rise to pathological fibrosis. Collagen is a super polymer made up of fibers containing large numbers of well organized cross-linked trimeric peptides. These assemblies have an enormous tensile strength and are the major insoluble, fibrous component of the extracellular matrix. Collagen is made up of three polypeptide pro-␣ chains, coiled together to form a triple helix (4 -6). Peptides in the trimeric unit are ϳ1050 residues in length and associate by formation of a right-handed triple helix. This con-formation is stabilized by a large number of Gly-Pro-X (where X represents any amino acid) repeats.
Nascent monomeric procollagen chains are synthesized with N-and C-terminal propeptides. Collagen assembly begins with the folding and association of C-terminal propeptides into trimeric assemblies (7) within the endoplasmic reticulum. This initial docking is aided by the formation of disulfides and is modulated by the presence of protein-disulfide isomerase (8,9). This association ensures that the proline-rich regions are close enough together for efficient and accurate procollagen triple helix formation. Trimeric procollagen units are then exported to the cell surface via the Golgi apparatus. Once in the Golgi the collagen trimers are able to associate to form procollagen fibrils (10). Throughout the synthesis and processing of collagen, the presence of proteins (11), such as heat shock protein 47 (Hsp47) 1 and prolyl 4-hydoxylase (12,13), appears to play a vital role assisting and/or controlling trimer assembly.
Hsp47 (also known as colligin or J6 protein), is described as a 47-kDa molecular chaperone protein that is constitutively expressed in collagen-synthesizing cells (14) and binds specifically and only to procollagen/collagen (15,16). There have been a number of roles and functions proposed for Hsp47 that include interacting with nascent procollagen polypeptide chains (17), unhydroxylated non-triple helical chains, and even hydroxylated helical trimers (18,19) in the endoplasmic reticulum. Furthermore, recent work by Tasab and co-workers (16) has demonstrated that Hsp47 will bind to fully formed helical trimers in vivo, perhaps thereby stabilizing the triple helix and preventing chain dissociations. This functional hypothesis is supported by a recent demonstration that Hsp47 binds specifically to stretches of non-hydroxylated Pro-Gly-X repeats (20), regions of low structural stability within the full-length collagen triple helices. After binding there is clear evidence that triple helical procollagen-Hsp47 complexes are eventually transported from the endoplasmic reticulum to cis-Golgi (21), where Hsp47 is dissociated and recycled back to the endoplasmic reticulum (18).
Although these observations make a compelling case for Hsp47 as a procollagen/collagen-specific binding protein, there is little evidence so far to suggest how Hsp47 might actually function as a molecular chaperone in these early stages of collagen biosynthesis. This early role must be significant. Increased Hsp47 expression has been linked to a number of diseases in which increased collagen deposition occurs, including arteriosclerosis (22,23), myocardial infarction (24), and hepatic (25), renal (26), and pulmonary fibrosis (27). Indeed Sunamoto et al. (28) successfully reduced the progression of glomerulonephritis in a rat model by reducing Hsp47 expression.
Hsp47 is an unusual heat shock protein belonging to a protein family that is very well known, namely the serine protease inhibitor (serpin) family. Serpin family members have a highly conserved secondary structure that is made up of a core of three ␤-sheets surrounded by nine ␣-helices (Fig. 1a) (29,30,34). The central ␤-sheet (A-sheet) typically exists in a 5-or 6-stranded form. The 5-stranded form occurs as a result of normal folding but is metastable. Proteolytic cleavage or partial protein denaturation leads to the formation of the 6-stranded form that is hyperthermostable (Fig. 1b). Work on the characteristics of the metastable nature of the serpin fold has shown that the 5-stranded A-sheet may also be filled by exogenous peptides to form a binary complex with a heterologous 6-stranded A-sheet (3) (Fig. 1c). In addition, serpins have the ability to self-assemble via a number of ␤-strand/␤-strand interactions to form homopolymers (32)(33)(34)(35). These polymers vary in biophysical characteristics depending upon the type of linkage. Polymers of ␣ 1 -antitrypsin are highly stabilized due to a linkage involving the filling of the center of ␤-sheet A (36, 37) (Fig. 1d). These polymers are pathogenic, and their accumulation within the liver leads to liver cirrhosis (32).
We describe the results of an intensive biochemical investigation of mature recombinant mouse Hsp47 (mrmHsp47).

Materials
Unless otherwise stated, all reagents were obtained from BDH Chemicals, Ltd., and Sigma. CNBr-activated Sepharose, MonoS ion exchange, and Superdex S200 column were purchased from Amersham Pharmacia Biotech. Collagen was provided from Cellon S.A., Strassen, Luxembourg. Active MrmHsp47 (from recombinant Escherichia coli DHlhind-pKS26 incorporating the mrmHsp47 gene) was a generous gift from Professor Kazuhiro Nagata (Department of Molecular and Cellular Biology, Kyoto University, Japan). The peptide (PPG) 10 was synthesized by Peptides International. Anti-MrmHsp47 monoclonal antibody was purchased from Stressgen Biotechnologies, Canada.

Purification of mrmHsp47
Recombinant E. coli DHlhind-pKS26 incorporating the mrmHsp47 gene was grown at 32°C in LB medium containing ampicillin (50 g/ml). The cells were then heat-shocked at 42°C and harvested by centrifugation (5000 rpm, 15 min) followed by resuspension in Tris-HCl, pH 8.0, containing 5 mM EDTA, 150 mM NaCl, and 30% (v/v) glycerol. Pepstatin and aprotinin were added along with lysozyme and Igepal (non-ionic detergent, 2.45 ml of 10% (v/v) solution in water). Cells were stirred for a further 60 min. The resulting suspension was sonicated for a total of 8 min (2-s pulses and 7-s intervals) at 4°C. Cell debris was pelleted by centrifugation (15,000 rpm, 1 h) at 4°C, and the supernatant was applied to a collagen-Sepharose 4␤ column at 4°C, previously equilibrated with 50 mM Tris-HCl, pH 8.0, containing 5 mM EDTA and 150 mM NaCl. The column was washed with 50 mM Tris-HCl, pH 8.0, containing 5 mM EDTA and 400 mM NaCl to remove any nonspecifically bound material. mrmHsp47 was eluted with 50 mM Tris-HCl, pH 5.8, containing 5 mM EDTA and 150 mM NaCl. Fractions containing mrm-Hsp47 were combined, and the pH was readjusted to pH 8.0. The protein was loaded onto a Superdex 200 gel filtration column, and the two peaks containing mrmHsp47 were collected. The oligomeric variants were also separated using a MonoS HR 10/10 column equilibrated in 50 mM Hepes, pH 8.0, containing 10 mM NaCl and 5 mM EDTA. mrmHsp47 was eluted with a 120-ml linear gradient of 10 -1000 mM NaCl, in 50 mM Hepes, pH 8.0, containing 5 mM EDTA. Protein preparations were stored at Ϫ80°C.

Native, Denaturing PAGE and Transverse Urea Gradient Electrophoresis Gels
Native PAGE-Samples were dissolved in NOVEX (UK) Tris glycine native sample buffer. Native PAGE was carried out on 14% Tris glycine homogenous gels run for 12 h at 4°C, using Tris glycine native running buffer (25 mM Tris base, pH 8.3, containing 192 mM glycine).
Denaturing PAGE-Samples were dissolved in 2% SDS sample buffer and heated at 90°C for 5 min before loading onto the gels. SDS-PAGE was carried out using 15% Tris glycine homogenous native gels. The gels were run at 125 constant voltage for 90 min at room temperature, using Tris glycine SDS running buffer (25 mM Tris base, pH 8.3, containing 192 mM glycine and 0.1% SDS).
Transverse Urea Gradient Gel Electrophoresis-6% w/v polyacrylamide gels were cast with using a gradient maker to give a linear gradient from 0 to 8 M urea left to right across the gel. The gels were made using a high pH non-denaturing PAGE buffer system (192 mM glycine, 25 mM Tris-HCl, pH 9.5). The proteins were visualized by staining with Coomassie Brilliant Blue or by silver staining.
Western Blotting-Proteins were transferred overnight from polyacrylamide gels to nitrocellulose (38). Nitrocellulose blots were probed with monoclonal mouse IgG 2b mrmHsp47-specific primary antibody and A0168 monoclonal anti-mouse IgG-horseradish peroxidase, conjugate peroxidase, and secondary antibody.

Size Exclusion Chromatography
Size exclusion chromatography was carried out using a Superdex 200 resin (Amersham Pharmacia Biotech) in an XK16/60 column with a total bed volume of 120 ml. The separation was performed using a flow rate of 1 ml/min, and samples were loaded (50 mM Hepes, pH 8.0, containing 150 mM NaCl and 5 mM EDTA) in a total volume not exceeding 2 ml.

Measurement of Intrinsic Fluorescence
Fluorescence measurements were made using a PerkinElmer Life Sciences LS 50B spectrofluorimeter using a method modified from Ref. 39. The solution was excited at a wavelength of 295 nm; photons emitted at 90°to the excitation beam were detected at 340 nm. Wherever possible the slits controlling the intensity of the excitation light source were kept at the minimum machine-permissible limit of 2.5 nm; any other values for these slit widths were as detailed in the text. Emission slit widths were varied between 2.5 and 15 nm dependent on the experimental conditions to give an optimal emission signal. The samples (in 20 mM Tris, pH 7.4) were incubated in a 0.5-ml cuvette with a path length of 1 cm on the excitation axis and 0.2 cm on the emission axis, and throughout all experiments the sample temperature was maintained at 25°C by a heated water jacket within the cuvette holder. The data were fitted to a single site weak binding hyperbola function using Grafit (version 3.00, 1992, Erithracus Software Ltd.).

Circular Dichroism
CD measurements were performed using a Jasco J-810 and J-720 spectropolarimeter in 50 mM phosphate buffer, pH 7.4, at 20°C using a 0.05-and 0.1-cm path length quartz cuvette for wavelength scans and titrations, respectively. All scans were performed over 20 repetitions with the presented data being the average. Melting points were obtained by monitoring the change in dichroic signal at a single wavelength, whereas the temperature of the sample was increased from 25 to 95°C at a constant rate of 1°C/min. Melting points were calculated using the method of Ref. 40 using an expression for a two-state transition (2) as shown in Equation 1, where obs is the observed CD signal, and N and D represent the native and denatured intercepts on the y axis, respectively. T m is the melting point, and ⌬H M is the Van't Hoff enthalpy change for unfolding, and m is a correction factor for linear base-line drift.

Differential Scanning Calorimetry
Calorimetry measurements were carried out on a Microcal VP-DSC instrument in 0.5156-ml cells at a heating rate of 20°C h Ϫ1 unless otherwise indicated. Protein concentration varied from 0.5 to 1.2 mg/ml. To analyze functions of excess heat capacity (⌬C p ), Microcal Origin 5 software package was used. The samples were prepared in 50 mM Tris-HCl, pH 8.0, containing 5 mM EDTA and 150 mM NaCl. Aliquots of the sample buffer were used as a DSC reference and for base-line corrections. Sample and reference solution were degassed for a total of 8 min under vacuum with gentle stirring before being loaded and were kept under 26 -27-mPa pressure during DSC to inhibit degassing and bubble formation at higher temperatures.

Homology Model Building
A homology model of mrmHsp47 was constructed using the program MODELLER (41). A basis set of homologous structures was constructed with reference to alignment of the mrmHsp47 sequence (45) against a data base of known structures (ovalbumin (code 1OVA), human plasminogen activator inhibitor II (code 1BY7), and human ␣ 1 -antitrypsin (code 1QLP)). These structures were combined with multiple sequence alignment data obtain using ClustalW (42) in conjunction with the primary amino acid sequence of mrmHsp47. These data were then used by the MODELLER routines to produce a homology model.

RESULTS
Purification of mrmHsp47-mrmHsp47 was initially purified by collagen-affinity column chromatography from a recombinant strain of E. coli (DH1ind Ϫ /pKS26) after heat shock induction (43,44). According to SDS-PAGE analysis, homogeneity was judged to be Ͼ98% (Fig. 2a, lane 3). However, we were concerned that contaminant peptides might be present (from collagen or else E. coli debris) that might complicate future biophysical analyses. Therefore, mrmHsp47 previously eluted from the collagen-affinity column (mrmHsp47c) was subjected to high resolution size exclusion chromatography ( Fig. 2b), revealing that contrary to initial expectations, mrmHsp47c was in fact a mixture of two molecular species, a monomer (ϳ45 kDa) and a trimer (ϳ147 kDa). The presence of mrmHsp47 in each of the two oligomeric species was confirmed by native gel/Western blotting analysis (Fig. 2c). Cation (Mo-noS) exchange chromatography was then used to separate monomeric Hsp47 from mrmHsp47c, providing a more rapid and convenient purification procedure. Previous work (20, 44) with mrmHsp47c has always been conducted on the basis that mrm-Hsp47 was only monomer in character. Therefore, the existence of the trimeric state of mrmHsp47 is an important revelation. The possibility that the trimeric form was merely an artifact caused by intermolecular disulfide linkage was rejected because the mrmHsp47 construct used contains just one cysteine (45). The absence of the cysteine at position 10 was confirmed by N-terminal sequencing (data not shown).
Monomeric mrmHsp47-Members of the serpin protein family exist in a large number of conformational sub-forms. For this reason we elected to characterize the behavior of monomeric mrmHsp47s. The CD spectrum of mrmHsp47s showed the protein to be a typical ␣/␤-protein consistent with the known secondary structure composition of other serpin family members (Fig. 3a). The stability of mrmHsp47s was assessed using a number of orthogonal techniques. Changes in the intensity of the ␣-helical CD minimum (⌬A 222 ) were observed as a function of temperature and found to undergo a sharp transition with midpoint inflection at 54.7°C (Fig. 3b). A similar sharp transition was also observed when mrmHsp47s were analyzed by differential scanning calorimetry (DSC). In this case the transition maximum was found to be 52.4°C (Fig. 3c). Such a sharp transition is highly indicative of a member of the serpin family in a metastable form (see Fig. 1a). This corresponds to a conformational state in which the main ␤-sheet (A-sheet) of the serpin fold is composed of only five strands and an optional sixth strand is missing, expelled to form a surface loop. In previous work (45), we had suggested that monomeric mrmHsp47 may reside primarily in a conformation comparable with that of latent inhibitory plasminogen activator inhibitor (PAI-1) (Fig. 1e) that possesses an A-sheet with six strands. However, the CD and DSC data provide categorical evidence that mrmHsp47s is in the metastable native state with a 5-stranded A-sheet (Fig. 1a).
CD and DSC experiments were followed by experiments involving fluorescence spectroscopy and transverse urea gradient PAGE. mrmHsp47 has 5 tryptophan residues and therefore has significant intrinsic tryptophan fluorescence. When mrmHsp47s were titrated with urea, the tryptophan fluorescence emission maximum, I 340 , was observed to undergo a progressive red shift (indicative of a progressive increase in tryptophan residue solvent accessibility) as a function of urea concentration. This transition was divisible into two transitions (Fig. 3d). The first transition was observed to occur between 1 and 3 M urea (9 nm red shift) and the second between 6 and 8 M urea (8 nm red shift). This first transition is consist- ent with an early unfolding event, leading to an intermediate state, and the second transition with a late unfolding event, leading to random coil. Similar transitions have been observed in urea titrations involving other members of the serpin family, most notably ␣ 1 -antitrypsin (46). These two consecutive unfolding events were observed independently in a comparable urea titration experiment performed with CD spectroscopy. In this case, a progressive increase in ⌬A 222 (indicative of a progressive loss in secondary structure) was observed to occur as a function of urea concentration, and this progressive increase was also divisible into two main transitions equivalent to those seen by fluorescence spectroscopy (Fig. 3d). Visualization of protein volume changes that occur during unfolding were achieved using transverse urea gel PAGE. These data show a change in volume occurring only at the second transition above 7 M urea (Fig. 3e). This is consistent with the formation of a molten globule state after the first transition that is usually accompanied by only minor increases in molecular volume. Hence, all these data presented here support the view that the monomeric active form of mrmHsp47 is in a metastable, sheetexposed conformation of the serpins with a 5-stranded A-sheet (Fig. 1a).
The Interaction between (PPG) 10 and mrmHsp47-The peptide (PPG) 10 is a well established procollagen mimic peptide. We therefore chose to use this peptide to characterize the nature and mechanism of the binding interactions between Hsp47 and procollagen. However, it is known that the oligomerization state of the peptide is dependent upon both its concentration and the temperature (1,2,47). Thus, prior to performing these experiments, the physical state of (PPG) 10 used during our experiments was determined by size exclusion chromatography. High resolution size exclusion chromatography of 20 M (PPG) 10 at 25°C (conditions consistent with those used during binding experiments) established that the peptide was actually monomeric (ϳ2.8 kDa).
Fluorescence and CD Binding Studies-The binding of monomeric (PPG) 10 to mrmHsp47s was measured by observing changes in the intrinsic tryptophan fluorescence of mrmHsp47s as a consequence of binding monomeric (PPG) 10 (39). Peptide binding was accompanied by a corresponding decrease in the tryptophan fluorescence emission intensity with a minimal change in the wavelength of the maximum. The titration of mrmHsp47s with (PPG) 10 peptide at pH 8.0, 25°C proceeded to saturation giving a binding isotherm that was curve-fitted with a single site, single affinity binding model. This is unlike the cooperative binding observed previously (39) for the interaction between Hsp47 and full-length collagen. This leads us to suspect that cooperative binding involving Hsp47 is induced by the presence of three collagen strands in close proximity, something that is not available in (PPG) 10 under these conditions. Analysis of these data provides an equilibrium dissociation constant, K d , of 700 nM (Fig. 4a); this is much weaker than that previously recorded using a comparable fluorescence-based assay for Hsp47 binding to collagen (K d of 140 nM) (39). These results for collagen are in disagreement with Koide and coworkers (20) who have determined K d values (300 -1100 nM) for the interaction between mrmHsp47 and full-length type I-IV collagen using surface plasmon resonance. It seems more reasonable to compare our results with McDonald and Bä chinger (39) as our methodologies are identical. Thus, our results, in agreement with McDonald and Bä chinger (39), suggest that binding to (PPG) 10 is less tight than for full-length collagen. This result is of particular interest as we have shown earlier that (PPG) 10 is monomeric, throughout the concentration range used during the binding assay.
CD spectroscopy was used to provide information on the conformational changes that occur upon formation of the (PPG) 10 -Hsp47 complex. The binding of monomeric (PPG) 10 to mrmHsp47s was monitored by observing changes in CD signal as a function of peptide concentration (Fig. 4a). Spectra of mrmHsp47s were acquired in the presence of different concentrations of (PPG) 10 peptide and were then processed by subtraction of concentration-matched (PPG) 10 spectra from a blank titration and an mrmHsp47 background spectrum (assuming mrmHsp47s secondary structure to be largely constant throughout the titration, an assumption that is reasonable in view of the CD data presented below, Fig. 4e). Values of ⌬⌬A 216 (216 nm was chosen in place of 225 nm because the change in signal was significantly larger at this wavelength, providing higher quality titration data) were also plotted as a function of (PPG) 10 concentration to give a saturation binding isotherm that was curve-fitted as above to give a K d value of 810 nM in good agreement with the value obtained by fluorescence titration (Fig. 4a). The fact that the titration proceeded to satura-tion with a positive increase at 225 nm, as observed from the spectra (Fig. 4b), is significant. The P II helix character of (PPG) 10 is clearly increasing with binding suggesting that binding promotes a conformational ordering of prolyl residue peptide bonds in (PPG) 10 .
Experiments to investigate the effect of pH on the ability of monomeric Hsp47 to prevent collagen fibrilization were also performed (Fig. 4c). These results agreed with those of Thomson and Ananthanarayanan (48) and demonstrated that antifibrilization activity is abolished at low pH. This concurs with the pH-dependent conformational change of Hsp47 observed previously by ourselves (44) and others (48). The binding of Hsp47 to (PPG) 10 over a range of pH was measured using the fluorescence-based assay and showed a marked reduction in affinity as the pH was reduced from 7 to 6 ( Fig. 4d). Thus, a more complete picture mechanism of pH mediated activity can be presented. A drop in pH between 7 and 6 induces a conformational change, and this in turn causes a 1000-fold reduction in affinity of Hsp47 for collagen, causing Hsp47 to dissociate, leading to an abolition of biological activity.
Trimeric mrmHsp47-A number of members of the serpin family have been reported to become homopolymeric. These include the polymers of ␣ 1 -antitrypsin, neuroserpin, ␣ 1 -antichyomtrypsin that are implicated in disease and the dimeric form of antithrombin. Size exclusion chromatography was used to purify small quantities of trimer from mrmHsp47c (Fig. 2b), and its conformational stability was investigated. Reanalysis of purified trimer by gel filtration after the original separation showed no reversion to the monomeric state (data not shown). This indicates that the rate of conversion between trimer and monomer is slow as expected for the reactive center loop to strand 4 ␤-sheet A linkage observed in serpin polymers. CD temperature titration of the trimer did not reveal a clear, sharp transition as would be expected for the unfolding of a protein but rather a gentle reduction in secondary structure between 50 and 95°C (Fig. 3b). Trimeric Hsp47 was also analyzed using transverse urea gradient gel and showed that trimeric Hsp47 is insensitive to urea concentrations in excess of 8 M (Fig. 3e). These results are consistent with those obtained for ␣ 1 -antitrypsin and ␣ 1 -antichymotrypsin polymers, polymers formed by a linkage involving the insertion of loop from one monomer into the vacant fourth strand position of the A-sheet of a second (37) (Fig. 1d).
Assessment of the Biological Activity of Trimeric Mrm-Hsp47-The biological activity of mrmHsp47 was assessed by observing its ability to form a complex with both full-length collagen and the (PPG) 10 collagen mimic peptide. Fixed aliquots of mrmHsp47c were incubated with different quantities of collagen type I and (PPG) 10 ; the resulting complex mixtures were then analyzed by native PAGE to determine the fate of mrmHsp47 trimer and monomer species (Fig. 5, a and b). mrmHsp47c, a solution containing both monomeric and trimeric forms of Hsp47, was used during these experiments to allow comparison between the activities of both forms. Bands attributed to both monomer and trimer were found to decrease in intensity as the quantity of collagen type 1 was increased with no evidence of substrate-induced inter-conversion between the two forms. This result for collagen type I is consistent with the sequestration of both species by collagen into large high molecular weight complexes unable to run into the gel, thus showing that both monomer and trimer clearly have the capacity to interact with and bind to collagen. The equivalent experiment carried out using (PPG) 10 showed a similar result to that obtained for full-length collagen. This is not consistent with the bandshift expected when a 47-kDa protein binds to a 2.7-kDa monomeric peptide. The only explanation consistent with this behavior is the intriguing possibility that Hsp47 may be acting to assemble the monomeric (PPG) 10 peptides into higher order complexes that then lead to such an extreme bandshift. To test this observation, mixtures of Hsp47 and (PPG) 10 were subjected to analysis by DSC (Fig. 5c). These results showed peaks corresponding to both free peptide (43.6°C), Hsp47 (55°C), and a third peak at 74.5°C. This third peak is similar in character to those observed by Frank and co-workers (47), when (PPG) 10 is forced into a triple helical conformation by the addition of artificial trimerization domains to the C terminus of the peptide. It thus seems plausible that this peak in the Hsp47/(PPG) 10 mixture is attributable to high order (PPG) 10 multimers that are assembled by the presence of Hsp47, unfolding at a much increased temperature.
Molecular Model Building-The structure of the serpin scaffold is highly conserved between members of the family whose sequences are quite divergent. Thus the use of homology modeling software to provide structural models is possible. We used three structures of monomeric serpins in the native state as templates for the construction of a model of mouse Hsp47. The sequence identities for these template structures when compared with Hsp47 were 25%. Initial analysis of the structure showed that the amino acid packing with the core was good with no clashes and few cavities. The stereochemistry and geometry of the structure were satisfactory when analyzed by Procheck (49).
In the first instance the model was used to gain an understanding of the mechanism of pH-dependent ligand binding. The 1000-fold reduction binding affinity that occurs between pH 6 and 7 is close to the pK a of histidine, suggesting that the conformational change may be triggered by a change in charge of one or more histidines (44). Thus the model was examined for histidine clusters, with three clusters of 3 histidines being Native PAGE showing mrmHsp47c (mixture of monomer and trimer) in the presence of increasing amounts of collagen type I. Lane 1, mrm-Hsp47 (10 g); lane 2, collagen (27 g); lanes 3-15, mrmHsp47 (10 g) with 0.5-6.5 g of collagen in 0.5-g increments. b, (PPG) 10 binding by mrmHsp47. Native PAGE showing mrmHsp47c (mixture of monomer and trimer) in the presence of increasing amounts of (PPG) 10 . Lane 1, mrmHsp47 (10 g); lane 2, (PPG) 10 (27 g); lanes 3-15, mrmHsp47 (10 g) with 0.5-6.5 g (PPG) 10 in 0.5-g increments. c, thermal unfolding by DSC. The heat capacity of mrmHsp47s, (PPG) 10 , and a mixture of both was measured using DSC.
identified. These were localized to one group under the base of the A-sheet (His-297, His-302, and His-335), one beneath the hinge region at the top of the A-sheet (His-197, His-198, and His-220), and a third group in the "gate" region at the end of the C-sheet (His-255, His-256, and His-368). All three of these histidine clusters reside in regions that are known to be involved in the conformational dynamics of serpins (50) and as such one or more are likely to be involved in the conformational switch.
The discovery of the trimeric form of Hsp47 along with the determination that the linkage is likely to be by reactive center loop of one molecule into the strand 4 position of ␤-sheet A of a second led to the possibility that a model of the trimer could be constructed. Previous work by Sivasothy et al. (52) on the structure of serpin polymers has demonstrated that a larger amount of biophysical data is required to allow a structure of linear serpin polymers to be determined (Fig. 6a). This is because of the inherent flexibility of such an arrangement. However, the existence of a discrete trimer in preference to a variety of polymeric species suggests a ring structure (Fig. 6b). Such configurations have been seen in the S iiyama variant of ␣ 1antitrypsin (51). The formation of such ring structures by the insertion of the reactive center loop (RCL) of one molecule into the A-sheet of a second leads, however, to some fairly rigid steric constraints. As result only a small number of trimeric Hsp47 configurations can exist. These configurations can be modeled, and all require that sheet A of each molecule points into the middle of the ring, leaving the helices that pack behind sheet A exposed to solvent. This region includes the proposed peptide binding cleft between helices hA and hG/hH (45) leading to the possibility that three collagen peptides may associate around the external face of the Hsp47 trimer.  (52). b, a closed "circlet" trimer formed by RCL (loop) insertion into A-sheet forming 6-stranded A-sheets. Diagrammatic representations are shown on the left-hand side. Individual monomers are colored red, green, and yellow, respectively, and a stick representation of the (PPG) 10 peptide has been placed in the putative binding site between ␣-helices hA and hG/hH in the closed circlet model, as described previously (45). c, a flying capstan, a piece of apparatus essential to maintain strand order during rope making.

DISCUSSION
Hsp47 is a molecular chaperone protein derived from the serpin family of proteins, essential for collagen biogenesis. Extensive evidence has suggested that Hsp47 may associate with nascent procollagen polypeptide chains (17), unhydroxylated non-triple helical chains, and even hydroxylated triple helices (16,18,19) in the endoplasmic reticulum during the early stages of collagen biosynthesis. However, there has been little evidence to date to suggest how Hsp47 might actually function as a molecular chaperone in these early stages of collagen biosynthesis. Tasab and co-workers (16) have suggested that Hsp47 acts by binding to fully formed hydroxylated triple helices to stabilize these triple helices and prevent chain dissociation in the endoplasmic reticulum. We present here data that combine the extensive studies of other members of the serpin superfamily with the already established behavior of Hsp47.
Implications of Peptide and Protein Datum State-In this type of study, it was important that the conformational and configurational states of the reactants be determined. In particular, members of the serpin family are highly conformationally mobile. It is also true that "model" peptides such as (PPG) 10 that are used throughout the study of collagen-binding proteins are not conformationally static (1,2,47). Indeed the oligomerization state of (PPG) 10 varies with both concentration and temperature. Thus we have begun our study by carefully defining these "datum" states using a range of biochemical techniques.
Hsp47 Conformation, Configuration, and Collagen-binding Site-Critical to the process of defining the functional capacity of Hsp47 has been the generation of well characterized protein preparations. Our attempts to establish such a preparation has led to the novel discovery that Hsp47 can adopt both monomeric and trimeric forms. Monomeric Hsp47 (mrmHsp47s) possesses all the classic hallmarks of metastable, sheet-exposed serpins conformation with a 5-strand A-sheet (Fig. 1a). CD, DSC, temperature, and urea titration/transverse gel data all agree to support this view (Fig. 3). This is in disagreement with our previous model that represented monomeric mrmHsp47 in a conformation comparable with that of latent non-inhibitory PAI-1 (Fig. 1e), possessing an A-sheet with six strands (45). The new proposal is also consistent with the amino acid sequence of the "hinge region" of mrmHsp47. This sequence in Hsp47 consists of a large number of residues with bulky side chains that preclude the formation of the hyperstable 6-stranded form characterized by the latent form of PAI-1 (Fig. 1a). The discovery that Hsp47 is in the 5-stranded metastable conformation may have implications for our past model of the collagenbinding site in a cleft between helices hA and hG/hH (45). It is well established that the metastable form of a serpin can bind peptides (which could conceptually include collagen) in the vacant strand 4 position of the A-sheet (Fig. 1c) (3) producing a hyperstable species. However, the observation by us and Thomson and Ananthanarayanan (48) that the collagen-Hsp47/ (PPG) 10 -Hsp47 complex does not alter the stability of Hsp47 (Fig. 4e) suggests that the collagen/peptide do not bind to the A-sheet.
The Hsp47 trimer identified as a component of mrmHsp47c is potentially fascinating in both structural and functional terms for both molecular chaperones and serpins. A number of serpin family proteins are known to undergo oligomerizations; however, the formation of only two discrete forms (i.e. monomer and trimer) is unusual. Only the serpin, antithrombin, has been described previously as forming a small set of discrete oligomers. Circulating antithrombin forms dimers by a linkage involving an interaction of the RCL of one molecule into ␤-sheet C of a second conformationally stabilized (latent state) mole-cule (3,33). However, the Hsp47 trimer does not conform to the antithrombin model as no stabilized monomeric species equivalent to the latent form was observed at any time. The structure of the trimer is, however, indicated by virtue its high stability. This indicates that the RCL of one molecule is at least partially inserted into the ␤-sheet A of a second, creating Asheets with six strands and consequently hyperstability. This configuration is the same as that seen for polymers ␣ 1 -antitrypsin (32,51,52). The observation that Hsp47 trimer appears to be as ready to bind collagen as the Hsp47 monomer (Fig. 5) has further implications on the position of the collagen-binding site. The filling of the vacant strand 4 position of the A-sheet to form the link within the trimer means that peptide binding in this position is no longer possible, and yet the protein retains activity. Thus it appears that our original proposal of a collagen-binding site between helices hA and hG/hH (45) seems all the more viable.
Implications of Hsp47 Mechanism in Biology-Our study of the self/self and self/substrate interactions of Hsp47 has allowed us to propose, in more detail than before, possible roles for Hsp47 in biology. We have demonstrated that Hsp47 is able to bind to a monomeric prolyl peptide inducing the formation of a polyproline type II conformation. We have also shown that in doing so Hsp47 is able to induce the formation of higher order assemblies of the peptide with increased stabilities similar to those observed by Frank and co-workers (47) for a stabilized (PPG) 10 trimer. Translation of these activities into biological function provides a number of possibilities for what is a multifunctional chaperone. It might be proposed that Hsp47 may bind to nascent monomeric polypeptide chains as observed by Sauk et al. (17), inducing a productive polyproline type II conformation that aids the formation of the triple helix. Equally, Hsp47 may bind to misfolded areas within fully formed collagen that are proposed to be the result of collagen folding (53), reducing possible intramolecular aggregation, while again inducing polyproline type II conformation and assembly. Which combination of these functions Hsp47 provides within the cell can only be determined by more detailed in vivo work.
The mechanism of collagen:Hsp47 association and dissociation appears somewhat less complex. The first stage of collagen assembly occurs in the lumen of the endoplasmic reticulum. In the neutral pH environment of this compartment (54), Hsp47 is able to bind to collagen chains. Correct formation of the collagen trimer leads to transportation to the cis-Golgi complete with bound Hsp47. This compartment has a reduced pH compared with the endoplasmic reticulum (31). We have shown that at this pH the affinity of Hsp47 for collagen is greatly reduced due to pH-induced trans-conformational changes. Thus Hsp47 is shed from the collagen molecule promoting like with like association with other collagen molecules to form protofibrils. The free Hsp47 should then be recycled back to the endoplasmic reticulum by virtue of its C-terminal KDEL transport signal.
Conclusions-The various biophysical and structural observations made in this paper have allowed us to make an integrated proposal for the role of Hsp47 as a procollagen/collagenspecific molecular chaperone protein. Furthermore, as collagen molecules are often described as "molecular rope," we are able to propose an apt analogy that describes our hypothesis. For centuries rope makers have used a tool called a flying capstan that acts to order the three coiled strands that make up rope. These strands are then able to intertwine in a controlled manner producing the final stable rope. If we take that analogy to a logical conclusion, then the Hsp47 trimer itself appears to have all the right characteristics to be analogous to a molecular flying capstan (Fig. 6c). Trimeric Hsp47 has three sites; it is able to bind single collagen strands and in some way induce the formation of higher order assemblies.