Xaa-Arg-Gly triplets in the collagen triple helix are dominant binding sites for the molecular chaperone HSP47.

HSP47 is an essential procollagen-specific molecular chaperone that resides in the endoplasmic reticulum of procollagen-producing cells. Recent advances have revealed that HSP47 recognizes the (Pro-Pro-Gly)(n) sequence but not (Pro-Hyp-Gly)(n) and that HSP47 recognizes the triple-helical conformation. In this study, to better understand the substrate recognition by HSP47, we synthesized various collagen model peptides and examined their interaction with HSP47 in vitro. We found that the Pro-Arg-Gly triplet forms an HSP47-binding site. The HSP47 binding was observed only when Arg residues were incorporated in the Yaa positions of the Xaa-Yaa-Gly triplets. Amino acids in the Xaa position did not largely affect the interaction. The recognition of the Arg residue by HSP47 was specific to its side-chain structure because replacement of the Arg residue by other basic amino acids decreased the affinity to HSP47. The significance of Arg residues in HSP47 binding was further confirmed by using residue-specific chemical modification of types I and III collagen. Our results demonstrate that Xaa-Arg-Gly sequences in the triple-helical procollagen molecule are dominant binding sites for HSP47 and enable us to predict HSP47-binding sites in homotrimeric procollagen molecules.

Procollagen folding in the endoplasmic reticulum (ER) 1 is a unique pathway involving unique post-translational modifications and the assistance of molecular chaperones. Nascent procollagen ␣-chains entering the lumen of the ER are modified by specific enzymes including prolyl 4-hydroxylase. After completion of translocation, three C-terminal propeptides selectively associate, and the trimer is stabilized by disulfide bridges. The long triple helix (300 nm for type I collagen) comprised of tandem Xaa-Yaa-Gly repeats is believed to propagate from the C-terminal nucleus, like fastening a zipper (1). HSP47 is only one heat-shock protein found in the ER, and it specifically interacts with various types of collagen (2)(3)(4). This protein has been recognized as a collagen-specific molecular chaperone that facilitates normal procollagen biosynthesis (5). HSP47 is essential to normal embryogenesis and normal collagen biosynthesis. Disruption of the hsp47 gene in both alleles causes an embryonic lethal phenotype in mice with severe impairment of collagen-based tissue structures (6). To unveil the mechanism of HSP47 function, the elucidation of the molecular basis of the chaperone-substrate interaction is required.
The recent investigations on substrate recognition by HSP47 provided information about primary and tertiary structures of procollagen required for interaction with HSP47. We (7) previously found that HSP47 binds to classical collagen model peptides with the sequence of (Pro-Pro-Gly) n and that the binding was negatively affected by replacements of Pro residues at Yaa positions with 4-hydroxyproline (Hyp) residues. Triple-helical conformation of the substrate was also shown to be important in recognition by HSP47 (8,9). However, we have not yet determined the actual HSP47-binding sites in procollagen molecules at the amino acid level. In this study, we describe the results of in vitro binding analysis elucidating the HSP47binding sequences in the collagen triple helix using a series of synthetic collagenous peptides. The residue-specific chemical modifications of native collagens also enabled us to identify essential amino acid residues in the substrate recognition by HSP47.
Solid-phase Binding Assay-Peptide immobilization onto CNBr-activated Sepharose (Amersham Biosciences, Inc.) and the solid-phase binding assay were performed as reported earlier (7). For the set of basic peptides, an extra Cys residue was added to the N terminus of each peptide, and the peptides were coupled to thiopropyl-Sepharose 6B (Amersham Biosciences, Inc.). Immobilized peptides were kept at 4°C for 2 days before the binding assay to refold the peptides. Escherichia * This work was supported partly by Grant-in-aid for Encouragement of Young Scientists 13780484 from the Ministry of Education, Culture, Sports, Science and Technology of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ coli lysate containing the GST-HSP47 fusion protein (100 l) was mixed with 100 l of the binding buffer and a 30-l bed of the affinity beads. The binding was carried out at 4°C for 1 h. After washing the bed, proteins retained on the beads were eluted by adding Laemmli's SDS sample buffer and separated by SDS-PAGE on 10% gels.
Competitive Elution of GST-HSP47-The competitive elution assay was performed as described previously (7). Briefly, GST-HSP47 fusion protein retained on (Pro-Pro-Gly) 10 beads (30 l bed) was mixed with 180 l of competitor solution in the binding buffer, and the mixture was shaken at 4°C for 1 h. Before use, competitor peptides were refolded at 4°C for 2 days. GST-HSP47 retained on the beads was eluted with Laemmli's SDS sample buffer and analyzed by SDS-PAGE.
Chemical Modification of Type I and III Collagen-For Arg-specific chemical modification, a 80-l bed of porcine type I or III collagen (Nitta Gelatin, Osaka, Japan) immobilized to NHS-activated Sepharose 4FF (Amersham Biosciences, Inc.) was treated with 0.2 M 2,3-butanedione in 160 mM sodium borate buffer (pH 8.3) containing 20% methanol at room temperature for 4 days. The Arg modification was confirmed by amino acid analysis of the corresponding 6 N HCl-hydrolysates. To acetylate Lys residues, the collagen beads were treated with 0.8 M acetic anhydride in saturated NaHCO 3 adjusted to pH 8.3, at room temperature for 4 h. An aliquot of the acetylated collagen beads was further treated with 150 mM 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide in 1 M ethanolamine-HCl (pH 7.4) for 8 h to modify carboxyl groups. For modification of His and Met residues, the collagen beads were mixed with 150 mM iodoacetic acid in 0.2 M Tris-HCl (pH 7.5), and the suspension was incubated at room temperature for 4 h. These residue-specific modifications were confirmed by shifting the collagen bands on 7% SDSpolyacrylamide gels visualized by silver staining. Before the binding assay, all of the modified collagen beads were washed and equilibrated with the binding buffer.
Peptide Binding Assays Based on Fluorescence Quenching-Chick HSP47 was purified from 13-day-old chick embryos according to the protocol described by Saga et al. (10). Fluorescence spectra were recorded on a Hitachi F-3010 spectrophotometer using a 1-cm path length quartz cuvette. The excitation wavelength was 295 nm. Fluorescence quenching was monitored at the emission wavelength of 340 nm. In the binding assays, stock solutions of type I collagen or collagenous peptides were added to a solution of chick HSP47 (31 g/ml) in 50 mM bis-Tris (pH 7.4), 150 mM NaCl, 1 mM EDTA. After each addition, the sample solution was equilibrated for 2 min. To ensure the triple-helical conformation of collagenous peptides, all measurements were performed at 4°C using an Eyela NCB-2200 circulating water bath. The spectra were corrected for dilution.

RESULTS
Introduction of Arg to Yaa Positions of (Pro-Pro-Gly) n Enhances the Binding of HSP47-In the previous study (9), we identified dozens of HSP47-binding sequences from a random DNA library designed to encode Gly-Pro-Pro-(Gly-Pro-Yaa) 6 -Gly-Pro-Pro in the yeast two-hybrid screening using HSP47 as a bait. At the Yaa positions in the selected peptides, Arg and Pro residues were highly enriched (6-and 5-fold enrichment, respectively) from the basal frequencies (9). To investigate the effect of Arg residues on the interaction with HSP47, we first analyzed the binding of the GST-HSP47 fusion protein produced in E. coli to synthetic Arg-containing collagenous peptides in vitro. Because only (Pro-Pro-Gly) n -peptides are known to interact with HSP47 in the in vitro binding assay (7), sequences of the Arg-containing peptides to be tested were designed to be (Pro-Pro-Gly) m -Pro-Arg-Gly-(Pro-Pro-Gly) n (Fig.  1A). In accordance with the previous data (7), specific binding of GST-HSP47 to (Pro-Pro-Gly) n was observed in a chain lengthdependent manner (Fig. 1B). The minimal length required for the HSP47 binding was shortened to five triplets (15 amino acid residues) when one of the Pro residues at the Yaa positions was replaced by an Arg residue (Fig. 1B, lane 9, R8-(PPG)5). The binding of HSP47 to the Arg-containing peptide was saturated at the 18-mer peptide R11-(PPG) 6 (lanes 10 and 11). Single substitution of the Pro 11 of (Pro-Pro-Gly) 8 for an Arg residue markedly enhances the affinity to GST-HSP47 (lanes 7 and 11). Taken together, the Arg residue incorporated in the Yaa positions of (Pro-Pro-Gly) n enhances the interaction with HSP47.
Pro-Arg-Gly Triplet Is a Binding Site for HSP47-Arg residues at the Yaa positions are known to stabilize the triplehelical conformation of collagen model peptides as well as Hyp residues (11). It is therefore necessary to clarify whether the Pro-Arg-Gly triplet forms an actual binding surface for HSP47 or the Arg residue merely stabilizes the triple-helical conformation of an adjacent HSP47-binding sequence of Pro-Pro-Gly. To solve this issue, we examined the effect of Arg incorporation to (Pro-Hyp-Gly) 8 that was previously shown to not interact with HSP47 (7). In parallel, the effect of incorporation of Pro, Gln, Met, Ile, Thr, and Ala, which were also enriched by the previous two-hybrid selection of HSP47-binding peptide, to the Yaa positions was also examined (9). Among the peptides of (Pro-Hyp-Gly) 3 -Pro-Yaa-Gly-(Pro-Hyp-Gly) 4 , only the Pro-Arg-Gly-containing peptide showed significant binding to GST-HSP47 ( Fig. 2A, lane 7). Other peptides containing a Pro-Pro-Gly, Pro-Gln-Gly, Pro-Met-Gly, Pro-Ile-Gly, Pro-Thr-Gly, or Pro-Ala-Gly triplet did not show detectable interaction with GST-HSP47 (lanes 6 and 8 -12). We also reconfirmed the previous result that HSP47 does not bind to Pro-Hyp-Gly repeats (7) (lane 5). This result indicates that HSP47 directly interacts with the Pro-Arg-Gly triplet in the triple helix and that the affinity to Pro-Arg-Gly is much stronger than the affinity to the previously identified HSP47-binding triplet Pro-Pro-Gly (lane 4). The difference in HSP47 binding affinity of these peptidic substrates was further confirmed by a competitive elution assay (Fig. 2B). GST-HSP47 bound to (Pro-Pro-Gly) 10 beads was eluted by the addition of refolded peptides, and the IC 50 values were estimated to be about 40 M for Arg 11 -(Pro-Hyp-Gly) 8 and Ͼ1 mM for (Pro-Pro-Gly) 8 (Fig. 2C).
The Arg Residue at the Yaa Position Is Important for HSP47 Binding-Because the Xaa and the Yaa amino acids are not spatially equivalent in the collagen triple helix, we further examined the positional effect of an Arg residue on the inter- action with HSP47 using similar (Pro-Hyp-Gly) 3 -Xaa-Yaa-Gly-(Pro-Hyp-Gly) 4 peptides. In the binding assay, four variations of the triplets, Pro-Hyp-Gly, Pro-Arg-Gly, Arg-Hyp-Gly, and Arg-Pro-Gly, were tested for GST-HSP47 binding. As shown in Fig. 3A, only the Pro-Arg-Gly triplet interacted with GST-HSP47, and peptides possessing Arg residues at Xaa positions did not show detectable interaction with GST-HSP47.
The possible contribution of side-chain basicity to the interaction with HSP47 was also examined by replacing the Arg residue with other basic amino acids. We prepared Cys-(Pro-Hyp-Gly) 3 -Pro-Yaa-Gly-(Pro-Hyp-Gly) 4 , to Yaa of which Arg, homoarginine, Lys, and Orn were incorporated, and these peptides were coupled to Sepharose beads through N-terminal disulfide bonds. The result of a similar solid-phase binding assay using these basic peptides is shown in Fig. 3B. Replacement of the Arg residue in the Pro-Arg-Gly triplet with a homoarginine residue resulted in a decrease of affinity to GST-HSP47 (Fig. 3B, lane 5). When the Arg was replaced with either a Lys or an Orn, the binding to GST-HSP47 was not detected (lanes 6 and 7).
These results indicate that the positive charge of the Yaa residue does not account for the specific binding to HSP47 and probably the guanidino group, and the length of the Arg side chain is a critical factor for the interaction. We concluded that HSP47 recognizes the Arg residue in a position-specific and side chain-specific fashion.
In the native collagen sequences, various residues are found in the Xaa positions. Among Xaa-Arg-Gly sequences, Gly-Glu-Arg (2.7%), Gly-Pro-Arg (2.6%), Gly-Asp-Arg (1.2%), and Gly-Ala-Arg (1.1%) are prominent, and other combinations are at less than 1% frequency each (12). To examine the effect of Xaa residues on the interaction with HSP47, we synthesized peptides containing an Xaa-Arg-Gly triplet and afforded to the GST-HSP47 binding assay. All of the tested triplets, Pro-Arg-Gly, Ala-Arg-Gly, Asp-Arg-Gly, and Glu-Arg-Gly, showed significant binding to GST-HSP47, suggesting that Xaa amino acids does not largely contribute to the binding to HSP47 (Fig.  3C).
Importance of Arg Residues of Native Collagens for the Interaction with HSP47-All of the data presented above strongly suggest that Xaa-Arg-Gly triplets in the triple helix form binding sites for HSP47 in collagen. To confirm this suggestion, homotrimeric porcine type III collagen was chemically modified by several residue-specific modification reagents, and the binding of GST-HSP47 to the modified collagens were examined. When type III collagen was modified with 2,3-butanedione to block the guanidino group of Arg residues, the binding of GST-HSP47 was abolished (Fig. 4, upper panel, lane 4). in accordance with the data using synthetic peptides. Other amino acid residues such as Lys, His, Met, Asp, and Glu were also revealed to not significantly contribute to HSP47-type III collagen binding. The effect of side-chain modification in heterotrimeric type I collagen on HSP47 binding was also examined, and very similar results as those shown for type III collagen were obtained (Fig. 4, lower panel).
Relative binding affinity of chick HSP47 to type I collagen and synthetic substrates was further estimated by means of fluorescence quenching of intrinsic Trp residues of HSP47 (13). As shown in Fig. 5, specific interaction of HSP47 with type I collagen, Arg 11 -(Pro-Hyp-Gly) 8 , and (Pro-Pro-Gly) 10 , whose binding was detectable in the previous solid-phase binding assay, was also detected in this spectrometric assay. The nonbinder, (Pro-Hyp-Gly) 10 , did not quench the fluorescence of HSP47. Comparison of the triple-helical triplet-based affinity revealed that the affinity of a Pro-Pro-Gly triplet in (Pro-Pro-Gly) 10 was weaker than that of the mean Xaa-Yaa-Gly triplet in type I collagen. When the concentration of type I collagen was converted to that of the Xaa-Arg-Gly triplet unit (34 homotrimeric Xaa-Arg-Gly triplets in type I collagen), the affinity was comparable with that of a Pro-Arg-Gly triplet in Arg 11 -(Pro-Hyp-Gly) 8 . This result also indicates that the interaction of HSP47 with type I collagen is attributed to the binding to the triple-helical Xaa-Arg-Gly triplets.

DISCUSSION
The indispensable role of HSP47 in collagen biosynthesis was recently demonstrated by the analysis of embryonic lethal hsp47 Ϫ/Ϫ mice (6). In the mouse embryos, various collagenbased tissue structures were severely disrupted, and cells derived from the mice produced abnormal molecules of at least types I and IV collagen. In addition, two-hybrid screening of cDNA using HSP47 as bait yielded cDNA sequences encoding proteins containing collagenous Xaa-Yaa-Gly repeats. 2 These findings were in good agreement with the fact that HSP47 interacts with various types of collagen (at least type I-V) (3). Taking this into consideration, HSP47 is expected to function through a specific interaction with common sequences in the domain(s) found in various types of collagen. In other words, HSP47-binding sites should exist in the triple helix-forming region of collagens. In our (7) previous search of HSP47-binding sequences, we identified the HSP47-binding sequence (Pro-Pro-Gly) n . The triple-helical conformation was also revealed to be important in recognition by HSP47 (8,9). However, in the biosynthetic pathway of procollagen, most of the Pro-Pro-Gly sequences are converted to Pro-Hyp-Gly, to which HSP47 does not bind, by the action of prolyl 4-hydroxylase before triple helix formation. This discrepancy led us to further search for HSP47-binding sequences other than (Pro-Pro-Gly) n . Here, we identified Xaa-Arg-Gly triplets as the HSP47-binding sequences by using various synthetic collagen model peptides (Figs. 2 and 3), and the dominance of this sequence in HSP47 binding to collagen was demonstrated utilizing a residue-specific chemical modification technique (Fig. 4). The affinity of HSP47 to the Pro-Arg-Gly triplet was also shown to be much higher than that to the previously identified (Pro-Pro-Gly) n sequence (Fig. 2). In the fluorescent quenching-based binding assay, the affinity of HSP47 binding to a synthetic Pro-Arg-Gly triplet was shown to be comparable with that to an Xaa-Arg-Gly triplet in type I collagen when other triplets were assumed to have negligible affinity (Fig. 5). This result emphasizes our conclusion that Xaa-Arg-Gly triplets in procollagen are dominant binding sites for HSP47. In addition, interactions of HSP47 with Pro-Pro-Gly triplets in (Pro-Pro-Gly) 10 are also detectable in the spectrometric assay performed at 4°C consistently with the previous data (7). However, the similar binding assay at 25°C failed to detect significant binding between HSP47 and (Pro-Pro-Gly) 10 as reported by Macdonald and Bä chinger (13). This temperature dependence in the HSP47 binding would be further evidence for the importance of triplehelical conformation of the substrates (8,9).
The information derived from the experiments shown in this study has enabled us to estimate the HSP47-binding sites in native homotrimeric collagens. For instance, HSP47-binding sites in the triple-helical region of bovine type III collagen can be estimated as shown in Fig. 6. Type III collagen possesses 32 possible binding sites for HSP47 having the identified sequences of Xaa-Arg-Gly (Xaa ϭ Glu, Pro, Asp, or Ala; Fig. 6, shadowed). Based on an extended assumption that any amino acid residues are tolerable in the Xaa positions, nine additional binding sites were predicted in type III collagen (Fig. 6, underlined). The Arg-containing sequence is also shown to be necessary in the recognition of heterotrimeric collagen such as type I collagen (Fig. 4, lower panel).
An Arg residue at the Yaa position in the Xaa-Yaa-Gly triplets is known to enhance the thermal stability of triple-helical conformation in a similar magnitude to a Hyp residue. This stabilizing effect was suggested to be attributed to hydrogen bond formation between the guanidino group and a backbone carbonyl group (11). We demonstrate that the basicity of the Arg side chain did not largely contribute to the interaction with HSP47 (Fig. 3B) and that high ionic strength buffer did not disrupt the binding of HSP47 to the Pro-Arg-Gly triplet (data not shown). This result coincides with the previous suggestion that a hydrophobic interaction would mainly contribute to the HSP47-collagen interaction (10,14). We suggest that a hydrophobic ring formed by the Arg side chain would directly interact with the hydrophobic cleft or patch of HSP47 molecule. This speculation explains the previous data that HSP47 binding to (Pro-Pro-Gly) n was abolished by the introduction of a hydroxyl group to the ␥-carbons of the hydrophobic proline rings at Yaa positions (7).
We proposed in the previous report that HSP47 would be involved in the quality control of procollagen molecules transported to the Golgi apparatus by surveying the level of prolyl 4-hydroxylation. This hypothesis was based on the result obtained from the peptide binding experiments that HSP47 interacts with unhydroxylated Pro-Pro-Gly sequences in the triple-helical structure (7). However, this hypothesis seems unlikely once Xaa-Arg-Gly was identified as the dominant binding site for HSP47. In addition, two other possible roles of HSP47 were recently proposed. One is for the stabilization of correctly folded triple-helical intermediates of procollagen that are otherwise unstable at body temperature (8,9). The other is the prevention of lateral association, or bundle formation, of procollagen in the ER (14). Our findings presented here support both of these hypothetical roles. HSP47 would stabilize triple-helical intermediates of procollagen by binding to Xaa-Arg-Gly triplets found approximately every 30 residues (Fig. 6). This stabilizing effect could be precisely evaluated by estimating the melting temperatures of Pro-Arg-Gly-containing triplehelical peptides in the presence and absence of HSP47 in vitro in the future. Although this hypothesis explains the heat-shock inducibility of HSP47, Macdonald and Bä chinger (13) recently reported that HSP47 has little effect on the thermal stability of collagen. HSP47 may also act as an inhibitor of the lateral association of procollagen in the ER by masking Arg residues because charged amino acid residues including Arg are presumed to be, at least in part, involved in the formation of the higher order collagen structures (15)(16)(17). The latter scenario is also supported by procollagen aggregate formation occurring at the Golgi apparatus where procollagen is liberated from the chaperone (4,18). Because excessive collagen production in pathophysiological conditions is always associated with the up-regulation of HSP47 (19 -21), a precise analysis of HSP47collagen interaction as shown in this study would provide information for the rational design of an HSP47 inhibitor, which may be used for the treatment of various fibrotic diseases, as well as for the elucidation of the molecular basis of HSP47 function.