Structural basis of redox-dependent substrate binding of protein disulfide isomerase

Protein disulfide isomerase (PDI) is a multidomain enzyme, operating as an essential folding catalyst, in which the b′ and a′ domains provide substrate binding sites and undergo an open–closed domain rearrangement depending on the redox states of the a′ domain. Despite the long research history of this enzyme, three-dimensional structural data remain unavailable for its ligand-binding mode. Here we characterize PDI substrate recognition using α-synuclein (αSN) as the model ligand. Our nuclear magnetic resonance (NMR) data revealed that the substrate-binding domains of PDI captured the αSN segment Val37–Val40 only in the oxidized form. Furthermore, we determined the crystal structure of an oxidized form of the b′–a′ domains in complex with an undecapeptide corresponding to this segment. The peptide-binding mode observed in the crystal structure with NMR validation, was characterized by hydrophobic interactions on the b′ domain in an open conformation. Comparison with the previously reported crystal structure indicates that the a′ domain partially masks the binding surface of the b′ domain, causing steric hindrance against the peptide in the reduced form of the b′–a′ domains that exhibits a closed conformation. These findings provide a structural basis for the mechanism underlying the redox-dependent substrate binding of PDI.

In the endoplasmic reticulum (ER) of eukaryotic cells, a number of molecular chaperones and folding enzymes assist the proper folding of newly synthesized polypeptide chains. Protein disulfide isomerase (PDI) is a major ER protein that operates as a molecular chaperone and a folding enzyme by catalyzing the formation, cleavage, and rearrangement of the disulfide bonds of unfolded or misfolded proteins [1][2][3] . After the first description of its enzymatic activity in 1963 4 , extensive structural and functional studies of PDI have been reported using PDI from various species, including human [5][6][7][8][9] , yeast 10,11 , and thermophilic fungus [12][13][14] . PDI consists of four tandem thioredoxin-like domains a, b, b′ , and a′ plus a C-terminal extension [1][2][3]15 , which are arranged into a U-shaped structure [9][10][11] . Among the four domains, a and a′ possess a catalytic CXXC motif, which is not shared by b and b′ . Cumulative biochemical data indicate that the b′ and a′ domains are primarily responsible for substrate recognition 3,13,16 . In particular, mutational and cross-linking analyses indicate that the b′ domain provides the principal peptide binding site in PDI 16 . The a′ domain is oxidized by the flavoprotein Ero1 and thereby acts as a disulfide donor for the PDI substrates 17,18 .
One unique property of this modular enzyme is that it undergoes conformational rearrangement of the b′ -a′ domains depending on the redox states of the a′ active site 13,14 . These two domains exhibit a closed conformation in the reduced form and are converted into an open conformation with the exposure of the hydrophobic surface upon oxidation of the a′ domain. This conformational transition is supposed to be associated with the redox-dependent substrate binding of PDI. However, no three-dimensional structural data have yet been reported for PDI ligand binding despite the long history of research on this topic. In view of this situation, we attempted to provide the structural basis of PDI substrate recognition by using an appropriate model ligand.
It has been reported that molecular chaperones actively contribute to the suppression of toxic aggregate formation of various amyloidogenic proteins associated with neurodegenerative disorders 19,20 . In particular, PDI is upregulated in the brain of patients with Parkinson disease and is found in Lewy bodies 21 , which are composed of the protein α -synuclein (α SN), an intrinsically unstructured protein consisting of 140 amino acid residues associated with other proteins. The increased expression of PDI was also observed in α SN transgenic mice 22 . Moreover, we have recently shown that α SN is capable of interacting with the bacterial chaperone GroEL 23 and archaeal chaperone PbaB 24 , serving as a useful probe for characterizing their molecular recognition by biophysical techniques, which include nuclear magnetic resonance (NMR) spectroscopy and small-angle neutron scattering. Hence, we undertook to examine the possible interaction of PDI with α SN and, based on the results, we executed structural analyses using X-ray crystallography in conjunction with NMR spectroscopy that focused on the substrate-binding b′ -a′ domains of PDI.

Results
Redox-dependent interaction of PDI with αSN. To investigate whether α SN can bind PDI, we performed NMR analyses assisted by stable isotope labeling. We prepared 15 N-labeled α SN and observed the heteronuclear single-quantum correlation (HSQC) spectral changes induced upon addition of the PDI b′ -a′ domains. The results indicate that the oxidized b′ -a′ domains caused significant perturbations in the HSQC peaks originating from the α SN segments Val37-Val40 and Val48-Gly51 (Fig. 1a,b), both of which contain hydrophobic (Hb) and aromatic (ϕ ) residues as Hb− Hb− ϕ triplets (Fig. 1c). Remarkably, the peaks from the former segment almost completely disappeared, indicating its extensive involvement in an interaction with the PDI b′ -a′ domains. On the basis of these data, we concluded that α SN is capable of interacting with PDI through its specific hydrophobic segment. Based on peak We confirmed the binding of this segment using 15 N-labeled PDI b′ -a′ domains and a synthetic α SN peptide, Gly-Lys-Thr-Lys-Glu-Gly-Val-Leu-Tyr-Val-Gly, which corresponds to the principal binding site of α SN (Fig. 1c). HSQC spectral data indicated that the peptide caused chemical shift perturbations largely for Gly268, His270, Ala271, and Asn273 in the b′ domain and, to a lesser extent, for their surrounding residues in the same domain and the residues proximal to the a′ active site in the oxidized b′ -a′ form ( Fig. 2 and Supplemental Fig. S1). Such spectral changes were much less pronounced in the reduced form of the b′ -a′ domains, indicating that peptide binding depends on the redox states of the a′ active site. These data are consistent with those previously obtained using mastoparan as the model ligand, which preferentially binds the oxidized form of the b′ -a′ domains 13 . The redox-dependent interaction was confirmed between the PDI b′ -a′ domains and full-length α SN ( Supplementary Fig. S2).
Crystal structure of the PDI b′-a′ domains in complex with the αSN peptide. To determine the interaction mode of PDI with α SN, we carried out X-ray crystallographic analysis using the oxidized form of the PDI b′ -a′ domains and the α SN peptide. We successfully crystallized their complex and determined the crystal structure at 1.60 Å resolution. The final model, refined to a resolution of 1.60 Å, had an R work of 18.4% and R free of 21.7% (Supplemental Table S1). The crystal belonged to space group P2 1 2 1 2 1 with one b′ -a′ molecule and one α SN peptide per asymmetric unit.
The PDI b′ -a′ construct we used for crystallization consisted of residues 208-449, and all residues were ordered in the electron density map. Even though the two-domain arrangement was extensively affected by the crystal packing, the b′ -a′ domains showed an open conformation in the oxidized state (Fig. 3a). Due to the crystal packing, the spatial domain arrangement of the α SN-bound oxidized PDI b′ -a′ domains was remarkably different from that of the unliganded form (PDB code: 3WT2) 25 , suggesting the dynamic nature of the interdomain substrate-binding region (Fig. 3b). Each domain structure of the complexed form was essentially identical to those of the apo form with the RMSD of 0.41 and 0.35 Å for the b′ and a′ domains, respectively. Concerning the bound α SN undecapeptide, all residues were clearly visible in the electron density map (Fig. 3c). Interestingly, the α SN undecapeptide adapts a β -hairpin structure in the crystal.
Because a crystallographically neighboring molecule was accommodated in contact with the two domains, two different interaction modes were observed between the PDI b′ -a′ domains and the α SN peptide (Fig. 3a). One interaction mode (termed contact-b′ ) was mediated through the b′ domain surface proximal to the a′ domain with a peptide-binding area of 391.6 Å 2 . The other interaction mode (termed contact-a′ ) gave a smaller interface area with 242.9 Å 2 exclusively on the a′ domain. In contact-b′ , the α SN peptide was recognized through several hydrophobic interactions involving Leu38 and Val40 of α SN and Ile213, Tyr218, Met222, and Phe267 of PDI (Fig. 3c). Furthermore, the main-chain amide group of Leu38 makes a hydrogen bond with His270 Nδ 1 atom. In contact-a′ , in addition to the hydrophobic interactions mediated by Tyr39 αSN , the peptide ligand was recognized through electrostatic interactions between the C-terminal carboxyl group of Gly41 αSN and Arg431 PDI (Supplemental Fig. S3). The extent of the interface area and the number of intermolecular interactions suggest that contact-b′ , rather than contact-a′ , primarily mediates the interaction.
To probe the peptide binding sites in solution, we examined possible spectral changes of isolated b′ and a′ domains upon addition of the α SN peptide. The results indicated that the b′ but not the a′ domain exhibited extensive chemical shift perturbations, consistent with observations of the connected b′ -a′ domains (Fig. 4). These data clearly indicate that the b′ domain provides the principal binding site of the hydrophobic segment of α SN.

Discussion
In the present study, we found that PDI can capture the hydrophobic segment of α SN primarily through its b′ domain and determined their binding mode in detail. The hydrophobic PDI-binding segment identified herein is also involved in interactions with GroEL 23 and PbaB 24 , suggesting that it displays a chaperone-philic binding motif that can be widely recognized as a mimic of the malfolded protein hallmarks. Hence, the α SN peptide employed in this study would offer a useful tool for probing chaperone interactions because of its potential broad reactivity with various molecular chaperones.
The α SN peptide contact site largely overlaps with the b′ surface involved in interactions with somatostatin and mastoparan, peptide inhibitors that compete with substrates, and with hydrophobic fluorescent probe ANS, which was previously characterized by NMR chemical shift perturbation experiments 3,5,13 . The present crystal structure successfully provides an atomic view of the molecular recognition of the substrate-binding site of PDI, which is primarily characterized by hydrophobic interactions (Fig. 3).
Our previous small-angle X-ray scattering data demonstrated that reduced-state PDI b′ -a′ domains adopt a closed conformation in which the hydrophobic ligand binding surface is supposed to be shielded from the solvent 13,14 . The crystal structure with a closed conformation of the b′ -a′ domains has been available only for human PDI with the reduced a′ active site 8,9 . Our structural model based on this crystal structure indicates that the a′ domain masks parts of the ligand binding surface of b′ and causes steric hindrance, with the α SN peptide accommodated on the b′ domain, which results in impaired interaction with the peptide in the closed conformation (Fig. 5a). This explains why this peptide preferentially binds the oxidized form of the PDI b′ -a′ domains (Fig. 5b). In this crystal structure, the peptide was stabilized in the compact β -hairpin conformation due to the crystal contacts. However, physiological substrates of PDI are generally more bulky and mobile in solution and therefore would cause more substantial steric clashes.
In summary, this study presents the first crystallographic snapshot of presumably dynamic PDI interactions with ligand peptides. Our findings provide a structural basis for the mechanisms underlying the redox-dependent substrate binding of PDI, which captures the hydrophobic segments of substrates through its hydrophobic surface that is exposed in the open conformation of the b′ -a′ domains in its oxidized form. Reduction of the a′ active site is coupled with the interdomain b′ -a′ interaction, resulting in release of the substrate with disulfide formation.

Methods
Protein expression and purification. Expression and purification of the PDI b′ -a′ domains (residues 208-449), b′ domain (residues 208-335), and a′ domain (residues 334-449) from Humicola insolens were performed as previously described 12,13,25 . To prepare the oxidized form, the purified protein (1 mg/ml) was dialyzed against 50 mM Tris-HCl (pH 8.0) containing 0.1 mM oxidized glutathione for a week. To prepare the reduced form, the protein was dissolved in a buffer containing 10 mM dithiothreitol (DTT). The expression and purification of 15 N-labeled α SN were performed as previously described 26 . Synthetic α SN peptide (Gly-Lys-Thr-Lys-Glu-Gly-Val-Leu-Tyr-Val-Gly) was purchased from Wako Pure Chemical Industries, Ltd. NMR measurements and analyses. NMR measurements were performed in 10 mM sodium phosphate buffer (pH 7.0) containing 100 mM KCl, and 10% (v/v) D 2 O using an AVANCE800 spectrometer (Bruker Biospin) equipped with a 5 mm triple-resonance cryogenic probe. To prepare the reduced form of the PDI proteins, 10 mM d-DTT was added to the buffer. The 1 H-15 N HSQC spectra were recorded at a 1 H observation frequency of 800.32 MHz with 256 (t 1 ) × 2048 (t 2 ) complex points. The spectral data of the PDI-derived proteins (at a concentration of 0.05 mM) were acquired at 303 K in the presence and absence of 0.2 mM full-length α SN or 0.2 mM α SN peptide. The spectral assignments of the PDI b′ -a′ domains, b′ domain, and a′ domain have been described previously 12 . The HSQC spectra of 15 N-labeled full-length α SN (at a concentration of 0.05 mM) were measured at 283 K in the presence and absence of 0.01-0.05 mM PDI b′ -a′ domains. The NMR assignments of α SN have been described previously 26 . Chemical shift perturbations were quantified as (0.04Δ δ N 2 + Δ δ H 2 ) 1/2 , where Δ δ H and Δ δ N are the observed chemical shift changes for 1 H and 15 N, respectively. The NMR data were processed and analyzed using TOPSPIN-2.1 (Bruker Biospin) and SPARKY 27 software. In NMR perturbation profiles, proline residues and the residues whose 1 H-15 N HSQC peaks could not be observed because of peak overlapping and/or broadening were shown by asterisks.
Protein crystallization, X-ray data collection, and structure determination. The crystals of the PDI b′ -a′ domains (10 mg/ml) complexed with α SN peptide (1:5 molar ratio) were grown in 0.1 M HEPES buffer (pH 7.5) containing 25% (w/v) PEG3350 for a week at 293 K. The crystals were directly transferred into the reservoir solution and flash-cooled in liquid nitrogen. The diffraction data set was collected using synchrotron radiation at BL44XU of SPring-8 (Japan), and was scaled and integrated using HKL2000 28 . Crystal parameters are summarized in Supplemental Table S1. The 1.60-Å resolution crystal structure of the PDI b′ -a′ domains complexed with the α SN peptide was solved by molecular replacement using the program MOLREP 29 with the isolated b′ and a′ domain coordinates derived from the crystal structure of H. insolens PDI b′ -a′ domain (oxidized form, 3WT2) 25 as search models. Model building into the electron density maps and refinement were performed using COOT 30 and REFMAC5 31 , respectively. The stereochemical quality of the final model was validated by PROCHECK 32 . The final refinement statistics are summarized in Supplemental Table S1. Molecular graphic figures were prepared using PyMOL (http://www.pymol.org/).