Redox-Dependent Domain Rearrangement of Protein Disulfide Isomerase Coupled with Exposure of Its Substrate-Binding Hydrophobic Surface

doi:10.1016/j.jmb.2009.11.049

J. Mol. Biol. (2010) 396, 361–374

Available online at www.sciencedirect.com

Redox-Dependent Domain Rearrangement of Protein Disulfide Isomerase Coupled with Exposure of Its Substrate-Binding Hydrophobic Surface Olivier Serve 1,2,3 †, Yukiko Kamiya 1,2,3 †, Aya Maeno 2 †, Michiko Nakano 2,3 †, Chiho Murakami 2 , Hiroaki Sasakawa 2,3 , Yoshiki Yamaguchi 2,4 , Takushi Harada 2 , Eiji Kurimoto 2 , Maho Yagi-Utsumi 2 , Takeshi Iguchi 5 , Kenji Inaba 6 , Jun Kikuchi 7 , Osamu Asami 8 , Tsutomu Kajino 8 , Toshihiko Oka 9 , Masayoshi Nakasako 9,10 and Koichi Kato 1,2,3 ⁎ 1

Okazaki Institute for Integrative Bioscience, National Institutes of Natural Sciences, 5-1 Higashiyama, Myodaiji, Okazaki, Aichi 444-8787, Japan 2

Graduate School of Pharmaceutical Sciences, Nagoya City University, 3-1 Tanabe-dori, Mizuho-ku, Nagoya 467-8603, Japan

3

Institute for Molecular Science, National Institutes of Natural Sciences, 5-1 Higashiyama, Myodaiji, Okazaki, Aichi 444-8787, Japan

4

Structural Glycobiology Team, Systems Glycobiology Research Group, Chemical Biology Department, RIKEN Advanced Science Institute, 2-1 Hirosawa, Wako-shi, Saitama 351-0198, Japan

Protein disulfide isomerase (PDI) is a major protein in the endoplasmic reticulum, operating as an essential folding catalyst and molecular chaperone for disulfide-containing proteins by catalyzing the formation, rearrangement, and breakage of their disulfide bridges. This enzyme has a modular structure with four thioredoxin-like domains, a, b, b′, and a′, along with a C-terminal extension. The homologous a and a′ domains contain one cysteine pair in their active site directly involved in thiol–disulfide exchange reactions, while the b′ domain putatively provides a primary binding site for unstructured regions of the substrate polypeptides. Here, we report a redox-dependent intramolecular rearrangement of the b′ and a′ domains of PDI from Humicola insolens, a thermophilic fungus, elucidated by combined use of nuclear magnetic resonance (NMR) and small-angle X-ray scattering (SAXS) methods. Our NMR data showed that the substrates bound to a hydrophobic surface spanning these two domains, which became more exposed to the solvent upon oxidation of the active site of the a′ domain. The hydrogen–deuterium exchange and relaxation data indicated that the redox state of the a′ domain influences the dynamic properties of the b′ domain. Moreover, the SAXS profiles revealed that oxidation of the a′ active site causes segregation of the two domains. On the basis of these data, we propose a mechanistic model of PDI action; the a′ domain transfers its own disulfide bond into the unfolded protein accommodated on the hydrophobic surface of the substrate-binding region, which consequently changes into a “closed” form releasing the oxidized substrate. © 2009 Elsevier Ltd. All rights reserved.

5

Bioscience Research Laboratory, Fujiya Co., Ltd., Hadano, Kanagawa 257-0031, Japan

6

Medical Institute of Bioregulation, Kyushu University, 3-1-1 Maidashi, Higashi-ku, Fukuoka 812-8582, Japan *Corresponding author. Okazaki Institute for Integrative Bioscience, National Institutes of Natural Sciences, 5-1 Higashiyama, Myodaiji, Okazaki, Aichi 444-8787, Japan. E-mail address: [email protected]. † O.S., Y.K., A.M., and M.N. contributed equally to this work. Present address: E. Kurimoto, Faculty of Pharmacy, Meijo University, 150 Yagotoyama, Tempaku-ku, Nagoya 468-8503, Japan. Abbreviations used: PDI, protein disulfide isomerase; SAXS, small-angle X-ray scattering; ANS, 1-anilino-8naphthalenesulfonate; SPR, surface plasmon resonance; RNase A, ribonuclease A; DTT, dithiothreitol; HSQC, heteronuclear single-quantum coherence; NOESY, fnuclear Overhauser enhancement spectroscopy.

0022-2836/$ - see front matter © 2009 Elsevier Ltd. All rights reserved.

362

Redox-Dependent Conformational Rearrangement of PDI

7

RIKEN Plant Science Center, 1-7-22 Suehiro-cho, Tsurumi-ku, Yokohama 230-0045, Japan

8

Toyota Central Research & Development Laboratories, Inc., Nagakute, Aichi 480-1192, Japan

9

Department of Physics, Faculty of Science and Technology, Keio University, 3-14-1 Hiyoshi, Kohoku-ku, Yokohama 223-8522, Japan

10

RIKEN Harima Institute, Sayo-cho, Sayo-gun, Hyogo 679-5148, Japan Received 16 October 2009; received in revised form 19 November 2009; accepted 19 November 2009 Available online 26 November 2009

Edited by P. Wright

Keywords: protein disulfide isomerase; NMR; SAXS; molecular chaperone; domain rearrangement

Introduction Maturation of proteins in cells is assisted by molecular chaperones, which inhibit aggregation of unstructured proteins, and by enzymes that accelerate protein folding.1,2 Protein disulfide iso-

merase (PDI), a major protein in the lumen of the endoplasmic reticulum and a member of the thioredoxin superfamily, operates as an essential folding catalyst and a molecular chaperone.3,4 This enzyme introduces disulfide bonds into protein substrates and catalyzes the rearrange-

Fig. 1. Structural and functional asymmetry of the PDI domains. (a) Domain organization of H. insolens PDI based on the crystal structure of yeast PDI. (b) 13C NMR spectra of the PDI labeled with 13C selectively at the carbonyl carbon atoms of cyst(e)ine residues. The spectra were measured in the absence (top) and presence (bottom) of 0.8 molar equivalent of reduced glutathione. The residue numbers are displayed on top of the peaks. (c) SPR analyses of the interactions of the full-length PDI (red), a–b (green), and b′–a′ (blue) with disulfide-scrambled (top) and native RNase A (bottom). (d) Fluorescence spectra of ANS in the presence of b′–a′oxi, b′–a′red, a–boxi, and a–bred.

Redox-Dependent Conformational Rearrangement of PDI

363

Table 1. Structural statistics of the b′ and a′ domains of thermophilic fungal PDI

thioredoxin fold.9–12 The mutational and crosslinking analyses indicate that the b′ domain provides the principal peptide-binding site in PDI, but the other domains are a prerequisite for binding to the larger proteins.13 The crystal structure of yeast PDI has revealed that the four thioredoxin domains are arranged in a twisted “U” shape, with the active site of the a and a′ domains facing each other. 10 In the a and a′ domains, hydrophobic patches surround their active sites. Furthermore, hydrophobic patches of the individual four domains, displayed at the same relative position, form a continuous hydrophobic area on the inside surface of the “U.” The hydrophobic area, which accommodates the b domain of a symmetry-related molecule in the crystal, possibly provides a platform for interaction of PDI with unfolded or misfolded proteins to suppress their aggregation. Tsai et al.14 demonstrated that PDI acts as a redox-dependent chaperone to unfold the cholera toxin A1 chain. The reduced PDI molecule binds the substrate, and the oxidized form of PDI releases the substrate. The authors have suggested that redox-dependent conformational changes in PDI regulate the catch and release of the substrate, on the basis of the observation that PDI becomes more susceptible to protease digestion upon oxidation. However, such redox-dependent regulation of polypeptide-binding activity of PDI is still controversial,15 and the putatively proposed conformational change has been poorly characterized. In this study, we focused on redox-dependent conformational alteration of PDI from Humicola

Restraints used in structure calculation No. of distance restraints No. of hydrogen bonds No. of torsion angle restraints ϕ ψ Geometric statistics r.m.s.d.s from the mean structure (Å)a Backbone atoms All heavy atoms Ramachandran analysis (%) Most favored regions Additional allowed regions Generously allowed regions Disallowed regions

b′

a′

2247 35

2199 31

70 70

96 96

0.56 1.95

0.66 1.34

80.9 16.5 7.0 0.6

80.7 18.9 0.4 0.0

a Mean coordinates were obtained by averaging coordinates of the 10 calculated structures.

ment of the incorrect disulfides during the process of protein folding. The chaperone function of PDI is essential during the catalysis of disulfide bond isomerization.5,6 PDI has a modular structure with four globular domains, a, b, b′, and a′, along with a C-terminal acidic extension (Fig. 1a).7 The homologous a and a′ domains have thioredoxin folds, each of which contains a cysteine pair in a WCGHCK active-site sequence motif directly involved in the thiol exchange reaction.8 Although the homologous b and b′ domains share no obvious sequence similarity to either a or a′ domain and contain no active-site sequence motif, these two domains also adopt a

Fig. 2. The converged structures of the b′ and a′ domains of the thermophilic fungal PDI. The structural models are illustrated as an ensemble of the final 10 lowest-energy structures. These images were prepared using MOLMOL.22

364

Redox-Dependent Conformational Rearrangement of PDI

insolens, a thermophilic fungus, in solution.16,17 We considered the fragment comprising the b′ and a′ domains, which are connected to each other through a 17-residue segment referred to as loop “x” (Fig. 1a and Supplementary Fig. 1). Our nuclear magnetic resonance (NMR) data enabled us to map amino

acid residues involved in substrate binding on the structural models of the two domains. In addition, the NMR and small-angle X-ray scattering (SAXS) data indicate a remarkable exposure of the substrate-binding hydrophobic surface of the PDI regulated through intramolecular domain rear-

Fig. 3. Redox-dependent structural change of b′–a′ probed by NMR. (a) A part of the 1H–15N HSQC spectra of b′–a′oxi (black) and b′–a′red (red) recorded at the 1H observation frequency of 920 MHz. (b) Chemical shift changes on reduction of the cleavage of the disulfide bridge in the a′ domain. The data are shown according to the equation (0.04δ2N + δH2)1/2, where δN and δH represent the change in nitrogen and proton chemical shifts, respectively. Chemical shift change values are mean ± SD of three independent experiments. The asterisks indicate (0.04δ2N+ δH2)1/2 N 0.3 ppm for Ala362 (1.02 ppm), Cys365 (2.81 ppm), Cys368 (3.09 ppm), and Lys369 (0.62 ppm). (c) Redox-dependent changes in the profiles of hydrogen– deuterium exchange rates of amide protons and (d) mapping of residues on the 3D model of b′–a′red according to the Koxi/Kred values with the color gradient indicating the magnitude of the enhancement of the exchange rate upon the oxidation. The dots indicate Koxi/Kred N 20. The residues whose Koxi/Kred value could not be determined because of rapid hydrogen–deuterium exchange in b′–a′red (Kred N 2.7 × 10−2 min−1) are shown by gray bars. Proline residues and the residues whose 1H–15N HSQC peaks could not be used as a probe because of peak overlapping and/or broadening are shown by asterisks. (e) 15N R2 relaxation rates of b′–a′red and b′–a′oxi and (f) mapping of 15N R2 values on the 3D model of b′–a′red with a linear color gradient, the scale is from 0 (blue) to 70 Hz (red). The side chains are displayed for the residues exhibiting a larger R2 value (N 40 Hz ), while the residues whose R2 value could not be determined are shown in gray. The asterisks indicate R2 N 50 Hz for Lys265 (56 Hz) and Phe267 (70 Hz).

365

Redox-Dependent Conformational Rearrangement of PDI

Fig. 3 (legend on previous page)

rangement coupled with the redox states of the a′ active site.

Results Functional nonequivalence of the homologous domains of PDI We first compared the redox states of the two active sites using the 13C NMR signals originating from the cyst(e)inyl carbonyl groups of the fulllength PDI protein as spectroscopic probes. The 13C NMR spectral data show that Cys30 and Cys33 forming the a domain active site are exclusively reduced upon addition of 0.8 molar equivalence of reduced glutathione, while the disulfide bridge between Cys365 and Cys368 in the a′ domain active site remains intact (Fig. 1b). These data clearly demonstrate that the two active sites exhibit different redox properties in the full-length PDI molecule. The a active site tends to be reduced, while that of the a′ domain prefers the oxidized form. The observed asymmetry of the redox states of thermophilic fungal PDI, along with the evidence for yeast and mammalian PDIs reported thus far,18,19 suggests a functional nonequivalence of the a domains in the N-terminal region and the a′ domains in the C-terminal region of this molecule.

To address this issue, we expressed and characterized the two truncated fragments of PDI separately: one comprising the N-terminal a and b domains (termed a–b) and the other comprising the Cterminal b′ and a′ domains (termed b′–a′). Using disulfide-scrambled ribonuclease A (RNase A) as a ligand, we subjected these fragments to a surface plasmon resonance (SPR) analysis to identify the principal substrate-binding region of PDI.5 The b′–a′ fragment, but not the a–b fragment, interacts with the scrambled RNase A under oxidizing conditions at a comparable level as with the full-length PDI, while the native RNase A binds with neither of these constructs (Fig. 1c). This indicates that the b′–a′ region is primarily responsible for the recognition of misfolded proteins, consistent with what has been reported for human PDI based on a cross-linking experiment.13 Since PDI has been postulated to interact with the substrate through its hydrophobic patches,10 the degrees of exposure of hydrophobic surfaces of a–b and b′–a′ were examined in the presence and absence of 10 mM dithiothreitol (DTT) with a widely used hydrophobic probe, 1-anilino-8-naphthalenesulfonate (ANS), which becomes fluorescent when bound to nonpolar sites of proteins.20 A dramatic enhancement of ANS fluorescence was induced by b′–a′, particularly in the oxidized condition. This clearly demonstrates that the hydrophobic surfaces of b′–a′ are more solvent-exposed than a–b and are further

366

Redox-Dependent Conformational Rearrangement of PDI

Fig. 3 (legend on page 364)

expanded upon formation of the disulfide bridge at the a′ domain active site (Fig. 1d). Hereafter, the oxidized and reduced forms of the b′–a′ fragment will be referred to as b′–a′oxi and b′–a′red, respectively. Similar notations will be used for a–b.

Solution structures of the substrate-binding domains of PDI For a detailed analysis of the redox-dependent conformational change of b′–a′ detected by the ANS

Redox-Dependent Conformational Rearrangement of PDI

367

Fig. 4. Chemical shift perturbations upon binding with mastoparan. (a) Chemical shift changes in b′–a′oxi and b′–a′red upon binding with mastoparan. The data are shown according to the equation in Fig. 3B. For Trp364, the indol 1H–15N HSQC peak was used as a spectroscopic probe (orange bar) instead of the backbone amide peak, which was not detectable due to broadening in b′–a′oxi. (b) Mapping on the 3D model of b′–a′red of residues exhibited chemical shift perturbations [(0.04δ2N + δH2)1/2 N 0.02 ppm] upon addition of one molar equivalent of mastoparan in b′–a′red (top) and b′–a′oxi (bottom). Red gradient indicates the strength of the perturbation. The proline residues and the residues whose 1H–15N HSQC peak could not be used as a probe due to broadening and/or overlapping are shown in gray. The catalytic cysteine residues appear in green. Structural images were prepared with PyMOL (http://www.pymol.org/).

368

Redox-Dependent Conformational Rearrangement of PDI

fluorescence experiment, we carried out NMR analyses of these substrate-binding domains. The a′ domain dissolved at a concentration of 1 mM under reducing condition, but precipitated upon oxidation. The instability of the isolated a′ domain under oxidizing conditions was also reported for human PDI based on the circular dichroic data.21 Hence, we determined the NMR structures of the isolated a′ and b′ domains in the presence of 10 mM DTT with sufficient statistics (Table 1 and Fig. 2) on the basis of the NMR spectral assignments reported previously.23

Although the solution structures of these domains are similar to those of their counterparts in the crystal structure of yeast PDI (r.m.s.d. of 1.8 Å for b′ domain and 1.9 Å for a′ domain), the b′ domain of thermophilic fungal PDI lacks the short helical segment (α5) preceding the last helix observed for yeast PDI (Supplementary Fig. 1). The recently reported structures of human PDI b′ domain also lack the corresponding helical segment.11,12 The structures of His367 and Cys368 forming the active site are not converged well because of the significant

Fig. 5 (legend on next page)

Redox-Dependent Conformational Rearrangement of PDI

369

line broadening of 1H–15N heteronuclear singlequantum coherence (HSQC) peaks of these residues. A similar phenomenon has been reported for the a domain of human PDI.8

Intriguingly, the magnitudes of the chemical shift changes differed between b′–a′oxi and b′–a′red for some of the amino acid residues. One of the most remarkable examples was the indole 1H–15N HSQC peak of Trp364, which exhibited virtually no perturbation under reducing condition, but showed a significant chemical shift change upon binding to mastoparan. These data suggest that the redoxdependent conformational alteration of b′–a′ affects its ligand-binding mode.

Redox-dependent behaviors of b′–a′ characterized by NMR We performed further NMR analyses of b′–a′ to characterize its redox-dependent conformational changes. A comparison of 1H–15N HSQC spectra between b′–a′oxi and b′–a′red indicated that the cleavage of the disulfide bridge at the active site of the a′ domain induced chemical shift perturbations in the peaks from the residues not only in the a′ domain, but also, although to a lesser extent, in a part of the b′ domain (Fig. 3a and b). The oxidation of the a′ domain active site resulted in dramatic increases in hydrogen–deuterium exchange rates of the backbone amide of the b′ domain (Fig. 3c and d). Furthermore, the relaxation data showed a general increase in transverse relaxation rate R2 upon oxidation. More specifically, we can note the remarkable increases in the R2 values of Lys265 and Phe267 in the b′ domain along with those of the a′ residues Ile348, Val349, Gly379, Ala394, and Phe439, which are most likely due to substantial conformational exchange (Fig. 3e and f). These data indicate that the redox state of the active site of the a′ domain influences the dynamic behavior of the neighboring b′ domain. It is intriguing to examine whether the redoxdependent domain interaction affects substrate recognition by full-length PDI. Using the NMR peaks as spectroscopic probes, we investigated the interactions of b′–a′ with mastoparan, which is often used as a competitive peptide inhibitor against the substrates.24,25 Upon titration of mastoparan, chemical shift perturbations were observed in the residues located in a surface area spanning the domain–domain interface including hydrophobic residues (Fig. 4 and Supplementary Fig. 2). Although the chemical shift perturbations observed for the b′ domain are consistent with those reported for the b–b′ fragment of human PDI,12 significant chemical shift changes were also induced in the amino acid residues located in the a′ domain.

Domain rearrangement in b′–a′ revealed by SAXS In order to examine the redox-dependent conformational changes of the b′–a′ fragment, the structural parameters and the molecular shape of the fragment were investigated by SAXS under reducing and oxidizing solvent conditions. For comparison, we also measured the SAXS of the a–b fragment, which was expected to display absence of significance conformational changes from the spectroscopic measurements. The SAXS profiles of the fragments were roughly similar with those calculated from the corresponding regions of the crystal structure of yeast PDI (Fig. 5a and b). However, it should be noted that the experimentally obtained profiles of b′–a′oxi and b′– a′red displayed small but significant differences (Fig. 5a and Supplementary Fig. 3), while those from the a–b fragment were almost independent of the redox conditions (Fig. 5b and Supplementary Fig. 3). The monodispersive properties of a–boxi, a–bred, b′–a′oxi, and b′–a′red in the concentration range measured so far were confirmed from their Guinier plots approximated by single regression lines and also from the linear concentration dependencies of the radii of gyration and the zero-angle scattering intensities (Fig. 5c). The apparent molecular mass, the radius of gyration (Rgmol), and the maximum dimensions (Dmax) of the b′–a′oxi were larger than those of the b′–a′red (Table 2). The slightly larger value of apparent molecular mass of the b′–a′oxi than that of b′–a′red may be explained by the association of solvent molecules. The molecular structures of the a–boxi, a–bred, b′– a′oxi, and b′–a′red fragments were restored from

Fig. 5. Redox-dependent structural changes of a–b and b′–a′ probed by SAXS. (a) SAXS profiles of the b′–a′red (red symbols) and the b′–a′oxi (blue dots) are plotted in a semilogarithmic form and shifted appropriately along the ordinate for clarity. In the insets, the low-resolution molecular models predicted by GASBOR26 are shown as fishnets representing the molecular envelops of the averaged models for b′–a′red (red) and b′–a′oxi (blue). The molecular shapes are compared with the b′–a′ region in the crystal structure of yeast PDI [Protein Data Bank (PDB) accession number: 2B5E]. The scattering profiles of GASBOR-predicted structural models are shown as black lines in the upper two profiles. The profiles at the bottom are the experimentally obtained SAXS profile of b′–a′red (red symbols) and simulated profiles from the b′–a′ regions of the crystal structures of yeast PDI crystallized at 277 K (black dots) and 295 K (green). (b) SAXS profiles and the restored structural models of the a–bred (orange symbols) and the a–boxi (green dots) are shown as in (a). The predicted molecular shapes are compared with the a–b region in the crystal structure of yeast PDI. The bottom profiles are the experimentally obtained SAXS profile of a–bred (orange symbols) and profiles computed from the b′–a′ regions of the crystal structures of yeast PDI at 277 K (black dots) and 295 K (purple). (c) Guinier plots of a–bred (orange symbols), a–boxi (green), b′–a′red (red), and b′–a′oxi (blue). The inset shows the concentration-dependent variations of C/I(0,C) and Rg(C) of a–bred (orange), a–boxi (green), b′–a′red (red), and b′–a′oxi (blue). In the inset, the continuous regression lines are calculated by the least-squares method assuming the linear concentration dependence of I(0,C) and Rg. (d) Guinier plots of a–bred (orange), a–boxi (green), b′–a′oxi (blue), and b′–a′red (red) extended to higher scattering angle to estimate Rgdom. The Rgdom values are calculated from the region indicated by arrows by the least-squares method.

370

Redox-Dependent Conformational Rearrangement of PDI

Table 2. Structural parameters of a–b and b′–a′ fragments from SAXS Molecular mass (kDa) from Amino acid sequence

SAXS

Rg(C = 0) (Å)

Rgdom (Å)

ddom (Å)

Dmax (Å)

24.8

21.4 ± 1.2 22.1 ± 0.3 26.0 ± 1.1 24.0 ± 0.6

22.7 ± 0.3 22.4 ± 0.5 24.9 ± 0.4 23.1 ± 0.2

13.5 ± 0.3 13.1 ± 0.2 14.4 ± 0.2 14.8 ± 0.2

36.5 ± 1.2 36.3 ± 1.4 40.6 ± 1.2 35.5 ± 0.9

69 ± 3 70 ± 3 76 ± 3 72 ± 3

30.6 35.7

73.2 73.1

a–boxi a–bred b′–a′oxi b′–a′red

24.8

Crystal structure of the yeast PDI (PDB accession number 2B5E) a–b b′–a′

20.0 22.0

13.0 (a)/12.5 (b) 13.9 (b′)/13.1 (a′)

Structural parameters of a, b, b′, and a′ domains and fragments of yeast PDI are calculated by using the program CRYSOL.26 These ddom values were almost independent of temperatures in crystallization.

their SAXS profiles with the GASBOR program and appeared as dumbbell shapes resembling each other. The two globular clusters of dummy residues were interpreted to correspond to the thioredoxin domains. Because the GASBOR program does not provide a unique solution for the three-dimensional (3D) structure, the molecular structures are obtained by the alignment and averaging for models from 10 independent calculations. The b′–a′ fragment displays small but significant redox-dependent structural changes, while the a–b is independent of the redox conditions of their buffer. The structural differences were quantitatively evaluated by the structural parameters obtainable under the assumption that the molecular shape of either fragment is a dumbbell shape composed of two identical spheres with the radii of gyration Rgdom separated by a connector of length ddom. In such a case,27 the Rgmol is approximated as R2g = R2gdom + d2dom = 4 The Rgdom values were calculated from Guinier approximations for higher scattering-angle regions. In all cases, the Rgdom values of each fragment were independent of the redox conditions of the sample solution and close to those calculated from the crystal structures of yeast PDI (Table 2). The calculated ddom of the b′–a′ is increased by 5 Å under the oxidizing condition, while the ddom of the b′–a′red is close to those of the b′–a′ region in the crystal structure of yeast PDI as well as those of the a–boxi and the a–bred. Based on the theoretical scattering curves from dumbbell-shaped particles with various structural dimensions,28 the increase in the ddom reflects the separation of the dumbbell spheres and/or the swell of the connector region. The changes in the Dmax and the ddom of the b′–a′ suggest redox-dependent rearrangements of domains and the structural nonequivalence between a–boxi and b′–a′oxi.

Discussion Nonequivalent roles of the four homologous domains of PDI in its enzymatic and chaperone functions have been postulated on the basis of

biochemical and mutational data. For instance, the a′ domain is more involved in apparent steadystate substrate binding, while the a domain contributes more to catalysis at saturating concentration of the substrate. 18 Additionally, the b′ domain functions as a substrate-binding site more importantly than does the b domain.10,12,13,29,30 The recent crystallographic studies of yeast PDI have provided a structural framework for further investigation of the underlying mechanisms of the actions of this modular enzyme.10,31 On the basis of these previous studies, we have used the combined approach of NMR and SAXS to provide insights into the dynamics of PDI with relevance to its biological functions. First, 13C NMR spectral data have shown that the disulfide bond of the a′ active site is more stable than that of the a active site (Fig. 1b). In addition, SPR and ANS fluorescence experiments clearly demonstrated that b′–a′ exhibits an exposed hydrophobic surface and forms a primary substrate-binding site under oxidizing conditions. These data are consistent with the functional asymmetry of the four domains in the PDI molecule. Moreover, the present NMR data demonstrate that the solvent-accessible surface spanning from the hydrophobic patch of the b′ to the active site of the a′ domains is responsible for peptide binding (Fig. 4). The substrate-binding site thus identified constitutes the continuous hydrophobic area of the inner surface of the U-shaped structure of PDI, supporting the crystal structurebased model in which the substrate is putatively sequestered in the hydrophobic cleft and, thereby, subjected to the attack of the two active sites.10 Recently, Tian et al. reported that yeast PDI crystallized at different temperatures, that is, 4 °C and 22 °C, exhibited significant change in relative positions and orientation of the a domain, and to a lesser extent the a′ domain, with respect to the b and b′ domains, suggesting a flexible property of this molecule.31 They also showed that restriction of domain movements by artificially introduced interdomain disulfide bonds results in impairment of enzymatic activity of yeast PDI in vitro and in vivo.31 These data suggest the functional importance of domain movement during the enzymatic action of PDI. The present SAXS data demonstrate that oxidation of the a′ domain causes domain rear-

Redox-Dependent Conformational Rearrangement of PDI

371

rangement of b′–a′ in solution. The molecular shapes of both b′–a′oxi and b′–a′red restored from the profiles are approximated as dumbbells with two globular clusters corresponding to the b′ and a′ domains. However, the details of the shapes, such as the thickness of the connecter region of two globular clusters, are significantly different between b′–a′oxi and b′–a′red (Fig. 5a and Supplementary Fig. 3). This shape alteration is in accordance with the interpretation of D dom in the scattering theory of a dumbbell-shaped particle:28 Ddom reflects the center-to-center distance between the dumbbell spheres and/or the thickness of the connector region. The SAXS profiles also showed that the apparent molecular mass value of b′–a′oxi was slightly larger than that of b′–a′red, suggesting an increased binding of solvent molecules. This is probably because of the changes in the surface structures, that is, the interface between the globular clusters becomes more open to expose the surface in b′–a′oxi. These SAXS data along with the spectroscopic data indicate that oxidation of the a′ active-site induces segregation of the b′ and a′ domains, resulting in the exposure of a hydrophobic surface. The present study also demonstrates that a redox-dependent domain rearrangement does not occur in a–b, providing the molecular basis of the functional nonequivalence of a–b and b′–a′ in the PDI molecule. On inspection of these data, we suggest that the a′ domain transfers its own disulfide bond into the unfolded protein accommodated on this hydrophobic surface and consequently changes into a “closed” form liberating the oxidized substrate. A misfolded protein with incorrect disulfide bridge(s) might also be trapped by this hydrophobic surface and thereby subjected to the attack presumably by the thiol group of the a domain for disulfide breakage. Although the redox potentials of the a and a′ active sites of PDIs can be different among different species,32,33 the cysteine pair in the a′ domain has been shown to be selectively oxidized by the flavoprotein Ero1p.34 All these data are consistent with the proposed functional asymmetry of PDI molecules. The a and a′ domains serve opposing functional roles as a disulfide isomerase and a disulfide oxidoreductase, respectively. It is intriguing to address the issue regarding the microenvironmental change of the active site, which, triggered by the cleavage of the disulfide bridge, leads to the rearrangement of the substratebinding domains of PDI. PDI contains a histidine residue between the two cysteine residues in each active site, and this residue appears to destabilize the oxidized form, particularly in the a′ domain, through a histidine–cysteine ion pair.35 In the crystal of yeast PDI, these residues do not interact with each other, and the histidine residue in question is involved in the crystal contacts with the b domain of a symmetry-related molecule.10 The present NMR data suggest that the active site of the a′ domain exhibits conformational fluctuation in solution. For the next stage of the research, 3D structures of the oxidized as well as the reduced

forms of PDI at atomic resolution should be determined, taking into account their dynamic properties in solution. It is also intriguing to examine the conformational alteration of b′–a′ in the full-length PDI molecule. In conclusion, our findings demonstrate that PDI exhibits an intramolecular domain–domain rearrangement in a redox-dependent manner, which is coupled with shielding and exposure of substratebinding hydrophobic surface. This redox-dependent domain rearrangement would provide a structural basis for the mechanism of catch and release of the substrate by PDI and possibly by other members of the PDI family.

Experimental procedures Reagents Mastoparan and RNase A were obtained from Sigma. Scrambled RNase A was prepared as described previously.36 Deuterated and undeuterated DTT were purchased from Cambridge Isotope Laboratories and Wako, respectively, while ANS was purchased from Nacalai Tesque. Protein expression and purification Expression and purification of the b′ (208–335) and a′ (334–449) domains of H. insolens PDI were performed as described previously.23 Full-length PDI (1–485) and a Cterminally truncated mutant a–b fragment (1–219) were expressed with an N-terminal hexahistidine tag and purified by Ni2+–NTA affinity chromatography, while an N-terminally truncated mutant b′–a′ fragment (208– 449) was produced with an N-terminal glutathione Stransferase tag and purified by glutathione–Sepharose chromatography. After removal of the tags, the proteins were further purified as described previously.23 Oxidized PDI and b′–a′oxi were prepared by dialyzing the sample against 50 mM Tris–HCl (pH 8.0), including 0.1 mM oxidized glutathione, at 277 K for a week. NMR measurements and structure determination The NMR assignments of the b′ and a′ domains have been described previously.23 For structure determination of the b′ and a′ domains, distance restraints were obtained from the 2D nuclear Overhauser enhancement spectroscopy (NOESY), 3D 15N-edited NOESY, and 13C-edited NOESY spectra, with a mixing time of 100 ms. The CYANA software package37 was used for NOE assignments and initial structure calculations. The final structural models were obtained using a standard simulated annealing protocol with CNS 1.238 using hydrogenbonding data as conformational restraints. The generated structures have been subsequently refined in the presence of water molecules. NMR spectra of b′–a′ were acquired at 303 K with Bruker DMX500, Avance600, DRX800, and JEOL ECA-920 spectrometers. The protein solution subjected to heteronuclear NMR measurements39 contained 0.1–1 mM b′–a′ in 10 mM sodium phosphate buffer (pH 6.0) containing 100 mM KCl, 0.05% (w/v) NaN3, 10% 2H2O, and 10 mM

372

Redox-Dependent Conformational Rearrangement of PDI

[2H10]DTT, which was omitted in the experiments using b′–a′oxi. The backbone assignments for b′–a′ were performed using 3D HNCA and HN(CO)CA and confirmed by amino-acid-selective labeling methods. In hydrogen–deuterium exchange experiments, the time course of 1H–15N HSQC spectral change was monitored on a Bruker DMX500 spectrometer after addition of 2H2O to the protein lyophilized in the solution. The measurement period for each spectrum was approximately 60 min. To determine the exchange rates of b′–a′oxi and b′–a′red (designated as Koxi and Kred, respectively), plots of the peak volume versus time were fitted to a single-exponential decay. The Koxi/Kred ratio was calculated for each amide proton. 15 N relaxation measurements were recorded at 920 MHz of 1H observation frequency at a temperature of 303 K. R2 data sets were recorded in a randomized order regarding relaxation delays. Visualization of spectra, and measurements of peaks intensities were performed with Sparky.40 Estimation of R2 was performed with the Relax 1.3.4 program.41 1D 13C NMR spectra of PDI were measured at 310 K on an AMX400 spectrometer as described previously,42,43 using a 10-mm NMR sample tube containing 0.25 mM PDI dissolved in 10 mM sodium phosphate buffer containing 100 mM KCl and 0.05% (w/v) NaN 3 (pH 7.3). The assignments of carbonyl 13C signals were carried out by the double-labeling method and spectral comparison between the full-length PDI and its truncated mutants a–b and b′–a′.42–44

SAXS experiments and analyses

Modeling of the b′–a′ structure The solution structures of b′ and a′ were superimposed onto the corresponding domains in the crystal structure of the yeast PDI using PyMOL's plugin CEAlign.45 Then, the b′ and a′ domains were merged into one polypeptide through the linker generated by replacing amino acid residues constituting the linker region of yeast PDI with those of H. insolens PDI. Subsequently, the structure was energy-minimized with CNS 1.2. SPR analyses SPR-based binding assay was performed at 298 K using BIAcore 2000 (Biacore AB). HBS buffer [10 mM Hepes, 150 mM NaCl, 3.4 mM ethylenediaminetetraacetic acid, 0.005% (v/v) Tween 20, pH 7.4] was used as a running buffer at a flow rate of 20 μl min−1. The full-length PDI and its truncated fragments a–b and b′–a′ were individually immobilized on CM5 sensor chips (Biacore AB) according to the manufacturer's instructions. The binding response was measured after injection of native RNase A or scrambled RNase A solution. Fluorescence measurements Fluorescence spectra were recorded at 293 K using a 1cm cuvette on a Hitachi F-4500 spectrofluorometer. A solution of a–boxi, a–bred, b′–a′oxi, or b′–a′red at a concentration of 2 μM was mixed with a concentrated solution of ANS in 10 mM phosphate buffer containing 100 mM KCl (pH 6.0) and incubated at 310 K for 1 h. The final ANS concentration was 140 μM. The redox state of PDI was monitored by tryptophan fluorescence measurement as described in the literature.46

SAXS data were collected at the BL40B2 station of SPring-8 using the procedure described previously.47 All protein solutions were prepared in a 10 mM Hepes buffer (pH 6.0) containing 0.1 M KCl. The SAXS profiles of b′–a′oxi and b′–a′red were collected in the concentration range of 4.0–10.1 and 4.0–10.0 mg/ml, respectively, with an increment of ca 2.0 mg/ml. Prior to the measurements of b′–a′red, the solutions of b′–a′oxi and DTT (final concentration, 10 mM) were mixed and incubated at 303 K for 30 min. Scattering profiles were obtained by circularly averaging the 2D-recorded scattering patterns after background subtraction. The profiles in the small-angle region were analyzed by Guinier's approximation.48 The scattering intensity I(S,C), as a function of scattering vector S and protein concentration C, is expressed by the forward scattering intensity, I(0,C), and the radius of gyration, Rg (C), as h i I ðS; CÞ = I ð0; CÞexp −4k2 = 3 × Rg ðCÞ2 S2 ; S = 2 sin u = E where 2θ is the scattering angle and λ is the X-ray wavelength. The analyses were carried out according to the procedure described previously.47 When the sample solution is monodispersive, the I(0,C) and Rg(C) values vary linearly depending on the concentration: The I(0, C = 0) gave an estimate of apparent molecular mass of solute, and those of the fragments were estimated using the concentration-dependent variation of lysozyme as a reference sample. In the Guinier plot for a dumbbell-shaped molecule comprising two globular domains with closely similar structures and sizes such as a–b or b′–a′, there are two regions to be approximated by straight lines.28 The slope in the conventional Guinier region gives Rgmol of the whole molecule as described above, and that of the second straight region in the high scattering-angle gives an averaged radius of gyration of the globular domains (Rgdom). The molecular structures were restored by applying the GASBOR26 program to the SAXS profiles extrapolated to the dilution limits and represented as an assembly of dummy residues. The discrepancy is monitored with the χ2 values defined as v2 = 1=ðN −1Þ

X         2 c Sj Iexp Sj − KImodel Sj =j Sj j

where N is the number of experimental data points, c(Sj) is a correction factor, K is a scaling factor, and σ(Sj) is the statistical error of the experimental scattering profile Iexp (Sj) at the scattering vector Sj. Imodel(Sj) is the profile of the predicted structural model. Because molecular models restored are not unique, 10 independent calculations were performed for each scattering profile. Finally, the molecular model is presented as an averaged structure of superimposed 10 models. PDB accession numbers The atomic coordinates of a′ and b′ have been deposited in the PDB. The accession numbers are 2KP1 and 2KP2, respectively.

Redox-Dependent Conformational Rearrangement of PDI

Acknowledgements We thank Dr. Katsuaki Inoue for his support in the SAXS experiments with the SPring-8 BL40B2, which was carried out under an approved proposal (2005B0380). We acknowledge Dr. Charles Schwieters (NIH) for his useful advice. We are grateful to Dr. Uno Tagami (Ajinomoto Co., Inc.) for her help with the construction of the 3D model. We thank Kiyomi Senda, Kumiko Hattori, Yukie Kito, and Yumiko Wada for their help in the preparation of the recombinant proteins for use in NMR spectroscopy. M.Y.-U. is a recipient of the Japan Society for the Promotion of Science Research Fellowships for Young Scientists. This work was supported in part by grants-in-aid (15032249, 17028047, 15076210, 18870023, 20050030, 20059030, 20107004, 21370050, and 21870052), by Protein 3000 Project and Nanotechnology Network Project from the Ministry of Education, Culture, Sports, Science and Technology of Japan, and by NINS Cooperative Project Biomolecular Sensor and the Joint Studies Program of the Institute for Molecular Science from National Institute of Natural Sciences.

Supplementary Data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/ j.jmb.2009.11.049

References 1. Hartl, F. U. & Hayer-Hartl, M. (2002). Molecular chaperones in the cytosol: from nascent chain to folded protein. Science, 295, 1852–1858. 2. Buchner, J. (1996). Supervising the fold: functional principles of molecular chaperones. FASEB J. 10, 10–19. 3. Wilkinson, B. & Gilbert, H. F. (2004). Protein disulfide isomerase. Biochim. Biophys. Acta, 1699, 35–44. 4. Freedman, R. B., Klappa, P. & Ruddock, L. W. (2002). Protein disulfide isomerases exploit synergy between catalytic and specific binding domains. EMBO Rep. 3, 136–140. 5. Zheng, J. & Gilbert, H. F. (2001). Discrimination between native and non-native disulfides by proteindisulfide isomerase. J. Biol. Chem. 276, 15747–15752. 6. Winter, J., Klappa, P., Freedman, R. B., Lilie, H. & Rudolph, R. (2002). Catalytic activity and chaperone function of human protein-disulfide isomerase are required for the efficient refolding of proinsulin. J. Biol. Chem. 277, 310–317. 7. Edman, J. C., Ellis, L., Blacher, R. W., Roth, R. A. & Rutter, W. J. (1985). Sequence of protein disulphide isomerase and implications of its relationship to thioredoxin. Nature, 317, 267–270. 8. Kemmink, J., Darby, N. J., Dijkstra, K., Nilges, M. & Creighton, T. E. (1996). Structure determination of the N-terminal thioredoxin-like domain of protein disulfide isomerase using multidimensional heteronuclear 13C/15N NMR spectroscopy. Biochemistry, 35, 7684–7691.

373 9. Kemmink, J., Dijkstra, K., Mariani, M., Scheek, R. M., Penka, E., Nilges, M. & Darby, N. J. (1999). The structure in solution of the b domain of protein disulfide isomerase. J. Biomol. NMR, 13, 357–368. 10. Tian, G., Xiang, S., Noiva, R., Lennarz, W. J. & Schindelin, H. (2006). The crystal structure of yeast protein disulfide isomerase suggests cooperativity between its active sites. Cell, 124, 61–73. 11. Nguyen, V. D., Wallis, K., Howard, M. J., Haapalainen, A. M., Salo, K. E., Saaranen, M. J. et al. (2008). Alternative conformations of the x region of human protein disulphide-isomerase modulate exposure of the substrate binding b′ domain. J. Mol. Biol. 383, 1144–1155. 12. Denisov, A. Y., Määttänen, P., Dabrowski, C., Kozlov, G., Thomas, D. Y & Gehring, K. (2009). Solution structure of the bb′ domains of human protein disulfide isomerase. FEBS J. 276, 1440–1449. 13. Klappa, P., Ruddock, L. W., Darby, N. J. & Freedman, R. B. (1998). The b′ domain provides the principal peptide-binding site of protein disulfide isomerase but all domains contribute to binding of misfolded proteins. EMBO J. 17, 927–935. 14. Tsai, B., Rodighiero, C., Lencer, W. I. & Rapoport, T. A. (2001). Protein disulfide isomerase acts as a redox-dependent chaperone to unfold cholera toxin. Cell, 104, 937–948. 15. Lumb, R. A. & Bulleid, N. J. (2002). Is protein disulfide isomerase a redox-dependent molecular chaperone? EMBO J. 21, 6763–6770. 16. Kajino, T., Miyazaki, C., Asami, O., Hirai, M., Yamada, Y. & Udaka, S. (1998). Thermophilic fungal protein disulfide isomerase. Methods Enzymol. 290, 50–59. 17. Harada, T., Kurimoto, E., Tokuhiro, K., Asami, O., Sakai, T., Nohara, D. & Kato, K. (2001). Disulfide bond formation in refolding of thermophilic fungal protein disulfide isomerase. J. Biosci. Bioeng. 91, 596–598. 18. Lyles, M. M. & Gilbert, H. F. (1994). Mutations in the thioredoxin sites of protein disulfide isomerase reveal functional nonequivalence of the N- and C-terminal domains. J. Biol. Chem. 269, 30946–30952. 19. Westphal, V., Darby, N. J. & Winther, J. R. (1999). Functional properties of the two redox-active sites in yeast protein disulphide isomerase in vitro and in vivo. J. Mol. Biol. 286, 1229–1239. 20. Weber, G. & Laurence, D. J. (1954). Fluorescent indicators of adsorption in aqueous solution and on the solid phase. Biochem. J. 56 [(325th Meeting) xxxi (abstr.)]. 21. Darby, N. J. & Creighton, T. E. (1995). Functional properties of the individual thioredoxin-like domains of protein disulfide isomerase. Biochemistry, 34, 11725–11735. 22. Koradi, R., Billeter, M. & Wüthrich, K. (1996). MOLMOL: a program for display and analysis of macromolecular structures. J. Mol. Graphics, 14, 51–55, 29–32. 23. Nakano, M., Murakami, C., Yamaguchi, Y., Sasakawa, H., Harada, T., Kurimoto, E. et al. (2006). NMR assignments of the b′ and a′ domains of thermophilic fungal protein disulfide isomerase. J. Biomol. NMR, 36 (Suppl. 1), 44. 24. Morjana, N. A. & Gilbert, H. F. (1991). Effect of protein and peptide inhibitors on the activity of protein disulfide isomerase. Biochemistry, 30, 4985–4990. 25. Klappa, P., Hawkins, H. C. & Freedman, R. B. (1997). Interactions between protein disulphide isomerase and peptides. Eur. J. Biochem. 248, 37–42. 26. Svergun, D. I. (1999). Restoring low resolution

374

27.

28. 29.

30.

31.

32.

33.

34.

35.

36.

structure of biological macromolecules from solution scattering using simulated annealing. Biophys. J. 76, 2879–2886. Fujisawa, T., Ueki, T. & Inoko, Y. (1987). X-ray scattering from a troponin C solution and its interpretation with a dumbbell-shaped-molecule model. J. Appl. Crystallogr. 20, 349–355. Kaya, H. (2004). Scattering from cylinders with globular end-caps. J. Appl. Crystallogr. 37, 223–230. Pirneskoski, A., Klappa, P., Lobell, M., Williamson, R. A., Byrne, L., Alanen, H. I. et al. (2004). Molecular characterization of the principal substrate binding site of the ubiquitous folding catalyst protein disulfide isomerase. J. Biol. Chem. 279, 10374–10381. Darby, N. J., Penka, E. & Vincentelli, R. (1998). The multi-domain structure of protein disulfide isomerase is essential for high catalytic efficiency. J. Mol. Biol. 276, 239–247. Tian, G., Kober, F. X., Lewandrowski, U., Sickmann, A., Lennarz, W. J. & Schindelin, H. (2008). The catalytic activity of protein-disulfide isomerase requires a conformationally flexible molecule. J. Biol. Chem. 283, 33630–33640. Wilkinson, B., Xiao, R. & Gilbert, H. F. (2005). A structural disulfide of yeast protein-disulfide isomerase destabilizes the active site disulfide of the Nterminal thioredoxin domain. J. Biol. Chem. 280, 11483–11487. Darby, N. J. & Creighton, T. E. (1995). Characterization of the active site cysteine residues of the thioredoxin-like domains of protein disulfide isomerase. Biochemistry, 34, 16770–16780. Kulp, M. S., Frickel, E. M., Ellgaard, L. & Weissman, J. S. (2006). Domain architecture of protein-disulfide isomerase facilitates its dual role as an oxidase and an isomerase in Ero1p-mediated disulfide formation. J. Biol. Chem. 281, 876–884. Kortemme, T., Darby, N. J. & Creighton, T. E. (1996). Electrostatic interactions in the active site of the Nterminal thioredoxin-like domain of protein disulfide isomerase. Biochemistry, 35, 14503–14511. Hillson, D. A., Lambert, N. & Freedman, R. B. (1984). Formation and isomerization of disulfide bonds in proteins: protein disulfide-isomerase. Methods Enzymol. 107, 281–294.

Redox-Dependent Conformational Rearrangement of PDI

37. Güntert, P., Mumenthaler, C. & Wüthrich, K. (1997). Torsion angle dynamics for NMR structure calculation with the new program DYANA. J. Mol. Biol. 273, 283–298. 38. Brunger, A. T. (2007). Version 1.2 of the Crystallography and NMR system. Nat. Protoc. 2, 2728–2733. 39. Clore, G. M. & Gronenborn, A. M. (1994). Multidimensional heteronuclear nuclear magnetic resonance of proteins. Methods Enzymol. 239, 349–363. 40. Goddard, T. D. & Kneller, D. G. SPARKY 3, University of California, San Francisco, CA. 41. d'Auvergne, E. J. & Gooley, P. R. (2008). Optimisation of NMR dynamic models I. Minimisation algorithms and their performance within the model-free and Brownian rotational diffusion spaces. J. Biomol. NMR, 40, 107–119. 42. Kato, K., Matsunaga, C., Igarashi, T., Kim, H., Odaka, A., Shimada, I. & Arata, Y. (1991). Complete assignment of the methionyl carbonyl carbon resonances in switch variant anti-dansyl antibodies labeled with [1-13C]methionine. Biochemistry, 30, 270–278. 43. Matsunaga, C., Kato, K. & Arata, Y. (1991). A 13C NMR study of the hinge region of a mouse monoclonal antibody. J. Biomol. NMR, 1, 379–390. 44. Kainosho, M. & Tsuji, T. (1982). Assignment of the three methionyl carbonyl carbon resonances in Streptomyces subtilisin inhibitor by a carbon-13 and nitrogen-15 double-labeling technique. A new strategy for structural studies of proteins in solution. Biochemistry, 21, 6273–6279. 45. Shindyalov, I. N. & Bourne, P. E. (1998). Protein structure alignment by incremental combinatorial extension (CE) of the optimal path. Protein Eng. 11, 739–747. 46. Lundstrom, J. & Holmgren, A. (1990). Protein disulfide-isomerase is a substrate for thioredoxin reductase and has thioredoxin-like activity. J. Biol. Chem. 265, 9114–9120. 47. Nakasako, M., Iwata, T., Matsuoka, D. & Tokutomi, S. (2004). Light-induced structural changes of LOV domain-containing polypeptides from Arabidopsis phototropin 1 and 2 studied by small-angle X-ray scattering. Biochemistry, 43, 14881–14890. 48. Guinier, A. & Fournet, G. (1955). Small-Angle Scattering of X-rays. John Wiley, New York.