Solution structure of the bb¢ domains of human protein disulfide isomerase Alexey Y. Denisov*, Pekka Ma ¨ a ¨ tta ¨ nen*, Christian Dabrowski, Guennadi Kozlov, David Y. Thomas and Kalle Gehring Department of Biochemistry, McGill University, and Groupe de Recherche Axe ´ sur la Structure des Prote ´ ines (GRASP), Montre ´ al, Canada The endoplasmic reticulum (ER) is the cell compart- ment where membrane and secretory proteins fold. The rate-limiting step for the folding of many proteins is the formation of disulfide bonds. As polypeptides are synthesized, their cysteine thiols enter the oxidizing environment of the ER and form covalent intramolec- ular and intermolecular disulfide links. Although this oxidative folding process occurs spontaneously [1], non-native disulfide-bonded intermediates often occur, acting as kinetic traps along the folding pathway [2,3]. To avoid these, the ER contains a large family of enzymes called protein disulfide isomerases (PDIs) that catalyze both disulfide bond formation and the rear- rangement of incorrect disulfide bonds [4–7]. PDI family members are loosely defined by homology to thioredoxin and ER localization. There are at least 17 PDI family proteins in humans, 13 of which contain CXXC active-site motifs, and 9 have been shown to catalyze disulfide-exchange reactions [4,5]. The best studied and most abundant member of the family is PDI, a ubiquitous enzyme found at very high concentrations in the ER. Its concentra- tion has been estimated to be 10 lm in dog pancre- atic microsomes [8], the highest of all ER resident proteins. PDI has four thioredoxin-like domains, a-b-b¢-a¢, where the two a domains contain catalytic CGHC motifs, and the two b domains lack the conserved cysteine residues and are noncatalytic. The linkers between the domains are generally short. The longest is a stretch of 19 amino acids between the b¢ and a¢ domains, referred to as the x-linker [9]. Keywords chaperone; endoplasmic reticulum; NMR solution structure; protein disulfide isomerase family Correspondence K. Gehring, Department of Biochemistry, McGill University, 3655 Promenade Sir William Osler, Montreal, QC H3G 1Y6, Canada Fax: +1 (514) 398 7384 Tel: +1 (514) 398 7287 E-mail: kalle.gehring@mcgill.ca *These authors contributed equally to this work (Received 16 November 2008, revised 30 December 2008, accepted 30 December 2008) doi:10.1111/j.1742-4658.2009.06884.x Protein disulfide isomerase is the most abundant and best studied of the disulfide isomerases that catalyze disulfide bond formation in the endoplas- mic reticulum, yet the specifics of how it binds substrate have been elusive. Protein disulfide isomerase is composed of four thioredoxin-like domains (abb¢a¢). Cross-linking studies with radiolabeled peptides and unfolded pro- teins have shown that it binds incompletely folded proteins primarily via its third domain, b¢. Here, we determined the solution structure of the sec- ond and third domains of human protein disulfide isomerase (b and b¢, respectively) by triple-resonance NMR spectroscopy and molecular model- ing. NMR titrations identified a large hydrophobic surface within the b¢ domain that binds unfolded ribonuclease A and the peptides mastoparan and somatostatin. Protein disulfide isomerase-catalyzed refolding of reduced ribonuclease A in vitro was inhibited by these peptides at concen- trations equal to their affinity to the bb¢ fragment. Our findings provide a structural basis for previous kinetic and cross-linking studies which have shown that protein disulfide isomerase exhibits a saturable, substrate- binding site. Abbreviations ER, endoplasmic reticulum; GST, glutathione S-transferase; HSQC, heteronuclear single-quantum correlation; PDI, protein disulfide isomerase; RDC, residual dipolar coupling; RNase A, ribonuclease A. 1440 FEBS Journal 276 (2009) 1440–1449 ª 2009 The Authors Journal compilation ª 2009 FEBS The structure of yeast PDI has been determined in two crystal forms [10,11]. In both structures, the protein adopts a U shape with the catalytic a and a¢ domains on the same side of the protein. Compari- son of the two structures shows that considerable flexibility exists in the interdomain linkers. The larg- est difference is a twist of over 120° in the relative orientations of the a and b domains. In one struc- ture, the catalytic cysteines face each other; in the other, the catalytic residues of the a domain face away from the a¢ domain. The crystal structures also revealed the presence of a hydrophobic pocket (which faces inwards at the base of the U) in the b¢ domain. This b¢-domain pocket was postulated to be the site for binding of incompletely folded proteins, along with adjoining contiguous portions of the a, b and a¢ domains [12]. Cross-linking studies with radio- labelled model peptides identified a homologous, hydrophobic binding site on the b¢ domain of human PDI [13]. Sequence identity between human and yeast PDI for the b and b¢ domains is < 10% (Fig. 1), making it dif- ficult to compare the two proteins accurately. The structures of individual a [14], b [15], a¢ (Protein Data Bank entry code 1X5C) and b¢ [16] domains of human PDI have been solved by NMR or X-ray crystallogra- phy. The overall shape of full-length human PDI has been investigated by small-angle X-ray scattering and shown, at low resolution, to adopt a flat annular arrangement [17]. Here, we reported the solution structure of the bb¢ fragment of human PDI and addressed the question of how PDI recognizes unfolded proteins. Using NMR titrations, we mapped the region of PDI that binds unfolded proteins and we showed that peptides which bind to this region inhibit the PDI-catalyzed refolding of ribonuclease A (RNase A). Results Spectra of the human PDI-bb¢ domains Dimerization of PDI fragments containing the hydro- phobic b¢ domain has complicated structural studies for more than a decade [11]. NMR spectra of the isolated b¢ domain of human PDI showed broad lines and mul- tiple peaks as a result of the presence of a mixture of monomeric and dimeric forms (Fig. S1). These forms can be separated by gel filtration but rapidly exchange at 20–30 °C( 20% of dimers in 2 h and more than 50% of dimers in 12 h, starting with pure 0.5 mm b¢ monomer at 30 °C). The 1 H- 15 N heteronuclear single- quantum correlation (HSQC) spectrum of PDI-b¢ was significantly better upon the addition of hydrophobic compounds, such as peptide ligands or detergents (e.g. 0.3% Triton X-100), that dissociate the dimer (Fig. S1). Alternatively, the b domain moderates the tendency of the b¢ domain to dimerize and significantly slows inter- conversion. The monomeric form of the PDI-bb¢ frag- ment converts into < 10% of dimers in 12 h and < 25% of dimers in 3 days, starting with 0.5 mm of monomers at 30 °C. Most of the 1 H- 15 N HSQC signals of the dimeric form of PDI-bb¢ coincide with the signals of the monomeric form or are weak as a result of the high molecular weight of the dimer. The monomeric form gives good-quality spectra required for structural studies. The 1 H- 15 N spectrum of the 25 kDa PDI-bb¢ fragment shows signal dispersion typical for a well- folded protein and allows determination of the back- bone and side-chain NMR signal assignments [18]. Protein structure and comparison The solution structure of the human PDI-bb¢ fragment was calculated based on  2200 NMR-derived con- Fig. 1. Structure-based sequence alignment of human and yeast PDI-bb¢ showing the positions of the a-helices and the b-strands. Color shading represents the size of the amide chemical shift changes in human PDI-bb¢ upon the binding of unfolded RNase A (red, Dd > 0.10; yellow, 0.10 > Dd > 0.05 p.p.m.). Residues are numbered from the initiator methionine in the signal sequence. A. Y. Denisov et al. Structure of the bb¢ domains from human PDI FEBS Journal 276 (2009) 1440–1449 ª 2009 The Authors Journal compilation ª 2009 FEBS 1441 straints, including 1 H- 15 N residual dipolar couplings (RDCs) (Fig. S2). The set of best structures is pre- sented in Fig. 2A and the structural statistics are shown in Table 1. The mean rmsd obtained from the average structure was 0.7 A ˚ for backbone atoms. The greatest uncertainties were in modeling helix a3 in the b domain and the loop between b4¢ and b5¢ in the b¢ domain. A ribbon representation of the PDI-bb¢ structure is pre- sented in Fig. 2B. The structures of both b and b¢ domains corresponded to a babababba thioredoxin- like fold, where the central five-stranded b-sheet is sur- rounded by a-helices on both sides. Heteronuclear 15 N{ 1 H} NOEs were in the range of 0.6–0.9 (Fig. S2), indicating the absence of a flexible interdomain linker. Contacts between the b and b¢ domains could be observed as long-range NOEs between protons H a (His231) ⁄ H a (Gly251), H e1 (His231) ⁄ H N (Gly251), Me(Val155) ⁄ H N (Leu234) and H e1 (Phe209)⁄ H N (Leu236). Analysis of RDCs for the two domains yielded the same degree of alignment and rhombicity, which fur- ther confirms the rigid structure of the bb¢ domain frag- ment (Fig. S2). It is interesting to note that the protein surface and electrostatic potential is quite different for the b and b¢ domains (Fig. 2C). The pairwise C a -atomic coordinate rmsd between human PDI-bb¢ and the crystal structure of yeast PDI [10] was 3.5 A ˚ for 198 structurally equivalent amino acids (DALI Z-factor = 14.6). The principal differ- ence between the protein fold of human and yeast PDI bb¢ domains was an extra helix, a3, in the b domain of the human protein and an extra a-helix in the b¢ domain from yeast (Figs 1 and 3B). In this sense, the fold of human PDI-bb¢ is more similar to the fold of human ERp57-bb¢. ERp57 is a disulfide isomerase that has the same domain architecture as PDI but shares very low sequence identity with PDI and is glycopro- tein specific via interaction with calnexin or calreticu- lin. The rmsd between human PDI-bb¢ and the crystal structure of human ERp57-bb¢ [19] is 4.5 A ˚ for 209 amino acids (Z = 14.9). A comparison of the b domain in our PDI-bb¢ structure with the reported solution structure of the isolated b domain [15] gave an rmsd of 1.6 A ˚ for 101 amino acids (Z = 16.1), showing that the structure is not changed significantly by interaction with the b¢ domain. The pairwise C a -atomic coordinate comparison of our solution structure of human PDI-bb¢ and the crys- tal structure of the I289A mutant of the human PDI- b¢x fragment [16] showed differences with an rmsd of 2.6 A ˚ for 116 amino acids (DALI Z-factor = 12.1). Superimposition of these structures is shown in Fig. 3C. The differences could result from (a) an effect of the second b domain in our PDI-bb¢ structure, (b) the I289A mutation, or (c) the presence of the x-linker in the crystal structure. In the b¢x structure, the hydro- phobic x-linker folds back and binds to the b¢ domain in the region that we identified here as the hydropho- bic peptide-binding pocket (vide infra). Binding site for unfolded ligands Analysis of the PDI-bb¢ electrostatic surface revealed a highly hydrophobic region within the b¢ domain (Fig. 2C). Previous work has demonstrated that the amphipathic peptides mastoparan and D-somatostatin can bind directly to PDI, and that this interaction is Tri- A B C Fig. 2. The human PDI-bb¢ fold. Stereoview of the backbone superposition for 10 low- energy structures (A); ribbon representation of the solution structure of PDI-bb¢ (B); and color-coded surface of PDI-bb¢, with red indicating negative electrostatic potential and blue indicating positive potential (C). Structure of the bb¢ domains from human PDI A. Y. Denisov et al. 1442 FEBS Journal 276 (2009) 1440–1449 ª 2009 The Authors Journal compilation ª 2009 FEBS ton X-100 sensitive [13]. This finding was confirmed by NMR titrations of PDI-bb¢ by mastoparan and somato- statin peptides and unfolded RNase A protein. Compar- ison of 1 H- 15 N HSQC spectra of PDI-bb¢ in the absence or presence of unfolded ligands (Fig. S1) was indicative of strong shifts of the NMR signals for residues in heli- ces a1¢, a3¢ and all five b-strands of the b¢ domain. The most strongly shifted HSQC signals were Thr241, Ala245, Phe249, Gly250, His256, Asp297, Glu322, Met324 when titrated with mastoparan (Fig. 4A), Thr241, Gln243, Ile248, Gly250, Asp297, Arg300, Ile318, Thr325 when titrated with somatostatin and Thr241, Gly251, His256, Ile318, Thr319 and Glu321 when titrated with unfolded RNase A (Figs 1 and 4C). The chemical shift changes were plotted throughout the PDI-bb¢ sequence, and affected residues were mapped onto the protein backbone trace (Fig. 4B,D). A close-up view of the hydrophobic pocket in the b¢ domain is shown in Fig. 3A. The binding pocket is large and could accommodate multiple hydrophobic residues. The sig- nals identified by NMR belong to hydrophobic residues of the binding pocket or neighboring residues, which could be influenced by steric contacts with the side chains of the hydrophobic residues and small changes in the conformation of the b¢ domain. From the NMR titrations, the dissociation constant (K d ) was 130 ± 30 lm for mastoparan and 35 ± 15 lm for both somatostatin and unfolded RNase A (Fig. S2). The higher affinity of somatostatin and unfolded RNase A compared with mastoparan is probably a result of the larger number of hydrophobic residues with aromatic side chains. In control NMR titrations, folded RNase A showed essentially no binding to PDI-bb¢ ( K d >2mm), as hydrophobic patches of RNase A are not exposed to solvent in the folded state. The structure of the bb¢ domains of the glycopro- tein-specific PDI homolog, ERp57, did not reveal a similar hydrophobic-binding pocket [19]. ERp57 instead relies on substrate recruitment by the lectin-like chaperones calnexin and calreticulin, which bind the ERp57 b¢ domain on the surface opposite to the corre- sponding hydrophobic surface in PDI [5]. Nonetheless, many of the hydrophobic residues in the PDI-b¢ pocket (shown in Fig. 3A) are similar in other PDI family Table 1. Structural statistics for PDI-bb¢. Restraints for structure calculations Total restraints used 2191 Intraresidue NOEs 682 Sequential NOEs 531 Medium and long-range NOEs 294 Hydrogen bonds 92 / and wbackbone angles 386 NH RDCs 206 Final energies (kcalÆmol )1 ) Etotal 560 ± 35 Ebond 26 ± 7 Eangle 128 ± 26 Eimpr 20.8 ± 5.0 Erepel 217 ± 39 Enoe 12.6 ± 2.6 Ecdih 13.7 ± 3.3 Esani 36 ± 8 rmsd from idealized geometry Bond (A ˚ ) 0.0026 ± 0.0004 Bond angles (°) 0.35 ± 0.04 Improper torsions (°) 0.26 ± 0.04 rmsd for experimental restraints Distances (A ˚ ) 0.010 ± 0.001 Dihedral angles (°) 0.54 ± 0.07 RDCs rmsd (Hz) 1.35 ± 0.05 Q-value 0.089 ± 0.004 Coordinate rmsd from the average structure (A ˚ ) a Backbone atoms (N,C a ,C¢) 0.72 ± 0.06 All heavy atoms 1.23 ± 0.05 Ramachandran analysis (%) Residues in most favored regions 84.0 ± 2.2 Residues in additional allowed regions 12.8 ± 3.0 Residues in generously allowed regions 3.2 ± 1.2 a For residues 137–350. AB C Fig. 3. (A) View of the peptide-binding hydrophobic pocket in the human PDI-b¢ domain with the residues displayed in stick representation. (B) Superimposition of the solution structure of the human PDI-b¢ (blue) with the crystal structure of yeast PDI-b¢ (red, Protein Data Bank entry code 2B5E). (C) Superimposition of the solution structure of human PDI-b¢ (blue) with the crystal structure of the human PDI-b¢x I289A mutant (green, Protein Data Bank entry code 3BJ5). The x-linker tail of PDI-b¢x is shown in red. A. Y. Denisov et al. Structure of the bb¢ domains from human PDI FEBS Journal 276 (2009) 1440–1449 ª 2009 The Authors Journal compilation ª 2009 FEBS 1443 members (Fig. S3). It is likely that PDIp, PDILT, ERp27 and ERp44 share features with PDI concerning how they bind unfolded proteins [20–22]. Residue-specific interactions The importance of individual amino acids in human PDI for the binding of D-somatostatin was previously investigated [23], but many of the reported mutations were not in the region of the PDI-b¢ binding pocket. In that study it was reported that the mutation of resi- due Ile289 (numbered as Ile272 without the PDI signal sequence), which is located at the bottom of the b¢ hydrophobic pocket (Fig. 3A), significantly reduced cross-linking with D-somatostatin. To explore the role of individual amino acids in the b¢ domain, we prepared two mutants (I289A and F240E) in both the bb¢ fragment and the full-length protein. Surprisingly, NMR titration experiments of the binding of somato- statin and mastoparan to the PDI-bb¢ I289A mutant did not show a significant effect in comparison with wild-type bb¢ domains (data not shown). We also investigated the effect of the I289A mutant on the PDI-catalyzed refolding of RNase A using a continu- ous spectroscopic assay of 2¢3¢ cCMP hydrolysis (Fig 5A). In agreement with the NMR titration, the I289A mutant did not diminish the foldase activity of the PDI. By contrast, the mutation of F240E strongly decreased PDI-catalyzed refolding. This mutation destablized the b¢ domain (most 1 H- 15 N HSQC signals of the b¢-domain of the PDI-bb¢ F240E mutant were shifted in comparison with wild-type protein and strongly broadened) and prevented peptide binding in the context of the bb¢ fragment. To identify peptide residues involved in binding to PDI, we carried out reciprocal NMR titrations by observing changes in the signals for mastoparan following the addition of PDI-bb¢ protein (Fig. 6). At the lowest concentration of PDI-bb¢, at least half of the 14 mastoparan signals were significantly shifted by binding to the b¢ domain of PDI. No strong selectivity in residue binding was found. At a protein ⁄ mastoparan ratio of 1 : 15, practically all of the mas- toparan signals (except for those of the terminal amino acids) were strongly broadened as a result of binding to the PDI-bb¢ protein. Further work is necessary to A B D C Fig. 4. Mapping residues involved in ligand binding. Magnitude of amide chemical shift changes in the primary sequence of PDI-bb¢ and backbone trace of PDI-bb¢ colored according to the magnitude of the chemical shift changes upon binding mastoparan (A, B) and unfolded RNase A (C, D). Structure of the bb¢ domains from human PDI A. Y. Denisov et al. 1444 FEBS Journal 276 (2009) 1440–1449 ª 2009 The Authors Journal compilation ª 2009 FEBS determine the precise roles of substrate residues and residues in the b¢ domain, but our preliminary results indicate that the binding reaction involves multiple redundant interactions. Inhibition of PDI in RNase A refolding In order to understand better the contribution of the b¢ binding site to PDI activity, the inhibitory influence on the refolding of RNase A caused by the peptides binding b¢ was examined. In the assay, incubation of unfolded RNase with PDI led to RNase activity, which was measured by the hydrolysis of tRNA. In the absence of an inhibitor or in the presence of a highly charged peptide, RNase A was rapidly refolded in 10 min, leading to the disappearance of tRNA (Fig. 5B). Addition of the hydrophobic peptides, mas- toparan or D-somatostatin, inhibited RNase A refold- ing at concentrations similar to their affinity to the bb¢ fragment. Mastoparan completely inhibited RNase A A B Fig. 5. PDI-catalyzed RNase-refolding assays. (A) Mutagenesis of the b¢ domain reduces the efficiency of PDI-catalyzed refolding of RNase A in a simultaneous refolding and cCMP hydrolysis assay. The refolding rate of the F240E mutant was 50% lower than that of the wild-type PDI or the I289A mutant relative to spontaneous refolding in the absence of PDI. A small increase in absorbance was observed in the absence of RNase A. (B) Peptides that bind to the b¢ domain inhibit PDI refolding of RNase A in a dose-dependent manner. Folding reactions were carried out with the indicated con- centrations of a control peptide (KEKEKVKQIPKAPK), mastoparan, or D-somatostatin, and the activity of RNase A was measured in a gel assay of tRNA hydrolysis. In the presence of the control pep- tide, PDI rapidly refolded RNase A, leading to the complete degra- dation of the substrate tRNA. Both mastoparan and D-somatostatin blocked refolding. Fig. 6. NMR titrations of 2 mM mastoparan by human PDI-bb¢ protein and a plot of the magnitude of changes in mastoparan proton chemical shifts at a protein ⁄ mastoparan ratio of 1 : 15. A. Y. Denisov et al. Structure of the bb¢ domains from human PDI FEBS Journal 276 (2009) 1440–1449 ª 2009 The Authors Journal compilation ª 2009 FEBS 1445 refolding at 120 lm, whereas D-somatostatin blocked refolding at concentrations between 30 and 60 lm. These are similar to the K I of 80 lm reported for the inhibition of PDI glutathione-insulin transhydrogenase activity by the peptide somatostatin [24]. Control experiments with di(o-aminobenzyl)-labeled oxidized glutathione showed no inhibition of PDI oxido-reduc- tase activity by D-somatostatin (data not shown). A systematic study of the kinetics of PDI-mediated RNase A refolding showed that the refolding rate of RNase A is saturable with increasing concentrations of unfolded RNase A [25]. The K m measured, 7 lm,is close to the affinity measured by NMR for unfolded RNase A binding to the isolated bb¢ domains. The sec- ondary importance of other domains for the binding of large protein substrates has been previously demon- strated [13,26]. Mutational analysis of PDI revealed that loss of the two cysteines in the C-terminal a¢ domain increased the K m to 30 lm, and loss of an additional cys- teine in the a domain resulted in an increase of the K m to 50 lm [25,27]. On the other hand, the role of the b domain seems to be to act simply as a spacer to allow room for the a and a¢ domains to interact with substrate thiols. By NMR, we detected no interactions between unfolded RNase A and the b domain. Discussion There are many examples of chaperone proteins that bind unfolded protein segments via hydrophobic patches. The best known is cytosolic Hsp70, which binds and releases, through cycles of ATP binding and hydrolysis, short stretches of hydrophobic polypeptides that are in an extended conformation [28]. In Escheri- chia coli, the ClpA–ClpP chaperones disaggregate and unfold proteins in order to degrade them. ClpA binds to substrates with low affinity, but broad specificity, via a hydrophobic surface formed by two helices in its N-terminal domain [29]. Multisubunit GroEL binds in vivo to more than 10% of newly synthesized poly- peptides [30] via a groove between two alpha helices that is lined with hydrophobic residues [31]. Neverthe- less, hydrophobic binding is not a universal mechanism of chaperone function, and other chaperones use charged and polar residues for interactions between the chaperone and the substrate [31,32]. The relatively weak binding of PDI-bb¢ to peptides and unfolded RNase A, and the large size of the binding pocket, is consistent with a low degree of specificity for hydrophobic ligands. High specificity is not expected because PDI acts on many substrates with different primary sequences. It is also important that substrate proteins are released from PDI after disulfide bond formation and protein folding. A large, multivalent hydrophobic binding site is an effective way to bind a variety of substrates when unfolded and to release them once they acquire their native conformation with fewer hydrophobic residues exposed. To conclude, structural analysis of the bb¢ fragment of PDI has revealed a large hydrophobic surface that inter- acts with peptides and unfolded RNase A. This site appears to be responsible for the saturable kinetics observed for RNase A folding by PDI, and blocking the site strongly inhibits the activity of PDI. Structural anal- ysis of the substrate-binding sites of other disulfide isomerases should shed more light on their substrate specificities and help to explain why such a large variety of disulfide isomerases is found in the mammalian ER [5]. Experimental procedures Sample preparation PDI was cloned from cDNA derived from human bronchial epithelial cells. The bb¢ (residues P135–S357) and b¢ (residues L236–S357) fragments were subcloned into pGEX-6P-1 (Amersham Pharmacia Biotech, Piscataway, NJ, USA) and expressed in E. coli BL21 (DE3) as glutathi- one S-transferase (GST) fusion proteins. To provide iso- tope-labeled samples for NMR, cultures were grown at 37 °C on minimal M9 medium supplemented with 15 N ammonium chloride and [ 13 C]-glucose (Cambridge Isotopes Laboratory, Andover, MA, USA) to produce uniformly 15 N- or 15 N, 13 C-labeled proteins. The protein was purified by GST-affinity chromatography on a Glutathione Sepha- rose 4B column (Amersham). PreScission protease (Amer- sham) was used to cleave the fusion protein from GST. The resulting proteins contained five extraneous N-terminal resi- dues (GPLGS). Further purification was carried out using gel-filtration chromatography on a Superdex-75 column. Mass spectral analysis confirmed the sequence composition of human PDI- bb ¢. The NMR samples contained 0.1-1 mm protein in 90% H 2 O ⁄ 10% D 2 O, 25 mm sodium phosphate buffer (pH 7.0), 70 mm NaCl, 0.5 mm EDTA and 5 mm dithiothreitol. Unlabeled 14 amino acid mastoparan INLKALAALAK KIL, D-somatostatin AGSKNFFWKTFTSS and charged KEKEKVKQIPKAPK peptides were chemically synthe- sized at EZBiolab (Westfield, IN, USA) and additionally purified by reverse-phase HPLC. Somatostatin AGCKN FFWKTFTSC (‡ 97% pure by HPLC) was purchased from Sigma (St Louis, MO, USA). Bovine pancreatic RNase A from Sigma was unfolded and reduced for 20 min at room temperature in 0.1 m Tris ⁄ HCl (pH 8.0) containing 6 m guanidine ⁄ HCl and 20 mm dithiothreitol [33]. Unfolded Structure of the bb¢ domains from human PDI A. Y. Denisov et al. 1446 FEBS Journal 276 (2009) 1440–1449 ª 2009 The Authors Journal compilation ª 2009 FEBS RNase A was desalted in 0.1% formic acid on a NAP-5 column (Amersham) and lyophilized. The maximum solubility of the unfolded RNase A in the NMR phosphate buffer was  0.2 mm. Mutagenesis of PDI Point mutants of full-length PDI and the bb¢ domains were prepared in the vectors used for the expression of the wild- type proteins using QuickchangeÔ site-directed mutagenesis (Stratagene, La Jolla, CA, USA) with mismatched primers and were verified by DNA sequencing. NMR spectroscopy NMR spectra were recorded at 30 °C on Bruker DRX 600 MHz and Varian Unity Inova 800 MHz spectrometers equipped with triple-resonance cryoprobes and pulsed-field gradients. Proton homonuclear NOEs were obtained from 15 N-edited and 13 C-edited NOESY spectra recorded at 800 MHz with a mixing time of 80 ms. Amide heteronuclear 15 N{ 1 H} NOEs were measured to determine high-mobility regions of protein [34]. 1 H- 15 N RDCs with precision ± 1 Hz were extracted from in-phase/anti-phase-HSQC experiments [35] on an isotropic sample and on a sample containing 12 mgÆmL )1 of Pf1 phage. NMR spectra were processed using nmrpipe [36] and xwinnmr (Bruker Biospin, Milton, Canada) software, and then analyzed using xeasy [37] and nmrview [38]. Detailed analysis of ligand binding to PDI-bb¢ was carried out by comparison of chemical shifts for back- bone amide signals in 1 H- 15 N HSQC spectra. HSQC spectra were recorded at 1 : 2, 1 : 1, 2 : 1, 4 : 1 and 8 : 1 peptide to protein ratios. The magnitude of amide chemical shift changes was calculated as [(D 1 H shift) 2 +(D 15 N shift · 0.2) 2 ] 1 ⁄ 2 , in p.p.m. Values of dissociation constants were obtained by monitoring the chemical shift changes as a function on ligand concentration using a simple binding model. A least-squares search was performed by varying the values of K d and the chemical shift of fully saturated protein. Standard deviations were derived for each K d value by com- paring different cross-peaks in the HSQC spectra. Assignments of the amide proton signals of mastoparan were determined using 2D NOESY with a mixing time of 200 ms and TOCSY experiments on a 2 mm sample at 10 °C. Structure calculations Regions of a-helical or b-strand secondary structure were determined based on C a -chemical shifts [39] and the NOE patterns [40]. ARIA-assigned [41] and manually verified NOEs were collected from 15 N- and 13 C-edited NOESY spectra. Backbone angles were estimated from the chemical shifts using the TALOS database [42]. The starting struc- ture was generated with modeller [43] using the yeast PDI crystal structure (Protein Data Bank entry code 2B5E) and was in agreement with manually assigned NOEs. The pro- tein structure was refined using the standard protocol in CNS version 1.1 [44], and the structural statistics for the 10 best structures is shown in Table 1. The atomic coordinates have been deposited as the Protein Data Bank entry 2K18. The pairwise coordinate rmsd comparisons between differ- ent proteins were obtained using dali [45]. module software [46] was used for comparison of the RDCs with their back- calculated values. Structural figures were generated using py- mol [47] and molmol [48]. protskin (C. Deprez and K. Gehring; http://www.mcgnmr.mcgill.ca/ProtSkin) soft- ware was used for mapping chemical shift changes onto pro- tein backbone traces. procheck-nmr software [49] was used to check the protein stereochemical geometry (Table 1). Refolding of bovine RNase A by PDI PDI-catalyzed refolding of RNase A was measured in two assays: a continuous spectroscopic assay of 2¢3¢ cCMP hydrolysis; and a gel-based assay of RNA degradation. PDI and unfolded RNase A were prepared as described above. The first assay monitored the absorbance change at 296 nm and was carried out as previously described [50] with the following modifications. Refolding was carried out in 25 mm Hepes, pH 8.0, containing 0.5 mm oxidized gluta- thione, 2 mm reduced glutathione, 0.75 mm CaCl 2 and 100 mm NaCl. The concentration of reduced RNase A in the refolding reaction was 4.2 lm, and the concentration of PDI was 0.6 lm. In the second assay, 0.18 lm RNase A was refolded with 0.3 lm PDI in 25 mm Hepes, pH 8.0, containing 0.5 mm oxidized glutathione, 2 mm reduced glu- tathione, 0.75 m m CaCl 2 and 100 mm NaCl. Samples were removed during folding and free thiols were blocked with an equal volume of 0.5 m iodoacetamide. The RNase A activity at each time-point was assayed by incubation with 10 lg of yeast tRNA (Sigma) for 15 min at 25 ° C followed by electrophoresis in a 1% agarose gel containing ethidium bromide for visualization. Gels were exposed to UV light and photographed using an Alpha Innotech Alpha Imager. Control experiments with di(o-aminobenzyl)-labeled oxi- dized glutathione (a gift of Bulent Mutus) were carried out as described previously [51]. Acknowledgements The authors are grateful to Lloyd Ruddock for sharing data and helpful discussions and to Tara Sprules for assistance in running experiments at the Quebec-East- ern Canada High Field NMR Facility. This work was funded by operating grants to D. T. and K. G. from the Canadian Institutes of Health Research (CIHR). P. M. was supported by a CIHR Canada Graduate Scholarships Doctoral Award. A. Y. Denisov et al. Structure of the bb¢ domains from human PDI FEBS Journal 276 (2009) 1440–1449 ª 2009 The Authors Journal compilation ª 2009 FEBS 1447 References 1 Anfinsen CB, Haber E, Sela M & White FH Jr (1961) The kinetics of formation of native ribonuclease during oxidation of the reduced polypeptide chain. Proc Natl Acad Sci USA 47, 1309–1314. 2 Creighton TE (1997) Protein folding coupled to disul- phide bond formation. Biol Chem 378, 731–744. 3 Weissman JS (1995) All roads lead to Rome. The multi- ple pathways of protein folding. Chem Biol 2, 255–260. 4 Ellgaard L & Ruddock LW (2005) The human protein disulphide isomerase family: substrate interactions and functional properties. EMBO Rep 6, 28–32. 5 Maattanen P, Kozlov G, Gehring K & Thomas DY (2006) ERp57 and PDI: multifunctional protein disul- fide isomerases with similar domain architectures but differing substrate-partner associations. Biochem Cell Biol 84, 881–889. 6 Hatahet F & Ruddock LW (2007) Substrate recognition by the protein disulfide isomerases. FEBS J 274, 5223– 5234. 7 Appenzeller-Herzog C & Ellgaard L (2007) The human PDI family: versatility packed into a single fold. Bio- chim Biophys Acta 1783, 535–548. 8 Bies C, Guth S, Janoschek K, Nastainczyk W, Volkmer J & Zimmermann R (1999) A Scj1p homolog and fold- ing catalysts present in dog pancreas microsomes. Biol Chem 380, 1175–1182. 9 Freedman RB, Gane PJ, Hawkins HC, Hlodan R, McLaughlin SH & Parry JW (1998) Experimental and theoretical analyses of the domain architecture of mam- malian protein disulphide-isomerase. Biol Chem 379, 321–328. 10 Tian G, Xiang S, Noiva R, Lennarz WJ & Schindelin H (2006) The crystal structure of yeast protein disulfide isomerase suggests cooperativity between its active sites. Cell 124, 61–73. 11 Tian G, Kober FX, Lewandrowski U, Sickmann A, Lennarz WJ & Schindelin H (2008) The catalytic activ- ity of protein disulfide isomerase requires a conforma- tionally flexible molecule. J Biol Chem 283, 33630– 33640. 12 Gruber CW, Cemazar M, Heras B, Martin JL & Craik DJ (2006) Protein disulfide isomerase: the structure of oxidative folding. Trends Biochem Sci 31, 455–464. 13 Klappa P, Ruddock LW, Darby NJ & Freedman RB (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 Kemmink J, Darby NJ, Dijkstra K, Nilges M & Creigh- ton TE (1996) Structure determination of the N-terminal thioredoxin-like domain of protein disulfide isomerase using multidimensional heteronuclear 13 C ⁄ 15 N NMR spectroscopy. Biochemistry 35, 7684–7691. 15 Kemmink J, Dijkstra K, Mariani M, Scheek RM, Penka E, Nilges M & Darby NJ (1999) The structure in solution of the b domain of protein disulfide isomerase. J Biomol NMR 13, 357–368. 16 Nguyen VD, Wallis K, Howard MJ, Haapalainen AM, Salo KEH, Saaranen MJ, Sidhu A, Wierenga RK, Freed- man RB, Ruddock LW et al. (2008) Alternative confor- mations of the x region of human protein disulfide- isomerase modulate exposure of the substrate-binding b¢ domain. J Mol Biol 383, 1144–1155. 17 Li SJ, Hong XG, Shi YY, Li H & Wang CC (2006) Annual arrangement and collaborative actions of four domains of protein-disulfide isomerase: a small angles X-ray scat- tering study in solution. J Biol Chem 281, 6581–6588. 18 Denisov AY, Maattanen P, Sprules T, Thomas DY & Gehring K (2007) 1 H, 13 C and 15 N resonance aassign- ment of the bb¢ domains of human protein disulfide isomerase. Biomol NMR Assign 1, 129–130. 19 Kozlov G, Maattanen P, Schrag JD, Pollock S, Cygler M, Nagar B, Thomas DY & Gehring K (2006) Crystal structure of the bb¢ domains of the protein disulfide isomerase ERp57. Structure 14, 1331–1339. 20 Anelli T, Alessio M, Bachi A, Bergamelli L, Bertoli G, Camerini S, Mezghrani A, Ruffato E, Simmen T & Sitia R (2003) Thiol-mediated protein retention in the endoplasmic reticulum: the role of ERp44. EMBO J 22, 5015–5022. 21 van Lith M, Hartigan N, Hatch J & Benham AM (2005) PDILT, a divergent testis-specific protein disul- fide isomerase with a non-classical SXXC motif that engages in disulfide-dependent interactions in the endo- plasmic reticulum. J Biol Chem 280 , 1376–1383. 22 Wang L, Wang L, Vavassori S, Li S, Ke H, Anelli T, Degano M, Ronzoni R, Sitia R, Sun F et al. (2008) Crystal structure of human ERp44 shows a dynamic functional modulation by its carboxy-terminal tail. EMBO Rep 9 , 642–647. 23 Pirneskoski A, Klappa P, Lobell M, Williamson RA, Byrne L, Alanen HI, Salo KEH, Kivirikko KI, Freed- man RB & Ruddock LW (2004) Molecular character- ization of the principal substrate binding site of the ubiquitous folding catalyst protein disulfide isomerase. J Biol Chem 2004, 10374–10381. 24 Morjana NA & Gilbert HF (1991) Effect of protein and peptide inhibitors on the activity of protein disulfide isomerase. Biochemistry 30, 4985–4990. 25 Lyles MM & Gilbert HF (1994) Mutations in the thior- edoxin sites of protein disulfide isomerase reveal func- tional nonequivalence of the N- and C-terminal domains. J Biol Chem 269, 30946–30952. 26 Darby NJ, Penka E & Vincentelli R (1998) The multi- domain structure of protein disulfide isomerase is essen- tial for high catalytic efficiency. J Mol Biol 276, 239–247. 27 Walker KW, Lyles MM & Gilbert HF (1996) Catalysis of oxidative protein folding by mutants of protein Structure of the bb¢ domains from human PDI A. Y. Denisov et al. 1448 FEBS Journal 276 (2009) 1440–1449 ª 2009 The Authors Journal compilation ª 2009 FEBS disulfide isomerase with a single active-site cysteine. Biochemistry 35, 1972–1980. 28 Young JC, Agashe VR, Siegers K & Hartl FU (2004) Pathways of chaperone-mediated protein folding in the cytosol. Nat Rev Mol Cell Biol 5, 781–791. 29 Xia D, Esser L, Singh SK, Guo F & Maurizi MR (2004) Crystallographic investigation of peptide binding sites in the N-domain of the ClpA chaperone. J Struct Biol 146, 166–179. 30 Houry WA, Frishman D, Eckerskorn C, Lottspeich F & Hartl FU (1999) Identification of in vivo substrates of the chaperonin GroEL. Nature 402, 147–154. 31 Gomez-Puertas P, Martin-Benito J, Carrascosa JL, Willison KR & Valpuesta JM (2004) The substrate recognition mechanisms in chaperonins. J Mol Recognit 17, 85–94. 32 Hubbard TJP & Sander C (1991) The role of heat- shock and chaperone proteins in protein folding: possi- ble molecular mechanisms. Protein Eng 4, 711–717. 33 Hillson DA, Lambert N & Freedman RB (1984) Forma- tion and isomerization of disulfide bonds in proteins: pro- tein disulfide-isomerase. Meth Enzymol 107, 281–294. 34 Peng JW & Wagner G (1994) Investigation of protein motions via relaxation measurements. Meth Enzymol 239, 563–596. 35 Ottiger M, Delaglio F & Bax A (1998) Measurement of J and dipolar couplings from simplified two-dimensional NMR spectra. J Magn Reson 131, 373–378. 36 Delaglio F, Grzesiek S, Vuister GW, Zhu G, Pfeifer J & Bax A (1995) NMRpipe: a multidimensional spectral processing system based on UNIX pipes. J Biomol NMR 6, 277–293. 37 Bartels C, Xia TH, Billeter M, Guntert P & Wu ¨ thrich K (1995) The program XEASY for computer-supported NMR spectral analysis of biological macromolecules. J Biomol NMR 6, 1–10. 38 Johnson BA & Blevins RA (1994) NMR View: a com- puter program for the visualization and analysis of NMR data. J Biomol NMR 4, 603–614. 39 Wishart DS & Sykes BD (1994) Chemical shifts as a tool for structure determination. Meth Enzymol 239, 363–392. 40 Wu ¨ thrich K (1986) NMR of Proteins and Nucleic Acids. John Wiley & Sons, New York. 41 Nilges M, Macias MJ, O’Donoghe SI & Oschkinat H (1997) Automated NOESY interpretation with ambiguous distance restraints: the refined NMR solution structure of the pleckstrin homology domain from beta-spectrin. J Mol Biol 269, 408–422. 42 Cornilescu G, Delaglio F & Bax A (1999) Protein backbone angle restraints from searching a database for chemical shift and sequence homology. J Biomol NMR 13, 289–302. 43 Sali A & Blundell TL (1993) Comparative protein modelling by satisfaction of spatial restraints. J Mol Biol 234, 779–815. 44 Bru ¨ nger AT, Adams PD, Clore GM, DeLano WL, Gros P, Grosse-Kuntsleve RW, Jiang JS, Kuszewski J, Nilges M, Pannu NS, et al. (1998) Crystallography and NMR system: a new software suite for macromolecular structure determination. Acta Crystallogr D (Biol Crystallogr) 54, 905–921. 45 Holm L & Sander C (1995) Dali: a network tool for protein structure comparison. Trends Biochem Sci 20, 478–480. 46 Dosset P, Hus JC, Marion D & Blackledge M (2001) A novel interactive tool for rigid-body modeling of multi-domain macromolecules using residual dipolar couplings. J Biomol NMR 20, 223–231. 47 DeLano WL & Hus JC (2002) The PyMOL Molecular Graphics System. DeLano Scientific, San Carlos, CA. 48 Koradi R, Billeter M & Wu ¨ thrich K (1996) MOLMOL: a progam for display and analysis of macromolecular structures. J Mol Graph 14, 51–55. 49 Laskowski RA, Rullmann JA, MacArthur MW, Kaptein R & Thornton JM (1996) AQUA and PRO- CHECK-NMR: programs for checking the quality of protein structures solved by NMR. J Biomol NMR 8, 477–486. 50 Lyles MM & Gilbert HF (1991) Catalysis of the oxidative folding of ribonuclease A by protein disulfide isomerase: dependence of the rate on the composition of the redox buffer. Biochemistry 30, 613–619. 51 Raturi A, Vacratsis PO, Seslija D, Lee L & Mutus B (2005) A direct, continuous, sensitive assay for protein disulphide-isomerase based on fluorescence self-quench- ing. Biochem J 391, 351–357. Supporting information The following supplementary material is available: Fig. S1. Comparison of 1 H- 15 N HSQC spectra of human PDI-b¢ (A) and PDI-bb¢ (B) in the absence (black) and presence (red) of mastoparan (at 8 : 1 pep- tide-protein ratio). Fig. S2. Values of the 15 N{ 1 H} heteronuclear NOE for backbone amides in human PDI-bb¢, the correlation between the observed and back-calculated RDCs for solution structure of human PDI-bb¢, and changes of chemical shifts in PDI-bb¢ Asp297 versus peptide concentrations. Fig. S3. Multiple sequence alignment of the b¢ domains for human PDI family. This supplementary material can be found in the online version of this article. Please note: Wiley-Blackwell is not responsible for the content or functionality of any supplementary materials supplied by the authors. Any queries (other than missing material) should be directed to the corre- sponding author for the article. A. Y. Denisov et al. Structure of the bb¢ domains from human PDI FEBS Journal 276 (2009) 1440–1449 ª 2009 The Authors Journal compilation ª 2009 FEBS 1449 . folded proteins primarily via its third domain, b¢. Here, we determined the solution structure of the sec- ond and third domains of human protein disulfide isomerase. between the protein fold of human and yeast PDI bb¢ domains was an extra helix, a3, in the b domain of the human protein and an extra a-helix in the b¢ domain