High-resolution Structures of Escherichia coli cDsbD in Different Redox States: A Combined Crystallographic, Biochemical and Computational Study

doi:10.1016/j.jmb.2006.02.030

J. Mol. Biol. (2006) 358, 829–845

High-resolution Structures of Escherichia coli cDsbD in Different Redox States: A Combined Crystallographic, Biochemical and Computational Study Christian U. Stirnimann1, Anna Rozhkova2, Ulla Grauschopf2 Rainer A. Bo¨ckmann1,3, Rudi Glockshuber2*, Guido Capitani1* and Markus G. Gru¨tter1 1

Biochemisches Institut Universita¨t Zu¨rich Winterthurerstrasse 190 8057 Zu¨rich, Switzerland 2

Institut fu¨r Molekularbiologie und Biophysik, Eidgeno¨ssische Technische Hochschule Ho¨nggerberg, 8093 Zu¨rich Switzerland 3

Theoretical and Computational Membrane Biology, Center for Bioinformatics Saar Universita¨t des Saarlandes 66041, Saarbru¨cken, Germany

Escherichia coli DsbD transports electrons from cytoplasmic thioredoxin to periplasmic target proteins. DsbD is composed of an N-terminal (nDsbD) and a C-terminal (cDsbD) periplasmic domain, connected by a central transmembrane domain. Each domain possesses two cysteine residues essential for electron transport. The transport proceeds via disulfide exchange reactions from cytoplasmic thioredoxin to the central transmembrane domain and via cDsbD to nDsbD, which then reduces the periplasmic target proteins. We determined four high-resolution structures ˚ resolution), chemically reduced (1.3 A ˚ ), photoof cDsbD: oxidized (1.65 A ˚ reduced (1.1 A) and chemically reduced at pH increased from 4.6 to 7. The ˚ resolution, the highest achieved so far latter structure was refined at 0.99 A for a thioredoxin superfamily member. The data reveal unprecedented structural details of cDsbD, demonstrating that the domain is very rigid and undergoes hardly any conformational change upon disulfide reduction or interaction with nDsbD. In full agreement with the crystallographic results, guanidinium chloride-induced unfolding and refolding experiments indicate that oxidized and reduced cDsbD are equally stable. We confirmed the structural rigidity of cDsbD by molecular dynamics simulations. A remarkable feature of cDsbD is the pKa of 9.3 for the active site Cys461: this value, determined using two different experimental methods, surprisingly was around 2.5 units higher than expected on the basis of the redox potential. Additionally, taking advantage of the very high quality of the cDsbD structures, we carried out pKa calculations, which gave results in agreement with the experimental findings. In conclusion, our wide-scope analysis of cDsbD, encompassing atomic-resolution crystallography, computational chemistry and biophysical measurements, highlighted two so far unrecognized key aspects of this domain: its unusual redox properties and extreme rigidity. Both are likely to be correlated to the role of cDsbD as a covalently linked electron shuttle between the membrane domain and the N-terminal periplasmic domain of DsbD. q 2006 Elsevier Ltd. All rights reserved.

*Corresponding authors

Keywords: high-resolution crystal structures; cDsbD; oxidative protein folding; redox properties; protein stability

Abbreviations used: TCEP, Tris(2-carboxyethyl) phosphine; GdmCl, guanidinium chloride; IAM, iodoacetamide; EC-Trx, E. coli thioredoxin; H-Trx, human thioredoxin; MD, molecular dynamics; PB, Poisson Boltzmann. E-mail addresses of the corresponding authors: [email protected]; [email protected] 0022-2836/$ - see front matter q 2006 Elsevier Ltd. All rights reserved.

830

Introduction Dsb proteins catalyze the formation and rearrangement of disulfide bonds in the periplasm of Escherichia coli. Two independent Dsb pathways are known: the oxidative DsbA/B and the reductive DsbC/D pathway.1,2 In the former, the strong dithiol oxidase DsbA introduces disulfide bonds into reduced polypeptides in a random and very fast way.3–5 Upon oxidation of substrate proteins, DsbA becomes reduced and is re-oxidized by the inner membrane protein DsbB, which transfers electrons from DsbA to ubiquinone.6–9 Since DsbA has no disulfide isomerase activity, rearrangement of wrong disulfide bonds in the scrambled polypeptides to the native conformation has to be catalyzed by another enzyme, the periplasmic disulfide isomerase DsbC.10–13 The inner membrane protein DsbD is required for maintenance of DsbC in the active, reduced state in the oxidizing periplasm. DsbD supplies two other periplasmic proteins with reducing equivalents: DsbG, a homolog of DsbC of unknown function,14–16 and CcmG, a specialized thiol reductase that is essential for cytochrome c maturation.17 DsbD consists of three domains: an N-terminal periplasmic domain (nDsbD, referred to as DsbDa in previous publications) with an immunoglobulinlike fold,18,19 a central transmembrane domain (tDsbD, referred to as TMD of DsbD or DsbDb in previous publications), composed of eight predicted transmembrane helices,20 and a C-terminal periplasmic thioredoxin-like domain (cDsbD, referred to as DsbDg in previous publications).21 Each domain has two conserved cysteine residues that are required for electron transport.22 Intramolecular electron transfer within DsbD is assumed to proceed exclusively through sequential disulfide exchange reactions between its three domains.23–26 The electron cascade starts with the reduction of tDsbD by thioredoxin, then cDsbD shuttles electrons from tDsbD to nDsbD, and finally reduced nDsbD passes electrons to DsbC, DsbG or CcmG. Two structural studies of cDsbD have been published so far: the first described the crystal structure of oxidized cDsbD (residues 423–546 of ˚ resolution and mature E. coli DsbD) at 1.9 A revealed a thioredoxin fold with an extended N-terminal stretch.21 In the second study, we determined the structure of the mixed disulfide ˚ complex between cDsbD and nDsbD (2.85 A resolution), which represents an important reaction intermediate in the catalytic cycle of DsbD.27 Many questions regarding the structural and biochemical properties of cDsbD, however, remain to be answered: what is the structure of the reduced form; are there conformational changes upon oxidation/reduction of the active site disulfide; what is the pKa of the active site; and what is the thermodynamic stability of the oxidized and reduced forms? 2 In the present study, we used a combination of high-resolution protein crystallography, biochemical experiments and

Structural Studies of cDsbD

computational approaches to answer these questions. We determined four high-resolution crystal structures of cDsbD, three at atomic or near-atomic resolution: (1) cDsbD chemically reduced by Tris(2-carboxyethyl) phosphine (TCEP) ˚ at pH 4.6 (termed cDsbDred from now on, 1.3 A structure), (2) cDsbD TCEP-reduced at pH 7 ˚ structure), (3) cDsbD photo(cDsbDpH7, 0.99 A ˚ ), and (4) cDsbD oxidized reduced (cDsbDpr, 1.1 A ˚ ). To our knowledge, cDsbDpH7 is (cDsbDox, 1.65 A the highest resolution structure determined so far for a member of the thioredoxin superfamily. cDsbDpH7 has a protonated active site Cys461. In order to determine the pKa value of Cys461, we studied the pH-dependent reactivity of Cys461 with the alkylating reagent iodoacetamide and the pH-dependent thiolate-specific absorbance at 240 nm. These experiments showed that the active site Cys461 has a pKa value of 9.3, unexpectedly high for a thioredoxin-like protein with a redox potential of K235 mV.27–29

Results and Discussion Crystallization of cDsbD Oxidized and chemically reduced cDsbD crystallized under the same conditions. The first dataset collected for cDsbDox was photo-reduced (termed cDsbDpr from now on). A dataset of cDsbD in an intact oxidized state (cDsbDox) was measured from a crystal grown under similar conditions (the mixture of ammonium and sodium acetate in the buffer in that case was replaced by sodium acetate only). In both cases, the crystals grew within three to five days. cDsbD formed plate-like crystals of about 100!650!30 mm that were arranged in bundles and star-like structures. Before being taken out of the drop, the crystal bundles were separated into single-crystal fragments. All crystals belong to the orthorhombic space group P212121 with similar cell parameters and contain one molecule per asymmetric unit. In comparison, the lower-resolution SeMet cDsbD crystals described by Kim et al. belonged to the same space group with ˚ ), very similar unit cell parameters (resolution 2.3 A while the native crystals had different parameters (same space group) and contained two molecules per asymmetric unit (PDB-code: 1UC7, resolution ˚ ).21 1.9 A X-ray structure determination of cDsbD and structural quality The first structure solved in this study was that ˚, from an oxidized cDsbD crystal diffracting to 1.1 A which was photo-reduced by synchrotron radiation (cDsbD pr). The crystal packing is extremely dense with a very low Matthews parameter of ˚ 3 DaK1. 1.6–1.7 A cDsbDpr was refined anisotropically to a resol˚ (Table 1). Given the high resolution of ution of 1.1 A

831

Structural Studies of cDsbD

Table 1. Crystallographic data and refinement statistics A. Data collection Radiation source Dataset ˚) Wavelength (A Space group Unit cell parameters ˚) a (A ˚) b (A ˚) c (A ˚) Resolution range (A No. reflections No. unique reflections Completeness (%) Rsym Average I/s Redundancy B. Refinement statistics ˚) Resolution (A No. reflections (test) No. atoms No. water molecules No. ions R factor Rfree rmsd from ideal ˚) Bond lengths (A ˚) Bond angles (A ˚) DPI value (A ˚ 2) Average B-factor (A Ramachandran plot regions Most favored (%) Additionally allowed (%) a b c d e

cDsbDox 0.7514 P212121

SLS Villigen, CH beamline X06SA cDsbDpr cDsbDred 0.8600 1.0000 P212121 P212121

cDsbDpH7 0.8555 P212121

30.3 46.0 73.8 20–1.65 88,321 13,029 99.5 (97.7)a 9.3 (44.7)a 19.0 (3.3)a 6.8

30.17 45.97 73.82 16–1.1 166,079 42,360 99.7 (99.5)b 12.4 (42.9)b 6.9 (3.0)b 3.9

30.36 46.08 74.14 20–1.3 138,741 25,864 98.2 (87.0)c 9.1 (28.8)c 20.8 (3.1)c 5.4

30.29 46.07 74.07 20–0.99 337,860 58,388 99.6 (96.9)d 7.9 (44.5)d 18.8 (2.6)d 5.8

20–1.65 11,933 (950) 967 173 9 0.165 (0.162)e 0.235 (0.219)e

16–1.1 41,402 (953) 967 176 0 0.149 (0.135)e 0.174 (0.163)e

15–1.3 25,054 (779) 977 164 9 0.170 (0.160)e 0.196 (0.187)e

15–0.99 56,874 (1437) 921 200 3 0.113 (0.105)e 0.146 (0.138)e

0.008 0.024 0.110 16.4

0.020 0.030 0.041 14.5

0.011 0.026 0.056 17.6

0.014 0.030 0.031 12.2

92.7 7.3

93.6 6.4

93.7 6.3

95.3 4.7

˚. Outermost shell: 1.71–1.65 A ˚. Outermost shell: 1.15–1.10 A ˚. Outermost shell: 1.35–1.30 A ˚. Outermost shell: 1.03–0.99 A FoO4s(Fo).

the data, hydrogen atoms could be included in the final stage of refinement. The model contains residues 427 to 543 of cDsbD and five partly disordered residues of the His6-tag (547–551). Residues 544 to 546 are very disordered and could not be included in the final model. The final R-factor is 0.149 and the Rfree is 0.174. To protect the catalytic disulfide bridge of oxidized cDsbD from radiation damage, a second dataset was measured using an aluminum filter that decreased the beam intensity to 70.3% of the ˚ ) was original value. A shorter wavelength (0.751 A chosen to reduce radiation damage.30 The crystal ˚ and the disulfide bridge diffracted to 1.65 A remained intact, with no positive or negative difference density being visible at its site. The cDsbDox structure, encompassing residues 428 to 549, was finally refined to an R-factor of 0.165 and Rfree of 0.235 (Table 1). Heavy ions (iodide and nickel) were refined anisotropically. This approach led to a decrease of the R-factor by 0.2% and Rfree by 0.3%. The cDsbDred dataset was collected at a wave˚ to a resolution of 1.3 A ˚ . The length of 1.000 A cDsbDred structure includes residues 428 to 550 and was refined isotropically (with the exception of the active site sulfur atoms and of some bound iodide ions, which were treated anisotropically) to an R-factor of 0.170 and Rfree of 0.196 (Table 1).

A fourth dataset (cDsbDpH7) was collected from a crystal of TCEP-reduced cDsbD, which was soaked in a solution at pH 7. The data collection ˚ and the diffraction limit wavelength was 0.856 A ˚ using the aluminum filter described reached 0.99 A for cDsbDox. cDsbDpH7 was refined anisotropically. Since almost all main-chain and several side-chain hydrogen atoms were already visible after the fourth refinement cycle, they were included in the refinement. In a final cycle, all visible hydrogen atoms were included, with the final R-factor and Rfree leveling at 0.113 and 0.146, respectively (Table 1). The model contains residues 428 to 543. Residues 544 to 546 and the C-terminal His6-tag are disordered and correspond to uninterpretable density, therefore they were not modeled. All four final atomic models exhibit excellent Ramachandran plot statistics with all residues found in most favored and additionally allowed regions (Table 1). Overall structure and comparison of different cDsbD structures cDsbD possesses a thioredoxin-like fold. Here, we follow the nomenclature for the secondary structure elements introduced by Kim et al.21 with the only difference being that those authors refer ˚ crystal structure to cDsbD as DsbDg. The 1.9 A

832 reported by Kim et al. contained two molecules per asymmetric unit (termed cDsbD ox1 and cDsbDox2)21. Both cDsbDox1 and cDsbDox2 encompassed residues 423 to 546, with the N-terminal residues being involved in a non-physiological dimer contact. Our cDsbD models (three of which were refined at atomic or near-atomic resolution) start with residue 427 (cDsbDpr) or 428 (cDsbDox, cDsbDred and cDsbDpH7). All four models correspond to excellent electron density up to residue 543. Figure 1(a) shows a close-up view of the cDsbDpH7 active site. Comparison of the four structures determined in this work shows that cDsbD is a very rigid domain. Except for the active site region, superposition of cDsbDox, cDsbDpr , cDsbDred and cDsbDpH7 reveals no significant structural change. Ca rmsd values for ˚ all pairs of structures lie in the range 0.12–0.24 A (calculated using residues His428 to Asp543). The Ca rmsd values of the two models described by Kim et al.21 (cDsbDox1 and cDsbDox2, PDB-code: 1UC7) from our four models lie in the range between ˚ and 0.60 A ˚ . Between them, cDsbDox1 and 0.53 A ˚ (see cDsbDox2 exhibit a Ca rmsd value of 0.39 A Supplementary Data, Table 2). Residues 428 to 543 were used for this calculation. Compared to the four

Structural Studies of cDsbD

structures reported here, the 1UC7 structure (cDsbDox1 and cDsbDox2) exhibits two minor differences close to the active site. Tyr470 is involved in a crystal contact in 1UC7,21 which influences the conformation of the last three residues of helix a2a (Glu468-Tyr470). In the structures of cDsbDox, cDsbDred, cDsbDpr and in cDsbDox2 of 1UC7, residue Trp460 is involved in a weak crystal contact with residues of the His6-tag. In cDsbDox2 of 1UC7 the side-chain of Trp460 is in a stacking interaction with the side-chain imidazole ring of His532 from a symmetry-related molecule. In cDsbDpH7 the His6-tag is disordered, so Trp460 appears to be free from crystal contacts. The conformation of the Trp460 side-chain is stabilized ˚ ) between the by a hydrogen bond (length 2.8 A indole nitrogen atom and the main-chain carbonyl group of Thr492. In addition, the loop between helix a1 and b-strand 2 as well as that between strand b4 and helix a4, form crystal contacts and show minor structural rearrangements when compared to 1UC7. The quality of the four structures was analyzed by calculating Cruickshank’s diffractioncomponent precision index (DPI),31 using SFCHECK.32 The DPI values for our four structures

˚ Figure 1. (a) Close-up view of the active site of cDsbDpH7 (grey and atom colors) showing a 2FoKFc map (0.99 A resolution) contoured at 1s. (b) Superposition of cDsbD from the nDsbD-SS-cDsbD complex (cDsbDco; blue)27 onto cDsbDpr (grey). The main-chain carbonyl group of Thr529 assumes a different conformation in cDsbDpr (grey and atom colors) and in cDsbDco (cyan and atom colors). This conformational change is forced, upon formation of the nDsbD-SScDsbD complex, by direct van der Waals interactions of the Thr529 carbonyl group with the Phe11 and Phe108 side˚ ) between the carbonyl chains of nDsbDco (cyan and atom colors). Yellow broken lines indicate the real distance (3.1 A oxygen atom of cDsbD Thr529 and the nearest atom of nDsbD Phe11 in the nDsbD-SS-cDsbD complex, and the distance ˚ ) that would result if the Thr529 carbonyl group retained the same conformation as in uncomplexed cDsbD. (2.4 A

833

Structural Studies of cDsbD

indicate a much lower coordinate error (DPI ˚ for cDsbDred, ˚ for cDsbDox, 0.056 A of 0.110 A ˚ for cDsbDpH7) ˚ 0.041 A for cDsbDpr and only 0.031 A than for the oxidized structures reported by Kim ˚ for cDsbDox1 and cDsbDox2). (DPI of 0.204 A Upon reduction, the sulfur atom of the buried ˚ active site cysteine residue (Cys464) moves by 0.6 A towards the interior of the protein (comparing cDsbDox and cDsbDred). This corresponds to a change of 68 in the c1 angle and K38 in the f angle of Cys464. For the accessible, nucleophilic Cys461, the f angle changes by K68 and the c1 angle by 68. The sulfur–sulfur distance in cDsbDred ˚. is 3.59 A Superposition of the cDsbD domain from the nDsbD-SS-cDsbD mixed disulfide complex (termed cDsbDco from now on)27 onto unliganded cDsbD reveals differences only for helix a2a (including the two active site cysteine residues) and for Thr529. In the complex, helix a2a is pulled slightly towards the Cap-loop region of nDsbD (Asp68-Gly72),18 probably as a consequence of the Cys461 (cDsbD)–Cys109 (nDsbD) disulfide bond formation. The main-chain carbonyl group of cDsbDco Thr529 is forced by direct van der Waals interactions with the two side-chains of Phe11 and Phe108 of nDsbD to take up a different conformation with respect to unbound cDsbD: 4 and f change by approximately 108 and 1008, respectively (Figure 1). This change of Thr529 corresponds to a shift in the Ramachandran plot from the additionally allowed a-region to the most favored b-region. In general, compared to nDsbD, which undergoes significant opening of the Cap-loop region upon complex formation,19,27,33 cDsbD exhibits only small structural rearrangements. cDsbDox and cDsbDred crystals grew (and were soaked in, in the case of cDsbDpH7) in the presence of sodium iodide. Several iodide-binding sites were identified during refinement (see Supplementary Data for a detailed description).

Figure 2. Mechanism of disulfide opening in crystals induced by synchrotron radiation (adapted from Weik et al.37 and Fauvodon et al.40).

˚ resolution was dent refinement with CNS41 at 1.4 A then carried out for each dataset. In all cases, van der Waals interactions between the two sulfur atoms of the active site cysteine residues were made to be ignored during refinement in CNS. Since each dataset provides a time and spaceaveraged structure of the crystal content, it was expected that the sulfur–sulfur active site distance would increase with progressive radiation damage and increasing percentage of the unit cells with damaged disulfides. Indeed, the refined sulfur– ˚ sulfur distances increased in each step from 2.44 A for the first reflection file based on 150 frames to ˚ for all 500 frames. Already, for the first 2.65 A reflection file, opening of the disulfide bridge is clearly visible. The observed sulfur–sulfur distance ˚, ˚ longer than that for cDsbDox (2.05 A is 0.39 A restrained). To examine the disulfide bridge state at the end of the measurement, two additional reflection files were created, encompassing frames ˚) 350 to 550 (refined sulfur–sulfur distance of 2.96 A and of frames 420 to 676 (refined sulfur–sulfur ˚ ). These values indicate that even distance of 2.99 A at the end of data collection, the disulfide bridge is not reduced in all unit cells (in comparison, ˚ , and cDsbDpH7 cDsbDred shows a distance of 3.59 A ˚ ). The analysis is shows a distance of 3.55 A summarized in Supplementary Data, Table 1.

Radiation-driven disulfide-reduction of cDsbDpr Disulfide bonds are the most radiation-sensitive moieties in proteins.34–36 The first step of the radiation-driven opening of a disulfide bond is the trapping of an electron (RSSR%K). Spontaneous and reversible bond rupture to RS% and RSK can then occur. Protonation of the RSSR%K radical leads to the formation of a RSSRH % species and the equilibrium shifts towards the broken state: upon bond rupture, a thiol (RSH) and a thiyl radical (RS%) are formed. Figure 2 shows an overview of the disulfide-opening mechanism.37–40 The availability of a high-resolution and very redundant radiation-damaged dataset of cDsbD (cDsbDpr) allowed a detailed analysis of this phenomenon. The dataset was partitioned into eight incremental reflection files. A total of 150 diffraction images were included in the first reflection file. For every additional reflection file, the number of diffraction frames included was increased by 50. An indepen-

Analysis of radiation damage in the four datasets The large amount of radiation damage suffered by the active site in cDsbDpr prompted us to carry out a general analysis of damage for all four cDsbD datasets, using isomorphous FoKFo difference Fourier maps (see Materials and Methods).34,42 In accordance with peaks observed in conventional FoKFo maps (Figure 3(a)), FoKFc maps of cDsbDpr ˚ , exposure time per image, 0.5 s) show (lZ0.8599 A clear and heavy radiation damage of the active site cysteine residues (Figure 3(b), contoured at 4s) and also of most glutamate and aspartate side-chains. Remarkably, even radiation damage of the mainchain (reviewed by O’Neill et al.43) is detected.43 To improve the quality of the structure, we used only the minimum of 200 frames for refining the model for cDsbDpr. As mentioned above, radiation damage of the disulfide bridge was minimized by

834

Structural Studies of cDsbD

Figure 3. Radiation damage in cDsbDpr. (a) Active site residues are depicted in grey and atom colors, the corresponding final 2FoKFc map (contoured at 1s) in blue, the final FoKFc difference map (contoured at 3s) in green and (b) the isomorphous FoKFo Fourier difference map (contoured at 4s) in magenta. Clear photo-reduction of the active site disulfide bond between Cys461 and Cys464 is visible in the FoKFc and in the FoKFo map.

collecting a cDsbDox dataset at very short wave˚ ) and using an aluminum filter (IZ length (0.7514 A 0.703I0). Weak radiation damage is visible in the active site also for this dataset. The pH-shifted structure (cDsbDpH7) is clearly radiation-damaged at the active site sulfur atoms and at several aspartate and glutamate side-chains. In addition to the FoKFo difference peaks, negative density appeared for the active site sulfur atoms when they were refined with full occupancy: occupancies were thus lowered in the final model to 0.9 and to 0.8 for the Cys464 and Cys461 sulfur atoms, respectively. Interestingly, the cDsbD red dataset does not exhibit significant radiation damage in FoKFo difference density maps, even at the active site sulfur atoms. This may be due to a lower intensity of the synchrotron beam when that dataset was collected, compared to cDsbDpH7 and cDsbDpr. Thermodynamic stability of oxidized and reduced cDsbD Guanidinium chloride (GdmCl)-induced) unfolding and refolding transitions of oxidized and reduced cDsbD at 25 8C and pH 7.0 were followed by measuring fluorescence at 345 nm (excitation at 280 nm). The transitions were cooperative and fully reversible (Figure 4(a)). Evaluation of the data according to the two-state model yielded DG values of K43.9(G2.9) kJ molK1 for cDsbDox and

K45.5(G5.0) kJ molK1 for cDsbDred, indicating that at pH 7.0 both redox forms are equally stable. pKa of the active site Cys461 and reactivity with DTT To determine the pKa of the cDsbD active site, we first investigated the pH-dependent reactivity of Cys461 with iodoacetamide (IAM).6–9 cDsbDred (5 mM) was mixed with excess of IAM (between 0.1 mM and 10 mM) in the pH range of 4–10. Samples were removed after incubation for different lengths of time, and the reaction was quenched with acid. The reaction products were separated by reversed-phase HPLC, and amounts of reduced and IAM-modified cDsbD were quantified by integration of the peak areas. Figure 4(b) shows the pH-dependence of the apparent second-order rate constant (kIAM) of the alkylation of Cys461 in reduced cDsbD by IAM. The transition appeared to be biphasic, as described for thioredoxin.6–9 Although this method did not provide an exact pKa value, we can conclude that the pKa of Cys461 is above 8.0. We then investigated the pH-dependence of the thiolate-specific absorbance at 240 nm.44–46 The protein solution (initial concentration of 30 mM) was titrated in the pH range of 4–12 and the absorbance at 240 nm and 280 nm was recorded. Far-UV CD measurements proved that the secondary structure of cDsbD is not changed in this pH range

Structural Studies of cDsbD

835

Figure 4. Biophysical properties of cDsbD. (a) GdmCl-induced folding and unfolding transitions of oxidized (circles) and reduced (squares) cDsbD at 25 8C and pH 7.0. Unfolding (filled symbols) and refolding (open symbols) were followed by measuring fluorescence at 345 nm (excitation at 280 nm). Original data were evaluated according to the two-state model and normalized. Continuous and broken lines represent the fit for cDsbDox and cDsbDred, respectively. The m-values of 15.9(G1.1) kJ molK1 MK1 and 17.5(G1.9) kJ molK1 MK1 and DG of K43.9(G2.9) kJ molK1 and K45.5(G5.0) kJ molK1 for the oxidized and reduced forms, respectively, are identical within experimental error. (b) pH-dependence of the active site Cys461 reactivity with IAM. The apparent second-order rate constants (kIAM) were determined by HPLC analysis as described in Materials and Methods. (c) pH-dependence of the ionization of active site thiol groups of cDsbD (circles) and of the single cysteine variant cDsbD-C464S (squares). The thiolate-specific absorbance was monitored at 240 nm and the fraction of the thiolate anion was calculated as described in Materials and Methods. The fit (continuous line) yields a pKa value of 9.4 for cDsbD and 9.3 for Cys461 in cDsbD-C464S. (d) Very low reactivity of cDsbD with DTTat 25 8C and pH 7.0. The reduction of cDsbD by DTT was performed under pseudo first-order conditions with initial concentrations of cDsbDox of 5 mM and DTTof 1 mM. The amounts of the reduced and oxidized cDsbD species after incubation for different lengths of time were quantified by reverse-phase HPLC. The fit (continuous line) yields a pseudo first-order rate constant of 0.0154 sK1, which corresponds to a second-order rate constant of 15.4 MK1 sK1.

(data not shown), in full agreement with the crystallographic results. The titration curves of cDsbDred (circles) and of its single cysteine variant cDsbD-C464S (squares) are depicted in Figure 4(c). Both cDsbDred and cDsbD-C464S showed a single transition, with a midpoint at pH 9.2–9.5. From these data, we can conclude that Cys461 has an unusually high pKa of about 9.3, which is similar to the pKa of a normal cysteine thiol group. It has been shown that the pKa of the nucleophilic active site thiol groups in DsbA and thioredoxin variants is lowered and correlates with the redox potential of these enzymes.45,47 cDsbD has a redox potential of K235 mV.27,28 Accordingly, the corresponding theoretical pKa for the exposed active site Cys lies between 6 and 7. Although the pKa determined for cDsbD Cys461 deviates significantly from theory, it

is entirely consistent with our results obtained by the IAM-modification method (Figure 4(b)) and with the identical stabilities of both redox forms at pH 7.44 We also studied the reactivity of cDsbDox with DTT at 25 8C and pH 7.0. Measurements were carried out under pseudo first-order conditions with initial concentrations of cDsbDox and DTT of 5 mM and 1 mM, respectively. Samples were removed after incubation for different lengths of time, and the reaction was quenched with acid. Reaction products were separated by reversedphase HPLC, and the amounts of cDsbDox and cDsbDred were quantified by integration of the peak areas. Figure 4(d) shows that the reduction of cDsbDox by DTT is (in contrast to electron transfer between other Dsb proteins) very slow, with an

836 apparent second-order rate constant of 15 MK1 sK1. This value is typical for uncatalyzed thiol–disulfide exchange between small organic molecules.48 pKa and molecular dynamics calculations of cDsbD pKa calculations using the crystal structure of cDsbDred are consistent with the experimentally determined high pKa of 9.3 for Cys461: for this residue, the calculated pKa value is 10.0. The difference between the calculated and measured (9.3) pKa values of Cys461 is presumably due to the water accessibility of some neighboring titratable groups, thereby increasing the dielectric constant locally. Initial calculations (yielding values of 11.2 for Cys461 and of O20 for Cys464) were refined by taking into account the greater flexibility of solventexposed side-chains compared to buried sidechains. This was achieved by introducing a second, larger protein dielectric constant (eZ16) for amino ˚ 2.49 acid residues with B-factors O12 A The intrinsic pKa values, i.e. values calculated excluding the influence of other titratable groups in the vicinity, are lower (9.8 for Cys461 and 11.6 for Cys464) than the full pKa values (using one dielectric constant 3Z8 for the protein). Our analysis shows that lowering in the pKa of Cys464 is caused mainly by the vicinity to Asp455. In contrast to the equivalent residue in E. coli thioredoxin (Asp26),50 the calculated pKa of Asp455 is 4.8 for cDsbDred (pKaO7 for Asp26 in thioredoxin). Interestingly, the predicted pKa of Asp455 is increased dramatically to around 9.9 within the nDsbD-SS-cDsbD complex, rendering the protonation of Asp455 in the intermediate nDsbD-SScDsbD complex highly probable. In order to evaluate possible flexibilitydependent fluctuations in the pKa of the active site

Structural Studies of cDsbD

cysteine residues, we performed molecular dynamics simulations (Figure 5) of cDsbDred with both Cys461 and Cys464 protonated (system A) and with Cys464 deprotonated (thiolate state, system B). The backbone rmsd converged for system A within 1 ns to ˚ and for system B within 3 ns to approximately 1.8 A ˚ 2.5 A. While the active site Cys sulfur atoms stay close ˚ in system A for the total within approximately 4.6 A simulation time, the sulfur–sulfur distance fluctuates considerably upon deprotonation of Cys464 (system ˚ and 9.3 A ˚ . The averaged pKa values B) between 6.2 A (see Materials and Methods) of Cys461 (10.9), Cys464 (19.0) and of Asp455 (3.8) for the reduced cDsbD (system A) are close to the respective values of the crystal structure. Deprotonating Cys464 (system B) shifts the pKa during the first 1–2 ns to values as low as 9–11; however, after about 4 ns, the initial high pKa with values O15 is restored. This increase is coupled to the pKa increase of Asp455 to approximately 5–6. The cDsbD fold displays a considerable rigidity in the simulations: excluding the termini from the analysis, the rmsd of cDsbD (residues 432–544, ˚. system A) is as low as 1.1–1.3 A Comparison of the cDsbD and DsbA active sites The active sites of cDsbD and DsbA are structurally similar but their redox potentials (K235 mV 1,27 for cDsbD and K122 mV 47 for DsbA, respectively) and pKa values (9.3 and 3.5,44,45 respectively) are strongly different. We compared the structures (oxidized and reduced forms) of the two proteins to find an explanation. As for cDsbD, there are no large conformational differences between oxidized and reduced DsbA.51 Both in cDsbD and DsbA the more C-terminal sulfur atom moves only upon reduction and remains totally buried in the interior of the protein. The sulfur atom of the solvent-accessible cysteine

Figure 5. Titration curves for snapshots (DtZ200 ps) of a simulation with both Cys461 and Cys464 protonated (A, system A) and of a simulation with Cys464 deprotonated (B, system B). The time-sequence is color-coded, starting from black to blue, green, yellow and red. The average over ten titration curves from the last nanosecond is shown as a broken blue line, a fit according to the Henderson–Hasselbalch equation to the average titration curve is shown as a broken red line.

837

Structural Studies of cDsbD

Figure 6. Reduced DsbA active site (green and atom colors) superimposed onto the active site of cDsbDred (magenta and atom colors).

˚ upon residue moves in both cases by around 1 A reduction. Guddat et al. described the network of hydrogen bonds and electrostatic interactions that stabilize the Sg thiolate of DsbA Cys30.51 The structure of reduced DsbA (PDB entry 1A2L) was determined at pH 5.6, a condition under which Cys30 is known to be deprotonated. Our biochemical results (Figure 4(c)) show that in cDsbDred, a structure determined at pH 4.6, the Sg of Cys461 is known to be protonated. We analyzed the hydrogen bonding pattern in the cDsbDred active site region with the program HBPLUS.52 The program did not find any stabilizing uncharged hydrogen bonds received by the Sg of Cys461. Instead, the algorithm assigned the thiol group of Cys461 as a hydrogen bond donor to the thiol group of Cys464. The sidechain of DsbA His32, which was proposed to electrostatically stabilize, in its protonated state, the Sg thiolate of Cys30, is replaced by Ala in the DsbD sequence. The role of DsbA His32 has been discussed extensively.51,53,54 Mutation analysis by Guddat et al. showed a destabilization effect of oxidized DsbA caused by His32.53 Schirra et al. concluded that the interaction of the Cys30 thiolate with the active site helix dipole is the main reason for the very low pKa value.54 Also, the conformations of DsbA Cys30 and of the corresponding Cys461 of cDsbD (the so-called accessible cysteine residues) are slightly different (Figure 6). The fact that the structures of DsbA and cDsbDred were ˚ for DsbA, 1.3 A ˚ solved at different resolution (2.7 A for cDsbDred) precludes conclusive statements based on the slight differences observed in their active sites: however, the difference in orientation between the side-chains of Cys30 and Cys461 is very likely to be related to the different conformations of the regions preceding the active site and may be related to the different protonation Table 2. Interatomic distances in the active site of cDsbD

Sg461–Sg464 Sg461–N464 Sg461–N463 Sg464–O461 Sg464–N458 Sg464–O458

cDsbDox

cDsbDred

cDsbDpH7

cDsbDpr

2.0 3.2 4.1 4.1 4.1 4.4

3.6 3.7 3.8 4.3 3.7 4.2

3.6 3.7 3.8 4.2 3.8 4.3

2.5 3.3 4.0 4.2 3.9 4.5

states of the two Cys side-chains in the respective experimental conditions. To test for conformational changes upon deprotonation of the Cys Sg atoms, we soaked the cDsbDred crystals in solutions at higher pH. The crystals were stable in solutions up to a pH value of 6.5, while at pH 7 they started to dissolve after 10–15 min (see comment in Supplementary Data). Crystals soaked in solutions with pH higher than 7 dissolved immediately. We were able to collect a dataset from a crystal that showed incipient signs of stress in the soaking solution at pH 7 (this appears to be an effect mediated by the C-terminal His6-tag; see Supplementary Data). The resulting crystal structure (cDsbDpH7) exhibits no significant conformational change in the active site compared to cDsbDred (Table 2). The refined positions of the active site sulfur atoms in cDsbDpH7 were validated by calculating a Bijvoet difference Fourier map, which featured a 3.5s peak for the Sg of Cys461 and a 2.5s peak for that of Cys464. Both corresponded to the refined atomic positions of the sulfur atoms. These results are fully consistent with the aforementioned biochemical and computational results, indicating pKaO9 for Cys461. Comparison of the cDsbD with the active sites of E. coli and human thioredoxin A DALI search55 against the Protein Data Bank† identified E. coli thioredoxin (EC-Trx) as the closest ˚ ). relative of cDsbD (Z-score of 14.7; rmsd of 1.9 A a The C -trace of the CGPC active site motif of ˚ )56 is nearly identical oxidized EC-Trx (2TRX, 1.68 A ˚ (rmsd 0.042 A) with that of cDsbDox (sequence CVAC). No crystal structure of reduced EC-Trx has been solved, but an NMR structure is available. The active site conformations of cDsbDred and of reduced EC-Trx (1XOB57: model 2 identified as the most representative using the OLDERADO server) 58 are very similar, with a C a rmsd ˚ . The distance between the two active of 0.155 A ˚. site sulfur atoms in reduced EC-Trx is 3.8 A The second closest relative to cDsbD is human thioredoxin (H-Trx) with a Z-score of 14.0 and an † http://www.pdb.org

838

Structural Studies of cDsbD

˚ . Both oxidized and reduced crystal rmsd of 2.1 A structures of H-Trx are available (PDB codes 1ERU and 1ERT, respectively).59 The active site of oxidized H-Trx (CGPC) is nearly identical (Ca ˚ ) with that of cDsbDox. In contrast, rmsd 0.046 A the active site of reduced H-Trx is more different in its Ca trace conformation compared to cDsbDred or ˚ and 0.132 A ˚ , respectcDsbDpH7 (rmsd of 0.110 A ively). Upon reduction of H-Trx, the side-chain of Trp31 (residue before the active site Cys32) under˚ .59 Such a movement is goes a translation by 1.3 A not observed in cDsbD for the corresponding residue (Trp460). This is probably due to the

˚ ) shared in aforementioned hydrogen bond (2.8 A cDsbD by the side-chain indole nitrogen atom of Trp460 and by the carbonyl oxygen atom of Thr492. In reduced H-Trx, the side-chain indole nitrogen atom of Trp31 forms only a weak hydrogen bond with the side-chain carboxyl ˚ ) and is not hydrogen group of Asp60 (3.6 A ˚ ). In bonded in oxidized H-Trx (distance: 4.0 A reduced H-Trx, the active site Cys sulfur atoms ˚ apart, a much greater distance than that are 3.9 A ˚ ). In light of these observed for cDsbDred (3.6 A observations, the active site of cDsbD appears to be significantly more rigid than that of Trx.

Figure 7 (legend next page)

Structural Studies of cDsbD

839

Figure 7. (a) Multiple sequence alignment of 14 cDsbD sequences from SwissProt60 (computed using CLUSTAL W61). Completely conserved residues are indicated in red, residues O80% conserved are indicated in green and O60% are indicated in blue. The relative alignment numbering is indicated at the top, the absolute numbering for all cDsbD sequences appears at the right. The secondary structure assignment is based on the E. coli cDsbD structure using the nomenclature of Kim et al.21 Abbreviations: EC: Escherichia coli cDsbD; ST: Salmonella typhimurium cDsbD (66% sequence identity and 72% similarity to EC); STi: Salmonella typhi cDsbD (67%, 72%); PC: Pantoae citrea cDsbD (59%, 72%); YP: Yersinia pestis cDsbD (55%, 67%); VP: Vibrio parahaemolyticus cDsbD (58%, 70%); VV: Vibrio vulnificus cDsbD (52%, 64%); VC: Vibrio cholerae cDsbD (49%, 60%); HI: Haemophilus influenzae cDsbD (45%, 60%); PM: Pasteurella multodica cDsbD (42%, 59%); RS: Ralstonia solanacearum cDsbD (42%, 58%); PS: Pseudonomas species cDsbD (33%, 47%); PA2: Pseudonomas aeruginosa cDsbD2 (30%, 45%); PA1: Pseudonomas aeruginosa cDsbD1 (31%, 45%); CM: Camptylobacter jejuni cDsbD (24%, 52%); NM: Nesseria meningitidis serotype A&B cDsbD (25%, 43%). (b) Stereo surface representation of cDsbD mapped by conserved residues. Completely conserved residues belonging to the binding interface are shown in red, other completely conserved residues are shown in light red, residues belonging to the binding interface and O80% conserved are shown in green, other residues O80% conserved are shown in aquamarine, other residues of the binding interface in light blue and the accessible Cys461 are shown in yellow. (c) Stereo surface representation of the cDsbD back side mapped by conserved residues (color code as above).

Comparison with cDsbD in other prokaryotic organisms We compared E. coli DsbD with its prokaryotic orthologs to determine the amount of sequence conservation and to map it onto the structure. A SwissProt database60 search† (date 05/11/16) identified 19 distinct DsbD orthologs. Identical sequences from different E. coli, Neisseria meningitidis or Vibrio vulnificus strains were ignored for the alignment. In all, 16 different sequences were used as input for a multiple sequence alignment in CLUSTAL W,61 using a Gonnet 250 matrix.62 The alignment (Figure 7(a)) shows that E. coli cDsbD is the shortest member, and that the overall degree of conservation is rather high. If one excludes Campylobacter jejuni cDsbD, the active site motif is † www.expasy.org/sprot/

YADWC(V,I)(A,S)CK. C. jejuni possesses the active site pattern: TASWCENCK. This agrees with the observation that C. jejuni cDsbD exhibits the lowest level of sequence identity and the second lowest level of similarity level of all sequences used (see the legend to Figure 7). A more distant relationship of C. jejuni cDsbD to the rest of the group was reported by Kimbal et al. in a phylogenetic analysis of the DsbD family.63 That tree was built using multiple alignments of the transmembrane domain of DsbD (which precedes cDsbD in the sequence) using homolog sequences from all Gram-negative (GK) Proteobacteria, G-bacteria, Gram-positive (GC) bacteria, Archaea, Thermotoga and Cyanobacteria. The multiple sequence alignment in Figure 7(a) was mapped onto the cDsbD structure in order to identify the position of the conserved residues on the surface. Figure 7(b) indicates that most residues located in the binding interface of cDsbD to nDsbD

840 are highly conserved. On the contrary, hardly any conserved residue can be found on the opposite side of the cDsbD surface (Figure 7(c)).

Conclusions The aim of this work was to combine highresolution protein crystallography with biochemical and computational analyses to obtain a complete characterization of cDsbD. As a result, cDsbD structures of unprecedented quality are now available. They reveal that upon reduction, only minor conformational changes take place, which highlights the structural rigidity of the domain (also upon pH change). The structural rigidity, confirmed also by molecular dynamics simulations, is consistent with a model in which cDsbD acts as a “stiff electron shuttle”: it is reduced by docking to a site in the tDsbD transmembrane domain and then, without relevant conformational changes, transfers electrons to the structurally adaptable N-terminal domain, nDsbD.19,27,33 Also the very similar thermodynamic stabilities of the oxidized and reduced forms of cDsbD reflect its structural stiffness. Another interesting feature is the unusually high pKa value of 9.3 of the exposed active site Cys461. Compared to most thioredoxin superfamily structures, its pKa value is 2.3–3.3 units higher than expected on the basis of the redox potential. These experimentally determined values were reproduced with reasonable accuracy by pKa calculations combining a Poisson–Boltzmann solver with a global optimization of the hydrogen-bond network in all protonation states. Previous studies of thioredoxin-like proteins with redox potentials similar to that of cDsbD, such as E. coli thioredoxin (E0ZK270 mV;64 pKaZ6.7–7.665–67), E. coli CcmG (E0ZK203 mV K210 mV;33,68 pKaZ6.768) or Trypanosoma brucei tryparedoxin (E0ZK249 mV;69 pKaZ 7.269) showed pK a values slightly lowered compared to that of 8.3 to 9.5 observed for a normal cysteine residue in a protein.70,71 To our knowledge, cDsbD is the first thioredoxin-like domain where the exposed active site cysteine residue exhibits a pKa that is not lowered and comparable to that of a normal cysteine side-chain, while lower values are common (e.g. for DsbA, E0ZK122 mV;72,73 pKaZ3.544,45 and for DsbC, E0Z140 mV;13,27 pKaZ 4.174). Inconsistencies between active site cysteine pKa values and redox potential in certain members of the thioredoxin superfamily (with a redox potential similar to that of cDsbD) had been observed by Roberts et al.75 in alkyl hydroperoxide reductase (AhpF), the experimentally determined pKa value was 5.1 and that expected from the redox potential (K265 mV76) was higher by around three units.75 The pKa of the active site cysteine residue of E. coli thioredoxin predicts that the protein should be about 40 mV more oxidizing than observed.45,77 An obvious challenge in the future analysis of cDsbD research will be the characterization of its interaction with tDsbD: so far, no tDsbD-SS-cDsbD

Structural Studies of cDsbD

mixed disulfide has been detected or isolated. Moreover, in the current “funnel model” of tDsbD, proposed by Katzen & Beckwith,78 both redox active cysteine residues of tDsbD are facing the cytoplasmic side of the membrane. How cDsbD, which displays comparatively low intrinsic reactivity, can access them from the periplasmic side remains to be determined.

Materials and Methods Protein purification (His)6-tagged cDsbD and its single-cysteine variant cDsbD-C464S (residue numbering corresponds to the mature full-length DsbD) were overexpressed in E. coli, purified, and protein concentration was determined as described.27 Crystallization of cDsbD Purified cDsbD was concentrated to 19.3 mg/ml. The sitting-drop, vapor-diffusion method was used for producing crystals. In the case of cDsbDpr and cDsbDred, 1.4 ml of protein solution was mixed with 2 ml of precipitant solution (0.1 M ammonium acetate, 0.2 M sodium acetate (pH 4.6), 0.1 M sodium iodide, 40% PEG 4K and for cDsbDred additionally 0.3 mM TCEP–HCl. The crystals used for cDsbDpH7 grew under the same conditions as those of cDsbDpr and cDsbDred, but the drop size was 2 ml (1 ml in protein C1 ml precipitant solution). The crystal used to collect cDsbDox was grown in 0.3 M sodium acetate (pH 4.6), 0.1 M sodium iodide and 40% PEG 4K; 0.8 ml of protein solution were mixed with 2 ml of precipitant solution. In the case of both cDsbDpH7 and cDsbDox, plate-like crystals grew at 4 8C within three to five days to maximum dimensions of 100 mm!650 mm!30 mm for a single crystal. Reduced cDsbD was obtained by treating oxidized cDsbD with 3 mM TCEP–HCl. Crystals grew under the conditions described for the oxidized form. Data collection and structure solution All cDsbD crystals were frozen directly in a nitrogen gas stream. No addition of cryoprotectant was necessary. The first diffraction data of oxidized cDsbD were collected at 90.9 K on beamline X06SA at the Swiss Light Source (Paul Scherrer Institut, Villigen, Switzerland) using an MAR CCD image plate at a wavelength of ˚ . This crystal had been soaked in saturated cis0.860 A Pt(NH2)2Cl2 for around 24 h before mounting. Data were ˚ . To obviate processed with XDS,79 to a resolution of 1.1 A a problem with the spindle axis acceleration at the beamline, the processed data were corrected with the program check_spindleshutter (Kai Diederichs, personal communication). The structure of cDsbD was solved by molecular replacement with AMoRe,80 using a truncated ˚ thioredoxin model (PDB code 2TRX)56 against 12–4 A data. The truncated search model was aptly modified by visual inspection of a published stereo picture of cDsbD superimposed on E. coli thioredoxin:21 the segments that appeared least similar to the published cDsbD structure (atomic coordinates had not been released at that time) were removed from the search model. The cDsbD structure was auto-built using ARP/wARP v681 and

841

Structural Studies of cDsbD

iteratively rebuilt in O.82 The structure was first ˚ using CNS41 and then isotropically refined to 1.4 A ˚ ) with anisotropically refined at full resolution (1.1 A SHELXL83 to an R-factor of 0.149 and Rfree of 0.174 (Table 1). SHELXL was used to refine all four structures described here. Since the first dataset of oxidized cDsbD exhibited photoreduction of the disulfide bridge (see Results and Discussion), a second dataset was collected at 98.2 K using an aluminum filter (reducing intensity to 70.3% of ˚ . Oxidized the original value) and a wavelength of 0.751 A cDsbD data (cDsbDox) were processed with DENZO and ˚ . The cDsbDox structure was SCALEPACK,84 to 1.65 A refined isotropically to an R-factor of 0.165 and Rfree of 0.235. Diffraction data for TCEP-reduced cDsbD (cDsbDred) ˚ . Data were collected at 100 K with a wavelength of 1.000 A were processed with DENZO and SCALEPACK to a ˚ and refined to an R-factor of 0.170 and resolution of 1.3 A Rfree of 0.196. A second TCEP-reduced crystal was soaked for ten minutes at room temperature in a solution obtained by mixing 300 ml of precipitant solution (0.1 M ammonium acetate, 0.2 M sodium acetate (pH 4.6), 0.1 M sodium iodide, 40% PEG 4K and 3 mM TCEP–HCl) and 167.5 ml of titration solution (0.3 M Hepes base, 0.1 M NaI, 40% PEG 4K and 3 mM TCEP–HCl), resulting in a final pH of 7. Diffraction data for TCEP-reduced and pH-shifted cDsbD (cDsbDpH7) were collected at 100 K with a ˚ . The dataset was processed wavelength of 0.856 A with DENZO and SCALEPACK to a resolution of ˚ and was refined to an R-factor of 0.113 and 0.99 A Rfree of 0.146. All cDsbD crystals belong to space group P212121 and contain one protein molecule per asymmetric unit (Table 1). The geometry and stereochemistry of all four cDsbD structures were checked with WHAT_CHECK85 and PROCHECK.86 The two structures refined in full anisotropic mode (cDsbDpH7 and cDsbDpr) were also validated with the Protein Anisotropic Refinement Validation and Analysis Tool (PARVATI87). All structural Figures were prepared using PyMOL†. Analysis of the radiation-driven disulfide opening The opening of the active site disulfide bridge was analyzed by partitioning the cDsbDpr-dataset into eight reflection files. In the first reflection file, 150 diffraction images were included. An increment of 50 frames was used for every additional reflection file. Every dataset was then refined independently with CNS,41 and van der Waals interactions between the two sulfur atoms of Cys461 and Cys464 were ignored. For all models, sulfur–sulfur distances were measured using the program O.82 Calculation of isomorphous FoKFo difference maps An initial and a final segment were chosen for each cDsbD-dataset and rescaled using SCALEPACK, resulting in two reflection files (initial and final) per dataset. The frames ranges were 1–110 and 90–180 for cDsbDox, 1–150 and 180–340 for cDsbDred, 1–180 and 460–676 for cDsbDpr, and 1–151 and 210–360 for cDsbDpH7. Each pair of reflection files was merged and scaled in CNS.41 Model phases were calculated from each final refined cDsbD † http://pymol.sourceforge.net/

model. Using model phases, an isomorphous Fo(initial)K Fo(final) Fourier difference map between the initial and final part of each dataset was computed with CNS.

Stability measurements GdmCl-induced unfolding and refolding of oxidized and reduced forms of cDsbD were performed at 25 8C in 20 mM Hepes–NaOH (pH 7.0), 170 mM NaCl, 0.1 mM EDTA. In the case of reduced cDsbD, 0.5 mM DTT was included. For unfolding, 2.5 mM protein was incubated for 15 h with various concentrations of GdmCl. For refolding, 125 mM protein was first denatured with 6 M GdmCl for 3 h at 25 8C, then diluted 1:50 (v/v) into refolding buffer with various concentrations of GdmCl and incubated for 15 h. The transitions were followed by measuring the fluorescence at 345 nm (excitation at 280 nm). Data were evaluated according to the two-state model with a six-parameter fit.88

HPLC measurements HPLC analysis was performed on an Agilent 1100 HPLC instrument equipped with a diode array detection system and an Agilent Zorbax 300 SB C18 reverse-phase column (5-Micron, 2.1 mm!150 mm). Different species of cDsbD were separated at 55 8C with a linear gradient from 35% to 55% (v/v) acetonitrile in 0.1% (v/v) trifluoroacetic acid, and detected by measuring absorbance at 220 nm and at 280 nm. The amounts of reduced, oxidized or IAMmodified forms of cDsbD were quantified by integration of the peak areas.

Determination of the pKa of the active site Cys461 All experiments were performed at 25 8C in reaction buffer (200 mM KCl, 10 mM di-sodium hydrogen phosphate, 10 mM boric acid, 10 mM succinic acid, 0.1 mM EDTA), adjusted with HCl or NaOH to pH 4–12. Reduced cDsbD was prepared by reduction with a 1000-fold molar excess of DTT at pH 8.0 for 30 min at room temperature, followed by removal of DTT by gel-filtration in degassed buffer. The ionization of Cys461 was measured by the pHdependent reactivity with IAM.44,65,89 Reactions were performed under pseudo first-order conditions with initial concentration of cDsbDred of 5 mM, and concentrations of IAM between 0.1 mM and 10 mM. Samples were removed after incubation for different lengths of time. The reaction was stopped by addition of 0.4 vol. 30% (v/v) formic acid (final pH !2), and HPLC separation of reaction products was performed as described above. Data were evaluated according to pseudo first-order kinetics. The apparent second-order rate constants (kIAM) were calculated and plotted against pH. The pKa of Cys461 in cDsbD or cDsbD-Cys464 was alternatively determined by measurement of the pHdependent thiolate-specific absorbance at 240 nm.44–46 The initial concentration of protein was 30 mM. The sample absorbance was measured against air and was corrected for dilution caused by pH adjustment and for the absorbance of a protein-free reference solution, titrated in the same manner. The data were evaluated according to the Henderson–Hasselbach equation as described.45

842 HPLC analysis of the reactivity with DTT Measurements were performed under pseudo firstorder conditions, with an initial concentration of oxidized cDsbD of 5 mM and 1 mM DTT at 25 8C in 100 mM sodium phosphate, 0.1 mM EDTA (pH 7.0). Samples were removed after incubation for different lengths of time, and the reaction was stopped by addition of 0.4 vol. 30% (v/v) formic acid (final pH !2). HPLC analyses were performed as described above. Data were evaluated according to pseudo first-order kinetics and the apparent second-order rate constants were calculated. pKa calculations and MD simulations Protein residue pKa calculations were performed on the crystal structures as well as on snapshots of molecular dynamics (MD) trajectories (see below) to capture effects on the pKa values from the inherent flexibility of the protein. The question of a possible stabilization of the thiolate state of Cys464 was addressed by carrying out an MD simulation with the reduced cysteine residue in position 464. The pKa calculations were based on the scheme proposed by Nielsen et al.,49,90 which combines finite difference solutions to the Poisson–Boltzmann equation (FDPB) with a global optimization of the hydrogen bond network in all protonation states. As for the MD simulations, all radii and charges of the atoms were consistently taken from the OPLS forcefield.91 The linearized PB equation was solved applying DELPHI II,92,93 with the parameters used by Nielsen & McCammon;49 80 for the dielectric constant of the water, 8 for the protein interior, a 65 cubed grid, a grid resolution of 3 grid ˚ for the desolvation energy and of 4 grid points/ points/A ˚ ˚ ionA for the background interaction energy, a 2.0 A exclusion layer, an ionic strength of 0.144 M, and a surface ˚. probe radius of 1.4 A For the MD simulations, cDsbDred (pH 4.6) was taken as a starting structure. The structure was first minimized (20 steps steepest descent), solvated with TIP494 water molecules at approximately 150 mM NaCl (plus additional ions to neutralize the total system) and minimized again (20 steps steepest descent). The periodic box was chosen such that the distance between the periodic images of the protein is about 3 nm. After minimization, the system was simulated for 100 ps with harmonic position restraints on all heavy atoms (force constant: 1000 kJ/mol per nm 2) in order to allow relaxation of the solvent molecules. The OPLS all-atom forcefield was applied.95 Two simulations were carried out, one with both cysteine residues protonated (system A) and one with a deprotonated charged Cys464 (system B). After deprotonating Cys464, system B was simulated with position restraints on all heavy atoms for another 100 ps. Both systems were finally equilibrated for 5 ns at a temperature of 310 K. The systems comprise more than 49,000 atoms. The MD simulations and part of the analysis were carried out using the GROMACS simulation suite (version 3.3).96 Application of Lincs (constraining the bond length to hydrogen atoms)97 and SETTLE98 allowed for an integration step size of 2 fs. Short-range electrostatic interactions between charged atoms were calculated explicitly, and long-range electrostatic interactions were calculated using the particle-mesh Ewald method.99 Lennard-Jones interactions were cut at a distance of 1 nm, a long-range correction for the energy and the pressure

Structural Studies of cDsbD

was applied. The systems were coupled to a temperature bath separately for the protein and the solvent (tTZ 0.1 ps) and to an isotropic pressure bath (tpZ1 ps). The residue pKa values were calculated by first averaging the individual site titration curves over snapshots of the trajectory,100separated each by 100 ps, and subsequent fitting of the averaged titration curve (over the last nanosecond) by a fit according to the Henderson– Hasselbalch equation. Protein Data Bank entry codes The atomic coordinates and structure factors have been deposited with the RCSB Protein Data Bank, with entry codes 2FWE (cDsbD ox ), 2FWF (cDsbD red ), 2FWG (cDsbDpr), and 2FWH (cDsbDpH7).

Acknowledgements This project was funded by the Schweizerische Nationalfonds, the ETH Zurich and the University of Zurich within the framework of the NCCR Structural Biology program. Data collection for this work was performed at the Swiss Light Source, Paul Scherrer Institut, Villigen, Switzerland. We thank the staff of beamline X06SA for excellent support in X-ray data collection. Ch.U.S. and G.C. are grateful to Beat Blattman for help in crystal screening, to Christophe Briand for help with synchrotron data collection and to Kai Diederichs for providing a program to correct the spindleshutter error in one of the data sets. Ch.U.S. thanks Daniel Frey and Heinz Gut for helpful discussions. R.B. thanks Jens Erik Nielsen for providing the pKa package and for help with the package.

Supplementary Data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/ j.jmb.2006.02.030

References 1. Collet, J. F. & Bardwell, J. C. (2002). Oxidative protein folding in bacteria. Mol. Microbiol. 44, 1–8. 2. Ritz, D. & Beckwith, J. (2001). Roles of thiol-redox pathways in bacteria. Annu. Rev. Microbiol. 55, 21–48. 3. Darby, N. J. & Creighton, T. E. (1995). Catalytic mechanism of Dsba and its comparison with that of protein disulfide-isomerase. Biochemistry, 34, 3576–3587. 4. Wunderlich, M., Otto, A., Seckler, R. & Glockshuber, R. (1993). Bacterial protein disulfide isomerase: efficient catalysis of oxidative protein folding at acidic pH. Biochemistry, 32, 12251–12256. 5. Zapun, A. & Creighton, T. E. (1994). Effects of Dsba on the disulfide folding of bovine pancreatic trypsininhibitor and alpha-lactalbumin. Biochemistry, 33, 5202–5211.

Structural Studies of cDsbD

6. Bader, M. W., Xie, T., Yu, C. A. & Bardwell, J. C. (2000). Disulfide bonds are generated by quinone reduction. J. Biol. Chem. 275, 26082–26088. 7. Bardwell, J. C., Lee, J. O., Jander, G., Martin, N., Belin, D. & Beckwith, J. (1993). A pathway for disulfide bond formation in vivo. Proc. Natl Acad. Sci. USA, 90, 1038–1042. 8. Dailey, F. E. & Berg, H. C. (1993). Mutants in disulfide bond formation that disrupt flagellar assembly in Escherichia coli. Proc. Natl Acad. Sci. USA, 90, 1043–1047. 9. Grauschopf, U., Fritz, A. & Glockshuber, R. (2003). Mechanism of the electron transfer catalyst DsbB from Escherichia coli. EMBO J. 22, 3503–3513. 10. Maskos, K., Huber-Wunderlich, M. & Glockshuber, R. (2003). DsbA and DsbC-catalyzed oxidative folding of proteins with complex disulfide bridge patterns in vitro and in vivo. J. Mol. Biol. 325, 495–513. 11. McCarthy, A. A., Haebel, P. W., Torronen, A., Rybin, V., Baker, E. N. & Metcalf, P. (2000). Crystal structure of the protein disulfide bond isomerase, DsbC, from Escherichia coli. Nature Struct. Biol. 7, 196–199. 12. Missiakas, D., Georgopoulos, C. & Raina, S. (1994). The Escherichia coli dsbC (xprA) gene encodes a periplasmic protein involved in disulfide bond formation. EMBO J. 13, 2013–2020. 13. Zapun, A., Missiakas, D., Raina, S. & Creighton, T. E. (1995). Structural and functional-characterization of Dsbc, a protein involved in disulfide bond formation in Escherichia coli. Biochemistry, 34, 5075–5089. 14. Andersen, C. L., Matthey-Dupraz, A., Missiakas, D. & Raina, S. (1997). A new Escherichia coli gene, dsbG, encodes a periplasmic protein involved in disulphide bond formation, required for recycling DsbA/ DsbB and DsbC redox proteins. Mol. Microbiol. 26, 121–132. 15. Bessette, P. H., Cotto, J. J., Gilbert, H. F. & Georgiou, G. (1999). In vivo and in vitro function of the Escherichia coli periplasmic cysteine oxidoreductase DsbG. J. Biol. Chem. 274, 7784–7792. 16. van Straaten, M., Missiakas, D., Raina, S. & Darby, N. J. (1998). The functional properties of DsbG, a thiol–disulfide oxidoreductase from the periplasm of Escherichia coli. FEBS Letters, 428, 255–258. 17. Fabianek, R. A., Hennecke, H. & Thony-Meyer, L. (1998). The active-site cysteines of the periplasmic thioredoxin-like protein CcmG of Escherichia coli are important but not essential for cytochrome c maturation in vivo. J. Bacteriol. 180, 1947–1950. 18. Goulding, C. W., Sawaya, M. R., Parseghian, A., Lim, V., Eisenberg, D. & Missiakas, D. (2002). Thiol– disulfide exchange in an immunoglobulin-like fold: structure of the N-terminal domain of DsbD. Biochemistry, 41, 6920–6927. 19. Haebel, P. W., Goldstone, D., Katzen, F., Beckwith, J. & Metcalf, P. (2002). The disulfide bond isomerase DsbC is activated by an immunoglobulin-fold thiol oxidoreductase: crystal structure of the DsbC– DsbDa complex. EMBO J. 21, 4774–4784. 20. Gordon, E. H., Page, M. D., Willis, A. C. & Ferguson, S. J. (2000). Escherichia coli DipZ: anatomy of a transmembrane protein disulphide reductase in which three pairs of cysteine residues, one in each of three domains, contribute differentially to function. Mol. Microbiol. 35, 1360–1374. 21. Kim, J. H., Kim, S. J., Jeong, D. G., Son, J. H. & Ryu, S. E. (2003). Crystal structure of DsbDg reveals the mechanism of redox potential shift and substrate specificity. FEBS Letters, 543, 164–169.

843 22. Stewart, E. J., Katzen, F. & Beckwith, J. (1999). Six conserved cysteines of the membrane protein DsbD are required for the transfer of electrons from the cytoplasm to the periplasm of Escherichia coli. EMBO J. 18, 5963–5971. 23. Chung, J., Chen, T. & Missiakas, D. (2000). Transfer of electrons across the cytoplasmic membrane by DsbD, a membrane protein involved in thiol– disulphide exchange and protein folding in the bacterial periplasm. Mol. Microbiol. 35, 1099–1109. 24. Krupp, R., Chan, C. & Missiakas, D. (2001). DsbDcatalyzed transport of electrons across the membrane of Escherichia coli. J. Biol. Chem. 276, 3696–3701. 25. Rietsch, A., Bessette, P., Georgiou, G. & Beckwith, J. (1997). Reduction of the periplasmic disulfide bond isomerase, DsbC, occurs by passage of electrons from cytoplasmic thioredoxin. J. Bacteriol. 179, 6602–6608. 26. Katzen, F. & Beckwith, J. (2000). Transmembrane electron transfer by the membrane protein DsbD occurs via a disulfide bond cascade. Cell, 103, 769–779. 27. Rozhkova, A., Stirnimann, C. U., Frei, P., Grauschopf, U., Brunisholz, R., Grutter, M. G. et al. (2004). Structural basis and kinetics of inter- and intramolecular disulfide exchange in the redox catalyst DsbD. EMBO J. 23, 1709–1719. 28. Collet, J. F., Riemer, J., Bader, M. W. & Bardwell, J. C. (2002). Reconstitution of a disulfide isomerization system. J. Biol. Chem. 277, 26886–26892. 29. Huber-Wunderlich, M. & Glockshuber, R. (1998). A single dipeptide sequence modulates the redox properties of a whole enzyme family. Fold. Des. 3, 161–171. 30. Schlichting, I., Berendzen, J., Chu, K., Stock, A. M., Maves, S. A., Benson, D. E. et al. (2000). The catalytic pathway of cytochrome P450cam at atomic resolution. Science, 287, 1615–1622. 31. Cruickshank, D. W. J. (1999). Remarks about protein structure precision. Acta Crystallog. sect. D, 55, 583–601. 32. Vaguine, A. A., Richelle, J. & Wodak, S. J. (1999). SFCHECK: a unified set of procedures for evaluating the quality of macromolecular structure-factor data and their agreement with the atomic model. Acta Crystallog. sect. D, 55, 191–205. 33. Stirnimann, C. U., Rozhkova, A., Grauschopf, U., Grutter, M. G., Glockshuber, R. & Capitani, G. (2005). Structural basis and kinetics of DsbD-dependent cytochrome c maturation. Structure, 13, 953–961. 34. Burmeister, W. P. (2000). Structural changes in a cryocooled protein crystal owing to radiation damage. Acta Crystallog. sect. D, 56, 328–341. 35. Ravelli, R. B. & McSweeney, S. M. (2000). The “fingerprint” that X-rays can leave on structures. Struct. Fold. Des. 15, 315–328. 36. Schrøder Leiros, H. K., McSweeney, S. M. & Smala˚s, A. O. (2001). Atomic resolution structures of trypsin provide insight into structural radiation damage. Acta Crystallog. sect. D, 57, 488–497. 37. Weik, M., Berges, J., Raves, M. L., Gros, P., McSweeney, S., Silman, I. et al. (2002). Evidence for the formation of disulfide radicals in protein crystals upon X-ray irradiation. J. Synchr. Radiat. 9, 342–346. 38. Armstrong, D. A. (1990). In Sulfur-centered Reactive Intermediates in Chemistry and Biology (Chatgilialoglu, C. & Asmus, K.-D., eds), NATO ASI Series. Series A, Life Sciences, Plenum Press, New York.

844 39. von Sonntag, C. (1990). In Sulfur-centered Reactive Intermediates in Chemistry and Biology (Chatgilialoglu, C. & Asmus, K.-D., eds), NATO ASI Series. Series A, Life Sciences, Plenum Press, New York. 40. Favaudon, V., Tourbez, H., Houee-Levin, C. & Lhoste, J. M. (1990). CO2.- radical induced cleavage of disulfide bonds in proteins. A gamma-ray and pulse radiolysis mechanistic investigation. Biochemistry, 29, 10978–10989. 41. Bru¨nger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P., Grosse-Kunstleve, R. W. et al. (1998). Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallog. sect. D, 54, 905–921. 42. Weik, M., Ravelli, R. B., Kryger, G., McSweeney, S., Raves, M. L., Harel, M. et al. (2000). Specific chemical and structural damage to proteins produced by synchrotron radiation. Proc. Natl Acad. Sci. USA, 97, 623–628. 43. O’Neill, P., Stevens, D. L. & Garman, E. F. (2002). Physical and chemical considerations of damage induced in protein crystals by synchrotron radiation: a radiation chemical perspective. J. Synchr. Radiat. 9, 329–332. 44. Nelson, J. W. & Creighton, T. E. (1994). Reactivity and ionization of the active site cysteine residues of DsbA, a protein required for disulfide bond formation in vivo. Biochemistry, 33, 5974–5983. 45. Grauschopf, U., Winther, J. R., Korber, P., Zander, T., Dallinger, P. & Bardwell, J. C. (1995). Why is DsbA such an oxidizing disulfide catalyst? Cell, 83, 947–955. 46. Kortemme, T. & Creighton, T. E. (1995). Ionisation of cysteine residues at the termini of model alphahelical peptides. Relevance to unusual thiol pKa values in proteins of the thioredoxin family. J. Mol. Biol. 253, 799–812. 47. Wunderlich, M. & Glockshuber, R. (1993). In vivo control of redox potential during protein folding catalyzed by bacterial protein disulfide-isomerase (DsbA). J. Biol. Chem. 268, 24547–24550. 48. Rothwarf, D. M. & Scheraga, H. A. (1992). Equilibrium and kinetic constants for the thiol–disulfide interchange reaction between glutathione and dithiothreitol. Proc. Natl Acad. Sci. USA, 89, 7944–7948. 49. Nielsen, J. E. & McCammon, J. A. (2003). On the evaluation and optimization of protein X-ray structures for pKa calculations. Protein Sci. 12, 313–326. 50. Dillet, V., Dyson, H. J. & Bashford, D. (1998). Calculations of electrostatic interactions and pKas in the active site of Escherichia coli thioredoxin. Biochemistry, 37, 10298–10306. 51. Guddat, L. W., Bardwell, J. C. & Martin, J. L. (1998). Crystal structures of reduced and oxidized DsbA: investigation of domain motion and thiolate stabilization. Structure, 6, 757–767. 52. McDonald, I. K. & Thornton, J. M. (1994). Satisfying hydrogen bonding potential in proteins. J. Mol. Biol. 238, 777–793. 53. Guddat, L. W., Bardwell, J. C., Glockshuber, R., Huber-Wunderlich, M., Zander, T. & Martin, J. L. (1997). Structural analysis of three His32 mutants of DsbA: support for an electrostatic role of His32 in DsbA stability. Protein Sci. 6, 1893–1900. 54. Schirra, H. J., Renner, C., Czisch, M., HuberWunderlich, M., Holak, T. A. & Glockshuber, R. (1998). Structure of reduced DsbA from Escherichia coli in solution. Biochemistry, 37, 6263–6276.

Structural Studies of cDsbD

55. Holm, L. & Sander, C. (1996). Mapping the protein universe. Science, 273, 595–603. 56. Katti, S. K., LeMaster, D. M. & Eklund, H. (1990). Crystal structure of thioredoxin from Escherichia coli ˚ resolution. J. Mol. Biol. 212, 167–184. at 1.68 A 57. Jeng, M.-F. & Dyson, H. J. (1994). High-resolution solution structures of oxidized and reduced Escherichia coli thioredoxin. Structure, 2, 853–868. 58. Kelley, L. A. & Sutcliffe, M. J. (1997). OLDERADO: on-line database of ensemble representatives and domains. Protein Sci. 6, 2628–2630. 59. Weichsel, A., Gasdaska, J. R., Powis, G. & Montfort, W. R. (1996). Crystal structures of reduced, oxidized, and mutated human thioredoxins: evidence for a regulatory homodimer. Structure, 4, 735–751. 60. Boeckmann, B., Bairoch, A., Apweiler, R., Blatter, M. C., Estreicher, A., Gasteiger, E. et al. (2003). The SWISS-PROT protein knowledgebase and its supplement TrEMBL in 2003. Nucl. Acids Res. 31, 365–370. 61. Thompson, J. D., Higgins, D. G. & Gibson, T. J. (1994). CLUSTAL W.: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, positions-specific gap penalties and weight matrix choice. Nucl. Acids Res. 22, 4673–4680. 62. Gonnet, G. H., Cohen, M. A. & Benner, S. A. (1992). Exhaustive matching of the entire protein sequence database. Science, 256, 1443–1445. 63. Kimball, R. A., Martin, L. & Saier, M. H., Jr (2003). Reversing transmembrane electron flow: the DsbD and DsbB protein families. J. Mol. Microbiol. Biotechnol. 5, 133–149. 64. Krause, G., Lundstrom, J., Barea, J. L., Pueyo de la Cuesta, C. & Holmgren, A. (1991). Mimicking the active site of protein disulfide-isomerase by substitution of proline 34 in Escherichia coli thioredoxin. J. Biol. Chem. 266, 9494–9500. 65. Kallis, G. & Holmgren, A. (1980). Differential reactivity of the functional sulfhydryl groups of cysteine-32 and cysteine-35 present in the reduced form of thioredoxin from Escherichia coli. J. Biol. Chem. 255, 10261–10265. 66. Mossner, E., Huber-Wunderlich, M. & Glockshuber, R. (1998). Characterization of Escherichia coli thioredoxin variants mimicking the active-sites of other thiol/ disulfide oxidoreductases. Protein Sci. 7, 1233–1244. 67. Chivers, P. T., Prehoda, K. E., Volkman, B. F., Kim, B. M., Markley, J. L. & Raines, R. T. (1997). Microscopic pKa values of Escherichia coli thioredoxin. Biochemistry, 36, 14985–14991. 68. Li, Q., Hu, H. Y., Wang, W. Q. & Xu, G. J. (2001). Structural and redox properties of the leaderless DsbE (CcmG) protein: both active-site cysteines of the reduced form are involved in its function in the Escherichia coli periplasm. Biol. Chem. 382, 1679–1686. 69. Reckenfelderbaumer, N. & Krauth-Siegel, R. L. (2002). Catalytic properties, thiol pK value, and redox potential of Trypanosoma brucei tryparedoxin. Biol. Chem. 277, 17548–17555. 70. Creighton, T. E. (1992). Proteins: Structures and Molecular Properties. 2nd edit., W. H. Freeman, New York. 71. Voet, D., Voet, J. G. & Pratt, C. W. (2006). Fundamentals of Biochemistry: Life at the Molecular Level. 2nd edit., Wiley, New York. 72. Wunderlich, M. & Glockshuber, R. (1993). Redox properties of protein disulfide isomerase (DsbA) from Escherichia coli. Protein Sci. 2, 717–726.

Structural Studies of cDsbD

73. Zapun, A., Bardwell, J. C. & Creighton, T. E. (1993). The reactive and destabilizing disulfide bond of DsbA, a protein required for protein disulfide bond formation in vivo. Biochemistry, 32, 5083–5092. 74. Sun, X. X. & Wang, C. C. (2000). The N-terminal sequence (residues 1–65) is essential for dimerization, activities, and peptide binding of Escherichia coli DsbC. J. Biol. Chem. 275, 22743–22749. 75. Roberts, B. R., Wood, Z. A., Jonsson, T. J., Poole, L. B. & Karplus, P. A. (2005). Oxidized and synchrotron cleaved structures of the disulfide redox center in the N-terminal domain of Salmonella typhimurium AhpF. Protein Sci. 14, 2414–2420. 76. Reynolds, C. M. & Poole, L. B. (2000). Attachment of the N-terminal domain of Salmonella typhimurium AhpF to Escherichia coli thioredoxin reductase confers AhpC reductase activity but does not affect thioredoxin reductase activity. Biochemistry, 39, 8859–8869. 77. Mossner, E., Iwai, H. & Glockshuber, R. (2000). Influence of the pK(a) value of the buried, active-site cysteine on the redox properties of thioredoxin-like oxidoreductases. FEBS Letters, 477, 21–26. 78. Katzen, F. & Beckwith, J. (2003). Role and location of the unusual redox-active cysteines in the hydrophobic domain of the transmembrane electron transporter DsbD. Proc. Natl Acad. Sci. USA, 100, 10471–10476. 79. Kabsch, W. (1993). Automatic processing of rotation diffraction data from crystals of initially unknown symmetry and cell constants. J. Appl. Crystallog. 26, 795–800. 80. Navaza, J. (1994). AMoRe: an automated package for molecular replacement. Acta Crystallog. sect. A, 50, 157–163. 81. Lamzin, V. S. & Wilson, K. S. (1993). Automated refinement of protein models. Acta Crystallog. sect. D, 49, 129–147. 82. Jones, T. A., Zou, J.-Y., Cowan, S. W. & Kjeldgaard, M. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallog. sect. A, 47, 110–119. 83. Sheldrick, G. M. & Schneider, T. R. (1997). SHELXL: high resolution refinement. Methods Enzymol. 277, 319–343. 84. Otwinowski, Z. & Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307–326. 85. Hooft, R. W. W., Vriend, G., Sander, C. & Abola, E. E. (1996). Errors in protein structures. Nature, 381, 272. 86. Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallog. 26, 283–291.

845 87. Merritt, E. A. (1999). Comparing anisotropic displacement parameters in protein structures. Acta Crystallog. sect. D, 55, 1997–2004. 88. Santoro, M. M. & Bolen, D. W. (1988). Unfolding free energy changes determined by the linear extrapolation method. 1. Unfolding of phenylmethanesulfonyl alpha-chymotrypsin using different denaturants. Biochemistry, 27, 8063–8068. 89. Yang, Y. & Wells, W. W. (1991). Identification and characterization of the functional amino acids at the active center of pig liver thioltransferase by site directed mutagenesis. J. Biol. Chem. 19, 12759–12765. 90. Nielsen, J. E. & McCammon, J. A. (2003). Calculating pKa values in enzyme active sites. Protein Sci. 12, 1894–1901. 91. Jorgensen, W. L. & Tirado-Rives, J. (1988). The OPLS potential functions for proteins. Energy minimizations for crystals of cyclic peptides and crambin. J. Am. Chem. Soc. 110, 1657–1666. 92. Rocchia, W., Alexov, E. & Honig, B. (2001). Extending the applicability of the nonlinear Poisson–Boltzmann equation: multiple dielectric constants and multivalent ions. J. Phys. Chem. B, 105, 6507–6514. 93. Rocchia, W., Sridharan, S., Nicholls, A., Alexov, E., Chiabrera, A. & Honig, B. (2002). Rapid grid-based construction of the molecular surface and the use of induced surface charge to calculate reaction field energies: applications to the molecular systems and geometric objects. J. Comput. Chem. 23, 128–137. 94. Jorgensen, W. L., Chandrasekhar, J., Madura, J. D., Impey, R. W. & Klein, M. W. (1983). Comparison of simple potential functions for simulating liquid water. J. Chem. Phys. 79, 926–935. 95. Kaminski, G., Friesner, R. A., Tirado-Rives, J. & Jorgensen, W. L. (2001). Evaluation and reparametrization of the OPLS-AA force field for proteins via comparison with accurate quantum chemical calculations on peptides. J. Phys. Chem. B, 105, 6474–6487. 96. Lindahl, E., Hess, B. & van der Spoel, D. (2001). GROMACS 3.0: a package for molecular simulation and trajectory analysis. J. Mol. Model. 7, 306–317. 97. Hess, B., Bekker, H., Berendsen, H. J. C. & Fraaije, J. G. E. M. (1998). LINCS: a linear constraint solver for molecular simulations. J. Comput. Chem. 18, 1463–1472. 98. Miyamoto, S. & Kollman, P. A. (1992). SETTLE—an analytical version of the shake and rattle algorithm for rigid water models. J. Comput. Chem. 13, 952–962. 99. Darden, T., York, D. & Pedersen, L. (1993). Particle mesh Ewald: an N log(N) method for Ewald sums in large systems. J. Chem. Phys. 98, 10089. 100. van Vlijmen, H. W., Schaefer, M. & Karplus, M. (1998). Improving the accuracy of protein pKa calculations: conformational averaging versus the average structure. Proteins: Struct. Funct. Genet. 33, 145–158.

Edited by P. Wright (Received 9 December 2005; received in revised form 10 February 2006; accepted 10 February 2006) Available online 28 February 2006