Protein production

Codon-optimized K. pneumoniae dsbA (GenBank® accession number ACI08793), lacking the sequence coding for the predicted signal sequence (19 aa), was cloned into a modified pMCSG7 (Midwest Center for Structural Genomics) vector compatible with ligation-independent cloning. This modified vector encoded a leader sequence consisting of an N-terminal His₆-tag followed by a linker containing the tobacco-etch virus protease (TEV) recognition sequence. KpDsbA was expressed in BL21(DE3)pLys cells using autoinduction medium [18] and purified with Talon® resin (Clontech, Australia). The His₆-tag was removed by TEV protease, leaving the engineered KpDsbA with two additional amino acids (S–1 and N0) at the N-terminus. A final size-exclusion chromatography step using a Superdex75 column (GE Healthcare, USA) yielded highly purified KpDsbA, as judged by SDS-PAGE. Oxidized or reduced KpDsbA was prepared using a 25-fold molar excess of copper-(II)-1,10-phenanthroline or DTT, respectively. Oxidizing/reducing agent was then removed and the protein buffer-exchanged into 10 mM HEPES, pH 7.4 in one step using GE-25 Sephadex desalting resin for crystallization and biochemical experiments.

Preparation of E. coli DsbA (CAA56736), S. enterica Typhimurium DsbA (AAB81592) and E. coli DsbC (AAA83074), lacking the periplasmic leader signal were purified as described for KpDsbA. For peptide oxidation experiments, E. coli DsbB (AAC74269) membrane extracts were prepared as described previously [19] and re-suspended in phosphate buffered saline (PBS, 137 mM NaCl, 2.7 mM KCl, Na₂HPO₄ 10 mM and KH₂PO₄, pH 7.4) containing 10 % glycerol.

KpDsbA Complementation of EcDsbA

The ability of KpDsbA to rescue non-motile E. coli dsbA^- null (JCB817) and dsbA^-/dsbB^- double-null (JCB818) strains was assessed in a cell-swarming assay as described previously [16]. The mature KpDsbA coding sequence was cloned into pBAD33 under an arabinose inducible promotor with the EcDsbA periplasmic signal sequence. A wild-type EcDsbA cloned into pBAD33 vector was used as a positive control. Non-motile E. coli dsbA^- deficient (JCB817) or dsbA^- / dsbB^- double-mutant (JCB818) [3] cells (2x10⁶) transformed with a KpDsbA or EcDsbA pBAD33 inducible vector were spotted onto the center of a soft M63 minimal agar plate containing 40 mg/mL of each amino acid (except L-cysteine). Plates were incubated at 37 °C and motility of cells monitored using a Molecular Imager® Gel Doc™ system from BIO-RAD (CA 94547, USA) after 3-7 h. Complementation experiments were repeated as biological triplicates.

KpDsbA Disulfide Reductase Activity

Under mild reducing conditions, DsbA proteins can reduce the intermolecular disulfide bonds formed between insulin chains A and B [3]. The rate of disulfide bond reduction can be spectroscopically followed at OD_650nm by an increase in turbidity resulting from production of the insoluble insulin chain B [20]. Samples were prepared in 1 cm cuvettes containing 10 μM of protein (KpDsbA, EcDsbA or EcDsbC), 0.33 mM DTT and 2 mM EDTA in 100 mM NaH₂PO₄ / Na₂HPO₄ titrated to pH 7.0. Catalysis was initiated by the addition of 0.131 mM insulin (I0516, Sigma-Aldrich, Australia) to the sample mixture. The assay was repeated three times and data were plotted showing standard deviations.

Measurement of KpDsbA Redox Potential

The standard redox potential of KpDsbA was measured using its intrinsic tryptophan fluorescence, as described previously for EcDsbA [6]. Oxidized KpDsbA was incubated for 12 h at 25 °C in degassed 100 mM NaH₂PO₄ / Na₂HPO₄ buffer (pH 7.0, 1 mM EDTA, 298K), containing 1 mM oxidized glutathione (GSSG) and varying concentrations of reduced glutathione (GSH) (0–2 mM). KpDsbA (200 µL) from each redox condition was dispensed into a 96-well plate (TPP AG, Switzerland #92096) and tryptophan fluorescence was measured (excitation at 280 nm, emission set to 332 nm) using a microplate reader (Synergy H1 and Gen5 2.0 software, Biotek, USA). Data were normalized and the redox potential was calculated as described for EcDsbA [6]. In brief, the equilibrium constant K_eq was calculated using the equation: Y = ([GSH]² / [GSSH])/(K_eq + ([GSH]² / [GSSH])), where Y is the fraction of reduced protein at equilibrium. The redox potential for KpDsbA was calculated from the Nernst equation: E^0’_KpDsbA = E^0’_GSH/GSSH - (RT/nF)lnK_eq where E^0’_GSH/GSSH = - 240 mV, R is the ideal gas constant 8.314 JK^-1mol^-1, T is the absolute temperature in K, n is the number of electrons transferred (n = 2), F is the Faraday constant 9.648x104 Cmol^-1 and K_eq is the equilibrium constant derived from the binding equation. All measurements were performed as biological triplicates. The graph shows a plot of the average values including error bars representing the standard deviation for the replicates.

KpDsbA Thiolate Anion pK_a Determination

The pH-dependent absorbance of the catalytic thiolate anion of KpDsbA was followed at 240 nm [21] using a CARY 50 UV/VIS spectrophotometer (Agilent Technologies, USA). The pH titration measurements of oxidized or reduced KpDsbA (40 μM) in 2 mL composite buffer (10 mM Tris, 10 mM sodium citrate, 10 mM K₂HPO₄, 10 mM KH₂PO₄, 200 mM KCl, and 1 mM EDTA) were conducted at 22 °C. Absorbance (λ = 240 and 280 nm) was measured between pH 6.5 and 2.0 in 0.25 increments. The pK_a value was calculated from the fitted curves of three replicates using the Henderson-Hasselbalch equation (pH = pK_a - log ([A240 ⁄A280]red ⁄ [A240 ⁄A280]oxid)). Experiments were repeated at least three times. Plotted data represent average values and error bars represent the standard deviations across the replicates.

Relative Stability of Oxidized and Reduced Forms of DsbA Enzymes

Temperature-induced unfolding of native SeDsbA and KpDsbA was determined as described previously [13] using a Jasco J-810 circular dichroism (CD) spectropolarimeter (Jasco, USA). The redox state of the protein was confirmed using Ellman’s reagent [22]. The largest difference in molar ellipticity for oxidized or reduced enzymes was calculated from initial far-UV CD spectra (from 250 nm to 190 nm) recorded at 25 °C and 95 °C, respectively. The unfolding of oxidized and reduced protein (SeDsbA_ox = 220 nm, SeDsbA_red = 220.5 nm and KpDsbA_ox = 211 nm, KpDsbA_red = 209.5 nm) was monitored at a heat rate of 1 K / min from 298 K to 368 K in a 1 mm quartz cuvette. All measurements were carried out with 10 µM protein in 100 mM NaH₂PO₄ / Na₂HPO₄, 1 mM EDTA at pH 7.0. Samples for measurement of reduced enzyme contained 0.75 mM DTT. Raw data were analyzed in Prism and fitted to a two-state unfolding model as described previously [23]. The standard deviation was measured from three replicates.

KpDsbA Dithiol Oxidation Activity

A peptide (CQQGFDGTQNSCK) with a 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTA) group amide-coupled to the N-terminus, and a methylcoumarin amide-coupled to the ε-amino group of the C-terminal lysine, was purchased from AnaSpec (Fremont, CA). Lyophilized peptide was re-suspended in 100 mM imidazole, pH 6, at a concentration of 2 mM. Europium trifluoromethanesulfonate (Sigma Aldrich, Australia) solution (100 mM) was added to the peptide at a molar ratio of 2:1 and incubated for 5 min at room temperature, to allow europium chelation. The peptide solution was then immediately aliquoted, flash frozen in liquid nitrogen and stored at -80°C. An increase in fluorescence occurs upon oxidation of the peptide cysteines to form a disulfide. Thus, fluorescence can be used to monitor the capacity of DsbA enzymes to catalyse dithiol oxidation.

Assays were conducted using a Synergy H1 multimode plate reader (BioTek, USA) with the excitation wavelength set to 340 nm and emission to 615 nm. A 150 μs delay before reading and 100 μs reading time were used for time-resolved fluorescence. The assay was performed in a white 384-well plate (Perkin Elmer OptiPlate-384, Part #: 6007290). The buffer consisted of 50 mM MES, 50 mM NaCl and 2 mM EDTA at pH 5.5. The reaction consisted of a 50 μL solution in each well, containing 160 nM EcDsbA, KpDsbA or SeDsbA, 1.6 μM EcDsbB (crude membrane extracts, containing ubiquinone) and 8 μM peptide substrate added last to initiate the reaction. Samples containing buffer and DsbA or buffer and peptide were used as controls. Data were measured for three replicates and are presented as mean values, with the standard error of the mean indicated by error bars.

KpDsbA Crystallization and Crystal Structure Determination

After initial screening using the UQ ROCX facilities, crystals of reduced KpDsbA were grown at 20 °C in VDXm 24-well plates (Hampton Research) using the hanging-drop vapor diffusion method. Screening plates were imaged and incubated in a RockImager 1000 (Formulatrix, MA, USA). Drops contained 0.5 μL of 180 mg/mL reduced KpDsbA and 0.5 μL of crystallization solution (0.1 M succinic acid pH 5.3, 25 % (w/v) polyethylene glycol 1500 and 15 % (v/v) 2-methyl-2,4-pentanediol). For diffraction data measurement, crystals were frozen in liquid nitrogen without additional cryo-protectant. Diffraction data were measured at the Australian Synchrotron micro-focus MX2 beamline using BlueIce software [24]. Reflections were processed in Mosflm [25] and XDS [26], analyzed and converted to MTZ in Pointless [27] and scaled in SCALA [27]. Phases were obtained by molecular replacement (MR) using PHASER [28] with EcDsbA as template (PDB code: 1DSB) . The initial model was improved by iterative model building in COOT [29] and refinement in PHENIX [30]. However, the progress of refinement was stalled with a high R-factor/Rfree of 25.7 % / 29.3 %. Diffraction data analysis in Phenix.xtriage indicated that the crystal was merohedrally twinned with a twinning fraction of 0.42. Further refinement cycles were performed using the twin target function as implemented in PHENIX with the twinning operator h,-h-k,-l. Two fold non-crystallographic symmetry (NCS) is present (which does not align with space group axes), though NCS was not used at any stage of refinement. The refinement finally converged after several TLS refinement cycles. No atoms were modeled into additionally spherical density located between chain D (L133) and chain B (T57) because it was not obvious what was bound. The stereochemical quality of the final model was assessed using MolProbity [31]. A summary of the data processing and refinement statistics are provided in Table 2.

Molecular figures were generated in PyMOL (The PyMOL Molecular Graphics System, Version 1.5.0.4 Schrödinger, LLC) and figures of the electrostatic potential were generated using APBS [32]. The surface, including the proportion of carbon atoms lining the hydrophobic groove in KpDsbA, was calculated using the CastP server [33], by averaging over all six molecules within the asymmetric unit. RMSD calculations and structural alignments were conducted using PyMOL as well as FATCAT [34].

NMR Structure Determination of Oxidized KpDsbA

A sample of uniformly ¹³C,¹⁵N labeled oxidized KpDsbA (1.3 mM) was prepared in 50 mM MES (pH 6.5, 10% ²H₂O and 90 % ¹H₂O). NMR experiments were conducted at 303 K on either 600 MHz or 800 MHz spectrometers equipped with cryogenically cooled probes. All spectra were acquired with standard pulse sequences and processed using TOPSPIN3.1 (Bruker BioSpin). H^N, N, C^α, C^α-1, C^β, C^β-1 peak lists were generated manually in CARA using 2D [¹⁵N,¹H]-HSQC, 3D HNCA, 3D CBCA(CO)NH and 3D HNCACB spectra and used as the input for automated backbone assignments using UNIO-MATCH. These assignments were refined manually and extended using 3D ¹⁵N-resolved [¹H,¹H]-NOESY. H^β, H^α assignments were obtained using a 3D HBHA(CBCACO)NH spectrum. H^N, N, C^α and C^β assignments together with H^β, H^α were provided as input for UNIO-ATNOS/ASCAN for automated side-chain assignments using 3D ¹⁵N-, ¹³C_ali - and ¹³C_aro - resolved [¹H,¹H] NOESY datasets [35,36]. Upper limits for distance restraints used in structure calculations were automatically generated from NOESY datasets using UNIO-ATNOS/CANDID and the structure of oxidized KpDsbA was determined using the torsion angle dynamics program CYANA3.0 [37]. Conformers with lowest CYANA target function values were energy minimized using OPALp and validated using structure validation tools (http:/www.pdb.org/ and http:/www.nihserver.mbi.ucla.edu/). Structures were inspected and analyzed with MOLMOL [38]. Table 3 summarizes the NMR statistics.

Binding Affinity of DsbA-Interacting Peptides

Crystal structures of the EcDsbA:EcDsbB complex revealed that the P2 loop region of EcDsbB interacts with EcDsbA [39,40]. Two peptides derived from the P2 loop sequences of EcDsbB and KpDsbB (Ec – PSPFATCD and Kp – PSPFQTCD) were synthesized by solid-phase methods using Fmoc deprotection on rink-amide MBHA resin (leading to C-terminal amidation) and capped by N-terminal acetylation. Amidation and acetylation ensure that there are no charges on the peptide termini, as these are not present in the native DsbB loop sequence. Binding affinity was measured using a MicroCal™ Auto-iTC₂₀₀ from (GE Healthcare, USA) at 25 °C. The sample cell was loaded with 200 μL of 100 μM KpDsbA or EcDsbA in 25 mM HEPES, 50 mM NaCl, pH 7.4, and DMSO 0.8 %. The peptide (3 mM) diluted in the same buffer was titrated with an initial injection of 0.5 μL into DsbA, followed by 19 consecutive injections (2.0 μL) offset by 180 s, while the solution was constantly stirred (1000 rpm). Data were fitted to a single-site binding model using MicroCal™ Origin 7.0 software (Origin 7 SR4 v7.0552). Experiments were conducted in triplicate and affinity and thermodynamic parameters are reported as means and standard deviations (Table 4).

Comparative Sequence and Structural Analyses

The sequence conservation of ten virulence factors previously identified [2] as substrates of DsbA were analyzed here. Sequences from published and validated DsbA substrate virulence factors were taken from the original literature and used to search the publicly available UniProt database [41] for potential homologues in E. coli, S. enterica Typhimurium and K. pneumoniae. Most of the 10 factors were originally identified in those three organisms except YscC and Caf1M, which were initially reported in Yersinia pestis. A protein-protein BLAST search was performed using the UniProt bacterial genome database with a threshold of P < 0.0001. Unless stated otherwise, homologues were identified in pathogenic strains, i.e. E. coli UPEC O6:K15:H31 and EPEC O127:H6 / O55:H7, S. enterica Typhimurium SL1344 and non-motile K. pneumonieae (hvKP1 / MGH 78578 / NTUH-K2044). Sequence identity between homologues was extracted from the UniProt protein BLAST results. All other sequence alignments reported herein (e.g. for Table 1) were conducted using ClustalW2 [42].

Binding Residues of EcDsbA are conserved in SeDsbA and DsbAs of Five Other Enterobacteriaciae

EcDsbA and SeDsbA share 86 % sequence identity and both have been characterized previously [14,43]. SeDsbA can complement EcDsbA [44] in a null mutant motility assay, indicating that SeDsbA is able to interact with the EcDsbA binding partner EcDsbB and with the EcDsbA substrate E. coli FlgI [45]. Both are weak disulfide reductants in the standard insulin reduction assay for redox enzymes [43]. Both are similarly oxidizing enzymes: the redox potentials of EcDsbA and SeDsbA are -122 and -126 mV, respectively [9,43], whereas the range for all DsbAs is -80 to -163 mV (Table 1). In both EcDsbA and SeDsbA the measured pK_a of the nucleophilic cysteine is 3.3 [7,43], though values vary across all DsbAs from 3.0 to 5.1 (Table 1). Although disulfide bonds generally stabilize folded proteins, the disulfide form of DsbA enzymes is destabilizing [6,7]. The melting temperatures of the oxidized and reduced forms of EcDsbA and SeDsbA are almost identical (reduced 350 K and 351 K; oxidized 341 K and 342 K, respectively) [13] (Figure S1), whereas the range of melting temperatures across all DsbAs varies considerably (Table 1). Importantly, the crystal structures of EcDsbA and SeDsbA can be superimposed with an RMSD of 0.8 Å for 176 Cα atoms, whereas across all structurally characterized DsbAs the RMSD with EcDsbA varies from 1.3 Å to 2.9 Å (for 122-167 Cα atoms) (Table 1) [14].

Two catalytically relevant EcDsbA complex structures have been described, a complex between EcDsbA and EcDsbB [39,40,46] and one between EcDsbA and a peptide segment of SigA, an autotransporter protein from Shigella flexneri [47]. Analysis of these structures revealed that the binding interface comprises the N-terminal regions of the active site helix H1, as well as loops L1 (the first of two loops connecting the thioredoxin and helical domains), L2 (the second of two loops connecting the thioredoxin and helical domains, also referred to as the cisPro loop) and L3-H7 (residues in the loop preceding and at the N-terminal region of helix H7) (Figure 1A). A hypothesis is that DsbAs sharing overall high sequence identity with EcDsbA and with highly conserved loop lengths and residues in these regions will share similar binding activities. As shown in Figure 1B, SeDsbA falls into this cluster as does Shigella flexneri (SfDsbA, P52235), Enterobacter cloacae (EnDsbA. E3G5L9), Citrobacter koseri (CkDsbA, A8AL80) and Cronobacter sakazakii (CsDsbA, I2ED40) and K. pneumoniae (KpDsbA) (Figure 1B). Of these, the DsbA with lowest sequence identity to EcDsbA is KpDsbA (81 %) encoded by an important human pathogen responsible for many antibiotic-resistant nosocomial infections [1,48,49]. To determine whether KpDsbA falls within the same class as EcDsbA, we investigated its structure, surface, redox and binding properties and compared them with EcDsbA.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Comparison of EcDsbA and SeDsbA.

A. Structural superposition of EcDsbA (magenta, PDB id: 1FVK) and SeDsbA (blue, PDB Id: 3L9S). N- and C-termini, helices (H1 - 7) and strands (β1-5) are indicated. In addition, surface loops (L1 – L3) predicted to be involved in binding EcDsbB periplasmic loop P2 or substrate are labeled in red. Active site cysteines are shown as orange spheres and the cisPro motif in the L2 loop is indicated by a yellow star. B. Sequences of EcDsbA loops that bind DsbB (blue/red) or SigA substrate (blue). Homologues with highly conserved loop sequences are shown: S. flexneri (SfDsbA, P52235), S. enterica Typhimurium (SeDsbA E1WE53), C. koseri (CkDsbA, A8AL80), E. cloacae (EnDsbA, E3G5L9), C. sakazakii (CsDsbA, I2ED40) and K. pneumoniae (KpDsbA B5XZJ6). Conserved residues are shown in grey, and variable residues in black.

https://doi.org/10.1371/journal.pone.0080210.g001

KpDsbA Complements EcDsbA in vivo

The E. coli protein FlgI is required for E. coli motility and, in turn, FlgI requires the DSB machinery of E. coli to function. FlgI function is impaired in E. coli dsbA^- deficient (JCB817) and dsbA^-/dsbB^- double-mutant (JCB818) strains due to the absence of EcDsbA mediated dithiol oxidase activity [50]. As a consequence, these E. coli strains are non-motile. Intriguingly, K. pneumoniae is non-motile and does not encode a FlgI homologue. We tested whether KpDsbA was able to catalyse disulfide bond formation of E. coli FlgI using an in vivo complementation strategy [3]. We demonstrated that KpDsbA – like SeDsbA [44] – can fully restore the motility of dsbA^- deficient strains, but not in the double dsbA^-/dsbB^- mutant cells (Figure S2). This experiment shows that KpDsbA is able to oxidize FlgI cysteines and this requires the presence of EcDsbB.

Some distantly related DsbAs do not complement EcDsbA in this assay, including Gram-negative Wolbachia pipientis α-DsbA1 [23] and Gram-positive Staphylococcus aureus DsbA [13]. However, rescue or partial rescue of motility has been observed for a wide range of DsbA homologues, some sharing quite low sequence identity with EcDsbA, such as VcDsbA (40 %), PaDsbA (30 %) and NmDsbA1 (23 %) [15-17,51]. Consequently, EcDsbA complementation may not be a suitable guide for categorizing DsbA enzymes into distinct classes.

KpDsbA has redox properties almost identical to those of EcDsbA and SeDsbA

EcDsbA exhibits weak insulin reductase activity in the presence of dithiothreitol [52] whereas the E. coli disulfide isomerase EcDsbC is highly active in this assay. Reduction of the intermolecular disulfide bonds between the A and B chains of insulin results in precipitation of the B chain and this can be monitored by measuring the OD_650nm. We found that purified recombinant KpDsbA has the same weak insulin reductase activity as EcDsbA (Figure 2A) and SeDsbA [43]. The activity of other characterized DsbA enzymes varies. NmDsbA1, for example, has a much weaker activity than that of EcDsbA [15], and DsbA from Mycobacterium tuberculosis (MtbDsbA) is inactive in this assay [53]. In contrast, TcpG (VcDsbA) from Vibrio cholerae catalyses insulin reduction much faster than EcDsbA [54].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. KpDsbA redox properties.

A. Disulfide bond reduction activity of KpDsbA (▲), EcDsbA (■) EcDsbC (●) and a control without enzyme (△) was monitored spectrophotometrically. SeDsbA activity has been published elsewhere [43]. B. Redox equilibria of KpDsbA with glutathione (GSH/GSSG). C. Determination of the nucleophilic Cys33 (CXXC) pK_a. The pH-dependent absorbance of the thiolate anion at 240 nm was fitted to the Henderson-Hasselbach equation D. Temperature induced unfolding of oxidized (ox, ■) and reduced (red, □) KpDsbA was determined by far-UV CD spectroscopy, showing that the reduced form is more stable than the oxidized form.

https://doi.org/10.1371/journal.pone.0080210.g002

We next determined the standard redox potential of KpDsbA relative to glutathione ([GSH]²/GSSG, E^0’ = -240 V). The equilibrium constant for KpDsbA was estimated from the [GSH]²/GSSG titration experiment to be 61.4 ± 0.1 µM (Figure 2B), which corresponds to a standard redox potential of -116 mV. This value falls very close to the values reported for EcDsbA (-122 mV [9]) and SeDsbA (-126 mV [43]) considering the wide range of values reported across all DsbA enzymes (-80 to -163 mV) [14].

The pK_a value of the nucleophilic cysteine in the active site CXXC motif is a key determinant of DsbA reactivity towards substrate proteins. We measured the pK_a value for the nucleophilic cysteine of KpDsbA using pH-dependent thiolate absorbance at λ = 240 nm (Figure 2C). The pK_a^Cys30 for KpDsbA was found to be 3.2, nearly identical to that of EcDsbA and SeDsbA (3.3) compared with the observed range for other DsbAs (3.0-5.1).

We also confirmed that reduced KpDsbA (T_m^red 347.1 ± 0.2 K) is more stable than oxidized KpDsbA (T_m^ox 335.8 ± 0.3 K) (Figure 2D). The melting temperatures fall between values reported previously for EcDsbA (T_m^red 350.9 ± 0.2 K, T_m^ox 341.7 ± 0.2 K [7]) and those for SeDsbA (T_m^red 351.2 ± 0.2 K, T_m^ox 342.8 ± 0.4 K) reported here (Figure S1). Again, the range reported for all DsbAs is much wider (T_m^red 337-357 / T_m^ox 331-341 K) [51,54].

We then tested the dithiol oxidase activity of KpDsbA using a fluorescently labeled peptide substrate. The activity was monitored by the increase in europium fluorescence resulting from cyclization of the substrate peptide through formation of an intramolecular disulfide bond. In the presence of EcDsbB, we found that the rate for KpDsbA and SeDsbA catalyzed disulfide bond formation was almost indistinguishable from that of EcDsbA measured at the same concentration of enzyme (Figure 3). This result suggests that KpDsbA (and SeDsbA) is able to interact in the same way as EcDsbA with the peptide substrate and with EcDsbB. TcpG has a similar activity to EcDsbA in this assay [54], whereas MtbDsbA is inactive in the presence of EcDsbB [53].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. In vitro peptide dithiol oxidation.

A. Dithiol oxidase activities of EcDsbA (■), SeDsbA (●) and KpDsbA (➉) were monitored using a fluorescently labeled peptide substrate. Samples lacking the partner protein EcDsbB (KpDsbA/peptide △), EcDsbA/peptide □, SeDsbA/peptide ○, or buffer alone ▼) showed no increase in signal over the same time period.

https://doi.org/10.1371/journal.pone.0080210.g003

Crystal structure of reduced KpDsbA

We determined the crystal structure of reduced KpDsbA (PDB: 4MCU) at 1.99 Å resolution by molecular replacement, using EcDsbA as the template. As expected, the structure is very similar to that of EcDsbA (Figure 4A). The asymmetric unit contains six KpDsbA molecules each adopting the typical DsbA fold. Structural superposition of these six independent copies yielded a root mean square deviation (RMSD) < 0.45 Å for 176 Cα atoms between residues Gly6 - Val181. Likewise, structural alignment of KpDsbA with EcDsbA (1FVK, 1.7 Å, molecule B) and SeDsbA (3L9S, 1.6 Å) gave RMSD values < 0.9 Å for the identical range of 176 Cα atoms. By comparison, high resolution crystal structures of distantly related DsbAs have much higher RMSDs covering a smaller range of equivalent Cα atoms (e.g. PaDsbA (PDB code 3H93) and EcDsbA (1FVK, molecule B), 161 Cα atoms RMSD of 2.4 Å) [16]. These higher values are a consequence of structural deviations including a truncated helix H7 and a shortened hydrophobic groove.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Crystal structure of KpDsbA.

A. Superposition of crystal structures of KpDsbA (cyan, PDB Id: 4MCU) and EcDsbA (magenta, PDB id: 1FVK). The N- and C-termini, helices (H1 - 7) and strands (β1-5) are indicated. Surface loops L1 – L3 are labeled in red, and active site cysteines are shown as orange spheres. B. Electron density in the active site region of KpDsbA indicates that the cysteines are reduced. The 2Fo - Fc map was created using Phenix (model-map correlations) [30] and is contoured at 1.0 σ C. Electrostatic surface representation of EcDsbA, SeDsbA and KpDsbA (left, middle, right). Positive and negative electrostatic potentials are contoured from blue (+7.5 kT/e) to red (-7.5 kT/e). The hydrophobic grooves of all three enzymes are indicated by a dashed oval [8,43].

https://doi.org/10.1371/journal.pone.0080210.g004

The structure of the catalytic site of KpDsbA is strictly conserved with that of EcDsbA, comprising the active site motif ³⁰Cys-Pro-His-Cys³³ located at the N-terminal end of helix H1 and the adjacent cisPro (Val-Pro¹⁵¹) L2 loop (Figure 4B). The cysteine residues (Cys30 and Cys33) are present in the reduced state in the crystal structure. A hydrophobic patch and a large groove surrounds the nucleophilic Cys30, as also occurs in EcDsbA and SeDsbA (Figure 4C). As expected, these surface features are lined with residues contributed from the L1, L2 and L3 loops.

The six independent copies of KpDsbA in the crystal structure allow an analysis of conformational variability of the loop residues forming the binding surface. This revealed that the side chains of His32, Phe63, Leu64, Gln147, Thr167 and Met170 adopt various rotamer conformations, whereas there is no evidence of conformational variability in Tyr29, Cys30, Pro31, Val149, Pro150, and Phe173 (Figure 5A). The side chain variations do not influence the surface accessibility of the hydrophobic groove, which was calculated to be 371 ± 32 Ų by CastP [33] across the 6 molecules. Moreover, the hydrophobic nature of the groove is unaffected by the side chain conformational variability as indicated by the proportion of carbon atoms lining this groove (69 ± 3 %) [33].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Conformational variability in X-ray and NMR structures of KpDsbA.

A Superimposition of the six KpDsbA molecules (blue) in the asymmetric unit shows the limited conformational variability in the side chains of active site and L1, L2, and L3 loop residues (stick representation). B. Cartoon representation of the KpDsbA crystal structure (Molecule D), with C_α atoms colored by temperature factor (B-factor). Molecule D was selected as its temperature factor distribution is the most pronounced due to minimal crystallographic contacts. In particular, the high B-factor of loop L3 indicates mobility in that region, consistent with the NMR data C. Stereo diagram of representative states of reduced (X-ray, cyan) and oxidized (NMR, yellow) structures of KpDsbA. Red arrows highlight differences in the structures at N-terminal and L3 loop regions.

https://doi.org/10.1371/journal.pone.0080210.g005

NMR Solution Structure of KpDsbA is Similar to the Crystal Structure

Previous studies have demonstrated that there are minimal differences between reported structures (crystal and NMR) of oxidized and reduced EcDsbA. To determine if this was also the case for KpDsbA, a semi-automated NMR approach was used to determine the structure of oxidized KpDsbA (PDB ID: 2MBS, BMRB ID: 19413). Following UNIO-ATNOS/ASCAN, manual verification and refinement enabled assignment of 89.2 % of the non-labile proton resonances in KpDsbA. These were used to generate the NOE-based distance constraints for final structure calculation. Twenty conformers with lowest target function and least violations of input restraints were chosen to represent the structure of oxidized KpDsbA (Figure S4 A/B). It was not possible to assign several backbone amide resonances corresponding to residues in the β1- β2 loop (Ile16, Gly18, Glu19, Gln21, Val22, Leu23), so that this region appears to be largely disordered in the NMR ensemble compared with the rest of the structure. The backbone (N, Cα, C’) and all-heavy atom RMSD for the 179 well-defined residues (1–15, 24-187) of the 20 KpDsbA conformers were 0.67 ± 0.17 Å and 1.03 ± 0.13 Å, respectively. Structural statistics are summarized in Table 3. As observed for other DsbA structures, the individual thioredoxin and helical domains can be superimposed with higher precision than the entire structure. This is most likely due to inter-domain motion, which has also been reported in the structures of EcDsbA [55] and VcDsbA [56]. Residues which fall into disallowed Ramachandran regions include the unassigned residues Glu19, Gln21, Val22, and His32, and residues in loop regions, i.e. Lys55, Phe63, Leu64, Asn155 and Met170.

The overall conformation of the NMR structure of oxidized KpDsbA is similar to that of the crystal structure of reduced KpDsbA (Figure 5C). For example, superposition of molecule A in the crystal structure of reduced KpDsbA with the first structure in the NMR ensemble of oxidized KpDsbA, yields an RMSD of 1.09 Å over 169 Cα atoms. To make a similar comparison, the crystal structures of oxidized (1FVK, molecule B) and reduced (1A2L, molecule B) EcDsbA have an RMSD of 0.45 Å (over 186 Cα atoms) and the crystal structure of oxidized EcDsbA (1FVK, molecule B) and the first structure in the NMR ensemble of reduced EcDsbA (1A24) have an RMSD of 1.95 Å over 181 Cα atoms [57,58].

The structures of the catalytic sites and hydrophobic surface features are similar, considering that the cysteines of the CXXC motif are oxidized in the NMR structure and reduced in the crystal structure (Figure S4C). As has been noted previously for other DsbA solution and crystal structures [56,59], L3 of KpDsbA is a relatively flexible part of the protein in both NMR and crystal structures (Figure 5B and C). Thus, overall the structures of oxidized and reduced KpDsbA are similar, notwithstanding the different conditions and approaches used for structure determination.

Binding Affinity of DsbB peptides is similar for KpDsbA and EcDsbA

The similar surface features and similar predicted binding residues of KpDsbA and EcDsbA suggested that these enzymes would interact with binding partners with similar affinity. The crystal structures of the EcDsbA:EcDsbB complex showed that the second periplasmic loop P2 of EcDsbB binds directly to EcDsbA [39,40]. The binding residues are 98-PSPFATCD-104 and these are highly conserved in KpDsbB (98-PSPFQTCD-104). These two P2 peptides were synthesized and isothermal titration calorimetry (ITC) was used to assess their binding affinity for both enzymes. KpDsbA and EcDsbA were found to bind to PSPFATCD and PSPFQTCD with similar affinities (K_d 11-18 µM, Table 4, Figure S3A). We investigated the interaction of KpDsbA with PSPFQTCD by structural superposition of KpDsbA onto the structure of EcDsbA in the EcDsbA:EcDsbB complex structure. Residue Ala of EcDsbB PSPFATCD was mutated in silico to PSPFQTCD, using the most commonly observed rotamer for glutamine. The superimposed model showed that the P2 loop matched the surface of KpDsbA very well, with no clashes apparent between the P2 residues and KpDsbA (Figure S3B).