doi:10.1016/j.jmb.2007.05.067 J. Mol. Biol. (2007) 371, 703–716
Probing the Flexibility of the DsbA Oxidoreductasefrom Vibrio cholerae—a 15N - 1H HeteronuclearNMR Relaxation Analysis of Oxidized andReduced Forms of DsbA
James Horne1, Edward J. d’Auvergne2, Murray Coles3, Tony Velkov1
Yanni Chin1, William N. Charman4, Richard Prankerd4
Paul R. Gooley2 and Martin J. Scanlon1⁎
1Department of MedicinalChemistry, Victorian College ofPharmacy, Monash University,381 Royal Parade, Parkville,VIC 3052, Australia2Department of Biochemistryand Molecular Biology,University of Melbourne,Melbourne, VIC 3010, Australia3Department of ProteinEvolution, Max-Planck-Institute for DevelopmentalBiology, Spemannstr. 35,72076 Tuebingen, Germany4Department of Pharmaceutics,Victorian College of Pharmacy,Monash University, 381 RoyalParade, Parkville, VIC 3052,Australia
Abbreviations used: NOE, nucleaenhancement; HSQC, heteronuclearcoherence; DSC, differential scanninE-mail address of the correspondi
We have determined the structure of the reduced form of the DsbAoxidoreductase from Vibrio cholerae. The reduced structure shows a highlevel of similarity to the crystal structure of the oxidized form and is typicalof this class of enzyme containing a thioredoxin domain with an inserted α-helical domain. Proteolytic and thermal stability measurements show thatthe reduced form of DsbA is considerably more stable than the oxidizedform. NMR relaxation data have been collected and analyzed using amodel-free approach to probe the dynamics of the reduced and oxidizedstates of DsbA. Akaike's information criteria have been applied both in theselection of the model-free models and the diffusion tensors that describethe global motions of each redox form. Analysis of the dynamics reveals thatthe oxidized protein shows increased disorder on the pico- to nanosecondand micro- to millisecond timescale. Many significant changes in dynamicsare located either close to the active site or at the insertion points betweenthe domains. In addition, analysis of the diffusion data shows there is a cleardifference in the degree of interdomain movement between oxidized andreduced DsbA with the oxidized form being the more rigid. Principalcomponents analysis has been employed to indicate possible concertedmovements in the DsbA structure, which suggests that the modeledinterdomain motions affect the catalytic cleft of the enzyme. Taken together,these data provide compelling evidence of a role for dynamics in thecatalytic cycle of DsbA.
The Dsb (disulfide bond-forming) family ofproteins are oxidoreductase enzymes found withinthe periplasm of Gram-negative bacteria. TheDsbA/B system is primarily responsible for the for-mation of new disulfide bonds in substrates within
r Overhausersingle quantumg calorimetry.ng author:
lsevier Ltd. All rights reserve
the periplasm. DsbA catalyzes the oxidation of awide range of substrate proteins via an efficientthiol-disulfide transfer mechanism. Reduced DsbA,which is formed in the reaction, is re-oxidized by acognate, membrane-bound partner, DsbB to com-plete the catalytic cycle. For substrates that containmore than one pair of cysteine residues, disulfidesmay be linked incorrectly, hence a second, comple-mentary system exists to catalyze disulfide isomer-isation; DsbC and its membrane-bound reductivepartner DsbD.1
The prototypical DsbA enzyme is comprised oftwo domains (Figure 1) including a largely α-helicaldomain that inserts into a thioredoxin-like domain
Figure 1. Cartoon of the crystal structure 1BED ofoxidized VcDsbA. Secondary structure elements arenumbered sequentially from the N terminus. Blacknumbers denote helices and grey numbers denote strands.The side-chain atoms of C30 and C33 are shown in CPKrepresentation.
704 NMR Relaxation Analysis of DsbA
at the end of a long helix (residues 143–148 inEscherichia coli DsbA (EcDsbA);2 residue numberingconforms to the E. coli protein sequence unlessotherwise noted) and a loop between strand 3 of thethioredoxin domain and helix 2 in the helicaldomain (residues 58–62). The relative orientationof the two domains can vary through simplerotations around the insertion points. By inferencefrom thioredoxin-substrate crystal structures,3 bind-ing of peptide substrate is predicted at a groovebetween helix 1 and helix 7 within the thioredoxindomain. The relative size of this groove is affectedby variations in the interdomain angle. Among thevarious DsbA enzyme structures solved usingcrystallographic and NMR methods there is con-siderable variation in the observed interdomainangle.4–6 However, for the EcDsbA there is asmuch variation in the interdomain angle withindifferent crystal forms of the oxidized protein asthere is between oxidized and reduced forms, andthe significance of the changes in domain orientationis not clear.The active site of DsbA lies in a cleft at the
interface of the two domains and comprises a highly
conserved primary sequence motif C-P-X-C at theN-terminal end of helix 1. In oxidized DsbA, adisulfide bond links C30 and C33. In the reducedform of EcDsbA a sulfhydryl at C33 and a thiolateanion at C30 remain after release of the oxidizedproduct post catalysis. Recent observations with amutant C33A form of EcDsbA,7 indicate thepossibility of cis-trans proline isomerization withinthe C-P-X-C motif common to many DsbAs. How-ever, mutation of the proline in this motif of theEcDsbA does not diminish the activity of theenzyme,8 suggesting that this isomerism is not arequirement for catalysis.Proteolysis experiments with EcDsbA indicated
marked differences for the reduced and oxidizedforms of DsbA in relative susceptibility to proteasedigestion.9 Oxidized DsbAwas found to be cleavedmore readily than reduced DsbA. This was inter-preted as an indication of possible higher degree offlexibility in the oxidized form, which may beimportant for accommodating substrate interac-tions, the more stable reduced form perhaps drivingthe release of the oxidized product.Thus, there are several lines of evidence suggest-
ing that there may be dynamic differences betweenthe oxidized and reduced forms of DsbA. However,EcDsbA is the only protein for which the structure ofthe reduced form of the protein has been deter-mined. As such, it is unclear if the changes indomain orientation are a general feature of this classof enzymes or if this observation is unique toEcDsbA. Furthermore, to date there has been nodirect measurement of protein dynamics to comple-ment observations that have been inferred from theavailable static structures of DsbA. Here, we presentthe structure of the reduced form of Vibrio choleraeDsbA (VcDsbA) as well as proteolytic data, thermalstability data and measurements of backboneheteronuclear 15N T1, T2 and {1H}-15N steady-stateNOEs for both reduced and oxidized forms ofVcDsbA. Model-free analyses of the NMR relaxationdata demonstrate clear differences in the dynamicproperties of the two oxidation states, both of localmotions and global movements of the thioredoxinand helical domains. We show that interdomainmotions for reduced and oxidized VcDsbA, whilesharing broadly similar mechanics, are clearlydifferent in amplitude and that local motions, bothon the pico- to nanosecond and micro- to milli-second timescale, are also different. We present aprincipal components analysis of potential modes ofinterdomain motions to supplement our discussionof the importance of these dynamic processes in thecatalytic cycle of DsbA enzymes.
Results
Stability of VcDsbA redox forms
Figure 2 shows a comparison of HPLC traces foroxidized and reduced VcDsbA after incubation for
Figure 2. Proteolytic and thermal stability data forreduced and oxidized VcDsbA. (a) The HPLC traces offragments generated by proteolytic cleavage of reduced(top) and oxidized VcDsbA. T18 represents the C-terminalresidues 165–181 and DsbA-T18 the residual. The y-axis isthe percentage of total ion count value during massspectrometric characterization. (b) The primary aminoacid sequence of VcDsbAwith cleavage sites indicated bythe dark arrows. Fragment numbering is from the Nterminus. (c) Excess specific heat capacity (Cp, kcal/Kmol) curves for oxidized (1) and reduced (2) VcDsbA. Thecalorimetric thermal transition values, transition tempera-ture mid-point (tm), enthalpy (ΔHcal) and entropy (ΔScal)are documented in the inset.
705NMR Relaxation Analysis of DsbA
24 h with trypsin (Figure 2(a)). Both redox formscleaved readily at the C-terminal site (K165). How-ever, the residual portion of the reduced protein was
much more resistant to further proteolysis over aperiod of 24 h as evidenced by the size of the peakdenoted DsbA-T18 in the HPLC trace. The HPLCpeak labels match the fragment sequences indicatedon the primary sequence for VcDsbA shown inFigure 2(b). Although the same fragments wereidentified in both samples, a greater proportion ofthe oxidized protein was fully proteolyzed. Thus,reduced VcDsbA is more resistant than oxidized toproteolytic degradation by trypsin.The thermodynamic stability of the reduced and
oxidized forms of VcDsbA was examined by dif-ferential scanning calorimetry. The thermal transi-tion curves obtained are shown in Figure 2(c) and thecorresponding thermodynamic parameters aredetailed in the Figure inset. The parameters obtainedfor the thermal transition of both redox forms areconsistent with values reported for EcDsbA.10,11
Reduction of VcDsbA increased the unfolding mid-point temperature, with an increase in both ΔHcalandΔScal. Thus, oxidation of the protein destabilizesVcDsbA resulting in a lowering of the meltingtemperature.
NMR data analysis
Localized changes in the 15N-1H heteronuclearsingle quantum coherence (HSQC) data wereobserved upon reduction of VcDsbA. Chemicalshift changes for residues in and around the catalyticsequence of the thioredoxin were marked, notablyfor C30 and C33 (Figure 3), a consequence of thereduction of the disulfide bond. Resonances adja-cent to this region also experienced significantchemical shift changes by a margin greater thanthe linewidth of the peaks. Reduced VcDsbAshowed no evidence of reoxidation over the periodof data acquisition. It has been shown that reducedEcDsbA is also resistant to aerial oxidation forperiods of greater than two months in bufferssimilar to those used in the current experiments.12
In addition to the chemical shift changes, therewere intensity variations within individual spectraof each form of the protein. Variations in peakintensity in the 1H-15N-HSQC spectra were notlimited to the peaks of residues adjacent to thecatalytic site. Residues 127–135 (helix 6) werereduced in intensity in both redox forms whencompared to the majority of other amide resonances.In addition, N64 in the loop before helix 2 and F127at the start of helix 6 were not observable above thespectral baseplane noise in either redox state andrelaxation data could not be recorded for these.
Structural analysis
The final calculated ensemble of structures forreducedVcDsbA is shown in Figure 4. The ensemblecontains 15 (from a total of 50 calculated) structures,which have the least violations of the experimentaldata and lowest energies. Structural statistics for theensemble are shown in Table 1. The Procheck_nmrRamachandran plot for backbone dihedral quality
Figure 3. The 15N-1H HSQC spectra for oxidized (left) and reduced VcDsbA recorded at 600 MHz and backboneamide chemical shift comparison. Residues undergoing large chemical shift changes are noted. A comparison of thecombined amide nitrogen and proton shifts ð ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðDNshift2 * 0:154Þ þ DHshift2
p) is shown beneath.64
706 NMR Relaxation Analysis of DsbA
indicates 91.6% of residues in most favored regionsand 8.3% in additionally allowed regions, none ingenerously allowed regions and 0.1% in disallowedregions. Residues 1-2, 63-64, 162–166 and 144–148were excluded from the definition of the structuredcore as they were dynamically mobile terminalresidues or unstructured loops. The calculated root-mean square deviation from the mean structure forsuperposition of the remainder is 0.32 Å for back-bone atoms and 0.75 Å for heavy atoms.Overall, the superposition over backbone C′, Cα
and N atoms is very tight with small differences inthe region of the catalytic motif in helix 1 and in helix3, which are maximal around residue H94. Aconsequence of the breakage of the disulfide bondis the movement of the side-chains of C30, H32, F53,M61, F90 and H94 with most other side-chainpositions being relatively conserved, as shown inFigure 5. The side-chains of C30 and M61 swing intoward each other bringing the Hε protons of M61closer to the core. H32 swings across the transverseplane of the hydrophobic pocket toward M61. H94moves outward away from the core and does notresolve to a precise position in the ensemble. Itwould appear likely to be relatively solvent-exposedin these reduced VcDsbA structures as is R97 at theC-terminal end of helix 3.
Model-free analysis
15N, R1, R2 and steady-state NOE values recordedfor the backbone amides of reduced and oxidizedVcDsbA are presented in Figure 6. The R1 values are
consistently lower in the oxidized form. Also shownare S2 and Rex values determined with Modelfree4.Analysis of the raw data demonstrates that differ-ences exist between the two states.Figure 7 contains a cartoon depiction of VcDsbA
in oxidized and reduced forms. A spline is used totrace the Cα atoms of the protein. The width of thespline is related to the degree of disorder asrepresented by the calculated order parameter S2
such that the width increases as S2 decreases fromunity. The narrowest splinewidth represents near-unity values and therefore a high degree of order forthe timescale of motion represented in the Figure.An additional color ramp is applied to highlight thesame information with change from white throughto red as the order parameter value decreases fromunity. S2 values for fast (τeb100 ps) and slow(τeN100 ps) motions, as determined from thecalculated τe value in Modelfree4 output, aremapped onto the structure in separate cartoons. Inthis representation, both redox forms displayincreased disorder of fast motions in the region ofhelix 6 and, to a lesser extent, between strand 3 andhelix 2. The decrease in the order parameter for fastmotions is greater for oxidized VcDsbA. Analysis ofthe slow motions also reveals similar profiles ofdisorder in both redox forms, but some discretedifferences are noted adjacent to the catalytic motif.Y29 is muchmore disordered in the oxidized form ofVcDsbA. However, more striking are the differencesin the micro- to millisecond motions as modeled bythe Rex terms. Reduced VcDsbA has many morechemical exchange terms (Rex) within the thiore-
Table 1. NMR and refinement statistics for 15 VcDsbAmodel structures
A. NMR distance and dihedral constraintsDistance constraintsTotal NOE
Lennard-Jones energy (kcal mol−1)a −862.21±21.72r.m.s.d. from the mean structure
Backbone (Å) 0.32±0.07Heavy (Å) 0.75±0.05
Pairwise r.m.s.d. was calculated among 15 refined structures.a Lennard-Jones energy values from CNS using protein.par
parameters from the Xplor-NIH distribution.
Figure 4. Ensemble of 15 structures calculated withXplor-NIH. (a) and (b) Backbone superposition overstructured regions for whole molecule. (c) and (d)Comparison with 1BED oxidized crystal structure (greycartoon ribbon).
707NMR Relaxation Analysis of DsbA
doxin domain, but the largest Rex terms have beenmodeled in the oxidized form proximal to thecatalytic motif (H57) and at the N-terminal ends ofhelix 2 and helix 6. The cartoon diagram in Figure 7shows Rex values mapped onto the backbonenitrogen atoms of the protein color-ramped fromwhite to red (for zero to 20 s−1).Three models of diffusion were applied within
Modelfree4: isotropic, oblate anisotropic and prolateanisotropic. Diffusion models were fitted to eitherthe whole VcDsbA molecule or to the individualdomains, thus allowing for the possibility ofindependent motion of the domains. Table 2 con-tains the AIC values (see Materials and Methods) forthe nine diffusion models based on the relaxationdata for oxidized and reduced VcDsbA that wereused to identify the best models for this dataanalysis. Fits of diffusion tensors to the individualdomains yield significantly better AIC values thanfor single diffusion tensor models. Table 3 containsthe optimized diffusion parameters for all diffusionmodels. Both redox forms are best fit by the ani-sotropic models; reduced VcDsbA by oblate diffu-sion axes and oxidized VcDsbA by a mixture ofoblate and prolate diffusion axes. The AIC values for
the two axially symmetric anisotropic diffusionmodels in the oxidized form are very similar, indi-cating a fully anisotropic diffusion model might bemore appropriate. In contrast, there is a large dif-ference between the same models fit to reducedVcDsbA and therefore a much clearer model selec-tion. The τm values for the helical and thioredoxindomains of oxidized VcDsbA (13.05 ns and 12.70 ns,respectively) are significantly longer than for re-duced VcDsbA (11.32 ns and 10.73 ns, respectively).
Modeling of VcDsbA motions
Figure 8 displays the first four modes with non-zero eigenvalues from the principal componentsanalysis of reduced VcDsbA (the first six modesinvolve global movement). Each panel in the Figurecontains orthogonal views of porcupine plots (seeMaterials and Methods), which conveniently repre-sent motion information in a static form. The base ofthe cones or spines is set at the Cα position, and theyextend with direction and amplitude described byrelevant components of the first eigenvector fromthe transformation matrix. The amplitude is scaledto allow for coincident presentation of structure andmotions. The first principal component of motion(Figure 8(a)) is predominantly a twisting or shearingmotion, with the helical and thioredoxin domainstwisting in opposite direction. The second compo-nent (Figure 8(b)) displays a more classical hingingmotion with variable extent of the cleft formedbetween both domains. Figure 8(c) and (d) exhibit amixture of twisting and flexing. In all cases the
Figure 5. Comparison of residues in the catalytic site for reduced and oxidized VcDsbA. In this divergent stereo viewrepresentation, the side-chains of residues 25–33, 50, 53, 57, 60, 61, 90, 94, 136, and 140 are shown. C30 and C33 are coloredyellow. The oxidized DsbA (1BED) side-chains are drawn with a thicker line.
708 NMR Relaxation Analysis of DsbA
hinge/twist points are centered in the regionsbetween strand 3 and helix 2, and in helix 6.
Discussion
The structures that are reported here for thereduced form of the VcDsbA are typical of thisoxidoreductase enzyme family. Figure 4 shows thecalculated NMR structure ensemble in 90° orthogo-nal projections and superimposed over the back-bone of the crystal structure of the oxidized from ofVcDsbA (1BED). Some of the most significantdifferences between redox forms are for residuesadjacent to the catalytic site, in particular aroundH94 in helix 3. We note that the observed differencescould be due to a number of phenomena, such asstructural underdetermination or crystal packingeffects, but in this case the density of NOE restraintsin the region of H94 is high. There are ten or moreinter-residue restraints per residue in helix 1, withonly P31 notably lacking in NOE information. Thisis lower than for other regions of the protein butsufficient to give reasonable local structural preci-sion. Other regions of sparse NOE informationinclude the end of strand 3 and the loop beforehelix 2, the short spacer between helix 5 and helix 6,and the loop between strand 5 and helix 7. Theequivalent loop in the EcDsbA NMR ensemble doesnot resolve to any consistent conformation and islikely to be mobile in solution.As is the case for reduced EcDsbA, the NMR
ensemble reveals that the helical and thioredoxindomains superimpose on themselves with slightlygreater precision compared with superposition ofthe entire molecule. There are relatively few NOEsacross the interface of this protein to define preciselythe relationship of the two domains with respect toeach other. Line broadening for residues in theregion of the catalytic helix 1 and in the long helix 6
and the Rex terms that have been modeled in therelaxation analysis provide a clue to a possiblemicro- to millisecond timescale dynamic mechanismunderpinning this observation.The difference in domain displacement between
the oxidized and reduced structures for VcDsbA isvery similar to that observed for the EcDsbA withthe reduced form being slightly more “open” thanthe oxidized crystal structure. Between the oxidizedand reduced VcDsbA structures, there are smalldifferences local to the active site region and thehydrophobic core proximal to the catalytic motif.But otherwise, within the limits of precision ofavailable NMR structure ensembles and the limits ofthe comparable crystal structures, there are very fewsignificant structural differences.The dynamics data demonstrate much more com-
pelling differences between the two redox forms ofVcDsbA on the pico- to millisecond timescale forlocal motions and for global motions. The patterns oforder and disorder in model-free analysis data areapproximately similar, as might be expected whenmuch of the overall structure remains unchanged.However, discrete differences between oxidized andreduced are clearly present. Oxidized VcDsbAshows greater disorder of fast motions in helix 2and helix 6, which link to the thioredoxin domain,and disorder of slow motions adjacent to thecatalytic motif. Reduced VcDsbA shows moreevidence of chemical exchange in the thioredoxindomain but oxidized VcDsbA has larger exchangeterms for residues near the catalytic motif and inhelix 6. In addition there are other importantdifferences in the global motions of the redoxforms, as shown from our analysis of the diffusiontensor data.TheAIC selection of rotational diffusionmodels for
reduced and oxidized VcDsbA showed significantlybetter fit of axially symmetric anisotropic diffusiontensors when applied individually to the helical and
Figure 6. R1 and R2 relaxationrates and steady-state NOE values,generalized order parameters S2 andRex values compared for reduced(right) and oxidized VcDsbA.
709NMR Relaxation Analysis of DsbA
thioredoxin domains. In Figure 9(a) and (b), agraphical representation of these tensors is shownpositioned at the center of mass of each domain. Thediffusion rate Å−1 is equal to 5.55e6 s−1 Å−1. Todistinguish between interdomain motions and theglobal tumbling of the molecule, Figure 9(c) presentsa representation of the selected helical and thior-edoxin domain diffusion tensors of reducedVcDsbA.All of the rotations about the long axis of the fullprotein (shown in red) can be accounted for by theglobal tumbling of the molecule. As the rotationaldiffusion rate about this axis is similar for both
domains it is not possible to discern if there is anyinterdomainmotion. For the orthogonal green axis ofthe schematic, the component of diffusion due to theglobal tumbling is limited by the diffusion rate of thehelical domain. If this rate is subtracted from thesame axis in the thioredoxin domain, then a residualdiffusion rate remains that is due to interdomainmobility. Similarly for the orthogonal blue axis, theglobal diffusion is limited by the thioredoxin domainand therefore there is residual interdomainmotion inthe helical domain. Thus, perpendicular interdomainmotions in both oxidized and reduced VcDsbA are
Figure 7. Comparison of “fast” (τeb100 ps) and“slow” (τeN100 ps) order terms and exchange terms foroxidized (left) and reduced VcDsbA (right). VcDsbA isrepresented by a spline drawn through backbone nitrogenatoms. The width of the spline is scaled by a factor of1.1-S2 for order parameter plots yielding a thin line forvalues near unity. The color ramp from white to red islikewise scaled. For the Rex comparison, the width of thespline is scaled to the normalized range of both sets of Rexvalues (RadiusAtom=2×normalized value in the MolMolmolecule display package) . The color ramp from white tored is likewise scaled. The numbering of secondarystructure elements from Figure 1 has been applied to thetop left panel.
710 NMR Relaxation Analysis of DsbA
clearly evident. Despite the different diffusion tensorselected for the helical domain of the oxidizedprotein, Figure 9 shows that the interdomainmotions of the two redox forms of VcDsbA aresimilar. However, the extent of the movement is lessin the oxidized form, implying that there is restric-tion of interdomain movement upon oxidation. This
restriction is supported statistically by the AIC valuedifferences. The sum of the AIC values of the twoselected diffusion tensors for the helical and thior-edoxin domains of the reduced form differs by 1073from the single diffusion tensor for the entire protein,suggesting that the two domains have significantindependent motion. On the other hand, thisdifference is 45 for the oxidized form,which indicatesa restriction of the interdomainmotions. It is possiblethat the decrease in order of fast motions in helix 6 inthe oxidized form is a compensatory response to thedecrease in interdomain flexibility. Direct correlationof dynamics on the different timescales with proteinstability are not trivial, with fast timescale motiondisorder linked directly to local disturbance of side-chain structure and micro to millisecond timescalechanges possibly indicating a shift in the relativepopulations of alternative conformations. The rela-tive priority of either for the observed stability is notclear.13
The principal components analysis provides avisualization of probable modes of interdomainmotion available to VcDsbA, as inferred in this caseform a CONCORD-derived ensemble of structures.The combination of twisting and flexing movementsfits nicely with a model of two partially independentdomains inwhich the relative degree of movement isaltered upon oxidation. Though we cannot resolvethe twisting motions in this case, it is an entirelyreasonablemodel for potential interdomainmotions.Within the catalytic cycle of DsbA, oxidized DsbA
displays specificity for a diverse range of reducedsubstrates,14 whilst reduced DsbA interacts primar-ily with DsbB.12 The similarity between the struc-tures of the different redox states of VcDsbAprovides no clear indication of the basis for thisdifference in substrate specificity. However, it haspreviously been suggested that dynamics may playa role.5 This observation is supported by ourrelaxation data analysis. Thus, oxidized VcDsbAshows increased disorder around the active site forboth fast and slow motions. This is consistent withthe suggestion that increased flexibility enables theoxidized form of VcDsbA to interact with a broadrange of substrates, whereas the loss of flexibility inreduced VcDsbA contributes to substrate release.5
In contrast, reduced VcDsbA shows greater inter-domain motions, which alter the dimensions of thehydrophobic groove on the surface of the enzyme.This groove has recently been shown to be the siteof interaction between reduced EcDsbA and aperiplasmic loop of DsbB.15 Therefore, it is con-ceivable that the interdomain motions that arepresent in reduced VcDsbA facilitate binding ofDsbB and contribute to the observed specificity ofthe reduced protein. This is analogous to recentobservations on the thiol disulfide oxidoreductaseResA, where there was found to be increasedaccessibility of a surface cavity, which contributedto a difference in substrate specificity in the reducedform of the protein.16
We propose that the hinging/shearing motion ofVcDsbA, a likely model for the entire family, is
The underlined numbers are the lowest AIC values for the three diffusion models fitted for that molecule or part thereof.
711NMR Relaxation Analysis of DsbA
achieved through motion at the N-terminal end ofhelix 6 and at the end of strand 3 as it then connectsto the helical domain via a short loop to helix 2. Thefacility to hinge is clearly present in both forms butour relaxation data suggest that these motions mayoccur at different rates or amplitudes in the differentredox states of the protein. The energetics ofstabilizing the thiolate anion or the disulfide maywell drive the dynamic and biochemical differencesthat are observed between reduced and oxidizedVcDsbA. Formation of the disulfide in the oxidizedVcDsbA results in a thermodynamic destabilizationof the protein, which is accompanied by an increasein some observed motions of the oxidized protein inboth the pico- to nanosecond and micro- to milli-second timescales and an increased rigidity ofinterdomain movements.The structures of oxidized and reduced VcDsbA
are remarkably similar. In contrast, there are cleardifferences in the dynamics and interdomainmotions of the different states of the protein. Thesedifferences in the dynamics may be the moreimportant factor in the accommodation of a diverserange of substrates, product release and recognitionby DsbB in the catalytic cycle of the enzyme.
Materials and Methods
Expression and purification of VcDsbA
The E. coli expression system forVcDsbAwas a gift fromRonald Taylor (Dartmouth University, New Hampshire).Uniformly 15N isotope-labeled protein was producedaccording to the method of Marley et al.17 and purifiedusing a modification of the protocol described by Hu et al.4
Table 3. Diffusion tensor parameters for the models chosen bysingle fit per molecule and individual domain fits
Monte-Carlo simulated error is shown in parentheses below the relev
An overnight streak culture was grown from glycerolstock on LB/agar plates containing 100 μg/ml ofampicillin. An isolated colony was picked and used toinoculate a 100 ml starter culture of LB/ampicillin andgrown overnight at 37 °C. The overnight starter culturewas inoculated into 1 l of LB medium and grown at 37 °Cuntil an absorbance at 600 nm wavelength (A600) of 0.6was measured, with respect to culture without cells. Cellswere pelleted by centrifugation at 5000g for 10 min,washed with M9 minimal salts solution and resuspendedin 0.25 vol. M9 labeling medium with [15N]NH4C1 as thesole source of nitrogen available to the bacteria for singlylabeled protein and [13C]glucose added for doublelabeling. This labeling culture was grown at 37 °C for1 h and then induced with 1 mM IPTG. The culture wasgrown for 4 h post induction, when the cells were pelletedby centrifugation at 5000g for 15 min at 4 °C. ExpressedVcDsbA was released from the harvested cells byperiplasmic fractionation using 20–30 ml of polymyxinB(1 mg/ml) in Tris–HCl (pH 8.0), 150 mM NaCl. Thesolution was incubated at 4 °C for 1 h and cell materialwas pelleted by centrifugation at 15,000g for 30 min. Solid(NH4)2SO4 was added to the supernatant to 1 M finalconcentration and the supernatant was clarified bycentrifugation at 15,000g for 30 min.Two chromatographic steps were used to purify the
labeled protein. First, the soluble fraction was applied toPhenyl HP 16/10 hydrophobic interaction column (Amer-sham Biosciences, Sydney, Australia) equilibrated with10 mM Hepes (pH 6.8), 1 M (NH4)2SO4. Proteins wereeluted with a linear gradient from 100% to 0% ammoniumsulfate at a flow rate of 3 ml/min over three columnvolumes. VcDsbA-containing fractions were pooled, con-centrated and diafiltered into buffer C (10 mM Hepes, pH6.8) using a centrifugal concentrator (Amicon 10 kDacutoff membrane). The resulting solution was applied to aMonoQ HR10/10 cation-exchange column (AmershamBiosciences, Sydney, Australia) and eluted with a gradientfrom 0 to 250 mM NaCl in buffer C over ten columnvolumes. VcDsbA-containing fractions were pooled,
AICmodel selection for reduced and oxidized VcDsbA as
Figure 8. (a)–(d) Porcupine plots for the first fourmodes with non-zero eigenvalues of principal componentanalysis of reduced VcDsbA. Orthogonal views for eachare shown with 90° rotation about the vertical axis. Theblue cones are based at the α carbon position and arescaled on the basis of the size of the eigenvector thatdescribes the motion in each mode. Twisting and hingingof the helical and thioredoxin domains with respect toeach other are clearly indicated in the displayed modes.Images were created with the Dynamite web server usinga low-energy structure from the refined NMR ensemble.
Figure 9. The separate diffusion tensors for the helical andselection and optimized using Modelfree4,22,49 and a represenand (b) reduced VcDsbA are shown in cartoon representatiodiffusion tensor representations, which are positioned at the cfiles created using the program relax,57 and are displayerepresentation of the diffusion tensor data. The schematic is aligthe end of each axis represent a positional probability distributof time. Because the long axis of the helical and thioredoxin doxidized and reduced forms, all of the rotations about the longrotational diffusion. However, for the blue axis, the contributiothe thioredoxin domain about that axis. Hence, a residual intevident. For the green axis, the contribution of global diffusiointerdomain motion of the thioredoxin domain along that per
712 NMR Relaxation Analysis of DsbA
analyzed by SDS-PAGE and electron spray ionizationmass spectrometry. VcDsbAwas reduced using a 100-foldexcess of dithiothreitol (DTT) or oxidized using 1.5 mMcopper phenanthroline. The reduction/oxidation reagentswere removed by passage through a Superdex75 HR10/30 size-exclusion column into buffer D (10 mMHepes (pH6.8), 50 mM NaC1, 1 mM EDTA). Samples for NMRanalysis were concentrated before addition of 10% (v/v)2H2O to give a final concentration of approximately300 μM protein in a volume of 300 μl. The samples weretransferred into Shigemi tubes for NMR data collection,which was performed at 27 °C for the relaxation analysesand at 47 °C for all other experiments. An additionalsample was prepared with 10% (w/v) [13C]glucose addedto the labeling protocol outlined above for stereospecificassignment of methyl groups of valine and leucineresidues.18
Proteolysis of VcDsbA
Samples of reduced and oxidized VcDsbA wereexchanged into 100 mM ammonium bicarbonate (pH 8.0)and concentrated to 1 mg/ml. Sequencing-grade trypsinprotease (Promega,USA) was added at an enzyme toVcDsbAmolar ratio of 1:50 and incubated at 37 °C for 24 h.Aliquots of 50 μl were removed and the proteolyticreaction were halted by acidification. Samples wereanalyzed using mass spectrometry.
thioredoxin domains of VcDsbA as chosen by AIC modeltation of whole versus interdomain diffusion. (a) Oxidizedn looking down the long axis of helix 3. The geometricentre of mass of each domain, were encoded within PDBd using PyMOL [http://pymol.sourceforge.net/]. (c) Aned approximately with the orientation of (a). The stubs ation of a vector parallel with that axis after a given amountomain diffusion tensors are approximately equal in bothred axis of VcDsbA could be accounted for by the globaln of global rotational diffusion is limited by the motion oferdomain motion in the helical domain about that axis isn is limited by the helical domain, hence there is residualpendicular axis.
713NMR Relaxation Analysis of DsbA
Differential scanning calorimetry (DSC) of VcDsbA
DSC measurements of the thermal stability were under-taken on freshly reduced or oxidized forms ofVcDsbA on amodel 6100 NII-DSC (Calorimetry Science Corporation,Spanish Fork, UT). For all DSC scans, VcDsbA wasdiafiltered into the sample buffer (10 mM Hepes (pH 6.8),50 mM NaC1, 1 mM EDTA; containing 1 mM DTT in thereduced sample) and a base-line was determined using thesame buffer. Before being loaded into the calorimetry cell,VcDsbA sampleswere centrifuged at 13,000g for 3min andfiltered to remove any particulate matter. A scan rate of 1deg.C/min was used for base-line and sample scans.Protein concentrations were 0.4–0.5 mg/ml. The originaldatawere converted to listings of excess specific heat. Afterconcentration normalization and base-line correction,analysis and fitting of the thermograms to a two-statetransition model were performed using the model 6100NII-DSC software package (Calorimetry Science Corpora-tion, Spanish Fork, UT), which is based on standarddeconvolution models.19 The thermodynamic stabilityparameters thermal unfolding midpoint temperature(tm), the calorimetric enthalpy (ΔHcal) and entropy(ΔScal) of unfolding were calculated from the transitioncurves.
NMR data acquisition and processing
15N- T1, T2 and {1H }15N steady-state NOE measure-ments were recorded at spectrometer fields of 500 MHz(Bruker DRX) and 600 MHz (Bruker DRX and VarianUnity 600) at a temperature of 300 K. T1 and T2 datasetswere recorded, interleaved and collected with rando-mized order of observation for the time increments tominimize the effects of spectrometer drift during datacollection. All data were recorded with 2048 data points inF1, 200 data points in F2, and 16 scans per data slice (24scans for T2 experiments) with spectral width of 30 ppmfor 15N and 15 ppm for 1H. Sensitivity enhanced 15N T1,T2, and steady-state NOE experiments used Echo/Anti-echo gradient selection with decoupling during theacquisition period. Steady-state NOE data were acquiredwith a saturation period of 5 s. Data were processed usingNMRPipe20 and XWINNMR. (2048×512K matrix dimen-sions; Gaussian apodization in the direct dimension;shifted, squared sine bell apodization in the indirectdimension; polynomial baseline correction in the indirectdimension; proton chemical shifts referenced relative toDSS at 300 K, 15N chemical shifts were referenced using anindirect referencing ratio.21 Visualization and measure-ment of cross-peak intensities in 2D transformed matriceswas performed using the software SPARKY (Goddard andKneller, UCSF). Errors for time-series T1 and T2 data werecalculated from the overall standard deviation forduplicate data points in the series. Errors for the NOEdata were estimated from measurements of the root-mean-square deviation of the base-plane noise in thosespectra.22 Levenberg-Marquardt non-linear least-squaresfitting of two-parameter exponential functions to time-series data was performed using the software CurveFit†.Errors were calculated by Monte-Carlo simulation usingthe covariance matrix method.Triple resonance CBCACONNH,23,24 HNCACB,
were recorded to establish a sequential assignment forVcDsbA through Cα, Cβ and C shift correlation to backboneamides. CCONH-/HCCONH-TOCSY (12 ms; 100×32×1024 complex data points)28–33 and HCCH-TOCSY (mixingtime of 16 ms) experiments (32×64×512 complex datapoints)34 were recorded to facilitate side-chain resonanceidentification in carbon and proton frequencies. 3D 15Nseparated NOESY-HSQC,35–39 and 3D 13C separatedHSQC-NOESY37–40 experiments were recorded (mixingtimes of 100-140 ms) and peak picked using a combinationof automated strip searching onheteronucleus and attachedproton frequencies and manual picking. Additional stereo-specific assignment information for valine and leucine side-chain methyl groups was obtained from observing cou-plings in 13C-HSQC spectra of a 10% 13C-labeled sample.Peak lists with peak height values and resonance as-
signment listswere used as input to the software Cyana-2.1for calculation using automated NOE assignment.41 Thesoftware TALOS was used to determine backbone ϕ/ψangle restraints based on chemical shifts for Cα, Cβ, Hα,HN and C′ nuclei.42 TALOS-derived ϕ/ψ restraints andvaline and leucine stereospecific assignments were addedto the calculation protocol. The input scripts and frequencytolerances were taken from the distributed example scriptsin the Cyana-2.1 academic software release. The final list ofupper limits for interproton distances was exported toXplor-NIH for refinement of the structure.43
Ensembles of structures were calculated using standardsimulated annealing protocols in Xplor-NIH along withmanual inspection of restraint violations, data errors andrecalculation.44 The final ensemble of structures wascalculated from 865 medium-range (|i–j|b4), 999 long-range (|i–j|N3), and 29 ambiguous long-range NOErestraints. An additional 270 dihedral angle restraintswere applied to backbone ϕ/ψ angle restraints and χ1angle restraints for stereospecifically assigned Leu and Valside-chain methyl groups. A total of 19 hydrogen bondrestraints were applied using a patch statement to theXplor parameter libraries as described.44 Fifty structureswere calculated in the final ensemble and filtered toinclude only those with no NOE violation N0.3 Å ordihedral violation N5° and good overall geometry asdetermined by Whatcheck45 and by Procheck-nmr.46,47
Model-free analysis
The principal axes of the diffusion tensor,Dx,Dy andDz,and their orientation relative to the inertia frame werecalculated using a least-squares fit of the spectral densityfunctions of the R1/R2 ratios to the experimental datausing the software TENSOR2.48 The crystal structure forVcDsbA was chosen as an initial template for the ex-perimental data. Diffusion parameters were subsequentlyoptimized during the calculations with Modelfree4.22,49
To avoid undue influence of slowmotions and exchangeprocesses on the tensor calculation, only residues with anNOE value in excess of 0.65 were chosen for fitting of thediffusion tensor.50–52 Within this subset, residues identi-fied to be within elements of secondary structure werechosen so that the diffusion frame was based on a core ofresidues undergoing broadly similar motion and thereforeless likely to be skewed by multiple motional regimes.48,53
Diffusion tensors were calculated for both the 600 MHzand 500 MHz relaxation data to check for consistency. Forisotropic models of motion, a correlation time value wascalculated and used as input for Modelfree4. For axiallysymmetric anisotropic models, D-parallel (D‖=Dz), D-perpendicular (D⊥=Dx=Dy), θ and ϕwere optimized andthen passed as input to Modelfree4.
The model-free formalism presented by Lipari andSzabo describes motions for a protein tumbling in solutionas a combination of global rotational motion and internalmotions on various timescales.54,55 Internal motions aredescribed by the amplitude-dependent square of thegeneralized order parameter S2 and the correlation timefor motions in the pico- to nanosecond timescale τe. Thisbasic model has been extended by Clore and co-workerswith fast motions described by the parameters Sf
2 and τf,and the slower by the parameters SS
2 and τs.56 For models
that incorporated such motions, the generalized orderparameter is of the form:
S2 ¼ S2f S2S
These terms are fit to the observed relaxation data withan additional, variable exchange term Rex fit to thetransverse relaxation rate constant to account for theinfluence of possible internal movements in the micro- tomillisecond timescale.The software Modelfree4.01 has five motional models
that can be fit to the available relaxation data. Thesecomprise the possible parameters S2, S2f, SS
2, τe, τs, and Rexin the following five combinations: {S2}, {S2, τe}, {S
2, Rex},{S2, τe, Rex}, and {Sf
2, SS2, τs} denoted models 1–5
respectively.During the final stages of the parameter fitting, the
diffusion tensors were optimized. and used as input forfurther rounds of calculation and optimization until anarbitrary convergence of tensor values was established.Iterations were suspended when values for the tensor ratioand correlation time did not change by more than 0.01%.Typically, this would involve some five or six rounds ofcalculation for isotropic models of diffusion and from 8–11iterations for the axially symmetric anisotropic models.The software relax57 was used for implementation of
Akaike's information criteria (AIC) for the selectionbetween the model-free models.58,59 Also, relax was usedto manage input and parsing of Modelfree4 parametersand the corresponding output. Calculated diffusiontensors were used as input and appropriate motionalparameter selections for Modelfree4 specified. AIC scoringof each of the five possible motional model-free modelswas then used to assist in model selection for the residue-specific modeling of relaxation data. The selectioncriterion was of the form AIC=χ2+2k, where χ2 is theminimized value and k is the number of parameters in themodel-free model.For VcDsbA, no assumption was made about the most
appropriate diffusional model for either reduced oroxidized form. Tensor2 was used to calculate isotropicand axially symmetric (oblate and prolate) sets ofdiffusion parameters; there was no suitable treatment fora fully anisotropic model of diffusion within Modelfree4.Each of the three sets of diffusion parameters was thenused as input for iterative calculation and refinementwithinModelfree4. A total of 500Monte-Carlo simulationswere performed for the final calculation once convergenceof the optimized diffusion parameters had been reached.Modeling of the proteins was performed with single
diffusion fits for the whole protein and with individual fitsfor the thioredoxin and helical domains separately.Independent motion of the domains may require multiplediffusion tensors to better describe the observed relaxationdata for the system. AIC values were again calculated foreach domain where individual diffusion parameters wereused, counted and compared with AIC values calculatedfor the single-domain fits. The AIC model-selecteddiffusion tensors were then used to generate molecular
surfaces describing the shape and magnitude of the globaldiffusion for each domain for comparison.
Modeling of VcDsbA motions
The web-based software Dynamite was used to in-vestigate and visualize likely modes of movement avail-able to VcDsbA.60–62 While less rigorous than carefullyparameterized molecular dynamics simulations, themethod used in the underlying CONCORD63 componentof Dynamite generates an ensemble of structures withinsensible configuration space for a given protein moleculefrom which an analysis of possible concerted motions canbe made. Two structures were submitted to the Dynamiteserver, the crystal coordinates of oxidized VcDsbA (1BED)and a lowest-energy NMR structure from our ensemble ofrefined structures.
Protein Data Bank accession code
The ensemble of 15 structures has been deposited in theProtein Data Bank identified by the accession code 2IJY.
Acknowledgements
This research was supported under the AustralianResearch Council's Linkage funding scheme (projectnumber LP0455508). The authors thank Mr StuartThomson for mass spectrometric data collection andassistance with data presentation, and Dr JamesSwarbrick for guidance with the calculation andanalysis of structures from NMR data.
References
1. Nakamoto, H. & Bardwell, J. C. A. (2004). Catalysis ofdisulfide bond formation and isomerization in theEscherichia coli periplasm. Biochim. Biophys. Acta, 1694,111–119.
2. Martin, J. L., Bardwell, J. C. A. & Kuriyan, J. (1993).Crystal structure of the DsbA protein required fordisulphide bond formation in vivo. Nature, 365,464–468.
3. Qin, J., Clore, G. M., Kennedy, W. P., Kuszewski, J. &Gronenborn, A. M. (1996). The solution structure ofhuman thioredoxin complexed with its target fromRef-1 reveals peptide chain reversal. Structure, 4,613–620.
4. Hu, S. H., Peek, J. A., Rattigan, E., Taylor, R. K. &Martin, J. L. (1997). Structure of TcpG, the DsbAprotein folding catalyst from Vibrio cholerae. J. Mol.Biol. 268, 137–146.
5. Guddat, L. W., Bardwell, J. C. & Martin, J. L. (1998).Crystal structures of reduced and oxidized DsbA:investigation of domain motion and thiolate stabiliza-tion. Structure, 6, 757–767.
6. Schirra, H. J., Renner, C., Czisch, M., Huber-Wunder-lich, M., Holak, T. A. & Glockshuber, R. (1998).Structure of reduced DsbA from Escherichia coli insolution. Biochemistry, 37, 6263–6276.
7. Ondo-Mbele, E., Vives, C., Kone, A. & Serre, L. (2005).Intriguing conformation changes associated with the
715NMR Relaxation Analysis of DsbA
trans/cis isomerization of a prolyl residue in the activesite of the DsbA C33A mutant. J. Mol. Biol. 347,555–563.
8. Huber-Wunderlich, M. & Glockshuber, R. (1998). Asingle dipeptide sequence modulates the redoxproperties of a whole enzyme family. Fold. Des. 3,161–171.
9. Vinci, F., Couprie, J., Pucci, P., Quemeneur, E. &Moutiez, M. (2002). Description of the topographicalchanges associated to the different stages of the DsbAcatalytic cycle. Protein Sci. 11, 1600–1612.
10. Moutiez, M., Burova, T. V., Haertle, T. & Quemeneur,E. (1999). On the non-respect of the thermodynamiccycle by DsbA variants. Protein Sci. 8, 106–112.
11. Couprie, J., Vinci, F., Dugave, C., Quemeneur, E. &Moutiez, M. (2000). Investigation of the DsbAmechanism through the synthesis and analysis of anirreversible enzyme-ligand complex. Biochemistry, 39,6732–6742.
12. Bader, M., Muse, W., Zander, T. & Bardwell, J. (1998).Reconstitution of a protein disulfide catalytic system.J. Biol. Chem. 273, 10302–10307.
13. Jarymowycz, V. A. & Stone, M. J. (2006). Fast timescale dynamics of protein backbones: NMR relaxationmethods, applications, and functional consequences.Chem. Rev. 106, 1624–1671.
14. Ruddock, L. W., Hirst, T. R. & Freedman, R. B. (1996).pH-dependence of the dithiol-oxidizing activity ofDsbA (a periplasmic protein thiol:disulphide oxidor-eductase) and protein disulphide-isomerase: studieswith a novel simple peptide substrate. Biochem. J. 315,1001–1005.
15. Inaba, K., Murakami, S., Suzuki, M., Nakagawa, A.,Yamashita, E., Okada, K. & Koreaki, I. (2006). Crystalstructure of the DsbB-DsbA complex reveals amechanism of disulfide bond generation. Cell, 127,789–801.
16. Colbert, C. L., Wu, Q., Erbel, P. J. A., Gardner, K. H. &Deisenhofer, J. (2006). Mechanism of substrate-speci-ficity in Bacillus subtilus ResA, a thioredoxin-likeprotein involved in cytochrome c maturation. Proc.Natl Acad. Sci. USA, 103, 4410–4415.
17. Marley, J., Lu, M. & Bracken, C. (2001). A method forefficient isotopic labeling of recombinant proteins.J. Biomol. NMR, 20, 71–75.
18. Neri, D., Szyperski, T., Otting, G., Senn, H. &Wuthrich, K. (1989). Stereospecific nuclear magneticresonance assignments of the methyl groups of valineand leucine in the DNA-binding domain of the 434repressor by biosynthetically directed fractional label-ing. Biochemistry, 28, 7510–7516.
19. Privalov, P. L. & Potekhin, S. A. (1986). Scanningmicrocalorimetry in studying temperature-inducedchanges in proteins. Methods Enzymol. 131, 4–51.
20. Delaglio, F., Grzesiek, S., Vuister, G. W., Zhu, G.,Pfeifer, J. & Bax, A. (1995). NMRPipe: a multidimen-sional spectral processing system based on UNIXpipes. J. Biomol. NMR, 6, 277–293.
21. Wishart, D. S., Bigam, C. G., Yao, J., Abildgaard, F.,Dyson, H. J., Oldfield, E. et al. (1995). 1H, 13C and 15Nchemical shift referencing in biomolecular NMR.J. Biomol. NMR 6, 135–140.
22. Palmer, A. G., 3rd, Rance, M. & Wright, P. E. (1991).Intramolecular motions of a zinc finger DNA-bindingdomain from Xfin characterized by proton-detectednatural abundance 13C heteronuclear NMR spectros-copy. J. Am. Chem. Soc. 113, 4371–4380.
23. Grzesiek, S. & Bax, A. (1993). Amino acid typedetermination in the sequential assignment procedure
of uniformly 13C/15N-enriched proteins. J. Biomol.NMR, 3, 185–204.
24. Muhandiram, D. R. & Kay, L. E. (1994). Gradient-enhanced triple-resonance three-dimensional NMRexperiments with improved sensitivity. J. Magn.Reson. 106, 203–216 (series B).
25. Grzesiek, S. & Bax, A. (1992). Improved 3D triple-resonance NMR techniques applied to a 31 kDaprotein. J. Magn. Reson. 96, 432–440.
26. Kay, L. E., Xu, G. Y. & Yamazaki, T. (1994). Enhanced-sensitivity triple-resonance spectroscopy with mini-mal H2O saturation. J. Magn. Reson. 109, 129–133.
27. Clubb, R. T., Thanabal, V. & Wagner, G. (1992). Aconstant-time three-dimensional triple-resonancepulse scheme to correlate intraresidue 1HN, 15N, and13C chemical shifts in 15N-13C-labelled proteins.J. Magn. Reson. 97, 213–217.
28. Montelione, G. T., Lyons, B. A., Emerson, S. D. &Tashiro, M. (1992). An efficient triple resonanceexperiment using carbon-13 isotropic mixing fordetermining sequence-specific resonance assignmentsof isotopically-enriched proteins. J. Am. Chem. Soc.114, 10974–10975.
29. Grzesiek, S., Anglister, J. & Bax, A. (1993). Correlationof backbone amide and aliphatic side-chain reso-nances in 13C/15N-enriched proteins by isotropicmixing of 13C magnetization. J. Magn. Reson. B, 101,114–119.
30. Lyons, B. A. & Montelione, G. T. (1993). An HCCNHtriple-resonance experiment using carbon-13 isotropicmixing for correlating backbone amide and side-chainaliphatic resonances in isotopically enriched proteins.J. Magn. Reson. 101, 206–209.
31. Logan, T. M., Olejniczak, E. T., Xu, R. X. & Fesik, S. W.(1993). A general method for assigning NMR spectraof denatured proteins using 3D HC(CO)NH-TOCSYtriple resonance experiments. J. Biomol. NMR, 3,225–231.
32. Clowes, R. T., Boucher, W., Hardman, C. H., Domaille,P. J. & Laue, E. D. (1993). A 4D HCC(CO)NNHexperiment for the correlation of aliphatic side-chainand backbone resonances in 13C/15N-labelled pro-teins. J. Biomol. NMR, 3, 349–354.
33. Carlomagno, T., Maurer, M., Sattler, M., Schwendin-ger, M. G., Glaser, S. J. & Griesinger, C. (1996). PLUSHTACSY: homonuclear planar TACSY with two-bandselective shaped pulses applied to Cα,C′ transfer andCβ, Caromatic correlations. J. Biomol. NMR, 8, 161–170.
34. Kay, L. E., Xu, G. Y., Singer, A. U., Muhandiram, D. R.& Formankay, J. D. (1993). A gradient-enhancedHCCH-TOCSY experiment for recording side-chain1H and 13C correlations in H2O samples of proteins.J. Magn. Reson. B, 101, 333–337.
35. Zuiderweg, E. R. P. & Fesik, S. W. (1989). Hetero-nuclear three-dimensional NMR spectroscopy ofthe inflammatory protein C5a. Biochemistry, 28,2387–2391.
36. Zhang, O., Kay, L. E., Olivier, J. P. & Forman-Kay, J. D.(1994). Backbone 1H and 15N resonance assignmentsof the N-terminal SH3 domain of drk in folded andunfolded states using enhanced-sensitivity pulsedfield gradient NMR techniques. J. Biomol. NMR, 4,845–858.
37. Palmer, I., A. G., Cavanagh, J., Wright, P. E. & Rance,M. (1991). Sensitivity improvement in proton-detectedtwo-dimensional heteronuclear correlation NMRspectroscopy. J. Magn. Reson. 93, 151–170.
38. Kay, L. E., Keifer, P. & Saarinen, P. (1992). Pureabsorption gradient enhanced heteronuclear single
716 NMR Relaxation Analysis of DsbA
quantum correlation spectroscopy with improvedsensitivity. J. Am. Chem. Soc. 114, 10663–10665.
39. Schleucher, J., Schwendinger, M., Sattler, M., Schmidt,P., Schedletzky, O., Glaser, S. J. et al. (1994). A generalenhancement scheme in heteronuclear multidimen-sionalNMRemploying pulsed field gradients. J. Biomol.NMR, 4, 301–306.
40. Davis, A. L., Keeler, J., Laue, E. D. & Moskau, D.(1992). Experiments for recording pure-absorptionheteronuclear correlation spectra using pulsed fieldgradients. J. Magn. Reson. 98, 207–216.
41. Herrmann, T.,Guntert, P.&Wuthrich, K. (2002). ProteinNMR structure determination with automated NOEassignment using the new software CANDID and thetorsion angle dynamics algorithm DYANA. J. Mol. Biol.319, 209–227.
42. Cornilescu, G., Delaglio, F. & Bax, A. (1999). Proteinbackbone angle restraints from searching a databasefor chemical shift and sequence homology. J. Biomol.NMR, 13, 289–302.
43. Schwieters, C. D., Kuszewski, J. J., Tjandra, N. &Clore,G.M. (2003). The Xplor-NIHNMRmolecular structuredetermination package. J. Magn. Res. 160, 66–74.
44. Truffault, V., Coles, M., Diercks, T., Abelmann, K.,Eberhardt, S., Luttgen, H. et al. (2001). The solutionstructure of the N-terminal domain of riboflavinsynthase. J. Mol. Biol. 309, 949–960.
45. Vriend, G. (1990). WHAT IF: a molecular modelingand drug design program. J. Mol. Graph. 8, 52–56.
46. Laskowski, R. A., MacArthur, M. W., Moss, D. S. &Thornton, J. M. (1993). PROCHECK: a program tocheck the stereochemical quality of protein structures.J. Appl. Crystallog. 26, 283–291.
47. Laskowski, R.A., Rullmann, J.A.C.,MacArthur,M.W.,Kaptein, R. & Thornton, J. M. (1996). AQUA andPROCHECK-NMR: programs for checking the qualityof protein structures solved byNMR. J. Biomol. NMR, 8,477–496.
48. Dosset, P., Hus, J. C., Blackledge, M. & Marion, D.(2000). Efficient analysis of macromolecular rotationaldiffusion from heteronuclear relaxation data. J. Biomol.NMR, 16, 23–28.
49. Mandel, A. M., Akke, M. & Palmer, A. G., 3rd (1995).Backbone dynamics of Escherichia coli ribonuclease HI:correlations with structure and function in an activeenzyme. J. Mol. Biol. 246, 144–163.
50. Szyperski, T., P., G., Stone, S. R. &Wüthrich, K. (1992).Nuclear magnetic resonance solution structure ofHirudin(1-51) and comparison with correspondingthree-dimensional structures determined using thecomplete 65-residue Hirudin polypeptide chain.J. Mol. Biol. 228, 1193–1205.
51. Kay, L. E., Torchia, D. & Bax, A. (1989). Backbonedynamics of proteins as studied by nitrogen-15inverse detected heteronuclear NMR spectroscopy:
application to staphylococcal nuclease. Biochemistry,28, 8972–8979.
52. Barbato, G., Ikura, M., Kay, L. E., Pastor, R. W. & Bax,A. (1992). Backbone dynamics of calmodulin studiedby 15N relaxation using inverse detected two-dimen-sional NMR spectroscopy: the central helix is flexible?Biochemistry, 31, 5269–5278.
53. Habazettl, J., Myers, L. C., Yuan, F., Verdine, G. L. &Wagner, G. (1996). Backbone dynamics, amidehydrogen exchange, and resonance assignments ofthe DNA methylphosphotriester repair domain ofEscherichia coli Ada uqsing NMR. Biochemistry, 35,9335–9348.
54. Lipari, G. & Szabo, A. (1982). Model-free approach tothe interpretation of nuclear magnetic resonancerelaxation in macromolecules. 1. Theory and rangeof validity. J. Am. Chem. Soc. 104, 4546–4559.
55. Lipari, G. & Szabo, A. (1982). Model-free approach tothe interpretation of nuclear magnetic resonancerelaxation in macromolecules. 2. Analysis of experi-mental results. J. Am. Chem. Soc. 104, 4559–4570.
56. Clore, G.M., Szabo, A., Bax,A., Kay, L. E., Driscoll, P. C.& Gronenborn, A. M. (1990). Deviations from thesimple two-parameter model-free approach to theinterpretation of nitrogen-15 nuclear magnetic relaxa-tion of proteins. J. Am. Chem. Soc. 112, 4989–4991.
57. dAuvergne, E. J. & Gooley, P. R. (2006). Model-freemodel elimination: A new step in the model-freedynamic analysis of NMR relaxation data. J. Biomol.NMR, 35, 117–135.
58. Akaike, H. (1973). Information theory and anextension of the maximum likelihood principle. InProceedings of the 2nd International Symposium onInformation Theory, pp. 267–281, Academiai Kiado,Budapest.
59. d'Auvergne, E. J. & Gooley, P. R. (2003). The use ofmodel selection in the model-free analysis of proteindynamics. J. Biomol. NMR, 25, 25–39.
60. Barret, C. P., Hall, B. A. & Noble, M. E. M. (2004).Dynamite: a simple way to gain insight into proteinmotions. Acta Crystallog. sect. D, 60, 2280–2287.
61. Lindahl, E., Hess, B. & Spoel, D. (2001). GROMACS3.0: a package for molecular simulation and trajectoryanalysis. J. Mol. Mod. 7, 306–317.
62. Humphrey, W., Dalke, A. & Schulten, K. (1996). VMD:visual molecular dynamics. J. Mol. Graph. 14, 27–38.
63. de Groot, B. L., van Aalten, D. M., Scheek, R. M.,Amadei, A., Vriend, G. & Berendsen, H. J. (1997).Prediction of protein conformational freedom fromdistance constraints. Proteins: Struct. Funct. Genet. 29,240–251.
64. Ayed, A., Mulder, F. A. A., Yi, G.-S., Lu, Y., Kay, L. E. &Arrowsmith, C. E. (2001). Latent and active p53 areidentical in conformation. Nature Struct. Biol. 8,756–760.
Edited by A. G. Palmer III
(Received 11 December 2006; received in revised form 18 April 2007; accepted 21 May 2007)Available online 31 May 2007
