Protein stabilization by compatible solutes Effect of diglycerol phosphate on the dynamics of Desulfovibrio gigas rubredoxin studied by NMR Pedro Lamosa 1 , David L. Turner 1,2 , Rita Ventura 1 , Christopher Maycock 1 and Helena Santos 1 1 Instituto de Tecnologia Quı ´ mica e Biolo ´ gica, Universidade Nova de Lisboa, Oeiras, Portugal; 2 Department of Chemistry, University of Southampton, UK Heteronuclear NMR relaxation measurements and hydro- gen exchange data have been used to characterize protein dynamics in the presence or absence of stabilizing solutes from hyperthermophiles. Rubredoxin from Desulfovibrio gigas was selected as a model protein and the effect of diglycerol phosphate on its dynamic behaviour was studied. The presence of 100 m M diglycerol phosphate induces a fourfold increase in the half-life for thermal denaturation of D. gigas rubredoxin [Lamosa, P., Burke, A., Peist, R., Huber, R., Liu, M.Y., Silva, G., Rodrigues-Pousada, C., LeGall, J., Maycock, C. & Santos, H. (2000) Appl. Environ. Microbiol. 66, 1974–1979]. A model-free analysis of the protein backbone relaxation parameters shows an average increase of generalized order parameters of 0.015 reflecting a small overall reduction in mobility of fast-scale motions. Hydrogen exchange data acquired over a temperature span of 20 °C yielded thermodynamic parameters for the struc- tural opening reactions that allow for the exchange. This shows that the closed form of the protein is stabilized by an additional 1.6 kJÆmol )1 in the presence of the solute. The results seem to indicate that the stabilizing effect is due mainly to a reduction in mobility of the slower, larger-scale motions within the protein structure with an associated increase in the enthalpy of interactions. Keywords: chemical exchange; compatible solutes; protein dynamics; rubredoxin; thermostability. Protein stability, activity and dynamics are interrelated issues with great importance not only in physiological processes but also in protein engineering. The evolution of protein structures towards extreme thermostability was vital for hyperthermophiles, microorganisms thriving near the boiling point of water. In general, the proteins of these organisms are intrinsically resistant to heat denaturation. However, hyperthermophiles also possess intracellular pro- teins that are not particularly stable, implying the existence of alternative strategies for their stabilization in vivo [1,2]. Hyperthermophiles accumulate high levels of charged organic osmolytes in response to supra-optimal growth temperatures, and this observation led to the hypothesis that these compounds play a role in thermoprotection of macromolecules in vivo [3,4]. This view is supported by in vitro studies showing that these osmolytes protect proteins against heat [1,5–8]. Nevertheless, the molecular basis for this well established stabilization phenomenon remains elusive. Several possible mechanisms for protein stabilization by osmolytes have been proposed [9–11]. Arakawa and Timasheff [12,13] proposed a preferential hydration model to explain protein stabilization by compatible solutes: solute molecules are excluded from the protein surface, thereby making denaturation entropically less favourable. In con- formity, exclusion factors have been measured for a variety of organic solutes and salts [14–16], however, the correlation between exclusion factors and the degree of protection a solute can bestow upon a particular protein is neither unequivocal nor general [17,18]. These apparent inconsis- tencies have sometimes been interpreted as being due to specific protein–solute interactions [7,19]. In fact, the magnitude of the stabilizing effect depends on the particular solute–protein pair examined [5,8,19]. Another approach, proposed by Bolen and coworkers, describes the stabilizing or destabilizing nature of inter- actions between solutes and exposed groups in the protein structure [20,21]. In this proposal, the stabilizing effect is attributed mainly to a large contribution from interactions with exposed backbone groups in a partially unfolded state, with side-chain interactions modulating the specificity of the effect. Overall, the interactions should cause a contraction of the protein structure with a concomitant decrease in internal mobility [21,22]. Indeed, the higher thermal stability of hyperthermophilic proteins has often been correlated with structure rigidification [23,24]. Structural data, both from X-ray and NMR, on series of homologous proteins show evidence for stronger local interactions and/or improved packing of the polypeptide chain, which would bring about a higher conformational rigidity [23]. More- over, the lower catalytic efficiency observed in hyperther- mophilic enzymes is usually explained by the decreased Correspondence to H. Santos, Instituto de Tecnologia Quı ´ mica e Biolo ´ gica, Apartado 127, 2780-156 Oeiras, Portugal. Fax: + 351 21 4428766, Tel.: + 351 21 4469828, E-mail: santos@itqb.unl.pt Abbreviations: DGP, diglycerol phosphate; RdDg, Rubredoxin from Desulfovibrio gigas. (Received 4 July 2003, revised 22 September 2003, accepted 2 October 2003) Eur. J. Biochem. 270, 4606–4614 (2003) Ó FEBS 2003 doi:10.1046/j.1432-1033.2003.03861.x flexibility of the active site, corroborated by the fact that mutations increasing thermostability while maintaining low-temperature activity are extremely rare [25]. In fact, this rigidification has been revealed by H–D exchange experiments [23,26]. However, this view was recently challenged by the observation of relatively fast exchange rates in the rubredoxin from Pyrococcus furiosus,themost stable protein known to date [27]. In this context, assessing the changes in the dynamic behaviour of proteins in the presence of solutes is expected to shed light on the stabilization phenomenon. Desulfovibrio gigas rubredoxin (RdDg), a small iron– sulfur protein with a hydrophobic core formed by the side chains of six invariant residues, a three-stranded b-sheet, and an exposed hairpin loop, was chosen as a model protein. Its NMR solution structure was recently obtained [28]. Also, RdDg is highly stabilized by diglycerol phosphate (DGP), a solute accumulated by the hyperthermophilic archaeon Archaeoglobus fulgidus [7,29]. Addition of 100 m M DGP yields a fourfold increase in the half-life for thermal denaturation of RdDg, measured by UV–visible spectros- copy at 90 °C[7]. We used NMR for these studies because it provides a wide range of time-scales for the dynamic analyses. Heteronuclear NMR relaxation data, from which general- ized order parameters can be derived, provides a tool to probe the dynamic behaviour of proteins in various conditions [30–32]. Hydrogen exchange rates of labile protons, such as the amide protons of protein backbones, can also provide dynamic information on longer time- scales. Amide protons that are buried inside the protein structure and/or involved in hydrogen bonds require a local structural opening to allow exchange with solvent protons [33,34]. Therefore, the measurement of amide exchange rates can be used to evaluate the relationship between stability and the rigidity of several parts of the protein structure [35,36]. Materials and methods Rubredoxin production Plasmid pRPPL1 [7] harbouring the RdDg gene was digested with NdeIandEcoRI restriction enzymes. The 175-bp DNA fragment obtained was purified from an agarose gel (2%) and inserted into vector pT7-7 [37] previously digested with the same restriction enzymes. The resulting construct was named pMSPL1. Escherichia coli strain BL21(DE) was transformed with pMSPL1 and grown in medium containing: KH 2 PO 4 ,4.5gÆL )1 ; K 2 HPO 4 ,10.5gÆL )1 ;NaCl,0.5gÆL )1 ;Mg 2 SO 4 Æ7H 2 O, 0.5 gÆL )1 ;FeCl 3 Æ3H 2 O6mgÆL )1 ;U 15 N-(NH 4 ) 2 SO 4 ,2gÆL )1 ; glucose, 4 gÆL )1 ; vitamin solution, 10 mLÆL )1 ; trace element solution 10 mLÆL )1 and ampicillin 100 mgÆL )1 . One litre of vitamin solution contained 500 mg aminobenzoic acid, 200 mg nicotinic acid, 100 mg pantothenic acid, 500 mg pyridoxine, 100 mg thiamine, 200 mg thioctic acid, 200 mg biotin, 100 mg folic acid and 100 mg riboflavin. The trace element solution contains per litre: CaCl 2 ,1.06g; MnSO 4 Æ5H 2 O, 50 mg; CuSO 4 Æ5H 2 O, 8 mg; ZnSO 4 Æ7H 2 O 40 mg; NaMoO 4 Æ2H 2 O, 8 mg; CoCl 2 Æ6H 2 O, 8 mg; H 3 BO 3 , 6mg. Transformed E. coli cells were grown until D ¼ 0.3 and RdDg production induced with isopropyl thio-b- D -galacto- side (IPTG; 25 lgÆL )1 final concentration). At this time the culture was supplemented with glycerol (4 mLÆL )1 )and ZnCl 2 (5 mgÆL )1 final concentration) and incubated for 8 h. Purification of the recombinant protein was performed as described previously [7]. A yield of approximately 10 mgÆL )1 of the zinc form of RdDg uniformly labelled with 15 Nwasobtained. Sample preparation Purified uniformly 15 N-labelled RdDg (Zn form) was concentrated and the buffer removed by ultrafiltration using a YM3 membrane (Amicon). Two samples were prepared in 10% 2 H 2 O at a final concentration of  4m M . In one sample, DGP (potassium salt) was added to a final concentration of 100 m M , while in the other sample KCl was added to the same concentration. The pH was adjusted to 6.9 in both samples and an antibiotic cocktail was added with 70 l M ampicillin, 50 l M kanamicin, 50 l M rifampicin and 50 l M chloroamphenicol. For the 1 H– 2 H exchange experiments RdDg (Zn form) was used at a final concentration of  1m M . KCl or DGP was added to the protein in 2-mL Eppendorf tubes to a final concentration of 100 m M , the pH was adjusted to 6 in the unlabelledsamplesandto5inthe 15 N-labelled RdDg, and the samples were freeze-dried. The dried samples were then dissolved in 2 H 2 O, the pH readjusted (if necessary), and placed in the spectrometer at the desired temperature. After allowing a period for temperature equilibration, series of 1D 1 H(or2D 1 H- 15 N HSQC for the labelled samples) spectra were acquired. NMR spectroscopy Unless otherwise stated all spectra were recorded at 303 K in a DRX500 Bruker spectrometer equipped with a 5-mm inverse detection probe head with internal B 0 gradient coils. (Bruker, Rheinstetten, Germany). Temperature was con- trolled using a Eurotherm 818 unit with a B-CU 05 cooling unit. One-dimensional 1 H spectra for the exchange experi- ments were acquired with 72 transients, and continuous low-power water saturation during the relaxation delay of 2.0 s. A series of 1 H– 15 N correlation spectra was acquired to measure the 15 N relaxation constants R 1 and R 2 ,and heteronuclear 1 H– 15 N NOE using the procedures outlined in Kay et al. [38], modified to include a Watergate 3-9-19 water suppression scheme [39]. Values of R 1 and R 2 were obtained by fitting the intensities (measured as peak- volumes) over time to a single exponential decay. NOE enhancements were taken from the mean value of three integrations of peak volumes in spectra recorded with and without proton saturation. The 2D 15 N– 1 H HSQC spectra were recorded with standard Bruker pulse programs. In these experiments 4096 1 H · 512 15 N data points were collected using a delay of 2.7 ms for evolution of magneti- zation in the INEPT transfer sequence. The 3D 15 N– 1 H HSQC-TOCSY spectrum (4096 1 H · 32 15 N · 64 1 Hdata points) was recorded using a delay of 2.7 ms evolution of magnetization in the INEPT transfer sequence and a TOCSY mixing time of 80 ms. The data were processed Ó FEBS 2003 Effect of DGP on rubredoxin dynamics (Eur. J. Biochem. 270) 4607 with standard BRUKER software (Bruker). Polynomial baseline corrections were applied in both dimensions of all 2D spectra. Results 1 H and 15 N chemical shifts The 2D 1 H- 15 N-HSQC spectrum of RdDg was assigned with the aid of a 3D 1 H– 15 N TOCSY-HSQC spectrum and published proton chemical shift data [28] (Fig. 1). Ambigu- ities in signal assignment due to overlap in the 1 Hdimension were solved through spin-system analysis of the 3D TOCSY-HSQC spectrum, but three signals with AMX type spin-systems could not be assigned unequivocally. The temperature dependence of the 1 Hand 15 Nchemical shifts was investigated in the presence of 100 m M KCl or DGP by acquiring a series of 1 H– 15 N HSQC spectra over a temperature span of 50 °C (from 30 to 80 °C). At 30 °C, the addition of DGP had little or no influence on the proton NH chemical shifts of RdDg. In fact, chemical shift displacement upon solute addition seems random, with most shift changes within experimental error, an average value of 0.004 p.p.m., and a maximum value of 0.087 p.p.m. (Phe30). The displacement of 15 N chemical shifts follows a similar pattern, with an average value of 0.031 p.p.m and a maximum value of 0.607 p.p.m. (Ala48). These results agree with previous findings [28], in which DGP addition caused no visible change in the proton spectrum. The chemical shifts of amide protons in RdDg present a small, linear dependence on temperature (up to 80 °C), both in the presence of 100 m M KCl and DGP. The variation of chemical shift with temperature seems random with average slope of )0.0029 ± 0.0028 p.p.m.ÆK )1 throughout the protein, with the error given as the standard deviation of the slopes. The segment 25–32 in the protein sequence shows the largest temperature dependences with an average of )0.0066 ± 0.0045 p.p.m.ÆK )1 . The addition of DGP does not significantly change this pattern. In fact, the difference in chemical shift temperature dependence with or without DGP is random and within the experimental error. Amide 15 N chemical shifts display both positive and negative correlations with temperature, which seem unrelated to protein sequence or residue type and present a relatively small range of values (from 0.018 p.p.m.ÆK )1 in Val5 to )0.047 p.p.m.ÆK )1 in Phe49). In some residues, such as Ile3, Tyr11, Gly23, Lys25, Phe30 or Ser45, the chemical shifts are temperature independent. Many of the plots of 15 N chemical shift against temperature are nonlinear. This also seems unrelated to protein structure or residue nature. Upon solute addition, all signals still exhibit little tempera- ture dependence and tend to maintain their positive or negative correlations. Relaxation data and dynamic parameters Relaxation parameters were measured at 30 °C for 42 of the 47 15 N amide nuclei present in the protein (Fig. 2), and analysed using the program Model-free v.4.01 [31,40]. The diffusion tensor (D) and the rotational correlation time (s m ) were evaluated prior to analysis. The software package R 1 R 2_ DIFFUSION [31,40] was used to translate the centre of mass of the mean structure of the NMR ensemble [28] to the origin of coordinates, and to estimate D from T 1 /T 2 ratios. Residues that might be undergoing conformational exchange were identified from the condition: (ÆT 2 æ ) T 2,n )/ÆT 2 æ ) (ÆT 1 æ ) T 1,n )/ÆT 2 æ >1.5r andexcluded[41].Here,T 2,n is the T 2 of residue n, ÆT 2 æ is the average T 2 ,andr is the standard deviation of (ÆT 2 æ ) T 2,n )/ÆT 2 æ ) (ÆT 1 æ ) T 1,n )/ÆT 2 æ. The axially symmetric diffusion model best fitted the experimental data, and the structure was rotated to its principal axis for use in the model-free analysis. The parameters, selected by extensive Monte-Carlo simulations as described by Mandel et al. [31], are summarized in Table 1. After model selection, both the correlation time and the axially symmetric diffusion tensor were optimized simultaneously with all other model-free parameters. In the presence of KCl, there were five residues that did not fit any model in the analysis; these are Tyr11, Tyr13, Leu33, Gly43, and Ala44. Five residues also failed to fit any model in the presence of DGP: Thr7, Val8, Ala16, Leu33, and Val41. The rotational correlation time, s m , determined in the final calculations, was 3.9 ± 0.2 and 4.6 ± 0.4 ns in the presence of 100 m M KCl and DGP, respectively. These values for s m are in agreement with the observed negative NOE values and the small size of the protein. Effective correlation times (s e ) in the range of 20–70 ps were found for 13 residues in 100 m M KCl (Fig. 3). In the presence of DGP, 10 residues required the determination of s e to fit the model. In both cases, most of these residues are located in the hairpin loop region. Only two residues (8 and 46) required an R ex term for adequate fitting in the presence of KCl, with values ranging from 0.8 to 4 s )1 . When DGP was present, six residues needed an R ex term (residues 24, 31, 32, 44, 49 and 51), but the fitted value is close to zero in all six cases. The values of the generalized order parameter, S 2 (Fig. 3) do not display any particular trend over the protein Fig. 1. 1 H– 15 NHSQCspectrumof 15 N-labelled RdDg (Zn-form) in the presence of 100 m M KCl at 30 °C. 4608 P. Lamosa et al. (Eur. J. Biochem. 270) Ó FEBS 2003 sequence except for the small values in residue 2, which agrees with the expected flexibility of the N-terminal region of the protein. The difference of S 2 values in the presence and absence of solute are shown in Fig. 4. Overall, the average S 2 values tend to be higher in the presence of DGP, but the average difference is only 0.015, and there is no obvious trend towards segmental rigidification of any part of the sequence. Instead, the whole protein (with the exception of residues 14–18 and 37–45) tends to display higher S 2 values in the presence of the solute (Fig. 4). 1 H– 2 H amide exchange To evaluate the relative mobility and exposure of the several segments of the protein sequence, 1 H– 2 H amide exchange rates of 15 N labelled RdDg were measured at 40 °Cand pH 5 by recording 2D HSQC spectra in 2 H 2 O as a function of time and fitting the peak volumes to single exponential decays (Fig. 5). The exchange rates of several amides were inaccessible under these experimental conditions: 26 resi- dues exchanged so rapidly that the signals were undetectable at the start of the spectral acquisition, and nine residues gave signals that remained almost constant throughout the experiment, indicating half-lives greater than 250 h. The slowly exchanging residues are clustered around the knuckle that contains the metal centre, while the central region of the b-sheet and the base of the hairpin loop display intermediate exchange rates (Fig. 6). The most rapidly exchanging residues are positioned in the hairpin loop, the protein termini and the less structured region of residues 34–36 [28], which is in agreement with a possible higher mobility of these regions. The addition of DGP produced a remarkable increase of half-lives in the 17 amide exchange rates that were measured both in its presence and absence at 40 °C, reflecting the structural stabilization provided by this solute. The EX2 exchange regime has been established in various rubredoxins (as in most globular stable proteins) [27,42]. In fact, EX1 reactions are rarely seen in stable proteins, occurring mostly under the conditions used in some protein refolding experiments [43–46]. Under the EX2 regime, the exchange rates are described by Eqn (1): k ex ¼ K op k ch ½Catð1Þ where K op is the equilibrium constant for structural opening reactions that expose the NH group [33]. The term k ch [Cat] can be calculated from exchange rates in unstructured peptides and used to obtain K op , and hence a value of DG for the opening reactions [47]. Assu- ming that the slowest exchanging residues (Val5, Fig. 2. 15 N amide relaxation parameters of RdDg as a function of residue number in the presence of 100 m M KCl (A–C), or 100 m M DGP (D–F). (A,D) Longitudinal relaxation time; (B,E) transverse relaxation time; (C,F) heteronuclear NOE. Table 1. Summary of parameters used to fit T 1 , T 2 and hNOE. S 2 is the square of the generalized order parameter characterizing the amplitude of the internal motions; s e is the effective correlation time for the internal motions; R ex , is the exchange contribution to T 2 ,andthe subscripts f and s indicate fast and slow time scales, respectively. Model Optimized parameters Fitted residues in the presence of KCl DGP 1S 2 23 24 2S 2 and s e 12 7 3S 2 and R ex 13 4S 2 , s e and R ex 13 5 S 2 s , S 2 f and s e 00 Not fit – 5 5 Ó FEBS 2003 Effect of DGP on rubredoxin dynamics (Eur. J. Biochem. 270) 4609 Cys6, Thr7, Val8, Cys9, Tyr11, Tyr13, Cys39, Val41, and Cys42), which are all located near the metal centre, exchange via a single opening reaction, it is possible to use the measured exchange rates at five temperatures (between 50 and 70 °C) at pH 6, to obtain the tempera- ture dependence for the DG of the structural opening Fig. 3. Estimated model-free parameters of RdDg as a function of residue number in the presence of 100 m M KCl (A–C), or 100 m M DGP (D–F). (A,D) Generalized order parameter; (B,E) effective correlation time; (C,F) chemical exchange rate. Fig. 4. Difference between the generalized or- der parameters in the presence of 100 m M KCl or DGP of RdDg. Only residues whose parameters were calculated in both cases with thesame(blackbars)orwithdifferent dynamic models (grey bars) are included. Fig. 5. Half-life values for the 1 H– 2 Hamide exchange reaction in RdDg measured at 40 °C in the presence of DGP (black bars) or KCl (grey bars) at 100 m M . The broken bars rep- resent the slowest exchanging residues with half-life values higher than 250 h, which were too long to be determined in the experimental time frame. 4610 P. Lamosa et al. (Eur. J. Biochem. 270) Ó FEBS 2003 around the metal centre, that allows the exchange of these amide residues to take place. In these residues, DG displayed a linear dependence on temperature. Con- sidering the experimental errors and the narrow range of temperatures investigated, it is reasonable to treat DHas constant, and use the Gibbs equation to obtain global enthalpies and entropies for the opening reaction. In the presence of 100 m M KCl the structural opening reactions presented an average DG at 60 °Cof28.8±0.9kJÆ mol )1 ,fromwhichaDHof143±23kJÆmol )1 and DS of 0.34 ± 0.07 kJÆK )1 Æmol )1 can be derived. In the presence of DGP an average DG of 30.4 ± 0.9 kJmol )1 was found, and a DH of 160 ± 23 kJmol )1 and DS of 0.39 ± 0.07 kJÆK )1 Æ mol )1 were calculated. Upon solute addition a stabilization of RdDg took place and an average positive DG of 1.6 kJÆmol )1 can be observed for the structural opening reactions that allow amide exchange around the metal centre to take place. Gibbs free energy vs. temperature plots result in two almost parallel lines in the presence of KCl or DGP (Fig. 7), which suggests that, in this case, the enthalpic contribution is more important to the stabilization phenomenon than the entropic term. Discussion Chemical shifts are sensitive probes of structural changes in the mean position of atoms within a defined structure or protein conformation. The presence of DGP caused no significant alteration in 1 Hor 15 N chemical shifts or in their temperature dependence. This means that, throughout the temperature range investigated (from 30 to 80 °C), solute addition leaves the mean protein structure unchanged. A similar lack of change was observed in chymotrypsin inhibitor 2 and in horse heart cytochrome c,wherethe presence of 2 M glycine produced no visible alteration in proton chemical shifts [36]. These results support the view that stabilizing compatible solutes exert their effect through changes in solvent structure and/or subtle changes in the dynamic properties of the protein rather than by changing the structure of the protein itself. In fact, protein function depends vitally on its structure and dynamics; if a stabilizing solute induced substantial changes, it might hamper enzyme activity, rendering the stabilizing effect useless. The solute DGP is an effective stabilizer of RdDg [7], however, like other solutes, it does not seem to alter protein structure. The increased stability of proteins from hyper- thermophiles, in comparison with their mesophilic homo- logues, has frequently been interpreted as a consequence of a number of mechanisms that improve internal attractive forces and result in increased rigidity of protein structure [48]. However, the terms rigidity or flexibility must be regarded with caution, for there is no single measure of flexibility. Proteins undergo a wide variety of motions on vastly different time-scales; a protein may be rigid in the nanosecond, and flexible on the millisecond time-scale. Bearing this in mind, it is interesting to note that the increased rigidity of RdDg upon addition of DGP is small in the more rapid, low amplitude motions. The effect is mostly in the longer time-scales and in the restriction of wider concerted motions, which encompass the structural reac- tions that allow protected amides to exchange with the solvent. In fact, solute addition caused relatively small changes in the relaxation parameters, in particular for T 1 , which remained practically unaltered. T 1 relaxation meas- urements carry information about motions with frequencies of about 10 8 )10 12 Hz, while T 2 and NOE enhancements also depend upon higher amplitude motional regimes in the micro- and millisecond time scale [38]. Thus, solute addition appears to have a greater impact on the wider motions, while leaving the small high frequency fluctuations unchanged. Recently, however, Wang et al. [49] argued that, in the case of ubiquitin, 15 N relaxation measurements alone underestimate the variations in backbone dynamics. If this proves to be a general effect, and not just a peculiarity of ubiquitin, it could mean that the variations in backbone Fig. 6. Backbone of RdDg showing different time ranges for the hydrogen exchange rates of amide groups: slow (dark-blue), medium (mid-blue), and fast (cyan). Cysteine sulphur atoms are depicted in yellow. The slowest exchanging residues, clustered around the metal centre, are indicated by residue number. Fig. 7. Temperature dependence of the average free energy of the opening reactions that determine hydrogen amide exchange, in the presence of 100 m M KCl (s)andDGP(d). Ó FEBS 2003 Effect of DGP on rubredoxin dynamics (Eur. J. Biochem. 270) 4611 dynamics we have measured may be underestimated, which in turn could explain why they are so small. Nevertheless, the faster time-scales would still be the least affected by solute addition, suggesting that the stabilizing effect oper- ates via a higher restriction of the wider motions. Most dynamic studies reported in the literature were performed on proteins with about 100 amino acids, and rotational correlation times of around 6–9 ns [30,38,50–52]. D. gigas rubredoxin has only 52 amino acids and the smaller correlation time (3.9 ns was estimated in 100 m M KCl), higher average R 2 values and lower (more negative) NOE enhancements [38,53] agree well with values from other studies in view of the smaller size of RdDg. Studies in which the dynamics of a protein are investigated in different conditions invariably find increases in correlation times upon ligand binding which are interpreted in the light of the higher mass or bigger size of the complex [32,52,54]. In this study, the presence of DGP was found to increase the correlation time by 0.7 ± 0.6 ns. It is difficult to interpret this effect as a mass variation, since binding of the solute should cause more dramatic changes in the chemical shifts than we observed. Although close to the experimental error, a slight increase in correlation time would be consistent with a higher solvent viscosity, brought about by solute addition. The measured generalized order parameters (S 2 )for RdDg are all above 0.7 (with the exception of Asp2), which denotes a relatively rigid protein structure. The least ordered parts of the NMR structure [28] are the N terminus and the loop region. DGP addition leaves the overall picture unaltered, but, with the exception of segment 37–45, there is a trend towards generalized rigidification of the protein (Fig. 4). Most residues that require the determination of s e to fit the model, with or without solute, are located in the loop. Similar behaviour has been observed in the loop regions or turns of E. coli topoisomerase I [52]. The average s e value for these high frequency motions is also left almost undisturbed by solute addition. The R ex term measures wider motions than S 2 or s e and reflects ÔslidingÕ or ÔbreathingÕ motions in protein structure [54,55]. In the presence of 100 m M KCl only two residues required this term for adequate fitting, which implies that these motions are limited in RdDg. The addition of DGP required the inclusion of R ex terms in six residues, but in all cases the fitted value is near zero, suggesting that the solute completely restricts concerted motions of groups on larger time-scales. The influence of DGP addition on the order parameters of RdDg is of the same magnitude as the effect of ligand binding reported in several studies. For instance, in ketosteroid isomerase from Pseudomonas testosteroni the active site residues show an average increase in order parameters upon ligand binding of only 0.03, while order parameters decrease in the rest of the protein [32]. In 4-oxalocrotonate tautomerase, binding of a competitive inhibitor causes almost no change in average order param- eters [54]. Most of the residues in E. coli topoisomerase I suffer a decrease in order parameters upon DNA binding while a small portion of the protein, directly involved in binding, experiences an average order parameter increase of 0.04 [52]. In these studies, the increased mobility in large sections of the protein was interpreted as a compensatory effect for the unfavourable entropy associated with binding site rigidification [32,54]. In this light, the generalized increase in order parameters in RdDg upon addition of DGP can be interpreted as a thermodynamically unfavour- able entropic process. This would lead to an increased stability of the native form of the protein, in agreement with the preferential exclusion model [13,15,56], only if the effect on the denatured form were still more unfavourable. However, in rubredoxins, denaturation occurs in an irreversible process with concomitant loss of the metal centre and therefore stability is determined by the rate of unfolding. In fact, kinetic stability may be as important, physiologically, as thermodynamic stability, particularly in hyperthermophilic organisms. Many denaturation proces- ses are irreversible at high temperature and, in those cases, a solute that leaves DG for unfolding unchanged but that is able to increase the activation energy of the denaturing reaction would be an effective stabilizer. This leads us to consider the possible effects of solute addition on the activation energy of the denaturing process and we look to the results of the amide exchange experi- ments to provide information about structural openings as an approximation of the transition state in the unfolding process. In fact, 1 H– 2 H exchange experiments on series of homologous proteins, or in the presence of stabilizing solutes, have shown a strong correlation between stability and exchange rates [23,26,36]. In RdDg the slowest amide proton exchange rates were found in the metal centre and in the b-sheet region, which reflects the rigidity of these regions [27,34,45]. Most of the rapidly exchanging amide protons are exposed to the solvent [28] and little can be said about flexibility on the basis of exchange rates alone. The most slowly exchanging residues are protected by the protein structure and require a structural opening reaction to exchange. This opening of protein structure to the solvent (although transient) has obvious parallels in the process of denatur- ation, and hence probes protein stability [26,44]. In the presence of 100 m M DGP and at 90 °C, we found an increase of 6% in DG (1.6 kJÆmol )1 ) for the structural opening reaction around the metal centre, affecting the slowest exchanging amide groups. Although the linearity of the temperature plots points towards a single opening reaction around the metal centre, a superposition of several opening reactions cannot be ruled out. In any event, the decrease in DG, the relative importance of the entropic and enthalpic terms, and therefore the general conclusions, would still hold true. Although the localized opening required for amide exchange does not lead to loss of the metal centre, this effect may be compared with a change of 4.2 kJÆmol )1 required by the Arrhenius equation to explain the fourfold increase in the half-life for thermal denatura- tion [7]. The temperature dependence of the free energy provides information about the stabilizing or destabilizing nature of enthalpic and entropic contributions. Assuming that DH is constant over the experimental temperature range, the almost parallel linear fits obtained in the presence of DGP or KCl indicate that the added stability, in this case, is in essence a consequence of an enthalpy increase for the opening reaction, with small contributions from the entropic term. This is in agreement with a small protein rigidification in response to solute addition, inferred from 4612 P. Lamosa et al. (Eur. J. Biochem. 270) Ó FEBS 2003 the dynamic behaviour of the protein. Thus, we envisage an improvement in favourable interactions brought about by a slight restriction of the small high-frequency motions, and a larger reduction in lower-frequency movements such as sliding or ÔbreathingÕ motions. Conclusion Protein stability is the result of marginal differences between various large stabilizing and destabilizing interactions, and is therefore an elusive subject in which small and subtle changes may result in considerable added stability. In fact, if taken alone, the dynamic data derived from the relaxation measurements on RdDg do not look very informative, because the differences are small and seem random. However, they do point towards some rigidification, particularly with respect to the slower, wider motions, which is in agreement with the reduction in the slower amide exchange rates. Taking the results of the two sets of experiments together, and in view of the strong stabilizing effect the solute confers upon this protein and the lack of structural alteration, a clearer picture begins to emerge. Thus, despite the uncertainties in the experimental values, it appears that the stabilizing effect of DGP is essentially enthalpic (with small contributions from the entropic term), involving improved internal attractive forces and promoting a tighter protein structure with restricted large-scale motions, without significantly altering the smaller, faster dynamic motional regimes or perturbing the average protein structure. Acknowledgements The SON large-scale facility at Utrecht is acknowledged for valuable support and the acquisition of several spectra. This work was supported by the European Commission, 5th Framework Programme contract QLK3-CT-2000-00640, Fundac¸ a ˜ oparaaCieˆ ncia e Tecnologia, PRAXIS XXI and FEDER, Portugal (POCTI/BME/35131/99, and PRAXIS/BIO/12082/98). References 1. Hensel,R.&Ko ¨ nig, H. (1988) Thermoadaptation of methano- genic bacteria by intracellular ion concentration. FEMS Micro- biol. Lett. 49, 75–79. 2. Hensel, R., Fabry, S., Biro, J., Bogedain, C., Jakob, I. & Siebers, B. (1994) Glyceraldehyde-3-phosphate dehydrogenases from archaea: objects for studying protein thermoadaptation. Biocatalysis 11, 151–164. 3. da Costa, M.S., Santos, H. & Galinski, E.A. (1998) An overview of the role and diversity of compatible solutes in Bacteria and Archaea. Adv. Biochem. Eng. Biotechnol. 61, 117–153. 4. Santos, H. & da Costa, M.S. (2001) Organic solutes from ther- mophiles and hyperthermophiles. Methods Enzymol. 334,302– 315. 5. Scholz, S., Sonnenbichler, J., Scha ¨ fer, W. & Hensel, R. (1992) Di-myo-inositol-1,1¢-phosphate: a new inositol phosphate isolated from Pyrococcus woesei. FEBS Lett. 306, 239–242. 6. Ramos,A.,Raven,N.D.H.,Sharp,R.J.,Bartolucci,S.,Rossi,M., Cannio, R., Lebbink, J., van der Oost, J., de Vos, W.M. & Santos, H. (1997) Stabilization of enzymes against thermal stress and freeze-drying by mannosylglycerate. Appl. Environ. Microbiol. 63, 4020–4025. 7. Lamosa, P., Burke, A., Peist, R., Huber, R., Liu, M.Y., Silva, G., Rodrigues-Pousada, C., LeGall, J., Maycock, C. & Santos, H. (2000) Thermostabilization of proteins by diglycerol phosphate, a new compatible solute from the hyperthermophile Archaeoglobus fulgidus. Appl. Environ. Microbiol. 66, 1974–1979. 8. Borges,N.,Ramos,A.,Raven,N.D.,Sharp,R.J.&Santos,H. (2002) Comparative study of the thermostabilizing properties of mannosylglycerate and other compatible solutes on model enzymes. Extremophiles 6, 209–216. 9. Baldwin, R.L. (1996) How Hofmeister ion interactions affect protein stability? Biophys. J. 71, 2056–2063. 10. Timasheff, S.N. (1998) In disperse solution, Ôosmotic stressÕ is a restricted case of preferential interactions. Proc. Natl Acad. Sci. USA 95, 7363–7367. 11. Davis-Searles, P.R., Saunders, A.J., Erie, D.A., Winzor, D.J. & Pielak, G.J. (2001) Interpreting the effects of small uncharged solutes on protein-folding equilibria. Annu. Rev. Biophys. Biomol. Struct. 30, 271–306. 12. Arakawa, T. & Timasheff, S.N. (1983) Preferential interactions of proteins with solvent components in aqueous amino acid solu- tions. Arch. Biochem. Biophys. 224, 169–177. 13. Arakawa, T. & Timasheff, S.N. (1985) The stabilization of pro- teins by osmolytes. Biophys. J. 47, 411–414. 14. Arakawa, T. & Timasheff, S.N. (1982) Preferential interactions of proteins with salts in concentrated solutions. Biochemistry 21, 6545–6552. 15. Arakawa, T. & Timasheff, S.N. (1982) Stabilization of protein structure by sugars. Biochemistry 21, 6536–6544. 16. Arakawa, T. & Timasheff, S.N. (1984) Mechanism of protein salting in and salting out by divalent cation salts: balance between hydration and salt binding. Biochemistry 23, 5912–5923. 17. Arakawa, T., Bhat, R. & Timasheff, S.N. (1990) Why preferential hydration does not always stabilize the native structure of globular proteins. Biochemistry 29, 1924–1931. 18. Lee, L.L. & Lee, J.C. (1987) Thermal stability of proteins in the presence of poly (ethylene glycols). Biochemistry 26, 7813–7819. 19. Lippert, K. & Galinski, E.A. (1992) Enzyme stabilization by ectoine-type compatible solutes: protection against heating, freezing and drying. Appl. Microbiol. Biotechnol. 37, 61–65. 20. Liu, Y. & Bolen, D.W. (1995) The peptide backbone plays a dominant role in protein stabilization by naturally occurring osmolytes. Biochemistry 34, 12884–12891. 21. Qu, Y., Bolen, C.L. & Bolen, D.W. (1998) Osmolyte-driven contraction of a random coil protein. Proc. Natl Acad. Sci. USA 95, 9268–9273. 22. Baskakov, I., Wang, A. & Bolen, D.W. (1998) Trimethylamine- N-oxide counteracts urea effects on rabbit muscle lactate dehy- drogenase function: a test of the counteraction hypothesis. Biophys. J. 74, 2666–2673. 23. Jaenicke, R. & Bo ¨ hm, G. (1998) The stability of proteins in extreme environments. Curr. Opin. Struct. Biol. 8, 738–748. 24. Jaenicke, R. (2000) Do ultrastable proteins from hyperthermo- philes have high or low conformational rigidity? Proc. Natl Acad. Sci. USA 97, 2962–2964. 25. Giver, L., Gershenson, A., Freskgard, P. & Arnold, F.H. (1998) Directed evolution of a thermostable esterase. Proc.NatlAcad. Sci. USA 95, 12809–12813. 26. Wagner,G.&Wu ¨ thrich, K. (1979) Correlation between the amide proton exchange rates and the denaturation temperatures in globular proteins related to the basic pancreatic trypsin inhibitor. J. Mol. Biol. 130, 31–37. 27. Hernandez,G.,Jenney,F.E.,Adams,M.W.&LeMaster,D.M. (2000) Millisecond time scale conformational flexibility in a hyperthermophile protein at ambient temperature. Proc. Natl Acad. Sci. USA 97, 3166–3170. Ó FEBS 2003 Effect of DGP on rubredoxin dynamics (Eur. J. Biochem. 270) 4613 28.Lamosa,P.,Brennan,L.,Vis,H.,Turner,D.L.&Santos,H. (2001) NMR structure of Desulfovibrio gigas rubredoxin: a model for studying protein stabilization by compatible solutes. Extremophiles 5, 303–311. 29. Martins, L.O., Huber, R., Huber, H., Stetter, K.O., da Costa, M.S. & Santos, H. (1997) Organic solutes in hyperthermophilic Archaea. Appl. Environ. Microbiol. 63, 896–902. 30. Buck, M., Boyd, J., Redfield, C., MacKenzie, D.A., Jeenes, D.J., Archer, D.B. & Dobson, C.M. (1995) Structural determinants of protein dynamics: analysis of 15 N NMR relaxation measurements for main-chain and side-chain nuclei of hen egg white lysozyme. Biochemistry 34, 4041–4055. 31. Mandel, A.M., Akke, M. & Palmer, 3rd. A.G. (1995) Backbone dynamics of Escherichia coli ribonuclease HI: correlations with structure and function in an active enzyme. J. Mol. Biol. 246, 144–163. 32. Yun,S.,Jang,D.S.,Kim,D.H.,Choi,K.Y.&Lee,H.C.(2001) 15 N NMR relaxation studies of backbone dynamics in free and steroid-bound Delta 5–3-ketosteroid isomerase from Pseudomonas testosteroni. Biochemistry 40, 3967–3973. 33. Englander, S.W. & Kallenbach, N.R. (1984) Hydrogen exchange and structural dynamics of proteins and nucleic acids. QRev. Biophys. 16, 521–655. 34. Englander, S.W., Mayne, L., Bai, Y. & Sosnick, T.R. (1997) Hydrogen exchange: the modern legacy of Linderstrøm-Lang. Protein Sci. 6, 1101–1109. 35. Betz, S.F., Marmorino, J.L., Saunders, A.J., Doyle, D.F., Young, G.B. & Pielak, G.J. (1996) Unusual effects of an engineered disulfide on global and local protein stability. Biochemistry 35, 7422–7428. 36. Foord, R.L. & Leatherbarrow, R.J. (1998) Effect of osmolytes on the exchange rates of backbone amide protons in proteins. Biochemistry 37, 2969–2978. 37. Tabor, S. & Richardson, C.C. (1985) A bacteriophage T7 RNA polymerase/promoter system for controlled exclusive expression of specific genes. Proc. Natl Acad. Sci. USA 82, 1074–1078. 38. Kay, L.E., Torchia, D.A. & Bax, A. (1989) Backbone dynamics of proteins as studied by 15 N inverse detected heteronuclear NMR spectroscopy: application to staphylococcal nuclease. Biochem- istry 28, 8972–8979. 39. Piotto, M., Saudek, V. & Sklenar, V. (1992) Gradient-tailored excitation for single-quantum NMR spectroscopy of aqueous solutions. J. Biomol. NMR 2, 661–665. 40. Palmer, 3rd, A.G., Rance, M. & Wright, P.E. (1991) Intramolecular motions of a zinc finger DNA-binding domain from Xfin characterized by proton-detected natural abundance carbon-13 heteronuclear NMR spectroscopy. J. Am. Chem. Soc. 113, 4371–4380. 41. Tjandra, N., Feller, S.E., Pastor, R.W. & Bax, A. (1995) Rota- tional diffusion anisotropy of human ubiquitin from 15 NNMR relaxation. J. Am. Chem. Soc. 117, 12562–12566. 42. Hernandez, G. & LeMaster, D.M. (2001) Reduced temperature dependence of collective conformational opening in a hyper- thermophile rubredoxin. Biochemistry 40, 14384–14391. 43. Englander, S.W. & Mayne, L. (1992) Protein folding studied using hydrogen-exchange labeling and two-dimensional NMR. Annu. Rev. Biophys. Biomol. Struct. 21, 243–265. 44. Englander, S.W., Sosnick, T.R., Englander, J.J. & Mayne, L. (1996) Mechanisms and uses of hydrogen exchange. Curr. Opin. Struct. Biol. 6, 18–23. 45. Hiller, R., Zhou, Z.H., Adams, M.W. & Englander, S.W. (1997) Stability and dynamics in a hyperthermophilic protein with melting temperature close to 200 degrees C. Proc.NatlAcad.Sci. USA 94, 11329–11332. 46. Hvidt, A. & Nielsen, S.O. (1966) Hydrogen exchange in proteins. Adv. Protein. Chem. 21, 287–386. 47. Bai, Y., Milne, J.S., Mayne, L. & Englander, S.W. (1993) Primary structure effects on peptide group hydrogen exchange. Proteins 17, 75–86. 48. Jaenicke, R. (1997) What ultrastable globular proteins teach us about protein stabilization. Biochemystry (Moscow) 63, 312– 370. 49. Wang, T., Cai, S. & Zuiderweg, E.R.P. (2003) Temperature dependence of anisotropic protein backbone dynamics. J. Am. Chem. Soc. 125, 8639–8643. 50. Cheng, J.W., Lepre, C.A. & Moore, J.M. (1994) 15 NNMR relaxation studies of the FK506 binding protein: dynamic effects of ligand binding and implications for calcineurin recognition. Biochemistry 33, 4093–4100. 51. Farrow, N.A., Muhandiram, R., Singer, A.U., Pascal, S.M., Kay, C.M., Gish, G., Shoelson, S.E., Pawson, T., Forman-Kay, J.D. & Kay, L.E. (1994) Backbone dynamics of a free and phosphopep- tide-complexed Src homology 2 domain studied by 15 NNMR relaxation. Biochemistry 33, 5984–6003. 52. Yu, L., Zhu, C.X., Tse-Dinh, Y.C. & Fesik, S.W. (1996) Backbone dynamics of the C-terminal domain of Escherichia coli topo- isomerase I in the absence and presence of single-stranded DNA. Biochemistry 35, 9661–9666. 53. Harris, R.K. (1983) Nuclear Magnetic Resonance Spectroscopy. Pitman Pubhishing Inc., London. 54. Stivers, J.T., Abeygunawardana, C. & Mildvan, A.S. (1996) 15 N NMR relaxation studies of free and inhibitor-bound 4-oxalo- crotonate tautomerase: backbone dynamics and entropy changes of an enzyme upon inhibitor binding. Biochemistry 35, 16036– 16047. 55. Palmer, 3rd. A.G. (2001) NMR probes of molecular dynamics: overview and comparison with other techniques. Annu. Rev. Bio- phys. Biomol. Struct. 30, 129–155. 56. Gekko, K. & Timasheff, S.N. (1981) Mechanism of protein sta- bilization by glycerol: preferential hydration in glycerol-water mixtures. Biochemistry 20, 4667–4676. Supplementary material The following material is available from http://blackwell publishing.com/products/journals/suppmat/EJB/EJB3861/ EJB3861sm.htm Table S1. Relaxation parameters for D. gigas rubredoxin at 303.15 K. Table S2. Model-free parameters for D. gigas rubredoxin at 303.15 K. Table S3. Logarithm of the exchange protection factors as a function of temperature. 4614 P. Lamosa et al. (Eur. J. Biochem. 270) Ó FEBS 2003 . Protein stabilization by compatible solutes Effect of diglycerol phosphate on the dynamics of Desulfovibrio gigas rubredoxin studied by NMR Pedro. results of the two sets of experiments together, and in view of the strong stabilizing effect the solute confers upon this protein and the lack of structural