The Effects of Ionic Strength on Protein Stability: The Cold Shock Protein Family

The Effects of Ionic Strength on Protein Stability: The Cold Shock Protein Family

doi:10.1016/S0022-2836(02)00259-0 available online at on w B J. Mol. Biol. (2002) 319, 541–554 The Effects of Ionic Stre...

397KB Sizes 0 Downloads 3 Views

doi:10.1016/S0022-2836(02)00259-0 available online at on

w B

J. Mol. Biol. (2002) 319, 541–554

The Effects of Ionic Strength on Protein Stability: The Cold Shock Protein Family Brian N. Dominy1, Dieter Perl2, Franz X. Schmid2 and Charles L. Brooks III1* 1

Department of Molecular Biology, TPC6, The Scripps Research Institute 10550 North Torrey Pines Road La Jolla, CA 92037, USA 2

Biochemisches Laboratorium Universitaet Bayreuth D-95440 Bayreuth, Germany

Continuum electrostatic models are used to examine in detail the mechanism of protein stabilization and destabilization due to salt near physiological concentrations. Three wild-type cold shock proteins taken from mesophilic, thermophilic, and hyperthermophilic bacteria are studied using these methods. The model is validated by comparison with experimental data collected for these proteins. In addition, a number of single point mutants and three designed sequences are examined. The results from this study demonstrate that the sensitivity of protein stability toward salt is correlated with thermostability in the cold shock protein family. The calculations indicate that the mesophile is stabilized by the presence of salt while the thermophile and hyperthermophile are destabilized. A decomposition of the salt influence at a residue level permits identification of regions of the protein sequences that contribute toward the observed salt-dependent stability. This model is used to rationalize the effect of various point mutations with regard to sensitivity toward salt. Finally, it is demonstrated that designed cold shock protein variants exhibit electrostatic properties similar to the natural thermophilic and hyperthermophilic proteins. q 2002 Elsevier Science Ltd. All rights reserved

*Corresponding author

Keywords: protein folding; electrostatics; salt effects; cold shock proteins; thermophilic proteins

Introduction Salt permeates biological systems and is crucial for life. Among its many roles, cellular control over salt concentration prevents osmotic pressure from lysing cells by balancing the internal solute concentration with the external ionic concentration.1 The intrinsic intracellular salt concentration also acts to shield ionic interactions affecting macromolecular stability as well as intermolecular binding reactions. The intracellular and extracellular concentrations of ionic species, such as sodium, potassium, or chloride ions, vary with the organism and environment but fluctuate near 1– 200 mM.1 At these concentrations, salt can have a significant impact on the stability of macromolecules such as highly charged nucleic acid Abbreviations used: PB, Poisson – Boltzmann; Bs-CspB, cold shock protein B from B. subtilis; Bc-Csp, cold shock protein from B. caldolyticus; Tm-Csp, cold shock protein from T. maritima. E-mail address of the corresponding author: [email protected]

structures and more weakly charged protein structures. Understanding the effects of salt is important for understanding the thermodynamic properties of biological systems. It has been long understood that salt has a significant impact on the stability of nucleic acid structures.2 The large net negative charge from the phosphate backbone destabilizes nucleic acid structures, however the inclusion of salt effectively screens the inherent repulsive interactions. Nucleic acid structures are thereby stabilized at increased ionic strength. In addition to changes in stability, salt also affects the conformational equilibrium in DNA. The common B-form DNA can be converted into A-form containing a smaller major grove and shorter phosphate – phosphate distances when high salt concentrations are present.3 Furthermore, the Z-form of DNA is stabilized by the presence of salt.4 The stability of proteins is also affected by salts. Monovalent salts, such as sodium chloride, often affect protein stability by modifying the ionic strength of the solution, which overall can be slightly stabilizing or destabilizing,5,6 depending

0022-2836/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved


on the nature of the specific charge distribution within the protein.7 Generalizations may be made regarding proteins from organisms that have evolved to live at high salt concentrations. These halophilic proteins typically adapt through an increase in the number of negatively charged residues displayed on their surface.8,9 Halophilic proteins are often destabilized when transferred into a medium of low salt concentration9,10 because the negative charges, which are screened in solutions containing salt, repel each other in environments containing little salt or no salt.9,10 In addition to halophiles, other extremophilic proteins adapt to their environments using electrostatic mechanisms. Psychrophilic (cold-adapted) proteins are known, in some cases, to adapt to low temperature by incorporating more negative charges on their surface.11 This is similar to the more commonly observed halophilic adaptations described above. Thermophilic and hyperthermophilic proteins often occur with increased numbers of both acidic and basic residues, which are assumed to create more salt bridges on their surface.12 The intrinsic desolvation penalty associated with including charged residues in a folded protein is somehow overcome by the enhanced Coulomb-driven residue pairing which stabilizes the folded state. These examples illustrate the significant role of electrostatics in extremophilic protein adaptations. Coulombic interactions at the protein surface are efficiently screened in the presence of salt. Their contributions to the thermodynamic stability of a protein can thus be determined experimentally by measuring unfolding transitions as a function of salt concentration. However, interpreting the results from such studies is not always straightforward given the complex nature of salt effects. In some cases, ions stabilize proteins by highaffinity binding to specific sites. This ligandinduced stabilization is ion-specific and usually observed in the range of 0 –0.2 M salt.13,14 Bulk ionic strength effects play a role in screening surface charge–charge interactions.10 Hofmeister effects, which occur at still higher salt concentrations, may strengthen the hydrophobic force by increasing the surface tension of the solvent, or stabilize peptide dipoles through specific ionic interactions.5,15 In the absence of site-specific ion binding, differential salt effects in the 0 –1 M range reflect primarily differences in Coulombic interactions, e.g. between natural or designed variants of a protein. The Hofmeister effects that occur in parallel are usually insensitive to local changes in amino acid sequence. Bulk ionic strength effects primarily related to electrostatic screening may be studied using the Poisson–Boltzmann formalism.10,16 The Poisson– Boltzmann (PB) equation relates the electrostatic potential of a system to the fixed charge distribution of that system and the mobile charge distribution of salt. The mobile charges are modeled through a Boltzmann distribution with respect to the electrostatic potential generated by the fixed charges.17 Using this approach, a linear-

Salt Effects and the Cold Shock Protein Family

ized form of the PB equation can be solved with numerical finite difference methods to determine the electrostatic energy of a residue in an unfolded reference state as well as in the folded protein. The difference in this energy (properly normalized) yields the residue-wise contribution to the electrostatic free energy of folding. By applying this model at a variety of salt concentrations, one can determine the effect of salt on the contribution of individual residues to the electrostatic stability of the protein. Here we explore the cold shock protein family as a model system to probe the role of salt in modulating protein stability, as well as to examine the contribution of electrostatics toward thermostability. Cold shock proteins are small, monomeric b-barrel proteins for which thermodynamic and structural studies have been performed on mesophilic, thermophilic, and hyperthermophilic homologs.18 – 20 Their small size in addition to the lack of disulfide linkages, cis-proline residues, tight-binding ligands, and their reversible twostate folding transitions make them convenient systems for studying a wide range of thermodynamic properties. In particular, experimental studies have been performed to elucidate differences in the folding landscapes of these proteins,18,19,21 as well as the origin of their differential thermostability.20 In accordance with a recent structural survey,12 the thermophilic and hyperthermophilic cold shock proteins appear to derive their enhanced stability in part through optimized electrostatic interactions. These proteins comprise a small, convenient system to study the electrostatic properties of fold stabilization. Here we develop a model for the residue-wise contribution to electrostatic stability, based on numerical finite difference solutions of the Poisson–Boltzmann continuum electrostatic theory. We examine the effects of salt on the mesophilic, thermophilic, and hyperthermophilic cold shock proteins from Bacillus subtilis (Bs-CspB), B. caldolyticus (Bc-Csp), and Thermotoga maritima (Tm-Csp). We explore the electrostatic properties of wild-type cold shock proteins from this thermophilic series, as well as designed thermophilic variants. Experimental studies are performed on the same systems to complement and confirm the theory. Using the resulting theoretical model, we provide an explanation for the salt effects observed in the experimental studies as well as generate novel predictions. We find that the change in stability with addition of salt is correlated with thermostability. In addition, an analysis of the residue-wise contribution to the salt response in the three wild-type proteins explains the effect of point mutations on the overall halophilic behavior observed in Bs-CspB, and demonstrates that it is due in large part to the significant net charge of this protein. The effect of net charge on the protein sequence also provides a rationalization for the relative insensitivity of Bc-Csp to point mutations (with regard to salt effects) and predicts that the


Salt Effects and the Cold Shock Protein Family

∆∆G (kJ/mol)

Results and Discussion



The electrostatic contribution to the folding free energy was computed using the Poisson – Boltzmann continuum electrostatic formalism to evaluate the electrostatic free energy in the folded cold shock proteins and corresponding models for the unfolded states. Applying this protocol at multiple salt concentrations allows one to evaluate the effect of bulk ionic strength on the electrostatic free energy contribution to protein stability.

4 2 0 -2 5

The cold shock protein family

∆∆ Gelec (kJ/mol)

(B) 0 -5 -10

Mesophile Thermophile Hyperthermophile

-15 10

∆∆ G (kJ/mol)

(C) 5


-5 0

0.2 0.4 0.6 0.8 Salt Concentration (M)


Figure 1. Electrostatic stability as a function of salt concentration. The values were computed using a twodielectric FDPB model with the internal dielectric of 4. The experimental results are shown in (a), two sets of theoretical calculations are shown in (b) and (c). The effect of ionic strength relative to a 0.1 M monovalent salt buffer is shown in (b). In (c) a simple cavity model is used to account for salting-out effects with regard to hydrophobic residues and is also shown relative to a 0.1 M monovalent salt buffer. Experimental data presented in (a) represent data at 60 8C.

majority of point mutations in Tm-Csp will weaken the sensitivity of this protein toward salt. Finally, we show that designed variants of cold shock protein B exhibit electrostatic properties similar to wild-type thermophiles. This is in agreement with experiment as well as the intentions of the design protocol (B.N.D., H. Minoux & C.L.B.,III, unpublished results). We demonstrate the general applicability of our model to understand the connection between the charge distribution present in biological macromolecules and the effect of ionic strength on the stability of these structures.

We focus our approach first on the wild-type proteins. We find that the mesophilic protein (Bs-CspB) is stabilized by the presence of salt while both the thermophile (Bc-Csp) and the hyperthermophile (Tm-Csp) are destabilized. The results from our theoretical model are in excellent agreement with experimental data.18,20 This is illustrated in Figure 1, where we compare the experimental salt profiles for the mesophilic and thermophilic protein with those from the theoretical model. Calculations regarding the hyperthermophilic cold shock protein are testable predictions since accurate experimental data have not yet been collected for this system. Figure 1(a) shows the experimental data20 while Figure 1(b) illustrates the theoretical findings. Our results are plotted relative to an ionic strength consistent with the 0.1 M sodium cacodylate buffer used in the experimental studies in order to maintain a pH of 7. As the salt concentration increases, the theoretical model predicts a saturating effect where no further electrostatic shielding is feasible. This is not evident in the experimental data; salt has a stabilizing effect at high concentrations. This difference can be attributed in part to Hofmeister effects, or specifically salting-out effects. As discussed previously, salt can increase the surface tension of an aqueous solution, thereby increasing the hydrophobic force and stabilizing the protein core. Thus at higher salt concentrations, although the electrostatic effects saturate, the increase in solvent surface tension stabilizes the folded or collapsed protein conformation. To illustrate this influence, we extend the continuum model with a surface area-based cavity term that crudely accounts for the enhanced hydrophobicity of the solvent with increasing salt. The inclusion of salting-out effects with this simple cavity model does not change the qualitative behavior for salt concentrations between 0 and 0.5 M, which most closely represents physiological conditions. However, it yields changes in stability versus increasing salt more qualitatively and quantitatively similar to experiment (Figure 1(c)). The qualitative agreement between theory and experiment demonstrated in Figure 1 is very good, however, the two sets of results are not expected to match perfectly. One reason is that specific ionic


Salt Effects and the Cold Shock Protein Family

model by Elcock.23 These findings are outlined in Methods. Another factor that can affect the quantitative accuracy of our model is the salt dependence of the solvent dielectric constant. A salt-dependent solvent dielectric constant was examined to explore this issue. However, very little change was observed at the low salt concentrations of interest here (data not shown).10,24 Deviations from quantitative accuracy are not likely to arise from changes in the solvent’s dielectric constant. The effect of salt and thermostability

Figure 2. The change in the electrostatic free energy of folding upon the addition of 0.1 M NaCl separated into contributions from individual residues. In correspondence with all Figures here, positive values indicate residues that contribute to the stabilization of the protein due to the addition of salt: (a) B. subtilis (mesophile), (b) B. caldolyticus (thermophile) and (c) T. maritima (hyperthermophile).

Figure 1 demonstrates that the response to changes in salt is correlated with thermostability. As the thermostability of the cold shock protein increases, so does the destabilizing effect of salt. In order to understand why this occurs for this protein family, it is helpful to consider a simpler model for the effects of bulk ionic strength. Case and co-workers25 recently described a hybridization of the generalized Born theory for continuum solvent polarization effects with the Debye – Hu¨ckel equation for ionic strength effects (equation (1)). This equation simply illustrates the effects of salt on electrostatic solvation within the continuum dielectric approximation:  1 0 ij exp 2k f X X D GB 21 @ 1 2 A qi qj DGsolv GB ¼ ij 2 i 1 1 water solute f j GB

interactions are not fully represented in the theoretical model;22 these effects are minimized by focusing on low, physiological salt concentrations. However, contributions from these components at low ionic strength will still impact the quantitative agreement between experiment and theory. In addition to the neglect of specific ion effects and Hofmeister influences, the unfolded reference state chosen in the theoretical model is not necessarily representative of the true denatured state ensemble for the proteins being studied. The unfolded state model used in our theoretical analysis is presumed to have no long-range structure, only sequential nearest neighbor residues interact. The thermally denatured proteins at 60 8C, where the experimental data were represented, are probably not extended random coils, as assumed in the calculations. Furthermore, the mesophilic and thermophilic cold shock proteins, which have significantly different midpoints of thermal unfolding, may differ in the distribution of conformations at 60 8C. Consequently, the theoretical model probably provides an upper bound on the experimentally determined effects of salt on protein stability. This conclusion is reinforced by additional theoretical calculations utilizing different models for folded and unfolded states, including thermally fluctuating native state models and a recently described “compact” unfolded state


where fGB

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi !ffi u u 2r2ij ¼ tr2ij þ ai aj exp 4ai aj ð1Þ

Using this model, the total electrostatic free energy is written as the sum of the generalized Born solvation term and the direct Coulomb energy of interaction between the charged sites in the protein (equation (2)): DGelec ¼ DECoul þ DGsolv GB


It is generally true that the interaction component of solvation between any pair of charges is smaller in magnitude and opposite in sign to the direct Coulomb interaction, and serves to screen the Coulomb interaction energy. The influence of salt can be modeled as an exponential factor that exaggerates this screening effect. This factor is a function of the salt concentration and the distance between charges. As the concentration of salt increases and/or the distance between charges increases, the screening effect of salt increases. Therefore, at low salt concentrations the Coulomb energy is screened, with a greater screening of more distant interactions. This is a key point in understanding, more generally, how salt and the protein charge distribution interact to affect specific halophilic or halophobic behavior.

Salt Effects and the Cold Shock Protein Family




Figure 3. The change in the electrostatic free energy of folding upon the addition of 0.1 M NaCl separated into contributions from individual residues and projected onto the protein surface: (a) B. subtilis (mesophile), (b) B. caldolyticus (thermophile) and (c) T. maritima (hyperthermophile). Blue regions indicate residues contributing the halophilic character to the protein and the red regions indicate residues that contribute to the destabilization of the protein in the presence of salt. Right-hand views represent 1808 rotation of left view.

Microscopic basis for the salt effect on the cold shock protein family With this conceptual picture for the influence of ionic strength on stability, we may examine the cold shock proteins more closely to identify the origins of differences in their sensitivity to salt. Two factors contribute toward the observed trends. The first factor is the optimization of salt bridges and hydrogen bonds, or short-range electrostatic interactions. These electrostatic components are highly favorable in the thermostable species, but less optimal in the mesophilic protein. The greater thermal sensitivity of the mesophilic protein has been shown to partly result from an unfavorable Coulombic interaction between E3 and E66.20 Our theoretical model also predicts this unfavorable interaction to be largely responsible for the halophilic nature of Bs-CspB. The dissection of the salt effect into per-residue contributions is shown in Figure 2 and graphically projected onto the protein surface in Figure 3. Added salt tends to screen the repulsive interaction between E3 and E66 and thereby stabilize the protein. Favorable interactions


between residues are also screened by salt, the strongest of which involve amino acid residues Lys5 and Lys7 near the N terminus. An electrostatic analysis of the wild-type cold shock proteins was reported recently by Sanchez – Ruiz and Makhatadze.26 Their work explored the consequences of direct electrostatic interactions without exploration of salt effects as described here and utilized the simple, approximate Tanford– Kirkwood model to evaluate the charge–charge interactions for each residue in these proteins. Qualitative aspects of their findings coincide with those we find, e.g. as presented in Figure 2. They identify the repulsive interactions between the N and C-terminal glutamate residues as being keys to the lower stability of mesophilic CspB as well as others. However, our approach, which directly probes the effect of ionic strength on protein stability, moves beyond a simple electrostatic interaction analysis and provides quantitative information regarding the influence of point mutations on salt-dependent stability that is directly testable via experiment. These results are discussed in detail below. The second factor contributing to the salt-dependent stability of the cold shock protein family is net charge, or long-range electrostatic interactions. On close inspection of Figures 2 and 3, it becomes apparent that residues contributing to the halophilic nature of Bs-CspB are always negatively charged, while positively charged residues contribute to the destabilization of the protein due to salt. Conversely, in the hyperthermophile both negatively and positively charged residues contribute to the destabilizing effect of salt on the protein’s stability. The origin of this difference in behavior can be explained based on the net charge of the two systems. The mesophilic protein, Bs-CspB, has a net charge of 2 6. While nearest neighbor interactions in proteins have often evolved to be favorable regardless of the acidity or basicity of the residue in question, long-range interactions may have developed to optimize other properties. In the negatively charged Bs-CspB, acidic residues experience repulsive Coulombic interactions from distant negatively charged residues. This can be significant with respect to the effects of salt for two reasons. First, the larger number of repulsive “long-range” interactions relative to nearestneighbor contacts may dominate the total Coulomb interaction energy of a given acidic residue. Second, since salt preferentially screens long-range interactions over short-range terms, it also alters the balance of favorable and unfavorable interactions and contributes to the influence of the overall protein charge on a residue’s contribution to stability as a function of salt. The converse applies to positive charges, since these will interact favorably with the net negative charge of Bs-CspB. The result is that the large overall charge of Bs-CspB leads to the halophilic character of acidic residues and the halophilicity of the protein.


Salt Effects and the Cold Shock Protein Family

The interplay of net charge and salt screening

Figure 4. The change in the electrostatic free energy of folding upon the addition of 0.1 M NaCl separated into contributions from individual residues versus the protein charge. We examine the (a) wild-type (WT) mesophile and the mesophilic protein where (b) þ4 and (c) þ6 charge has been “smeared” uniformly over all atoms in the protein. In the mesophile (WT) system the net charge is 26, in the mesophile (þ4) system the net charge is 2 2 (identical to the thermophile), and in the mesophile (þ6) system the net charge is 0 (identical to the hyperthermophile).

The cold shock protein from the hyperthermophile has a net charge of zero. The consequence of this is that long-range terms largely cancel out and contribute little with regard to the salt effect. Short-range terms dominate and, as mentioned previously, the evolutionary constraints upon proteins have forced them to have favorable shortrange interactions. Therefore, each charged residue contributes to destabilizing the protein with added salt. The effect is fairly uniformly distributed throughout the structure of the hyperthermophile, as seen in Figures 2(c) and 3(c). The thermophile represents an intermediate case. In the majority of the sequence, individual residues do not contribute to the salt-dependent stability of Bc-Csp. The net charge on this protein (2 2) is much smaller than the mesophile (2 6) but more negative than the hyperthermophile (0). The long-range terms are not as strong as in the mesophile and the short-range terms are not as strongly optimized as in the hyperthermophile. The result is an insignificant salt influence on the majority of the structure. Figures 2(b) and 3(b) illustrate the distribution of salt sensitivity throughout the structures of the three cold shock protein species. The salt-destabilized N terminus is the only significant feature in the thermophilic protein’s profile.

To more fully explore the role of net charge and long-range interactions in determining the salt dependence of protein stability, we artificially reduced the net charge on the mesophilic protein (Bs-CspB) from 2 6 to 2 2 and 0. This was accomplished by uniformly applying a small charge to every atom in the system to yield the desired molecular charge. This approach strongly influences long-range interactions while only minimally perturbing close interactions. The contributions of individual residues to the salt effect on the protein were then analyzed. The results shown in Figure 4 demonstrate that many of the qualitative differences in the salt behavior observed in the three cold shock protein B charge variants are due to the influence of net charge and longer range interactions. The mesophile with the net charge reduced to 2 2 shows characteristics very similar to the natural thermophile (Bc-Csp). Specifically, most of the sequence makes little contribution to the overall sensitivity of the protein to added salt. The strong short-range interactions in the N terminus and extreme C terminus are also still present. When the charge is completely neutralized, the resulting system behaves similarly to the hyperthermophile. Specifically, most of the residues contribute to an overall destabilizing effect due to salt. In fact, this protein is no longer halophilic, despite the remaining repulsive interaction involving the glutamic acid at position 3. We can also explore the interplay of long-range Coulombic interactions and salt by examining the correlation between these two terms in systems of varying net charge. To do this, we compute the net Coulombic contribution to protein stability for each residue as a function of how many neighboring interactions are included. The residue-wise neighbor contributions to the Coulomb energy are computed using a cut-off distance to filter the interactions, beyond which a given residue does not feel the influence of its neighbors. We also compute the residue-wise change in the protein stability due to the addition of an increment of salt (from 0 M to 0.1 M) using the Poisson– Boltzmann equation. The correlation (computed as a linear correlation coefficient) between the residuewise Coulombic contribution to stability and the change in stability per residue on addition of salt form the basis of our separation of short-range versus long-range effects in altering the salt-dependent stability of proteins. In general, this correlation should be negative since the effect of salt on favorable interactions is unfavorable with respect to protein stability. As is apparent from Figure 5 by the small values of the correlation coefficient for short neighbor cut-off distances, the correlation between the influence of salt (on a perresidue basis) and a residue’s Coulombic contribution to stability arising from its nearest neighbors is weak. As Coulomb effects from longerrange interactions are included, however, the corre-


Salt Effects and the Cold Shock Protein Family




Bs-CspB Bc-Csp Tm-Csp Bs-CspB (neutralized) Tm-Csp (Net -6)

-0.4 -0.6 -0.8 -1 6




25 No Cutoff

Cutoff (Å)

Figure 5. Correlation between the effect of salt on residue-wise protein stability and the Coulomb interaction energy between the residue and neighboring protein residues within a cut-off radius given as the abscissa.

lation increases dramatically in the case of the mesophilic cold shock protein. This is because the most significant Coulombic influences (for each residue) in this negatively charged protein arise from longer-range interactions that correlate most strongly (in a negative sense) with the longerrange screening effects of salt. This is also observed in the hyperthermophilic cold shock protein with a smeared negative net charge. In a system with zero net charge, long-range Coulombic interactions do not contribute much to the total interaction energy of a given residue. However, the effect of salt is still to screen long-range components more effectively than short-range terms. The result is that the salt effect remains less correlated in the hyperthermophilic protein and neutralized mesophilic protein, even when long-range Coulombic interactions are considered. This simple physical picture demonstrates that salt can have a more complex screening behavior than predicted simply based on Coulomb interaction energies. The role of net charge on thermostability Thermostability is linked to the optimization of both long-range and short-range electrostatic interactions. The correlation between thermostability and the number of potential salt bridge interactions has already been noted in a structural survey of 25 protein families containing mesophilic and thermophilic species.12 A more recent analysis of charge– charge interactions calculated using a Tanford– Kirkwood model applied to protein crystal structures also demonstrated a correlation between electrostatic interactions and thermostability.26 – 28 The penalty from solvation due to the self-polarization associated with charge burial in these systems is apparently outweighed by the direct Coulomb interaction energy arising from saltbridges and this serves to stabilize thermostable proteins.29 A consequence of stabilizing a protein through favorable electrostatic interactions is that salt will have a significantly destabilizing effect. We have also verified that this idea remains true in other thermophilic series of homologs, such as

chemotaxis protein Y (CheY) and RNAse H. In these systems (data not shown) the (hyper)thermophile is demonstrated to have the greatest sensitivity to added salt relative to mesophilic species. This suggests that the correlation between salt sensitivity and thermostability observed in the cold shock protein family may be a more general phenomenon. Although not previously noted, a trend also exists in which hyperthermophilic proteins are typically more electroneutral than homologous proteins of mesophilic origin. A recent paper by Szilagyi & Zavodszky describes a structural survey of 29 thermophilic protein subunits and 64 homologous mesophilic proteins.12 Of these 29 thermophilic structures, 13 are considered hyperthermophilic, originating from organisms with optimal growth temperatures over 70 8C. We note that 63% of the mesophilic proteins had an absolute net charge that was higher than the homologous hyperthermophilic protein. On average, hyperthermophilic proteins had an absolute net charge 2.75 electron units lower (closer to neutrality) than the corresponding mesophilic proteins. Although this trend is not as significant as the correlation between thermostability and salt bridge interactions discussed by Szilagyi & Zavodszky, it does suggest that the optimization of long-range electrostatic terms could be a possible evolutionary mechanism for enhanced thermostability. It should also be noted that neutrality and protein stability have already been linked.30 Pace and co-workers demonstrated that ˚ from the nearest removal of an acidic residue 8 A charged group in a net negatively charged protein enhanced the stability of that protein by 0.5 to 1.1 kcal/mol.31 In a more recent paper, Pace and colleagues have illustrated that net charge is not a strong factor in determining the stability of RNase SA.32 Rather, pKa shifts between the folded and unfolded conformations may play a more important role. Although these results suggest caution in the interpretation of the role of net charge on protein stability, the optimization of long-range electrostatic interactions may be indicated in the natural evolutionary process behind hyperthermophilic proteins. The effect of salt on cold shock protein point mutants We were further interested in comparing our results for salt-dependent stability to experimental data for point mutants of the cold shock proteins. The most significant single point mutation according to the data shown in Figure 2 is E3R in the Bs-CspB. This mutation alone is predicted to eliminate the halophilic response, or salt stabilizing effect, in this system. Further, our theoretical analysis suggests that this single mutation will cause the protein to be nearly as sensitive as the thermophile to added salt, as seen in Figure 6(a). Both of these suggestions are confirmed by experiment. The


Salt Effects and the Cold Shock Protein Family

4 ∆∆ Gelec (kJ/mol)

(A) 2 0 -2 -4

∆∆ G (kJ/mol)



Bc-Csp (wild type) Bc-Csp (R3E) Bc-Csp (R3L) Bs-CspB (wild type) Bs-CspB (E3R) Bs-Csp (E3L) Bs-Csp (E66L)

4 2 0 -2

∆∆ G (kJ/mol)



6 4 2 0 -2 0

0.5 Salt Concentration (M)


Figure 6. Calculated salt effects on the electrostatic contribution to stability in point mutants of the mesophilic (Bs-CspB) and thermophilic (Bc-Csp) cold shock proteins: (a) calculated using our theoretical model and (b) from experiment. Bs-CspB and Bc-Csp refer to structures that were treated in the same manner as the mutants. Residue 3 is minimized in the context of a rigid protein structure. Experimental data presented in (b) represent data at 60 8C.

electrostatic stability of these mutant proteins as a function of the salt concentration is shown (from theory) in Figure 6(a) and from experiment in Figure 6(b). Simply replacing E3 with a hydrophobic residue will significantly weaken the protein’s halophilic response. The electrostatic component of the salt-dependent stability is completely eliminated according to our theoretical model. The Hofmeister effects present in the experimental results, however, are weakly stabilizing in the presence of NaCl. The empirical surface area model demonstrates the same qualitative trend. The repulsive contact between E3 and E66 contributes significantly to the halophilic character of Bs-CspB, however these residues do not respond equally to added salt. From Figure 2, it is apparent that point mutations of E66 to hydrophobic residues will have a lesser impact with regard to salt-dependent stability changes, in contrast to mutations in E3. This mutation also illustrates that

the E3– E66 contact is not solely responsible for the halophilic behavior demonstrated by the mesophilic cold shock protein. According to the theoretical analysis, the E3L and E66L mutants are stabilizing via Coulombic interactions, however, the long-range interactions between the negative charges (present in the native protein) are preferentially screened by salt. This is an example where Coulombic energy terms do not resolve the effects of ionic strength on protein stability. This is again in-line with experimental findings shown in Figure 6(b). The experimental results also demonstrate that the E66L point mutant is more weakly stabilized by salt relative to the wild-type mesophilic cold shock protein. The effect of salt, however, is not attenuated as much as with the E3L mutant. In addition to the strong N and C-terminal repulsive interactions, a loop region (comprising residues 19– 25) in Bs-CspB is also suggested by the theoretical model to show exceptional halophilicity. There are four acidic residues (two glutamate residues and two aspartate residues) within this loop that contribute to the halophilicity of Bs-CspB. Mutations in D24 to the corresponding Bc-Csp serine residue are predicted to reduce the mesophilic protein’s halophilic nature. The alternate mutation in the thermophile (S24D) does not have a significant impact on the protein’s sensitivity toward salt (data not shown). Experiment again confirms these findings.20 This is because these residues contribute to the halophilicity of Bs-CspB through long-range interactions with the net charge of the mesophilic protein. Figure 4 demonstrates that when the mesophilic protein’s net charge is reduced to that of the thermophile (2 2) the halophilic contributions from this loop are significantly attenuated. Therefore, the mutations in the corresponding loop within the thermophile, without more significant changes to the net charge of the protein, will not impact the salt-dependent stability. Point mutations in the N-terminal region, where short-range interactions contribute significantly to the salt-dependent behavior of the cold shock protein family, do significantly impact the thermophilic protein’s sensitivity toward bulk ionic strength. This has been shown experimentally with R3 (see Figure 6(b)).20 The effects of K5 and K7 remain predictions of our theoretical model. Although no experimental data currently exist regarding the effects of salt on Tm-Csp stability, predictions may be made using our theoretical model. In all of the proteins from the cold shock family, the model predicts that the most significant mutations affecting the salt-dependent stability will be in the N terminus. In the hyperthermophile, however, the model predicts that hydrophobic substitution in many of the positions along the sequence will impact the sensitivity toward salt. In fact, these mutations are predicted to almost uniformly reduce the sensitivity of this protein toward bulk ionic strength.


Salt Effects and the Cold Shock Protein Family


Figure 7. Sequences corresponding to designed proteins.

The effect of salt on designed sequences To further assess the predictive power of the theoretical model, two mutant sequences, mutant A and mutant B, were designed starting from the mesophilic sequence with the aim of optimizing the electrostatic contributions to stability (B.N.D., H. Minoux & C.L.B.,III, unpublished results). In the design of these sequences, E3 was not altered, avoiding the trivial single point mutation that affects the halophilic character of Bs-CspB. Two variants of mutant B (MLE, MELE) were created based on an original designed sequence. Methionine is placed at the N terminus and an E25D mutation is made in both variants in order to improve expression levels. Mutant A and the MELE variant of mutant B have a net zero charge while the MLE variant of mutant B has a þ 1 net charge. The sequences of the mutants are shown in Figure 7. The genes for the designed proteins were assembled from mutated oligonucleotides by PCR, expressed in Escherichia coli, and purified. In 0.1 M sodium cacodylate buffer and at pH 7.0, all three mutants were more stable than wild-type Bs-CspB, with increases in tM ranging from 4.8 to 7.4 deg.C (Table 1). However, they are all strongly stabilized by electrostatics, as indicated by the salt-dependent stability curves plotted in Figure 8. Mutant A shows the same dependence on NaCl concentration as the thermophilic reference protein Bc-Csp. Mutant B variants are even more stabilized electrostatically, and thus their stability decreases more strongly with salt concentration between 0 and 0.5 M NaCl. As a consequence, all three variants are less stable than wild-type Bs-CspB in 2 M NaCl (Table 1). From the differences in the stabilities of the wild-type and mutant proteins at

0 and 2 M NaCl we can estimate that the chargeoptimized variants are stabilized by 6 – 9 kJ/mol of favorable Coulombic interactions relative to the wild-type protein. In each case, the sensitivity of the mutants to added salt was successfully altered from halophilic to halophobic as shown in Figure 8(a). This is reflected in both the theoretical and experimental results. The reduction in the net charge, in addition to mutations in E66, permit short-range, favorable Coulombic interactions to dominate. The E66K replacement in the variants relieves the repulsion between Glu3 and Glu66, which destabilizes wildtype Bs-CspB by 4 kJ/mol.20 The remaining 2– 5 kJ/mol observed in the experiment may be due to the reduction in the overall net charge, long-range repulsive interactions, or additional stabilizing salt bridge interactions. A per-residue separation of the salt effects on the stability of the designed sequences is shown in Figure 9. To summarize, we have demonstrated that thermostability is correlated with the sensitivity to added salt in the cold shock protein family. This is due in part to the screening of short-range interactions, which are generally found to be highly optimized in thermostable proteins. In addition, the glutamic acid at position 3 is found to be the single most important residue in providing the halophilicity of the mesophile. This residue has also been linked with the mesophilic protein’s reduced thermostability.20 We propose that the net negative charge on Bs-CspB is a significant determinant of each charged residue’s contribution to the salt dependence of protein stability. The dominant short-range interactions (which tend to be favorable) within the hyperthermophile are predicted to effectively optimize this protein’s sensitivity toward salt such that any mutation,

Table 1. Experimental characterization of the salt-dependent stability of wild-type cold shock proteins and designed homologs

Bs-CspB (wild-type) Bc-Csp (wild-type) Csp-Mutant A Csp-Mutant B (MLE variant) Csp-Mutant B (MELE variant)

tM (8C)

DH (tM) (kJ mol21)

DGD (60 8C) (kJ mol21)

DDGD (60 8C) (kJ mol21) (Mut.-WT)

tM in 2 M NaCl (8C)

DH (tM) in 2 M NaCl (kJ mol21)

DGD in 2 M NaCl (60 8C) (kJ mol21)

DDGD in 2 M NaCl (60 8C) (kJ mol21)

53.6 ^ 0.1

193 ^ 2


71.4 ^ 0.1

205 ^ 2


76.9 ^ 0.1

245 ^ 5



84.4 ^ 0.1

257 ^ 3



58.4 ^ 0.1

173 ^ 3



66.0 ^ 0.1

168 ^ 2



61.0 ^ 0.1

149 ^ 1



63.2 ^ 0.1

159 ^ 1



59.8 ^ 0.1

177 ^ 3



61.3 ^ 0.1

164 ^ 1




Salt Effects and the Cold Shock Protein Family


∆∆ G (kJ/mol)

6 4 2 0 -2 5 ∆∆Gelec (kJ/mol)

(B) Mesophile Thermophile Mutant A Mutant B (MELE) Mutant B (MLE) Hyperthermophile

0 -5

-10 -15 10

∆∆ G (kJ/mol)

(C) 5

Figure 9. The residue-wise contributions to protein stability with added salt in the three designed sequences. Calculations represent the change in free energy upon the addition of 0.1 M salt.


-5 0

0.2 0.4 0.6 0.8 Salt Concentration (M)


Figure 8. Electrostatic stability as a function of salt concentration comparing wild-type cold shock proteins with designed sequences (mutant A and mutant B variants). The experimental results are shown in (a), two sets of theoretical calculations are shown in (b) and (c). The effect of ionic strength relative to a 0.1 M monovalent salt buffer is shown in (b). In (c) a simple cavity model is used to account for salting-out effects with regard to hydrophobic residues and is also shown relative to a 0.1 M monovalent salt buffer. Experimental data presented in (a) were collected at 60 8C.

particularly among charged residues, is likely to reduce its sensitivity to increases in bulk ionic strength. We also predict that the thermophile, due to weaker short-range terms and a small negative net charge, will be insensitive to mutations in the middle of the sequence with regard to its salt sensitivity. Finally, we have shown that the designed thermophilic variants exhibit electrostatic properties consistent with the wild-type thermophilic and hyperthermophilic cold shock proteins as well as the intent of the design protocol.

Conclusions Here we have described a theoretical model to quantitatively evaluate the electrostatic effects of salt on protein stability. Using this model, we have elucidated the effect of salt on the stability of a variety of wild-type cold shock proteins, as well

as various point mutants and designed sequences. The results show good agreement with the corresponding experimental data. Our results suggest that salt-dependent changes in stability may be correlated with thermostability, offering further evidence for the electrostatic nature of thermophilic adaptation. Over the years, multiple theories have been offered to explain the ability of thermophilic and hyperthermophilic proteins to adapt to their extreme environment. These have ranged from better hydrophobic packing to more productive helix dipole interactions.8 Since the number of these protein structures has been growing, more general explanations are beginning to come to light. A survey of thermophilic protein structures and their corresponding mesophilic homologs discounted the suggestion that enhanced hydrophobic packing generally contributes to thermophilic stability.33 An independent survey suggested that the only structural property that was correlated with thermostability was the number of salt bridge pairs on the protein surface.12 These results taken together suggest that electrostatic mechanisms are likely to be at work in stabilizing thermostable proteins (B.N.D., H. Minoux & C.L.B.,III, unpublished results). We have also shown that the effect of salt on the contribution of an individual residue to the stability of its protein does not necessarily correlate with short-range Coulomb interactions. Shortrange interactions between a residue and its surroundings have evolved to be favorable in almost every case. However, our results suggest that each

Salt Effects and the Cold Shock Protein Family

negatively charged residue in the mesophilic protein is contributing toward the halophilicity of this sequence. Although short-range interactions are favorable, longer range repulsive interactions contribute significantly to the overall Coulombic interaction energy. This is seen from our evaluation of the residue-wise salt effects in mesophilic systems where the net charge has been attenuated. The results described here involved a collaborative effort between theory and experiment. This collaboration was critical for the evolution of this work. Based on critical testing of our models with the experimental findings, a robust theoretical model that can be used to accurately interpret and predict the effect of ionic strength on protein stability evolved. The success of this model was demonstrated at a variety of levels starting with the macroscopic effect of salt on the stability of wild-type cold shock protein sequences, moving to the effect of salt on single point mutants as well as the microscopic effect of salt on individual residues in the protein sequences, and finally to the effect of salt on designed cold shock protein variants. In each case, robust agreement between theory and experiment is shown, providing a strong connection between an important, physically observable phenomenon and a mathematical model.

Methods Model structures This study focuses on the electrostatic behavior of the cold shock protein family derived from a mesophilic bacterium B. subtilis (Bs ), a thermophilic bacterium B. caldolyticus (Bc ) and a hyperthermophilic bacterium T. maritima (Tm ). The crystal structure of the cold shock protein (1csp) from B. subtilis was used to represent the mesophilic protein and also as the basis for homology modeling the thermophilic and hyperthermophilic proteins. The high degree of sequence homology between the mesophilic and thermophilic, and mesophilic and hyperthermophilic proteins (82% and 62%, respectively) justify this modeling approach.34 Further, when the crystal structure of the thermophile became available18 it was evaluated for its stability as a function of the salt concentration. The quantitative differences were negligible when both of two crystal forms were compared to the homology model. Finally, we note that following the completion of our work additional structures were solved for hyperthermophile35 and several mutants,36 and these structures further confirm our modeling. The modeling procedure used the homology modeling package within Insight to initially place the mutated side-chains. These altered side-chains were then minimized in the context of a rigid protein environment in order to remove any van der Waals clashes and improve local electrostatic interactions. Theoretical model The model used to calculate the electrostatic free energy of folding involves numerical finite-difference solutions of the Poisson –Boltzmann equation (FDPB) and a thermodynamic cycle that involves structural


models for the folded state and a model for the unfolded state at two different salt concentrations.37 Salt is introduced through the Poisson – Boltzmann formalism in order to explore the effect of ionic strength on the stability of a given protein. The unfolded state model considers each residue surrounded by its two nearest (in sequence) neighbors on either side. In the five-residue segments representing the unfolded state, the local conformation is chosen to be that of the folded state. Energetically, this restricts the unfolded state to the electrostatic free energy of a given residue as well as the local interactions with neighboring residues. The electrostatic solvation free energy for a given residue is calculated in the context of its neighboring residues, where the charges on the neighboring residues are temporarily turned off. The electrostatic interaction energy between the residue in question and the four neighbors is calculated by multiplying the electrostatic potential determined in the previous calculation by the charges on the neighboring residues.17 The identical calculations are performed for each residue in the context of the entire protein. The difference between these energies is the contribution of the residue to the electrostatic free energy of protein folding. The above procedure is carried out on all residues and the properly normalized sum (preventing double counting of electrostatic interactions) is the total electrostatic free energy of folding. This protocol can be performed at a variety of salt concentrations in order to examine the effects of salt on the electrostatic contribution to folding. Finite difference Poisson – Boltzmann calculations17 were performed within the CHARMM molecular mechanics package38 using an interior dielectric of 1interior ¼ 4 for the protein and an exterior dielectric of 1solvent ¼ 80: Multiple dielectric models were examined for both the protein and solvent, but were not found to significantly influence the results. The value used for the dielectric constant of the solute cancels out in evaluating the effect of salt on stability (see equations (1) and (2) above) and thus a single value of 4 was used in all of the calculations described above. The CHARMM PB solver uses a finite difference approach. The grid parameters chosen included a grid ˚ 21, and a grid size of 50 A ˚ 3. More point density of 1 A finely resolved grid spacing as well as larger grid volumes was found to have a minimal impact on either the qualitative or quantitative results. Charges and van der Waals radii were assigned according the CHARMM param19 polar hydrogen potential.38 A cavity model was used to examine the qualitative effects related to “salting-out” at high ionic strength and to illustrate this phenomenon in the context of the electrostatic model. Further work would be required to make this a quantitative model. The cavity model employed to reproduce salting-out effects was implemented in the following manner. The hydrophobic contribution to solvation was included following Sitkoff et al.39 The model expresses the hydrophobic solvation component as the product of the hydrophobic accessible surface area of the solute (ASA) and the surface tension of the solvent (g) ðDGhydr ¼ g £ ASAÞ: The oil/water interface surface tension is approximately (72 cal) ˚ 2.40 The increment in surface tension as a 0.3 kJ/mol A function of NaCl concentration is 1.64 dynes/cm or ˚ 2.40 We reduced this value (by a factor of 9.9 J/mol A about 6) to more closely represent the molecular aspects of this phenomenum and the results obtained from the salt-dependent stability experiments for the wild-type cold shock proteins. This reduction compensates for


Salt Effects and the Cold Shock Protein Family



-5 -10

Bs-CspB (New Model) Bc-Csp (New Model) Tm-Csp (New Model) Bs-CspB Bc-Csp Tm-Csp

-15 10

∆∆G (kJ/mol)

(B) 5


∆∆Gelec (kJ/mol)




-10 Mesophile Thermophile Hyperthermophile

-15 8 6 ∆∆G (kJ/mol)

∆∆Gelec (kJ/mol)



4 2 0 -2 -4


-6 0.2 0.4 0.6 0.8 Salt Concentration (M)

-8 8


Figure 10. The salt-dependent electrostatic stability of cold shock proteins using a compact denatured state model for the “native state” and unfolded state. An analysis of the electrostatic stability of the compact denatured state itself (a) (using the local interaction model as the reference structure) reveals the same trends as seen using the native structure (Figure 1). The quantitative effects of salt are attenuated. Since the effects of salt vary predominantly in the compact models, subtracting the salt-dependent curve generated using the compact denatured model from that generated using the native state results in the salt-dependent stability obtained by replacing the local interaction model with the compact denatured state. The result is a weakened salt dependence with the same trends reported using the local interaction unfolded model. The results included Hofmeister salting-out effects are shown in (b).

quantitative free energy differences due to the model of the unfolded state as well as curvature effects inherent in relating a macroscopic surface tension to the microscopic size scale.39,41 The hydrophobic solvation term was simply added to the component related to the electrostatic solvation free energy to provide our estimate of the salting-out contribution to protein stability with increasing salt concentration. Alternative models of unfolded and folded states To further substantiate our observations regarding the influence of ionic strength on protein stability, we examine the effects of salt on compact denatured states derived from the native structures of the wildtype cold shock proteins. In addition, it has been suggested by experimental and theoretical studies that denatured states retain some long-range interactions.42,43 Here we demonstrate that the model chosen to represent the unfolded state does not impact the qualitative results obtained for the cold shock protein B family. Possible representations of compact denatured states may be generated using the energy minimization procedure outlined by Elcock, where electrostatic inter-

6 ∆∆G (kJ/mol)



4 2 0 -2 -4 -6 -8 0

0.2 0.4 0.6 0.8 Salt Concentration (M)


Figure 11. The effects of salt on snapshots from solvated molecular dynamics trajectories. Error bars indicate one standard deviation. Relaxation induced by the solvated molecular dynamics simulation improves electrostatic interactions within the mesophile and thermophile, thereby allowing salt to have a slightly more destabilizing effect on both structures relative to the crystal conformations. Qualitative trends are preserved both in the purely electrostatic effects of ionic strength (a) as well as when Hofmeister salting-out effects are included (b). We further note that increasing the surface tension parameter used to evaluate the Hofmeister effects can compensate for many of the quantitative differences between the molecular dynamics results and those obtained both from crystal structures and experiment (c).

actions are eliminated and exaggerated van der Waals repulsions are used to generate compact conformations lacking secondary structure.23 While there is no reason to believe these states are physically representative of thermally denatured cold shock proteins under the conditions of the experiments, they provide another set of (folded and) unfolded models and further bounds on our theoretical picture. When the electrostatic stabilities of these compact denatured states are examined, relative to our localized denatured state model, the same qualitative trends as observed above are produced (Figure 10(a)). Quantitatively, however, the effects of salt are uniformly attenuated. When these compact denatured states are substituted for our original unfolded model, the result is a weakened


Salt Effects and the Cold Shock Protein Family

salt dependence on protein stability (Figure 10(b)). Neither model may be judged in this context to be “more correct”, since the addition of Hofmeister effects may compensate for quantitative discrepancies between either theoretical model and experiment. Further theoretical development will be required to precisely account for these Hofmeister effects, thereby eliminating any flexible parameters in the model. For the purposes of this study however, the shortrange model for the unfolded state is reasonable for describing the qualitative effects of salt on protein stability. We further note that thermal fluctuations in structure do not affect the qualitative results reported here. This is expected given the qualitative similarity of the results even in the compact unfolded model. However, further analysis was performed on snapshots from molecular dynamics trajectories generated for the three wild-type cold shock proteins. The results demonstrate fluctuation in the quantitative effects of salt on protein stability, however the qualitative trends are robust (Figure 11(a) and (b)). As mentioned previously, the quantitative results are modulated by the surface tension parameter chosen to represent the Hofmeister salting-out effects. The effect of increasing the surface tension parameter is shown in Figure 11(c).

Construction and overexpression of protein variants Mutants A, B (MLE) and B (MELE) were constructed starting from the plasmid pCspB344 as a template. The gene encoding Bs-CspB was amplified by the polymerase chain reaction (PCR) using 50 and 30 appropriate primers, cut with Kpn I and Taq I and purified to obtain a DNA fragment encoding the first 18 amino acid residues of Bs-CspB. The DNA fragment coding amino acid residues 19 to 67 of the mutant proteins was generated by filling up two overlapping 66 and 139 base oligonucleotides, designed to introduce a Taq I site at the 50 site and a Eco RI restriction site at the 30 end of the gene, and amplification with terminal primers by PCR. These DNA fragments were cut with Eco RI and Taq I, purified and ligated, together with the Kpn I-Taq I DNA fragment, in the plasmid pBluescript II SK(2) from Stratagene (La Jolla, CA) cut with Kpn I and Eco RI. The additional residue Glu2 in mutant B (MELE) was introduced by site-directed mutagenesis of the plasmid encoding mutant B (MLE) using the QuikChange kit of Stratagene (La Jolla, CA). The mutations were confirmed by sequencing of the whole genes. The variants were overexpressed and purified using the T7 RNA polymerase promotor system as described for wild-type Bc-Csp18 with minor modifications.

Heat-induced equilibrium unfolding transitions Thermal unfolding transitions were measured in a Jasco J-600 spectropolarimeter equipped with a PTC-348 WI Peltier element. Protein concentrations were 4 mM in 100 mM sodium cacodylate (pH 7.0) buffer of the desired NaCl concentration. Samples were heated in 1 cm cells at a rate of 60 8C per hour. Transitions were monitored by the decrease in the CD signal at 222.6 nm upon unfolding and analyzed by a non-linear least squares fit according to a two-state model45 with a fixed heat capacity change DCp of 4 kJmol21 K21.

Acknowledgments The authors thank Herve´ Minoux for his involvement in the initial design of the thermostable mutants. Also, financial support from the National Institutes of Health (GM37554, GM48807 and RR12255) is greatly appreciated. Finally, we thank Deutsche Forschungsgemeinschaft (Schm 444/15-1) and the Fonds der Chemischen Industrie.

References 1. Alberts, B., Bray, D., Lewis, J., Raff, M., Roberts, K. & Watson, J. D. (1994). Molecular Biology of the Cell, 3rd edit., Garland Publishing, New York. 2. Schildkraut, C. & Shneior, L. (1965). Dependence of the melting temperature of DNA on Salt concentration. Biopolymers, 3, 195–208. 3. Piskur, J. & Rupprecht, A. (1995). Aggregated DNA in ethanol solution. FEBS Letters, 375, 174– 178. 4. Feigon, J., Wang, A. H. J., Van der Marel, G. A., Van Boom, J. H. & Rich, A. (1985). Z-DNA forms without an alternating purine – pyrimidine sequence in solution. Science, 230, 82 – 84. 5. Record, M. T., Jr, Zhang, W. & Anderson, C. F. (1998). Analysis of effects of salts and uncharged solutes on protein and nucleic acid equilibria and processes: a practical guide to recognizing and interpreting polyelectrolyte effects, Hofmeister effects, and osmotic effects of salts. Advan. Protein Chem. 51, 281– 353. 6. Von Hippel, P. H. & Wong, K. Y. (1964). Neutral salts. The generality of their effects on the stability of macromolecular conformations. Science, 145, 577 –580. 7. Kohn, W. D., Kay, C. M. & Hodges, R. S. (1997). Salt effects on protein stability: two-stranded alphahelical coiled-coils containing inter- or intrahelical ion pairs. J. Mol. Biol. 267, 1039– 1052. 8. Jaenicke, R. & Bo¨hm, G. (1998). The stability of proteins in extreme environments. Curr. Opin. Struct. Biol. 8, 738– 748. 9. Rao, J. K. M. & Argos, P. (1981). Structural stability of halophilic proteins. Biochemistry, 20, 6536– 6543. 10. Elcock, A. H. & McCammon, J. A. (1998). Electrostatic contributions to the stability of halophilic proteins. J. Mol. Biol. 280, 731– 748. 11. Gerday, C., Aittaleb, M., Arpigny, J. L., Baise, E., Chessa, J. P., Garsoux, G. et al. (1997). Psychrophilic enzymes: a thermodynamic challenge. Biochim. Biophys. Acta, 1342, 119 – 131. 12. Szilagyi, A. & Zavodszky, P. (2000). Structural differences between mesophilic, moderately thermophilic, and extremely thermophilic protein subunits: results of a comprehensive survey. Structure, 8, 493–504. 13. Goedken, E. R. & Marqusee, S. (2001). Co-crystal of Escherichia coli RNase H1 with Mn2þ ions reveals two divalent metals bound in the active site. J. Biol. Chem. 276, 7266– 7271. 14. Van Asselt, E. J. & Dijkstra, B. W. (1999). Binding of calcium in the EF-hand of Escherichia coli lytic transglycosylase Slt35 is important for stability. FEBS Letters, 458, 429– 435. 15. Baldwin, R. L. (1996). How Hofmeister ion interactions affect protein stability. Biophys. J. 71, 2056 –2063. 16. Vinod, K., Sharp, K. A., Friedman, R. A. & Honig, B. (1994). Salt effects on ligand– DNA binding. Minor groove binding antibiotics. J. Mol. Biol. 238, 245– 263.


Salt Effects and the Cold Shock Protein Family

17. Warwicker, J. & Watson, H. C. (1982). Calculation of the electric potential in the active site cleft due to a-helix dipoles. J. Mol. Biol. 157, 671– 679. 18. Mueller, U., Perl, D., Schmid, F. X. & Heinemann, U. (2000). Thermal stability and atomic-resolution crystal structure of the Bacillus caldolyticus cold shock protein. J. Mol. Biol. 297, 975– 988. 19. Perl, D., Welker, C., Schindler, T., Schro¨der, K., Marahiel, M. A., Jaenicke, R. & Schmid, F. X. (1998). Conservation of rapid two-state folding in mesophilic, thermophilic and hyperthermophilic cold shock proteins. Nature Struct. Biol. 5, 229– 235. 20. Perl, D., Mueller, U., Heinemann, U. & Schmid, F. X. (2000). Two exposed amino acid residues confer thermostability on a cold shock protein. Nature Struct. Biol. 7, 380– 383. 21. Shea, J. E. & Brooks, C. L., III (2001). From folding theories to folding proteins: a review and assessment of simulation studies of protein folding and unfolding. Annu. Rev. Phys. Chem. 52, 499– 535. 22. Perkyns, J. S., Wang, Y. & Pettitt, B. M. (1996). Salting in peptides: conformationally dependent solubilities and phase behavior of a tripeptide zwitterion in electrolyte solution. J. Am. Chem. Soc. 118, 1164– 1172. 23. Elcock, A. H. (1999). Realistic modeling of the denatured states of proteins allows accurate calculations of the pH dependence of protein stability. J. Mol. Biol. 294, 1051– 1062. 24. Hasted, J. B., Ritson, D. M. & Collie, C. H. (1948). Dielectric properties of aqueous ionic solutions. J. Chem. Phys. 16, 1 –11. 25. Srinivasan, J., Trevathan, M. W., Beroza, P. & Case, D. A. (1999). Application of a pairwise generalized born model to proteins and nucleic acids: inclusion of salt effects. Theor. Chem. Acc. 101, 426– 434. 26. Sanchez-Ruiz, J. M. & Makhatadze, G. I. (2001). To charge or not to charge? Trends Biotechnol. 19, 132– 135. 27. Loladze, V. V., Ibarra-Molero, B., Sanchez-Ruiz, J. M. & Makhatadze, G. I. (1999). Engineering a thermostable protein via optimization of charge– charge interactions on the protein surface. Biochemistry, 38, 16419– 16423. 28. Spector, S., Wang, M., Carp, S. A., Robblee, J., Hendsch, Z. S., Fairman, R. et al. (2000). Rational modification of protein stability by the mutation of charged surface residues. Biochemistry, 39, 872– 879. 29. Xiao, L. & Honig, B. (1999). Electrostatic contributions to the stability of hyperthermophilic proteins. J. Mol. Biol. 289, 1435 –1444. 30. Tanford, C. (1968). Protein denaturation. Advan. Protein Chem. 23, 121– 282. 31. Grimsley, G. R., Shaw, K. L., Fee, L. R., Alston, R. W., Huyghues-Despointes, B. M. P., Thurlkill, R. L. et al. (1999). Increasing protein stability by altering longrange coulombic interactions. Protein Sci. 8, 1843– 1849.

32. Shaw, K. L., Grimsley, G. R., Yakovlev, G. I., Makarov, A. A. & Pace, C. N. (2001). The effect of net charge on the solubility, activity, and stability of ribonuclease Sa. Protein Sci. 10, 1206– 1215. 33. Karshikoff, A. & Landenstein, R. (1998). Proteins from thermophilic and mesophilic organisms essentially do not differ in packing. Protein Eng. 11, 867– 872. 34. Sanchez, R. & Sali, A. (1997). Advances in comparative protein-structure modelling. Curr. Opin. Struct. Biol. 7, 206–214. 35. Kremer, W., Schuler, B., Harrieder, S., Geyer, M., Gronwald, W., Welker, C. et al. (2001). Sloution NMR structure of the cold-shock protein from the hyperthermophilic bacterium Thermotoga maritima. Eur. J. Biochem. 268, 2527– 2539. 36. Delbrueck, H., Mueller, U., Perl, D., Schmid, F. X. & Heinemann, U. (2001). Crystal structures of mutant forms of the Bacillus caldolyticus cold shock protein differing in thermal stability. J. Mol. Biol. 312, 359– 369. 37. Yang, A. S. & Honig, B. (1994). Structural origins of pH and ionic strength effects on protein stability. Acid denaturation of sperm whale apomyoglobin. J. Mol. Biol. 237, 602–614. 38. Brooks, B. R., Bruccoleri, R. E., Olafson, B. D., States, D. J., Swaminathan, S. & Karplus, M. (1983). CHARMM: a program for macromolecular energy, minimization, and dynamics calculations. J. Comput. Chem. 4, 187–217. 39. Sitkoff, D., Sharp, K. A. & Honig, B. (1994). Correlating solvation free energies and surface tensions of hydrocarbon solutes. Biophys. Chem. 51, 397– 409. 40. Arakawa, T. & Timasheff, S. N. (1982). Preferential interactions of proteins with salts in concentrated solutions. Biochemistry, 21, 6545– 6552. 41. Jackson, R. M. & Sternberg, M. J. (1994). Application of scaled particle theory to model the hydrophobic effect: implications for molecular association and protein stability. Protein Eng. 7, 371– 383. 42. Oliveberg, M., Vuilleumier, S. & Fersht, A. R. (1994). Thermodynamic study of the acid denaturation of barnase and its dependence on ionic strength: evidence for residual electrostatic interactions in the acid/thermally denatured state. Biochemistry, 33, 8826– 8832. 43. Shortle, D., Chan, H. S. & Dill, K. A. (1992). Modeling the effects of mutations on the denatured states of proteins. Protein Sci. 1, 201– 215. 44. Willimsky, G., Bang, H., Fischer, G. & Marahiel, M. A. (1992). Characterization of CspB, a Bacillus subtilis inducible cold shock gene affecting cell viability at low temperatures. J. Bacteriol. 174, 6326– 6335. 45. Mayr, L. M., Landt, O., Hahn, U. & Schmid, F. X. (1993). Stability and folding kinetics of ribonuclease T1 are strongly altered by the replacement of cisproline 39 with alanine. J. Mol. Biol. 231, 897– 912.

Edited by B. Honig (Received 8 August 2001; received in revised form 22 January 2002; accepted 22 March 2002)