Salt Bridges in the Hyperthermophilic Protein Ssh10b Are Resilient to Temperature Increases*

A double mutant cycle (DMC) approach was employed to estimate the effect of temperature on the contribution of two highly conserved salt bridges to protein stability in the hyperthermophilic protein Ssh10b. The coupling free energy were 2.4 ± 0.4 kJ/mol at 298 K and 2.2 ± 0.4 kJ/mol at 353 K for Glu-54/Arg-57, and 6.0 ± 0.2 kJ/mol at 298 K and 5.9 ± 0.6 kJ/mol at 353 K for Glu-36/Lys-68. The stability free energy of Ssh10b decrease greatly with increasing temperature, while the direct contribution of these two salt bridges to protein stability remain almost constant, providing evidence supporting the theoretical prediction that salt bridges are extremely resilient to temperature increases and thus are specially suited to improving protein stability at high temperatures. The reason for the difference in coupling free energy between salt bridges Glu-54/Arg-57 and Glu-36/Lys-68 is discussed. Comparing our results with published DMC data for the contribution of salt bridges to stability in other proteins, we found that the energy contribution of a salt bridge formed by two charged residues far apart in the primary sequence is higher than that of those formed between two very close ones. Implications of this finding are useful for engineering proteins with enhanced thermostability.

Hyperthermophiles are a group of microorganisms with an optimum growth temperature between 80 and 100°C. Remarkably, hyperthermophilic proteins and their mesophilic homologues typically show 40 -85% sequence similarity, and their three-dimensional structures are highly superimposable (1)(2)(3)(4). These facts indicate that the molecular basis of their extreme thermal tolerance is associated with the delicate balance of non-covalent interactions.
Considerable efforts have been invested in recent years to understand how proteins from hyperthermophiles can maintain stability at high temperatures. Several factors, such as improvement in packing density, strengthening of hydrophobic cores, and decreased length of surface loops, are thought to correlate with the increased thermostability of hyperthermophilic proteins (5)(6)(7)(8)(9). The most commonly cited feature, however, is an increased number of salt bridges and salt bridge networks distributed over the protein surface (10 -15). A comparison of 13 structural parame-ters derived from the tertiary structures of 64 proteins from mesophiles and 29 proteins from thermophiles and hyperthermophiles revealed that the only feature shared in common is an increase in the number of salt bridges with increasing growth temperatures (16).
However, experimental estimates of the free energy contribution of salt bridges to protein stability have led to conflicting conclusions, ranging from them having a stabilizing, insignificant, or even a destabilizing effect (17)(18)(19)(20)(21)(22)(23). The association of two charged residues to form a salt bridge is thought to incur a substantial desolvation penalty that is seldom completely compensated for by favorable interactions within the salt bridge and the rest of the protein (24 -26). A continuum solvation model was used to investigate how this same argument applies at the high temperatures hyperthermophiles experience and predicted that at such high temperatures the desolvation penalty for the formation of a salt bridge is markedly reduced (25). Furthermore, molecular simulations suggest salt bridges are extremely resilient to temperature increases and thus are specially suited to promoting protein stability at high temperatures (26). However, attempts to obtain thermodynamic information at high temperature have been plagued by irreversibility of unfolding and/or inaccessibility of the unfolding transition to physical measurements. So far there are no reports of experiments to measure the stability contribution of salt bridges at high temperatures and to provide evidence for these theoretical predictions.
The DNA-binding protein Ssh10b from the archaeon Sulfolobus shibatae is a member of the Sac10b family that is thought to be involved in chromosomal organization or DNA repair/recombination. Ssh10b is a highly thermostable dimeric protein composed of two identical subunits, each monomer consisting of 97 amino acid residues with no disulfide bonds (27,28). Ssh10b constitutes about 4 -5% of total cellular protein, and binds dsDNA without apparent sequence specificity. Ssh10b is also capable of constraining negative DNA supercoils in a temperature-dependent fashion (29). The crystal structure of Ssh10b reveals that the monomer is a mixed ␣/␤ structure comprised of four ␤-strands and two ␣-helices (Fig. 1), each monomer containing four isolated salt bridges: Glu-36/Lys-68, Glu-54/Arg-57, Asp-63/Lys-97, Glu-66/Arg-95, and an ionpair network, Lys-40/Glu-91/Arg-71/Glu-69 (30), evidently more than the average of about 5 salt bridges per 150 amino acid residues (31). However, only the Glu-36/Lys-68 and Glu-54/Arg-57 salt bridges are highly conserved across the Sac10b family (Fig. 2). All of these salt bridges and ion-pair networks are surface-exposed, located on the back of the Ssh10b dimer and thought to play an important role in bracing and stabilizing the structure (30).
Electrostatic interactions of salt bridges can be studied by several methods. A particularly useful approach is to employ a method called double mutant cycle (DMC) 2 analysis to obtain quantitative information about the stability contribution of a salt bridge by mutating the residues involved separately and concurrently (32)(33)(34)(35)(36), thus cancelling all interaction effects except the direct interaction between the two charged residues. A DMC gives the so-called coupling free energy of a salt bridge associated with the direct interaction between the two charged residues under ideal conditions.
In the present work, we analyzed the net strength of the two highly conserved salt bridges, Glu-36/Lys-68 and Glu-54/Arg-57, by DMC analysis. The coupling free energy were 2.4 Ϯ 0.4 kJ/mol at 298 K and 2.2 Ϯ 0.4 kJ/mol at 353 K for Glu-54/Arg-57, and 6.0 Ϯ 0.2 kJ/mol at 298 K and 5.9 Ϯ 0.6 kJ/mol at 353 K for Glu-36/Lys-68. The stability free energy of Ssh10b decrease greatly with increasing temperature, while the direct contribution of these two salt bridges to protein stability remain almost constant, providing evidence supporting the theoretical prediction that salt bridge interactions are extremely resilient to temperature increases and thus are specially suited to improving protein stability at high temperatures (26).

Mutagenesis, Expression, and Purification of Ssh10b and Its
Variants-All expression plasmids of Ssh10b variants were constructed from the parental plasmid pET11a-ssh10b. All mutations were introduced by site-directed mutagenesis through overlap extension PCR. The construct for each variant was verified by DNA sequencing.
Expression plasmids of Ssh10b and its variants were transformed into the Escherichia coli BL21 (DE3) host strain. For protein expression, a single colony was grown in 100 ml of LB media containing 100 g/ml ampicillin by shaking (ϳ220 rpm) at 37°C overnight. Cultures were diluted 1:50 in fresh LB antibiotic-containing media and shaken for about 3 h at 37°C. Target protein expression was then induced at A 600 ϭ 0.8 -1.0 by the addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 0.3 mM, and growth was allowed to continue for about 5 h at 37°C with constant shaking.
After harvesting by centrifugation (4,000 rpm for 30 min), cell pellets from 1-liter cultures were re-suspended in 25 ml of buffer A (20 mM Tris-HCl, pH 7.5) and then disrupted by ultrasonication on ice. The lysate was maintained at 60°C for 20 min to precipitate the E. coli proteins, and then centrifuged at 16,000 rpm for 30 min at 4°C. The supernatant was applied to a 20-ml Source 30S column, which had been equilibrated with buffer A, and proteins were eluted with a 0 -50% gradient (120 ml) of buffer B (1.5 M NaCl, 20 mM Tris-HCl, pH7.5). Fractions containing the target protein were identified by 15% SDS-PAGE and dialyzed overnight against buffer A. After centrifugation at 16,000 rpm for 30 min, the supernatant was applied to a 6-ml Resource-S column, which had been equilibrated with buffer A, and proteins were eluted with a 0 -50% gradient (60 ml) of buffer B. Fractions containing the target protein were identified by 15% SDS-PAGE, dialyzed overnight against deionized water, and then lyophilized.
Unfolding Studies-Unfolding of Ssh10b and its variants was studied by taking circular dichroism (CD) measurements with a *-pistar 180 spectrometer (Applied Photophysics Ltd, UK), performed in buffer H (10 mM HEPES, pH 7.0), at a protein monomer concentration of about 25 M. The CD signal was monitored using a rectangular quartz cuvette with a path length of 1 mm. For urea-induced unfolding, the urea solution was freshly prepared on the day of use. The samples containing various concentrations of urea were equilibrated at 25°C overnight and then measured by far-UV CD at 222 nm. For heatinduced unfolding, each sample was heated from 40 to 98°C and then cooled from 98 to 40°C using stepwise changes of 2°C, and the CD signal was recorded after equilibration for 2 min at each temperature point. All unfolding experiments were repeated 3-4 times.
Analysis of the Denaturation Data-Previous studies in our laboratory have shown that both denaturant-and heat-induced unfolding of Ssh10b are fully reversible and follow a two-state mechanism involving a native dimer and two denatured monomers (28). Therefore, in this work the thermodynamic properties of Ssh10b and its variants were calculated assuming a twostate denaturation process in Reaction 1.
The observed equilibrium constant (K obs ) and the corresponding free energy change (⌬G) at temperature T or denaturant concentration [D] were calculated according to Equations 1 and 2 (28), 2 The abbreviation used is: DMC, double mutant cycle.
where P t is the total protein concentration in monomer units; R is the gas constant; T is the absolute temperature; y is the experimentally measured signal value at a given temperature (T) or given denaturant concentration ([D]); y N and y U are the intercepts; and m N and m U are the slopes of the native and unfolded baselines, respectively. According to the linear free energy model (37)(38)(39)(40), changes in free energy (⌬G) that occur on unfolding are expected to vary linearly with denaturant concentration in Equation 3 ([D]), where ⌬G(H 2 O) represents the free energy change of unfolding in the absence of denaturant and m G is the slope of the transition for the free energy.
For thermal unfolding, assuming that the heat capacity change (⌬C p ) between the native and unfolded states of the system is relatively independent of temperature, gives us Equation 4 (41,42), where ⌬H m and ⌬S m are the enthalpy and entropy changes, respectively, of the protein at the transition midpoint, where Equation 4 can be simplified to the van't Hoff plot in Equation 5.
The temperature of the transition midpoint (T m ) can be calculated according to Equation 6.
The uncertainties of the data were represented as two-sided 95% confidence intervals, which are given by Equation 7, where b is the data (⌬G, ⌬H, and T m , etc.), se(b) is the standard error, t v,0.975 is the value of the t distribution with the number of degrees of freedom v for the two-sided 95% confidence interval.
The direct stability contribution of each salt bridge was represented by using the average values of the coupling free energy of the four different DMCs and uncertainties estimated by Equation 7.
Unfolding Studies-Ssh10b is resistant to urea-induced denaturation in phosphate buffer (28), but is more susceptible to urea-induced denaturation in monovalent ion buffers (43). And for any one of above proteins, when the urea-denatured protein solution was diluted to a lower concentration of denaturant, the CD spectrum of the renatured protein was identical to that of the native protein, indicating that the urea-induced unfolding processes of Ssh10b and its variants are fully reversible (data not shown). Fig. 3 shows representative urea-induced unfolding profiles for Ssh10b and its partial variants in buffer H. The linear free energy model was used to analyze the ureainduced unfolding profiles. Results are presented in Table 1 and reveal that the difference in stability between different variants is marked: relative to the Ssh10b wild-type, some variants show an increase in stability free energy at 298 K (⌬G (298) ) of up to 7 kJ/mol, while others show a decrease in ⌬G (298) of more than 10 kJ/mol. Ssh10b is high resistant to heat-induced denaturation in phosphate buffer (28), but we found that its thermostability decrease greatly in a monovalent ion buffer resulting in accessibility of the heat-induced unfolding transition to physical measurement. The heat-induced unfolding and refolding curves of Ssh10b and its variants in buffer H monitored by far-UV CD at 222 nm are nearly superimposable, and for each one the CD spectrum of the refolded protein is nearly identical to that of the native protein (data not shown), indicating that the processes of heat-induced unfolding of Ssh10b and its variants are fully reversible. Fig. 4 shows representative heat-induced unfolding profiles for Ssh10b and its partial variants in buffer H. The heat-induced unfolding profiles of Ssh10b and its variants were analyzed by using the two-state denaturation model described under "Experimental Procedures": the parameters ⌬H m and ⌬S m were obtained by fitting K obs to the van't Hoff plot, the parameter T m was then obtained by substituting ⌬H m and ⌬S m into Equation 6. The T m values of these variants were all around 353 K, and so the stability free energy values of these proteins at 353 K (⌬G (353) ) were calculated using the van't Hoff plot ( Table 1). As with the results for urea-induced unfolding, the difference between different variants in stability free energy at 353 K was also marked: the difference in ⌬G (353) between mutants and the wild-type ranged from Ϫ8.3 to 7.5 kJ/mol.

Estimating the Contribution of Salt Bridges to Stability by
Using DMC Analysis-DMC analysis was used to estimate the strength of the two conserved salt bridges (Glu-36/Lys-68 and Glu-54/Arg-57) in protein Ssh10b. Fig. 5 shows a general scheme for a DMC consisting of two single and one double mutant. A thermodynamic cycle is set up between the wild type, each of the single mutants, and the double mutant. The effects of single amino acid substitutions of each residue involved in the salt bridge and the effect of simultaneous replacement at both positions on the stability of the protein were measured. If the two charged residues do not interact with each other, the effect of substitution of either of the two residues will be independent of the replacement of the other, in other words, the effect on protein stability resulting from their simultaneous replacement will be equal to the sum of the effects of the two single mutations. By contrast, if the two residues do interact with each other, the effect of substituting either of the two residues will depend on the substitution of the other. By using the DMC method, the so-called coupling free energy of a salt bridge can be obtained. As shown in Fig. 5, the coupling free energy (⌬⌬G coup ) is defined in Equations 8 -11 (44). Here, because Ssh10b is a homodimer, the coupling free energy for each salt bridge in the molecule can be represented as Equation 13.   Fig. 6, a and b shows the results of different DMCs for the two conserved salt bridges (Glu-36/Lys-68 and Glu-54/Arg-57) in protein Ssh10b at 298 K. For each salt bridge, four different DMCs were constructed to evaluate the interaction strength between the two charged residues involved in the salt bridge. For Glu-36/Lys-68, the reference proteins (doubly substituted variants) for the four DMCs were Ala-36/Ala-68, Ala-36/Leu-68, Gln-36/Ala-68, and Gln-36/Leu-68. Though the reference protein for each DMC has a different stability, all four DMCs gave identical, within the experimental error, values for the coupling free energy of the salt bridge Glu-36/Lys-68 (5.9, 6.0, 6.2, and 5.8 kJ/mol, respectively) at 298 K. For Glu-54/Arg-57, the reference proteins for the four DMCs were Ala-54/Ala-57, Ala-54/Leu-57, Gln-54/Ala-57, and Gln-54/Leu-57. The four DMCs of Glu-54/Arg-57 also gave identical, within the experimental error, coupling free energy values (2.1, 2.7, 2.3, and 2.5 kJ/mol, respectively) at 298 K. Thus, we can assign favorable Gibbs free energy of 6.0 Ϯ 0.2 kJ/mol and 2.4 Ϯ 0.4 kJ/mol to the salt bridges Glu-36/Lys-68 and Glu-54/Arg-57, respectively, at 298 K. Fig. 6, c and d shows the results of different DMCs for the two conserved salt bridges in protein Ssh10b at 353 K. For Glu-36/ Lys-68, the four coupling free energy values at 353 K were 5.4, 5.9, 6.3, and 5.9 kJ/mol, and for Glu-54/Arg-57, the coupling free energy values of the four DMCs at 353 K were 2.0, 2.6, 2.1, and 2.1 kJ/mol. We can assign favorable Gibbs free energy of 5.9 Ϯ 0.6 kJ/mol and 2.2 Ϯ 0.4 kJ/mol to the salt bridges Glu-36/Lys-68 and Glu-54/Arg-57, respectively at 353 K. Remarkably, these results were highly consistent with those at 298 K.

DISCUSSION
Experimental estimates of the energy contribution of a salt bridge to protein stability have led to conflicting conclusions, ranging from them having stabilizing, insignificant, or even destabilizing effects (17)(18)(19)(20)(21)(22)(23). In some reports, electrostatic interactions of salt bridges were measured by mutating a charged residue to a non-charged residue and measuring the impact of this change on protein stability. This type of approach is unsuitable as such mutations not only remove charge-charge interactions but also alter a number of other interactions, including the desolvation penalty and the background interactions of the residue within the protein. However, DMC analysis provides an elegant method to estimate the strength of a salt bridge.
The contribution of a salt bridge (⌬⌬G bri ) to protein stability involving two charged residues can be represented as Equation 14, where ⌬⌬G dir is the direct contribution of the salt bridge to protein stability. In most situations, it is the direct chargecharge interaction, but it sometimes also includes other direct interactions, e.g. van der Waals interactions between atoms of  the two charged residues. The other two components are all indirect contributions of the salt bridges: ⌬⌬G desol is the desolvation penalty of the two charged residues, and ⌬⌬G backg is the stability contributions due to background interactions of the two charged residues with other components of the protein.
Generally, the desolvation penalty is regarded as the dominant component of the indirect contributions (44).
DMC analysis is designed to cancel all interactions except the direct interactions between the two residues involved in the salt bridge. However, the validity of the method is subject to the following two assumptions. First, all indirect interactions are simply additive in nature from the single to the double mutation. Secondly, there is no interaction between the two residues in the doubly substituted mutant. Then, in an ideal DMC, according to Equations 12 and 13, indirect contributions are first removed for each residue (Ϫ⌬G m1 Ϫ ⌬G m2 ) and then added back again by (ϩ⌬G wt ). Hence, in an ideal DMC, the coupling free energy is the direct contribution of the salt bridge. To test the above assumptions, some DMC analyses have been carried out alongside structural studies (34,35). Another effective method is to construct several independent DMCs for a specific salt bridge. If the coupling free energy is independent of different DMCs, the assumptions described above are likely to be correct (36,44).
Here, we used the DMC approach to estimate the net strength of the two highly conserved salt bridges, Glu-54/ Arg-57 and Glu-36/Lys-68 in protein Ssh10b. Four independent DMCs were generated for each salt bridge to test the feasibility of this method. The coupling free energies were 2.4 Ϯ 0.4 kJ/mol at 298 K and 2.2 Ϯ 0.4 kJ/mol at 353 K for Glu-54/Arg-57, and 6.0 Ϯ 0.2 kJ/mol at 298 K and 5.9 Ϯ 0.6 kJ/mol at 353 K for Glu-36/Lys-68. Our results demonstrate that, though the stability free energy of Ssh10b decrease greatly with increasing temperature, the direct stability contribution of these two salt bridges remain almost constant, providing evidence supporting the theoretical prediction that salt bridge interactions are extremely resilient to temperature increases and thus are specially suited to promoting protein stability at high temperatures (26). This may be the reason why thermophilic and hyperthermophilic proteins have increasing numbers of salt bridges with increasing growth temperature.
It has been shown that a higher transition temperature in a hyperthermophilic protein can be obtained in three theoretical ways: 1) by shifting the stability curve to increase the overall unfolding free energy at any temperature, 2) by decreasing the ⌬C p between the folded and unfolded states to flatten the stability curve, or 3) by shifting the stability curve toward higher temperatures (45,46). The Ssh10b structure is stabilized by eight isolated salt bridges and two ion-pair networks (each involves three salt bridges). Because salt bridge interactions are extremely resilient to temperature increases, increasing the number of salt bridges in Ssh10b should decrease ⌬C p between the folded and unfolded states of the protein and thus flatten its stability curve.
According to a previous report, Glu-54/Arg-57 and Glu-36/ Lys-68 form two salt bridges, which are located on the surface of the Ssh10b molecule (30). However according to solvent accessible surface area (ASA) calculations, the four residues involved in these salt bridges are only partially solvent exposed. The solvent ASA of side chains of these residues are: Glu-54: 5%, Arg-57: 28%, Glu-36: 40%, and Lys-68: 58%. The two residues of the Glu-54/Arg-57 salt bridge are both located in the second ␣-helix, while those of the Glu-36/Lys-68 salt bridge are far apart in primary sequence and located in two separate ␤-strands (Fig. 1). Because the dielectric constant inside a protein is lower than that of the protein surface, it is generally accepted that the direct stability contribution of a salt bridge buried in a protein should be higher than that of a salt bridge located on the surface (47). However, though the residues involved in the Glu-36/Lys-68 salt bridge are more exposed than those of the Glu-54/Arg-57 salt bridge, the coupling free energies of Glu-36/Lys-68 (6.0 kJ/mol at 298 K and 5.9 kJ/mol at 353 K) are much higher than those of Glu-54/Arg-57 (2.4 kJ/mol at 298 K and 2.2 kJ/mol at 353 K).
The contribution of a salt bridge to protein stability is defined as the free energy of unfolding the protein containing the salt bridge minus that of unfolding the protein without the salt bridge (26). If a salt bridge interaction could exist in both a native and unfolded state, the stability contribution of this salt bridge would be lower than that if the salt bridge only existed in the native state. In the primary sequence of Ssh10b, the Glu-54 and Arg-57 residues are separated by two hydrophobic resi-  favorable conformation should arise when Glu-54 and Arg-57  interact with water, while residues Val-55 and Ile-56 bind to  hydrophobic areas of the protein. This conformation would  increase the frequency of salt bridge contacts formed by Glu-54 and Arg-57. However, the situation for the Glu-36/Lys-68 salt bridge is different: the two residues are far apart in the primary sequence, so the frequency of salt bridge contacts formed by Glu-36 and Lys-68 in the unfolded state is much lower.
To determine whether conclusions reached from our experiments on salt bridges in Ssh10b are applicable to other proteins, we analyzed published DMC experimental data for the contribution of salt bridges to protein stability in other proteins ( Table 2). This confirmed that the average coupling free energy of intrahelical salt bridges (typically formed by two very close charged residues) is lower than that of salt bridges formed between residues far apart in sequence. Five of the proteins included in Table 2 have intrahelical salt bridges; coupling free energies (⌬⌬G coup ) for these salt bridges determined by DMC experiments are: 0.3 and 0.9 kJ/mol for Asp-12/Arg-16 and Glu-28/Lys-32 respectively in Barnase, 3.4 kJ/mol for Asp-14/ Arg-17 in Repressor, 0.4 kJ/mol for Asp-116/Arg-119 in T4 lysozyme, 1.2 and 2.1 kJ/mol for Lys-8/Glu-11 and Glu-11/Lys-15, respectively, in GCN4-p1, and 2.4 kJ/mol for Glu-54/Lys-57 in Ssh10b as reported in this work. The average coupling free energy of salt bridges located in an ␣-helix is only about 1.5 kJ/mol, much lower than that of salt bridges found in other structures (about 5.4 kJ/mol). This finding provides further insight into how to engineer proteins for improved stability at high temperatures; salt bridges introduced between two residues, which are close to each other in tertiary structure but far from each other in primary structure will make a greater contribution to protein stability. Further experimental evidence is required to substantiate these findings.