Thermodynamics of protein denaturation at temperatures over 100 °C: CutA1 mutant proteins substituted with hydrophobic and charged residues

Although the thermodynamics of protein denaturation at temperatures over 100 °C is essential for the rational design of highly stable proteins, it is not understood well because of the associated technical difficulties. We designed certain hydrophobic mutant proteins of CutA1 from Escherichia coli, which have denaturation temperatures (Td) ranging from 101 to 113 °C and show a reversible heat denaturation. Using a hydrophobic mutant as a template, we successfully designed a hyperthermostable mutant protein (Td = 137 °C) by substituting six residues with charged ones. Thermodynamic analyses of these mutant proteins indicated that the hydrophobic mutants were stabilized by the accumulation of denaturation enthalpy (ΔH) with no entropic gain from hydrophobic solvation around 100 °C, and that the stabilization due to salt bridges resulted from both the increase in ΔH from ion-ion interactions and the entropic effect of the electrostatic solvation over 113 °C. This is the first experimental evidence that has successfully overcome the typical technical difficulties.


Results
Hydrophobic mutants of EcCutA1 with no SH group. We constructed hydrophobic mutants with no SH groups (EcCutA1_0SH_S11V, EcCutA1_0SH_E61V, EcCutA1_0SH_S11V/E61V), which we expected to increase stability 30 . Figure 1A shows typical differential scanning calorimetry (DSC) curves of EcCutA1_0SH and its hydrophobic mutant without SH groups at pH 9.0. As shown in Fig. 1B, the reheating curve (second scan) of EcCutA1_0SH agrees completely with the first scan. The other two proteins also exhibited good reproducibility. These results indicate that the removal of SH groups facilitates excellent reversibility of heat denaturation under these conditions and that we can reliably determine the denaturation enthalpies of these proteins. The denaturation temperature (T d ) of EcCutA1_0SH decreased by 4.3 °C, relative to that (89.9 °C) of EcCutA1 with SH groups, whereas those of EcCutA1_0SH_S11V and EcCutA1_0SH_E61V were 103 and 101 °C, respectively, which were remarkably improved relative to the template. Furthermore, the T d of a double mutant, EcCutA1_0SH_S11V/E61V, was 113 °C, which is 28 °C higher than that of the template (Table 1). Hereafter, EcCutA1_0SH_S11V/E61V is abbreviated as Ec0VV. These changes in stability due to the hydrophobic mutations were comparable to those observed in mutant proteins with SH groups 30 . Ionic mutants of EcCutA1_0SH_S11V/E61V (Ec0VV). To examine the thermodynamic parameters of stabilization by ion-ion interaction at temperatures over 100 °C, we constructed several mutant proteins containing substitutions with charged residues, using Ec0VV as a template. Ionic mutants, whose denaturation temperatures are improved and whose DSC curves are suitable for thermodynamic analysis, were selected from our pool of stock mutants (Table 1). Typical DSC curves and reversibility curves are shown in Fig. 2 and Supplementary Fig. 2, respectively. Although the T d of a double mutant, Ec0VV_T17K/S48D, was lower than that of the template, it was selected because the T d was over 100 °C and higher than those of the original single mutants (Ec0VV_T17K and Ec0VV_S48D) ( Table 1). In the case of Ec0VV_A39D/S48K/H72K/S82K/Q87K/T88R, which is abbreviated as Ec0VV_6, the DSC curve was suitable for analysis ( Fig. 2), but the reversibility curve could not be properly obtained due to certain side reactions that occurred at high temperatures. The T d of Ec0VV_6 was 136.8 ± 0.9 °C, improved by 23.6 °C with the introduction of six charged residues. In acidic pH, negatively charged residues of a protein should be protonated, leading to a decrease in conformational stability. In the case of CutA1 from P. horikoshii, which is stabilized by many ionic interactions, the T d of 148.5 °C at pH 7.0 is drastically reduced to 75.6 °C at pH 2.5, whereas the T d of CutA1 from T. thermophilus changes from 112.8 °C at pH 7.0 to 86.6 °C at pH 2.5 28 . To confirm the stabilization resulted from ionic interactions, the stabilities of ionic mutants were examined under acidic conditions at pH 2-3. The T d values of the ionic mutants monotonically decreased as the pH was lowered, reaching a constant minimum at pH 2-2.5 (Supplementary Table 1). We plotted the T d shift (T d value at pH 9.0 vs. pH 2.0-2.5) versus the T d value at pH 9.0 for several ionic mutants (closed circles in Supplementary  Fig. 3). Clearly, the T d shift became greater as T d increased. These results suggest that the electrostatic interactions dominats the thermo-stabilization of the ionic mutant proteins.
Temperature dependence of denaturation enthalpy at higher temperatures. The denaturation heat capacity (Δ Cp) is generally assumed to be constant at temperatures below 80 °C 31 , but it gradually decreases at higher temperatures 14 . Therefore, it is important to elucidate the temperature function of Δ Cp at higher temperatures. To this end, we measured the Cp values of Ec0VV in the native state by DSC at temperatures up to 95 °C (Y2 of Supplementary Fig. 4A). Unfortunately, the temperature dependence of Cp values in the denatured state could not be determined experimentally due to the high reversibility of denatured Ec0VV. Alternatively, assuming that the heat-capacity contribution of amino-acid groups is additive, the heat capacity of proteins in the denatured state can be calculated from their amino-acid composition 32 . Y1 in Supplementary Fig. 4A also shows the temperature function of the heat capacity of Ec0VV in the denatured state, which was estimated from its amino-acid composition using the parameters in Table II of Makhatadze and Privalov 33 . Next, we were able to estimate the temperature function of the denaturation heat capacity (Δ Cp) for Ec0VV from these native and denatured Cp values (Y3 of Supplementary Fig. 4A). The temperature function obtained of Δ Cp can be expanded around T d as a second-order polynomial: Then, the temperature functions of denaturation enthalpy (Δ H) and denaturation entropy (Δ S) can be calculated by the following equations (2) and (3), respectively. Figure 3A shows the temperature function of Δ H for Ec0VV. The Δ H values of Ec0VV were higher than those at each denaturation temperature of other proteins (EcCutA1_0SH, EcCutA1_0SH_S11V, and EcCutA1_0SH_E61V). If we assume that the temperature function of Δ Cp is not largely affected by the constitution of the protein, this observation indicates that stabilization of hydrophobic mutants at residue positions 11 and 61 is mainly caused by enthalpic effects.
The temperature function of Δ Cp for Ec0VV_6 was also determined using the native Cp values ( Supplementary Fig. 4B), which were directly measured up to 110 °C. Figure 3B shows the temperature function of Δ H for Ec0VV_6 and the denaturation enthalpy values at the denaturation temperatures of several ionic Ec0VV mutants. This Figure indicates that the Δ H value of Ec0VV is similar to those of Ec0VV_S110R and Ec0VV_6 at each denaturation temperature, but remarkably higher than those of other mutants derived from Ec0VV by the further addition of charged residues.
The thermodynamic parameters of denaturation for EcCutA1_0SH mutants at the denaturation temperature (113.2 °C) of Ec0VV are listed in Table 2. The Δ G values, estimated using the Δ Cp temperature function obtained from Ec0VV, agreed well with those from Ec0VV_6, around the denaturation temperature of Ec0VV. Figure 4 also shows the temperature functions of Δ G, Δ H, and TΔ S for Ec0VV and Ec0VV_6 over a larger temperature range (between 280 and 420 K), indicating that Δ G values of Ec0VV_6 are positive over a broad range of temperatures.

Discussion
Forty years ago, Privalov et al. 14,31 developed highly qualified adiabatic differential micro-calorimeters for determining the thermodynamic parameters of protein denaturation. They reported that specific characters of amino-acid residues disappear during protein unfolding near 110 °C, and that all the observed entropy originates from the increase in conformational freedom of the polypeptide upon unfolding, because the specific Δ H and Δ S of unfolding for several proteins intersect at a single point near 110 °C. On the other hand, the thermodynamics of transfer of hydrocarbons to water provides a model for the temperature dependence of the hydrophobic interaction in protein folding. Baldwin 15 examined the solution thermodynamics of several liquid hydrocarbons in water. He found that the extrapolated temperatures at which the transfer Δ S reaches zero, around 112.8 °C, were similar for six hydrocarbons. The extrapolated temperature of the transfer Δ H is 22.0 °C. This means that at 113 °C, the hydrophobic interaction changes from being entropy-driven at 22 °C to being enthalpy-driven at 113 °C,  Table 1. The black curves represent the temperature function of Δ H upon denaturation using the temperature function of Δ Cp for Ec0VV obtained from Y3 of Supplementary Fig. 4A. The red curve represents TΔ S of Ec0VV. In the case of Ec0VV, the parameters A, B, and C of Δ Cp (in kJ mol −1 K −1 ) in equation (1) Table 2. Thermodynamic parameters of denaturation for EcCutA1_0SH mutants at the denaturation temperature (113.2 o C) of Ec0VV. The unit is kJ mol −1 . * a and b represent the calculated results using the temperature function of Δ Cp obtained from Ec0VV and Ec0VV_6, respectively. and the contribution of water to the entropy of protein unfolding (hydrophobic hydration) is removed 15 . Regarding these estimations, the heat capacity change (Δ Cp) is assumed to be constant against temperature. Later, Makhatadze and Privalov 33,34 reported that the temperature at which Δ S is zero approaches 145 °C when the decreasing nature of Δ Cp against temperature is taken into account, because the Δ Cp value of hydrocarbon hydration decreases with increasing temperature. In this study, the heat capacities of the native states of Ec0VV and Ec0VV_6 could be directly measured by DSC up to 95 and 110 °C, respectively, and the temperature functions of their Δ Cp values were estimated as shown in Supplementary Fig. 4A and Supplementary Fig. 4B, respectively.
Hydrophobic effects strongly contribute to ΔH at temperatures around 100 °C. The temperature function of Δ H of Ec0VV is depicted in Fig. 3A. Using the same temperature function of Δ Cp, the Δ H values of Ec0VV, EcCutA1_0SH_S11V, EcCutA1_0SH_E61V, and EcCutA1_0SH were estimated to be 1569, 1396, 1340, and 1175 kJ/mol at 113.2 °C, respectively ( Table 2). The increase in Δ H (394 kJ/ mol) of Ec0VV (EcCutA1_0SH_S11V/E61V) agrees well with the sum (386 kJ/mol) of the increases in Δ H of EcCutA1_0SH_S11V (221 kJ/mol) and EcCutA1_0SH_E61V (165 kJ/mol). On the other hand, the TΔ S values of Ec0VV (the red curve in Fig. 3A) were larger than the Δ H value (= TΔ S at T d ) at each T d of EcCutA1_0SH_S11V, EcCutA1_0SH_E61V, and EcCutA1_0SH, indicating that Ec0VV is entropically unfavorable compared with other proteins. That is, Ec0VV, which contains hydrophobic substitutions for hydrophilic residues in the interior of a molecule, is mainly stabilized by the enthalpic gain upon substitutions, but is partly destabilized by the entropic loss, probably due to the disruption of the hydrophilic solvation in the denatured state upon substitutions. These results at high temperatures around 100 °C are contrary to the well-accepted belief that the entropic gain from hydrophobic solvation can account for the stabilization effect of hydrophobic substitutions at lower temperatures 15 . Whereas, the estimations from the hydration of amino acids 34 are generally consistent with our results. This is the first experimental evidence pertaining to the hydrophobic effects on protein stability, which is obtained by direct measurement at temperatures around 100 °C.
Ec0VV_6 substituted with six charged residues is stabilized by both enthalpic and entropic effects around 137 °C. Below 100 °C, the ion-ion interaction (salt bridge) is driven entirely by entropic effects due to the release of water strongly bound to the ions of charged residues [35][36][37] . The proteins substituted with single charged residues, Ec0VV_H72K, Ec0VV_S82K, Ec0VV_Q87K, and Ec0VV_T88R, are stabilized by electrostatic interactions (Supplementary Fig. 3). The Δ H values of all of these proteins were drastically decreased relative to Δ H at each corresponding temperature on the temperature function of Ec0VV_6 (Fig. 3B). Because the changes in Δ H upon these mutations are unfavorable for folding, the observed improvements in stability were caused by entropic effects due to the release of water at the charged residues that we introduced (electrostatic solvation). The thermodynamic analyses also clearly confirmed stabilization resulted from entropic effects ( Table 2). The other single mutants, Ec0VV_A39D and Ec0VV_S48 K, whose mutation sites were located in the interior of the molecule (Table 3), demonstrated drastically decreased T d and Δ H values. However, a double mutant containing both of these single mutants, Ec0VV_A39D/S48 K, in which Asp39 forms a strong salt bridge with Lys48, was stabilized by an increase in Δ H relative to the single mutants (Fig. 3B), probably due to Coulomb's force resulting from salt bridge formation. A similar result was obtained in the other double mutant, Ec0VV_T17 K/S48D (Fig. 3B). The double mutant Ec0VV_Q87 K/T88R was also stabilized by an increase in Δ H relative to Ec0VV_Q87 K and Ec0VV_T88R. Because this double mutant does not have additional salt bridges, two individual thermo-stabilizations might work synergistically at around 120 °C to promote the desolvation of the ionic residues introduced, thereby reducing both the enthalpic loss and the entropic gain that are mutually attributed to the electrostatic solvation. In addition, the increase in Δ H in this double-ion mutant might have been caused by a hydrophobic interaction due to the alkyl groups of Lys87 and Arg88.
Ec0VV_6, with six additional charged residues, was stabilized by an increase in the Δ H value relative to each of the six original ionic mutants (Fig. 3B). The increase in Δ H might indicate that hydrophobic effects due to the alkyl groups of Lys or Arg and Coulomb's force still function effectively at these high temperatures. According to Coulomb's law, the strength of an electrical interaction is inversely proportional to the dielectric constant. The dielectric constant of water drops from 80 at 0 °C to 55 at 100 °C 38 . Furthermore, Elcock 39 found that increasing temperature decreases the electrostatic desolvation penalty incurred in forming a salt bridge, leading to an increase in salt bridge stabilization from his continuum solvation model 40 . The increase in Δ H of Ec0VV_6 might result mainly from the high degree of desolvation of the ionic residues in the denatured state at around 137 °C, in addition to the other effects described above.
The temperature dependence of Δ H for Ec0VV_6 has an intersection near T d (113 °C) of Ec0VV, as shown in Fig. 4, suggesting that the contribution of Δ H to the stability of Ec0VV_6 with additional 6 charged residues becomes favorable in the temperature region above 113 °C, compared with those of Ec0VV. Furthermore, Fig. 4 shows that the increase in Δ G of Ec0VV_6 results largely from the decrease in Δ S of Ec0VV_6 when compared with the template Ec0VV at temperatures below 113 °C. That is, the stabilization due to ionic mutations results mainly from both the enthalpic gain from ion-ion interactions in the native state and the entropic gain from the water release of ionic residues in the denatured state at temperatures over 113 °C.
A mutant, Ec0VV_S110R, whose substitution position is located in the C-terminus of α-helix and almost buried (Table 3), was stabilized by an increase in Δ H (Fig. 3B). In this case, due to local conformations, the decrease in Δ H due to water release might be suppressed by the effects of other stabilizing factors.

Conclusion
The rational design of hyper-thermostable proteins may be possible through the introduction of multiple salt bridges at high temperatures over 113 °C that can be reached through a preceding stabilization by hydrophobic substitutions.

Matrials and Methods
Mutagenesis, expression, and purification of CutA1 mutants from E. coli. The mutagenesis, expression, and purification of CutA1 mutants from E. coli were performed as described 30 with minor modifications. The homogeneity and identity of the purified samples were assessed using SDS-PAGE. The protein concentration was estimated from the absorbance at 280 nm, assuming E 1cm 1% = 14.96, based on the number of aromatic amino acids 41 .

Differential scanning calorimetry (DSC) experiments.
To measure the changes in stability due to mutations, DSC was performed using a scan rate of 60 °C/h on a VP-capillary DSC platform (Microcal, USA) for temperatures up to 130 °C at pressures below 60 psi, or a Nano-DSC 6300Y microcalorimeter (TA Instruments, USA) for higher temperatures up to 160 °C at a pressure of 88 psi. Protein concentrations were around 0.6 mg/ml in a 50 mM glycine buffer at pH 9.0 containing 2 mM EDTA or a 50 mM  Table 3. Burial rates of target residues of Ec0VV mutants * . * The burial rates (%) were estimated from the average of ASA values for nine structures during 40 ns MD at 300 K (in preparation).
Scientific RepoRts | 5:15545 | DOi: 10.1038/srep15545 glycine buffer at pH 2.0-3.5. All samples were dialyzed against the buffers overnight at 4 °C and then filtered through a membrane with 0.22-μ m pores. The denaturation temperature (T d ) is the temperature at which the area of the denaturation enthalpy (Δ H) is 0.5. The T d and Δ H values in this study represent the averages for at least six experiments.
To measure the heat capacity of mutant proteins in their native states, the protein concentrations were adjusted to around 2.0 mg/ml in a 50 mM glycine buffer at pH 9.0. Two different scan rates, 60 and 200 °C/h, were used. Each experiment comprised six cycles of reheating to the pre-denaturation temperatures: 95 °C and 110 °C for Ec0VV and Ec0VV_6, respectively. The partial specific volumes for the calculation of heat capacity were estimated from the amino-acid composition of each mutant protein 42 .