Role of charged residues in stabilization of Pyrococcus horikoshii CutA1, which has a denaturation temperature of nearly 150 °C

Authors


K. Yutani, RIKEN SPring-8 Center, RIKEN Harima Institute, 1-1-1 Kouto, Sayo, Hyogo 679-5148, Japan
Fax: +81 791 58 2917
Tel: +81 791 58 2937
E-mail: yutani@spring8.or.jp

Abstract

The CutA1 protein from Pyrococcus horikoshii (PhCutA1), a hyperthermophile, has an unusually high content of charged residues and an unusually high denaturation temperature. To elucidate the role of ion–ion interactions in protein stability, mutant proteins of PhCutA1 in which charged residues were substituted by noncharged residues were comprehensively examined. The denaturation temperatures (Td) for 13 of 53 examined mutant proteins were higher than that of the wild-type (148.5 °C at pH 7.0), among which E99Q had the highest Td at 154.9 °C. R25A had the largest decrease in Td among single mutants at ΔTd = −12.4 °C. The average decrease in Td of Lys or Arg mutants was greater than that of Glu or Asp mutants, and the average change in TdTd) of 21 Glu mutants was negligible, at 0.03 ± 2.05 °C. However, the electrostatic energy (−159.3 kJ·mol−1) of PhCutA1 was quite high, compared with that of CutA1 from Escherichia coli (−9.7 kJ·mol−1), a mesophile. These results indicate that: (a) many Glu and Asp residues of PhCutA1 should be essential for highly efficient interactions with positively charged residues and for generating high electrostatic energy, although they were forced to be partially repulsive to each other; (b) the changes in stability of mutant proteins with a Td value of ∼ 140–150 °C were able to be explained by considering factors important for protein stability and the structural features of mutant sites; and (c) these findings are useful for the design of proteins that are stable at temperatures > 100 °C.

Database

  • Structural data are available from the Protein Data Bank under the accession codes for 1v99 and 1naq of CutA1 proteins from P. Horikoshii and E. coli, respectively

Abbreviations
DSC

differential scanning calorimetry

EcCutA1

CutA1 protein from Escherichia coli, a mesophile

PhCutA1

CutA1 protein from Pyrococcus horikoshii, a hyperthermophile

Introduction

The proteins of hyperthermophiles, which have optimal growth temperatures > 80 °C, are significantly more stable than those of mesophilic organisms, which grow at moderate temperatures near 37 °C. Genomic analysis has shown that a large difference between the proportions of the numbers of charged residues (Asp, Glu, Lys, Arg) and polar residues (Asn, Gln, Ser, Thr), which is called the CvP-bias, is the dominant proteome characteristic of microorganisms adapted to hyperthermophilic growth [1]. Many reports have shown that an abundance of ion pairs and ionic networks formed by charged residues contributes to the stabilization of proteins from hyperthermophiles [2–13]. However, negative contributions of salt bridges to protein stability have been reported [14,15]. The conformational stability of proteins, even from hyperthermophiles, is a result of a delicate balance of different interactions, including hydrophobic effects, hydrogen bonds, electrostatic interactions, hydration effects and entropic effects. The thermodynamic energy of protein stability is marginal, around 50 kJ·mol−1 [16–18], which corresponds to only several moles of stabilizing factors such as hydrogen bonds and hydrophobic interactions [19–23]. However, how the abundance of ion pairs contributes to the conformational stability of proteins from hyperthermophiles remains unclear.

It has recently been reported that the CutA1 protein from the hyperthermophile Pyrococcus horikoshii (PhCutA1) has an unusually high stability, with a denaturation temperature (Td) of nearly 150 °C at pH 7.0, which is ∼ 30 °C greater than the highest record determined by differential scanning calorimetry (DSC) [13]. The CutA1 protein was originally identified as the product of the cutA gene locus of Escherichia coli, and it is involved in divalent metal tolerance [24]. The X-ray crystal structure (PDB ID; 1v99) of PhCutA1 (Fig. 1) is a tightly intertwined trimer, which clearly resembles that (PDB ID; 1naq) of CutA1 from Escherichia coli (EcCutA1). However, the amino acid compositions of the two proteins are quite different, especially in the proportions of charged and polar residues for the CvP-bias: 40.2% (8 Asp, 16 Glu, 11 Lys and 6 Arg in 102 residues) and 7.8% (1 Ser, 5 Thr, 2 Asn and 0 Gln) in PhCutA1; 18.8 and 23.2% in EcCutA1, respectively. Thus, the CvP-bias values of PhCutA1 and EcCutA1 are 32.4 and −4.4%, respectively. The CvP-bias of PhCutA1 is more than double the average (14.7%) of proteins from P. horikoshii [1], suggesting that the content of charged residues in PhCutA1 is unusually large even among proteins from hyperthermophiles. Therefore, PhCutA1 may represent a good model for investigating the role of charged residues in the stability of proteins at temperatures > 100 °C.

Figure 1.

 The trimer crystal structure of PhCutA1 (1v99). Different colors represent different chains. α and β represent α helix and β strand, respectively.

In this study, using PhCutA1, we performed systematic and exhaustive analysis of the stability of mutant proteins by DSC to elucidate the role of ion–ion interactions in protein stability. To confirm the contribution of charged residues to protein stability, the electrostatic energy of ion–ion interactions was evaluated by the computer algorithm, FoldX [25]. FoldX, available via a web-interface at http://foldx.crg.es/, can quantitatively estimate the factors contributing to protein stability and protein interactions. The FoldX energy function also includes stabilization factors that are important for protein stability. From the changes in stability of > 50 PhCutA1 mutants, we discuss the role of ion–ion interactions in hyperthermophile proteins in conformational stability at temperatures > 100 °C.

Results

Design of PhCutA1 mutants

All of the mutant variants of PhCutA1 in this study are listed in Table 1. The variants were constructed with the following intentions. (a) Previous analysis has shown that the increased number of ion pairs in the monomeric structure of PhCutA1 contributes to stabilization of the trimeric structure [13]. Therefore, Glu and Asp residues (Glu15, Glu24, Glu47, Asp60, Glu63, Glu64, Glu67, Glu71, Asp84 and Glu99) having ion pairs within 4 Å on the same subunit of PhCutA1 were substituted by Gln or Ala and Asn or Ala, respectively, in order to confirm the role of intrasubunit ion pairs on the same subunit in the stability of PhCutA1. (b) The electrostatic energies between the ion groups formed by each charged residue, evaluated by FoldX, are shown in Fig. S1 and are listed in order of favorable and unfavorable interactions (15 residues each) in Table 2. To elucidate how much the ion–ion interactions of charged residues contribute to protein stability, the charged residues listed in Table 2 were single-substituted by noncharged residues. (c) Double mutants of ion-paired residues and multiple mutants were also constructed (Table 1) in order to confirm the contribution of the charged residues as a pair to stability.

Table 1.   Denaturation temperatures (Td in °C) of PhCutA1 mutants and the difference in TdTd) between the wild-type and its mutant at pH 7.0. Td is the average value of more than two data. Positive values of ΔTd indicate the increase in stability due to mutations. The Td of the wild-type is 148.5 °C.
MutantsTdΔTdMutantsTdΔTdMutantsTdΔTd
  1. a Ph14S represents E12Q/K16A/K44A/E46Q/D48N/K49A/R58A/D60N/E63Q/E67Q/E71Q/D84N/E90Q/K94A.

E12Q149.20.7D10N149.40.9K19A145.9−2.6
E15A147.7−0.8D48N151.22.7K44A148.1−0.4
E15Q148.4−0.1D60A145.2−3.3K49A149.30.8
E24A145.5−3.0D60N145.0−3.5K66A142.0−6.5
E24Q146.8−1.7D76N149.40.9K70A145.1−3.4
E34Q147.1−1.4D84A147.0−1.5K101A144.1−4.4
E42Q153.44.9D84N148.0−0.5R25A136.1−12.4
E46Q152.13.6D86N151.02.5R33A139.3−9.2
E47A146.0−2.5D87N141.5−7.0R33M140.8−7.7
E47Q147.4−1.1D91A141.9−6.6R36A146.6−1.9
E59Q152.23.7D91N140.1−8.4R58A141.7−6.8
E63A149.00.5   R68A146.4−2.1
E63Q147.0−1.5   R82A138.5−10.0
E64A145.6−2.9   R25A/E99Q146.1−2.4
E64Q145.0−3.5   R58A/D60N141.7−6.8
E67A149.00.5   R68A/E24Q146.2−2.3
E67Q148.0−0.5   K70A/D91N143.4−5.1
E71A148.0−0.5   K66A/D87N138.6−9.9
E71Q147.0−1.5   K101A/E64Q145.3−3.2
E99A150.01.5   E15Q/K19A/R36A/E47Q145.7−2.8
E99Q154.96.4   Ph14Sa141.3−7.2
Table 2.   The energy of ion–ion interactions and changes in Td due to mutations at each charged residue of PhCutA1. Positive values of ΔTd indicate the increase in stability due to mutations. The energies of ion–ion interactions come from Table S1.
In the order of favored to unfavoredIn the order of unfavored to favored
PositionsEnergy of ion–ion intereaction (kJ·mol−1)MutantsΔTd (°C)PositionsEnergy of ion–ion interaction (kJ·mol−1)MutantsΔTd (°C)
ARG82−35.7R82A−10.0ASP8619.6D86N2.5
ARG68−28.0R68A−2.1GLU1211.6E12Q0.7
ARG58−21.8R58A−6.8ARG3310.9R33A−9.2
LYS66−21.5K66A−6.5ASP1010.6D10N0.9
LYS70−16.2K70A−3.4ASP4810.5D48N2.7
GLU24−15.6E24Q−1.7ASP847.2D84N−0.5
ARG36−13.5R36A−1.9GLU465.5E46Q3.6
ARG25−12.0R25A−12.4GLU675.2E67Q−0.5
GLU34−11.0E34Q−1.4GLU635.1E63Q−1.5
LYS49−10.2K49A0.8GLU153.8E15Q−0.1
GLU99−9.1E99Q6.4ASP763.3D76N0.9
ASP60−8.8D60N−3.5GLU473.1E47Q−1.1
GLU59−7.3E59Q3.7GLU422.5E42Q4.9
LYS101−6.7K101A−4.4LYS231.7 
GLU71−5.7E71Q−1.5ASP871.5D87N−7.0

The thermal stability of PhCutA1 mutants

To examine changes in stability due to mutations, the heat stability of mutant variants was measured using DSC. Typical DSC curves at pH 7.0 are shown in Fig. 2. As shown in the figure, the DSC curves seem to be slightly aggregated after heat denaturation; therefore, the apparent peak temperatures of the DSC curves were taken as the denaturation temperature (Td). The Td values listed in Table 1 are averages of more than two experiments. The Td value of E99Q was 154.9 °C, the highest among the examined mutant proteins, and 6.4 °C greater than that of the wild-type protein. In fact, the E99Q mutant set a new world record for the protein with the highest heat stability [13]. The Td values of 13 variants among the 45 single mutant proteins were higher than that of the wild-type protein, as shown in Table 1. These results suggest that even a protein with a Td of 150 °C has many sites that could be altered to further increase its stability. The greatest decrease in Td due to a single amino acid substitution was 12.4 °C for R25A. The Ph14S mutant is the PhCutA1 variant E12Q/K16A/K44A/E46Q/D48N/K49A/R58A/D60N/E63Q/E67Q/E71Q/D84N/E90Q/K94A, which was substituted at all 14 sites of intrasubunit ion interactions within 4 Å, a pair of which is closely located in the sequence (Fig. 3). The ΔTd of Ph14S was −7.2 °C.

Figure 2.

 Typical DSC curves of PhCutA1 mutant proteins at pH 7.0. Heating rates of DSC were 60 °C·h−1. The perpendicular line shows the Td of the wild-type protein.

Figure 3.

 Intra/intersubunit ion pairs within 5 Å. Solid and broken lines represent intra- and intersubunit ion pairs, respectively.

The changes in Td of Glu and Asp mutant variants (at Glu15, Glu24, Glu47, Asp60, Glu63, Glu64, Glu67, Glu71, Asp84 and Glu99) forming intrasubunit ion pairs on the same subunit of PhCutA1 ranged from −3.5 (E64Q and D60N) to 6.4 °C (E99Q) (Table 1). The average value of these 20 samples forming intrasubunit ion pairs on the same subunit was −1.0 ± 1.5 °C. However, the average changes in Td of all 21 Glu mutants in Table 1 was 0.03 ± 2.05 °C, while the average changes of 11 Asp and 13 positively charged (Lys and Arg) mutants were −2.17 ± 3.27 °C and −5.14 ± 3.38 °C, respectively. These results suggest that the Glu residues of PhCutA1 contribute less to protein stability than other ionic residues.

Discussion

Figure 1 shows the crystal structure of PhCutA1, which clearly resembles the structures of CutA1 proteins from other sources. The monomeric structure consists of three α helices and five β strands. The monomers are assembled into a trimer through interactions between the edges of three β strands. This tightly intertwined interaction seems to contribute to the stabilization of the trimer structure for CutA1 proteins [26]. There are 43 charged residues (8 Asp, 16 Glu, 11 Lys, 2 His, 6 Arg) in each subunit of PhCutA1 (102 residues). The total energies of ion–ion interactions for each charged residue evaluated by FoldX are shown in Fig. S1, and Table S1 lists all values of each pair of interactions. It might be important to elucidate how electrostatically favorable (or unfavorable) residues contribute to protein stability. For example, PhCutA1 has 30 intrasubunit ion pairs within 5 Å whereas EcCutA1 has only one. By contrast, PhCutA1 has 16 intersubunit ion pairs within 5 Å, whereas EcCutA1 has 17 [13] even though the number of intersubunit ion pairs for oligomeric proteins seems to increase with thermostability [2,3,10]. Figure 3 shows intra/intersubunit ion pairs within 5 Å of PhCutA1 [13]. Solid and broken lines represent intra- and intersubunit ion pairs, respectively. In this study, the stability of PhCutA1 was extensively examined using mutant proteins in which charged residues were changed to noncharged residues in order to elucidate the contribution of electrostatic interactions to protein stability at temperatures > 100 °C. The changes in stability can be elucidated based on relationships with their local structures as follows.

The role of intrasubunit ion pairs in protein stability

Eight PhCutA1 Glu residues at positions 15, 24, 47, 63, 64, 67, 71 and 99, which form intrasubunit ion interactions within 4 Å with positively charged residues, were substituted by Gln and Ala. Two Asp residues at positions 60 and 84 were substituted by Asn and Ala. The Td values of the Glu24, Glu64 and Asp60 variants decreased by ∼ 3 °C. Changes in stability of the other mutations, except for Glu99, were slight (< 3 °C). These results seem to indicate that intrasubunit interactions at the examined positions do not play a very important role in stability, contrary to the expectations in the previous report [13], because changes in Td due to the deletion of charged residues were small and changes in Td due to Ala substitutions were similar to those of Gln or Asn substitutions.

The energy of ion–ion interactions estimated by FoldX suggests that charged residues at positions 24, 60, 71 and 99 have favorable interaction (Fig. S1 and Table S1). The decrease in stability of mutant proteins at positions 24 and 60 corresponded to the FoldX estimates, but the Td of E99Q was 154.9 °C (ΔTd = 6.4 °C), showing that Glu99 is unfavorable for stability even though the electrostatic energy estimated by FoldX is favorable at this position. Glu99 is located at the C-terminus of the α3 helix and forms a salt bridge with Arg25 in loop 2 (Fig. 4A). The changes in Td of R25A and R25A/E99Q were −12.4 and −2.4 °C, respectively, indicating that the decrease in stability of the double mutant is largely compensated by the E99Q substitution. The other charged residues near Arg25 and Glu99 provide only limited effect on stability compared with the electrostatic energy between Arg25 and Glu99 (by the analysis of FoldX) (Table S1). Arg25 forms favorable interactions with Glu99 (−12.0 kJ·mol−1) and Glu98 (−1.3 kJ·mol−1), and an unfavorable interaction with Arg58 (1.4 kJ·mol−1). Glu99 forms favorable interactions with Arg25 (−12.0 kJ·mol−1) and Arg58 (−1.1 kJ·mol−1), and unfavorable interactions with Glu42 (1.3 kJ·mol−1) and Glu98 (2.9 kJ·mol−1). Therefore, the increase in the stability of E99Q is not primarily caused by changes in electrostatic interactions. Protein α helices have been reported to have dipole moments due to the alignment of peptide bond dipoles; a negatively charged group at the C-terminus of an α helix can therefore destabilize the native structure of a protein [27–29]. The mutant E99Q should be stabilized by deletion of the negatively charged group at the C-terminus of the α3 helix in addition to elimination of the repulsive interactions with Glu42 and Glu98. Furthermore, FoldX suggests that the ‘side-chain hydrogen bonds’ and ‘solvation of polar atoms’ energies in Eqn (1) in addition to ‘helix dipole’ are favorable for ΔΔG of E99Q. This may be because the carboxyl group of Glu99 is partially buried. These positive factors for E99Q might largely surpass the strong electrostatic energy of the interaction with Arg25.

Figure 4.

 The structure of PhCutA1 at specific sites. (A) Near the Arg25–Glu99 ion pair. (B) The Glu47–Lys19–Glu15–Arg36 ion network. (C) Ion networks of centrally located Arg82 and Lys66. (D) The Arg68–Glu24 ion pair. (E) Ion pairs of Arg58. (F) The ion group of Arg33 seems to be repulsive to those of Arg33 in the other chains in the interior of the molecule. (G) Buried Arg33 favorably interacts with other residues. (H) Apparently unfavorable charged residues, Glu42, Glu46, and Asp48.

The role of intersubunit ion pairs in protein stability

The Glu residues at positions 15 and 47 form intersubunit ion pairs in addition to intrasubunit pairs, forming an ion network with Lys19 and Arg36 (Figs 3 and 4B). Single mutants of paired residues and a quadruple mutant of all of the residues in the network were constructed. The differences in TdTd) of E15Q, K19A, R36A, E47Q and E15Q/K19A/R36A/E47Q were −0.1, −2.6, −1.9, −1.1, and −2.8 °C, respectively (Table 1). Figure 4B shows the location of these four charged residues, which seem to strengthen the interactions between segments of different subunits, the α1 helix of the A chain and the β2 (or β3) strand of the C chain. The change in stability of the quadruple mutant is close to the highest value of the single mutants, indicating that this network contributes only slightly to stability. The ionic energies of Lys19 and Arg36 (−3.6 and −13.5 kJ·mol−1, respectively) contribute favorably to the stability, but the ionic energies of Glu15 and Glu47 (3.8 and 3.1 kJ·mol−1) contribute negatively, as estimated by FoldX (Fig. S1), resulting in a slight contribution regardless of the ion network between different subunits. Table S1 indicates that Glu15 forms favorable interactions with Arg36 (−5.4 kJ·mol−1) and Lys19 (−1.7 kJ·mol−1) and unfavorable interactions with Glu47 (4.9 kJ·mol−1), Asp10 (1.5 kJ·mol−1), Glu12 (3.9 kJ·mol−1) and Glu34 (1.4 kJ·mol−1). Glu47 forms favorable interactions with Arg36 (−3.8 kJ·mol−1) and Lys19 (−3.1 kJ·mol−1) and unfavorable interactions with Glu46 (2.1 kJ·mol−1), Asp48 (2.0 kJ·mol−1) and Glu15 (4.9 kJ·mol−1) (Table S1). These multiple negative and positive interactions might originate from the unusually high number of charged residues in a small protein (102 residues).

However, substitutions for Asp residues at positions 87 and 91 dramatically affected the Td, even though changes in Td for other Glu and Asp substitutions were less than about 3 °C, as described (Table 1). Asp87 in the β5 strand of the A chain forms intersubunit ion pairs with Arg82 in the β4 strand and Lys66 in the α2 helix of the C chain, and Asp91 at the N-terminus of the α3 helix of the A chain interacts with Lys 70 in the α2 helix of the C chain (Fig. 4C). The ΔTd values of D87N, D91N, K66A, K70A and R82A were −7.0, −8.4, −6.5, −3.4 and −10.0 °C, respectively. ΔTd values of the K66A/D87N and K70A/D91N double mutants were −9.9 and −5.1 °C, respectively (Table 1). The dramatic decrease in Td for the D91N mutant compared with the other Glu and Asp mutants can be explained by decreases in both electrostatic energy and helix dipole moment at the N-terminus of the α3 helix. The change in Td for D87N might be caused by the close connection of Asp87 with two counter-pairs (Arg82 and Lys66) as described below.

Changes in denaturation temperature of mutant variants at the top 15 electrostatically favorable and unfavorable sites

It is important to elucidate how much the ion–ion interactions of charged residues contribute to protein stability. Table 2 lists the energies of the ion–ion interactions of charged residues in the order of favored to unfavored (and unfavored to favored), evaluated by FoldX. The most favored energy is −35.7 kJ·mol−1 for Arg82 and the most unfavored is 19.6 kJ·mol−1 for Asp86. The favored and unfavored residues listed in Table 2 were substituted by noncharged residues. Some charged residues provided favorable electrostatic energies and contributed to stability, such as Arg82. However, the stability of the obtained mutant proteins did not change as expected. For example, the change in Td of R68A was small even though the electrostatic energy of Arg68 was quite favorable, and the decrease in Td of R33A was large even though the energetically unfavorable charged group of Arg33 was eliminated. Next, changes in stability of these mutant proteins will be discussed on the basis of structural features near the mutation sites.

Arg82 is located in the β4 strand (Fig. 4C), and has the most favorable energy of all the charged residues (Table 2). Arg82 forms strong salt bridges with Asp84 (electrostatic energy of −15.6 kJ·mol−1), Glu59 (−2.7 kJ·mol−1) and Glu63 (−4.0 kJ·mol−1) in the same chain, and with Asp86 (−6.2 kJ·mol−1) and Asp87 (−4.9 kJ·mol−1) in the other chain (Table S1). Because these ion–ion interactions are eliminated when Arg82 is replaced by Ala, the Td of R82A decreased by 10.0 °C. Lys66 in the α2 helix close to Arg82 in the β4 strand also has high electrostatic energy (Table 2). Lys66 forms strong salt bridges with Glu63 (−1.3 kJ·mol−1) and Glu67 (−3.0 kJ·mol−1) in the same α2 helix, and with Asp87 (−7.2 kJ·mol−1), Glu90 (−9.0 kJ·mol−1) and Asp91 (−1.5 kJ·mol−1) in another subunit (Fig. 4C). Replacement of Lys66 by Ala resulted in a significant decrease in Td (6.5 °C). Arg82 links Lys66 via Asp87 in a centrally located ion network that contributes to the unusually high stability of PhCutA1.

By contrast, Asp86, Asp84, Glu67 and Glu63, which are involved in the ion network of Arg82 and Lys66, have unfavorable electrostatic energies (Table 2). Substitutions at positions D84N, E67Q and E63Q showed only slight decreases in Td, suggesting that the repulsive energy of these charged residues hardly affected protein stability. In the case of Asp86, the repulsive energy was great, because the Asp86 residues in the A and C chains are close to each other as shown in Fig. 4C. Therefore, D86N had a slightly higher Td (Table 2) due to elimination of a charged group. The residues in the least favored 13 of electrostatic energy in Table 2 are negatively charged (Glu and Asp), except for Arg33. These results indicate that the number of negatively charged residues, Glu and Asp, surpasses that of positively charged residues, Lys and Arg, so the ion–ion interactions of many Glu and Asp residues should be forced to be partially repulsive to each other.

The electrostatic energies of Arg68, Glu24 and Glu71 are −28.0, −15.6 and −5.7 kJ·mol−1, respectively, higher than those of the other residues (Table 2). However, the decreases in Td for R68A, E24Q and E71Q are small, −2.1, −1.7 and −1.5 °C, respectively. As shown in Fig. 4D, Arg68 and Glu24 form a salt bridge between the α1 and α2 helices, and Glu71 forms a salt bridge with Arg68 in the same helix. The favorable electrostatic energies of Arg68 with Glu24, Glu71 and Glu64 were evaluated by FoldX to be −16.6, −11.4 and −2.2 kJ·mol−1, respectively. Arg68 has unfavorable interactions with Lys16 (1.2 kJ·mol−1) and Lys101 (1.0 kJ·mol−1). Regardless of the high favorable electrostatic energy of Arg68, the decrease in Td of R68A was only 2.1 °C. According to ΔΔG analysis by FoldX, this compensation is mainly caused by ‘solvation of polar atoms’ and ‘entropy effect of side chain’ energies in Eqn (1). Glu24 and Glu71, paired with Arg68, also have favorable electrostatic energies of −15.6 and −5.7 kJ·mol−1, respectively (Table 2). The E24Q and E71Q mutant variants showed slight decreases in Td of 1.7 and 1.5 °C, respectively. Glu24 and Glu71 are located at the C-terminus of the α1 helix and in the second position from the C-terminus of the α2 helix, respectively (Fig. 4D). Because the substitution of Glu with noncharged Gln at the C-terminus of a helix enhances the energy of the helix dipole, the decrease in stability due to the elimination of ion pairs might be compensated by the increase in the helix dipole in the stability of both mutant proteins.

Arg58 and Asp60 have high favorable electrostatic energies of −21.8 and −8.8 kJ·mol−1, respectively (Table 2). Arg58, which is at the turn between the β3 strand and the α2 helix, strongly interacts with Asp60 at the N-terminus of the α2 helix (Fig. 4E). The electrostatic energy was −15.4 kJ·mol−1 between them (Table S1). Furthermore, Arg58 interacts weakly with Glu59 (−2.5 kJ·mol−1) and the carboxyl group of the C-terminal Lys102 residue (−3.2 kJ·mol−1). The decreases in Td for the R58A, R58A/D60N, D60N and D60A variants were 6.8, 6.8, 3.5 and 3.3 °C, respectively. Comparably large changes in stability due to substitution of Asp60 with nonionizable residues might result from the sum of the elimination of ion–ion interactions and the decrease in helix dipole moment. The ΔTd of the R58A/D60N double mutant hardly changed, compared with R58A, because of the elimination of repulsive interactions of Asp60 with Glu59, Glu63, Glu64 and C-terminus of Lys102 (Table S1). Glu59 also showed high electrostatic energy (−7.3 kJ·mol−1) (Table 2), and given its position in the N-terminal region of the α2 helix, should also stabilize the wild-type protein by the helix dipole moment. However, the Td of E59Q increased by 3.7 °C contrary to expectations. This might be explained by the following. (a) The electrostatic energy of Glu59 was primarily evaluated from the interaction with the amino group of the N-terminal Met (−12.8 kJ·mol−1) at pH 7.0. This value might be overestimated because the amino group may not be fully ionized at pH 7.0 [30], the value at which the stability of the mutant protein was measured by DSC. (b) Because there are many repulsive residues (Asp60, Asp84, Asp86, Asp87 and Glu63) around Glu59 (Table S1), the elimination of these unfavorable interactions may result in the increase in stability of the mutant protein.

Arg33 is in the β2 strand, and is almost completely buried in the interior of the molecule (accessible surface area of 5.3%). This residue is close to and ionically repulsive to the Arg33 residues of the other two chains (Fig. 4F) with a repulsive energy of 12.7 kJ·mol−1. Arg33 also favorably interacts with Glu34 (−2.9 kJ·mol−1) and Glu50 (−2.5 kJ·mol−1) (Table S1). The stabilities of the R33A and R33M variants decreased by 9.2 and 7.7 °C, respectively, due to substitution of the repulsive residue (Table 1). If burial polar residues and repulsive ionic interactions are eliminated, mutant proteins should generally increase in stability. However, the stability of both mutant proteins decreased. This discrepancy may be explained from the crystal structure of PhCutA1 (1V99), which indicates that NH1 and NH2 of Arg33 contact ND1 of His35 at a distance of 2.86 Å and OH of Tyr5 at a distance of 3.20 Å, respectively (Fig. 4G). These interactions should stabilize the native structure of the wild-type protein, whereas the mutant proteins (R33A and R33M) were destabilized by the elimination of these interactions. It has been reported that completely buried, nonion-paired glutamic acid contributes favorably to the conformational stability of pyrrolidone carboxyl peptidases from hyperthermophiles because of the formation of a hydrogen bond [31]. It is possible that Arg33 is not ionized in the interior of the molecule [32].

The locations of Glu42, Glu46 and Asp48, high in the order of unfavorable electrostatic interactions, are shown in Fig. 4H. The three mutants at these sites, E42Q, E46Q and D48N, increased Td values by 4.9, 3.6 and 2.7 °C, respectively. Glu42 mainly interacts with Lys44 (−1.6 kJ·mol−1), Glu46 (1.8 kJ·mol−1) and Glu99 (1.3 kJ·mol−1) (Table S1), resulting in an unfavorable electrostatic energy of 2.5 kJ·mol−1 (Table 2). Glu46 has strong unfavorable interactions with Asp48 (5.1 kJ·mol−1), Glu47 (2.1 kJ·mol−1), Glu42 (1.8 kJ·mol−1) and Glu50 (1.0 kJ·mol−1), and a favorable interaction with Lys44 (−3.9 kJ·mol−1). Asp48 has strong unfavorable interactions with Glu46 (5.1 kJ·mol−1), Glu50 (3.1 kJ·mol−1), Glu47 (2.0 kJ·mol−1) and Asp10 (1.2 kJ·mol−1) and a slightly favorable interaction with Arg36 (−1.1 kJ·mol−1) (Table S1). The increase in stability of these mutants can be explained by decreased repulsive ionic interactions.

Relationships between ΔTd and ΔΔG of unfolding and between ΔTd and ΔΔG of electrostatic energy

Figure 5 shows the relationship between changes in ΔTd due to mutations of PhCutA1 and ΔΔG of unfolding estimated by FoldX (Eqn 1). The correlation coefficient was 0.737 except for the R33M, R58A and R58A/D60N variants. In the cases of R58A and R58A/D60N, FoldX estimates unusually high values of ΔΔG due to decreases in the ‘side chain hydrogen bonds’ category to 58.8 and 57.1 kJ·mol−1, respectively. The FoldX values for the ‘Van der Waals’ and ‘Solvation of polar atoms’ categories for R33M were more favorable than those for R33A. These results indicate that the estimated changes in stability for mutant proteins modeled by FoldX are reasonably reliable, although there are a few outliers. On the other hand, the difference between the electrostatic energies of ion–ion interactions for charged residues of the wild-type and each mutant protein (ΔΔG of electrostatic energy) was estimated by FoldX. The correlation between ΔTd and ΔΔG of electrostatic energy was poor, having a correlation coefficient of 0.437 (not shown).

Figure 5.

 Relationship between ΔTd of PhCutA1 variants and ΔΔG of unfolding obtained from FoldX. ΔTd values were obtained as in Table 1. ΔΔG of unfolding for each mutant protein was obtained from Eqn (1). The structures of mutant proteins were constructed by FoldX. Broken and solid lines represent linear regression of all data points with and without the three open circle mutant proteins, respectively. Correlation coefficients of the broken and solid lines were −0.639 and −0.737, respectively.

As described above, these studies show that introduction of charged residues into proteins contributes to stability in various ways, including attractive or repulsive ionic interactions, as well as positive or negative effects on other important factors for protein stability, and the effects due to these other factors sometimes surpass the energy of ion–ion interactions. Therefore, the introduction of ion–ion interactions does not directly contribute to protein stability, but ΔΔG of unfolding, estimated by considering almost all interactions as shown in Eqn (1), correlates with the experimentally measured stability (Fig. 5). The FoldX webpage suggests that FoldX is accurate for relative energies, although absolute energies are not precise (http://foldx.crg.es/examples.jsp). As described, this work indicates that FoldX is useful for evaluating changes in stability of mutant proteins [33,34].

The role of charged residues of PhCutA1 in protein stability

The large number of charged residues having intrasubunit ion pairs in the same subunit (30 pairs in PhCutA1 against 1 pair in EcCutA1) was anticipated to play an important role in the thermostability of PhCutA1 at high temperatures around 150 °C [13]. To confirm this hypothesis, single mutant variants replacing Glu or Asp residues with noncharged residues were constructed and their stabilities were measured using DSC. The changes in stability of the mutant proteins were not great: the largest change in ΔTd was −3.5 °C and the average ΔTd was −1.0 ± 1.5 °C, suggesting that intrasubunit interactions at the examined positions do not play a very important role in stability. The ΔTd of the Ph14S mutant, a variant with replacement of all residues forming intrasubunit ion interactions within 4 Å, was −7.2 °C. This value is similar to the ΔTd (−6.8 °C) of the R58A mutant, which is included in the Ph14S mutant. The great decrease in Td of R58A might originate from loss of the interaction with Asp60 and many other neighboring negatively charged ones (Fig. 4E and Table S1). The ΔTd values of R58A and R58A/D60N were not different (Table 1). These results suggest that all ion pairs, which are deleted in Ph14S, hardly contribute to protein stability except for the interaction between Arg58 and Asp60. On the other hand, the average changes in Td of all Glu mutants examined (Table 1) was 0.03 ± 2.05 °C, suggesting that the Glu residues of PhCutA1 are weak contributors to the protein stability of PhCutA1. However, the electrostatic energy (−159.3 kJ·mol−1) of PhCutA1 is quite high, compared with that of EcCutA1(−9.7 kJ·mol−1). The pI values for PhCutA1 and EcCutA1 are 4.91 and 4.85, respectively. In the case of PhCutA1, unusually large numbers of negatively charged residues (especially Glu) might be essential in order to make highly efficient interactions with positively charged residues and to generate high electrostatic energy, resulting in a situation in which negatively charged residues surrounding positively charged residues are forced to be partially repulsive to each other.

Design of stable proteins by the introduction of ion pairs

Ion pairs are effective in stabilizing proteins from mesophiles and thermophiles, when they are exposed in solvent. However, when fully buried, each charged residue represents a desolvation penalty during folding [35]. Recently, the Makhatadze group has developed a computational method for the rational design of stable proteins by optimization of surface ion–ion interactions [36,37]. In this method, the energies of ion–ion interactions on the protein surface are calculated using the Tanford–Kirkwood model corrected for solvent accessibility [37,38]. There are several successful examples including α-lactalbumin [39], ubiquitin [36] and acylphosphatase [40]. However, our study suggests that the structural features of mutation sites should be carefully considered, and particularly residues that are involved in helix dipole moment, even if the residues are on the surface. Therefore, the present findings might be useful for designing stable proteins by the introduction of ion pairs.

Conclusions

To examine the role of ion–ion interactions in the stability of PhCutA1, a protein with a large number of charged residues and a high denaturation temperature, mutant proteins in which charged residues were substituted by noncharged residues were comprehensively examined. The following features were found.

The average decrease in Td of Lys or Arg mutants was greater than that of Glu or Asp mutants, suggesting that several negatively charged residues do not directly contribute to stability. However, many negatively charged residues should be essential in order to make efficient interactions with positively charged residues and to generate high electrostatic energy, because the energy of ion–ion interactions (−159.3 kJ·mol−1) of PhCutA1 with pI of 4.91 was much larger than that of EcCutA1 (−9.7 kJ·mol−1) with pI value of 4.85. Therefore, negatively charged residues were forced to be partially repulsive to each other.

The changes in stability of mutant proteins with Td of around 140–150 °C could be explained by considering factors important for protein stability and the structural features of mutant sites, suggesting that stabilizing factors already reported are valid on mutant proteins with Td around 140–150 °C. These findings are useful for the design of proteins that are stable at temperatures > 100 °C.

The Td of the E99Q mutant of PhCutA1 was higher by 6.4 °C than that of the wild-type protein (148.5 °C). The world record for heat stability was updated.

FoldX was useful for evaluating changes in stability of mutant proteins.

Experimental procedures

Construction, expression and purification of mutant proteins

Mutant versions of PhCutA1 were constructed by site-directed mutagenesis [41], expressed and purified as described previously [13]. All purified mutant proteins ran as single bands in SDS/PAGE. Protein concentrations were determined using an absorption coefficient of inline image at 280 nm, which is based on number of aromatic amino acids [42]. All chemical reagents were of analytical grade.

Measurement of thermal stability

Thermal denaturation of proteins was measured using VP-DSC/ETR microcalorimeter (Microcal LLC, Northampton, MA, USA) for temperatures up to 150 °C or Nano-DSC 6300Y microcalorimeter (TA Instruments, North Lindon, UT, USA) for temperatures up to 160 °C. Protein samples were dialyzed for > 20 h against 50 mm potassium phosphate (pH 7.0) including 2 mm EDTA. Samples were filtered through 0.22-μm pore membranes following dialysis and were degassed before measurements. Protein concentrations were typically around 1.0 mg·mL−1. The heating rate (scan rate) of DSC measurements was 60 °C·h−1.

Estimation of unfolding Gibbs energy and electrostatic energy due to ion–ion interactions from the tertiary structure of a protein using FoldX

The computer algorithm, FoldX [25], can quantitatively estimate the factors contributing to protein stability and protein interactions. FoldX is available via a web-interface at http://foldx.crg.es/. The FoldX energy function at pH 7.0 includes stabilization factors that are important for protein stability. The difference in unfolding Gibbs energy (ΔG) between wild-type and mutant proteins (ΔΔGFoldX) is evaluated on the basis of the following factors by FoldX.

image(1)

where ΔΔGBH, ΔΔGSH, ΔΔGVW, ΔΔGEl, ΔΔGSol,P, ΔΔGSol,H, ΔΔGVW,cla, ΔΔGEnt,s, ΔΔGEnt,m, ΔΔGTor,cla, ΔΔGBac,cla, ΔΔGHd and ΔΔGEl,sub represent the energies of the backbone hydrogen bonds (BH), side-chain hydrogen bonds (SH), Van der Waal’s interactions (VW), electrostatic interactions (El), solvation of polar atoms (Sol,P), solvation of hydrophobic atoms (Sol,H), Van der Waal’s clashes (VW,cla), entropy of the side chains (Ent,s), entropy of the main chain (Ent,m), torsional clashes (Tor,cla), backbone clashes (Bac,cla), helix dipoles (Hd) and electrostatic interactions between the subunits (El,sub), respectively. FoldX recommends that the RepairPDB command should be run prior to energy calculation, because the crystal structures are models based on electron density and have errors produced during refinement that will result in nonstandard angles or distances.

In the reported structure of wild-type PhCutA1, Cys29 was replaced by cysteine-s-dioxide (PDB entry 1v99). Therefore, we generated a standard structure of PhCutA1 by modeling a replacement of cysteine-s-dioxide with cysteine for energy calculations of the mutant proteins. Models of this standard structure (wild-type protein) and the necessary mutant structures were built using FoldX. ΔΔGFoldX values for the mutant proteins were generated using FoldX. The electrostatic energy of each residue between charged residues was sorted out from ‘AllAtoms_Electro’ file in FoldX. The electrostatic energies of the wild-type PhCutA1 are shown in Table S1.

Acknowledgements

This work was supported in part by a Grant-in-Aid for Scientific Research (C) (No. 20589003 to YM and No. 21570173 and 22570166 to KY) by JSPS of Japan.

Ancillary