Selected Mutations in a Mesophilic Cytochrome c Confer the Stability of a Thermophilic Counterpart

Right This research was originally published in the Journal of Biological Chemistry. Jun Hasegawa, Susumu Uchiyama, Yuko Tanimoto, Masayuki Mizutani, Yuji Kobayashi, Yoshihiro Sambongi and Yasuo Igarashi. Selected Mutations in a Mesophilic Cytochrome c Confer the Stability of a Thermophilic Counterpart. J. Biol. Chem. 2000; 275(48):37824-37828. © the American Society for Biochemistry and Molecular Biology.

Heat-stable proteins from thermophilic bacteria usually exhibit main chain foldings similar to those of mesophilic counterparts. Mutational studies on mesophilic proteins modeled with respect to thermophilic counterparts have proved that specific side chain interactions in the thermophiles are partially responsible for the higher stability (1)(2)(3). Thermodynamic analysis has also indicated that a thermophilic protein can be stabilized through global interaction throughout the molecule (4). It remains enigmatic as to how many amino acid substitutions contribute to the stability of a natural thermophilic protein (5). In some cases, a mesophilic protein only acquires the stability of the thermophilic counterpart after substantial exchanges of a linear sequence; groups of individual mutations are not sufficient (6). Multiple mutations in mesophilic proteins that completely increase the stability to the levels of thermophilic counterparts would provide important information about relationships between local side chain interactions and overall protein stability and demonstrate that the thermophilic character can depend on a limited number of strong noncovalent interactions.
Cytochrome c is a powerful tool for characterizing protein stability because structural information on a variety of cytochromes c is available, and heterologous expression systems for holopoteins have been established (7). Cytochrome c 551 (PA c 551 ) 1 from a mesophile, Pseudomonas aeruginosa, and cytochrome c 552 (HT c 552 ) from a thermophile, Hydrogenobacter thermophilus, are 82-and 80-amino acid proteins, respectively, each with a covalently attached heme. These proteins exhibit 56% sequence identity (8) and almost the same main chain foldings (9), but HT c 552 exhibits much higher stability compared with PA c 551 (10). On a structural comparison between HT c 552 (9) and PA c 551 (11), we identified three distal regions responsible for the higher stability of the former (9). The single mutation Val-78 to Ile (V78I), and two double mutations Phe-7 to Ala/Val-13 to Met (F7A/V13M) and Phe-34 to Tyr/Glu-43 to Tyr (F34Y/E43Y), chosen with reference to HT c 552 , in these three regions of PA c 551 have each been shown to increase protein stability (1).
In the present study, the five mutations were introduced together into PA c 551 . Thermodynamic analysis showed that the quintuple mutant of PA c 551 was as stable as HT c 552 . In order to provide a molecular basis for understanding protein stabilization, we have determined the solution structure of the quintuple mutant. We discuss factors contributing to protein stability in conjunction with structural analyses of HT c 552 , and the wild-type and quintuple mutant PA c 551 proteins.

EXPERIMENTAL PROCEDURES
Protein Preparations-Mutations were introduced into the PA c 551 gene with a polymerase chain reaction-based kit, Mutan-Super Express Km (Takara, Kyoto, Japan), as described previously (1). Transformed Escherichia coli JCB7120 cells harboring PA c 551 genes were harvested from an anaerobic culture, and the proteins used in this study were purified as described previously (1,8). The concentrations of the purified protein solutions were determined spectrophotometrically using extinction coefficients of ⑀ 551 ϭ 25,200 cm Ϫ1 M Ϫ1 and ⑀ 552 ϭ 20,400 cm Ϫ1 acid mixture (98.2%), and a glycerol-13 C 3 and 13 C/ 15 N-labeled Algal amino acid mixture (97.5%) as nitrogen and carbon sources. The labeled compounds were obtained from Shoko Co., Ltd. (Tokyo, Japan).
Guanidine Hydrochloride (GdnHCl) Denaturation-Proteins (10 g/ ml) were incubated in diluted HCl water (pH 5.0) with various concentrations of GdnHCl at 25°C for 2 h before measurements in order to equilibrate the proteins with the denaturant. The CD ellipticity at 222 nm of the protein solutions was measured using a 1-cm path length cuvette at 25°C with a JASCO J-720 CD spectrophotometer with a PT343 thermoelectric temperature controller. The data were fitted by nonlinear least-squares analysis with KaleidaGraph 3.0 (Synergy Software) employing the Marquart-Levenberg algorithm using a linear extrapolation model as described previously (12). C m was the concentration of GdnHCl at which the free energy change value, ⌬G, became 0.
Thermal Denaturation-The temperature dependence of the CD ellipticity at 222 nm was monitored using a 1-cm path length cuvette with a JASCO J-720 spectrophotometer with a PT343 thermoelectric temperature controller as described previously (1). Protein solutions (ϳ10 g/ml), pH 5.0, containing 1.5 M GdnHCl were heated from 15°C to 90 or 100°C at a heating rate of 1 K/min. Under these conditions, the thermal transition was highly reversible (95%). The van't Hoff enthalpy, ⌬H van , was determined from the CD denaturation curve according to a two-state mechanism with a temperature-independent heat capacity change, ⌬C P (13). The ⌬C P values used were obtained by differential scanning calorimetric (DSC) measurements at pH 5.0 in the presence of 1.5 M GdnHCl for each protein. The slope of the CD base line for HT c 552 in the denatured state was assumed to be the same as that in the native state.
DSC measurements were carried out with the VP-DSC developed by MicroCal Inc. (14). Degassed protein solutions, with a concentration of ϳ1 mg/ml, were loaded into the calorimeter cell, and each sample was heated from 10 to 125°C under approximately 3 atmospheres, with a heating rate of 1 K/min. Thermodynamic parameters, i.e. T m , ⌬H, ⌬S, ⌬C P , and ⌬G, were estimated from the heat capacity curve using nonlinear least square fitting based on the previously described method (15,16). The fitting was performed with MATHEMATICA 3.0 (Wolfram Research Inc.) employing the Marquart-Levenberg algorithm.  (20), and HNHA (21). Protein side chain assignments were made through HCCH-total correlation spectroscopy experiments (22). Stereospecific assignments of the ␥-methyl protons of valines and ␤-methylene protons were made by analyzing HNHB (23), 15 N-edited NOESY-HSQC (24), and DQF-COSY (25) spectra. All proton signals from the heme moiety were assigned according to the procedure of Keller and Wü thrich (26). The signals of carbons attached to heme protons were assigned with a constant time 13 C-1 H HSQC spectrum (27). All data were processed using the software NMRPipe (28), and the data analysis was assisted by the software PIPP (29). The 1 H, 13 C, and 15 N resonance assignments of the quintuple mutant have been deposited in the BioMagResBank, under accession number 4578.
Structure Calculation-Approximate interproton distances were obtained from simultaneous 13 C/ 15 N-edited NOESY-HSQC (30), 15 N-edited NOESY-HSQC, and NOESY (31) spectra. The mixing time was 60 ms for all NOESY experiments. The distance restraints were grouped into four classes: 1.8 -2.7, 1.8 -3.3, 1.8 -5.0, and 1.8 -6.0 Å, corresponding to strong, medium, weak, and very weak NOE cross-peak intensities, respectively. The NOEs including backbone amide protons were grouped into the four classes of 1.8 -2.9, 1.8 -3.5, 1.8 -5.0, and 1.8 -6.0 Å. The backbone coupling constants, 3 J NH␣ , were measured in a HNHA experiment. The and -dihedral angle restraints were derived from 3 J NH␣ coupling constants and chemical shift indices. Values of Ϫ60 Ϯ 30°and Ϫ40 Ϯ 30°were used for the and -dihedral angles, respectively, for ␣-helical regions. Hydrogen bond restraints were obtained by analyzing the H/D exchange rates and NOE patterns characteristic of ␣-helices. Two distance restraints,   (34), with a probe radius of 1.4 Å. The Protein Data Bank accession numbers for the three-dimensional structures of the native proteins were 451C and 1AYG for PA c 551 and HT c 552 , respectively. Surfaces were classified into polar and nonpolar components by regarding carbon and sulfur atoms as nonpolar and oxygen and nitrogen as polar (35). G hN was calculated according to the method proposed by Oobatake and Ooi (36).

RESULTS AND DISCUSSION
Stability against GdnHCl Denaturation-The far-ultraviolet CD spectrum of the quintuple mutant of PA c 551 (F7A/V13M/ F34Y/E43Y/V78I) was nearly identical to that of the wild type (data not shown). A GdnHCl-induced denaturation curve of the quintuple mutant obtained by CD showed that its C m value (4.39 M) was highly elevated relative to that of the wild type ( Fig. 1A and Table I) and essentially the same as that of HT c 552 (C m ϭ 4.47 M). Since PA c 551 and HT c 552 exhibit a total of 35 amino acid differences, it is noteworthy that only five of these residues are responsible for the enhanced overall stability against chemical denaturation.
Stability against Thermal Denaturation-To address the question of how both the quintuple mutant PA c 551 and HT c 552 are thermodynamically stabilized, we performed thermal denaturation experiments using CD and DSC. Fig. 1B shows thermal denaturation curves of the quintuple mutant and HT c 552 together with the wild-type PA c 551 observed by CD. The ⌬H van values estimated from CD measurements for the three proteins were almost identical to the calorimetric enthalpy change at the denaturation temperature (T m ), ⌬H m , obtained from DSC measurements (see below and Table I), indicating that thermal denaturation of these proteins proceeded in a two-state manner. Fig. 2A shows the heat capacity curves of the three proteins measured by DSC. From these curves, thermodynamic parameters were obtained as a function of temperature. We further obtained the ⌬C P value of the wild-type PA c 551 from T m -dependent ⌬H(T m ) measurements at pH 3.6, 3.8, and 4.0 in the absence of GdnHCl (Fig. 2, B and C) (37). The ⌬C P value

TABLE II
Statistics of the 20 structures of the quintuple mutant protein ͗SA͘ represents the 20 individual structures calculated with the X-PLOR program. ͗SA͘r is the refined structure obtained by energy minimization of the mean structure obtained by simple averaging of the coordinates of the SA structures. F NOE and F tor were calculated using force constants of 50 kcal/mol/Å 2 and 200 kcal/mol/rad 2 , respectively. F vdw was calculated using a final value of 4 kcal/mol/Å 2 with the van der Waals hard sphere radii set to 0.75 times those in the parameter set PARALLHSA supplied with the X-PLOR program.  (38). The DSC measurements ( Fig. 2A) showed that the quintuple mutant had an increased T m value of 32.9°C and enhanced thermodynamic stability (⌬⌬G) of 5.86 kcal/mol at the T m value of wild-type PA c 551 (Table I). These values were nearly the same as those of HT c 552 . The quintuple mutant exhibited a large increase in ⌬H compared with the wild-type PA c 551 (Table I), suggesting that the mutant was enthalpically stabilized. In contrast, HT c 552 was stabilized by a small ⌬S rather than by an enthalpic factor (Table I). This is obvious from the heat capacity curve ( Fig. 2A), since the peak height and area (nearly representing ⌬H m ) of HT c 552 were smaller than those of the quintuple mutant although their T m values were nearly the same. These results suggest that the enhanced T m values of HT c 552 and the quintuple mutant are mainly due to the five residues (Ala-7, Met-13, Tyr-34, Tyr-43, and Ile-78, numbered as in the quintuple mutant of PA c 551 ); however, the stabilizing factors differ in the two stable proteins.
Additivity of Thermal Stabilization-We estimated ⌬⌬G values for the reported PA c 551 mutants (1) having F7A/V13M, F34Y/E43Y, and V78I substitutions, respectively, by DSC measurements (data not shown). The estimated values were as follows: F7A/V13M, 2.39; F34Y/E43Y, 2.52; and V78I, 0.82 kcal/mol. The ⌬⌬G value of the quintuple mutant obtained in this study was almost identical to the value for the hypothetical difference in free energy change, ⌬⌬G hyp , i.e. the sum of the ⌬⌬G values for the mutant proteins with the F7A/V13M, F34Y/ E43Y, and V78I substitutions, respectively (Table I). This indicates that the mutations in each of the three regions contribute in an additive manner to the enhanced overall stability. The three mutated regions in the quintuple mutant do not interact with each other; thus, they may behave independently without nonlocal structural perturbations.
Structure of the Quintuple Mutant Protein-We next determined the solution structure of the quintuple mutant PA c 551 protein using 1545 NOE-based distance restraints (comprising 592 intraresidue and intraheme, 300 sequential, 257 medium range, 396 long range including NOEs between the heme and polypeptide chain), supplemented with 102 dihedral and 46 hydrogen bond restraints (Fig. 3A). The best 20 structures for the quintuple mutant satisfied the experimental constraints with small deviations from the idealized covalent geometry (Table II). The stereochemical quality of the 20 structures was determined using PROCHECK-NMR (39). Ignoring all glycines and prolines,  98.5% of the remaining residues fell into the most favored and additional allowed regions of and spaces. The average atomic root mean square (r.m.s.) deviations for heavy atoms of residues 3-80 were 0.40 Ϯ 0.06 Å for the main chain atoms and 0.84 Ϯ 0.05 Å for all atoms. These values indicate that the determined structures were well converged and that the restrained energyminimized structure could be used as a representative for comparison with those of the wild-type PA c 551 and HT c 552 . The main chain folding of the quintuple mutant was similar to those of the wild-type PA c 551 and HT c 552 ( Fig. 3B; backbone r.m.s. deviation values were 0.84 Å for residues 3-80 of the wild-type PA c 551 and 0.99 Å for residues 3-78 of HT c 552 , respectively). This indicates that the introduction of five mutations into PA c 551 does not alter the main chain folding. Structure Comparison between HT c 552 and the Quintuple Mutant-The structure of the quintuple mutant showed that the side chains of the introduced Ala-7 and Met-13 filled a small cavity found in the wild-type (Fig. 4A). The F7A/V13M mutations in this region also changed the Ile-18 side chain conformation to the favorable gauche plus form from the wildtype gauche minus one (Fig. 4A). The r.m.s. deviation value for the Ala-7, Met-13, Tyr-27, and Trp-77 heavy chain atoms and the residues 5-20, 25-29, and 75-79 main chain atoms was 1.28 Å when these atoms were superimposed on the corresponding atoms of HT c 552 .
The two introduced Tyr-34 and Tyr-43 aromatic side chains in the quintuple mutant were adjacent, as found in HT c 552 , and suggested to undergo a hydrophobic interaction and/or Ϫ interaction with one another (Fig. 4B). The r.m.s. deviation value for the heavy atoms of the introduced Tyr-34 and Tyr-43 and the main chain atoms of residues 34 -44 in the quintuple mutant was 0.91 Å when these atoms were superimposed on the corresponding atoms of HT c 552 . Molecular modeling of PA c 551 with the F34Y mutation predicts that the oxygen atom of the introduced Tyr-34 forms a hydrogen bond with the guanidyl base of Arg-47; this was also indicated by our previous thermodynamic analysis (1). However, it was not clear from the NMR data whether an extra hydrogen bond exists in the quintuple mutant, because the guanidyl base of Arg-47 in the mutant was not well defined in the NMR structure.
The mutant structure also showed that the introduced Ile-78 filled a cavity around the heme, which was found in the wildtype (Fig. 4C). The r.m.s. deviation value for the Ile-48, Leu-74, Ile-78, and heme heavy atoms and residues 72-80 main chain atoms in the quintuple mutant was 0.91 Å when these atoms were superimposed on the corresponding atoms of HT c 552 .
These comparisons of side chain interactions in the three mutated regions of the three proteins clearly showed that the regions in the quintuple mutant became more like those in HT c 552 .
Difference in Accessible Surface Area-We further evaluated, using the structural data, the effects of the five mutations in the three regions on the total ASA (accessible surface area). The quintuple mutant and HT c 552 in their native states had larger ASA values compared with that of the wild-type PA c 551 ; this was due to the larger polar ASA (ASA pol ) value in both cases (Table III). Thus, undefined hydrophilic and polar groups may be more exposed to the solvent. Consequently, the quintuple mutant and HT c 552 in the native states exhibited greater negative G hN compared with the wild-type PA c 551 . The negative G hN values for these two proteins may contribute to the enhanced stability, which is consistent with the results of recent statistical analyses of proteins from thermophiles (40).
Conclusion-Our successful design of a mesophilic protein is, to the best of our knowledge, the first example of protein stabilization to the level of a natural thermophilic counterpart by means of limited amino acid substitutions (2,41,42). The formation of extra side chain interactions and exposure of hydrophilic and polar groups of the quintuple PA c 551 mutant caused the overall elevated stability, which was partly reflected by the increased enthalpy change. The stabilizing strategy for the mutant differed from that in the case of the cold shock protein (43), in which the protein stabilization was mainly achieved through improvement of the electrostatic interaction on the molecular surface.
Our way of carefully comparing the structures of thermophilic and mesophilic homologous proteins and combining selected mutations is valuable for elucidating the relationship between local side chain interactions and overall protein stability. Now that this has been achieved for the first time, it will be worthwhile exploring the possibility of altering other proteins, especially those of industrial and medical interest, in the same manner.