Stabilization of Pseudomonas aeruginosa Cytochromec 551 by Systematic Amino Acid Substitutions Based on the Structure of Thermophilic Hydrogenobacter thermophilus Cytochrome c 552 *

A heterologous overexpression system for mesophilic Pseudomonas aeruginosa holocytochromec 551 (PA c 551) was established using Escherichia coli as a host organism. Amino acid residues were systematically substituted in three regions of PA c 551 with the corresponding residues from thermophilic Hydrogenobacter thermophilus cytochromec 552 (HT c 552), which has similar main chain folding to PA c 551, but is more stable to heat. Thermodynamic properties of PAc 551 with one of three single mutations (Phe-7 to Ala, Phe-34 to Tyr, or Val-78 to Ile) showed that these mutants had increased thermostability compared with that of the wild-type. Ala-7 and Ile-78 may contribute to the thermostability by tighter hydrophobic packing, which is indicated by the three dimensional structure comparison of PA c 551 with HTc 552. In the Phe-34 to Tyr mutant, the hydroxyl group of the Tyr residue and the guanidyl base of Arg-47 formed a hydrogen bond, which did not exist between the corresponding residues in HT c 552. We also found that stability of mutant proteins to denaturation by guanidine hydrochloride correlated with that against the thermal denaturation. These results and others described here suggest that significant stabilization of PAc 551 can be achieved through a few amino acid substitutions determined by molecular modeling with reference to the structure of HT c 552. The higher stability of HT c 552 may in part be attributed to some of these substitutions.

Proteins isolated from thermophilic organisms are usually stable to heat, indicating that these proteins must themselves embody most of the determinants of protein thermostability. Comparative studies of homologous proteins from mesophiles and thermophiles have provided ideas to explain elevated ther-mostability which include relatively small solvent-exposed surface area (1), increased packing density (2)(3)(4) and core hydrophobicity (5,6), decreased length of surface loops (4), and generations of ion pairs or hydrogen bonds between polar residues (7,8). Some recent site-directed mutagenesis studies have indicated that significant stabilization occurs in proteins as a result of mutations to reduce the entropy of the unfolded state (9,10).
Cytochrome c is characterized by covalent attachment of the heme to the polypeptide chain. This protein has proved useful as a model system for studying the relationship between protein structure and stability because (i) primary and threedimensional structures of cytochromes c from a wide variety of organisms (both mesophiles and thermophiles) are available, and (ii) heterologous expression systems of both prokaryotic and eukaryotic holocytochromes c have been established (11,12), which facilitate site-directed mutagenesis studies.
Cytochrome c 552 (HT c 552 ) 1 from a thermophilic hydrogen oxidizing bacterium, Hydrogenobacter thermophilus that grows optimally at 70°C, is an 80-amino acid protein with a heme. HT c 552 has 56% sequence identity to an 82-amino acid monoheme cytochrome c 551 (PA c 551 ) from mesophilic Pseudomonas aeruginosa (13), and the main chain foldings of these proteins are almost the same (14). As expected from the optimal growth temperatures of H. thermophilus and P. aeruginosa, HT c 552 is more stable to heat than PA c 551 (15). The genes encoding both proteins have been cloned (16,17) with a view to identifying (by site-directed mutagenesis) amino acid residues that contribute to the higher stability of HT c 552 compared with PA c 551 . PA c 551 and HT c 552 are thus very suitable proteins for identifying substitutions of amino acid residues that endow stability.
Here we report that holo-PA c 551 , which in terms of visible absorption spectra and thermostability is indistinguishable from the native protein, could be expressed in the periplasm of Escherichia coli. Using this expression system, site-directed mutagenesis studies were performed to show that the stability of PA c 551 could be significantly increased through selected mutations, which had been chosen by molecular modeling with reference to corresponding amino acid residues in HT c 552 . We discuss the structural origins of higher stability of HT c 552 .

EXPERIMENTAL PROCEDURES
Bacterial Strain, Plasmids, and Growth Condition-The EcoRI-PstI gene fragment CP1, encoding the 22-amino acid signal sequence and * This work was supported in part by grants from the Japanese Ministry of Education, Science and Culture. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ To whom correspondence and requests for reprints may be addressed: Daiichi Pharmaceutical Co. Ltd., 1-16-13 Kita-Kasai, Edogawa-ku, Tokyo 134-8630, Japan. Fax: 81-3-5696-8336; E-mail: haseg7li@daiichipharm.co.jp. ** Supported by a fellowship from the Japan Society for the Promotion of Science. § § Supported by a fellowship from the Japan Society for the Promotion of Science and by grants from the Naito Foundation and the Wellcome Trust. To whom correspondence and requests for reprints may be addressed. Present address: Institute of Scientific and Industrial Research, Osaka University, Ibaraki, Osaka 567-0047, Japan; E-mail: sambongi@sanken.osaka-u.ac.jp. 82-amino acid mature protein of PA c 551 (18), was inserted into the corresponding restriction sites of pKK223-3 (Amersham Pharmacia Biotech) to generate pKPA1. The pKPA1-based plasmids carrying mutated PA c 551 gene fragments (see below) were transformed by standard methods into E. coli JCB7120 strain in which the expression of c-type cytochromes is unusually high. 2 The transformed E. coli cells were grown anaerobically in minimal media in the presence of glycerol, nitrite, and fumarate (19) supplemented with casamino acid (2 mg/ml), tryptophan (20 g/ml), and ampicillin (50 g/ml) at 37°C for the production of wild-type and mutant PA c 551 proteins.
Introduction of Mutations into PA c 551 Gene-Two methods for sitedirected mutagenesis were used to introduce a series of mutations in PA c 551 . The first method was the PCR overlap extension technique (20,21) that was used for the mutations of Phe-7 to Ala/Val-13 to Met (F7A/ V13M), Phe-34 to Tyr/Gln-37 to Arg/Glu-43 to Tyr (F34Y/Q37R/E43Y), and Val-78 to Ile (V78I). The conditions for the PCR were as follows: incubation at 94°C for 2 min followed by 28 cycles of 94°C for 1 min, 56°C for 1 min, and 72°C for 1 min, with a final step of 5 min at 72°C. Mutations of Phe-7 to Ala (F7A), Val-13 to Met (V13M), Phe-34 to Tyr (F34Y), Gln-37 to Arg (Q37R), Glu-43 to Tyr (E43Y), F34Y/Q37R, F34Y/ E43Y, and Q37R/E43Y were introduced by the PCR-based kit, Mutan-Super Express Km (Takara Shuzo). All the PCR products were purified and digested with EcoRI and PstI, and then ligated into the corresponding restriction sites of pKK223-3. The DNA sequences of the entire PCR products were confirmed by the dideoxy-chain termination method using ABI Prism model 310 DNA sequencer.
Production of PA c 551 Proteins-The transformed E. coli cells harboring the wild-type or mutant PA c 551 gene were harvested from the anaerobic culture. Periplasmic protein fractions were recovered by cold osmotic shock, and membrane and cytoplasmic fractions were obtained as described previously (11). Cytochromes c were specifically detected on sodium dodecyl sulfate-polyacrylamide gel electrophoresis gels by a heme staining procedure (22). The expressed PA c 551 proteins in the periplasmic fraction were purified by Hi Trap Q or Mono Q column chromatography (Amersham Pharmacia Biotech), both eluted by 10 mM Tris-HCl buffer (pH 8.0) with an NaCl concentration gradient (0 -100 mM), followed by a Superdex 75 column equilibrated and eluted with 50 mM ammonium acetate buffer (pH 7.0). Protein concentrations were determined by the Bio-Rad protein assay kit with bovine serum albumin as a standard. The N-terminal amino acid sequence of the purified wild-type PA c 551 expressed in the E. coli periplasm was determined by automatic sequencer (Applied Biosystems model 470A).
UV-visible and Circular Dichroic (CD) Spectra-The UV-visible and CD spectra were measured on Hitachi U-3300 and Jasco J-720 machines, respectively. The protein samples were dissolved in water (pH 5.0 adjusted with HCl), the same conditions as used for thermal denaturation experiments.
Protein Thermostability-The wild-type and mutant PA c 551 proteins (10 g/ml protein concentration in water, pH 5.0 adjusted with HCl) were subjected to the thermal melting profile analysis by monitoring the changes of CD spectra at 222 nm as described previously (15) with slight modifications. The temperature of the protein solution was continuously raised from 25°C to 100°C at the rate of 50°C per 60 min in the absence of guanidine hydrochloride (GdnHCl) and from 15°C to 90°C at the same rate in the presence of 1.5 M GdnHCl. The temperature in the cell was controlled using a Jasco PT343 thermoelectric temperature controller.
Denaturation with GdnHCl-The stability against GdnHCl denaturation of wild-type PA c 551 or three mutants with F7A/V13M, F34Y/ E43Y, or V78I substitutions was assayed. The proteins (10 g/ml) were incubated in the diluted HCl water (pH 5.0) with varying concentrations of GdnHCl at 25°C for 2 h before the measurement in order to equilibrate the proteins with the denaturant. The CD ellipticity at 222 nm of the protein solutions was measured at 25°C.
Materials-Restriction enzymes, Taq polymerase, T4 DNA ligase, and other reagents for DNA handling were purchased from Takara Shuzo or Toyobo. The authentic PA c 551 protein from P. aeruginosa and GdnHCl were purchased from Sigma. All other chemicals used were of the highest grade commercially available.

RESULTS
Expression of PA c 551 in the Periplasm of E. coli-The wildtype PA c 551 protein expressed in E. coli JCB7120 strain was fully recovered in the cold osmotic shock fluid containing the periplasmic protein fraction, but not in the membrane and cytoplasmic fractions. The expressed protein had covalently attached heme as judged by heme staining following separation by SDS-polyacrylamide gel electrophoresis (data not shown). The production yield of holo-PA c 551 in E. coli reached 8 mg/1 liter of culture (equal to 30% of the total periplasmic protein) 2 J. Cole, personal communication. The NMR data were processed using the software nmrPipe and nmrDraw (38), and the analysis was assisted by the software Pipp (39). Backbone signal assignment on 1 H-15 N HSQC spectrum was achieved based on the sequence-specific NOE connectivity (40). The cross-peaks are labeled according to the sequential assignment. calculated by using a millimolar coefficient for the ␣ band at 551 nm (21 mM Ϫ1 cm Ϫ1 ). All the mutant proteins tested in this study could be obtained in the E. coli periplasm at almost the same yield as the wild-type.
The N-terminal amino acid sequence of the wild-type PA c 551 protein expressed in the E. coli periplasm was determined as Glu-Asp-Pro-Glu-Val-Leu-Phe-Lys-Asn-Lys-Gly, which was identical to that of the authentic protein purified from the native organism.
Spectroscopy-The UV-visible (400 -600 nm) spectrum of dithionite-reduced wild-type PA c 551 protein expressed in E. coli showed absorption maxima at 417, 521, and 551 nm, which are characteristic features of authentic PA c 551 protein (Fig. 1A). The far-ultraviolet CD (200 -250 nm) spectrum of the air-oxidized form of the wild-type was also the same as that of the authentic protein, having an absorption peak at 222 nm (Fig.  1B). The same properties in UV-visible and CD spectra were obtained from all the mutant PA c 551 proteins used in this study (data not shown). Furthermore, the 1 H-15 N HSQC spectrum of the dithionite-reduced wild-type protein was essentially the same as that of authentic protein (Fig. 2) (25). These results together suggested that the wild-type PA c 551 expressed in E. coli folded correctly and the mutant proteins did not markedly differ in terms of the three-dimensional structure.
Assay Condition for Thermostability-The wild-type PA c 551 from the native organism and E. coli both had the same cooperative melting transition with a T m value of 50.4°C in the presence of 1.5 M GdnHCl at pH 5.0 (Fig. 3A). The T m values of all the mutants could be determined in the presence of the same concentration of the denaturant (Fig. 3A and Table I). Therefore, we carried out the thermal denaturation assays under these conditions throughout the present study. By contrast, in the absence of GdnHCl, the T m values of the mutant proteins could not be determined because they did not reach a completely denatured state even at 100°C (Fig. 3B).
Substitutions of Phe-7 and Val-13-In PA c 551 , a small cavity exists around the side chains of Phe-7 and Val-13, which correspond to Ala and Met, respectively, in HT c 552 . The threedimensional structure of HT c 552 shows that the side chains of the Ala and Met fill this cavity (14). The double mutation F7A/V13M in PA c 551 caused increased thermostability compared with the wild-type (⌬T m : 12.0°C; Fig. 3A and Table I). Each single mutation, F7A and V13M, enhanced the thermostability essentially in an additive manner, the individual ⌬T m values being 9.5 and 3.2°C, respectively ( Fig. 3A and Table I Table I). The simultaneous mutation (F34Y/E43Y) caused enhanced thermostability, which was contributed by each single mutation in a cumulative manner (⌬T m : 20.3°C, ⌬T m hyp : 21.1°C, Fig. 3A and Table I). Although the single Q37R mutation in PA c 551 reproducibly made a small contribution to the increased stability (⌬T m value was 4.3°C, Table I Table I) were each significantly lower than those with F34Y, E43Y, and F34Y/ E43Y, respectively.
Substitution of Val-78 -The region around Val-78 in PA c 551 , interacting with heme hydrophobically, should become more hydrophobic if Val were substituted by Ile as in HT c 552 ; this is because Ile has one additional methyl group in the side chain compared with that of Val. As expected, the V78I mutation in PA c 551 caused an 8.4°C elevation of the T m value compared with that of the wild-type ( Fig. 3A and Table I).
Stability against GdnHCl Denaturation-We also tested whether F7A/V13M, F34Y/E43Y, and V78I mutations in PA c 551 stabilize the structure against GdnHCl denaturation. Fig.  4 shows GdnHCl-induced denaturation curves and plots of ⌬G versus GdnHCl concentration around the midpoint of denaturation (C m ). The denaturation curve of the wild-type showed that 1.5 M GdnHCl did not denature the protein structure. Thus, in the thermal denaturation experiments, the temperature of the protein samples can be ramped from the nondenatured condition in the presence of 1.5 M GdnHCl. The C m values of the mutant proteins (F7A/V13M, F34Y/E43Y, and V78I) were elevated as compared with that of the wild-type (⌬C m : 0.52, 0.73, and 0.19 M, respectively; Table II). Among these mutants, F34Y/E43Y was the most stable, followed by the F7A/V13M, and then the V78I protein, as judged by the comparison of ⌬C m and ⌬⌬G H2O values. This order of the stability  The temperature of the midpoint of the transition (T m ) and the enthalpy change during unfolding at T m (⌬H) were calculated from curve fitting of the resulting CD values versus the temperature data on the basis of van't Hoff analysis. This curve fitting was achieved using the function of a least-square analysis in the software MATHEMATICA (Wolflam Inc.). The entropy change during unfolding at T m (⌬S) was calculated using the equation, ⌬S ϭ ⌬H/T m . The differences in the free energy changes of unfoldings between the mutant proteins and wildtype at the wild-type T m (⌬⌬G m ) were calculated using the equation given by Becktel and Schellman (41), ⌬⌬G m ϭ ⌬T m * ⌬S (wild-type), where ⌬T m is the difference in T m values between the mutant and wild-type proteins, and ⌬S m (wild-type) is the entropy change of the wild-type protein at the T m . The hypothetical ⌬T m value (⌬T m hyp ) was calculated for each mutant protein with multiple mutations assuming that the effect of each amino acid replacement on the protein stability is independent and cumulative. was same as that from the heat denaturations of these mutants.

DISCUSSION
In this study we developed an expression system for PA c 551 using E. coli JCB7120 strain as a host organism. The correctly processed wild-type PA c 551 expressed in the periplasm of E. coli has the same spectral properties in CD, UV-visible, and NMR, and also has the same thermostability as the authentic protein, indicating that the protein is in "native" state. In previous studies, the PA c 551 gene on extra-chromosomal plasmids could be expressed heterologously in Pseudomonas putida (26) and in the original organism (18). The holo-PA c 551 formation was also observed in the periplasm of other E. coli strains. 3 The expression level in the present E. coli strain was the highest relative to any previous studies that demonstrated heterologous expression of cytochromes c in other E. coli strains (16, 27, 28), 4 The efficient production level and easy purification procedure from the E. coli periplasm enabled us to obtain 15 N-labeled PA c 551 for heteronuclear NMR spectroscopy (detailed structural analysis is in progress). Furthermore, this system will facilitate other studies requiring large amounts of the protein sample such as differential scanning calorimetry and x-ray crystallography. The JCB7120 strain was chosen because it expresses higher levels of endogenous c-type cytochromes than other strains (29). 2 The basis for this is not known, but it is seemingly not a consequence of enhanced expression of c-type cytochrome biogenesis genes (30) because the JCB7120 strain also produces high level of cytoplasmically expressed HT c 552 (31), a process that is independent of the biogenesis genes.
Although 35 amino acid residues are substituted between HT c 552 and PA c 551 , we have postulated, from structure comparison between the proteins, that a few amino acid residues in the three regions formed by Phe-7/Val-13, Phe-34/Gln-37/Glu-43, and Val-78 ( Fig. 5; see Fig. 7 in Ref. 14 for the detailed structures) are important for stability. Therefore, we systematically introduced the mutations in these regions of PA c 551 modeled by the corresponding residues in HT c 552 . All the mutations tested in this study resulted in the increased stability, although HT c 552 was still more stable than these mutants; its T m and C m values were 91.8°C and 4.49 M, respectively, assayed under the same condition as used for the PA c 551 proteins. 5 It is notable that the present findings with a mainly ␣ helical protein contrast with those made with the ␤-sheet rubredoxins from Pyrococcus furiosus (thermophile) and Clostridium pasteurianum (mesophile) (32). For these iron-sulfur proteins substantial exchanges of linear sequence, rather than individual mutations, have been shown to be required to enhance thermostability, implying that many small interactions cumulatively contribute to large increases in the stability.
The increased stability of the PA c 551 with F7A, V13M, F7A/V13M, or V78I mutations indicates that the void spaces in PA c 551 formed by the side chains of the original residues destabilize the protein structure. Therefore, the mutations designed to fill the void space were effective in increasing stability. It has been suggested in other proteins that higher stability can be achieved when Val is substituted by Ile, having one additional methyl group (33,34) as found in the PA c 551 V78I mutant. The stabilization by the E43Y mutation in PA c 551 could also be attributed to tighter hydrophobic packing between the introduced Tyr residue and Phe-34, Ala-40, or Leu-44, among which the latter two residues are conserved in HT c 552 .
The ⌬⌬G m value for the F34Y mutant is 1.9 kcal/mol, which is comparable to the free energy of the hydrogen bond (2ϳ4 kcal/mol). Consistent with this calculation, three-dimensional

TABLE II Parameters characterizing GdnHCl denaturation
The difference in free energy change between the folded and unfolded states (⌬G) was calculated as described by Pace (42). The free energy change in H 2 O (⌬G H2O ) and the dependence of ⌬G on the GdnHCl concentration (m) were determined by a least-squares fit of the data from the transition region using the equation: ⌬G ϭ ⌬G H2O Ϫ m[GdnHCl] (42). The midpoint of the GdnHCl denaturation (C m ) was the concentration of GdnHCl at which the ⌬G value became 0. The differences in C m (⌬C m ) and ⌬G H2O (⌬⌬G H2O ) between the wild-type and mutant proteins were calculated by subtracting the values of the wildtype from those of mutants.  molecular modeling of PA c 551 with the F34Y mutation suggest that the oxygen atom of the introduced Tyr-34 forms a hydrogen bond with the guanidyl base of Arg-47 in PA c 551 : the distance between these atoms is estimated to be 3.2 Å. However, Lys-45 in HT c 552 (corresponding to the PA c 551 Arg-47) does not form a hydrogen bond with the "original" Tyr residue. If Lys-45 could be substituted by Arg in HT c 552 , much higher thermostability should be obtained in the thermophilic protein.
The ⌬T m value for the PA c 551 F34Y mutant (16°C) is the largest among those of the single amino acid mutants of PA c 551 , and this thermostabilization is one of the most dramatic observed for a single substitution in any protein. It is equivalent to that of a yeast iso-1-cytochrome c mutant, which exhibits the highest elevation of T m ever observed (35).
The NMR solution structure of HT c 552 indicates that Arg-35 and the two Tyr residues (corresponding to Gln-37, Phe-34, and Glu-43 in PA c 551 , respectively) form aromatic-amino interactions (14). These interactions have been suggested to cause the higher thermostability of HT c 552 . However, the Q37R mutation negatively affected thermostability of the three PA c 551 proteins with F34Y, E43Y, or F34Y/E43Y mutations. This observation clearly indicates that the aromatic-amino interaction(s) are formed between the introduced Arg and Tyr residues as in HT c 552 , but these interactions may disturb the hydrogen bond formation between Tyr-34 and Arg-47, and/or the hydrophobic interactions between Tyr-43 and Phe-34, Ala-40, Tyr-41, or Leu-44.
In this study, successful enhancement of protein stability has been achieved by filling small void spaces, increasing hydrophobicity, and generation of a hydrogen bond in the three local regions of PA c 551 . Hydrogenophilus thermoluteolus (formerly Pseudomonas hydrogenothermophila; Ref. 36), which grows optimally at 52°C, has a homologous cytochrome c 552 (HP c 552 , having 65% amino acid identity to HT c 552 ), although the partial sequence (60 amino acids) is only available. HP c 552 has been shown to have high thermostability like HT c 552 (37). It should be noted that, in the HP c 552 protein, corresponding amino acid residues to PA c 551 Phe-7, Val-13, and Phe-34 are, respectively, Ala, Met, and Tyr as found in HT c 552 . These findings may support the proposition that the substitutions of Phe to Ala, Val to Met, and Phe to Tyr at these positions play important roles in the protein stability, assuming that HP c 552 protein has a three-dimensional structure similar to that of HT c 552 and PA c 551 . Our study strongly indicates that the HT c 552 can be used as an ideal model protein, it is 26 residues smaller than the yeast iso-1-cytochrome c, for elucidating the roles of amino acid residues in protein stability by mutating the mesophilic homologue, PA c 551 protein.