[J. Biochem. Vol. 130, pp. 107-118 (2001), Regular paper; © 2001 by The Japanese Biochemical Society]
Hideaki Tanaka,* Masanobu Chinami,† Tsunehiro Mizushima,* Kyoko Ogasahara,* Motonori Ota,‡ Tomitake Tsukihara,* and Katsuhide Yutani*,1
*Institute for Protein Research, Osaka University, 3-2 Yamadaoka, Suita, Osaka, 565-0871; †Department of Nutritional Science, Kyushu Women's University, 1-1 Jiyugaoka, Yahatanishi, Kitakyushu, Fukuoka, 807-8086; and ‡National Institute of Genetics, Yata Mishima, Shizuoka 411-8540
Received March 12, 2001; accepted April 20, 2001
In order to elucidate the mechanism of the thermostability of proteins from hyperthermophiles, X-ray crystalline structures of pyrrolidone carboxyl peptidase from a hyperthermophile, Pyrococcus furiosus (PfPCP), and its mutant protein with Ser substituted at Cys142 and Cys188 were determined at 2.2 and 2.7 Å resolution, respectively. The obtained structures were compared with those previously reported for pyrrolidone carboxyl peptidases from a hyperthermophilie, Thermococcus litoralis (TlPCP), and from a mesophile, Bacillus amyloliquefaciens (BaPCP). The PfPCP structure is a tetramer of four identical subunits similar to that of the TlPCP and BaPCP. The largest structural changes among the three PCPs were detected in the C-terminal protrusion, which interacts with that of another subunit. A comparison of the three structures indicated that the high stability of PfPCP is caused by increases in hydrophobic interactions and hydrogen bonds, the formation of an intersubunit ion-pair network, and improvement to an ideal conformation. On the basis of the structures of the three proteins, it can be concluded that PfPCP does not have any special factors responsible for its extremely high stability and that the conformational structure of PfPCP is superior in its combination of positive and negative stabilizing factors compared with BaPCP.
Key words: hyperthermophile, Pyrococcus furiosus, pyrrolidone carboxyl peptidase, thermal stability, X-ray structure.
[email to author] • [Related articles in JB] • [Home page]
The conformations of general globular proteins from mesophiles are marginally maintained by a combination of many positive (such as hydrogen bonds and hydrophobic interaction) and negative (such as entropic factors and steric hindrance) factors for stabilization (1). Studies on extremely thermostable proteins from hyperthemophiles that have denaturation temperatures above 100°C would be expected to help in understanding the stabilization mechanism of general globule proteins. Recently, the X-ray structures of many proteins from hyperthermophiles have been determined and several factors responsible for extremely thermostable proteins have been proposed, including an increase in the number of ion-pairs and hydrogen bonds (2-11), an increase in core hydrophobicity (12, 13), decreases in the length of surface loops and peptide chains (11, 14), increases in packing density (14, 15), and the oligomerization of several subunits (16-21). However, the origin of extremely high thermostability at the molecular level has not yet been determined. This might be because the stability results in a combination of many factors and the contribution of each factor to thermostability cannot yet be estimated quantitatively.
The pyrrolidone carboxyl peptidase [EC 3.4.19.3] (PfPCP) from a hyperthermophilic archaeon, Pyrococcus furiosus, which removes amino-terminal L-pyroglutamic acid from peptides and proteins, is a suitable protein for the physicochemical exploration of the stabilization mechanism of hyperthermophile proteins, because the thermal denaturation of PfPCP is completely reversible under some conditions and can be thermodynamically analyzed (22, 23). In the case of hyperthermophile proteins, the thermal denaturation is generally irreversible and only a few small proteins are reported to be reversibly denatured (24, 25). In order to elucidate the mechanism of the thermal stability of PfPCP, its three-dimensional structure is needed. Thus we have analyzed the X-ray structure of this protein.
Recently, the X-ray structures of pyrrolidone carboxyl peptidase (TlPCP) from the hyperthermophilic archaeon, Thermococcus litoralis (26), and (BaPCP) from a mesophile, Bacillus amyloliquefaciens (27), have been reported. From the TlPCP structure, the highly hydrophobic insertion that forms the core of the PCP tetramer and the presence of intersubunit disulphide bonds between Cys190 residues in each subunit seem to be responsible for the stability (26). The Cys190 in TlPCP corresponds to Cys188 in PfPCP, but there is no corresponding Cys in BaPCP. Cys188 in PfPCP hardly forms a disulphide bond in solution, but the stability increases remarkably when Cys188 is oxidized (23). This has been confirmed by DSC (Differential Scanning Calorimetry) experiments using a Cys-free mutant PfPCP (PfC142/188S) (23). The Cys-free mutant is very suitable for calorimetry because Cys residues introduce complexity into the heat capacity curve in DSC.
Therefore, in this study, we determined the X-ray structures of PfPCP at 2.2 Å resolution and its Cys-free mutant protein (PfC142/188S) at 2.7 Å. The origin of the extremely high stability of proteins from hyperthermophiles will be discussed on the basis of the structures of the newly determined PfPCPs, and those already reported for TlPCP and PfPCP.
The wild type PCP (PfPCP) from P. furiosus was expressed in Escherichia coli strain JM109/pPCP3. The Cys mutant protein of PfPCP with Ser substituted at positions Cys142 and Cys188 (PfC142/188S), was constructed by site-directed mutagenesis (22). PfPCP and PfC142/188S were purified as described previously (22). Crystallization experiments were carried out at 25°C using the hanging-drop vapor diffusion method. The concentrations of both purified proteins were adjusted to 10 mg/ml in 80 mM Na-acetate buffer, pH 4.6 [including 4 mM EDTA and 4 mM dithioerythritol (DTE) in the case of the PfPCP]. Crystals of PfPCP were obtained from a reservoir solution containing 80 mM Na-acetate buffer, pH 4.6, including 4 mM EDTA and 4 mM DTE and 4% (w/v) PEG 4 K. Crystals of the PfC142/188S mutant were obtained from a reservoir solution containing 80 mM Na-acetate, pH 4.6, 0.5 M ammonium sulfate and 30% (v/v) 2-methyl-2,4-pentanediol (MPD). Both crystals, the wild type and its Cys mutant, belong to space group P21 with cell dimensions of a = 57.9 Å, b = 105.0 Å, c = 78.5 Å, and b = 90.7°, and a = 49.0 Å, b = 105.8 Å, c = 105.8 Å, and b = 96.2°, respectively.
The X-ray intensity data for PfPCP were collected on Rigaku R-AXIS IIC imaging plates using graphite-monochromated CuKa radiation. The data for PfC142/188S were measured in Mac Science DIP2030 nickel-filtered double-mirror focused CuKa-radiation. A Rigaku RU-200 rotating-anode X-ray generator was operated at 40 kV and 100 mA. Crystals of wild-type and mutant PCPs diffracted up to 2.2 and 2.5 Å resolution, respectively. The intensity data were processed and integrated by the program DENZO and scaled by the program SCALEPACK (28). Data collection statistics are summarized in Table I.
The structure of PfPCP was solved by the molecular replacement (MR) method using the X-PLOR package with the X-ray structure of PCP from Bacillus amyloliquefaciens (BaPCP, PDB entry No. 1AUG). The initial sigma-a weighted electron density map was then refined by solvent flattening (29), histogram mapping (30), and the non-crystallographic symmetry (NCS) averaging technique (31, 32) using the program DM from the CCP4 suite (33). The modified electron density map showed the main chain and most of the side chains. PCP was a homotetramer of four independent monomers related by noncrystallographic 222 symmetry. One molecule was built into the electron density map using the program TURBO-FRODO version 5.5 (34), and other molecules in the asymmetric unit were generated by the NCS operation. Because the electron density of the N-terminal and C-terminal loop was not clear, 2Fo-Fc omit maps were used to built structural models of these regions. Simulated annealing, positional and B-factor refinement were carried out using the diffraction data in the resolution range between 10 to 2.2 Å with the program X-PLOR (35). The crystal structure was restrained under NCS operations during the refinements. The final round of B-factor refinement used the program CNS (36) by including bulk-solvent scaling (37). The structure of the mutant protein, PfC142/188S, was determined by the MR technique using the obtained tetrameric model of PfPCP. Structure refinement of the mutant PCP was performed by following that of the wild-type. Refinement statistics of both structures are summarized in Table II.
Secondary structures of both PCPs were inspected by the program PROCHECK. Interatomic contacts with a donor-to-acceptor distance of less than 3.2 Å were assigned as hydrogen bonds. Superposition of the two molecules and estimation of rms deviations between them were performed by the program Biocon.
The refinement of PfPCP was converged to an R factor of 19.5% and an Rfree of 23.2% for reflections in the resolution range 10-2.2 Å. The Cys-free mutant (PfC142/188S) was refined to an R factor of 21.0% and an Rfree of 24.3% for reflections from 10 to 2.7 Å. Both current models include all 832 residues, 208 residues in each of four subunits, and 168 water molecules for PfPCP and 124 water molecules for PfC142/188S. The refined structure model of PfPCP converged well with rms deviations from the ideal bond length and the angles of 0.017 Å and 3.09°, respectively. 90.6% of the non-glycine residues were in the most favorable region of the Ramachandran plot, 9.2% in the additionally allowed region, and no residues in the disallowed region. Pro159 in each subunit was in a cis conformation. The rms deviations from the ideal bond length and the angles for PfC142/188S were 0.025 Å and 3.68°, respectively. 89.7% of the non-glycine residues were in the most favorable region of the Ramachandran plot, 10.3% in the additionally allowed region, and no residues in the disallowed region. The final coordinates have been deposited in the Protein Data Bank (accession No. 1IOF for PfPCP and No. 1IOI for PfC142/188S).
The structure of the PfPCP monomer is depicted by a ribbon drawing in Fig. 1a. Monomeric structures of BaPCP and TlPCP superposed on that of PfPCP by the program O (39) were drawn by stereoscopic Ca traces using TURBO-FRODO (34). The PfPCP monomer folded into a single a/b globular domain. The central b-sheet comprised four-stranded parallel b sheets (b3, b1, b4, and b9), two short two-stranded anti-parallel b sheets (one with b2 and b3, and the other with b5 and b9) and a short parallel b-sheet (b5 and b8). The central b-sheet was sandwiched within five a-helices; two helices (a2 and a4) on one side and three helices (a1, a6, and a3) on the other side. Polypeptide segments from Arg80 to Ser113 and from Tyr169 to Cys188 protruded from the core a/b domain, which are termed the L-shaped protrusion and C-terminal protrusion, respectively, in this paper. The C-terminal protrusion included a 2.5 turn a-helix of a5 and an extended strand. The L-shaped protrusion consisted of a short two-strand anti-parallel b-sheet (b6 and b7) at the horizontal line of L and a distorted anti-parallel b-sheet at the vertical line which was assigned to a loop by the program PROCHECK. Both the C-terminal and the L-shaped protrusions have higher temperature factors than the core domain.
PfPCP exhibits a tetrameric structure similar to that of TlPCP and BaPCP. Individual subunits of PfPCP, termed A, B, C, and D, superpose well on the other subunits related by 222 symmetries within rms displacements less than 0.023 Å for all the main-chain atoms. Three twofold axes were designated the P-, Q-, and R-axes as shown in Fig. 2a. The P-axis relates A to C and D to B, the Q-axis, A to D and C to B, and the R-axis, A to B and C to D. Each subunit of PfPCP is in contact with two other subunits at two different interfaces. One interface, through which the P-axis runs, was designated the P-interface, and the other, around the Q-axis, the Q-interface. The intersubunit contacts in the P-interface are mostly hydrophobic interactions, while those at the Q-interface include ion bonds and hydrophobic interactions.
The hydrophobic contacts between two C-terminal protrusions related by the P-axis are the main intersubunit interactions in the P-interface. Each C-terminal protrusion has an additional intersubunit interaction with a C-terminal end of a3, a loop between an a3 and a b8, a strand of b8, a loop between b4 and b5, and an N-terminal end of a6. Hydrophobic interactions appeared prominent, and no ion bond was detected at the P-axis interface.
The L-shaped protrusion is responsible for intermolecular interactions at the Q-interface. This protrusion interacts with an other L-shaped protrusions, the N-terminal end of a2, the N-terminal end of a3, the C-terminal end of b5, the loop between b5 and b6, the C-terminal end of b8, the loop between b8 and a4, and the N-terminal end of a4. Ten ion bonds were detected at this interface. They were Asp87-Arg80, Asp100-Lys118, Glu99-Lys118, Asp100-Arg80, Asp100-Lys118, and their twofold symmetry-related pairs. In addition to these charged amino acids, Ile81, Val83, Ala85, Thr109, Ala110, Tyr111, Phe112, Ser136, and Leu139 were included in the hydrophobic interactions around the Q-axis.
In TlPCP, Cys190 locates close to the end of the C-terminal protrusion, that is, at the beginning of the C-terminal helix (a6) in each monomer, forms a disulfide bond (A-C and B-D) (26). Although PfPCP has Cys188 located at the equivalent position, the existence of an inter-subunit disulfide bond was not observed. The distance of Cys188 Sg between A and C was 3.77 Å.
PfPCP has two cysteine residues, Cys142 and Cys188, one of which (Cys142) is completely conserved in all known PCP sequences. This conserved residue plays a key role in PCP activity. In PfPCP, a nucleophilic Cys142 is located in the beginning of a4 near the central sheet (b5 and b9). A schematic drawing of the PfPCP active site is shown in Fig. 3.
His166 located in b9 has also been suggested to be a second component of the Cys-His catalytic dyad. Cys142 Sg forms a hydrogen bond with His166 Ne1 with a distance of 3.49 Å. Glu79, the third component of a catalytic triad, is located in the loop between b5 and b6. The Glu79 Oe forms a hydrogen bond with His166 Nd with a distance of 2.72 Å. The sequence of this catalytic triad, Cys-His-Glu, is highly conserved in all known PCP sequences, and its three-dimensional arrangement in PfPCP is consistent with those of other PCP structures.
The overall X-ray structure of the Cys-free mutant PCP is essentially identical to that of the wild-type PCP. The rms deviations for the main chain atoms between PfPCP and PfC142/188S are 0.266 Å. Slight deviations were found in the flexible loops (b1-a1, b2-b3, and a5-a6) as shown in Fig. 4a. The distance (4.74 Å) of Ser188 Og between A and C in PfC142/188S was also close to that (3.77 Å) of Cys188 Sg between A and C in PfPCP. These results indicate that the Cys mutation hardly perturbs the conformation of the wild type PCP, so that PfC142/188S is a very suitable protein for studying the mechanism of the extremely high stability of PfPCP using DSC, because no complexities are introduced into the heat capacity curve in DSC by the Cys residues.
As shown in Fig. 2, large structural variations were detected in the C-terminal protrusion among three PCPs, in which 5 residues were inserted into PfPCP and 6 residues into TlPCP (Fig. 5). Excluding the amino acid residues without equivalent residues among the three PCPs, the rms deviations of the Ca atoms were evaluated. The deviations of PfPCP to TlPCP and BaPCP were 0.726 and 1.399 Å, respectively (Fig. 4). This suggests that PfPCP is closer to TlPCP than to BaPCP in terms of its main-chain folding, corresponding to the sequence homology. Sequence alignments of the three PCPs and a schematic drawing of the secondary structure of PfPCP are shown in Fig. 5. The amino acid sequence of PfPCP shows 39.4% (82/208) homology with that of BaPCP and 56.3% (117/208) homology with the TlPCP sequence.
On the other hand, remarkable differences in the structures between PfPCP and TlPCP were detected in six regions along the sequence, the b1-a1, b2-b3, and b3-a2 turns, the L-shaped protrusion, the C-terminal protrusion, and the C-terminal coil. The largest structural change between PfPCP and TlPCP was detected in the C-terminal protrusion as shown in Fig. 2. This region of PfPCP contacts another subunit at the P-interface, while that of TlPCP interacts with two different subunits at the P- and R-interfaces. The structural difference in the L-shape protrusion between PfPCP and TlPCP is larger than that between PfPCP and BaPCP. This region in all PCPs interacts with another subunit related by Q-axis symmetry. Large structural differences between PfPCP and BaPCP were detected in the b1-a1, b2-b3, b3-a2, a2-b4, and a4-b9 turns, the C-terminal protrusion, and the C-terminal coil. PfPCP has a five-residue insertion compared with BaPCP in the C-terminal protrusion. This insertion produces a remarkable structural difference in this region.
The number of hydrogen bonds per residue for PfPCP, TlPCP, and BaPCP is 2.13, 2.16, and 2.00, respectively. The two hyperthermophile PCPs have more hydrogen bonds than the mesophile PCP. The ion pairs of the three PCPs within 4 Å are listed in Table III. Figure 6 shows the ion pairs between subunits A and D for the three proteins. There are ten ion pairs at each interface in the case of PfPCP, while TlPCP and BaPCP contain 6 and 4, respectively.
In PfPCP, a polar amino acid, Glu192, was found in the interior of the molecule (Fig. 7). This Glu192 is completely surrounded by hydrophobic residues, Pro18, Ile22, Ile76, Val167, Pro168, Ile170, Met187, Ala195, and Val196. Charged residues are usually located on the surface of a protein, but Glu192 was completely buried in the interior, suggesting the destabilization effect. But the distance between the carboxyl oxygen atom of Glu192 and the oxygen atom of Pro168 is 2.95 Å, and the atoms can form hydrogen bonds if the carboxyl is protonated. This buried polar residue is conserved in TlPCP (Glu194) in the hydrophobic surrounding, and its three-dimensional arrangement is also similar to that in PfPCP. On the other hand, the corresponding residue in BaPCP is Ile189, and it is also in the hydrophobic interior.
The thermal stability of PfPCP is very high compared with that of BaPCP: the peak temperatures of the DSC curves at pH 9.5 are 104.5°C for PfPCP and 62.9°C for BaPCP (23). If Cys188 in PfPCP forms a disulfide bond between subunits, the stability further increases: the peak temperature of the DSC curve is 116.5°C at pH 9.5. On the other hand, TlPCP seems to be less stable than PfPCP (a half-life of 2 h at 75°C (26), although it comes from a hyperthermophile archaeon. The stabilization mechanism of PfPCP was investigated on the basis of structural differences among the three proteins, PfPCP, TlPCP, and BaPCP.
Hydrophobic interactions in the interior of a protein are among the important stabilizing forces (40). In experiments using a series of hydrophobic mutants of human lysozyme, Takano et al. (41) have found a general rule for the relationship between a hydrophobic effect and the conformational stability of a protein. The change in the unfolding Gibbs energy (DG) due to a hydrophobic effect between wild-type and mutant proteins (DDGHP) can be expressed using changes in the accessible surface area (ASA) of non-polar and polar atoms due to denaturation as follows (42):
DDGHP = aDDASAnon-polar + bDDASApolar (1)
where DDASAnon-polar and DDASApolar represent the differences between wild-type and mutant proteins in DASA of the non-polar and polar atoms, respectively, of all residues upon denaturation. Using the stability/structure database for a series of mutant human lysozymes, the parameters a and b have been determined to be 0.178 and -0.013 kJ mol-1 Å-2, respectively (42). ASA values of proteins in the native state can be calculated using X-ray crystal structures, and those in the denatured state using their extended structures. For the calculation of ASA, C/S atoms in residues were assigned to ASAnon-polar and N/O to ASApolar.
We tried to estimate the contribution of hydrophobic interactions in PfPCP, TlPCP, and BaPCP to the stabilization (DDGHP). From the X-ray crystal structures of the three proteins in the native state, ASA values were calculated by the procedure of Connolly (43). The ASA values in the denatured states were calculated from the extended structures of each protein, which were generated from the native structures using Insight II. As shown in Table IV, the DG values due to hydrophobic interactions (DGHP) of PfPCP were greater than those of BaPCP: the DDGHP values between PfPCP and BaPCP were 63 and 43 kJ/mol per monomer, assuming a monomer in the native state and a tetramer in the native state, respectively. However, the DDGHP values for PfPCP were smaller than those of TlPCP. These results indicate that the higher stabilities of PfPCP and TlPCP are caused by hydrophobic interactions. The contents of hydrophobic residues (total of Ile, Val, Leu, Phe, Cys, Met, and Ala) are 41.9, 40.0, 37.7% for PfPCP, TlPCP, and BaPCP, respectively. Although the content in TlPCP is not the highest, the higher hydrophobic effect of TlPCP might come from buried hydrophobic ethylene groups on Lys and Arg, which contribute to the hydrophobic interactions.
Ota et al. (44) have proposed that the changes in conformational stability due to a single amino acid substitution can be calculated by SPMP (stability profiles of mutant proteins) using the X-ray structures. In SPMP, the pseudo-energy potential (DDGSPMP) developed for evaluating structure-sequence compatibility in the structure prediction method is employed, consisting of four elements: side-chain packing (DDGSP), hydration (DDGHyd), local structure (DDGLC), and back-bone-side-chain repulsion (DDGBR) (44, 45).
DDGSPMP = DDGSP + DDGHyd + DDGLC + DDGBR (2)
The pseudo-energy potential provides a fitness score for each residue type at a site in a native structure. The estimation of DDG for a replacement of residue X by residue Y is as follows.
DG(X) = <G(X)> - G(X) = GD(X) - GN(X) (3)
DDG(X—>Y) = DG(Y) - DG(X) = [GD(Y) - GN(Y)] - [GD(X) - GN(X)] = [GD(Y) - GD(X)] - [GN(Y) - GN(X)] (4)
where G(X) and <G(X)> represent the fitness scores for residue X at a site in the native structure calculated by the potential and the reference score, respectively, corresponding to the Gibbs energy in the native and denatured states for residues X, GN(X), and GD(X), respectively.
The conformational stabilities of the three PCP structures were analyzed by SPMP. The individual stability score of four terms are summarized in Table V. The difference in the total score of the PCPs from hyperthermophiles and the mesophile, e.g., Tl-Ba and Pf-Ba, explains the significant stability of both hyperthermophile PCPs very well; however, the contributions of each score term differ slightly. TlPCP appears to be stabilized mainly by the side-chain packing term, while the hydration and local structural terms are the contributors to PfPCP stabilization. Although it is not easy to interpret the side-chain packing and hydration terms individually because both originate from hydrophobic interactions, it seems that the packing pattern of PfPCP is simpler than that of TlPCP: hydrophobic and hydrophilic residues are located on the interior or exterior of the protein structure, respectively, and TlPCP has a more preferable side-chain arrangement than PfPCP. On the basis of a tetramer conformation, the stability score of PfPCP was also the highest, as shown in Table Vb.
The fitness of the PCP sequence at each position was estimated by the knowledge-based function (SPMP). Comparing the stability score of each residue of PfPCP with that of the corresponding residue in BaPCP, the residues important for the conformational stability of PfPCP can be determined. The top 10 differences are listed in Table VI, suggesting which residues of PfPCP contribute to the stability as compared with BaPCP.
The largest difference in the stability scores is between Ile31 of PfPCP and Glu33 of BaPCP. Ile31 is involved in a b strand and Ile exhibits strong b intensity. Also, Ile 31 interacts with Ile29, Val36, Ile200, and Ile204, and they constitute a small hydrophobic cluster near the protein surface. On the other hand, Glu33 of BaPCP interacts only with Lys194 and they form a salt-bridge. Tyr169 of PfPCP (Tyr171 of BaPCP) increases the score by interactions with Gln14, Pro71, and Lys177. On the other hand, Tyr171 and Gly73 appear too close in the BaPCP structure. Pro105 of PfPCP (Gln107 of BaPCP) is very suitable for the local structure around this site: extended coil (position 104)-extended coil (position 105)-left-handed helix (position 106). Tyr132 of PfPCP (Ala134 of BaPCP) in the A chain interacts with Ala75 and Ser77 in the same chain and Val183 in the C chain. Moreover, the b propensity of Tyr is preferable to that of Ala. Ile78 of PfPCP (Pro80 of BaPCP) is in a hydrophobic environment: Ile171, Ile120, Met121, Leu124, Ile113, and Ile165. Around Pro80 in BaPCP, there are also Ile119, Ile122, and Ile167, yet the interactions are less than in the former case. Glu205 of PfPCP (Val202 of BaPCP), on the helix C-terminal and partially buried, is more suitable than Val. Leu124 of PfPCP (I126 in BaPCP) interacts with more hydrophobic residues (Ile76, Ile78, Ile120, Ile129, Ala131, Ala195, and Val198) than Ile126 in BaPCP (Ile78, Ile131, Ala133, and Ile195).
SPMP provides stability scores for each residue at every site. In the case of PfPCP (208 amino acid residues), the stability changes upon mutation. DDG values for 208 x 19 mutants were predicted by SPMP (44) using the crystal structure of PfPCP. 30.3% of the native residues of PfPCP rank at the top of 20 kinds of residues, while the percentage in BaPCP is 27.9%. The average ranking of all native residues among 20 amino acids for PfPCP and BaPCP is 4.95 and 5.74, respectively. These results indicate that the sequence of the hyperthermophile protein (PfPCP) is more fitted to an ideal structure than that of the mesophile protein (BaPCP).
In hyperthermophile proteins, ion pair networks on the protein surface are usually assumed to be stabilizing (4, 9, 46). Table III lists the ion pairs formed within 4 Å in the three PCPs. The number (0.10) of ion pairs per residue for PfPCP is similar to that (0.09) for BaPCP but greater than that (0.07) for TlPCP. All values are greater than the average number (0.04) in mesophile proteins (47). In PfPCP, 31 residues are involved in the formation of 20 ion pairs: 6 ion pairs with 6 residues form an ion pair network between subunits (A and D, or B and C) (Fig. 6a). In BaPCP, 5 ion pairs involving 6 residues form an intramoleculer ion network. TlPCP has only a small ion network between subunits (4 ion pairs involving 5 residues). An intramoleculer ion-network in a mesophilic protein (BaPCP), which is almost buried in the interior of the molecule, disappears in hyperthermophile proteins (PfPCP and TlPCP). Ion pairs between subunits A and D are partly conserved in the three proteins, but those in PfPCP form an intersubunit network and seem to contribute to the higher stability of the tetramer (Fig. 6).
Hydrogen bonds are ubiquitous in proteins, and their contribution to conformational stability is of fundamental importance, as are hydrophobic interactions. Recently, the net contribution of an intramolecular hydrogen bond has been estimated to be 8.5 kJ/mol for a 3 Å-long hydrogen bond as determined using a series of hydrogen bond mutants of human lysozyme (48). The average number of hydrogen bonds per residue in PfPCP, TlPCP, and BaPCP is 2.086, 2.064, and 1.986, respectively. The total number of hydrogen bonds in PfPCP (434/208 residues) is greater than in BaPCP (417/210 residues) but comparable to that in TlPCP (454/220 residues). The increase in 17 hydrogen bonds corresponds to a stabilization of 144 kJ/mol. These results suggest that the hydrogen bonds of both hyperthermophile proteins contribute considerably to their higher thermostabilities compared with the mesophile protein. In addition to intrasubunit hydrogen bonds, there are 10, 14, and 2 intersubunit hydrogen bonds in PfPCP, TlPCP, and BaPCP, respectively. This indicates that both hyperthermophile proteins are also more stabilized by subunit interactions through hydrogen bonding than the mesophile protein.
Important stabilizing factors include entropic effects when the conformational entropy of a protein is decreased in the denatured state due to substitutions or deletion. Entropic effects correspond to the introduction of disulfide bonds, substitution of other residues with Pro, and the shortening of polypeptides. The X-ray structure of PfPCP indicates that Cys188 in PfPCP does not form a disulfide bond, although the corresponding Cys190 in TlPCP does (26). PfPCP is expressed in an E. coli strain that clones only the PCP gene from P. furiosus. Cys188 is partly oxidized during the purification process, and the resulting oxidized PfPCP shows greater stability than the not-oxidized PfPCP (23). The X-ray structure of PfPCP also indicates that Cys188 is capable of forming a disulfide bond between subunits. Therefore, the PfPCP expressed under physiological conditions in P. furiosus at around 100°C might form an intersubunit disulfide bond at Cys188 that stabilizes the conformation.
The number of Pro residues is 15 or 16 in the three proteins, suggesting that there is no difference in stability due to Pro content. The number of residues in PfPCP, TlPCP, and BaPCP is 208, 220, and 215, respectively. The entropic effects of denaturation can be calculated from their amino acid compositions using thermodynamic parameters proposed by Oobatake and Ooi (49): the denaturation entropy for PfPCP, TlPCP, and BaPCP is 0.67, 0.67, and 0.40, kJ mol-1 K-1, respectively. This indicates that neither hyperthermophile protein is stabilized by its entropic effects.
The increase in cavity volume in the interior of a protein decreases the conformational stability. The contribution (DG) of the changes in the cavity volume to protein stability has been estimated to be -0.100 and -0.073 kJ mol-1 Å-3, using mutant T4 lysozymes (50) and data for both mutant human and T4 lysozymes (51), respectively. For tetramers of the three PCPs, the cavity volume was determined by attempting to insert a probe sphere with a radius of 1.4 Å (assuming a water molecule) (43). The total volume of the cavity for PfPCP, TlPCP, and BaPCP is 973, 622, and 1,129 Å3, respectively. This result indicates that TlPCP is stabilized considerably by the decrease in cavity volume, but PfPCP is comparable to BaPCP.
Ionized amino acids in globular proteins are almost exclusively on the outside exposed to the solvent. They are believed to form ion pairs or play essential roles in function (including catalysis and stability) when in the interior of a protein. Glu192 in PfPCP is 100% buried in the hydrophobic core, and the corresponding Glu194 in TlPCP is also completely buried, suggesting that the buried Glu residue is conserved in hyperthermophiles. On the other hand, the corresponding residue in the mesophile protein BaPCP is a hydrophobic residue, Ile195, which might play an important role in hydrophobic interactions. Both Glu residues in the hyperthermophiles, which do not have any ion pairs, seem to contribute to destabilizing the protein. If the Glu residues are replaced by hydrophobic residues, the mutant protein should be stabilized. However, all PfPCPs with hydrophobic mutations at Glu192 are less stable than the wild-type protein (in this case Cys-free PfPCP was used) (in preparation). The OE1 atom of Glu192 in PfPCP is close to the O atom of Pro168 (2.95 Å): the OE2 atom of Glu194 in TlPCP is also close to the O atom of Pro169 (2.62 Å). This suggests that if Glu192 (Glu194) is protonated, the protonated Glu might form a hydrogen bond with the O atom of Pro. However, at present, it is difficult to explain how the conserved buried Glu residues in hyperthermophiles play an important role in stabilization.
The crystal structures of pyrrolidone carboxyl peptidases from P. furiosus and its Cys-free mutant were determined at 2.2 and 2.7 Å, respectively. The extremely high stability of the protein could be explained on the basis of its structure and those of two homologous structures from a hyperthermophile and a mesophile. The high stability of PfPCP is due to increases in hydrophobic interactions and hydrogen bonds, the formation of an inter-subunit ion pair network, and improvement to an ideal conformation. On the other hand, the stabilization of TlPCP is mainly caused by a remarkable increase in hydrophobic interactions and a decrease in denaturation entropy due to the formation of disulfide bonds. From these results and those previously reported, we can conclude that a hyperthermophile protein does not have special factors responsible for its extremely high stability, and that the conformational structure is superior in the combination of positive and negative factors for stabilization compared to a mesophile protein.