Domain-Swapped Dimer of Pseudomonas aeruginosa Cytochrome c 551: Structural Insights into Domain Swapping of Cytochrome c Family Proteins

Cytochrome c (cyt c) family proteins, such as horse cyt c, Pseudomonas aeruginosa cytochrome c 551 (PA cyt c 551), and Hydrogenobacter thermophilus cytochrome c 552 (HT cyt c 552), have been used as model proteins to study the relationship between the protein structure and folding process. We have shown in the past that horse cyt c forms oligomers by domain swapping its C-terminal helix, perturbing the Met–heme coordination significantly compared to the monomer. HT cyt c 552 forms dimers by domain swapping the region containing the N-terminal α-helix and heme, where the heme axial His and Met ligands belong to different protomers. Herein, we show that PA cyt c 551 also forms domain-swapped dimers by swapping the region containing the N-terminal α-helix and heme. The secondary structures of the M61A mutant of PA cyt c 551 were perturbed slightly and its oligomer formation ability decreased compared to that of the wild-type protein, showing that the stability of the protein secondary structures is important for domain swapping. The hinge loop of domain swapping for cyt c family proteins corresponded to the unstable region specified by hydrogen exchange NMR measurements for the monomer, although the swapping region differed among proteins. These results show that the unstable loop region has a tendency to become a hinge loop in domain-swapped proteins.


Introduction
In domain swapping, a protein molecule exchanges its secondary or tertiary structural unit with the corresponding unit of another molecule of the same protein [1,2]. Approximately 5% of protein families based on protein structural classification [3] have been found to contain domain-swapped structures [4]. Domain swapping has also been reported in human pathology- structurally weak regions in the protein [38] have been suggested to correlate with the swapping region. However, it is still difficult to predict the swapping region in proteins. In this study, we show that PA cyt c 551 forms oligomers, in which the region containing the Nterminal α-helix and heme are swapped. The swapping region in cyt c family proteins is shown to correlate with the less stable local structure, based on the comparison of the structure and folding properties between the monomer and domain-swapped dimer.

Materials and Methods
Preparation of cytochrome c 551 Enzymes for site-directed mutagenesis were obtained from Takara Shuzo Co. (Kyoto, Japan). Oligonucleotide primers were purchased from Sigma-Aldrich Japan (Tokyo, Japan). The E. coli expression system of PA cyt c 551 was gifted from Prof. Sambongi [39]. Amino acid substitution of Met61 was performed by PCR-based in vitro mutagenesis of the original plasmid vector using PA-M61A-F and PA-M61A-R primers (S1 Table). Mutated DNA was purified using the QIAprep spin Mini prep kit (QIAGEN, Venlo, Netherlands). DNA sequencing was carried out with the BigDye Terminator v3.1 cycle sequencing kit (Applied Biosystems, Inc., Foster City, CA) and an ABI 3100 Avant generic analyzer (Applied Biosystems, Inc.). Recombinant wildtype (WT) and M61A PA cyt c 551 were overproduced in E. coli JCB387 cells [40]. The cells were grown at 37°C in 5 L flasks containing 2 L of LB broth (Sigma-Aldrich, St. Louis, USA) (20 g/L) for 12 hours and harvested. Oxidized PA cyt c 551 was purified by the previous method [39]. The purity of oxidized WT and M61A PA cyt c 551 was confirmed by the absorption spectrum and elution curve of gel chromatography. The absorption coefficients of oxidized monomeric WT and M61A PA cyt c 551 were estimated as ε 409 = 106.1 mM -1 cm -1 and ε 401 = 142.0 mM -1 cm -1 , respectively, by the pyridine hemochrome method [41]. The absorption coefficient of oxidized dimeric WT PA cyt c 551 was estimated as ε 409 = 109.4 mM -1 cm -1 from the absorbance change by dissociation of the dimer to monomers by heating at 70°C for 10 min. The concentrations of oxidized WT and M61A PA cyt c 551 were calculated from the absorbance at 409 and 401 nm, respectively, and adjusted to desired concentrations.

Optical absorption and CD measurements
Absorption spectra were measured with a UV-2450 spectrophotometer (Shimadzu, Japan) using a 1-cm path-length quartz cell. CD spectra were measured with a J-725 circular dichroism spectropolarimeter (Jasco, Japan) using a 0.1-cm path-length quartz cell.

X-ray crystallography
Crystallization of domain-swapped dimeric PA cyt c 551 was carried out at 277 K using the sitting drop vapor diffusion method with Crystal Screen 1 (Emerald Biosystems Inc., Bainbridge Island, USA). The protein concentration was adjusted to 9.2 mg/ml in 50 mM potassium phosphate buffer, pH 7.0. The droplets prepared by mixing 2 μl of the protein solution with 2 μl reservoir solution were equilibrated. The best reservoir solution was found to be 0.1 M HEPES-NaOH buffer, pH 7.5, containing 1.4 M sodium citrate tribasic dehydrate. A crystal was observed in the protein solution after incubation at 4°C for 5 weeks.
The diffraction data were collected at the BL38B1 beamline of SPring-8, Japan. The crystal was mounted on a cryo-loop without an additional cryoprotectant, and flash-frozen at 100 K in a nitrogen cryo-system. The detector was Quantum315 (ADSC). The crystal-to-detector distance was 250 mm and the wavelength was 0.8 Å. The oscillation angle was 0.5°and the exposure time was 3 sec per frame. The total number of frames was 180. The diffraction data were processed using the program HKL-2000 [42].
The preliminary structure was obtained by the molecular replacement method (MOLREP [43]) using the atomic coordinates of the monomer structure of PA cyt c 551 (PDB ID: 351C) as a starting model. There were two protomers, each from a different domain-swapped dimer molecule, in the asymmetric unit of the crystal. The structure refinement was performed using the program REFMAC [44]. The molecular model was manually corrected, and water molecules were picked up in the electron density map using the program COOT [45]. The data collection and refinement statistics are summarized in S2 Table. Small angle X-ray scattering measurements All samples were prepared in 50 mM potassium phosphate buffer, pH 7.0. Small angle X-ray scattering (SAXS) measurements were carried out using a rotating anode X-ray generator, UltraX18 (Rigaku, Tokyo, Japan), in which a monochromatic X-ray with a wavelength of 1.54 Å was focused through a confocal Max-Flux mirror (Rigaku). Scattering profiles were collected using an X-ray image intensifier CCD detector (Hamamatsu Photonics K.K., Shizuoka, Japan). The path length of the sample cell was 1 mm, and its temperature was controlled to 20°C. A series of monomer dilutions were measured to extrapolate the scattering intensity to zero protein concentration and eliminate the inter-particle interference.

Differential scanning calorimetry measurements
Differential scanning calorimetry (DSC) thermograms of oxidized monomeric and dimeric PA cyt c 551 were measured at a scan rate of 1°C/min with VP-DSC (MicroCal, GE Healthcare) in 50 mM potassium phosphate buffer, pH 7.0.

Electrochemistry
Cyclic voltammetry responses were obtained with ALS-612DN (BAS Inc., Tokyo, Japan). An Au electrode was used as a working electrode, and Pt wire and Ag/AgCl (3 M NaCl) were used as counter and reference electrodes, respectively. The redox potentials were calculated with respect to the normal hydrogen electrode (NHE). Modification of the surface of the Au electrode was performed by the following procedure [46]. The surface of the Au electrode was polished with 0.05 μm alumina water slurry and subsequently rinsed with pure water. To remove residual organic compounds from the electrode surface, the Au electrode was cleaned by electrochemical oxidation/reduction treatment. The Au electrode was dipped in a methanol solution containing 1 mM 4-mercaptopyridine (Wako, Osaka, Japan) for 30 s, and then rinsed with pure water. Cyclic voltammograms of oxidized monomeric (~100 μM, heme unit) and dimeric (~200 μM, heme unit) PA cyt c 551 were recorded in 50 mM potassium phosphate buffer, pH 7.0, containing 200 mM NaCl (Wako). All the measurements were performed at room temperature after degassing with a vacuum line, and flowing Ar gas for at least 5 min to remove oxygen dissolved in the solution.

Results
Oligomer formation of cytochrome c 551 Dimeric, trimeric, and tetrameric WT PA cyt c 551 were produced from oxidized monomeric cyt c 551 by an addition up to 80% (v/v) ethanol, subsequent lyophilization, and resolvation with buffer (S1 Fig). No oligomers larger than the tetramer formed, similar to the case of HT cyt c 552 [18], whereas high order oligomers have been produced in horse cyt c by the treatment with ethanol [15]. These results suggest that a structural restriction and/or electrostatic repulsion suppresses formation of high order oligomers for PA cyt c 551 . The WT PA cyt c 551 dimer was stable at 4°C, although it converted to monomers when heated at 70°C for 10 min (S1 Fig).

Structure of dimeric cytochrome c 551
The absorption spectrum of oxidized dimeric WT PA cyt c 551 was similar to that of its monomer (S2 Fig). The 695-nm band characteristic for the Met-heme iron coordination was observed in the spectrum of the dimer, and exhibited a similar intensity as that in the monomer spectrum (S2 Fig). The intensities of the negative 208-nm and 222-nm CD bands in the dimer spectrum were also similar to those in the monomer spectrum (S3 Fig). These results indicate that the heme environment, especially the His and Met coordination to the heme iron and the secondary structures of PA cyt c 551 did not change by the dimerization. These spectral properties were similar to those of HT cyt c 552 [18]. For horse cyt c, the intensities of the 695-nm absorption band and the 208-nm and 222-nm CD bands were decreased and increased, respectively, by the dimerization [15], suggesting that the effect of domain swapping on the secondary structures of PA cyt c 551 is similar to that of HT cyt c 552 but different from that in horse cyt c.
We solved the X-ray crystal structure of dimeric WT PA cyt c 551 at 1.5 Å resolution to elucidate its detailed structure (Fig 2). There were two independent cyt c 551 molecules with similar tertiary structures in an asymmetric unit of the crystal. Both cyt c 551 molecules exhibited domain-swapped structures, where the N-terminal Glu1-Met22 (helix 1 and loop 1) and the heme were relocated from the original position observed in the monomer. The produced vacant area was occupied by Glu1-Asp19 (helix 1 and half of loop 1) from the other cyt c 551 molecule. The hinge loop was constructed with only three amino acids; Thr20, Lys21, and Met22 (S4 Fig). Interestingly, the active site structure of the dimer was formed by the same amino acids as that of the monomer, but the heme axial ligands (His and Met) belonged to different protomers (Fig 3). The swapped region (Glu1-Asp19) and active site structure of dimeric PA cyt c 551 were similar to those of dimeric HT cyt c 552 [18], but different from those of dimeric horse cyt c, in which the Met80-heme iron bond was disrupted [15].
We have calculated the root-mean-square deviation (rmsd) for the Cα atoms of the regions from the N-terminus to Asp19 and from Val23 to the C-terminus between the structures of the monomer and each protomer (protomer 1 and protomer 2) of the dimer (S1 Table). The rmsd values of both regions were less than 0.9 Å (S5 Fig). These results indicate that the structures of both regions were similar between the monomer and each protomer of the dimer. The hydrogen bond network was also similar between the monomer and those of the dimer. The positions of the side-chains forming the proposed folding nucleus in cyt c family proteins (Pro3, Phe7, Leu74 and Trp77 in PA cyt c 551 ) [47] did not change by the dimerization. There were four major hydrogen bonds (< 3.2 Å between heavy atoms; Cys15CO/Gly24NH, His16N δ / Pro25CO, Ala17CO/Tyr27NH, and Ile18CO/Lys28NH) between the N-terminal region (Glu1-Asp19) and the rest of the protein in the monomer (PDB ID: 351C) (S6 Fig). These  (Fig 4A), indicating that the α-helical content of M61A PA cyt c 551 decreased slightly compared to that of the WT protein. The radii of gyration were obtained as 13.7 and 13.9 Å for WT and M61A PA cyt c 551 , respectively, by SAXS measurements (Fig 4B). Although the size of the global structure of PA cyt c 551 did not change significantly by the removal of Met61, the secondary structures were slightly perturbed (Fig 4). It has been reported that carboxylmethylation of Met61 of PA cyt c 551 destabilizes its folded state [48], indicating that the α-helical structure of PA cyt c 551 is stabilized by the Met-heme coordination. The amount of dimer produced by the treatment with ethanol decreased to less than 5% and no trimer or tetramer was No signal was observed below 82°C in the thermogram of the dimer, as well as that of the monomer. According to the results of gel chromatography (S1 Fig), the dimer dissociated to monomers when heated at 70°C for 10 min. These results show that the PA cyt c 551 dimer dissociates to monomers without a significant enthalpy change (ΔH =~0 kcal/mol). In the case of horse cyt c, the ΔH value for dimer-to-monomer transition was -40 ± 2 kcal/mol. Since Met did not dissociate from the heme iron in dimeric PA cyt c 551 but dissociated in dimeric horse cyt c, we may attribute the difference between PA cyt c 551 and horse cyt c in the enthalpy change on dimer dissociation to the difference in the heme coordination structure of the dimers.

Redox potential of dimeric cytochrome c 551
We measured the redox potential of dimeric PA cyt c 551 by cyclic voltammetry to investigate the effect of domain swapping on the function. The midpoint redox potentials of monomeric and dimeric WT PA cyt c 551 were obtained as 274 ± 5 and 242 ± 5 mV (vs NHE), respectively, at pH 7.0 in the presence of 200 mM NaCl (S10 Fig). The value for the monomer was similar to the reported value (276 ± 2 mV (vs NHE) in pH 7.0 at 25°C) [49]. Although the redox potential of the dimer decreased by about 30 mV compared to that of the monomer, the dimer exhibited a relatively high redox potential, which is characteristic for cyt c family proteins [50,51,52].

Discussion
Met-heme coordination contributes to the stability of the structure and the ability of electron transfer in cyt c family proteins [48,53,54,55]. Although the optical absorption spectra and redox potentials were similar between monomeric and dimeric WT PA cyt c 551 , heme-ligating His and Met originated from different protomers in the dimer (Fig 3), similar to the case of dimeric HT cyt c 552 [18]. In the case of dimeric horse cyt c, Met-heme coordination was perturbed and a hydroxide ion or a water molecule was coordinated to the heme iron [15]. The difference in the heme coordination structure between dimeric PA cyt c 551 and dimeric horse cyt c may be due to the differences in the stability of the Met-heme coordination bond [26,27,28] and the rigidity of the loop containing the heme-ligating Met [20,29]. According to DSC measurements, ΔH for the dissociation of dimeric horse cyt c to monomers exhibited a large, negative value (-40 kcal/mol) [15], whereas the ΔH values for the dissociation of dimeric PA cyt c 551 and dimeric HT cyt c 552 were~0 and +14 kcal/mol, respectively (S9 Fig) [18]. These results show that the coordination of Met to the heme contributes to stabilization of the dimer enthalpically.
Since carboxymethylation of Met61 destabilizes considerably the native state of PA cyt c 551 [48], one may expect that removal of Met61 destabilizes dimeric PA cyt c 551 and thus leads to an increase in high order oligomers. However, formation of dimers was minimal and no trimer was detected by the treatment of M61A PA cyt c 551 with ethanol ( S8 Fig). The α-helical structure of M61A PA cyt c 551 was partially perturbed by the disruption of the Met-heme coordination bond (Fig 4A), although its global structure did not unfold completely (Fig 4B). Therefore, the intermolecular interactions through the N-and C-terminal α-helices and formation of high order oligomers may be suppressed by the perturbation of the α-helical structures in M61A PA cyt c 551 . The interaction between the N-and C-terminal α-helices in dimeric PA cyt c 551 was similar to that of the monomer (Fig 2), although the swapping regions of PA cyt c 551 was different from that of horse cyt c. We have shown that domain-swapped oligomers are generated by intermolecular hydrophobic interaction between the N-and C-terminal α-helices at the early stage of folding for horse cyt c [36]. These results indicate that the swapping region is defined subsequent to formation of the intermolecular interaction between the terminal α-helices which occurs during folding in cyt c family proteins. Moreover, the contacts between the N-and C-terminal α-helices are essential for not only protein folding but also domain swapping.
It has been suggested that PA cyt c 551 folds through multiple transition states separated by a high energy intermediate [23], whereas HT cyt c 552 transiently populates a compact obligatory intermediate during folding [56]. It has also been proposed that the propensity to form elements of stable secondary structures controls the process of folding in proteins [57,58]. It has been shown that the region with the highest helical propensity is helix 3 for PA cyt c 551 according to the calculation by the program AGADIR [59], whereas it is helix 4 (C-terminal helix) for HT cyt c 552 [25]. Although the highest helical propensity region differed between PA cyt c 551 and HT cyt c 552 , the swapping regions in their dimers were similar, suggesting that the helical propensity does not define the swapping region.
It has been suggested by molecular dynamics simulations that the native topology generally determines the domain-swapped structure [60]. Although the topology of the tertiary structure (Fig 5A and 5C) and folding features, such as the burst-phase collapse in the folding process, are similar between PA cyt c 551 and horse cyt c, the swapping region was different between these proteins (Fig 5B and 5D); the hinge loops of PA cyt c 551 and horse cyt c were Thr20-Met22 (at loop 1) and Thr78-Ala83 (at loop 3), respectively. According to hydrogen exchange NMR measurements, the thermodynamic property of foldons is different between PA cyt c 551 and horse cyt c [20,21,24]. A foldon with low energy (small ΔG HX ) loses its secondary structure at the early stage of unfolding, showing that the region corresponding to the low energy foldon has low structural stability. The region containing loop 1 and helix 2 of PA cyt c 551 is low in stability [21], whereas that of loop 3 of horse cyt c is low [20,24]. Interestingly, the hinge loops of the domain-swapped structure in PA cyt c 551 and horse cyt c correspond to the low stability regions of their monomers. Unfolding simulations have suggested that the hinge loop are 'hotspots', around which proteins tend to locally unfold prior to complete unfolding [60]. These results show that the region with low stability in the monomer correlates to the hinge loop in domain swapping.
In summary, we found that the region containing the N-teminal α-helix and heme was exchanged in the domain-swapped dimer of PA cyt c 551 . The active site structure of the dimer was formed by the same amino acids as that of the monomer, but the heme axial ligands (His and Met) belonged to different protomers. By comparison of the domain-swapped structures of cyt c family proteins, we propose that the stability of the local structure may determine the position of the hinge loop in domain swapping, and thus the knowledge on protein folding may help to predict the structural features of domain swapping.