The Electron Transfer Complex between Cytochrome c552 and the CuA Domain of the Thermus thermophilus ba3 Oxidase

The structural analysis of the redox complex between the soluble cytochrome c552 and the membrane-integral cytochrome ba3 oxidase of Thermus thermophilus is complicated by the transient nature of this protein-protein interaction. Using NMR-based chemical shift perturbation mapping, however, we identified the contact regions between cytochrome c552 and the CuA domain, the fully functional water-soluble fragment of subunit II of the ba3 oxidase. First we determined the complete backbone resonance assignments of both proteins for each redox state. Subsequently, two-dimensional [15N,1H]TROSY spectra recorded for each redox partner both in free and complexed state indicated those surface residues affected by complex formation between the two proteins. This chemical shift analysis performed for both redox states provided a topological description of the contact surface on each partner molecule. Remarkably, very pronounced indirect effects, which were observed on the back side of the heme cleft only in the reduced state, suggested that alterations of the electron distribution in the porphyrin ring due to formation of the protein-protein complex are apparently sensed even beyond the heme propionate groups. The contact residues of each redox partner, as derived from the chemical shift perturbation mapping, were employed for a protein-protein docking calculation that provided a structure ensemble of 10 closely related conformers representing the complex between cytochrome c552 and the CuA domain. Based on these structures, the electron transfer pathway from the heme of cytochrome c552 to the CuA center of the ba3 oxidase has been predicted.

possibly because the electrostatic attractions would rather be weakened at the high temperatures these bacteria are exposed to. This different specificity between the reaction partners is also supported by the fact that cytochrome c 552 from P. denitrificans does not interact with the Cu A domain of T. thermophilus (18).
Under steady-state turnover conditions at 25°C, molar redox activities with k max ϭ 250 s Ϫ1 have been reported between cytochrome c 552 and the ba 3 oxidase of T. thermophilus (5). The complex therefore has to be short lived to ensure efficient electron transport (ET). 5 This transient nature of the redox interaction precludes the detection of any intermolecular NOE connectivities to define the contact region between the proteins. However, the highly sensitive amide resonances allow the observation of chemical shift changes as a result of transient alterations in the local environment due to the presence of the redox partner, as previously demonstrated with other systems such as plastocyanin/cytochrome c, plastocyanin/cytochrome f, cytochrome c peroxidase/iso-1cytochrome c, and Cu A domain/cytochrome c 552 from P. denitrificans for example (19,(21)(22)(23).
We therefore employed two isolated, soluble components, i.e. cytochrome c 552 and the Cu A domain, to determine the biologically relevant ET complex of the T. thermophilus system. Contrary to the Cu A domain from P. denitrificans, which was not sufficiently stable for prolonged NMR data collection at room temperature, the Cu A domain from T. thermophilus proved highly stable. Both proteins were complexed under uniform redox conditions that precluded ET; but the transient complex interaction apparently still took place, as in the P. denitrificans system (19). Interestingly, analogous to the previous P. denitrificans study, in the case of reduced cytochrome c 552 from T. thermophilus we again detected the most pronounced shifts at residues located in the protein interior behind the heme ring. These indirect effects are an indication for redox state-dependent alterations of the electron delocalization in the porphyrin system. Based on chemical shift perturbation mapping, protein-protein docking calculations subsequently yielded the first structural characterization of the ET complex between cytochrome c 552 and the Cu A domain that is founded on experimental data. Using this information, the shortest ET pathway from the heme iron to the Cu A center was calculated based on the "pathway model," revealing an involvement of Phe 88 , which however does not play such a crucial role as the corresponding Trp 121 residue in the P. denitrificans system (24). Two alternative ET scenarios, matching our experimental mutagenesis data, will be discussed.

EXPERIMENTAL PROCEDURES
Sample Preparation-Cytochrome c 552 (133 amino acid residues; 14,405 Da including the heme cofactor) and the Cu A domain (136 residues; 15,062 Da including the two copper atoms) of T. thermophilus were both expressed heterologously in Escherichia coli and subsequently purified as described previously (18,25). For 15 N enrichment, both proteins were expressed in M9 medium. In the case of cytochrome c 552 , heme maturation was achieved by co-transformation of the E. coli cells with the ccmABCDEFGH gene cluster (26) present on the pEC86 plasmid. Copper atoms were introduced into the apo-Cu A domain by addition of Cu(His) 2 after cell lysis. The NMR resonance assignments revealed that the soluble cytochrome c 552 protein carried an alanine-tothreonine point mutation in position 123; subsequent activity tests, however, showed the same functionality as the wild-type protein. The Cu A domain, the water-soluble fragment of the ba 3 oxidase, also was fully functional as revealed by redox spectroscopy (18).
For the resonance assignments, NMR samples of 2 mM protein concentration were prepared for each redox partner, containing 20 mM potassium phosphate buffer (pH 6.0), 0.15 mM 2,2-dimethyl-2-silapentane-5-sulfonate (DSS) as internal chemical shift reference (Cambridge Isotope Laboratories, Andover, MA) and 5% D 2 O. Depending on the redox state to be investigated, 5 mM sodium ascorbate or 5 mM potassium hexacyanoferrate(III) were added to obtain a fully reduced or oxidized sample, respectively.
For the chemical shift perturbation study, the protein samples were prepared according to the following procedures: 1) two samples both containing 0.5 mM 15 N-labeled cytochrome c 552 in 20 mM potassium phosphate buffer (pH 6.0), 0.15 mM DSS, and 5% D 2 O were treated in parallel. Non-labeled Cu A was added in 4-fold molar excess to one of the two samples. The pH values of both samples were adjusted precisely. To obtain cytochrome c 552 sample pairs under reduced or oxidized conditions, the solutions included 4 mM sodium ascorbate or 5 mM potassium hexacyanoferrate(III), respectively. 2) Two samples both containing 0.5 mM 15 N-labeled Cu A fragment in 20 mM potassium phosphate buffer (pH 6.0), 0.15 mM DSS, and 5% D 2 O were treated likewise. Non-labeled cytochrome c 552 was added in 4-fold excess to only one of the two samples and pH values checked for both solutions. To obtain Cu A domain sample pairs under reduced or oxidized conditions, the solutions included 4 mM sodium ascorbate or 5 mM potassium hexacyanoferrate(III), respectively.
NMR Experiments and Analysis-For the resonance assignments of both proteins, NMR spectra were acquired at 25°C using Bruker DMX 500 and DMX 600 spectrometers operating at 500.13 and 600.13 MHz proton resonance frequencies, respectively, and both equipped with 5-mm triple-resonance 1 H{ 13 C/ 15 N} probes that have XYZ-gradient capability. The following spectra were collected to assign the resonances of cytochrome c 552 and the Cu A domain in both redox states, as described previously (27) In the homonuclear one-and two-dimensional 1 H experiments, the water signal was suppressed by selective presaturation during the relaxation delay, with the carrier placed in the center of the spectrum on the water resonance. All heteronuclear experiments made use of pulsed field gradients for coherence selection and artifact suppression and utilized gradient sensitivity enhancement schemes (29). Quadrature detection in the indirectly detected dimensions was obtained either by the States-TPPI (time proportional phase incrementation) or by the echo/ antiecho method. All NMR spectra were acquired and processed on Silicon Graphics computers using the program XWINNMR 3.5 (Bruker Bio-Spin, Rheinstetten, Germany). A 90°phase-shifted squared sinebell function was used for apodization in all dimensions. Polynomial base-line correction was applied to the processed spectra in the directly detected 1 H dimension. Peak picking and data analysis of the transformed spectra were performed using the AURELIA 2.5.9 (Bruker Bio-Spin) software package. The chemical shifts were referenced to internal DSS to ensure consistency among all spectra (30). 5

The Cytochrome c 552 ⅐Cu A Complex from T. thermophilus
For the chemical shift perturbation mapping, two-dimensional [ 15 N, 1 H]TROSY spectra were recorded, as described previously (19), at 25°C and pH 6 on a Bruker Avance 900 spectrometer, operating at 900.13 MHz proton resonance frequency and equipped with a 5 mm cryogenic z-gradient 1 H{ 13 C/ 15 N} triple-resonance probe. Data acquisition and processing were performed as described above. The backbone amide peaks were picked with the program FELIX 2000 (Accelrys Inc., San Diego, CA). Chemical shift differences in the amide proton (⌬␦ 1HN ) and nitrogen (⌬␦ 15N ) resonances of the free and complexed protein forms were combined for each residue by using the expression [(⌬␦ 1HN ) 2 ϩ (⌬␦ 15N /6.5) 2 ] 1/2 (31). These combined chemical shift differences were illustrated with the program MOLMOL (32) by colorcoding each respective surface residue.
Docking Calculations-The structure of the cytochrome c 552 ⅐Cu A domain complex was determined with the program HADDOCK (high ambiguity driven protein-protein docking) (33) that has been implemented in the program CNS (34), making use of python scripts derived from ARIA (35) for automation. HADDOCK employs biophysical interaction data, such as chemical shift perturbation resulting from NMR titration experiments, that are introduced into the structure calculation as ambiguous interaction restraints (AIRs) to drive the docking process. In our application, four independent sets of amide proton chemical shift perturbation data were available: 15 N-labeled cytochrome c 552 titrated with non-labeled Cu A domain, both in the reduced and oxidized state, and 15 N-labeled Cu A domain titrated with non-labeled cytochrome c 552 , again both in the reduced and oxidized state. From each independent data set, the residues with the strongest chemical shift perturbations were selected as "active AIRs" for the HADDOCK calculations; residues with less than 20% accessible surface area, however, were rejected. In addition, surface residues located next to the selected active AIRs were chosen as "passive AIRs", if their relative surface accessibility was also above 20%. The x-ray coordinates of cytochrome c 552 (PDB ID code 1DT1) and the ba 3 oxidase (PDB ID code 1EHK) were used both to calculate the surface accessibilities with the program NACCESS (36) and for the subsequent docking calculations.
First, 100 structures of the cytochrome c 552 ⅐Cu A domain complex were calculated using the rigid docking protocol of HADDOCK. Next, the 20 structures showing the lowest AIR violations were further energy minimized with the side chains of the active and passive residues left flexible. Finally, the 10 best structures were minimized once more in a 8 Å shell of explicit TIP3P (37) water molecules (for details, see Ref. 33). AIR violations, interaction energies, and buried surface areas of the final structures were compared. In addition, the complexes were characterized in terms of electron transfer by the evaluation of electron-tunneling coupling factors using the program GREENPATH (38). In the latter program, the highly interconnected network of bonded and nonbonded contacts within the protein matrix is searched to specify the pathways that maximize the electron-tunneling coupling between the electron donor and acceptor (i.e. the iron atom of the cytochrome c 552 heme moiety and the copper center of the Cu A domain, respectively). This is achieved using the theories and methods developed by Beratan, Onuchic, and Betts (39), which quantify the ET (without interference) using the pathway model. A pathway is defined as a combination of interacting atoms that link the donor with the acceptor via covalent bonds, hydrogen bonds, and through-space jumps. Rates of non-adiabatic ET reactions can be defined by the expression k ET ϭ (2/h -) ͉T DA ͉ 2 (FC). Thereby, the term T DA describes the donor-acceptor interaction associated with electron tunneling, while the FC (Franck-Condon) term contains the free energy dependence (reorganization and reaction free energy) related to nuclear motion.

RESULTS AND DISCUSSION
To allow chemical shift perturbation mapping, the full set of backbone amide resonance values had to be determined for both proteins, i.e. cytochrome c 552 and the Cu A domain from T. thermophilus. The resonance assignments were performed according to the classical strategy based on NOE connectivities between adjacent residues (40), with 15 N labeling used to achieve a better signal dispersion. The resulting 1 H and 15 N assignments of cytochrome c 552 and the Cu A domain in both redox states have been deposited at the BioMagResBank data base.
Both redox partners investigated in this study had been shown earlier to be fully functional when expressed in a heterologous host organism (18). Moreover, the T. thermophilus proteins, in particular the Cu A domain, displayed a higher stability compared to the homologous proteins from P. denitrificans, which had been employed in an earlier NMR study (19).
NMR Resonance Assignments-The reduced (i.e. diamagnetic) cytochrome c 552 was fully assigned (BMRB-6966); heme proton resonances were determined based on NOE data that agreed with interproton distances in the x-ray structure (PDB ID code 1DT1). Several residues showed highly unusual chemical shift resonances because of ring-current effects; the backbone amide proton resonance of Leu 116 , for example, is located upfield of the water signal at 4.35 ppm (Fig. 1). The oxidized (i.e. paramagnetic) cytochrome c 552 was assigned nearly completely (BMRB-6967), except for the heme ring and three protein residues: His 15 and Met 69 , the two axial ligands of the iron atom, and Cys 14 , which is bound covalently to the heme ring. Only a few of the heme proton resonances could be identified by using NOE information from the two-dimensional and three-dimensional NOESY data. Certain resonances showed strong line broadening due to the proximity of the paramagnetic center.
The reduced (i.e. diamagnetic) Cu A domain had been previously assigned using triple-resonance experiments (25); however, several aromatic ring proton resonances have been additionally identified in the present work based on homonuclear two-dimensional TOCSY and NOESY spectra (update of BMRB-5819). Interestingly, the 1 H resonances of the Phe 88 ring, which is situated close to the Cu A center and has been postulated to play a role in the ET (13), were the only aromatic ring resonances that could not be identified. In the case of the oxidized (i.e. paramagnetic) Cu A domain, the assignment of the 1 H and 15 N resonances was again not complete (BMRB-6965), due to the paramagnetic Cu A center. Moreover, as reported elsewhere (41), several resonances are extremely shifted, such as an amide group at 300 ppm and certain C␤ protons at 30 ppm. Nevertheless, except for 9 residues (i.e. the three N-terminal amino acids Met 33 -Tyr 35 , Gly 115 , and Cys 149 -Cys 153 ), all other backbone amide groups could be identified for the oxidized Cu A domain.
Chemical Shift Perturbation Mapping-To obtain structural data on the transient complex between cytochrome c 552 and the Cu A domain, two-dimensional [ 15 N, 1 H]TROSY spectra comparing the free and the complexed state of each protein were collected. Employing 15 N-labeled protein samples, the chemical shift changes upon addition of 4-fold molar excess of the non-labeled reaction partner provided crucial information about the residues that are affected by the formation of the complex. The non-labeled redox partner was always added in excess, to shift the equilibrium toward the associated complex state. Nevertheless, the observed effects were rather small, presumably due to the very short-lived nature of the cytochrome c 552 ⅐Cu A domain complex. The chemical shift changes in the 1 H and 15 N dimensions between the free and complexed protein form were subsequently combined for each residue (31), as indicated in Fig. 2.
In the case of cytochrome c 552 , for both redox states the most pronounced shift perturbations upon addition of the Cu A domain were seen in residues located around the heme cleft (Fig. 3), indicating that the "front side" of the protein is the contact surface during the interaction with the redox partner, similar to the corresponding P. denitrificans system (19). Moreover, also analogous to the effects noted for P. denitrificans, the largest shifts in the reduced state of cytochrome c 552 from T. thermophilus were observed on the back side of the heme cleft, at residues Ala 34 and His 32 (Fig. 4); since both residues are not exposed at the protein surface, these dominant shifts must be indirect effects that are apparently relayed from the contact surface through the heme pocket. Interestingly, these indirect effects (at Ala 34 and His 32 in T. thermophilus and at Gly 54 , Gly 55 , and Asp 56 in P. denitrificans) stand out only in the reduced but not in the oxidized state of cytochrome c 552 . This finding suggests that these chemical shift perturbations in the back of the heme moiety are a result of the electronic differences between the two redox states. More precisely, His 32 and Arg 125 form a hydrogen bonding network with the propionate A chain at the back of the heme ring (Fig. 5, bottom), resembling the arrangement of Trp 57 , Arg 36 , and propionate A in the P. denitrificans system (Fig. 5, top). We therefore propose that the electronic state of the heme is propagated through the propionate A substituent and across the hydrogen bond to the aromatic ring, i.e. His 32 in T. thermophilus or Trp 57 in P. denitrificans. In the case of P. denitrificans, for example, fluorescence spectra of reduced and oxidized cytochrome c 552 (see Fig. S1 in the supplemental data) had shown a 20-nm shift of the tryptophan band (42). This is apparently due to an alteration in the electronic structure of Trp 57 , since the protein conformation is identical in both redox states as confirmed by both x-ray and NMR structure analysis (42,43), thus excluding an explanation that is based on conformational changes in the protein structure. Hence, the only distinction that could explain this redox state-dependent effect in the fluorescence spectrum of P. denitrificans cytochrome c 552 is the additional electron delocalized across the porphyrin system. This is a clear indication that the electronic state of the heme system is also sensed in the protein region located beyond the propionate groups. Presumably, the electrons of the heme iron show an effective delocalization toward the periphery of the porphyrin ring including its substituents, as previ- ously suggested by Johansson et al. (44,45): the actual change of the central iron charge in the redox reaction is only about 0.1-0.2 electrons, despite the unit difference in the formal oxidation state. This relatively small difference in electron probability at the heme iron implies a considerable electron delocalization into the periphery of the porphyrin system, which seems to be very important for both the ET rates and the accommodation of the charged heme moiety in a low dielectric enviroment such as the interior of a protein (44,45). In our chemical shift perturbation study, the reduced cytochrome c 552 is additionally complexed with the reduced Cu A domain. Hence, the electron delocalization in the heme porphyrin ring of the protein complex may be distributed even further into the back of the heme pocket, to minimize unfavorable Coulomb interactions that arise because of increased electron repulsion in the combined heme-Cu A system, as both redox centers are fully occupied with electrons. This shift in the electron density upon complex formation apparently is sensed by the ring systems of Trp 57 in P. denitrifians or His 32 in T. thermophilus via the hydrogen bond connection to propionate A, thereby in turn presumably affecting their respective local environments (see Fig. 5).
In case of the Cu A domain, the most pronounced shift perturbations upon addition of cytochrome c 552 occurred in different regions (Fig. 6). For the docking calculations, however, several of the affected residues could be excluded because of either low surface accessibility or location at the interface to subunit I of the ba 3 oxidase, as described below. The contact region relevant for the ET is located near the Cu A center, at the surface residues Ala 87 , Phe 88 , Gln 158 , and Asn 159 .
Theoretically, in the fully oxidized state pseudocontact shifts (see Ref. 46 and references therein) could occur in residues of cytochrome c 552 that are closest to the paramagnetic copper center of the Cu A domain, and vice versa; such shifts, however, would hardly be distinguishable from those due to "true" intermolecular contacts. They would arise at the interface between the copper and iron metal centers where most of the intermolecular contacts occur, and thus both effects on the chemical shift would superpose. The impact on the structure calculation using AIRs would therefore be negligible, as indicated also by the consistency of our calculations (see below).
Docking Calculations-To perform docking calculations between cytochrome c 552 and the Cu A domain, it was necessary to make a reasonable selection among the residues affected in the chemical shift per- turbation experiments, based on their surface accessibility and location in the molecule relative to the redox center.
In the case of reduced cytochrome c 552 , residues Ala 34 , His 32 , Gly 24 , Ser 70 , Gln 16 , His 15 , Leu 116 , Leu 29 , Cys 14 , and Lys 98 (in this order) showed the largest combined chemical shift perturbations (⌬␦ Ն 0.008 ppm). Some of these residues were excluded as possible contact partners for the following reasons: Ala 34 , His 32 , His 15 , Leu 116 , Leu 29 , and Cys 14 were rejected because of a too low surface accessibility (Ͻ20%); Lys 98 could be neglected due to its location on the back side of the molecule. Thus, residues Gly 24 and Ser 70 (both with over 40% relative surface accessibility) were chosen as active AIRs, whereas Gln 16 with only 28.7% relative surface accessibility was classified as passive AIR.
In the oxidized cytochrome c 552 , residues Lys 115 , Gln 57 , Ala 113 , Gly 56 , Ala 105 , Gln 119 , Asn 18 , Gln 120 , Gly 13 , Gly 24 , and Val 68 (in this order) showed the largest combined chemical shift perturbations (⌬␦ Ն 0.0124 ppm). Ala 105 was excluded because of its positition on the back side of the molecule. Gly 56 (with 23.4% relative surface accessibility) was classified as passive AIR. All the other affected residues show over 40% relative surface accessibility and were therefore accepted as active AIRs.
In the case of the reduced Cu A domain, residues Gly 120 , Arg 141 , Ile 45 , Glu 51 , Arg 52 , Glu 126 , Leu 50 , Phe 88 , Asn 159 , Gln 158 , Arg 146 , Lys 140 , and His 157 (in this order) showed the largest combined chemical shift perturbations (⌬␦ Ն 0.010 ppm). Gly 120 , Ile 45 , and Glu 126 could be excluded because of their position at the interface to subunit I of the ba 3 oxidase  Fig. 2A), with the color intensity normalized to a maximum of 100% for the residue that was most strongly affected upon complex formation. The affected residues are shown in red for the reduced and in blue for the oxidized protein; those residues showing the most pronounced shifts are labeled. For clarity, not just the backbone amides are highlighted but rather the entire residues have been colored. The molecules on the left and right are rotated by 180°about the vertical axis relative to each other. The front side (left picture) with the heme (green) in the center represents the contact surface in the complex with the Cu A domain. (see Fig. S2 in the supplemental data). Arg 141 , Arg 52 , Leu 50 , and Lys 140 were also neglected, since these residues are only accessible in the soluble Cu A fragment, while in the full ba 3 oxidase their side chains should be immersed into the lipid membrane. Glu 51 was not taken into account, since it is located at the opposite side of the Cu A domain relative to the copper center. His 157 and Gln 158 were rejected because of too low surface accessibilities (Ͻ20%). The remaining residues Phe 88 , Arg 146 , and Asn 159 were chosen as active AIRs.
In the oxidized Cu A domain, residues Asn 122 , Val 112 , His 157 , Asn 159 , Gly 115 , Gly 156 , Val 127 , Ala 85 , His 117 , and Ala 87 (in this order) showed the largest chemical shift perturbations (⌬␦ Ն 0.020 ppm). Val 112 , His 157 , Gly 115 , Val 127 , and Ala 85 could be excluded because of too low surface accessibilities (Ͻ20%). Asn 122 and His 117 were also neglected because they are located at the interface to subunit I of the ba 3 oxidase. The remaining residues Ala 87 , Gly 156 , and Asn 159 were accepted as active AIRs.
All residues that were thus chosen for the docking calculations as active AIRs (10 and 5 for cytochrome c 552 and the Cu A domain, respectively) are listed in Table 1. Consequently, neighboring residues with relative surface accessibility above 20% were selected as passive AIRs for the calculations. Based on these AIRs, 100 rigid structures of the cytochrome c 552 ⅐Cu A domain complex were calculated with the HAD-DOCK program (33). The 20 structures with the lowest interaction energies were further energy minimized by keeping the side chains of the active and passive residues flexible. Finally, the 10 lowest energy structures were minimized once more in a shell of explicit water molecules. Listed in Table 2 are the energy terms, buried surface areas, ET distances, ET pathway lengths, and ET efficiencies of the 10 final structures. The distance between the electron donor (i.e. the iron atom of cytochrome c 552 ) and acceptor (i.e. the copper atom CU2 of the Cu A domain) varies between 15.6 and 16.8 Å. The estimated electron-tunneling coupling factor (log ͉T DA 2 ͉) ranges from Ϫ11.2 to Ϫ12.8 and the electron pathway length from 19.6 to 24.5 Å. The total interaction energy varies in the ensemble between Ϫ124 and Ϫ79 kcal/mol.
As the contact surfaces of the protein molecules are rather flat, and since the AIRs allow different contact combinations between the active and passive residues of the two redox partners, no single preferred solu-  T. thermophilus (B). In both cases, the carboxylate group of propionate A forms hydrogen bonds to an arginine side chain and an aromatic ring. The amide groups with the largest chemical shift perturbations upon complex formation in the reduced state (i.e. Gly 54 , Gly 55 , and Asp 56 in P. denitrificans, respectively, His 32 and Ala 34 in T. thermophilus) are always situated beyond the aromatic ring.  Fig. 2B), with the color intensity normalized to a maximum of 100% for the residue that was most strongly affected upon complex formation. The affected residues are shown in red for the reduced and in blue for the oxidized protein; those residues showing the most pronounced shifts are labeled. For clarity, not just the backbone amides are highlighted, but rather the entire residues have been colored. Molecules on the left and right are rotated around the vertical axis by 180°relative to each other. The Cu A domain has a slightly elongated form, with the ET-relevant contact surface located at the end where the side chains of residues Ala 87 , Phe 88 , Gln 158 , and Asn 159 (purple arrows) protrude at the surface close to the mostly occluded Cu A center (yellow atoms marked by a full circle in the "front view" or by a broken circle in the "back view"). The appendix at the other end denotes the start of the membrane anchor; the Cu A domain itself rests on the membrane-embedded subunit I, as indicated by the gray bar. tion was expected from these docking calculations (33). In the present work, however, superposition of the Cu A backbone atoms revealed an ensemble of complex structures where the backbone r.m.s.d. values of the cytochrome c 552 conformers ranged between 1.57 and 4.54 Å relative to structure 1 (Fig. 7). In other words, the cytochrome c 552 positions display only a moderate variation, thus indicating that all 10 lowest energy conformers essentially belong to the same complex structure cluster. Structures 1-3 (see Table 2), which display the closest proximity between the electron donor and acceptor atoms (Ͻ16 Å), moreover possess the most favorable total interaction energies (ϽϪ100 kcal/mol) and exhibit an identical ET pathway; they were therefore selected as most representative of the cytochrome c 552 ⅐Cu A domain complex and their atom coordinates deposited at the Brookhaven Data Bank under PDB ID code 2FWL. (The cytochrome c 552 backbone r.m.s.d. between the superposed complex structures 2 and 3 is 2.55 Å).
None of the calculated complexes was able to fully compensate the potential energy that is associated with the AIRs; its contribution, however, remains significantly smaller than the van der Waals or electrostatic terms. More importantly, intermolecular contacts with the partner molecule were shown either directly by the residues classified as active AIRs or at least by one of the respective neighboring residues representing passive AIRs. The intermolecular contacts in structure 1, as displayed in Fig. 8, are therefore in agreement with the experimental picture; this has been achieved with a set of high-quality AIRs, derived from four independent experiments. Moreover, in agreement with the postulated hydrophobic/non-ionic character of the cytochrome c 552 ⅐Cu A domain interaction in the T. thermophilus system (18,20), about 40% of the contact surfaces are composed of hydrophobic residues. This is due to a large number of nonpolar intermolecular interactions (see Fig. 8), for example by Ile 22 and Val 68 (both in cytochrome c 552 ) as well as Phe 88 and Leu 155 (both in the Cu A domain), whereas only two charged residues (Lys 115 in cytochrome c 552 and Arg 146 in the Cu A domain) are found within the protein-protein contact zone, in comparison to an inner ring of four positively charged lysine residues encircling the heme cleft in P. denitrificans cytochrome c 552 (19).
An additional consideration regarding the quality of the complex structures involves the Cu A domain that was used in the chemical shift perturbation experiments. This Cu A domain represents merely the solvent-exposed part of the entire ba 3 oxidase. It is reasonable to assume, however, that the complete ba 3 oxidase forms the same type of complex with cytochrome c 552 like the free Cu A domain. To test whether this is true for complex structures 1-3, the Cu A domain coordinates were reattached to subunit I of the ba 3 oxidase (Fig. 9). Subsequent analysis for steric overlap with the corresponding cytochrome c 552 molecule displayed only few addi-   tional close contacts, mainly between the side chains of Gln 455 (subunit I) and Gln 119 (cytochrome c 552 ) in structures 1 and 2 and between Trp 559 (subunit I) and Gln 57 (cytochrome c 552 ) in structure 3. However, no backbone-to-backbone contacts occurred in any of these cases. The same type of complex as obtained in structures 1-3 can thus be formed in vivo by cytochrome c 552 and the complete ba 3 oxidase.
The Electron Pathway-The program GREENPATH (38) was used to compare, based on the pathway model, the ET efficiencies within the complex structures by calculating the electron-tunneling coupling factors. The Franck-Condon term (see "Experimental Procedures") contains the free energy dependence related to nuclear motion; this is difficult to quantify, since it requires the evaluation of reorganization energies, but should be approximately constant for different complex conformations of the same partners. The results obtained with this method have been quite successful in the prediction of ET properties of proteins (see Refs. 47 and 48 and references therein), although the semiempirical formulation behind it is a simplification of more complete descriptions (49). The shortest electron pathway proposed for the cytochrome c 552 ⅐Cu A domain complex was found in structure 1. The electron originates at the porphyrin system in cytochrome c 552 , formally traveling from the iron center along the heme NC, C4C, C3C, CAC, and CBC atoms, and crosses over to the Ala 87 backbone oxygen in the Cu A domain. The electron continues along the Ala 87 -Phe 88 peptide bond to the amide proton of Phe 88 , where another jump occurs to the imidazole ring of His 114 , a direct ligand of the Cu A center. Continuing along the His 114 ring atoms HE1, CE1, and ND1, the electron eventually reaches the copper atom CU2 (Fig. 10).
This pathway involves residues Ala 87 and Phe 88 of the Cu A domain, which both showed significant perturbations of the amide 1 H and 15 N chemical shifts upon titration with cytochrome c 552 . Moreover, the involvement of Phe 88 in the ET was confirmed by site-directed mutagenesis experiments: replacement of Phe 88 by a leucine residue did not completely abolish the ET rates in stopped-flow kinetics (see Fig. S3 in the supplemental data) but diminished its efficiency significantly to ϳ68% of the apparent bimolecular rate constant in the physiological direction compared with the wild-type (WT) protein (k forward (WT): 5.0 ϫ 10 6 M Ϫ1 s Ϫ1 , k forward (F88L): 3.4 ϫ 10 6 M Ϫ1 s Ϫ1 ). This situation is different to the P. denitrificans system, where the corresponding Trp 121 residue in the Cu A domain has a key role in the ET to cytochrome c 552 (24,50,51): substitution of Trp 121 , e.g. by glutamine, rendered the enzyme inactive. The fact that the F88L mutant of T. thermophilus still shows 68% ET activity therefore suggests that one or more alternative pathways may exist. The neighboring Phe 86 , corresponding to Tyr 122 in P. denitrificans, could be excluded as possible ET component for several reasons: first, the F86L mutant was fully functional like the wild-type protein. Second, the F86L/F88L double mutant showed the same reduction in the ET activity as the F88L mutation alone. And finally, the Phe 86 ring is too far off the line connecting the heme with the Cu A center to warrant an efficient ET.
As a consequence, the influence of Phe 88 on the ET can be narrowed down to two possible scenarios. Either the ET pathway proposed by the GREENPATH program is the only biologically relevant route the electron can take, in which case the effect of the F88L mutation on the ET activity must be due to the resulting decrease of the hydrophobic portion in the contact surface and/or changes in the reorganization energy; or as we assume more likely, the electron can principally take two alternative paths both involving position 88 of the Cu A domain (Fig. 10). The through-bond pathway of ϳ19 -20 Å length, as proposed by GREEN-PATH, has only short through-space jumps of 1.83 Å (between heme HBC and Ala 87 O) and 2.95 Å (between Phe 88 HN and His 114 HE1). In this case, the closest edge-to-edge distance between the conjugated donor and acceptor systems (i.e. the heme ring and the His 114 imidazole ring, respectively) is 10.9 Å. This distance can be bridged easily and efficiently by a tunneling electron, as the majority of known ET reac-  Table 2), showing the number of atom-to-atom contacts closer than 2.8 Å for each residue involved in the complex formation. This scheme was prepared with the program nmr2st (27).  Table 2). The heme c moiety (orange) of cytochrome c 552 (green ribbon) approaches the binuclear Cu A center (magenta) of subunit II (yellow) for subsequent electron transfer to occur. The electrons are then passed on to the cofactors heme b and heme a 3 (both in red) and Cu B (magenta) in subunit I (cyan; residues 496 -500 in the loop between transmembrane helices 12 and 13 are missing in the x-ray structure). Subunit IIa is colored in blue. This picture was created with the program GRASP (54). tions between natural redox centers occur over distances of 14 Å or less (52). These ET reactions are remarkably rapid and specific with favorable electron-tunneling coupling factors, since the coupling via covalent bonds and hydrogen bonds is much stronger than that across van der Waals gaps. According to Gray and Winkler (53), in a protein environment electrons will tunnel a distance of ϳ11 Å on the nanosecond to subnanosecond time scale; hence, it may be concluded that the ET is not the limiting factor in the turnover rate between cytochrome c 552 and the Cu A domain from T. thermophilus, which typically ranges between 100 and 250 s Ϫ1 (5).
Alternatively, the electron may also travel from the heme moiety either to the His 114 imidazole ring or directly to the Cu A center entirely by through-space jumps via the Phe 88 ring (see dashed arrows in Fig. 10), thereby bridging a total distance of around 13.5 Å. However, in this case the transfer rates should be rather low, as electrons tunnel "through space" from one center to another with a rate that decreases exponentially with distance. Although this edge-to-edge distance is still within the productivity limit for ET through space, it should be less efficient compared with the "through-bond" path outlined above, as indicated by the fact that the F88L mutation reduced the ET activity by not more than 32%. Hence, both pathways appear possible within biologically relevant time scales according to current ET theories (52,53). Substitution of the phenyl ring by an aliphatic side chain in the F88L mutant would therefore eliminate only one of the possible ET pathways. In fact, the through bond pathway along the backbone of Ala 87 and Phe 88 might even represent a rational solution from an evolutionary point of view, since in this case the ET will not be significantly affected by spontaneous point mutations that could otherwise possibly render the system inactive by eliminating an essential side chain.