Nanosecond electron tunneling between the hemes in cytochrome bo₃

Abstract

Biological electron transfer (eT) between redox-active cofactors is thought to occur by quantum-mechanical tunneling. However, in many cases the observed rate is limited by other reactions coupled to eT, such as proton transfer, conformational changes, or catalytic chemistry at an active site. A prominent example of this phenomenon is the eT between the heme groups of mitochondrial cytochrome c oxidase, which has been reported to take place in several different time domains. The question of whether pure eT tunneling in the nanosecond regime between the heme groups can be observed has been the subject of some experimental controversy. Here, we report direct observations of eT between the heme groups of the quinol oxidase cytochrome bo₃ from Escherichia coli, where the reaction is initiated by photolysis of carbon monoxide from heme o₃. eT from CO-dissociated ferrous heme o₃ to the low-spin ferric heme b takes place at a rate of (1.2 ns)⁻¹ at 20°C as determined by optical spectroscopy. These results establish heme–heme electron tunneling in the bo₃ enzyme, a bacterial relative to the mitochondrial cytochrome c oxidase. The properties of eT between the closely lying heme groups in the heme–copper oxidases are discussed in terms of the reorganization energy for the process, and two methods for assessing the rate of electron tunneling are presented.

The membrane-bound heme–copper oxidases catalyze respiratory O₂ consumption in the mitochondria of eukaryotic cells as well as in aerobic bacteria. A structurally and functionally common denominator for all enzymes in this large family is a binuclear heme–copper site of O₂ binding and activation and another “electron-queuing” low-spin heme (1, 2) that lies edge-to-edge to the O₂-binding heme and almost at van der Waals distance from it (1, 2). It is known that the low-spin heme delivers electrons to the binuclear site in a controlled fashion that is coupled to proton translocation across the membrane (3). The rate of this electron transfer (eT) has been studied extensively, in particular by photodissociation of carbon monoxide (CO) from the oxygen-binding heme in the so-called mixed-valence form of the enzyme, where only the binuclear site metals are initially reduced (4–6). In such a system, the “reversed” eT from photodissociated heme to the low-spin heme can be conveniently monitored by time-resolved optical spectroscopy. This reaction was originally reported to consist of two phases (5, 6): a ≈3-μs phase of eT between the heme groups was followed by slower electron equilibration with an additional copper site (Cu_A) that lies at the membrane interface in some of the heme–copper oxidases but that is lacking in others (7, 8). More recently, it was shown that a third phase of interheme eT (50 μs to milliseconds) that is coupled to proton release from the binuclear center can be observed after CO photolysis at high pH (9–11). However, with the quinol-oxidizing cytochrome bo₃ from Escherichia coli, even more phases were discerned (12).

The 3-μs phase of interheme eT was long thought to be limited by pure electron tunneling, especially because this rate is independent of pH and substitution with heavy water (5, 9) and could therefore not be ascribed to be linked to proton transfer. This rate is, however, some 3 orders of magnitude slower than predicted by the empirical eT treatment advanced by Moser et al. (13, 14) on the basis of the short edge-to-edge distance between the heme groups, the driving force (<0.1 eV in the forward direction), and a proposed generic value for the reorganization energy, λ (0.7 eV). However, applications of electron pathway analysis (15) gave results consistent with a rate of ∼(3 μs)⁻¹ (16, 17), and thus a major discrepancy was evident. In 1993, Morgan et al. (12) reported on heme–heme eT in the quinol oxidase cytochrome bo₃ from E. coli, another member of the heme–copper oxidase family. Several phases were observed in addition to the 3-μs event described earlier for the aa₃-type cytochrome c oxidases. One of them was faster than 3 μs, kinetically unresolved, but proposed because the difference between the optical spectra of the mixed-valence and fully reduced enzyme indicated eT from the oxygen-binding heme to the low-spin heme in the former case. Together with the prediction of the Moser–Dutton treatment, this finding prompted further studies, and in 2001 a submicrosecond, but still kinetically unresolved heme–heme eT phase was reported in cytochrome c oxidase from bovine heart (18). Although Namslauer et al. (19) could not confirm such fast eT in cytochrome oxidase from the same source, Pilet et al. (20) subsequently showed heme–heme eT with a rate of ≈8 × 10⁸ s⁻¹ in the mitochondrial enzyme. In agreement with the latter study, a computational pathway analysis by Tan et al. (21) yielded a rate of ≈3 × 10⁸ s⁻¹ after taking into account the molecular dynamics of the system.

Here, we have returned to the quinol oxidase cytochrome bo₃ with adequate time-resolved methodology to test whether nanosecond eT between the heme groups may indeed be a general property of the heme–copper oxidases and thus present also in the bacterial enzymes. Despite the strong structural similarity, data have been lacking on whether the fast kinetic component seen in the mitochondrial enzyme also occurs in the bacterial system (22). Furthermore, the hemes of cytochrome bo₃ are different from those of cytochrome aa₃ both chemically and spectroscopically, which makes the bo₃ enzyme an excellent candidate to test the universality of this eT mechanism. We indeed find interheme eT in this enzyme with a rate of ≈8 × 10⁸ s⁻¹ and discuss this observation in the light of the empirical Moser–Dutton treatment (13) and the report that the reorganization energy for this reaction is <0.2 eV (23).

Results and Discussion

Morgan et al. (12) studied electron redistribution after photolysis of CO in the mixed-valence state of cytochrome bo₃ and discovered that the electron is not redistributed in a single process but rather in a number of waves. Processes with characteristic times of 3 μs, 100 μs, and 10 ms were observed, and all of these processes had the unique spectral signature of eT between the high- and low-spin hemes. A faster eT phase was also found in this investigation but remained unresolved at the time resolution available (see above). In both the fully reduced and mixed-valence states of the enzyme, at delay times up to ≈30 ps after CO photolysis, we observed spectral phases that can be attributed to heme photophysics, as described earlier for other heme proteins, including oxidases (24, 25). In the fully reduced CO state, the absorbance after this delay did not change up to 4 ns. By contrast, in the mixed-valence CO state, we found a significant difference between the spectra at 50 ps and 4 ns (Fig. 1). The kinetics of this difference in the Soret region can be fitted with a single exponential with a time constant of 1.21 ± 0.09 ns. The time constants measured in the α-band region are within experimental error of this value (Fig. 2).

Fig. 1.

Spectral differences upon photolysis of CO from mixed-valence CO cytochrome bo₃. Solid and dashed lines show the absorbance changes at, respectively, τ = 50 ps and τ = 4 ns after photolysis, in the Soret (Left) and the α-band (Right) regions of the spectrum.

Fig. 2.

Time course of absorbance changes near the isosbestic points for CO dissociation in cytochrome bo₃. Experimental data at 423 nm (Left) and 563 nm (Right) after CO photolysis in the mixed-valence (circles) and fully reduced (triangles) states are shown. The fits (solid lines) yield time constants of 1.21 ± 0.09 ns (423 nm) and 1.6 ± 0.4 ns (563 nm).

The spectrum associated with the 1.2-ns phase was obtained from a global fit and is shown in Fig. 3 (solid lines). It consists of characteristic minima ≈420, 440, 550, and 585 nm and marked maxima at 430 and 568 nm. The spectrum in the α-band resembles that predicted for submicrosecond eT (12). The full spectra (Fig. 3) also compare well with both the calculated (b²⁺o₃³⁺ minus b³⁺o₃²⁺) difference spectrum, and the spectra reported for the 3-μs phase (12). By analogy to the case in mitochondrial aa₃ oxidase (23), the 3-μs phase reflects the change in redox equilibrium between hemes o₃ and b upon CO release by Cu_B. The offset in the calculated steady-state difference spectrum <410 nm may be caused by a baseline drift. The spectrum of the fast eT appears somewhat shifted in the α-band. This shift can be ascribed to a photoselection effect caused by the necessary use of perpendicularly polarized pulse pairs in this region (see Materials and Methods). Such effects can be expected for two-heme systems (26). Taking the Soret band normalization of the calculated steady-state spectra and the 1.2-ns phase as a measure of the extent of eT, and the initial photolysis spectrum with respect to the steady-state unliganded minus CO-liganded spectrum as a measure of CO photolysis, we estimate that the 1.2-ns phase corresponds to ≈6% oxidation of CO-dissociated heme o₃. Thus, on the nanosecond time scale, the free-energy difference between the states b²⁺o₃³⁺ and b³⁺o₃²⁺ amounts to ≈70 meV. This difference becomes smaller during further phases of eT as the reaction proceeds (12).

Fig. 3.

Spectrum of the nanosecond phase. Solid line, the nanosecond phase; dashed line, calculated steady-state heme b²⁺o₃³⁺ minus heme b³⁺o₃²⁺ spectrum (from ref. 12); triangles, experimental points of the microsecond phase (from ref. 12). Data in the Soret and α-band regions are normalized at 418 nm and 562 nm, respectively. The difference between the spectrum of the nanosecond phase and the calculated spectrum in the α-band region can be ascribed to polarization effects (see Results and Discussion).

CO photolysis from mixed-valence cytochrome bo₃ is associated with several phases attributable to eT from heme o₃ to heme b, but here we focus only on the fastest phase that was kinetically unresolved. The other phases reflect the multiphasic protein relaxation after CO photolysis in this enzyme (27), and the 3-μs phase can be attributed to CO dissociation from the Cu_B center (18, 23), which lies next to the oxygen-binding heme and is the first site of CO binding after photolysis (28, 29). Therefore, only the nanosecond phase can be ascribed to true electron tunneling.

The rate observed here for eT from heme o₃ to heme b (8.3 × 10⁸ s⁻¹) is the sum of the forward and backward rate constants. The ≈6% amplitude of the reaction yields the equilibrium constant, thus it follows that the forward rate constant of (exergonic) eT from heme b to heme o₃ is ≈8 × 10⁸ s⁻¹, which is remarkably similar to the rate observed in the mitochondrial enzyme (20) and consistent with the conservation of cofactor disposition and environment among the heme–copper oxidases despite the differences in heme structure. Our present results thus indicate that the mechanism of nanosecond interheme eT is universal in the family of heme–copper oxidases.

The empirical Moser–Dutton ruler is a widely used framework to describe eT reactions in proteins (13). This treatment is based on theoretical (30) and experimental (31–34) work that elucidated the free energy, temperature, and distance dependences of eT rates, with the nuclear term taken from classical Marcus theory (35). A correction term assumes nuclear tunneling with a single generic characteristic frequency to account for quantum effects (30, 36), which is necessary because the high-temperature regime does not apply. The protein between the electron donor and acceptor is treated as bulk medium. The optimal rate is predicted from the closest distance r between the macrocycles of the donor and acceptor, taking, in the simplest form (13), a generic average bulk coefficient β, describing the decrease in the electronic coupling with distance, proportional to exp(−βr). In a refined version, β is modulated according to the “packing density” ρ, an average measure of the volume in the interreactant space occupied by protein atoms (36). The actual rate at a given temperature is calculated by taking into account the activation barrier calculated from the driving force ΔG and the reorganization energy λ, for which a generic value of 0.7 eV was advocated (14). Another much used approach to describe biological eT is the pathway analysis method advanced by Beratan et al. (15, 37, 38), where discrete tunneling pathways between donor and acceptor are identified in the three-dimensional structure. The two approaches have been compared (39, 40) and found to yield similar results in most but not all cases. It is clear that the few cases where there is disagreement are the most interesting ones with respect to understanding the mechanism. We believe that the case presented here is one of them (see below).

Using the Moser–Dutton treatment, with the observed driving force, the generic reorganization energy (λ) of 0.7 eV, and the shortest edge-to-edge distance of 7.8 Å between the two heme macrocycles in cytochrome bo₃ (41), k_eT is predicted to be ≈5 × 10⁸ s⁻¹. This result is in excellent agreement with the observed rate of 8 × 10⁸ s⁻¹. In the refined version, a ρ of 0.82 was suggested for cytochrome c oxidase [instead of the generic value of 0.76 (14, 42)]. With the observed rate constant, the value for λ would then become 0.72 eV, again in good agreement with the Moser–Dutton treatment. However, this analysis is at variance with the conclusion by Jasaitis et al. (23) that λ for the interheme eT is <0.2 eV, based on a quantum mechanical analysis of the temperature dependence of the eT rate and the equilibrium constant. Recently, Moser et al. (42) argued that a value of λ = 0.7 eV could be maintained for this reaction on the basis of the temperature dependence of the rate, but this reasoning ignored the experimental temperature dependence of ΔG and required a value for the frequency of the relevant protein modes two times higher than the advocated generic value. Moreover, indications for reorganization energies well <0.7 eV have been reported for a variety of low-driving force intraprotein reactions (43–46).

The question arises, therefore, as to how the observation of λ <0.2 eV (23) can be accommodated to the current data and whether the Moser–Dutton treatment can be used in this case. If we set a limit value of λ = 0.2 eV into the Moser–Dutton formulation, with the eT rate constant and the heme–heme distance at known values, we arrive at ρ = 0.36, much lower than the values used by Moser et al. (42). Considering the known crystal structures of the heme–copper oxidases (1, 2, 41), one can immediately see that there is little “protein” between the adjacent heme edges where the eT most likely takes place. The computational work by Tan et al. (21) indeed suggested that the major route, or pathway, of interheme eT is “through space” between the methyl substituents at the closest approach between the two heme edges. These observations appear to contradict the proposed high ρ value (42). According to the Moser–Dutton treatment (36), ρ is calculated by averaging protein density encountered along all lines connecting all atoms on the donor macrocycle and all atoms on the acceptor macrocycle. For the case of the two close-lying hemes in the heme–copper oxidases this procedure includes a large volume relatively far away from the closest contact points that may be irrelevant for actual eT. Extracting a more relevant, unambiguous ρ for the interheme space is not straightforward. Among other parameters, it depends on which, if any, of the heme macrocycle substituents are considered to be part of the donor and acceptor. Using different assignments and restricting the analysis volume to a box located between the near-parallel heme edges, we find a value for ρ of ≈0.39–0.56 for the bo₃ enzyme. Even the high-end density is much lower than the values used previously (42), which cannot be ascribed to our slightly different methodology for determining ρ (see Materials and Methods).

Further, ρ exhibits large local variations. On longer-length scales, the local variations usually cancel, enabling the use of the average ρ as a parameter in the Moser–Dutton treatment. In systems where the donor–acceptor distance is short, the variations in local ρ become more important (cf. 14), possibly breaking down the justification of averaging. In bo₃, the edge-to-edge distance between the near-parallel heme edges varies from 7.8 to 9.8 Å. Simply taking an average ρ, the 7.8-Å distance would thus seem to be the dominating and defining parameter for eT efficiency. However, studying the local structure more closely reveals that ρ is significantly higher between the longer approach of the heme edges. Here, because of the intervening isoleucine (Ile-424), the local ρ is between 0.69 and 0.75, whereas it is only 0.23–0.51 at the shortest approach between the hemes, where methyl substituents from each heme are at a 5.7-Å C–C distance from one another. The approximations used have to be kept in mind, however. The ρ, as defined by standard van der Waals radii, with all atoms mediating electron transfer equally, is a simplification of the true situation. The choice of which macrocycle substituents contribute to an “eT-active density” is also important.

Conclusions

The problem of a straightforward application of the Moser–Dutton ruler at short distances becomes apparent from our analysis. We suggest that the Moser–Dutton treatment with generic values for λ and ρ is well applicable for eT at relatively long distances, where there are multiple possible eT pathways. However, when the donor–acceptor distance becomes short, as is the case for the heme–heme eT in the heme–copper oxidases, this general treatment may fail because the number of effective eT pathways and ρ are drastically reduced (cf. 39, 40). With less packing one also expects the reorganization energy to decrease, as reported (23). Therefore, it appears that the agreement between the observed tunneling rate and the rate predicted by application of the Moser–Dutton treatment in its general form may be fortuitous for heme–heme electron tunneling in the heme–copper oxidases insofar as both λ and ρ may simultaneously take values much lower than those applied in the treatment. We note that the heme–heme eT rate computed by Tan et al. (21) using pathway analysis combined with molecular dynamics was the optimum rate at −ΔG = λ. The fact that this rate closely corresponds to the experimentally observed rate at −ΔG ≈70 mV reported here also supports the low reorganization energy for this reaction deduced earlier (23). We conclude that the heme–heme electron transfer in cytochrome bo₃ is essentially activationless and that a true atomistic approach is required to understand the mechanism of short-distance eT processes.

Materials and Methods

Sample Preparation.

The enzyme was purified as a non-His-tagged version, as described in ref. 47. The sample was thoroughly degassed in a gas-tight vessel and transferred to a degassed gas-tight optical cell (117.007 QS, optical pathlength, 1 mm; Hellma) sealed with a rubber septum. The mixed-valence carboxylated state of cytochrome bo₃ was obtained by overnight incubation of the protein <1 atm CO, at 4°C. For the preparation of the fully reduced carboxylated enzyme, the sample was also thoroughly degassed, 20 mM sodium dithionite was added as a reductant, and the sample was exposed to 1 atm CO. The redox and ligation state of the proteins was followed by means of absorption spectroscopy with a Shimadzu 1601 UV-visual spectrophotometer. The sample concentrations were 25 μM and 45 μM for experiments in the Soret and α-band spectral regions, respectively.

Experimental Procedures.

Multicolor transient absorption pump probe spectroscopy (48) was performed with a 55-fs pump pulse centered at 590 nm and a <30-fs white light continuum probe pulse, at a repetition rate of 30 Hz as described in ref. 20. Briefly, the continuum was generated by using the fundamental beam of the laser system, which was centered at ≈615 nm. For this reason, the required low noise could not be obtained ≈615 nm, and the α-band data could be used only in the spectral region <606 nm. The polarization of the pump pulse could be rotated with respect to that of the continuum probe pulse with a half-wavelength plate. The measurements in the Soret region were performed under magic-angle configuration to avoid polarization photoselection effects. In the α-band experiments, noise caused by scattering of the pump pulse was minimized by setting the polarization of the pump beam perpendicular to that of the probe beam and eliminating the scattered light with a polarizer. The system was equipped with a delay line allowing temporal delays up to 4 ns (corresponding to an optical pathlength difference of 1.2 m). The alignment of the delay line was verified with CO myoglobin. Full spectra of the probe beam were recorded by using a combination of a polychromator and a CCD camera. All experiments were performed at 20°C.

Calculation of Packing Density.

The packing density between the hemes was calculated as the ratio of the space occupied by the united-atom van der Waals radii (49) of the intervening atoms to the whole volume. The space was defined as an asymmetric box with a thickness of 2 Å following the proximal edges of the two hemes. The choice of thickness admittedly includes some ambiguity, but tests with thinner and thicker boxes showed that the qualitative conclusions about the packing densities remain the same. For the even more local densities at the short, medium, and long distances between the two heme edges, this box was further divided into three regions, corresponding to local channels between the two macrocycle rings facing each other.

Data Analysis.

Basic data matrix manipulations and presentation were done with Matlab (Mathworks). The absorbance changes from the instrument were data arrays of different size. To find common processes in the whole data array, data were treated by using the SPLMOD algorithm (50), the Matlab interface for which was developed in the Helsinki Bioenergetics Group, as described in ref. 47.