Net Proton Uptake Is Preceded by Multiple Proton Transfer Steps upon Electron Injection into Cytochrome c Oxidase*

Background: The coupling mechanism of proton and electron transfer in the redox-linked proton pump cytochrome c oxidase (COX) is still not understood. Results: Both H+ uptake and release steps during single-electron injection into oxidized COX precede net H+ uptake. Conclusion: The first H+ uptake coincides with electron input into CuA at the opposite membrane side. Significance: This suggests efficient H+ uptake mechanisms, such as proton-collecting antennae. Cytochrome c oxidase (COX), the last enzyme of the respiratory chain of aerobic organisms, catalyzes the reduction of molecular oxygen to water. It is a redox-linked proton pump, whose mechanism of proton pumping has been controversially discussed, and the coupling of proton and electron transfer is still not understood. Here, we investigated the kinetics of proton transfer reactions following the injection of a single electron into the fully oxidized enzyme and its transfer to the hemes using time-resolved absorption spectroscopy and pH indicator dyes. By comparison of proton uptake and release kinetics observed for solubilized COX and COX-containing liposomes, we conclude that the 1-μs electron injection into CuA, close to the positive membrane side (P-side) of the enzyme, already results in proton uptake from both the P-side and the N (negative)-side (1.5 H+/COX and 1 H+/COX, respectively). The subsequent 10-μs transfer of the electron to heme a is accompanied by the release of 1 proton from the P-side to the aqueous bulk phase, leaving ∼0.5 H+/COX at this side to electrostatically compensate the charge of the electron. With ∼200 μs, all but 0.4 H+ at the N-side are released to the bulk phase, and the remaining proton is transferred toward the hemes to a so-called “pump site.” Thus, this proton may already be taken up by the enzyme as early as during the first electron transfer to CuA. These results support the idea of a proton-collecting antenna, switched on by electron injection.

Cytochrome c oxidase (COX), 2 the terminal enzyme of the respiratory chains of mitochondria and many aerobic prokaryotes, catalyzes electron transfer from cytochrome c to molecular oxygen, reducing the latter to water. Cytochrome c, which binds to COX on the positively charged side (P-side) of the membrane (the extramitochondrial or periplasmic side of bacteria), injects electrons into the bimetallic Cu A center, which in turn donates electrons (1 at the time) to the low-spin heme a (Fig. 1). From there, electrons are passed on to the high-spin heme a 3 -Cu B binuclear center, the binding site for oxygen. The protons required for water formation originate from the opposite, negatively charged side (N-side; the matrix side in the case of mitochondria or the cytoplasmic side in the case of bacteria) of the membrane. This redox reaction is coupled to translocation of additional protons across the membrane ("proton pumping") to further increase the electrochemical proton gradient, which is the driving force for the ATP synthesis by the ATPase. To reduce 1 molecule of oxygen, 4 electrons are taken up from cytochrome c. Extensive studies have been performed to elucidate the mechanisms by which the enzyme translocates protons and couples this process with the chemical reaction (1)(2)(3). It is generally accepted that there is an overall involvement of 8 protons during the catalytic cycle: 4 "substrate" protons to complete the reaction (water formation) and 4 to be translocated ("pumped") across the membrane.
Based on the crystal structure (4,5) and mutagenesis studies (6 -8), two proton pathways have been suggested for the bacterial COXs from Paracoccus denitrificans and Rhodobacter sphaeroides, leading from the N-side toward the heme-copper site: the K-pathway and the D-pathway. The K-pathway includes the conserved amino acid Lys 354 and may be involved in the delivery of the first 1 or 2 protons during the reduction of the oxidized enzyme. The D-pathway, including Asp 124 , is likely to be involved in the uptake of both "chemical" and pumped protons in the F 3 O state transition. It appears to be the only pathway required when the fully reduced COX reacts with molecular oxygen (9 -12).
The assignment of proton uptake and proton pumping to the individual steps of the catalytic cycle of COX is a matter of controversy (10,(13)(14)(15)(16)(17)(18)(19)(20)(21)(22). Injection of 1 electron into the fully oxidized O state leads to formation of the 1 electron-reduced E state. During this step of the catalytic cycle, proton uptake was proposed to take place from the N-side of the membrane via the K-pathway and to be linked to the reduction of heme a (23). This idea is based on the fact that the slower electrogenic phase ( ϳ 180 s) observed in voltage measurements showed a clear kinetic deuterium isotope effect, indicative of the transfer of a proton from the N-side toward the P-side. However, the existence of this protonic phase was questioned (24) and is still a matter of debate (22,25).
Resolving this debate is of crucial importance not only for proton uptake linkage in the O 3 E step but also for the determination of the role of proton transfer pathways through COX in the different reactions of the catalytic cycle. Our experiments are designed to directly determine the proton uptake and release in COX by nanosecond time-resolved absorption spectroscopy in combination with pH indicator dyes. To address the role of the D-and K-pathways, COX variants with mutations in the respective pathway were investigated.
Our results presented here clearly show that the first proton uptake from the aqueous environment already takes place in the 1-s time range and coincides with electron input into the Cu A center (26,27). At later times, no proton uptake was observed. Instead and most surprisingly, electron transfer to heme a was accompanied by proton release. These data challenge and extend the current models of coupling proton and electron transfer (5,17,(22)(23)(24)28). We present a model that contains important aspects of each of the previous models and is, in a sense, a unification of these hypotheses.
Sample Preparation-Enzyme preparation of COX and variants from P. denitrificans strain AO1 was performed as described (29).
Proteoliposomes were prepared by the cholate dialysis method as described (30) using asolectin, which was further purified as described (31), at a concentration of 40 mg/ml. Lipids were dried under vacuum and resuspended in 100 mM HEPES/KOH (pH 7.3), 10 mM KCl, and 2% (w/v) cholate. The suspension was stirred on ice under argon for 1-2 h and sonicated to clarity with a Branson sonifier. COX was added to a concentration of 4 M. Subsequently, asolectin vesicles were dialyzed against buffer without cholate and subsequent reduction of HEPES/KOH (10-kDa cutoff). In the last dialysis step, no buffer was present.
The respiratory control ratio was determined as the ratio of the rates of cytochrome c oxidation in the coupled and uncoupled states, respectively (32). Reduced cytochrome c was used at a concentration of 40 M in 10 mM HEPES/KOH, 50 mM KCl, and 50 mM sucrose (pH 7.3); the rate of its oxidation was measured by following the change in absorbance at 550 nm after the addition of 1.2 nM reconstituted COX. Uncoupling was achieved by the addition of 5 M valinomycin and 10 M carbonyl cyanide m-chlorophenylhydrazone. The turnover number of COX was determined to be ϳ500 electrons/s, and the respiratory control ratio was 8.5-9, i.e. the enzyme has sustained no damage during preparation. The COX concentration was determined from the reduced-minus-oxidized optical difference spectrum with ⑀ 605-630 nm ϭ 11.7 mM Ϫ1 cm Ϫ1 (33).
Flash Spectroscopy-Flash spectroscopy was performed with a homemade flash photolysis spectrometer (34). Prior to the experiments, samples of WT COX and functional variants were incubated overnight with potassium ferricyanide to ensure a fully oxidized enzyme. Ferricyanide was then quickly removed by gel filtration (GE Healthcare PD-10 columns). In general, preparation and measurements were carried out in the dark to prevent the enzyme from prereducing (24).
Samples containing 10 M COX in 0.05% LM, 25 M Ru 2 D as an electron donor, 10 mM aniline as a sacrificial donor for ruthenium, 1 mM 3CP to prevent acidification due to proton release from aniline (35), and 50 mM KCl (pH 7.5) were excited with 10-ns pulses of 10 -15 mJ of energy at 492 nm. Under these conditions, up to ϳ10% of the COX becomes photoreduced. Electron uptake was monitored at 605 nm, as a rise in absorbance at this wavelength indicates the reduction of heme a. To reduce the amount of scattered light from the exciting laser flash, a cutoff filter (OG515) was placed in front of the entrance slit of the monochromator in the monitoring path. The signalto-noise ratio obtained in a single-flash experiment was sufficient for data analysis.
Proton Uptake and Release Measurements-Proton concentration changes were recorded via the absorbance change in the soluble pH indicator dye phenol red (50 M) at 558 nm with and without 50 mM Tris-HCl. Typically 5-10 time traces from single-flash experiments were averaged for the protonation kinetics accompanying the O 3 E transition.
The pK a of phenol red observed in the presence of COX and 50 mM salt was 7.95 compared with the value of 7.8 in the absence of COX. This small change in pK a indicates that the phenol red molecules may interact, at least partially, with the COX/detergent micelle, leading to a shift in pK a due to the protein surface potential (36,37). The time constant of proton release from bacteriorhodopsin/LM micelles measured with phenol red ( ϳ 70 s) agrees with the time constant for proton release from bacteriorhodopsin (36) measured with a covalently bound pH indicator dye facing the detergent shell or residing at the cytoplasmic surface (opposite the proton release side), thus supporting our assumption. When phenol red interacts with the COX micelle, i.e. resides in the membrane-water interfacial layer, fast proton release and uptake events are detectable, in contrast to measurements with pH indicator dyes residing entirely in the aqueous bulk phase (36). Proton concentration changes in proteoliposomes were observed by the addition of 50 M phenol red in the medium outside of the liposomes (P-side of COX).
Proton uptake stoichiometry was calculated according to Equation 1, where ⌬A H ϩ signal denotes the absorbance change at 558 nm (phenol red), ⑀ red 605 nm is the extinction coefficient of the reduced heme a (21,600 M Ϫ1 cm Ϫ1 (38)), d is the diameter of the cuvette used, ⌬A/⌬c H ϩ is the proton calibration factor (determined as the absorbance change for a defined proton concentration change), and ⌬A 605 nm is the absorbance change resulting from the reduction of heme a.

RESULTS
Spectral Characterization of WT COX-WT COX shows clear spectral differences between the oxidized and reduced forms ( Fig. 2A). The shift of the strong absorption band from 425 to 440 nm is attributed to the reduced form of both heme groups (hemes a and a 3 ), whereas the two absorption bands at 598 nm (oxidized form) and 605 nm (reduced form) are derived from heme a to almost 90% (marked by black and red arrows in Fig. 2A).
To investigate proton uptake or release, we used the soluble pH indicator dye phenol red, which has a pH-dependent absorption band at 558 nm (marked by an arrowhead in Fig.  2A). At this wavelength, the absorption spectrum of COX displays only marginal changes during reduction or oxidation; thus, the detected pH-dependent absorption change is almost solely due to the pH indicator dye.
Time-resolved Measurements of Electron Transfer-To observe the kinetics of electrons transferred to heme a in a single-electron photochemical reduction of the enzyme from the light-reactive electron donor Ru 2 D, the latter was excited with a single laser flash, and the absorbance change at 605 nm was recorded (Fig. 2B, inset). The lower the ionic strength, the better the binding of the ruthenium complex to COX and the subsequent electron transfer (Fig. 2B). Optimal conditions for maximum electron transfer were found at pH 7.5. However, we observed aggregation below 50 mM salt. Thus, we set the salt concentration to 50 mM. A representative time trace is presented in Fig. 3A. For a better comparison with results in the literature, the trace is presented with a linear time scale, although our data were recorded with 50-ns time resolution and sampled on a logarithmic time scale with 100 data points per decade. The time-dependent absorbance changes at 605 nm were fitted with two time constants of 1 ϭ 1.5 Ϯ 0.1 s and 2 ϭ 13.2 Ϯ 0.7 s. The first rise time was assigned to the relaxation of Ru 2 D, which correlates with the kinetics of electron uptake by Cu A (26). The second time constant describes the kinetics of the electron transfer from Cu A to heme a. The average time constant of electron transfer under these optimal conditions is ϭ 13.7 Ϯ 2.4 s (mean value of five single-flash experiments at 492 nm excitation, 22°C, and 50 mM KCl (pH 7.5)). No further absorbance changes were observed, in agreement with the formation of the 1 electron-reduced E state.
Proton Uptake and Release Kinetics of WT COX-Proton uptake from the aqueous bulk phase was detected with the pH indicator dye phenol red in the O 3 E step of the WT COX catalytic cycle. In Fig. 3 (A and C), the kinetics of electron transfer (⌬A at 605 nm, "electron transfer signal") are compared with the kinetics of proton concentration changes as detected with the pH indicator dye (⌬⌬A at 558 nm, "proton signal"). Both measurements were performed under the same conditions (20°C, pH 7.5, and 10 M COX in 0.05% LM, 25 M Ru 2 D, 10 mM aniline, 1 mM 3CP, and 50 mM salt/buffer). The proton signal was calculated as the difference between the two phenol red time traces obtained with and without buffer (Fig. 3B). To discriminate between true proton uptake by COX and possible transient protonation changes caused by the electron donor system, we measured a control proton signal using a covalent ruthenium-cytochrome complex. Under our experimental conditions, no contribution of the electron donor system to the transient proton signal was observed (Fig. 3E). The proton signal (Fig. 3C) contains both proton uptake and release phases as seen by the positive and negative absorption changes. A fit of the proton signal (Fig. 3C) required three exponentials, marked by the respective arrows. The proton uptake time of 1 ϭ 1.2 Ϯ 0.1 s correlates with the first phase of the electron transfer signal ( 1 ϭ 1.5 Ϯ 0.1 s), i.e. with the time constant for the electron transfer to Cu A . Surprisingly, the proton uptake is followed by proton release. The decay of the proton signal contains two components, 2 ϭ 11.5 Ϯ 2.8 s and 3 ϭ 249 Ϯ 18 s. The 11.5-s proton release component correlates with the electron transfer from Cu A to heme a ( 2 ϭ 13.7 Ϯ 2.4 s). However, the sum of the amplitudes of the decay components (proton release) is smaller than the amplitude of the rise component (proton uptake), therefore resulting in a positive net amplitude, i.e. in a net proton uptake.
Proton Kinetics of WT COX Incorporated in Liposomes-To ascertain the sidedness of the above observed proton uptake and release steps, proton measurements were also performed using COX-containing liposomes. The overall shape of the time trace (Fig. 3D) resembles that of the solubilized enzyme (Fig.  3C), with fast proton uptake in the beginning ( 1 ϭ 0.9 Ϯ 0.1 s), followed by two consecutive proton release steps with 2 ϭ 6.7 Ϯ 1.4 s and 3 ϭ 192 Ϯ 28 s. In contrast to the kinetics of proton uptake and release observed in solubilized COX, no net proton uptake was detected in COX-containing liposomes with phenol red residing at the P-side of the enzyme (Fig. 3D). Thus, net proton uptake occurs from the N-side.
Stoichiometry of Proton Uptake by COX and Its Variants K345M and D124N-The analysis of the two proton signals measured for the solubilized enzyme and COX incorporated in liposomes (Fig. 3, C and D) in terms of the numbers of protons taken up per injected electron using Equation 1 clearly shows that, for each injected electron, more than 1 proton is taken up by the enzyme: ϳ1 proton from the N-side and, on average, ϳ1.5 H ϩ from the P-side (Table 1). Simultaneously with electron transfer from Cu A to heme a, ϳ1 proton is released from the P-side. The last proton release step with a time constant of ϳ200 s takes place on both sides of the enzyme. The number of remaining protons in the enzyme at the N-side is 0.4 H ϩ /COX on average. Time constants and stoichiometries are summarized in Table 1.
Measurements of COX variants blocked in either the K-pathway (K354M) or the D-pathway (D124N) showed that the blockage results in reduced initial proton uptake but does not abolish net proton uptake (Table 1). In both variants, ϳ1 proton is taken up in the first step, which kinetically correlates with the initial electron injection into Cu A as observed in the

DISCUSSION
The coupling of electron injection and proton uptake is still elusive. Although several studies investigating the proton uptake by COX have been published, the results are not consistent (10,(13)(14)(15)(16)(17)(18)(19)(20)(21)(22). In particular, for the O 3 E transition, where 1 electron is injected into the oxidized enzyme, the published results are controversial (22)(23)(24)(25). Electron transfer from Cu A to heme a followed by proton transfer through the K-pathway in the O 3 E transition with kinetics of ϳ200 s has been found in voltage measurements (23). In disagreement with that, a signal consistent with the absence of proton uptake was observed upon reduction of heme a, also using voltage measurements (24), although a more recent study indicated that proton uptake occurs with a time constant of 150 s (25). To explain the differences in the various studies, it was suggested (22) that the proton uptake observed previously (23) did not correlate to the O 3 E transition but rather to electron transfer from heme a to the heme a 3 -Cu B binuclear center due to a not fully oxidized sample or differences in preparation of the fully oxidized COX sample leading to partial transfer of the first electron from heme a to the binuclear center. In our experiments, care was taken to use a fully oxidized COX sample (see "Experimental Procedures"), and the kinetics of the electron transfer steps were monitored.
In this study, using time-resolved absorption spectroscopy and pH indicator dyes, we directly observed proton concentration changes equivalent to proton uptake and release by COX during the O 3 E step. A careful optimization of the parameters affecting the proton signal allowed us to gain fundamental insights into the proton transfer steps associated with the injection of 1 electron into the oxidized enzyme.
A striking feature is the observation that proton uptake by COX occurs already in the 1-s time range, when electron injection into Cu A occurs. This fast proton uptake is followed by a gradual proton release. As the amplitude of the proton release signal is smaller than the amplitude of the proton uptake component, a final net proton uptake from the N-side is determined for the O 3 E step and becomes apparent at times slower than 300 s (Fig. 3C).
The proton uptake stoichiometry was calculated according to Equation 1 and amounts to an initial value of 2.6 H ϩ /COX per electron input into the WT enzyme. Comparison of proton uptake by the solubilized enzyme and that by COX incorporated into liposomes unambiguously shows that protons are taken up from both sides of the enzyme, ϳ1.5 H ϩ from the P-side and 1 H ϩ from the N-side ( Fig. 3 and Table 1). This large number of protons taken up is clearly surprising. It is very unlikely, however, that the electron donor system is responsible for the observed proton excess. Rapid proton release by aniline upon re-reduction of Ru 2 D, an effect that even opposes the

TABLE 1 Time constants of proton uptake and release and proton stoichiometries
Time constants of proton uptake and release as well as of electron transfer in the O 3 E step and proton stoichiometries for the individual proton transfer reactions triggered by the injection of 1 electron into the oxidized enzyme are given below. observed decrease in proton concentration in the aqueous bulk phase as a result of proton uptake by COX, was eliminated by the addition of 3CP and further tested in our experiments by monitoring the pH indicator dye absorption changes upon photoreduction of cytochrome c by using a covalent ruthenium complex. However, even more surprising is the finding that electron transfer to Cu A already leads to proton uptake at the opposite membrane surface. One might have expected the proton uptake accompanying electron injection into Cu A on the P-side by titratable residues nearby to electrostatically balance the negative charge of the extra electron at Cu A , but obviously, the organization of the protein in the low membrane dielectric allows pK changes leading to proton uptake at the opposite side of the membrane via long-range interactions. These findings are in accordance with the electroneutrality principle (39), which states that each electron transfer into the hydrophobic interior of COX is charge-compensated by the uptake of a proton. In addition, long-range effects may result in protonation dynamics at the opposite membrane surface.
For proton uptake from the N-side, the idea of a protoncollecting antenna has been widely discussed (see Ref. 40 for a detailed review). This idea is supported by our observations: as initial excess proton uptake occurs not only from the P-side but also from the N-side and net proton uptake accompanying formation of the E state is abolished in neither the D124N nor K354M variant, some of the residues in the vicinity of the proton transfer pathways probably play a role as the primary proton acceptor(s). Our experimental result differs from an earlier theoretical prediction (41), where all protonation changes upon reduction of Cu A are accompanied by proton uptake mostly by residues located on the cytosolic side (N-side) of the membrane. On the basis of our data, we propose that important transmembrane charge compensation occurs already during electron transfer to the Cu A center and not primarily during reduction of heme a (Fig. 4). Reduction of Cu A may, via longrange interactions, influence and prepare proton transfer via the D-pathway.
From the primary proton acceptor sites at the N-side as well as from the P-side, partial proton release occurs within ϳ200 s, leaving ϳ0.4 H ϩ on the N-side of the enzyme (Figs. 3 and 4). This time constant correlates with the time constant observed for the deuterium-sensitive phase (ϳ180 s) in voltage measurements (23). Combining these results, the following picture emerges: at the same time as the last proton release step occurs from the enzyme, the remaining proton is transferred toward the hemes, observed as a shift of a positive charge from the N-side toward the membrane interior (Fig. 4). The number of 0.4 H ϩ /COX also agrees with the results for redox-linked proton uptake of ϳ0.2-0.4 H ϩ /heme a-Cu A pair in carbon monoxide-treated COX (42).
In summary, we have shown that multistep proton transfer reactions take place during the single-electron transfer in the O 3 E step of the catalytic cycle of COX. Excess proton uptake from both sides of the membrane is coupled to electron input into Cu A and precedes proton transfer from the N-side to the hemes. The former suggests the existence of efficient proton uptake mechanisms, such as proton-collecting antennae at the protein surface (40), which were also discussed for proton uptake by bacteriorhodopsin (43)(44)(45) and green fluorescent protein (46). The observed crosstalk of the two enzyme surfaces, probably mediated by long-range electrostatic interactions, and the consecutive protonation-deprotonation reactions may constitute a common mechanism linking the proton and electron transfer reactions in the different stages of the catalytic cycle of COX.