Kinetic Resolution of a Tryptophan-radical Intermediate in the Reaction Cycle of Paracoccus denitrificans Cytochrome c Oxidase*

The catalytic mechanism, electron transfer coupled to proton pumping, of heme-copper oxidases is not yet fully understood. Microsecond freeze-hyperquenching single turnover experiments were carried out with fully reduced cytochrome aa3 reacting with O2 between 83 μs and 6 ms. Trapped intermediates were analyzed by low temperature UV-visible, X-band, and Q-band EPR spectroscopy, enabling determination of the oxidation-reduction kinetics of CuA, heme a, heme a3, and of a recently detected tryptophan radical (Wiertz, F. G. M., Richter, O. M. H., Cherepanov, A. V., MacMillan, F., Ludwig, B., and de Vries, S. (2004) FEBS Lett. 575, 127-130). CuB and heme a3 were EPR silent during all stages of the reaction. CuA and heme a are in electronic equilibrium acting as a redox pair. The reduction potential of CuA is 4.5 mV lower than that of heme a. Both redox groups are oxidized in two phases with apparent half-lives of 57 μs and 1.2 ms together donating a single electron to the binuclear center in each phase. The formation of the heme a3 oxoferryl species PR (maxima at 430 nm and 606 nm) was completed in ∼130 μs, similar to the first oxidation phase of CuA and heme a. The intermediate F (absorbance maximum at 571 nm) is formed from PR and decays to a hitherto undetected intermediate named FW*. FW* harbors a tryptophan radical, identified by Q-band EPR spectroscopy as the tryptophan neutral radical of the strictly conserved Trp-272 (Trp-272*). The Trp-272* populates to 4-5% due to its relatively low rate of formation (t½ = 1.2 ms) and rapid rate of breakdown (t½ = 60 μs), which represents electron transfer from CuA/heme a to Trp-272*. The formation of the Trp-272* constitutes the major rate-determining step of the catalytic cycle. Our findings show that Trp-272 is a redox-active residue and is in this respect on an equal par to the metallocenters of the cytochrome c oxidase. Trp-272 is the direct reductant either to the heme a3 oxoferryl species or to Cu 2+B. The potential role of Trp-272 in proton pumping is discussed.

The superfamily of heme-copper oxidases comprises the cytochrome oxidases, which catalyze the reduction of molecular oxygen to water and the NO reductases that catalyze the reduction of NO to N 2 O (1-6). Cytochrome oxidases (CcOs) 2 are the final electron acceptors in the respiratory chains of bacteria, archaea, and mitochondria. Cytochrome aa 3 from Paracoccus denitrificans, is a Type A oxidase based on the structure of its D-and K-proton pathways (7,8). The reduction of oxygen (Reaction 1) generates a proton electrochemical gradient across the cytoplasmic membrane. Four protons are used for the formation of water, and four are pumped across the membrane according to, where H C ϩ are protons taken up from the cytoplasm and H P ϩ protons are those ejected to the periplasm (9 -13).
The crystal structures of cytochrome aa 3 from bovine heart mitochondria, P. denitrificans, and Rhodobacter sphaeroides have been solved previously (8, 14 -18). P. denitrificans cytochrome aa 3 is a four-subunit membrane complex. Subunit one harbors heme a and the heme a 3 -Cu B binuclear reaction center where reduction of oxygen takes place. Subunit two contains the docking site for cytochrome c (19,20) and the Cu A mixedvalence binuclear center with two copper atoms separated by 2.5 Å (21). Electrons from cytochrome c enter CcO at the Cu A site and are further transferred via heme a to heme a 3 and Cu B . Protons from the cytoplasm enter the enzyme via the D-or K-proton pathways (10 -13, 22, 23). These pathways connect the aqueous cytoplasmic phase with the conserved Glu-278 3 in the interior of the enzyme (D-pathway) or with the binuclear center (K-pathway). The proton exit route to the periplasm is less well defined. Water is expelled to the periplasm via the Mg 2ϩ or Mn 2ϩ bound at the interface of subunits I and II (24 -27).
The oxygen-reduction cycle of CcO has been studied by a great variety of kinetic techniques such as the flow-flash method monitored by UV-visible spectroscopy (10 -13, 23, 28 -31) or resonance Raman scattering (32)(33)(34)(35)(36)(37)(38). Collectively these studies have led to a general understanding of the catalytic mechanism in terms of oxygen chemistry, electron transfer, and proton translocation. The catalytic cycle can be initiated from the fully reduced enzyme (R) or from the mixed-valence form (MV) in which only heme a 3 and Cu B are reduced or by single electron injection (11,39).
When the reaction is started with the fully reduced CcO, the enzyme cycles through a series of intermediates designated as R 3 A 3 P M 3 P R 3 F 3 O H (see Fig. 7 below). The first detectable intermediate after mixing R with O 2 , A, is formed within ϳ10 s. A is the oxy-ferrous complex (Fe 2ϩ -O 2 ) of heme a 3 (29 -31, 34 -38). Subsequently, the OϭO bond is broken yielding P M . P M accumulates (ϳ150 s) when the reaction is started with MV. However, when the reaction is initiated from R, P R accumulates (completed in ϳ100 s). Direct evidence for OϭO bond splitting in both P M and P R was provided by resonance Raman spectroscopy, which identified the specific vibrations of the oxoferryl (Fe 4ϩ ϭO) state of heme a 3 (34 -38).
In P M and P R , the Cu B 1ϩ has been oxidized to Cu B 2ϩ -OH Ϫ . The oxygen atom in Cu B 2ϩ -OH Ϫ is derived from molecular oxygen (40). The breaking of the OϭO bond requires donation of four electrons and a proton. Because this also occurs when the reaction is started with MV, whereas heme a 3 and Cu B provide only three electrons, an amino acid, Tyr-280 (P. denitrificans numbering), was proposed to act as the donor of the fourth electron (plus the proton) (33). Tyr-280 is covalently linked to His-276, a ligand to Cu B , and sufficiently close to the binuclear center to act as a rapid reductant (8, 14, 16 -18). Indeed, some Tyr-280 had been converted to the radical Tyr-280* in P M (41). In P R , in contrast to P M , the Tyr-280* is absent (42) presumably because it has been reduced by heme a to the anion (Tyr-280 Ϫ ) or to the protonated Tyr-280. Even though heme a oxidation in the sequence R 3 A 3 P M 3 P R is completed within ϳ100 s, the true rate of electron transfer from heme a to heme a 3 (and to Tyr-280*?) is ϳ1 ns (43), which would explain why P M (and Tyr-280*) does (do) not accumulate to a measurable extent when the reaction is started from R. Instead of Tyr-280, the conserved Trp-272 has been proposed recently as the amino acid residue involved in OϭO bond breaking (44).
The P R 3 F transition is not associated with electron transfer but with proton binding and proton translocation (9 -13). The slowest step of the oxidative part of the catalytic cycle (1-1.5 ms) is the F 3 O H transition, which is linked to proton translocation as well (9 -13). In this step an electron is transferred from Cu A /heme a to Fe 4ϩ ϭO yielding Fe 3ϩ -OH Ϫ (38). At this stage, the enzyme is completely oxidized, but in a metastable "high energy state," O H . Reduction of O H , but not of the resting enzyme (O), leads to two successive proton-pumping events in the reaction sequence O H 3 E 3 MV (11,45).
To resolve enzyme catalytic mechanisms on the microsecond time scale, we have developed a microsecond freeze-hyperquenching mixing/sampling device (MHQ) (5, 46 -48). MHQ is an extension of the rapid-freeze quench technique (49) in which the instrument dead time has been reduced from 5-7 ms to 60 -80 s (46,47). The great advantage of MHQ and rapidfreeze quench is that the resulting frozen powder containing trapped intermediates can be analyzed by a variety of spectro-scopic techniques, including EPR spectroscopy, an invaluable tool in the study of metallo-redox enzymes. MHQ experiments with cytochrome bo 3 from Escherichia coli and cytochrome aa 3 from P. denitrificans showed the formation of a tryptophan radical (Trp*) after ϳ200 s, which was weakly magnetically coupled to the Fe 4ϩ ϭO state of the heme a 3 (46). The EPR properties of this transient radical differ from the Tyr-167* and the proposed porphyrin cation and/or tryptophan radicals obtained by incubation with H 2 O 2 (50 -52).
In the work presented in this report we used MHQ to determine the kinetics of the Trp* found previously and assign its role in the catalytic cycle. Q-band EPR spectroscopy identifies the radical as the neutral radical form of Trp-272. The kinetics of Trp-272*, Cu A , heme a, and heme a 3 were determined and simulated with a single set of rate constants in a model including the new intermediate F W *, which harbors the Trp-272*. The Trp-272* is formed in the second part of the catalytic cycle and Trp-272 is proposed as the electron donor to heme a 3 or Cu B . An additional role for Trp-272 in proton pumping is discussed.

EXPERIMENTAL PROCEDURES
Enzyme Purification-P. denitrificans cytochrome aa 3 was purified as previously described (53,54) and contained 0.2 mol Mn 2ϩ per mole of enzyme as determined by EPR spectroscopy. Mn 2ϩ -depleted enzyme was obtained by decreasing the Mn 2ϩ concentration to 0.5 M in the growth medium (55,56).
Microsecond Freeze-hyperquenching and Sample Handling-The MHQ setup (dead time of 60 -80 s and effective reaction temperature of 10 Ϯ 1°C.) and the sample handling procedures are identical to those described before (5,46) except that the aluminum rotating cold plate was coated with a layer of ϳ5 m of 99.999ϩ % tungsten. Tungsten was applied in 50 cycles of physical vapor deposition. The tungsten coating of the cold plate resulted in lower impurities (e.g. Fe 3ϩ trapped in corundum) in the samples analyzed by EPR spectroscopy. For each time point 0.25 ml of pulsed cytochrome aa 3 (100 -280 M) was used (5,46). The reduced enzyme (incubated with 10 mM ascorbate and 1 M phenazine ethosulfate) was mixed 1:1 with an O 2 -saturated (1.3 mM at 20°C) buffer (50 mM HEPES, pH 7.2, 0.1% laurylmaltoside). Sample preparation for the low temperature UV-visible spectroscopy and normalization and analysis of the spectra are described in (46). Sample packing is described previously (47).
UV-visible Spectroscopy-UV-visible spectra were recorded with an Olis upgraded Aminco DW2000 scanning spectrophotometer equipped with a custom-made liquid N 2 -flow system, to maintain the temperature during the measurement stable at 90 K (48). The spectrophotometer was calibrated with a holmium oxide filter to an accuracy of 0.2 nm.
Data Analysis-Data were processed and analyzed with the IGOR Pro software package (Wavemetrics). Q-band EPR spectra were simulated using a home-written simulation program in Pascal for the Macintosh computer. The program allows for non-linear g-and A-tensors. The kinetic data were fitted to a model of six consecutive irreversible reactions. The analytical solution of this set of homogeneous first-order differential equations has been added to the supplemental information.
EPR Spectroscopy-X-band EPR spectroscopy was performed on a Bruker ER200D spectrometer, Q-band EPR spectra were recorded on a Varian E9 spectrometer. Both spectrometers were equipped with a home-built helium-flow system (57). EPR signals were quantitated with respect to a 10 mM CuClO 4 standard. Differences in sample packing were corrected using the Mn 2ϩ signal as internal standard in case Mn 2ϩ -containing CcO was used (see supplemental Fig. S2). For Mn 2ϩ -free CcO no correction for sample packing was applied, only for differences in starting concentrations of the CcO. The overall dilution of the enzyme as present in the EPR tube is 6-to 10-fold with respect to the starting concentration. This dilution arises from the 1:1 mixing, condensation of water vapor during sample handling, and the loose sample packing owing to the fine nature of the frozen powder. The uncertainty in sample packing was determined at 1 Ϯ 0.20 (n ϭ 25) and is indicated with error bars in the figures. However, the relative concentrations of Cu A , heme a and the Trp* are accurate to 1 Ϯ 0.05, because they are determined in the same sample. The g values of the Trp* at Q-band frequency were determined with 2,2-diphenyl-1-picrylhydrazyl and the Mn 2ϩ signal as an internal standard and are accurate to Ϯ0.0002.

RESULTS
Low Temperature UV-visible Spectroscopy of MHQ Samples-MHQ samples were prepared by reacting fully reduced cytochrome aa 3 with O 2 for various times between 83 s and 6 ms. Fig. 1 shows the low temperature absolute and difference UVvisible spectra of a selection of these samples. In the first 130 s of the reaction the absorbance of the Soret band at 444 nm decreased, and a new band appears at 430 nm (Fig. 1A). The difference spectra (Fig. 1B) indicate a loss of intensity of ϳ50% at 444 nm. Concomitantly, the ␣-band shifted from 603 to 606 nm ( Fig. 1A) or, in the difference spectra (Fig. 1B), to 608 nm while losing some intensity. The peaks at 430 and 606 nm in the absolute spectra are characteristic for the oxoferryl state (Fe 4ϩ ϭO) of heme a 3 , and the P-state, specifically P R (32)(33)(34)(35)(36)(37)(38). The spectra obtained after 220 and 355 s indicated a small blue shift of the ␣-band from 608 to 603-604 nm without a change in the Soret region (Fig. 1B). This blue shift is consistent with the P R 3 F transition (10 -13, 23, 28 -31, 33, 35, 38). More direct evidence for the formation of F came from control experiments using H 2 O 2 (data not shown) to generate F at pH 6 and 7.2, the latter being the pH in our experiments. These experiments were performed exactly as in a previous study (44). Samples were first monitored at room temperature to check for formation of F and subsequently frozen for analysis at 90 K by UV-visible spectroscopy. The low temperature UV-visible difference spectra (H 2 O 2 minus oxidized enzyme) indicated the formation of a broad band with a maximum at 571 Ϯ 1 nm instead of 580 nm at room temperature (42,44,58,59). The amplitude of the 571 nm band amounted to 7 and 11% (at pH 7.2 and 6, respectively) of the ␣-band intensity of the absolute spectrum of oxidized enzyme (compare with Ref. 59). Fig. 1B shows a shift in the ␤-band position from 565 nm at t ϭ 0 to 571 nm indicating formation of F. The 571 nm maximum persisted up to 3 ms, which suggests rapid formation and relatively slow breakdown (1-1.5 ms) of F. The kinetics of formation and FIGURE 1. Low temperature absolute (A) and difference (B) visible spectra of cytochrome aa 3 from P. denitrificans recorded at 90 K. Fully reduced pulsed cytochrome aa 3 was rapidly mixed with an O 2 -saturated buffer and reacted for various times (indicated in microseconds). "Red" refers to fully reduced cytochrome aa 3 with maxima at 444 nm in the Soret region and at 603 nm in the ␣-band region. A, the Soret maximum of the oxoferryl form of heme a 3 is at 430 nm. Formation of P R (606 nm) after 130 s is indicated by an arrow. The peak shifts to 603/604 nm after 220 s and longer times. After 6 ms the enzyme is oxidized (O H ); the Soret maximum is 427 nm. "Ox" refers to the as-isolated oxidized form of cytochrome aa 3 displaying maxima at 424 nm and 600 nm in the Soret and ␣-band regions, respectively. B, the spectrum obtained after 6 ms was subtracted from those in A. The Soret maximum of reduced hemes (a ϩ a 3 ) is 444 nm. Formation of P R (608 nm) after 130 s is indicated by an arrow. The peak shifts to 603/604 nm after 220 s and longer times. Formation of F is indicated by the shift of the ␤-band from 565 nm to 571 nm (0 -130 s). Decay of F is seen as a decrease in intensity of the 571 nm band (355 s to 3000 s). After the formation of F the Soret maximum at 430 nm shifts to 427 nm (completed after 3-6 ms), corresponding to the absorbance maximum of the oxidized (pulsed) enzyme (Fig.  1A). The disappearance of the Soret maximum at 444 nm also points to complete oxidation of hemes (a ϩ a 3 ) after 3-6 ms (Fig. 1B). The UV-visible spectra did not resolve the concomitant reduction of heme a 3 (Fe 4ϩ ϭO to Fe 3ϩ -OH Ϫ ) and oxidation of heme a, neither in the Soret region nor in the ␣-band. The intensity of the ␣-band decreased (355 s to 6 ms) due to oxidation of heme a. Control stopped-flow experiments indicated that after 4 -6 ms no further optical changes occurred up to several minutes (data not shown).
When the oxidation of the hemes (a ϩ a 3 ) was calculated as the optical absorbance difference at 444 -462 nm (supplemental Fig. S1), the characteristic apparent biphasic oxidation was observed, similar to that for the bovine heart mitochondrial cytochrome aa 3 oxidase (28).
EPR Spectroscopy of MHQ Samples-Representative EPR spectra of the Mn 2ϩ -free CcO samples are shown in Fig. 2. The magnetic field range displayed is suitable for the determination of the redox states of heme a, Cu A , Cu B , and for the detection of the radical described previously (5,46). The figure shows rapid oxidation of heme a and Cu A to ϳ50% after 83 s and 355 s, whereas the remainder was oxidized on the millisecond time scale. The Trp* was developed maximally after 355 s and disappeared within a few milliseconds concomitant with the sec-ond slow heme a/Cu A oxidation phase (Fig. 2). After 6 ms another radical was observed, called the "6-ms radical" (Fig. 3).
The X-band EPR spectrum of the Trp* (355 s) in Mn 2ϩdepleted CcO (Fig. 3) was slightly of a different form than that reported previously; in that work the contribution from Mn 2ϩ had to be subtracted (see Fig. 2 in Ref. 46). In particular, the low field line (g ϭ 1.985) was poorly resolved but present in both the cytochrome aa 3 and cytochrome bo 3 EPR spectra. The Trp* X-band EPR signal consists of four lines with apparent g values of g ϭ 2.036, 2.018, 2.004, and 1.985. We have previously attributed the lines at g ϭ 2.036 and g ϭ 2.004 to a Trp* weakly magnetically coupled (J x,y ϭ Ϫ4.9 GHz) to the heme a 3 Fe 4ϩ ϭO state (46). The lines at g ϭ 2.018 and g ϭ 2.005 could originate from a second Trp* (with J x,y ϭ Ϫ2.1 GHz) or from a Tyr* also including a low field feature now clearly resolved at g ϭ 1.985. Q-band EPR spectroscopy (Fig. 4) leads to a revision of this assignment regarding the Tyr*.
The EPR line shape of the four-line radical signal is fairly constant in the MHQ samples obtained up to 940 s. However, after 2 ms and longer reaction times, a sharp and at X-band frequency axial signal with g Ќ ϭ 2.003 and g ʈ ϭ 1.998 had clearly developed, which is the 6-ms radical spectrally overlapping with the Trp* (Figs. 2 and 3). In particular, part of the sharp feature at g ϭ 2.004 of the Trp* is due to a contribution of the 6-ms radical (Fig. 3).
EPR spectroscopy of the 355-s and 6-ms MHQ samples at Q-band frequency allowed a more solid assignment of the radical (Fig. 4). The four-line spectrum of the Trp* at X-band from g ϭ 2.04 to g ϭ 1.98 is apparently confined to a narrow magnetic FIGURE 2. Representative X-band EPR spectra of cytochrome aa 3 from P. denitrificans rapidly mixed with O 2 and reacted for various times (indicated in microseconds). The four vertical lines indicate the peaks of the Trp* (spectrum after 355 s). The g y resonance of heme a and the g Ќ of Cu A were used to determine their redox states. EPR conditions: frequency, 9.42 GHz; modulation amplitude, 1.0 millitesla; microwave power, 2 milliwatts; temperature, 14 K. The spectra are normalized correcting for differences in gain and enzyme concentrations.  OCTOBER 26, 2007 • VOLUME 282 • NUMBER 43 field region around g ϭ 2. The Zeeman interaction at Q-band frequency (35 GHz) is now much larger than the weak magnetic dipolar/exchange coupling (Ϫ4.9 GHz) leading to a simplification of the four-line spectrum seen at X-band frequencies. The Q-band spectrum of the 6-ms radical (Fig. 4) was better resolved than the X-band spectrum (Fig. 3) displaying a threeline rhombic signal. The rhombic signal represents ϳ0.5% of the CcO concentration. The 6-ms radical is not an ascorbate or PES radical. The absence of resolved hyperfine structure in its spectrum and the g values (g x,y,z ϭ 2.0022, 1.9965, and 1.9994) close to the free electron g value suggested an organic radical, perhaps a main-chain radical. The structural characterization and the possible function of this radical must await further experimentation.

Tryptophan Radical in Cytochrome Oxidase
The Q-band EPR spectrum of the radical obtained after 355 s can be simulated as an S ϭ 1 ⁄ 2 system with simulation parameters (Table 1) characteristic for Trp* radicals (60 -62), in agreement with our previous assignment (46). The hyperfine constants determined by the simulation can be used to calculate the dihedral angles of the ␤-methylene protons with respect to the indole ring (60 -62). Because the dihedral angles are known from the crystal structure (8, 14, 16 -18), the radical was assigned as residing at Trp-272 (see Table 1 and "Discussion"). Tyrosine radicals have much larger g-anisotropy (56,(62)(63)(64)(65) and are absent from the spectrum. The Trp-272* EPR spectrum contained a small contribution from the 6-ms radical (indicated by the arrow in Fig. 4), which was not reproduced in the simulation of Trp-272*.
The kinetics of the Trp-272* calculated from the X-band EPR spectra is displayed in Fig. 5. The maximum amount of the Trp-272* was formed after 300 -500 s and amounted to 4 -5% of the CcO. The transient was fitted with the same rate constants as the Cu A and heme a traces (Fig. 6) and apply to the model shown in Fig. 7. The relatively slow formation of the  Fig. 4 was transferred to Q-band EPR tubes. The 6-ms radical contributes slightly to the Trp* in the spectrum of the 355-s sample. This is seen most clearly in the positive part of the right line, which is relatively sharp. In this region (indicated by an arrow) the fit to the experimental spectrum is somewhat less. The simulation parameters for the Trp* are listed in Table 1. Those for the 6-ms radical are: g x,y,z ϭ 2.0022, 1.9965, and 1.9994. EPR conditions: frequency, 34.972 GHz; modulation amplitude, 1.0 millitesla; microwave power, 2.5 and 5 milliwatts (6-ms sample); temperature, 16 K. Each spectrum is an average of 85 scans. a ␦a 3 -C 3 , ␦Cu B -C 3 , and ␦Ca-C 3 : shortest distance from the C3 atom of the respective tryptophan residue to heme a 3 , Cu B , and heme a, respectively. About 50% of the spin density is located at the C3 atom. b Not conserved in other oxidases. c Y in S. acidocaldarius. d The dihedral angles 1 and 2 are calculated from the crystal structure of P. denitrificans CcO (PDB entry 1QLE). The dipolar/exchange coupling of the oxoferryl-Trp-119 radical in cytochrome c peroxidase (Ϫ4.9 GHz) is the same as that observed for the Trp-272*. The coupling of Ϫ4.9 GHz yields for both radicals a peak at g z ϳ 2.04. The Trp* and Tyr* described in Refs. 61-65 with similar couplings/g z -values are all located within 9 Å from the oxoferryl. The dipolar contribution to the magnetic coupling is dominant over the exchange coupling at distances Ͼ3-4 Å and falls off with the third power of the distance. Trp-358, at 16.1 Å, is expected to yield a coupling of Ͻ1 GHz or g z ϳ 2.009, which is not seen in the X-band EPR spectrum. Both Trp-164 and Trp-272 are at distances from the oxoferryl consistent with the g z ϳ 2.04. e "Standard deviation," the value for the ratio H␤ 2 /H␤ 1 yielding EPR spectra quite similar to the experimental spectrum. See supplemental Fig. S3 and text for further explanation.
Trp-272* (1200 s) and its rapid breakdown (60 s) are consistent with the total accumulation to 4 -5% as determined by EPR.
The oxidation kinetics of Cu A and heme a determined by EPR are presented in Fig. 6. Cu A and heme a are oxidized in two kinetic phases. The kinetics of Cu A and heme a suggest very similar reduction potentials for the two cofactors. The calculated equilibrium constant, K eq ϭ 1.20 Ϯ 0.23, corresponds to a 4.5 Ϯ 5.4 mV (n ϭ 9) lower midpoint potential for Cu A relative to heme a (Fig. 6, inset), a value in good agreement with pulse radiolysis experiments monitored optically (66). Assuming redox equilibrium, the slightly lower midpoint potential of Cu A led to the slightly higher apparent rate of oxidation of Cu A (apparent t1 ⁄2 ϭ 50 s for 25% oxidation) relative to heme a (apparent t1 ⁄2 ϭ 65 s for 25% oxidation) (Fig. 6, inset). The apparent half-lives for the second phase (each component oxidized for 75%) were 1.22 and 1.62 ms for Cu A and heme a, respectively. Fig. 6 further shows that Cu A plus heme a acts as a redox pair donating a  Redox states of heme a and Cu A were calculated from EPR spectra as shown in Fig. 2. The lines through the data are simulations using the six rate constants shown in the model of Fig. 7 and further applying the K eq ϭ 1.   OCTOBER 26, 2007 • VOLUME 282 • NUMBER 43

JOURNAL OF BIOLOGICAL CHEMISTRY 31585
single electron to the binuclear center in each oxidation phase.
The typical four-line EPR signal of Cu B (40,42) was not detected in any of the MHQ samples, even though the temperature was varied between 6 and 100 K, and the microwave power was between 20 microwatts and 200 milliwatts. All samples including the time-zero sample, showed the g ϭ 6 a signal of Fe 3ϩ high-spin heme a 3 , but its intensity was low (Ͻ3% of the CcO) and hardly changed (though slightly increased) between t ϭ 0 and 6 ms (data not shown). These g ϭ 6 data were not further analyzed.
The EPR spectra of MHQ samples obtained with Mn 2ϩcontaining enzyme are shown in supplemental Fig. S2. For these samples quantitation of the radical intensity and Cu A was more difficult due to overlap of the Mn 2ϩ EPR spectrum. However, the Mn 2ϩ signal served as a good internal standard to determine sample reproducibility even though the Mn 2ϩ spectrum is dependent on the redox state of Cu A (27). The sample preparation reproducibility (n ϭ 25) was determined at 1 Ϯ 0.20. The kinetic behavior of Mn 2ϩ -containing or Mn 2ϩ -depleted cytochrome aa 3 was found to be indistinguishable, consistent with similar turnover numbers (180 -190 O 2 s Ϫ1 ) for both types of oxidase preparations. There-fore, Figs. 1, 5, 6, 8, and S1 display data obtained for both types of enzymes.

DISCUSSION
In the oxidative part of the catalytic cycle of P. denitrificans cytochrome aa 3 two radicals are formed as determined by MHQ in conjunction with X-band and Q-band EPR spectroscopy. Simulation of the Q-band EPR spectrum identifies one of the radicals as the catalytically competent tryptophan-neutral radical of the strictly conserved Trp-272 (Trp-272*). Formation of Trp-272* constitutes the rate-limiting step of the catalytic cycle. The current finding that the Trp-272 radical is neutral demonstrates that this residue couples electron transfer to proton movements. We will discuss below how oxidoreduction of Trp-272 can provide the driving force for the transmembrane movement of protons ("proton pumping") through its participation in a proton-relay network. Our findings underscore the general importance of amino acid side chains in coupling electron transfer to proton transfer reactions, alongside the well known metallo-redox centers.
In this report a full kinetic profile in a time window of 83 s to 6 ms has been determined for the oxidation-reduction kinetics of Cu A , heme a, heme a 3 , and the Trp-272*. EPR spectroscopy has the great advantage over UV-visible and resonance Raman spectroscopy in that the concentrations of these components can be determined without mutual spectral interference and without assumptions about the (relative) extinction coefficients or the resonance enhancement. The assignment of the Trp-272* and its possible function in catalysis are discussed within the framework of the model presented in Fig. 7. This model describes the oxidation route of the fully reduced enzyme by the reaction sequence R 3 A 3 P M 3 P R 3 F 3 F W * 3 O H in which the new intermediate F W * contains the Trp-272*. Simulations of the kinetic traces (Figs. 5 and 6) were performed with a single set of rate constants (half-lives) shown in Fig. 7. The appearance and accumulation of the various intermediates is depicted in Fig. 8.
Scope and Limitations of the MHQ Setup-The formation of A (t1 ⁄ 2 ϭ 16 s at 0.65 mM O 2 (29)) could not be resolved, because the instrumental dead time of the MHQ setup is 60 -80 s. The oxygen-binding rate has been established with the flow-flash setup monitored by UV-visible spectroscopy (dead time ϳ 1.5 s determined by the CO dissociation rate (29)), and the structural assignment of A (Fe 2ϩ -O 2 ) is based on resonance Raman The calculated populations of A, which could not be resolved experimentally, and of P R , which was observed (Fig. 1B), were calculated as indicated in the legend to Fig. 7 and are in excellent agreement with previous studies (28,38,68,69). Note that the intermediate P M does not accumulate, due to its rapid (1 ns) conversion to P R . The vertical line represents the MHQ dead time (60 -80 s), positioned at 70 s. The symbols are experimentally determined values. Squares, Trp-272* (and F W *); circles, Cu A plus heme a; triangles, F. F was measured at 571 nm (cf. Fig. 1B). ϳ25 s (32-38)). We did observe the formation of P R (Fe 4ϩ ϭO) indicated by the ␣-band absorbance shift to 606 nm (608 nm in the difference spectra) concomitant with the formation of the Soret absorbance at 430 nm and the disappearance of ϳ50% of the Soret intensity at 444 nm ( Fig. 1  and supplemental Fig. S1). The formation of F at 571 nm was also detected. In addition we could resolve and analyze by EPR spectroscopy rather than by UV-visible spectroscopy part of the initial oxidation phase of heme a and Cu A . Transfer of the first electron to the binuclear center after OϭO bond splitting was completed in ϳ130/200 s (Figs. 6 -8).

spectroscopy (dead time
The heme (a ϩ a 3 ) oxidation kinetics follow the characteristic biphasic pattern of the fully reduced enzyme (supplemental Fig. S1 (28)). In our work the UV-visible spectral data were obtained from independently frozen samples, in contrast to the "continuous" flow-flash methods, and have to be normalized to compare the redox states of the hemes between different samples (5,46,48). Because our data could be simulated with a similar set of kinetic and spectral parameters as for the bovine heart enzyme (Ref. 28 and supplemental Fig. S1), we conclude that the normalization procedure is adequate. A multicomponent analysis of the UV-visible spectra to determine the spectra of the intermediates A, P R , or F proved too difficult at present and has not been pursued.
Formation of P R , F, and O H -The major intermediate accumulating to 80 -90% after 200 -400 s is F (28,35,36,38), formed by rapid protonation of P R (Figs. 7 and 8). Although the optical spectra of P R and F were indistinguishable in the Soret region (maximum at 430 nm due to Fe 4ϩ ϭO of heme a 3 ), F absorbed at 571 nm in the low temperature UV-visible spectrum and P R at 608 nm (606 nm in the absolute spectrum). We could monitor the shift in the ␣-band to 608 (606) nm (at 130 s, Fig. 1, A and B) signifying P R formation, and at slightly later times the formation of F at 571 nm. The decay of F was relatively slow (1-1.5 ms, Fig. 1B) and was analyzed from the absorbance change at 571-580 nm (Fig. 8). Except for the first 100 -200 s, in which these wavelength pairs might contain significant contributions from A and P R , the time course of F was satisfactorily reproduced. The half-life for the P R 3 F protonation was simulated as 27 s, identical to the value in a previous study (38). The value of 27 s was, however, not determined directly from the time course of F, but was constrained to adequately fit the Cu A /heme a and Trp* kinetics (Figs. 5, 6, and 8). The decay rate of F, the F 3 O H transition, can also be estimated from the oxidation of heme a in the ␣-band at 603 nm and in the Soret region at 444 nm (Fig. 1, A and B). Furthermore, the 430 nm maximum shifted with a similar rate (t1 ⁄2 ϭ 1.2 ms) yielding O H , characterized by the maximum at 427 nm after 3-6 ms (Fig. 1, A  and B). Resonance Raman spectroscopy showed that in the F 3 O H transition heme a 3 Fe 4ϩ ϭO was reduced by heme a/Cu A to Fe 3ϩ -OH Ϫ (36,38). The simultaneous oxidation of heme a and reduction of heme a 3 is very difficult to analyze by UV-visible spectroscopy. The rate of the F 3 O H transition was calculated from the second oxidation phase (t1 ⁄2 ϭ 1.2 ms) of heme a and Cu A and from the formation rate of the Trp-272* all three monitored by EPR spectroscopy (Figs. 5, 6,  and 8).
The majority of the high spin heme a 3 Fe 3ϩ (-OH Ϫ ) was EPR-silent under all conditions examined. Likewise the characteristic four-line EPR signal of Cu B 2ϩ -OH Ϫ was not observed in any of the MHQ samples even though the experimental conditions were optimized for its detection. This specific Cu B 2ϩ -OH Ϫ state has been observed in low temperature kinetics ("triple trapping experiments") of the bovine heart CcO and has been assigned to the P R intermediate (42). The Cu B EPR signal is absent in F (67). The four-line EPR spectrum was suggested to disappear due to e.g. protonation of Cu B 2ϩ -OH Ϫ or of another base close to the binuclear center, which would slightly change the magnetic interaction with Fe 4ϩ ϭO to a value rendering Cu B 2ϩ EPR invisible (42). Thus, although our UV-visible spectra indicate formation of P R , our EPR data did not seem to support this. A possible explanation might be that the magnetic interaction in the P. denitrificans CcO differs from the bovine heart enzyme yielding an EPR-silent Cu B 2ϩ in P R . We consider it, however, more likely that the protonation of Cu B (and other) equilibria are somewhat different at the low temperatures employed in the triple-trapping method (40,42,67) compared with 10°C in our experiments. This could lead to accumulation of different intermediates and rendering Cu B 2ϩ EPR-silent. Such a shift in equilibria might also explain why the Trp-272* has not been observed in the triple trapping experiments (40,42,67).
Heme a and Cu A Kinetics and Equilibrium-The biphasic oxidation kinetics of heme a and Cu A determined by EPR (Fig.  6) displayed an initial phase completed within ϳ130/200 s (apparent t1 ⁄ 2 ϭ 55 s), whereas the decay half-life of the second phase equaled 1.2 ms. This biphasic time course and the halflives were in perfect agreement with the kinetics of the two electrogenic events of the P. denitrificans CcO (68,69). In each oxidation phase a single electron from the heme a/Cu A redox pair was donated to the binuclear center ( Figs. 6 and 8).
Heme a oxidation proceeded on the nanosecond time scale (43) in the P M 3 P R transition (Fig. 7), but the apparent t1 ⁄2 values for Cu A and heme a were 50 s and 65 s, respectively (Fig. 6, inset). These latter apparent half-lives are upper limits with respect to the true Cu A N heme a electron transfer rates, because the preceding formation of R 3 P M takes ϳ30 -50 s both for the bovine heart CcO (28,35,36,38) and the P. denitrificans enzyme (68,69). The half-life for Cu A 3 heme a electron transfer has been determined at t1 ⁄2 ϭ 24 and 35 s for the R. sphaeroides and P. denitrificans CcOs, respectively (66,70). The actual freezing time of the MHQ is 30 -40 s and because the heme a 3 heme a 3 electron transfer rate is in nanoseconds, the finding of similar degrees of reduction for Cu A and heme a strongly suggested that they are in electronic equilibrium at all measured reaction times (Fig. 3). The calculated equilibrium constant indicated a 4.5-mV lower midpoint potential for Cu A relative to heme a in good agreement with a previous study (66). The lower reduction potential of Cu A led to a slightly faster apparent oxidation of Cu A relative to heme a. In contrast, in the bovine heart CcO the oxidation of heme a is apparently faster than of Cu A (38).
The second oxidation phase of Cu A /heme a was slow (t1 ⁄ 2 ϭ 1.2 ms) and gated by the slow F 3 F W * reaction (Fig. 7). The reduction of Trp-272* occurred with t1 ⁄ 2 ϭ 60 s (Figs. 5 and 7) by electron transfer from Cu A /heme a. The value of 60 s represents most likely electron transfer from Cu A to heme a. So the rates of Cu A N heme a electron transfer are actually very similar in the two oxidation phases, t1 ⁄ 2 ϭ ϳ30 s in the first phase and 60 s in the second. Both rates are similar to measured (70) and calculated (71) electron transfer rates for Cu A to heme a. The approximate 2-fold difference of t1 ⁄ 2 in the two phases might reflect small differences in the effective reduction potentials and reorganization energies (totaling ϳ20 mV) for each oxidation phase.
Identification of Trp* as Trp-272*-EPR spectroscopy revealed two different radicals (Figs. [2][3][4][5]. The origin of the 6-ms radical could not be established. Likewise, its functional assignment was difficult even though it was being formed on the time scale of turnover. The kinetics of the 6-ms radical, which accumulated to 0.5% of the CcO, could not be accurately established due to spectral overlap with the Trp-272*. However, the observation that two radicals are formed on the time scale of turnover indicates rapid radical migration within CcO (46).
The Q-band spectrum of the radical formed maximally after 300 -500 s can be firmly assigned as a Trp*. The EPR and electron nuclear double resonance properties of Trp radicals are well understood. All Trp radicals have similar hyperfine constants for the indole ring protons (H5 and H7) and the indole nitrogen and show similar small g-anisotropies (60,62). The major differences in Trp* EPR spectra are caused by variations in the angles of the two ␤-methylene protons (H␤ 1 and H␤ 2 ) with respect to the indole ring, which strongly affects their hyperfine values (62). The relation between the ratio of the hyperfine values of the two ␤-methylene protons and the angles can be calculated with the McConnnel relation and thus permits assignment of the Trp* residue in case the crystal structure is known (62). Furthermore, EPR can distinguish between a neutral Trp* and a protonated Trp*.
The hyperfine constants for the indole ring protons and nitrogen used for the Trp* simulation ( Fig. 4 and Table 1) are very similar to those of other Trp* (62). The best fitting values for the two ␤-methylene protons (25 and 7.5 Gauss) correspond to angles of 3.1°and 123.1°(calculated from the ratio H␤ 1 /H␤ 2 ) characterizing the radical as that from Trp-272 (Table 1). However, given the signal-to-noise level in the Q-band spectrum (Fig. 4), which required averaging of 85 spectra, because the absolute Trp* concentration was only 1.4 M, we need to discuss possible other Trp residues as the origin of the Trp*. In particular, simulated EPR spectra in which the ratio H␤ 2 /H␤ 1 would be close to zero (Trp-358 and Trp-431, Table 1) are not entirely inconsistent with the experimental spectrum (see supplemental Fig. S3). Our observation that the Trp* is also formed in the E. coli cytochrome bo 3 (5,46), implies that it concerns a Trp residue conserved in the P. denitrificans and E. coli oxidases. There are 12 Trp residues in subunit I conserved between these two enzymes. Table 1 lists nine Trp residues conserved in the Type A1 oxidases, their distances to the metal sites, and their predicted ratios of the ␤-methylene proton hyperfine con-stants. The highly conserved Trp-164 was ruled out because the radical was found to be present in W164F and W164T mutants (data not shown). Trp-323, although close enough to heme a 3 (and Cu B ) was ruled out on the basis of its predicted EPR spectrum, which would show a strong central line (cf. supplemental Fig. S3), in contrast to experimental observations. Residues Trp-22, Trp-136, Trp-431, and Trp-532 are judged too far from any of the metal centers to produce the magnetic coupling dominating the X-band EPR spectral features. In terms of its predicted EPR spectrum (supplemental Fig. S3) and the 15-to 16-Å distance to heme a 3 , which might lead to a weak magnetic coupling, Trp-358 is the only other serious candidate apart from Trp-272. However, Trp-358 is a phenylalanine residue in the Aeropyrum pernix oxidase and is not conserved in Type A2, Type B, and Type C oxidases (7). We ascribe the observed radical to the Trp-272*, because 1) the Q-band EPR spectrum is optimally simulated as a Trp-272*, 2) Trp-272 is close to all three metal centers in subunit I, and 3) Trp-272 is strictly conserved in all cytochrome oxidases. The Q-band EPR spectrum could not be simulated as a protonated Trp-272* (72), and we therefore conclude that the species represents the neutral Trp-272*. A weak H-bond interaction of the indole-N with another residue, however, can not be ruled out. In fact the phenol-OH of Tyr-167 and the indole-N of Trp-272 are within H-bonding distance (3,8).
In contrast to the Q-band spectrum, the four-line X-band EPR spectrum of Trp-272* (Fig. 4) could not be simulated. With respect to our previous analysis (46) we know now, on the basis of the Q-band spectrum, that this spectrum does not contain contributions from a Tyr* as suggested at the time, although it does contain a small contribution from the 6-ms radical (Figs. 3  and 4). The rapid relaxation of the Trp* (46) and the increased resolution at Q-band indicate that the Trp-272* is weakly magnetically coupled as proposed earlier. Simulation of the X-band EPR spectrum would require a full diagonalization of the spin Hamiltonian matrix, because the various magnetic interactions, including the Zeeman interaction, are in the same order of magnitude. This most complicated EPR-simulation scenario is outside the scope of this report.
The Function of the Trp-272* in the Catalytic Cycle-Our current view on the catalytic cycle of cytochrome oxidases is depicted in Fig. 7. The half-lives of the various intermediates indicated were used for simulation of Cu A /heme a and Trp-272* kinetics (Figs. 5, 6, and 8) with values for R 3 P M /P R taken from the literature. Maximal Trp-272* formation occurs after 300 -500 s, to an extent of 4 -5% of the oxidase. This low extent is due to an ϳ20-fold lower rate of formation (t1 ⁄ 2 ϭ 1.2 ms) than rate of breakdown (60 s). The low formation rate may seems to preclude a direct role of Trp-272 as electron donor in the OϭO bond splitting reaction (A 3 P M ) as suggested recently (44). However, any radical formed in this reaction (Tyr-280* or Trp-272*) would most likely accumulate to undetectable levels (like P M , Fig. 8) given the ϳ1-ns rate of electron transfer from heme a to heme a 3 . Thus while Trp-272 might nevertheless play a role in the OϭO bond breaking (cf. Ref. 44), our data suggest another role in the catalytic cycle.
The intermediate F accumulates to ϳ88% of the CcO concentration ( Fig. 8 and Ref. 38), because it was formed with an apparent t1 ⁄ 2 ϳ 100 s and decays in 1.2 ms, approximately the opposite of the kinetic parameters calculated for Trp-272*. Unless the quantitation of the X-band EPR spectrum of Trp-272* is off by a factor of 20, the accumulation of Trp-272* to 4 -5% implies that Trp-272* is not present in F (Fig. 8) and, in addition, that it is formed later than F. We therefore propose the new sequence F 3 F W * 3 O H (Fig. 7). The simulation of the kinetics allows for ϳ88% accumulation of F and ϳ4% of F W * (Fig. 8). The low accumulation of F W * might explain why it has not been detected hitherto by UV-visible or resonance Raman spectroscopy, even though several authors have proposed additional intermediates in the F 3 O H reaction sequence (28,51,73,74).
Formation of the neutral Trp-272* with t1 ⁄ 2 ϭ 1.2 ms represents the major rate determining step in the CcO catalytic cycle. The Trp-272* is obtained according to Reaction 2.

REACTION 2
Which are the electron and proton acceptors for Trp-272? Potential electron acceptors are heme a 3 Fe 4ϩ ϭO and Cu B 2ϩ -OH Ϫ . Regarding Cu B , its reduction to Cu B 1ϩ -OH 2 would explain its EPR silence. In the next step, Cu B 1ϩ would reduce Fe 4ϩ ϭO while Trp-272* is re-reduced to Trp-272 by the electron residing in the Cu A /heme a redox pair. This sequence of events would explain why heme a 3 and Cu B are EPR-silent in F W * and O H . In the alternative route (shown in Fig. 7) Trp-272 is suggested as the direct reductant to heme a 3 Fe 4ϩ ϭO yielding Fe 3ϩ -OH Ϫ . The Trp-272* is subsequently re-reduced to Trp-272 by electron transfer from Cu A /heme a. In both scenarios, the metal ions can accept the proton (Reaction 2) upon their reduction. Our data and those availing in the literature do not appear to distinguish between these two alternatives in part due to the spectroscopic silence of Cu B . However, resonance Raman spectroscopy indicates the formation of Fe 3ϩ -OH Ϫ prior to Fe 3ϩ (38). According to the literature, the electron donors Cu A / heme a reduce heme a 3 Fe 4ϩ ϭO to Fe 3ϩ -OH Ϫ . The source of the proton in this reaction remains unknown. In Fig. 7, Trp-272 is indicated to act as the electroneutral reductant to Fe 4ϩ ϭO. However, instead of Trp-272 the conserved Glu-278 might be the direct proton donor in the reduction (by Trp-272) of Fe 4ϩ ϭO to Fe 3ϩ -OH Ϫ (in F 3 F W *), whereas the proton from Trp-272 (Reaction 2) is expelled to the periplasm, according to Reaction 3.

REACTION 3
Reprotonation of Glu-278 Ϫ occurs via the D-pathway by proton uptake form the cytoplasm. The short distance of Trp-272 to Cu B or heme a 3 (Table 1) would in any case ensure submillisecond to millisecond electron transfer rates even when the reduction potentials of the redox partners differ by 0.4 V (71). The 1.2-ms rate of Reaction 2 thus suggests a reduction potential of 0.7-0.8 V for Trp-272, a value lower than the ϳ0.9 V and ϳ1.1 V for free Trp in solution (pH 7) or buried, respectively (75), but adequate to play a role in OϭO bond breaking and as reductant to heme a 3 or Cu B .
In the reverse of Reaction 2 the strong base Trp-272 Ϫ (pK a Ͼ 15, cf. Ref. 75) is formed initially by electron transfer from Cu A / heme a. The Trp-272 Ϫ anion might subsequently be rapidly protonated to Trp-272 by a proton en route from the cytoplasm to the periplasm, thus providing directionality to proton translocation (in F 3 O H ). The conserved Glu-278 might serve as the direct proton donor to the Trp-272 Ϫ anion according to Reaction 4.

REACTION 4
Reactions 3 and 4, combined, describe (part of) a protonrelay network in which protons are translocated from the cytoplasm to the periplasm thermodynamically driven by the oxidoreduction of Trp-272 and the strong basicity of the Trp-272 Ϫ anion.
Electrometric and proton translocation measurements have provided ample evidence for proton pumping in the F 3 O H transition (9,12,39,58,68), which we here propose to occur actually in the F W * 3 O H transition involving Trp-272. For the P R 3 F transition, a proton acceptor was suggested to be located close to or at the heme a 3 propionates, but not at the heme a 3 -Cu B binuclear center (9,12,39,58,68). The Trp-272 Ϫ anion fits both the proton acceptor properties and the proposed location.
In view of our findings, the concept of cytochrome oxidase as a redox-linked proton pump (9, 68) might thus be extended. Although the various metal centers are engaged in oxidationreduction-linked deprotonation-protonation reactions, specific aromatic residues like the strictly conserved Trp-272 (and possibly Tyr-280) also change their redox-and protonation states during the catalytic cycle and are likewise involved in proton binding, proton release, and proton translocation. In contrast to the metal ions of the binuclear center, the aromatic residues are not directly involved in the binding (and activation) of O 2 . According to the model presented here, formation of the Trp-272 Ϫ anion provides the driving force for proton binding and even translocation in the F W * 3 O H transition and, thus, constitutes an integral part of a protonrelay network in the cytochrome oxidases. Whether a similar mechanism applies to the P R 3 F transition and to the two proton-pumping events in the reductive part of the catalytic cycle, and whether it would involve Trp-272 as well, are subject to future experimentation.