Oxoferryl-Porphyrin Radical Catalytic Intermediate in Cytochrome bd Oxidases Protects Cells from Formation of Reactive Oxygen Species*

Background: Cytochrome bd oxidases are proposed to reduce O2 to H2O via a peroxide intermediate. Results: Kinetic studies detected, however, an oxoferryl-porphyrin radical intermediate and established insignificant production of reactive oxygen species. Conclusion: Cytochrome bd oxidases, like heme-copper oxidases, reduce O2 in a single four-electron transfer reaction. Significance: Both classes of terminal oxidases converged independently to minimize the production of reactive oxygen species. The quinol-linked cytochrome bd oxidases are terminal oxidases in respiration. These oxidases harbor a low spin heme b558 that donates electrons to a binuclear heme b595/heme d center. The reaction with O2 and subsequent catalytic steps of the Escherichia coli cytochrome bd-I oxidase were investigated by means of ultra-fast freeze-quench trapping followed by EPR and UV-visible spectroscopy. After the initial binding of O2, the O–O bond is heterolytically cleaved to yield a kinetically competent heme d oxoferryl porphyrin π-cation radical intermediate (compound I) magnetically interacting with heme b595. Compound I accumulates to 0.75–0.85 per enzyme in agreement with its much higher rate of formation (∼20,000 s−1) compared with its rate of decay (∼1,900 s−1). Compound I is next converted to a short lived heme d oxoferryl intermediate (compound II) in a phase kinetically matched to the oxidation of heme b558 before completion of the reaction. The results indicate that cytochrome bd oxidases like the heme-copper oxidases break the O–O bond in a single four-electron transfer without a peroxide intermediate. However, in cytochrome bd oxidases, the fourth electron is donated by the porphyrin moiety rather than by a nearby amino acid. The production of reactive oxygen species by the cytochrome bd oxidase was below the detection level of 1 per 1000 turnovers. We propose that the two classes of terminal oxidases have mechanistically converged to enzymes in which the O–O bond is broken in a single four-electron transfer reaction to safeguard the cell from the formation of reactive oxygen species.

Cytochrome bd oxidases are membrane-bound heterodimeric terminal oxidases consisting of CydA (57 kDa) and CydB (43 kDa) (1). These oxidases occur in bacteria and archaea and catalyze the oxidation of ubiquinol or menaquinol (2). This reaction is coupled to the generation of a protonmotive force because the four chemical protons consumed per O 2 are taken from the cytoplasmic side of the membrane, whereas the QH 2 2substrate protons are ejected into the periplasm (3,4). Cytochrome bd oxidases bear no sequence homology to heme-copper oxidases (1) and, because they do not pump protons, have a lower bioenergetic efficiency than heme-copper oxidases (3,4). Cytochrome bd oxidases generally have a high affinity for oxygen and are suggested to act further as oxygen scavengers and as a protection against H 2 O 2 and NO stress (5)(6)(7).
Although three-dimensional structures are lacking for cytochrome bd oxidases, studies suggest that its three heme groups are all located in CydA. The low spin heme b 558 is coordinated by His 186 -Met 393 , the high spin heme b 595 by His 19 , and the chlorin heme d to Glu 99 depending on the redox state (1, 8 -15). The heme normals of b 558 and b 595 are parallel to the plane of the membrane (16), whereas the heme normal of heme d makes an angle of ϳ55°with those of the two other hemes (16,17). A quinone-binding domain has also been identified (1,18,19) that stabilizes a semiquinone (20). Spectroscopic studies suggest that hemes d and b 595 are within 10 Å (21) and form a functional binuclear active site that receives electrons from heme b 558 , proposed as the direct electron acceptor of QH 2 (13,17,(22)(23)(24)(25). Raman spectroscopy has identified heme d 2ϩ -O 2 (Oxy 1 ) 3 and heme d 4ϩ ϭ O (F, oxoferryl or compound II) intermediates indicating that heme d is the site for binding and conversion of O 2 (26 -28).
The current catalytic mechanism, which has been proposed on the basis of flow-flash and stopped-flow kinetic experiments of the reaction between fully reduced enzyme and oxygen, suggests an initial binding of O 2 to heme d to form the Oxy 3 or A * This work was supported by ECHO Grant 700. 54.003 from the Netherlands Organization for Scientific Research (to S. d. V.). This paper is dedicated to the memories of Profs. Thomas M. Loehr and Joann Sanders-Loehr, both pioneers in the field of resonance Raman spectroscopy on metallo-enzymes, including the cytochrome bd oxidase. S. de Vries cherishes the collaborations in science and their warm friendship. □ S This article contains supplemental Figs. S1 and S2 and additional references. 1 To whom correspondence should be addressed. Tel.: 31152785139; Fax: 31152782355; E-mail: s.devries@tudelft.nl.
state (29 -31). Oxy 3 is subsequently converted to a peroxy intermediate, P, with heme b 595 and heme d oxidized to their ferric states, while heme b 558 remains reduced. In the next step (P 3 F) electron transfer from heme b 558 and heme d leads to scission of the O-O bond followed by H 2 O release (29,30). The further donation of one electron and a proton to the active site would restore the enzyme to its fully oxidized form O 0 , with a hydroxo-bound heme d iron. However, this form of the enzyme is probably not part of the normal catalytic cycle (31). Instead, and under physiological conditions, it is more likely that the two-electron donor QH 2 reduces F to Oxy 1 (heme d 2ϩ -O 2 ) followed by reduction by a second QH 2 (yielding Oxy 3 ) to provide the necessary electrons to enter the next catalytic cycle. The as isolated or resting enzyme is usually a mixture of F and heme d 2ϩ -O 2 (26 -28). According to the mechanism described above, the O-O bond is broken in two sequential two-electron transfer steps via a peroxy intermediate. This mechanism differs fundamentally from that of the functionally equivalent heme-copper oxidases, which catalyze a single four-electron O-O bond splitting without a peroxy intermediate (32). The physiological advantage of the latter mechanism is the possible prevention of ROS. Indeed the production of ROS by the heme-copper oxidases has been found to be undetectably low (33,34); in fact (all mitochondrial) ROS production is due to side reactions with O 2 of other respiratory enzymes in their reduced state, notably complex I and complex III (33)(34)(35)(36)(37). Whether cytochrome bd oxidases produce ROS is not known. If they do so, how much ROS is produced and would this be due in consequence to the formation of a peroxy intermediate?
The assignment of a catalytic peroxy intermediate was based solely on the UV-visible spectrum (29) and lacks a solid biophysical underpinning further preventing conclusions about its possible structure as a side-on, end-on, or heme-bridged peroxy species. To characterize the structure of P, the catalytic mechanism of the cytochrome bd-I oxidase from Escherichia coli was investigated using an ultrafast mixing and freezequenching technique (MHQ) that in addition to UV-visible enables EPR spectroscopic analyses (38,39).
Our results indicate that cytochrome bd oxidases split the O-O bond like the heme-copper oxidases in a single four-electron transfer reaction. However, in cytochrome bd oxidases a compound I intermediate is formed, unlike the heme-copper oxidases. The amount of ROS produced by cytochrome bd oxidase was below the detection level of 1 per 1000 turnovers. We propose that both classes of terminal oxidases have convergently evolved to enzymes in which the O-O bond is broken in a single four-electron transfer reaction to minimize the cellular production of ROS.

EXPERIMENTAL PROCEDURES
Overexpression of Cytochrome bd-I Oxidase from E. coli-The cytochrome bo and bd-II knock-out strain MB30 was a donation by M. Bekker (40). MB30 was transformed with plasmid pACYC177 containing the E. coli CydAB operon overproducing cytochrome bd-I. Precultures were grown aerobically in LB medium with ampicillin (50 g/ml) in a shaking incubator at 37°C (ϳ175 rpm). Liter flasks of basal glycerol/fumarate minimal medium (41) containing ampicillin (50 g/ml) were inoculated with 5% of the LB culture, filled to the rim, and closed, creating semi-anaerobic conditions. Cells were allowed to grow at 37°C in a shaking incubator for ϳ20 h. These starter cultures were used to inoculate (4%) four 25-liter glass vessels with basal glycerol/fumarate medium and ampicillin (5 g/ml). Cells were grown under hypo-aerobic conditions after nitrogen flushing, while stirring at 30°C for ϳ65 h.
Purification of Cytochrome bd-I Oxidase-After a 25-fold concentration of the cell cultures in a cross-flow filtration system, the cells were harvested by centrifugation (4°C, 10 min, 9000 ϫ g) and washed once with 50 mM Tris-HCl buffer, pH 8. The washed cell pellets were resuspended in the same buffer prior to cell disruption at 1.8 kbar. The resulting suspension was centrifuged (4°C, 10 min, 3000 ϫ g) to remove cell debris. The supernatant was then centrifuged in a Beckman ultracentrifuge (4°C, 1 h, 100,000 ϫ g) to spin down the cell membranes containing the bd-I oxidase. Membranes were resuspended in 25 mM MOPS, pH 6.8, 1 mM EDTA and washed once or twice. The enzyme was extracted from the E. coli membranes by addition of 1% lauryl maltoside to the solution and incubating while stirring on ice for 15 min. Purification of the membrane-extracted enzyme consisted of a single column chromatography step (Q-Sepharose FastFlow) with 25 mM MOPS buffer, pH 6.8, as the running buffer. Diluted fractions were pooled by activity, concentrated, and stored at Ϫ80°C.
Freeze-quench Experiments-MHQ, EPR, UV-visible experiments, and kinetic simulations were performed as described previously (38,39) using the IGOR Pro software from Wavemetrics, Inc. The MHQ setup was modified just before the mixer entry with a stainless steel tubing extension immersed in ethylene glycol at Ϫ5°C to bring the reaction temperature to 1 Ϯ 1°C. For kinetic experiments, purified enzyme (150 or 300 M) in 50 mM sodium phosphate buffer, pH 7.8, 5 mM EDTA, 0.05% lauryl maltoside was made anaerobic, reduced with 2 mM sodium dithionite, and subsequently mixed with the same buffer saturated with O 2 . EPR spectra were normalized at the intensity of heme b 595 and in separate experiments using an internal CuClO 4 (0.1 mM) standard in the oxygenated buffer before mixing. Data in Fig. 5 represent the average of four independent experiments. UV-visible averaged spectra were corrected for scatter and base line as described previously (38,39) and normalized as follows. Normalization is necessary because for the UV-visible experiments, the amounts of cold freezequenched powder in the low temperature cuvette is variable. Low temperature reference spectra (not prepared by MHQ) of fully reduced and "as isolated" enzyme (10 M) were recorded in buffer, and the Soret band maxima relative to 490 nm were determined (1.0 and 0.46, respectively). The major difference in the measured maximal amplitude of the Soret band absorbance is due to the relatively sharp peak of reduced heme b 558 , in particular for samples after 100 s. The relatively broad Soret peaks of the oxidized hemes contribute mainly to the difference 450 -490 nm, in particular for samples after 100 s. With these two parameters, the relative intensities of the spectra shown in Fig. 4 were calculated. From these, the fractional amount of reduced heme b 558 was calculated from the spectra in the ␣-band region. The error in this calculation amounts to ϳ0.1 heme b 558 per enzyme. For the absorbance at 680 nm, the error was 0.35 per enzyme. The maximal absorbance at 680 nm was taken the same as that of oxidized heme d (cf. Ref. 29).
Determination of ROS-Spin trapping assays were performed with 25 mM DEPMPO (42) in the same buffer as above. The reaction was started by addition of 0.1 M bd-I oxidase or 200 M dQH 2 , both, or both in the presence of either catalase (1 unit) or superoxide dismutase (1 unit). The 200 M dQH 2 is fully oxidized in 20 s. Superoxide was prepared from solid KO 2 in 1 M NaOH. The DEPMPO superoxide adduct has a half-life time of 17 min (42). Room temperature EPR spectra were recorded in a 100-l aqueous sample cell 120 s after addition of the reagents and subsequently after 240 and 360 s. The spectra in Fig. 7 are the average of these three spectra. At longer reaction times, a background DEPMPO radical developed in the presence of dQH 2 . Similar experiments with 400 M ferrous cytochrome c ϩ 0.1 M Paracoccus denitrificans cytochrome aa 3 oxidase produced a background signal after ϳ200 s, limiting the detection level of the assay with cytochrome aa 3 oxidase to one ROS per 250 turnovers.
The Amplex Red hydrogen peroxide/peroxidase assay was performed according to the manufacturer's protocol (Invitrogen). The final assay volume of 80 l each consisted of 50 l of the Amplex Red reagent/HRP working solution; to this 30 l of buffer was added yielding the same final enzyme and reagent concentrations as used for EPR. The formation of resorufin was monitored at 550 nm in an HP Agilent 8453 diode array spectrophotometer in 1-min intervals after manual mixing of the enzymatic solution with the reagent working solution. All assays were performed at room temperature, and a background trace was recorded for each assay. The background reaction with dQH 2 limits the sensitivity of the assay to ϳ1 H 2 O 2 per 1000 turnovers of the cytochrome bd oxidase. The assay could not be performed successfully with cytochrome aa 3 oxidase because of spectral overlap of resorufin and ferrous cytochrome c.

Heme d Oxoferryl Porphyrin -Cation Radical Intermediate (Compound I) Detected by EPR Spectroscopy-
To study the mechanism of O-O bond splitting by the cytochrome bd oxidase, single turnover experiments were performed at 1°C to slow down the reaction. After the reaction between reduced enzyme and O 2 , intermediates were trapped by means of freeze-quenching at times Ն100 s and analyzed by EPR and low temperature UV-visible spectroscopy.
The EPR spectrum of as-isolated cytochrome bd oxidase displays resonances from two high spin heme species, the axial heme d (g ϳ6) and the rhombic heme b 595 (g x ϳ6.2 and g y ϳ5.7) and a third signal from the low spin heme b 558 (g z ϭ 3.58) (cf. Fig. 1) (24,43). The reduced enzyme is EPR-silent. After reacting for 100 s, heme b 595 became fully oxidized (Fig. 1, middle trace), whereas only ϳ0.1-0.2 heme d had converted to the ferric state. The middle trace in Fig. 1 further shows a previously undetected intermediate at 3100 -3500 G, which is argued below to be a compound I derivative of heme d. The new EPR signal ( Fig. 2 and supplemental Fig. S1) consists of three overlapping signals arising from three rhombic S ϭ 1 ⁄ 2 spin sys-tems when recorded at 4.2 K. At higher temperatures the line shapes of the three EPR signals change and coalesce at 77 K into a single rhombic signal with g values that are the average of the Top, as isolated enzyme ("infinite time"); middle, after reaction for 100 s with O 2 ; bottom, fully reduced enzyme ("zero time"). The (partial) upper spectrum shows the g z resonance of the low spin heme b 558 at g ϭ 3.58 from a 25-fold concentrated as isolated enzyme solution; the peak at g ϭ 3.58 is too weak to be detected in freeze-quenched samples. The peak at g ϭ 4.3 is from adventitious iron, the broad peak around 3200 G from adventitious Cu 2ϩ in the cavity, and the sharp signal at ϳ3380 G is due to the freeze-quench procedure. EPR conditions are as follows: Microwave frequency, 9.45 GHz; modulation amplitude, 0.5 millitesla; microwave power, 20 microwatts; temperature, 4.2 K. Full traces are displayed at the same gain. individual signals (Table 1 and supplemental Fig. S1). At 4.2 K the integrated intensity of the three signals together accounts for 0.75-0.85 spins per enzyme. The unusual temperature dependence of the EPR signals is due to a magnetic dipolar interaction between oxidized heme b 595 and the compound I, a conclusion that will be explained below.
The magnetic properties of compound I are well understood (supplemental material) (44,45). Briefly, compound I comprises an S ϭ 1 heme oxoferryl center (Fe 4ϩ ϭO) that is magnetically coupled to a S ϭ 1 ⁄ 2 porphyrin -cation radical. The coupling of the two spins yields three Kramer's doublets (cf. Fig.  3B) yielding either an S ϭ 1 ⁄ 2 or S ϭ 3/2 ground state, which depends on the relative magnitudes and signs of the Heisenberg exchange interaction (J) between the Fe 4ϩ ϭO and the porphyrin radical and further on the zero-field splitting (D) of the S ϭ 1 species. The finding here that the three g values are close to g ϭ 2 indicates a total spin of S ϭ 1 ⁄ 2 for the ground state of the compound I. The Kramer doublets are separated in energy by an amount ⌬ (ϳD) and J. Because ⌬ is usually quite small, 20 -40 cm Ϫ1 , compound I species follow a two-phonon Orbach relaxation mechanism. The presence of a low-lying first excited state will also result in significant loss of spin intensity at temperatures greater than ⌬ (i.e. above ϳ30 K). Hence, to validate the assignment as compound I, both the relaxation behavior and the ground state population were determined (Fig. 3). Both these experiments should yield a similar value for ⌬ (46).
The increase of the relaxation rate upon increasing the temperature follows an Orbach relaxation mechanism at T Ͼ4.2 K (Fig. 3A) for a first excited state at ⌬ ϭ 36.8 Ϯ 4.8 cm Ϫ1 . The decrease of spin intensity corresponds to the presence of excited states that are 32.2 Ϯ 10.4 cm Ϫ1 (⌬) and 33.9 Ϯ 11.7 cm Ϫ1 (⌬ ϩ J), respectively, above the ground state (Fig. 3B). The latter two values indicate a small value for J of ϳ2 cm Ϫ1 or ⌬ Ͻ0.1 J. Such a small value for J (either negative or positive) relative to ⌬ (or D) is consistent with g values close to g ϭ 2 (Table 1), and in fact is quite similar to those calculated for the isolated S ϭ 1 Fe(IV) system for which J ϭ 0 (47). The value g z ϭ 1.973 determined here is consistent with a calculated value for D (or ⌬) of ϳ30 cm Ϫ1 (47) and close to that determined here. The decrease of the ground state spin population rules out that the EPR signal is derived from a ferric heme peroxy center for which the first excited state lies at Ͼ700 cm Ϫ1 , determined by the strength of the crystal field (supplemental material) (48). In addition, the g values would be very unusual for low spin heme centers.
The observation that the compound I EPR signal is split into three signals (Figs. 1 and 2 and supplemental Fig. S1) with similar intensities suggests that compound I is coupled to a nearby anisotropic magnet for which at 4.2 K the relaxation is much slower than that of compound I, whereas at higher tempera-tures the reverse holds. At T Ͼ 60 K, the relaxation of this magnet is so fast that the splitting averages out resulting at 77 K in a compound I signal with g values that are the average of those at 4.2 K (Table 1). Previous studies have provided evidence for magnetic interactions between ferric heme d and heme b 595 (9,24,43). We therefore propose an anisotropic magnetic dipolar interaction between the heme d-derived compound I and heme b 595 . At high temperatures, the relaxation of the ferric heme b 595 is much faster than that of compound I, consistent with the detection of an EPR signal of the latter at 77 K but not of heme b 595 (see supplemental material). Interestingly, the splitting is much more pronounced in the g x,y resonances than in the g z peak. The g z is directed along the Fe 4ϩ ϭO bond, perpendicular to the plane of heme d. Because the angle between the heme d and heme b 595 normals is ϳ55° (16, 17),   MARCH 16, 2012 • VOLUME 287 • NUMBER 12

JOURNAL OF BIOLOGICAL CHEMISTRY 8833
which is close to the magic angle of 54.7°at which the magnetic dipolar coupling is zero, the small splitting in the g z resonance is consistent with this angle determined by other methods. We conclude that the new EPR signal is from a heme d oxoferryl porphyrin -cation radical in dipolar magnetic interaction with heme b 595 3ϩ . Fig. 4 shows low temperature UV-visible spectra of the reaction between cytochrome bd-I oxidase and O 2 . After 100 s, the peak of heme d at 624 nm has shifted to 641 nm and broadened considerably, whereas heme b 558 has remained largely reduced (75-85%). In agreement with the EPR spectra (Fig. 1), heme b 595 is completely oxidized after 100 s indicated by the disappearance of the broad absorbance around 595 nm and the appearance of a negative peak in the Soret region (439 nm) in the "0 -100-s" difference spectrum (49). This difference spectrum further indicates the appearance of a broad absorbance at 404 nm. The peak at 404 nm (and that at 641 nm) is ascribed to that of the heme d compound I intermediate. In agreement with this are the blue shifts from ϳ430 nm for the ferrous state to 404 nm and the low extinction, ϳ25-30% of the intensity of the Soret band of heme b 595 , two features also observed for compound I from horseradish peroxidase (50). The compound I absorbance is also directly visible in the absolute spectra of Fig. 4 as a shoulder at 404 nm on the Soret peaks of hemes b 595 and b 558 . This shoulder disappears as the reaction proceeds. Difference spectra calculated for times Ͼ100 s did not resolve the 404-nm band as well as after 100 s because of spectral interference from hemes b 558 and d, the latter changing to the ferric state at longer reaction times. In contrast, the intermediate state obtained after 100 s is quite pure, i.e. full oxidation of heme b 595 , Ͼ75% change of heme d 2ϩ to heme d 4ϩ ϭO, Ͻ25% heme d 3ϩ , and Ͻ25% change of heme b 2ϩ 558 to b 3ϩ 558 . Kinetic Analysis-The time-dependent redox changes calculated from UV-visible and EPR spectra are shown in Figs. 4 and 5 and supplemental Fig. S2. The formation of compound I within 100 s is followed by the slower oxidation of heme b 558 (Fig. 4) and by optical changes in the Soret region and around 640 nm because of formation of heme d 3ϩ , also observed by an increase of the signal at g ϳ 6 (supplemental Fig. S2). In the same time window, the EPR signal of compound I disappears (supplemental Fig. S2). The absorbance changes at 680 nm (F) are small and difficult to analyze due to sloping base lines. F is estimated to accumulate to ϳ0.1 per enzyme and did not show the transient behavior expected for a true intermediate. The lack of this transient and the remaining 15-20% reduction of heme b 558 after 2 ms are explained by the slight excess of reductant present (sodium dithionite), which renders the oxidation kinetics not pure single turnover; instead, the enzyme reaches a quasi steady state. Here, heme d is ϳ80% oxidized with the remainder present as F and a small amount of heme d 2ϩ -O 2 represented by the absorbance at 646 nm (cf. 26 -28). This electronic distribution is in agreement with experiments that show that F and Oxy 1 are dominant steady-state species (51). Note that the 680 nm band is also present in the enzyme as isolated   Table 2. Because the system evolves to a quasi-steady state (see text), the traces of b 558 2ϩ and d 3ϩ (Oxy 1 ) are calculated for a total change of 0.9 and 0.8 per enzyme, respectively. For the same reason the kinetics of CpdII were calculated by the "approach to steady-state method" employing the relevant equations in Ref. 39 using the rate constants in Table  2. Data for heme b 595 , heme d, and CpdI are from EPR (e.g. supplemental Fig.  S2) and those for CpdII and heme b 558 from the UV-visible spectra in e.g. Fig. 4. (Fig. 4) (26 -28) but that heme d 2ϩ -O 2 is absent in our preparation.

Compound I Species Detected by UV-visible Spectroscopy-
EPR spectroscopy (supplemental Fig. S2) shows that as compound I disappeared and heme d 3ϩ is formed, the line shape of heme b 595 , in particular the g ϳ 5.7 derivative-like resonance, shifts by ϳ 10 G. The small shift is interpreted as a change in magnetic interaction between heme b 595 and heme d, as the latter changes from the compound I state to the ferric state. After 2 ms, the EPR spectrum of the high spin heme centers is similar to that of the "as-isolated enzyme." The kinetic profiles of the various intermediates determined from the UV-visible and EPR spectra are shown in Fig. 5, and the calculated rate constants are listed in Table 2. The rates of oxygen binding, heme b 595 oxidation, and compound I formation were too fast to be determined directly in this study, even at the reaction temperature of 1°C, where these reactions appear completed within 100 s. Flow-flash experiments (29) indicate a 10-fold lower rate of oxidation of heme b 558 than the preceding reactions suggesting a (combined) rate of ϳ20,000 s Ϫ1 for oxygen binding, oxidation of heme b 595 , and compound I formation (cf. Table 2). The accumulation of compound I to 0.75-0.85 per enzyme is consistent with the ϳ10-fold higher rate of its formation than its rate of decay (Table 2).
Significantly, the rates of compound I decay and oxidation of heme b 558 are the same, but the formation of heme d 3ϩ is slower ( Table 2). The similar rates of compound I decay and heme b 558 oxidation are consistent with the view that electron transfer from heme b 558 leads to direct reduction of the porphyrin -cation radical, thus producing the ferryl form of heme d or F. In the subsequent reaction F, which barely accumulates, is rapidly reduced further, by excess reductant, yielding heme d 3ϩ . The rate of compound II reduction is calculated at ϳ7300 s Ϫ1 based on the experimental time delay between compound I decay and heme d 3ϩ formation and the accumulation to 0.1 per enzyme estimated from the 680 nm absorbance.

DISCUSSION
Catalytic Mechanism-The reduction of O 2 by reduced cytochrome bd oxidase will in general not yield clean (pseudo-) first-order traces because the complete reaction needs four electrons, and the enzyme can store only three. In our experiments, the small excess reductant leads to rapid net 4 -5 electron transfer leaving some reduced enzyme after 2 ms that is slowly oxidized in a quasi-steady state in which excess reductant and remaining oxygen are exhausted. In the flow-flash experiments, F was formed almost stoichiometrically in 47 s, apparently corresponding to a net three-electron reaction (29). In the next step (ϳ1.1 ms), F was converted to a mixture of oxidized and oxygenated enzyme as observed here. The transient kinetics of F in the flow-flash experiments show that it is a true intermediate. The nontransient kinetics of F in our experiments might suggest that it is not part of the main catalytic pathway but, for example, in rapid equilibrium with another/ unknown intermediate. However, because the reaction proceeds to a quasi-steady state, also a true intermediate may show nontransient kinetics.
The optical and kinetic properties of the intermediate formed after 100 s with peaks at 404 and 641 nm are the same as those observed for the peroxy intermediate P formed after 4.5 s (peak at 635 nm at 20°C (29)). The assignment as a peroxy intermediate by the authors was based solely on UVvisible spectroscopy and on the notion that formation of compound I "is not very likely, because it would require oxidation of a nearby amino acid residue or the porphyrin ring that is energetically unfavorable in the presence of the reduced heme b 558 in the proximity of the catalytic center" (29). Here, we provide both EPR and UV-visible spectroscopic evidence that the intermediate labeled P in Ref. 29 is in fact compound I. We therefore propose a new reaction mechanism for the cytochrome bd oxidase (Fig. 6). Accordingly, after binding of O 2 to the fully reduced enzyme (Red 3 ) in which Oxy 3 is formed, the O-O bond is split in a single four-electron reaction producing within 100 s the compound I intermediate (CpdI or F ؉ *) without formation of a peroxy state. Oxygen bond breaking is accomplished in an apparently concerted electron transfer reaction from heme b 595 , heme d (two electrons), and the heme d porphyrin moiety. The obligatory proton donor needed for O-O bond splitting remains unknown. The subsequent internal electron transfer from heme b 558 (525 s) converts CpdI into CpdII (or F). Rapid electron transfer from dithionite or from endogenous QH 2 (1-2 electrons per enzyme) produces a largely oxidized enzyme after 2 ms and a mixture of some heme b 558 2ϩ , Oxy 1 , and CpdII. The formation of kinetically competent compound I and II intermediates by the cytochrome bd oxidases resembles the mechanism of plant peroxidases and eukaryotic catalases. The catalytic mechanism of cytochrome bd oxidases is similar to that of heme-copper oxidases, which also break the O-O bond in a single four-electron transfer. However, in the heme-copper oxidases, P M rather than Cpd1 is formed initially, i.e. a heme a 3 oxoferryl intermediate plus an amino acid radical, most likely a tyrosine radical (52).
ROS Production-Breaking the O-O bond in a single fourelectron transfer represents a mechanism to protect the cell from ROS production during aerobic growth. Possible production of ROS was determined by the EPR spin trapping assay with DEPMPO (42) and the UV-visible based Amplex Red reagent assay. The sensitivity of the Amplex Red assay is limited by a slow background reaction and the spin trapping method by the DEPMPO-derivative stability, its EPR detection limit, and the dQH 2 solubility of 200 M limiting the number of turnovers. Both methods indicated that the production of ROS by cytochrome bd oxidase was below the detection limit of Ͻ1 per 1000 turnovers, i.e. less than 0.1 M O 2 Ϫ or H 2 O 2 produced per 100 M O 2 consumed (cf. Fig. 7). The production of ROS by purified P. denitrificans cytochrome aa 3 oxidase was determined at Ͻ1 per 250 turnovers, a relatively high value because  MARCH 16, 2012 • VOLUME 287 • NUMBER 12

Compound I in Cytochrome bd Oxidase Prevents ROS Formation
of the presence of a background signal. However, the production of ROS by the similar mitochondrial cytochrome aa 3 oxidase has previously been estimated to be much lower than 1/250 turnovers (33)(34)(35)53). In fact ROS production is too low to detect mainly due to the contribution by other respiratory enzymes. In vivo ROS production by mitochondria is estimated at ϳ1 per 1000 turnovers of the respiratory chain (53)  . equilibrium in the initial oxy-derivative. Resonance Raman spectroscopy on the two classes of terminal oxidases and oxymyoglobin indeed shows considerable superoxide character for the oxy-complexes (27,55,56). The release of O 2 . by oxymyoglobin is, however, very slow (t1 ⁄ 2 ϳ10 h (57)) among others because of its high midpoint potential (ϳ0.1 V) (58). In the terminal oxidases the midpoint potentials of the hemes are even higher (ϳ0.3 V) (43,52), which would most likely result in even slower release of O 2 . . In addition, the oxy-complexes of the terminal oxidases are rapidly (ϳϽ10 s) converted to Cpd1 or to P M yielding a calculated O 2 . production of Ͻ1 per 10 9 turnovers, far below current detection levels and far below that of other respiratory complexes. It appears that the cytochrome bd oxidases and heme-copper oxidases have evolved independently to minimize if not prevent production of ROS by very rapidly breaking the O-O bond in an apparently concerted single four-electron transfer and protonation reaction. To do so, both classes of terminal oxidases harbor a compact bi-metallic center integrated with a nearby proton donor. This bi-metallic center contains one metal ion able to attain the Fe 4ϩ state and an additional redox center, either the porphyrin ring itself, or a nearby amino acid as donor of the fourth electron. In the single heme-containing peroxi-  ؊ or other (e.g. OOH ؊ ) adducts during the reaction. None are seen in the complete reaction (bd/ dQH 2 ), or in the controls with superoxide dismutase (bd/dQH 2 /SOD), or catalase (bd/dQH 2 /Cat), or in the presence of only enzyme (bd) or substrate (dQH 2 ). The DEPMPO-O 2 Ϫ adduct at 1.5 M was prepared as described under "Experimental Procedures." The traces of 0.4 and 0.1 M were calculated from the 1.5 M spectrum by multiplication of 0.27 and 0.067, respectively, and then adding a random noise function with the same noise as the experimental spectrum. The detection DEPMPO-O 2 Ϫ limit is below ϳ0.1 M, because in this (calculated) spectrum the S/N ratio is ϳϽ1. Microwave frequency, 9.79 GHz, modulation amplitude, 0.1 millitesla, microwave power, 20 milliwatt. dases, catalases, and cytochrome P 450 enzymes, the same natural variation is observed that the porphyrin ring or a nearby amino acid such as Trp or Tyr donates the extra electron to cleave the peroxy O-O bond (44, 45, 59 -61).