Quantitative Reevaluation of the Redox Active Sites of Crystalline Bovine Heart Cytochrome c Oxidase*

Approximately 30% of the iron contained in a bovine heart cytochrome c oxidase preparation was removed by crystallization, giving a molecular extinction coefficient 1.25–1.4 times higher than those reported thus far. Six electron equivalents provided by dithionite were required for complete reduction of the crystalline cytochrome c oxidase preparation. The fully reduced enzyme was oxidized with 4 oxidation equivalents provided by molecular oxygen, giving an absorption spectrum slightly, but significantly, different from that of the original fully oxidized form. Four electron equivalents were required for complete reduction of the O2-oxidized enzyme. The O2-oxidized form, when exposed to excess amounts of O2, was converted to the original oxidized form which required 6 electrons for complete reduction. A slow reduction of the O2-oxidized form without any external reductant added indicates the existence of internal electron donors for heme irons in the enzyme. These results suggest that the 2 extra oxidation equivalents in the original oxidized form, compared with the O2-oxidized form, are due to a bound peroxide produced by O2 and electrons from the internal donors, consistently with a peroxide at the O2 reduction site in the crystal structure of the enzyme (Yoshikawa, S., Shinzawa-Itoh, K., Nakashima, R., Yaono, R., Yamashita, E., Inoue, N., Yao, M., Fei, M. J., Peters Libeu, C., Mizushima, T., Yamaguchi, H., Tomizaki, T., and Tsukihara, T. (1998)Science 280, 1723–1729).

Cytochrome c oxidase catalyzes the reduction of O 2 to water as the terminal oxidase of cell respiration. The O 2 reduction is coupled with proton pumping. The enzyme contains two redox active copper sites and two heme iron sites (1). The x-ray crystallographic structure of the enzyme isolated from bovine heart has been determined at 2.3-Å resolution in the fully oxidized state and at 2.35-Å resolution in the fully reduced state (2). The x-ray structure shows that one of the copper sites (Cu A ) is dinuclear but likely to be a 1-electron accepting site and that a peroxide is bridged between iron and copper in the O 2 reduction site in the fully oxidized state. The two heme A sites, the two copper sites, and the bridging peroxide may accept 6 electron equivalents in total. The prediction is not consistent with the widely accepted conclusion that 4 electron equivalents are required for complete reduction of the fully oxidized enzyme, based on metal analysis and redox titration experiments (1,3). However, difficulties in purification and incomplete removal of O 2 from the membrane protein solution are likely to decrease the accuracy in the metal analysis and in the redox titration experiment. Thus, we reexamined the metal content of the enzyme using crystalline bovine heart cytochrome c oxidase and the electron equivalents required for complete reduction of the fully oxidized form and for complete oxidation of the dithionite reduced form. The results obtained here are consistent with the enzyme structure in the fully oxidized state containing 1 equivalent of peroxide and four metal sites, each receiving 1 electron equivalent.

EXPERIMENTAL PROCEDURES
Cytochrome c oxidase was purified from bovine heart muscle by the method of Yoshikawa et al. (4). The detergent, Tween 20, was replaced with C 12 E 8 1 (Nikko Chemicals, Japan), C 12 E 23 (Pierce), or DM (Anatrace). For crystallization, the enzyme solution was concentrated with an Amicon Diaflow apparatus. Metal content was analyzed with a model Z-6100 Hitachi polarized Zeeman atomic absorption analyzer. The enzyme samples were first wet-ashed with nitric acid, hydrogen peroxide, and perchloric acid. About 20 mg of the enzyme sample was used for a single determination. Thus, at least 80 mg of the enzyme sample was required for a single set of analyses for four metals (iron, copper, magnesium, and zinc). Cytochrome c oxidase activity was determined by following the aerobic oxidation of 15 M ferrocytochrome c at pH 6.0.
The redox titration of cytochrome c oxidase was performed under anaerobic conditions, using sodium dithionite or NADH-PMS system for reductive titration and O 2 for oxidative titration, using a modified anaerobic titration system of Burleigh et al. (5). The system included a Thumberg type cuvette (Fig. 1, A and B), a vacuum system for exchanging the gas phase in the cuvette with O 2 -free N 2 (Fig. 1C), and a flask for dithionite solution (data not shown). The Thumberg type cuvette was made of Pyrex glass (Fig. 1, A and B) with three ports. The port for titrant consisted of a male joint stoppered with a rubber septum (Fig.  1A, b). A female joint, fused to a male joint (Fig. 1A, c) stoppered with a silicon rubber septum (Fig. 1A, d), was attached to the male joint of the titrant port (Fig. 1A, a). A female joint mounted with a gas-tight syringe (fitted with a needle) was attached to the male joint stoppered with the silicon rubber septum (Fig. 1, A, d, and B). The brim of the silicon rubber septum was trimmed for effective sealing between the two joints. The space inside the male-female joint attached to the titrant port was filled with N 2 -saturated water. The rubber septum for sealing the titrant-port (Fig. 1A, b) had been degassed by keeping it under vacuum overnight. The port for connecting the cuvette to the vacuum system was a two-way stopcock with a male joint at the end (Fig. 1A, g). The stopper of the two-way stopcock was especially ground for close fitting and had its connecting channel running from the bottom through the center in order to minimize leakage. A fairly large space was provided between the optical cuvette and the port for connection to the vacuum line to allow for foaming from the protein solution in the cuvette during evacuation. The last port (Fig. 1A, e) was used for introducing the enzyme solution and was sealed with a glass stopper. All of the tapered ground joints were sealed with Apiezon L high vacuum grease, whereas Apiezon N was used for the two-way stopcock.
A gas-tight syringe with a screw-driven piston (Hamilton, no. 1750) was used to add titrant to the enzyme preparation in the Thumberg type cuvette (Fig. 1B). The syringe had a long stainless steel needle fixed permanently for gas-tightness with epoxy resin. A sketch of the cuvette mounted with the titrant syringe is given in Fig. 1B.
The vacuum system (Fig. 1C) consisted of three stainless steel tubing lines (3 mm in diameter; Takahama, Himeji), for connecting the Thumberg type cuvette, for supplying O 2 -free N 2 , and for evacuation with a vacuum pump. The three stainless steel tubes were connected with a T-shaped stainless steel tube (Fig. 1C, c), and swage locks were used for all steel tube-steel tube connections. A female glass joint for connecting the cuvette was fixed with epoxy resin on the end of the steel tubing line (Fig. 1C, a), which had an adjacent spiral steel tubing (Fig. 1C, b) for mechanical flexibility. The second branch was connected by rubber vacuum tubing (Fig. 1C, i) to a vacuum pump (Fig. 1C, h) with a liquid nitrogen trap (Fig. 1C, g) in between. The third branch was connected in tandem to a tube containing a catalyst (Oxytrap, Alltech) (Fig. 1C, e) for complete removal of trace amounts of O 2 in N 2 gas, and to a N 2 tank containing ultrapure N 2 (99.9999%; Takahama, Himeji) (Fig. 1C, f). The second and third branches were equipped with high vacuum two-way metal stopcocks (Nupro, SS-2H) (Fig. 1C, d) for switching the connections to the Thumberg type cuvette.
Dithionite solution was prepared in, and withdrawn from, a 130-ml flask having a port at the side similar in construction to the titrant port of the Thumberg type cuvette (data not shown), with the method of Burleigh et al. (5). The flask was filled with 90 ml of 0.1 M pyrophosphate buffer, pH 9.0, and after placing a magnetic stirring bar coated with glass, the flask was sealed with a glass stopper equipped with a two-way stopcock. The stopcock was connected to the vacuum system with a rubber vacuum tube, and the buffer solution in the flask was deaerated by five cycles of evacuation-equilibration with N 2 . After adding 40 mg of solid sodium dithionite to the buffer solution through the top with a small funnel, the flask was again stoppered and deaerated with three cycles of evacuation-N 2 saturation. The reducing equivalent contained in the dithionite solution was calibrated using potassium ferricyanide as the standard and the Thumberg type cuvette. The dithionite solution remained stable for at least 7 days without any detectable change in concentration of the reducing equivalent. The air-saturated water was prepared in a flask open to the air at 20°C with gentle overnight stirring. The O 2 concentration was obtained from a standard table (6).
For reductive titration of bovine heart cytochrome c oxidase with dithionite, the Thumberg type cuvette, containing 3 ml of 7.5 M cytochrome c oxidase in 0.1 M sodium phosphate buffer, pH 7.4, and 1 ml of deionized water to compensate for water loss during evacuation, was connected to the vacuum system and deaerated through three cycles of the evacuation-N 2 equilibration. Precaution was taken to minimize foaming in the enzyme solution during the evacuation. Each evacuation took 20 -30 min, and at the end of the final evacuation, the volume of the enzyme solution was adjusted to 3.0 ml by estimating the height of the solution in the cuvette. The amount of the enzyme was determined from the extinction coefficient of the fully reduced enzyme after reductive titration. Comparison of amount of enzyme before and after deaeration revealed that about 0.6% of the protein was removed by each evacuation-N 2 saturation procedure.
After securing the syringe within the port (Fig. 1B), the dithionite solution was dispensed onto the L-shaped arm of the Thumberg type cuvette (Fig. 1A, a) and mixed with the enzyme solution by gently tilting the cuvette. For the oxidative titration, the air-saturated water was dispensed directly into the enzyme solution, which was placed in the L-shaped arm in order to avoid release of O 2 to the gas phase in the cuvette.
Anaerobiosis produced inside the Thumberg type cuvette was examined by oxidation of ferrocytochrome c on addition of a catalytic amount of the fully oxidized enzyme (0.02-0.2 M). The amount of ferrocytochrome c oxidation was less than half the oxidation equivalents carried by the enzyme added and essentially proportional to the amount of the enzyme added. It has been shown that this enzyme quickly receives 2 electron equivalents per enzyme molecule from ferrocytochrome c under anaerobic conditions (1). Thus, the above result indicates essentially complete anaerobiosis inside the cuvette.
For the reductive titration with NADH-PMS system, NADH (Sigma, preweighed vial) and PMS solutions in 0.1 M sodium phosphate buffer, pH 7.4, were added using a 10-l Hamilton gas-tight syringe to the enzyme solution deaerated in the Thumberg type cuvette, as described above. Both solutions were prepared daily. O 2 in both solution was not removed, since the oxidation equivalents carried by O 2 introduced into the Thumberg type cuvette with these solutions are negligible compared with the oxidation equivalents in the enzyme to be titrated. The calculated amount of O 2 introduced by the NADH solution for complete reduction of 4.6 oxidation equivalents in the enzyme contains 0.3 oxidation equivalents. However, as described below, the present work showed that O 2 in small volume of the air-saturated water placed inside the cuvette was readily diffused into the gas phase in the cuvette if the water was not directly added into the solution containing enough amount of reactant to trap O 2 . The diffused O 2 at extremely low level in the gas phase did not readily reacts with cytochrome c oxidase. Thus, most of O 2 in the NADH solution are likely to be diffused into the gas phase when the NADH solution is dispensed on the bottom of the titrant arm ( Fig. 1A, a), before mixing the NADH solution with the enzyme solution. And the volume of the NADH solution introduced indicates that the possible level of O 2 due to the diffused O 2 from the NADH solution is too low for the enzyme to react readily with. Therefore, the titration results were not corrected for O 2 introduced by the NADH solution. The maximal amount of O 2 carried by PMS solution corresponds to 0.01 oxidation equivalent to the enzyme in the cuvette. The method for the reductive titration of highly concentrated enzyme solution was described previously (7).

RESULTS
Metal Analysis-The effect of repeated crystallization on the metal content of cytochrome c oxidase was examined. The purified preparation, stabilized with C 12 E 8 before crystallization, was washed with 2 mM EDTA in 50 mM sodium phosphate buffer, pH 7.4, using an Amicon Diaflow apparatus five times and wet-ashed for the determination of iron, copper, magnesium, and zinc atoms by atomic absorption spectrometry. The amount of enzyme in the sample was evaluated by the ␣-band absorption of the dithionite reduced form. The contents of copper, magnesium, and zinc were not significantly affected by the crystallization, whereas the iron content decreased significantly on repeated crystallization (Table I). The atomic ratio of iron to copper, which was 1.0 before crystallization, decreased to 0.69 after the second crystallization. Further crystallization did not affect the iron content, indicating absence of contaminant iron. In Table I, the sample crystallized three times is presumed to be free from contaminant iron and contains 2 iron atoms per enzyme molecule. Closely similar results were obtained for the preparation stabilized with C 12 E 23 . Averaged metal composition for seven different preparations of twice crystallized enzyme, stabilized with C 12 E 8 or C 12 E 23 (Table I), indicates the atomic ratio of Fe:Cu:Mg:Zn in the enzyme preparation to be 2:3:1:1. The iron content gives a molecular extinction coefficient (per two hemes) for the absorption of the ␣-band region of the fully reduced form (⌬⑀ 604 -630 nm red ) of 46.6 mM Ϫ1 cm Ϫ1 , with a standard error of 1.16 mM Ϫ1 cm Ϫ1 (Table I). The wet-ashing and the large quantity of the sample (20 mg of protein for each determination) are critical for the metal analysis at this high accuracy. The spectra of the fully oxidized and reduced forms of the enzyme free from contaminant metal are given in Fig. 2.
A preparation stabilized with DM allows for more effective purification by crystallization. The metal content of the oncecrystallized preparation was about the same as that for the twice-crystallized sample stabilized with C 12 E 8 or C 12 E 23 .
The iron contaminant is likely to be associated with a contaminant protein, since repeated crystallization effectively removes contaminant peptide bands found in SDS-polyacryl-amide gel electrophoresis patterns (data not shown). The spectral difference between a noncrystallized preparation and a once-crystallized preparation stabilized with C 12 E 8 reveals that about 40% of the contaminant iron removed by the initial crystallization is due to contamination by cytochrome bc 1 complex. On the other hand, the contaminant iron left in the once-crystallized sample does not have strong absorption in the visible region.
The enzyme preparations described herein were highly active (250 s Ϫ1 in terms of molecular activity), and no significant effect on the enzymic activity was detectable by the crystallization.

FIG. 1. The anaerobic titration system including the Thumberg type cuvette.
A, the cuvette shown in two sides. a, a port for titrant; b, a rubber septum on the end of the titrant port; c, a female joint with male joint on one end; d, a silicon rubber septum; e, a port for introducing the enzyme solution to be titrated; f, a Pyrex glass cuvette; g, a two-way stop cock for connection to the vacuum system given in C. The space inside the femalemale joint is filled with deaerated water. B, a sketch for the cuvette with the titrant syringe. C, the vacuum line system for the Thumberg type cuvette. a, a female joint for placing the Thumberg type cuvette; b, spiral stainless steel tubing; c, a T-shaped stainless steel tube; d, metal two-way stop cocks; e, a tube containing catalyst for complete removal of trace amount of O 2 in N 2 ; f, a nitrogen tank for ultra pure nitrogen gas; g, a water trap; h, a vacuum pump; i, vacuum rubber tubing.
Titration with Dithionite-The once crystallized preparation stabilized with C 12 E 8 was titrated with dithionite under the anaerobic conditions as described above. The absorbance change was completed within 20 min. The spectrum was also taken at 30 min to confirm that no further change had taken place. Fig. 3 shows the spectral changes in the Soret, ␣-band, and near infrared regions during reductive titration using the once crystallized enzyme solution stabilized with C 12 E 8 . The fractional changes in absorption spectra due to reduction are essentially constant regardless of the wavelength region as indicated by the titration curves (Fig. 3, insets) and show that 6.2 electron equivalents are required for complete reduction of the enzyme. The slope of each titration curve in the initial 1-2 electron equivalents is about 1/3 of that in the latter part of the titration curve. As given in Table II, the average of the end point values for eight determinations is 6.3 electron equivalents, with a standard error of 0.27. The smaller absorbance changes in the initial part of the reductive titration are not due to incomplete anaerobiosis of the titration system, since the titration curves were independent of the number of evacuation-N 2 saturation cycles between 3 and 10 times. On the other hand, when the cycle of evacuation and N 2 saturation was applied only once or twice, both slopes of the biphasic titration curve decreased to give higher end point value.
After the reductive titration, the fully reduced enzyme was reoxidized with air-saturated water (Fig. 4A). The oxidative titration curve showed an initial lag phase where no oxidation of the enzyme was detectable. The oxidation equivalent consumed for the lag phase corresponded to the excess dithionite in the initial reductive titration. The oxidation proceeded monophasically with increasing amounts of O 2 added, in contrast to reductive titration. Compared with the spectrum before the reductive titration, the final reoxidized spectrum had a slightly weaker Soret band with a slightly higher absorbance in the region from 440 nm to 460 nm and a slightly higher ␣-band, giving peaks at 605 nm and 444 nm and a trough at 420 nm in the difference spectrum (Fig. 4C). The O 2 -oxidized enzyme was fully reduced monophasically with 5.2 electron equivalents of dithionite (Fig. 4B, Table II). The end point value was significantly smaller (about 1 electron equivalent) than that of the initial reductive titration. Titration curves for the Soret region (⌬A 444 -480 nm ) and near infrared region (⌬A 750 -805 nm ) coincided with that of the ␣-band spectrum given in Fig. 4, within the experimental accuracy.
The O 2 -oxidized form as well as partially reduced form during the oxidative titration showed a slow and small spectral change indicating reduction of the enzyme, which corresponded to about a 3% reduction in 1 h under anaerobic conditions. The slow reduction proceeded further, up to a 15% reduction in 24 h. The spectral change on addition of O 2 in the oxidative titration was significantly faster than that on addition of dithionite in the reductive titration. A stable spectrum was obtained within 10 min after each addition of O 2 -saturated water, so that it took about 1 h to obtain the whole oxidative titration  curve. Thus, the "autoreduction" was negligible in the oxidative titration.
When the O 2 -oxidized sample was equilibrated with air by opening the port used to introduce the enzyme solution (Fig.  1A, e), the spectrum gradually moved toward the original spectrum preceding the initial reductive titration. The spectral change was essentially completed within 30 min after addition of excess O 2 . The resulting spectrum in the visible-Soret region closely resembled that of the enzyme before the initial reductive titration. The O 2 -oxidized enzyme, exposed to air for 2 h, provided a reductive titration curve with dithionite identical with that in the reductive titration of the fully oxidized form as prepared. As is well known, bovine heart cytochrome c oxidase as prepared is in the fully oxidized form, which shows significantly lower reactivities to ferrocytochrome c and cyanide compared with the fully oxidized form under turnover conditions. This fully oxidized form is alternatively called "resting oxidized form," and the oxidized form under turnover conditions is called "oxygen-pulsed form" (8). In this paper, the fully oxidized orm as prepared denotes the resting oxidized form.
The difference spectrum between the O 2 -oxidized form and the oxidized form as prepared (Fig. 4C) is similar to the redox difference spectrum of heme a, which is characterized by a stronger ␣ band versus Soret band compared with that for heme a 3 (9). However, the O 2 -oxidized form is also in a fully oxidized state since the form does not react with O 2 in the time scale for the enzymic turnover. Thus, this spectral difference must be due to a difference in the coordination structure of the O 2 reduction site.
Quantitative addition of O 2 to the fully reduced enzyme was possible only when the air-saturated water was added directly to the reduced enzyme solution without exposure to the gas phase. When the air-saturated water was dispensed on the L-shaped arm of the titrant port, followed by the addition of reduced enzyme solution to the titrant solution (about 10 l), the reproducibility and the spectral change due to oxidation of the enzyme were definitely more reduced than when the titrant was added directly. This result indicates that a significant part of the O 2 in the air-saturated water dispensed to solution of the enzyme in the O 2 -oxidized state on the titrant arm diffuses to the gas phase in the Thumberg type cuvette. The residual O 2 in the gas phase would react slowly with any reducing equivalent introduced as in the case of the reductive titration under incomplete anaerobiosis. Thus, the second reductive titration has no initial phase for consuming the excess O 2 (Fig. 4, A and B) and the residual O 2 in the gas phase may interfere with progress of the reduction with dithionite. Some part of the extra O 2 could react with O 2 -oxidized form to form the fully oxidized form, equivalent to that before initial reductive titration. These two factors are likely to yield a slightly higher end point value as opposed to the true end point value for the reductive titration of the O 2 -oxidized form. The number of electron equivalents required for complete reduction of the O 2 -oxidized form therefore should be identical with the number of oxidation equivalents required for producing the O 2 -oxidized form from the fully reduced form.
The apparatus for highly concentrated protein solutions in which no evacuation is required (7) was used for anaerobic titration of the preparation stabilized with DM, since extra DM added to the medium (0.2%) for stabilizing the micelles including the enzyme molecules interferes with complete removal of O 2 by the above procedure. Enzyme preparations stabilized with C 12 E 8 and C 12 E 23 require no extra detergent in the medium, because the critical micelle concentration is low enough. The reductive titration results are fully consistent with those for the preparation stabilized with C 12 E 8 obtained by the system including the Thumberg type cuvette (Table II). The preparation solubilized with C 12 E 8 at 0.7 mM gave titration results identical to those at 7.5 M. These results indicate that the redox properties of bovine heart cytochrome c oxidase are independent of the detergent species and of the protein concentration (Table II).
Absorption Spectral Changes during Redox Titration-As shown in Fig. 3, changes in the difference in the absorbance between 444 and 480 nm during reductive titration with dithionite parallels quite well the increase in absorption of the ␣-band and the decrease in the near infrared region. However, the fractional decrease in the absorbance at 416 nm is slightly, but significantly, larger than the fractional increase in absorbance at 604 nm at any electron equivalent added, except for the initial segment of the titration curve (Fig. 5A). The fractional increase in absorbance at 444 nm is significantly smaller than that at 605 nm in the initial half portion of the titration curve (Fig. 5A). On the other hand, the fractional changes at these The arrows indicate the directions of the spectral changes with increasing amount of dithionite added. The absolute spectra are given for the Soret and ␣-band regions, while the difference spectra versus the spectrum in the fully reduced state are given in the near infrared region for correction of the back ground. Insets in the panels show the changes in the absorbance difference between two wavelengths as given in each inset.
wavelengths coincide well with each other in the oxidative titration with O 2 (Fig. 5B) and in the rereductive titration of the O 2 -oxidized form (data not shown). The fractional absorbance changes in the near infrared region paralleled well that at 604 nm in all the above cases.
The shape of the spectral change, induced by the initial 1 electron equivalent in the reductive titration of the fully oxidized form as prepared, was significantly different from that induced by the last 1 electron equivalent before the end point was reached (Fig. 5, C and D). The most obvious difference was in the ratio of the maximum intensity at 444 nm to the minimum intensity at 420 nm. The ratios were approximately unity for the absorbance change in the initial 1 electron equivalent and 0.4 for the absorbance change in the final 1 electron equivalent. The difference was consistent with the absence of an isosbestic point in the regions near 432 and 560 nm, where the spectra in different oxidation states intersect.
The spectral changes induced by the addition of various amounts of O 2 to the fully reduced form, or of dithionite to the O 2 -oxidized form were essentially the same as those induced by the addition of the last 1 electron equivalent in the reductive titration of the fully oxidized form as prepared (Fig. 5D).
Titration with NADH-PMS System-The reductive titration curve using NADH and a catalytic amount of PMS indicated that 4.6 electron equivalents were required for complete reduction of the fully oxidized enzyme as prepared (Table I) (Fig. 6). However, the O 2 -oxidized form that had been equilibrated with air for at least 30 min reacted with cyanide as slowly as the fully oxidized form as prepared. As stated above, a spectrum closely resembling that of the O 2 -oxidized form was obtained by treatment of the fully oxidized form as prepared with 1 M PMS. The PMS-treated form reacted with 200 M cyanide as fast as the O 2 -oxidized form (Fig. 6).
The enzyme solution in 77% oxidized state in the oxidative titration reacted just as fast with cyanide at 200 M as the O 2 -oxidized form, indicating that the reducing equivalent (1 electron/enzyme) does not influence the reactivity of the cyanide binding site, i.e. the O 2 reduction site.
Effect of Contaminant Metals on Redox Titration-As seen from Table I, the once crystallized enzyme contains contaminant iron at a level of 10% of total iron. However, the twice crystallized enzyme, which is free from contaminant iron, gave the same reductive titration curve with dithionite as the once crystallized enzyme which was used for the present titration experiments. The kinetics of the spectral change after each addition of titrant was not affected by the removal of the residual contaminant iron. However, the kinetics of the spectral change was significantly influenced by the contaminant iron in the non-crystalline preparation. A non-crystalline sample stabilized with C 12 E 8 showed a biphasic absorbance increase in the ␣-band, with a rapid initial increase within 10 min followed by a slow increase in absorbance. No stable spectrum was obtained even 1 h after the addition of dithionite. An approximate reductive titration curve, drawn from the spectrum obtained at 70 min after each addition of dithionite solution was monophasic without any initial lag phase (data not shown). The end point was significantly (1.0 electron equivalent) higher than that of the crystalline preparation. These results indicate that at least part of the contaminant iron is redox active, which is consistent with a spectrum of the contaminant fraction showing the presence of cytochrome bc 1 complex, as described above. Furthermore, fractional spectral change in near infrared region showed a sigmoidal titration curve, giving a definitely smaller percentage reduction in the initial half of the reductive titration (at most, in 16% reduction) than those in ␣ and Soret regions. Spectral changes in the visible-Soret region were essentially parallel with each other. The delay in reduction of Cu A is likely to be due to a redox interaction between the contaminant metalloproteins and the enzyme. Table III summarizes the extinction coefficients reported thus far. The extinction coefficients based on iron content of non-crystalline preparations of the enzyme (4, 9 -13) are significantly lower than the value determined in the present study, suggesting that non-crystalline preparations from other laboratories also contain contaminant iron.

Purity and Molecular Extinction Coefficient of Cytochrome c Oxidase-
Anaerobic Redox Titration System-Cytochrome c oxidase is extremely reactive to molecular oxygen (8,13). However, complete removal of O 2 from such a membrane protein preparation stabilized in aqueous medium with detergent as this enzyme is extremely difficult. Thus, one of the objectives of the study was to improve the system originally designed by Burleigh et al. (5) for cytochrome c oxidase solubilized with detergent. The following three improvements are critical for the accuracy and the reproducibility of the titration experiments. The rubber septum for inserting the needle of the titrant syringe in the titrant port of the Thumberg type cuvette was covered with deaerated water. Rubber material was not used in any of the vacuum lines for introducing pure N 2 except for the line connecting the vacuum pump and the water trap. The glass-joints in the system were sealed with Apiezon grease.
A Peroxide in the Fully Oxidized Form as Prepared-The present results indicate that the overall oxidation state of the O 2 -oxidized form is 2 oxidation equivalents lower than that of the fully oxidized form as prepared. However, the O 2 -oxidized form has no reducing equivalent, i.e. all of the redox active metal sites are in the oxidized state since the reactivity of the form to O 2 was negligible in the time scale for the enzymic turnover. The fully oxidized state as prepared was regenerated when the O 2 -oxidized form was exposed to an excess of O 2 . Thus, the ligand that has 2 oxidation equivalents in the fully oxidized form as prepared is most likely to be an O 2 derivative, e.g. O 2 2Ϫ . The electrons for reduction of O 2 to O 2 2Ϫ may be donated via internal electron transfer to the metal sites, which were detected in the O 2 -oxidized form under anaerobic conditions. The internal electron transfer is much slower (3% reduction in 1 h) than the rate of formation (t 1/2 of about 20 min) of the fully oxidized form as prepared. However, O 2 , a strong oxidant, could enhance the electron transfer.
The regenerated fully oxidized form (i.e. the O 2 -oxidized form exposed to air) requires only 6 electron equivalents (not 8 electron equivalents) for complete reduction. The result indicates that the electron donors, after donating electrons to O 2 to form O 2 2Ϫ , do not receive electrons from dithionite, i.e. the redox potentials of these electron donors are apparently far lower than that of dithionite. However, some amino acids such as tryptophan, tyrosine, and lysine could serve as a 2-electron donor to provide a stable non-radical product, which is not reducible with dithionite. The 2-electron donation process to O 2 provides an irreversible modification of the amino acid residue at the electron donor site. The chemical modification could induce serious damages in the enzyme function. The stability of the fully oxidized form as prepared, which is the O 2 2Ϫ -bound form, indicates that the O 2 2Ϫ stops the internal electron transfer from the protein moiety. Thus, a possible physiological role of the stable O 2 2Ϫ -bound form is to prevent modification of amino acid residues when electron transfer is limited from the upstream of this enzyme in the respiratory chain. The electrons for forming O 2 2Ϫ under physiological conditions could come from an external reductant, in which no chemical modification would be caused in the protein.
As described above, in the presence of PMS, 4 electron equivalents from NADH are enough for complete reduction of the enzyme. This result indicates that PMS stimulates an internal electron transfer process to reduce (or remove) O 2 2Ϫ in the fully oxidized form as prepared, and that the 2-electron process provides a non-radical oxidative derivative of an unidentified amino acid residue which is non-reducible with NADH. Many amino acid residues in the enzyme, though unidentified, could be the 2-electron donor sites, as in the case of the formation of O 2 2Ϫ from O 2 . No significant modification in the absorption spectrum as well as the enzymic function has been detected in the PMS-treated enzyme, compared with the O 2 -oxidized form. Similarly, the fully oxidized form regenerated by treatment of the O 2 -oxidized form with excess O 2 has the absorption spectrum and the function strictly identical with the fully oxidized form as prepared. Furthermore, the effects of PMS and O 2 appear fairly slowly (in the time scale of 30 min or so). These findings suggest that the unknown modification sites are far apart from the active center of the enzyme, at least, after the first treatment by PMS or O 2 . This enzyme may have a pool of such amino acids for reducing radical species accidentally produced near the O 2 reduction sites. The direct electron donors to the O 2 reduction site stimulated by PMS or O 2 could be very near the active site. However, the radical species produced could be readily reduced with amino acids in the pool remote from the active site, for preserving the integrity of the active site.
The peroxide is most likely to be situated between Cu B and Fe a3 as a bridging ligand, which has been shown recently in the crystal structure of the fully oxidized enzyme (2). However, the electron density at 2.3-Å resolution does not exclude the possibility that O 2 instead of O 2 2Ϫ is bridged between the two metals. Thus, the present study excludes the possibility that it is O 2 .
Absorption Spectral Changes-The fractional change in absorption in the transition between the fully reduced form and the O 2 -oxidized form is independent of wavelength, i.e. two spectrally independent species are sufficient to account for the spectral transition. This result indicates the following two possibilities: (a) the enzyme system in any oxidation state between the O 2 -oxidized state and the fully reduced state contains only the two extreme forms in various ratios, and (b) all of the redox active metal sites in an overall oxidation state have an identical redox potential, and the potential depends on the oxidation state. The reactivity of cyanide to the enzyme in 77% oxidized state suggests the absence of the fully oxidized form as prepared in the partially reduced state. On the other hand, it has been shown that the number of cyanide sensitive site is independent of the oxidation state between the fully oxidized state and the 3-electron-reduced state (7). This result indicates the absence of fully reduced form in the partially reduced preparation since the fully reduced form has much weaker reactivity to cyanide than those of the fully oxidized and partially reduced forms. These results indicate that possibility b given above is the case, suggesting an extremely tight negative cooperativity between the redox-active metal sites, which is consistent with the results of x-ray structure studies showing close proximity in the locations of the four redox active metal sites (14 -16). The equipotential state of the four metal sites in any overall oxida- tion state suggests an extremely facile electron transfer between these metals.

Comparison of the Present Results with the Redox-coupled Spectral Changes
Reported thus Far-The spectral changes of bovine heart cytochrome c oxidase in relation to the oxidation state in the visible-Soret and near infrared regions, have been reported by several groups (7,12,(17)(18)(19)(20)(21). All of them concluded that 4 electron (or oxidation) equivalents are required for complete reduction (or oxidation) of the fully oxidized (or reduced) form. However, using our extinction coefficient, their results give 5 equivalents in stead of 4 equivalents. Five electron equivalents by NADH-PMS system and by ferricyanide (17,18,21) seem slightly higher than our results: 4.6 by NADH-PMS system and 4.5 equivalents by O 2 , respectively (Table II). The small differences may be due to contaminant iron in their non-crystalline preparation. The reported end point value of a reductive titration with dithionite (19), recalculated with our extinction coefficient, gives 5 electron equivalents, which is significantly lower than our value, 6.3 Ϯ 0.27 equivalents. Their titration curve has no initial lag phase. The inconsistency is likely to be due to incomplete occupancy of the bridging peroxide on the O 2 -reduction site of their enzyme preparation.
PMS provides the fully oxidized form corresponding to the O 2 -oxidized form. Thus, redox mediators used in potentiometric titrations are also likely to give the O 2 -oxidized form. Thus, all of the redox titrations reported thus far except for the dithionite titration (19) correspond to the present titration between the fully reduced form and the O 2 -oxidized form. None of the reported titrations shows parallel fractional absorbance changes, in contrast to the present results, indicating weaker interactions among the four metal sites in non-crystallizable enzyme preparations than in crystalline preparation. The weaker interactions could be caused by modification of the intrinsic three-dimensional structure of the enzyme. Crystallization is effective in removing partially denatured protein, if present, from isolated cytochrome c oxidase preparation. Our preparation before crystallization showed a delay in fractional absorbance change in the near infrared region and monophasic reductive titration curves for the absorbance changes in the visible-Soret region, suggesting that contaminant metalloproteins could influence the titration curve of the integral (or crystallizable) enzyme. Thus, redox titration of cytochrome c oxidase in mitochondria or submitochondrial particles could be influenced by coexisting various metalloproteins to provide non-parallel titration curves.
It should be noted that the difference between dithionite and NADH-PMS titrations has never been recognized until this work. After the discovery by Antonini et al. (8) that the fully oxidized form as prepared is not involved in the enzymic turnover, many models for the structure of the form has been proposed. However, none of the proposed models (22)(23)(24)(25)(26)(27) is consistent with the peroxide bridge between the two metals in the O 2 reduction site.