Two sites of interaction of anions with cytochrome a in oxidized bovine cytochrome c oxidase.

An interaction between cytochrome a in oxidized cytochrome c oxidase (CcO) and anions has been characterized by EPR spectroscopy. Those anions that affect the EPR g = 3 signal of cytochrome a can be divided into two groups. One group consists of halides (Cl-, Br-, and I-) and induces an upfield shift of the g = 3 signal. Nitrogen-containing anions (CN-, NO2-, N3-, NO3-) are in the second group and shift the g = 3 signal downfield. The shifts in the EPR spectrum of CcO are unrelated to ligand binding to the binuclear center. The binding properties of one representative from each group, azide and chloride, were characterized in detail. The dependence of the shift on chloride concentration is consistent with a single binding site in the isolated oxidized enzyme with a Kd of approximately 3 mm. In mitochondria, the apparent Kd was found to be about four times larger than that of the isolated enzyme. The data indicate it is the chloride anion that is bound to CcO, and there is a hydrophilic size-selective access channel to this site from the cytosolic side of the mitochondrial membrane. An observed competition between azide and chloride is interpreted by azide binding to three sites: two that are apparent in the x-ray structure plus the chloride-binding site. It is suggested that either Mg2+ or Arg-438/Arg-439 is the chloride-binding site, and a mechanism for the ligand-induced shift of the g = 3 signal is proposed.

The respiratory heme-copper oxidases constitute a superfamily of terminal oxidases in both prokaryotic organisms and the mitochondria of eukaryotic cells. Cytochrome c oxidases (CcO) 1 catalyze the reduction of molecular oxygen to water using the reducing equivalents supplied by ferrocytochrome c. Four redox centers of CcO are involved in promoting electron transfer (ET) from cytochrome c to dioxygen. Cu A and cytochrome a are primary electron acceptors, and electrons from these sites are delivered to the binuclear center of CcO, consisting of cytochrome a 3 and Cu B , where oxygen is reduced to water.
ET is coupled to the generation of transmembrane proton gradient. Two different processes contribute to the formation of this gradient. The first is the oxidation of cytochrome c from the cytosolic side with the protons required for water formation taken from the matrix side. The second process involves active proton translocation from the matrix to the cytosolic side and is referred to as proton pumping.
Substantial insight into the catalytic mechanism of the enzyme has been obtained from studies of the interaction of ligands with CcO. Most of these studies are devoted to the direct interaction of external molecules with the redox active centers. However, there are sites distant from the redox centers that may well be involved in catalysis. One such site is a surface segment from Gly-49 to Asn-55 on the cytosolic side of subunit I of bovine CcO. It has been established that this region undergoes a redox-coupled conformational change (1,2), and a surface site, Asp-51 of subunit I, sensitive to the redox state of cytochrome a, was identified as essential for proton translocation (3,4). A second intriguing distant site is the nonredoxactive magnesium located at the bottom of a well connected to the cytosolic side of the membrane. Magnesium has been suggested to participate in the exit of water from CcO (5). This suggestion is consistent with the present evidence showing that water departs the catalytic center via a discrete pathway involving magnesium (6,7).
The experimental identification of distant sites that may have a physiologically relevant interaction with the redox center(s) brings a new aspect for both the consideration and exploration of these sites in the catalytic mechanism. At least one site has been implicated by the modulation of the EPR signal of oxidized cytochrome a by several anions (8 -12). Because the binuclear center, composed of cytochrome a 3 plus Cu B , is the established site where external ligands are bound directly, it led to the idea that the EPR signal of cytochrome a is affected from this center. However, our previous work on ligand binding suggested the interaction of cytochrome a with azide is not mediated by the cytochrome a 3 -Cu B center and that there has to be at least one additional binding site (13). This site was recently located in the crystal structure of the enzyme-azide complex and shown to be in the proximity of cytochrome a on the hydrophobic surface of the enzyme within the membrane (14).
In this work we have used EPR spectroscopy to classify anions into three groups via their effect on the low field g-value of cytochrome a in oxidized CcO. One group consists of very weakly or noninteracting anions (phosphate, sulfate). The anions from the other two groups induce opposite shifts of the g ϭ 3 signal. A detailed characterization of the interaction with chloride and azide shows there is a novel chloride-binding site in native oxidized CcO.
Bovine heart mitochondria and CcO were isolated using the method of Soulimane and Buse (15) with a minor modification (16). Briefly, mitochondria were subjected to protein extraction twice with Triton X-100 and K 2 SO 4 . The second extract contained solubilized CcO and was further purified on a Sepharose Q fast flow column. In this study, we made some additional changes. EDTA (1 mM) and histidine (1 mM) was added to the original extraction medium (10 mM Tris, pH 7.6, 250 mM sucrose) to minimize binding of transition metals; the EDTA was omitted for the second extraction. After loading on the column the second extract was washed with 1 liter of 10 mM Tris, pH 7.6, 1 mM histidine, and 0.1% Triton X-100 before eluting with a sulfate gradient. (Either K 2 SO 4 or Na 2 SO 4 can be used without any noticeable differences in the binding of anions to isolated CcO.) The preparation of mitochondria depleted of cytochrome c was based on the earlier methods (17,18). Mitochondria were isolated from fresh beef hearts (15) and washed twice with 10 mM Tris, pH 7.6, and 250 mM sucrose. The washed mitochondria were diluted to a protein concentration of about 3.3 mg/ml with cold 10 mM Na 2 SO 4 , stirred at 4°C for 30 min, and centrifuged at 26,000 ϫ g for 30 min. Cytochrome c was extracted by suspension of the sediment in cold solution of 100 mM Na 2 SO 4 and 10 mM K 3 Fe(CN) 6 . This suspension was stirred at 4°C for 10 min and centrifuged at 48,000 ϫ g for 25 min. To ensure that extraction of cytochrome c was complete, this step was repeated four times using 100 mM Na 2 SO 4 with ferricyanide omitted and three times with 10 mM Hepes, pH 7.6, containing 50 mM Na 2 SO 4 . These mitochondria were used directly for chloride binding measurement at pH 7.6. For experiments at pH 6.8, the pH of the mitochondrial suspension was lowered by small additions of dilute H 2 SO 4 .
Enzyme concentration was determined at pH 8.0 from the absorbance of isolated oxidized enzyme at 424 nm using A ϭ 158 mM Ϫ1 cm Ϫ1 . Enzyme concentration in the mitochondria was determined on mitochondria clarified with 0.1% dodecyl maltoside from the absorption of dithionite-reduced CcO⅐CN minus oxidized CcO⅐CN using ⌬A 606Ϫ620 ϭ 18.6 mM Ϫ1 cm Ϫ1 for cytochrome a (19).
To determine the dependence of CcO catalytic activity on NaCl concentration, both purified enzyme and mitochondria solubilized with 0.1% DM were used. The oxidation of 10 M ferrocytochrome c by 4.5 nM CcO was monitored at 550 nm and 22°C (10 mM Tris, pH 7.6, 0.1% DM). The ionic strength of the solution was held constant (I ϭ 210) using mixtures of Na 2 SO 4 and NaCl.
The midpoint potential of cytochrome a was estimated in the CcO⅐CN under argon atmosphere using cytochrome c present as a redox indicator (E obs 0 ϭ 264 mV at pH 7.7 and 23°C) (20). The midpoint potential of cytochrome a was calculated from the equilibrium ratios of (cytochrome a 3ϩ )/(cytochrome a 2ϩ ) and (cytochrome c 3ϩ )/(cytochrome c 2ϩ ). These ratios were obtained from optical spectra after addition of 12.8 M ferrocytochrome c to 2.6 M CcO⅐CN in 10 mM Tris, pH 7.6, and 0.1% Triton X-100. The CcO⅐CN had been depleted of free cyanide by gel filtration.
To determine the effect of chloride on the rate of cytochrome a reduction by ferrocytochrome c at constant and high ionic strength, the CcO⅐CN complex was used once more. Typically, 3 M CcO⅐CN was mixed with 29 M reduced cytochrome c in the Gibson-Durrum stoppedflow instrument at 13°C. The buffer was 10 mM Hepes, pH 7.6, and 0.1% DM, and the ionic strength was maintained at I ϭ 460 by NaCl and Na 2 SO 4 . The high ionic strength and reduced temperature were necessary to detect the electron transfer kinetics by stopped flow (21).
To determine the effect of chloride on intramolecular electron transfer from cytochrome a to cytochrome a 3 , 3.7 M oxidized CcO was rapidly mixed with 40 M ferrocytochrome c plus 0.1 M dithionite in the stopped-flow instrument at 23°C under argon. Under these conditions, a large fraction of Cu A and cytochrome a was reduced in the dead time of the instrument, and, predominantly, the kinetics of electron transfer to the binuclear center was observed. The solution contained 10 mM Ches buffer, pH 9.0, 0.1% DM, and 48 mM NaCl. In the control samples, chloride was replaced by 16 mM Na 2 SO 4 . The stopped-flow measurements were made about 20 min after dissolving stock CcO in the buffer.
All samples for EPR were frozen by quick immersion in a methanol dry ice bath at a temperature of Ϫ58°C and then transferred to liquid nitrogen and stored at this temperature. EPR spectra were recorded at 10 K on a Varian E6 spectrometer. The conditions for measurements were: frequency, 9.26 GHz; power, 3 mW; modulation amplitude, 10 G; and the modulation frequency, 100 kHz.
All dissociation constants were established from the spectral shift of the g ϭ 3 signal in the EPR spectrum of oxidized CcO induced by the binding of anions. Chloride binding to isolated CcO was studied at three pH values: at pH 5.7 (20 mM Mes, 100 mM Na 2 SO 4 , 0.1% Triton X-100), at pH 7. 6 (10 mM Hepes, 0.1% Triton X-100), and at pH 8.8 (10 mM Ches, 0.1% Triton X-100). At pH 5.7, a high ionic strength was necessary to avoid enzyme precipitation. The dependences of spectral shifts on the ligand concentration were normalized before the fit was applied. To normalize the amplitudes, the following assumptions based on the resolutions of the g ϭ 3 peak into two gaussian curves were employed. In mitochondria, the whole population of CcO is in the chloride-bound state at a chloride concentration of 1 M. For purified CcO, about 80% of CcO is in this chloride-bound state in the presence of 20 mM NaCl. In the presence of 20 mM azide, 1 M NaCl converts ϳ70% of purified CcO to the chloride-bound state.

RESULTS
At liquid helium temperatures, the EPR spectrum of oxidized "rapid" CcO only exhibits signals caused by low spin cytochrome a and Cu A . The absorption-like g ϭ 3 signal of oxidized cytochrome a of purified CcO (Fig. 1A) is unaffected by sodium ions over the range 10 -400 mM at pH 7.6. This peak is also insensitive to the addition of 100 mM Na 2 SO 4 or 100 mM sucrose, brief sonication of the enzyme under a stream of nitrogen on ice, repeated freezing and thawing, and storage on ice for 14 h in 10 mM Tris, pH 7.6, 100 mM Na 2 SO 4 , 0.1% Triton X-100 (w/v). In addition, the complex between cytochrome c and CcO prepared by stoichiometric addition of oxidized cytochrome c to CcO at low ionic strength (10 mM Hepes, pH 7.6, 0.1% Triton X-100) did not affect the peak at g ϭ 3. A downfield shift of about 5 G and slight broadening of the peak occurred following the addition of 150 mM sodium phosphate (Table I).
However, we have found anions that markedly influence the g ϭ 3 signal of oxidized CcO. On the basis of the observed spectral shifts, all ligands tested were consequently divided into one of three groups (Table I). The first group, nitrogencontaining ligands (cyanide, azide, nitrite, and nitrate), induces a low field shift of the g ϭ 3 peak of ϳ30 -35 G. The second group contained halides (chloride, bromide, and iodide),  (0)) and in the presence of 1 and 20 mM NaCl. B, the data from the same experiment presented as the difference spectra between CcO samples containing 1, 4, and 20 mM NaCl minus untreated (control) CcO. Buffer: 10 mM Hepes, pH 7.6, 0.1% Triton X-100.
which shifted the g ϭ 3 peak by ϳ17-30 G to a higher magnetic field. The third group was the large anions, phosphate and sulfate, which caused little, if any, shift.
The presence of two opposing spectral effects suggested the presence of two separate binding sites for these ligands. To test this possibility we selected, for further characterization, a representative from each of the first two groups: azide, because it has been located in the vicinity of cytochrome a in the crystal structure of the CcO-azide complex (14), and chloride, because it may have some relevance for CcO catalysis (22,23).
Chloride Binding-It has been demonstrated previously that chloride can bind to the binuclear center of CcO (24 -30). Thus, the observation that the g ϭ 3 peak of cytochrome a is sensitive to chloride raises the question of whether or not the influence of chloride on the EPR spectrum of cytochrome a is a consequence of chloride binding to the catalytic center. That this is not the case was established by the following observations. First, the rate of chloride binding to the site that interacts with cytochrome a is much faster than the rate of chloride binding to the binuclear center (29). The reaction with cytochrome a was complete in Ͻ1 min, the time needed for manually mixing the enzyme and chloride and freezing the EPR sample in a methanol/dry ice slush. Longer incubation of CcO with chloride at room temperature had no additional affect on the EPR spectrum. By contrast, the binding of chloride to the binuclear center occurs on a time scale of hours at pH 7.6 (29). Second, the g ϭ 3 signal observed in the presence of chloride is restored to the signal of the original untreated CcO immediately following the removal of chloride by gel filtration, whereas release of chloride from the binuclear center is an extremely slow process (29). Third, the binding of chloride to the binuclear center is associated with the changes in the optical spectrum (12,27,29) that were not observed when the EPR spectrum of cytochrome a was already modified by chloride. Finally, the influence of chloride and all other ligands tested on the cytochrome a EPR signal was also observed in the complex of oxidized CcO with cyanide bound in the binuclear center.
Increasing the concentration of sodium chloride in samples of purified oxidized CcO led to an increasing amount of enzyme with its g ϭ 3 peak shifted upfield by about 24 gauss (Fig. 1A). Thus, in a sample containing 20 mM NaCl, the EPR spectrum was resolved into two gaussian curves with about 80% of CcO having the upfield shifted signal (chloride-bound) and the remainder chloride-free. Between 0 and 20 mM NaCl the binding process is associated with a decrease of EPR intensity of unreacted enzyme with a parallel increase of intensity because of chloride-bound enzyme. This process can be nicely visualized from the difference EPR spectra of chloride-treated minus untreated enzyme (Fig. 1B).
At concentrations of sodium chloride between 20 mM and 1 M, the effect of chloride was reversed, and the normalized amplitude in the difference spectra slightly decreased ( Fig. 2A). A similar dependence was obtained from normalized areas under gaussian curves ( Fig. 2A). For this purpose, the g ϭ 3 peak at each chloride concentration was resolved to two gaussian curves and the relative area under the curve, corresponding to the chloride-bound state, calculated.
The dependence of normalized amplitudes can be fitted assuming a single chloride-binding site with a K d of 3.5 mM at pH 7.6 ( Fig. 2B). This K d was insensitive to increasing the ionic strength by the addition of 50 mM Na 2 SO 4 , to replacing the detergent Triton X-100 with 0.1% DM, and to changing of the buffer from 10 mM Hepes to 10 mM Tris. The estimated K d for bromide and iodide are 2 and 15 mM, respectively (Table I).
Measurements with chloride at both pH 5.7 and 8.8 gave almost identical results. In both cases, the binding of chloride proceeded in two phases, with the first phase having K d values of 3.4 and 2.5 mM at pH 5.7 and 8.8, respectively. At all three pH values (pH ϭ 5.7, 7.6, and 8.8), raising the concentration of NaCl to 1 M did not convert the g ϭ 3 peak of isolated CcO to a homogenous EPR signal. At this high concentration of NaCl, ϳ70% of the enzyme population exists in the high field-shifted state.
Chloride has a similar effect on the EPR signal of oxidized CcO present in mitochondrial membranes (8). Because the EPR spectra of cytochrome c and cytochrome a overlap in the g ϭ 3 region, we first depleted the mitochondria of cytochrome c as  2. The dependence of the high field shift of the g ‫؍‬ 3 signal on chloride concentration. A, the chloride-induced high field shift for two CcO preparations. Dependence of normalized amplitudes (f) obtained from difference spectra is as described for Fig. 1B. Dependence of the normalized areas under the curve of the chloride-bound state (E) is shown; each g ϭ 3 peak was resolved into two gaussian curves, and the relative amounts of chloride-bound and chloride-free were calculated. Conditions are the same as in Fig. 1 described under "Experimental Procedures." With membranebound CcO, we noticed two differences in the behavior of the g ϭ 3 signal compared with the isolated enzyme. First, there was only the high field shift of the g ϭ 3 signal upon chloride binding; the second phase of chloride binding is missing at both pH 7.6 and 6.8. Second, the dissociation constant is about 4 -5-fold larger than that found with the purified enzyme. From the fits to the data for CcO in mitochondria, a single binding site with K d values of 12.5 and 15.6 mM at pH values of 7.6 (Fig. 2B) and 6.8, respectively, were determined.
The presence of a second phase in chloride binding with isolated CcO, but not with the membrane-bound enzyme, suggests the purification procedure leads to some modification of the enzyme. To examine this possibility, we collected samples of enzyme after each purification step and measured the EPR spectra of the untreated sample and the sample treated with 200 mM NaCl. The data suggested that CcO is modified during chromatography on Sepharose Q. Thus, the longer the time the enzyme spends being washed on the Sepharose Q column (with either 70 mM K 2 SO 4 or Na 2 SO 4 in 10 mM Tris, pH 7.6, and 0.1% Triton X-100), the greater the fraction of CcO that reacts with chloride in two phases.
Competition between the Binding of Azide and Chloride-The crystal structure of the azide complex of bovine cytochrome oxidase reveals a second binding site for azide that is close to cytochrome a and on the surface of the enzyme (14). To test whether this site could also bind chloride, we determined the K d for chloride in the absence and in the presence of 20 mM NaN 3 (Figs. 2B and 3A); we have also compared the affinity of CcO for azide with and without 50 mM NaCl (Fig. 4).
After the addition of 20 mM NaN 3 , the g ϭ 3 signal of isolated CcO showed two closely positioned maxima (Fig. 3A), with a fraction of the signal unchanged and the remainder shifted downfield by about 35 G; this curve shape (as shown in Fig. 3) persists up to an azide concentration of 1 M. The fraction of CcO with an EPR spectrum almost unchanged from that of untreated enzyme could be interpreted as a population of enzyme that does not react with azide (Fig. 3A). That this is not the case was readily demonstrated. The addition of 10 mM sodium chloride to azide-free enzyme produced the expected upfield shift of the EPR spectrum (Fig. 3B), whereas addition of this same concentration of sodium chloride to the enzyme pretreated with sodium azide left the azide-modified EPR spectrum unchanged (Fig. 3A).
The response of CcO to azide was again indicative of the presence of two populations of cytochrome oxidase, observed in the interaction of chloride with CcO and attributed to the modification of enzyme during isolation. As before, enzyme samples were collected at each stage of the purification and reacted with 200 mM azide. It was again observed that it was mainly chromatography on Sepharose Q that led to the development of this inhomogeneous response to ligand binding. It can be seen that the presence of azide decreases the affinity of CcO for chloride (Fig. 3). To estimate quantitatively the effect of azide, we again measured the high field shift of the g ϭ 3 signal produced by raising the concentration of NaCl. Fitting the normalized plot of the signal increase on the high field side of the g ϭ 3 peak resulted in a single binding site with a K d of 70 mM (Fig. 2B). The same value was determined from a plot of the signal decrease at the low field side of the g ϭ 3 peak. The result indicated that the two populations of CcO, characterized by the distinct maxima observed in the presence of 20 mM azide, showed the same affinity for chloride. This affinity was about 20-fold smaller than that determined in the absence of azide.
At this point it would appear that our data can be explained in two ways: (i) there is only one ligand-binding site, and chloride and azide compete for it, or (ii) there are two different sites, but there is an interaction between them such that when azide occupies its specific site, the affinity of the second site for chloride is decreased 20-fold.
To examine these alternatives we compared the affinity of CcO for azide in the presence and absence of 50 mM NaCl (Fig.  4). Surprisingly, at this concentration of NaCl, there was no reduction in the affinity of CcO for azide, as revealed in the spectra of enzyme treated with 10 mM azide and with and without 50 mM chloride (Fig. 4, A and B). In both cases, azide induced a low field shift of the g ϭ 3 signal. Moreover, it appears the fraction of CcO that underwent the low field shift following reaction with 10 mM NaN 3 was enhanced by the presence of NaCl (Fig. 4, A and B).
The normalized plots of either the signal increase at the low field side of the g ϭ 3 peak (Fig. 4C) or the signal decrease measured at the position of the initial peak maximum gave the same result. Both dependences were fitted well by assuming a single binding site for azide; this site had a K d of 15.6 mM in the absence of chloride (Table I) and 7.5 mM when 50 mM NaCl was present (Fig. 4C).
A relationship similar to that described above for the competition between azide and chloride was also observed between cyanide and chloride. The K d for cyanide binding to oxidized CcO (Table I) was almost unaffected by the presence of 40 mM NaCl. However, the affinity for chloride was substantially decreased when 10 mM NaCN was present.
Chloride Effect on Electron Transfer Rates-To assess whether chloride binding to a site other than the binuclear center might affect the catalytic mechanism in some way, we characterized the effect of chloride on (i) the midpoint potential of cytochrome a, (ii) the catalytic activity of CcO, (iii) the rate of ET from ferrocytochrome c to cytochrome a, and (iv) the rate of intramolecular ET from cytochrome a to cytochrome a 3 . We found the midpoint potential of cytochrome a, estimated at pH 7.6 in the CcO⅐CN complex under argon, was 283 Ϯ 3 mV and was unchanged by the presence of 50 mM NaCl.
At constant ionic strength, both the catalytic activity and the rate of reduction of cytochrome a by cytochrome c decreased with increasing concentration of NaCl. However, in both cases the decrease was a monotonic function of chloride concentration and does not correlate with the K d of 3.5 mM for chloride binding that we determined by EPR spectroscopy. The catalytic activity of solubilized mitochondria decreases almost linearly with increasing concentration of NaCl at constant ionic strength (I ϭ 210); at pH 7.6, the rate constant for oxidation of cytochrome c decreased from 60 s Ϫ1 at 0 mM NaCl to ϳ30 s Ϫ1 at 210 mM NaCl.
A similar effect of NaCl was observed for the rate of reduction of cytochrome a in the CcO⅐CN complex. At high ionic strength, the reduction proceeded in two phases with the faster phase accounting for 70% of the enzyme (21). The fast phase decreased in rate from ϳ33 s Ϫ1 at 0 mM chloride to 10 s Ϫ1 at 480 mM NaCl. The rate constant of the slow phase, 1.9 s Ϫ1 , was found to be practically independent of NaCl concentration.
The rate of ET from cytochrome a to cytochrome a 3 , followed from the absorbance increase at 445 nm, can be described by two exponentials (31). At pH 9.0 and in the absence of chloride, 80% of the enzyme reacts with a rate of 103 s Ϫ1 ; the remainder reacts with a rate of 12.5 s Ϫ1 . In the presence of 48 mM NaCl, the rates were 84 and 12.4 s Ϫ1 , respectively. From the amplitude of the spectral changes, we estimated that about 43% of cytochrome a was also reduced in the first phase in both experiments. We stress here that occupancy of the binuclear center by chloride, which might be considered as the reason for the suppression of the rate of ET, can be excluded under these conditions. At pH 9.0, K d for chloride binding to the binuclear center was sufficiently high, and also the rate of chloride binding (compared with the time required for a realization of the measurement) was too slow to form the complex of chloride with the cytochrome a 3 -Cu B center (29).

DISCUSSION
Identification of the Chloride-binding Site-In the following discussion we will focus on the binding of chloride to cytochrome oxidase that occurs over and above that at the binuclear center, which we and others have characterized in some detail (12,25,26,29,30). Unless explicitly stated, all references to "site" imply the site(s) different from the binuclear center.
The ability of chloride to shift the g ϭ 3 signal of cytochrome a to higher field was noted in earlier work with bovine enzyme (8,10). At pH 7.6, we find that this high field shift can be described by assuming a single binding site with a dissociation constant of 3.5 mM (Table I). This value is identical to that reported previously for the binding of chloride to the bacterial bo ubiquinol oxidase at pH 7.4 (23).
A secondary effect, in which ϳ30% of the high field-shifted g ϭ 3 EPR signal returns to its original position, occurs at very high concentrations of NaCl. This observation implies that purified CcO must have at least two binding sites for chloride that influence the g ϭ 3 signal of cytochrome a. However we have not studied this secondary process in detail, because it is absent in CcO embedded in the mitochondrial membrane (Fig.  2B), and we assume that this very low affinity site develops in a portion of the enzyme during isolation.
The comparison of chloride binding to purified and membrane-embedded CcO shows the native enzyme contains but a single binding site external to the binuclear center, and occupancy of this site induces only a high field shift of the g ϭ 3 peak of cytochrome a (Fig. 2B). In mitochondria, the apparent dissociation constant of this site is about 4-fold larger than that determined for isolated enzyme. However, in both cases, K d is unaffected by changes in pH, and both oxidases exhibit the same high field shift in EPR induced by chloride. We thus conclude that these properties of the binding site are not modified by the purification protocol. The apparent difference in the dissociation constants can be due to several factors. First is the influence of the membrane environment on CcO, second is the alteration of the site during isolation, and third is the thermodynamic activity of chloride in the hydrocarbon phase of the mitochondrial membrane being less than in aqueous buffer. Any of these factors can result in an apparent increase in the K d for chloride binding to the mitochondrial enzyme relative to the purified protein.
Where is this high affinity chloride-binding site located? The only certain fact is that it is exterior to the binuclear center.
The available x-ray structures of the bacterial and bovine oxidases (1, 32-36) provide no direct evidence for chloride binding, but it would almost certainly be confused with one or more of the water molecules that are included in the x-ray refinement. Consequently, we have to rely on indirect evidence in attempting to locate the binding site.
Because azide reduces the affinity for chloride (Figs. 2B and 3), it would seem conceivable the chloride site overlaps with, or is identical to, the second (nonbinuclear center) azide-binding site (14). Azide bound to this site is visible in a space-filling model of the enzyme, and hence the site is accessible from the enzyme surface, although it is at the level of the center of the hydrocarbon phase of the membrane (14). This surface was reported to be covered with detergent (14) (although no detergent molecules are present in the archived Protein Data Bank file (1ocz)), and this led to the conclusion that only hydrazoic acid (HN 3 ) can access the binding site (14). Infrared spectra of azide bound to oxidized CcO (37) indicate that even though it is HN 3 that reacts with the enzyme, azide is bound in the deprotonated state. In the crystal structure of the CcO-azide complex, azide is hydrogen-bonded to Tyr-379 and Asn-422, with His-429 possibly functioning as a proton acceptor (14). Because Tyr-379 is connected to His-378, one of the axial ligands to heme a (31), this suggests an obvious path by which the binding of azide can influence the spectral properties of cytochrome a (see below).
By analogy, we might expect the binding of chloride to Tyr-379 and Asn-422 implies that HCl is the reactive species. However, HCl has a pK a of Ϫ7 (38), so the concentration of the neutral acid at pH 7 would be 10 Ϫ14 M. Even assuming the rate of chloride binding is diffusion-controlled, it would still take Ͼ10 3 s for the reaction to proceed to completion. In reality, the reaction is complete in Ͻ1 min, the time needed to make the EPR sample. It is thus much more credible to assume that it is the chloride anion that is the reactive species; this is also consistent with our finding that the apparent dissociation constant for chloride binding is practically pH-independent, both for the isolated enzyme and for the enzyme in mitochondria.
However, we speculate that the low affinity chloride-binding site is identical with the azide site. As noted, this chloride site develops during purification on the Sepharose Q column and increases with increased washing on this column. It is possible that this washing removes lipid molecules from the hydrophobic surface of the enzyme. This results in a decrease of the hydrophobic barrier, allowing the access of chloride to the azide site, which is otherwise inaccessible in mitochondria.
The above discussion argues that the second azide-binding site identified in the crystal structure is not identical with the higher affinity chloride site. This conclusion is supported by the observation that the downfield shifts in the g ϭ 3 EPR signal induced by the addition of azide to the native enzyme are almost unaffected by the presence of 50 mM NaCl. Thus, both the enzyme pretreated with chloride and the chloride-free enzyme respond to azide in essentially the same way (Fig. 4C), implying the x-ray-defined azide-binding site and the high affinity chloride site are different.
This, in turn, raises the question as to how azide can decrease the affinity of chloride for the chloride-binding site. The most straightforward explanation is that azide can bind to both sites. This explanation is not surprising, because multiple sites for the binding of azide to isolated CcO have been detected by Fourier transform infrared spectroscopy (37).
The existence of two azide-binding sites on the surface of the enzyme also provides an explanation for the larger fraction of the enzyme shifted to the low field by 10 mM azide when high concentrations of chloride are present (Fig. 4, A and B). In the presence of 50 mM NaCl, the chloride site is not accessible to azide because it is already occupied by chloride anion. Azide can bind only to one site, the x-ray-defined azide site, and binding to this site results in a low field shift of the g ϭ 3 EPR signal. However, in the absence of NaCl, azide can bind to both sites. Binding of azide to its site induces the low field shift of the EPR signal, but the binding of azide to the chloride site produces the opposite effect, i.e. a high field shift. The net effect results in a smaller fraction of the g ϭ 3 signal shifted to the low field position compared with the situation when chloride is present.
The following interesting questions arise. What is this high affinity chloride site? What is the mechanism for the upfield shift of the g ϭ 3 signal of cytochrome a induced by chloride?
The observation that the binding of chloride is almost independent of pH implies it is the chloride anion that is bound to the protein and suggests the binding site is positively charged over the pH range 5.7-8.8. As the site is also accessible to chloride added to the mitochondria, it would seem that access to this site has to be from the cytosolic side of the membrane. Although chloride, bromide, and iodide all cause an upfield shift of the g ϭ 3 signal, sulfate and phosphate are very weak or noninteracting anions. This suggests the site has some size selectivity. In addition, the dissociation constant for chloride binding is unchanged by the presence of 50 mM Na 2 SO 4 . Thus, we believe the high affinity chloride site is partially buried within the protein and connected to the exterior by a sizeselective hydrophilic channel.
There are two possible sites that satisfy the above restrictions. The first consists of the conserved Arg-438 and Arg-439 residues of subunit I, and the second is the bound Mg 2ϩ located at the interface between subunits I and II. Both Arg-438/Arg-439 and Mg 2ϩ have a strong electrostatic interaction with the propionate substituents of heme a (39,40), and there is also a strong interaction between these two arginine residues and Mg 2ϩ (39,40). The Mg 2ϩ lies at the bottom of the proposed water/proton channel (32) and is coordinated by Asp-369 and His-368 from subunit I by Glu-198 from subunit II and by a water molecule (32). We suggest the water molecule can be displaced by chloride, and the binding of chloride to Mg 2ϩ is communicated to the propionate side chains of heme a. The feasibility of water displacement is supported by the rapid deuterium exchange at the Mg 2ϩ site observed with the Rhodobacter sphaeroides CcO (6). Likewise, in the bacterial aa 3 oxidases from Rhodobacter and Paracoccus denitrificans, the Mg 2ϩ site can be occupied by Mn 2ϩ when the culture medium is supplemented with Mn 2ϩ (41)(42)(43). The EPR spectrum of the manganese changes in the presence of H 2 17 O (41), again indicative of solvent access to this metal ion. Our view on the interaction of chloride and azide with cytochrome a in bovine CcO is summarized in Fig. 5. Azide, the location of which is defined by the x-ray structure (14), interacts with iron through Tyr-379 and His-378. Chloride is proposed to interact with heme a propionates from the site located at the cytosolic surface of CcO. The chloride site is, however, also accessible to azide, but the azide site in native CcO does not bind chloride.
The participation of the magnesium ion in catalysis has yet to be established. However, evidence is accumulating that Mg 2ϩ might be involved in the water and/or proton exit pathway from the binuclear center (6,7). According to our suggestion, chloride can replace the water molecule bound to Mg 2ϩ , and, in this circumstance, the competition between water and chloride for this site would be relevant for the mechanism of water and/or proton exit. This idea is supported by two observations. First, the chloride concentration within a cell is probably close to 25 mM (44); this is sufficiently high to keep the site in oxidized CcO occupied by Cl. Second, it seems likely that chloride is not permanently bound to CcO during the catalytic cycle. It has been observed that when fully reduced bovine CcO was reoxidized by oxygen in the presence of chloride, the g ϭ 3 EPR signal initially observed was characteristic of the chloridefree enzyme; however, this rapidly changed to that of the chloride-bound enzyme (10). Based on our data and those of Ref. 10, we estimate the rate constants for chloride binding and dissociation from this site to be at least 700 M Ϫ1 s Ϫ1 and ϳ2 s Ϫ1 , respectively. With 25 mM chloride present in the cell, the rate of binding will be about 18 s Ϫ1 . These rates are comparable with the turnover of CcO reported to be between 0.7 and 14 s Ϫ1 in coupled mitochondria (45)(46)(47) or in the cell (48).
From the available compilations on the stability constants of metal-ion complexes (49 -51), it can be seen that the stability constants of fluoride, chloride, bromide, and iodide with Mg 2ϩ , under comparable conditions, decrease in the order F Ϫ Ͼ Cl Ϫ Ͼ Br Ϫ Ͼ I Ϫ . Thus, with the increasing size of the coordinated anion, the stability of the Mg 2ϩ complex decreases. Because we have determined the one dissociation constant for chloride, whereas the constants for Br Ϫ and I Ϫ were simply estimated, the ranking order of stability of halides with Mg 2ϩ in oxidized oxidase could be the same as it is for inorganic complexes.
This order of halides is coincidental with the order of anions in the Hofmeister series (52). However, we have raised the following arguments for the exclusion of the Hofmeister effect in the spectrum of cytochrome a induced by halides. First, the Hofmeister effect is a result of multiple interactions of salts at the waterprotein interface. This is in contrast to our observation indicating that only one site in oxidase interacts with chloride. Second, the Hofmeister effect on protein structure and function becomes important at moderate and high (0.01-1.0 M) salt concentrations. However, we can observe the chloride-induced shift of the g ϭ 3 signal readily at 0.5 mM NaCl. A further point is that there is typically a sign inversion of the Hofmeister effect at about NaCl. It means the chloride effect on proteins is neutral (52). In our case, however, chloride is effectively inducing the change of the EPR spectrum of oxidized oxidase.
The observed shift of the g ϭ 3 signal by iodide is smaller than the shift produced by chloride and bromide. It is feasible that the extent of the shift corresponds to the strength of interaction between a particular ion and the binding site. Thus, the smaller shift of the g ϭ 3 signal might well reflect the decreased affinity of the site for iodide.
It was reported that the rate of intramolecular ET during the reaction of fully reduced bacterial quinol bo oxidase with dioxygen (22, 23) is substantially higher in the presence of chlo-ride than in its absence. A K d of 3 mM was found for this novel chloride-binding site (23), a value almost identical to that which we find for bovine CcO. Our findings also show the rate of ET is dependent on chloride concentration. However, there was no obvious correlation between the dissociation constant for chloride binding as determined by EPR spectroscopy and the variation in the rates of the assessed ET processes. Our measurements showed the overall catalytic activity, the rate of ET from cytochrome c to cytochrome a, and also the rate of intramolecular ET between cytochromes a and a 3 are sensitive to the presence of chloride (at constant ionic strength). In all three cases, a monotonic decrease both in turnover and ET rates occurs upon increasing the concentration of NaCl.
Possible Mechanisms for the Shift in the Low Field g-Value of Cytochrome a-The EPR spectra of low spin bisimidazole hemeproteins, such as cytochrome a and cytochrome b 5 , are well described by the model of Griffith (53). This model assumes the unpaired electron of the ferric ion resides in an orbital composed primarily of d yz , with lesser contributions from d xz and d xy , respectively; the other two d orbitals are assumed to be irrelevant. The three participating d orbitals are ordered energetically as shown in Fig. 6, with d xy being the most stable and d yz the least stable. The separation between d xy and the center of gravity of d yz and d xz is called ⌬. Its value is determined primarily by the field strength of the axial ligands; bisimidazole ligation ⌬ is in the neighborhood of 1200 cm Ϫ1 . The separation between d yz and d xz is called V; its value is determined by the ligand type, and if the two axial ligands are planar, as in bishistidine cytochromes, by the angle subtended by their respective planes ().
In the axial limit ϭ 90°, V is zero, and the g-values are determined exclusively by ⌬. Were ⌬ infinitely large, the low field g-value (g z ) would be 4. For more realistic values of ⌬, the largest value for g z is around 3.8. For example, in the bc 1 complex from yeast, cytochrome b L has a that is very close to 90°, and g z is 3.76 (54,55), whereas cytochrome b H has a close to 80°, and g z is 3.6 (54,55).
At the other extreme, the two imidazole rings are close to parallel. For example, for cytochrome b 5 and cytochrome a, is in the range 5-10°, and V can be as large as 0.5⌬; the low field g-value falls in the range 2.91-3.07 (55).  (53) for the d(t 2g ) 5 configuration of low spin ferric heme, assuming a proper axis system (57). The unpaired electron resides in an orbital that typically has the composition 3, 10, and 87%, respectively, of these three d orbitals. For bisimidazole coordination, ⌬ is typically 3 ( s the spin orbit coupling constant; its value is 460 cm Ϫ1 for the free ferric ion and falls in the range 300 -400 cm Ϫ1 in covalent compounds, such as heme). V can range from 0 to 0.5⌬ (see text for details).
Thus, shifts in the location of g z can occur by one of two mechanisms, changes in the Fe-N his bond strengths or changes in the value of . Increases in bond strength and/or decreases in will lead to high field shifts in g z ; conversely a decrease in bond strength and/or an increase in will lead to a low field shift in g z .
From data mining on the large number of crystal structures of hemeproteins bearing at least one axial histidine, Zaric et al. (56) concluded there were two important factors that determine the orientation of the imidazole ring(s) in hemeproteins.
First, and almost without exception, the plane of the histidine is oriented such that N ␦ points to the propionate side chains. In bovine cytochrome oxidase, the two axial histidines are His-378 and His-61. His-378 is oriented with its N ␦ pointing to the methine bridge located between the two propionate side chains. The sole exception to this conclusion of Zaric et al. (56) is that His-61 is rotated by almost 180°with respect to His-378.
Second, the plane of the imidazole ring is usually perpendicular to the plane defined by C␣-C␤-C␥ of the histidine side chain. This is found to be the case for both His-61 and His-378. Parenthetically, hydrogen bonding to N ␦ does not seem to be important in determining histidine orientation (56).
From the above information, one can deduce that structural changes at both the azide-binding site and the Mg 2ϩ site could affect the low field g-value of cytochrome a. In the first case, the binding of azide to Tyr-379/Asn-422 could affect the protein backbone in that locale, resulting in a movement of the C␣-C␤-C␥ side chain of His-378 with an associated movement of its imidazole ring and an increase in the dihedral angle subtended by His-378 and His-61.
In the second case, replacing the water molecule bound to the Mg 2ϩ by a chloride anion reduces the charge on the metal ion and, thus, the electrostatic interaction of this site with the propionate side chains of heme a. This might occur either directly or via Arg-438 and Arg-439. In either case, the adjustment of the propionate side chains is transmitted to one or both imidazole rings, with a net reduction in the mutual dihedral angle, .
In conclusion, our data show anions that perturb the EPR spectrum of cytochrome a can be divided into two groups, and this suggests the presence of (at least) two anion-binding sites in oxidized CcO. Of these two sites, the more intriguing is the chloride site, and we intend to pursue the extent to which this site may be of importance in the catalytic steps of this enzyme.