The pH dependence of cytochrome a conformation in cytochrome c oxidase.

The pH dependence of the conformation of cytochrome a in bovine cytochrome c oxidase has been studied by second derivative absorption spectroscopy. At neutral pH, the second derivative spectra of the cyanide-inhibited fully reduced and mixed valence enzyme display two Soret electronic transitions, at 443 and 451 nm, associated with cytochrome a. As the pH is lowered these two bands collapse into a single transition at approximately 444 nm. pH titration of the cyanide-inhibited mixed valence enzyme suggests that the transition from the two-band to one-band spectrum obeys the Henderson Hasselbalch relationship for a single protonation event with a transition pKa of 6.6 +/- 0.1. No pH dependence is observed for the spectra of the fully reduced unliganded or CO-inhibited enzyme. Tryptophan fluorescence spectra of the enzyme indicate that no major disruption of protein structure occurs in the pH range 5.5-8.5 used in this study. Resonance Raman spectroscopy indicates that the cytochrome a3 chromophore remains in its ferric, cyanide-bound form in the mixed valence enzyme throughout the pH range used here. These data indicate that the transition observed by second derivative spectroscopy is not due simply to pH-induced protein denaturation or disruption of the cytochrome a3 iron-CN bond. The pH dependence observed here is in good agreement with those observed earlier for the midpoint reduction potential of cytochrome a and for the conformational transition associated with energy transduction in the proton pumping model of Malmström (Malmström, B. G. (1990) Arch. Biochem. Biophys. 280, 233-241). These results are discussed in terms of a model for allosteric communication between cytochrome a and the binuclear ligand binding center of the enzyme that is mediated by ionization of a single group within the protein.


The pH Dependence of Cytochrome a Conformation in Cytochrome c
Oxidase* (Received for publication, June 13, 1991) Naoko Ishibe$, Stephen R. Lynch$, and Robert A. CopelandB The pH dependence of the conformation of cytochrome a in bovine cytochrome c oxidase has been studied by second derivative absorption spectroscopy. At neutral pH, the second derivative spectra of the cyanide-inhibited fully reduced and mixed valence enzyme display two Soret electronic transitions, at 443 and 461 nm, associated with cytochrome a. As the pH is lowered these two bands collapse into a single transition at -444 nm. pH titration of the cyanide-inhibited mixed valence enzyme suggests that the transition from the two-band to one-band spectrum obeys the Henderson Hasselbalch relationship for a single protonation event with a transition pK, of 6.6 f 0.1. No pH dependence is observed for the spectra of the fully reduced unliganded or CO-inhibited enzyme. Tryptophan fluorescence spectra of the enzyme indicate that no major disruption of protein structure occurs in the pH range 5.5-8. 6 used in this study. Resonance Raman spectroscopy indicates that the cytochrome u3 chromophore remains in its ferric, cyanide-bound form in the mixed valence enzyme throughout the pH range used here. These data indicate that the transition observed by second derivative spectroscopy is not due simply to pH-induced protein denaturation or disruption of the cytochrome a3 iron-CN bond. The pH dependence observed here is in good agreement with those observed earlier for the midpoint reduction po- These results are discussed in terms of a model for allosteric communication between cytochrome a and the binuclear ligand binding center of the enzyme that is mediated by ionization of a single group within the protein.
The coupling of electron transfer reactions to the formation of transmembrane electrochemical gradients is the fundamental mechanism by which respiratory energy transduction occurs in mitochondria and aerobic bacteria (1). One of the primary sites of this coupling is cytochrome c oxidase (ferrocytochrome-c:oxygen oxidoreductase, EC 1.9.3.1), a metallo-*This work was supported by a Syntex Scholars and a Merck Academic Faculty Development Award (both to R. A. C.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
These authors contributed equally to the work reported here. § To whom correspondence should be addressed Dept. of Biochemistry and Molecular Biology, University of Chicago, 920 E. 58th St., Chicago, IL 60637. enzyme which has been the focus of intense study for over 30 years (2). Cytochrome c oxidase functions as an electron transfer-driven proton pump, translocating protons from the mitochondrial matrix to the cytosol in response to electron transfer among its four redox active metal cofactors. Two of these metal centers, cytochrome a and CuA, serve as the initial electron acceptors from the physiological reductant cytochrome c. The remaining two metals, cytochrome a3 and CUB, together form a binuclear center for exogenous ligand binding (2).
Which of these metal centers functions as an energy transducer is still a matter of debate. Models of proton pumping have been put forth in which reduction of cytochrome a (3), CUA (4), or both (5) provides the driving force for proton transport. Recent studies by Wikstrom (6) suggest that the oxygen binding site either plays a direct role in proton translocation or is in allosteric communication with the primary site of energy transduction.
While the identity of the site of energy transduction remains unresolved, theoretical studies have suggested certain properties for the site of coupling in cytochrome c oxidase. It is generally agreed that the energy-transducing metal center must have two conformations available to it in both its oxidized and reduced valence states. Furthermore, one expects that the reduction potential (and perhaps conformation) of the energy transducer should be pH-dependent if it is to couple proton movement to electron transfer. This latter expectation has led some workers to propose that cytochrome a is the site of coupling because it displays a pH dependence of its midpoint reduction potential (7). Blair et al. (8) have pointed out, however, that the site of energy tranduction need not necessarily display a steep pH dependence, and in fact the pH dependence of the cytochrome a reduction potential is not as dramatic as earlier studies had suggested. Some pH dependence of the reduction potential of the energy transducing metal is nevertheless expected.
To better follow the structural transitions of the heme cofactors of cytochrome c oxidase, our group has recently introduced the use of second derivative absorption spectroscopy (9). We find that the individual electronic transitions of cytochrome a and cytochrome a3 can be resolved by this method. Using various stable forms of the enzyme we have shown that the Soret transition of ferrocytochrome a splits into two bands at -443 and 451 nm when exogenous ligands are bound to cytochrome a3; when the ligand binding site is unliganded one observes only a single ferrocytochrome a Soret transition at 444 nm. These data suggest that cytochrome a undergoes a conformational switch upon ligand binding at the distal cytochrome a3 center. We have further shown that this conformational switching at cytochrome a occurs during steady state turnover of the bovine enzyme, suggesting that it is part of the normal catalytic mechanism of cytochrome c oxidase (10).
In this study we have made use of the high resolution power of the second derivative absorption method to evaluate the pH dependence of the cytochrome a cofactor in stable forms of the bovine enzyme. The results reported here suggest that the conformation of cytochrome a is influenced by ionization of a protein component that titrates with a pK, of 6.6 when the enzyme is inhibited by cyanide but is independent of pH in the unliganded or CO-inhibited enzyme.

MATERIALS AND METHODS
All reagents were the highest grades commercially available, and all solutions were prepared with doubly glass-distilled water. Bovine cytochrome c oxidase was isolated from cardiac muscle as previously described and stored at -80 "C until use (11). Protein concentration was determined by the Bradford colorimetric assay using bovine serum albumin as a standard (12). Heme A concentration was determined from the spectrum of the fully reduced enzyme at 605 nm and utilizing the extinction coefficient 39.6 mM" cm" with respect to the enzyme (13). Enzyme activity was measured spectrophotometrically by following the loss of absorbance at 550 nm for ferrocytochrome c solutions as described by Smith (14). The purified enzyme used in these studies contained 9.9 nmol of heme A/mg of protein and displayed a maximum turnover rate of 142 SKI.
The cyanide-inhibited (CN) enzyme was prepared by incubating a 300 p~ solution of enzyme with 5 mM KCN at 4 "C for 48 h. The extent of cyanide binding was checked by observing the red shift of the Soret band of the enzyme from 421 to 429 nm. The CN-enzyme was then diluted to 10 PM into buffer containing either 50 mM HEPES' or 50 mM MES (depending on the final pH), 100 mM KCl, and 0.1% Tween 20 (Pierce Chemical Co.). The final enzyme concentration was determined spectrophotometrically using and extinction coefficient of 156 mM", cm" at 429 nm (13). The pH of the enzyme solution was determined at 25 "C using a Corning model 220 pH meter which was calibrated daily. For samples below pH 6.0 the CNenzyme was first diluted into pH 7.4 buffer and then dialyzed against 100 volumes of the final pH buffer, containing 1 mM KCN, for 24 h.
The CN-inhibited mixed valence enzyme was prepared from the above solutions as follows. The CN-enzyme solution was placed in an anaerobic cuvette equipped with a stopcock and septum seal. The atmosphere above the sample was exchanged for vacuum and then nitrogen a minimum of five times. Either sodium dithionite or sodium ascorbate and N,N,N,N'-tetramethyl-p-phenylene diamine were then added through the septum, and the sample was cycled through vacuum and nitrogen several more times. The sample was then incubated at 4 "C for 20 min to allow complete reduction to occur. After spectral data acquisition, the septum was removed from the cuvette, and the pH of the solution was confirmed as described above.
The fully reduced, fully reduced CN-bound, and fully reduced CObound enzyme forms were prepared in analogous ways using previously described procedures (9).
Optical spectra were recorded at 25 "C with a Cary 14 UV-Vis-near IR spectrophotometer. The instrument was interfaced to an IBMcompatible computer which was used to control data acquisition and for digital storage of the data (OLIS, Jefferson, GA). For each experiment a buffer baseline was recorded and digitally subtracted from all subsequent spectra. Spectra of the enzymes were recorded in 1-nm steps with a spectral bandpass of 0.5 nm. Each reported spectrum is the average of 10 such scans. The second derivatives of the spectra were obtained as previously described (9).
For a single component absorption spectrum, the second derivative of absorbance with respect to wavelength displays a sharp negative extreme at the wavelength maximum of the electronic transition; this negative feature is bordered on the high and low energy sides by positive "wings." Such a pattern is observed, for example, in the second derivative spectrum of the fully reduced unliganded enzyme (9). For such a single component system, an accurate measure of the intensity of the electronic transition can be obtained by drawing a line to connect the two positive wings and then integrating the area enclosed by the roughly triangular shape described by this tangent line and the band envelope itself. We have used a similar strategy to estimate the relative intensities of the 443-and 451-nm transitions in the second derivative spectra here. Using an integration program The abbreviations used are: HEPES, 4-(2-hydroxyethyl)-l-piperazineethanesulfonic acid; MES, 2-morpholinoethanesulfonic acid; Tween 20, polyoxyethylenesorbitan monolaurate. supplied by OLIS (Jefferson, GA) we drew a line between the positive features at 435 and 446 nm and thus integrated the area under the 443-nm feature. Likewise we drew a line from 447 to 457 nm to integrate the area under the 451-nm transition. At high pH it is clear that these wavelengths correspond to the positive wings described above. It is less clear where these wings occur at lower pH, but in the interest of consistency we have chosen to use the same wavelengths for all spectra. This process is likely to underestimate the relative contribution of the 443 nm feature to the low pH spectra; however, for the type of comparative measurements we wish to make here we believe this procedure is appropriate.
Fluorescence spectra of the CN-enzyme at various pH values were obtained with a Photon Technology International alphascan spectrofluorometer (PTI, South Brunswick, NJ). Samples (1 PM in enzyme) were contained in a 1-cm quartz cuvette and excited at 280 nm. The emission was scanned from 290 to 425 nm with the emission polarizer set at 54.75". Spectra were obtained with an integration time of 1 s/ nm and a spectral bandpass of -5 nm.
Resonance Raman spectra of the fully reduced unliganded and the mixed valence CN-enzyme were obtained with 441.6-nm excitation from a He-Cd laser (Liconix, Sunnyvale, CA). The laser light was focused onto a capillary tube that was embedded in an aluminum block heat exchanger. A water/ethylene glycol solution was flowed through the heat exchanger from a refrigerated circulating bath to maintain the sample temperature at 15 "C. Samples were circulated through the capillary at a rate of 2 ml/min by a peristaltic pump (LKB Microplex S). Throughout the experiments the samples were maintained anaerobic by flowing nitrogen gas over the sample reservoir. Scattered light was collected at 90" from the excitation beam and focused onto the entrance slit of a Spex 1403 double monochromator, which was driven by an IBM-compatible computer (Spex, Metuchen, NJ). All spectra were obtained at a scan rate of 1 cm"/s and a spectral bandpass of 5 cm". The reported spectra are each the average of 10 scans. derivative spectra obtained for this form of the enzyme at three representative pH values. (Note that the spectrum of the enzyme at pH 7.4 displays a weaker contribution from the 429-nm ferricytochrome a3 band than that originally reported by Sherman et al. We have determined that this discrepancy results from incomplete reduction of cytochrome a in the originally reported spectrum. In terms of the ferrocytochrome a transitions, however, the present spectrum agrees well with the previously reported one.) As seen in Fig. 1, there is a broadening out of the 429-nm feature of cytochrome a3 as one lowers the pH from 7.4 to 5.8. This effect most likely arises from some increased heterogeneity in the ligand binding pocket of the enzyme at low pH. The most striking effect, however, is that the longer wavelength band at 451 nm is greatly diminished as the pH is lowered, so that by pH 5.8 one observes essentially a single ferrocytochrome a Soret band at 443 nm.
If one fully reduces the enzyme prior to cyanide addition, one forms the fully reduced CN-inhibited enzyme (i.e. a", aj'+-CN). The effect of pH on the second derivative spectrum of this enzyme form is similar to that for the mixed valence state, as illustrated in Fig. 2. In contrast to the case of the cyanide-inhibited enzyme forms, the second derivative spectra of the fully reduced unliganded (i.e. a'+, a3") and fully reduced CO-inhibited (i.e. a'+, a3'+-CO) enzyme were insensitive to pH over the 5.5-8.5 pH range (data not shown). Likewise, the low pH-induced changes in the spectra of the cyanide-inhibited forms of the enzyme could not be mimicked by calcium ions, although this cation has been shown to mimic protoninduced changes in the difference spectra of cytochrome c oxidase within mitochondria (15).
The pH-induced transition from the two-band-to singleband spectrum appears to be fuIIy reversed when the pH is readjusted to -7.4. To test this we prepared a solution of the fully reduced CN-inhibited enzyme in 50 mM MES, 100 mM KCl, 0.1% Tween 20, pH 5.9, as described above. The second derivative spectrum of this sample displayed a strong transition at 443 nm and a weak shoulder at 451 nm. The ratio of the integrated intensities of the 451/443-nm bands for this spectrum was 0.19. This soIution was then titrated to pH 7.3 The single-band spectra obtained at low pH in the cyanideinhibited enzymes are similar to that seen for the fully reduced unliganded enzyme (9). We were therefore concerned that the present results might reflect pH-induced disruption of protein structure leading to rupture of the iron-CN bond and subsequent reduction of cytochrome a3. To address this issue we studied the intrinsic tryptophan fluorescence spectrum of the enzyme and the resonance Raman spectrum of the heme cofactors. Fig. 3 illustrates the effects of pH on the intrinsic tryptophan fluorescence spectrum of CN-inhibited cytochrome c oxidase. As first shown by Hill et al. (16) the tryptophan fluorescence maximum occurs at 328 nm in the native enzyme at neutral pH. Lowering the pH of the enzyme solution to 5.0 does not significantly affect the fluorescence wavelength maximum or quantum yield. Further lowering the pH to 2.5 leads to a significant increase in quantum yield and a red shift of the emission wavelength maximum as the protein unfolds and the otherwise buried tryptophan residues become more solvent exposed. These results suggest that over the pH range of interest, there is no major disruption of protein structure. Fig. 4 illustrates the resonance Raman spectra from 175 to 225 and from 1325 to 1425 cm" of fully reduced unliganded cytochrome c oxidase at pH 7.4, and the mixed valence CNinhibited enzyme at pH 7.4 and 5.5. In the reduced unliganded enzyme, one observes a band at 212 cm" assigned to the ironhistidine stretch of five coordinate ferrocytochrome a3. This out of plane iron-nitrogen stretch is not observed in six coordinate low spin hemes, such as the CN-bound ferricytochrome a3 (17). As expected, this band is not present in the mixed valence CN-inhibited enzyme at either pH 7.4 or 5.5, indicating that cyanide remains bound to cytochrome a3 at low pH. This inference is also supported by the high frequency resonance Raman data. The spectral region between 1300 and 1400 cm" contains the totally symmetric stretching mode, v4, of the porphyrin ring system, This band is sensitive to the oxidation state of the central iron atom of the heme; for heme A the band occurs at 1355 cm" for the ferrous iron valence state and at 1365 cm" for the six coordinate low spin ferric heme (17). Thus, if cyanide is bound to cytochrome a3 in the mixed valence enzyme one expects to observe two u4 bands in the resonance Raman spectrum. As seen in Fig. 4, this expectation is met for the enzyme at both pH 7.4 and 5.5, confirming the fact that at low pH cytochrome a3 remains oxidized and six coordinate. Spectra obtained in the 1500-1600 cm" region of the spectrum also confirm that cytochrome a3 is low spin in the mixed valence CN-inhibited enzyme at both pH 7.4 and 5.5 (data not shown).
Having established that the pH effect observed in the second derivative spectra is not simply due to protein denaturation, we investigated the pH dependence of the mixed valence CN-inhibited enzyme in greater detail. We have recorded the second derivative spectrum of this form of the enzyme at a number of pH values between 5.5 and 8.5 and measured the integrated areas for the 443 and 451 nm bands. Fig. 5 illustrates the effect of pH on the ratio of the integrated areas under the 4511443 nm bands. These data appear to follow the Henderson Hasselbalch relationship for a single protonation event, which is described by the following equation.
where y is the experimental value at any given pH, yHA is the experimental value for the fully protonated form of the molecule, and yA-is the experimental value for the fully deprotonated form of the molecule. The data in Fig. 5 were fit to this equation using the nonlinear curve-fitting program NFIT (Island Products, Galveston, TX). The curve drawn through the data points in Fig. 5 represents the least squares best fit obtained in this way, and yields an estimate of the transition pK, of 6.61 k 0.13 (x2 = 0.0153). DISCUSSION The data presented here provide clear evidence that the conformation of ferrocytochrome a in the CN-inhibited enzyme is sensitive to proton binding at a group within the enzyme that titrates with a pK, of 6.6. We observe this pH dependence only in the CN-inhibited forms of the enzyme, and not in the unliganded or CO-inhibited species. Interestingly, this is exactly the behavior observed in studies of the pH dependence of the midpoint reduction potential of cytochrome a. Thus, Artzatbanov et al. (18) have shown that the reduction potential of cytochrome a is sensitive to pH in the CN-inhibited enzyme but much less so in the unliganded species (18). This group localized the pH effect to the mitochondrial matrix side of the respiratory membrane, suggesting that the protonatable group was in closer proximity to the binuclear center ( i e . cytochrome a3-CuB) than the cytochrome a cofactor. Similarly the related terminal oxidase from Escherichia coli, cytochrome 0, displays a pH dependence for the reduction potentials of its heme groups, and evidence of hemeheme interactions (19). Several groups have shown that spectral features of the CN-inhibited ferricytochrome a3 site display pH sensitivity that titrate with pK, values between 6.5 and 6.9 (20, 21), while Papadopoulos et al. (22) have shown that the spectroscopic features of ferricytochrome a3 titrate with an apparent pK, of 7.8 in the resting enzyme. The similarity between the pK. values obtained in these studies and that observed here suggests allosteric communication between a proton binding group in the cytochrome U~-CUB pocket and cytochrome a, rather than direct proton binding in the vicinity of cytochrome a.
Recently Malmstrom (5) has proposed a model for redoxcoupled proton pumping by cytochrome c oxidase in which electrons enter cytochrome a and CuA in one conformational state of the protein, El, but can only migrate to the binuclear ligand binding center after a conformational transition to the E2 state; proton translocation occurs during the El to EP transition in this model. Malmstrom's group has studied the pH dependence of cytochrome c oxidation (23) and steady state levels of cytochrome a reduction (24) for cytochrome c oxidase within proteoliposomes and has found that the transition from El to Ez conformation is associated with protonation of an acid-base group with pK, 6.4 that is protonated from the matrix side of the membrane. Taken together, these studies suggest that the differences in cytochrome a conformation observed by second derivative spectroscopy and the differences in cytochrome a reduction potential may both be manifestations of a pH-induced E, to E2-like transition in the CN-inhibited enzyme.
What is the structural basis of the proposed communication between the ligand binding site of the enzyme and cytochrome a? The transition pK, values observed here and by others fall within the range 6.4-6.9, which is the expected pK, value for the side chain nitrogen of the amino acid histidine. Although cytochrome c oxidase is a multisubunit enzyme, biochemical (25) and genetic (26) data suggest that cytochrome a, cytochrome a3, and CUB are all contained within subunit I of the enzyme. It seems likely that the structural components which facilitate allostery between cytochrome a and the binuclear center would also be contained within this subunit. Hydropathy mapping suggests that subunit I is composed of 12 membrane-spanning CY helices, labeled I-XII. Comparing the amino acid sequences of subunit I from a variety of eukaryotic and prokaryotic cytochrome c oxidases and E. coli cytochrome o one finds that there are only six highly conserved histidine residues within the transmembrane helices of subunit I (26); all six of these are expected to be directly involved in metal ligation. From a variety of molecular modeling and site directed mutagenesis studies on cytochrome c oxidase and the related cytochrome o enzyme, a consensus view of the identities of the individual metal ligands is emerging. In this view one of the conserved histidine residues of helix X (His-410) provides an axial ligand of cytochrome a, and the second conserved histidine of this helix is ligated to CUB; the remaining conserved histidines (His-91, His-273, His-322, and His-323) serve as binuclear site ligands of either cytochrome a3 or CUB and provide the second axial ligand of cytochrome a (27).2 None of these histidines would be available for acid-base chemistry while ligated to a metal center.
Thus, for a subunit I histidine to be involved in the observed pH dependence of the enzyme would require a deligation from one of the binuclear site metals to occur. It is interesting to note in this regard that the pH dependence observed here is limited to the CN-inhibited enzyme. Recent Fourier transform infrared studies indicate that cyanide, unlike carbon monoxide, binds to both cytochrome a3 and CUB at ambient temperatures (28). It is possible then that exogenous ligand binding at CUB might result in the displacement of a histidine ligand which would then be available for acid-base titration. However, to our knowledge there is no direct evidence for such a ligand displacement reaction at the binuclear center. Alternatively one must consider other components of the enzyme that might serve as the ionizable group. Under certain conditions other amino acid side chains, notably carboxylate residues, will display pK, values near 6.6 (29). There are three highly conserved aspartate residues (Asp-121, Asp-177, and Asp-401) and two conserved glutamate residues (Glu-275 and Glu-528) within subunit I; one or more of these could potentially play a role in allostery within the enzyme. In fact, the titratable group need not be an amino acid at all. The heme A cofactors of the enzyme each contain two propionate substituients which are potential candidates for the ionizable group. While the pK, of propionic acid in aqueous solution is 4.87 (30), this value could easily be elevated to 6.6 by covalent attachment to the porphyrin system and by the influence of the polypeptide matrix of the enzyme. While we cannot unambiguously define the molecular basis for allostery at this time, it is clear that such allostery occurs in cytochrome c oxidase. What the mechanistic significance of this allosteric communication is for the catalytic activities of cytochrome c oxidase remains an issue to be clarified by further experimental work.