THE ELECTRONIC STATE OF HEME IN CYTOCHROME OXIDASE MAGNETIC CIRCULAR DICHROISM OF THE ISOLA TED ENZYME AND ITS DERIVATIVES

Magnetic circular dichroism (MCD) spectra h_~e been recorded for beef heart cytochrome oxidase and a number of its inhibitor co=?lexes. The resting enzyme exhibits a derivative shape Faraday ~ term in the Soret region, characteristic of low-spin ferric heme, which accounts for 50.% of the total oxicase heme a. The remaining heme a (SO%) ,, is assigned to the high-spin state. MCD temperc:.~ure studies, comparison with the MCD . spectra of heme a - imidazole model compotinds ar!d ligand binding (cyanide, formate) J" studies are consistent with these spin state assig~ents in the oxidized enzyme. Further-more, the ligand binding properties and correlations between optical and MCD parameters 3+ indicate that in the resting enzyme the low spin heme a is aue solely to cytochrome ~ and the high spin heme~ to cytochrome a~+.

Introduction \J U I. 7 7 -1-Although it is now firmly established that the active center of cytochrome oxidase ·contains .two moles of heme a and two atons of copper (1)(2)(3) there is considerable uncertainty regarding the functional relationships between these four metal components (4,5).
As originally proposed by Keilin and Hartree (6) the two heme moieties of the oxidase can be distinguished by their ligand binding properties. One of them, named cytochrome a 3 , was postulated to react with -common heme ligands such as cyanide and carbon monoxide, while the other, called cytochrome a, was unavailable to such reagents.· This heterogeneity with respect to ligand binding has been amply confirmed in a number of laboratories and is most· simply interpreted in terms of the so-called "classical model" for the enzyme. In this model the two hemes are asserted to be chemically and spectrally distinct species. Cytochrome a 3 is presumed to be high spin (S = 5/2 in the ferric state, S = 2 in the ferrous state) to account ~or its ready reaction with external ligands and is expected to exhibit weak absorbance in the a band region. Cytochrome a, on the other hand, is assigned as a typical hexacoordinate low-spin heme protein (S = 1/2 in the ferric state, S = 0 in the ferrous state) witn an a band intensity roughly four times that of cytochrome a 3 • Cytochrome a 3 is presumed to be the.co~ponent which reacts with oxygen while cytochrome ~ is postulated to be the site for reaction with cytochrome £· This classical model has been summarized most forcefully in a review by Lemberg (1).
An alternative hypothesis minimizes any intrinsic spectral differences between the two hemes. In its most extreme form cytochroEes a and a 3 are postulated to be indistinguishable and low-spin in the oxidized enzyme and that addition of ligands or reducing ~quivalents is required for the observed heterogeneity in the properties of the two hemes (7). In this model some form of heme-heme interaction exists such that the redox and/or li~and binding state of one of the he~es a has a pronounced effect on the spectral properties of the second heme~ (8)(9)(10)(11).
In a third model the two heroes a of cytochrome oxidase have.independent spectral properties; However, heme-heme interactions are manifested by interdependent redox . -2potentials such that reduction or ligand binding by one of the hemes ~ alters the reduction potential of the second heme (12). The arguments in favor of each of these models have been reviewed by Nicholls and Chance (4).
These models make distinct predicti~ns of the coordination geometry around the Fe atoms of cytochromes a and ~3 so that a crucial test for differentiating between these extremes is a determination of the heme ~ spin-states in the enzyme in various redox states and after reaction with external ligands. EPR spectroscopy has been used extensively for this purpose and while the technique has been most fruitful in the overall characterization of the enzyme, resolution of the above questions has been complicated by the following observations: (i) in the oxidized enzyme only ·so% of the heme iron contributes to the low-spin resonance at g = 3 and only ~· 40% of the copper can be accounted for by the resonance at g = 2 (13,14). (ii) During the course of a·reductive titration both axial and rhombic high-spin EPR s}.gnals appear at g = 6; these resonances account for ca. 15-25% of the total iron although under some conditions the value approac:hes 35-45% (14)(15)(16). These observations have led to contradictory assignments with Wilson and coworkers (9,11) identifying the low-spin species as cytochrome a 3 by virtue of it having the most positive potential while Hartzell et al. (15), noting that the g = 3 resonance appears to be the site of reduction during reaction with ferrocytochrome £, assign this resonance to cytochrome ~· Magnetic circular dichroism spectroscopy (MCD) is emerging as a powerful tool in the study of iron spin-states in heme-proteins. The versatility of the method lies in two recent observations: (i) heme iron in both ferric (S = 1/2 and 5/2) and ferrous (S = 4/2) valence states show temperature dependent MCD intensity in the Soret region .which is indicative of the paramagnetic ground states of these species (17)(18)(19)(20)(21) and (ii) t~e intensity Qf this MCD in ferric heme proteins is correlated with the fraction of iron in the low-spin state irrespective of the nature of the axial ligands (19) •. Thus \ MCD spectroscopy provides spin-state data for both comcon valence states of heme iron.
This information both extends and complements the results available from EPR. Moreover, .. 0 Q D ~"cJ 6 u ';.U I --; 8 l;": -I MCD has the particular advantage that data are routinely obtained close to ambient temperature in contrast to the extremely low temperatures (ca. 10° K) required for the satisfactory observation of all the species contributing to the EPR spectrum of this enzyme.
\~e have previously reported preliminary results on the HCD of cytochrome oxidase (22,23) and these data have recently been confirmed (24). Based on our observations we have proposed a model for the enzyme which is similar in many respects to the classical model described above, but has as added hypotheses (i) an antiferromagnetic interaction between cytochrome a 3 and one of the two copper moieties in the resting enzyme and (ii) the occurrence of a conformational change subsequent to partial reduction of the enzyme. This model accounts for the available EPR and magnetic susceptibility data. In the experiments described in this paper we present further data in support of this model and in a subsequent paper we will describe the behavior of the enzyme during reductive ti'trations.

Materials and Methods
Solubilized beef heart cytochrome oxidase was obtained by three different isola~ was varied down to 77° K, a low-temperature optical dewar and a 3.0 mm cell with a copper constantan thermocouple incorporated were used; the applied magnetic field was 0.9 Tesla.
All spectra are presented as the difference in extinction for left versus right circularly polarized light (ae = eL -eR) on a per heme ~ basis and are normalized to an applied magnetic field of 1 Tesla (10 kG), i.e., as 6e/Tesla. The use of these units allows di"rect comparison of all HCD spectra normalized to a 1 gauss field by the obtained. The data can be converted to [8] units nm. This fine structure has been interpreted as arising from charge-transfer bands which exhibit MCD transitions that are highly dependent on the nature of the axial ligand (19).
In the Soret region a derivative type curve is also observed along with a shoulder at shorter wavelengths near 400 nm. The crossover for the S-shaped curve is at 427 nm with positive and negative extrema at 420.5 and 434 nm respectively; it.~s considerably red-shifted when compared to the absorption maximum of the oxidized enzyme at 418 nm.
This red shift is anomalously large when compared with other. heme proteins which exhibit this derivative shape curve (19,20,32,33)~ In these the 'HCD approximates the first derivative of the absorption spectrum and the MCD zero crossing corresponds closely to the maximum in the absorption spectrum. In addition }1CD transitions of this type have been shown to be temperature dependent, therefore corresponding to Fa~aday £terms, and typical of ferric heme proteins in which alY or part of the heme iron is in the S a 1/2, low-spin state (17,19,20). Similarly if the derivative shaped MCD curve obse_rved in the Soret region for cytochrome oxidase (Fig. 1) is due to an S = 1/2 paramagnetic ground state, strong temperature dependence for the intensity of this band is expected. The results o~ Fig. 2 show that this is indeed the case for the Soret MCD of resting cytochrome oxidase. At both'421 and 435 nm the amplitude of the }1CD spectrum is directly proportional to 1/T (Fig. 2, inset) demonstrating for cytochrome oxidase that this MCD 1 band arises from Faraday f terms. The data of Fig. 2 were obtained only to -145° C; however, in a second set of experiments with oxidase in which cytochrome bc 1 contamination interfered at wavelengths less than 415 nm, we were able to obtain MCD spectra to -196° C, which were consistent with the results of Fig. 2 (19), the insensitivity of the HCD spectrum to for.mate binding indicates .that the formate. binding site is high-spin both in the absence and presence of this 'inhibitor.
. 3+ Again ass~ing that formate b~nds solely to cytochrome ~3 , we conclude that both the * * .

3+
The high spin to low spin transition observed by ~lCD for CN-binding to cytochrome ~3 in the resting enzyme offers very strong support for these assignments.

Reduced cvtochrone oxidase
The visible and Soret ~lCD and absorption spectra for reduced cytochrome oxidase are and we observe very little MCD intensity in the spectral range from 500-560 nm except for a negative shoulder at 568 nm and weak but complex structure between 500 and 530 nm.
The weak derivative curve centered at 550 nm is due to a slight contamination of the oxidase with complex III (cytochromes b-c 1 ). The MCD of reduced band~ type cytochromes is extremely intense (~~/Tesla ~ 200-300, depending on bandwidth) (20) and this region of the spectrum provides a very.sensitive test for the presence of Complex III contaminat ion.
The change in MCD upon reduction is much more dramatic in the Soret region with the spectru~ changing sign and growing about four-fold in intensity (cf. Fig. 1). The crossover at 440.5 nm is slightly blue-shifted from the absorption maximum at 443 nm ' in the optical spectrum of the reduced protein. In general the MCD spectrum of .reduced oxidase bears a striking resemblance to the spectra obtained with deoxymyoglobin (19), de_oxyhemoglobin (18) and ferro-horseradish peroxidase (36). All three of these proteins are ferrous high-spin and exhibit an asymmetric derivative type Soret MCD with the trough occurring at higher energy than the larger amplitude peak; again the cross-over is blue shifted with respect to the absorption maximum. In addition all three protoheme proteins show a negative shoulder some 15-30 nm to shorter wavelength of the mnjor trough; this secondary feature appears at 413.5 nm in the MCD spectrum of re~uced cyto--10chrome oxidase. T~e principal difference observed with oxidase compared to the'other three proteins is the larger amplitude obse~ed for the principal trough; for cytochrome oxidase the peak-to-trough ratio is 1.95, while for deoxyhemoglobin'it is 3.45 (18).
The temperature dependence of the 440 nn band of reduced oxidase (Fig. 6a) establishes the paramagnetic origin (Faraday f. ter-_s) of the Soret HCD for the reduced protein. In addition to an increase in intensity as the temperature is lowered there is also a change in the shape of the spectrum; this is most clearly seen in the difference spectrum between the }!CD spectra recorded at -145 and 0° C (Fig. 6b). The peak.present at 447 nm at 0° has shifted to 446 nm and the peak to trough ratio has increased from 1.95 to 2.45 over this temperature range. The corresponding CD spectra, recorded simultaneously (31), exhibited a temperature independent peak at 446 nm with only slight (<5%) changes in ampl~tude indicating that any intrinsic band narrowing'must be small and cannot be the origin of the increase in MCD intensity. Plots of MCD intensity~ 1/T (Fig. 6a, inset) are linear at the four wavelengths shown. However, these plots show marked deviations from behavior expected from a simple, temperature dependent Boltzmann distribution. This effect is most pronounced at 447 and 452 nm: the Boltzmann tu:/Tesla (128°) factor between 273° K and 128° K is 2.13, the intensity increase ( 6 e/Tesla ( 273 o)) is 2.42 I at 447 nm but only 1.66 at 4.52 nm. This behavior suggests the contribution of Faraday A and/or B terms to the spectra in the region around 450 nm. However, due to the possibility of small band shifts and narrowing as the temperature is lowered (37), and the resulting uncertainties in extrapolation to infinite temperature, we have not carried out this analysis in more detail.
Reduced cytochrome oxidase: carbon monoxide and cyanide complexes The MCO spectra for reduced cytochrome o~idase and its inhibitor complexes with CO and CNare shown in Fig. 7. In the presence of CO the Soret MCD peak decreases in intensity by about one-half and is red-shifted by 5.5 nm compared to the untreated enzyme; the peak is at 452.5 nm with the crossover at 447 nm, a negative shoulder at approximately 434 nm, a second zero-crossing at 428 nrn and a small positive peak at 423 nm. The optical absorption spectrum of this derivative has a peak at 430 nm which is 2+ .
classically assigned to the a 3 ·co complex on the basis of photochemical action spectra (4).
. i+ In addition, there is a shoulder at 442.5 nm which is assigned to cytochrome·~ On the basis of this absorption spectrum we assign that portion of the MCD curve with a 2+ peak at 423 nm, crossover at 428 nm and shoulder at 434 nm to the cytochrome a 3 •CO complex. This weak, derivative shape curve is qualitatively similar to the spectra of low-spin ferrous complexes of hemoglobin (18) and myoglobin (19) but the intensity is much smaller than that observed for the protoheme proteins. The more intense, redshifted }~D curve with a peak at 452.5 is .assigned to cytochrome a 2 + These assignments ·are borne out by the cyanide .and formate derivatives reported below as well as by reductive titrations to be reported in a subsequent paper.
The MCD spectrum of the CN complex of r~duced cytochrome oxidase (Fig. 7) resembles the CO complex closely. The peak occurs at 452.5 nm and the zero-crossing occurs at 446 nm. The absorption spectrum of the co~lex s~owed a peak at 443 nm with a slight shoulder at ~ 430 nm; this shoulder may correspond to incomplete reduction of cyto-3+ chrome oxidase· prior to addition of the CN-leading to the formation of the a 3 ·CN complex. This would result in a greater intensity between 420-435 nm (vide infra) and may account for the increased MCD around 440 nm compared to the CO complex.
Partially reduced cytochrome oxidase: cyanide, formate, sulfide and azide compk~es Complexes of partially reduced cytochrome oxidase with a number of inhibitors can be prepared in the aerobi.c steady-state. Under these conditions, cytochrome a is largely reduced while the cytochrome a 3 • inhibitor complex is primarily oxidized (4,3,8).
The MCD and absorption spectra of two of these derivatives, cyanide and formate, are shown in Fig. 8. • formate_ compound and also the 25 run separation in absorption maxima for the two heme species, we conclude that the derivative MCD curve (442 nm trough, 447 nm· 2+ crossover, 451.5 nm peak) in Fig. 8 is due solely to cytochrome~ will be used subsequently in calculating ~~CD difference spectra.

This assignment
We have also observed the MCD spectra of cytochrome ox.idase in the presence of sulfide and azide with reductants (TMPD a~d ascorbate) added subsequent to.the inhibitor. The .spectra exhibit the characteristics of ferrous cytochrome a with peak at 452 nm (A£/Tesla 40), ·a crossover at 447 nn and a trough at 438 nm •. As observed by Nicholls (38), we also see partial redt1ction of the a 3 ·inhibitor complex under these conditi.ons so that interpretation of the spectra at wavelengths less than 435 nm is difficult.
"Oxygenated" cytochrome oxidase We have prepared the ''oxygenated" derivative of cytochrome oxidase by oxygenating the dithionite reduced enzyme. The :HCD spectrum recorded 30 min after oxygenation is shown in Fig. 9. This spectrum is identical to that cf the resting enzyme, apart from some additional intensity at 452 nm which ~e attribute to incomplete reoxidation of  The calculated MCD spectrum for ~3 has been obtained by 3 different subtraction procedures: (1) fully reduced oxid~se minus partially reduced + forrn3te (2) fully reduced oxidase minus fully reduced + cyanide (3) fully reduced oxidase minus fully · -14reduced + carbon monoxide. Fig. lOa shows the MCD spectra for reduced cytoc~rome oxidase and the formate complex of the partially reduced enzyme, the difference spectrum i's shown in Fig. lOb. Figs. lOc and d show the difference spectra for the CO and CN derivatives respectively (see also Fig. 7). In each case the shape of the resulting spectrum shows very ~trong si~ilarity to the direct ~~CD spectrum of deoxyhemoglobin 1 deoxy-myoglobin~ The most dramatic aspect 6£ this similarity is the anomalous sign of the derivative ~ype spectrum (peak to longer wavelengths) observed for all three of these proteins. Treu'and Hopfield (18)   t. as the product Of (full Width at nalf height) 2 and amplitude) for deoxyhemoglobin "'and r.'rJY•dJ, 2+ cytochrome a 3 we calculate that, on a per hene ~basis, about 45% of the heme in cytochrome oxidase contributes to the dif~erence spectrum.
Of the three derivatives used in obtaining the difference spectra of Fig. 10, that 2+ of partially reduced enzyme plus formate yields the spectrum of cytochrome ~ in its purest form (vide supra). The effect of both CO and CN-on cytochrome oxidase appears to be analogous to the effect of CO on myoglobin and hemoglobin, viz the conversion of 2+ high spin ferrocytochrome a 3 to a low-spin. ferrous state.
We have proposed above that the C\-co~?lex of ferric cytochrome a 3 is low spin.
Accordingly, we expect a fairly strong . . , derivative shape Soret HCD band for this derivative.  Soret ~~D intensity which appears to be highly dependent on second-order effects.

Discussion
The MCD spectra of cytochrome oxidase and its derivatives are most simply interpreted in. terms of the classical scheme of Keilin and Hartree (6) with each heme ~ existing in a different chemical environment in both redox states of the enzyme.
Although the theoretical bases for the origin of MCD spectra is under active investigation currently the most promising application of the method depends upon coniparisons and correlations drawn from spectra of compounds of known chemical composition, · valence and spin-state. By this means we have established that about 50% of the heme a in the resting enzyme is low-spin with the balance presumably in the high-spin state.
Recent EPR data has established that 'the species with g-values of 3.03, 2.21 and 1.45, typical of low-spin heme, has an intensity equivalent to one 10\v-spin heme/mole enzyme (14,16). Beinert and coworkers (13,15,16,42) have assigned this low-spin resonance to 3+ cytochrome ~ since it appears to be the site for ferrocytochrome ~ oxidation. Correlating these data we conclude that the MCD derivative curve centered at 427 nm in the resting enzyme arises solely from low-spin cytochrome a 3 +, the contribution from high -' > 3+ spin cytochrome a 3 being too small to be identified. Moreover, the lineprity of the peak and trough HCD intensities as a function of 1/T (Fig. 2)   . 3+ the EPR detectable copper (CuA ) are magnetically isolated while cytochrome a 3 and the EPR undetectable copper (Cu~+) are antiferromagnetically coupled to give an S = 4/2 ground state; it was sLggested that the antiferromagnetic coupling is mediated by an imidazole histidine in superoxide dismutase (48). \-lith these assignments for the metals it is possible to account for the published magnetic susceptibility (49,50) of the enzyme and the behavior of the various, EPR species observed during a reductive titration of the enzyme (16). The model predicts a value of 32 forJA!ff which decrease~ only slightly, to 24-29, on full reduction.
Because of its even spin the antiferro~agnetically coupled conventional operating conditions; addition of.cyanide converts pa:Lr w'ill not exhibit EPR 3+ a 3 to a low-spin species undE but need not eliminate the antiferromagnetic coupling between the iron and copper. The apparent discrepancy bet\veen the_ typical ferric low-spin NCD spectrum ob_served for cytochrome a;+•CN and the lack of a corresponding low-spin EPR signal in the CN-treated oxidized enzyme may reside in the extremely low temperatures required for the EPR measurements, i.e.

3+
2+ at 0° C the coupling bet~veen low spin ~3 and CuB may be sufficiently weak so that both the S = 0 and S = 1 coupled states are occupied, thus allowing detection of a paramagnetic state (S = 1) by MCD. However, the lmv temperature of the EPR measurement may result in extensive occupancy of only the S = 0 state. We are currently exploring this possibility by MCD. Upon partial reduction of the cyanide treated enzyme both MCD and EPR measurements 3+ detect the low spin cytochrome a 3 •CN complex presumably since the CuB has been reduced M t' under these c~nditions and a 3 •CN now behaves as an isolated and typical S =_1 2 spin system.
Our temperature dependence data on the MCD of the oxidized enzyme was obtained over K the ra~ge (77°-273~) and overlaps the magnetic susceptibility results of Tsudzuki and .; Okunuki (48) and the EPR data of Beinert and coworkers (10°-77°) (13,16). All of the , data are consistent with the absence of a thermal spin state equilibrium in the range of interaction is considerably larger than 200 em This conclusion coupled Hith the recent data on the intensity of the low-spin EPR (14,16) argue strongly against a model in which the t~.;ro heroes of cytochrome oxidase and/or one of the coppers are coupled to give rise to a g = 3 signal (5,7)~ The MCD spectrum of the reduced enzyme is nost simply interpreted as the sum of two 2+ components; an intense hemoglobin-like spectr m from high-spin ferrous cytochrome a 3 with a positive peak at 446 nm, and a weaker, low-spin ferrous cytochrome-like spectrum Faraday £ term for the high-spin cytochrome a 3 (446 nm) and a Faraday A term from the diamagnetic low-spin ferrous cytochrome a 2 + (452 nm). Analogous temperature dependence studies on deoxymyoglobin shm-1 that the HCD intensity for this S = 2 system increases with the Boltzmann factor whereas ferrocytochrorae _£ (S = 0) shows only slight temperature effects (lq -21).
Several recent· models for cytochrome oxidase have invoked a strong heme-heme interaction to explain certain redox and ligand binding properties of the enzyme (4). These interactions are seen most clearly upon·ligand binding in which changes of up to 70% in extinction, coefficients for cytochromes ~and ~3 are postulated to occur (8). The data we have presented here argue against the existence of any heme-heme interactions which modify spectral properties to any great extent. This is seen most clearly for the MCD , and further weakens the credibility of "heme-heme" interaction manifested by extensive interdependence of spec~ral properties.
The interpretation of CD spectra of oxidase and its derivatives is controversial.
Myer (47) arguing from the lack of any exiton-type resonance interact:J_on in the observed ! CD spectra could find no evidence·for heme-h~e interaction, while Tiesjema and van Gelder (46) using difference techniques modeled after Yonetani (43) find small differences in both wavelength extrema and ellipticity for difference spectra obtained with various ligands and co~clude that heme-heme interaction must exist. However, it should be noted that CD is remarkably sensitive to variations in the symmetry of the heme environment, . and that the origin of rotational strength may arise from coupling with transitions in chromophoric groups distant from the heme moiety (51), even in different subunits (52); MCD n the other hand reflects the electronic structure of the heme group itse~f and hence is sensitive only to those perturbations which affect the iron or porphyrin orbitals directly.
In addition, some heterogeneity in ligand binding properties are invariably present with purified preparations of cytochrome oxidase (53). The small differences in both CD and }1CD spectra of cytochrome_oxidase are nore si~?ly attributable to effects such as these, rather than to the more profound alternative of heme-heme interaction. We do not mean      MCD (upper) and absorption (lower) spectra for oxidized cytochrome oxidase.
The enzyme was dissolved in 0.1 N potassium phosphate buffer, pH 7.4, containing 0. 5% Tween 80. The temperature was maintained near 0° C.     subtracting the HCD spectrum recorded at 0° C from that observed at -145° C.
The smoothing routine was used to enhance signal to noise .
. · Fiit:__]__. .MCD spectra of reduced cytochrome oxidase (---) and the CO ( · • • • ·) and CN-( ) complexes of the reduced enzyme. The buffer system consisted of 0.1 M potassium phosphate, pH 7.4, containing 0.5% potassium chelate. The enzyme was reduced with a few crystals of dithionite; CO gas or neutralized potassium \ cyanide (6 mH) were added to form the respective inhibitor complexes.  Reduction was achieved using a few crystals of dithionite; reoxidation by bubbling air thru the enzyme solution. The smoothing routine was used to enhance signal to noise.