Cytochrome aa3 of Rhodobacter sphaeroides as a Model for Mitochondrial Cytochrome c Oxidase PURIFICATION, KINETICS, PROTON PUMPING, AND SPECTRAL ANALYSIS* for site-directed mutational

Aerobically grown Rhodobacter sphaeroides synthe- sizes a respiratory chain similar to that of eukaryotes. We describe the purification of the aa3-type cytochrome c oxidase of Rb. sphaeroides as a highly active ( Vmax 2 1800 s-’), three-subunit enzyme from isolated, washed cytoplasmic membranes by hydroxylapatite chromatography and anion exchange fast protein liquid chromatography. The purified oxidase exhibits bi- phasic kinetics of oxidation of mammalian cytochrome c, similar to mitochondrial oxidases, and pumps protons efficiently (H+/e- = 0.7) following reconstitution into phospholipid vesicles. A membrane-bound cytochrome c is associated with the aa3-type oxidase in situ, but is removed during purification. The EPR spec- tra of the Rb. sphaeroides enzyme suggest the presence of a strong hydrogen bond to one or both of the histidine ligands of heme a. In other respects, optical, EPR, and resonance Raman analyses of the metal centers and their protein environments demonstrate a close correspondence between the bacterial enzyme and the struc-turally more complex bovine cytochrome c oxidase. The results establish this bacterial oxidase as an excellent model system for the mammalian enzyme and pro- vide the

)I To whom correspondence should be addressed. Tel.: 517-355-while the remaining 10 are of nuclear origin (3). The complexity of its physical and genetic structure has made structure/function analysis at a molecular level difficult. A number of aerobic bacteria synthesize cytochrome c oxidases that are simpler in structure, but have significant homology to the mitochondrial subunits of the eukaryotic enzymes (4, 5 ) . These bacterial oxidases can be more readily analyzed by sitedirected mutagenesis. Clearly, it is desirable to study a bacterial oxidase with a high degree of structural and functional homology to mitochondrial cytochrome oxidase. In addition, most rapid progress will result if the host cell can be easily manipulated by the techniques of molecular genetics.
We have chosen to develop Rhodobacter sphueroides as a model system for the structural and mutational analysis of cytochrome c oxidase. As a member of the cy subgroup of the purple bacteria from which mitochondria arose (6), aerobically grown Rb. sphaeroides synthesizes a respiratory electron transfer chain similar to that of mitochondria (7), including an aaa-type cytochrome c oxidase (8,9). Studies of this bacterium's photosynthetic apparatus and bcl complex have led to the development of a variety of tools for engineering its genome and for the introduction and expression of genes (10).
Three genes with a high degree of homology to the mitochondrial genes for subunits I, 11, and I11 of cytochrome c oxidase have been cloned from the Rb. sphaeroides genome and sequenced (11,12,86). The DNA-derived amino acid sequences of these genes show 52% identity with bovine subunit I ( l l ) , 39% with subunit I1 (12), and 49% with subunit I11 (86). The arrangement of the genes and the level of homology with eukaryotic cytochrome U U :~ is much like that of the aas-type cytochrome c oxidase of Paracoccus denitrificans (13,14). When the gene for subunit I or subunit I1 is interrupted or deleted from the Rb. sphaeroides genome, it causes the disappearance of cytochrome aa3 from the cytoplasmic membrane (11,86). Replacement of the wild-type genes on freely replicating plasmids restores oxidase synthesis to normal levels.
Here we report the characterization of a highly active, three-subunit cytochrome c oxidase from Rb. sphaeroides, purified by anion exchange chromatography in lauryl maltoside. This enzyme pumps protons efficiently, and its optical, EPR, and resonance Raman spectra demonstrate its structural homology to bovine cytochrome c oxidase. A previously reported isolation yielded a two-subunit enzyme that appeared incapable of proton pumping (15). Our results clearly establish the oxidase of Rb. sphaeroides as an excellent bacterial model for examining the structural basis of the catalytic function of the more complex eukaryotic enzyme. Indeed, we have recently used this system to define the specific ligands of heme a and the heme as-CuB center by site-directed mutagenesis (16).
Bacterial Growth-Rb. sphaeroides CY91, a derivative of strain Ga in which the gene for cytochrome bhW2 has been inactivated (obtained from C. Yun and R. B. Gennis, University of Illinois), was grown in the dark in a 12-liter fermentor at 30 "C on medium A of Sistrom (20) plus 25 pg/ml kanamycin. The cultures were vigorously stirred, sparged with air supplemented to 30% 02, harvested at late exponential phase, and stored frozen at -70 "C.
T h e cells were broken by two passages through a French pressure cell at 20,000 Ib/in'. Whole cells and debris were removed by centrifugation at 20,000 X p for 20 min. Following the addition of PMSF to 0.4 mM, and pepstatin and leupeptin to 1 pg/ml each, the membranes were pelleted by centrifugation a t 200,000 X g for 1.5 h. The membranes were resuspended in KH,PO,/EDTA plus protease inhibitors as above, homogenized with a glass homogenizer and Teflon pestle, and centrifuged as above. The membrane pellet was resuspended in KH,PO,/EDTA, 5% sucrose, 0.1 mM PMSF, 0.5 pg/ml pepstatin, and leupeptin, homogenized, and layered onto sucrose step gradients composed of 8 ml of 25% sucrose over 13 ml of 50% sucrose (w/v) in KH,PO,/EDTA. Following centrifugation a t 150,000 X g for 4 h, cytoplasmic membranes were collected from the 25%/50% sucrose interface, and the sedimented outer membrane was discarded. The membrane suspension was diluted in several volumes of KH,PO,/ EDTA, centrifuged at 200,000 X g for 1.5 h, and the pellets were stored frozen a t -70 "C. Typically, these preparations yield 0.8-1 mg of membrane-bound cytochrome aa3 per liter of culture (based on reduced minus oxidized spectra) a t a concentration of about 1 nmol of heme A/mg of protein.
Chaotropic Salt Wash of Cytoplasmic Membranes-Membrane pellets were resuspended to a concentration of 1.5 mg of protein/ml in KH,PO,/EDTA, 0.5 mM PMSF, 2 pg/ml leupeptin, and pepstatin and homogenized as above. Solid NaBr was added to 2 M (w/v) and dissolved by inversion, and the resuspension was passed twice through a French pressure cell a t 16,000 lbs/in2. The suspension was diluted with an equal volume of KH2P04/EDTA and centrifuged at 200,000 X p for 2 h. The membranes were washed free of NaBr by resuspension and homogenization in KHsPO,/EDTA, followed by centrifugation as above. of protein/ml. Lauryl maltoside was added to a final concentration of 4%; the solution was stirred for 10 min at 0 "C, followed by the addition of CHAPS to 4% final concentration, and continued stirring for 10 min. Undissolved material was removed by centrifugation a t 40,000 X g for 20 min. The solubilized membranes were loaded onto a Bio-Gel H T P (Bio-Rad) column (3.3 ml of packed hydroxylapat,ite per mg of cytochrome aa3) which had been pre-equilibrated with 2.5 volumes of 10 mM KH2P04, pH 7.0, 0.1 M KC1, 1 mM EDTA, 0.2% lauryl maltoside, and 0.6% CHAPS. After loading, the column was washed with the above buffer to remove less tightly bound material (including photosynthetic pigments). The cytochromes were eluted with 0.3 M KH2P0,, pH 7.0, in column buffer a t 0.5 ml/min. The cytochrome-containing fractions were made 1% in lauryl maltoside and exchanged into 10 mM KHzPO,, pH 7.0, 1 mM EDTA, 0.05 mM PMSF, 0.2 pg/ml leupeptin, and pepstatin using a Centriprep 30 concentrator (Amicon).
Isolation of Cytochrome aa3-The concentrated cytochrome fraction, containing approximately 6 mg of cytochrome aa:?, was made 3% in lauryl maltoside, stirred for 30 min a t 0 "C, and applied to tandem DEAE-5PW columns (7.5 mm X 7.5 cm, Tosoh Corp.) using an FPLC system (Pharmacia LKB Biotechnology Inc.). The columns were preequilibrated with 10 mM KH,PO,, pH 7.6, 1 mM EDTA, 0.2% lauryl maltoside, and the bound proteins were eluted with a 48-ml linear gradient of 0-0.4 M KC1 in the same buffer at a flow rate of 0.4 ml/ min. Peak fractions of cytochrome aa,, determined by optical spectra, were combined, brought to 3% in lauryl maltoside, stirred for 30 min at 0 "C, diluted to 100 mM KC1 with the column buffer, and reapplied to the DEAE-5PW columns, and the bound proteins were eluted with a 36-ml linear gradient of 0.1-0.4 M KC1 at 0.4 ml/min. Peak fractions were collected and stored a t -70 "C.
Reconstitution of Cytochrome c Oxidase into Phospholipid Vesicles-All glassware was rinsed with ethanol. Asolectin was suspended to 40 mg/ml by sonication in 2% cholate, 75 mM HEPES-KOH, pH 7.4, a t 0 "C under argon. A Heat Systems-Ultrasonics sonicator (Model W-225), equipped with a microtip, was used a t a power setting of 5 for intervals of 30 s on, 30 s off, until clarity was reached. The suspension was centrifuged for 15 min a t 12,000 X g to remove titanium particles. Purified cytochrome c oxidase in 0.2% lauryl maltoside was added to the phospholipid suspension to a final concentration of 2 nmol/ml and dialyzed with rapid stirring at 4 "C ix Spectrapor dialysis tubing (number 25225/204, 12-14,000 M , cutoff) by the following protocol: Gel Filtration-A GF-450 HPLC column (Du Pont) on the FPLC system was equilibrated and developed with 0.2 M KH,PO,, pH 7.6, 1 mM EDTA, 0.2% lauryl maltoside.
Protein and Heme A Determinations-Protein concentrations were determined by the bicinchoninic acid method (21) and a modified Lowry assay (22). Dithionite-reduced minus ferricyanide-oxidized pyridine hemochrome spectra were obtained after adding pyridine to 10% and NaOH to 0.2 N to an oxidase sample. The concentration of heme A was determined by the method of Berry and Trumpower (23).

RESULTS AND DISCUSSION
Purification of Rb. sphaeroides Cytochrome c Oxidase-Aerobically grown Rb. sphaeroides 2.4.1 cells synthesize approximately four times more heme b and c containing cytochromes than cytochrome uu3, in contrast to mitochondrial inner membranes where the amounts of each are nearly equal (Fig. 1, C and D). During a screen of several genetically engineered strains (obtained from R. B. Gennis and colleagues, University of Illinois), it was discovered that Rb. sphaeroides CY91, a strain in which the gene for cytochrome bSe2 is interrupted by an antibiotic resistance gene, synthesizes approximately twice as much cytochrome ua3 as 2.4.1 (0.09 mg of aa,/g of wet weight cells in 2.4.1; 0.2 mg of aa3/g in CY91; Fig. 1B). Because of this increased expression of cytochrome aaa, we use CY91 as the source of the oxidase. Cytochrome aa3 prepared from CY91 and from 2.4.1 appear identical. Bubbling with oxygen-supplemented air during the growth of CY91 allows for increased synthesis of cytochrome aa3 and de- creased synthesis of the photosynthetic light harvesting and reaction center complexes. The latter bind tightly to the anion exchange column and, when present in sufficient quantities, interfere with the resolution of the aa:,-type oxidase.
Purified cytoplasmic membranes, prepared as under "Experimental Procedures," are washed with NaBr in order to remove the F, component of the ATP synthase, which otherwise persists as a significant contaminant. The cytochrome content of the washed membranes is only slightly altered (Fig.  IA). Urea, guanidine HCI, and high concentrations of EDTA are not as effective as NaBr in the removal of F1. We attempted to selectively solubilize and remove the bcl complex using low levels of various detergents, including short and long chain alkyl glycosides and bile salt derivatives. However, both cytochromes bcl and aus were solubilized to similar extents by these procedures, perhaps because of the presence of a membrane-bound cytochrome c that associates with the two complexes (see below).
Separation of the cytoplasmic membrane proteins on the anion exchange column appears to depend upon prior removal of excess lipid and residual pigment-containing proteins. We accomplished this by solubilization of the membranes in high concentrations of lauryl maltoside and CHAPS, followed by hydroxylapatite chromatography. The inclusion of CHAPS appears to be necessary to promote efficient binding of cytochrome U U :~ to hydroxylapatite. While this step does not separate the various cytochrome species under the conditions described, it enhances the resolution of the subsequent DEAE-5PW chromatography.
The key features of the first DEAE-chromatography step are the continued use of high levels of lauryl maltoside, loading at low ionic strength, and the use of tandem columns. The bc, complex elutes at 260 mM KCl, the cytochrome aa:, peak at 290 mM KC1, and a cb-type cytochrome c oxidase at 340 mM KC1, as a small shoulder (Fig. 2).
Some of the bcl remains associated with the aa:,-type oxidase, and this mixture elutes at 275 mM KC1, immediately before the aa:{ peak. A repeat chromatography on DEAE of the peak fraction shown in Fig delivered to the top of the column. Since protein (280 nm) is monitored at the bottom of the columns, a void volume of -5 ml must he subtracted from the value on the x-axis in order to determine the KC1 concentration that initiates the elution of a peak. The cytochrome c oxidase fractions collected for further purification or final analysis are shaded.
an overall yield, from cytoplasmic membranes to final product, of approximately 30% (Table I).
A slightly less pure enzyme is prepared without hydroxylapatite chromatography by solubilizing the membranes in 5% lauryl maltoside (at IO mg/ml total protein), loading onto the DEAE-5PW column, and then washing with 3% lauryl maltoside in phosphate buffer prior to the salt gradient. Both enzyme preparations are fully active in electron transfer and proton translocation (see below), and the yield of either procedure is similar.
Values of 19.3 and 17.9 nmol of heme A/mg of protein are obtained for the purified oxidase when protein is determined by the bicinchoninic acid (21) and modified Lowry (22) assays, respectively, and the heme A content is obtained from the pyridine hemochrome spectrum. These values are 121 and 113% greater than a calculated maximum value of 15.9 nmol of heme A/mg of protein, using the molecular weights predicted by the gene sequences and assuming two hemes A per monomer. Since polyacrylamide gels show no apparent loss of any of the subunits during the purification, we conclude that the protein assays are underestimating the actual amount of protein.
Subunit Composition and Size-Polyacrylamide gel electrophoresis of the purified oxidase shows three subunits with apparent molecular masses of 49,37, and 22 kDa (Fig. 3). The predicted molecular masses, calculated from the DNA-derived amino acid sequences (11,12,86), are 62.6, 32.9, and 30.1 kDa for subunits I, 11, and 111, respectively. As noted for the mitochondrial oxidase, subunits I and 111 electrophorese at an anomalously rapid rate due to their extreme hydrophobicity.  Limited N-terminal amino acid sequencing (not shown) confirmed the assignment of the 49-kDa peptide as subunit I, by comparison to the predicted amino acid sequence. The 37-kDa peptide is identified as subunit I1 based on its reaction with antisera to subunit I1 of yeast cytochrome c oxidase (data not shown). It has previously been shown that this peptide reacts with antibodies to subunit I1 of cytochrome c oxidase from P. denitrificans (15). Cleavage of a -2-kDa fragment from subunit I1 is not completely prevented by the use of a broad spectrum of protease inhibitors during preparative procedures. Chromatography on DEAE-5PW resolves oxidase fractions containing mostly cleaved or mostly intact subunit 11, with the enzyme containing the full-length peptide eluting ahead of that with the cleaved peptide. We find no differences in the electron transfer activity of these fractions. The shorter peptide (35 kDa), previously assigned as subunit I11 (15), also reacts with antisera to yeast subunit 11. One interesting possibility is that the larger subunit I1 peptide (37 kDa) retains its presumed signal sequence (12) after insertion into the membrane. The signal sequence has overall basic character, and thus its presence would account-for the earlier elution of the oxidase with the full length peptide from the anion exchange column.
The peptide with an apparent molecular mass of 22 kDa is 9.7 23.3 30 specifically labelled with ['"CIDCCD under the same conditions as subunit I11 of the mitochondrial oxidase (Fig. 3), establishing its relationship to eukaryotic subunit I11 (29) and eliminating the possibility that this peptide is the membranebound cytochrome c which appears to be present in the bacterial membrane (see below). Membrane-bound cRP2 of P.
denitrificans has a molecular mass of 22 kDa and can remain associated with this bacterium's oxidase during purification (30). Variable amounts of a contaminating 68-kDa peptide are also seen in SDS gels. Gel filtration of the purified oxidase, with its associated lipid and detergent, gives an estimated molecular mass between 350 and 450 kDa depending on the conditions and the standards used (data not shown). A predicted molecular mass of a monomer, with -100 kDa of associated detergent, would be 225 kDa. Whether the observed species is a monomer with high levels of associated lipid, or a dimer, cannot be ascertained from gel filtration alone, since anomolous behavior of membrane proteins is often observed (31).
Optical Spectra-The visible spectra of the purified oxidase (Fig. 4) strongly resemble those of beef heart cytochrome oxidase (32, 33), as can be seen if spectra of the two enzymes are overlaid on the same spectrophotometer. This indicates similar protein environments for the hemes of the two enzymes. There is, however, a small red-shift of the absorbance maxima such that the spectrum of the reduced bacterial oxidase shows peaks a t 444.5 and 606 nm (Fig. 4). The magnitude of the shifts, coupled with the fact that the height and width of the peaks are identical to those of the beef heart oxidase spectrum, suggests that the absorbance of both hemes of the reduced enzyme is shifted, resulting in a 1-nm shift in the Soret and a 2-nm shift in the a-band relative to bovine oxidase (32, 33). The Soret to a-band ratio (A443..190nm/AROR. ) of the reduced bacterial oxidase is 5.8, as for the bovine enzyme. The CO difference spectrum (Fig. 4, inset) shows peaks at 430.5 and 591.5 nm, also red-shifted in comparison to a beef heart spectrum. The a peak of the oxidized enzyme appears a t 599 nm, while the Soret peak varies between 421 and 424 nm in different preparations. A ratio of Azw.., to Ad?.+ " , , , of 2.2 is obtained for the purified bacterial oxidase; the higher values of 2.6-3.0 observed for beef heart cytochrome oxidase (31, 34) are consistent with its higher molecular weight. The measured extinction coefficient values (At (606-650 nm, reduced) = 41 cm" mM"; A< (444-480 nm, reduced) = 213 cm" mM") are essentially the same as those derived for beef heart cytochrome oxidase (32, 33). Somewhat lower extinction coefficients have been reported for the homologous cytochrome oxidase of P. denitrificans (35); however, these differences are due to the choice of the extinction coefficient used to calculate the heme A concentration from pyridine hemochrome spectra.
Actiuity-The turnover number of the purified oxidase, using 30 PM horse heart cytochrome c, is approximately 1600 s"; the omission of soybean phospholipids from the assay

sphaeroides. Dithionite-reduced (solid lines)
and air-oxidized (dashed line) absolute spectra of the purified enzyme were recorded as in Fig. 1 in a solution containing 50 mM KH2P04, pH 7.2, 1 mM EDTA, and 0.2% lauryl maltoside. To obtain the CO difference spectrum (a"a:r"-CO minus a'+,:;'+; inset), 1 ml of CO was bubbled over 5 min into 0.5 ml of dithionite-reduced cytochrome ua3 a t 25 "C i n dim light. This protocol results in the complete saturation of the oxidase with CO. decreases this value by approximately 40%. At earlier stages of the purification, turnover numbers up to 2000 s-' are obtained in the absence of added phospholipids. Thus, the loss of lipids during the purification likely contributes to the decrease in activity.
This enzyme is unique in that it shows the highest electron transfer activity of any purified cytochrome c oxidase, especially when comparing the ability of the bacterial oxidases to utilize mammalian cytochrome c as a substrate. A previous preparation of the Rb. sphaeroides oxidase as a two-subunit enzyme in Triton X-100 (15) had significantly less activity (V,,, = 500 s-') than the three-subunit enzyme presented here and a preparation of the homologous oxidase of Paracoccus, as a three-subunit enzyme, also shows lower activity (36). Our own preparation of Paracoccus cytochrome aa3, using the method presented here, shows a turnover of 800 s" with 30 h~ horse heart cytochrome c. While oxidase turnover appears rarely to be limiting for bacterial growth, the maintenance of high turnover during purification is an important indicator of a n intact enzyme.
Kinetics of Interaction with Soluble Cytochrome c-The purified bacterial oxidase exhibits biphasic kinetics of cytochrome c oxidation (Fig. 5) similar to beef heart cytochrome c oxidase (37). The estimated K,, and Km2 values for the high and low affinity reactions in 50 mM KH2P04, pH 6.5, are 0.026 and 3.7 PM respectively, nearly identical to those of beef heart cytochrome c oxidase measured under the same reaction conditions (Fig. 5, inset). However, VmaXl and Vmaxa (96 and 1800 s-l) are three to four times greater.
These biphasic kinetics indicate that this three-subunit enzyme is capable of high and low affinity interactions with 0 1000

TN (SI)
FIG. 5. Eadie-Hofstee plots of the kinetics of oxidation of cytochrome c by purified Rb. sphaeroides and beef heart cytochrome c oxidase. Oxygen consumption was measured polarographically using a Gilson Model 5/6H oxygraph a t 25 "C in a reaction medium containing 50 mM KH2P04, pH 6.5, 0.05% lauryl maltoside, 1.1 mg/ml asolectin, 2.8 mM ascorbate, 0.55 mM TMPD, and 2.2 nM Rb. sphaeroides cytochrome c oxidase (main plot) or 7.8 nM beef heart cytochrome c oxidase (inset). The concentration of horse heart cytochrome c was varied from 0.02 to 45 p~. The very slow rate of ascorbate oxidation seen in the absence of enzyme was subtracted from each assay before the turnover numbers ( T N ) were calculated as described in  soluble cytochrome c, supporting the idea that nuclear encoded subunits of the eukaryotic enzymes are not essential for biphasic behavior (38). Other investigators have reported both biphasic (36) and monophasic (39) kinetics with the homologous enzyme of P. denitrificans, under different experimental conditions, while Fukomori and Yamanaka (40) found distinctly biphasic kinetics for the weakly active, two-subunit aa3-type oxidase of Nitrobacter ugilis. There is, as yet, no definitive interpretation of this kinetic behavior; two binding sites for cytochrome c (37) or two forms of the oxidase with different cytochrome c reactivity (41) are among the hypotheses proposed.
Interaction with Membrane-bound Cytochrome c-While the purified enzyme oxidizes TMPD only very slowly in the absence of soluble cytochrome c (T, = 7 s-'), cytoplasmic membrane vesicles catalyze a substantial rate of TMPD oxidation (Fig. 6A). This reaction is apparently due to the presence of a membrane-associated cytochrome c (cmem) that is bound to cytochrome aa3 in situ, but lost upon purification. In addition, approximately 40% of the TMPD oxidase activity is due to the presence of a cb-type cytochrome c oxidase in these membranes. This enzyme, which contains an attached cytochrome c even after purification, is an efficient TMPD oxidase, as determined by our experiments with purified cytochrome cb and with membranes of a Rhodobacter strain (JS100) from which cytochrome aa3 is genetically deleted (11).
Cytochrome aa3 in situ also supports high rates of electron transfer from added mitochondrial cytochrome c (Fig. 6B). The rate of this reaction is diminished by high concentrations of KC1, similar to the interaction of mitochondrial cytochrome c and cytochrome c oxidase. In contrast, TMPD-mediated electron transfer through cmem (Fig. 6A) is unaffected by KC1, suggesting tight, nonelectrostatic binding between c, , , and cytochrome aa3.
Thus, the association, in situ, of a membrane-bound cyto-chrorL>e c (c,,,) with Rb. sphaeroides cytochrome aa3 is indicated by 1) the high rates of oxygen reduction supported by TMPD alone in membrane vesicles, 2) the relative insensitivity of this rea-tion to ionic strength, and 3) the loss of TMPD oxidase activity during purification. In P. denitrificans, mem- Effect of ionic strength on the rate of electron transfer from TMPD or soluble cytochrome c to the membranebound cytochrome c oxidase. Oxygen consumption catalyzed by cytoplasmic membranes of Rb. sphueroides CY91 was measured as in Fig. 5 in a reaction medium containing 50 mM KH,PO,, p H 6.5, 2.8 mM ascorbate, 0.083 mM TMPD ( A ) , or 0.083 mM TMPD plus 7.8 p~ horse heart cytochrome c ( B ) and a quantity of intact membranes containing 16.3 pmol of cytochrome uu3. Membrane pellets (see "Experimental Procedures") were resuspended to 5 mg protein/ml in 50 mM KH,PO,, pH 7.2, 1 mM EDTA, 5% sucrose, 0.2 mM PMSF. A subsaturating concentration of TMPD was used in this experiment; 2.2 mM TMPD supports a rate of 150 nmol 02/min. In order to calculate the rate of electron transfer due to only soluble cytochrome c ( R ) , the rate of oxygen consumption catalyzed by 0.083 mM TMPD alone has been subtracted. Separate experiments (not shown) indicate that under these conditions the contributions of the cb-type cytochrome c oxidase, also present in these membranes, to the TMPD and cytochrome c oxidation reactions are 40 and 20%, respectively. These contributions have not been subtracted.
brane-bound cytochrome css2 apparently mediates electron flow between cytochromes bcl and a u 3 , and can be isolated in a ternary complex of same (30). Given the high degree of homology between the respiratory systems of P. denitrificans and Rb. sphaeroides, cmem is likely to have a similar role. Conversely, soluble cytochrome cz may not be required for respiratory electron flow (42,43), as indicated by the fact that the gene for soluble cq of Rb. sphaeroides can be deleted without affecting chemoheterotrophic growth (44).
Proton Translocation-Purified Rb. sphaeroides cytochrome aa:3 can be reconstituted into phospholipid vesicles showing high respiratory control (respiratory control ratio = 4-10). The addition of reduced horse-heart cytochrome c leads to immediate acidification of the external medium, followed by a slow alkalinization due to the consumption of protons during water formation (Fig. 7). A proton ejection stoichiometry of 0.7 H+/e-is achieved with 1.6 turnovers of the enzyme (0.63 nmol of reduced cytochrome c added to 0.1 nmol of au3). The addition of 5 WM CCCP, which allows rapid equilibration of protons across the lipid bilayer, greatly accelerates the rate of alkalinization, which exhibits a stoichiometry of 1.0 OH-/e-. Since the uncoupler allows access to the additional buffering inside the cytochrome oxidase vesicles (COVs), separate additions of the acid standard are used to quantitate proton ejection and net alkalinization. With attention to the concerns set forth by Casey et ul.(27) for the validity of a proton pumping experiment, we performed a series of control experiments with the following results. First, two consecutive pulses of cytochrome c to the uncoupled COVs give the same extent of alkalinization. Second, the addition of sodium azide eliminates any measurable proton ejection and inhibits alkalinization by greater than 99%, indicating that the observed acidification is not due to proton release upon binding of Following their addition to a stirred cell, 110 pl of COVs, containing 0.1 nmol of cytochrome uu3, were equilibrated with 3.2 KM valinomycin and 0.4 nM CCCP until the base line stabilized. Immediate acidification of the medium occurred upon injection of 0.63 nmol of reduced horse heart cytochrome c ( A ) ; the pH of the cytochrome c solution was previously adjusted using blank vesicles to avoid artifactual absorbance changes. A second addition of cytochrome c ( B ) , injected after the complete uncoupling of the COVs with 5 PM CCCP, initiated rapid alkalinization of the medium. Additions of 0.05 nmol of HC1 were used to calibrate the absorbance change both before and after the addition of CCCP. In an otherwise identical experiment (not shown), reconstituted beef heart cytochrome c oxidase pumped protons with a stoichiometry of 0.7 H+/e-. cytochrome c to the COVs. Third, examination of the kinetics of cytochrome c oxidation by the COVs, followed as the decrease in absorbance at 550 nm, shows rapid and complete oxidation of the added cytochrome c, while control vesicles give an insignificant rate.
The H+/e-value of the reconstituted bacterial oxidase (0.7) is similar to that measured for bovine cytochrome c oxidase (27,28). An earlier preparation of Rb. sphueroides cytochrome aa:3, lacking subunit 111, reconstituted with good respiratory control but could not be demonstrated to pump protons (15). Possibly relevant differences in the current preparation include the retention of subunit 111, the high turnover of the enzyme and the use of lauryl maltoside.
EPR Spectroscopy-The EPR spectrum of oxidized cytochrome c oxidase of Rb. sphaeroides is presented in Fig. 8. The g = 2 region is similar to that of mitochondrial oxidases (45,46), once the presence of two additional signals (discussed below) are taken into account. Along with the 830-nm absorption band in the optical spectrum of the oxidized enzyme (not shown), the EPR spectrum indicates the presence of a typical CuA site, consistent with the presence of the conserved histidines, cysteines, and methionine thought to bind CUA in subunit I1 (12).
The first of the additional signals, at g = 2.19, is present at 10 K but disappears upon warming to 108 K (Fig. 8, inset).
This behavior indicates it is not a characteristic of the CuA spectrum, but its origin is unknown. A set of signals appears at g = 2.15, 2.10, 1.92, and 1.86, and remains present at 108 K. These signals, the most obvious of which is the g = 2.15 resonance (Fig. 8, inset), are due to the presence of tightly bound Mn'+ (47, 48) that is not removed by treatment with 50 mM EDTA. A typical six-line hexaqua Mn'+ spectrum (not shown) is obtained upon acidification of an oxidase sample, from which a Mn2+ content of approximately 1 per 20 oxidase monomers is estimated.
Substoichiometric Mn'+ appears ubiquitous among bacte- FIG. 8. EPR spectra of purified Rb. sphaeroides and beef heart cytochrome 003. EPR spectra of 50 PM Rb. sphueroides or 100 FM beef heart samples were recorded at X-band using a Bruker ER2OOD spectrometer equipped with a TE,,, cavity. An Oxford Instruments ESR-900 liquid helium cryostat or a Wilmad N2 flow system was used to maintain sample temperature. The g values were determined from direct measurements of the magnetic field strength and microwave frequency using a Bruker ER035M gaussmeter and a Hewlett-Packard 5245/5255 frequency converter and counter. The modulation amplitude was 12.5 gauss (peak-to-peak), the sweep time was 200 s, and the time constant was 0.5 s. The main plot is an average of four scans of the Rb. sphaeroides enzyme at 10 K, 2milliwatt microwave power at 9.7344 GHz. The inset compares the E P R spectra of Rb. sphaeroides (top; six scans) and beef heart (bottom; four scans) cytochrome c oxidase obtained a t 108 K with 20-milliwatt microwave power at 9.5944 GHz. Relevant g values are indicated by the arrows. The amplitude and area of the g = 2.83 signal of Rb. sphaeroides heme a and the g = 2.0 region of CuA correspond closely t o those of beef heart cytochrome c oxidase, indicating identical stoichiometries of heme a and CuA.
rial cytochrome c oxidases, although published spectra show that amounts vary (49-51). Atomic absorption and EPR spectroscopy of P. denitrificam cytochrome aa3 and the caa3-type oxidase of Bacillus subtilus show that these preparations contain up to four times the amount of Mn2+ present in the Rb. sphaeroides enzyme (47,48,51). In work to be presented elsewhere, we have recently identified and Asp4'* of Rb. sphaeroides subunit I as probable ligands of the Mn2+ (52).
The signals at g = 2.83, 2.31, and 1.62 (Fig. 8, 10 K spectrum) arise from low spin heme a and are considerably shifted from the g values of 3.03, 2.21, and 1.45 assigned to heme a of beef heart cytochrome oxidase (45,53). The homologous oxidase of P. denitrificam also shows these shifted g values (54, 55). These two bacterial enzymes apparently differ from bovine cytochrome c oxidase with respect to the local environment of heme a, even though many of the residues surrounding the axial ligands of heme a, recently identified as and H i P of Rb. sphaeroides subunit I (16), are highly conserved. This difference is not universal for bacterial aas-type oxidases, since the caa3-type cytochrome c oxidases of Thermus thermophilus and Bacillus species have g, values similar to bovine heme a (49,51,56). Extensive studies of low spin hemes in both model compounds and hemoproteins (57-61) allow us to conclude that the g values obtained for Rb. sphaeroides heme a result from an increase in electron donation by the axial histidines. Calculations of the ligand field parameters of Rb. sphueroides heme a (by the method of Taylor (62) using the axis system of Peisach et al. (58)) yield a tetragonality value ( A / h ) of 3.32 and a rhombicity value (./A) of 0.66. These values fall within the H region of the crystal field diagram of low spin hemes (58), a region that contains hemes known to be ligated by deprotonated imidazole or histidine. More specifically, when compared to model systems using heme A ligated by two neutral (gz = 2.96, A/h = 2.88) or two deprotonated (gz = 2.71, A/h = 4.08) imidazoles (53), the Rb. sphaeroides oxidase is intermediate between the two. One interpretation of these results is that heme a of Rb. sphaeroides is ligated by one neutral and one deprotonated histidine in the oxidized bacterial enzyme. Indeed, histidine-imidazolate derivatives of the hemes of soybean leghemoglobin a (63) and met-myoglobin (64) have g values nearly identical to those of Rb. sphueroides and Paracoccus heme a. However, the existence of a fully deprotonated histidine at pH 7.6 suggests an unusually low pK. for the nonliganded nitrogen of the imidazole ring; such a deprotonated structure may require a stabilizing ionic interaction (63,64). Since in all cytochromes of known structure the nonliganded nitrogen of each of the coordinating histidines is hydrogen bonded to the polypeptide chain (65)(66)(67)(68)(69), this is likely to be the case in cytochrome c oxidase. Therefore, another possibility is that one or both of the histidine ligands of Rb. sphueroides heme a forms a stronger hydrogen bond with a neighboring amino acid side chain than the corresponding hydrogen bond in beef heart cytochrome oxidase. Strong hydrogen bonding has been shown to have an effect similar to deprotonation in model compounds of low spin hemes (59, 60). Possible candidates for the hydrogen acceptor of the proposed hydrogen bond(s) are conserved residues Asp4'' and Asp407, both located in an interhelix loop that appears to interact with heme a (70).
In an attempt to discriminate between strong hydrogen bonding or deprotonation, we examined the effect of pH on the EPR spectrum. At pH 5.5 (the lowest pH at which the bacterial oxidase retains near normal activity, indicating the retention of native structure) we might expect an imidazolate side chain to be protonated, with an increase in g,, while a hydrogen bond to either of the histidines should be unaffected. In fact, the spectra obtained after the oxidase had been reduced, using ascorbate/TMPD, and reoxidized, as well as after extensive incubation at this low pH, were identical to that taken at pH 7.6. This result is more consistent with the presence of a strong hydrogen bond(s), since the turnover conditions should allow proton access and such a low pK, for the histidine-histidinate conversion seems unlikely. This apparent lack of proton dissociation from the imidazole side chains may also eliminate the attractive possibility that the unique g values of Rb. sphaeroides and Paracoccus heme a are related to one of the two heme a conformers described by Copeland and colleagues (71), since these conformers interconvert with a pK, of 6.6.
We have also considered, but rejected, the possibility that orientation of the axial histidines accounts for the difference in the heme a EPR spectra of Rb. sphaeroides and beef heart cytochrome c oxidase. First, the rhombicity values (0.87 for beef heart (41); 0.66 for Rb. sphueroides) suggest that the planes of the histidines are near parallel in both enzymes (59,72). Second, even a completely parallel orientation of the bacterial heme a ligands would not explain the data, since model compound studies indicate a lower limit value of g, = 2.9 for this geometry (53,59,72).
The strong hydrogen bond(s) suggested by the EPR data can also be invoked to explain the red shift of the heme a visible spectrum: the increased basicity of the histidine ligands may allow the the optical transitions of the heme to occur at lower energy. Small optical red shifts accompany the deprotonation of histidine or imidazole ligands in hemoproteins (73, 74) and model compounds (60,75,76). Of particular interest is the fact that both hemes a and a3 of the bacterial oxidase appear red-shifted. It may be that a single structural change affects both hemes, especially considering the proximity of the metal centers (16, 77).
Resonance Raman Spectroscopy-Overall, the resonance Raman (RR) spectrum of dithionite-reduced Rb. sphaeroides cytochrome c oxidase (Fig. 9A) is remarkably similar to that of beef heart oxidase (Fig. 9B); all but one of the bands can be assigned on the basis of the well characterized beef heart spectrum (78, 79). Of particular importance is the examination of those RR modes that reflect interactions of the protein with substituent groups on the hemes. The protein environment of 5-coordinate heme a3 is shown to be similar to that of mitochondrial cytochrome c oxidase by the presence of the Fe-Nhip. stretch at 214 cm", the ring bending mode at 365 cm", and the formyl stretch at 1662 cm" in the spectrum of the reduced bacterial enzyme. The same is true for the envi- FIG. 9. Resonance Raman spectra of purified Rb. sphaeroides and beef heart cytochrome c oxidase. RR spectra were obtained using Spex 1401 and 1877 spectrometers with photomultiplier and photodiode array detectors, respectively. Excitation a t 441.6 nm was provided by a Liconix HeCd laser operated at 10-milliwatt power. Cytochrome c oxidase samples (35 pM), in 100 mM KH,PO, p H 7.6, 1 mM EDTA, and 0.5-1% lauryl maltoside, were reduced with sodium dithionite under an argon flow and placed in capillary tubes for spectra acquisition. Temperature was maintained between 1 and 7 "C with a stream of cold N P gas and optical spectra were recorded before and after the RR experiment to ensure sample integrity and maintenance of the correct redox state. The RR spectrum of Rb. sphaeroides cytochrome nag ( A ) is compared to that of the beef heart enzyme ( B ) . Assignments of the peaks are in wave numbers (cm-I).
T o produce the mixed valence sample (U2+a3''+-CN-; spectrum C), the Rb. sphaeroides oxidase was incubated with sodium cyanide (5 mM) for 6 h before reduction with sodium dithionite. ronment of heme a, as shown by the presence of the formyl and vinyl stretches a t 1610 and 1624 cm", respectively. The enhanced intensity of the 1610 cm" mode may indicate that the formyl group of heme a conjugates more strongly with the porphyrin ?r system than the corresponding formyl of beef heart cytochrome oxidase (see Ref. 80).
Minor differences from the beef heart oxidase spectrum include a shift in a porphyrin bending mode, which appears at 342 cm" in beef heart but at 338 cm" in Rb. sphaeroides, and intensity changes at 239 and 214 cm". A unique mode at 1473 cm" appears in the spectrum of the reduced bacterial oxidase. This mode apparently arises from heme a since it remains present in the spectrum of the cyanide-bound enzyme (see below).
The spectrum of the cyanide-bound mixed valence oxidase (u'+u~~'-CN-; Fig. 9C) represents only heme a, since CNbinding shifts the absorbance maximum of heme a3 to 428 nm, well away from the excitation wavelength. As expected, we see the loss of the three modes specific to heme a3: at 214, 365, and 1662 cm" (78, 79). The 1610 cm" band is unaltered in the spectrum of the CN--bound enzyme; therefore, its enhanced intensity reflects a change in heme a and not a3.
The RR spectra of fully oxidized cytochrome aa3 and the oxidized cyanide-bound form (not shown) are nearly identical to those of mammalian oxidase (78, 81). One difference appears: the spin state marker of heme a, which occurs at 1590 cm" in beef heart oxidase, shifts to 1585 cm" in Rb. sphaeroides.
The heme a formyl stretch appears at 1648 cm" in the oxidized enzyme, as in beef heart cytochrome c oxidase (82,83), but unlike plant cytochrome oxidases in which this band is shifted to 1657 cm" due to a weaker hydrogen bond (80). In the reduced form of all three enzymes, the formyl stretch has the same frequency, 1610 cm", indicating a similar strength of hydrogen bond. The peak of the a-band absorbance, however, is different in each of these oxidases. Although a major determinant of the a-band position may be the strength of the formyl hydrogen bond (82,83), this result suggests that other factors must be involved, such as the increased basicity of the heme a ligands as discussed above. CONCLUSIONS We have purified the aa3-type cytochrome c oxidase of Rb. sphaeroides as a highly active, three subunit enzyme. The preparative procedure is well suited for examining mutant forms of cytochrome c oxidase prepared in Rb. sphaeroides since it allows high yield from a reasonable amount of cells (40-60 g from a 12-liter growth) and provides sufficient material for spectroscopic and biochemical analysis. Indeed, we have used it successfully to purify cytochrome c oxidase from a number of mutant strains of Rb. sphueroides (16,52). The procedure may be scaled up simply by increasing the FPLC column size.
In addition to presenting an efficient purification procedure, this work demonstrates the appropriateness of Rb. sphaeroides cytochrome ua3 as a model of the more complex cytochrome c oxidase of eukaryotes. Functional homology with the mitochondrial enzymes is seen in the rapid rate of electron transfer from mammalian cytochrome c, the biphasic kinetics of cytochrome c oxidation and the ability of the bacterial oxidase to translocate protons with the same efficiency as beef heart cytochrome c oxidase. The overall structural homology of the metal centers and their protein environment is evidenced by the remarkable similarity of the optical, resonance Raman, and EPR spectra of Rb. sphaeroides to those of beef heart cytochrome c oxidase. Recent analysis of the heme a3-CuB center of the Rb. sphaeroides enzyme by Fourier transform IR spectroscopy corroborates this result (84). A high degree of homology is predicted by genetic analyses (11,12,86); in fact, Rb. sphaeroides cytochrome aa3 shows almost as much identity with the core subunits of beef heart cytochrome c oxidase as does yeast cytochrome c oxidase, and essentially the same degree of similarity as P. denitrificans to eukaryotic cytochrome c oxidases (85). The development of the purification procedure and the in-depth characterization of the properties of this bacterial oxidase, combined with the deletion and reintroduction of the oxidase genes (11, 86), provides the basis for site-directed mutational analysis of the energy transducing functions of cytochrome c oxidase.