INTRODUCTION

Terminal oxidases of all oxygen-respiring organisms are members of a superfamily of membrane-residing heme-copper oxidases (1-3). Depending on the specific organism, either reduced quinones or cytochrome c may serve as electron donors for the reduction of oxygen to water. The reaction is coupled to the electrogenic translocation of protons across the membrane, thus serving as a primary energy converter. Independent of their complexity and subunit composition, all terminal oxidases appear to have closely related core structures with respect to subunits I, II, and III. This clearly emerges from the three-dimensional structures recently resolved on an atomic scale of the enzymes from Paracoccus denitrificans (4) and from beef heart (5, 6), respectively. Subunit I carries two heme centers and Cu_B. The latter forms the binuclear heme-copper site for binding and reduction of oxygen. Thus, a high spin heme with variable coordination geometry is always present as well as a hexacoordinated low spin heme acting as the primary electron acceptor. Only in cytochrome c oxidases, subunit II bears a binuclear Cu_A site, which is missing in subunit II of quinol oxidases (2, 7). The functional role of subunit-III in electron transport is unknown.

In previous reports, we described the properties of the first archaebacterial terminal oxidase isolated in functional form (8-11). This quinol oxidase from the extreme thermoacidophilic archaeon S. acidocaldarius has been reconstituted into archaebacterial tetraether lipid vesicles, which were shown to develop a proton motive force across the membrane upon energization by reductant pulses (12). Although generation of a steady state membrane potential and/or a pH gradient could be verified with this system, the proton to electron stoichiometry (H⁺/e) could not be determined simply because of the random orientation of the quinol oxidase molecules in the membrane and the lack of a membrane-impermeable reductant permitting unidirectional energization. Moreover, by genetic analysis it was demonstrated that in the membrane, the terminal oxidase of Sulfolobus is composed of a complex including an additional cytochrome a (cytochrome a₅₈₇) (13) and, in contrast to the previously investigated "single entity" form (8), may combine functional properties of terminal oxidases with those of respiratory complexes III (for review see Refs. 14-16). On the basis of its unusual structure, it has been speculated that this respiratory complex, termed SoxABCD (according to its subunit composition), may exhibit proton pumping activity by a Q cycle-like mechanism (13), which would be an extraordinary novelty for terminal oxidases.

The necessity of an efficient proton-translocating device in the membrane of S. acidocaldarius is obvious, because this archaebacterium thrives at 75-80 °C (17) and pH values between 2 and 3, maintaining a large pH gradient across the plasma membrane. In this work, we report on a method for isolation of the intact SoxABCD complex from membranes of S. acidocaldarius as well as for its successful reconstitution as a proton pump into archaeal lipid vesicles. This has been achieved by co-reconstitution with the SoxL Rieske-FeS protein from Sulfolobus (18), thereby permitting construction of a system that allowed energization of the membranes by reductant pulses from only one side; thus, the problem of randomized orientation of pumps could be overcome, and H⁺/e stoichiometries could be determined. Based on comparative sequence analyses, the data suggest that the single entity form of the enzyme is unlikely to pump protons, whereas the integrated SoxABCD complex is a proton pump very likely operating by an alternate pumping mechanism, eventually involving a Q cycle.

EXPERIMENTAL PROCEDURES

S. acidocaldarius (DSM 639) was grown heterotrophically in a mineral salt medium (17) as described by Anemüller and Schäfer (8) at the Gesellschaft für Biotechnologic Forschung (Braunschweig, Germany). Cells were suspended in 50 mM MES,¹ 1 mM EDTA, pH 5.5, at 4 °C, centrifuged for 15 min at 7.800 × g, resuspended in the same buffer containing 50% (v/v) glycerol, and stored frozen at 70 °C. Cells were disrupted, and membranes were prepared essentially as described by Anemüller and Schäfer (8). For chaotropic preextraction, the membranes were diluted to a concentration of 7.5 mg of protein in 50 mM Tris-Cl, 30 mM Na₄P₂O₇, pH 7.5, at 20 °C. After 1 h, the suspension was centrifuged (120,000 × g, 1 h, 20 °C), and the membrane pellets were finally resuspended in 50 mM Tris-Cl, pH 5.5, to give a protein concentration of 45 mg/ml and stored at 20 °C. Typically, these preparations yield 30-40 mg of membrane protein/liter of culture at a concentration of 0.8-1.1 nmol of heme a/mg of protein.

The preextracted membranes (800 mg) were suspended to a protein concentration of 8.5 mg/ml in a buffer containing 50 mM Tris-Cl, 500 mM ammonium sulfate (AS), pH 7.3, and solubilized using 20 mM n-dodecyl--D-maltoside (1.2 g of DM/1 g of protein) for 1.5 h at 20 °C. Unsolubilized components were removed by centrifugation at 120,000 × g for 1 h at 20 °C. All of the following purification steps were performed at 4 °C. A saturated AS solution was slowly added to the supernatant until 50% saturation. This solution was applied onto tandem columns containing propyl-agarose (Sigma P-5268;  = 1.5 cm, length = 30 cm) and hexyl-agarose (Sigma H-1882, = 2.5 cm, length = 6 cm). Both columns had been equilibrated with 50% saturated AS, 25 mM Tris-Cl, 0.5 mM DM, pH 7.3, at 0.5 ml/min. After loading, they were washed with the above buffer to remove less tightly bound material. The hexyl-agarose column, which contained only a-type cytochromes, was eluted with 30% saturated AS, 25 mM Tris-Cl, 1 mM DM, pH 7.3, at a flow rate of 0.5 ml/ml. This intensely green fraction was desalted by dialysis twice against 3 liters of 25 mM Tris-Cl, pH 7.3, overnight, and concentrated by ultrafiltration on a PM-30 membrane (Amicon, Beverly, MA) to 3-5 ml.

The concentrated cytochrome fraction containing approximately 30 mg of SoxABCD was applied to FPLC Mono-Q (10 mm × 10 cm, Pharmacia, Germany), which was equilibrated with 25 mM Tris-Cl, 0.5 mM DM, pH 7.3. Bound proteins were eluted with a step gradient of MgSO₄ in the same buffer at a flow rate of 1 ml/min. Peak fractions of SoxABCD determined by optical spectroscopy were collected, concentrated to a heme concentration of 60-70 nmol of heme a/ml by ultrafiltration on a PM-30 membrane, and stored at 70 °C.

Tetraetherlipids from S. acidocaldarius were isolated and fractionated from freeze-dried cells as described by Elferink et al. (19) and stored under nitrogen at 4 °C until use. The liposomal forming lipid fraction was dried by rotary evaporation and dispersed in 50 mM KH₂PO₄, pH 6.5 at 15 mg/ml in the presence of 45% (w/w) n-octyl--D-glucopyranoside. The suspension was sonicated using a probe-type sonicator (Branson Sonifier, 40 watts, 30% duty cycle) until a clear solution was obtained and centrifugated for 10 min at 13,000 × g. Sonication was performed at 0 °C, and the sample was flushed with nitrogen. Purified SoxABCD (20 nmol of heme a) was added to the tetraether lipid suspension to a final concentration of 2.5 nmol/ml and mixed extensively. The suspension was dialyzed four times at room temperature against a 500-fold volume of 50 mM KH₂PO₄, pH 6.5. Proteoliposomes were stored at 70 °C. Before use, a small sample of the proteoliposomes was slowly thawed at room temperature and extruded through 200-nm polycarbonate filters (Avestin, Ottawa, Canada), using a small volume extrusion apparatus (LiposoFast^TM Basic, Avestin). This technique has the advantage of producing monolayer vesicles of homogeneous size with a small range (20, 21).

For proton pumping experiments, SoxABCD was co-reconstituted with a purified Rieske-FeS protein from S. acidocaldarius, which was isolated as described by Schmidt et al. (18). Tetraetherlipid from Ther- moplasma acidophilum was isolated as described (22) and suspended by sonication (see above) to 20 mg/ml in 45% (w/w) n-octyl--D-glucopyranoside, 75 mM HEPES-KOH, 14 mM KCl, pH 7.4. SoxABCD (3 nmol) and Rieske-FeS protein (6 nmol) were added to the tetraether lipid suspension and dialyzed at room temperature in a dialysis cassette (Slide-A-Lyzer^TM, Pierce, catalog number 66425, cut-off 10 kDa) according to the following protocol: 5 h in 200 volumes of 75 mM HEPES-KOH, 14 mM KCl, 0.1% n-octyl--D-glucopyranoside, pH 7.4; 17 h in 200 volumes of 75 mM HEPES-KOH, 14 mM KCl, pH 7.4; 5 h in 200 volumes of 50 mM HEPES-KOH, 24 mM KCl, 15 mM sucrose, pH 7.4; 2 × 7 h in 300 volumes of 1 mM HEPES-KOH, 45 mM KCl, 44 mM sucrose, pH 7.4. Monolayer vesicles were prepared by using the extruder (see above).

Protein concentrations were determined by the modified Lowry method in the presence of detergents (23) using bovine serum albumin as a standard. Polyacrylamide gel electrophoresis was carried out in the presence of SDS by the Laemmli procedure (24) on 15% gels. Proteins were visualized by Coomassie Brilliant Blue staining. Heme A concentrations were determined utilizing the pyridine hemochrome method as described (25).

Quinol oxidase activity was measured spectrophotometrically at various temperatures by oxidation of N,N,N,N-tetramethyl-1,4-phenylenediamine (TMPD) or caldariella quinol as substrate in an air-saturated 10 mM citrate buffer, pH 6.5 (12). Cytochrome c oxidase activity was measured spectrophotometrically by following horse heart cytochrome c oxidation at 540-550 nm using a differential extinction coefficient (_550-540) of 19.5 mM¹ cm¹. Horse heart cytochrome c (Boehringer Mannheim) was reduced with NaBH₄, and excess NaBH₄ was removed by aliquots of 0.1 N HCl.

Absorption spectra were recorded either in a Kontron Uvikon 810 or a HP 8452-A diode array spectrophotometer at room temperature. EPR spectra were recorded with an X-band Bruker ER 200 D-SRC spectrometer equipped with an ESR 910 continuous flow helium cryostat from Oxford Instruments. Fluorescence measurements were performed using a fluorometer type SLM 4800S.

Particle size distribution of the isolated detergent-oxidase complex was determined by using an ALV 5000^TM system (Langen (Hessen), Germany) operating at 632.8 nm (neon-helium laser) and 50 milliwatts. Protein solutions were diluted in 25 mM Tris-Cl, pH 7.3 to give a final concentration of 10 µM and centrifuged for 30 min at 13,000 × g to remove any aggregated material. Measurements were performed at 20 °C.

Purified SoxABCD complex in 25 mM Tris-Cl, 250 mM KCl, 0.2 mM dodecyl maltoside, pH 7.3 was concentrated to approximately 1 mM using Microcon ultrafiltration cells (100 kDa). The optically transparent thin layer electrochemical (OTTLE) cell (path length set to 5-10 µm) used for redox titration was described previously (26). The gold grid working electrode was surface-modified by dipping it in a saturated solution of pyridine-3-carboxyaldehyde thiosemicarbozon as described by Hill et al. (27). After 15 min, excess pyridine-3-carboxyaldehyde thiosemicarbozon was thoroughly removed with deionized water, and the gold grid was dried. The reaction mixture consisted of 1 mM purified SoxABCD complex in 25 mM Tris-Cl, 0.2 mM DM, 250 mM KCl, pH 7.3, as conducting electrolyte. To accelerate the redox reaction, mediators were added to a final concentration of 40 µM: 1,1-ferrocenedicarboxylic acid, potassium ferrocyanide, 1,1-dimethylferrocene, tetrachloro-p-benzoquinone, 2,6-dichlorophenolindophenol, hexamineruthenium(III) chloride. 6-8 µl of the reaction mixture were sufficent to fill the cell. The potential in the spectroelectrochemical cell was controlled by a potentiostat built to our design (Sevenich-Gimbel type I). All potentials quoted refer to the silver, AgCl, 3 M KCl electrode (for potentials versus normal hydrogen electrode, add 208 mV). Spectra were recorded in a remodeled Cary 14 spectrophotometer at room temperature. The photometer and the potentiostat were controlled by data acquisition and treatment software (MSPEK) developed by S. Grzybek (Institut für Physikalische Chemie, Erlangen, Germany).

A series of potentials in the range of 100-365 mV were applied to the spectroelectrochemical cell. After equilibration of the cell contents at each applied potential, a spectrum was recorded in the range of 380-700 nm. A plot of the amplitude of the difference bands against the applied potential allowed us to fit the experimental data to a calculated Nernst curve yielding the characteristic data of different cofactors: the midpoint potentials (E_m) and the numbers (n) of electrons transferred.

RESULTS

S. acidocaldarius cells synthesize heme a- and heme b-containing cytochromes, exhibiting -bands in reduced minus oxidized difference spectra at 562-566, 587, and 605 nm. They do not, however, synthesize c-type cytochromes (11). The cytochrome composition of the membranes changes according to oxygen tensions, and the heme a/b ratio ranges from 2.2 ± 0.4 at low oxygen to 4.8 ± 1.2 at high oxygen. Cytochrome b_558/566 is only expressed at low oxygen, and its function is unknown. The absorption peak at 587 nm is the most prominent one at any oxygen tension. At least two electron transport components with different redox potentials but with identical optical absorptions contribute (28); one is the soxG gene product of the SoxM gene cluster (29), and the other is the SoxC component of the SoxABCD complex (13). Both are b-type apocytochromes, hosting, however, heme a.

To isolate the intact SoxABCD terminal oxidase complex, purified cytoplasmic membranes were washed with Na₄P₂O₇, which removed loosely bound proteins, such as the F₁ component of the ATP synthases or residual cell wall polypeptides. For solubilization of the cytochromes, several commonly used ionic and nonionic detergents like n-octyl glycoside, Chaps, CHAPSO, Mega-10, and Thesit were tested, which had either little solubilizing efficiency or led to spectral impairments. Most of the membrane-bound cytochromes (over 75%) could be subilized with dodecyl maltoside at a ratio of 1.2:1 (w/w; detergent/protein). This detergent does not induce any spectral changes of a- and b-type cytochromes and does not influence the oxidase activity.

The intact SoxABCD complex was purified by combining hydrophobic interaction with anion exchange chromatography. Crucial for the efficiency of purification was the use of two hydrophobic interaction columns in a tandem array. At high ionic strength, most of the solubilized membrane proteins such as the Rieske-FeS proteins and all b-type cytochromes remain bound to propyl-agarose, whereas the SoxABCD complex remains bound to hexyl-agarose, which is more hydrophobic. This hexyl-agarose fraction displays two -bands in the difference spectrum (see below) and shows high oxidase activity.

Further purification was achieved by FPLC anion exchange on a Mono-Q column; a typical elution profile is shown in Fig. 1. The substantial amount of SoxABCD eluted at 35 mM MgSO₄, with an overall yield of about 31% based on activity with TMPD, but only about 1% with reference to total membrane protein (Table I).

Table I. Purification protocol of the SoxABCD complex from S. acidocaldarius Fractions Membranes (after P2 extraction) Detergent extract Propyl-/Hexyl-agarose Mono-Q Protein (mg) 700 295 37 7.2 Heme a contenta (nmol) 885 625 416 182 Heme b contenta (nmol) 264 243 TMPD oxidase (units) 3950 3030 2230 1260 Cytochrome c oxidase (units) 593 TMPD specific activity (units/mg) 5.6 10.3 60.3 175 Cytochrome c specific activity (units/mg) 0.85 Purification factorb 1 1.7 8.9 20 Yield (TMPD activity)c (%) 97 74 55 31 a Determined from pyridine ferrohemochrome spectrum. b Purification factor is based on the specific heme a content. c Yield is based on crude membranes (not P2-extracted).

The final preparation contains only a-type hemes, as revealed by pyridine hemochrome analysis (c.f. Fig. 3A). Interestingly, it resembles the previously described single entity form of this oxidase complex (8) by the presence of a highly stabilized EPR-detectable semiquinone species. This sharp signal at g = 2.002 (not shown) is present in each state of the preparation and has been assigned to partially reduced tightly bound caldariella quinone. From spin integration of the EPR signals, the single entity aa₃ preparation, and especially the respective quinol oxidase from Acidianus ambivalens (30), were found to contain presumably stoichiometric amounts of bound quinone.² However, attempts to quantify caldariella quinone by chemical extraction according to conventional methods (31, 32) fail due to instability of this thiophenoquinone under the respective assay conditions. In addition, it is insoluble in ethanol/H₂O mixtures as normally used for its spectroscopic determination. Other quinol oxidases have definitely been shown to contain superstoichiometric amounts of tightly bound quinones (33-35). It has even been proposed that tightly bound quinones are a general feature of quinol oxidases (34).

It should be added that also the preparations of the Sulfolobus Rieske FeS-protein used for reconstitution (see "Experimental Procedures" and below) routinely contain the same radical species also ascribed to some tightly bound quinone.³

The polyacrylamide gel electrophoresis of purified SoxABCD under usual conditions shows only three protein bands (Fig. 2, lane 1). The diffuse band with an apparent molecular mass of 38 kDa is heterogeneous and contains both SoxB (subunit I) and SoxC (cytochrome a₅₈₇). It has previously been shown that this protein band cross-reacts with antisera against both subunits, which was further confirmed by protein sequencing (16). The presence of isobutyl alcohol (1%, v/v) allowed us to separate these associated peptides into two distinct bands of 39 (SoxB) and 64 kDa (SoxC), respectively (Fig. 2, lane 4), which could be verified by immunoblotting. The 27- and 14-kDa bands correspond to SoxA (subunit II) and SoxD, respectively (16), thus demonstrating the presence of all four polypeptides encoded by the sox operon (13). Analytical gel filtration (high pressure liquid chromatography) of the purified SoxABCD with its associated lipid and detergent molecules gave an estimated molecular mass of 280 ± 20 kDa. The identical elution profiles of the oxidase monitored at A₂₇₈ (protein) and at A₄₂₆ (heme a) suggested that the complex is homogenous. A predicted molecular mass on the basis of an equimolar 1:1 stoichiometry of all subunits (SoxA-D) of 144.2 kDa (without cofactors) and associated detergents in a roughly 1:1 weight ratio (16) suggest that the isolated detergent-protein complex is monomeric.

Interestingly, the oligomeric state of the finally isolated SoxABCD preparation depends critically on the ionic strength during the detergent solubilization, which could be confirmed by light scattering. When the membrane solubilization was carried out at low salt concentration, an oligomeric state of the oxidase was obtained with an apparent molecular mass of 570 ± 40 kDa. The high ionic strength applied in this work probably prevented the electrostatic surface charge attraction between monomers.

Difference spectra (reduced minus oxidized) of the purified SoxABCD (Fig. 3A) partially resemble those of the mitochondrial cytochrome c oxidase. The absorbance peak at 605 nm is typical of the aa₃ center, which is located in SoxB (subunit I). The peak at 587 nm is atypical for other terminal oxidases and results from two low spin hemes bound to the SoxC component. The pyridine hemochrome spectrum (Fig. 3C) as well as the Soret maximum at 440 nm indicate that the oxidase complex contains only cytochrome a. A constant -band ratio (A₅₈₇/A₆₀₅) of 1.72 ± 0.06 of different preparations demonstrates the reproducibility of the purification procedure. The differential absorption coefficients (based upon pyridine hemochrome) are 20.6 mM¹ cm¹ (587-545 nm) for heme a₅₈₇ and 12.1 mM¹ cm¹ (605-630 nm) for heme a₆₀₅ (Fig. 3A). The latter value corresponds to the heme a of other cytochrome c oxidases (36, 37), whereas the extinction coefficient of the heme a₅₈₇ is unusually high. In low temperature difference spectra (77 K) the absorbance peak at 587 nm is downshifted to 582 nm (with a shoulder at 573 nm).

The absolute spectra of both air-oxidized and dithionite-reduced SoxABCD, along with their apparent molar absorptivities are given in Fig. 4 and indicate a characteristic fine structure in the Soret region. The spectrum of the oxidized complex shows maxima at 426 nm for the Soret band and 601 nm for the -band region. The shoulder at 587 nm results from partial reduction of the air-oxidized sample. In the presence of ferricyanide and catalytic amounts of TMPD, a symmetric -peak appears at 598 nm. A ratio of A₂₇₈ to A₄₂₆ of 1.25 ± 0.05 was obtained for the purified complex; higher values are reported for other bacterial oxidases (36). In the reduced state, the Soret maximum shifted to 436 nm, and two -band maxima at 587 and 603 nm were observed.

The CO difference spectrum of the reduced SoxABCD (Fig. 5) is very similar to the corresponding spectrum of the single entity cytochrome aa₃ (8) with the -peak at 596 nm and the -peak at 548 nm caused by the a₃-CO complex. The Soret region has its maximum at 432 nm and its minimum at 447 nm. The concentration of heme a₃, which was calculated from this spectrum using a differential (432-447 nm) absorption coefficient of 82 mM¹ cm¹ (38), was approximately 25% of the total heme concentration determined by the pyridine hemochrome method in agreement with a ratio of 3:1 heme a/heme a₃.

Five different redox centers are present in the intact terminal oxidase: two heme a centers from cytochrome a₅₈₇ (SoxC) plus the centers heme a and heme a₃/Cu_B from cytochrome aa₃ (SoxB). All heme centers could be detected by spectroelectrochemical redox titrations at different wavelengths. A titration at 605 nm (not shown) yields values of 200 and 400 ± 30 mV, confirming previous titrations with the single entity form of the enzyme (8); these latter values of 220 mV (heme a) and 370 mV (heme a₃), respectively, had been obtained by both chemical redox titrations and EPR titration. Here for the first time the midpoint potentials of the heme a centers of cytochrome a₅₈₇ have been determined. An example of a spectroelectrochemical redox titration is shown in Fig. 6, evaluating the absorbance change at 587 nm. Under the applied conditions, two redox transitions within cytochrome a₅₈₇ could be detected with an apparent midpoint potential of E_m¹ = +210 ± 10 mV and E_m² = +270 ± 10 mV. Each heme a₅₈₇ contributed about 50% to the total absorbance change in the respective -band. Interference by the 605-nm transition is less than 10%. Values in the range of n = 0.9 for the slope of the Nernst plot indicate that one electron is transferred during each redox transition. Deviations from the theoretical value of n = 1 are presumably due to a narrow distribution around a mean value of the midpoint potential, which is even observed for ultrapure proteins.⁴ Thus four redox cofactors can be differentiated by potentiometric titrations in accordance with the four heme centers and the prediction from the polypeptide structure of the SoxABCD complex.

The EPR spectra in the air-oxidized state of purified single entity cytochrome aa₃ and of the SoxABCD complex of S. acidocaldarius are presented in Fig. 7. The former (spectrum A) exhibits one low spin ferric signal at g_zyx = 3.04, 2.21, and 1.45, arising from heme a₆₀₅, which is identical to that of mitochondrial oxidases (39). The EPR spectrum of SoxABCD (Fig. 7B), however, differs considerably from that of single entity cytochrome aa₃; instead of one g_z value at 3.04, there are three different low spin ferric heme signals at g_z = 2.93, 2.80, and 2.75. While these g_z signals are partially overlapping, three g_y signals are clearly resolved at 2.21, 2.28, and 2.37. The missing g_z value at 3.04 in spectrum B suggests that the local microenvironments around heme a₆₀₅ are different between the two oxidase preparations (see "Discussion"). The EPR spectra of both cytochrome aa₃ and SoxABCD complex exhibit in the air-oxidized state a large high spin heme signal at g_max = 6.10, which is higher than that of the low spin heme signals.

SoxABCD is a quinol oxidase; accordingly, a Cu_A site in subunit II (SoxA) is absent. Neither the typical EPR signals nor the 830-nm absorption band in the spectra of the oxidized enzymes could be detected; this is in line with predictions from the amino acid sequence of subunit II (13). A sharp signal at g = 2 (not shown) originates from a stabilized radical most likely from partially reduced tightly bound caldariella quinone (see above).

The purified SoxABCD oxidase does not react with cytochrome c or blue copper proteins such as halocyanine or azurine as electron donors. However, considerable turnover numbers were found with TMPD or caldariella quinol, confirming that in vivo the enzyme is acting as a quinol oxidase. Table II summarizes the kinetic data. Detergent-solubilized SoxABCD exhibits a significantly lower activity with the artificial electron donor TMPD than when reconstituted into tetraether lipids from S. acidocaldarius, which more closely resemble the natural membrane environment. This activation could not be observed with caldariella quinol, although it is most probably the natural substrate. Hence, the amphiphilic character of TMPD allows its rapid diffusion into the lipid membrane or into detergent micelles. In contrast, the lipophilic caldariella quinol had to be applied in a Triton-stabilized micellar form (see Ref. 8 for details). This decreases the mobility of caldariella quinol toward the binding site of the oxidase in liposomes. Nevertheless, apparent K_m values within the same range were found for both forms of the enzyme (Table II). Due to the limited availability of caldariella quinol, TMPD was used as a suitable electron donor for routine assays.

Table II. Catalytic properties of the SoxABCD complex Experimental details of the enzyme assays are described under "Experimental Procedures." SoxABCD Substrate KM(app) Turnover kcat/KM µM s1 M/s In detergent micelle (dodecyl--D-maltoside) QCal 36 393  (50 °C) 1.1 107 TMPD 67 157  (40 °C) 2.3 106 In liposomes (tetraether lipids) QCal 26 239  (50 °C) 9.2 106 TMPD 46 404  (40 °C) 8.7 106

Determinations of the oxidase activity in lipid vesicles were performed in the presence of valinomycin (400 nM) and nigericin (500 nM). The oxidation of TMPD and caldariella quinol was completely inhibited by cyanide (2 mM). The highest activity of the SoxABCD complex reconstituted into liposomes (1300 s¹ at 65 °C) was approximately 3 times greater than the maximum activity found for the reconstituted single entity form of cytochrome aa₃ (530 s¹) in previous investigations under the same conditions (12). An activation energy of 59.6 kJ/mol was determined within a temperature range from 30 to 75 °C.

By reconstitution of the single entity cytochrome aa₃ into liposomes, both a and a pH could be measured when the oxidase was energized by TMPD/ascorbate (12). pH measurements were performed with a fluorescent pH indicator entrapped inside the liposomes; the outside pH was kept constant. Fig. 8 compares the generation of proton motive force between liposomes containing either the single entity form of the oxidase (Fig. 8A) or the integrated SoxABCD complex (Fig. 8B). The addition of the electron donor TMPD/ascorbate caused a slow internal alkalinization of the former; a pH gradient equivalent to 32 mV was generated by the cytochrome aa₃ liposomes. Valinomycin, by dissipation of the electrical potential, increased the pH to 79 mV. With proteoliposomes containing the integrated SoxABCD complex, a rapid internal alkalinization was detected already in the absence of valinomycin up to 58 mV; after its addition, the pH stabilized at 85 mV. In both cases, it was completely collapsed when nigericin was added. With identical amounts of terminal oxidase in terms of cytochrome aa₃ present in each type of liposomes (adjusted by the 605-nm absorption) the activity of the integrated SoxABCD complex was again considerably higher than that of the single entity cytochrome aa₃; this is clearly illustrated by the time scale of the individual traces in Fig. 8.

Determination of pH, generated by cytochrome aa₃ (A) and SoxABCD (B) in proteoliposomes by measurements of pyranine fluorescence.

Thus, it could be established that the SoxABCD complex can generate a significant proton motive force under steady state conditions. However, the experiments do not allow us to distinguish between vectorial proton pumping and a "chemical" proton gradient produced by scalar processes. As such, the release of protons from ascorbate on the outside and the proton consumption by water formation inside the vesicles must be taken into account.

The determination of H⁺/e stoichiometries with quinol oxidases reconstituted in liposomes is generally hampered in two ways: 1) the reconstitution procedure usually leads to a random orientation of oxidase molecules in the liposome membrane, and 2) both orientations can be energized by TMPD due to its significant membrane permeability leading to proton translocation in both directions. Thus, it was necessary to construct a system overcoming these problems.

It has been shown that in isolated form one of the two Rieske-FeS proteins from Sulfolobus membranes (18, 40) can readily equilibrate electrons between cytochrome c and ubiquinone analogs. Further, we have demonstrated that the reduced Rieske-FeS protein can be reoxidized by the detergent-solubilized SoxABCD complex, which inherently contains bound caldariella quinone. On that basis, an artificial electron transport system was established by co-reconstitution of the Sulfolobus Rieske-FeS protein with the SoxABCD complex into tetraether lipid vesicles (Fig. 9A). This Rieske-FeS·oxidase complex showed a significant turnover (V_max = 187 s¹) with reduced cytochrome c (80 µM) as substrate and displayed Michaelis-Menten kinetics. An apparent K_m of 11.6 ± 0.3 µM was found. The reaction was coupled to stoichiometric oxygen uptake and was 100% cyanide-inhibited. A respiratory control ration of 2.9-3.3 could be measured (Fig. 9B). The orientation of incorporated complexes was determined spectrophotometrically, resulting in 65% accessibility (on average) to reduction by cytochrome c (cyanide present), indicating that 35% of the oxidase complexes were inversely oriented; full reduction was achieved by membrane-permeable TMPD.

The co-reconstituted system has two advantages; it can be unidirectionally energized by pulses of reduced cytochrome c, and it does not release scalar protons. Thus, it could be used for single turnover experiments monitoring the appearance of protons by an outside optical probe (41-43). Fig. 10 shows a typical experiment, demonstrating the rapid transient acidification following addition of 1.9 nmol of reduced horse heart cytochrome c (Fig. 10B). The subsequent slow alkalinization is caused by back-diffusion and equilibration with the intravesicular pH, which became alkaline due to H⁺ consumption by water formation. In fully uncoupled vesicles, only the respective alkalinization was observed (Fig. 10C). For quantitation of proton ejection and/or alkalinization, the system was calibrated by HCl aliquots after each experiment. In the coupled system, net H⁺/e > 1 was routinely determined; on average, a net H⁺/e of 1.2 ± 0.03 was found, definitely identifying the co-reconstituted SoxABCD complex as a vectorial proton pump.

A series of control experiments was performed to ascertain the validity of the proton translocation experiment (41-43) with the following results. First, in the absence of proteoliposomes, the addition of reduced cytochrome c to the assay mixture did not induce any absorption change (Fig. 10A). Second, the reduction of solubilized Rieske-FeS protein with cytochrome c did not cause a redox-dependent pH change, indicating that the observed acidification is neither due to proton release after binding of cytochrome c to the Rieske protein nor to a redox-Bohr effect. Third, fully uncoupled liposomes gave only the proton uptake due to water formation with a H⁺/e ratio of -1.08 ± 0.04, perfectly in agreement with the chemical reaction stoichiometry. Finally, identical experiments performed with cytochrome c oxidase from P. denitrificans (a kind gift from Prof. B. Ludwig, University of Frankfurt, Frankfurt, Germany) showed an analogous proton ejection with a net H⁺/e of 0.6 ± 0.05.

DISCUSSION

The thermoacidophilic archaebacterium S. acidocaldarius thrives at pH ~2 and has been shown to conduct respiration-coupled proton ejection (44), thereby keeping its cytosolic pH near neutral. Two terminal oxidases have been identified in its plasma membrane (8, 45, 46) and genetically characterized (13, 29). Both are sharing an unusual complexity compared with other procaryotic cytochrome c or quinol oxidases (for a review see Refs. 47 and 48). While the supercomplex "SoxM" (45) could not be isolated in catalytically sufficiently active form, the SoxABCD complex described here has been identified in previous works as a highly active quinol oxidase, using the genus-typical caldariella quinol (49) as its natural substrate. Here we have demonstrated for the first time that this integrated SoxABCD complex represents a respiratory proton pump, while evidence can be derived from structural properties that its cytochrome aa₃ moiety itself has no proton pumping capacity (see below).

Sulfolobus cytochrome aa₃ is the product of the soxB gene and was studied previously in purified form as so-called single entity quinol oxidase (8, 50). Its spectroscopic properties essentially resemble those of other aa₃-type terminal oxidases with the exception that resonance-Raman spectra indicate only very weak, if any, hydrogen bonding of the formyls of heme a and heme a₃ (51) as well as an equilibrium between 6cHS and 5cHS heme a₃.

In the integrated SoxABCD complex cytochrome a₅₈₇, the product of the soxC gene (13), forms an additional di-heme redox component also hosting the archaetypical heme A_s (52). Thus, a total of five redox-active metal sites are present, comprising four heme a sites and Cu_B. Subunit II of the qinol oxidase is devoid of metal sites (in contrast to Cu_A in cytochrome c oxidases) but may be involved in tight as well as in reversible binding of caldariella quinone (34, 53).

While recent resonance-Raman studies of the integrated SoxABCD complex confirm the features found for isolated SoxB even in a more pronounced form (54), significant differences of the EPR spectra could be detected (Fig. 7B). Three different low spin signals were identified with down-shifted g_z and g_y values as compared with SoxB. The shifted g values may reflect an increase in electron donation by axial histidine. This has been shown by extensive studies of low spin hemes in both model compounds and hemoproteins (55, 56). The effect cannot be caused by deprotonated imidazole, because in that case the signals of histidine-imidazolate derivatives would occur in the MCD spectrum of SoxABCD, which is not the case (57). MCD spectra are very sensitive to the chemical nature of axial heme ligands (55, 58). Therefore, one or both of the histidine ligands of heme a₆₀₅ in SoxABCD most likely form a stronger hydrogen bond with a neighboring amino acid side chain than the corresponding hydrogen bond in isolated SoxB. In fact, in model compounds of low spin hemes, strong hydrogen bonds have shown an effect similar to the deprotonation of the histidine ligands (59, 60). Furthermore, the strong hydrogen bond can also explain the red shift of the Soret band of SoxABCD relative to SoxB; the increased basicity of the histidine ligands may allow the optical transition of the heme to occur at lower energy.

As reported elsewhere, MCD spectra support the conclusion that in SoxABCD one of the two low spin hemes in its cytochrome a₅₈₇ moiety (SoxC) displays a His-Met ligandation (57). This can also be readily derived from the primary sequence of SoxC, which resembles an apocytochrome b, however with the option to use Met instead of His at one of the heme binding motifs. This observation together with the redox titration reported here solves an interesting problem. Actually, two pools of cytochromes absorbing at 587 nm but exhibiting different redox potentials have been detected in membranes of S. acidocaldarius (11, 28, 47). The present study unequivocally assigns the high potential heme a₅₈₇ to the SoxC gene product as a component of the SoxABCD terminal oxidase complex. The span between the midpoint potentials of its hemes (+210 and +270 mV) may be of significant functional importance for the process of proton pumping (see below). From this assignment it follows also that the previously described low potential cytochrome a₅₈₇ with an average midpoint potential of about +80 to +100 mV (11) corresponds to the product of the soxG gene (29) which is a component of the SoxM alternate oxidase complex.

The preparation of the SoxABCD complex reported here allows a complete separation from the SoxM complex by hydrophobic interaction chromatography as revealed by removal of all b-type cytochromes during this step (Table I); it yields a 20-fold enrichment of the purified product (based on heme a content), indicating an ~5% abundance of the complex in the intact membrane. Despite the lack of cytochrome c (11, 14) Sulfolobus membranes oxidize cytochrome c from horse heart in a cyanide-sensitive manner. Since it was shown that the membrane-residing Rieske-FeS proteins of Sulfolobus can accept electrons from cytochrome c (18, 61), it is very likely that this activity is artifactual; as shown previously (61), the Rieske-FeS centers are reoxidized by the terminal oxidases also in intact membranes under aerobic conditions. Consequently, the cytochrome c oxidase activity vanishes during the purification of the SoxABCD quinol oxidase.

The catalytic activity of the detergent-solubilized complex is highest with caldariella quinone and comes close to a purely diffusion-controlled reaction, as documented by the k_cat/K_m value (Table II). Compared with the purified SoxB single entity quinol oxidase, a significantly higher maximum turnover could be measured under various conditions. A reasonable explanation may be that the quinone binding site is much better preserved in the integrated complex than in the single entity form. Anticipating the mechanism discussed below, the presence of cytochrome a₅₈₇ may be necessary to provide a completely intact substrate binding site. Conformational differences resulting from polypeptide interactions within the complex are suggested also by the above mentioned differences of the EPR spectra between isolated cytochrome aa₃ (SoxB) and SoxABCD. Although they are functionally classified as quinol oxidases, none of the classical inhibitors interacting with quinol binding sites, either in bc₁ complexes or with quinol oxidases of the bo type, were found to inhibit this enzyme (data not shown). However, the persisting appearance of a strong g = 2 radical signal in oxidase samples "as prepared" suggested the presence of a tightly bound partially reduced quinone.

In fact, recent studies of various quinol oxidases support the view of tightly bound quinones as being a general feature of these membrane enzymes (33-35, 53). As in case of SoxABCD, in the resting state of the preparations, a highly stabilized quinone radical was observed also in Escherichia coli bo₃ oxidase (35). From hyperfine structure in the EPR spectrum, the presence of a second quinone molecule in a distance of ~1.5 nm was postulated. Interestingly, also for the Q_o site of bc₁ complexes, a dual occupancy by quinones has been postulated (62), forming a quinhydrone state that represents a rather stable semiquinone radical as well as a perfect single electron donor/acceptor system. Thus, similar structures may be postulated to function in the SoxABCD complex, which in analogy to bc₁ contains a di-heme cytochrome structurally homologous to cytochrome b.

Regarding the co-reconstituted liposomal system, the observed catalytic turnover is considerably below that of real cytochrome c oxidases (about 1/5-1/3). This is not surprising in a totally artificial system employing a quasireversed electron flow from reduced cytochrome c via a Rieske-FeS protein to SoxABCD-bound caldariella quinone. Although the Sulfolobus Rieske protein (SoxL (18)) presumably has no genuine cytochrome c binding domain, a sufficient rate of redox equilibration between both components is obviously possible (18). The back-transfer of electrons from the Rieske protein to quinone is thermodynamically compensated for by the large free energy change of the quinol oxidase reaction catalyzed by the integrated SoxABCD complex. As revealed by the results illustrated in Fig. 10, the catalytic activity of the system is high enough to cope with the rate of diffusional proton back-flow into the liposomes, and thus the proton pumping capability of the co-reconstituted system could be demonstrated.

Actually, it was the predominant aim of this investigation to examine whether or not the terminal oxidase complex of Sulfolobus could perform respiration-coupled proton pumping exceeding purely chemical charge separation. Although this could be shown, for interpretation of the presented data a discussion of the reconstituted model system as well as of structural properties of cytochrome aa₃ from S. acidocaldarius is of crucial importance. Can subunit SoxB itself, the cytochrome aa₃ moiety of the complex, act as a proton pump? With a high degree of certainty, the answer is no. Sequence alignments strongly suggest that the core structure of the catalytic subunit I of all heme-copper oxidases has been conserved throughout all aerobic organisms (47, 48, 63, 64). Besides the six invariant histidine residues in helices II, VI, VII, and X involved in heme-iron and Cu_B binding, a number of other residues have been identified by site-directed mutagenesis to be intimately linked with proton pumping (41, 65, 66). Convincing evidence has been presented that two separate proton transfer pathways exist in subunit I of proton pumping heme-copper oxidases: one for protons consumed for water formation at the binuclear reaction center (chemical H⁺) and one for vectorially translocated protons (pumped H⁺) (41, 67, 68). This is strongly supported by the recently presented three-dimensional structures of the enzymes from P. denitrificans (4) and from beef heart (4-6). The residues on helix IIX delineating the putative chemical proton channel are present in SoxB (as in all other oxidases), whereas the critical residues for the pumping channel are missing. These are the -NX₁₀DX₆N- motif in the loop connecting transmembrane helices II and III and a highly conserved glutamic acid of the -GHPEVY- motif in transmembrane helix VI. In proton pumping oxidases, irrespective of whether cytochrome c or quinoles serve as electron donors, the aspartate at the interhelical connection is located at the putative entrance to the pumping channel, while the glutamic acid appears to be involved in proton conduction at the output side of a mechanism involving the so-called histidine cycle for pumping (4, 69, 70). In SoxB this very glutamate is replaced by apolar valine, and the essential residues of the loop motif are replaced by nonprotonatable ones. Thus, a proton pumping activity of SoxB is indeed very unlikely. Actually, multiple alignments and phylogenetic analysis strongly suggest that two lines of oxidases already separated during early evolution (for details see Refs. 47, 48, and 71), with and without the essential residues for the H⁺ pumping channel, respectively. Nevertheless, recent mutagenesis experiments with R. sphaeroides have shown that alternate intermediate pathways may be used for pumped protons; however, the crucial aspartate motif (loop II-III) at the H⁺ uptake side was indispensable (72).

Since proton pumping by the integrated SoxABCD complex has been shown in the present work, new mechanisms involving SoxC, the diheme cytochrome a₅₈₇, have to be considered. In analogy to the function of b-type cytochromes in bc₁ complexes, it had been hypothesized that the SoxABCD complex might involve an intramolecular Q cycle (13). The applied reconstituted vesicle system contains tightly bound quinone (as discussed above) but does not contain an excessive quinone pool as required for steady state proton translocation by a Q cycle mechanism in native membranes. This, however, is not necessary in single turnover experiments as described here, presumably involving direct reduction of tightly bound quinone of the integrated Rieske-FeS·SoxABCD complex. Operation of a Q cycle is characterized by a H⁺/e >1 (theoretically 2). Taking into account that experimentally measurable H⁺/e ratios usually fall short due to proton leaks or insufficient coupling of reconstituted liposomal systems, values of >1.2 H⁺/e as found in the above experiments are well in line with a Q cycle. In this context, it is important to notice that by the applied method scalar protons are not produced outside the lipid vesicles, and only pumped protons are monitored.

Possible proton pumping mechanisms of terminal ubiquinol oxidases have been controversially discussed (7, 73-75). The linkage of quinol oxidation to proton transduction has been hypothesized also in alternate models to involve redox loops between two quinone binding sites at the terminal oxidase (7). The net pumping stoichiometry for the proposed Q redox loops, however, would be H⁺/e = 1, which is below the ratio of a Q cycle and the above observed results with SoxABCD of >1, respectively.

Of course, any details of such a mechanism remain speculative as long as no high resolution structure of this particular enzyme complex or related enzyme complexes becomes accessible. The two hemes of SoxC separated by 60 mV between their midpoint potentials could indeed assume a similar role as b-type cytochromes for disproportionation of Q radicals. However, nothing particular is known about the Q binding site(s) on SoxABCD, especially whether equivalents to the Q_i and Q_o sites on bc₁ complexes exist; this issue may be approached in further studies as well as the search for site-specific inhibitors or quinone analogs.

A rigorous alternate assumption could be that cytochrome a₅₈₇ of SoxABCD simply serves as an electron storage device facilitating the transition of electrons from a two-electron donor (quinol) into a single electron pathway of the terminal oxidase. In that case, a novel pumping mechanism would have to occur, which appears unlikely in view of the conserved structure of oxidase subunits I (4, 5) and of the observed pumping stoichiometry.

In conclusion, therefore, it should be emphasized that the measured stoichiometry does not definitely prove the existence of a Q cycle-like mechanism, but it excludes the mechanism of "classical" aa₃- or bo₃-type cytochrome c oxidases.