Mechanism of activation of the chloroplast ATP synthase. A kinetic study of the thiol modulation of isolated ATPase and membrane-bound ATP synthase from spinach by Eschericia coli thioredoxin.

The mechanism of thiol modulation of the chloroplast ATP synthase by Escherichia coli thioredoxin was investigated in the isolated ATPase subcomplex and in the ATP synthase complex reconstituted in bacteriorhodopsin proteoliposomes. Thiol modulation was resolved kinetically by continuously monitoring ATP hydrolysis by the isolated subcomplex and ATP synthesis by proteoliposomes. The binding rate constant of reduced thioredoxin to the oxidized ATPase subcomplex devoid of its epsilon subunit could be determined. It did not depend on the catalytic turnover. Reciprocically, the catalytic turnover did not seem to depend on thioredoxin binding. Thiol modulation by Trx of the epsilon-bearing ATPase subcomplex was slow and favored the release of epsilon. The rate constant of thioredoxin binding to the membrane-bound ATP synthase increased with the protonmotive force. It was lower in the presence of ADP than in its absence, revealing a specific effect of the ATP synthase turnover on thioredoxin-gamma subunit interaction. These findings, and more especially the comparisons between the isolated ATPase subcomplex and the ATP synthase complex, can be interpreted in the frame of the rotational catalysis hypothesis. Finally, thiol modulation changed the catalytic properties of the ATP synthase, the kinetics of which became non-Michaelian. This questions the common view about the nature of changes induced by ATP synthase thiol modulation.

F 0 F 1 1 proton ATPase (or ATP synthase) is responsible for ATP synthesis in the inner membrane of mitochondria, the thylakoid membrane of chloroplasts, and the cytoplasmic membrane of bacteria (for review, see Ref. 1 and all papers in the same issue). The catalytic sector F 1 , with subunits ␣, ␤, ␥, ␦, and ⑀ (stoichiometry ␣ 3 ␤ 3 ␥ 1 ␦ 1 ⑀ 1 ) is a water-soluble entity bear-ing catalytic and noncatalytic nucleotide binding sites (for review, see Ref. 2). The X-ray structure of the major part of this subcomplex has been elucidated in the case of the bovine heart mitochondrial species (3). The F 0 sector is membranous and forms a proton-channeling device converting the energy of the electrochemical proton gradient ⌬ H ϩ into mechanical energy. In chloroplasts, it contains the four subunits I, II, III, and IV in the probable stoichiometry I 1 II 1 III 9 -12 IV 1 . The number of c subunits (subunit III equivalent) per complex was recently found to be 10 in crystals of the yeast mitochondrial ATP synthase (4). ATP hydrolysis probably results in the rotation of the ␥ subunit within the ␣ 3 ␤ 3 crown (5)(6)(7)(8) coupled to the rotation of the oligomer of c subunits (9). The whole F 0 F 1 complex is thought to act as a rotatory proton-driven motor (for reviews, see Refs. 10 and 11), the stator containing subunits I, II, IV, ␦, ␣, and ␤, and the rotor subunits III, ␥, and ⑀ in the case of the chloroplast enzyme.
It has been known for many years that the electrochemical proton gradient is necessary to activate CF 0 CF 1 in addition to supplying it with energy for ATP synthesis (12,13). In addition, reduction by a dithiol of a specific disulfide bridge located on the ␥ subunit (14) diminishes the magnitude of the ⌬ H ϩ required to activate the enzyme (15,16). This process is called thiol modulation. Thiol modulation itself is energy-dependent (13,17). It is not yet possible to locate the two cysteines involved (␥Cys 199 -␥Cys 205 ) in the tridimensional structure because the only F 1 structure that has been published until now is that of the mitochondrial enzyme (3,4), in which there is no domain homologous to that containing ␥Cys 199 and ␥Cys 205 . In fact, these two cysteines are included in an insertion specific of CF 1 (18). In vivo, the ␥ subunit reduction is achieved via a thioredoxin (Trx) by electrons diverted from photosystem I (19,20). Two different Trxs, called f and m, are present in chloroplasts. On the basis of slightly better catalytic efficiency, Trx f has been proposed to be the physiological reductant of CF 1 . However, Trx m is also able to reduce CF 1 with a good efficiency (21). Escherichia coli Trx, which is more similar in primary structure to Trx m than to Trx f, is more efficient than the former toward CF 1 reduction and has already been used for in vitro studies of its interaction with CF 1 (21)(22)(23). Binding properties of E. coli Trx to the isolated CF 1 subcomplex, devoid or not of its ⑀ inhibitory subunit, and in different redox states, have been investigated using fluorescent techniques (23). The kinetics of thiol modulation of CF 0 CF 1 by different Trxs in energized thylakoid membranes has been analyzed and the rate constant of binding determined in the presence of a high protonmotive force (21).
Because the domain of interaction of CF 1 with Trx belongs to the ␥ subunit, it is expected to rotate during the enzyme turn-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ To whom correspondence should be addressed: Section de Bioénergétique, Bâ timent 532, CEA Saclay, F91191 Gif-sur-Yvette Cedex, France. Tel.: 33-1-6908-9891; Fax: 33-1-6908-8717; E-mail: haraux@ dsvidf.cea.fr. 1 The abbreviations used are: F 0 F 1 , ATP synthase complex; F 0 , membranous sector of ATP synthase; F 1 , soluble subcomplex of ATP synthase; CF 0 CF 1 , chloroplast ATP synthase complex; CF 1 , soluble subcomplex of chloroplast ATP synthase containing the ⑀ subunit; CF 1 -⑀, soluble subcomplex of chloroplast ATP synthase devoid of ⑀ subunit; Trx, thioredoxin; MOPS, 4-morpholinepropanesulfonic acid; DTT, dithiothreitol; NEM, N-ethylmaleimide; FCCP, carbonyl cyanide p-trifluoromethoxyphenylhydrazone; ⌬ H ϩ, electrochemical proton gradient or protonmotive force; ⌬pH, transmembrane pH difference; ⌬, membrane potential. over. It is then of particular interest to study the Trx interaction in the isolated CF 1 subcomplex and in the membranebound CF 0 CF 1 and to investigate carefully the effect of the protonmotive force and of the enzyme turnover on this interaction. The present paper is a kinetic analysis of the interaction of E. coli Trx with isolated spinach CF 1 , devoid or not of the ⑀ subunit, and with complete CF 0 CF 1 reconstituted into liposomes. The role of different effectors was investigated. In the whole complex, but not in the isolated ATPase, the enzyme turnover proved to have a specific effect on this interaction, antagonistic to that exerted by the membrane energization. The consequences for the mechanism of thiol modulation were examined in relation to the rotatory mechanism hypothesis. The catalytic properties of oxidized and thiol-reduced CF 0 CF 1 were also investigated, and it was shown that thiol modulation does not simply consist of an increase in the proportion of active ATP synthases at a given magnitude of the protonmotive force.

EXPERIMENTAL PROCEDURES
Purification of CF 1 , CF 1 -⑀, and CF 0 CF 1 from Spinach Leaves-The soluble chloroplast ATPase (CF 1 ) was extracted from spinach (Spinacia oleracea L.) leaves and purified under its two forms, containing or not the ⑀ subunit (24). Protein concentration was determined by UV absorption spectroscopy assuming an extinction coefficient of 0.48 cm 2 mg Ϫ1 at 278 nm (25). The enzymes were stored at 5°C in 34% ammonium sulfate. The CF 0 CF 1 ATP synthase complex was extracted and purified from spinach leaves following the procedure described in Ref. 26. Its concentration was determined using the BCA protein assay reagent from Pierce. The purified complex was kept frozen in small vials under liquid nitrogen before use in reconstitution experiments.
Purification of Thioredoxin from E. coli-The E. coli thioredoxin overexpression plasmid pFPI (27) was used to transform a DH5a E. coli strain (Life Technologies). Protein purification to homogeneity was performed with 24-h bacterial cultures, essentially as described in (27). The main steps were: disruption of bacteria with a French pressure cell, heat-shock (3 min, 80°C), ammonium sulfate fractionation, size exclusion chromatography on G-50, and ion exchange on DEAE-Sephacel (both from Amersham Pharmacia Biotech).
Purification of Bacteriorhodopsin-Purple membrane from Halobacterium salinarium was isolated according to Oesterhelt and Stoeckenius (28) and stored for months at Ϫ20°C. Bacteriorhodopsin at a final concentration of 4 -5 mg ml Ϫ1 in the reconstitution buffer was monomerized by incubation in presence of 2% (31 mM) Triton X-100 for 20 h at 4°C in the dark. The concentration was determined from the absorption at 560 nm (⑀ ϭ 52, 800 M Ϫ1 cm Ϫ1 ).
Coreconstitution of Bacteriorhodopsin and CF 0 CF 1 into Liposomes-Unilamellar liposomes were made by reverse-phase evaporation from a mixture of 18 mM egg phosphatidylcholine, 2 mM egg phosphatidic acid, and 11 mM cholesterol in the reconstitution medium containing 50 mM MOPS, pH 7.3, 50 mM Na 2 SO 4 , and 50 mM K 2 SO 4 . Then, 80 l of 200 mg ml Ϫ1 Triton X-100 was mixed with 60 g of CF 0 CF 1 . 2 min later, 100 l of 4 mg ml Ϫ1 bacteriorhodopsin was added to the mixture, which was then supplemented with reconstitution buffer (final volume 1.5 ml). 0.5-ml liposomes were added and incubated for 10 min at room temperature. The detergent was then removed with three successive additions of SM-2 Bio-Beads (first one, 80 mg; second one, 80 mg 1 h later; third one, 160 mg 1 h after the second). Bio-Beads were then eliminated by rapid filtration on glass cotton. Proteoliposomes were stored on ice and in darkness. Their activity was stable for 1 day (for details, see Ref. 29).
ATP Hydrolysis Measurements with the Isolated ATPase Complex-CF 1 or CF 1 -⑀ was diluted at a concentration of 10 nM (unless otherwise indicated) or 1.3 nM, respectively, in a 1-ml spectrophotometric microcuvette, stirred, and thermostatted at 37°C. The reaction medium (1 ml) contained 50 mM Tris-SO 4 , 40 mM KHCO 3 , pH 8.0, 4 mM MgSO 4 , 1 mM P-enolpyruvate, 0.45 mM NADH, 20 units/ml pyruvate kinase, 50 units/ml lactate dehydrogenase. KHCO 3 was omitted when indicated. The absorbance at 340 nm was monitored continuously. The time response of the coupled system was checked by ADP injections. The ATPase reaction was triggered by injecting 2 mM MgATP in the reaction medium. For studies of the ATPase reduction by Trx, a small aliquot of Trx preincubated with 2 mM DTT was injected into the reaction medium supplemented with 0.1 mM DTT. In some experiments, prereduced Trx was injected in the reaction medium devoid of MgATP, and the reaction was initiated later by injecting ATP. The first 10 s following each addition were discarded in the data analysis.
Steady-state ATP Synthesis Measurements-Proteoliposomes were diluted 4-fold in the reconstitution buffer, pH 7.3, supplemented with 2 mM MgSO 4 , 1 mM ADP, and 10 M diadenosine pentaphosphate. The 0.4-ml sample was contained in a transparent vessel, stirred, and thermostatted at 20°C. It was illuminated continuously with a strong yellow light, obtained by filtering the light produced by a halogen lamp (filter band pass 500 -650 nm, beam intensity 1.5 kW m Ϫ2 ). After 10 min of illumination, ATP synthesis was initiated by adding 50 mM K 2 HPO 4 preadjusted at pH 7.3. Small aliquots were taken up at different times, quenched with trichloroacetic acid, and assayed for ATP concentration using the luciferin-luciferase technique as in Ref. 29.
Energy-induced thiol modulation of CF 0 CF 1 was carried out by adding 3 M Trx and 0.1 mM DTT before starting the illumination. Experiments with NEM-inactivated Trx were carried out as follows. Trx and DTT were mixed in the reaction medium 10 min before the addition of 0.25 mM NEM. 5 min later, NEM was neutralized by the addition of 0.2 mM DTT. Liposomes were then added, illuminated, and assayed for ATP synthesis as described above.
Thiol modulation of CF 0 CF 1 by Trx in darkness was achieved as follows. Liposomes were diluted in the reaction medium containing 3 M Trx and 0.1 mM DTT. The reaction was stopped by adding 0.25 mM NEM at a given time. 5 min later, 0.2 mM DTT was added to neutralize NEM, and the experiment of ATP synthesis was carried out as described above.
Time-resolved Measurement of ATP Synthesis-Proteoliposomes were diluted 5-fold in the reconstitution buffer, pH 7.3, supplemented with 50 mM KH 2 PO 4 , 4 mM MgSO 4 , 10 M diadenosine pentaphosphate, 4 mM glucose, 2 mM NADP ϩ , 20 units/ml hexokinase, and 20 units/ml glucose-6-phosphate dehydrogenase. The 1-ml sample was contained in a spectrophotometric microcuvette, stirred, and thermostatted at 20°C. The absorbance at 340 nm was monitored continuously. The time response and the proportionality between the rate of ATP production and the rate of NADP ϩ reduction were checked using myokinase at different concentrations in the presence of ADP and absence of diadenosine pentaphosphate. A small positive drift was detected in the absence of a known ATP-producing reaction. This drift was constant for 1 h; it did not depend on the presence of the liposomes and was not affected by the illumination. It was subtracted from the measured rates of NADP ϩ reduction. The sample was illuminated from the top of the cuvette with an optical fiber equipped with a yellow filter (filter band pass 500 -650 nm, light intensity 0.8 kW m Ϫ2 ). The setup allowed injections into the cuvette to be made without interrupting the actinic light. The windows of the spectrophotometer were protected from the actinic light with highly selective filters transmitting at 340 nm. 10 min after the beginning of illumination, ATP synthesis was initiated by adding 1 mM MgADP to the liposome suspension. It was verified that the rate of ATP synthesis was practically constant for periods as long as 30 min. The reaction was allowed to proceed for 5 min, and then prereduced Trx was added in the presence of 0.1 mM DTT. ATP synthesis was then followed for the required time. In some experiments, prereduced Trx was injected in the reaction medium devoid of MgADP, and the reaction was initiated later by injecting MgADP. In studies at variable ADP concentrations, successive additions of substrate were made, separated by 4 -5-min intervals. The first 10 s following each addition were discarded in all data analysis.
⌬pH Measurements-Proteoliposomes were diluted 20-fold in the reconstitution buffer, pH 7.3, supplemented with 50 mM KH 2 PO 4 , 1 mM MgADP (or 1 mM MgATP), and 4 M 9-aminoacridine. The sample was contained in a fluorometric microcuvette, stirred, and thermostatted at 20°C. Fluorescence ( excitation ϭ 400 nm, emission ϭ 450 nm) was recorded continuously. Illumination was provided by the same setup as for spectrophotometric measurements of ATP synthesis. 0.1 M valinomycin was added during the illumination.
Data Analysis-Data were fitted to theoretical models by a nonlinear least squares iterative procedure based on Marquardt's algorithm, using Microcal Origin 5.0.
Reagents-All reagents were of analytical grade. Egg phosphatidylcholine and egg phosphatidic acid were purchased from Avanti Polar Lipids. SM2 Bio-Beads came from Bio-Rad. FCCP, valinomycin, DTT, NEM, diadenosine pentaphosphate, and cholesterol were obtained from Sigma. ATP, ADP, P-enolpyruvate, NADH, NADP ϩ , lactate dehydrogenase, and pyruvate kinase were obtained from Roche Molecular Biochemicals Ltd. Hexokinase and glucose-6-phosphate dehydrogenase were obtained from Calbiochem. The ATP-monitoring kit came from Bio-Orbit. 9-Aminoacridine was purchased from Fluka. Fig. 1 shows how the rate of ATP hydrolysis catalyzed by isolated, oxidized CF 1 -⑀ changes after the addition of prereduced Trx into the reaction medium (a, direct curves; b, first derivatives). The ATPase starts to increase immediately after the addition of Trx and reaches a stable new value after a few minutes. Kinetics such as those displayed in Fig. 1b could be fitted satisfactorily with the equation

Rate of Activation of Isolated Oxidized CF 1 -⑀ by reduced thioredoxin-
where V(t) is the rate of ATP hydrolysis at a given time t after the addition of Trx, V 0 the rate of ATP hydrolysis at zero time (just before the addition of Trx), and V eq is the final rate of ATP hydrolysis. V 0 (as a control), V eq , and k app are plotted in Fig. 2 as a function of the concentration of reduced Trx. It can be seen that V eq is practically independent of the concentration of reduced Trx. This indicates that thiol activation is irreversible under these conditions, as expected from the presence of 0.1 mM DTT to rereduce oxidized Trx (DTT alone did not reduce CF 1 -⑀ in this time range). The reduced form of CF 1 -⑀ has a catalytic turnover that is approximately doubled with respect to the oxidized form. Fig. 2 also shows that k app is proportional to the concentration of reduced Trx, which demonstrates that the activation of CF 1 -⑀ here is kinetically controlled by the formation of the supercomplex between oxidized CF 1 -⑀ and reduced Trx. The slope of the plot gives the value of the rate binding constant (k on ) of reduced Trx to oxidized CF 1 -⑀, where [Trx red ] is the concentration of reduced Trx. The value of k on here is 1.2 ϫ 10 5 M Ϫ1 s Ϫ1 . Data of Figs. 1 and 2 were obtained in the presence of 40 mM bicarbonate ions, which activate the isolated ATPase. We have repeated the experiments in the absence of bicarbonate. The ATPase activity of oxidized CF 1 -⑀ was diminished by about 80%, but the residual activity was still doubled by thiol modulation. The rate of formation of the CF 1 -⑀⅐Trx supercomplex, as defined by the k on constant, was unchanged (results not  Fig. 1. Panel a, rate constant of CF 1 -⑀ activation (k app ) as a function of the concentration of reduced Trx. k app was calculated from fits as those displayed in Fig. 1a. The slope of the graph gives k on ϭ 1.2 ϫ 10 5 M Ϫ1 s Ϫ1 . Panel b, rate of ATP hydrolysis before the addition of Trx (f) and extrapolated at infinite time after the addition of Trx (Ⅺ). The latter (V eq ) was calculated from the same fit as k app . These two rates represent the activity of fully oxidized and fully reduced CF 1 -⑀, respectively. shown). At first sight, this binding rate does not depend on the catalytic turnover rate of CF 1 -⑀.
We got more decisive evidence that the rate of formation of the supercomplex does not depend on the catalytic turnover. 25 nM reduced Trx was added to oxidized CF 1 -⑀ in the absence of ATP. The substrate was injected at different times after the addition of Trx, and the rate of ATP hydrolysis was measured immediately after the addition of ATP. Fig. 3, closed circles, shows how the initial rate of ATPase activity increases with the time of incubation of oxidized CF 1 -⑀ with reduced Trx. On the same graph is also displayed the continuous measurement of the ATPase activity after the addition of 25 nM reduced Trx in the presence of ATP. The two kinetics of activation are identical, showing that the working enzyme is reduced by Trx at the same rate as the resting enzyme.
Thioredoxin-induced Activation of Isolated CF 1 with the ⑀ Subunit-We have investigated the effect of the presence of the ⑀ subunit on the rate of activation of CF 1 by reduced Trx. Experiments were carried out exactly as previously, except that CF 1 containing its inhibitory ⑀ subunit was used. As expected, the rate of ATP hydrolysis by the oxidized enzyme, before the addition of Trx, was very low and probably represents a small subpopulation of CF 1 -⑀ because intact CF 1 has been proposed to be fully inactive (30,31). After the addition of 2.5 M reduced Trx, the rate of ATP hydrolysis increased progressively, and the kinetics of activation critically depended on the CF 1 concentration (Fig. 4a). Because the activation was more pronounced at low CF 1 concentrations, it can be deduced, in accordance with previous proposals (30,31), that the reduction of the ␥ subunit decreases the affinity of ⑀ for the ATPase subcomplex. Fig. 4b shows the specific ATPase activity before the addition of Trx and at the end of the activation process, as a function of the CF 1 concentration. In the initially oxidized state, the activity is independent of the enzyme concentration, which confirms that it is probably due to a small fraction of ⑀-depleted enzymes. Comparison of this activity with that of the oxidized CF 1 -⑀ (Fig. 2b) suggests that this fraction represents about 5% of the total. After reduction, the specific activity decreases with the CF 1 concentration. The data could be correctly fitted with the equation where V is the rate of ATP hydrolysis at a given ATPase  5.6 Ϯ 1.9 s Ϫ1 , and the two other parameters became V m ϭ 68.2 Ϯ 3.9 s Ϫ1 and K d ϭ 2.3 Ϯ 0.7 nM. In both cases the specific activity of the reduced CF 1 -⑀ so extrapolated (V m ) is close to the value determined in Fig. 2b. The values of the dissociation constant determined here for the reduced form of CF 1 are close that previously reported by Soteropoulos et al. (K d ϭ 4 nM) (31).
Because of the complexity of the process of activation of intact oxidized CF 1 by reduced Trx, it is not possible to describe its time course with simple kinetic equations. It is then difficult to know which steps limit the rate of formation of reduced CF 1 -⑀ from oxidized CF 1 in the presence of reduced Trx. Since the rate of activation depends on CF 1 concentration even at the beginning of the kinetics (Fig. 4b), the dissociation and reassociation of the ⑀ subunit are probably fast, and other factors probably control the rate of formation of reduced CF 1 -⑀, among them the binding of reduced Trx to CF 1 . Fig. 4c shows the kinetics of activation of 10 nM CF 1 by reduced Trx at different concentrations (from 0.5 to 5 M). It appears qualitatively that in this range, the concentration of Trx at least partially controls the rate of activation. In the initial phase of the kinetics, and for any Trx concentration below 3M, the instantaneous rate of ATP hydrolysis was indeed a unique function of the (time ϫ Trx concentration) product (not shown). The analysis of this function 2 allowed calculation of a gross value for the rate constant of binding of reduced Trx to oxidized CF 1 . It was estimated to be 3 ϫ 10 3 M Ϫ1 s Ϫ1 , about 40 times lower than for CF 1 -⑀.
Thioredoxin-induced Activation of CF 0 CF 1 in Proteoliposomes- Fig. 5a shows the light-induced formation of ATP driven by bacteriorhodopsin-CF 0 CF 1 proteoliposomes, in control conditions or a few minutes after the addition of 3 M reduced Trx prior to illumination. The rate of ATP synthesis was more than doubled after Trx treatment, likely because the amount of activated CF 0 CF 1 at a given magnitude of the protonmotive force was increased (16). It is then possible to follow the thiol reduction of the membrane-bound CF 0 CF 1 by measuring the rate of ATP synthesis. Fig. 5b shows the rate of light-induced ATP synthesis measured with proteoliposomes treated with reduced Trx for different times in darkness (the reaction was stopped by derivatizing Trx active-site thiols with NEM, and the excess of NEM was neutralized further by a small addition of DTT). The activity increased progressively after the addition of Trx, confirming that in the absence of protonmotive force, the reduction of the ␥ disulfide bond by Trx is effective, but slow.
The method used here to measure the rate of reduction of CF 0 CF 1 in darkness cannot be applied to study its reduction in the light because the arrest of the reaction by NEM is not fast enough. We have then developed a method similar to that used in the case of the isolated complex, based here on the continuous monitoring of photophosphorylation. This was achieved by coupling the reaction of ATP synthesis to the reduction of NADP ϩ via the hexokinase/glucose-6-phosphate dehydrogenase system (33). As for ATP hydrolysis, the instantaneous reaction rate was computed directly from the first derivative of the spectrophotometric measurement at 340 nm. Fig. 6 shows the time course of ATP-dependent NADP ϩ reduction in the presence of illuminated proteoliposomes before and after the addition of reduced Trx (a) and the time-dependent rate of ATP synthesis after the addition of Trx (b). The kinetics of reduction of ATP synthase were found to be monophasic, and the rate constant of the process was proportional to the concentration of reduced Trx (Fig. 6c). The thiol activation of the membranebound ATP synthase can therefore be considered limited by the binding of Trx, which allows the determination of the rate binding constant k on (Equations 1 and 2). Here, k on ϭ 6.4 ϫ 10 3 M Ϫ1 s Ϫ1 . Using different proteoliposome preparations, its value ranged between 6 ϫ 10 3 and 7.5 ϫ 10 3 M Ϫ1 s Ϫ1 .

Effect of Partial Uncoupling and Enzyme Turnover on the Rate of Energy-induced Thiol Modulation of the Membranebound ATP Synthase-
We have checked the sensitivity of the rate of CF 0 CF 1 thiol modulation to a limited decrease of the electrochemical proton gradient. This was achieved by monitoring the rate of ATP synthesis before and after the addition of reduced Trx, as in Fig. 6, but at a unique Trx concentration and in the presence of FCCP at different concentrations. The progression curves were submitted to the same numerical analysis as above (Equation 1), and three parameters were drawn from each experiment: the initial rate of ATP synthesis (by CF 0 CF 1 in its oxidized state), the final rate of ATP synthesis (by CF 0 CF 1 in its reduced state), and the rate constant k app of the transition between these two forms. These three parameters are plotted as a function of the FCCP concentration in Fig. 7. The rate constant k app decreases when the uncoupler concentration increases. Its decrease is less pronounced than the loss of activity of the oxidized CF 0 CF 1 but more pronounced than the loss of activity of the reduced CF 0 CF 1 .
In the data of Figs. 6 and 7, the reduction of the membrane-2 By plotting the expression log (V eq Ϫ V) as a function of the (time ϫ Trx concentration) product, V eq being the final rate of ATP hydrolysis and V its instantaneous value, one obtains a linear graph, the slope of which is the binding rate constant k on . The linearity is observed here only at the beginning of the kinetics. In the case of isolated CF 1 -⑀, the activation of which is monophasic, this analysis can be applied to the whole kinetics. It gives the same value of k on as that calculated from Equations 1 and 2, which demonstrates its validity. bound ATP synthase by Trx proceeded during continuous ATP synthesis. To try to discriminate between the respective roles of the electrochemical proton gradient and of the enzyme turnover in the exposure of the ␥-disulfide bond, we also studied this process in the absence of ADP. The principle was identical to that used previously to investigate the possible role of enzyme turnover in the thiol modulation of isolated CF 1 -⑀ (see Fig. 3). Here, reduced Trx (0.4 M) was added to illuminated proteoliposomes, and then ADP was added at different times. The rates of ATP synthesis revealed after ADP addition were compared with those obtained at the same time after the addition of Trx, but with ADP present from the beginning. The results are presented in Fig. 8a, which shows the rates of ATP synthesis as a function of the time after the addition of Trx under the different conditions. Obviously, the different kinetics cannot be superimposed. When the time between the addition of Trx and ADP is increased, it becomes clear that in the absence of ADP the thiol modulation of ATP synthase was accelerated. For each curve, the rate of ATP synthesis was extrapolated to the time of the addition of ADP (Fig. 8a, sym-bols). Its time dependence allowed estimation of the rate constant of thiol modulation in the absence of ADP. It was about double of that obtained in the presence of ADP. Because the rate of Trx binding depends on the protonmotive force, we have verified that ADP did not affect its magnitude. This was achieved by comparing the light-induced quenching of 9-aminoacridine, related to the ⌬pH (34), in the presence of 1 mM ATP and 1 mM ADP (Fig. 8b). No difference was observed, in FIG. 7. Effect of FCCP on ATP synthesis and on the rate of thiol modulation in illuminated proteoliposomes. Conditions were as described under "Experimental Procedures." The apparent rate constant k app was measured as in Fig. 6. Trx concentration, 0.5 M. All values are normalized to those measured without FCCP. q, rate of ATP synthesis with CF 0 CF 1 in the oxidized form; f, rate of ATP synthesis with CF 0 CF 1 in the reduced form; ‚, k app of the transition between the oxidized and the reduced form. Controls: activity of oxidized CF 0 CF 1 , 150 nmol min Ϫ1 mg Ϫ1 of protein; activity of reduced CF 0 CF 1 , 376 nmol min Ϫ1 mg Ϫ1 of protein; k app , 6.3 ϫ 10 Ϫ3 s Ϫ1 . the absence and in the presence of valinomycin, which shows that neither ⌬ nor ⌬pH was affected by ADP. We can then conclude that contrary to the isolated CF 1 -⑀, the activated membrane-bound ATP synthase is thiol-modulated at a rate that is modified by the enzyme turnover.
Kinetic Properties of Oxidized and Thiol-modulated CF 0 CF 1 in Proteoliposomes-For the sake of simplicity, it is generally assumed that oxidized and thiol-reduced CF 0 CF 1 differ only in their ability to be activated by the electrochemical proton gradient and have the same catalytic properties. A simple way to check this proposal is to examine parameters that do not depend on the amount of active enzyme, e.g. Michaelis constants. Fig. 9 shows the result of a typical experiment where the light-induced rate of ATP synthesis catalyzed by oxidized and thiol-reduced CF 0 CF 1 was plotted as a function of ADP concentration. In the case of the oxidized CF 0 CF 1 , the data were fitted satisfactorily with simple Michaelis-Menten kinetics from 1 to 1,000 M. The case of the reduced enzyme was different. The data could not be fitted with simple Michaelis-Menten kinetics, but rather with a model with two independent classes of sites ( Fig. 9, dotted curve). This does not prove that the thiol-modulated CF 0 CF 1 works with two independent classes of catalytic sites because a lot of different models could probably accommodate the data. However, it is clear that under our conditions, the thiol-modulated CF 0 CF 1 works under a different mode than the oxidized CF 0 CF 1 . If one restricts the kinetic analysis to the 1-100 M range and imposes a Michaelis-Menten fit (Fig.  9, solid curves), one finds that the K m (ADP) of the reduced form is approximately the double of the K m (ADP) of the oxidized form. Actually, the K m /V max ratio is approximately the same for both forms. Table I shows that this pattern was confirmed with different preparations of proteoliposomes. Even though some variations were found in the kinetic parameters V max and K m (calculated in the 1-100 M ADP concentration range), K m was always higher in the reduced form than in the oxidized form. Moreover, in the 1-1,000 M concentration range, the saturation curve of the oxidized form was always monophasic, and the saturation curve of the reduced form biphasic (not shown). These experiments show that thiol modulation modifies not only the proportion of active CF 0 CF 1 , but also its catalytic properties.

DISCUSSION
Thiol Modulation of the ATPase Subcomplex and of the CF 0 CF 1 Complex by Thioredoxin-To get a clear picture of the mechanism of CF 0 CF 1 thiol modulation in situ, a prerequisite is to study the interaction between Trx and the ␥ subunit in the isolated ATPase subcomplex and in the whole, membranebound, ATP synthase complex. The present study is a new step in this investigation. Surprisingly, it reports for the first time kinetic data about the interaction between Trx and the isolated CF 1 (with or without its ⑀ subunit). Previous data included qualitative studies of the interaction between CF 1 and different Trxs (22,23), quantitative studies of this interaction at equilibrium using fluorescent probes (23), and kinetic studies of this interaction in thylakoids (21). These different approaches are complementary. From the rate of activation of oxidized CF 1 -⑀ by reduced Trx, we were able to measure the rate constant of formation of the supercomplex between these two species. It was about 10 5 M Ϫ1 s Ϫ1 at pH 8 and 37°C. Because of the quasi-irreversibility of the reaction, neither the dissociation constant of this supercomplex nor the characteristics of further steps of ATPase reduction could be determined.
The K d of the ATPase⅐Trx supercomplex was estimated previously to be about 1 M, regardless of the redox state of CF 1 and Trx (23), but changes in kinetic binding parameters with the redox state remain a priori possible. We have found that k on ϭ 1.2 ϫ 10 5 M Ϫ1 s Ϫ1 if Trx is initially reduced and CF 1 -⑀ initially oxidized. If K d is close to 1 M, k off can be estimated to 0.1 s Ϫ1 . We have tried to know if the occupancy of the binding site of oxidized CF 1 -⑀ by oxidized Trx (up to 4 M) could delay its reduction by reduced Trx. No delay was observed within our resolution time (a few seconds, data not shown), consistent with a k off of 0.1 s Ϫ1 , or higher, for the release of oxidized Trx from the oxidized CF 1 -⑀. At the present time, it seems reasonable to consider that, like the K d , the kinetics of formation and dissociation of the ATPase⅐Trx supercomplex are redox-independent. At first sight, this should exclude large redox-dependent conformational changes of the interacting domains of ATPase and Trx. Such changes are limited indeed in the case of E. coli Trx (35).
If one assumes a K d of 1 M for the ATPase⅐Trx supercomplex  same redox state (data not shown). Binding of Trx has apparently no effect in itself on the ATPase activity, although comparisons of k off (0.1 s Ϫ1 ) and catalytic turnover rate (40 to 80 s Ϫ1 ) indicate that the ATPase⅐Trx supercomplex remains stable during several hundreds of enzyme turnovers. This situation is depicted in Scheme 1a. This rules out possible contacts or collisions of Trx with any part of the stator during the rotation of the ␥-Trx entity in CF 1 -⑀. This is also consistent with the lack of effect of the enzyme turnover on the rate of binding of Trx to isolated CF 1 -⑀. It was also reported that the presence of the ⑀ subunit increases the distance between ␥-Cys 322 and Trx in the ATPase⅐Trx supercomplex without changing its dissociation constant (23). In the present work, the kinetic study of the reduction of intact CF 1 by Trx was complicated by the release of the ⑀ subunit. However, the kinetics obtained at limited Trx concentrations suggests that the binding of reduced Trx to CF 1 is slower than its binding to CF 1 -⑀.
Observations about the activation of the isolated ATPase complex are relevant to the mechanisms only if studies are also made with the membrane-bound ATP synthase complex. Until now, thiol modulation of the membrane-bound CF 0 CF 1 was always revealed by the rate of ATP hydrolysis immediately after the collapse of the protonmotive force (16,36,21). The present work shows that it is also possible to follow the thiol modulation by measuring the increase of the ATP synthesis capacity. When a coupled enzymatic assay is used, this approach allows a continuous monitoring of the ATP synthase activation. It was also relatively easy, using this method, to investigate the effect of ADP on the rate of thiol modulation. In thylakoids, the unmasking of the thiol-sensitive domain of the ␥ subunit by the protonmotive force is well known (13,17,21) and was confirmed here using an artificial system. The role of the enzyme turnover had not been investigated yet, and we have brought the first evidence that Trx binds less rapidly to the working ATP synthase than to the ATP synthase maintained potentially active by the protonmotive force. We have shown that this cannot be attributed to some decrease of the protonmotive force by ATP synthesis. ATP synthesis indeed did not affect the protonmotive force, which was not surprising because of the very low number of CF 0 CF 1 molecules/liposome and their moderate turnover rate. It is tempting to speculate about the opposite effects of protonmotive force and catalytic turnover on the exposure of the ␥ thiol-sensitive domain. This can be done in the frame of the rotational catalysis hypothesis, and a plausible model is depicted in Scheme 1b. In the resting CF 0 CF 1 (deenergized membranes), the thiol-sensitive domain could be hidden by some part of the stator, for example the external stalk formed by subunits I and II. In the presence of a protonmotive force (but in the absence of the substrates of ATP synthesis), a fraction of turn would be achieved by the rotor, leading to the exposure of the regulatory domain of ␥. The balance between the protonmotive force and the return torque would determine the statistical number of enzymes in this activated state. In the presence of protonmotive force, ADP and phosphate, the continuous rotation of ␥ in the activated enzymes would lead to a periodic unmasking of the thiol-sensitive domain. With respect to the simply activated CF 0 CF 1 , the probability of interaction with Trx should be decreased, leading to a decrease of the k on rate constant of Trx binding to CF 0 CF 1 .
Enzyme Activation, Enzyme Turnover, Thioredoxin Binding, and Lifetime of the CF 0 CF 1 ⅐Thioredoxin Supercomplex-In the presence of ADP, the maximal value of the k on rate constant of binding of Trx to membrane-bound CF 0 CF 1 reported here is about 6.5 ϫ 10 3 M Ϫ1 s Ϫ1 . Considering (Fig. 8) that this value is doubled in the absence of ADP, one finds that the maximal value of k on in illuminated proteoliposomes is about 1.3 ϫ 10 4 M Ϫ1 s Ϫ1 . This value should be compared with the maximal value obtained by Schwarz et al. (21) with E. coli Trx in energized thylakoids, about 2.4 ϫ 10 4 M Ϫ1 s Ϫ1 . The latter value, only two times higher, was obtained at a very high protonmotive force, estimated to more than 4 pH units. Although we did SCHEME. 1. Trx-ATP synthase interaction and the rotational catalysis hypothesis. Panel a, isolated CF 1 -⑀ subcomplex. Panel b, membrane-bound CF 0 CF 1 . Left, ATPase (a) and ATP synthase (b) complexes free of Trx. Right, the same complexes with bound Trx. Some values of the rate constants of binding (k on ) and release (k off ) of Trx are indicated in the middle. Rotation rates of the ␥ subunit during enzyme turnover (a, ATP hydrolysis; b, ATP synthesis) are indicated below the complexes. Binding and release of Trx are assumed to be independent of the redox state of Trx and ␥ subunit. In panel a, the isolated ATPase complex and the ATPase⅐Trx supercomplex, both present at M concentrations of Trx, hydrolyze ATP at the same rate. The rotation rates (in turns/s) were deduced directly from the enzyme turnover of the reduced state of CF 1 -⑀ in Fig. 2. The rate of formation of the ATPase⅐Trx supercomplex (k on ϭ 10 3 M Ϫ1 s Ϫ1 ) comes from the data of Fig. 2. The rate of dissociation of the ATPase⅐Trx supercomplex (k off ) was calculated from k on and K d , assuming K d ϭ 1 M (23). The comparison between the rotation rate and the dissociation rate of the ATPase⅐Trx supercomplex indicates that this supercomplex may experience a lot of turns without being dissociated. The quasi-permanent presence of Trx bound to the rotor of CF 1 -⑀ does not affect its rotation. Reciprocally, the rotation induced by ATP hydrolysis does not change the rate of Trx binding to CF 1 -⑀, as shown in Fig. 3. In panel b, the membrane-bound ATP synthase complex exists under three different functional states: top, resting (without proton gradient); middle, proton-activated (with proton gradient but without substrates); bottom, working (driving ATP synthesis). The rate constant k on of Trx binding is different for the three states. Its approximate values, drawn from the present report, are indicated. Trx binding site on the ␥ subunit is represented by a spot. In the resting state, this site is partially hidden, and k on is very low. In the activated state, it is permanently unmasked, therefore k on is increased. In the working state, it is unmasked only a fraction of the time, which leads to an intermediate value of k on . In the case of membrane-bound CF 0 CF 1 , assuming a K d of 1 M for Trx would have two consequences: 1) the CF 0 CF 1 ⅐Trx supercomplex would represent a significant fraction of the ATP synthase in the presence of Trx in the M range, and then it should have the same rotation rate as the Trx-free CF 0 CF 1 (1 or 2.5 s Ϫ1 , depending on the redox state, in the conditions of this report); and 2) the dissociation constant k off should be about 5 ϫ 10 Ϫ3 s Ϫ1 . As developed under "Discussion," these two proposals are in fact mutually exclusive. not carry out absolute measurements of the protonmotive force in our proteoliposomes, we know that it is much lower (37). This is confirmed by the low maximum rates of ATP synthesis (200 -300 nmol of ATP min Ϫ1 mg Ϫ1 of oxidized CF 0 CF 1 ), 30 -50 times lower than that observed when these liposomes are submitted to an artificial protonmotive force (29). These comparisons suggest that the rate of binding of Trx to membrane-bound CF 0 CF 1 saturates at moderate values of the protonmotive force, contrary to the rate of ATP synthesis. Accordingly, when the protonmotive force is lowered with FCCP (Fig. 7), the binding rate diminishes, but less than the rate of ATP synthesis catalyzed by oxidized CF 0 CF 1 . If the decrease of k on with the protonmotive force directly reflects the diminution of the number of activated CF 0 CF 1 (21), this would mean that the more pronounced decrease of the rate of ATP synthesis by oxidized CF 0 CF 1 is due to the combined lowering of activation and catalytic turnover. 3 The situation is actually complex because k on itself is affected by enzyme turnover (Fig. 8). Understanding in more quantitative terms the modulating effect of the catalytic turnover, at different magnitudes of the protonmotive force, is a challenge for future studies.
It is of interest to have an estimation of the rate of dissociation of the CF 0 CF 1 ⅐Trx supercomplex and to correlate it to the enzyme turnover. If the rotatory mechanism is exact, it seems difficult indeed to imagine that the lifetime of this supercomplex exceeds the time of rotation of the enzyme because of the presence of the external stalk. The center-to-center distance between the external stalk and the ␥⑀ internal stalk seems to be about 3 nm (38). The space between these two stalks is probably too narrow to be readily crossed by the bound Trx at each revolution of the rotor (the diameter of Trx is not far from 3 nm). If the K d of the CF 0 CF 1 ⅐Trx supercomplex is 1 M, as for the isolated ATPase (23), since the k on is about 6 ϫ 10 3 M Ϫ1 s Ϫ1 (present work), the k off should be 6 ϫ 10 Ϫ3 s Ϫ1 , which is actually much lower than the apparent catalytic turnover (2 s Ϫ1 in the present conditions). In this case, the supercomplex should be very stable. Consequently, it should be unable to rotate and inactive. This is depicted in Scheme 1b. With K d ϭ 1 M, 33% of CF 0 CF 1 should be inactivated by 0.5 M Trx, 50% by 1 M Trx, and 67% by 2 M Trx. Because this was not observed, it is probable that the K d is much higher than 1 M in the case of CF 0 CF 1 or that the binding parameters depend on the redox state of the partners. This problem obviously deserves further investigations.
Catalytic Properties of Oxidized and Thiol-reduced CF 0 CF 1 -We have compared the ATP synthesis activity of oxidized and thiol-reduced CF 0 CF 1 at various ADP concentrations. Surprisingly, this is the first time, to our knowledge, that such a comparison is made without ambiguity. Because of the negligible contribution of ATP synthesis to the total proton flow in the present conditions, our measurements were made at a constant value of the protonmotive force, which avoids side effects generally encountered in this kind of study, neglected by most investigators, but stressed by others (39 -41). For the sake of simplicity, it was assumed previously that the functional characteristics of the catalytic reaction did not depend on the redox state of the enzyme, and it was concluded from this postulate that contrary to the maximum activation, the maximum rate of catalytic turnover of the oxidized CF 0 CF 1 was reached at a rather low protonmotive force magnitude (16). The present data do not allow a refined kinetic study, but they are sufficient to show that oxidized and thiol-reduced CF 0 CF 1 actually have different catalytic properties. Because they respond in a different way to the ADP concentration, there is no reason to believe that they respond in an identical way to variations of the protonmotive force, and the conclusions based on this postulate should be questioned. The non-Michaelian pattern observed here in the case of thiol-reduced CF 0 CF 1 differs from previous observations in thylakoids where ATP synthesis catalyzed by the thiol-reduced ATP synthase obeyed simple Michaelis-Menten kinetics (41,42). Although different reasons could be invoked to explain this discrepancy (different magnitudes of the protonmotive force or different pH, for example), at the present time the actual reason is not clear, and a detailed comparative kinetic study, out of the present scope, would be necessary to solve this problem.