ATP synthesis by the F0F1-ATPase from the thermophilic Bacillus PS3 co-reconstituted with bacteriorhodopsin into liposomes. Evidence for stimulation of ATP synthesis by ATP bound to a noncatalytic binding site.

F-type ATPase from the thermophilic Bacillus PS3, TF0F1, which was essentially free of bound nucleotides after isolation and purification, was co-reconstituted into liposomes with the light-driven proton pump bacteriorhodopsin. The time course of the light-induced ATP synthesis was biphasic; an initial slow phase accelerated to a final steady-state rate two to three times faster. Adding ATP before initiating the reaction suppressed the slow phase, suggesting that the state of occupancy of specific sites by ATP regulated the synthetic activity of TF0F1. Incubating the purified TF0F1 with ADP and ATP revealed one ADP and two ATP binding sites that were stable to gel filtration. We analyzed the time courses of light-induced ATP synthesis for the enzyme with different nucleotide content, after co-reconstitution into liposomes with bacteriorhodopsin. The two ATP sites were identified to have regulatory function. A complex containing TF0F1•ADP, 1:1, was co-reconstituted with various quantities of ATP to obtain a range of molar ratios of TF0F1•ADP:ATP of between 1:0 and 1:1.7. It was found that the initial rate of ATP synthesis increased with the level of ATP bound to the enzyme. After binding one ATP, a stimulation of ATP synthesis by a factor of 2 was observed. The second ATP site also exhibited regulatory properties. It stimulated ATP synthesis but to a much smaller extent; the stimulation did not exceed 20%. Binding of the photoreactive analogues 2-azido-[α-32P]ADP and 2-azido-[α-32P]ATP to the TF0F1 and their effects on the rate of ATP synthesis are described further. Importantly, after covalent labeling of the enzyme, tryptic digestion, and high performance liquid chromatography purification, the label was found associated with the βY364-containing tryptic peptide in all cases. βY364 is in the region of conserved residues GXEHYXXA, which is in the β subunit and known to be part of the noncatalytic site.

The proton-translocating ATPase from the thermophilic Bacillus PS3, TF 0 F 1 , 1 is a transmembrane protein that catalyzes ATP synthesis coupled with proton flux across the membrane (1,2). It belongs to the class of F-type ATPases that all have in common a membrane-embedded F 0 part, which mediates transmembrane proton conduction, and a hydrophilic F 1 part, which contains the nucleotide binding sites (3)(4)(5). The F 1 part is composed of five subunits with the stoichiometry ␣ 3 ␤ 3 ␥␦⑀ regardless of the organism (6 -8). F 1 has been studied extensively with respect to the nucleotide binding sites, and the structure at 2.8 Å resolution has been reported recently for beef heart F 1 (9). It is generally accepted that there are six nucleotide binding sites, three of them being potentially catalytic and the other three noncatalytic (10 -12). In this context the nucleotide sites of the detached TF 1 (13), the core complex (14), as well as the isolated ␣ and ␤ subunits (15) of the thermophilic ATPase have been characterized in some detail. Interestingly, the TF 1 complex has no endogenously bound adenine nucleotides (16), a characteristic that distinguishes it from the F 1 of other species that contain 1-4 endogenously bound adenine nucleotides/enzyme (17,18). This is an important fact that essentially simplifies the study of the nucleotide binding sites so that TF 1 can be seen as a model system with respect to the binding sites. However, despite all of this information little is known about the binding sites in the intact TF 0 F 1 and more surprisingly nothing about their roles during the synthesis of ATP.
To understand the exact functional and/or mechanistic role of the nucleotide binding sites during ATP synthesis, we have co-reconstituted the whole enzyme TF 0 F 1 with the light-driven proton pump bacteriorhodopsin into liposomes. Upon illumination, relatively high rates of ATP synthesis between 200 and 700 nmol of ATP/mg of TF 0 F 1 /min were observed. Although the rate of synthesis was limited by the light-induced pH gradient attainable with bacteriorhodopsin, the co-reconstituted system has various advantages. (i) It is chemically well defined, i.e. only two purified enzymes and a defined lipid composition. (ii) It is physically well defined, i.e. proteoliposomes are homogeneous in size and in protein distribution and orientation. (iii) It provides a stable and constant transmembrane electrochemical potential gradient for many hours.
In this work we have used a previously described procedure (for review see Ref. 19) for co-reconstitution of TF 0 F 1 with bacteriorhodopsin to analyze in more detail the kinetics of ATP synthesis and the role of the nucleotide binding sites. Three tight binding sites that were stable to gel filtration were identified and characterized. The ATP binding sites were identified to be responsible for an acceleration of ATP synthesis by a factor of 2-3, demonstrating for the first time that enzymebound nucleotides directly affect synthesis of ATP by intact F-type ATPase. This is in agreement with previous reports on the soluble F 1 part (20, 21). Use of the photoaffinity analogues 2-N 3 [␣-32 P]ADP and 2-N 3 [␣-32 P] ATP allowed us to identify the three binding sites all of which contained the derivatization of ␤Y364, which is in the region of conserved residues GXE-HYXXA. This peptide is common in all F-type ATPases and is located on a loop that reaches into the noncatalytic site on the ␣ subunit (9).
Enzyme Preparation and Assays-TF 0 F 1 from the thermophilic Bacillus PS3 was in part a kind gift from Prof. Bä uerlein (Martinsried, Germany). The enzyme was isolated and purified as described by Kagawa and Yoshida (25); for all calculations, a molecular weight of 550,000 g/mol was used. Protein concentration was determined with a protein assay (Bio-Rad) using bovine serum albumin as a standard assuming an extinction coefficient for bovine serum albumin of (⑀ 279 ϭ 0.667 mg Ϫ1 ml cm Ϫ1 ). ATPase hydrolytic activity was determined spectrophotometrically using the enzyme-linked assay of Froud et al. (26) in which ADP production was detected by measurement of NADH oxidation.
Analysis of Enzyme-bound Nucleotides-Binding of adenine nucleotides to TF 0 F 1 was done following incubation of the protein (2-3 mg/ml) for 15 min at 40°C in a buffer containing 50 mM K 2 SO 4 , 50 mM Na 2 SO 4 , 1 mM MgSO 4 , 25 mM KH 2 PO 4 (pH 7.3), 5 mg/ml Triton X-100, and the indicated amounts of nucleotide. TF 0 F 1 -nucleotide complex was separated from medium nucleotides by three consecutive centrifuge elutions on a 1-ml syringe column of Sephadex G-50 equilibrated with the same buffer (27). The protein concentration was measured after the columns (about 70% of the protein was recovered) (28). Free and bound nucleotides were determined as described earlier (29).
Photolabeling-TF 0 F 1 (2-3 mg/ml) was preincubated for 15 min at 40°C in the dark with the indicated concentration of 2-N 3 [␣-32 P]ATP in 25 mM KH 2 PO 4 buffer (pH 7.3) and 50 mM K 2 SO 4 , 50 mM Na 2 SO 4 , 2 mM MgSO 4 , and 1 mg/ml Triton X-100. Following consecutive centrifugation columns, the enzyme⅐2-N 3 [␣-32 P]ATP complex was filled in a quartz cuvette and exposed to UV light for 30 s. The cuvette was placed in the focused light beam with a 10-cm distance from an SLM Aminco UV lamp (Urbana, IL) equipped with a 450-watt high pressure xenon lamp. To measure the fraction of the covalently bound nucleotides, the enzyme was precipitated and washed with trichloroacetic acid (30 g/liter). After solubilization in 200 mM NaOH and 0.1% sodium dodecyl sulfate the radioactivity and protein content were measured. Although control experiments indicated no significant damage to the protein by UV irradiation, it was applied in all controls when comparing the effects of prior incubation with 2-N 3 [␣-32 P]ATP, 2-N 3 [␣-32 P]ADP, ADP, or ATP.
Identification of the Binding Site-The labeled enzyme was trypsin treated as described by Xue et al. (30). The peptides were then purified by HPLC using first an ion exchange column and subsequently a reverse phase column as described by Wise et al. (31). The ion exchange column (Whatman Partisil PX S25/SAX10) was eluted with a linear gradient starting from 100% solvent A (29:71, acetonitrile, 0.01 M NaH 2 PO 4 (pH 4.0) to 100% of eluent B (29:71, acetonitrile, 0.4 M NaH 2 PO 4 (pH 3.0)) starting 10 min after loading and lasting 50 min. The reverse phase column (Vydac C4) was eluted with a nonlinear gradient as shown in Fig. 7, B and D. Eluent A was 0.1 volume % trifluoroacetic acid and eluent B 0.1 volume % trifluoroacetic acid and 90 volume % acetonitrile. The labeled peptides were identified according to their radioactivity as determined by Cerenkov counting. Nterminal sequencing using the classical Edman degradation was performed using a pulsed liquid Sequencer 474A from Applied Biosystems.
Co-reconstitution of TF 0 F 1 and Bacteriorhodopsin-Proteoliposomes containing bacteriorhodopsin and TF 0 F 1 were reconstituted according to a recently published procedure (for review see Ref. 19). Unilamellar liposomes were prepared by reverse phase evaporation using a mixture of phosphatidylcholine and phosphatidic acid (molar ratio 9:1) and resuspended at a lipid concentration of 4 mg/ml, i.e. 5 mM. Then Triton X-100 was added under vortexing to a final concentration of 8 mg/ml. Presolubilized bacteriorhodopsin and TF 0 F 1 were added to obtain weight ratios of 20:1 and 133:1, respectively. n-Octyl ␤-D-glucopyranoside was added to a final concentration of 20 mM, and the mixture was incubated for 5 min. The detergent was removed by four successive additions of 80 mg/ml washed Bio-Beads SM-2; additions were made every hour. All reconstitutions were performed at room temperature in 25 mM KH 2 PO 4 -KOH (pH 7.3), 50 mM K 2 SO 4 , and 50 mM Na 2 SO 4 . A detailed analysis of the optimization of the co-reconstitution will be published elsewhere. 2 Light-induced ATP Synthesis-After detergent removal, the proteoliposomes were diluted in the same buffer used for their preparation and supplemented with 2 mM ADP or 2 mM Mg 2ϩ . This mixture was incubated at 40°C and preilluminated for 15 min. The ATP synthesis was started by adding 2 mM ADP when 2 mM MgSO 4 was present, or alternatively by 2 mM MgSO 4 when 2 mM ADP was present. For some experiments ATP synthesis was initiated by light after preincubation in the dark with ADP and MgSO 4 . Aliquots of the illuminated samples were taken at different reaction times, quenched with an equal volume of trichloroacetic acid (40 g/liter), and measured for their ATP content using luciferin-luciferase assay (29). When ATP was removed continuously from the reaction medium, the proteoliposomes were incubated at 40°C in a 1-ml spectrophotometer cuvette containing the buffer supplemented with 1 mM NADP, 1 mM glucose, 2 units of hexokinase, and 2 units of glucose-6-phosphate dehydrogenase. At intervals the optical density of the mixture was measured at 340 nm. Fig. 1 illustrates a typical experiment in which the light-induced ATP synthesis by co-reconstituted bacteriorhodopsin-TF 0 F 1 proteoliposomes is measured. Proteoliposomes (lipid/bacteriorhodopsin ϭ 20:1 w/w; lipid/TF 0 F 1 ϭ 133:1 w/w) were first incubated at 40°C in the presence of 2 mM ADP. After a 15-min preillumination period which ensured generation of a stable transmembrane pH gradient, the reaction was started by the addition of Mg 2ϩ at time zero. Aliquots of the illuminated sample were taken at different reaction times, and their ATP content was measured with the luciferin-luciferase assay. The most FIG. 1. Light-driven ATP synthesis by bacteriorhodopsin-F 0 F 1 proteoliposomes. Proteoliposomes were reconstituted at 25°C from phospholipid-bacteriorhodopsin-TF 0 F 1 -Triton X-100 micellar solutions supplemented with 20 mM octyl glucoside before detergent removal. Final concentrations: lipid (4 mg/ml), bacteriorhodopsin (200 g/ml), TF 0 F 1 (30 g/ml), Triton X-100 (8 mg/ml) in a medium containing 50 mM Na 2 SO 4 , 50 mM K 2 SO 4 , 25 mM KH 2 PO 4 (pH 7.3). CF 0 F 1 was co-reconstituted with bacteriorhodopsin under similar experimental conditions except that octyl glucoside was omitted. After detergent removal by successive addition of Bio-Beads SM-2, proteoliposomes were resuspended in the same medium, supplemented with 2 mM ADP, and preilluminated for 15 min at 40°C. At time zero ATP synthesis was initiated by adding 2 mM MgSO 4 . Aliquots were analyzed as a function of time for their ATP content using luciferin-luciferase assay. Shown are time courses of ATP synthesis by TF 0 F 1 proteoliposomes in the light (trace a), in the presence of carbonyl cyanide p-trifluoromethoxyphenylhydrazone (trace b), and in the dark (trace c). The time course of ATP synthesis by CF 0 F 1 -bacteriorhodopsin proteoliposomes in the light is shown in trace d.

Time Course of Light-induced ATP Synthesis-
striking feature of the data presented in Fig. 1, trace a, is that TF 0 F 1 synthesized ATP in two distinct kinetic phases. Initially the rate of ATP synthesis was slow; then the rate increased progressively until, after about 20 min, a steady state was reached which remained constant for at least 3 h of illumination. In all of the experiments performed, independent of the amplitude of the preformed pH gradient, the lipid to protein ratios, or the concentration and the order of addition of ADP and Mg 2ϩ , this biphasic behavior was always observed with a final steady-state ATP synthesis rate about two to three times faster than the initial rate. Increasing the preillumination period demonstrated that the time lag between the initial and final rates of ATP synthesis was not related to a slow buildup of ⌬] H ϩ across the proteoliposome membrane. Fig. 1 also shows that in the presence of the uncoupler carbonyl cyanide p-trifluoromethoxyphenylhydrazone (trace b) or in the dark (trace c) the light-induced ATP synthesis was inhibited completely (Ͼ98%), excluding any contribution of some adenylate kinase activity in the TF 0 F 1 preparations. For comparison the eukaryotic F-type ATPase from chloroplasts was co-reconstituted using a similar procedure (19,32) and analyzed in the same way. Trace d in Fig. 1 shows that in contrast to TF 0 F 1 , CF 0 F 1 exhibits a monophasic behavior, and ATP synthesis proceeds with a constant rate as soon as the reaction is initiated (similar monophasic behavior was also observed in proteoliposomes co-reconstituted with the F 0 F 1 -ATPase from pig heart mitochondria (data not shown)).
Thus it appears that the biphasic behavior observed during ATP synthesis by TF 0 F 1 is an intrinsic property of this ATPase. Another characteristic that in part distinguishes TF 0 F 1 from F-type ATPases from other sources such as chloroplasts and mitochondria is that it can be isolated without any nucleotide endogenously bound to any of the six sites/enzyme (16). This prompted us to check whether the acceleration occurring during ATP synthesis through TF 0 F 1 could be related to saturation of some ATP binding sites by newly synthesized ATP.
Effect of ATP on Biphasic Behavior-To check whether the increase in the ATP synthesis rate was due to the accumulation of ATP in the reaction medium, bacteriorhodopsin-TF 0 F 1 proteoliposomes were preilluminated for 15 min in the presence of Mg 2ϩ , and the reaction was started by the addition of ADP together with variable amounts of ATP. Fig. 2 shows the time courses of light-induced ATP synthesis as a function of the concentration of ATP initially present together with ADP. As the initial concentration of ATP was increased from 0 to 10 M, the initial rate of ATP synthesis increased with a concomitant decrease in the time lag needed to reach the final high steadystate rate of ATP synthesis. When 10 M ATP was added together with 2 mM ADP, light-induced ATP synthesis proceeded immediately to the rapid kinetic phase. Interestingly, the final steady-state rates were found to be independent of the presence of ATP. It can also be noted that in the control experiment, i.e. in the absence of added ATP, the fast rate occurred when about 10 M ATP had accumulated.
The role of ATP in determining the kinetic behavior of ATP synthesis was investigated further by analyzing the time courses of the light-induced ATP synthesis in experimental conditions where newly synthesized ATP was continuously removed from the reaction medium. In Fig. 3 the rate of ATP synthesis by bacteriorhodopsin-TF 0 F 1 proteoliposomes was measured through the formation of NADPH in a reaction medium containing glucose, hexokinase, glucose-6-phosphate dehydrogenase, and NADP (closed symbols). Under these conditions the rate of ATP synthesis was not accelerated as compared with the control experiment in which glucose was omitted and the synthesized ATP was measured by the luciferin-luciferase assay (open symbols). Throughout the illumination period, the rate of ATP synthesis remained low, whereas in the control experiment the biphasic behavior was again observed with an acceleration by a factor of 2.5 when about 10 M ATP had been accumulated in the medium. 3 The experiments depicted in Figs. 2 and 3 demonstrated that, depending upon the concentration of ATP present in the reaction medium, TF 0 F 1 can exist in two different states, slow and a fast, with respect to the rate of ATP synthesis.
K m for ADP and P i -To investigate whether the different nucleotide content has an effect on the K m and the V max of ATP synthesis we measured the rates of ATP synthesis for the two different states as a function of ADP and P i concentrations. Fig.  4A shows the results of the dependence of the rates of ATP synthesis on ADP concentrations. Slow rates (closed symbols) were determined from the slopes of ATP synthesized during the 3 In the previously reported co-reconstitution studies of TF 0 F 1 with bacteriorhodopsin (33,34), light-induced ATP synthesis were measured using the glucose-hexokinase assay. This may explain in part the very low rates of ATP synthesis reported by these authors as compared with those reported in our reconstituted systems. Bacteriorhodopsin-TF 0 F 1 proteoliposomes were reconstituted as described in Fig. 1. After reconstitution, they were resuspended in a medium supplemented with 1 mM NADP, 2 units/ml hexokinase, 2 units/ml glucose-6-phosphate dehydrogenase, and with or without 1 mM glucose. After preillumination for 15 min in the presence of 2 mM MgSO 4 , 2 mM ADP was added, and ATP synthesis was measured as a function of time with luciferin-luciferase (E) or with the coupled assay when glucose was present (q). first 10 min after starting the reaction, whereas fast rates (open symbols) were measured after more than 10 M ATP had accumulated. The concentration dependence of both rates can be described by Michaelis-Menten kinetics. From these results the K m values were found to be 300 M and V max to be 124 nmol/ mg/min for the slow rates. The fast rates exhibited the same K m , whereas the V max was 302 nmol/mg/min, which is a factor 2.4 faster than for the slow state. Similar behavior was observed with various concentrations of inorganic phosphate. The K m for both slow and fast rates was approximately 10 mM; however, the V max for the fast state was two to three times faster than that of the slow rate (Fig. 4B). The solid lines in Fig.  4, A and B, were computer-fitted to obtain the K m and V max .
Role of Sites for Tightly Bound Nucleotides-The 2-3-fold stimulation of the ATP synthesis rate by ATP suggested on a molecular level that ATP binding sites on the TF 0 F 1 could have an important role in the regulation of the catalytic process. We thus analyzed in detail the possible relationships between nucleotide binding to ATPase and rate of ATP synthesis.
First we determined the nucleotide content of our purified TF 0 F 1 preparation. The amounts of ADP and ATP endogenously bound were found to be 0.4 and 0.1 mol/mol of TF 0 F 1 , respectively (see Table I), confirming previous reports on purified TF 1 which has been isolated almost free of tightly bound nucleotides (35). In a second set of experiments, the purified solubilized TF 0 F 1 preparation was incubated in the presence of nucleotides. To mimic the experimental conditions used for measuring light-induced ATP synthesis in proteoliposomes, TF 0 F 1 was first incubated for 15 min at 40°C in the presence of ADP and Mg 2ϩ , followed by a second incubation in the presence of various amounts of ATP. Incubation of TF 0 F 1 in the presence of 100 M ADP and 1 mM Mg 2ϩ , followed by gel filtration through Sephadex columns equilibrated in the same medium but without ADP, indicated that 1 mol of ADP was tightly bound per mol of enzyme. Increasing the incubation time or the ADP concentration did not lead to further ADP binding, and a 1:1 TF 0 F 1 ⅐ADP complex was obtained. Preincubation of TF 0 F 1 as above with ADP and Mg 2ϩ , followed by a second incubation in the presence of ATP, revealed, in addition to one tightly bound ADP, about two tightly bound ATP. As shown in Fig. 5 the binding of ATP to TF 0 F 1 is biphasic with respect to the ATP concentration in the second incubation medium. 100 M ATP led to 1 mol of ATP/mol of TF 0 F 1 , and at 1 mM ATP to about 1.7 mol of ATP tightly bound per mol of enzyme. 4 Interestingly, although the unbound ATP could be hydrolyzed during the incubation, the TF 0 F 1 ⅐ADP⅐ATP complexes were found to be stable during repeated gel filtration indicating, that the bound ATPs were not hydrolyzed even in the presence of 1 mM Mg 2ϩ . These findings suggest the presence of distinct nucleotide binding and nucleotide hydrolyzing sites (see also Ref. 35).
To investigate the effect of the tightly bound ATP, TF 0 F 1 , pretreated as described above, was co-reconstituted with bacteriorhodopsin, and the resulting rates of light-induced ATP synthesis were measured. ATP synthesis was started by adding 2 mM ADP simultaneously with illumination of the proteoliposomes. After a time lag of about 2 min, allowing the generation of a stable transmembrane pH gradient, the rate of ATP synthesis was monitored over a period of 10 min. Fig. 6 gives the rates of ATP synthesis as a function of ATP tightly bound to TF 0 F 1 . The rates of ATP synthesis increased from 58 nmol of ATP/mg/min when no ATP was tightly bound to 100 nmol of ATP/mg/min when 1 ATP was tightly bound to the TF 0 F 1 . A further increase in tightly bound ATP produced a small stimulation of ATP synthesis of 10 -20%.
Another important observation was that the stimulation through ATP was not observed when the proteoliposomes were preilluminated. Under conditions in which light-induced ATP synthesis was started by adding ADP after a 15-min preillumination of proteoliposomes, the stimulation was no longer observed, and ATP synthesis proceeded with a rate of 58 nmol of ATP/mg/min (only after 20 min of illumination the rate accelerated by a factor of 2-3). This observation indicated that the tightly bound ATP may have been released or changed to another binding site under energized conditions. Also important is that when purified TF 0 F 1 was incubated with ATP without prior incubation with ADP, the same nucleotide to enzyme ratios were found after gel filtration (i.e. 1 ADP and 1.7 ATP/enzyme). However co-reconstitution of this treated 4 The concentration of added ATP is not constant throughout the incubation period because of ATP hydrolysis. Thus, the average ATP concentrations during incubation are much lower than that reported in the abscissa of Fig. 5. This also explains why the results from this figure were not analyzed in terms of Scatchard plots. On the other hand, at high ATP concentrations (above 2 mM), because of the difficulties in removing the free nucleotides, the estimation of the tightly bound fraction is not very accurate.

FIG. 4. Rate of ATP synthesis by bacteriorhodopsin-TF 0 F 1 proteoliposomes as a function of ADP or P i concentrations.
Proteoliposomes were reconstituted as described in Fig. 1 (except for a lipid/ bacteriorhodopsin of 10:1 w/w) in a buffer containing 50 mM Na 2 SO 4 , 50 mM K 2 SO 4 , and 25 mM Pipes (pH 7.4). After detergent removal, the proteoliposomes were resuspended in the same buffer, supplemented with 2 mM MgSO 4 , and 1-30 mM potassium phosphate (pH 7.4). After preillumination for 15 min, ATP synthesis was initiated by adding 12.5-5,000 M ADP. Panel A, rates of ATP synthesis as a function of ADP concentration. For all samples 25 mM phosphate buffer was used. q, initial rates of ATP synthesis measured up to 10 min after the reaction was initiated. E, steady-state rates of ATP synthesis measured 20 min after the reaction was initiated. Panel B, rates of ATP synthesis as a function of P i concentration. For all samples 2 mM ADP was used. q, initial rates of ATP synthesis measured up to 10 min after the reaction was initiated. E, steady-state rates of ATP synthesis measured 20 min after the reaction was initiated. enzyme into liposomes led to a slow initial ATP synthesis rate of 58 nmol/mg/min. This observation suggested that different sites were occupied depending upon the order of incubation with ADP and ATP during the pretreatment conditions of TF 0 F 1 .
Photochemical Labeling of TF 0 F 1 by 2-Azidonucleotides-To locate the nucleotide binding sites on TF 0 F 1 we used the photoaffinity nucleotide analogues 2-N 3 [␣-32 P]ADP and 2-N 3 [␣-32 P]ATP for covalent labeling. First we analyzed in detail the stoichiometry of labeling by the 2-azido analogues. To quantify the amount of 2-azido nucleotides covalently bound to the enzyme we incubated the solubilized enzyme at 40°C in the dark in a medium containing 100 M 2-N 3 [␣-32 P]ATP and Mg 2ϩ . After a 15-min incubation the TF 0 F 1 -nucleotide complex was separated from unbound labeled nucleotides by gel filtration through Sephadex columns, irradiated, and analyzed for its nucleotide content. Under these conditions we found 1.2 labeled nucleotides/enzyme, which were stable to acid precipitation, i.e. covalently bound (Table I, fourth row). Using unlabeled nucleotides, Table I (third row) shows that 100 M ATP led to 0.8 ADP and 1 ATP tightly bound per enzyme. Taking into account the initial amount of tightly bound nucleotides, which were 0.4 ADP and 0.1 ATP (Table I, first and second rows), it can be estimated that the ATP treatment led to an additional 1.3 nucleotides, comprising 0.4 ADP and 0.9 ATP. In another set of experiments we preincubated the enzyme with 100 M ADP before the addition of 100 M 2-N 3 [␣-32 P]ATP. After Sephadex columns and subsequent irradiation we found 0.6 azido nucleotides/ enzyme which were stable to acid precipitation. In the control 0.7 additional ATP was found (Table I, fifth and sixth rows).
Second we investigated the effects of covalent labeling of TF 0 F 1 by 2-azidonucleotides on the kinetics of ATP synthesis after co-reconstitution of the labeled TF 0 F 1 with bacteriorhodopsin. After treating with 100 M ADP, 100 M 2-N 3 [␣-32 P]ATP, and photoirradiation before reconstitution, ATP synthesis was monophasic (Table I,   3), 1 mM Mg 2ϩ , and 5 g/liter Triton X-100 with the indicated nucleotides (in fifth and sixth rows ADP incubation was followed by a second incubation of 15 min at 40°C with either 100 M ATP or 100 M 2-N 3 [␣-32 P]ATP). After Sephadex columns the amounts of tightly bound ADP and ATP were measured. Covalently bound 2-N 3 [␣-32 P]AXP was measured from the radioactivities of the acid precipitates. After co-reconstitution of the different treated TF 0 F 1 with bacteriorhodopsin, the proteoliposomes were incubated for 15 min at 40°C under constant illumination, and light-induced ATP synthesis reactions were initiated with 2 mM ADP. In the experiments designated by a, ATP synthesis was started by adding 2 mM ADP simultaneously with illumination. Rates of ATP synthesis (average of at least three experiments) were normalized in regard to the rates of the initial slow phase measured in the controls, that is, in proteoliposomes co-reconstituted with the untreated enzyme. The rates of ATP synthesis were lower as compared with the controls in all cases. This can be due to inhibition of fractions of enzyme or a chemical modification that is unfavorable for catalysis.  6. Correlation between ATP tightly bound to TF 0 F 1 and stimulation of ATP synthesis by bacteriorhodopsin-TF 0 F 1 proteoliposomes. Purified TF 0 F 1 preparations treated as described in Fig.  5 were recovered after gel filtration and contained 0.08 -1.7 mol of ATP/mol of TF 0 F 1 . The different treated enzymes were co-reconstituted with bacteriorhodopsin as described in Fig. 1. ATP synthesis was then initiated directly with light in the presence of 2 mM ADP, and aliquots were analyzed for their ATP content every 3 min for 15 min. The figure gives the rate of light-induced ATP synthesis as a function of ATP tightly bound to TF 0 F 1 . incubated with 2-N 3 [␣-32 P]ATP was not photoirradiated before reconstitution (not shown) or 2-N 3 [␣-32 P]ATP was replaced by unlabeled ATP. Under these conditions monophasic behavior was found to correspond to the fast rate of ATP synthesis (Table I, fifth row). However, when TF 0 F 1 was pretreated with 100 M 2-N 3 [␣-32 P]ATP without prior incubation with ADP, ATP synthesis was monophasic with a very slow rate (Table I, third and fourth rows). These results confirmed that the nucleotide pattern of TF 0 F 1 by ATP was dependent upon the order of incubation with ADP and ATP.
When TF 0 F 1 was incubated with 2-N 3 [␣-32 P]ADP and photoirradiated, the light-induced ATP synthesis (although halved) exhibited a biphasic behavior as in the control with ADP (Table I, seventh and eighth rows). It is unlikely that the acceleration in ATP synthesis is a result of the release of inhibitory ADP as suggested for the mitochondrial ATPase (20, 21, 36) since (i) in mitochondria the ADP is released from a catalytic site whereas in our case it is a noncatalytic site (see next paragraph), and (ii) the ADP analogue is bound covalently.
It should be noted that in all of these experiments the rates of ATP synthesis with the covalently photolabeled enzyme were lowered as compared with the controls. This could be due to inhibition of fractions of enzyme (37) or to a chemical modification which is in turn unfavorable for catalysis (15,38).
Identification of the Binding Sites-TF 0 F 1 labeled with 2-N 3 [␣-32 P]ATP in the presence of 1 mM Mg 2ϩ was examined by sodium dodecyl sulfate-gel electrophoresis. Autoradiography of the gel revealed exclusive labeling of the ␤ submits (data not shown).
To identify the amino acid sequence of the ADP and ATP binding sites we trypsin treated the labeled enzyme and purified the fragments by HPLC using ion exchange and a reverse phase column. After preincubation with 100 M ADP followed by 100 M 2-N 3 [␣-32 P]ATP and photoirradiation, the ion exchange column revealed two peaks, an ADP and an ATP peak (Fig. 7A). The ATP peak was collected and applied to a reverse phase column. It resulted in a single peak that eluted at 28% of eluent B (Fig. 7B). Incubating with 100 M 2-N 3 [␣-32 P]ADP followed by photoirradiation resulted in a single ADP peak after the ion exchange (Fig. 7C). We collected this peak and applied it to a reverse phase column. The resulting single peak eluted at the same place as the ATP, at 28% of eluent B (Fig.  7C). Amino acid sequencing of the fragment isolated by reverse phase HPLC gave, in both cases, the same sequence, ALA-PEIV. . . . This sequence was identified in the ␤ subunit (39) as 353-ALAPEIVGEEHYQVA-367 with the Tyr-364 as the labeled amino acid (see also Ref. 15). Tyrosine 364 is contained in the sequence GXEHYXXA, which is a highly conserved sequence in F-type ATPases from bovine mitochondria, spinach chloroplast, and Escherichia coli (30) and is part of the noncatalytic or regulatory site. DISCUSSION The studies described in this paper represent an effort to understand the mechanisms and particularly the complex patterns of nucleotide binding on F-type ATPase involved during ATP synthesis. The most interesting finding emerging from this report is that ATP synthesis by the F-type ATPase from the thermophilic Bacillus PS3 is shown to be significantly accelerated upon ATP binding to the protein. On a more general level, this is the first report for an intact F-type ATPase effect on the regulation of the catalytic process, i.e. ATP synthesis, by an enzyme bound nucleotide. In earlier reports regulatory effects were only described for the mitochondrial F 1 in hydrolysis (20, 21). TF 0 F 1 possesses fundamental similarities to other energytransducing ATP synthases of F-type. In particular there is an agreement that F 1 , the water-soluble catalytic moiety of this protein, contains six nucleotide sites that fall in two categories: (i) three binding sites associated with a regulatory function, and (ii) three catalytic sites. Despite a large number of valuable reported studies on their binding properties as well as on their functional and structural roles, a comprehensive classification of the six sites is still lacking. There is now agreement about their exact location; the catalytic sites are on the ␤ subunits and the noncatalytic on the ␣ subunits. Although the structure is well known, i.e. the ␤Y from the conserved sequence GXE-HYXXA is on a loop that reaches into the ␣ subunit to the noncatalytic binding site (9), the roles of the nucleotides on the ␣ subunits remain unknown (10,11,40). It should be stressed that most investigations have been made with the isolated F 1 , i.e. during hydrolysis of Mg-ATP, and only few with the holoenzyme, although it is of primary importance since only the TF 0 F 1 is capable of ATP synthesis, which is the normal physiological response.
The use of bacteriorhodopsin-TF 0 F 1 co-reconstituted proteoliposomes was very advantageous as it allowed investigation of the mechanism of ATP synthesis and regulation by enzymebound nucleotides. They are powerful tools for analyzing detailed mechanisms inaccessible to study in complex native membranes and allow synthesis through repeated enzyme cycles over a long time of experimental observation. For comparison it should be pointed out that using artificially imposed ⌬] H ϩ transitions (⌬pH/⌬ jumps), the reaction is terminated within a few seconds (41), rendering detailed kinetic evaluations difficult. The only limitation of the co-reconstituted systems is related to the small ⌬] H ϩ generated by bacteriorhodopsin which limits the turnover rate of the H ϩ -ATP synthase. However, for the purpose of the work presented here, this limitation is an advantage since it allows a slowdown in the process of ATP synthesis making the study of its kinetic behavior easier.
It is clear from this study that TF 0 F 1 is subject to regulation during ATP synthesis which depends on the state of occupancy of some specific sites by ATP. The first experimental evidence came from the observation that the rate of ATP synthesis by TF 0 F 1 was dependent upon the concentration of ATP in the reaction medium. When purified TF 0 F 1 was co-reconstituted into proteoliposomes with bacteriorhodopsin, and a transmembrane pH gradient had been established, ATP synthesis proceeded in two phases: an initial slow one that gradually accelerated to a final steady-state rate. The steady-state rate was about two to three times higher than the initial slow rate and commenced when approximately 10 M newly synthesized ATP had accumulated in the assay medium. Continuous removal of synthesized ATP led to a slow monophasic ATP synthesis (Fig.  3). Conversely, adding exogenous ATP before initiating the reaction decreased the transition time between the slow and the rapid phases. In the presence of 10 M exogenous added ATP, ATP synthesis proceeded only at the rapid steady-state rate (Fig. 2). It was further shown that the acceleration in ATP synthesis did not influence the affinity for ADP and P i (Fig. 4). These observations indicated that TF 0 F 1 contained specific ATP binding sites which, once occupied, allowed ATP synthesis to proceed at a high rate.
Little is known about the nucleotide binding sites on TF 0 F 1 since the nucleotide binding properties are investigated mainly in the disrupted TF 1 moiety and the isolated subunits. There is a general agreement that TF 0 F 1 as well as TF 1 can be isolated and purified nearly free of nucleotides (35); however, the results concerning adenine nucleotide binding to the TF 1 are confusing. Using circular dichroism, Ohta and co-workers (35) first reported that in the presence of Mg 2ϩ , TF 0 F 1 contained three ADP or alternatively three ATP sites stable to gel filtration and located on both ␣ and ␤ subunits. Yoshida and Allison (13), using [ 3 H]ADP, described only one ADP site stable to gel filtration which was located on the ␤ subunit. Using 3Ј-O-(4benzoyl)benzoyl ADP, Bar-Zvi et al. (42) demonstrated inhibition of the rate of ATP hydrolysis by covalent binding to TF 1 ; complete inactivation occurred upon binding of 2 mol of Bz-ADP/mol TF 1 , and the label was found to bind exclusively to the ␤ subunit. It has also been reported recently that there is a 95% inhibition of TF 1 ATPase activity in the presence of 2-N 3 [␤,␥-32 P]ATP and Mg 2ϩ (15). On comparison between the labeling on the TF 1 and ␣ or ␤ subunits it was proposed that the catalytic sites of F 1 resided exclusively on ␤ subunit, whereas noncatalytic sites were located mostly on ␣ subunit and that the adenine ring of ADP or ATP interacted with side chains contributed by ␤ subunits. Furthermore the literature (43)(44)(45) has analyzed the role of the enzyme-bound nucleotides in terms of hydrolytic activity. Any information about ATP binding sites on TF 0 F 1 and their role in ATP synthase activity is lacking.
The results presented in this paper are a first insight into the characterization of the ATP binding sites of the thermophilic F 0 F 1 -ATPase. TF 0 F 1 can be isolated and purified nearly free of bound nucleotides. When ADP was added to the isolated enzyme in the presence of Mg 2ϩ , a 1:1 TF 0 F 1 ⅐ADP complex was formed which was stable to gel filtration. Binding of ATP, after preincubation with ADP, revealed two additional sites for ATP which did not dissociate upon gel filtration. Analysis of the binding of ATP after ADP incubation revealed that one site saturated at 0.1 mM ATP, whereas the other saturated above 1 mM ATP (Fig. 5). Thus the main finding of these data is that TF 0 F 1 contained, in addition to one ADP binding site, two ATP binding sites that were stable to gel filtration.
The role of these ATP sites in regulating the ATP synthesis process was analyzed after reconstitution into liposomes (Fig.  6). The initial rate of the incorporated TF 0 F 1 ⅐ADP⅐ATP 1:1:1 complex was stimulated 2-3-fold over that of the TF 0 F 1 ⅐ADP 1:1 complex, and the initial rate of the TF 0 F 1 ⅐ADP⅐ATP 1:1:1.7 complex was further stimulated by 10 -20%.
To characterize further the ATP binding sites of TF 0 F 1 , we used the photoreactive analogue 2-N 3 [␣-32 P]AT(D)P which fulfilled conditions for successful photoaffinity labeling experiments. In particular it is a real substrate (i.e. it is hydrolyzed), and the affinities for the adenine nucleotides and their 2-azido analogues are nearly identical for F-type ATPases (46,47). When TF 0 F 1 was first incubated with 100 M ADP, followed by a second incubation with 100 M 2-N 3 [␣-32 P]ATP, we found 0.7 2-N 3 [␣-32 P]adenine nucleotides/enzyme. In the control experiment the azido compound was replaced by ATP, and two nucleotides, 1.2 ADP and 0.8 ATP, were tightly bound (Table I, fifth and sixth rows). This clearly indicated that 2-N 3 [␣-32 P]ATP was mainly incorporated in the ATP site. When coreconstituted into liposomes with bacteriorhodopsin, this covalently labeled enzyme led to fast monophasic behavior. Since fixing the analogue to the ATP site was shown to fix the regulatory function it could be concluded that the ATP sites were regulatory. We have also used 2-N 3 [␣-32 P]ATP to identify the amino acid sequence of the binding site. After photolabeling, tryptic digestion, and ion exchange HPLC, the ATP peak was collected and applied to a reverse phase HPLC. The amino acid sequence then gave the information that ␤Y364 was labeled. This site is in the conserved region which is commonly named noncatalytic or regulatory (48).
Finally our data provide some information about the ADP site in the TF 0 F 1 complex. The observation that after preincubation with ADP 2-N 3 [␣-32 P]ATP was found mainly on the ATP sites strongly suggested that the hydrolyzing 2-N 3 [␣-32 P]ATP did not chase the ADP from the ADP site, i.e. the ADP site, which is stable to gel filtration, is not part of the catalytic (hydrolytic) cycle. This was clearly confirmed by the identification of the amino acid labeled by 2-N 3 [␣-32 P]ADP which was ␤Y364. This can make a difference to the TF 1 since Yoshida and Allison (13) provided evidence that on the ADP site, stable to gel filtration, enzyme-bound ADP was converted to bound ATP, which would point toward a catalytic site. However, the experiments were performed in the presence of dimethyl sulfoxide, i.e. under extreme conditions. Thus an important feature of this work is that it demonstrates that the three nucleotides that bound tightly to TF 0 F 1 are noncatalytic, binding to one ADP and two ATP sites.
In addition to this, several interesting observations are derived from this study. (i) ATP bound to its sites did not hydrolyze, suggesting that these sites are different from the sites for ATP hydrolysis. (ii) In the absence of exogenous ADP, the ATP sites are not stable to an electrochemical H ϩ gradient. Preillumination of the proteoliposomes with the TF 0 F 1 ⅐ADP⅐ATP complex in the absence of exogenous ATP results in a biphasic kinetics of ATP synthesis, as opposed to a monophasic fast kinetic observed when ATP synthesis was started by illumination in the presence of exogenous ADP. A possible explanation could be that in the presence of an electrochemical proton gradient, tightly bound ATP is released and/or shifted to another site. (iii) It was also observed that first binding ADP followed by ATP binding is essential to gain a direct acceleration of ATP synthesis to the fast phase. TF 0 F 1 preparations treated with ATP without prior ADP incubation showed the same tightly bound nucleotide pattern (i.e. a 1:1:2 TF 0 F 1 ⅐ADP⅐ATP complex) but led to biphasic ATP synthesis. This can be interpreted assuming the existence of a first and second site both able to bind ADP as well as ATP. However to gain rapid ATP synthesis the nucleotides must each be bound to a specific site. (iv) The observation that the 1:1 TF 0 F 1 ⅐2-N 3 [␣-32 P]ADP complex in which 2-N 3 [␣-32 P]ADP was covalently bound, exhibited the same biphasic kinetic behavior as the TF 0 F 1 :ADP complex implies that the 2-3-fold acceleration in ATP synthesis cannot be related to the release of inhibitory ADP by newly synthesized ATP. (v) The ATP sites are true regulatory sites in the sense that they are able to regulate the rate of ATP synthesis.
In conclusion our studies allowed us to gain further insight into the function of nucleotide binding sites during ATP synthesis through TF 0 F 1 . Other than the various reports about nucleotide regulation of ATP hydrolysis of the disrupted enzyme, i.e. the F 1 part (47,49,50) and nucleotides that play a role in the activation (51), this is the first report on the regulation of the catalytic process, i.e. acceleration of ATP synthesis by enzyme-bound nucleotide for an F-type ATPase. The fact that the significant regulation of ATP synthesis by such sites has never been reported with F-type ATPases from other sources may be related to the fundamental difference between TF 0 F 1 and F 0 F 1 from other sources, which is that the former can be isolated and purified free of bound nucleotides. For instance, one can speculate that in the case of CF 0 F 1 from spinach chloroplast, which contains 1 ADP and 1 ATP tightly bound after isolation and purification (12,29), only the fast rate of ATP synthesis can be observed due to the fixed nucleotides.