Catalytic site occupancy during ATP hydrolysis by MF1-ATPase. Evidence for alternating high affinity sites during steady-state turnover.

The mechanism of ATP hydrolysis by the solubilized mitochondrial ATPase (MF1) has been studied under conditions where catalytic turnover occurs at one site, uni-site catalysis (obtained when enzyme is in excess of substrate), or at two sites, bi-site catalysis (obtained when substrate is in excess of enzyme). Pulse-chase experiments support the conclusion that the sites which participate in bi-site catalysis are the same as those which participate in uni-site catalysis. Upon addition of ATP in molar excess to MF1, label that was bound under uni-site conditions dissociates at a rate equal to the rate of bi-site catalysis. Similarly, when medium ATP is removed, label that was bound under bi-site conditions dissociates at a rate equal to the rate of uni-site catalysis. Evidence that a high affinity catalytic site equivalent to the one observed under uni-site conditions participates as an intermediate in bi-site catalysis includes the demonstration of full occupancy of a catalytically competent site during steady-state turnover at nanomolar concentrations of ATP. Improved measurements of the interaction of ADP at a high affinity catalytic site have lead to the revision of several of the rate constants that define uni-site catalysis. The rate constant for unpromoted dissociation of ADP is equal to that for Pi (4 X 10(-3) s-1). The rate of binding ADP at a high affinity chaseable site (Kd = 1 nM) is equal to the rate of binding ATP (4 X 10(6) M-1 s-1). The rate of catalysis obtained when substrate binding at one site promotes product release from an adjacent site (bi-site catalysis) is up to 100,000-fold faster than unpromoted product release (uni-site catalysis).

The mechanism of ATP hydrolysis by the solubilized mitochondrial ATPase (MF1) has been studied under conditions where catalytic turnover occurs at one site, uni-site catalysis (obtained when enzyme is in excess of substrate), or at two sites, bi-site catalysis (obtained when substrate is in excess of enzyme). Pulse-chase experiments support the conclusion that the sites which participate in bi-site catalysis are the same as those which participate in uni-site catalysis. Upon addition of ATP in molar excess to MFI, label that was bound under uni-site conditions dissociates at a rate equal to the rate of bi-site catalysis. Similarly, when medium ATP is removed, label that was bound under bi-site conditions dissociates at a rate equal to the rate of unisite catalysis. Evidence that a high affinity catalytic site equivalent to the one observed under uni-site conditions participates as an intermediate in bi-site catalysis includes the demonstration of full occupancy of a catalytically competent site during steady-state turnover at nanomolar concentrations of ATP.
Improved measurements of the interaction of ADP at a high affinity catalytic site have lead to the revision of several of the rate constants that define uni-site catalysis. The rate constant for unpromoted dissociation of ADP is equal to that for Pi (4 X s-'). The rate of binding ADP at a high affinity chaseable site (& = 1 nM) is equal to the rate of binding ATP (4 X lo6 M" s-'). The rate of catalysis obtained when substrate binding at one site promotes product release from an adjacent site (bi-site catalysis) is up to 100,000-fold faster than unpromoted product release (uni-site catalysis).
MF,' has a subunit structure of (~~@~y B c .
Catalytic sites are located on the &subunits or at interfaces between cy-and @subunits. The catalytic sites show negative cooperativity in binding substrate and strong positive catalytic cooperativity. There are a total of six adenine nucleotide-binding sites on MF,. Nucleotide bound at three of the sites exchanges rapidly with medium nucleotide during catalytic turnover. These sites are likely to have a catalytic function. The remaining three * This work was supported by Research Grant GM 23152 from the National Institutes of Health, United States Public Health Service. 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.
The abbreviations used are: MF,, soluble ATPase portion of beefheart mitochondrial ATP synthase; MF, [+y], MF, containing x mol of ANP at noncatalytic sites and y mol of ANP at catalytic sites/mol of enzyme; SHPMgA, SHPMg buffer containing in addition 1 mg of defatted bovine serum albumin/ml; Hepes, 4-(2-hydroxyethyl)-l-piperazineethanesulfonic acid. binding sites do not exchange readily with medium nucleotide. These sites are referred to as noncatalytic sites (for reviews see Boyer, 1987;Senior, 1988).
The fact that there are three exchangeable nucleotidebinding sites and three copies of the P-subunitlMF, have led a number of investigators to consider it likely that the enzyme has three functional catalytic sites. Evidence supporting this possibility comes from studies of the kinetic properties of the enzyme. In one such study (Grubmeyer et al., 1982;, it was concluded that turnover can occur at one site (Kd = 10"' M, uni-site catalysis), two sites (K, = 30 pM, bisite catalysis), or three sites ( K , = 150 pM, tri-site catalysis), depending on the substrate concentration.
Two laboratories have recently questioned the idea that the site responsible for uni-site catalysis is a normal site capable of multi-site catalysis at high substrate concentration (Bullough et al., 1987;Milgrom and Murataliev, 1987). Here, we have re-examined the properties of catalysis under uni-site and bi-site conditions and find that when medium ATP is removed, sites undergoing bi-site catalysis revert to uni-site rates. We also show that sites undergoing uni-site catalysis switch to bi-site rates when excess substrate is added. The results support the conclusion that the same sites participate in both activities. In the course of this work, we further characterized the interaction of ADP at a high affinity catalytic site. A preliminary report of some of this work has appeared (Cross et al., 1984(Cross et al., , 1986.

EXPERIMENTAL PROCEDURES
Preparation of MFl-MFl was isolated from beef heart submitochondrial particles and stored as an ammonium sulfate suspension (Knowles and Penefsky, 1972). Prior to each experiment, an aliquot of the suspension was centrifuged in a Beckman Airfuge. The supernatant was carefully removed, and the enzyme pellet was dissolved pH 7.8 or 8.0, 1 mM K'-Pi, and 1 mM MgSO, (SHPMg buffer) and at 4-6 & I in buffer containing 150 mM sucrose, 10 mM K'-Hepes, desalted by passage through a Sephadex centrifuge column (Penefsky, 1977) equilibrated in the same buffer. The MF1 concentration was adjusted to approximately 1 PM with SHPMg buffer, and the actual concentration was determined by a modified Lowry procedure (Peterson, 1977) as described (Nalin and Cross, 1982). MF, that is prepared, stored, and desalted as described above has two endogenous nucleotides bound at noncatalytic sites and 1.0-1.5 bound at catalytic sites (MF,[2,1], Kironde and Cross, 1986). Recognizing the importance of having enzyme with well defined nucleotide site occupancy available for both presteady-state kinetic measurements and nucleotide binding studies, we previously developed a procedure for preparing MF, with the three noncatalytic sites filled and the three catalytic sites empty (referred to as MF, [3,0]). The procedure employed pyrophosphate at one step to displace adenine nucleotide from catalytic sites (Kironde and Cross, 1986). Here we describe an improved procedure in which MgGTP is used in place of pyrophosphate.
Step 1: native MF, is incubated at approximately 6 PM in SHPMg buffer with 8 mM MgATP for 2 min. Unbound nucleotide is removed by passage through a centrifuge column.
Step 2: MF, is incubated at 3 p M with 4 mM MgGTP for 30 s, and the sample is passed through a centrifuge column equilibrated in SHPMg buffer with K+-Pi added to adjust the Pi concentration to 50 mM and the pH to 7.0.
Step 3: after 10 min of incubation, free nucleotide is removed by passage through a third centrifuge column also equilibrated with SHPMg buffer adjusted to 50 mM Pi. Following a second 10-min incubation at high phosphate concentration, unbound ligand is removed on a fourth centrifuge column equilibrated in normal SHPMg buffer.
The nucleotide content of MF, at each step in this procedure has been measured using [3H]GTP and a luciferin-luciferase assay (Lundin et al., 1976) of KOH-neutralized perchloric acid extracts pretreated with pyruvate kinase and phosphoenolpyruvate. In the first step, ADP produced by ATP cleavage fills the single unoccupied noncatalytic site. With passage through the first centrifuge column, MF, retains a total of 4.5 adenine nucleotides with three bound at noncatalytic sites and 1.5 bound at catalytic sites (MF,[3,1.5]). In the second step, adenine nucleotide bound at catalytic sites is chased during MgGTP cleavage. With passage through the second centrifuge column, MF, retains three adenine nucleotides at noncatalytic sites and one GDP at a catalytic site. In the third step, GDP, unlike ADP, is readily displaced from the catalytic site by incubating in the presence of high Pi concentration to give MF, [3,0].
MF,[2,0] was prepared as described above, except that the first step was omitted.
Ligand Binding Measurements-Incubations were initiated by mixing equal volumes of radiolabeled nucleotide with MF,. Enzymebound ligand was separated from free ligand on Sephadex centrifuge columns. Each incubation time reported includes 20 s beyond the time of initiation of the centrifugation step. This allows for the time required to separate protein and unbound ligand on the centrifuge column (Grubmeyer et al., 1982).
A number of experiments in this study were conducted either at very low MF, concentration or with a very large molar ratio of radiolabeled ligand to enzyme. The level of recovery of MF, and the resolution of bound from unbound ligand required in such experiments exceeded previously tested limits for the Sephadex centrifuge column technique. For example, to measure catalytic site occupancy during steady-state turnover it was necessary to use at least a 10-fold molar excess of substrate over enzyme. At the lowest ATP concentration tested, binding measurements were performed on samples containing 3 X IO-" M MF, (see Fig. 2). Preliminary tests with [3H] acetyl-MF, demonstrated that at all MF, concentrations used, approximately 95% of the MF, was recovered in centrifuge column effluents when the application volume was between 80 and 130 pl, and defatted bovine serum albumin was included at 1 mg/ml in reaction mixtures and in the SHPMg buffer (SHPMgA) used to equilibrate the columns (Cross and Nalin, 1982). At high ATP concentrations, the MF, concentration was limited to 1 nM in order to avoid substrate depletion during the time required to mix the reactants and complete the binding assays. Binding measurements at 10 pM ATP and 1 nM MF, thus required a minimal resolution of 1 part bound from 10,000 parts unbound. As previously reported, the nonspecific leakage of unbound ligand through a 1-ml column is approximately 1 part in 1,000 when the sample volume is 100 p1 (Penefsky, 1977). However, we find that resolution improves considerably as the sample volume is decreased to 50 pl, where nonspecific leakage is less than 1 part in 200,000. Further reduction of the sample volume causes protein recovery to become variable and to drop below 85%.
Column effluents were collected directly in plastic liquid-scintillation vials, and MF1-bound ligand was determined by scintillation counting. In order to obtain sufficient label for counting at very low MF, concentrations (10"' to lo-" M), three 100-p1 aliquots of each sample were applied to three separate 1-ml columns, and the effluents were collected in a single vial. Alternatively, a larger aliquot was applied to a 3-or 5-ml column. Corrections were made for recovery of MF, as determined by control experiments with [3H]acetyl-MF1. Additional corrections were made for the small amount of label in column effluents of controls lacking MF,.
,'P; Assay-"Pi was determined by the triethylamine/ammonium molybdate precipitation method of Sugino and Miyoshi (1964) as described in (Grubmeyer and Penefsky, 1981) and further modified here. Time consuming transfers were eliminated and recovery was improved by precipitating 32Pi samples directly in the plastic minivials used for liquid scintillation counting. Molybdate reagent (1.0 ml, 0 "C) containing 1% triethylamine, 1% bromine water, and 3% ammonium molybdate in 0.067 N HCl is added to an equal volume of acid-quenched sample containing 0.5 M perchloric acid, 1.0 mM Na'-Pi, and 5 p~ ATP. The vials are stored 10 min on ice to allow complete precipitation of the phosphomolybdate complex. Samples are centrifuged for 10 min at 4000 X g in a Beckman J-6 centrifuge (0 "C) using a 5-4.2 rotor with bucket adapters for 14-mm tubes.
Supernatants are decanted, and the pellets containing '*Pi are washed once at 0 "C by vortexing with 1.0 ml of a freshly prepared mixture containing equal volumes of molybdate reagent and 0.5 M perchloric acid, 50 p~ Pi. Following a second centrifugation, the supernatants are carefully decanted, and the pellets are dissolved in 150 pl of 1.0 N NaOH. Samples are then acidified by addition of 150 pl of 1.5 N acetic acid, and 32P is determined by liquid scintillation counting as described above. Corrections were made for the recovery of "Pi standards which ranged from 94 to 99%.
Calculations-The solid line given in Fig. 2 predicts the total number of moles of 32P bound either as [Y-~'P]ATP or 32Pi at catalytic sites/mole of MF1 during steady-state hydrolysis. The line is calculated using the following equation: , is the fraction of total enzyme with bound "P, k, is the rate of 32Pi release (1.2 X s-'; Grubmeyer et al., 1982), and kl is the rate constant for ATP binding (4 X lo6 M" s-', Fig. 3).

Dissociation of bound [Y-~'P]ATP does not affect the calculation
since it occurs at a rate three orders of magnitude slower than '*Pi dissociation. Rebinding of 32Pi is prevented by use of medium Pi as an isotope trap. Finally, since the rate of substrate-promoted product release during bi-site catalysis  is more than an order of magnitude faster than the turnover rate at the highest ATP concentration tested in Fig. 2, the enzyme will spend a negligible fraction of its time in a form with two 32P bound.
Other Methods-Adenine nucleotide concentrations were determined from the absorbances of solutions at 259 nm using an extinction coefficient of 15,400 cm" M-'.
["]ADP was freed of contaminating ATP by treatment with hexokinase and glucose as described (Grubmeyer et al., 1982). Procedures." Times were selected to allow one or more turnovers without significant depletion of substrate. Where indicated by the arrow, 4 mM MgATP was added, and the nonchaseable "P (A) was measured 45 and 60 s later. catalytic site. Addtion of a large excess of unlabeled ATP results in rapid dissociation of greater than 95% of the bound 32P (Fig. 1, triangles), and identical results are obtained to those shown in Fig. 1 when MF1[3,0] is used.

Catalytic Site Occupancy as a Function of ATP
Experiments similar to the one shown in Fig. 1 were repeated over a wide range of [y3'P]ATP concentrations (3 x 10"" to M, Fig. 2, circles). The results demonstrate full occupancy of one site during steady-state hydrolysis at substrate concentrations down to 3 nM, well below published K,,, values.
A model based on the binding change mechanism for Fl-ATPases (Boyer, 1979) was used to derive an equation for predicting catalytic site occupancy as a function of ATP concentration (see "Experimental Procedures"). The solid line shown in Fig. 2 was generated by substituting values for the ATP concentration, and the rate constants for ATP binding (Fig. 3) and P, release (Grubmeyer et al., 1982). The model predicts that at ATP concentrations above lo-' M, one catalytic site/MF, will be fully occupied by 32P-labeled substrate and product. At these ATP levels, turnover results exclusively from bi-site catalysis, where substrate binding at one catalytic Aliquots were removed at various times and quenched by addition of an equal volume of 1 M perchloric acid containing 1 mM Pi and 10 p~ ATP. The amount of 32P, formed was determined as described under "Experimental Procedures." Initial reaction velocities are plotted against ATP concentration. site promotes rapid release of product from an adjacent site. Since the highest concentration of ATP reported in Fig. 2 is still well below the K , for bi-site catalysis, substrate binding will be rate limiting, and the enzyme will spend only a small fraction of its time in a form with two sites occupied. Hence, the stoichiometry is not expected to raise above 1.0.
With ATP concentrations at or below IO-' M ATP, the model predicts (Fig. 2, solid line) that the rate of [y"P]ATP binding at a second site will become sufficiently slow to allow time for spontaneous dissociation of 32Pj by uni-site catalysis. Dissociation of ATP does not contribute to the decrease in level of bound 32P since interconversion of bound substrate and product is rapid, and the rate of dissociation of ATP is three orders of magnitude slower than that of Pi (Grubmeyer et al., 1982). The results not only demonstrate the existence of a high affinity site during steady-state bi-site catalysis, but the excellent agreement of the data with the calculated plot also provides independent confirmation of the rate constants for ATP binding and Pi release.
It was previously reported in single turnover experiments that the rate of ADP dissociation is 10-fold slower than Pi.
This leads to the prediction that with decreasing concentrations of [3H]ATP, the drop in level of 3H-label bound/MF1 should be shifted one log unit to the left of that shown in Fig.   2 for 32P. However, this prediction was not verified. The ATP concentration dependence of catalytic site occupancy during steady-state hydrolysis of [3H]ATP (Fig. 2, triangles) was identical to that detected using [y"P]ATP (Fig. 2, circles). The results suggest that the rates of ADP and P, dissociation are equal. This is demonstrated directly by experiments reported in a later section.

The Rate of ATP Binding during Bi-site Catalysis-At ATP
concentrations above lo-' M ( Fig. 2) but below M , catalytic turnover will be due exclusively to bisite catalysis. Since the rate-limiting step is ATP binding, the velocity should show a linear dependence on substrate concentration. A rate constant for ATP binding at a second site was previously determined over a narrow range of ATP concentrations (5-60 nM, Cross et al., 1982). Experiments reported in Fig. 3 extend these data to show a linear dependence of velocity over a 104-fold range in substrate concentration. A second-order rate constant of 4 X lo6 "' s" is calculated from the slope of the plot in Fig. 3. This value is in good agreement with the rate constant reported previously and is very similar to the rate of ATP binding to the first site on MF, (6 X lo6 M" s-', Grubmeyer et al., 1982).

Replacement of Tightly Bound "P-Labeled Ligand by Medium Substrate during Steady-state Bi-site Catalysis-An al-
ternating site mechanism predicts that under steady-state conditions at substrate concentrations well below the K,, for bi-site catalysis, bound 32P-labeled product will dissociate at a rate equal to the rate of ATP binding. Results presented in Fig. 4 verify this prediction. When enzyme having one fully occupied site during turnover at 10 nM [y"'P]ATP is diluted 11-fold into medium containing the same concentration of unlabeled ATP, bound "P chases rapidly (Fig. 4A, open circles). In a control, where enzyme is diluted into buffer containing 32P-labeled substrate at the same specific activity, bound 32P remains constant at 1 mol/mol as expected (A, triangles), When another sample is diluted 33-fold into buffer containing nonradioactive ATP at 3-8 nM, 85% of the 32P label chases in a pseudo-first-order fashion (B). In each case, dividing the observed rate of loss of bound label by the ATP concentration gives a value equal to the rate constant for ATP binding (5-6 X lo6 M" s-l). The results graphically demonstrate that release of product from a very high affinity site on MF, is promoted by the binding of substrate at an adjacent site.
Promoted Release of Product from a Site Loaded under Unisite Conditions-When MF, is mixed with a substoichiometric amount of [yr2P]ATP, rapid binding occurs at a single high affinity site (Kd = 10"' M, Grubmeyer et aL, 1982). Binding is followed by a rapid equilibration between bound substrate and products and a slow rate-limiting release of products (unisite catalysis). The experiments reported in Fig. 5 were performed to determine whether the high affinity site loaded under uni-site conditions is capable of participating in normal bi-site catalysis. The results show that when 20 nM [y-"P] ATP is mixed with 100 nM MF, and the resulting complex is diluted 400-fold with buffer containing 3-8 nM nonradioactive ATP to allow bi-site catalysis, label is chased in a pseudofirst-order manner. Again, dividing each observed rate by the ATP concentration yields a value equal to the rate constant for ATP binding (4-5 X lo6 M" s-'). Thus, release of label initially loaded under uni-site conditions is promoted by medium substrate at the same rate as bi-site catalysis.
Dissociation Constants for Product Release during Uni-site Catalysis and Unpromoted Release from Steady-state Bi-siteloaded Enzyme-As mentioned above, our previous attempt to measure the rate of ADP dissociation during uni-site catalysis underestimated its true value. When the MF,-['HH] ANP complex, formed by incubating ["HIATP with an excess of MF,, is diluted to 14 nM MF,, a dissociation rate of 1.2 X s-' is measured (Grubmeyer et al., 1982). However, if the reaction mixture is further diluted to 0.2 nM MFI, a faster dissociation rate, equal to 1.4 X lo-? s-I, is observed (Fig. 6A,   triangles). This result indicates that the slower rate obtained at higher enzyme concentration may reflect some other proc-ess which occurs during reversible binding of ADP at a site having a much higher affinity than previously suspected. This explanation is consistent with the demonstration in a later section of ADP binding at a catalytic site having a Kd = 1 nM.
Results presented in Fig. 6A also show that under the more dilute conditions, i3H]ADP dissociates at the same rate as ["PIP, (circles). Dissociation of ATP does not contribute to the observed rates since it is three orders of magnitude slower (Grubmeyer et al., 1982).
If bi-site catalysis occurs by substrate-promoted product release from a tight site characteristic of uni-site catalysis, then removal of medium substrate should result in product dissociation rates returning to those observed for uni-site catalysis. This prediction is verified by the experiments reported in Fig. 6B. MF1 was loaded during bi-site catalysis with [3H]ATP or [y3'P]ATP. Following removal of medium substrate and dilution to 0.09 nM MF,, the rates of dissociation of ADP and Pi are measured as the rates of dissociation of 3H (triangles) and 32P (circles). These rates are found to be equal at 2.2 x s-'. A similar slowing of the rate of Pi dissociation upon removal of medium ATP was observed in studies with the chloroplast enzyme, CF, (Wu and Boyer, 1986).
Since bound substrate and products are in rapid equilibrium (Grubmeyer et al., 1982), the fraction of label present as product must be taken into account in calculating the intrinsic dissociation rate constants for product release. Results presented in Table I show that the ratio of product to substrate for enzyme loaded under bi-site conditions is slightly higher than that for enzyme loaded under uni-site conditions (1.4 compared with 0.53). When the appropriate corrections are made, the dissociation rate constants for unpromoted ADP and Pi release from both uni-site-and bi-site-loaded enzyme are identical at 4 x s-'.
Tight Binding of ADP at Catalytic Sites-The need to dilute MFl to subnanomolar concentrations in order to accurately measure the rate of dissociation of ADP suggests that ADP binding at a single catalytic site on MF, is much tighter than previously reported. Scatchard analysis of direct binding measurements using [3H]ADP and MF, [3,0] shows that 0.7 mol of ADP bind/mol of MFI at a rapidly chaseable site having a Kd of 1 nM (Fig. 7 B ) and 0.4 mol of ADP bind/mol of MF, at a site resistant to rapid chase having a Kd of 7 nM ( C ) . Release of ligand from the chaseable site is promoted by nanomolar concentrations of ATP at the same ATP-concentration-dependent rate as bi-site catalysis (5 X lo6 M" s-' , data not shown). Dissociation rates for the chaseable and nonchaseable sites in the absence of medium nucleotide were  Fig. 6 for unisite and bi-site catalysis. At the times indicated, unbound 32P was removed on a centrifuge column. Column effluents were collected directly in perchloric acid containing carrier Pi and ATP. An aliquot of the quenched effluent was counted for total 32P bound and another was assayed for ["P]Pi as described under "Experimental Procedures." Bound ATP was calculated as the difference between total bound 32P and bound 13*P]Pi.

Time after
Bound determined to be 2 x s" and 6 x IOb4 s-I, respectively (data not shown). The chaseable ADP site appears to be the same site as that which binds ATP tightly under conditions for uni-site or bisite catalysis. Medium ATP promotes ADP release from the sites at identical rates, and the rate of unpromoted ADP dissociation from each site is the same. The nonchaseable site shown in Fig. 7C may be the same fractional nonchaseable catalytic site as that reported previously by Kironde and Cross (1986). Association rate constants for the high affinity ADP sites can be calculated from their measured & values and dissociation rate constants to be 2 X IO6 M-l s" and lo5 M" s-' for the chaseable and nonchaseable sites, respectively. Attempts to measure the association rates directly using native MF, or MF1[3,0] were unsuccessful. Second-order rate constants varied with ADP concentration, decreasing with increasing ADP. The results suggest that ADP binding is a two-step process with initial binding at a loose site followed by a conformational change that increases the stability of bound ligand and allows its retention through a centrifuge column. Attempts to measure the first-order process indicate that it proceeds at less than 10" s-l. Experiments were performed to determine the cause of heterogeneity in the ADP-binding site. It was found not to be dependent on the initial nucleotide content, since MF1[3,0], MFl[2,1], and MFl[2,0] all gave similar Scatchard plots. The results with MF, [3,0] suggest that a noncatalytic site is not involved. In addition, the K d for binding at the vacant noncatalytic site on MF1[2,1] is 50-fold higher than that for the chaseable ADP site characterized in Fig. 7 and the association rate is a 100-fold slower (Kironde and Cross, 1987). The heterogeneity was also found not to result from a timedependent denaturation of MF, at very dilute concentrations.
MgSO, at 20 mM completely eliminated the nonchaseable fraction and only slightly decreased the chaseable fraction. Conditions reported by Senior (1981) to remove heterogeneity of M e content of MFI, i.e. catalytic turnover with ATP in excess to Mg2+, resulted in an increase in chaseable ADP (to 0.8 mol/mol) with a near complete elimination of the nonchaseable fraction.

DISCUSSION
The data presented support the conclusion that the catalytic sites of MF, which participate in bi-site catalysis are the same sites as those which participate in uni-site catalysis. The mode of catalysis exhibited by these sites appears to be strictly dependent on the availability of medium ATP to accelerate product release. This is supported by the following observations: 1) the existence of a high affinity site during steady-state bi-site catalysis is demonstrated by the full-site occupancy attained at nanomolar concentrations of ATP ( Figs. 1 and 2). These concentrations are far below the K,,, for bi-site catalysis (30 FM, Cross et al., 1982). 2) When enzyme undergoing steady-state bi-site catalysis at nanomolar ATP concentrations is subjected to pulse-chase measurements, bound label dissociates at a rate predicted by the rate of ATP binding (Fig. 4). This shows that the high affinity site detected under these conditions (Fig. 2) is kinetically competent for bi-site catalysis. 3) Upon addition of medium ATP in molar excess to MF,, uni-site-loaded label dissociates at a rate predicted by the rate of ATP binding (Fig. 5). This demonstrates that sites loaded under conditions for uni-site catalysis are capable of switching to bi-site catalysis when medium substrate is available. 4) The intrinsic rate constants for unpromoted product release are identical for both uni-site ADP that had been treated to remove ATP as described under "Experimental Procedures." Concentrations of ADP varied from 0.4 to 50 nM. A, after 1 h to allow equilibrium binding, total 13H] ADP bound was measured by centrifugecolumn assays as described under "Esperimental Procedures." B, the amount of ADP bound at chaseable sites was calculated as the difference between the total 3H bound ( p a n e l A ) and the nonchaseable 3H bound ( p a n e l C ) . C, prior to passage through a centrifuge column, aliquota of the reaction mixture were incubated with 4 mM MgATP for 40 s. The 3H-label retained by the enzyme represents nonchaseable bound ADP. and bi-site loaded enzyme (Fig. 6, Table I, and text). This again indicates that the same sites participate in both activities.
Two laboratories have disputed the idea that the catalytic site characterized under uni-site conditions is capable of participating in multi-site catalysis when excess ATP is added (Bullough et al., 1987;Milgrom and Murataliev, 1987). Under the conditions used by these investigators, uni-site-loaded label did not chase at a sufficiently rapid rate or to a sufficient extent. Recently, Penefsky (1988) has repeated some of these experiments. In contrast, he finds that 80% of the uni-siteloaded label chases in a kinetically competent manner. In the experiments reported here, we have used enzyme and substrate concentrations that are far below those used in the other studies. We find that at least 85% of uni-site-loaded label chases in a kinetically competent manner (Fig. 5 ) . Penefsky (1988) has offered several explanations for the failure to observe rapid chase of uni-site-loaded label under certain conditions. To these, we would add the possibility that differences in the Mg2' content of MF, purified by the different laboratories might contribute to the differences noted. This suggestion is based on the effect that M e has on the ability of medium ATP to chase tightly bound ADP (see "Results").
Several previous studies have suggested that uni-, bi-, and tri-site catalysis all occur in the micromolar concentration range for ATP. Hatefi's laboratory reports three K , values (Gresser et ai., 1982). From the data presented in Fig. 2, it would appear that uni-site catalysis will occur only at picomolar concentrations of ATP when substrate is added in excess of enzyme. Thus, all observations made by Wong et al. (1984) and Gresser et al. (1982) likely resulted from bi-site and tri-site catalysis. One of the three K,,, values measured by Wong et al. (1984) might reflect a transition between low and high activity forms of MFI obtained with the filling of the single vacant noncatalytic site that is present on native enzyme (Kironde and Cross, 1986). The measured by Gresser et al. (1982) for the oxygen exchange reaction might reflect a kinetic partitioning between resynthesis of ATP at the catalytic site, which is required for exchange, and substratepromoted product release, which terminates the exchange.
A second contribution of the current studies is the improved measurements of the interaction of ADP at a high affinity catalytic site. Scheme 1 summarizes the rate constants for In previous studies (Grubmeyer et al., 1982), MF, was diluted to 14 nM. Unbeknownst at the time, this concentration was well above the K d for binding at the high affinity chaseable site. Thus, under these conditions, a small amount of medium nucleotide would have been in equilibrium with bound ligand. The slow rate of loss of bound label measured under these conditions (1.2 x s-', Grubmeyer et al., 1982) apparently reflects some process other than the rate of dissociation of ADP from the catalytic site. One such process might be a slow dissociation of endogenous nucleotide which would reduce the specific activity of medium [3H]ADP. This explanation is consistent with the rate of dissociation of ADP (4 X s-') from a noncatalytic site having a K d of 50 nM (Kironde and Cross, 1987).
Measurement of the rate of ADP binding is complicated by the fact that it appears to be a two-step process (see "Results"). Instead, equilibrium binding of ADP was measured, yielding a Kd of 1 nM at a chaseable catalytic site (K4 in Scheme 1). This is more than 100-fold lower than a value measured previously (Hilborn and Hammes, 1973). An association rate constant can be calculated using the measured Kd and the dissociation rate constant ( K , and k,, Scheme 1).
This gives a value of 4 X lo6 M" s" for (Scheme 1). This is essentially the same rate as that for binding ATP (kl, Scheme 1). The difference in K d values for ATP and ADP binding (l/Kl and K4, Scheme 1) is due to a 1000-fold slower rate of dissociation of ATP from the catalytic site ( k -, uersus kq, Scheme 1).
The binding of Pi at a catalytic site containing bound ADP is not measurable with MF1 in aqueous solution, However, using the relationship KT = Kl. K 2 . K3. K4 where KT is the overall equilibrium constant for ATP hydrolysis (4 X lo5) and assuming that Kl is not larger than 1OI2 M", a value of 8 X 10' M can be calculated for K3. From I C 2 and k3, a value of 5 X M-' s-l is calculated for the rate of Pi binding at a site containing bound ADP ( L 3 , Scheme 1).
During ATP synthesis, ADP and Pi must bind readily at the same site. Overall, the two substrate-binding steps would be thermodynamically favored if they were linked (K, -K4 = 8 X M). This may occur on the energized membrane but not with de-energized membrane or soluble MFl. Several studies suggest that energization of the membrane is necessary to bind both ADP and Pi at the same site (Rosing et al., 1977;Shoshan and Strotmann, 1980). It is also possible that ATP must be bound at an adjacent catalytic site for rapid binding of both ADP and Pi (Nalin and Cross, 1982).
In light of the revised value for the rate-limiting step in uni-site catalysis, the rate enhancement obtained when substrate binding promotes product release is 100,000 rather than 1 million as reported earlier (Grubmeyer et al., 1982). The revised value is closer to the values obtained for Escherichia coli BF, (Wise et al., 1984) and from oxygen exchange measurements (O'Neal and Boyer, 1986). It should also be noted that since the dissociation rate constants for Pi and ADP are now known to be equal, product release may not be ordered as shown in Scheme 1.
Final comment should be made regarding the question of whether two or three catalytic sites on MF1 participate during multisite catalysis at saturating substrate concentration. The simultaneous participation of three sites might normally be assumed for an enzyme having three copies of the catalytic subunit. However, it has been suggested that the interaction of the single-copy subunits with one ap pair might render one catalytic site nonequivalent or nonfunctional (Amzel et al., 1982). Data presented here strongly supports the presence of at least two interacting sites. Over a wide range of ATP concentrations, the rate of product release is equal to the rate of ATP binding at an adjacent site (Figs. 3-5). This feature of catalysis was detected earlier in oxygen exchange studies (Kayalar et al., 1977;Choate et al., 1979) and in measurements with TNP-ATP (Grubmeyer and Penefsky, 1981). Triphasic kinetic behavior observed in the micromolar concentration range for ATP was originally interpreted as support for the participation of three catalytic sites (Gresser et al., 1982;Wong et aL, 1984). As noted above, this evidence may now warrant reinterpretation. This would leave the kinetic properties which we have referred to as uni-, bi-, and tri-site catalysis as the main support for participation of three sites . Evidence presented here confirms turnover of a single site at picomolar concentrations of ATP (Kd = M, Grubmeyer et al., 1982) and turnover of two sites at nanomolar to micromolar concentrations ( K , = 30 pM, Cross et al., 1982). A second K , observed at 150 p~ ATP  could reflect half-maximal turnover at three sites. However, nucleotide binding at an exchangeable site, perhaps a "nonequivalent" catalytic site which functions as a regulatory site, might be responsible for the 2-fold rate enhancement obtained in going from bi-site to tri-site concentrations of ATP. Although we favor the view that three sites can function simultaneously (Cross, 1981(Cross, , 1984(Cross, , 1988, this alternative remains a viable possibility.