Effect of the ε-Subunit on Nucleotide Binding to Escherichia coli F1-ATPase Catalytic Sites*

The influence of the ε-subunit on the nucleotide binding affinities of the three catalytic sites ofEscherichia coli F1-ATPase was investigated, using a genetically engineered Trp probe in the adenine-binding subdomain (β-Trp-331). The interaction between ε and F1was not affected by the mutation. K d for binding of ε to βY331W mutant F1 was ∼1 nm, and ε inhibited ATPase activity by 90%. The only nucleotide binding affinities that showed significant differences in the ε-depleted and ε-replete forms of the enzyme were those for MgATP and MgADP at the high-affinity catalytic site 1.K d1(MgATP) andK d1(MgADP) were an order of magnitude higher in the absence of ε than in its presence. In contrast, the binding affinities for MgATP and MgADP at sites 2 and 3 were similar in the ε-depleted and ε-replete enzymes, as were the affinities at all three sites for free ATP and ADP. Comparison of MgATP binding and hydrolysis parameters showed that in the presence as well as the absence of ε, K m equalsK d 3. Thus, in both cases, all three catalytic binding sites have to be occupied to obtain rapid (V max) MgATP hydrolysis rates.

F 1 is the catalytic sector of F 1 F 0 -ATP synthase, the enzyme responsible for ATP synthesis in the last step of oxidative phosphorylation. F 1 may be isolated in soluble form and is an active ATPase ("F 1 -ATPase") with a subunit stoichiometry of ␣ 3 ␤ 3 ␥␦⑀. The three catalytic nucleotide-binding sites are located on the three ␤-subunits (1)(2)(3). For Escherichia coli F 1 , the minimal subunit composition able to achieve significant ATP hydrolysis rates is ␣ 3 ␤ 3 ␥. The small subunits ␦ and ⑀ are required for functionally competent binding of F 1 to F 0 (4 -7).
The ⑀-subunit is part of the central stalk connecting F 1 to F 0 , the membrane sector of the ATP synthase complex. ⑀ is tightly associated with the ␥-subunit (8,9). It appears to interact also with (one of) the c subunits of F 0 , and it can be cross-linked to the C-terminal domains of ␣ and ␤ (10 -13). High-resolution structural models of isolated ⑀ have been obtained based on NMR spectroscopy as well as x-ray crystallography (14 -16).
In isolated F 1 , the ⑀-subunit acts as an inhibitor. F 1 containing ⑀ has ϳ10% of the ATPase activity of the ⑀-depleted form of the enzyme (5,8,17). How this inhibition is brought about is an interesting question, especially considering that ⑀ is at least 35 Å away from the phosphate-binding pocket of the catalytic site.
To gain more insight into the effects of ⑀ on the function of F 1 , we determined in this study the nucleotide binding affinities of the three catalytic sites in the absence and presence of the ⑀-subunit. The enzyme we used for this purpose was E. coli ␤Y331W mutant F 1 . This enzyme, which shows normal functional properties, has a Trp genetically engineered into the adenine-binding domain of the catalytic site. The fluorescence response upon nucleotide binding makes this residue, ␤-Trp-331, a very sensitive probe for the occupancy of the catalytic site and allows the determination of thermodynamic and kinetic ligand binding parameters (18 -20).

EXPERIMENTAL PROCEDURES
Protein Preparation and Analysis-Wild-type F 1 was from strain SWM1 (21), and ␤Y331W mutant F 1 was from strain SWM4 (18). Preparation of F 1 was as described (22). ␤Y331W mutant F 1 was depleted of the ⑀-subunit by immunoaffinity chromatography (23). Western blots (24) confirmed that the preparation was ϳ99% depleted of ⑀. The ⑀-subunit was prepared as described (25). The concentration of F 1 was determined using the Bio-Rad protein assay (26) with bovine serum albumin as the standard. The concentration of ⑀ was measured by either the Bio-Rad protein assay or the method of Lowry et al. (27), both of which gave similar results.
ATPase Activity Assay-The dependence of ATPase activity on the concentration of ⑀ was determined after preincubation of F 1 (40 nM) with varied concentrations of ⑀ in 50 mM Tris/H 2 SO 4 , pH 8.0, for a minimum of 15 min at room temperature. The ATPase assay was started by adding 0.1 volume of a MgATP solution containing 50 mM Tris/H 2 SO 4 , 40 mM MgSO 4 , and 100 mM NaATP, pH 8.0. After a 1-min incubation at room temperature, the reaction was stopped by addition of SDS (5% final concentration). Released P i was determined colorimetrically (28).
The dependence of ATPase activity on MgATP concentration (see Fig.  1) was measured in 50 mM Tris/H 2 SO 4 , pH 8.0, at room temperature. F 1 concentrations were 30 nM in the presence of ⑀ and 2.5 nM in its absence. The concentration of ⑀, when present, was 120 nM. The reaction was started by simultaneous addition of NaATP and MgSO 4 in a ratio of 2.5:1. Samples were withdrawn at 10-or 20-s intervals for a total time of 1 or 2 min. Released P i was determined colorimetrically (29). Hydrolytic activities were determined from initial linear rates. MgATP concentrations were determined as described (30).
Fluorescence Measurements-All fluorescence measurements were performed in a buffer containing 50 mM Tris/H 2 SO 4 , pH 8.0, at room temperature. The spectrofluorometers used were Spex Fluorolog 2, Aminco-Bowman Series 2, and SLM 4800. The excitation wavelength was 295 nm. F 1 concentration was 30 -50 nM. For MgATP titration, two different conditions were used. To make the results directly comparable to those obtained in the hydrolysis assay, NaATP and MgSO 4 were added in a constant ratio of 2.5:1 to a solution of protein in buffer (see Fig. 3). Alternatively, the cuvette contained protein in buffer plus 2.5 mM MgSO 4 , and NaATP was added (see Fig. 5). In both cases, maximally two data points were acquired in a single experiment to avoid interference by the hydrolysis product MgADP. For MgADP titration, the buffer contained 2.5 mM MgSO 4 , and NaADP was added. For titration with free (uncomplexed) ATP or ADP, buffer plus 0.5 mM EDTA was used, and NaATP or NaADP was added. Background signals were subtracted, and inner filter and volume effects were corrected for by performing parallel titrations with wild-type F 1 . Nucleotide binding stoichiometries were determined from the decrease in ␤-Trp-331 fluorescence at 360 nm (18,19). Nucleotide binding parameters were calculated by fitting theoretical binding curves to the measured data points. The binding models are given below or in the figure legends; for the fitting algorithms, see Ref. 31.

RESULTS
Interaction of ⑀-Depleted ␤Y331W Mutant F 1 with Isolated ⑀-Subunit-Inhibition of ␤Y331W mutant F 1 -ATPase activity by ⑀ was used as a signal to determine the affinity between the two proteins. ⑀-Depleted ␤Y331W F 1 showed specific activities of 80 -100 units/mg at 30°C and 30 -40 units/mg at 23°C. Incubation of the ⑀-depleted enzyme with a 10 -20-fold concentration of ⑀ reduced activity by Ն90%, to 5-7 units/mg at 30°C and 3-4 units/mg at 23°C. Titration of ⑀-depleted F 1 with ⑀ gave a curve indicating a K d of ϳ1 nM (data not shown). Based on these results, we calculated that for the experiments described below, incubation of 30 nM ⑀-depleted F 1 with 120 nM ⑀ would give an enzyme that was ϳ99% ⑀-replete.
Substrate Concentration Dependence of ATPase Activity of ␤Y331W Mutant F 1 in Absence and Presence of ⑀-The ATPase activities of ⑀-depleted and ⑀-replete ␤Y331W F 1 as a function of MgATP concentration are shown in Fig. 1. Ignoring the ϳ10-fold difference in V max , the curves look remarkably similar. In both cases, the substrate concentration dependence followed simple Michaelis-Menten kinetics, with a single K m value of 27 M in the absence of ⑀ and 21 M in its presence.
Fluorescence Properties of ␤Y331W Mutant F 1 in Absence and Presence of ⑀-⑀-Depleted ␤Y331W F 1 has a Trp fluorescence maximum at 335 nm, and the ⑀-replete enzyme has one at 339 nm; both spectra are red-shifted as compared with wild-type F 1 ( Fig. 2A). ⑀ itself contains no Trp and has no significant Trp fluorescence ( Fig. 2A). As described previously (18), saturation with nucleotide virtually completely quenches the fluorescence of the three introduced ␤-Trp-331 residues in ␤Y331W F 1 prepared by standard procedures. The same result was obtained here with both ⑀-depleted and ⑀-replete ␤Y331W F 1 . After saturation with nucleotide, the spectra of both enzyme forms were indistinguishable from the one for wild-type F 1 shown in Fig.  2A, indicating that the spectral differences between ⑀-depleted and ⑀-replete ␤Y331W F 1 are due to an effect of ⑀ on the fluorescence of the introduced ␤-Trp-331 residues. 1 Fig. 2B shows the Trp fluorescence spectra of ␤-Trp-331 in the absence and presence of ⑀. In the absence and presence of ⑀, the maxima are at 345 and 350 nm, respectively. Thus, in an empty catalytic site, the environment of ␤-Trp-331 is highly polar, especially in the ⑀-containing enzyme. Analysis of spectra generated in an MgADP binding experiment with both types of enzyme showed no significant shift of the respective ␤-Trp-331 spectrum during the titration (data not shown). Thus, the influence of ⑀ on ␤-Trp-331 fluorescence is similar in each of the three catalytic sites.
Catalytic Site Occupancy during Steady-state Hydrolysis in Absence and Presence of ⑀-Subunit-Catalytic site occupancy in ⑀-depleted and ⑀-replete ␤Y331W F 1 is plotted in Fig. 3 as a function of substrate MgATP concentration. The conditions in this experiment were the same as those under which the hydrolysis data in Fig. 1 were obtained, i.e. ATP and Mg 2ϩ were present in a constant ratio of 2.5:1. Under these conditions, maximal hydrolysis rates are reached (32). In both cases, a binding model assuming three sites of different affinity gave a good fit, with K d1 ϭ 0.02 M, K d2 ϭ 0.7 M, and K d3 ϭ 21 M for ⑀-depleted ␤Y331W F 1 and K d1 Ͻ 0.01 M, K d2 ϭ 0.4 M, and K d3 ϭ 21 M for the ⑀-replete enzyme. For both enzyme forms, 1 In Ref. 8, quenching (by 16%) of the Trp fluorescence of the ␥-subunit by ⑀ was reported. The apparent absence of influence of ⑀ on the fluorescence of wild-type F 1 suggested by the data reported here indicates that either (a) the Trp residue(s) in ␥ that responds to ⑀ does not contribute significantly to the fluorescence of F 1 or (b) once ␥ is incorporated into F 1 , its fluorescence is no longer effected by ⑀. K d3 correlates well with K m (MgATP), demonstrating that all three catalytic sites have to be occupied to obtain rapid (V max ) hydrolysis rates. Fig. 4 (symbols) shows relative specific ATPase activity plotted as a function of the fraction of catalytic sites occupied, using the data from Figs. 1 and 3. Of course, the occupancy data on the abscissa are the average over all enzyme molecules. The lines are theoretical curves assuming that only enzyme molecules with all three catalytic sites filled achieve measurable hydrolysis rates (i.e. the catalysis rate of enzyme molecules with only one or two sites filled is zero). These theoretical curves clearly provide very good fits to the actual behavior of both ⑀-depleted and ⑀-replete enzymes. A fit of theoretical curves to the measured data points assuming a non-zero activity for enzyme molecules with two filled catalytic sites resulted in nearly identical curves, with a "bi-site" activity of 0% of V max in the absence of ⑀ and 1.2% in its presence.
Effect of ⑀-Subunit on Nucleotide Binding Affinities-As described above, ⑀ appears to have little effect on the MgATP affinities of catalytic sites 2 and 3. Under the conditions used in Fig. 3 (a constant ATP/Mg 2ϩ ratio of 2.5:1 to achieve maximal activity), the exact calculation of submicromolar MgATP concentrations is difficult because of uncertainty in assessing the true Mg 2ϩ concentration and dependence on the value of K d for the Mg 2ϩ ⅐ATP complex. To overcome this, we repeated the MgATP binding experiments using a constant 2.5 mM Mg 2ϩ concentration, where virtually all ATP is in the metal-complexed form. Fig. 5 shows that at low MgATP concentrations, ⑀-replete ␤Y331W F 1 has significantly higher affinity for MgATP than the ⑀-depleted enzyme. The K d values derived from Fig. 5 indicate that ⑀ increases the affinity of site 1 for MgATP by a factor of 6 (K d1 ϭ 0.12 and 0.02 M in the absence and presence of ⑀, respectively), whereas there is only a small effect on site 2 (K d2 ϭ 2.8 and 1.4 M, respectively) and no effect on site 3 (K d3 ϭ 23 M in both cases).
The influence of the ⑀-subunit on the affinities for MgADP is shown in Fig. 6. Again, at low ligand concentrations, the ⑀-replete enzyme has a higher affinity for MgADP than the ⑀-depleted form. The data were fitted well using a model with two classes of binding sites. The calculated K d values for the ⑀-replete enzyme were K d1 ϭ 0.06 M and K d2 ϭ 19 M, with 1.0 sites of class 1 and 1.7 sites of class 2. The respective values for ⑀-depleted F 1 were K d1 ϭ 0.7 M (0.9 sites) and K d2 ϭ 25 M (1.9 sites). Thus, ⑀ increases the MgADP binding affinity of catalytic site 1 by ϳ10-fold, whereas the affinities of sites 2 and 3 are not significantly affected. sites of F 1 -ATPase, using a specific fluorescence probe for catalytic site occupancy, ␤-Trp-331. Previously, we had shown that the ␤Y331W mutation did not affect the enzymatic properties; ␤Y331W F 1 and F 1 F 0 are fully competent in ATP hydrolysis and synthesis (18,20). Here we first demonstrated that the interaction between ␤Y331W F 1 and the ⑀-subunit, with K d ϳ 1 nM, is very similar to that found in the wild-type enzyme (8,33) and that, as in wild-type F 1 (8), ⑀ reduced the ATPase activity of ␤Y331W F 1 by ϳ90%.
Effect of ⑀ on Catalytic Site Nucleotide Binding-⑀ did not affect the binding of free ATP or ADP to any of the catalytic sites. For magnesium-nucleotide binding, the only pronounced effects were at the high-affinity catalytic site 1, where the affinity for MgADP and MgATP was ϳ10-fold higher in the ⑀-replete than in the ⑀-depleted form of the enzyme (Figs. 5 and 6). One reason for this difference could be a reduced off-rate for the magnesium-nucleotide complex in the presence of ⑀. Interestingly, previous work from one of our laboratories (34) showed that release of product P i from ⑀-replete F 1 in "uni-site catalysis", i.e. under conditions where only the high-affinity site is filled with substrate, was reduced by a similar factor compared with that from the ⑀-depleted enzyme. The results presented here support the conclusion that the ⑀-subunit decelerates a conformational change step that is necessary to release ligands from the high-affinity binding site.
Further Evidence That MgATP Binding at Site 3 Determines Overall Rate of Hydrolysis-In ⑀-depleted and ⑀-replete ␤Y331W F 1 , a single K m value of 20 -30 M was sufficient to describe the substrate concentration dependence of enzymatic activity (Fig. 1), and in both cases, K m clearly corresponded to K d3 , i.e. the K d for MgATP at the low-affinity site 3. A re-plot of the data (Fig. 4) confirmed that in both ⑀-depleted and ⑀-replete ␤Y331W F 1 , only enzyme molecules that have all three catalytic sites filled contribute significantly to the measured hydrolytic activity ("tri-site catalysis").
In earlier reports using wild-type E. coli F 1 , a second, lower K m was found (34,35), which was ascribed to turnover due to enzyme molecules with two filled catalytic sites. However, neither the higher nor the lower K m was significantly affected by the presence or absence of ⑀ (34), consistent with the findings here that K d3 and K d2 for MgATP are similar in the ⑀-depleted and ⑀-replete enzymes. The fact that substantial bi-site activity was not detected in the current work or in a previous study (18) is likely due to differences in experimental protocol. Bi-site catalysis was evident when we investigated the hydrolysis of MgTNP-ATP 2 by ␤Y331W F 1. For this substrate, two K m values were required to describe the enzymatic activity adequately, and the lower K m corresponded extremely well to K d2 (36). What should be emphasized, however, is that even in the cases where bi-site activity is described, tri-site activity is clearly the physiologically relevant working mode of the enzyme in the ⑀-replete or ⑀-depleted enzyme.
Relevance of Current Studies to Inhibitory Properties of ⑀-It is intriguing that the ⑀-induced increase in affinity and the reduction of the product release rate from the high-affinity site are by a factor (ϳ10-fold) similar to that for the decrease in the multi-site hydrolysis rate in the presence of ⑀, suggesting that the two events might be closely connected. According to the model for multi-site hydrolysis recently proposed by two of us (3,37), MgADP release from the low-affinity site 3 is ratelimiting. The fact that the ⑀-subunit influences the fluorescence spectrum of the ␤-Trp-331 residues in all three catalytic sites in a similar manner (Fig. 2) shows that the long-range interactions between ⑀ and the catalytic sites are not restricted to any single site, but are of a more global nature. Also, it may be noted in this context that according to Grü ber and Capaldi (38), the ␤-subunit to which ⑀ can be cross-linked is neither the one that carries catalytic site 1 nor the one that carries site 3, but the one containing the medium-affinity site 2. Thus, ⑀ may affect product release from the low-affinity site 3 by a mechanism similar to that responsible for reduction of uni-site product release. As K d3 (MgADP) is essentially identical in the absence and presence of ⑀, from K d ϭ k off /k on it follows that a 10-fold increase in the product off-rate in the absence of ⑀ would have to be accompanied by a similar increase in the on-rate. Such parallel changes in kinetic constants are not unprecedented in F 1 . K d1 for binding of MgADP and MgAMP-PNP is ϳ0.1 M; however, k on (and consequently, k off ) for MgAMP-PNP is ϳ2 orders of magnitude less that that for MgADP (18). Further experiments will be necessary to determine the kinetics of each of the individual reaction steps in multi-site catalysis and the detailed inhibitory mechanism of ⑀.
Heterogeneity of Preparations with Respect to ⑀ Content Does Not Affect Conclusions Regarding Overall Mechanism-Finally, it is worthwhile to reexamine previous experimental data obtained with ␤Y331W F 1 (reviewed in Ref. 3) in light of the results of the present study. As judged from the specific activities (5.9 units/mg at 23°C and 13 units/mg at 30°C) (18), from the MgATP and MgADP binding affinities at site 1 (0.028 and 0.08 M at 2.5 mM Mg 2ϩ ) (39,40), as well as from the wavelength position of the ␤-Trp-331 fluorescence spectrum ( max ϭ 349 nm) (18), the enzyme prepared by the standard procedure in our laboratory is largely (Ͼ90%) ⑀-replete, and due to the relatively high concentration of enzyme used (Ն50 nM), it remains ⑀-replete in both hydrolysis and nucleotide binding assays. On the other hand, it appears that some degree of ⑀-depletion can occur as a result of the nucleotide depletion procedure involving Sephadex G-50 gel filtration in 50% (v/v) glycerol-containing buffer. Such nucleotide-depleted ␤Y331W F 1 has a specific activity of 9.4 units/mg (at room temperature) and K d1 (MgADP) ϭ 0.14 M (18), indicating that up to 20% of the enzyme molecules may not contain ⑀.
However, in general, such partial ⑀-depletion does not appear to present a problem in experiments studying multi-site hydrolysis, as the results described here show that the mechanism of the ⑀-depleted and ⑀-replete enzymes is fundamentally the same. Thus, recent criticism (41) of the Trp fluorescence approach to correlate hydrolysis and nucleotide binding data is FIG. 6. MgADP binding to catalytic sites in absence and presence of ⑀. OE, ⑀-depleted ␤Y331W mutant F 1 ; E, ⑀-replete ␤Y331W mutant F 1 . The lines are theoretical curves fitted to the experimental data points assuming a model with two types of binding sites. Further details are given under "Experimental Procedures." unjustified. 3 Of wider relevance, previous reports in the literature using E. coli F 1 where mixtures of ⑀-depleted and ⑀-replete enzyme molecules might have occurred can nevertheless be interpreted with confidence.