Conformational changes of the mitochondrial F1-ATPase epsilon-subunit induced by nucleotide binding as observed by phosphorescence spectroscopy.

Changes in conformation of the ϵ-subunit of the bovine heart mitochondrial F1-ATPase complex as a result of nucleotide binding have been demonstrated from the phosphorescence emission of tryptophan. The triplet state lifetime shows that whereas nucleoside triphosphate binding to the enzyme in the presence of Mg2+ increases the flexibility of the protein structure surrounding the chromophore, nucleoside diphosphate acts in an opposite manner, enhancing the rigidity of this region of the macromolecule. Such changes in dynamic structure of the ϵ-subunit are evident at high ligand concentration added to both the nucleotide-depleted F1 (Nd-F1) and the F1 preparation containing the three tightly bound nucleotides (F1(2,1)). Since the effects observed are similar in both the F1 forms, the binding to the low affinity sites must be responsible for the conformational changes induced in the ϵ-subunit. This is partially supported by the observation that the Trp lifetime is not significantly affected by adding an equimolar concentration of adenine nucleotide to Nd-F1. The effects on protein structure of nucleotide binding to either catalytic or noncatalytic sites have been distinguished by studying the phosphorescence emission of the F1 complex prepared with the three noncatalytic sites filled and the three catalytic sites vacant (F1(3,0)). Phosphorescence lifetime measurements on this F1 form demonstrate that the binding of Mg-NTP to catalytic sites induces a slight enhancement of the rigidity of the ϵ-subunit. This implies that the binding to the vacant noncatalytic site of F1(2,1) must exert the opposite and larger effect of enhancing the flexibility of the protein structure observed in both Nd-F1 and F1(2,1). The observation that enhanced flexibility of the protein occurs upon addition of adenine nucleotides to F1(2,1) in the absence of Mg2+ provides direct support for this suggestion. The connection between changes in structure and the possible functional role of the ϵ-subunit is discussed.

The ATPase (ATP synthase) is the enzyme responsible for ATP synthesis during oxidative phosphorylation in all energytransducing membranes. It is composed of two main parts: F 0 , capable of proton transport across the membrane and the catalytic part; F 1 , bound to F 0 through a "stalk" segment (for reviews see Refs. [1][2][3][4][5]. F 1 of eukaryotes and prokaryotes are similar in subunit composition; they are composed of five dif-ferent subunits ␣ through ⑀, in order of decreasing molecular weight, with the stoichiometry 3, 3, 1, 1, and 1 (2,6). However, these enzymes are not identical mainly because of differences in two of the subunits; ␦and ⑀-subunits of bacteria (and chloroplasts) are homologous to oligomycin sensitivity conferring protein and ␦ of beef heart, respectively. Therefore, the ⑀-subunit of mitochondria lacks a counterpart in bacteria and chloroplasts. The two major subunits ␣ and ␤ of F 1 contain at their interfaces a total of six nucleotide binding sites, which are characterized by different properties (7)(8)(9)(10)(11)(12)(13). Nucleotide sites have been classified as catalytic or noncatalytic according to their ability to exchange bound ligand rapidly during hydrolysis of Mg-ATP (13). According to this definition, there are three catalytic and three noncatalytic sites. The functional role of the latter sites has yet to be elucidated. However, several authors have suggested a regulatory function (11, 14, 16 -18). Only two noncatalytic sites and one catalytic site are fully occupied on desalted F 1 ; these sites are regarded as tight binding sites (19), and this state of occupancy is described by the symbol F 1 (2,1) 1 according to Kironde and Cross (20). The vacant noncatalytic site is also regarded as the exchangeable noncatalytic site (21). Catalytic sites have been shown to exhibit magnesium dependence and a relatively broad nucleotide specificity (2,19,22), whereas noncatalytic sites have a significant preference for adenine nucleotides (19,20,(23)(24)(25).
Both ATP hydrolysis and ATP synthesis catalyzed by the F-type ATPases are cooperative processes now thought to involve ligand-induced and energy-dependent conformational changes, which modulate the affinity of catalytic sites for substrates and products (25,26). The mechanism and the particular domain (or subunit) of F 1 and F 0 involved in each individual step of the processes are unknown, in particular the mode of the transmission of conformational signals between domains of the protein. Approaches that have been used to monitor conformational changes in F 1 and in F 0 F 1 -ATPase preparations, including chemical labeling of certain amino acid residues of the protein (27,28), hydrogen exchange (29), binding of inhibitors (30), and protease digestion experiments (31,32), are invasive, thus providing information regarding a structurally altered protein. Methods are needed to selectively monitor conformational changes of specific domains or subunits of the unmodified protein.
Trp phosphorescence measured at room temperature has shown considerable potential for the study of protein structure in solution (33). The sensitivity of the triplet-state lifetime of * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ To whom correspondence should be addressed. Tel.: 39-50-883320; Fax: 39-50-883215. the indole nucleus to the flexibility of its surrounding matrix (34) has been extremely useful with respect to revealing subtle conformational changes induced in proteins by binding of substrates, coenzymes, inhibitors, or interacting macromolecules (35)(36)(37). We have recently used the phosphorescence of the sole tryptophan residue of the mitochondrial F 1 complex as an internal probe of the ⑀-subunit (38,39), and the lifetime () measurements have revealed conformational changes of the nucleotide-depleted enzyme as a consequence of Mg-ATP binding at low temperature.
The high complexity of nucleotide binding sites of F 1 and the existence of temperature-dependent conformational states of the enzyme prompted us to further investigate on possible alterations in the dynamic structure of the ⑀-subunit induced by selective nucleotide-site occupancy. We have analyzed the phosphorescence decay kinetics of F 1 at room temperature in the presence or absence of nucleoside di-and triphosphates associated with loose or tight, catalytic and noncatalytic nucleotide binding sites. These noninvasive studies of intrinsic phosphorescence provide evidence that the conformation of the ⑀-subunit in situ is affected differently by the binding of nucleoside di-or triphosphates to the various nucleotide binding sites of F 1 .

MATERIALS AND METHODS
ATP, ADP, GTP, GDP, phosphoenolpyruvate, Hepes, Tris, and NADH were obtained from Sigma as were pyruvate kinase and lactate dehydrogenase in glycerol-containing buffers. Sephadex G-50, Sephacryl S-300, Blue Sepharose CL-6B, and standard marker proteins were obtained from Pharmacia Biotech Inc. F 1 was prepared from bovine heart mitochondria according to Penefsky (40). We have observed that at this stage the enzyme preparation contained minor contaminants: a protein with an apparent molecular mass of about 48 kDa and the ATPase inhibitor protein. All of the contaminants were removed as follows. By affinity chromatography on Blue Sepharose CL-6B, 16 mg of protein was loaded onto a column (4 cm ϫ 1 cm, inner diameter) in 0.2 M NaCl, 1 mM EDTA, 1 mM ␤-mercaptoethanol, 1 mM ATP, and 20 mM Tris-Cl, pH 8 (41), at a 4 ml/h flow rate. 13 mg of the non-retained protein were concentrated to 15 mg/ml by ultrafiltration with a Diaflo XM-300 (Amicon) membrane. The concentrate was then chromatographed on Sephacryl S-300 (40 cm ϫ 1.6 cm, inner diameter) in 25 mM Tris-Cl, 0.25 M sucrose, and 1 mM ATP, pH 8; a typical elution profile is shown in Fig. 1A. Fractions containing the 32-40-ml elution volume (9 mg of protein) were characterized by a constant specific activity, indicating the presence of a single molecular species that was used in the experiments. SDS-polyacrylamide gel electrophoresis confirmed this assertion since even overloading the gel, the typical five-subunit pattern of F 1 was observed (Fig. 1B). The enzyme solution was stored at 5°C as a suspension at 50% ammonium sulfate saturation in the presence of 4 mM ATP (pH 8). The enzyme activity was stable for several weeks in this state. Since the technique used in our experiments evaluates the single Trp of F 1 , it is extremely important that no contaminating protein is present in the enzyme preparation to avoid Trp signals other than that of F 1 . F 1 (2,1) was obtained from the above enzyme suspension by centrifugation, and the sedimented enzyme was dissolved at 4 -6 M in a buffer containing 150 mM sucrose, 1 mM KH 2 PO 4 , 1 mM MgSO 4 , 10 mM K ϩ -Hepes, pH 8; it was desalted on a Sephadex G-50 centrifuge column (42) equilibrated with the same buffer. F 1 (3,0) was prepared from the ammonium sulfate suspension removing unbound nucleotide and desalting by passage through a centrifuge column equilibrated with 150 mM sucrose, 10 mM Hepes, 1 mM MgSO 4 , pH 8; it was then followed by a procedure based on the displacement of nucleotides from the catalytic sites by a brief exposure (1 min) to pyrophosphate and removal of unbound nucleotides by gel filtration (20). F 1 (3,0) was also prepared from F 1 (2,1) by substituting pyrophosphate with Mg-GDP, according to Ref. 43. Nd-F 1 was prepared from submitochondrial particles by gel permeation chromatography in the presence of 50% glycerol (v/v) as previously described (38). All of the F 1 preparations were pure and had the typical subunit stoichiometry as evidenced from SDS-polyacrylamide gel electrophoresis (Fig. 1B). This is of particular importance, since it has been hypothesized that the ⑀-subunit might be substoichiometric with respect to the rest of the protein (12), which could result in misinterpretation of the phosphorescence data. The percent volume calculated for the protein bands of a typical F 1 , using the Molecular Analyst PC image analysis software for the Bio-Rad densitometer (model GS-670) was ␣ ϩ ␤ ϭ 85.81, ␥ ϭ 9.49, ␦ ϭ 3.41, and ⑀ ϭ 1.29.
The nucleotide content of each enzyme preparation was determined by reverse-phase HPLC, following nucleotide extraction according to Di Pietro et al. (44). The HPLC analysis of the extracts was performed on a Nova-Pak TM C 18 column (150 ϫ 3.9 mm, 4-m Nova-Pak TM packing material, Waters) equilibrated in 30 mM KH 2 PO 4 , pH 5.4, containing 3 mM tetrabuthylammonium sulfate. Nucleotides were eluted with a 0 -30% acetonitrile gradient over 25 min at a flow rate of 1 ml/min. Detection was by absorbance at 260 nm. The nucleotide/F 1 molar ratios observed were 0.4 Ϯ 0.1 for Nd-F 1 , 2.8 Ϯ 0.4 for F 1 (2,1), and 2.6 Ϯ 0.5 for F 1 (3,0) using a molecular mass of 370 kDa-for the enzyme (6) in all stoichiometry calculations.
SDS-polyacrylamide gel electrophoresis was carried out according to Laemmli (45) using a polyacrylamide gradient from 14 to 25% containing 0.1% SDS. The procedure has been previously described in detail (46).
The ATPase activity was determined with an ATP regenerating system by following the decrease of NADH absorption at 340 nm in a 7850 model Jasco spectrophotometer. The assay was carried out at substrate-saturating concentration (steady state) as previously reported (38). The specific activity of the enzyme was 80 -100 units/mg protein at 20°C.
Protein concentrations of enzyme solutions were determined by the method of Lowry et al. (47).
Phosphorescence spectra and decay measurements were obtained with a phosphorimeter, constructed in this institution, as previously described (48). The photons were generated by a Cernax xenon lamp (LX 150UV, ILC Technology), and the wavelength was set with a 250-nm grating monochromator (Jobin-Yvon, H25). The emission was detected with an EMI 9635 QB photomultiplier. Phosphorescence decay in fluid solution at room temperature was monitored with a homemade apparatus suitable for lifetime measurements in the s-ms range described in detail elsewhere (75). Pulsed excitation( ex ϭ 292 nm) was generated by a frequency-doubled flash-pumped dye laser (UV500 M Candela) with a pulse duration of 1 s and an energy/pulse typically of 1-10 mJ. The sample, placed in a vacuum-proof quartz cuvette that allows excitation of the solution from above, is extensively deoxygenated prior to analysis. The emitted light is measured at 90°from the excitation light and selected by a filter combination in the window between 420 and 480 nm. The photomultipliers are protected from the intense excitation light and fluorescence pulse by a high speed chopper blade that closes the slits during laser excitation. The minimum lag time of the apparatus is about 10 s. The decay signal was digitized by a computerscope system (ISC-16, RC Electronics) capable of averaging multiple sweeps. Subsequent analysis of decay curves in terms of the sum of exponential components was carried out by a nonlinear least squares fitting algorithm implemented by the program Global Analysis (Global Unlimited, LFD University of Illinois, Urbana).
For each sample, the phosphorescence decay was measured three times, and samples were prepared at least four times. The standard error of preexponential terms and lifetime components are better than Ϯ 10%. It should be noted, however, that the variability of these parameters can be even somewhat greater when one compares different preparations of the protein. Such variability in the decay kinetics can be traced down to different amounts of quenching impurities present in organic solvents (glycerol), and glasswares (75). For this reason, comparisons are always made between samples obtained from the same enzyme preparation. Fig. 2 shows the selected examples of phosphorescence intensity decay of Nd-F 1 incubated with adenine nucleotides 1 mM at 293 K. The lifetimes () and the preexponential parameters (␣) derived from a biexponential curve fitting are shown in Table I. Incubation of Nd-F 1 with adenine nucleotides at 1 mM in the presence of magnesium changes the intrinsic phosphorescence decay parameters of Nd-F 1 . ATP decreases the average phosphorescence lifetime from 2.8 to 1.9 ms. This lifetime decrease indicates an enhanced flexibility of the polypeptide chain surrounding the chromophore (35,49,50). Thus, the addition of ATP to Nd-F 1 induces conformational changes of the protein, resulting in a more flexible environment for the Nterminal segment of the ⑀-subunit, where Trp is located at position 4 (6).

Nucleotide Effects on Nd-F 1 Phosphorescence Decay-
ADP has an effect opposite that of ATP; it enhances the protein rigidity since the intrinsic av of the protein increases from 2.8 of Nd-F 1 to 3.4 ms upon nucleotide binding.
The opposite effect induced on the ⑀-subunit conformation by ATP and ADP binding suggests that different conformations of the nucleotide binding sites are induced by the two nucleotides and that different allosteric effects are then transmitted to the Trp environment of the ⑀-subunit. This would be consistent with the idea of several authors who, on the basis of inhibition studies (18,(51)(52), speculated that the F 1 -ATPase complex may exist into two different conformations, E s and E h , favored by ADP and ATP binding, respectively.
When Nd-F 1 is incubated with stoichiometric amounts of ADP in the presence of Mg 2ϩ , it results in inhibition. It has been shown that the inhibitory ADP is bound in a catalytic site (52)(53)(54). To establish whether the binding of ADP to this high affinity catalytic site affects the conformation of the ⑀-subunit, ADP has been added stoichiometrically to Nd-F 1 . The decay parameters (Table I) do not change significantly with respect to control, indicating a lack of influence of ADP filling the high affinity catalytic site on the conformation of the protein at the ⑀-subunit level. Also, the addition of stoichiometric ATP to Nd-F 1 (i.e. conditions for unisite catalysis as in Ref. 55) does not affect the Trp phosphorescence decay. Therefore, independent of the adenine nucleotide tested, it would appear reasonable to conclude that filling the high affinity catalytic site does not result in conformational changes of the protein involving the N-terminal segment of the ⑀-subunit.
To further investigate which nucleotide site(s) of F 1 have to be occupied to induce ⑀-subunit conformational changes, preparations of F 1 at different levels of nucleotide occupancy of sites have been studied. F 1 (2,1) Phosphorescence Decay Changes Induced by Nucleotides-Preliminarily, we obtained the phosphorescence spectra of the F 1 (2,1) form in glycerol-phosphate buffer glass at 140 K (not shown), and we observed that the spectra are essentially identical to that of Nd-F 1 previously reported (38). This implies that the nucleotide binding to the high affinity sites does not change the physico-chemical environment of the ⑀-subunit Trp.
Incubation of F 1 (2,1) with 1 mM Mg-ADP, 1 mM Mg-ATP, or 0.5 mM Mg-GTP at 293 K in 150 mM sucrose, 1 mM MgSO 4 , 10 mM Hepes, pH 8, induces consistent changes of the phosphorescence decay (Fig. 3). Table I shows the phosphorescence decay parameters of a typical experiment. The data clearly display a significant decrease of the average lifetime from 4.3 to 3.3 and 3.5 ms when ATP or GTP are bound, respectively. This indicates an increased flexibility of the chromophore environment upon occupancy of the vacant nucleotide binding sites by NTP, whereas ADP has an opposite effect since it enhances av to 4.6 ms (Table I). Thus, the results are similar to those observed on Nd-F 1 , providing further support to the conclusion that 1) occupation of loose binding site(s) is responsible for the ⑀-subunit structural change and 2) binding of the nucleoside triphosphate increases the flexibility of the Trp environment, whereas the binding of nucleoside diphosphate enhances its rigidity (Mg 2ϩ present).
It has been shown that addition of ATP or ADP plus P i in the presence of magnesium results in a reactivation of the AMP-PNP-inhibited ATP hydrolysis activity of the enzyme (56). The similar behavior of ATP or ADP plus P i has prompted investigation with respect to similar effect of the ligands on the F 1 conformation. Thus, experiments designed to determine whether inorganic phosphate added together with ADP might have the capability to influence the ⑀-subunit conformation  Ϫ , 0.5 mM EDTA, and 10% glycerol, pH 8. F 1 (2,1) (4 M) was in 150 mM sucrose, 1 mM MgSO 4 , and 10 mM K ϩ -Hepes, pH 8. F 1 (3,0) (3 M) was in 150 mM sucrose, 1 mM MgSO 4 , 10 mM Hepes, pH 8. Triplet state lifetimes ( i ) and preexponential terms (␣ i ) are derived from a biexponential fitting of the phosphorescence decay (P(t) ϭ ␣ 1 e Ϫt/1 ϩ ␣ 2 e Ϫt/2 ) of F 1 form added with Mg 2ϩ (2 mM) and with nucleotide. av is the average phosphorescence lifetime ( av ϭ 1 ␣ 1 ϩ 2 ␣ 2 ). Phosphorescence decays ( ex ϭ 292 nm) were measured at 20°C.    Table I. similarly to ATP were initiated. For this purpose, F 1 was incubated in the presence of ADP ϩ Mg 2ϩ over a range of 0.1-10 mM P i (i.e. the physiological range, according to Refs. 57 and 58), with care being taken to exclude adventitious P i from enzyme and buffers. This treatment did not alter significantly the effect of ADP alone on the phosphorescence decay parameters of the enzyme (Table I), indicating that the activation effect of P i on the Mg-ADP-F 1 complex cannot be related to the ⑀-subunit conformation. F 1 (3,0) Phosphorescence Decay Changes Induced by Nucleotide Binding-To establish whether the effects observed are the consequence of nucleotide binding to either or both the catalytic and the noncatalytic sites left vacant in the F 1 (2,1) form, F 1 (3,0), the enzyme containing three filled noncatalytic sites and three vacant catalytic sites, was prepared following two different procedures using either pyrophosphate or GDP in the presence of Mg 2ϩ to displace selectively nucleotides from the catalytic sites. The results obtained using one or the other of the two preparations were similar. Fig. 4 shows a typical decay curve obtained incubating at 293 K the enzyme in the presence of Mg-ATP. The addition of both ATP and GTP in the presence of Mg 2ϩ results in increased av , which shifts from 2 ms for F 1 (3,0) to 2.5 and 2.3 ms, respectively (Table I). Therefore, filling catalytic sites with nucleoside triphosphates induces a significant tightening of the ⑀-subunit conformation.
The comparison of these data with those obtained on Nd-F 1 suggests that the effects observed on addition of nucleotides to both Nd-F 1 and F 1 (2,1) are the sum of two distinct effects; the binding of the exchangeable noncatalytic site induces an increased flexibility of the ⑀-subunit N-terminal domain, which likely overwhelms the tightening of the same domain induced on filling the loose catalytic binding sites. F 1 (2,1) Phosphorescence Decay in the Absence of Mg 2ϩ under Different Nucleotide Conditions-Noncatalytic nucleotide binding sites of F 1 have a preference for uncomplexed nucleotides (i.e. in the absence of magnesium ions) (21, 59 -61). Here, we describe the effects of incubating the enzyme with adenine nucleotides, in the absence of Mg 2ϩ , on the ⑀-subunit conformation. Control experiments show that the phosphorescence emission decay of F 1 (2,1) with excess EDTA has a similar biexponential behavior as in the presence of excess Mg 2ϩ . This indirectly indicates that the binding of Mg 2ϩ to F 1 (2,1) does not affect the Trp environment of the ⑀-subunit. Moreover, the phosphorescence decay parameters of F 1 (2,1), to which was added EDTA ϩ P i , do not show significant changes of the ⑀-subunit conformation (Table II).
Addition of ATP in the presence of 2 mM EDTA reduces the av value from 4.3 to 3.7 ( Fig. 5 and Table II). This result was expected on the basis of above results where the filling of the exchangeable noncatalytic nucleotide binding sites could reduce av , indicating an enhanced flexibility of the protein. The enhanced flexibility associated with the filling of the exchangeable noncatalytic site and the enhanced rigidity associated to the filling of loose catalytic sites might be additive or not. In fact, it seems they are not since if they were, one would have expected an even larger increase of the protein flexibility, for the filling of the catalytic sites under the present experimental conditions is not favorable. This divergence from additivity might be explained if one considers that Mg 2ϩ changes the structure of the nucleotide binding sites as recently shown (62) and that the binding of sites in the absence of Mg 2ϩ might result in structural changes transmitted to the ⑀-subunit different from those observed in the presence of the metal.
Similar considerations could explain the effect of ADP on the dynamic properties of F 1 in the absence of magnesium ions, as monitored by Trp phosphorescence. In fact, addition of the nucleoside diphosphate in the absence of Mg 2ϩ reduces av from FIG. 5. Decay of phosphorescence intensity of F 1 (2,1) in the presence of ANP (Mg 2؉ absent) at 293 K. a, F 1 (2,1) was in 150 mM sucrose, 10 mM K ϩ -Hepes, 2 mM EDTA, pH 8; b, 1 mM was added to F 1 (2,1); and c,1 mM ADP was added to F 1 (2,1).  Table I.  (2,1) phosphorescence decay in the absence of Mg 2ϩ ions F 1 (2,1) (4 M) was in 150 mM sucrose and 10 mM K ϩ -Hepes, pH 8. EDTA (2 mM) was added to all samples. Nucleotide concentration was 1 mM. Potassium phosphate effect on the enzyme lifetime was checked in the range 0.1-10 mM. Experimental conditions and parameters are detailed in the legend to Table I 4.3 to 2.8 ms (Table II), indicating an increased flexibility of the protein, whereas an enhancement of rigidity was observed in experiments in which ADP was added along with Mg 2ϩ to the F 1 -ATPase complex. Finally, addition of 0.1-10 mM P i along with (or following) ADP does not change significantly the effect of ADP when added alone. Therefore, as in the presence of Mg 2ϩ , in its absence the addition of inorganic phosphate to the ADP-F 1 complex is without effect on the ⑀-subunit conformation.

DISCUSSION
The ⑀-subunit of the mitochondrial F 1 -ATPase complex is the polypeptide with the lowest molecular mass (5.5 kDa) of the protein, and its function is unknown. However, it is thought that it plays a role in the coupling between F 1 and F 0 (12,63). Consistent with such a role, although experimental evidence of the ⑀-subunit location is not available, most of the F 1 and F 0 F 1 models show the polypeptide located at the interior of the ␣-␤-subunit core, facing F 0 , possibly contributing to the stalk region (3,5,12,51,64,65).
Previously, through the investigation of the intrinsic phosphorescence of the mitochondrial F 1 -ATPase complex, we showed that the addition of Mg-ATP to the enzyme bearing vacant nucleotide binding sites resulted in large conformational changes of the protein surrounding the N-terminal domain of the ⑀-subunit (38,39). However, our experiments were performed at 273 K or below in the presence of 50% glycerol as a stabilizing co-solvent to operate under optimal conditions for the measurement of the phosphorescence signal. Given these conditions, it is not possible to interpret our results in terms of in vivo function, since the kinetic properties and the conformation of F 1 in media characterized by high viscosity and below 16°C are quite different from those shown between 20 and 37°C (28,66).
In the present study, we have overcome the problem since we were able to carry out the experiments at 20°C by modifying equipment for phosphorescence measurements, and we extended the investigations analyzing the effect of filling with different nucleotides several enzyme forms characterized by different nucleotide content and configuration.
This research has revealed a few intriguing features concerning conformational changes of the ⑀-subunit upon binding of nucleotides to the mitochondrial F 1 complex. 1) The binding of nucleotides to the loose sites is solely responsible for the conformational changes observed on the ⑀-subunit. This contention is based on the following observations: first, addition of Mg-ATP or Mg-GTP in large molar excess to both the Nd-F 1 and the enzyme containing the tightly bound nucleotides, F 1 (2,1), produces a net shortening of the phosphorescence lifetime, comparable in the two enzyme preparations; second, incubation of Nd-F 1 with unistoichiometric Mg-ATP, which loads a single, high affinity catalytic site, does not affect the average phosphorescence lifetime of F 1 ; third, phosphorescence spectra of F 1 (2,1) are identical to those of Nd-F 1 .
2) The comparative analysis of the results obtained with F 1 (2,1), F 1 (3,0), and Nd-F 1 (this in the presence of unistoichiometric Mg-ATP) strongly suggests that the increased flexibility of the Nd-F 1 and F 1 (2,1) forms upon binding of Mg-ATP or Mg-GTP is in fact the result of two different effects and that the filling of the vacant noncatalytic site is responsible for the large increased flexibility of the ⑀-subunit. These conclusions are supported by results of the experiment carried out in the absence of magnesium, a condition favoring the binding of nucleotides to the noncatalytic sites. Under this condition, the average phosphorescence lifetime of F 1 (2,1) markedly decreases upon incubation with adenine nucleotides. 3) Mg-ADP addition to the ATPase complex consistently and greatly enhances the rigidity of the Trp mi-croenvironment, showing an opposite effect with respect to Mg-ATP. Thus, our results provide evidence for two opposite conformational changes of the mitochondrial F 1 -ATPase ⑀-subunit whether ATP ϩ Mg 2ϩ or ADP ϩ Mg 2ϩ is added. Moreover, addition of P i along with (or following) ADP ϩ Mg 2ϩ did not significantly alter the effect of ADP ϩ Mg 2ϩ only, suggesting that once Mg-ADP is bound, P i can not influence the ADP-F 1 structure as probed at the ⑀-subunit level.
The markedly different conformational changes obtained on binding of Mg-ATP, as compared to Mg-ADP, might have several possible explanations. Thus, F 1 might assume two conformational states, depending whether ADP or ATP is the ligand. This possibility has been proposed by Boyer (51) on the basis of various experimental observations reporting significant differences in the ATPase behavior, whether the enzyme has been examined in the presence of ADP or ATP. Interestingly, an x-ray crystallographic study of a Ras protein catalyzing GTP hydrolysis has shown substantial structural differences whether Mg-GTP or Mg-GDP was the ligand. These differences seemed to be caused by a different coordination of the active site Mg 2ϩ ion (67). Since F 1 shares with the Ras protein the conserved phosphate-binding loop (68), Mg 2ϩ might have a similar role with F 1 on binding Mg-ATP or Mg-ADP. A second possibility is that different types of metal-nucleotide diastereoisomers could be the true ligands for binding Mg-ADP or Mg-ATP to F 1 (69). It has indeed been shown that different metalnucleotide epimers of ADP and ATP are the substrates for a number of F 1 -ATPases (70). The consequence might be that different structural signals could be transmitted from the nucleotide binding sites of F 1 to the ⑀-subunit, depending on the particular type of stereoisomer bound.
If the ⑀-subunit is not in close proximity of the nucleotide binding sites, as it is believed, our results demonstrate that conformational changes of F 1 upon substrate binding are transmitted over long distances. Thus, the ⑀-subunit might be involved in the propagation of signals to deeper regions of the F 0 F 1 complex.
Finally, observations of certain similarities between the present study and those carried out on F 1 from other sources should be cited. Evidence for a correlation between occupation of the nucleotide binding sites, catalysis, and conformational changes of the ⑀-subunit were obtained in studies with both Escherichia coli and chloroplast F 1 (31,32,72,73).
The enzyme from the different sources has the functional core ␣ 3 ␤ 3 ␥ composed of homologous subunits and has two additional small, single-copy subunits, which appear to play a role in the energy-coupling mechanism and are physically close to one another (15,64,71,74). The mitochondrial F 1 ⑀-subunit, which has no counterpart in other species, might share with the ⑀-subunit of F 1 from other sources the involvement in the coupling mechanism and/or in its regulation. Nevertheless, the molecular events involving conformational changes of the ⑀-subunit of F-type ATPases from different energy-transducing membranes might differ from one another, as it can be envisaged if one considers several different effects observed on small subunits of mitochondria and E. coli F 1 upon filling of the nucleotide sites. Mendel-Hartvig and Capaldi (31) found a direct relationship between P i binding and ⑀-subunit conformation of E. coli F 1 , whereas it was not possible to find any such relationship with the mitochondrial enzyme. Moreover, both Mg-ADP and Mg-ATP could induce a less tight structure of the ⑀-subunit of E. coli, whereas in our study Mg-ADP and Mg-ATP could increase or decrease, respectively, the tightness of the mitochondrial protein.
In conclusion, our results provide information on dynamic aspects of the enzyme structure and function and provide the first noninvasive, extensive experimental evidence for a change in the conformation of the mitochondrial F 1 -ATPase ⑀-subunit in situ in response to the filling of different classes of nucleotide binding sites with several substrates.