Mitochondrial F0F1 ATP Synthase

Studies reported here were undertaken to gain greater molecular insight into the complex structure of mitochondrial ATP synthase (F0F1) and its relationship to the enzyme's function and motor-related properties. Significantly, these studies, which employed N-terminal sequence, mass spectral, proteolytic, immunological, and functional analyses, led to the following novel findings. First, at the top of F1 within F0F1, all six N-terminal regions derived from α + β subunits are shielded, indicating that one or more F0 subunits forms a “cap.” Second, at the bottom of F1 within F0F1, the N-terminal region of the single δ subunit and the C-terminal regions of all three α subunits are shielded also by F0. Third, and in contrast, part of the γ subunit located at the bottom of F1 is already shielded in F1, indicating that there is a preferential propensity for interaction with other F1 subunits, most likely δ and ε. Fourth, and consistent with the first two conclusions above that specific regions at the top and bottom of F1 are shielded by F0, further proteolytic shaving of α and β subunits at these locations eliminates the capacity of F1 to couple a proton gradient to ATP synthesis. Finally, evidence was obtained that the F0 subunit called “F6,” unique to animal ATP synthases, is involved in shielding F1. The significance of the studies reported here, in relation to current views about ATP synthase structure and function in animal mitochondria, is discussed.

Nature's rationale for providing ATP synthases with this unique molecular architecture has become apparent only in recent studies where evidence that these fascinating enzymes contain two motors has been mounting (31)(32)(33). One of the proposed motors is F 1 , which is driven by ATP binding and/or hydrolysis (31), whereas the other is contained within F 0 and driven by a proton gradient (33). During ATP synthesis, the proposed motor within F 0 , composed of 10 -12 subunit c molecules (30,34) and a single subunit a, is believed to drive the F 1 motor in reverse (35) via a central rotor, the F 1 -␥ subunit. This subunit extends from a ring of subunit c molecules in the F 0 basepiece (19), through the central stalk region, and finally through the center of F 1 where it interacts differently with each of the three ␣␤ pairs (26 -29). During one 360°rotation of the ␥ subunit, the binding of ADP and inorganic phosphate, the synthesis of ATP, and the release of this ATP are believed to occur on each ␣␤ pair as proposed by the "binding change mechanism" (2,36). During this process, the ⑀ subunit in bacteria, or its ␦ subunit equivalent in animals, is also believed to rotate (32), presumably at the bottom of F 1 within the central stalk region.
Although much has been learned about the structure and function of ATP synthases as summarized above, a three-dimensional structure has not been obtained for the complete enzyme (F 0 F 1 ) from any source. In fact, of the 16 different subunit types within the mitochondrial ATP synthase of animal cells, only two and a half of these (␣, ␤, and about half of ␥) have been solved at atomic resolution (26,28). Consequently, there are still many important questions that remain unanswered about structure/function relationships in ATP synthases, and how, within these remarkably complex machines, energy coupling between the proposed proton gradient-driven motor within F 0 is coupled to the reversal of the ATP-driven F 1 motor. Specifically, one important question relates to the identity of exterior subunit regions at the top and bottom of F 1 that are shielded by F 0 in the complete ATP synthase, because these regions may include F 0 /F 1 contact sites essential for energy coupling during ATP synthesis. Although information about these regions, particularly at or near the top of F 1 , is gradually being generated for the Escherichia coli and chloroplast enzymes (12,23), little information is available for animal ATP synthases.
In studies reported below, we used a variety of approaches to identify exterior subunit regions at the top and bottom of rat liver F 1 that are shielded by F 0 . We then assessed the requirement of these regions for coupling a proton gradient to ATP synthesis. In addition, we obtained evidence for the involvement of an F 0 subunit unique to animal ATP synthases.

Materials
Rats (Harlan Sprague-Dawley CD , white males) were obtained from Charles River Breeding Laboratories. ATP, MgCl 2 , 2 EDTA, DTT, 1 tricine, Sephadex G-25 (coarse), soybean trypsin inhibitor, oligomycin, and DCCD were from Sigma. SDS, acrylamide, and bisacrylamide were from Bio-Rad, PVDF membranes from Millipore, and Western blot reagents from Amersham Pharmacia Biotech. Modified trypsin (sequencing grade), pyroglutamate aminopeptidase, hexokinase, and glucose-6-phosphate dehydrogenase were from Roche Molecular Biochemicals. The detergent CHAPS was obtained from Anatrace, whereas sucrose, urea, and potassium phosphate were from J.T. Baker. Coomassie Blue dye binding reagent for protein determination was from Pierce, AEBSF was from Calbiochem, and 3,5-dimethoxy-4-hydroxy-cinnamic acid and ␣-cyano-4-hydroxy-cinnamic acid were from Aldrich. Centriprep filtration units for protein concentration were from Amicon. Venturicidin was from BDH Chemicals. The antibodies to F 0 components were generous gifts from Drs. Y. Hatefi and Akemi Matsuno Yagi. All other products were of the highest purity or grade commercially available.

Methods
Purification of Mitochondrial ATP Synthase (F 0 F 1 )-This complex was purified from rat liver mitochondria by a modification of a previously described method developed in this laboratory (6). A washed submitochondrial particle fraction, referred to previously as 3X membranes (6), was stored at ϳ15 mg/ml and Ϫ20°C in TA buffer (50 mM Tricine, 1 mM ATP, 25 mM EDTA, 0.5 mM DTT, and 5% ethylene glycol, pH 7.9), thawed, and solubilized at 4 mg/ml in 0.6% CHAPS for l h on ice with occasional stirring. After centrifugation for 1 h at 48,000 rpm at 4°C in a 70.1-Ti rotor in a Beckman Optima LE 80K ultracentrifuge to remove unsolubilized material, the supernatant was concentrated using Amicon's Centriprep filtration unit (molecular mass cutoff ϭ 50 kDa). The concentrate was diluted with TA buffer to give a CHAPS concentration of 0.1% and frozen in dry ice and ethanol, and then stored at Ϫ20°C for 3 h. The sample (ϳ20 ml) was then thawed, and 5-ml aliquots (2.0 -2.5 mg/ml) were placed on 20 ml of 25% sucrose in TA buffer without CHAPS and centrifuged 12 h and 45 min at 4°C in the same centrifuge described above but using the SW-28 rotor. The pellet, F 0 F 1 (7-10 mg/ml), judged to be Ͼ90% pure, was stored at Ϫ80°C in TA buffer. The specific activity was 14 -15 mol of ATP hydrolyzed ϫ min Ϫ1 ϫ mg Ϫ1 protein based on the ATPase and protein assays described below.
Purification of the F 1 Moiety of Mitochondrial ATP Synthase-F 1 with a specific activity of about 25 mol ϫ min Ϫ1 ϫ mg Ϫ1 in TrisCl buffer was purified from isolated rat liver mitochondria exactly as described by Catterall and Pedersen (37) with one modification. The terminal Sephadex G-200 step was replaced by chromatography on a 2.1-ϫ 13-cm Sephadex G-25 column packed on top of a 1.5-cm layer of sea sand and a 1-cm layer of glass wool. It was then stored or prepared for use as described by Ko et al. (38).
Preparation of Inner Mitochondrial Membrane Vesicles and F 1 -depleted Vesicles, and Reconstitution of F 1 with the F 1 -depleted Vesicles-All steps were carried out exactly as described previously by Pedersen and Hullihen (39).
Treatment of F 0 F 1 with Pyroglutamate Aminopeptidase-To remove the N-terminally blocked group on the ␣ subunit, F 0 F 1 (1 mg/ml) was placed in a 200 l of digestion buffer containing 100 mM sodium P i , pH 8.0, 10 mM EDTA, 5 mM DTT, 5% glycerol, 0.6% CHAPS, and pyroglutamate aminopeptidase (0.019 mg/ml). The enzyme was then incubated in this mixture first for 12 h at 4°C and then for 2 h at 25°C, after which the entire reaction mixture was added to SDS sample buffer and analyzed by SDS-PAGE. The ␣ subunit band was then subjected to N-terminal sequence analysis exactly as described below. Treatment of F 1 and F 0 F 1 with Trypsin-F 1 and F 0 F 1 (each 160 g) were subjected to digestion for various incubation times at 25°C in a final volume of 0.1 ml containing 250 mM potassium P i , pH 7.5, 5 mM EDTA, and 10 g of trypsin. Trypsin inhibitor (50 g), or AEBSF at a final concentration of 5 mM, was then added at various incubation times to stop the reaction. Following the addition of ammonium sulfate to the mixture to give 65% saturation, the resultant suspension was centrifuged at 13, 000 rpm in a Sorvall SS 34 rotor for 1 h at 25°C.
Assay for ATPase Activity-ATPase activity was assayed exactly as described previously (37) by a spectrophotometric procedure in which ADP formed was coupled to the pyruvate kinase and lactic dehydrogenase reactions.
Assay for ATP Synthesis-ATP synthesis was monitored spectrophotometrically at 340 nm and 25°C by coupling the production of ATP to the reduction of NADP ϩ via the hexokinase and glucose-6-phosphate dehydrogenase reactions. The assay mixture contained, in a final volume of 1 ml, 10 mM potassium P i , 10 mM succinate, 50 mM Tris acetate, 1 mM glucose, 1 mM NADP ϩ , 2.0 mM MgCl 2 , 2 mg/ml bovine albumin, 200 mM sucrose, 20 IU hexokinase, 0.25 IU glucose-6-phosphate dehydrogenase, and 100 -200 g of inner mitochondrial membrane vesicles. The reaction was started by the addition of 1 mM ADP.
SDS-PAGE-This was carried out as indicated either by the method of Laemmli (40) or the more sensitive method of Schä gger and von Jagow (41) for separating proteins in the molecular mass range of from 1 to 100 kDa.
Western Analysis-After conducting SDS-PAGE, the proteins on the gel were transferred electrophoretically at 4°C in 10 mM CAPS, 10% methanol transfer buffer, pH 11, onto a PVDF membrane (1 h at 100 V and 0.2 amp). The membrane was then blocked for l h with 2% bovine albumin plus 5% non-fat dry milk in phosphate-buffered saline T buffer (80 mM Na 2 HPO 4 , 20 mM NaH 2 PO 4 , 100 mM NaCl, 0.1% Tween 20, pH 7.5), incubated for 1 h at 23°C with polyclonal antibodies as indicated in the figure legends, and then further incubated for 1 h at 23°C with secondary antibody (horseradish peroxidase-conjugated anti-rabbit IgG). The immunoreactive bands were detected by an enhanced chemiluminescence (ECL) system.
N-terminal Sequence Analysis-F 1 or F 0 subunits were transferred from SDS-PAGE gels onto PVDF membranes by electroblotting. Transfer conditions were for 1 h in 10 mM CAPS buffer, 10% methanol, pH 11, at 4°C. The peptides were then excised and subjected to N-terminal sequencing (42) using an Applied Biosystems 475A protein sequencing system (43).
Determination of Protein-Three different methods were used to determine protein, Lowry et al. (44), biuret (45), and Coomassie dye (Pierce). In all cases bovine albumin was used as the standard.

RESULTS
Objectives and Systems Employed-The first objective of this study was to identify exterior regions on F 1 that are shielded by F 0 within a fully active, complete ATP synthase (F 0 F 1 ) preparation from an animal mitochondrial source. The regions on F 1 of most interest (Fig. 1A) were its top where the N-terminal regions of ␣ and ␤ subunits reside (26 -28) and its bottom where the C-terminal regions of these same subunits (26 -28) reside together with part of the ␥ subunit (26,28), the ␦ subunit (25,30), and likely also the ⑀ subunit. The second objective was to assess the importance of these regions for coupling an electrochemical proton gradient to ATP synthesis. This information is generally lacking for animal ATP synthases, which are much more complex in their F 0 subunit composition than related enzymes from E. coli and chloroplasts. The third objective was to identify at least one F 0 subunit unique to the animal enzyme involved in shielding F 1 , because such a subunit would represent a candidate for part of the stator or second stalk believed to stabilize the F 1 motor. To complete these objectives, it was essential to have at hand, in addition to our extensively studied rat liver F 1 preparation (25,28,(37)(38)(39), an intact, highly purified preparation of the complete ATP synthase from the same source. This was accomplished by introducing several modifications (see "Methods") into the CHAPS-based procedure previously developed in this laboratory (6). The resultant prepa-FIG. 1. A, structure of rat liver F 1 at 2.8 Å (28) depicting those regions examined in this study for possible shielding by subunits of F 0 . The ribbon representation of the F 1 structure was constructed using the atomic coordinates of the rat liver enzyme (1mab (28)) using the program Quanta. The ␣-helical, ␤-sheet, and loop structures are colored in green, yellow, and purple, respectively. Nucleotides are shown in red. Regions examined in this study are located at the top and bottom of F 1 . These include the N-terminal regions of the ␣ and ␤ subunits (top) and the C-terminal regions of these same subunits (bottom), together with the central stalk ("stem") region (bottom). The central stalk region includes two sections, one that projects about 20 Å out of the F 1 headpiece as revealed by x-ray crystallography (28) and shown in the figure, and another section (not shown) that is not yet resolvable by x-ray crystallography and is predicted to extend about another 30 Å (30). Regions of the ␥ subunit that were resolved by crystallography include residues 4 -48, 209 -273, and 77-90 (28). The ␦ subunit (not shown) is believed to be located also within the region of the central stalk, both on the basis of biochemical studies (25) and on the basis of the recent structure at 3.9-Å resolution of a yeast F 1 -(subunit c) 10 complex (30). B, SDS-PAGE gels of F 1 and F 0 F 1 preparations used in this study. F 1 and F 0 F 1 were purified and subjected to SDS-PAGE as described under "Methods." The presence of all known F 0 F 1 subunits (16) was confirmed by N-terminal sequence analyses ("Methods") and/or by Western analyses ("Methods") using specific antibodies. C, results of ATPase assays demonstrating that the F 0 F 1 preparation used in this study is inhibited by agents known to act at the level of F 0 . ATPase assays were carried out in a 1-ml system as indicated under "Methods." In all cases, 14 g of F 0 F 1 was used to initiate the assay, and where shown the following inhibitors were added: oligomycin (2.5 g), DCCD (5 g), or venturicidin (5 g). ration exhibits a specific activity of 14 -15 mol of ATP hydrolyzed ϫ min Ϫ1 ϫ mg Ϫ1 protein, contains all 16 subunits (Fig. 1B) previously reported for the bovine heart ATP synthase (20), and is markedly inhibited by oligomycin, DCCD, and venturicidin (Fig. 1C), agents known to bind to F 0 (46,47).
Evidence That the N-terminal Regions of the ␣ and ␤ Subunits Located at the Top of F 1 Are Shielded by One or More F 0 Subunits in the Complete ATP Synthase-Here, studies focused on the N-terminal regions of the three ␣ and three ␤ subunits as they are known from the atomic resolution structures of bovine heart and rat liver F 1 (Fig. 1A) to project from the top of the enzyme (26,28). To determine whether one or more of the six N-terminal regions derived from these subunits are shielded by F 0 , we compared the extent to which F 1 , both alone and as part of the complete ATP synthase complex (F 0 F 1 ), is shaved by endogenous proteases during purification. This necessitated carrying out N-terminal sequence analysis on both the ␣ and ␤ subunits of preparations of rat liver F 1 and F 0 F 1 purified as described under "Methods." Therefore, following purification, these subunits were transferred from SDS-PAGE gels onto PVDF membranes by electroblotting, excised, and subjected to sequence analyses, also as described under "Methods." In these experiments, the N-terminal regions of ␣ subunits derived from the F 0 F 1 preparation could be sequenced only following prior treatment with the enzyme pyroglutamate aminopeptidase (48), which removes from the blocked N termini pyrrolidone carboxylic acid ( Fig. 2A, inset), a cyclic form of glutamine.
The results of these experiments revealed that the N-terminal regions of all ␣ and ␤ subunits remain intact in F 0 F 1 during its purification from mitochondria, because the amino acid sequences obtained (Fig. 2, A and B) correspond to those predicted for the mature sequences of these subunits. In contrast, the same subunits are shaved during the purification of F 1 (Fig.  2, C and D). Thus, two amino acids, Q and K, are shaved from ␣ subunits, whereas the 6-amino acid peptide, AAQSSA, is shaved from ␤ subunits (Fig. 2E). The degree of shaving remained the same in a number of F 1 preparations examined. In all cases the N-terminal sequence data obtained was homogeneous, thus ruling out the possibility of differences among the N-terminal regions of each of the three ␣ subunits and each of the three ␤ subunits. These results indicate that, when within the complete ATP synthase, all six N-terminal regions of ␣ ϩ ␤ subunits located at the top of F 1 are shielded during purification from specific proteases endogenous to the mitochondria. Moreover, they strongly implicate one or more F 0 subunits as comprising this shield, which appears to "cap" the top of F 1 rather than being confined to a limited region. Consistent with these conclusions, in experiments not reported here, treatment of purified F 0 F 1 with trypsin for 1.5 h was unable to cleave the N-terminal regions of either the F 1 ␣ or ␤ subunits despite the presence of potential cleavage sites (R and/or K) in both cases.
Evidence That, of the Three Remaining Subunits (␥, ␦, and ⑀), Only the N-terminal Region of ␦, Located at the Bottom of F 1 , Is Shielded by One or More F 0 Subunits in the Complete ATP Synthase-A comparison of amino acid sequencing data presented in Figs. 3A and 3B, and summarized in 3G, shows that the N-terminal region of the ␥ subunit has not been shaved by proteases during isolation of F 1 . Rather, its N-terminal sequence is identical to that found in purified F 0 F 1 , and to the known sequence for this subunit. This is an expected result, which serves as a control. Thus, it is known from the atomic resolution structures of bovine heart and rat liver F 1 preparations (26,28) that the N-terminal region of the ␥ subunit is tucked deep inside the central cavity of the F 1 headpiece where it is well shielded. In contrast, sequencing data for the ␦ sub-unit presented in Figs. 3C and 3D, and summarized in 3G, show that in F 1 purified alone a 4-amino acid peptide, AQAA, is shaved from its N-terminal region. However, when the ␦ subunit is purified as part of the complete F 0 F 1 complex, the AQAA peptide is retained. Finally, a comparison of N-terminal sequencing data presented in Figs. 3E and 3F, and summarized also in Fig. 3G, show that the ⑀ subunit, just as the ␥ subunit, has not been shaved during isolation, indicating that it may be shielded also within F 1 . To date, little is known about the precise location and role of the ⑀ subunit in mitochondrial ATP synthases, although it has been reported to subfractionate with a ␤␦⑀ complex (49) and to bind to the ␦ subunit (50).
The studies described here, taken together with the above studies, indicate that, within the complete liver mitochondrial ATP synthase, regions of F 0 located at both the top and bottom of F 1 shield the N-terminal regions of three of its subunit types (␣, ␤, and ␦), whereas the remaining two subunit types (␥ and ⑀) are shielded or tightly folded within F 1 .
Limited Treatment of Isolated F 1 with Trypsin Results in Further Shaving of Its ␣ and ␤ Subunits While Leaving the Smaller Subunits ␥, ␦, and ⑀ Unaltered-Results presented in Fig. 4 (A and B) show that treatment of isolated F 1 with trypsin for as long as 90 min has no effect on the staining intensities or mobilities of the ␥, ␦, and ⑀ subunits of F 1 in SDS-PAGE gels but does noticeably affect the mobility of the ␣ subunit. Two different gel systems were used in these experiments. One   FIG. 2. A and B, N-terminal sequences obtained for the ␣ and ␤ subunits within purified preparations of F 0 F 1 . The sequence of the ␣ subunit was obtained following removal of the cyclic glutamine residue (A, inset) blocking the N terminus. Removal of the cyclic glutamine residue, preparation of samples for N-terminal sequence analyses, and the sequencing method used are described under "Methods." C and D, N-terminal sequences obtained for the ␣ and ␤ subunits within purified preparations of F 1 . See "Methods" for procedural details. E, summary of A-D. The summary diagram emphasizes that, when F 1 is isolated in the absence of F 0 , endogenous proteases cleave 2 amino acids, Q and K from the N-terminal region of ␣ subunits and a 6 amino acid peptide, AAQSSA, from the N-terminal region of ␤ subunits. contained 7.5% acrylamide (Fig. 4A), which separates best the ␣, ␤, and ␥ subunits but not the ␦ and ⑀ subunits, which migrate with the dye front. The second contained 15% acrylamide (Fig.  4B), which separates the ␥, ␦, and ⑀ subunits better than the ␣ and ␤ subunits.
When all five F 1 subunits from the SDS-PAGE gels were subjected to N-terminal sequence analyses, it was found that a 13-amino acid peptide (TGTAEMSSILEER) had been cleaved from the N-terminal region of the ␣ subunit (Fig. 4C) and a 3-amino acid peptide (APK) from the N-terminal region of the ␤ subunit (Fig. 4D). In data not presented here, the N-terminal regions of the ␥, ␦, and ⑀ subunits were found to be unaltered. Because the decrease in mass of the ␣ subunit appeared by inspection of SDS-PAGE gels (Fig. 4, A and B) to be greater than the mass of the 13-amino acid peptide cleaved from its N terminus, this suggested that trypsin may have cleaved also at the C terminus. For this reason we subjected both isolated F 1 and trypsin-treated F 1 to MALDI-TOF-DE mass spectral analysis as described under "Methods." Consistent with SDS-PAGE and N-terminal sequence analyses, these studies confirmed that the mass of the ␥, ␦, and ⑀ subunits were unaltered and showed that the mass of each ␤ subunit had decreased by only 341 Da (Fig. 5A), as expected from the loss of the APK peptide. In contrast, however, the mass of each ␣ subunit decreased by 4425 Da, of which only 1406 Da could be accounted for by the 13-amino acid peptide cleaved from the N-terminal region (Fig.  4C). Significantly, the remaining decrease in mass, which must come from the C-terminal region, corresponds most closely to a 26-amino acid ␣-helix resulting from trypsin cleavage between amino acid Arg-484 and Ser-485 (Fig. 5B).
These results, while demonstrating that limited trypsin treatment of isolated F 1 causes further shaving of the exterior N-terminal regions of both the ␣ and ␤ subunits located at the top of the molecule, and the exterior C-terminal helix of the ␣ subunit located at its bottom, also show that all three of the smaller subunits (␥, ␦, and ⑀) are resistant to limited trypsin treatment. Significantly, in other studies it was shown that trypsin treatment of the complete ATP synthase (F 0 F 1 ), in contrast to treatment of F 1 , had no detectable effect on the electrophoretic mobility of ␣ subunits in SDS-PAGE (Fig. 5C), indicating that F 0 shields the C-terminal regions of all three of these subunits near the bottom of the F 1 headpiece.
Trypsin-treated F 1 , Although Active as an ATPase and Capable of Rebinding to F 0 within F 1 -depleted Inner Mitochon- 3. A and B, N-terminal sequences obtained for the ␥ subunit within purified preparations of F 0 F 1 (A) and F 1 (B). Preparation of samples for N-terminal sequence analyses and the sequencing method used are described under "Methods." C and D, N-terminal sequences obtained for the ␦ subunit within purified preparations of F 0 F 1 (C) and F 1 (D). Preparation of samples for N-terminal sequence analyses and the sequencing method used are described under "Methods." E and F, N-terminal sequences obtained for the ⑀ subunit within purified preparations of F 0 F 1 (E) and F 1 (F). Preparations of samples for N-terminal sequence analyses and the sequencing method used are described under "Methods." G, summary of A-F. The summary diagram emphasizes that the N-terminal regions of the ␥ and ⑀ subunits are not altered by endogenous proteases during purification of either F 0 F 1 or F 1 , whereas the 4-amino acid peptide AQAA is cleaved from the N-terminal region of the ␦ subunit in F 1 but not F 0 F 1 .   FIG. 4. A and B, SDS-PAGE pattern of F 1 following limited treatment with trypsin. F 1 was previously treated with trypsin exactly as described under "Methods" for the times indicated at the top of the gel and then subjected to SDS-PAGE in gels containing 7.5% (A) or 15% (B) polyacrylamide. The 7.5% gels resolved best the ␣, ␤, and ␥ subunits of F 1 but not the smaller subunits ␦ and ⑀, which migrated with the dye front. In contrast the 15% gels resolved the subunits ␥, ␦, and ⑀ better than the ␣ and ␤ subunits. C and D, N-terminal sequences obtained for the ␣ subunits (C) and the ␤ subunits (D) following treatment of F 1 with trypsin for l h. The summary diagram at the bottom of C shows that trypsin shaves 13 amino acids more off the N-terminal region of the ␣ subunit than the 2 amino acids (Q and K) shaved off by endogenous proteases during purification (Fig. 2E). The summary diagram at the bottom of D shows that trypsin shaves only 3 amino acids more off the N-terminal region of the ␤ subunit than the 6 (AAQSSA) shaved off by endogenous proteases during purification (Fig. 2E). Fig. 6A show that those conditions under which trypsin further shaves the Nterminal regions of both ␣ and ␤ subunits at the top of rat liver F 1 , and the C-terminal region of the ␣ subunits at its bottom (Figs. 4 and 5), are without effect on the capacity of the enzyme to catalyze ATP hydrolysis. Moreover, similar to isolated F 1 , trypsin-treated F 1 is able to restore to F 1 -depleted inner membrane vesicles (IMVs) the capacity to catalyze high rates of ATP hydrolysis inhibited by oligomycin (Fig. 6B, left) and DCCD (not shown). This shows that trypsin-treated F 1 is capable of binding appropriately to the basepiece of F 0 where the sites of action of oligomycin (subunits c and a) and DCCD (subunit c) are located (46,47). However, unlike isolated F 1 , which is able to restore the capacity of F 1 -depleted, actively respiring IMVs to catalyze ATP synthesis inhibited by oligomycin, DCCD (not shown), and the protonophore 2,4-dinitrophenol, trypsintreated F 1 is not effective in this capacity (Fig. 6B, right). (IMVs of rat liver mitochondria, prior to treatment with urea to deplete F 1 and provide F 1 -depleted IMV, exhibit specific activities for ATP hydrolysis in the range of 2000 -3000 nmol of ATP hydrolyzed ϫ min Ϫ1 ϫ mg Ϫ1 protein and specific activities for ATP synthesis in the range of 100 -250 nmol ϫ min Ϫ1 ϫ mg Ϫ1 protein. ) These results indicate that one or more of the exterior, tryp-sin-accessible regions on F 1 defined above are critical for coupling a proton gradient to ATP synthesis in the inner mitochondrial membrane, although they are of little or no importance to the enzyme for catalyzing ATP hydrolysis or for binding to F 0 . Evidence for an Involvement of the Unique F 0 Subunit Called F 6 in Shielding F 1 -Having acquired information that regions on rat liver F 1 , at its top and bottom, are shielded by F 0 , and that one or both of these locations on F 1 may be necessary for coupling a proton gradient to ATP synthesis, our attention turned to the involvement of F 0 . Here the major goal was to identify one or more F 0 subunits unique to the animal system involved in shielding F 1 within the complete F 0 F 1 ATP synthase. The F 0 subunits named "F 6 ," "OSCP," and "d" were considered to be likely candidates, because these subunits are not integral membrane proteins (18,51,52). The strategy employed took advantage of our knowledge that the N-terminal regions of each ␣ subunit shielded by F 0 at the top of F 1 has its N terminus blocked by pyrrolidone carboxylic acid. Therefore, it was rationalized that, by monitoring the ability of a deblocking enzyme to gain entrance to the N terminus of the blocked ␣ subunit in F 0 F l after the addition of a mild dissociating agent, while monitoring the release of any F 0 subunits, we might be able to identify one or more F 0 subunits involved in shielding F 1 .

drial Membranes, Is Incapable of Coupling a Proton Gradient to ATP Synthesis-Results presented in
Based on the above strategy and considerable preliminary work, it was found that the deblocking enzyme (pyroglutamate aminopeptidase) gains access to the N terminus of the ␣ subunit after treatment of F 0 F 1 with 0.6% CHAPS for 14 h (12 h at 4°C and then 2 h at 25°C). Thus, the N-terminal region of the FIG. 5. A, summary of molecular masses obtained for F 1 subunits before and after limited treatment with trypsin. F 1 was subjected to limited treatment with trypsin for 1 h exactly as described under "Methods" and then analyzed in a MALDI-TOF-DE mass spectrometer, also as described under "Methods." The masses obtained from two different experiments varied less than 1%. B, predicted trypsin cleavage site within the C-terminal region of the ␣ subunit. Taking into consideration the large total mass loss of the ␣ subunit, and the known loss of mass from its N-terminal region, the mass loss arising from its Cterminal region was calculated and found to correspond precisely to loss of the 26-amino acid peptide shown (see text for discussion). C, lack of effect of trypsin on the mobility of the ␣ and ␤ subunits within purified F 0 F 1 . F 0 F 1 was treated for up to 1.5 h with trypsin as described under "Methods" and then subjected to SDS-PAGE, also as described under "Methods. "   FIG. 6. A, effect of trypsin on the ATPase activity of rat liver F 1 as a function of time. F 1 was treated with trypsin exactly as described under "Methods" for the times indicated and then assayed for ATPase activity by the spectrophotometric method, also described under "Methods." B, assessment of the capacity of trypsin-treated F 1 to bind to F 1 -depleted inner mitochondrial membrane vesicles and restore ATPase and ATP synthetic activities. Preparation of F 1 -depleted inner membrane vesicles by treating inner membrane vesicles with 3.2 M urea, and reconstitution of F 1 (300 g) or trypsin-treated F 1 (300 g) with the depleted vesicles (0.5 mg), was carried out exactly as described previously (39). ATPase and ATP synthetic activities were assayed as described under "Methods." Where indicated, oligomycin (2.5 g/ml), DCCD (5 g/ml), or 2,4-dinitrophenol (100 M) was added. Standard deviations are based on three different experiments. (In experiments not presented here, DCCD (5 g/ml) inhibited by Ͼ90% the rates of ATP hydrolysis and ATP synthesis in all cases.) ␣ subunit following SDS-PAGE and electroblotting onto a PVDF membrane could now be sequenced (Fig. 7A), verifying that deblocking had occurred. Significantly, 1 M urea, when included under the above conditions, actually stabilized F 0 F 1 and prevented the deblocking enzyme from gaining access to the N terminus of the ␣ subunit (Fig. 7B). When both the supernatant and pellet following treatment of F 0 F 1 with CHAPS, but without urea, was analyzed by immunoblotting with an antibody specific for F 6 , all of this subunit was recovered in the supernatant (Fig. 7C, Condition 1). In contrast, when 1 M urea was present, F 6 was not detected in the supernatant (Fig. 7C, Condition 2) but remained with the pellet. Experiments conducted in an identical manner but using antibodies specific for OSCP, showed that this subunit is also released into the supernatant (Fig. 7D, Condition 1). However, in this case 1 M urea fails to prevent its release (Fig. 7D,  Condition 2). These results indicate that F 6 , but not OSCP, plays a primary role in making the top of F 1 inaccessible to the ␣ subunit-deblocking enzyme. Results obtained with an antibody to subunit d not presented here, although less clear, mimicked most closely those obtained with F 6 , indicating that subunit d may also shield in part the top of F 1 .

DISCUSSION
The novel set of studies reported here were undertaken to gain greater insight into the predicted motor-related features of ATP synthases from animal cells. The objectives were 3-fold and focused on (a) identifying exterior regions at the top and bottom of the F 1 motor that are shielded by F 0 ; (b) determining the importance of these regions for coupling an electrochemical proton gradient via the motor in F 0 to the synthesis of ATP on F 1 ; and (c) identifying one or more F 0 subunits unique to animal ATP synthases involved in shielding F 1 . Each of these objectives was achieved either fully or in part.
Clearly, as revealed by N-terminal sequence analyses ( Figs.  1 and 2), exterior regions at both the top and bottom of rat liver F 1 within the complete ATP synthase are shielded from limited proteolysis arising from endogenous proteases during purification. Shielded regions include the N-terminal regions of all ␣ and ␤ subunits, which are known from the atomic resolution structure of rat liver F 1 to reside at its top (28), and the N-terminal region of the single ␦ subunit, which is inferred from biochemical studies to reside at its bottom (25). Treatment of the complete ATP synthase (F 0 F 1 ) with trypsin is also without effect on the N-terminal regions of these same subunits, providing additional support that they are shielded by F 0 . These finding, although identifying potential F 0 /F 1 contact regions in the complete rat liver ATP synthase, also implicate F 1 as being "capped" at its top by one or more subunits derived from F 0 and shielded at its bottom.
Significantly, the exterior N-terminal regions of F 1 , which are shielded from limited proteolysis during isolation of F 0 F 1 , i.e. the QK dipeptide of the ␣ subunit (Fig. 2E), the AAQSSA hexapeptide of the ␤ subunit (Fig. 2E), and the AQAA tetrapeptide of the ␦ subunit (Fig. 3G), are not required for F 1 to restore ATP synthesis to F 1 -depleted inner membrane vesicles (Fig.  6B). However, if the endogenously shaved F 1 preparation is shaved further at both its top and bottom by exogenously added trypsin, the enzyme is unable to catalyze ATP synthesis when reconstituted with F 1 -depleted inner membrane vesicles (Fig.  6B). This loss of physiological function occurs despite the fact that the more closely shaved F 1 fully retains both its catalytic ATPase activity (Fig. 6A) and its capacity to bind in a normal manner to F 1 -depleted inner membrane vesicles (Fig. 6B). Specifically, trypsin removes from the ␣ subunit's N-terminal region an additional 13 amino acids, TGTAEMSSILEER (Fig.  4C), and from its C-terminal region a 26-amino acid ␣-helix, SDGKISEQSDAKLKEIVTNFLAGFEP (Fig. 5B), while removing also the tripeptide APK from the N-terminal region of the ␤ subunit (Fig. 4D). Taken together, these findings summarized in Fig. 8A indicate that one or more of these trypsin-cleaved exterior regions, which are also shielded by F 0 when within the complete ATP synthase (Figs. 2, 3, and 5), are critical for the capacity of F 1 to couple an electrochemical proton gradient to the synthesis of ATP.
Additional results presented here provide new information about the relationship of F 0 in the complete ATP synthase to the central stalk region of F 1 , i.e. the region that extends from the bottom of the F 1 headpiece to the basepiece of F 0 , a distance of about 50 Å (30). From SDS-PAGE, N-terminal sequencing, and mass spectral studies following treatment with trypsin ( Figs. 4 and 5), it is clear that ␥ subunit is already well protected in the central stalk region within F 1 . This shield is likely provided, at least in part, by the ␦ subunit of F 1 , a view that is supported by previous work in this laboratory on rat liver F 1 (25) and from the low resolution structure of yeast F 1 in complex with a decamer of the subunit c of F 0 (30). The ⑀ subunit of F 1 , which binds to the ␦ subunit (50), may also comprise, in part, the shield that protects the ␥ subunit within F 1 . Because   FIG. 7. A, N-terminal sequence of the F 1 ␣ subunit obtained after prior treatment of F 0 F 1 with the detergent CHAPS and then with the deblocking enzyme pyroglutamate amino peptidase. F 0 F 1 (200 g) was incubated for 12 h at 4°C and then for 2 h at 25°C in a 0.2-ml system containing 100 mM sodium P i , pH 8.0, 10 mM EDTA, 5 mM DTT, 5% glycerol, and 0.6% CHAPS. Following SDS-PAGE, the ␣ subunit band was subjected to N-terminal sequence analysis exactly as described under "Methods." B, absence of N-terminal sequence data for the ␣ subunit upon treating F 0 F 1 as in A but in the presence of urea. All conditions were exactly as in A with the exception that 1 M urea was present. In control studies, urea was shown to be without effect on the activity of the deblocking enzyme. C, Western blots obtained with an antibody to the F 0 subunit "F 6 " following SDS-PAGE of detergent extracts of F 0 F 1 treated as in A or as in B. F 0 F 1 samples were treated exactly as in A with or without urea and then centrifuged at 14,000 rpm for 15 min in a microcentrifuge at 25°C. The supernatant was removed and boiled for 20 min and then centrifuged at 14,000 rpm for 30 min at 25°C. The final supernatant was then removed and subjected to SDS-PAGE/Western blot analysis exactly as described under "Methods" using a polyclonal antibody to F 6 . D, experiments were carried out exactly as described in C using a polyclonal antibody to OSCP rather than F 6 . studies reported here implicate F 0 as shielding the N-terminal region of the ␦ subunit (Fig. 3, C and D), it would appear that in the intact ATP synthase complex, one or more components of F 0 may reside sufficiently near the central stalk region to form contacts, but that these contacts may not be directly in contact with the ␥ subunit. In other words, the ␥ subunit in the central stalk region may be doubly shielded, with the first layer involving the ␦ and ⑀ subunits of F 1 , and the second shield one or more subunits of F 0 .
Finally, studies reported here show that those conditions that make the N-terminally blocked residue (Q*) of the ␣ subunit at the top of F 1 accessible in F 0 F 1 to a deblocking enzyme also result in the release of the F 0 subunit called F 6 (Fig. 7). Moreover, conditions that prevent the N-terminally blocked residue from becoming accessible to the deblocking enzyme also prevent the release of F 6 (Fig. 7). These findings are consistent with a location, or partial location, of F 6 at the top of F 1 in the complete ATP synthase complex of rat liver, although they do not exclude its interaction or partial interaction at other loca-tions. Significantly, F 6 , a small 8.93-kDa protein, required for ATP synthesis in inner mitochondrial membrane vesicles (52), is not found among the F 0 subunits in the simpler ATP synthase preparations derived from E. coli and chloroplasts (1)(2)(3). Therefore, the novel finding reported here about F 6 is of special interest, and in future studies may help reveal why the F 0 unit of animal ATP synthases are much more complex than the F 0 units of the E. coli and chloroplast enzymes.
When conclusions from all the studies described here are considered, we envision an overall structural model for the rat liver ATP synthase in which one or more components of F 0 extend from its basepiece to the top of the headpiece of F 1 (Fig.  8B). At the bottom of F 1 , components of F 0 are considered to associate closely or "hug" the central stalk region forming a "collar" that shields the C-terminal regions of the three ␣ subunits. At the top of F 1 , the extension of this "second stalk" is considered to cover completely the N-terminal regions of all ␣ and ␤ subunits, thus providing a "cap." In addition, the cap may be composed, at least in part, by the F 0 subunit F 6 , unique to animal ATP synthases. This model, derived from using a variety of approaches and methodologies, is consistent with current views about ATP synthase structure, which envision the presence of a stator connecting two motors, one within the F 1 headpiece and another within the F 0 basepiece (31)(32)(33)35). In addition, this model and the experimental work on which it is based provide biochemical support for the recent electron microscopic study of bovine heart ATP synthase, showing a collar around the central stalk and a mass at the top of the F 1 (11), and support for a similar study conducted on the E. coli ATP synthase in which a mass was also observed on the top of F 1 (12). Taken together, these studies suggest that earlier models (53)(54)(55) depicting the stator or second stalk as interacting with only one ␣␤ pair and failing to extend from the basepiece of F 0 to the top of F 1 may require revision.
Significantly, the extension of the stator to the top of F 1 may have important functional consequences, because not one but all three ␣␤ pairs containing the catalytic sites must be stabilized during ATP synthesis. Moreover, that part of the stator that resides at the top may be important also in facilitating formation of those critical contacts recently suggested to be important in the assembly of the ␤-barrel domains (56). FIG. 8. A, summary of exterior regions of F 1 subunits shielded by F 0 that are nonessential for mitochondrial ATP synthesis, and those of which one or more is essential. Blue italicized letters designate those exterior amino acid residues that are not essential, while red letters inside dotted boxes designate those of which one or more is essential. B, model of rat liver mitochondrial ATP synthase incorporating conclusions derived from the novel findings reported here. Experimental data reported here are consistent with a model for animal ATP synthases in which (a) one or more F 0 subunits completely shields the top of F 1 forming a cap; (b) one or more F 0 subunits also shields the bottom of the F 1 headpiece, perhaps forming a collar; and (c) components of F 1 , most likely ␦ and ⑀, shield the ␥ subunit within the region of the central stalk at the bottom of F 1 . The data reported are consistent also with a role for the unique F 0 subunit F 6 in helping shield F 1 , most likely at its top, although another location cannot be completely ruled out.