Rotation of a γ-ε Subunit Domain in the Escherichia coli F1F0-ATP Synthase Complex

A triple mutant of Escherichia coliF1F0-ATP synthase, αQ2C/αS411C/εS108C, has been generated for studying movements of the γ and ε subunits during functioning of the enzyme. It includes mutations that allow disulfide bond formation between the Cys at α411 and both Cys-87 of γ and Cys-108 of ε, two covalent cross-links that block enzyme function (Aggeler, R., and Capaldi, R. A. (1996) J. Biol. Chem. 271, 13888–13891). A cross-link is also generated between the Cys at α2 and Cys-140 of the δ subunit, which has no effect on functioning (Ogilvie, I., Aggeler, R., and Capaldi, R. A. (1997) J. Biol. Chem. 272, 16652–16656). CuCl2 treatment of the mutant αQ2C/αS411C/εS108C generated five major cross-linked products. These are α-γ-δ, α-γ, α-δ-ε, α-δ, and α-ε. The ratio of α-γ-δ to the α-γ product was close to 1:2, i.e. in one-third of the ECF1F0 molecules the γ subunit was attached to the α subunit at which the δ subunit is bound. Also, 20% of the ε subunit was present as a α-δ-ε product. With regard to the δ subunit, 30% was in the α-γ-δ, 20% in the α-δ-ε, and 50% in the α-δ products when the cross-linking was done after incubation in ATP + MgCl2. The amounts of these three products were 40, 22, and 38%, respectively, in experiments where Cu2+ was added after preincubation in ATP + Mg2+ + azide. The δ subunit is fixed to, and therefore identifies, one specific α subunit (αδ). A distribution of the γ and ε subunits, which is essentially random with respect to the α subunits, can only be explained by rotation of γ-ε relative to the α3β3 domain in ECF1F0.

F 1 F 0 -type ATPases are found in the plasma membrane of bacteria, the inner membrane of mitochondria, and the thylakoid membrane of chloroplasts. These enzymes can both use a proton gradient to synthesize ATP and in the reverse direction hydrolyze ATP to establish a proton gradient for subsequent substrate and ion transport processes (1)(2)(3). The F 1 part of the enzyme from Escherichia coli, ECF 1 F 0 , is composed of ␣, ␤, ␥, ␦, and ⑀ subunits in the stoichiometry 3:3:1:1:1. This part is linked by a narrow stalk to the F 0 part that is composed of a, b, and c subunits in the stoichiometry 1:2:9 -12 (3)(4)(5)(6). The stalk contains a part of the ␥ and ⑀ subunits (6,7). Two other subunits, ␦ and b, are required for linkage of the F 1 and F 0 parts. These two subunits may provide a second separate connection (8,9).
Electron microscopy (10) and, more recently, a high resolution structure of the mitochondrial F 1 (11) have shown that the ␣ and ␤ subunits alternate in a hexagonal arrangement around a central cavity. These two large subunits have a similar fold, each with three domains, an N-terminal ␤ barrel domain on top and away from the F 0 , a middle nucleotide-binding domain, and a C-terminal ␣-helical domain (11). Three catalytic sites are present and located predominantly on ␤ subunits. The other three nucleotide binding sites on the ␣ subunits appear to have mostly a structural role. A part of the ␥ subunit is found within the ␣ 3 ␤ 3 barrel and organized as two ␣ helices. A third short ␣ helical segment of this subunit was also resolved in the crystal structure of mitochondrial F 1 (11). This lies under the C-terminal domain of one of the ␤ subunits where it interacts with the so-called "DELSEED" region (the sequence of residues of this part).
The ⑀ subunit in ECF 1 is a two-domain protein (12). The N-terminal 10-stranded ␤ barrel region interacts with the ␥ subunit through the length of the stalk (7,13) and with the c subunit oligomer at one end (14). The helix-loop-helix C-terminal domain of ⑀ lies under the ␣ 3 ␤ 3 subunit barrel and interacts with the DELSEED region (15) of a different ␤ subunit from the one that binds to the short ␣ helix of ␥ (16).
There is now considerable evidence that F 1 F 0 -type ATPases are highly cooperative enzymes with all three catalytic sites involved (2,17). This cooperativity of both ATP synthesis and ATP hydrolysis is currently best described by the alternating site model (18). In this model at any time during functioning, one catalytic site is involved in the bond cleavage reaction and is essentially closed; a second is opening to release the tightly bound product, and the third is closing to trap the substrate.
An important tenet of the binding change mechanism is that catalytic sites are sequentially linked to the proton channel for energy coupling by a rotation of the small subunits. Early evidence for such rotation came from cryoelectron microscopy studies on ECF 1 that showed the ␥ subunit distributed at all three ␤ subunits rather than fixed at one site (19,20). Consistent with the idea of rotation of ␥ within the ␣ 3 ␤ 3 domain, cross-linking of this subunit to ␣ or ␤ subunits was found to fully inhibit the functioning of ECF 1 (16,21).
Additional evidence for rotation of the ␥ subunit relative to the ␣ 3 ␤ 3 domain has been provided more recently by Duncan et al. (22). These authors isolated a complex containing a (unlabeled) ␤ and ␥ subunit, stably linked by a disulfide bond between Cys-87 of ␥ and a Cys introduced at position 380 of ␤. They mixed this ␤-␥ complex with a 35 S-labeled ␤ subunit along with ␣, ␦, and ⑀ subunits to regenerate a functional F 1 . This reconstituted enzyme was then shown to undergo subunit switching when the disulfide bond was broken and MgCl 2 ϩ ATP was added, i.e. the ␥ moved from unlabeled to labeled ␤ subunits. Sabbert et al. (23) have also provided evidence of * This research was supported by National Institutes of Health Grant HL 24526. 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.
However, all of the above studies have focused on the movements of the ␥ subunit in F 1 , and in terms of functional relevance it remains important to show that rotation of the ␥ subunit in conjunction with other subunits (e.g. the ⑀ subunit) occurs in the intact F 1 F 0 . Also, for an understanding of the coupling mechanisms, it is necessary to establish which subunits are moving with the ␥ subunit and which are fixed with respect to the ␣ 3 ␤ 3 subdomain. One attempt to examine rotation in ECF 1 F 0 has been reported recently by Cross and colleagues (25) using the same cross-linking and reconstitution approach they used before for F 1 . They showed an ATP-driven and dicyclohexylcarbodiimide-sensitive scrambling of ␥ relative to ␤ subunits, although the level of this scrambling was lower than would be expected if all enzyme molecules were active.
We have previously observed that disulfide bonds can be formed from an ␣ subunit via a Cys at position 2 to the ␦ subunit (in the mutant ␣Q2C (26)) and via a Cys at residue 411 to the ␥ and ⑀ subunits (in the mutant ␣S411C/⑀S108C (21)). Here, we report cross-linking studies with ECF 1 F 0 from the mutant ␣Q2C/␣S411C/⑀S108C where cross-linking from the ␣ subunit to all three small subunits can be obtained at the same time. The combinations of cross-linked products obtained provide information about which subunits have to be moving in the ATP synthase and which do not.
Other Methods-ECF 1 F 0 was reconstituted in egg lecithin on a Sephadex G-50 column (medium, 1.5 ϫ 60 cm) in 50 mM Tris, pH 7.5, 2 mM MgCl 2 , 2 mM DTT 1 and 10% glycerol, and cross-linking of the enzyme was carried out with CuCl 2 in 50 mM MOPS, pH 7.0, 2 mM MgCl 2 , 2 mM ATP, 10% glycerol as described by Aggeler et al. (16). Cross-linked products were separated by electrophoresis on SDS-containing polyacrylamide gels according to Laemmli (33). Two-dimensional SDS-polyacrylamide gel electrophoresis was carried out by resolving cross-linked products in a first dimension without prior treatment with reducing agents on an 8% polyacrylamide gel. A portion of a lane was cut out and exposed to 50 mM DTT in dissociation buffer for 2 h at room temperature. The gel piece was rotated 90°and positioned with agarose on a stacking gel of a 10 -18% polyacrylamide gel for the second dimension. Protein concentration was determined with the BCA protein assay from Pierce. Gels were stained with Coomassie Brilliant Blue R (34). Crosslinked products were identified with Western blotting, using monoclonal antibodies against F 1 subunits (35).

RESULTS
The experiments described here utilize the mutant ␣Q2C/ ␣S411C/⑀S108C. The ATP hydrolysis rates of this mutant were in the range of 25-30 mol of ATP hydrolyzed per min per mg, which is the same as for wild-type enzyme. Also, the ECF 1 F 0 displayed efficient proton pumping activity as determined by ATP-dependent 9-amino-6-chloro-2-methoxyacridine fluorescence quenching in inner membranes (result not shown). Therefore, the introduction of three different mutations did not disrupt the functioning of the enzyme. Previous work with ECF 1 and ECF 1 F 0 from the mutant ␣S411C/⑀S108C has shown that CuCl 2 treatment induces cross-linking between the Cys at 411 of one ␣ subunit and the intrinsic Cys-87 of the ␥ subunit and between the Cys-411 of a second ␣ subunit and Cys-108 of ⑀ (21). Cross-linking of ␣ to ␥, as obtained in an ␣S411C mutant, or of ␣ to ⑀ obtained at low CuCl 2 concentrations in the ␣S411C/⑀S108C mutant fully inhibited ATPase activity. Studies using the ECF 1 and ECF 1 F 0 from the mutant ␣Q2C have shown that CuCl 2 causes full yield cross-linking between an ␣ and the ␦ subunit involving the Cys at position 2 of ␣ and the two intrinsic Cys residues of ␦, i.e. Cys-64 and Cys-140. The cross-linking of ␣ to ␦ is without significant effect on either ATPase activity or ATP-coupled proton translocation as measured by the 9-amino-6-chloro-2-methoxyacridine fluorescence quenching method (26).
The effect of CuCl 2 treatment on ECF 1 F 0 isolated from the triple mutant ␣Q2C/␣S411C/⑀S108C is shown in Fig. 1. These experiments were conducted using 50 or 200 M CuCl 2 with incubation for 1 h at room temperature. All of the subunits of the complex are resolved in Fig. 1A. The ␦ and ⑀ subunits are each essentially missing, and the ␥ subunit is much reduced in the gel, having been cross-linked with the ␣ subunit. The band of the ␣ subunit is also greatly reduced (by around 85%). In Fig.  1B, the cross-linked products are optimally resolved. Each was identified by immunoblotting with monoclonal antibodies against each of the subunits. Five main cross-linked products were generated; in order of decreasing apparent molecular weight, these are ␣-␦-␥, ␣-␥, ␣-␦-⑀, ␣-␦, and ␣-⑀, respectively.  2 and 4). 60 mM N-ethylmaleimide was added and followed by dissociation buffer without reducing agent. 53-g samples were loaded on 10 -18% polyacrylamide gels. B, 120-l aliquots were treated as above, and 26 g of protein was loaded on a 8% polyacrylamide gel. The gels in panels A and B are stained with Coomassie Brilliant Blue. C, 53 g of cross-linked protein as in lane 2 was applied on a 5-cm wide lane on a 8% polyacrylamide gel. After transfer onto a nitrocellulose membrane monoclonal antibodies against the F 1 subunits were used (␦ is shown).
Other cross-linked products were obtained but only in small amounts. These include an ␣-␣ product that has also been resolved after cross-linking of the ␣Q2C mutant and probably involves a disulfide bridge from Cys-2 of one ␣ and Cys-90 of a second ␣ subunit.
In several different experiments the cross-linked products involving the ␦ subunit, i.e. ␣-␦-␥, ␣-␦-⑀, and ␣-␦, appeared to be in almost equal amounts based on immunostaining of the Western blots with the anti-␦ monoclonal antibody (e.g. Fig. 1C). To better quantitate the yields of these cross-linked products, CuCl 2 -treated enzyme was subjected to two-dimensional SDSpolyacrylamide gel electrophoresis with separation in the first dimension in the absence of DTT and in the second dimension in the presence of DTT. A typical result is shown in Fig. 2A. It is clear from these data that the amounts of ␥, ␦, and ⑀ not in the major products identified in the figure are very small, i.e. less than 5% for each subunit. A scan of the three peaks of ␦ subunit released from ␣-␦-␥, ␣-␦-⑀, and ␣-␦ products by DTT dissociation is shown in Fig. 2B. The relative percentage of these three peaks was 30, 20, and 50% (averages of two experiments), respectively. Similar quantitation of percentages of the ␥ subunit in the ␣-␦-␥ and ␣-␥ bands gave 31 and 69%, respectively, and for the ⑀ subunit in the ␣-␦-⑀ and ␣-⑀ gave 19 and 81%, respectively. The experiment in Fig. 2 involved crosslinking of ECF 1 F 0 after preincubation in MgCl 2 ϩ ATP. When cross-linking was performed after preincubation in MgCl 2 ϩ NaN 3 ϩ ATP the distribution of ␣-␦-␥, ␣-␦-⑀, and ␣-␦ was 40.3 Ϯ 1.0, 21.7 Ϯ 0.9, and 38.0 Ϯ 0.2% (averages of three experiments), respectively.

DISCUSSION
The question of whether there is rotation of the ␥ subunit between the three ␣-␤ pairs during the functioning of F 1 has been convincingly answered as reviewed in the Introduction. ATP hydrolysis clearly drives a movement of the ␥ subunit such that it visits all three ␣-␤ subunit pairs during cooperative catalysis (23,24). A priori, this rotational movement could be an artifact of a freedom of the ␥ subunit that is allowed only when the F 1 is dissociated from the F 0 . However, the results of Zhou et al. (25) and the data presented in this study are evidence that this is not the case. The key observations here are that in ECF 1 F 0 from the mutant ␣Q2C/␣S411C/⑀S108C there is cross-linking of ␥ and ⑀ subunits each separately to the same ␣ subunit that binds the ␦ subunit (Fig. 3). However, the ␥ and ⑀ subunits are never bound to the same ␣ subunit. The significance of these results is clear when the activity effects of cross-linking of the ␥, ␦, or ⑀ subunits to ␣ subunits are considered. Covalent cross-linking of the ␦ to an ␣ subunit has been found to have little or no effect on either cooperative ATP hydrolysis or on the proton pumping function of ECF 1 F 0 (26). This is in contrast to the cross-linking of ␥ or ⑀ to ␣ subunits, which completely blocks functioning (21). The conclusion from these activity data is that movements of ␥ and ⑀ but not ␦ are an essential part of the functioning of the enzyme. It follows that the ␦ subunit must be fixed with its interaction thereby identifying one of the three ␣ subunits (␣ ␦ ) that can be visited by both ␥ and ⑀ since both ␣-␦-␥ and ␣-␦-⑀ cross-linked products were obtained.
The distribution of cross-linked products observed is understandable when it is considered that the experiments reported here involve a population of ECF 1 F 0 molecules that are not synchronized. Thus, at any time during enzyme turnover after ATP hydrolysis has stopped, around one-third of any rotating subunits should be at each of the three ␣-␤ pairs. This is approximately the observed distribution of both ␥ and ⑀ subunits in our experiments whether enzyme activity was ended by substrate depletion or by addition of azide. In previous studies it was shown that cross-linking of ⑀ to the ␥ subunit did not block activity (7,27), indicating that these two subunits move together as a mobile domain. In electron microscopy studies (19), we have seen movements of the ␥ subunits relative to ⑀, but this is because the antibody fragments to ␣ and ⑀ used to tag specific subunits release ⑀ from ␥ so that it is fixed at one ␣-␤ subunit pair.
In summary, the scrambling of ␥ and ⑀ subunits with respect to the three ␣ subunits, one of which is clearly distinguished by interaction of the ␦ subunit, is evidence for rotational movements of the main stalk forming subunits in ECF 1 F 0 . The only other explanation of the data, that ECF 1 F 0 assembles with a fixed random distribution of the small subunits, does not seem tenable. The ␥ and ⑀ subunits appear to move as one domain, although there may be small movements of the two relative to one another as part of the coupling between catalytic sites and proton channel functioning (6).
For rotational movements of the ␥ and ⑀ subunits to occur within ECF 1 F 0 , the ␣ 3 ␤ 3 domain must be fixed relative to the F 0 part by a stator. Recent evidence suggests that this stator is contributed by the ␦ with the b subunits (9, 26, 36). For cou-FIG. 2. Resolution and quantitation of subunits involved in CuCl 2 induced cross-link of ECF 1 F 0 from ␣Q2C/␣S411C/⑀S108C by two-dimensional SDS-polyacrylamide gel electrophoresis. ECF 1 F 0 in egg lecithin vesicles in 50 mM MOPS, 2 mM MgCl 2 , 2 mM ATP, and 10% glycerol was cross-linked at a concentration of 0.46 mg/ml with 100 M CuCl 2 for 1 h at room temperature. The reaction was stopped with 7 mM EDTA, and 20 mM N-ethylmaleimide was added before the dissociation buffer without reducing agent. For the first dimension, 70 g of protein was applied on a 8% polyacrylamide gel. After electrophoresis the top 12.5 cm of the 18-cm long resolving gel of a lane was cut out, soaked in 12 ml of dissociation buffer with 50 mM DTT for 2 h, and placed on a 10 -18% polyacrylamide gel for the second dimension. A, the bands were visualized with Coomassie Brilliant Blue. B, relative intensities were determined by scanning the gel with Adobe Photoshop and NIH Image. The cross-linked products giving rise to the peaks are indicated. pling, ATP hydrolysis-driven movements of the ␥-⑀ domain must be linked to proton translocation. It has been established that both the ␥ and ⑀ (Refs. 7 and 14, respectively) interact directly with the c subunit oligomer of the F 0 subunit. The covalent cross-linking of ␥ or ⑀ to the c subunit ring does not block ATP hydrolysis (7,14), which implies that the rotatory element in ECF 1 F 0 is a ␥-⑀-c oligomer domain moving relative to the ␣ 3 -␤ 3 -␦-a-b 2 complex.