Nucleotide-dependent movement of the epsilon subunit between alpha and beta subunits in the Escherichia coli F1F0-type ATPase.

Mutants of ECF1-ATPase were generated, containing cysteine residues in one or more of the following positions: αSer-411, βGlu-381, and εSer-108, after which disulfide bridges could be created by CuCl2 induced oxidation in high yield between α and ε, β and ε, α and γ, β and γ (endogenous Cys-87), and α and β. All of these cross-links lead to inhibition of ATP hydrolysis activity. In the two double mutants, containing a cysteine in εSer-108 along with either the DELSEED region of β (Glu-381) or the homologous region in α (Ser-411), there was a clear nucleotide dependence of the cross-link formation with the ε subunit. In βE381C/εS108C the β-ε cross-link was obtained preferentially when Mg2+ and ADP + Pi (addition of MgCl2 + ATP) was present, while the α-ε cross-link product was strongly favored in the αS411C/εS108C mutant in the Mg2+ ATP state (addition of MgCl2 + 5′-adenylyl-β,γ-imidodiphosphate). In the triple mutant αS411C/βE381C/εS108C, the ε subunit bound to the β subunit in Mg2+-ADP and to the α subunit in Mg2+-ATP, indicating a significant movement of this subunit. The γ subunit cross-linked to the β subunit in higher yield in Mg2+-ATP than in Mg2+-ADP, and when possible, i.e. in the triple mutant, always preferred the interaction with the β over the α subunit.

There has been significant recent progress in determining the structure of the F 1 part of the F 1 F 0 -type ATP synthase, a key enzyme in oxidative phosphorylation and photo-phosphorylation (Senior 1988(Senior , 1990Boyer, 1993). The F 1 , which can be detached from the F 0 and studied separately, is a complex of five different types of subunits ␣, ␤, ␥, ␦, and ⑀ in the molar ratio 3:3:1:1:1. Electron microscopy has shown that the ␣ and ␤ subunits are hexagonally arranged, and alternate around a central cavity in which the ␥ subunit is located (Gogol et al. 1989a(Gogol et al. , 1989bBoekema and Böttcher, 1992). Biochemical studies place the ␦ and ⑀ subunits (nomenclature for the Escherichia coli enzyme) at the bottom of the ␣ 3 ␤ 3 ␥ core complex (Beckers et al., 1992;Dallmann et al., 1992;Capaldi et al., 1994Capaldi et al., , 1995 in the stalk region which is a 40 -45-Å long structure that links the F 1 to the F 0 part (Gogol et al., 1987;Lü cken et al., 1990). The recently published high resolution structure of a major part of the F 1 molecule confirms the above described arrangement of the ␣, ␤, and ␥ subunits relative to one another and adds important details (Abrahams et al., 1994). The ␣ and ␤ subunits have a similar overall fold, each made up of three domains: an NH 2 -terminal, predominantly ␤-sheet domain, a nucleotide-binding domain of 9 ␤ strands and 9 ␣ helices, and a COOH-terminal, predominantly ␣-helical domain, that provides the binding site for the ⑀ subunit, probably the ␦ subunit, as well as the b subunit of the F 0 . The part of the ␥ subunit within the ring of ␣ and ␤ subunits is arranged as two ␣ helices: a long COOH-terminal helix extending from the top NH 2 -terminal domain of the ␣ and ␤ subunits into the stalk region, and a short NH 2 -terminal helix running from the catalytic site domain into the stalk. These two ␣ helices form a coiled coil. A third short ␣-helix of the ␥ subunit (residues 82-99 in the E. coli enzyme) is inclined at about 45°to the two larger helices at the bottom of the F 1 as it becomes the stalk. The remainder of the ␥ subunit is unresolved in the structure, presumably because it was disordered in the crystal form. A key feature of F 1 from a functional standpoint is its asymmetry, identified earlier and clearly revealed in the x-ray structure determination. In the crystal form examined, three ␣-␤ pairs can be distinguished based upon nucleotide occupancy of catalytic sites and by interactions of the ␥ subunit. One ␤ subunit (␤ TP ), with Mg 2ϩ AMP-PNP 1 in its catalytic site, is linked to the short ␣-helix of the ␥ subunit via the COOH-terminal domain at the sequence DELSEED (which is highly conserved in all F 1 -ATPases). In a second ␤ subunit, the catalytic site is empty (␤ E ). The ␤ subunit and its partner ␣ subunit (␣ E ) have several contacts with the two long ␣-helices of the ␥ subunit. The catalytic site in the third ␣-␤ pair (␣ DP -␤ DP ) contains Mg 2ϩ -ADP. Overall, the interactions between the ␥ subunit and the ring of ␣ and ␤ subunits are few, and relatively nonspecific. The major contact is at the top of the molecule. Here, the COOHterminal helix of ␥ is slotted into the continuous ring provided by hydrophobic loops under the ␤-sheet region of the six ␣ and ␤ subunits. Abrahams et al. (1994) have likened this region to a molecular bearing which could allow the ␥ subunit to rotate relative to the ␣ and ␤ subunits. Evidence for movements of the ␥ subunit relative to ␣-␤ subunit pairs has been accumulating for several years, beginning with the electron microscopy studies of Gogol et al. (1990) and has more recently included crosslinking  and fluorescence data . There is also clear evidence of movements of the ⑀ subunit during the working of the enzyme (Mendel-Hartvig and Capaldi, 1991;Aggeler et al., 1992Aggeler et al., , 1995. Here, we describe experiments showing that the ⑀ subunit can be bound at either ␣ or ␤ subunits, a switching that would appear to require significant rotational movements of this subunit within the ␣-␤ subunit ring.

EXPERIMENTAL PROCEDURES
Construction of Plasmids Containing Mutations in uncA, uncD, and uncC-Site-directed mutagenesis was carried out according to Kunkel et al. (1987) using CJ236 (New England Biolabs). For routine cloning procedures (Maniatis et al., 1982;Davis et al., 1986) XL1-Blue (Strat-* This work was supported by National Institutes of Health Grant HL24526. 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. ‡ To whom correspondence should be addressed. Tel.: 541-346-5881; Fax: 541-346-4854. agene) was used.
The mutation in the ␣ subunit at position 411 replacing a serine residue with cysteine was created by using the oligonucleotide TCAGTTTGCATGCGACCTTGA and M13mp18, that contained in the SmaI site the 986-base pair SnaBI fragment with the part of uncA encoding the COOH-terminal half of the ␣ subunit, obtained from pRA100 (Aggeler et al. 1992). Successful introduction of the mutation was determined by cleavage with restriction enzyme SphI. The mutation was then inserted in a plasmid containing the unc operon: (i) the 493-base pair PmlI/SfuI fragment was incorporated in pRA14 , containing the 5.8-kilobase XhoI/NsiI fragment of pRA102, which encodes the mutant ⑀ subunit with a cysteine replacing serine residue 108 (Aggeler et al., 1992); (ii) the 5.8-kilobase XhoI/NsiI fragment was inserted in pRA100 producing the plasmid pRA140, which contained the mutations ␣S411C and ⑀S108C. A double and triple mutant were created by inserting the 2.8-kilobase XhoI/SacI fragment of pRA140 in the plasmid pRA133 or pRA134 , to obtain pRA142 or pRA143 with the mutations ␣S411C/␤E381C and ␣S411C/␤E381C/⑀S108C, respectively.
Cross-linking with CuCl 2 -Mutant enzymes were obtained from AN888 (unc Ϫ ) containing plasmids pRA134, pRA140, pRA142, and pRA143, respectively. ECF 1 and ECF 1 F 0 were isolated as described by Gogol et al. (1989a) and Aggeler et al. (1987). Cross-linking between subunits was induced by CuCl 2 as described in Aggeler et al. (1995) in buffer containing 50 mM MOPS, pH 7.0, 10% glycerol, 2.0 -2.5 mM MgCl 2 , and 2 mM nucleotide. Cross-link products were analyzed by polyacrylamide gel electrophoresis (Laemmli, 1970). Gels were stained with Coomassie Brilliant Blue R (Downer et al., 1976). The identity of the subunits involved in cross-links was revealed by use of monoclonal antibodies on Western blots (Mendel-Hartvig and Capaldi, 1991). ATPase activity was measured in a regenerating system according to Lötscher et al. (1984). Protein concentrations were determined with the BCA protein assay from Pierce.
Cross-linking Studies with the Mutant ␤ E381C/⑀S108C-Previous studies had shown that addition of CuCl 2 to ECF 1 isolated from the mutant ␤E381C/⑀S108C catalyzed disulfide bond formation between Cys-381 of one ␤ subunit and the intrinsic Cys-87 of the ␥ subunit (see "Discussion"), as well as a cross-link between Cys-381 of a second ␤ subunit and Cys-108 of ⑀ . Either cross-link inhibited ATPase activity essentially fully. Additional cross-links were also formed in lower yield between an as yet unidentified intrinsic Cys of ␣ and the intrinsic Cys-140 of ␦, and between Cys-381 of the third ␤ subunit and Cys-140 of ␦. It was shown that neither of the cross-links involving the ␦ subunit affected activity. The cross-linking between ␤ and ⑀, and ␤ and ␥ subunits in the mutant ␤E381C/⑀S108C is both CuCl 2 concentration dependent and sensitive to which nucleotides are present in catalytic sites, as shown in Fig. 1. With Mg 2ϩ AMP-PNP (a non-cleavable analog of ATP) bound in the three catalytic sites (concentration 2 mM), the yield of cross-linking of ␤ to ⑀ is lower at all CuCl 2 concentrations than with Mg 2ϩ ADP ϩ P i in catalytic sites. At the same time, the yield of disulfide bond formation between ␤ and ␥ is higher in Mg 2ϩ AMP-PNP than in Mg 2ϩ ADP ϩ P i .
Cross-linking Studies with the Mutant ␣S411C/⑀S108C-The ␣ and ␤ subunits of F 1 -ATPases have a very similar fold (Abrahams et al., 1994) and residue 411 of the ␣ subunit is the equivalent of Glu-381 in the DELSEED region of the ␤ subunit. Treatment of ECF 1 from the mutant ␣S411C/⑀S108C with CuCl 2 induced disulfide bond formation between an ␣ subunit and the ⑀ subunit in high yield, along with a small amount of cross-linking between an ␣ and the ␥ subunit ( Fig. 2A). There was inhibition of ATPase activity in proportion to the yield of cross-linking (result not shown) as in the case of the mutant ␤E381C/⑀S108C . Fig. 2A shows the clear nucleotide dependence of the yield of cross-linking of ␣ to ⑀. The ⑀ subunit was linked to ␣ with much higher yield at any CuCl 2 concentration with Mg 2ϩ AMP-PNP in catalytic sites than when Mg 2ϩ ADP ϩ P i was present. This is the opposite of what
Cross-linking Studies with the Mutant ␣S411C/␤E381C/ ⑀S108C-The observations with mutants ␤E381C/⑀S108C and ␣S411C/⑀S108C suggest that the ⑀ subunit is able to move between two positions in a nucleotide dependent manner, one position close to ␣, the other close to a ␤ subunit. This was examined further in the mutant ␣S411C/␤E381C/⑀S108C. As shown in Fig. 3A, the predominant cross-linked product involving the ⑀ subunit is with the ␤ subunit in Mg 2ϩ ADP ϩ P i and with the ␣ subunit in Mg 2ϩ AMP-PNP. These nucleotide dependent interactions of the ⑀ subunit were clearly observed at 5 M CuCl 2 within 5 min. However, in order to obtain maximal cross-linking yields, incubation times longer than 30 min were necessary. CuCl 2 treatment of the mutant ␣S411C/␤E381C/ ⑀S108C also created a cross-link product between an ␣ and ␤ subunit that was not seen in either mutant ␣S411C/⑀S108C or ␤E381C/⑀S108C, indicating that it involves disulfide bond formation between the Cys-411 of ␣ and Cys-381 of ␤. This disulfide bond formation between the ␣ and ␤ subunits inhibited ATPase activity in proportion to yield, based on studies with the mutant ␣S411C/␤E381C in which this is the exclusive cross-linked product at low CuCl 2 concentrations (result not shown). At high CuCl 2 concentrations, it was possible to obtain very high yields of the ␣-␤ subunit product and, at the same time, to get essentially full cross-linking of ␥ and ⑀ in products with an ␣ or ␤ subunit (Fig. 3B). The only significant triple subunit cross-link observed involved ␣ ϩ ␤ and the ␦ subunit. Therefore, the ␣-␤ pair that becomes internally cross-linked is different from the ␣-␤ pair interacting with ⑀, or that pair interacting with the ␥ subunit. It may be the ␣-␤ pair that interacts internally is the one that has the ␦ subunit bound. However, it is not clear that the ␦ subunit is bound at its physiological binding site in isolated F 1 , which complicates the analysis.
The effect of CuCl 2 treatment on ECF 1 F 0 isolated from the mutant ␣S411C/␤E381C/⑀S108C is shown in Fig. 4A. The same ␣-␤ cross-linked product, along with cross-links involving ␣ and ␤ with ␥ and ⑀, were obtained in the intact ATP synthase that were seen in isolated ECF 1 . Moreover, the nucleotide dependent shifting of the ⑀ subunit between an ␣ and ␤ subunit was also observed in ECF 1 F 0 (Fig. 4, B and C).
In ECF 1 F 0 , there was less than full cross-linking of ␥ and ⑀ to ␣ and ␤ subunits. At the same time, there appeared to be a higher yield of the ␣-␤ cross-linked product than can be accounted for by only one ␣-␤ pair being involved. This suggests that there is competition between cross-linking of ⑀ and ␥ to ␣ and ␤ versus cross-linking between ␣ and ␤ subunits, which is in favor of the latter product in the intact ATP synthase.

DISCUSSION
The results of CuCl 2 -induced disulfide bond formation in mutants that have a Cys at position 381 of the ␤ subunit and at the equivalent position, 411, in the ␣ subunit, along with a Cys in the ⑀ subunit at position 108, provide important insight into the dynamics of the structure of ECF 1 and ECF 1 F 0 . They show clearly that the ⑀ subunit can move such that the COOHterminal domain of this subunit interacts with the equivalent regions of either an ␣ or ␤ subunit. The shifting of the ⑀ subunit between these subunits is nucleotide dependent. With Mg 2ϩ   FIG. 3. Cross-linking of ECF 1 -ATPase from ␣S411C/␤E381C/ ⑀S108C. A, ATPase was precipitated in 70% ammonium sulfate and passed through two centrifuge columns in Sephadex G-50 in 50 mM MOPS, pH 7.0, 0.5 mM EDTA, and 10% glycerol. ECF 1 at a concentration of 0.7 mg/ml in 50 mM MOPS, pH 7.0, 0.1 mM EDTA, and 10% glycerol was supplemented with 2.5 mM MgCl 2 ϩ 2 mM ATP (lanes 1, 3, 5, 7, and 9) or 2.5 mM MgCl 2 ϩ 2 mM AMP-PNP (lanes 2, 4, 6, 8, and 10). After incubation for 10 min at room temperature, samples were treated for 2 h at a CuCl 2 concentration of 1 M (lanes 3 and 4), 2 M (lanes 5 and 6), 5 M (lanes 7 and 8), and 10 M (lanes 9 and 10). Controls containing no CuCl 2 are shown in lanes 1 and 2. 32 g of protein was loaded per lane. B, ATPase at 0.55 mg/ml was incubated for 2 h after adding 2.5 mM MgCl 2 ϩ 2 mM ATP (lanes 1 and 3) and 2.5 mM MgCl 2 ϩ AMP-PNP with 100 M CuCl 2 (lanes 2 and 4). After cross-linking, no dithiothreitol (lanes 1 and 2) and 20 mM dithiothreitol (lanes 3 and 4), respectively, was added and incubation carried out for 3 h at room temperature. 40 g of protein was loaded per lane. ADP ϩ P i in catalytic sites, the linkage of ⑀ from the Cys at 108 is predominantly with Cys-381 of a ␤ subunit, while in Mg 2ϩ -ATP (by using Mg 2ϩ AMP-PNP) the ⑀ subunit is linked to the equivalent position in the ␣ subunit. The cross-linking results described here expand on our previous cryoelectron microscopy studies which had shown a shifting of the ⑀ subunit by 20 -25 Å between adjacent ␣ and ␤ subunits. In addition to interactions with ␣ and ␤ subunits, the ⑀ subunit is also bound to the ␥ subunit (Dunn, 1982;Aggeler et al., 1992) 2 and to the c subunit ring of the F 0 part of the ATP synthase complex (Zhang and Fillingame, 1995). We suggest, therefore, that the movements of the ⑀ subunit seen here, and in Wilkens and Capaldi (1994), are involved in energy coupling between the F 1 and F 0 parts.
Our data also show a nucleotide dependent movement of the ␥ subunit in concert with the movement of the ⑀ subunit. The ␥ subunit can be cross-linked to a ␤ subunit readily with ATP bound, less readily with Mg 2ϩ -ADP bound in catalytic sites. In what may be its less tightly bound state, in Mg 2ϩ -ADP, the ␥ subunit appears to be able to reside somewhat closer to an ␣ subunit and can be trapped there by prolonged incubation in high concentrations of CuCl 2 . This interaction between ␥ and ␤ subunits involves Cys-87 based on the close positioning of this residue and ␤ Glu-381 from the crystallographic data (Abrahams et al., 1994), and based on recent studies from Duncan et al. (1995) in which cross-linking between the ␤ (DELSEED region) and ␥ subunit was abolished in the mutant ECF 1 ␤D380C/␥C87S. Cross-linking between the ␣ and ␥ subunits most likely involves Cys-87 as well, but the possibility that Cys-112 is involved has not been ruled out.
In the mutant ␣S411C/␤E381C/⑀S108C, there is high yield cross-linking of ␣ to ␤ subunits by CuCl 2 treatment involving the Cys residues introduced into these subunits. This requires the rotation of the COOH-terminal ␣-helical region of ␣ and/or ␤ toward each other. That such a rotation can occur is evident in the recent high resolution structure of MF 1 (Abrahams et al., 1994). In the form of the enzyme crystallized, ␣ E is arranged differently with respect to ␣ DP or ␣ TP in that the region involving Ser-411 is rotated so that it is very close to the region of ␤ DP including E381, i.e. the distance between these two residues is 6.8 Å from ␣ backbone carbon to ␣ backbone carbon. The results presented here show that rotation of ␣ subunits with respect to ␤ subunits occurs not only in F 1 , but also in F 1 F 0 , implying that catalytic sites can be open at all times even in the presence of high concentrations of ATP or ADP.
In terms of the mechanism of cooperative ATP hydrolysis (and ATP synthesis) it is possible that the switching of the ⑀, and possibly the ␥ subunit, shown schematically in Fig. 5, is an oscillation between two adjacent subunit (pairs). If this switching is not back and forth, but continues clockwise or anticlockwise, the rotation of the small subunit(s) could then "alternate" the different nucleotide affinities of the three catalytic sites.