Characterization of the Interface between and Subunits of Escherichia coli F-ATPase

The interaction faces of the and subunits in the Escherichia coli F-ATPase have been explored by a combination of cross-linking and chemical modification experiments using several mutant subunits as follows: S10C, H38C, T43C, S65C, S108C, and M138C, along with a mutant of the subunit, T106C. The replacement of Ser-10 by a Cys or Met-138 by a Cys reduced the inhibition of ECF by the subunit, while the mutation S65C increased this inhibitory effect. Modification of the Cys at position 10 with N-ethylmaleimide or fluoroscein maleimide further reduced the binding affinity of, and the maximal inhibition by, the subunit. Similar chemical modification of the Cys at position 43 of the subunit (in the mutant T43C) and a Cys at position 106 of the subunit (T106C) also affected the inhibition of ECF by the subunit. The various subunit mutants were reacted with TFPAM3, and the site(s) of cross-linking within the ECF complex was determined. Previous studies have shown cross-linking from the Cys at positions 10 and 38 with the subunit and from a Cys at position 108 to an α subunit (Aggeler, R., Chicas-Cruz, K., Cai, S. X., Keana, J. F. W., and Capaldi, R. A.(1992) Biochemistry 31, 2956-2961; Aggeler, R., Weinreich, F., and Capaldi, R. A.(1995) Biochim. Biophys. Acta 1230, 62-68). Here, cross-linking was found from a Cys at position 43 to the subunit and from the Cys at position 138 to a β subunit. The site of cross-linking from Cys-10 of to the subunit was localized by peptide mapping to a region of the subunit between residues 222 and 242. Cross-linking from a Cys at position 38 and at position 43 was with the C-terminal part of the subunit, between residues 202 and 286. ECF treated with trypsin at pH 7.0 still binds purified subunit, while enzyme treated with the protease at pH 8.0 does not. This identifies sites around residue 70 and/or between 202 and 212 of the subunit as involved in subunit binding.

Proton translocating F 1 F 0 ATPases (also ATP synthases) are found in the membranes of bacteria, chloroplasts, and mitochondria, where they function in oxidative phosphorylation or photophosphorylation to couple a transmembrane proton electrochemical gradient to ATP synthesis. These complex enzymes are reversible and can use ATP hydrolysis to establish a proton gradient for subsequent use in substrate and ion transport processes (reviewed in Cross (1988), Senior (1990), and Boyer (1993)).
As the name implies, F 1 F 0 ATPases are made up of two parts, an extrinsic F 1 part that contains the catalytic sites and a membrane-intrinsic F 0 part that contains the proton channel. Cryo-electron microscopy shows these two parts separated by a stalk of 40 -45 Å in length (Gogol et al., 1987;Lü cken et al., 1990). In the bacterium Escherichia coli, the F 1 part (ECF 1 ) is made up of five different subunits (␣, ␤, ␥, ␦, and ⑀) in the molar ratio 3:3:1:1:1. The F 0 part contains three different subunits (a, b, and c) in the ratio 1:2:9 -12 (Senior, 1990;Fillingame, 1992).
The structure of the F 1 part in the enzyme from bovine mitochondria has been obtained recently to 2.8 Å resolution by x-ray crystallography (Abrahams et al., 1994). This shows three ␣ and three ␤ subunits arranged in a hexagon, with the ␥ subunit extending the length of the cavity within the hexagon, projecting from one end of the structure into an extension that is part of the stalk. The ␦ subunit of MF 1 , which is the equivalent of the ⑀ subunit in bacterial and chloroplast enzymes, is also in the stalk but, in the crystal form studied, is probably disordered and not resolved.
There is now considerable evidence that the ␥ and ⑀ subunits play an important role in coupling catalytic site events with proton translocation (e.g. Gogol et al., 1990;Nakamoto et al., 1993;Abrahams et al., 1994;Capaldi et al., 1994;Zhang et al., 1994) and that this coupling involves conformational changes and, probably, translocations of one or both subunits (Aggeler et al., , 1993Capaldi, 1994a, 1994b;Wilkens and Capaldi, 1994). It is important, therefore, to obtain the detailed structures of, and interaction sites between, these two subunits.
We have recently determined the structure of the ⑀ subunit in solution by NMR spectroscopy (Wilkens et al., 1995). Here, we describe experiments that identify the ␥ subunit binding site on the ⑀ subunit, and that locate regions of the ␥ subunit involved in the ⑀ subunit binding.

EXPERIMENTAL PROCEDURES
Enzyme Preparations-Mutants ⑀S10C, ⑀H38C, and ⑀S108C have been described previously Skakoon and Dunn, 1993;Aggeler et al., 1995a). the mutation ⑀M138C was created in the unc operon-containing plasmid, pRA100 , for isolation of ECF 1 and ECF 1 F 0 . Mutations T43C and S65C were introduced into the uncC gene on plasmid pEX2 (Skakoon and Dunn, 1993) for overexpression and isolation of the altered ⑀. Pure ⑀ subunit was isolated as described by Patel et al. (1990). ECF 1 depleted of ␦ and ⑀ subunits (ECF 1 *) was prepared by one passage through an anti-⑀ subunit monoclonal antibody affinity column according to Dunn (1986) but in a buffer containing 0.2% N,N-dimethyldodecylamine N-oxide, followed by one passage through a Sephacryl S200 (Pharmacia LKB Biotechnology Inc.) column to remove N,N-dimethyldodecylamine Noxide. Trypsin-treated ECF 1 was prepared as described by Tang et al. (1994), with a trypsin digestion time of 1 h for pH 7.0 samples and 4 h for pH 8.0 samples, respectively, followed by one Sephacryl S200 sizing column to separate proteolytically treated ECF 1 from other components. Mutants ␥S8C, ␥T106C, and ␥V286C are described by . ECF 1 isolated from these mutants was labeled with CM 1 or NEM as described by Tang et al. (1994) before being treated with trypsin in MOPS buffer (50 mM MOPS, pH 7.0, 0.5 mM EDTA, and 10% glycerol), plus additional 5 mM ATP at room temperature for 1 h. The monoclonal antibodies used in this study were prepared and characterized as described by Aggeler et al. (1990).
Titration of ECF 1 Preparations with ⑀ Subunit-ECF 1 * or trypsintreated enzyme for the mutant ␥T106C at 4 -5 g were pre-incubated with specified amounts of pure ⑀ subunit or maleimide-modified ⑀ subunit containing the mutation S10C in MOPS buffer containing additional ATP (5 mM) and Mg 2ϩ (5.5 mM). Samples were analyzed for ATPase activity at 37°C using an ATP regenerating system .
Cross-linking Experiments-ECF 1 isolated from mutants ⑀S10C and ⑀M138C was cross-linked with 200 M TFPAM3 under different nucleotide conditions, essentially as described by . Crosslinking with mutant ⑀ subunits ⑀H38C, ⑀T43C, and ⑀S65C was conducted by first reacting the isolated subunit with TFPAM3 in the dark, removing excess label by passage through one Sephadex G25 centrifuge column, binding of the subunit to trypsin-treated ECF 1 in MOPS buffer with 5 mM ATP for 0.5 h at room temperature, and then photolysis of the sample for 2 h. Samples were analyzed for cross-linking by SDSpolyacrylamide gel electrophoresis, and the subunits involved in crosslinked products were identified by immunoblotting as described before .
Labeling Studies-Purified ⑀ subunit from the mutant ⑀S10C (4 mg/ml) was reacted with 1 mM NEM or FM at room temperature for 30 min before 20 mM cysteine was added, and the labeled protein was isolated by a single passage through a Sephadex G25 centrifuge column.
Sucrose Gradient Centrifugation Experiments-ECF 1 treated with trypsin at pH 7.0 or pH 8.0 as described by Tang et al. (1994) was incubated with 10-fold excess of pure wild-type ⑀ subunit in 200 l of a buffer containing 50 mM MOPS (pH 7.0), 5 mM ATP, 5 mM P i , 5 mM Mg 2ϩ at room temperature for 0.5 h. Samples were applied onto a 10 -40% sucrose step gradient in the same buffer and run at 4°C for 18 h in a Beckman SW 50.1 rotor at 40,000 rpm. Fractions were collected from the bottom of the tubes and assayed for ATPase activity. The fractions containing ATPase activity were pooled, and sodium deoxycholate (0.3%) and trichloracetic acid (8%) were added to precipitate the protein. Immunoblotting following SDS-polyacrylamide gel electrophoresis was used to analyze the protein.
Mapping of Cross-linking Sites-ECF 1 from the mutant ⑀S10C that had been cross-linked by TFPAM3 was incubated with trypsin in MOPS buffer plus 5 mM ATP at room temperature for 1 h, and trypsin digestion was stopped by addition of freshly made phenylmethylsulfonyl fluoride (4 mM). The digestion products were loaded on a 10 -18% SDS-polyacrylamide gel for electrophoresis to separate the truncated ␥-⑀ cross-linked product from other polypeptides. The 19,000-dalton band was excised and electroeluted from the gel using an ISCO electrophoretic concentrator (Isco, Inc.). The cross-linked product was purified further by reverse-phase high pressure liquid chromatography using a Brownlee Aquapore RP-300 (C 18 , 2.1 ϫ 3.0 mm) column on an Applied Biosystems model 130A. Mobile phase A contained 0.1% trifluoroacetic acid, while mobile phase B had 0.08% trifluoroacetic acid and 70% acetonitrile. The gradient of mobile phase B was set to 0 -35% in the first 35 min and then 35-45% for 20 min before 45-75% in the last 15 min. Peak fractions were detected at 200 nm. The ␥-⑀ crosslinked product appeared at 58 min and was concentrated with a Speed-Vac Concentrator (Savant) before addition of an equal volume of a buffer containing 100 mM Tris (pH 8.1), 10% acetonitrile, and 1 mM CaCl 2 . Trypsin (0.6 g) was added to complete digestion at 37°C overnight. Tryptic fragments were isolated by using the same reverse-phase high pressure liquid chromatography system and conditions as mentioned above. Each fragment was identified by sequencing the first four N-terminal amino acids. The fragment containing the N terminus Ala-Met-Thr-Tyr was analyzed further. Peptide sequencing was performed on a protein-sequenator (Applied Biosystems model 470A) with a phenylthiohydantoin analyzer (Applied Biosystems model 120A).
Other Methods-Protein concentrations were measured using the BCA protein assay from Pierce. Coomassie Brilliant Blue R staining of gels was carried out as described by Downer et al. (1976).

RESULTS
A number of mutants were used in the present study. Mutants ⑀S10C , ⑀H38C (Skakoon and Dunn, 1993), ⑀S108C , and ⑀M138C (this study) were created in the unc operon containing plasmid pRA100 . These mutants each showed wild-type growth on limiting glucose. Mutants T43C and S65C were created in the plasmid pEX2 containing the uncC, i.e. ⑀ subunit, gene (Skakoon and Dunn, 1993). The inhibitory effect of the ⑀ subunit when bound to isolated ECF 1 (Sternweis and Smith, 1980) was used as a convenient measure of the binding of the ⑀ subunit to the core complex (␣ 3 ␤ 3 ␥). Table I summarizes the activity effects of the various ⑀ subunit mutations. It includes data for ECF 1 isolated from strains, in which the mutation was in the unc operon, and data from experiments in which wildtype ECF 1 has been depleted of endogenous ⑀ subunit and then reconstituted with an excess of the mutant ⑀ subunit. The ATPase activity of wild-type ECF 1 was 10 Ϯ 0.5 units/mg (mol of ATP hydrolyzed/min/mg of enzyme), under our assay conditions at pH 7.5, with 2 mM ATP and 5 mM Mg 2ϩ . The activities of ECF 1 isolated from the mutants ⑀H38C and ⑀S108C were very similar to that of wild-type, while that of enzyme from mutants ⑀S10C and ⑀M138C were around 4-and 2-fold higher than the wild-type, respectively. The ATPase activity of ␦ and ⑀-free ECF 1 was 51 Ϯ 3 units/mg, an activation of around 5-fold, which could be reduced to a basal of around 7 units/mg by addition of excess wild-type ⑀ subunit. Addition of purified ⑀ subunit carrying the H38C or S108C mutations had similar effects to the addition of wild-type subunits. In all cases, excess of the ⑀ subunit inhibited activity below that of isolated ECF 1 , suggesting that there is some loss of ⑀ subunit during isolation of the intact ECF 1 complex. Rebinding of ⑀S10C gave a minimal activity of 13 units/mg. Addition of a 9-fold excess of ⑀ subunit with the mutation T43C also failed to inhibit the core ECF 1 complex (␣ 3 ␤ 3 ␥) to levels found using wild-type, consistent with the mutant having a reduced affinity for the core complex. Addition of mutant ⑀S65C gave a higher inhibition, giving a final ATPase activity that was only 70% of wild-type enzyme. The activity data, therefore, indicate that mutations S10C and T43C reduce, while the mutation S65C increases, the inhibitory effect of the ⑀ subunit on the core ECF 1 complex.
The ⑀ mutants S10C and T43C were examined further in experiments in which the introduced Cys was reacted with various maleimides, and the effect of this modification on activity was monitored. Fig. 1 shows the concentration dependence of the inhibition of ATPase activity of the mutant ⑀S10C before and after modification with different maleimides. It can be seen that both the concentration for half-maximal inhibition and the absolute extent of inhibition of this mutant were altered significantly by NEM and dramatically by modification a Units are mol of ATP hydrolyzed per min. b In molar ratio. c The mutation has not been introduced into the unc operon. d The pure ⑀ subunit has not been overexpressed.
with FM. Modification of the mutant T43C with FM had a much smaller, although significant, effect on the binding affinity and maximal inhibition (results not shown). These results indicate that the regions of the enzyme around residue 10 and, to a lesser extent, around residue 43 are important for interaction of the ⑀ subunit with the core ECF 1 and the resulting inhibition of enzymatic activity. Cross-linking of the Mutant ⑀ Subunit in the ECF 1 Complex-Previous studies have shown that the ⑀ subunit can be cross-linked to the ␥ subunit by reaction of the Cys introduced at positions 10 or 38, with the maleimide group of the photoaffinity reagent, TFPAM, and subsequent UV photolysis to activate the tetrafluorophenylazide (TFPA) group to a nitrene (Aggeler et al., , 1995b). Using the same approach, the ⑀ subunit in the mutant ⑀S108C was found to cross-link to an ␣ subunit . Fig. 2 shows a cross-linking experiment involving ECF 1 from the mutant ⑀M138C. Reaction of the mutant with TFPAM3 generated one major cross-linked product of M r approximately 75,000, which monoclonal antibody blots (not shown) revealed to be a cross-linked product between the mutant ⑀ subunit and a ␤ subunit. As shown in Fig. 2, the yield of this cross-linked product was nucleotide dependent, highest with ADP in catalytic sites and low with ATP ϩ EDTA (or AMP⅐PNP ϩ Mg 2ϩ ) bound.
The location of residues 43 and 65 were examined by reacting ⑀ subunit with TFPAM3 in the dark, removing excess reagent and then binding to ECF 1 , treated to remove endogenous ⑀ subunit with trypsin, followed by photolysis to activate the TFPA group. A cross-linked product between the ⑀ and ␥ subunits was observed in ECF 1 containing the T43C mutation (see below). No cross-link was observed with the S65C mutant.
Localization of the Sites of Cross-linking from Residue 10 of ⑀ in the ␥ Subunit-The site(s) of TFPAM cross-linking from Cys-10 in the mutant ⑀S10C to the ␥ subunit was determined by peptide mapping and protein sequencing. The cross-link was formed as described previously in ECF 1  and then the enzyme treated with trypsin (1:25) for 1 h at room temperature. Trypsin cleavage under these conditions has been found to generate two fragments of the ␥ subunit, labeled as ␥ A and ␥ D (Tang et al., 1994). The protease treatment of the TFPAM-treated ECF 1 from mutant ⑀S10C (Fig. 3A) gives these same products by cleavage of non-cross-linked ␥ subunit and, in addition, generated a product of M r approximately 19,000, which reacted with mAbs to the C-terminal part of the ␥ sub- unit (Fig. 3B), tentatively identifying it as a cross-linked product of the N terminus of ⑀ (via Cys-10) and the ␥ D fragment (residues 202-286) of the ␥ subunit.
The approximate M r 19,000 band was excised from polyacrylamide gels, protein was collected by electroelution, purified by HPLC, and then digested further with trypsin. The small fragments, so generated, were separated by high pressure liquid chromatography, and the peaks were resolved and then analyzed by N-terminal amino acid sequencing. One peak contained a peptide of ⑀ containing Cys-10, present in equimolar amounts with a fragment of the ␥ subunit identified by its sequence as including residues 222-242.
As shown in Fig. 4, the sequence of the ⑀ fragment continued beyond Cys-10, as expected, because the cross-link involves the side chain of the Cys, and Edman degradation is not prevented. In contrast, sequencing of the ␥ fragment stopped abruptly before tyrosine 228, consistent with cross-linking via the tetrafluorophenylazide into the backbone of the polypeptide between residues 227 and 228. We have seen a similar insertion of TFPAM into the backbone of the ␤ subunit next to a trypto- FIG. 4. Sequencing to determine the site of cross-linking from ⑀S10C to the ␥ subunit. Complete trypsin digestion of the truncated ␥-⑀ cross-linked product generated a tryptic fragment containing the N-terminal sequence of the ⑀ subunit, i.e. AMTY. Further Edman degradation of this tryptic fragment allowed identification of two sequences, one the N-terminal amino acid sequence of ⑀ (upper plot) and the second identified as the sequence of the ␥ subunit beginning at residue ␥222.
FIG. 5. Localization of the site of TFPAM3 cross-linking between ⑀H38C and the ␥ subunit. Pure ⑀H38C subunit, labeled with TFPAM3 in the dark, was added to trypsin-treated and CM-labeled ECF 1 from the mutants ␥S8C and ␥V286C. After photolysis, the samples were analyzed by 10 -18% SDS-polyacrylamide gel electrophoresis. The gel was later visualized on a UV light box. Lanes 1 and 2, crosslinking using ECF 1 from the mutant ␥V286C; lanes 3 and 4, crosslinking using ECF 1 from the mutant ␥S8C. Lanes 1 and 3, unphotolyzed; lanes 2 and 4, photolyzed.
FIG. 6. Sucrose gradient centrifugation to measure the binding of ⑀ subunit to trypsin-treated ECF 1 . ECF 1 that had been treated with trypsin at pH 7.0 or 8.0 was incubated with a 10-fold molar excess of pure wild-type ⑀ subunit, and the mixtures were applied on a 10 -40% sucrose step gradient. Samples from the gradient were concentrated and then loaded onto a 10 -18% SDS-polyacrylamide gel for electrophoresis followed by electroblotting onto a polyvinylidene difluoride membrane. The Western blot was later analyzed with anti-⑀I mAb. Lane 1, pure ⑀ subunit (applied on the gel as control); lane 2, ECF 1 treated with trypsin at pH 8.0; lanes 3 and 4, ECF 1 (pH 8.0) ϩ ⑀; lane 5, ECF 1 , treated with trypsin at pH 7.0 ϩ ⑀.

FIG. 7. Effects of chemical modification of a Cys at position 106
of the ␥ subunit on ⑀ subunit inhibition. ECF 1 from the mutant ␥T106C modified with NEM or CM was digested with trypsin to remove the ⑀ subunit. Purified ⑀ subunit (wild type) was added at different molar ratios with respect to ECF 1 , and the ATPase activity was measured. phan residue in experiments to identify the interaction site between ␥ and ␤ (Aggeler et al., 1993). It appears that the TFPA moiety tends to stack against aromatic residues, such as tyrosines and tryptophans, in an orientation that causes insertion of the reactive nitrene into the backbone.
Identification of the Sites of Cross-linking from Residues 38 and 43 of ⑀ into the ␥ Subunit-Attempts to use the same general approach described above for ⑀S10C to identify interaction sites on ␥ from the Cys at residue 38 or 43 of the ⑀ subunit failed because of the low yield of cross-linking observed with these mutations, i.e. 10 -20% compared with 50% for the ⑀S10C mutant. Instead, the location of the cross-linking in these mutants was assessed by tagging different parts of the ␥ subunit for identification after the covalent linkage with the ⑀ subunit.
An experiment involving the mutant ⑀H38C is shown in Fig.  5. Isolated ⑀ subunit containing the mutation H38C was reacted with TFPAM3 in the dark, and then the modified ⑀ subunit reacted with ECF 1 that had been treated with trypsin to remove endogenous ␦ and ⑀ subunits and at the same time cleave ␥ to ␥ A and ␥ D . Two forms of ECF 1 were used, one containing the mutation ␥S8C, the other ␥V286C . These sites were labeled with CM to identify the N-terminal ␥ A fragment or C-terminal ␥ D fragment on gels. As shown in Fig. 5, cross-linking from the Cys at 38 to the ␥ subunit with TFPAM3 gave a product of approximate M r 22,000, which contained Cys-286 but not Cys-8 of the ␥ subunit and is, therefore, a covalent product of the ⑀ subunit at position 38 with the C-terminal region 202-286 of ␥.
Similar experiments were then used to localize the site of interaction of the ⑀ subunit from residue 43 with the ␥ subunit. These show cross-linking of the ⑀ fragment to the same Cterminal part of the ␥ subunit (results not shown).
Binding Sites for the ⑀ Subunit on the ␥ Subunit from Protease Digestion Studies-As described above, trypsin digestion of ECF 1 at pH 7.0 leads to generation of two fragments, ␥ A and ␥ D . Trypsin cleavage is more efficient at pH 8.0 when the ␥ subunit yields three fragments, ␥ C involving residues 1 to ap- FIG. 8. Structural models of the ␥ and the ⑀ subunits from ECF 1 . A, a recently refined structure of the N-terminal ␤ barrel domain of the ⑀ subunit shown down the axis through the ␤ sheet sandwich, kindly provided by Dr. Stephan Wilkens. The open labeled spheres are the ␣ carbons of the residues changed to Cys in the experiments described here. The shaded spheres are side chains of the residues 15,9,79,77,68, and 42 that provide a hydrophobic patch on the side of ⑀ that binds the ␥ subunit. B, hypothetical secondary structure of the ␥ subunit based on 1) the secondary structure prediction methods of Chou and Fasman (1978) and Garnier et al. (1978), 2) the crystal structure of bovine mitochondrial F 1 -ATPase; and 3) our previous trypsin digestion studies (Tang et al., 1994) along with chemical modification and cross-linking studies reported here. Column, ␣ helix; arrow, ␤ sheet; line, ␤ turn or undefined region. The residue number at the ends of the ␣ helix/␤ sheet are labeled, and the trypsin cleavage sites are identified along with residue Thr-106. proximately 70, ␥ B from residues 71 to about 202, and ␥ E , residues 212-286 (Tang et al., 1994). The fragment, or fragments, that includes 11 residues between 202 and 211 is not resolved on gels. Trypsin cleavage at pH 7.0 or 8.0 activates the ECF 1 complex to the same level by removing the ⑀ subunit (final activity of about 50 unit/mg, cf. 10 units/mg for untreated enzyme). Addition of a 6-fold excess of pure ⑀ was found to fully inhibit the activity of ECF 1 treated with trypsin at pH 7.0 (i.e. the activity was reduced to less than 10 units/mg). However, there was no inhibition of sample that had been treated with the protease at pH 8.0. The inability of enzyme cleaved by trypsin at pH 8.0 to be inhibited by isolated ⑀ subunit resulted from its low affinity for the ⑀ subunit, as shown by sedimentation studies in Fig. 6. Taken together, these data indicate that the cleavage of ␥ at residue 70 and/or cleavage of the subunit at 212 with probable removal of residues 202-211 alters ⑀ binding to ECF 1 .
Chemical Labeling of the Mutant ␥T106C Places This Residue in the ␥ Subunit Binding Region-Insertion of a Cys for Thr at position 106 of the ␥ subunit has no effect on the activity of ECF 1 . This introduced Cys residue could be reacted with NEM without affecting the binding of, or the inhibition by, the ⑀ subunit added to ⑀-free ECF 1 from the mutant (␥T106C). However, as shown in Fig. 7, reaction of the Cys at position 106 with the more bulky maleimide, CM, altered both the binding, as judged by the half-maximum concentration of ⑀ for inhibition, and the absolute level of inhibition, obtained on binding wild-type ⑀ subunit.

DISCUSSION
Studies presented here focus on the interaction faces of the ⑀ and the ␥ subunits for each other. The results are summarized in Fig. 8, which shows the recently obtained structure of the N-terminal domain of the ⑀ subunit involving residues 1-86 in part A (Wilkens et al., 1995) and the predicted folding pattern of the ␥ subunit in part B. Genetic studies have predicted that the ⑀ subunit is a 2-domain protein (Kuki et al., 1988), and this is evident in the NMR structure determination of the isolated ⑀ subunit in solution. Residues 1-86 form a 10-stranded ␤ barrel, or sandwich, while the C-terminal 48 residues are arranged as an ␣-helix-loop-␣-helix structure. In solution, the helix-loophelix domain binds back on the ␤ barrel at one end.
The ⑀ subunit interacts with the ␣ and ␤ subunits of the F 1 part through the C-terminal domain, as indicated by crosslinking from Ser-108 to Glu-381 in the DELSEED region of the ␤ subunit by 1-ethyl-3-[3-dimethylamino)propyl]carbodiimide (Dallmann et al., 1992), by disulfide bond formation from a Cys at position 108 of ⑀ to a Cys in place of Glu-381 in ␤ (Aggeler et al., 1995b), and by disulfide bond formation between Cys at 108 of ⑀ and a Cys at position 411 in the ␣ subunit (this is the equivalent residue to Glu-381 of ␤). 2 Finally, as shown here, there is cross-linking of the ⑀ subunit to a ␤ subunit by TF-PAM3 from a Cys replacing Met-138 at the C terminus of ⑀.
In addition to interacting with ␣ and ␤ subunits, the ⑀ subunit is now known to interact with the c subunits of the F 0 (Zhang et al., 1994Watts et al., 1995), and this linkage is via the opposite end of the ␤ sheet sandwich from that which binds the helix-loop-helix domain (Fig. 8A).
The ⑀ subunit also binds to the ␥ subunit in ECF 1 (Dunn, 1986;Skakoon and Dunn, 1993). The homologous subunit in chloroplasts (⑀) also binds to the ␥ subunit in CF 1 (Suss, 1986;Soteropoulos et al., 1994). The results presented here identify Ser-10, His-38, and Thr-43 as close to, or involved in, this reaction based on mutagenesis to Cys and then cross-linking from these sites and/or by chemical modification of the introduced Cys residue and consequent steric effects on ⑀ binding. These three residues are on one face of the ␤ barrel and, as shown in Fig. 8, are close to a patch of hydrophobic residues, which may play a key role in the binding to the ␥ subunit.
The cross-linking results presented here indicate a role of residues between 202 and around 240 in ⑀ subunit binding. TFPAM cross-linking from the Cys-10 of ⑀ is with a tyrosine at residue 228 of the ␥ subunit. Cross-linking from Cys-38 or Cys-43 in the mutants ⑀H38C and ⑀T43C is within the region of ␥ from 202 to 286. The C terminus from residues 222-286 is organized as a long ␣ helix that, from around residue 240 to the very C terminus, is intercalated within the cavity formed by the hexagonally arranged ␣ and ␤ subunits (Abrahams et al., 1994). Our data indicate that the ⑀ subunit binds at, or close to, this ␣ helix as it extends from the ␣ 3 ␤ 3 barrel through the stalk region and makes contact with the c subunits of the F 0 part somewhere between residues 202 and 229 (Watts et al., 1995).
The protease digestion data focus attention on the region of ␥ around residue 70, as well as on the region between residues 202 and 212 in binding the ⑀ subunit. Residue 70 is in the epitope for a monoclonal antibody to the ␥ subunit, described by Dunn and colleagues, which reacts with ECF 1 only when the ⑀ subunit is first removed (Skakoon and Dunn, 1993). Residues 202-212 are close to the region of the ␥ subunit in which there is an insertion in the ␥ subunit of chloroplasts that contains two Cys residues, which regulate CF 1 activity by reduction and oxidation reactions involving the protein thioredoxin (reviewed by Soteropoulos et al. (1994)). The ⑀ subunit has been shown to affect the oxidation and reduction reaction of these Cys residues and protect this region from proteolytic digestion in CF 1 (Schumann et al., 1985).
In addition to the region around residue 70 and the Cterminal part from residues 202 to around 240, our chemical modification studies also place Thr-106 of ␥ in the ⑀ subunit binding site. The recent x-ray structure determination shows that the ␥ subunit has three segments in contact with ␣ and ␤ subunits, an N-terminal ␣-helix of residues 1-50, a short central ␣-helical residue region 83-99 in the numbering system of E. coli, as well as the C-terminal ␣-helix (Abrahams et al., 1994). It is noteworthy that the remainder of this subunit, including several of the sites identified here as in, or close to, the ⑀ binding site, is predicted to be in the ␤ sheet and turn structure. It seems likely, then, that the ␤ barrel of the ⑀ subunit binds in part to an equivalent structure formed by much of the region of the ␥ subunit, as well as binding to the extension of the C-terminal helix region.