The Subunit δ-Subunit b Domain of the Escherichia coli F1F0 ATPase

The δ and b subunits are both involved in binding the F1 to the F0 part in theEscherichia coli ATP synthase (ECF1F0). The interaction of the purified δ subunit and the isolated hydrophilic domain of the b subunit (bsol) has been studied here. Purified δ binds to bsol weakly in solution, as indicated by NMR studies and protease protection experiments. On F1,i.e. in the presence of ECF1-δ, δ, and bsol interact strongly, and a complex of ECF1·bsol can be isolated by native gel electrophoresis. Both δ subunit and bsol are protected from trypsin cleavage in this complex. In contrast, the δ subunit is rapidly degraded by the protease when bound to ECF1 when bsol is absent. The interaction of bsol with ECF1 involves the C-terminal domain of δ as δ(1–134) cannot replace intact δ in the binding experiments. As purified, bsol is a stable dimer with 80% α helix. A monomeric form of bsol can be obtained by introducing the mutation A128D (Howitt, S. M., Rodgers, A. J.,W., Jeffrey, P. D., and Cox, G. B. (1996) J. Biol. Chem.271, 7038–7042). Monomeric bsol has less α helix,i.e. only 58%, is much more sensitive to trypsin cleavage than dimer, and unfolds at much lower temperatures than the dimer in circular dichroism melting studies, indicating a less stable structure. The bsol dimer, but not monomer, binds to δ in ECF1. To examine whether subunit b is a monomor or dimer in intact ECF1F0, CuCl2 was used to induce cross-link formation in the mutants bS60C, bQ104C, bA128C, bG131C, and bS146C. With the exception of bS60C, CuCl2 treatment resulted in formation of b subunit dimers in all mutants. Cross-linking yield was independent of nucleotide conditions and did not affect ATPase activity. These results show the b subunit to be dimeric for a large portion of the C terminus, with residues 124–131 likely forming a pair of parallel α helices.

An F 1 F 0 type ATPase is located in mitochondrial, chloroplast, and bacterial membranes, where it catalyzes the terminal step in oxidative-and photo-phosphorylation. In Escherichia coli, the enzyme contains five different subunits in the F 1 part, ␣, ␤, ␥, ␦, and ⑀, in the stoichiometry 3:3:1:1:1, and three different subunits in the F 0 part, a, b and c, in the ratio 1:2:9 -12. The F 1 part contains three catalytic sites on ␤ subunits and is an ATPase when released from the F 0 , while the F 0 part forms a proton pore (1)(2)(3).
It is now generally accepted that the F 1 and F 0 parts are joined by a relatively narrow stalk of 40 -45 Å in length (4,5) that is constituted by the ␥ and ⑀ subunits (3,6). Two other subunits, ␦ of the F 1 part and the two copies of the b subunit of the F 0 part, are also involved in binding the F 1 to the F 0 (7,8). They do not appear to be a part of the main stalk, and it has been suggested that they form a second connection, a stator that fixes the ␣ 3 ␤ 3 subdomain to the a subunit to allow rotation of a ␥-⑀-c subunit subdomain during functioning (9).
We have recently obtained a structure for a major part of the ␦ subunit by NMR (9). The polypeptide forms two domains. The N-terminal domain, composed of residues 1-105, is a six-helix bundle. The C-terminal domain of residues 106 -176 contains at least one ␣ helix (residues 118 -129), which can pack into a cleft in the N-terminal part, but this domain is partly unfolded in the isolated ␦ subunit. Our recent cross-linking studies in ECF 1 F 0 1 (10) and the work of Lill et al. (11) in chloroplasts place the ␦ subunit near the top of the F 1 on the outside of the complex, interacting with one of the three ␣ subunits.
How the b subunits are arranged in ECF 1 F 0 is less well understood. The b subunits of bacterial F 1 F 0 -ATPase, and the equivalent polypeptides in the mitochondrial enzyme, are characterized by an N-terminal membrane intercalated region and a large hydrophilic C-terminal domain (12). Based on trypsin digestion studies (7,13,14), it is this cytoplasmic domain of the b subunit that is involved in the binding of F 1 to F 0 . A truncated form of the b subunit containing the C-terminal domain has been obtained genetically and purified (15,16). The construct generated by Dunn (15) included residues  and an Nterminal octapeptide extension derived from the vector pUC8. The polypeptide produced by Howitt et al. (16), purified initially as a C-terminal fusion with glutathione S-transferase, includes residues 29 -156, with Gly-Ser introduced at the N terminus to allow cleavage of the b subunit cytoplasmic domain from the fusion by thrombin. Both forms of the C-terminal domain, here called b sol as proposed by Dunn (15), are highly soluble in aqueous buffer. Sedimentation velocity centrifugation and circular dichroism spectroscopy studies show that b sol is an elongated dimer with a high ␣-helical content (15,16).
Recent mutational analysis has established that residues Val 124 , Ala 128 , and Gly 131 lie on one face of an ␣ helix formed by a conserved hydrophobic region (Val 124 to Ala 132 ) near the C terminus of the b subunit (16). The mutation bG131D had previously been shown to allow assembly of the b subunit into the membrane but to prevent assembly of the whole F 1 F 0 ⅐ATPase complex (17). Replacement of Ala 128 by Asp is now known to disrupt dimer formation: b sol with the mutation A128D is a stable monomer (16). Whether the dimeric structure seen in b sol is representative of the structural organization of the b subunit in the F 1 F 0 complex is not known.
Here, we have explored the arrangement of ␦ and the b subunits in ECF 1 F 0 in some detail. Clear evidence that the b subunit binds to F 1 via the ␦ subunit is reported, and it is shown that the dimer is both required for, and represents, the arrangement of subunit b in the complex.
Media and Growth of Organisms-Bacterial strains used in the overexpression of glutathione S-transferase fusion proteins (16) were grown in Luria broth supplemented with 33 mM glycerol. Wild-type and mutant (A128D) forms of b sol were purified as described previously (16), using glutathione-linked agarose resin and thrombin obtained from Sigma. Strains used for the preparation of F 1 F 0 -ATPase were grown in minimal medium with supplements (21).
CD Spectroscopy-Circular dichroism spectra were obtained using a Jasco J-720 spectropolarimeter. For the determination of the relative percentages of secondary structure, wild-type b subunit cytoplasmic domain (0.242 g/liter) and the corresponding mutant A128D (0.255 g/liter) were dissolved in 5 mM MgSO 4 , 10 mM sodium phosphate, pH 7.0. The concentration of protein was determined using amino acid analysis. The samples were placed in 0.1-mm path length cells at 20°C, and spectra were acquired using a scan speed of 20 nm/min, response time of 1 s, bandwidth of 1 nm, and resolution of 0.5 nm. For each sample, 10 acquisitions were collected between 260 and 200 nm and co-added, and 50 acquisitions were collected between 200 and 182 nm and co-added. The two sets were concatenated and the spectrum of the buffer, collected in the same way, was subtracted. The raw data were converted to mean residue ellipticity ([] MRW ) using a mean residue weight of 109.5 calculated from the amino acid sequence. Data were analyzed using the computer program VARSLC1 (22) to obtain estimates of the percentages of ␣ helix, parallel and antiparallel ␤ sheet, turns, and "other" structure.
For the temperature dependence of unfolding, both wild-type cytoplasmic domain (0.21 g/liter) and the mutant A128D (0.114 g/liter) were placed in 1-mm path length cells. The cells were placed in the jacketed cell holder of the spectropolarimeter. The temperature in the cell was controlled through the Jasco software and a Neslab RTE-111 water bath. The water bath temperature was raised at a rate of 1°C/min over the range 7 to 95°C, and data were collected at 222 nm every 0.2°C as a measure of the loss of ␣ helix. Corrections for the difference in the temperature between the water bath and the cell were made based upon previous calibrations using a thermocouple probe placed in the cell.
Plots of [] MRW at 222 nm were analyzed assuming a two-state transition between the native state and the unfolded state. The melting temperature, t m , was estimated using nonlinear regression of the plots with the following equation where the parameters 0 and ϱ are the starting and ending ellipticity values, respectively, a and b are the slopes of the initial and final linear parts of the plots, respectively, d is the dispersion of the data, and t is the temperature at a given ellipticity, t . The fitting program returned estimates of the values of the parameters (including t m ) Ϯ the standard errors.
CuCl 2 Induced Cross-linking of ECF 1 F 0 -ECF 1 F 0 was isolated and reconstituted into egg-lecithin vesicles as described in Aggeler et al. (23). ATP synthase-containing vesicles collected from the Sephadex G-50 column were pelleted by centrifugation for 1 h at 45,000 rpm at 4°C in a Beckman Ti60 rotor. The pellets were resuspended in 50 mM Tris-HCl, pH 7.5, 2 mM MgCl 2 , 2 mM dithiothreitol, and 10% glycerol and stored in liquid nitrogen. Prior to cross-linking experiments with CuCl 2 , the reducing agent was removed by pelleting the enzyme at 175,000 ϫ g for 30 min at 4°C in a Beckman TLA100.2 rotor. The sample was then washed twice by successive resuspension and centrifugation steps in 50 mM MOPS, pH 7.0, 2 mM MgCl 2 , and 10% glycerol. Final resuspension was in the same buffer at a protein concentration of 1 mg/ml. Cross-linking was carried out at 22°C for 2 h using concentrations of between 5 M and 200 M CuCl 2 . All cross-linking reactions were stopped by the addition of 5 mM EDTA, and ATPase activities (24) were measured with and without prior incubation of the samples with 20 mM dithiothreitol for 2 h at 22°C.
Trypsin Cleavage Experiments-In experiments comparing cleavage patterns of wild-type and mutant b sol , these proteins (1 mg/ml) were prepared in 50 mM Tris-HCl, pH 7.0, 1 mM EDTA, and 2 M glucose. The ratio of trypsin to b sol (w/w) was 1:3000. The reaction, conducted at 13°C, was stopped by adding 4 mM phenylmethylsulfonyl fluoride from a freshly prepared stock solution. Trypsin cleavage products were analyzed using 16% SDS-polyacrylamide gel electrophoresis gels prepared and run according to the method of Schä gger and von Jagow (25). N-terminal sequences of tryptic fragments excised from Coomassie stained SDS-polyacrylamide gel electrophoresis gels were determined by Dr. Denis Shaw, Australian National University, Canberra, after passive transfer from the gel pieces onto polyvinylidene difluoride membrane. Experiments comparing cleavage of b sol , the ␦ subunit and ECF 1 , alone or in combination, were carried out in 50 mM Tris-HCl, pH 7.4, 1

FIG. 1. Analysis of binding of the b subunit cytoplasmic domain (b sol ) to ECF 1 and ECF 1 (-␦).
Mixtures of polypeptides were first analyzed by native agarose gel electrophoresis (A) through a gel containing 100 mM glycine, 10 mM Tris acetate, pH 7.0, 86 g/liter sucrose, and 1% agarose. Lanes from left to right:  . Bands were excised from the agarose gel and separated on a 10 -18% gradient of polyacrylamide in SDS (B). The individual lanes 1-9 in the SDS-polyacrylamide gel are for the bands indicated in the agarose gel. mM EDTA, 10% glycerol. In experiments comparing cleavage of b sol and ECF 1 in the presence and absence of b sol , the ratio of trypsin to b sol (w/w) was 1:5000. In experiments comparing cleavage of b sol and ␦ in the presence and absence of b sol , the ratio of trypsin to b sol (w/w) was 1:300. These reactions were conducted at 22°C and were stopped by adding 4 mM phenylmethylsulfonyl fluoride.
NMR Studies-The ␦ subunit and ␦ (1-134) were produced from pJCI kindly provided by Dr. Stanley Dunn (University of Western Ontario). Polypeptides were purified, and NMR spectra were obtained as described previously (9).
Other Methods-Atebrin fluorescence quenching activities were assayed as described by Hatch et al. (26). Protein concentrations were determined using the BCA protein assay from Pierce, with bovine serum albumin as a standard. F 1 -ATPase was prepared from membranes of strain AN1460 (27) as described by Cox et al. (28). Stripped membranes were prepared from strain AN2840 (29) using the method described by Wise et al. (30).

RESULTS
The b sol used here was obtained by thrombin release from a fusion protein that includes glutathione S-transferase at the N terminus (16). It contains residues 25-156 with a Gly-Ser Nterminal extension. Both intact ␦ and ␦ (1-134) were obtained as described previously (9). All of the isolated subunits and subunit fragments were pure, based on SDS-polyacrylamide gel electrophoresis (Fig. 1). F 1 -␦-b sol Interactions Determined by Native Gel Electrophoresis-Interactions between F 1 -␦, ␦, ␦  , and b sol were examined first by native gel electrophoresis. In a typical experiment, such as shown in Fig. 1A, fractions were mixed and then electrophoresed through agarose. Protein was detected on the native gel by Coomassie Brilliant Blue staining. Bands containing protein were excised, dissolved in SDS, and then subjected to SDS-polyacrylamide gel electrophoresis (Fig. 1B). As shown in Fig. 1A, F 1 -␦, ␦, ␦  , and b sol each migrated very differently based on the combination of size and charge, while complexes of F 1 -␦, F 1 ϩ ␦, F 1 ϩ ␦  , and even F 1 ϩ ␦ ϩ b sol migrated similarly.
No complex formation was observed between b sol and either pure ␦ or ␦ (1-134) in native gel electrophoresis (Fig. 1A, lane IV). However, both ␦ and ␦ (1-134) bound to F 1 -␦ to form a complex that was retained through the electrophoresis step (Fig. 1A,  lanes VII and IX). Moreover, a complex could be formed be-  (9)). Resonances for residues 135-176 of the ␦ subunit have not been assigned. B, the same sample as above plus 0.14 mM b sol . tween F 1 -␦, ␦, and b sol (Fig. 1A, lane VIII), but not between F 1 -␦ and b sol (Fig. 1A, lane VI), or between F 1 ϩ ␦ (1-134) ϩ b sol (Fig.  1A, lane X). These results indicate that b sol binds tightly only to F 1 when ␦ is present, and that the C-terminal 43 residues are important for this interaction.
Subunit ␦-b sol Interactions Based on NMR Studies-The native gel electrophoresis experiments rule out a strong interaction between ␦ and b sol in the absence of F 1 , but do not preclude that there was weak interaction between the two subunits that was destabilized by the electrophoresis step. We have recently reported a structure determination of the ␦ subunit by NMR (9). All of the backbone resonances of residues 1-134 were assigned. As a second approach to examining interactions of ␦ and b sol in solution, pure 15 N-labeled ␦ subunit was titrated with purified b sol , and the resonance broadening of the ␦ sub-unit spectrum was monitored. Many of the individual resonances of the ␦ subunit progressively disappeared as the ratio of b sol to ␦ was increased (Fig. 2) as clearly evident for Ser 23 , Gly 72 , Gly 83 , and Asn 86 , four residues that are widely distributed in the N-terminal domain. These results indicate that an interaction occurs between the two polypeptides with a slow to intermediate rate of exchange between a free ␦ subunit and ␦ that is bound to b sol . The binding constant between ␦ and b sol from these NMR studies must be larger than 2 M. This is consistent with the interaction between the two polypeptides being weaker in solution than on F 1 , where a binding affinity of 2 M was estimated by Dunn (15). The disappearance of resonances, then, represents the lowing of the tumbling rate of the ␦ subunit by binding of b sol . The approach does not pinpoint the site of interaction, except that when the experiment was repeated with ␦ (1-134) , there was significantly less broadening of resonances of the N-terminal part, as would be expected if the interaction between ␦ and b sol is mainly through the C-terminal 42 residues of the ␦ subunit. Other individual subunits of ECF 1 failed to alter the NMR spectrum of the ␦.
Interaction of ␦ and b sol on F 1 Protects Both Polypeptides from Protease Digestion-Both the ␦ subunit and b sol are highly Atebrin fluorescence quenching was measured as described previously (21). Atebrin was added to give a final concentration of 4 M, NADH to 2 mM, NaCN to 2.5 mM, ATP to 1 mM, and carbonylcyanide m-chlorophenylhydrazone (CCCP) to 2 M. 100 l of F 1 -ATPase preparation (16 mg/ml) was used for each reconstitution. A, 500 l of wild-type form of b sol (1.5 mg/ml) was incubated with F 1 -ATPase prior to the assay. B, 500 l of the mutant (A128D) form of b sol (1.5 mg/ml) were incubated with F 1 -ATPase prior to the assay. C, control in which 500 l of buffer alone was incubated with F 1 -ATPase prior to the reconstitution assay. protease sensitive in pure form (shown for b sol in Fig. 3A). Moreover, the ␦ subunit is still highly protease-sensitive when bound to the core ECF 1 (Fig. 3C). However, as shown in Fig.  3B, when b sol is reacted with ECF 1 , both the ␦ subunit and b sol are protected from trypsin cleavage. These results confirm the role of the ␦ subunit in binding b sol . Fig. 3E presents data for the trypsin cleavage of the ␦⅐b sol complex formed in solution. It is evident that there is some protection from cleavage of both ␦ and b sol by complex formation, but this interaction is clearly weak as proteolysis still proceeds quite rapidly.
Dimer of b sol Is Required for Its Binding to F 1 -To examine if b sol was binding to F 1 at its functionally important site, the ability of the polypeptide to block the reconstitution of coupled F 1 F 0 ATPase activity was measured by atebrin fluorescence quenching assays. In the absence of added b sol , rebinding of F 1 to the membrane preparation gave ATP-dependent quenching that was 53% of the original fluorescence level (Fig. 4C) compared with 51% after preincubation of ECF 1 with monomeric b sol at the same concentration (Fig. 4B). When b sol was incubated with F 1 and then the complex reconstituted, ATP-dependent quenching was only 12% of the original fluorescence level (Fig. 4A). Thus, the binding of b sol blocks the functional rebinding of F 1 to F 0 .
Dimerization Stabilizes the Structure of b sol -A CD spectrum of the wild-type (dimeric) form of the b sol over the wavelength range 182-260 nm is presented in Fig. 5A. The data indicate 80% Ϯ 3% (SD) ␣ helix. The CD spectrum of the mutant (A128D) form of b sol is also shown in Fig. 5A, from which 58 Ϯ 2% ␣-helix was estimated. The loss of ␣-helical structure observed for the mutant protein is compensated largely by an increase in "other" (random coil) structure and parallel ␤ sheet.
The trypsin digestion patterns of the dimeric and monomeric forms of b sol are additional evidence of loss of secondary structure in the monomer. Thus, with trypsin to b sol at a ratio of 1:5000 (cf. 1:3000 in Fig. 3) as well as with 2 M glucose present to stabilize the protein, there was only a single cut of b sol in 30 min, the product of which had the N-terminal sequence QKE-IAD (cutting after residue Arg 36 ). Under equivalent conditions, the monomeric protein was cleaved at Arg 36 , Arg 49 , Lys 58 , Arg 83 , and Lys 100 , indicating increased accessibility of the protease, not only in the vicinity of the mutation but at a number of points throughout the protein (results not shown).
The increased susceptibility of monomeric b sol to proteolytic cleavage correlates with decreased thermal stability as shown by CD melting studies. CD scans at a fixed wavelength of 222 nm over the temperature range 6 -82°C (Fig. 5B) showed a clear secondary structure transition with midpoint at 40°C for the wild-type dimeric b sol . In this transition the polypeptide goes from a folded conformation to an unstructured state. The unfolding of the mutant monomeric b sol is centered at a significantly lower temperature (32°C).

Arrangement of b Subunits in the ECF 1 F 0 Complex Based on Disulfide Cross-link Formation by Mutant b Subunits-
The stability of the dimer, and importance of the dimer in binding to F 1 , supports the idea that b is a dimer in the ECF 1 F 0 complex. To test this more directly, mutants were constructed in which a cysteine was introduced in place of the following b subunit residues, Ser 60 , Gln 104 , Ala 128 , Gly 131 , and Ser 146 . With the exception of the mutant bS60C, which grew poorly, all of the strains grew well on solid succinate minimal medium. All of the strains exhibited growth yields in limiting (5 mM) glucose similar to the wild-type control (data not shown). The specific activities of ECF 1 F 0 purified from each of the mutants were similar to the wild-type enzyme, i.e. in the range 22-27 mol of ATP hydrolyzed/min/mg. Mutant enzyme preparations were treated with CuCl 2 to induce formation of disulfide bonds. Experiments were carried out under both ATP conditions (following preincubation with AMP⅐PNP) and ADP conditions (addition of ATP ϩ Mg 2ϩ followed by enzyme turnover), over a range of CuCl 2 concentrations (5-200 M). A cross-linked product in the range 32-35 kDa was observed with ECF 1 F 0 from mutants Q104C, A128C, G131C, and S146C. No cross linking was obtained with mutant S60C. Data for the mutant bQ104C under ATP conditions are shown in Fig. 6A in the form of a concentration dependence of CuCl 2 on cross-linking yield. The maximal yields obtained with this mutant were in the range of 70%. Fig. 6B shows data for all mutants using the minimum CuCl 2 concentration required for maximum cross-link formation under ATP conditions. The appearance of cross-linked product seen in Fig. 6, A and B, was in each case accompanied by a commensurate loss of the b subunit band. That this new band was a b subunit dimer was confirmed by antibody blotting. It reacted with anti-b subunit mAbs but not with antibodies to any of the other subunits of the complex (data not shown). Note the altered migration of both the monomeric and cross-linked dimeric forms of the mutant S146C b subunit. There was no nucleotide dependence of the yield of cross-linking for any of the mutants examined. Subunit b dimer formation had only a small effect on ATPase activity with any of the mutants, e.g. for bQ104C, ATPase activity was 25.1 mol/min/mg before and 19.2 mol/min/mg after cross-linking in 70% yield. This is the same loss of ATPase activity obtained by treating wild-type enzyme with CuCl 2 .

DISCUSSION
There is emerging evidence that the F 1 and F 0 parts of the ATP synthase are joined not only by the narrow and 40 -45-Å long stalk seen in electron micrographs (5), and now known to be made up of the ␥ and ⑀ subunits (3), but also by a second stalk provided by the ␦ and b subunits (9). There is clear evidence from cross-linking experiments that the ␦ subunit is bound at the outside of the ␣ 3 ␤ 3 subunit barrel near the top of the F 1 and away from the F 0 (10,11). The ␦ subunit is a two-domain protein, with an N-terminal domain of around 105 residues, which is roughly globular and contains a six ␣-helical bundle (9) and a less ordered C-terminal domain. The C-terminal domain must be in close contact with the N-terminal domain in the complex as a cross link is readily formed between intrinsic Cys 64 and Cys 140 (31). A number of studies have shown that the ␦ subunit and its equivalent in the mitochondrial enzyme, OSCP, are involved in the interaction of F 1 with the F 0 part. For example, ECF 1 does not bind to F 0 in the absence of ␦. The key sites for this interaction appear to be in the C terminus of ␦. C-terminal truncations of as few as 4 -6 residues from either ␦ or OSCP prevent the rebinding of ECF 1 or MF 1 , respectively, to F 0 (32,33).
The studies presented here show conclusively that the ␦ subunit binds to b subunits, and that this interaction involves mainly the C-terminal domain of ␦. This arrangement had been speculated upon based on cross-linking of ␦ to subunit CF 0 I in the chloroplast enzyme (34). However, the data reported here are the first direct binding experiments. It is shown that ␦ subunit is required for the interaction of the cytoplasmic domain of subunit b with F 1 . This binding requires the C-terminal domain of ␦, as the interaction was lost when ␦  was used in the reconstitution experiments. A complex was obtained between F 1 ϩ ␦ ϩ b sol which was stable to native gel electrophoresis. No such stable complex was formed between ␦ and b sol in the absence of the F 1 . It is likely that the ␣ 3 ␤ 3 ␥ domain helps stabilize the interactions between the N-and C-terminal domains of ␦, which are required for the tight bind- Both the monomeric and cross-linked dimeric forms of the mutant S146Cb subunit migrate more rapidly in the gel than equivalents from the other mutants or wild-type (see lane 5). Preliminary experiments using monoclonal antibodies raised against the b and ␦ subunits have established that the band in this lane marked with an arrow is a novel cross-link product containing these proteins.
ing of b subunits. Weak interaction between ␦ and b sol in solution was observed by NMR and in protease digestion studies. Both the ␦ and b sol are protected from trypsin digestion in the ECF 1 ϩ ␦ ϩ b sol complex, while the protection of ␦ and b sol by mixing the two in solution is much less. The binding of b sol blocks rebinding of F 1 ϩ ␦ to F 0 , consistent with the functional interaction between the cytoplasmic domain of b and ␦ in the reconstitution experiments.
Purified wild-type b sol is a stable dimer, and only a dimer, not monomer, is able to block F 1 ϩ ␦ binding to F 0 . A priori, the two copies of subunit b could provide separate connections between the F 1 and F 0 parts. However, the results presented here argue against this. First, dimer formation appears to stabilize the secondary structure of b sol . In the monomer form created by introducing an Asp for Gly at position 128, b sol is much less ␣-helical, because it is highly protease sensitive and denatures at lower temperatures.
More direct evidence that b subunits are close in F 1 F 0 was sought by cross-linking studies. Mutants were created with Cys residues at several sites along the C-terminal domain. On the addition of the oxidizing agent Cu 2ϩ , disulfide bonds were generated in high yield between Cys at positions 104, 128, 131, and 146, but not at 60. This could happen only if the two b subunits are paired for a significant length. As the ␦ subunit is near the top of the F 1 (10,11), the obvious arrangement of the b subunits is an extended one running up one side of the F 1 F 0 structure. In several of the mutants described here, it is possible to react the introduced Cys with bulky maleimides. This should allow us to label the b subunits, e.g. with gold particles, and then visualize them in side views of F 1 F 0 by cryoelectron microscopy.