The F₀ complex of the proton-translocating F-type ATPase of Escherichia coli

The membrane-bound ATP synthase (F-type ATPase) of Escherichia coli (EFiF₀) catalyzes the synthesis of ATP in a reaction that is driven by an electrochemical gradient of protons generated by respiration. The enzyme can also function in the reverse direction, generating an electrochemical proton gradient by ATP hydrolysis. Other bacteria, mitochondria (MF₁F₀) and chloroplasts (CF₁F₀) contain similar enzyme complexes (Schneider and Altendorf, 1987; Fillingame, 1990; Senior, 1990). However, in Propionigenium modestum an F-type ATPase (PF₁F₀) is present, which utilizes an electrochemical gradient of sodium ions for the synthesis of ATP from ADP and inorganic phosphate (Laubinger and Dimroth, 1988). The enzyme complexes are composed of two structurally and functionally distinct entities, designated F₁ and F₀. The peripheral F₁ part carries the catalytic centres of the enzyme and can be easily dissociated from the membrane. It consists of five different subunits with a stoichiometry of α₃, β₃, γ₁, δ₁, ϵ₁. The membrane-integrated F_o complex functions as a proton channel. In E. coll the Fo complex (EF₀) is composed of the three subunits a, b and c, with relative molecular masses of 30276, 17265 and 8288, respectively, and a stoichiometry of 1:2:10±l for a:b:c. The EF₀ complex can be isolated and reconstituted into liposomes in a functional form. In addition, reconstitution is also possible starting from the three individual subunits. Furthermore, analyses of deletion strains and reconstitution experiments with purified EF₀ subunits have revealed that all three subunits are necessary to obtain an EF₀ complex functional in proton translocation and Fi binding (Schneider and Altendorf, 1987; and references therein).

Previous studies with mutant strains (Cox el al. 1981) led to the proposal that subunits α and βof EF1 are necessary for the assembly of EF₀. However, more recent experiments have revealed that plasmid-encoded EF₀ subunits form a functional EF₀ complex in the absence of coordinated synthesis of EFi (Fillingame et al. 1986), while a ‘nonconducting’ EFo complex is assembled in the absence of any EF1 subunit (Pati and Brusilow, 1989; Pati et al. 1991). In the latter case, addition of EF1 brings about the ‘education’ of EF₀ to translocate protons. These results indicate that, after binding of EF1i to membranes and the subsequent depletion of those EFi complexes, the EF₀ complex has been transformed from the closed state to an open proton-translocating channel (Pati and Brusilow, 1989). In addition, although reconstitution of EF₀ sectors expressed in the presence or absence of EF₁ subunits revealed major differences in their ability to translocate protons, the rebinding of EF₁ is similar in both cases (Pati et al. 1991).

Subunit c, also referred to as proteolipid, is the most extensively studied subunit of the EFo complex (Schneider and Altendorf, 1987; Fillingame, 1990; Senior, 1990). The protein has been sequenced from various sources and found to possess a number of common features. Subunit c consists of two hydrophobic stretches separated by a central polar loop (Fig. 1). Within the C-terminal hydrophobic region, a conserved acidic residue, Asp-61, exists; this reacts with dicyclohexylcarbodiimide (DCCD), resulting in an inhibition of H⁺ translocation through F_o. The view that this carboxyl group plays an important role is supported by the observation that proton translocation is abolished by substitution of Asp-61 with glycine or asparagine (Schneider and Altendorf, 1987; Fillingame, 1990; Senior, 1990; and references therein). Furthermore, studies with a slow-growing revertant strain, in which in addition to the original Asp-61→Gly mutation Ala-24 has been replaced by aspartic acid, indicated that a carboxyl group in the middle of the membrane is absolutely required for H⁺ translocation, but that it can be moved from the C-terminal hydrophobic helix to the N-terminal one (Miller et al. 1990).

Observations supporting a hairpin-like structure of the E. coli subunit c within the membrane (Fig. 1) are the following: (i) analyses of DCCD-resistant mutants indicate that lle-28 and Ala-24 are close to Asp-61; (ii) analyses of revenant strains, which show partial suppression of the original mutations affecting proton translocation, indicate that the amino acid residues Asp-61/Ala-24 and Pro-64/Ala-20 are close to each other; (iii) the hydrophobic regions Leu-4 to Leu-19 and Phe-53 to Phe-76 are labelled with 3-trifluoromethyl-3-(m-[^l25l]iodophenyl)diazirine (TID), suggesting that these regions are membrane-integrated; (iv) the C-terminal region is located at the periplasmic side of the membrane, as shown by chemical modification studies; and (v) studies with peptidespecific antibodies reveal that the polar loop region is exposed to the cytoplasm (Schneider and Altendorf, 1987; Fillingame, 1990; Senior, 1990; and references therein; Fillingame et al. 1991; Hensel et al. 1990; Miller et al. 1990).

Analyses of mutant strains revealed that point mutations in the hydrophobic regions of subunit c produced effects on the assembly of EF₀ or affected proton translocation, whereas the hydrophilic loop region seemed to function in the coupling of proton translocation through EF₀ to ATP synthesis in EF₁ (Schneider and Altendorf, 1987; Fillingame, 1990; Senior, 1990; and references therein). Although the amino acid residues Ala-40, Arg-41, Gln/Asn-42 and Pro-3 of the polar loop region are highly conserved in bacteria, mitochondria and chloroplasts, it can be concluded from oligodeoxynucleotide-directed mutagenesis studies that Arg-41 is the only residue in this region that is absolutely required for function (Fraga and Fillingame, 1991).

Expression of chloroplast subunit III (homologous to subunit c of E. coli) together with subunits a and b of E. coli did not produce an active F₀ hybrid complex (Burkovski et al. 1990), although immunoprecipitations with antibodies against CF₀III indicate that an F₀ complex is assembled (A. Burkovski, G. Deckers-Hebestreit and K. Altendorf, unpublished results). We are therefore constructing hybrid c subunits, consisting of parts of both proteolipids, to determine those parts of subunit c that are essential for a functional interplay with subunits a and b.

It has been proposed that subunit a, which is an extremely hydrophobic protein, is involved in proton translocation together with subunit c. The analyses of missense mutants, which exhibit a reduction in H⁺ translocation without a concomitant effect on the assembly of EF₀, revealed that the amino acid residues Arg-210, Glu-219 and His-245 are critical for proton translocation, whereas other residues, such as Asp-119, Glu-196, Gly-197, Ser-206 and Asn-214, are thought to affect proton translocation indirectly (Schneider and Altendorf, 1987; Fillingame, 1990; Senior, 1990; and references therein). In addition, characterization of the double mutant Glu-219→His/His-245→Glu, in which the function is less impaired than in each of the single mutants, has suggested a linked functioning of these two residues in subunit a (Cain and Simoni, 1988). Furthermore, the suggested involvement of Arg-210, Glu-219 and His-245 in proton translocation is strikingly similar to the ‘proton relay’ system proposed for the lac permease (Lee et al. 1989).

Subunit a exhibits structural homology to its counterpart in other F₁F₀ complexes (Fig. 1) with strong amino acid homology in two regions towards the C-terminal part (Pro-190 to Leu-220 and Phe-244 to Tyr-263). Models containing 5–8 membranespanning helices have been predicted, but convincing evidence favouring one of those is still lacking (Fillingame, 1990; and references therein; Lewis et al. 1990; Bjorbaek et al. 1990). In addition, studies with alkaline phosphatase (PhoA) fusion proteins resulted in controversial conclusions about the localization of the N and the C termini of the protein with regard to the membrane (Lewis et al. 1990; Bjorbaek et al. 1990).

However, the following experimental evidence is useful in evaluating the different models: (i) Arg-210, Glu-219 and His-245, which are proposed to be critical for H⁺ translocation, are located in transmembrane helices; (ii) analysis of the double mutant Glu-219→His/His-245→Glu indicated that both residues are close to each other in the native protein; (iii) a second-site suppressor mutation in subunit a (Pro-240→Ala or Leu) partially corrects defects of the Gly-9→Asp mutation in subunit b; (iv) analyses of subunit a/PhoA and subunit α/ β-galactosidase fusion proteins revealed that the N terminus of subunit a is located in the periplasm; (v) subunit o/PhoA fusion proteins exhibiting high phosphatase activities have their fusion joints located in two regions of subunit a: between Val-110 and Leu-136 and between Leu-229 and Ile-246. Both regions are predicted by hydropathy analyses to be intramembranous. Therefore, from the theory of alkaline phosphatase fusions, both regions should be membrane-spanning, coming from the periplasm and going to the cytoplasm (Schneider and Altendorf, 1987; Fillingame, 1990; Senior, 1990; and references therein; Bjorbaek et al. 1990; Lewis et al. 1990; Lewis and Simoni, 1992).

To determine the localization of both termini of subunit a, a small DNA fragment has now been inserted into the Acyl site at the 5′ or the 3′ end of the uncB gene producing subunit a proteins, which carry at Thr-7 or at Ala-267 of the polypeptide chain an additional epitope of 11 amino acids (Asp-Asn-Pro-Ala-Ser-Thr-Thr-Asn-Lys-Asp-Lys; Charbit et al. 1986). These modified uncB genes are still able to complement an uncB deletion strain (H. Jäger, G. Deckers-Hebestreit and K. Altendorf, unpublished results). Binding studies with a monoclonal antibody against this epitope and with membrane vesicles of different orientation are now under way to disclose the location of this epitope.

The primary sequence of subunit b suggests that it is a hydrophilic protein, which is anchored in the cytoplasmic membrane by its apolar N-terminal region (about 25 amino acids), whereas the hydrophilic, highly charged part protrudes from the membrane in a predominantly a-helical structure. Labelling experiments with TID, hydrophobic nitrenes and a hydrophobic maleimide support this view. The TID labelling pattern and the analyses of mutants in which Cys-21 or Lys-23 have been substituted indicate that the region around Cys-21 is exposed to the lipid phase with little or no contact with other polypeptide chains. In contrast, the characterization of revertant strains, which partially suppress the Gly-9→Asp mutation of subunit b, as described above, supports the notion that the region around Gly-9 is in close contact with subunits a and c. In addition, the absence of TID labelling in the region Asn-2 to Gin-10 also suggests that these residues are part of a protein contact area. Studies with proteases and subunit-b-specific antibodies revealed that the hydrophilic part is exposed to the cytoplasm (Fig. 1). However, although subunit b is protected from proteolysis when EF₁ is bound to EF₀, polyclonal anti-h antibodies are able to bind to subunit b even in the presence of EF₁. The binding produced a partial removal of EF₁ from EF₀, whereas the residual membrane-bound ATPase activity is less DCCD-sensitive (Schneider and Altendorf, 1987; Fillingame, 1990; Senior, 1990; and references therein; Kauffer et al. 1991; Deckers-Hebestreit et al. 1992).

Treatment of F₁-stripped everted membrane vesicles or F₀ liposomes with trypsin, which cleaves subunit b from the C terminus, resulted in impairment of Fi binding, whereas the proton translocation remained unaffected. In contrast, if the degradation of subunit b with trypsin occurs prior to the isolation and reconstitution of ‘F₀’, H⁺ translocation is also impaired. However, both functions can be restored by addition of intact subunit b (Steffens et al. 1987). In addition, cells carrying amber mutations within the uncF gene, which cause premature termination of subunit b biosynthesis, are not able to assemble a functional F₀ complex even when only Glu-155 and Leu-156 are missing at the C terminus (Porter et al. 1985; Takeyama et al. 1988). These studies support the view that an intact subunit b is necessary for the correct assembly of F₀, but that, once F₀ is formed, the C-terminal region is dispensable for proton translocation.

Very little information exists about the interplay of all three subunits (stoichiometry of aifocioii) in the formation of an F_o complex functional in proton translocation and Fi binding. However, subunit a has a strong cooperative effect on the structure of the 10± I copies of subunit c in the F_o complex, since the TID labelling pattern of subunit c is largely reduced when subunit c is reconstituted into liposomes together with subunit a or when it is present in the F₀ complex compared to only subunit c in the membrane (Steffens et al. 1988). Two different models have been proposed for the arrangement of all three subunits in the EF₀ complex. The first model (Fig. 2B,D) exhibits a ring of c subunits surrounding subunits a and b (Cox et al. 1986). However, this model does not fit very well with the TID labelling data. In contrast, Hoppe and Sebald (1986) and Schneider and Altendorf (1987) have suggested that the a₁b₁ complex is located outside a ring of c subunits and that it interacts only at one side with the subunit c oligomer (Fig. 2A,C). In addition, this model is much better at explaining the relative ease of extraction of subunit b from F_o (Schneider and Altendorf, 1984).

Application of the electron spectroscopic imaging mode in electron microscopy, which allows detection of small particles with optimal density and high resolution, made it possible to distinguish different regions of high density in EF₀ complexes dissolved in cholate-containing buffer. Although single subunits of EF₀ could not be identified, the overall shape of the single enzyme complexes fits much better with the F₀ model proposed by Schneider and Altendorf (1987) and Sebald and Hoppe (1986) (Fig. 2A,C). In addition, electron micrographs of EF₀ complexes immunodecorated with monoclonal antibodies against subunits a, b and c indicate a peripheral arrangement of all three subunits in an EF₀ complex seen from the F₁ binding site (R. Birkenhäger, M. Hoppert, G. Deckers-Hebestreit, F. Mayer and K. Altendorf, unpublished results).

A ‘proton relay’ system or ‘proton wire’, in which protons cross the membrane by sequential interaction with a series of charged amino acids, has been proposed as a mechanism of proton translocation (Nagle and Tristam-Nagle, 1983; Cox et al. 1986). Amino acid residues involved include Arg-210, Glu-219 and His-245 of subunit a and Asp-61 of subunit c. However, one copy of subunit a must interact with an oligomer of 10±l subunits c. Furthermore, one molecule of DCCD is able to abolish proton translocation through F₀ by reaction with just one copy of subunit c (Hermolin and Fillingame, 1989). Rotational models were therefore suggested to account for these features (Cox et al. 1986; Schneider and Altendorf, 1987). However, only protons can be moved by such a mechanism. Alternatively, a transmembrane water-filled channel formed by subunits a and c could be envisaged, which would allow specific H⁺ translocation, when equipped with an H⁺-specific filter or gate. Studies with amphiphatic peptides, which have been constructed to mimic ion channels, indicate that differences in the number of polar groups have an influence on the number of water molecules bound in the channel and therefore also have an influence on the ion specificity (Lear et al. 1988). In addition, bacteriorhodopsin-bound water molecules are also involved in H⁺ translocation (Henderson et al. 1990). For H⁺ translocation through F_o such an ion transport mechanism has to be favoured, especially in the light of the recently discovered F-type ATPase of Propionigenium modestum, which uses an electrochemical gradient of sodium ions for the synthesis of ATP instead of a proton gradient (Laubinger and Dimroth, 1988). Furthermore, in the presence of low sodium concentrations, reconstituted PF₁F₀ switches from a Na⁺-translocating to a H⁺-pumping mode (Laubinger and Dimroth, 1989).

The translocation rate for protons by EF₀ reconstituted into liposomes was measured to be 20 H⁺ per EF₀ per second, which is in agreement with values obtained using F|-stripped everted membrane vesicles from E. coli. In contrast, Schönknecht et al. (1986) and Lill et al. (1986) calculated the proton translocation rate through CF₀ in thylakoid membranes from chloroplasts to be between 6.2×10³ and 2X10⁵ H⁺ per CF₀ per second measured by flash kinetic spectrophotometry. However, these high rates could only be observed for a fraction of the CF₀ complexes available and only for a limited period after removal of CF₁ (Lill et al. 1986). In the case of E. coli F₀, the proton translocation rates could be increased up to 100-120 H⁺ per EF₀ per second by using, for the purification of EF₀, an E. coli strain that extensively overproduces the ATP synthase (Moriyama et al. 1991) and by applying the pH-indicating dye Phenol Red instead of using a pH electrode, which allows proton translocation measurements on a very short time scale (in the millisecond range) (G. Deckers-Hebestreit, G. Althoff, W. Junge and K. Altendorf, unpublished results). Although the apparent proton translocation rate of EF₀ has been increased at least by a factor of five, the discrepancy between the values obtained for CF₀ and EF₀ still exists. Therefore, the question remains as to whether the differences depend only on the experimental conditions used for the measurement of H⁺ translocation or whether the translocation rate in everted membranes and in EF₀ liposomes represents only some residual activity due to unrecognized structural changes within EF₀ caused during preparation or due to structural changes upon removal of F₁ (Nelson, 1980). Alternatively, the enzyme in E. coli may work much slower than CF₁F₀.