Main

The bacterium Escherichia coli contains the simplest form of the enzyme, EC F1FO. It comprises eight different subunits, five in the F1 part (α3, β3, γ, δ, ɛ) and three in the FO part (a1, b2, c9-12) (ref. 3). The proton pore is formed at the interface between the single-copy ‘a’ subunit and the ‘c’ subunits4, which are arranged as a ring5.

The γ subunit sits within a ring of alternating and hexagonally arranged α and β subunits in the F1 part6. It extends from F1 and, in association with the ɛ subunit, forms the 40-45-Å-long stalk that has been seen by electron microscopy to link the F1 and FO parts7. Both the γ and ɛ subunits bind to the c subunit ring, and the emerging mechanism of energy coupling1,8 involves the γ, ɛ and c subunit ring rotating as a unit relative to the fixed α3 β3 domain of the F1 and ab2 domain of FO.

Biochemical evidence supports the idea of a second connection9 between F1 and FO, but this feature had not been observed directly. Previous electron microscopy studies used membranous EC F1FO embedded in a thin layer of ice. These studies could have missed a narrow second stalk, because the density difference between protein and solvent water is low and consequently the images are noisy, so many must be averaged to see any features clearly. Also, the membranous nature of the sample often resulted in superimposition of individual complexes in projection, making interpretation of weak features more difficult.

We used detergent-solubilized EC F1FO, prepared and examined after negative staining to enhance the contrast between protein and solvent. Staining with heavy metals had previously been avoided because this treatment led to release of F1 from FO; however, we found that reaction of the enzyme with dicyclohexylcarbodiimide helped to prevent this release.

We observed two stalks linking the F1 to the FO (Fig. 1) in about 40% of all images obtained. The fatter, more central stalk, which had been observed previously, contains the γ and ɛ subunits. The second connector is at the side of the molecule, clearly evident extending down from the F1, and a corresponding density rises up from the FO. This second stalk probably includes both the δ and the b subunits. Density in the middle of the stalk is weak, as would be expected if this region derives from one or two α-helices from each of the b subunits.

Figure 1: Electron microscopy images of EC F1FO in side view.
figure 1

a-c, Different views based on image analysis and classification of a data set of 139 single molecules; a is a mirror image of c; b appears to be a projection at around 60° to that in a and c. d, Our interpretation of the arrangement and composition of the second stalk, and the extra density on top of the F1 molecule.

Full size image

At the top of the F1 is a cap that extends in the direction of the more asymmetrically placed stalk. Such a feature is lacking in the X-ray structure reported for bovine F1 (ref. 6). It is probable that this cap is formed by the amino-terminal 30 or so residues of the α subunits — which were unresolved in the X-ray experiments — together with a part of the δ subunit.

There is also a clear asymmetry of the FO part, consistent with a substructure in which the c subunits form a ring with the a and b subunits outside5,10. This asymmetry has been observed in enzyme solubilized with amphipol (Fig. 1), as well as in lysolecithin and laurylmaltoside (results not shown). So, although detergent binding to the FO part may contribute to the effect, it is likely that the observed asymmetry is a feature of the protein.

The presence of the second stalk — which is also evident in electron micrographs of the less well-defined V1VO-type ATPase from another bacterium, Clostridium fervidus11 — has important functional implications. It could provide the stator against which the γ-ɛ subunit crankshaft rotates1. Thus, ATP hydrolysis in one direction and ATP synthesis in the other would rotate the γ-ɛ-c subunit domain, thereby mechanically coupling nucleotide binding in catalytic sites with proton translocation.