The structure, from Walker and co-workers (Abrahams et al. 1994), shows a radial, 3-fold symmetry, with each 120 ° sector, although composed of an common a-b heterodimer, containing a different ligand binding site associated with the catalytic subunit, b. The shape of each ligand binding site appears to be determined by the asymmetry of the single g subunit, which forms a spindle-like structure inserted through the central core of the roughly cylindrical a₃b₃ domain, where the points of interaction between g and a₃b₃ are large, hydrophobic amino acid side chains. From the structure alone, the irresistible conclusion is that F1 is a bearing. Rotation of g within the central axis of F₁ would obviously induce sequential changes in the conformation of each a-b heterodimer, a structural basis for Boyer's binding-change mechanism. Each heterodimer binds ADP and phosphate loosely; then ADP and phosphate tightly; and, finally, ATP, which is seen in situ in the crystal structure in the form of a non-hydrolysable ATP analogue.

The idea that the g subunit acts as a camshaft is currently supported by two quite independent lines of evidence. A direct and visually compelling demonstration has been provided by Yoshida and co-workers (Noyi et al., 1997), who successfully tethered the g subunit to an actin filament and the a₃b₃ headgroup to an inert, metal surface: upon addition of ATP, some actin filaments were observed, in a light microscope, to rotate. During ATP hydrolysis, the angular velocity of the actin filament depended on its length, but all rotations were anticlockwise. Junge and co-workers (Sabbert et al., 1996) used a fluorescence tag (eosin) on the g subunit of an immobilised chloroplast (C)F₁ in order to study movement of the tag by polarised absorption relaxation after photobleaching. The conclusion is that ATP induces rotation of g relative to the hexagonal a₃b₃ array. One complete ATP-induced rotation takes 100 ms: pt_s = 1.0.