The swimming of bacteria such as Salmonella typhimurium and Escherichia coli is driven by the rotation of flagella, no more than 0.25 μm in diameter but as much as 60 times that in length. The bacteria alternate between 'running' in a straight line and chaotic 'tumbling', while the rotary motor at the base of a flagellum changes from anticlockwise rotation to clockwise and back again. Reporting in Nature, Keiichi Namba and colleagues at the Protonic Nanomachine Project in Kyoto, Japan, have begun to work out the subtle atomic-level changes that couple these two events.

The bacterial flagellum consists over almost all of its length of a single protein, flagellin. Thousands of flagellin molecules form a hollow tube composed of 11 simple polymer threads, known as protofilaments. Electron microscopy of flagella has revealed that the protofilaments can exist in two forms, the L and R forms. The L form is slightly longer, with its flagellin subunits being 0.8 Å more elongated. A flagellum made up entirely of protofilaments of either type is straight and, although good for structural studies, it makes a poor propeller. Mixing the two forms in a single flagellum, however, means that the difference in lengths sets up tensions that can be resolved only by supercoiling the flagellum into a corkscrew shape.

When bacteria are swimming in a straight line, the flagella usually have nine L-type and two R-type protofilaments, producing a left-handed corkscrew. These flagella can bundle together to form a coordinated propulsion unit. When the flagellar motors reverse direction, a number of L-type protofilaments change to R-type protofilaments through a cooperative change in each flagellin molecule, right along the flagellum's length. Right-handed supercoiled flagella are produced, breaking up the flagellar bundles and leaving the individual flagella to push in different directions — this produces the tumbling motion. A very small change in flagellin's structure is thus at the heart of the bacterium's change in behaviour.

Proteins that form polymers pose a particular problem to structural studies; rather than forming well-ordered crystals, they tend to produce poorly ordered aggregates of their polymers. Consequently, Namba and colleagues studied a version of flagellin from Salmonella lacking 52 amino acids from its amino terminus and 44 amino acids from its carboxyl terminus. On solving its structure at 2.0-Å resolution, the authors found that the flagellin proteins were arranged as if single protofilaments were running throughout the crystals. This high-resolution crystal structure could be easily fitted into lower-resolution electron microscope data to produce a model for the straight flagella formed by R-type protofilaments alone (as in the figure).

To understand how the R-type protofilament converts into the L type, the Japanese group used computer modelling to stretch their structure. Three flagellin subunits were taken, arranged as in a protofilament. The end subunits were kept rigid while being pulled apart in 0.1-Å steps, and the central subunit was allowed to relax to its lowest energy state. At first there were no major changes in structure, but then, over a 0.2-Å stretch, the central flagellin underwent a subtle conformational change. A β-hairpin shifted to allow the 0.8-Å expansion required to lengthen the protofilament.

The bacterial flagellum has been one of the most intensely studied structures in biology, and what has piqued the curiosity of biochemists, biophysicists and engineers alike is its changes between forward swimming and reverse tumbling. At long last we are beginning to see exactly how subtle this switch in gear really is.