Kinesin motors power many motile processes by converting ATP energy into unidirectional motion along microtubules. The force-generating and enzymatic properties of conventional kinesin have been extensively studied; however, the structural basis of movement is unknown. Here we have detected and visualized a large conformational change of an approximately 15-amino-acid region (the neck linker) in kinesin using electron paramagnetic resonance, fluorescence resonance energy transfer, pre-steady state kinetics and cryo-electron microscopy. This region becomes immobilized and extended towards the microtubule 'plus' end when kinesin binds microtubules and ATP, and reverts to a more mobile conformation when gamma-phosphate is released after nucleotide hydrolysis. This conformational change explains both the direction of kinesin motion and processive movement by the kinesin dimer.

Kinesin and myosin are motor proteins that share a common structural core and bind to microtubules and actin filaments, respectively. While the actomyosin interface has been well studied, the location of the microtubule-binding site on kinesin has not been identified. Using alanine-scanning mutagenesis, we have found that microtubule-interacting kinesin residues are located in three loops that cluster in a patch on the motor surface. The critical residues are primarily positively charged, which is consistent with a primarily electrostatic interaction with the negatively charged tubulin molecule. The core of the microtubule-binding interface resides in a highly conserved loop and helix (L12/alpha5) that corresponds topologically to the major actin-binding domain of myosin. Thus, kinesin and myosin have developed distinct polymer-binding domains in a similar region with respect to their common catalytic cores.

Members of the kinesin superfamily share a similar motor catalytic domain yet move either toward the plus end (e.g., conventional kinesin) or the minus end (e.g., Ncd) of microtubules. The structural features that determine the polarity of movement have remained enigmatic. Here, we show that kinesin's catalytic domain (316 residues) in a dimeric construct (560 residues) can be replaced with the catalytic domain of Ncd and that the resultant motor moves in the kinesin direction. We also demonstrate that this chimera does not move processively over many tubulin subunits, which is similar to Ncd but differs from the highly processive motion of conventional kinesin. These findings reveal that the catalytic domain contributes to motor processivity but does not control the polarity of movement. We propose that a region adjacent to the catalytic domain serves as a mechanical transducer that determines directionality.