The microtubule cytoskeleton is a dynamic structure in which the lengths of the microtubules are tightly regulated. One regulatory mechanism is the depolymerization of microtubules by motor proteins in the kinesin-13 family 1 . These proteins are crucial for the control of microtubule length in cell division [2][3][4] , neuronal development 5 and interphase microtubule dynamics 6,7 . The mechanism by which kinesin-13 proteins depolymerize microtubules is poorly understood. A central question is how these proteins target to microtubule ends at rates exceeding those of standard enzyme-substrate kinetics 8 . To address this question we developed a single-molecule microscopy assay for MCAK, the founding member of the kinesin-13 family 9 . Here we show that MCAK moves along the microtubule lattice in a one-dimensional (1D) random walk. MCAK-microtubule interactions were transient: the average MCAK molecule diffused for 0.83 s with a diffusion coefficient of 0.38 mm 2 s 21 . Although the catalytic depolymerization by MCAK requires the hydrolysis of ATP, we found that the diffusion did not. The transient transition from three-dimensional diffusion to 1D diffusion corresponds to a "reduction in dimensionality" 10 that has been proposed as the search strategy by which DNA enzymes find specific binding sites 11 . We show that MCAK uses this strategy to target to both microtubule ends more rapidly than direct binding from solution.Kinesin-13 motor proteins act at microtubule ends, where they are thought to force protofilaments into a curved conformation 12,13 , which is a likely structural intermediate in the depolymerization process 14 . Classically, kinesin motor proteins reach microtubule ends by ATP-dependent translocation along microtubules. However, LETTERSFigure 1 | MCAK-dependent microtubule depolymerization. a, Diagram of the in vitro assay depicting a microtubule (red) immobilized above the glass surface by anti-tubulin antibodies (dark blue). Excitation by total internal reflection allows the detection of single molecules (namely MCAK-GFP in green) in the evanescent field (shown in blue). b, Epifluorescence images of immobilized microtubules at the times shown in minutes. MCAK dimer (8 nM) was added at t ¼ 2 min. c, Plot of microtubule depolymerization rate against MCAK concentration. Error bars are s.d. Data fitted to Hill equations (lines plotted) yielded K m ¼ 3.9 nM and K m ¼ 6.1 nM for MCAK and MCAK-GFP, and n ¼ 2.4 and n ¼ 2.2, respectively. Red squares, MCAK-His 6 ; green circles, MCAK-His 6 -EGFP. d, Shortening of four microtubules from a mean length of 8.4 mm to 7.8 mm (black line) after the addition of 5 nM MCAK (red line). The depolymerization rate approached steady state with a time constant of 3.8 s (green fitted line). (Fig. 1a) in which microtubules were immobilized on coverslips by means of surface-adsorbed anti-tubulin antibodies. Individual rhodamine-labelled microtubules and single MCAK-GFP molecules were revealed by epifluorescence and total-internalreflection fluorescence (TIRF) illumina...

Motor proteins in the kinesin-8 family depolymerize microtubules in a length-dependent manner that may be crucial for controlling the length of organelles such as the mitotic spindle. We used single-molecule microscopy to understand the mechanism of length-dependent depolymerization by the budding yeast kinesin-8, Kip3p. We found that after binding at a random position on a microtubule and walking to the plus end, an individual Kip3p molecule pauses there until an incoming Kip3p molecule bumps it off. Kip3p dissociation is accompanied by removal of just one or two tubulin dimers (on average). Such a cooperative mechanism leads to a depolymerization rate that is proportional to the flux of motors to the microtubule end and accounts for the length dependence of depolymerization. This type of feedback between length and disassembly may serve as a model for understanding how an ensemble of molecules can measure and control polymer length.

MCAK belongs to the Kin I subfamily of kinesin-related proteins, a unique group of motor proteins that are not motile but instead destabilize microtubules. We show that MCAK is an ATPase that catalytically depolymerizes microtubules by accelerating, 100-fold, the rate of dissociation of tubulin from microtubule ends. MCAK has one high-affinity binding site per protofilament end, which, when occupied, has both the depolymerase and ATPase activities. MCAK targets protofilament ends very rapidly (on-rate 54 micro M(-1).s(-1)), perhaps by diffusion along the microtubule lattice, and, once there, removes approximately 20 tubulin dimers at a rate of 1 s(-1). We propose that up to 14 MCAK dimers assemble at the end of a microtubule to form an ATP-hydrolyzing complex that processively depolymerizes the microtubule.

SUMMARYNon-centrosomal microtubule bundles play important roles in cellular organization and function. Although many diverse proteins are known that can bundle microtubules, biochemical mechanisms by which cells could locally control the nucleation and formation of microtubule bundles are understudied. Here, we demonstrate that the concentration of tubulin into a condensed, liquid-like compartment composed of the unstructured neuronal protein tau is sufficient to nucleate microtubule bundles. We show that, under conditions of macro-molecular crowding, tau forms liquid-like drops. Tubulin partitions into these drops, efficiently increasing tubulin concentration and driving the nucleation of microtubules. These growing microtubules form bundles, which deform the drops while remaining enclosed by diffusible tau molecules exhibiting a liquid-like behavior. Our data suggest that condensed compartments of microtubule bundling proteins could promote the local formation of microtubule bundles in neurons by acting as non-centrosomal microtubule nucleation centers and that liquid-like tau encapsulation could provide both stability and plasticity to long axonal microtubule bundles.