An atomic-level mechanism for activation of the kinesin molecular motors

Abstract

Kinesin cytoskeletal motors convert the energy of ATP hydrolysis into stepping movement along microtubules. A partial model of this process has been derived from crystal structures, which show that movement of the motor domain relative to its major microtubule binding element, the switch II helix, is coupled to docking of kinesin’s neck linker element along the motor domain. This docking would displace the cargo in the direction of travel and so contribute to a step. However, the crystal structures do not reveal how ATP binding and hydrolysis govern this series of events. We used cryoelectron microscopy to derive 8–9 Å-resolution maps of four nucleotide states encompassing the microtubule-attached kinetic cycle of a kinesin motor. The exceptionally high quality of these maps allowed us to build in crystallographically determined conformations of kinesin’s key subcomponents, yielding novel arrangements of kinesin’s switch II helix and nucleotide-sensing switch loops. The resulting atomic models reveal a seesaw mechanism in which the switch loops, triggered by ATP binding, propel their side of the motor domain down and thereby elicit docking of the neck linker on the opposite side of the seesaw. Microtubules engage the seesaw mechanism by stabilizing the formation of extra turns at the N terminus of the switch II helix, which then serve as an anchor for the switch loops as they modulate the seesaw angle. These observations explain how microtubules activate kinesin’s ATP-sensing machinery to promote cargo displacement and inform the mechanism of kinesin’s ancestral relative, myosin.

Cytoskeletal motors are integral to all forms of eukaryotic life, with roles ranging from intracellular transport to cell division. X-ray crystal structures are known for the functional motor domains of kinesin and myosin, the two best-characterized cytoskeletal motors, but both motors substantially change conformations upon binding to their respective filaments (1, 2), and available structural methods have been unable to obtain atomic-resolution descriptions of the motor-filament complexes. Consequently, available structural models do not account for many essential properties of the cytoskeletal motors.

In microtubule-attached kinesin, the binding of ATP triggers a conformational change of the neck linker, to which cargo is tethered, leading to cargo displacement along the microtubule (3, 4). This behavior has been linked to a hypothetical mechanism for part of the activity (5), which we will refer to as the switch II helix scheme. In this scheme, inferred through comparisons of crystal structures of kinesin, ATP-triggered movements of the switch II helix would control the conformation of the neck linker through a steric coupling mechanism. CryoEM studies have confirmed key aspects of this scheme, directly resolving ATP analog-induced changes in the neck linker and coincident movement of kinesin’s core domain relative to the switch II helix, both consistent with the scheme (6, 7).

Two key aspects of the switch II helix scheme, however, have remained poorly defined: the role of microtubules, and the role of ATP itself. CryoEM and probe studies indicate that, when ATP binds to the motor-microtubule complex, the switch loops change conformation (6, 8–11). ATP-triggered movement of the switch loops, in turn, could promote movement of the switch II helix. However, structural studies have not resolved the detailed conformation of the switch loops in any state of the motor-microtubule complex. Moreover, in crystal structures of motile kinesins, neither of the switch loops respond to the presence of ATP analogs in the active site, and movement of the switch II helix appears to be decoupled from movement of the switch loops (12). This nonfunctionality is consistent with the observation that kinesin requires microtubules for catalytic activation (13) but leaves a substantial void in the mechanism. How could ATP binding trigger movement of the switch II helix and consequent docking of the neck linker? How do microtubules engage this triggering behavior?

Here we present four cryoEM reconstructions describing kinesin’s microtubule-attached catalytic cycle, each with a resolution (8–9 Å) exceeding the highest reported in prior cryoEM studies of kinesin-microtubule complexes. By synthesizing preexisting structural information from crystal structures with these maps, we derived atomic-level structural models describing ATP-induced conformational changes by key functional residues at kinesin’s nucleotide site. These models supply a detailed mechanism accounting for ATP-triggered movement of the switch II helix, involving the formation of a “phosphate tube” conformation of the switch loops seen previously only in crystal structures of myosin. The models also indicate a probable allosteric mechanism by which microtubule attachment activates kinesin’s ATP-sensing machinery. Collectively, these details provide a detailed molecular explanation for kinesin’s microtubule-attached power stroke. Furthermore, the specific arrangement of nucleotide-site residues we find in the ATP state of kinesin appears to predict a previously undescribed, functionally relevant structural state of myosin.

Results

We decorated microtubules with monomeric constructs of conventional kinesin and used cryoEM to characterize these complexes in four nucleotide states (ADP-bound, no-nucleotide, and ATP analog-bound states) spanning the entire catalytic cycle. Data processing statistics for these reconstructions indicate a nominal resolution equal to or exceeding 9 Å in every case (Table S1 and Fig. S1). Consistent with these statistics, our maps reveal distinct, clearly defined features for all major helices and β-sheets of both tubulin and kinesin (see Figs. 1 and 2 and SI Text).

Fig. 1.

Conformation of the switch loops and switch II helix near the active site. (A) Overview of the nucleotide pocket face of the motor domain, with labeled components. (B)–(D) close-up views of the nucleotide pocket in our ADP, no-nucleotide, and ADP•Al•F_x maps, respectively, with crystal structure models of K349 superposed. In (D), our phosphate tube model for the switch loop residues (Green) has been substituted for the crystallized K349 switch loops. Location of the “switch pocket” (Panel I ) is indicated. Movie S6 depicts the hybrid atomic model for our K349 construct, as seen in this panel, juxtaposed with our ADP•Al•F_x map. (E) Network of stabilizing interactions formed following formation of the phosphate tube and occupation of the switch pocket by the switch II helix extension. Side chains of residues E250 and Ile 254 are rendered as red van der Waals spheres, while conserved switch loop residues 201–205 (from switch I) and 233–236 (from switch II) are rendered as ball-and-stick diagrams. Vantage point is the same as (A)–(D). Movie S5 illustrates this arrangement in 3D. (F) Molecular surface formed by the switch loops in crystal structure 1MKJ. Surface is rendered in the CPK coloring scheme; view is from the interior of the microtubule, looking out at the nucleotide pocket. (G) Molecular surface formed by the phosphate tube conformation of the switch loops, showing the “switch pocket” complementary to E250 and Ile 254 from the switch II helix extension. The entry point into the phosphate tube, which leads to the γ-phosphate, is visible as a fenestration in the middle of the switch pocket.

Fig. 2.

Predicted interactions of modeled switch loop conformations in no-nucleotide and ADP•Al•F_x maps. Similar results were obtained for the ADP and AMPPNP maps (see Fig. S5). The positions of residues R203 (switch I; main chain atoms only, Orange Sphere depiction), A233 (switch II; Red Sphere depiction), and Ile 254 from the switch II helix extension (Dashed Outline, Red Sphere depiction) are indicated. Beta strands connected to the switch I and switch II loops, running through the motor domain’s central β-sheet, are rendered as ribbon diagrams, and the nucleotide is rendered as a ball-and-stick diagram. A thin cross section of the density map isosurfaces, aligned with the depicted β-strands and running through the predicted position of R203 and Ile 254, is also shown, as is density for the switch II helix. (A) Analysis of the no-nucleotide state, predicting complementary packing between the side chain of Ile 254 from the switch II helix extension and residues R203, A233 from the fitted K349 crystal structure model. Relative orientation of the catalytic core is indicated by the blue bars. (B) Substituting the phosphate tube loop conformations into the fitted K349 structure from (A) generates a hydrophobic cavity between Ile 254 and the switch pocket. Dashed line indicates rotation of the switch II helix required to occupy the cavity. (C) Analysis of the ADP•Al•F_x map predicts steric interference between Ile 254 and residues R203 and A233 from the fitted K349 structure. The black and white striped area indicates the predicted region of steric overlap. Dashed line indicates rotation of the switch II helix required to relieve the overlap. (D) Modeling of the phosphate tube loop conformations in the ADP•Al•F_x-bound map relieves the steric overlap found in the (C).

ATP-Driven Tilting and Docking.

Consistent with previous studies, our maps indicated that ATP analog binding causes the motor domain to tilt while the switch II helix remains fixed on the microtubule surface, simultaneously docking the neck linker (Figs. S2–S4). Thus, our maps strongly support the presence of “ADP-like” and “ATP-like” orientations of the motor domain in the absence and presence of ATP analogs, respectively, as predicted by the switch II helix scheme. Also consistent with prior studies (2, 11), we observed that the N terminus of the switch II helix is extended by several turns relative to crystal structures in all nucleotide states examined (Fig. S2 and Fig. S5).

Conformational Changes at the Active Site.

Consistent with previous EM observations (6, 10), our ADP-bound map featured an open nucleotide pocket, with the nucleotide site itself relatively exposed to solvent, whereas our no-nucleotide map indicated that the switch I loop intrudes into the nucleotide pocket (14). Also consistent with prior observations (6, 10, 11), our ATP analog-bound maps indicate that the switch I loop coordinates the nucleotide in a tight complex, in striking contrast to crystal structure models of kinesin’s ATP-like state where the switch I loop is positioned away from the nucleotide site. In contrast to earlier work, however, the higher resolution of our maps reveals that the switch I density element in the ATP analog states possesses a sheet-like structure, with a distinctive arch (Fig. 1A and D; Fig. S5C–E; SI Movies 3, 4, and 6). This density is not consistent with the conformation of switch I found in published kinesin crystal structures.

Atomic Model Describing Kinesin’s ATP State: The Phosphate Tube.

An atomic model has previously been proposed for the closed switch loop conformation of kinesin (15), based on homology modeling with a crystal structure of kinesin’s ancestral relative, myosin. In this switch loop model, closure of the loops around the nucleotide γ-phosphate forms a “phosphate tube,” with the triphosphate group from the nucleotide inserted into one end of the tube, while the other end of the tube is open. This phosphate tube model has not previously been considered within the context of cryoEM descriptions of the kinesin-microtubule complex. When we generated a phosphate tube model using the methods of (15) (see Methods), we discovered excellent agreement between the predicted phosphate tube conformation and our ATP analog maps, particularly in the ADP•Al•F_x map where the switch I density lobe was best defined (Fig. 1D and Fig. S6C–E).

In our phosphate tube model, the open end of the tube forms a prominent, relatively hydrophobic cavity, which we will term the “switch pocket,” facing toward the microtubule surface (Fig. 2D and G). In the modeling study of (15), the switch pocket was exposed to solvent; in our density maps, however, the switch pocket is evidently occluded by the switch II helix extension (Fig. 2D). Crystal structure models of kinesin’s ATP-like state failed to account for the extension density for our map, having a disordered loop in place of the extension. However, it was possible to generate a model of the extended switch II helix for our ATP analog maps using an extended conformation of the helix found in the crystal structure of ADP-bound KIF1A (PDB ID 1I5S; an ADP-like conformation) (Fig. 2; SI Methods).

Our assignment of atomic coordinates for the switch loops and the extended switch II helix together yielded a nearly complete atomic model of the microtubule-attached, ATP-bound state of kinesin. To further validate our ATP-state model, we generated a synthetic density map using the atomic coordinates and filtered the result to 8 Å resolution to simulate the signal loss in our EM imaging experiments. The resulting map of the motor domain was consistent in essentially every detail with our ADP•Al•F_x map (SI Movies 6 and 7).

Atomic Models for the ADP and No-Nucleotide States.

As described previously (14), the switch II helix coordinates derived from the fitted K349 crystal structure were slightly misaligned relative to density in the no-nucleotide map and also lacked the switch II helix extension. We corrected these discrepancies by propagating the extended helix model determined above from our ADP•Al•F_x map into the no-nucleotide map, preserving the position and orientation of the helix relative to tubulin (see Methods). This operation, which we repeated for the ADP map, yielded excellent alignments between the modeled switch II helix and observed density (Fig. 2 and Fig. S5). Furthermore, the resulting arrangement of the switch II helix and core domain was nearly identical to that seen in crystal structures of the KAR3 kinesin, which is ADP-like and exhibits an extended switch II helix conformation (16) (Fig. S5C). In this configuration, backbone atoms of three conserved γ-phosphate-sensing residues in the switch loops (R203, A233, G234) exhibit complementary packing against the conserved I254 side chain, which is located within the switch II helix extension. These switch loop residues possess highly conserved conformations in kinesin crystal structures (backbone RMSD < 1 Å, when the nucleotide-coordinating P-loops are aligned), despite considerable conformational variation evident in the remainder of the switch loops and/or switch II helix extension in these structures. We therefore retained the coordinates of R203/A233/G234 found in our K349 crystal structure model without modification for our MT-bound models of the ADP-bound and no-nucleotide states.

In contrast to the R/A/G residue triad, whose crystallized conformations were positioned within observable density features in our maps, the remainder of the switch I loop in our K349 crystal structure models (residues 190–202) failed to match the density features seen in our ADP-bound/no-nucleotide maps. Furthermore, none of the other available crystallized conformations of the switch I loop, from any kinesin variant, rectified this mismatch. We therefore omitted these unaccounted-for switch I loop residues from our models and from the analysis that follows.

Analysis of the Modeled Coordinates.

We performed modeling experiments to test whether the switch loop conformations we derived for kinesin’s ATP state were compatible with motor domain/switch II helix arrangements found in the no-nucleotide/ADP maps, and vice versa, by superposing the respective models of these states onto each other. When we superposed our ADP/no-nucleotide models of the switch loops onto the ATP-state models, we found that the R203/A233/G234 γ-phosphate sensor residues were placed in severe steric overlap with the absolutely conserved I254 side chain from the switch II helix extension (Fig. 2C). Because our maps rule out a compensatory reorientation or displacement of the switch II helix, this analysis therefore demonstrated that the ADP/no-nucleotide orientation of the core domain (relative to the switch II helix) is incompatible with the closed conformation of the switch loops found in our ATP model.

Conversely, when we superposed our ADP/no-nucleotide models of the core domain onto the ATP-state model, we found that a substantial hydrophobic void was formed (Figs. 2B and 3B). This void is a consequence of the motor domains’ ∼14° tilt away from the switch II helix extension when assuming the ADP/no-nucleotide orientation; this movement displaces I254 (in the extension) out of the switch pocket. Thus, these modeling experiments predict substantial energetic penalties for microtubule-attached kinesin configurations if the motor domain orientation is “ATP-like” but switch loop conformation is “ADP-like” or vice versa.

Fig. 3.

Summary of our atomic models describing microtubule-activated cargo displacement and catalysis by the kinesin motor domain. (A) Close packing between the switch loops and the switch II helix extension prevents leftward tilting and premature docking of the neck linker. The approximate path of the “nucleotide-ejecting” conformation of the switch I loop, inferred from cryoEM density (14), is indicated by the orange dashed line. Bulky hydrophobic residues, which we identify as the fulcrum of seesaw movement by the motor domain, are depicted by blue spheres. (B) Formation of the phosphate tube/switch pocket, caused by retraction of the switch loops toward the nucleotide pocket. In the absence of the microtubule-stabilized switch II helix extension, this retracted switch loop conformation would not be fully stabilized and is expected to rapidly revert to conformation(s) seen in kinesin crystal structures. The state depicted here is unlikely to be highly populated and is presented here to highlight the features that are caused by retraction of the switch loops. (C). Leftward tilting of the catalytic core domain leads to occupation of the switch pocket by residues E250 and Ile 254. This arrangement stabilizes the phosphate tube conformation of the switch loops, triggering catalysis and simultaneously triggering a power stroke by the neck linker (Magenta) on the opposite side of the catalytic core. In conventional kinesins, the cargo (and/or dimer partner) is attached to the magenta helix at the end of the neck linker and so is translocated toward the plus end of the microtubule in this step. The axis of rotation defined by our molecular fits of the K349 motor domain (no-nucleotide to ADP•Al•F_x transition) is depicted by bulls-eye symbol where the axis runs through the fulcrum residue F84 (see Fig. S3).

Discussion

Here we have derived a set of models describing detailed interactions of the switch loop/switch II helix assembly in three distinct states of the kinesin-microtubule complex. While our models of the ADP-bound/no-nucleotide states are essentially equivalent to crystallized conformations that have already been described, the conformation we have modeled for the ATP-bound state has neither been observed nor fully anticipated. As we now discuss, comparison of these atomic models indicates that a conceptually simple “seesaw” mechanism underlies kinesin’s force-generating behavior. Furthermore, microtubule attachment engages this mechanism, which is not observable in crystal structures, primarily through a single structural element: the switch II helix extension.

How Microtubule Attachment Nucleates the Switch II Helix Extension.

Our density maps provide clear evidence that the switch II helix extension is omnipresent in all nucleotide states of the microtubule-attached motor. In contrast, observation of the switch II helix extension in crystal structures is relatively rare, with these residues most frequently not visible in electron density maps (and thus assigned as part of loop L11, rather than the helix). The extension is likely stabilized by a direct microtubule contact at its junction to the remainder of the switch II helix, at the locus of universally conserved N255 of kinesin (14) (see Fig. 2). However, a second characteristic of the kinesin-microtubule interaction, steric occlusion of the disordered loop L11 by the microtubule surface, also likely stabilizes the extension. By prohibiting many of the conformations normally explored in the disordered state of L11, the proximity of the microtubule surface would favor condensation of L11 into the helical extension on purely entropic grounds. Thus, microtubule attachment likely stabilizes the switch II helix extension, by either or both of these mechanisms.

How the Extension Stabilizes the Pre-Power Stroke State.

Our models embody the switch II helix scheme as a seesaw movement (Figs. 3 and 4), in which the switch II helix forms the stationary base of the seesaw, fixed with respect to the microtubule surface. Most of the remainder of the catalytic core domain may be viewed as the seesaw’s moveable platform. Movement of the seesaw platform is facilitated by specific side chains positioned between the platform and the base. Of particular note are two highly conserved residues from the catalytic core, F82 and Y84, which directly contact side chains of highly conserved L258 and L261 from the switch II helix. The bulky character of these side chains, which are positioned roughly at the center of the seesaw interface, has the effect of separating the seesaw platform from the base, so creating space for the platform to tilt leftward or rightward. This arrangement also suggests that this cluster of residues might serve as the pivot point, or fulcrum, of the seesaw; in agreement with this interpretation, our geometric analysis of the seesaw movement (during the no-nucleotide to ATP analog transition) places the axis of platform rotation directly through the side chain of Y84 (Fig. 3C).

Fig. 4.

Cartoon depiction of microtubule activation. (A) Disabled mechanism of kinesin in the absence of microtubules represented by crystallized conformations of the conventional motor domain. When detached from the microtubule, kinesin was previously shown to fluctuate between these states (12). Docking of the neck linker, to which cargo is attached, accompanies movement of the central core domain (Large Blue Rectangular Shape) relative to the switch II helix (Red Elongated Rectangle). Movement of the core domain is coupled to neck linker docking, due to a collision between α6 (to which the neck linker is connected) and the switch II helix (Left Cartoon). This collision is indicated by the dashed outline at the end of α6. The red dashed line depicts a ∼15–20 residue segment (L11) that connects the switch II loop to the switch II helix but is disordered in crystal structures. Circles labeled R and A depict residues R203 and A233. (B) Cartoon depiction of the states presented in Fig. 3. The N terminus of the switch II helix (Left Side) is extended by several turns, stabilized by a microtubule contact (represented by the Small Red Circle embedded within the microtubule surface). In the stroke state of the motor (state 1), this extension props the seesaw to the right, forcing the neck linker to remain in the disordered, stroke state. Following ATP binding, the switch loops retract toward the γ-phosphate, causing the seesaw to tilt to the left, accompanied by docking of the neck linker, which constitutes a power stroke that would translate the cargo of conventional kinesin toward the microtubule plus end (state 3). Residue I254 in the switch II helix extension is depicted as the red circle labeled I.

Prior to microtubule attachment, the switch II helix freely oscillates through a wide range of orientations, independent of nucleotide state (12). Microtubule-induced stabilization of the switch II helix extension, however, severely restricts the motional freedom of the switch II helix, through steric interference between the extension (notably including the conserved side chains of E250 and I254) and conserved amino acid residues in the nucleotide pocket (including R203 from switch I, and A233 from the switch II loop). Thus, the extension effectively props the seesaw to one side, forcing the neck linker/α6 and switch II helix elements into an ADP-like arrangement (Figs. 3 and 4). In this way, microtubule attachment would constrain the neck linker of an ADP-bound or no-nucleotide motor domain in a disordered conformation, biasing the cargo position toward the microtubule minus end.

Controlling the Power Stroke and Catalysis.

The stabilization of the phosphate tube afforded by ATP binding provides an ideal mechanism to drive docking by kinesin’s neck linker, because the resulting retraction of the switch loops toward the nucleotide pocket naturally attenuates the propping effect we identified in the pre-power stroke motor states (Fig. 3B). Hence, formation of the phosphate tube promotes leftward tilting of the seesaw and concomitant docking of the neck linker (Fig. 3C and Fig. 4B, state 2). Furthermore, the surfaces found at the interface between the phosphate tube and the switch II helix extension are highly complementary, not only in shape but also in their overall hydrophobic character. Of particular note at this interface are numerous hydrophobic contacts between conserved residues R203, A233, and E236 in the switch pocket and conserved residues E250 and I254 in the switch II helix extension. Thus, the ATP-bound conformation we have derived appears to be highly stable, thereby promoting a docked conformation of kinesin’s neck linker.

In our structure models, both ATP and the microtubule would supply significant free energy components stabilizing the phosphate tube: ATP supports the phosphate tube structure through numerous hydrogen bonds between the γ-phosphate and the switch loops, while the microtubule promotes formation of the switch II helix extension, which in turn supports the phosphate tube through direct interactions with the switch pocket. We therefore propose that both bound ATP as well as the microtubule-stabilized helix extension are required in order to fully stabilize the phosphate tube. This requirement would explain why the tube has not yet been observed in crystal structures of kinesin nor in the microtubule-attached states that lack ATP. Thus, the post-power stroke conformation of kinesin would only be strongly stabilized in the simultaneous presence of both ATP and the microtubule.

The phosphate tube geometry seen in our ATP-state model appears highly sensitive to mutagenesis of certain conserved sequence elements, in line with experimental studies. For example, the G234A mutant was found to be trapped in stroke state while attached to the microtubule (3), an effect that our model explains by a clear disruption of the geometry of the phosphate tube, so interfering with the post-power stroke seesaw configuration. Based on the same logic, we predict that bulkier substitutions at the adjacent, conserved A233 site would yield similar effects as G234A. Our mechanism also predicts substantial functional defects upon mutagenesis of the conserved I254 site in the switch II helix extension, although the phenotype would be unclear given the heavy involvement of this residue in both pre- and post-power stroke conformations.

The phosphate tube also defines the presumed catalytically active conformation of kinesin’s nucleotide pocket (15) by placing residues implicated in catalysis in close proximity with the γ-phosphate. Consequently, our models provide a qualitative explanation for why microtubules are required to fully stimulate catalysis (17): microtubule attachment is required to stabilize the switch II helix extension, so completing the array of stabilizing binding partners for the phosphate tube. We note that our ATP structure model exhibits a salt bridge pair between the highly conserved residues R203 and E236 (from the switch I and switch II loops, respectively), in line with prior expectations (18). However, the solvent-exposed placement of this salt bridge in our model, at the periphery of the phosphate tube, suggests that its deletion would not strongly destabilize kinesin’s power stroke. The E236 site is implicated in the catalytic cleavage step in myosin (and by association, kinesin) (19). Thus, our structure models explain why R203A or E236A substitutions lead to ATP-bound, microtubule-attached motors stalled in the post-stroke conformation (3, 16, 20, 21): ATP binding would still trigger the seesaw action, but direct or indirect disruption of the E236 site would inhibit the subsequent hydrolysis step.

Another feature of our ATP model is that, following catalysis, the hydrolytic products would be very well contained within the active site by the phosphate tube. This highly enclosed active-site conformation is strikingly different from the crystallized conformations of detached kinesin, where the active site is highly exposed to solvent. The active site in our kinesin model is even more fully enclosed than in the myosin structures where the phosphate tube was originally identified, because the switch II helix extension in our model effectively caps the open end of the tube. The tightly sealed nature of the active site in our model suggests that phosphate release from the active site could be delayed following the catalytic cleavage step, and that this delay might be especially significant in the microtubule-attached enzyme where the phosphate tube experiences additional stabilizing interactions. In accord with this idea, the rate of phosphate release in microtubule-attached kinesin was found to be ∼50/s (22), which is considerably slower than the cleavage step itself (> 100/s) (23, 24), and indicates a substantial dwell time of the cleaved phosphate product in the nucleotide pocket.

Containment of ADP•Pi in the phosphate tube is especially relevant to the energetics of the ATPase reaction itself, because containment would lead to increased probability of resynthesis, so attenuating the free energy lost during the actual hydrolysis step. This attenuation makes additional free energy available for kinesin’s force-generating step (which accompanies ATP binding), rather than releasing the stored energy of ATP in the hydrolysis step, where force is presumably not generated (25). In this way, the phosphate tube may be closely linked to kinesin’s overall efficiency. Reversibility of hydrolysis is a signature characteristic of myosin (26) and is accounted for by the closed phosphate tube structure seen in x-ray crystallography studies (27). Reversibility of kinesin’s ATPase step is also supported by experimental studies but only in kinesin’s microtubule-attached state (25). Our structural models account for the microtubule specificity of this effect, again because the phosphate tube would only be fully stabilized in the microtubule-attached case.

Relevance to Myosin.

While kinesin is known to have structural and functional parallels with its ancestral relative, myosin (28), the extent of this commonality has remained unclear. Our results indicate that the catalytically active switch loop arrangement, found in these motors’ ATP states, is very similar. However, we discovered a striking difference in the orientation of the switch II helix in crystallized ATP analog-bound states of myosin (1) when compared with our ATP-state model of kinesin. In these myosin crystal structures, the switch II helix extension is displaced out of the switch pocket, despite the presence of identically conserved residues equivalent to conventional kinesin’s E250/I254. The resulting helix/core domain arrangement in myosin closely resembles the “transient” (i.e., presumed unstable) conformation of ATP-bound, microtubule-attached kinesin depicted in state 2 of Fig. 4B. If the seesaw action we have characterized in kinesin were to be replicated in these myosin crystal structures, not only would this action fill the switch pocket (so capping the phosphate tube), but it would also result in closure of myosin’s actin binding cleft through the associated movement of myosin’s lower 50 kD domain (Fig. S8). Closure of this cleft is not observed in crystal structures of myosin, except in structures thought to represent the nucleotide-free rigor state, where a decidedly different domain arrangement is found compared to our model (29, 30). Cleft closure in the ATP state of myosin, as found in our kinesin-derived model, is one expected feature of the tightly attached pre-stroke state of myosin, which has so far eluded structural characterization (1). It also seems possible that the tightly capped phosphate tube seen in our hypothetical myosin conformation represents the true catalytically active state. Kinesin may therefore be poised to inform the mechanism of its larger and more complex, ancient cousin.

Materials and Methods

Preparation of the K349-microtubule complex was similar to our earlier published protocol (14). Final concentrations of nucleotides were: ∼2 mM ADP, ∼2 mM AMPPNP, and ∼2 mM ATP/2 mM AlCl₃/8 mM NaF, for the ADP-bound, AMPPNP-bound, and ADP•Al•F_x nucleotide states, respectively. Samples were imaged on a JEOL 4000 microscope operating at 400 kV. Films were scanned and processed using custom methods that treated short microtubule segments as asymmetric single particles, as described (14), with some modifications. For crystal structure fitting, the Fit in Map feature of UCSF Chimera (32) was used to generate synthetic maps at 8 Å resolution from atomic coordinates of crystal structures and identify transformations that maximized voxelwise cross correlation with our experimentally determined maps. Additional details may be found in SI Text.

Note Added in Proof.

A newly reported crystal structure of the Eg5 kinesin motor domain (31), complexed with AMPPNP, independently confirms the specific switch loop/switch II helix arrangement seen in our ATP-state model.