Coordination between motor domains in processive kinesins.

Among cellular functions that kinesins perform, their ability to walk along the microtubules transporting specific cargoes is the most fascinating (1, 2). The complexity of this phenomenon and tremendous scientific efforts put toward dissecting the mechanisms underlying it have led to conflicting models explaining how kinesins might travel long distances without dissociating from their biological track (e.g. move “processively” (3)). Although these models differ, for conventional kinesin dimer the favored one predicts that the two kinesin heads bind alternatively to the track, taking 80-Å steps along the microtubule while hydrolyzing one ATP molecule per step (4–6). This “hand-over-hand” model (7, 8) also predicts that the two kinesin heads remain enzymatically “out of phase” (e.g. at different stages of the ATP hydrolysis cycle), ensuring that, at any given tug, at least one of the two motor domains remains strongly attached to the track. Many technologically marvelous papers (9–18) illustrate the importance of coordination between the motor domains of kinesin during its processive movement. We review the phenomenon of kinesin processivity from a complementary perspective by considering specific structural features of the motor domains that underlie their coordination.


Structural Features That Transform a Protein into a Processive Motor
All kinesins share the conserved catalytic core (residues Asn 8 -Ala 322 in human kinesin (19)), which consists of a central ␤-sheet sandwiched between six ␣-helices (Fig. 1A, cream) and a topologically conserved smaller lobe (called the ␤-domain here) (Fig. 1A, peach) with three additional ␤-strands. The core has both the nucleotide- (Fig. 1A, green) and microtubule-binding sites (the major site, loop L12 and ␣4, is indicated in Fig.  1B) and with the microtubule executes nucleotide hydrolysis cycle powering kinesin motility (20,21). Depending on its nucleotide state (␥-phosphate either present (ATP/ADP⅐P i ) or absent (ADP/no nucleotide)), the catalytic core "switches" between different conformations (6,22), controlling affinity of kinesin for the microtubule and the positions of the forcegenerating mechanical element of the motor, the neck region (effectively a transmission (20,21,(23)(24)(25)). In conventional kinesin, the neck (19 -21, 26) is positioned C-terminally to the core and consists of the neck linker (residues Lys 323 -Thr 336 in human kinesin) (Fig. 1A, blue) and the following neck coiled coil (Fig. 1A, gray), which maintains the dimeric state of the motor. Fig. 1, B and C, shows how the major switching region, "switch II cluster" (22,27), rising above the central ␤-sheet in the presence of ␥-phosphate (blue) and collapsing back in its absence (red) controls the conformations of the neck by, respectively, either facilitating docking of the neck linker and its base, Ile 325 , along the core or sterically precluding this docked position. In the microtubule-bound motor, the power stroke is facilitated by repositioning the neck linker from its nucleotidefree conformation into the docked, ATP-like state, an assembly that propels the partner motor subunit toward the plus end of the microtubule (20,25).
Because of their proximity and inseparable functional roles, the catalytic core and the adjacent neck have been termed the motor domain (19,20). In the conventional kinesin, the nucleotide-induced conformational transitions in one motor domain are amplified into 80-Å step movements by the partner motor domain, which is connected through the neck coiled coil and the following stalk domain (6,20,25).
The alternating force-generating repositioning of the neck in kinesin motor domains is achieved when one of the two domains is firmly attached to the track (between the strongly bound nucleotide-free and ATP states (6,25,28)) and, therefore, can use the produced force to displace the partner motor domain forward. The coordinated control over the affinity of kinesin for the track is achieved by the nucleotide-dependent restructuring of the microtubule-binding sites of the motor, which are positioned either directly within the nucleotide-responsive switch II cluster (L12 and ␣4, Fig. 1B) or close to it (27).
In addition to the coordination between the affinity of the motor for the track and its power-generating movements, processivity of conventional kinesin requires special properties of its neck. Stability of the neck coiled coil, which does not unwind during movement of kinesin (29), keeps the two motor domains in register. At the same time, flexibility of the neck linker (25) allows the motor domains to propel over each other, optimizing their positions while searching for the new binding site. The ability of the neck linker to extend allows bridging the 80-Å distance between neighboring binding sites on the track (30). In non-processive kinesins, the necks are evolved for performing different functions (for optimizing the plus end-directed power stroke in the mitotic spindle kinesin Eg5 (31) or for producing the oppositely directed force in the minus end-directed motor Ncd (32)(33)(34)(35)(36)).

Cooperativity between Motor Domains Drives Processive Movement of Kinesin
Structural (30) and biophysical (15,18,37) data show that at a critical point of its movement conventional kinesin adopts a configuration on the microtubule with both motor domains bound to the track. In this "bridged" state ( Fig. 2A), the trailing motor is firmly attached to the microtubule, has ATP/ADP⅐P i in its nucleotide site, and adopts the ATP-like conformation with the neck linker docked alongside the core and pointing (from N to C termini) toward the plus end of the microtubule (30) (Fig. 2A, blue). This configuration allows the leading motor to attach to the next site on the protofilament, releasing ADP and adopting the nucleotide-free conformation with the neck linker pointing in the opposite direction (30) ( Fig. 2A, red). Kinetic data show that the leading motor cannot bind ATP until the trailing motor hydrolyzes ATP and releases one of the hydrolysis products, P i (18). This release weakens attachment of the motor core to the microtubule, and the trailing motor detaches from the track, loosing constraints placed onto the leading motor and allowing it to bind ATP and enter its nucleotide hydrolysis cycle (18).
The inability of the leading motor to enter a new hydrolysis cycle until the trailing motor reaches the end of the previous cycle and is prepared to bind to the next site on the track is critical for processive movement (9,13,15,18). This kinetic coordination guarantees that the leading motor would not prematurely enter and finish its ATP hydrolysis together with the trailing motor domain detaching from the microtubule and causing the motor dimer to lose its grip with the track.

Interactions of Kinesin with the Microtubule Track
Although kinetic cooperativity between the kinesin motor domains is proven and its functional importance is well under-stood, the structural mechanism underlying this phenomenon is fuzzy.
With eight structures of plus end-directed kinesin motor domains determined and some observed in different conformational states (22, 31, 38 -42), the character and the amplitude of conformational motions that kinesin would undergo during its movement can be analyzed. By coupling the crystal structures with electron microscopy images at improved resolution (22,43), the kinesin motor domain can be positioned on its track in different nucleotide states.
We analyzed conventional kinesin in its bridged state (Fig.  2B), looking for determinants that would explain the cooperativity between the two motor domains. For this analysis, the trailing motor with the docked neck linker (Fig. 2B, blue) was modeled from the structure of kinesin in the ATP-like conformation (42). Its position on the track was taken to be equivalent to that determined for KIF1A in the ATP-like state (22). The position of the neck coiled coil (Fig. 2B, gray) was adopted from the ATP-like structure of the kinesin dimer (40). The leading motor core was modeled by the ADP-like structure of kinesin (38). For its position on the microtubule, we considered two orientations. In the first dimer (not shown), the position of the leading core was assumed to be identical to that of the trailing core. This assumption derives from electron microscopy data that did not reveal changes in the position of the core between the nucleotide-free and ATP-like states (30,44,45). The position of the leading core in the second dimer (Fig. 2B) was adopted from the alternative orientation observed for KIF1A in the ADP/no nucleotide state (22). The leading core in FIG. 1. Structural organization of conventional kinesin. A, the catalytic core, ␤-domain, and bound nucleotide of conventional kinesin (3kin) are, respectively, in cream, peach, and green. The neck linker and coiled coil are blue and gray, respectively. B, mechanism of conformational switching is exemplified using superimposed atomic models for kinesin in ADP (1bg2) and ATP-like (1mkj) states. Switch II cluster and neck linker are blue for ATP-like and red for ADP states. C, mechanism underlying the nucleotide-dependent conformational transitions of the kinesin neck.

FIG. 2. Bridged state of conventional kinesin.
A, schematic representation of the bridged state. The catalytic core is cream, the ␤-domain is yellow/red, and the oppositely directed neck linkers of the trailing and leading motors are blue and red arrows, respectively. The neck coiled coil is drawn as a braid. The trailing motor (with either ATP (T) or ADP⅐P i (DP)) and the leading core (nucleotide-free) are strongly attached (indicated by anchors) to the microtubule. B, an atomic model for the bridged state of conventional kinesin. The coloring scheme is consistent with Fig. 1A and A above. The countermovement of the leading core is indicated by arrows. Polypeptide chains for ␣and ␤-tubulin (1tub) are in dark and light gray.

Minireview: Coordination in Kinesin Motors 15708
this dimer configuration is translated toward the minus end of the microtubule (by ϳ3 Å) and rotated clockwise (by ϳ10 degrees, Fig. 2B) relative to the position of the trailing core bound to the equivalent tubulin dimer. The mechanism underlying this changed orientation has been explained by the compensatory countermovement of the motor relative to its switch II cluster, which changes its position relative to the core between the nucleotide-free and ATP-bound states (Fig. 1B) (22,27). In the microtubule-bound motor, its movement relative to the microtubule might be restrained by anchoring interactions with the track in both strongly attached states (22,27).
For both dimers, we modeled the neck linker of the leading motor domain in the orientation opposite to that of the trailing motor (Fig. 2B, red). In placing the leading linker, we used two constraints. The first was the position of the N terminus of the leading neck coiled coil that marks the end point of the leading neck linker. The second was the observation that helix ␣6 preceding the neck linker can unwind only at its C terminus (two residues C-terminal to Arg 321 ). In all kinesin structures, the rest of the helix remains stable because of packing interactions with the core. Thus, for both kinesin dimers the start and the end of the leading neck linker have been defined.
In both kinesin configurations, the leading neck linker must adopt an extended conformation to bridge the preceding helix ␣6 and the following neck coiled coil. In both dimers, the leading linker runs alongside the ␤-domain of the leading core (Fig. 2B, peach), making contacts with this structural element unavoidable. However, in the first dimer (not shown) because of the orientation of the core and the position of the small lobe, the neck linker abrades the ␤-domain. Attempts to avoid this clash resulted in the linker being stretched beyond the geometry suitable for the polypeptide chain, making this configuration of the double head-bound kinesin unlikely.
Reorientation of the leading motor in the second dimer by the countermovement of the core (Fig. 2B) positions the ␤-domain compatibly with the conformation of the leading neck linker. Remaining extended, the linker runs alongside the ␤-domain in this bridged kinesin dimer, making side chain contacts with one of the ␤-strands of the small lobe, ␤1c. Notably, the neck linker of kinesin motor Eg5 has been observed in a similar conformation. Furthermore, this conformation correlated with the ADP/no nucleotide-like state of the motor (31).

A Structural Model for Coordination between Kinesin Motor Domains
A structural model for kinetic and mechanical cooperativity between the motor domains of conventional kinesin is proposed in Fig. 3. This model explains why the leading motor is unable to bind a new molecule of ATP (Fig. 3B) until the trailing motor hydrolyzes the previous one and, having ADP in its nucleotide site, becomes either weakly bound or detached from the microtubule (Fig. 3C). Only after detachment of the trailing motor (Fig. 3C) the leading neck linker relaxes and allows the leading core to reorient itself on the microtubule. This reorientation likely brings the nucleotide-binding elements of the motor (switch I and switch II regions (27)) into the proper position relative to each other and to the stabilizing microtubule surface, facilitating new nucleotide binding. The resulting reorientation of the leading neck linker into the alternative, ATPlike position (Fig. 3D) further stabilizes the ATP-bound conformation of the motor, allowing nucleotide hydrolysis. In a sense, the extended leading neck linker in the constrained, bridged state of kinesin (Fig. 3B) acts as a structural sensor that coordinates work of the enzymatic active sites in two kinesin cores keeping them "out of phase." We favor the suggested model because it explains the observed cooperativity between kinesin motor domains and is built using crystal structures (38,42) and experimentally observed orientations of the motor on the microtubule (22). Furthermore, for both the leading and trailing motors, these conformations and orientations are consistent with the general switching mechanism proposed for kinesins (6,22,27). The model also explains the existence of the distinct and topologically conserved feature of kinesin, the ␤-domain, whose functional role remained mysterious. In its newly proposed role for conventional kinesin, ␤-domain would function similar to the extended ␤-strands of the central core, stabilizing and pointing the neck linker either toward plus or minus end directions in the ATP/ADP⅐P i and nucleotide-free states of the motor, respectively. In non-processive plus end-directed kinesins, the ␤-domain would serve a similar function, stabilizing the ADP/nucleotide-free conformation of the neck linker and defining the amplitude of the power stroke of the motor (31). Recent structural studies (35,36) of the minus end-directed motor Ncd suggested that the amplitude of the oppositely directed power stroke might be defined by interdomain interactions modulated by the ␤-domain.
Packing interactions of the conventional neck linker with the ␤-domain in the nucleotide-free state are less extensive compared with interactions of the linker with the core in the ATP-like state. For this reason, the docked conformation of the linker in the nucleotide-free motor may be less stable. In the bridged state of kinesin, this conformation is enforced by the trailing motor tugging the leading neck in the minus end direction. Consistent with this idea, the neck linker of the nucleotide-free microtubule-bound monomeric kinesin is flexible and adopts multiple conformations (25,30). However, some of these conformations seen both in nucleotide-free and ADP-bound motor (25) are consistent with the neck linker being oriented backwards and docked along the ␤-domain.
The function of the ␤-domain in conventional kinesin likely extends beyond its role in supporting the nucleotide-free conformation of the leading neck. Upon binding ATP, the counter-

Minireview: Coordination in Kinesin Motors
15709 movement of the leading core and resulting repositioning of the ␤-domain would dislodge the neck linker from the small lobe (similar to dislodging the neck from its docked position on the central core by the collapsing switch II cluster upon nucleotide hydrolysis (6,22,27)). Furthermore, the ATP-induced countermovement of the core would bring the N-terminal base of the neck (conserved small hydrophobic residue, Ile 325 in human kinesin, Fig. 1C) closer to the hydrophobic pocket underneath the switch II cluster (6,22,27). In the extended conformation of the linker docked on the ␤-domain, Ile 325 is placed ϳ10 Å from its position in the ATP-like state (Fig. 1C) and, therefore, is unlikely to move into the pocket because of hydrophobic interactions. The assisted repositioning of the base of the neck would facilitate docking of Ile 325 , triggering docking of the entire neck linker along the core in the ATP state (25). Because the transition between two alternative docked conformations of the neck linker would not rely on brownian motion only but would be forced by the countermovement of the core, placement of the detached trailing motor toward the new site on the track could be achieved more efficiently. Importantly, the countermovement of the leading motor upon binding ATP would also prevent the trailing motor from rebinding to the same site on the microtubule. Electron microscopy and crystallographic data (25,31) suggest that the backward orientation of the neck linker could be achieved not only in the nucleotide-free but also in the ADP state of kinesin. These observations support the proposed model (Fig. 3), which predicts that even transient docking of the neck linker along the ␤-domain in the unattached ADPbound motor domain (Fig. 3A) would make repositioning of this domain toward the next binding site more energetically favorable. The free-energy gain would come from favorable enthalpy changes balancing out unfavorable entropy changes associated with straightening of the leading neck linker (Fig. 3, A and B) during forward stepping (46).

Future Challenges
High-resolution structures of the kinesin motor domain complexed with tubulin polymer at different stages of the mechanochemical cycle of the motor would, no doubt, clarify many of the questions raised in this review. So far, moderate resolution of the available electron microscopy images of the kinesin bound to its track prevent us from dissecting the exact structural mechanisms underlying the processive movement of the motor.