Neck Rotation and Neck Mimic Docking in the Noncatalytic Kar3-associated Protein Vik1*

Background: Kar3Vik1 is a heterodimeric kinesin with one catalytic subunit (Kar3) and one noncatalytic subunit (Vik1). Results: Vik1 experiences conformational changes in regions analogous to the force-producing elements in catalytic kinesins. Conclusion: A molecular mechanism by which Kar3 could trigger Vik1's release from microtubules was revealed. Significance: These findings will serve as the prototype for understanding the motile mechanism of kinesin-14 motors in general. It is widely accepted that movement of kinesin motor proteins is accomplished by coupling ATP binding, hydrolysis, and product release to conformational changes in the microtubule-binding and force-generating elements of their motor domain. Therefore, understanding how the Saccharomyces cerevisiae proteins Cik1 and Vik1 are able to function as direct participants in movement of Kar3Cik1 and Kar3Vik1 kinesin complexes presents an interesting challenge given that their motor homology domain (MHD) cannot bind ATP. Our crystal structures of the Vik1 ortholog from Candida glabrata may provide insight into this mechanism by showing that its neck and neck mimic-like element can adopt several different conformations reminiscent of those observed in catalytic kinesins. We found that when the neck is α-helical and interacting with the MHD core, the C terminus of CgVik1 docks onto the central β-sheet similarly to the ATP-bound form of Ncd. Alternatively, when neck-core interactions are broken, the C terminus is disordered. Mutations designed to impair neck rotation, or some of the neck-MHD interactions, decreased microtubule gliding velocity and steady state ATPase rate of CgKar3Vik1 complexes significantly. These results strongly suggest that neck rotation and neck mimic docking in Vik1 and Cik1 may be a structural mechanism for communication with Kar3.

Eukaryotic cells rely on nanometer-sized motors called kinesins to transport cellular components along microtubules (1) or to help build the mitotic spindle and distribute chromosomes between daughter cells (2,3). Recent studies have shown that dynamic interactions between the neck and a short region of either the N or C terminus of the motor domain form a structure responsible for force generation by the neck (4 -7) and that its conformation and interactions with the motor domain core, or regulatory proteins, is linked to the nucleotide-and microtubule-binding state of the motor (8,9). In kinesin-1, this region forms an N-terminal extension of the motor domain, called the "cover strand" (5), and in kinesin-14 motors this region is at the C terminus, after the ␣6 helix, and has been dubbed the "neck mimic" (8).
Kar3 is a kinesin-14 that plays essential roles in mitosis, meiosis, and karyogamy in Saccharomyces cerevisiae and Candida albicans (10 -13). These include cross-linking, stabilizing, and sliding spindle pole microtubules, as well as depolymerizing microtubules (10,14). To accomplish this array of functions, Kar3 associates with two discrete regulatory subunits, Cik1 and Vik1 (14 -16), whose motor homology domain (MHD) 2 can bind microtubules but not ATP (17,18). Amazingly, the affinity of the Cik1 and Vik1 subunits in their respective complexes with Kar3 is reduced when ADP is exchanged in Kar3 for AMP-PNP (a nonhydrolyzable ATP analog), indicating that Kar3 controls disengagement of Cik1 and Vik1 from microtubules (17,18). This form of intramolecular coordination between a catalytic and noncatalytic subunit-containing complex has not been previously observed in other motor proteins.
The recently determined structure of a truncated version of ScKar3Vik1 by Rank et al. (19) shows that the stalk and neck domain of Kar3 and Vik1, like Drosophila Ncd, forms a continuous coiled coil, nearly to the point of insertion of their neck into the motor (homology) domain (supplemental Fig. S1A). Also shown was that Kar3 and Vik1 interact laterally over two adjacent protofilaments when Kar3 is ADP-bound but that the complex transitions to a single-head bound state involving only Kar3 when ADP is exchanged for AMPPNP. Uptake of AMP-PNP also caused a minus-end-directed rotation of the stalkneck element, which resembles the power stroke of Ncd (20,21). In both these putative "pre-power stroke" and "post-power * This work was supported by grants from the Natural Sciences and Engineering Research Council of Canada and Canadian Institutes of Health Research (to J. S. A.). □ S This article contains supplemental Figs. S1-S8, Movies S1-S4, and additional references. The atomic coordinates and structure factors (codes 4GKP,4GKQ,and 4GKR) have been deposited in the Protein Data Bank (http://wwpdb.org/). 1  stroke" configurations, the Vik1 subunit is held in a similar conformation as the previous structure of the ScVik1MHD monomer (17), and thus conformational changes required to transition Vik1 from a high to low affinity binding state for the microtubule remain uncertain. Orthologs of Cik1 and Vik1 can be found in numerous fungal relatives of S. cerevisiae (22,23). By studying the structure and function of these proteins, we hypothesized that it may be possible to identify functionally conserved elements that underlie a general mechanism for communication with Kar3. This study focuses on Kar3 and Vik1 proteins from the human fungal pathogen, Candida glabrata. Its genome encodes a single 692amino acid Kar3 homolog (CAGL0D04994g) that shares 53% sequence identity with ScKar3, and a 584-amino acid Vik1 ortholog (CAGL0H00638g) that shares 23% identity with ScVik1. We have determined the x-ray structures of the motor domain regions of both CgKar3 and CgVik1, the latter of which crystallized in three unique conformations that were not previously observed in this protein. Two of the conformations exhibit rotation of the neck element in a manner similar to Ncd motors in different nucleotide states, suggesting that they may represent discrete intermediates during Kar3Vik1's motile cycle. To determine the importance of neck rotation in CgVik1, we mutated the residue that appears to facilitate neck rotation, as well as one of the neck residues involved in a specific network of interactions with the MHD core. Analysis of these mutants in complex with wild-type CgKar3 showed markedly reduced rates of microtubule gliding and ATP hydrolysis. In these structures, we also observed docking of the C terminus of CgVik1 onto the central ␤-sheet and its interaction with residues at the neck-core junction in a conformation similar to the C terminus of the ATP-bound form of Ncd (7), the neck mimic of the calcium-regulated kinesin-14 KCBP (8), and the neck linker of kinesin-1 (24). Interestingly, as seen in a separate CgVik1 structure, this docking of the C terminus seems to be completely abolished when the neck loses its ␣-helical structure and is stabilized in a state where it is undocked from the MHD. Based on this new information, we propose that the neck of Vik1 must change positions during the motile cycle of Kar3Vik1, and interactions between the neck and C terminus of Vik1 may provide the critical link for coupling release of Vik1 and Cik1 from microtubules to microtubule binding and subsequent neck rotation in Kar3 during nucleotide exchange.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-Plasmids encoding truncated C. glabrata Kar3 and Vik1 proteins ( Fig. 1A and supplemental Figs. S5A and S7A) were constructed by amplifying relevant regions of C. glabrata genomic DNA (ATCC 2001D-5) by PCR and ligating them into either pET24d (Novagen) or a modified version of pET16b (Novagen) in which the Factor Xa cleavage site was replaced with an recombinant tobacco etch virus protease cleavage site. CgVik1 mutagenesis was performed using QuikChange (Stratagene). All plasmids were transformed into Escherichia coli BL21-CodonPlus (DE3)-RIL cells (Stratagene) for protein expression in LB or M9 minimal media. Cells transformed with the pET24d vector encoding CgKar3-NϩMD (MA-Glu 324 -Asn 692 ) (supplemental Fig. S7A) were grown in LB media supplemented with 50 g/ml kanamycin and 50 g/ml chloramphenicol to A 600 ϳ0.8 and were then induced with 1.0 mM isopropyl 1-thio-␤-D-galactopyranoside and incubated at 20°C for a further 16 h. Cells were harvested by centrifugation and lysed by flash freezing and sonication, and the CgKar3-NϩMD protein was purified by ion-exchange chromatography (DEAE SP-Sepharose Fast Flow, GE Healthcare) as described previously (25). Cells expressing CgVik1-sNϩMHD (MGH 10 SSGRENLYFQGHM-Leu 313 -Lys 584 ) or CgVik1-NϩMHD (MGH 10 SSGRENLYFQGHM-Thr 302 -Lys 584 ) were grown in M9 minimal media supplemented with 50 g/ml ampicillin and 50 g/ml chloramphenicol to A 600 ϳ0.8 and were then induced with 1.0 mM isopropyl 1-thio-␤-D-galactopyranoside and incubated at 25°C for a further 16 h. Both constructs were purified and digested with the recombinant tobacco etch virus protease to remove the polyhistidine tag as described previously for ScVik1MHD (17). Cells co-transformed with plasmids encoding CgVik1-CCϩNϩMHD (MGH 10 SSGRENLYFQGHM-Asp 152 -Lys 584 ) and CgKar3-CCϩNϩMD (Met 198 -Asn 692 ) were grown in LB media supplemented with 50 g/ml kanamycin, 50 g/ml ampicillin, and 50 g/ml chloramphenicol to A 600 ϳ0.8 and induced with 1.0 mM isopropyl 1-thio-␤-D-galactopyranoside. After continued growth at 20°C for 16 h, cells were lysed by sonication, and the dimer was purified as described previously (17) followed by gel filtration on a HiLoad Superdex 200 26/60 column (GE Healthcare). All proteins were concentrated and flash-frozen in liquid nitrogen for storage.
Protein Crystallization, Data Collection, and Structure Determination-Crystals of CgVik1-sNϩMHD grew at 4°C in 0.1 M HEPES, pH 7.5, 0.075 M NaAc, pH 5.5, 13% PEG6000 and 5% isopropyl alcohol by hanging drop vapor diffusion. Data were collected at the X6a beamline at National Synchrotron Light Source (Brookhaven National Laboratory). CgVik1-NϩMHD crystals grew by hanging drop in 0.1 M Tris, pH 8.5, 12% PEG 8000, 0.05 M MgCl 2 , 0.15 M NaCl, 5% ethylene glycol, and 1 mM tris(2-carboxyethyl)phosphine. Data were collected at the GM/CA-CAT 23-ID-B beamline at Advanced Photon Source (Argonne National Laboratory). CgKar3-NϩMD crystals grew by hanging drop at room temperature in 0.1 M MIB buffer, pH 7.0, 25% PEG 1500, 1 mM tris(2-carboxyethyl)phosphine, and 1 mM ATP. Data were collected at the A1 beamline at Cornell High Energy Synchrotron Source (Cornell University). XDS was used to integrate and scale the reflection data of CgVik1-sNϩMHD and CgVik1-NϩMHD, and HKL2000 was used for that of CgKar3-NϩMD (26,27). All three structures were solved by molecular replacement with PHENIX AutoMR (28). For CgVik1-sNϩMHD, the motor homology domain region of S. cerevisiae Vik1 (Protein Data Bank (PDB) code 20OA) was used as the initial search model (17). For CgVik1-NϩMHD, the CgVik1-sNϩMHD structure was used as the initial search model. For CgKar3-NϩMD, the motor domain of S. cerevisiae Kar3 (PDB code 3KAR) was used as the initial search model (29). Manual building of the structures were done in COOT and refined in REFMAC to generate the final models 4GKP (CgVik1-sNϩMHD), 4GKQ (CgVik1-NϩMHD), and 4GKR (CgKar3-NϩMD) (30,31). Diffraction data collection and structure refinement statistics are presented in Table 1.
Analysis of Kinesin Motility-Motility assays were performed using WT and mutant CgKar3Vik1 complexes at ϳ0.3 M. Rhodamine-labeled bovine tubulin (Cytoskeleton Inc.) was mixed with unlabeled tubulin purified from bovine brain at a molar ratio of 1:4, and the mixture was polymerized, centrifuged, and resuspended in BRB80 (80 mM PIPES, pH 6.8 (KOH), 1 mM MgCl 2 , 1 mM EGTA). Polarity-marked microtubules were made with rhodamine-labeled microtubules and HiLyte TM 488 microtubules (Cytoskeleton Inc.) by adapting a protocol as described previously by Walter et al. (32). Homemade perfusion chambers were constructed by sandwiching 22 ϫ 60-and 22 ϫ 22-mm glass coverslips using double-sided tape. Microtubules were shredded to generate short filaments before imaging. Anti-histidine antibodies (Fisher) were used to attach the kinesins to the glass surface, which was then blocked with 1 mg/ml BSA. Imaging was performed on spinning disc confocal fluorescence microscope (Leica) in the presence of an oxygenscavenging system (OSM; BRB80, 1.5 mM MgAc, 1 mg/ml BSA, 200 g/ml glucose oxidase, 175 g/ml catalase, 25 mM glucose, 2 mM ␤-mercaptoethanol), 20 M taxol, 1 mM AMPPNP, and 1 mM ATP as described previously (25). To ensure sufficient ATP was present during the imaging process, ATP was regenerated by supplementing the OSM with 0.3 g/ml phosphocreatine kinase and 2 mM phosphocreatine (25). Quorum WaveFX software (Quorum Technologies Inc.) was used to process the images captured and compile them into movies. Microtubule movement was tracked using Image-Pro Plus (Media Cybernetics Inc.).
Steady State ATPase Activity-ATPase kinetics of the WT and mutant CgKar3Vik1 motors were determined using a pyru-vate kinase/lactate dehydrogenase-coupled NADH oxidation reaction (33) as described previously (34).
Steady State Microtubule Binding Affinity-Purified CgKar3-NϩMD or CgKar3Vik1 (6 M final concentration) was incubated with microtubules in ATPase buffer (20 mM HEPES, 5 mM magnesium acetate, 0.1 mM EGTA, 0.1 mM EDTA, 25 mM potassium acetate, 1 mM DTT, 40 M Taxol, pH 7.2) with either 1 mM MgAMPPNP or MgADP, and bound and unbound motors were separated by centrifugation at 55,000 rpm in a TLA100 ultracentrifuge and analyzed by SDS-PAGE as described previously (34).

Crystal Structures of the CgVik1 Motor Homology Domain-
The structures of two different constructs comprising different portions of the neck region (N) and the entire MHD of CgVik1 were determined by x-ray crystallography (Fig. 1A). The structure of the first construct, CgVik1-sNϩMHD was solved to 2.42 Å by molecular replacement using the ScVik1MHD crystal structure (PDB code 2O0A) as the search model (17). The solution showed two molecules in the asymmetric unit whose intermolecular interactions are insufficient to indicate complex formation. Similar to ScVik1MHD, the CgVik1MHD structural motif is an ␣/␤ fold that lacks a nucleotide binding pocket (Fig.  1B). Unlike ScVik1MHD, the neck in molecule A of CgVik1-sNϩMHD is completely disordered N-terminally from Arg 325 (supplemental Fig. S2C), and in molecule B it is extended away (undocked) from the MHD core without defined secondary structure (supplemental Fig. S2D). The extended neck confor- where R work refers to the R factor for the data utilized in the refinement and R free refers to the R factor for 5% (models 4GKR and 4GKP) or 10% (model 4GKQ) of the data that were excluded from the refinement. mation is stabilized by symmetry contacts with the ␤5a/b strands of another molecule in the crystal lattice. Crystals for the CgVik1-NϩMHD construct, which possesses 11 additional residues of the neck (Fig. 1A), also showed two molecules in its asymmetric unit after refinement of its structure to 2.99 Å. In contrast to the CgVik1-sNϩMHD molecules, the neck of both molecules of this construct is almost entirely helical, and each differs in their alignment along the MHD core by a ϳ37°rotation about a conserved glycine residue (Fig. 1, B and C). The neck of molecule A is oriented downward and away from the small ␤-lobe (␤1a, ␤1b, and ␤1c) at the edge of the MHD in a manner that is surprisingly similar to several previous Ncd structures (PDB codes 1CZ7, 2NCD, and 1N6M) (supplemental Fig. S3A) (7,35,36). In molecule B, the neck is rotated upward relative to molecule A and is in the same position as the ScVik1 neck in the ScVik1MHD and cross-linked SHD-Kar3Vik1 crystal structures (PDB codes 2O0A and 4ETP) (supplemental Figs. S1B and S3B) (17,19).
Interactions between the Neck and Motor Homology Domain Core of CgVik1-Core residues in the MHD that interact with the neck of the CgVik1-NϩMHD construct are found in the ␤1 strand, the C-terminal half of helix ␣1, loop L13, and the N terminus of the ␤8 strand (Fig. 2). These residues form discrete networks of electrostatic, hydrogen bonding, and van der Waals interactions to stabilize the neck against the core in molecule A and B and cover 317 and 372 Å 2 , respectively. In molecule A, Glu 319 of the neck forms salt bridges with Arg 325 from strand ␤1 and Lys 552 from ␤8 ( Fig. 2A). Also, Ser 316 forms a hydrogen bond with Arg 388 from ␣1, and a single bridging water molecule forms an additional reinforcement between Asn 315 and Arg 388 . In conformation B, The MHD is colored from the N to C terminus with a continuous spectrum of rainbow colors (blue to red), and the coiled-coil regions are in yellow (44). B and C, ribbon representations of the CgVik1-sNϩMHD structures (molecules A and B) and CgVik1-NϩMHD structures (molecules A and B) are colored the same as the MHD in the bar diagram in A. Secondary structure elements are numbered consecutively from the N terminus of the construct, adopting the system used previously by Allingham et al. (17). Loops with poor electron density were not built into the final models. The "pivot glycine" residue (Gly 322 ) is indicated in each structure and shown as a sphere. This figure was prepared with PyMOL (42).
the side chain of Glu 319 undergoes a 72°rotation to form salt bridges of different geometries with Lys 552 and Arg 325 (Fig.  2B). The water-mediated hydrogen bond between Asn 315 and Arg 388 was also disrupted in favor of a new hydrogen bond between Asn 315 and Arg 325 . These residues, and others immediately surrounding the neck-core junction of CgVik1, exhibit the highest degree of sequence conservation between different Vik1 orthologs and kinesin-14 motors (supplemental Fig. S4) (37,38). The residue allowing the neck to pivot between these interfaces, Gly 322 , is completely conserved among all kinesins and sequenced Vik1 and Cik1 orthologs (supplemental Fig. S4C). Additional stabilizing interactions for each neck conformation of CgVik1-NϩMHD are formed by contacts with symmetry-related molecules. Structural Changes in the MHD Core of CgVik1 That Coincide with Neck Isomerization-A major difference in the MHD core that coincides with structural changes in the neck involves the C terminus of CgVik1 (Fig. 2, C-F). In molecules A and B of the CgVik1-NϩMHD crystal, and in molecule A of the CgVik1-sNϩMHD crystal, five residues beyond the end of helix ␣6 can be modeled, three of which were not previously observed in ScVik1MHD (Fig. 2, C--E). However, in molecule B of the CgVik1-sNϩMHD crystal, the C terminus is completely disordered, and part of the ␣6 helix melts (Fig. 2F).
When the C-terminal region is visible, we observe it docking onto the central core of the MHD and forming a short twostranded ␤-sheet with the neck-core junction. This configuration is structurally analogous to the neck mimic of kinesin-14 family members, which has been observed in recent structures of Ncd (PDB code 3L1C) (7) and the calcium-regulated plant kinesin KCBP (PDB code 1SDM) (8).
In both molecules of the CgVik1-NϩMHD construct, stabilizing hydrogen bonds between the neck and C terminus are formed between the backbone amide of Ala 324 and the carbonyl of Ile 578 and between the backbone carbonyl of Gly 322 and the amide of Asn 580 (Fig. 2, C and D). In this conformation, the side chain of Ile 578 is inserted into a hydrophobic pocket formed by Ala 324 , Thr 530 , Phe 548 , Phe 555 , and Leu 575 within the MHD core, similar to the interaction observed in the ATP-like conformation of KCBP (8,9). Sequence alignment of the C terminus of other Vik1 and Cik1 proteins shows that Ile 578 is moderately conserved (supplemental Fig. S4A), and it also shows some conservation in kinesin-14 motors (supplemental Fig. S4D) (6). Additionally, the side chain of Asn 580 is sandwiched directly between Lys 552 (␤8) and the pivot point of the neck and is only 4.0 Å away from the Glu 319 -Arg 325 and Glu 319 -Lys 552 salt bridges. In this position, it could potentially destabilize either of these interactions and allow the neck to transition from one conformation to the other more readily.
The basis for complete disorder of the C terminus and melting of part of the ␣6 helix in molecule B of CgVik1-sNϩMHD may be related to the repositioning of the side chain of Arg 325 in relation to Asn 580 when the neck is stabilized in an extended, non-␣-helical conformation (Fig. 2F). In this configuration, Arg 325 moves toward Ser 553 from strand ␤8 and into the pocket that was occupied by Asn 580 . Combined with the dramatic change in position of Ala 324 and Gly 322 , which provide hydrogen bond acceptors and donors for aligning the C terminus against the neck-core junction, we speculate that these structural changes may be the basis for undocking the neck mimic from the MHD core. Interestingly, when the C terminus is undocked, Phe 577 takes the place of Ile 578 , perhaps as a means of stabilizing the core.
The fact that the C-terminal segment in molecule A of CgVik1-sNϩMHD is nearly identical to that of molecules A and B in the CgVik1-NϩMHD crystal is more difficult to reconcile given the nonlocalized position of its neck (Fig. 2E). Similar to molecule B of CgVik1-sNϩMHD, Arg 325 torsions toward Asn 580 and Ser 553 (Fig. 2F), but in this case it hydrogen bonds with Asn 580 . Asn 580 also interacts with a nearby water molecule that is bridged to the backbone amide of Ser 553 and is within hydrogen bonding distance with the side chain of Glu 506 from a symmetry-related molecule. Perhaps these interactions, along with hydrophobic interactions between Ile 578 and Pro 581 of the C terminus and the MHD core, are sufficient to achieve stabilization of the neck mimic in molecule A of the CgVik1-sNϩMHD crystal.
Mutations in the Neck of Vik1 Decrease the Microtubule Gliding Velocity of CgKar3Vik1-We were curious to understand the functional significance of CgVik1 neck isomerization in the motile cycle of CgKar3Vik1. To investigate this, mutations were created in two of the residues that enable the conformational changes observed. Glu 319 was chosen for mutation to alanine because we suspected that its involvement in salt bridge formation with the highly conserved MHD core residues Arg 325 and Lys 552 functions to stabilize each neck conformation and may create a neck position "sensor" for the MHD. The other residue chosen for mutagenesis was Gly 322 . It was mutated to alanine and proline to probe the importance of rotational freedom in the neck on the motile cycle of CgKar3Vik1. These mutations were generated in CgVik1 constructs that included the MHD, the neck, and an extended length of the native coiled coil to allow dimerization with similar constructs of CgKar3 (supplemental Fig. S5A). Based on the arrangement of subunits in the cross-linked synthetic heterodimer of ScKar3Vik1 (19), none of the mutants should disrupt the inter-helical interactions required for coiled-coil formation between Kar3 and Vik1. Indeed, co-expression of CgKar3-CCϩNϩMD (Met 198 -Asn 692 ) and CgVik1-CCϩNϩMHD (Asp 152 -Lys 584 ) proteins from E. coli and co-purification of their resulting wild-type (WT) and mutant complexes by nickel-nitrilotriacetic acid affinity chromatography produced stable heterodimers according to analytical size-exclusion chromatography (supplemental Fig. S5, B and C).
In microtubule-gliding assays, the CgKar3Vik1 G322A and G322P mutants were nearly 2-and 7-fold slower, respectively, than CgKar3Vik1 WT complexes (4.79 Ϯ 0.022 m/min), which migrated with plus-ends leading (Fig. 3, A-C, and supplemental Fig. S6 and supplemental Movies S1-S3). The larger impact of the G332P substitution on motility is consistent with the additional restrictions proline would place on possible neck conformations relative to alanine. This indicates that conformational freedom of CgVik1's neck at Gly 322 is important for motility of the CgKar3Vik1 motor. With an average microtubule gliding speed of 4.34 Ϯ 0.047 m/min, the E319A mutant had a more modest effect on motility compared with the G322A and G322P mutants; however, it did show a slightly higher number of microtubule "stalling" events than WT (Fig. 3D and supplemental Movie S4). This suggests that the CgKar3Vik1 motor is sensitive to changes in interactions between the neck and core of CgVik1 and that disruption of the Glu 319 -Arg 325 and Glu 319 -Lys 552 salt bridges may allow the neck to rotate and acquire conformations that are not always conducive to appropriate Kar3-MT interactions or force production events. Notably, in none of the motility studies did we observe obvious dissociation of microtubules from the coverslips, indicating that the CgVik1 subunit was interacting with MTs and that the mutations were not having a negative effect on microtubule binding of the CgKar3Vik1 complex.

JOURNAL OF BIOLOGICAL CHEMISTRY 40297
To confirm that both CgKar3 and CgVik1 participate in microtubule binding, equilibrium co-sedimentation experiments were performed on our wild-type CgKar3Vik1 dimer construct and a monomeric construct of CgKar3 (CgKar3-NϩMD) (supplemental Figs. S7 and S8). Unfortunately, the CgVik1 monomers used for our crystallization studies pelleted in the centrifuge tube regardless of the presence of microtubules, presumably due to aggregation, and thus they could not be used reliably for direct measurement of dissociation constants. Instead, CgVik1's microtubule binding affinity can be inferred from the difference in K d,MT of the CgKar3Vik1 WT dimer relative to the monomeric CgKar3-NϩMD construct in the presence of ADP, which confers a weak microtubule-binding state in kinesins. From the binding curves in supplemental Fig. S8, C and D, we were able to calculate that the K d,MT for the ADP complex of the CgKar3Vik1 WT dimer is 4.8 Ϯ 1 M, whereas CgKar3-NϩMD is nearly twice as weak at 7.6 Ϯ 2.7 M. This result indicates that in the ADP state, CgVik1 is contributing to the microtubule binding affinity of the CgKar3Vik1 complex. Moreover, similar K d,MT values for the AMPPNP complexes of CgKar3-NϩMD and CgKar3Vik1 WT dimer (3.05 Ϯ 0.8 and 3.65 Ϯ 0.8 M, respectively) suggests that, in the presence of ATP, the CgKar3 subunit must form the majority of the interactions between CgKar3Vik1 complexes and microtubules, whereas CgVik1 contributes little to microtubule binding.
Vik1 Neck Mutants Exhibit Reduced Microtubule-activated ATPase Activity-Even though the mutations we engineered were introduced into the noncatalytic subunit of the CgKar3Vik1 complex, we wanted to learn their effect on the biochemical function of Kar3. Compared with the WT dimer (0.95 s Ϫ1 ), the steady state ATPase kinetics studies show a 1.8and 11-fold decrease in k cat for the CgKar3Vik1 G322A (0.54 s Ϫ1 ) and G322P (0.088 s Ϫ1 ) mutants, respectively (Fig. 4). These data suggest that the slower rate of motility of the mutants is related to their creation of defects in the ability of CgKar3's ATPase activity to be activated by microtubules. This could be a consequence of the mutants impeding rotational freedom of the CgVik1 neck that is obligatory for appropriate CgKar3-microtubule interactions. An alternative possibility is that ATP binding or hydrolysis by CgKar3 is impeded by the CgVik1 neck rotation defects in a microtubule-independent manner, perhaps as a consequence of disruption of important interactions between the neck and motor domain core of CgKar3.
Similar to but much more dramatic than its effect on the microtubule-gliding assays, the E319A mutant lowered the k cat of CgKar3Vik1 by 3.7-fold (0.26 s Ϫ1 ) (Fig. 4), providing further evidence that the interactions between the neck and MHD core of Vik1 form a determinant of the mechanochemical cycle of the Kar3Vik1 motor. As is reflected by their similar K 0.5, MT for ATPase activation (Fig. 4C), none of the mutants dramatically lowered the affinity of the CgKar3Vik1 complex for microtubules. However, they did exhibit different trends in their ATP concentration-dependent activities; CgKar3Vik1 G322P and E319A showed lowered K m, ATP values relative to WT, whereas

Neck Rotation and Neck Mimic Docking in Vik1
the K m, ATP value for CgKar3Vik1 G322A increased. This may reflect differences in the effect of each mutant on Kar3's ability to undergo ATP-promoted isomerization during the mechanochemical cycle.

DISCUSSION
The mechanism by which kinesin motor assemblies coordinate the microtubule interactions and force production cycles of their subunits has mainly been studied in kinesins with two identical ATP-binding heads. In both processive and nonprocessive forms of these kinesins, the neck region is observed to be either docked onto or undocked from the motor core in crystal structures, and these states of the motor are generally determined by the nucleotide state of the motor domain.
Recent structural studies of the heterodimeric ScKar3Vik1 motor have shown that Kar3 and Vik1 bind side-by-side to adjacent protofilaments when Kar3 is in an ADP-bound state, and there is an ATP-dependent rotation of the neck of Kar3 that is nearly identical to Ncd (19). Unfortunately, high resolution structural information could not be provided to describe in detail the configuration of Vik1 in this state nor any of the other early states leading up to its release from microtubules when ADP is exchanged for ATP in Kar3. Our crystal structures of CgVik1 provide information about previously unseen states of Vik1 that may be relevant to the initial association of Kar3Vik1 with the microtubule. Through mutational analysis, we also provide evidence of head-head communication in the heterodimeric Kar3Vik1 motor assembly, as well as mechanistic details about the communication route.
Our studies show that the helical neck of CgVik1 can rotate with respect to the MHD, allowing separate clusters of conserved electrostatic, hydrophobic, and hydrogen bond-forming residues in the neck and core to interact. This rotation occurs by a change in the backbone torsion angles at the conserved Gly 322 , which appears to be critical for the microtubule-activated ATPase activity and motility of CgKar3Vik1. Although motility analysis of the constrained ScKar3Vik1 complex used for structural determination by Rank et al. (19) (cross-linked SHD-Kar3Vik1) suggested that rotation of the Vik1 neck is not critical for the Kar3Vik1 power stroke, a reduction in the rate of motility of this construct relative to the noncross-linked construct (noncross-linked SHD-Kar3Vik1) was observed.
We propose that neck rotation observed in our unconstrained CgVik1-NϩMHD construct represents early Vik1-microtubule binding states that facilitate proper Kar3 contacts on the adjacent microtubule protofilament (Fig. 5). Specifically, we suggest that during initial collision of Kar3Vik1 with the microtubule (Fig. 5, Step 1), before Kar3 binds, the neck of Vik1 is in the "up" position (i.e. molecule B of CgVik1-NϩMHD crystals). Next, rotation of neck to the "down" position (i.e. molecule A of CgVik1-NϩMHD) could help orient the head of Kar3 into closer proximity with the available microtubule binding position on the adjacent protofilament (Fig. 5, Step 2) in accord with the reverse orientation of Kar3 and Vik1 suggested by Rank et al. (19). The resulting two-headed binding state of Kar3Vik1, and the contemporaneous release of ADP, may cause some separation of the heads, placing strain on the neck of Vik1 sufficient to melt its helical structure in a manner analogous to molecule B of CgVik1-sNϩMHD (Fig. 5, Step 3). This is supported by the observation by Rank et al. (19) that cross-linking Kar3Vik1 at the base of the neck in a manner restricting such head separation led to a change in the microtubule-binding stoichiometry and motility of the motor (17). Without sufficient interactions between the neck-core junction and C terminus of Vik1 to maintain the small two-stranded ␤-sheet formed by these elements, undocking of the neck mimic could ensue. By analogy of the microtubule-binding and motility defects created by mutation or deletion of the neck mimic region in Ncd (6,9,39), and given the relationship between neck mimic docking and the ATPase state of KCBP (9,40,41), such movements of the C terminus of Vik1 could lead to dissociation of Vik1 from the microtubule. In this state, the coiled coil of Kar3Vik1

Neck Rotation and Neck Mimic Docking in Vik1
NOVEMBER 23, 2012 • VOLUME 287 • NUMBER 48 would be free to rotate toward the microtubule minus end as Kar3 binds ATP, completing the power stroke (Fig. 5, Step 4).
We therefore propose that the residues involved in the interactions between the neck, MHD core, and neck mimic in Vik1 constitute part of the communication route between Kar3 and Vik1 (Fig. 5) and that isomerization of the Vik1 neck may be the actuator that "flips" a molecular switch controlling Vik1's release from the microtubule to allow motility. As noted by Khalil et al. (4), most kinesins, and approximately half of all single-domain proteins in the Protein Data Bank (43), have interacting N-and C-terminal elements that experience dynamic interconversion between different types of secondary structure, or undergo disorder-to-order transitions. It appears that Vik1 may have adapted this behavior for regulation of its microtubule binding interactions by Kar3 in lieu of nucleotide binding and enzymatic function. Step 1 Step 2 Step 3 Step 4 Step 5 ADP FIGURE 5. Schematic relating observed Vik1 conformations to Kar3Vik1 microtubule binding and movement.
Step 1 represents the initial collision of a Kar3Vik1 complex with MTs where Vik1 is bound to the MT protofilament with its neck in the up conformation (molecule B of CgVik1-NϩMHD, shown in Fig. 1D) and its neck mimic (NM) docked against the MHD. In this step, Kar3 is in a low affinity state for MTs. Amino acid interactions that stabilize this conformation of Vik1 are shown in the right panel.
Step 2, the Vik1 neck rotates downward (molecule A of CgVik1-NϩMHD, shown in Fig. 1C) and positions the Kar3 head closer to the MTs.
Step 3, Kar3 binding to MTs and release of ADP induces strain in the coiled coil that may melt the Vik1 neck and cause it to disengage (undock) from the MHD core (molecule B of CgVik1-sNϩMHD, shown in Fig. 1B). To illustrate this, Vik1 has been displaced slightly from its putative microtubule-binding site on the ␣-tubulin subunit lateral to the Kar3 interaction site. Accompanying this is the repositioning of Arg 325 and contemporaneous undocking of the neck mimic, resulting in Vik1 release from MTs (right panel).
Step 4 represents the completion of the power stroke by Kar3's neck rotation. When Vik1 is free from the microtubule, relief of strain in the coiled coil could permit the neck to reform a helical structure.
Step 5 shows detachment of Kar3Vik1 complex from the microtubule following ATP hydrolysis and product release.